Sensing of SARS-CoV-2 by pDCs and their subsequent production of IFN-I contribute to macrophage-induced cytokine storm during COVID-19

pDCs are at the eye of the storm

In severe COVID-19, macrophages induce cytokine storms, which can lead to poor patient outcomes. However, macrophages are not directly infected by SARS-CoV-2, so how this cytokine storm is induced remains unclear. Here, Laurent et al. used COVID-19 patient databases and cell culture to identify that the macrophage-induced cytokine storm was linked to IFN-I signaling in patient lungs. Plasmacytoid dendritic cells (pDCs) were the main producers of IFN-I, because they were directly infected with SARS-CoV-2, which triggered TLR7 activation. This IFN-I made macrophages more responsive to environmental stimuli, thus triggering the production of multiple cytokines. Thus, the authors present a mechanism whereby pDCs are infected by SARS-CoV-2, subsequently producing IFN-I, and stimulating a macrophage-mediated cytokine storm during SARS-CoV-2 infection.

Abstract

Lung-infiltrating macrophages create a marked inflammatory milieu in a subset of patients with COVID-19 by producing a cytokine storm, which correlates with increased lethality. However, these macrophages are largely not infected by SARS-CoV-2, so the mechanism underlying their activation in the lung is unclear. Type I interferons (IFN-I) contribute to protecting the host against SARS-CoV-2 but may also have some deleterious effect, and the source of IFN-I in the lungs of infected patients is not well defined. Plasmacytoid dendritic cells (pDCs), a key cell type involved in antiviral responses, can produce IFN-I in response to SARS-CoV-2. We observed the infiltration of pDCs in the lungs of SARS-CoV-2–infected patients, which correlated with strong IFN-I signaling in lung macrophages. In patients with severe COVID-19, lung macrophages expressed a robust inflammatory signature, which correlated with persistent IFN-I signaling at the single-cell level. Hence, we observed the uncoupling in the kinetics of the infiltration of pDCs in the lungs and the associated IFN-I signature, with the cytokine storm in macrophages. We observed that pDCs were the dominant IFN-α–producing cells in response to the virus in the blood, whereas macrophages produced IFN-α only when in physical contact with infected epithelial cells. We also showed that IFN-α produced by pDCs, after the sensing of SARS-CoV-2 by TLR7, mediated changes in macrophages at both transcriptional and epigenetic levels, which favored their hyperactivation by environmental stimuli. Together, these data indicate that the priming of macrophages can result from the response by pDCs to SARS-CoV-2, leading to macrophage activation in patients with severe COVID-19.

INTRODUCTION

The coronavirus disease 2019 (COVID-19) pandemic is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which has already infected hundreds of millions of people worldwide and is responsible for millions of deaths (1–3). Using single-cell profiling of bronchoalveolar lavage (BAL) fluids taken at different stages of the disease and from the lungs of recently deceased patients with COVID-19 (4–8), we and others have described the presence of a large set of proinflammatory cytokines produced by macrophages (4, 6, 8–13), so-called cytokine storm, although this term remains debated (14–16). These observations highlight a dysregulation of myeloid cells in COVID-19 with an accumulation of inflammatory macrophages that associates with disease severity (4, 12, 17–19). Given that only a small fraction (less than 10%) of lung macrophages are SARS-CoV-2 positive (8), how hyperactivation of macrophages occurs is still unclear, suggesting that these cells must receive additional signals when reaching the lungs.

The role of type I interferon (IFN-I) in protection from viral dissemination of SARS-CoV-2 is well documented, because patients with defects in IFN-I or IFN-III responses (12, 20, 21) or who have autoantibodies to IFN-I cytokines (22–24) are susceptible to SARS-CoV-2 infection and prone to progress to life-threatening COVID-19. Moreover, giving IFN-α in the early stage of COVID-19 is beneficial for infected patients (25). However, the role of IFN-I may be more complex (26, 27). Similarly to what has been observed with SARS-CoV-1 (28) and MERS (Middle East respiratory syndrome) (29), the blockade of IFN-I in humanized mice at the chronic stage of SARS-CoV-2 infection attenuates the inflammatory response by macrophages (30). Moreover, targeting the sustained IFN production in the late phase of SARS-CoV-2 infection in mice by blocking the cGAS-STING pathway reduces severe inflammation in the lung (31), and mice injected with IFN-α have increased lethality when infected with SARS-CoV-2 due to the induction of inflammatory cell death in macrophages (32). Furthermore, mice expressing human angiotensin-converting enzyme 2 (ACE2) infected with SARS-CoV-2 reveal an inflammatory role of IFN-I, leading to immune infiltration by cells such as monocyte-derived macrophages and T cells (33). Hence, mortality after SARS-CoV-1 infection is prevented in IFNAR-deficient mice or by using anti-IFNAR monoclonal antibodies (mAbs), without an increase in viral load (28). These data indicate that IFN-I is beneficial at the early stage of infection with SARS-CoV-1/2 but can also have a nefarious role in later stages of the disease. Understanding the source and kinetics of the IFN-I response to SARS-CoV-2 is thus needed and still unclear.

Plasmacytoid dendritic cells (pDCs) are a key cell type involved in antiviral responses because of their unparalleled ability to secrete IFN-I in response to Toll-like receptor 7 (TLR7) and TLR9 signaling (34, 35). pDCs produce IFN-α in response to SARS-CoV-1 and MERS-CoV (36, 37) and can protect mice in a model of mouse hepatitis virus (38). The depletion of pDCs in the murine model of SARS-CoV-1 protects the mice from lethal lung injury (28). pDCs produce high amounts of IFN-I in response to SARS-CoV-2, with a role for TLR7 (39–43). The number of circulating pDCs decreases in the blood of patients (12) and is inversely correlated with the disease severity (43). In contrast, pDCs infiltrate the lungs of patients with COVID-19, with their abundance evolving with the severity of the disease (4). Together, these data highlight a protective role of pDCs during the course of infection, although whether pDCs contribute to disease is not explored. Here, we observed that pDCs were a dominant producer of IFN-I after the sensing of SARS-CoV-2 by TLR7. We also observed that IFN-I produced by SARS-CoV-2–infected pDCs primed macrophages, leading to an inflammatory response with deleterious consequences for patients.

RESULTS

The cytokine storm produced by macrophages is associated with IFN-induced signaling in the lungs of patients with COVID-19

We first analyzed lung samples from patients with terminal COVID-19 (6, 8) and cells from BAL of patients with mild or severe forms of COVID-19 (4). In the BAL samples, we observed a large population of macrophages compared with other cell subsets (Fig. 1, A and B, and fig. S1, A and B), which segregated by disease status (Fig. 1B) but with comparable abundance between patients with mild or severe disease (Fig. 1C). In macrophages and myeloid dendritic cells (mDCs) from patients with mild disease, we observed a strong IFN-I response signature expression (Fig. 1, D and E; fig. S1C; and table S1), whereas the IFN-I response was reduced in macrophages from patients with severe COVID-19 (Fig. 1E and fig. S1C). Of note, the expression of fibrotic genes followed the same pattern as that of the IFN-I–inducible genes, with strong induction in patients with mild disease and reduction in patients with severe disease (Fig. 1, D and E; fig. S1C; and table S1). In contrast, an inflammatory gene signature associated with COVID-19 was very pronounced and sustained only in macrophages of patients with severe disease (Fig. 1, D and E; fig. S1C; and table S1). We analyzed the relationship between IFN-I and inflammatory responses in macrophages at the single-cell level and found that the two programs negatively correlated in controls but positively correlated in patients with COVID-19. Of note, macrophages from patients with severe disease showed the highest correlation (Fig. 1F), because the response in patients with mild disease was more dominated by the IFN-I response (Fig. 1F). This was confirmed by an orthogonal approach using differential gene expression between the three populations of subjects and gene set enrichment. The enriched terms reinforced our conclusion that IFN-I was the main driver of the disease (fig. S2, A to C).

Next, we analyzed the transcriptional profile of total lung cells obtained from recently deceased patients with severe COVID-19 (6, 8) for the presence of genes associated with IFN-I response, inflammation, or fibrosis (Fig. 1G and table S1). Similar to our findings in the BAL data, we observed in the macrophage cluster the presence of a strong IFN-I response and higher expression of genes associated with inflammation and fibrosis (Fig. 1G). These signatures were present in other subsets as well but to a lesser degree than in macrophages (Fig. 1G) and with higher expression in cells from infected patients as compared with control (fig. S3A and table S2), suggesting disease-dependent induction of expression. Of note, there was a strong correlation between the IFN-I response and the presence of profibrotic signals in macrophages (Fig. 1H), which highlights the potential link between IFN-I and the promotion of fibrosis, something that has been suggested in other fibrotic diseases (44). These correlations were present in more than one cellular subset, including other myeloid cells such as DCs or monocytes, but were the highest in the macrophage cluster (Fig. 1H). Similarly to our BAL data, macrophages in lungs had highly significant positive correlation between the presence of IFN-induced signaling and the expression of proinflammatory cytokines in macrophages (Fig. 1I). Hence, these data demonstrate that the inflammatory response by macrophages in the lungs of patients with COVID-19 is tightly associated with the presence of an IFN-I signature.

The presence of an IFN-I response could also be linked to other pathways, such as what we observed with fibrotic genes, and we investigated coagulation cascades and complement activation, which have been reported in patients with COVID-19 (45–47). We observed the presence of coagulation pathways in the lungs at the single-cell level (fig. S4A), with enrichment mostly in non-immune cell types, such as smooth muscle, fibroblasts, and epithelial cells, whereas complement pathways were also pronounced in immune cells, in particular, macrophages (fig. S4B). Hence, correlation analysis highlighted the relationship between IFN-I response and both coagulation and complement pathways in these cell types (fig. S4C). When we reanalyzed the single-cell BAL fluid from control, mild, and severe COVID patients, we also saw enrichment for an inflammation signature in several cell types but only see enrichment in the macrophage population (fig. S5, A and B). Pairwise correlation between IFN-I, complement, and coagulation pathways across single cells for macrophages was evaluated and revealed a relationship between IFN-I and coagulation. When the relationship between IFN and inflammation pathways was compared for all cell types, the enrichment for macrophages became even more clear (fig. S5, C and D). These data suggest the involvement of the coagulation and complement pathways in relation to IFN-I, in particular, in the macrophage subsets. Overall, these data demonstrate an intimate relationship between the presence of an IFN-I signature and key factors associated with the inflammatory status of lung macrophages, including a set of not only proinflammatory genes but also profibrotic, coagulation, and complement pathways.

The infiltration of pDCs in the lungs of SARS-CoV-2–infected patients coincides with the peak of IFN-I response

To understand the dynamics of the IFN-I response in the lungs of patients, we analyzed the BAL dataset and observed the infiltration of the BAL by pDCs in patients with mild disease (Fig. 2A). The number of pDCs was significantly reduced in patients with severe COVID-19 (Fig. 2A), consistent with the observation by Liao et al. (4). These pDCs expressed IFN-I–regulated genes (Fig. 2B), suggesting an activated phenotype, although we could not detect transcripts of IFN-I. These data indicated that in the lungs of SARS-CoV-2–infected individuals with mild disease, up to 1% of the BAL cells were activated pDCs.

We thus hypothesized that the heterogeneous information contained in the single-cell RNA sequencing (scRNA-seq) data can be used to infer a pseudo-temporal continuum of the molecular response to SARS-CoV-2 infection in macrophages. Patients at different stages of the disease could give an indication of the progression of the disease in SARS-CoV-2–infected patients. Using diffusion maps on the transcriptome data of macrophages, we inferred a joint representation and used as the first dimension the dynamics of disease progression (Fig. 2C). Along this pseudotime axis, we observed a high concentration of cells from control patients with low IFN-I and inflammation followed by cells from mild patients with a peak of IFN-I signature, and last, macrophages predominantly from severe patients with high IFN-I and inflammation signatures (Fig. 2D). This is similar to what has been described during influenza virus (Flu) infection, where macrophages respond to IFNs, are influenced by products of Flu infection, and also are a major producer of other cytokines, which can lead to severe Flu (48–50). This unsupervised, dynamic view of the macrophage response reinforces the idea of sequential stages of activation of the macrophages during COVID-19, associated with the early induction of IFN-I at the peak of pDC infiltration (Fig. 2A), which primes macrophages for hyperinflammatory activation in a subset of patients who develop severe disease (4).

The cellular composition and activation status of lung cells in recently deceased patients after SARS-CoV-2 infection have been described (6, 8), but the detailed contribution of pDCs or the role of IFN-I in the disease pathogenesis is unclear. By conducting subcluster analysis in the DC subpopulation in these total lung samples, we identified pDCs (fig. S3, B to G, and table S3), and in comparison with control lungs, the number of pDCs was significantly reduced (fig. S3, H and I, and table S4), although the remaining lung-infiltrating pDCs expressed genes of the IFN pathways as a sign of their activation status in the lungs of the patients (Fig. 2E and fig. S3, J and K). Next, we investigated subsets of macrophages and observed that although both resident alveolar macrophages and infiltrating monocyte-derived macrophages expressed a COVID-19–related inflammatory response, the IFN-I response and fibrotic response were mostly restricted to the monocyte-derived macrophage subset (Fig. 2, F and G). These data indicate that during the course of the disease, pDCs can infiltrate the lung of an infected patient and their presence correlated with a strong IFN-I response in lung macrophages.

pDCs sense SARS-CoV-2 via TLR7 and are the dominant IFN-I– and IFN-III–producing cells in PBMCs in response to the virus

Although IFN-I is critical for the clinical response to SARS-CoV-2 infection, the cellular source of IFN-I is not well defined. Purified pDCs had a robust IFN-α response to both live and ultraviolet-inactivated SARS-CoV-2 (Fig. 3A and fig. S6, A and B), consistent with earlier findings (40–42), and we observed that the virus was able to efficiently infect the pDCs (Fig. 3B) and replicate in pDCs (Fig. 3C). We then evaluated the relative contribution of pDCs to the overall IFN-I response to SARS-CoV-2 by peripheral blood mononuclear cells (PBMCs). Although total PBMCs could produce significant amounts of IFN-α when incubated with either live or inactivated virus, the production of IFN-α in response to live or inactivated SARS-CoV-2 by PBMCs depleted of the pDCs (pDC-depleted PBMCs; prepared by removing BDCA4-positive cells from PBMCs using microbeads) was negligible (Fig. 3D and fig. S6, C to E). This was in contrast to similar conditions using Flu, where the depletion of pDCs only partially reduced the overall IFN-α response (Fig. 3E and fig. S6F). Consistent with these data, the amount of replicating SARS-CoV-2 in pDC-depleted PBMCs was also negligible (Fig. 3, F and G). These data demonstrated that, in contrast to Flu, where not only pDCs but also other cells could produce IFN-α, pDCs are the dominant producers of IFN-α in PBMCs in response to SARS-CoV-2.

To better characterize the response by pDCs to SARS-CoV-2, we studied the kinetics of pDC activation by SARS-CoV-2. pDCs were incubated with inactivated SARS-CoV-2 or CpG, as a positive control, for 3, 6, 10, and 18 hours. Inactivated SARS-CoV-2 induced IFN-α and interleukin-6 (IL-6) and showed a progressive curve, which was higher at 18 hours than at the 6- to 10-hour time points, as for the TLR9 agonist CpG (Fig. 4, A and B). However, the nature of the response by pDCs to SARS-CoV-2 was similar to what is known for TLR7/9 signaling (51), with the induction of all subtypes of IFN-I (Fig. 4, C and D, and fig. S7A) and of a series of chemokines that may contribute to the migration of immune cells into the lungs of patients (fig. S7, B and C). The difference in the kinetics of the response by pDCs between inactivated SARS-CoV-2 and CpG is likely due to the delay in entry of the virus (fig. S7D).

Because SARS-CoV-2 is a single-stranded RNA virus, we explored the mechanism of activation of the pDCs by nucleic acid–sensing pathways. First, we excluded a contribution of ACE2, because these cells have little to no expression of ACE2 (42), and adding an ACE2 inhibitor had no effect on the IFN-α production by SARS-CoV-2–infected pDCs (fig. S8A). In contrast, blocking TLR7 (52) or phosphatidylinositol 3-kinase δ (PI3Kδ), which is key to TLR7-induced IFN-α in pDCs (53), led to inhibition of the IFN-α response (Fig. 4E and fig. S8, A and B). Consistent with our observation that pDCs were the dominant producer of IFN-α in the blood, we observed a complete inhibition of IFN-α production by the TLR7 inhibitor in PBMCs as well (Fig. 4F). Although it is documented that pDCs sense Flu via TLR7 (52), this is different when using Flu, because the inhibition of TLR7 in PBMCs only partially reduced the IFN-α response to Flu (fig. S8C) and had no effect when using pDC-depleted PBMCs (fig. S8D). This finding is consistent with the observation that pDCs from TLR7-deficient patients have a poor response to SARS-CoV-2 (41), although the loss of the overall response to SARS-CoV-2 by pDCs isolated from the TLR7-deficient patients was not complete (41). This may be due to redundancy developed by these cells with germline mutations of TLR7 to virus sensing, as was observed with IRAK4-deficient patients [see (54)]. It is also likely that other cell types that bear TLR7 may be involved in the response to SARS-CoV-2, in particular, in the tissue environment. Nucleic acid–sensing TLRs are located in endosomal compartments (34, 35, 55), and we observed that the entry in pDCs of inactivated SARS-CoV-2 and the subsequent induction of IFN-α required clathrin-mediated endocytosis (Fig. 4, G and H, and fig. S8E). These data thus demonstrated that SARS-CoV-2 could enter pDCs using clathrin-mediated entry and were sensed by TLR7, which signals pDCs to trigger IFN-I production.

Lung macrophages are not directly infected by SARS-CoV-2 but can uptake SARS-CoV-2 by phagocytosis of infected epithelial cells

We recently described that less than 10% of lung macrophages are infected with SARS-CoV-2 (8). However, macrophages produce some IFN-I due to the activation of the cGAS-STING pathway (31), because the deletion of STING in a mouse model of SARS-CoV-2 infection partially reduced IFN-I and ISGs (interferon-stimulated genes) expression. These authors did not observe that macrophages can be directly stimulated by SARS-CoV-2. Here, we observed that neither CD14+ monocytes, pluripotent stem cell (PSC)–derived macrophages, monocyte-derived macrophages, nor primary alveolar macrophages isolated from human lungs could directly be infected or stimulated by live SARS-CoV-2 (Fig. 5, A to E, and fig. S9, A and B), and inactivated SARS-CoV-2 induced little to no tumor necrosis factor (TNF) or IL-6 in macrophages (fig. S9, C and D). We observed a similar lack of response when alveolar macrophages were cultured with epithelial cells in the presence of live SARS-CoV-2 in a transwell system (Fig. 5, F to J). However, when cultured in combination with epithelial cells (Fig. 5K), genomic SARS-CoV-2 E (Fig. 5L) and SARS-CoV-2 subgenomic N (Fig. 5, M and N) were detected in the macrophages, which then expressed ISGs (Fig. 5O). This suggests that the source of viral RNA and proteins in macrophages is likely not direct infection but phagocytosis of infected cells. Combined with our in vitro data of live SARS-CoV-2–inoculated pDCs, these data support a scenario where IFN-I was coming from at least two different sources: pDCs by direct sensing of the live virus and macrophages by interacting with epithelial cells infected by the virus.

The production of IFN-I by pDCs in response to SARS-CoV-2 exacerbates macrophage responses to environmental stimuli

An increase in bacterial infections in patients with COVID-19, resulting in higher levels of bacterial products [bacterial DNA/RNA, lipoproteins, and lipopolysaccharide (LPS)] in intensive care unit (ICU) patients, has recently been reported (56). We therefore investigated whether the activation of macrophages by bacterial or environmental products could be influenced by IFN-I produced by pDCs activated by SARS-CoV-2. Hence, we incubated macrophages overnight with the supernatant of SARS-CoV-2–activated pDCs, and the cells were then cultured in the presence of various pathogen products (Fig. 6A). First, the supernatants of SARS-CoV-2–activated pDCs had little effect when used alone (Fig. 6, B to E). Similarly, supernatants from SARS-CoV-2–activated pDCs had little effect on the ability of the macrophages to respond to SARS-CoV-2 itself (fig. S9, E and F). However, these supernatants drastically amplified the production and expression of proinflammatory cytokines, such as TNF and IL-6, by macrophages in response to not only LPS (Fig. 6B and fig. S10A) but also Pam3Cys (an agonist of another transmembrane TLR) (Fig. 6C and fig. S10B), poly I:C (polyinosinic:polycytidylic acid) (Fig. 6D and fig. S10C), and the TLR8 agonist ORN8L (both RNA-sensing TLRs, which are endosomal) (Fig. 6E and fig. S10D). As control, we used the supernatant of pDCs that were left unstimulated, which had little to no effect on macrophage responses to these stimuli (Fig. 6, B to E). CXCL10 was induced in macrophages by SARS-CoV-2–infected pDCs, likely due to the presence of IFN-I in the pDC supernatants, and the addition of LPS had little effect (fig. S10E). However, live virus in the context of the inflammatory milieu of the lung may enhance the sensing and inflammatory response by macrophages. In addition, although lung macrophages are not abundantly infected by SARS-CoV-2 (8), we showed that macrophages could be activated by the phagocytosis of infected cells (Fig. 5O) and could also be activated via cell-to-cell transfer of viruses by pDCs as reported in the context of HIV (57). pDCs could respond differently to free particles versus infected cells as previously suggested (40, 58).

Because pDCs produced large amounts of IFN-I in response to SARS-CoV-2 (Fig. 3A), we incubated macrophages with titrated amounts of IFN-α followed by LPS. We observed a dose-dependent effect of IFN-α on the LPS-induced response in macrophages with induction of TNF, IL6, IL1B, IL12B, and IFNB after treatment with high concentrations of IFN-α, comparable to that observed with supernatants from SARS-CoV-2–infected pDCs (Fig. 7A and fig. S11). We also explored the effect of TNF, which could have a synergistic or antagonistic effect on IFN-α. However, blocking TNF or the TNF receptor had no significant influence on macrophages primed with supernatants from SARS-CoV-2–infected pDCs in response to LPS (fig. S12), although it is possible that TNF may play a role in the context of inflamed lung, where TNF is abundantly secreted by the macrophages. Baricitinib inhibits Janus kinase 1/2 (JAK1/2), which is essential, although not restricted, to IFNAR signaling (59), and early evaluation in combination with remdesivir in hospitalized patients with COVID-19 yielded promising results, with significantly reduced mortality (60). In the absence of LPS, baricitinib prevented the induction of not only CXCL10 but also TNF and IL-6 secretion in macrophages cultured with the supernatant of SARS-CoV-2–infected pDCs (fig. S13, A and B). Inhibiting JAK1/2 prevented the induction of TNF and IL-6 by LPS (Fig. 7B and fig. S14A). Blocking IFNAR had a similar effect (Fig. 7C and fig. S14B), suggesting that the main effect was due to IFN-I, although it is also possible that other factors secreted by pDCs, beyond IFN-I or IFN-III, may contribute at some level to the modulation of macrophage response. A similar observation was made using Pam3Cys (Fig. 7D and fig. S14C), poly I:C (Fig. 7E and fig. S14D), or ORN8L (Fig. 7F and fig. S14E) in macrophages exposed to SARS-CoV-2–infected pDC supernatants (Fig. 7, B to F, and fig. S14, A to D). We conclude that macrophages produced exacerbated amounts of proinflammatory cytokines in response to environmental stimuli when exposed to IFN-I from SARS-CoV-2–infected pDCs.

The production of IFN-I by pDCs in response to SARS-CoV-2 mediates epigenetic and transcriptional changes in macrophages

To obtain a comprehensive understanding of how IFN-I and SARS-CoV-2–infected pDC supernatants influenced macrophage activation, we performed transcriptomic analysis using RNA-seq to evaluate TLR4 responses in macrophages exposed to IFN-α or SARS-CoV-2–infected pDC supernatants. Principal components analysis (PCA) showed that pDC supernatants and IFN-α conditions closely clustered together (Fig. 8A); IFN-α– and SARS-CoV-2–infected pDC supernatant–induced genes [differentially expressed genes (DEGs), false discovery rate (FDR) < 0.05 and fold induction > 2] highly overlapped with commonly induced genes (Fig. 8, B and C). Similarly, macrophages preincubated with either IFN-α or SARS-CoV-2–infected pDC supernatants and stimulated with LPS tightly clustered together, separated from the LPS-alone condition (Fig. 8A); 92% of DEGs induced by LPS in IFN-α– or SARS-CoV-2–infected pDC supernatant–treated macrophages were common (Fig. 8C). The data showed that both LPS and IFN-α/SARS-CoV-2–infected pDC supernatants contributed to the changes in gene expression (Fig. 8C). Several chemokines are present in the lungs of SARS-CoV-2–infected patients (61). We observed that IFN-α or supernatants from SARS-CoV-2–activated pDCs up-regulated a series of chemokine receptors, including CCR2, CCR1, CCR5, and CXCR2 (fig. S15A). These recognize CCL2, CCL5, CCL8, and CXCL8, which are induced by SARS-CoV-2 or IFN-α (fig. S7, B and C), suggesting a role for IFN-α in promoting the infiltration of macrophages to the lungs of patients. We conducted a K-means clustering based on genes with more than a twofold change in expression, which segregated DEGs into seven groups based on patterns of expression (Fig. 8B). The resulting clusters faithfully followed the culture conditions, except for the IFN-α + LPS and SARS-CoV-2–infected pDC supernatant + LPS, which, as expected, yielded similar gene patterns (see top part of Fig. 8B). We conducted pathway analysis of each cluster that segregated the clusters with a strong inflammatory bias (see fig. S15B). Clusters 3 and 4 were dominated by genes induced by IFN-I, whereas clusters 6 and 7 were dominated by genes down-regulated by LPS. Cluster 1 was composed of LPS-inducible genes whose expression was exacerbated by either IFN-α or SARS-CoV-2–infected pDC supernatants (similar to TNF and IL6). A more in-depth analysis indicated that cluster 1 included many genes encoding proinflammatory cytokines and chemokines (Fig. 8D and fig. S15B) and fibrosis-related genes (fig. S15C), which are implicated in COVID-19 pathogenesis (6, 26, 27). Additional bioinformatic analysis comparing macrophages stimulated with SARS-CoV-2–infected pDC supernatants versus LPS alone revealed enrichment for inflammatory and immune pathways (Fig. 8E and fig. S15D). Ingenuity Pathway Analysis suggested a role for IFN regulatory factor (IRF) and nuclear factor κB (NF-κB) family transcription factors in the activation of inflammatory genes by IFN-α or SARS-CoV-2–infected pDC supernatants with LPS (fig. S15E), which is in line with our previous report (62). Clusters 3 and 4 showed enrichment of IFN signaling (fig. S15B) and were composed of canonical ISGs (fig. S15F). Overall, these results support a role for pDC-derived IFN-I in the exacerbation of TLR4-mediated inflammatory response by macrophages. Accordingly, inhibition of IFN-I production using either the TLR7 inhibitor IRS661 or the PI3Kδ inhibitor CAL-101 reduced inflammatory gene induction upon challenge with LPS (Fig. 8F and fig. S16A). IFN-α had minimal effects on TLR4-induced IκBα degradation or activation of mitogen-activated protein kinases (MAPKs) ERK (extracellular signal–regulated kinase) and p38 (fig. S16B), which is in accord with previous work (62). However, cells primed with IFN-α followed by LPS treatment significantly increased chromatin accessibility of IL6 and TNF promoters (Fig. 8G). Hence, these data demonstrate that the IFN-I produced by pDCs in response to SARS-CoV-2 mediates transcriptional and epigenetic changes in macrophages, which exacerbates their production of inflammatory mediators in response to environmental triggers.

DISCUSSION

The immune response to SARS-CoV-2 evolves over time in patients with COVID-19 (3). Here, we showed an unexpected role of pDCs during the course of the disease. pDCs were the main producers of IFN-I in the blood and directly sensed SARS-CoV-2 via TLR7. We also observed that macrophages could produce IFN-I due to the phagocytosis of SARS-CoV-2–infected epithelial cells. The entry of SARS-CoV-2 in pDCs required clathrin-mediated endocytosis, and the IFN-I produced by pDCs markedly exacerbated the response of macrophages to various innate stimuli. In the lungs of patients, a peak of pDC infiltration was associated with an initial wave of IFN-I in the macrophages of patients that did not produce a cytokine storm. In contrast, in patients with severe or even fatal COVID-19, IFN-I and proinflammatory signals were both present in the lung-infiltrating macrophages. Our data thus support a model (see fig. S17) by which the IFN-I produced by pDCs due to the direct sensing of SARS-CoV-2 could epigenetically prime lung macrophages to induce a cytokine storm. This model is consistent with the recent findings that macrophages can induce IFN-α via the cGAS-STING pathway but only in response to SARS-CoV-2–infected cells (31). Another model is that an initial defect in pDC response can favor the infection of epithelial cells, which, in turn, can activate the macrophages. This is supported by data in IFN- or TLR7-defective patients (21, 41) and by some studies linking the IFN-I produced by macrophages to disease (31, 63). The macrophage priming could also be done by the presence of IFN-γ, which, by complementing by TLR signaling, promotes caspase-8 pathways leading to cell death and increased severity of SARS-CoV-2 disease (64). Together, this would indicate a redundancy in the response to SARS-CoV-2 by different cell types stimulated via different nucleic acid sensors and the contribution by both IFN-I and IFN-II.

It is difficult to assess the precise amount of IFN-I in the lungs of patients, but pDCs are the highest blood producers at the per-cell level of IFN-α in response to viruses (34, 35, 65), which may create a local environment where very high concentrations of IFN-I are present, even with a limited number of cells (66). Low concentrations of IFN-I (200 antiviral units/ml) induced antiviral ISG expression (“IFN signature” and, by inference, an antiviral response) and did not augment subsequent TLR responses, whereas higher concentrations of IFN-α prevented TNF-induced tolerance of a subset of TLR4-inducible genes (62). Here, extraordinarily high concentrations of IFN-I produced by SARS-CoV-2–infected pDCs reprogrammed macrophages for an augmented hyperinflammatory TLR response. In other words, low concentrations of IFN-I induce antiviral responses without inflammatory toxicity, but high amounts of IFN-I produced by pDCs additionally promote cytokine storm. Because the precise pathogenesis of COVID-19 is still unresolved, it is likely to vary between patients with respect to the source of IFN-I in the lungs. However, our data indicate that pDCs sense SARS-CoV-2 through TLR7, and they were the dominant producer of IFN-I in the blood, a property not shared with other viruses, such as Flu. Using a controlled setting in vitro, we demonstrated that macrophages exposed to SARS-CoV-2–infected pDCs had an exacerbated response to multiple stimuli, directly linking the IFN-I produced by SARS-CoV-2–infected pDCs and the observed hyperactivation of macrophages in patients with COVID-19. Because we showed that the IFN-I response was concentrated in monocyte-derived macrophages in the lungs of patients, our data suggest that, in addition to being primed in the lungs, some macrophages may be primed in the blood before the cells enter the lungs of infected individuals. After IFN priming, exacerbation of activation was observed after stimulation by different TLR ligands. Increased bacterial infections in ICU patients (56) and increased presence of intestinal mucosal damage in patients with COVID-19 (67, 68) may be responsible for the presence of LPS and other TLR ligands and thus induce hyperactivation. It seems that there is no clear correlation between viral loads (often quantified in the blood) and the inflammatory response in the lung [see the review (69)].

Our study has some limitations in part due to the difficulties in working with samples from infected patients. The lung samples were taken from recently deceased individuals with severe diseases, which may explain the low abundance of pDCs (70), consistent with the BAL data (4). These patients are at different time points in their COVID-19 illness, and samples can only be obtained at the time of death. A proper longitudinal time course would be ideal to address the question of pDC numbers over time but could not be done here. Furthermore, we used chemical inhibitors to study the role of TLR7 and the clathrin-mediated pathways, which brings some limitations to our system, including potential off-target effects and redundancy in the pathways.

Our findings have potential translational significance, and the physiological relevance of the pDC-macrophage circuit is supported by the efficacy of the JAK inhibitor baricitinib in decreasing mortality in patients with COVID-19 (60), likely by attenuating the cytokine storm. Our data describe a strong fibrotic signature particularly in BAL macrophages of patients with mild disease, which correlated with IFN-I response. This is consistent with the observation that macrophages have a profibrotic phenotype in patients with COVID-19 (71). Because pDCs can promote fibrosis in systemic sclerosis, it is possible that the pDC-macrophage interaction contributes to this phenotype in COVID-19 as well. In IFN-primed macrophages, JAK inhibitors suppress the expression of canonical inflammatory target genes such as TNF and IL-6, whereas in control macrophages, JAK inhibitors only suppress ISG expression (66). Thus, we propose that one mechanism of action of baricitinib in COVID-19 is direct suppression of the inflammatory cytokine genes that drive the cytokine storm. It remains possible that additional factors produced by pDCs, or direct pDC-macrophage interactions, contribute to the complex macrophage phenotype in COVID-19. Furthermore, SARS-CoV-2 activation of inflammasomes is associated with COVID-19 severity, and viral components produce a variegated response with early effects on inflammasome priming and later changes resulting in inflammasome activation and the production of inflammatory factors (72, 73). In addition, the infection by SARS-CoV-2 triggers the inflammasome in macrophages, leading to IL-1β and IL-18 release in the lungs and contributing to pulmonary inflammation (30). These findings suggest that inhibition of the pDC-macrophage circuit could be beneficial for patients with COVID-19.

When pDCs are chronically activated, they produce sustained levels of IFN-I that lead to negative consequences, such as CD4 T cell depletion in HIV-infected patients or the promotion of autoimmunity by promoting skin lesions (34, 35, 74, 75). Our data now suggest that a similar concept may apply during COVID-19, where the chronic activation of pDCs and the IFN-I that they produce can prime uninfected macrophages to produce a cytokine storm in the lungs of patients with COVID-19. Although we are showing that pDCs could be activated by direct sensing of SARS-CoV-2, in the lungs, the cells could also be activated by virus-infected cells (58). The sensing of infected cells by pDCs could lead to higher levels of IFN, which could compensate for the low number of pDCs present in the lungs of patients with severe disease. In addition, blood pDCs from patients with severe COVID-19 have a reduced response to SARS-CoV-2 (76), which is reminiscent of HIV, where the number of pDCs is reduced (77), and of autoimmune indications such as lupus or systemic sclerosis (34) and is an indication that the activated cells may have migrated to the tissue.

Although the cytokines storm produced in the lungs by macrophages is responsible for respiratory distress and poor outcome in SARS-CoV-2–infected patients, the underlying mechanism leading to the activation of macrophages is not well defined. Furthermore, whether IFN-I contributes to the inflammatory response in the lung is also unclear. Our data describe that pDCs and the IFN-I that they produce in response to SARS-CoV-2 can exacerbate the response of macrophages to innate stimuli that these cells can encounter in the lungs. This identifies the pDC-macrophage circuit as critical for the pathogenesis of COVID-19 and identifies the TLR7 sensing of SARS-CoV-2 by pDCs as a central element of the immune response to this virus. Hence, pDCs and TLR7 are needed to provide an adequate response to control the virus, but the IFN-I induced in pDCs can also exacerbate macrophage response, with deleterious consequences. These data also have translational potential because they identify pDCs and the IFN-I pathway as potential therapeutic targets for critically ill patients with COVID-19.

As Omicron rages on, scientists have no idea what comes next

SCIENCE Magazine: INSIDER HEALTH

A rapid succession of subvariants is the new normal—but a completely new variant could still emerge

19 JUL 2022, 4:00 PM BY KAI KUPFERSCHMIDT

A nurse prepares a COVID-19 vaccine in Guwahati, India, on 10 April. A new subvariant named BA.2.75 that was first detected in India has surfaced in many other countries. ANUPAM NATH/AP IMAGES

In the short history of the COVID-19 pandemic, 2021 was the year of the new variants. Alpha, Beta, Gamma, and Delta each had a couple of months in the Sun.

But this was the year of Omicron, which swept the globe late in 2021 and has continued to dominate, with subvariants—given more prosaic names such as BA.1, BA.2, and BA.2.12.1—appearing in rapid succession. Two closely related subvariants named BA.4 and BA.5 are now driving infections around the world, but new candidates, including one named BA.2.75, are knocking on the door.

Omicron’s lasting dominance has evolutionary biologists wondering what comes next. Some think it’s a sign that SARS-CoV-2’s initial frenzy of evolution is over and it, like other coronaviruses that have been with humanity much longer, is settling into a pattern of gradual evolution. “I think a good guess is that either BA.2 or BA.5 will spawn additional descendants with more mutations and that one or more of those subvariants will spread and will be the next thing,” says Jesse Bloom, an evolutionary biologist at the Fred Hutchinson Cancer Research Center.

But others believe a new variant different enough from Omicron and all other variants to deserve the next Greek letter designation, Pi, may already be developing, perhaps in a chronically infected patient. And even if Omicron is not replaced, its dominance is no cause for complacency, says Maria Van Kerkhove, technical lead for COVID-19 at the World Health Organization. “It’s bad enough as it is,” she says. “If we can’t get people to act [without] a new Greek name, that’s a problem.”

Even with Omicron, Van Kerkhove emphasizes, the world may face continuing waves of disease as immunity wanes and fresh subvariants arise. She is also alarmed that the surveillance efforts that allowed researchers to spot Omicron and other new variants early on are scaling back or winding down. “Those systems are being dismantled, they are being defunded, people are being fired,” she says.

The variants that ruled in 2021 did not arise one out of the other. Instead, they evolved in parallel from SARS-CoV-2 viruses circulating early in the pandemic. In the viral family trees researchers draw to visualize the evolutionary relationships of SARS-CoV-2 viruses, these variants appeared at the tips of long, bare branches. The pattern seems to reflect virus lurking in a single person for a long time and evolving before it emerges and spreads again, much changed.

More and more studies seem to confirm that this occurs in immunocompromised people who can’t clear the virus and have long-running infections. On 2 July, for example, Yale University genomic epidemiologist Nathan Grubaugh and his team posted a preprint on medRxiv about one such patient they found accidentally. In the summer of 2021, their surveillance program at the Yale New Haven Hospital kept finding a variant of SARS-CoV-2 called B.1.517 even though that lineage was supposed to have disappeared from the community long ago. All of the samples, it turned out, came from the same person, an immunocompromised patient in his 60s undergoing treatment for a B cell lymphoma. He was infected with B.1.517 in November 2020 and is still positive today.

By following his infection to observe how the virus changed over time, the team found it evolved at twice the normal speed of SARS-CoV-2. (Some of the viruses circulating in the patient today might be qualified as new variants if they were found in the community, Grubaugh says.) That supports the hypothesis that chronic infections could drive the “unpredictable emergence” of new variants, the researchers write in their preprint.

Other viruses that chronically infect patients also change faster within one host than when they spread from one person to the next, says Aris Katzourakis, an evolutionary biologist at the University of Oxford. This is partly a numbers game: There are millions of viruses replicating in an individual, but only a handful are passed on during transmission. So a lot of potential evolution is lost in a chain of infections, whereas a chronic infection allows for endless opportunities to evolve.

But since Omicron emerged in November 2021, no new variants have appeared out of nowhere. Instead, Omicron has accumulated small changes, making it better at evading immune responses and—together with waning immunity—leading to successive waves. “I think it’s probably harder and harder for these new things to emerge and take over because all the different Omicron lineages are stiff competition,” Grubaugh says, given how transmissible and immune-evading they already are.

If so, the U.S. decision to update COVID-19 vaccines by adding an Omicron component is the right move, Bloom says; even if Omicron keeps changing, a vaccine based on it is likely to provide more protection than one based on earlier variants.

But it’s still possible that an entirely new variant unrelated to Omicron will emerge. Or one of the previous variants, such as Alpha or Delta, could make a comeback after causing a chronic infection and going through a bout of accelerated evolution, says Tom Peacock, a virologist at Imperial College London: “This is what we would call second-generation variants.” Given those possibilities, “Studying chronic infections is now more important than ever,” says Ravindra Gupta, a microbiologist at the University of Cambridge. “They might tell us the kind of mutational direction the virus will take in the population.”

BA.2.75, which was picked up recently, already has some scientists concerned. Nicknamed Centaurus, it evolved from Omicron but seems to have quickly accumulated a whole slew of important changes in its genome, more like an entirely new variant than a new Omicron subvariant. “This looks exactly like Alpha did, or Gamma or Beta,” Peacock says.

BA.2.75 appears to be spreading in India, where it was first identified, and has been found in many other countries. Whether it’s really outcompeting other subvariants is unclear, Van Kerkhove says: “The data is superlimited right now.” “I certainly think it’s something worth keeping a close eye on,” says Emma Hodcroft, a virologist at the University of Bern.

Keeping an eye on anything is getting harder, however, because surveillance is decreasing. Switzerland, for example, now sequences about 500 samples per week, down from 2000 at its peak, Hodcroft says; the United States went from more than 60,000 per week in January to about 10,000. “Some governments are anxious to cut back on the money they dedicated to sequencing,” Hodcroft says. Defending the expense is a “hard sell,” she says, “especially if there’s a feeling the countries around you will continue sequencing even if you stop.”

Even if a variant emerges in a place with good surveillance, it may be harder than in the past to predict how big a threat it poses, because differences in past COVID-19 waves, vaccines, and immunization schedules have created a global checkerboard of immunity. That means a new variant might do well in one place but run into a wall of immunity elsewhere. “The situation has become even less predictable,” Katzourakis says.

Given that Omicron appears to be milder than previous variants, surveillance efforts should aim to identify variants that cause severe disease in hospitalized patients, Gupta says. “I think that that’s where we should be focusing our efforts, because if we keep focusing on new variants genomically, we may get a bit fatigued, and then kind of drop the ball when things do happen.”

Many virologists acknowledge that SARS-CoV-2’s evolution has caught them by surprise again and again. “It was really in part a failure of imagination,” Grubaugh says. But whatever scenario researchers can imagine, Bloom acknowledges the virus will chart its own course: “I think in the end, we just kind of have to wait and see what happens.” 

About the Author: Kai is a contributing correspondent for Science magazine based in Berlin, Germany. He is the author of a book about the color blue, published in 2019.

doi: 10.1126/science.ade0197

The molecular epidemiology of multiple zoonotic origins of SARS-CoV-2

Authors Info & Affiliations

JONATHAN E. PEKAR and his colleagues have published their work today on the genetics and nucleotide reversions of Lineage A and Lineage B strains of the coronavirus from the Wuhan market in late 2019.

SCIENCE, 26 Jul 2022 First Release

DOI: 10.1126/science.abp8337

Figure 1. Maximum likelihood phylogeny of the early SARS-CoV-2 pandemic, showing nucleotide reversions and putative candidates for the ancestral haplotype at the most common recent ancestor (MRCA).

Putative ancestral haplotypes are identified with colored shapes. Reversions from the Hu-1 reference genotype to the recCA are colored. Blue represents C-to-T reversions and black indicates all other reversions. The tree is rooted on Hu-1 to show reversion dynamics to the recCA.

Abstract

Understanding the circumstances that lead to pandemics is important for their prevention. Here, we analyze the genomic diversity of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) early in the coronavirus disease 2019 (COVID-19) pandemic. We show that SARS-CoV-2 genomic diversity before February 2020 likely comprised only two distinct viral lineages, denoted A and B. Phylodynamic rooting methods, coupled with epidemic simulations, reveal that these lineages were the result of at least two separate cross-species transmission events into humans. The first zoonotic transmission likely involved lineage B viruses around 18 November 2019 (23 October–8 December), while the separate introduction of lineage A likely occurred within weeks of this event. These findings indicate that it is unlikely that SARS-CoV-2 circulated widely in humans prior to November 2019 and define the narrow window between when SARS-CoV-2 first jumped into humans and when the first cases of COVID-19 were reported. As with other coronaviruses, SARS-CoV-2 emergence likely resulted from multiple zoonotic events.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for the coronavirus disease 19 (COVID-19) pandemic that caused more than 5 million confirmed deaths in the two years following its detection at the Huanan Seafood Wholesale Market (hereafter the ‘Huanan market’) in December 2019 in Wuhan, China (1–3). As the original outbreak spread to other countries, the diversity of SARS-CoV-2 quickly increased and led to the emergence of multiple variants of concern, but the beginning of the pandemic was marked by two major lineages denoted ‘A’ and ‘B’ (4).

Lineage B has been the most common throughout the pandemic and includes all eleven sequenced genomes from humans directly associated with the Huanan market, including the earliest sampled genome, Wuhan/IPBCAMS-WH-01/2019, and the reference genome, Wuhan/Hu-1/2019 (hereafter ‘Hu-1’) (5), sampled on 24 and 26 December 2019, respectively. The earliest lineage A viruses, Wuhan/IME-WH01/2019 and Wuhan/WH04/2020, were sampled on 30 December 2019 and 5 January 2020, respectively (6). Lineage A differs from lineage B by two nucleotide substitutions, C8782T and T28144C, which are also found in related coronaviruses from Rhinolophus bats (4), the presumed host reservoir (7). Lineage B viruses have a ‘C/T’ pattern at these key sites (C8782, T28144), whereas lineage A viruses have a ‘T/C’ pattern (C8782T, T28144C). The earliest lineage A genomes from humans lack a direct epidemiological connection to the Huanan market, but were sampled from individuals who lived or had recently stayed close to the market (8). It has been hypothesized that lineages A and B emerged separately (9), but ‘C/C’ and ‘T/T’ genomes intermediate to lineages A and B present a challenge to that hypothesis, as their existence suggests within-human evolution of one lineage toward the other via a transitional form.

Questions about these lineages remain: if lineage B viruses are more distantly related to sarbecoviruses from Rhinolophus bats, (i) why were lineage B viruses detected earlier than lineage A viruses and (ii) why did lineage B predominate early in the pandemic?

Answering these questions requires determining the ancestral haplotype, the genomic sequence characteristics of the most recent common ancestor (MRCA) at the root of the SARS-CoV-2 phylogeny. In this study, we combined genomic and epidemiological data from early in the COVID-19 pandemic with phylodynamic models and epidemic simulations. We eliminated many of the haplotypes previously suggested as the MRCA of SARS-CoV-2 and show that the pandemic most likely began with at least two separate zoonotic transmissions starting in November 2019.

Results

Erroneous assignment of haplotypes intermediate to lineages A and B

There are 787 near-full length genomes available from lineages A and B sampled by 14 February 2020 (data S1 and S2). However, there are also 20 genomes of intermediate haplotypes from this period containing either T28144C or C8782T but not both mutations: C/C or T/T, respectively.

We identified numerous instances of C/C and T/T genomes sharing rare mutations with lineage A or lineage B viruses, often sequenced in the same laboratory, indicating these intermediate genomes are likely artifacts of contamination or bioinformatics (10), similar to findings from our analysis of the emergence of SARS-CoV-2 in North America (11) (fig. S1 and supplementary text). We confirmed that a C/C genome from South Korea sharing three such mutations had low sequencing depth at position 28144 (≤10x), a T/T genome sampled in Singapore had low coverage at both 8782 and 28144 (≤10x), and three T/T genomes sampled in Wuhan had low sequencing depth and indeterminate nucleotide assignment at position 8782 (table S1). Further, the authors of eleven C/C genomes sampled in Wuhan and Sichuan confirmed that low sequencing depth at position 8782 led to the erroneous assignment of intermediate haplotypes.

C/C and T/T genomes continue to be observed throughout the pandemic as a result of convergent evolution, including T/T aboard the Diamond Princess cruise ship outbreak and subsequent COVID-19 waves in New York City and San Diego (fig. S2 to S5 and supplementary text). Instances of convergent evolution are identifiable because SARS-CoV-2 phylogenies exist in ‘near-perfect’ tree space where topology can be inferred with high accuracy (12). These findings cast doubt on the claim that transitional C/C or T/T haplotypes between lineages A and B circulated in humans, reopening the door to the hypothesis that lineages A and B represent separate zoonotic introductions.

Progenitor genome reconstruction

To better understand SARS-CoV-2 mutational patterns, we reconstructed the genome of a hypothetical progenitor of SARS-CoV-2. Using maximum likelihood ancestral state reconstruction across 15 non-recombinant regions of SARS-CoV-2 and closely related sarbecovirus genomes sampled from bats and pangolins (13), we inferred the genome of this recombinant common ancestor (“recCA”) (figs. S6 and S7 and supplementary text). The recCA differed from Hu-1 by just 381 substitutions, including C8782T and T28144C. It is more informative than an outgroup sarbecovirus because it accounts for the closest relative across all recombinant segments (figs. S8 to S14 and supplementary text) (14), and, as an internal node on the phylogeny, is more genetically similar to SARS-CoV-2 than any extant sarbecovirus.

Reversions across the early pandemic phylogeny

The ubiquity of SARS-CoV-2 reversions (i.e., mutations from Hu-1 toward the recCA) indicates that genetic similarity to related viruses is a poor proxy for the ancestral haplotype. We observe 23 unique reversions and 631 unique substitutions (excluding reversions) across the SARS-CoV-2 phylogeny from the COVID-19 pandemic up to 14 February 2020 (Fig. 1). Substitutions were overrepresented at the 381 sites separating the recCA from Hu-1 (23/381 = 6.04%), compared with substitutions at all other sites (631/29,134 = 2.17%).

Most reversions were C-to-T mutations (19/23 = 82.6%), matching the mutational bias of SARS-CoV-2 (15–17). Genomes with C-to-T reversions can be found within lineage A, including C18060T (lineage A.1; e.g., WA1) and C29095T (e.g., 20SF012), as well as C24023T, C25000T, C4276T, and C22747T in mid-late January and February 2020. Hence, triple revertant genomes, like WA1 and 20SF012, are neither unique nor rare. We also identified a lineage A genome (Malaysia/MKAK-CL-2020-6430/2020), sampled on 4 February 2020 from a Malaysian citizen traveling from Wuhan whose only four mutations from Hu-1 are all reversions (lineage A.1+T6025C) (Fig. 1). Therefore, no highly revertant haplotype can automatically be assumed to represent the MRCA of SARS-CoV-2, especially when these reversions are most often the result of C-to-T mutations. In fact, we continue to observe these reversion patterns throughout the pandemic, including in the emergence of WHO-named variants (figs. S15 and S16).

Inferring the MRCA of SARS-CoV-2

To infer the ancestral SARS-CoV-2 haplotype, we developed a non-reversible, random-effects substitution process model in a Bayesian phylodynamic framework that simultaneously reconstructs the underlying coalescent processes and the sequence of the MRCA of the SARS-CoV-2 phylogeny. The random-effects substitution model captures the C-to-T transition and G-to-T transversion biases (fig. S17 and supplementary text). Using this model, referred to as the unconstrained rooting (fig. S18A), we inferred the ancestral haplotype of the 787 lineage A and B genomes sampled by 14 February 2020.

Our unconstrained rooting strongly favors a lineage B or C/C ancestral haplotype and shows that a lineage A ancestral haplotype is inconsistent with the molecular clock [Bayes factor (BF) = 48.1] (Table 1). Lineage B exhibits more divergence from the root of the tree than would be expected if lineage A were the ancestral virus in humans (figs. S19 and S20). The T/T ancestral haplotype was also disfavored (BF>10), likely because of the C-to-T transition bias (fig. S17). We acknowledge that the timing of the earliest sampled lineage B genomes associated with the Huanan market could bias rooting inference toward lineage B haplotypes; however, lineage A was still disfavored after excluding all market-associated genomes (BF=11.0).

Even though sequence similarity to closely related sarbecoviruses alone is insufficient to determine the SARS-CoV-2 ancestral haplotype, this similarity can inform phylodynamic inference. Rather than rely on outgroup rooting [fig. S18B and (18)], we developed a rooting method that assigns the recCA as the progenitor of the inferred SARS-CoV-2 MRCA (fig. S18C). As opposed to the unconstrained rooting, the recCA root favored a lineage A haplotype over lineage B, although support for C/C was unchanged (Table 1). Our results were insensitive to the method of breakpoint identification in the recCA (supplementary text).

The A.1 and A+C29095T proposed ancestral haplotypes were strongly rejected by all the phylodynamic analyses, even when rooting with recCA or bat sarbecovirus outgroups, which include both C18060T and C29095T (Table 1 and data S3). Hence, WA1-like and 20SF012-like haplotypes cannot plausibly represent the MRCA of SARS-CoV-2 as previously suggested (19–21): the similarity of these genomes to the recCA is due to C-to-T reversions. Haplotypes not reported in Table 1 were similarly rejected (data S3).

We inferred the tMRCA for SARS-CoV-2 to be 11 December 2019 (95% HPD: 25 November–12 December) using unconstrained rooting. It has been suggested that a phylogenetic root in lineage A would produce an older time of most recent common ancestor (tMRCA) than a lineage B rooting (21). Therefore, we developed an approach to assign a haplotype as the SARS-CoV-2 MRCA and inferred the tMRCA (i.e., A, B, C/C, A.1 or A+C29095T) (fig. S18D). The tMRCA was consistent with the recCA-rooted and fixed ancestral haplotype analyses (table S2 and supplementary text).

We infer only three plausible ancestral haplotypes: lineage A, lineage B, and C/C. However, the inability to reconcile the molecular clock at the outset of the COVID-19 pandemic with a lineage A ancestor without information from related sarbecoviruses (e.g., the recCA) requires us to question the assumption that both lineages A and B resulted from a single introduction.

Separate introductions of lineages A and B

We next sought to determine whether a single introduction from one of the plausible ancestral haplotypes (lineage A, lineage B, or C/C) is consistent with the SARS-CoV-2 phylogeny. We simulated SARS-CoV-2-like epidemics (22, 23) with a doubling time of 3.47 days [95% highest density interval (HDI) across simulations: 1.35–5.44] (24–26) to account for the rapid spread of SARS-CoV-2 before it was identified as the etiological agent of COVID-19 (figs. S21 and S22, tables S3 and S4, and supplementary text). We then simulated coalescent processes and viral genome evolution across these epidemics to determine how frequently we recapitulated the observed SARS-CoV-2 phylogeny.

Lineages A and B comprise 35.2% and 64.8% of the early SARS-CoV-2 genomes, and each lineage is characterized by a large polytomy (i.e., many sampled lineages descending from a single node on the phylogenetic tree), with the base of lineages A and B being the two largest polytomies observed in the early pandemic (Fig. 1). Furthermore, large polytomies are characteristic of SARS-CoV-2 introductions into geographical regions at the start of the pandemic (e.g., fig. S23) (11, 27–29) and would similarly be expected to occur after a successful introduction of SARS-CoV-2 into humans. Congruently, the most common topology in our simulations is a large basal polytomy (with ≥100 descendent lineages), present in 47.5% of simulated epidemics (Fig. 2A).
In contrast, a topology corresponding to a single introduction of an ancestral C/C haplotype, characterized by two clades, each comprising ≥30% of the taxa, possessing a large polytomy at the base, and separated from the MRCA by one mutation (Fig. 2B), was only observed in 0.1% of our simulations. Further, a topology corresponding to a single introduction of an ancestral lineage A or lineage B haplotype, characterized by a large basal polytomy and a large clade, comprising between 30% and 70% of taxa, two mutations from the root with no intermediate genomes, was observed in only 0.5% of our simulations (Fig. 2C, see supplementary text for details).

Our epidemic simulations do not support a single introduction of SARS-CoV-2 giving rise to the observed phylogeny. We therefore quantified the relative support for two introductions resulting in the empirical topology. By synthesizing posterior probabilities of inferred ancestral haplotypes, frequencies of topologies in epidemic simulations, and the expected relationships between these haplotypes and topologies, we infer strong support favoring separate introductions of lineages A and B (BF=61.6 and BF=60.0 using the recCA and unconstrained rooting, respectively; see Methods). This support is robust across shorter and longer doubling times, varying ascertainment rates, and minimum polytomy size (tables S4 and S5).

If lineages A and B arose from separate introductions, then the MRCA of SARS-CoV-2 was not in humans, and it is the tMRCAs of lineages A and B that are germane to the origins of SARS-CoV-2 (i.e, not the timing of their shared ancestor). Rooting with the recCA, we inferred the median tMRCA of lineage B to be 15 December (95% HPD: 5 December to 23 December) and the median tMRCA of lineage A to be 20 December (95% HPD: 5 December to 29 December) (Fig. 3A). The tMRCA of lineage B consistently predates the tMRCA of lineage A (Fig. 3B). These results are robust to using unconstrained rooting, fixing the ancestral haplotype, and excluding market-associated genomes (Fig. 3, A and B; table S2; and supplementary text).

Timing the introductions of lineages A and B

The primary case, the first human infected with a virus in an outbreak, could precede the tMRCA if basal lineages went extinct during cryptic transmission (23, 30, 31). The index case, the first identified case, is rarely also the primary case (32, 33). We next used an extension of our previously published framework combining epidemic simulations and phylodynamic tMRCA inference [see Methods; (23, 30, 31)] to infer the timing of the lineage B and lineage A primary cases, accounting for both the index case symptom onset date and earliest documented COVID-19 hospitalization date.

The earliest unambiguous case of COVID-19, with symptom onset on 10 December and hospitalization on 16 December, was a seafood vendor at the Huanan market. Unfortunately no published genome is available for this case (8). Nonetheless, we can reasonably assume this individual had a lineage B virus (supplementary text), as an environmental sample (EPI_ISL_408512) from the stall this vendor operated was lineage B. The earliest lineage A genome (IME-WH01) is from a familial cluster where the earliest symptom onset is 15 December and earliest hospitalization is 25 December (34). Accounting for these dates and using the recCA rooting, we inferred the infection date of the lineage B primary case to be 18 November (95% HPD: 23 October to 8 December) and the infection date of the primary case of lineage A to be 25 November (95% HPD: 29 October to 14 December). The lineage B primary case predated that of lineage A in 64.6% of the posterior sample, by a median of 7 days (Fig. 3D and table S6).

Our lineage A and B primary case inference is robust to rooting on the recCA and fixing the plausible ancestral haplotype to lineage A, lineage B, or C/C, as well as different index case dates, accounting for only hospitalization dates, and varying growth rates and ascertainment rates (tables S7 to S10 and supplementary text). Therefore, our results indicate that lineage B was introduced into humans no earlier than late-October and likely in mid-November 2019, and the introduction of lineage A occurred within days to weeks of this event.

We then inferred the number of ascertained infections and hospitalizations arising from these separate introductions. We find that an earlier introduction of lineage B leads to a faster rise in lineage B-associated infections, dominating the simulated epidemics (Fig. 4) and recapitulating the predominance of lineage B observed in China in early 2020 (35). Similarly, simulated lineage B hospitalizations are more common than those from lineage A through January 2020 (fig. S24). We observe these patterns regardless of rooting strategy (unconstrained or recCA), ancestral haplotype (B, A, or C/C) (Fig. 4 and tables S11 and S12), and doubling time (figs. S25 to S28).

Minimal cryptic circulation of SARS-CoV-2

We do not see evidence for substantial cryptic circulation before December 2019 (Fig. 4), even if we assume a single introduction (fig. S29 and supplementary text). Our simulated epidemics have a median of three (95% HPD 1-18) cumulative infections at the tMRCA, with 99% of simulated epidemics resulting in at most 33 infections (table S13 and supplementary text). Further, it is unlikely there were any COVID-19 related hospitalizations before December (36), as the simulated epidemics show a median of zero (95% HPD: 0–2) hospitalizations by 1 December 2019. These results are in accordance with the lack of a single SARS-CoV-2-positive sample among tens of thousands of serology samples from healthy blood donors from September to December 2019 (37) and thousands of specimens obtained from influenza-like illness patients at Wuhan hospitals from October to December 2019 (34). Therefore, there was likely extremely low prevalence of SARS-CoV-2 in Wuhan before December 2019. Even when we simulated epidemics with a longer doubling time, resulting in an earlier timing of the primary cases (tables S8 and S10), there were still few infections prior to December 2019 (table S13).

Additional introductions

The extinction rate of our simulated epidemics (i.e., simulations that did not produce self-sustaining transmission chains) indicate there were likely multiple failed introductions of SARS-CoV-2. Similar to our previous findings (23), 77.8% of simulated epidemics went extinct. These failed introductions produced a mean of 2.06 infections and 0.10 hospitalizations; hence, failed introductions could easily go unnoticed. If we treat each SARS-CoV-2 introduction, failed or successful, as a Bernoulli trial and simulate introductions until we see two successful introductions, we estimate that eight (95% HPD: 2–23) introductions led to the establishment of both lineage A and B in humans.

Limitations

Our analysis of the putative intermediate haplotypes suggests there remain lineage assignment errors between lineages A and B, particularly of genomes sampled in January and February of 2020, which could influence the precision of the phylogenetic topology and tMRCA inference. Importantly, we lack direct evidence of a virus closely related to SARS-CoV-2 in non-human mammals at the Huanan market or its supply chain. The genome sequence of a virus directly ancestral to SARS-CoV-2 would provide more precision regarding the timing of the introductions of SARS-CoV-2 into humans and the epidemiological dynamics prior to its discovery. Although we simulated epidemics across a range of plausible epidemiological dynamics, our models represent a timeframe prior to the ascertainment of COVID-19 cases and sequencing of SARS-CoV-2 genomes and thus prior to when these models could be empirically validated.

Discussion

The genomic diversity of SARS-CoV-2 during the early pandemic presents a paradox. Lineage A viruses are at least two mutations closer to bat coronaviruses, indicating that the ancestor of SARS-CoV-2 arose from this lineage. However, lineage B viruses predominated early in the pandemic, particularly at the Huanan market, indicating that this lineage began spreading earlier in humans. Further complicating this matter is the molecular clock of SARS-CoV-2 in humans, which rejects a single-introduction origin of the pandemic from a lineage A virus. Here, we resolve this paradox by showing that early SARS-CoV-2 genomic diversity and epidemiology is best explained by at least two separate zoonotic transmissions, in which lineage A and B progenitor viruses were both circulating in non-human mammals prior to their introduction into humans (figs. S30 and S31).

The most probable explanation for the introduction of SARS-CoV-2 into humans involves zoonotic jumps from as-yet undetermined, intermediate host animals at the Huanan market (34, 38, 39). Through late-2019 the Huanan market sold animals that are known to be susceptible to SARS-CoV-2 infection and capable of intra-species transmission (40–42). The presence of potential animal reservoirs, coupled with the timing of the lineage B primary case and the geographic clustering of early cases around the Huanan market (39), support the hypothesis that SARS-CoV-2 lineage B jumped into humans at the Huanan market in mid-November 2019.

In a related study (39), we show that the two earliest lineage A cases are more closely positioned geographically to the Huanan market than expected compared with other COVID-19 cases in Wuhan in early 2020, despite having no known association with the market. This geographic proximity is consistent with a separate and subsequent origin of lineage A at the Huanan market in late-November 2019. The presence of lineage A virus at the Huanan market was confirmed by Gao et al. (43) from a sample taken from discarded gloves.

The high extinction rate of SARS-CoV-2 transmission chains, observed in both our simulations and real-world data (44), indicates that the two zoonotic events establishing lineages A and B may have been accompanied by additional, cryptic introductions. However, such introductions could easily be missed, particularly if their subsequent transmission chains quickly went extinct or the introduced viruses had a lineage A or B haplotype. Failed introductions of intermediate haplotypes are also possible. Critically, we have no evidence of subsequent zoonotic introductions in late-December leading up to the closure of the Huanan market on 1 January 2020. By then, the susceptible host animals that had been documented at the market during the previous months were no longer found in the Huanan market (34).

Other coronavirus epidemics and outbreaks in humans, including SARS-CoV-1, MERS-CoV, and, most recently, porcine deltacoronavirus in Haiti, have been the result of repeated introductions from animal hosts (45–47). These repeated introductions were easily identifiable because human viruses in these outbreaks were more closely related to viruses sampled in the animal reservoirs than to other human viruses. However, the genomic diversity within the putative SARS-CoV-2 animal reservoir at the Huanan market was likely shallower than that seen in SARS-CoV-1 and MERS-CoV reservoirs (45, 46, 48). Hence, even though lineages A and B had nearly identical haplotypes, their MRCA likely existed in an animal reservoir. The ability to disentangle repeated introductions of SARS-CoV-2 from a shallow genetic reservoir has previously been shown in the early SARS-CoV-2 epidemic in Washington state, where two viruses, separated by two mutations, were independently introduced from, and shared an MRCA in, China (figs. S23 and S30 and supplementary text) (11).

Successful transmission of both lineage A and B viruses after independent zoonotic events indicates that evolutionary adaptation within humans was not needed for SARS-CoV-2 to spread (49). We now know that SARS-CoV-2 can readily spread after reverse-zoonosis to Syrian hamsters (Mesocricetus auratus), American mink (Neovison vison), and white-tailed deer (Odocoileus virginianus), indicating its host generalist capacity (50–55). Furthermore, once an animal virus acquires the capacity for human infection and transmission, the only remaining barrier to spillover is contact between humans and the pathogen. Thereafter, a single zoonotic transmission event indicates the conditions necessary for spillovers have been met, which portends additional jumps. For example, there were at least two zoonotic jumps of SARS-CoV-2 into humans from pet hamsters in Hong Kong (56) and dozens from minks to humans on Dutch fur farms (52, 53).

We show that it is highly unlikely that SARS-CoV-2 circulated widely in humans earlier than November 2019 and that there was limited cryptic spread, with, at most, dozens of SARS-CoV-2 infections in the weeks leading up to the inferred tMRCA, but likely far fewer. By late-December, when SARS-CoV-2 was identified as the etiological agent of COVID-19 (8), the virus had likely been introduced into humans multiple times as a result of persistent contact with a viral reservoir.
Materials and methods summary

Materials and methods described in full detail can be found in the supplementary materials.

Sequence data

We queried the GISAID database (57), GenBank, and National Genomics Data Center of the China National Center for Bioinformatics (CNCB), for complete high-coverage SARS-CoV-2 genomes collected by 14 February 2020, resulting in a dataset of 787 taxa belonging to lineages A and B and 20 taxa with C/C or T/T haplotypes. Genomes were aligned using MAFFT v7.453 (58) to the SARS-CoV-2 reference genome (Wuhan/Hu-1/2019) and 388 sites were masked at the 5′ and 3′ ends and at sites based on De Maio et al. (59). All genome accessions are available in data S1 and S2.

Progenitor genome reconstruction and reversion analysis

We reconstructed the progenitor of SARS-CoV-2, the recombinant common ancestor (the recCA). We (i) inferred a maximum likelihood tree of 31 sarbecovirus genomes (SARS-CoV-2 and 30 closely related sarbecoviruses sampled from bats and pangolins) across 15 predefined non-recombinant regions (13) with IQ-TREE v2.0.7 (60), (ii) inferred the sequence of the ancestor of SARS-CoV-2 in each tree with TreeTime v0.8.1 (61), and (iii) concatenated the resulting sequences. We next inferred a maximum likelihood tree of the 787 SARS-CoV-2 taxa with IQ-TREE and performed ancestral state reconstruction with TreeTime to identify substitutions that were reversions from Wuhan-Hu-1 to the recCA across the SARS-CoV-2 phylogeny.

Phylodynamic inference and epidemic simulations

We performed phylodynamic inference using ​​BEAST v1.10.5 (62) with the 787-taxa dataset to infer the ancestral haplotype and the tMRCA of SARS-CoV-2 (and the tMRCAs of lineages A and B), employing a non-reversible random-effects substitution model and exploring unconstrained rooting, recCA-rooting, fixing the ancestral haplotype as a root, and outgroup rooting. SARS-CoV-2-like epidemics were simulated with FAVITES-COVID-Lite v0.0.1 (22, 63) using a scale-free network of 5 million individuals and a customized extension of the SAPHIRE model (64), producing coalescent trees on which we simulated mutations. We calculated the Bayes factor comparing the support of two introductions of SARS-CoV-2 to one introduction by considering the posterior probabilities of the four most likely ancestral haplotypes from the phylodynamic inference (Lineage A, Lineage B, C/C, and T/T), the frequencies of the phylogenetic structures associated with introductions of these haplotypes in the epidemic simulations, and equal prior probabilities for each ancestral haplotype and one versus two introductions.

We connected the phylodynamic inference and epidemic simulations via a rejection sampling-based approach (23), accounting for the tMRCAs of lineages A and B and the earliest documented COVID-19 illness onset and hospitalization dates. We then inferred the timing of the introductions of lineages A and B and the infections and hospitalizations for each lineage. The proportion of epidemic simulations that went extinct (i.e., no onward transmission by the end of the simulation) was used to approximate the number of SARS-CoV-2 introductions needed to result in two introductions with sustained onward transmission.

Acknowledgments

We gratefully acknowledge the authors from the originating laboratories and the submitting laboratories, who generated and shared via GISAID the viral genomic sequences and metadata on which this research is based (data S1) (57). We are greatly appreciative toward Lu Chen, Di Liu, and Yi Yan for providing insight into the putative intermediate genomes and clarification regarding the relative sequencing depth at positions 8782 and 28144, Marc Eloit and Sarah Temmam for sharing their sarbecovirus dataset and recombination analysis results, and Matthew Kuehnert for general feedback. Figure S30 was created with Biorender.com.

Funding: This project has been funded in whole or in part with Federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH), Department of Health and Human Services, under Contract No. 75N93021C00015 (MW). JEP acknowledges support from the NIH (T15LM011271). NM acknowledges support from the National Science Foundation (NSF) (NSF-2028040). JIL acknowledges support from the NIH (5T32AI007244-38). JOW acknowledges support from the NIH (R01AI135992 and R01AI136056). RFG is supported by the NIH (R01AI132223, R01AI132244, U19AI142790, U54CA260581, U54HG007480, OT2HL158260), the Coalition for Epidemic Preparedness Innovation, the Wellcome Trust Foundation, Gilead Sciences, and the European and Developing Countries Clinical Trials Partnership Programme. MAS and AR acknowledge the support of the Wellcome Trust (Collaborators Award 206298/Z/17/Z – ARTIC network), the European Research Council (grant agreement no. 725422 – ReservoirDOCS) and the NIH (R01AI153044). KGA is supported by the NIH (U19AI135995, U01AI151812, and UL1TR002550). ECH is funded by an Australian Research Council Laureate Fellowship (FL170100022). JL, HP, and MSP acknowledge support from the National Research Foundation of Korea, funded by the Ministry of Science and Information and Communication Technologies, Republic of Korea (NRF-2017M3A9E4061995 and NRF-2019R1A2C2084206). TIV acknowledges support from the Branco Weiss Fellowship. We thank AMD for the donation of critical hardware and support resources from its HPC Fund that made this work possible. This work was supported (in part) by the Epidemiology and Laboratory Capacity (ELC) for Infectious Diseases Cooperative Agreement (Grant Number: ELC DETECT (6NU50CK000517-01-07) funded by the Centers for Disease Control and Prevention (CDC). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of CDC or the Department of Health and Human Services.

Author contributions: Conceptualization: JEP, MAS, KGA, MW, JOW; Methodology: JEP, AM, NM, MAS, KGA, MW, JOW; Software: JEP, AM, NM, KG, MAS; Validation: JEP, AM, KI, KG, MAS; Formal analysis: JEP, AM, EP, KI, JLH, KG, JOW; Investigation: JEP, AM, EP, KI, JLH, KG, JOW; Resources: MAS, KGA, JOW; Data Curation: JEP, EP, KG, MZ, JCW, SH, JL, HP, MP, KCZY, RTPL, MNMI, YMN, JOW; Writing - original draft preparation: JEP, MW, JOW; Writing - review and editing: All Authors; Visualization: JEP, JLH, KG, LMMS; Supervision: MAS, KGA, MW, JOW; Project administration: MAS, KGA, MW, JOW; Funding acquisition: MAS, KGA, MW, JOW.

Competing interests: JOW has received funding from the CDC (ongoing) via contracts or agreements to his institution unrelated to this research. MAS receives contracts and grants from the US Food and Drug Administration, the US Department of Veterans Affairs and Janssen Research and Development unrelated to this research. RFG is co-founder of Zalgen Labs, a biotechnology company developing countermeasures to emerging viruses. MW, ECH, AR, MAS, JOW, and KGA have received consulting fees and/or provided compensated expert testimony on SARS-CoV-2 and the COVID-19 pandemic.

Data and materials availability: Genome accessions are available in data S1 and S2, and raw data for two genomes were deposited to NCBI SRA (PRJNA806767 and PRJNA802993). Code is available on Zenodo (65). The following data are available on Data Dryad (66): recCA sequence, BEAST phylogenetic inference output, and simulation and rejection sampling output for the primary analysis. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.

Safety, immunogenicity, and reactogenicity of BNT162b2 and mRNA-1273 COVID-19 vaccines given as fourth-dose boosters

Safety, immunogenicity, and reactogenicity of BNT162b2 and mRNA-1273 COVID-19 vaccines given as fourth-dose boosters following two doses of ChAdOx1 nCoV-19 or BNT162b2 and a third dose of BNT162b2 (COV-BOOST): a multicentre, blinded, phase 2, randomised trial

Figure 2 Solicited adverse events within 7 days following fourth-dose vaccination in participants who received a study vaccine (A) Severe (grade 3–4) local and systemic adverse events. (B) Moderate or severe (grade 2–4) local and systemic adverse events. For each solicited adverse event, the highest severity within the first 7 days after fourth-dose vaccination at an individual level was used to draw the plot.

Summary

Background

Some high-income countries have deployed fourth doses of COVID-19 vaccines, but the clinical need, effectiveness, timing, and dose of a fourth dose remain uncertain. We aimed to investigate the safety, reactogenicity, and immunogenicity of fourth-dose boosters against COVID-19.

Methods

The COV-BOOST trial is a multicentre, blinded, phase 2, randomised controlled trial of seven COVID-19 vaccines given as third-dose boosters at 18 sites in the UK. This sub-study enrolled participants who had received BNT162b2 (Pfizer-BioNTech) as their third dose in COV-BOOST and randomly assigned them (1:1) to receive a fourth dose of either BNT162b2 (30 μg in 0·30 mL; full dose) or mRNA-1273 (Moderna; 50 μg in 0·25 mL; half dose) via intramuscular injection into the upper arm. The computer-generated randomisation list was created by the study statisticians with random block sizes of two or four. Participants and all study staff not delivering the vaccines were masked to treatment allocation. The coprimary outcomes were safety and reactogenicity, and immunogenicity (anti-spike protein IgG titres by ELISA and cellular immune response by ELISpot). We compared immunogenicity at 28 days after the third dose versus 14 days after the fourth dose and at day 0 versus day 14 relative to the fourth dose. Safety and reactogenicity were assessed in the per-protocol population, which comprised all participants who received a fourth-dose booster regardless of their SARS-CoV-2 serostatus. Immunogenicity was primarily analysed in a modified intention-to-treat population comprising seronegative participants who had received a fourth-dose booster and had available endpoint data. This trial is registered with ISRCTN, 73765130, and is ongoing.

Findings

Between Jan 11 and Jan 25, 2022, 166 participants were screened, randomly assigned, and received either full-dose BNT162b2 (n=83) or half-dose mRNA-1273 (n=83) as a fourth dose. The median age of these participants was 70·1 years (IQR 51·6–77·5) and 86 (52%) of 166 participants were female and 80 (48%) were male. The median interval between the third and fourth doses was 208·5 days (IQR 203·3–214·8). Pain was the most common local solicited adverse event and fatigue was the most common systemic solicited adverse event after BNT162b2 or mRNA-1273 booster doses. None of three serious adverse events reported after a fourth dose with BNT162b2 were related to the study vaccine. In the BNT162b2 group, geometric mean anti-spike protein IgG concentration at day 28 after the third dose was 23 325 ELISA laboratory units (ELU)/mL (95% CI 20 030–27 162), which increased to 37 460 ELU/mL (31 996–43 857) at day 14 after the fourth dose, representing a significant fold change (geometric mean 1·59, 95% CI 1·41–1·78). There was a significant increase in geometric mean anti-spike protein IgG concentration from 28 days after the third dose (25 317 ELU/mL, 95% CI 20 996–30 528) to 14 days after a fourth dose of mRNA-1273 (54 936 ELU/mL, 46 826–64 452), with a geometric mean fold change of 2·19 (1·90–2·52). The fold changes in anti-spike protein IgG titres from before (day 0) to after (day 14) the fourth dose were 12·19 (95% CI 10·37–14·32) and 15·90 (12·92–19·58) in the BNT162b2 and mRNA-1273 groups, respectively. T-cell responses were also boosted after the fourth dose (eg, the fold changes for the wild-type variant from before to after the fourth dose were 7·32 [95% CI 3·24–16·54] in the BNT162b2 group and 6·22 [3·90–9·92] in the mRNA-1273 group).

Interpretation

Fourth-dose COVID-19 mRNA booster vaccines are well tolerated and boost cellular and humoral immunity. Peak responses after the fourth dose were similar to, and possibly better than, peak responses after the third dose.

Funding

UK Vaccine Task Force and National Institute for Health Research.

Introduction

With the emergence of highly transmissible SARS-CoV-2 variants, such as omicron (B.1.1.529), many high-income countries have rapidly deployed third doses of COVID-19 vaccines to their populations. Third-dose boosters increase humoral and cellular immunity1 and provide more short-term protection against symptomatic infection with variants of concern, including omicron,2, 3 compared with a two-dose schedule. However, protection against symptomatic infection wanes rapidly following the second4 and third2 doses of COVID-19 vaccines. As of March, 2022, some countries, such as Israel and Germany, started to offer fourth-dose booster vaccines to their populations, and the UK rolled out fourth doses for clinically vulnerable populations in April, 2022.5

Observational data from Israel have shown a boosting effect on immunogenicity6 and moderate protection against symptomatic infection from a fourth dose of mRNA COVID-19 vaccines administered approximately 4 months after a third dose.7, 8 The clinical need, timing, and dose of the fourth COVID-19 vaccine remain uncertain,9 as does the gain in vaccine effectiveness compared with a third dose. Given the urgent need for data to inform policy on additional booster doses, the COV-BOOST trial1 of third-dose booster vaccines for COVID-19 was extended to investigate the safety, reactogenicity, and immunogenicity of fourth-dose boosters against COVID-19 administered approximately 7 months following a third dose of BNT162b2 (Pfizer-BioNTech).

Methods

Study design and participants

The COV-BOOST trial is a multicentre, blinded, phase 2, randomised controlled trial1 done at 18 sites in the UK. For the main COV-BOOST study,1 we enrolled participants aged 30 years or older who had received two doses of BNT162b2 or ChAdOx1 nCoV-19 (Oxford-AstraZeneca) and randomly assigned them to receive either a third-dose booster of one of seven COVID-19 vaccines (in ten schedules) or a meningococcal vaccine control. Details of the main study design have been described previously,1 and the full inclusion and exclusion criteria can be found in the protocol (appendix 2 pp 36–38). The statistical analysis plan is provided as appendix 3. This study is a randomised sub-trial nested within the main COV-BOOST trial. Participants who received a third-dose BNT162b2 booster in the COV-BOOST trial during June, 2021, were eligible for inclusion in this sub-study unless they had a previous severe adverse reaction to mRNA vaccines or had acquired an additional COVID-19 vaccine outside of the study since enrolling. Based on site location and participant availability, a subset of participants (around 25 per group) were enrolled into an immunology cohort to collect cellular immunology samples at 14 days after the fourth dose. The trial was reviewed and approved by the UK National Health Service (NHS) Research Ethics Service (21/SC/0171). All participants provided written informed consent.

Research in context

Evidence before this study

We searched PubMed for randomised controlled trials in non-immunocompromised adults published between database inception and March 31, 2022, using the search terms “(COVID) AND (vaccin*) AND (booster OR fourth dose)” with no language restrictions. We identified no clinical trials including fourth-dose COVID-19 vaccine boosters. One observational study following fourth doses of full-dose BNT162b2 (Pfizer-BioNTech) or half-dose mRNA-1273 (Moderna) in Israel in people who had received three previous doses of BNT162b2 found that humoral immunity after the fourth dose was boosted above peak levels after the third dose. A preprint of a small observational study of fourth-dose boosters from Germany found a boost to humoral immunity from baseline and the activation of T cells, which was weakly correlated with serum anti-spike protein antibody titres.

Added value of this study

To our knowledge, this study is the first to report a randomised trial of fourth-dose COVID-19 boosters. These data suggest that, after a period of approximately 7 months following third-dose boosters with BNT162b2, an additional dose of a COVID-19 mRNA vaccine can boost humoral anti-spike protein IgG titres and cellular responses to, or higher than, levels seen at 28 days after a third dose. Some participants with high levels of humoral and cellular responses before the fourth dose had limited boosting from the fourth dose, indicating that there could be a vaccine-specific ceiling effect. There might be additional antibody and T-cell boosting from heterologous mRNA fourth vaccine doses.

Implications of all the available evidence

More than 6 months after third-dose boosters, fourth doses of COVID-19 mRNA vaccines provide large increases in anti-spike protein antibody titres, although these increases will probably wane rapidly, as has been observed after third doses. People with high antibody titres are unlikely to gain much boosting from additional doses. This study provides important data to guide policy makers who might be considering the deployment of further booster doses of COVID-19 vaccines to the clinically vulnerable or whole populations.

Randomisation and masking

Eligible participants were randomly assigned (1:1) to receive either BNT162b2 or mRNA-1273 (Moderna) as a fourth dose. The computer-generated randomisation list was created by the study statisticians with random block sizes of two or four, and randomisation was done with the electronic data capture system REDCap (version 10.6.13) by trained site staff. Allocation concealment was maintained by REDCap, in which the final randomisation list was only accessible by the IT manager and trial statistician. Randomisation was stratified by the initial two-dose vaccine schedule (ChAdOx1 nCoV-19 plus ChAdOx1 nCoV-19 vs BNT162b2 plus BNT162b2), study site, age (<70 years vs ≥70 years), and cohort (general vs immunology). Participants, laboratory staff, and the clinical study team not delivering the vaccines, including those assessing adverse events, were masked to treatment allocation. Data analysts were not masked to treatment allocation. Participant masking was maintained by concealing randomisation pages, preparing vaccines out of sight, and applying masking tape to vaccine syringes to conceal dose, volume, and appearance.

Procedures

Procedures for the main study have been previously described.1 Two COVID-19 vaccines were used in this sub-study. Both BNT162b2 and mRNA-1273 are lipid nanoparticle-formulated, nucleoside-modified mRNA vaccines encoding trimerised SARS-CoV-2 spike glycoprotein. Administered by appropriately trained trial staff at the trial sites, participants received either BNT162b2 (30 μg in 0·30 mL; full dose) or mRNA-1273 (50 μg in 0·25 mL; half dose) via intramuscular injection into the upper arm. Participants were observed for at least 15 min after vaccination.

Blood samples for immunogenicity were taken at day 0 (before the fourth dose), day 14 (after the fourth dose), and day 84. Immunological assays are described in appendix 1 (p 12). Briefly, we measured SARS-CoV-2 anti-spike protein IgG concentrations by ELISA (Nexelis; Laval, QC, Canada) for all participants at all timepoints and cellular immune responses by ELISpot (Oxford Immunotec; Abingdon, UK) at day 0 for all participants and at day 14 for the immunology cohort only. Anti-SARS-CoV-2 nucleocapsid IgG status was analysed at Porton Down, Public Health England, by an electrochemiluminescence immunoassay (Cobas platform, Elecsys assay; Roche Diagnostics; Rotkreuz, Switzerland). Safety endpoints were followed up by use of electronic diaries completed by participants daily for the first 7 days and then on an ad hoc basis and by direct solicitation in person at the day 14 follow-up visit. The study visits will be completed by May, 2022.

Outcomes

The coprimary outcomes were the safety and reactogenicity, and immunogenicity, of fourth-dose booster vaccination with full-dose BNT162b2 or half-dose mRNA-1273. Safety was assessed by sites, reactogenicity was self-reported, and immunogenicity was assessed centrally by different commercial laboratories. Safety and reactogenicity were characterised by the occurrence of solicited local and systemic adverse events within 7 days of the fourth dose, unsolicited adverse events within 28 days of the fourth dose, medically attended adverse events up to 3 months following the fourth dose, adverse events of special interest, and serious adverse events. Serious adverse events and adverse events of special interest (appendix 2 p 73) were recorded throughout the study. The severity of clinical and laboratory adverse events was assessed according to scales based on the toxicity grading scales of the Food and Drug Administration for healthy adult volunteers enrolled in preventive vaccine clinical trials. Immunogenicity was defined as anti-spike protein IgG antibody titres (and live virus neutralising antibody titres, data for which are not reported here due to laboratory delays but will be reported at the first opportunity) for all participants and cellular immune responses for participants in the immunology cohort (appendix 1 p 12). To accelerate the data being available for policy decision making, and because maximum anti-spike protein IgG responses had been seen before day 28 following a third dose in the initial analysis,1 we used day 14 as the primary outcome timepoint. A secondary outcome was immunogenicity at day 84 following the fourth dose; because these assays have not yet been processed, we do not report this outcome and it will be reported elsewhere.

Statistical analysis

Our aim was to investigate the boosting in immunological endpoints following two mRNA fourth-dose booster vaccines administered after ChAdOx1 nCoV-19 plus ChAdOx1 nCoV-19 plus BNT162b2 or BNT162b2 plus BNT162b2 plus BNT162b2 (the most commonly deployed COVID-19 vaccination schedules in the UK). As hypothesis testing between the two fourth-dose mRNA vaccines was not the primary aim of the main study, no power or formal sample size calculations were done.

Safety and reactogenicity were analysed in the per-protocol population, which comprised all participants who received a fourth-dose booster, regardless of their history of SARS-CoV-2 infection and anti-nucleocapsid IgG serostatus before the fourth dose. The proportions of participants with at least one severe (grades 3–4) or one severe or moderate (grades 2–4) adverse event are presented by initial vaccine schedules (ChAdOx1 nCoV-19 plus ChAdOx1 nCoV-19 vs BNT162b2 plus BNT162b2) by use of radial plots. Unsolicited adverse events were coded according to the Medical Dictionary for Regulatory Activities and tabulated at System Organ Class level across vaccine groups. Adverse events of special interest and serious adverse events are reported up to the data cutoff date of March 2, 2022, by line listing.

The primary immunogenicity outcomes were analysed in the modified intention-to-treat seronegative population, which comprised participants who received a fourth-dose booster, were seronegative before receiving the fourth dose (defined by the Roche Elecsys anti-SARS-CoV-2 nucleocapsid assay at all study visits before the fourth dose, including days 0 and 84 of the third dose and day 0 of the fourth dose), did not have SARS-CoV-2 infection before or within 7 days of the fourth dose (self-reported PCR or lateral flow tests following community testing), and had available endpoint data. The main analyses included all participants regardless of their initial two-dose vaccine schedules, with prespecified subgroup analyses split by the initial two-dose schedules (two doses of ChAdOx1 nCoV-19 vs BNT162b2) and age (<70 years vs ≥70 years).

In the immunogenicity analysis, we compared anti-spike protein IgG and T-cell responses at 14 days after the fourth dose versus 28 days after the third dose (data previously reported).1 For each paired data from one participant, the fold change was calculated by dividing immunogenicity values at day 14 after the fourth dose by those at day 28 after the third dose. As the fold change has a log-normal distribution, the geometric mean of the fold change between the two timepoints with 95% CIs are reported, with no adjustment for multiplicity. Absolute levels of immune responses and fold changes before (day 0) versus 14 days after the fourth dose are summarised by geometric means and 95% CIs. The immunogenicity analyses were also repeated in the seropositive modified intention-to-treat population, which comprised participants who received a fourth-dose booster and who had evidence of SARS-CoV-2 infection before the fourth dose (defined by the Roche Elecsys anti-SARS-CoV-2 nucleocapsid assay or via self-reported PCR or lateral flow test) or within 7 days of the fourth dose (self-reported PCR or lateral flow test).

All analyses were done by use of R, version 4.1.1. This trial is registered with ISRCTN, 73765130. An independent data safety monitoring board reviewed safety data regularly, and local trial site physicians provided oversight of all adverse events in real time.

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Among 215 participants who received a third dose of BNT162b2 in June, 2021, 166 people volunteered and were screened from Jan 11 to Jan 25, 2022, for the fourth dose sub-study (figure 1). All participants were eligible and were randomly assigned to receive either full-dose BNT162b2 (n=83) or half-dose mRNA-1273 (n=83) as a fourth-dose vaccine (figure 1). 88 participants had previously received two doses of ChAdOx1 nCoV-19 plus a third dose of BNT162b2 and 78 participants had previously received three doses of BNT162b2 (figure 1, Table 1, Table 2). The median age of the entire cohort was 70·1 years (IQR 51·6–77·5). Among those who had received two doses of ChAdOx1 nCoV-19 plus a third dose of BNT162b2, the baseline characteristics were well balanced between the two fourth-dose groups (table 1). For participants who had received three doses of BNT162b2, those in the fourth-dose BNT162b2 group were younger (median age 67·2 years vs 73·2 years) and had a shorter interval between the second and third doses (median time 96·0 days vs 106·0 days) than those in the fourth-dose mRNA-1273 group (table 1). The median interval between the third and fourth doses was similar for the four groups and was 208·5 days (IQR 203·3–214·8) for the entire cohort (table 1). 166 participants received a fourth-dose vaccination and were included in the safety and reactogenicity analysis. We excluded 29 participants from the main immunogenicity analysis who were seropositive or had self-reported COVID-19 before the fourth dose and four who did not attend the day 14 visit after the fourth dose. 133 people were included in the modified intention-to-treat immunogenicity analysis of the seronegative population, of whom 66 received full-dose BNT162b2 and 67 received half-dose mRNA-1273 (figure 1; table 2). Where the population does not total 133, there are missing data, the reason for which is still being investigated.

Pain was the most common solicited local adverse event for participants receiving full-dose BNT162b2 and those receiving half-dose mRNA-1273 booster doses and was mostly mild or moderate in severity (appendix 1 pp 2–3). Fatigue, headache, malaise, and muscle ache were the most common solicited systemic adverse events (appendix 1 pp 2–3). One (3%) of 39 participants who received four doses of BNT162b2, two (3%) of 39 participants who received three doses of BNT162b2 and one half-dose of mRNA-1273, four (9%) of 44 people who received two doses of ChAdOx1 nCoV-19 and two doses of BNT162b2, and three (7%) of 44 people who received two doses of ChAdOx1 nCoV-19, one dose of BNT162b2, and one half-dose of mRNA-1273 had any severe (grades 3–4) local and systemic solicited adverse event within 7 days of the fourth dose (figure 2).

Up to the data extraction cutoff date of March 2, 2022, three serious adverse events, all in recipients of BNT162b2 as a fourth dose, were reported, none of which were related to the study vaccine (appendix 1 pp 9–11). 16 adverse events were reported after fourth-dose BNT162b2 and 18 adverse events were reported after fourth-dose mRNA-1273 (including unsolicited adverse events within 28 days, medically attended adverse events within 3 months, and all other adverse events reported up to data lock). Four adverse events of special interest were reported in the group who received three doses of BNT162b2 and one half-dose of mRNA-1273, all of which were unrelated to the study vaccine (appendix 1 pp 9–11).

In the group who received BNT162b2 as their fourth dose, geometric mean anti-spike protein IgG concentration at day 28 after the third dose was 23 325 ELISA laboratory units (ELU)/mL (95% CI 20 030–27 162), which increased to 37 460 ELU/mL (31 996–43 857) after the fourth dose, representing a significant fold change (geometric mean 1·59, 95% CI 1·41–1·78; table 2; figure 3). Similarly, there was a significant increase in geometric mean anti-spike protein IgG concentration from 28 days after the third dose (25 317 ELU/mL, 95% CI 20 996–30 528) to 14 days after a fourth dose of mRNA-1273 (54 936 ELU/mL, 46 826–64 452), with a geometric mean fold change of 2·19 (1·90–2·52; table 2; figure 3). This increase in anti-spike protein IgG titres between these two timepoints was observed regardless of initial vaccine schedule or age group (Table 2, Table 3; appendix 1 pp 4–5). There was a considerable decay of anti-spike protein IgG titres during approximately 7 months from day 28 after the third dose to just before the fourth dose (day 0; figure 3), leading to a geometric mean fold change between day 0 and day 14 of the fourth dose ranging from 11·14 to 20·26 (table 2).

Among participants with cellular response data available, similar T-cell responses were seen at day 14 after the fourth dose compared with day 28 after the third dose across tested variants for participants who received two doses of ChAdOx1 nCoV-19 plus two doses of BNT162b2, two doses of ChAdOx1 nCoV-19 plus one dose of BNT162b2 plus one half dose of mRNA-1273, or four doses of BNT162b2 (table 2; figure 3; appendix p 7). However, among participants who received three doses of BNT162b2 and one half-dose of mRNA-1273, T-cell responses were significantly increased 14 days after the fourth dose compared with 28 days after the third dose (table 2; figure 3). Similar to anti-spike protein IgG titres, a decay of cellular response was also seen from 28 days after the third dose to day 0 of the fourth dose (figure 3), resulting in a significant boosting effect on cellular response in most groups after the fourth dose (fold change ranging from 3·46 to 11·07; table 2).

For participants with evidence of SARS-CoV-2 infection before or within 7 days of the fourth dose, there were 4·89-fold (95% CI 4·35–5·50; n=13) and 4·63-fold (4·04–5·29; n=15) increases in anti-spike protein IgG titres from day 0 of the fourth dose to day 14 after the fourth dose for full-dose BNT162b2 and half-dose mRNA-1273, respectively (appendix 1 p 8). A boost effect on T-cell responses in this population was also seen between day 0 and day 14 relative to the fourth dose, although the sample size was small (appendix 1 p 8).

Discussion

To our knowledge, we present the first data from a randomised trial on the safety, reactogenicity, and immunogenicity of full-dose BNT162b2 and half-dose mRNA-1273 COVID-19 vaccines given as fourth-dose boosters in healthy adult populations who had previously received different vaccine schedules. These data show that a fourth dose of COVID-19 mRNA vaccines is well tolerated and can provide a substantial boost to both humoral and cellular immunity approximately 7 months after a third-dose booster, with anti-spike protein IgG titres at day 14 following the fourth dose higher than those at day 28 after the third dose for both BNT162b2 and mRNA-1273.

The peak anti-spike protein IgG concentration after a fourth vaccine dose was also higher than after a third dose for full-dose BNT162b2 and half-dose mRNA-1273 among participants in an Israeli observational study who had previously received three doses of BNT162b2 and had low anti-spike IgG titres before the fourth dose.6 The fold changes before and after the fourth dose in the Israeli study were lower to those found in our study, probably due to the shorter interval between third and fourth doses in the Israeli study as a longer duration between vaccine doses is recognised to increase immunogenicity.10, 11 A large increase from baseline in neutralising antibody titres after a fourth dose of mRNA COVID-19 vaccine was also observed in a German observational study, although neutralising capacity against omicron subvariants remained low.12

In our study, the fold change in anti-spike protein IgG titres between day 0 and day 14 of the fourth dose ranged from 11·14 to 20·26. There are two possible explanations for such a large fold change: first, the vaccines remain strongly immunogenic, and, second, the boost is from a relatively low baseline following waning of immunity after the third dose. Baseline anti-spike protein IgG concentrations before the fourth dose (day 0) were similar to baseline concentrations before the third dose (day 0). Some participants in our study maintained high levels of humoral and cellular responses even before the fourth dose and had limited boosting from the fourth dose. This finding was replicated in participants with a history of SARS-CoV-2 infection, indicating that there might be a ceiling or maximum anti-spike protein IgG titre and T-cell response and that the fourth dose might not boost humoral and cellular responses if the baseline response is high. These individual data are important for policy makers as the benefit of a fourth dose might be less in people who already have high levels of immune responses from recent infection or vaccination. In addition, this ceiling effect could be dependent on vaccine type and dose. If this ceiling effect is replicated in other datasets, it could be due to host immunity, vaccine type, or vaccine dose, which needs to be explored in further trials and analyses.

Our results for immunogenicity are also consistent with the little observational evidence on vaccine effectiveness available from Israel, which indicates increased protection against symptomatic infection and severe illness from a fourth-dose booster.6, 7 In our study, half-dose mRNA-1273 appeared to have higher immunogenicity than full-dose BNT162b2, which was also seen in the Israeli study,6 although the two groups in the Israeli study were not randomised. This result might be due to a heterologous schedule effect or the vaccine dose. For third doses given in the main COV-BOOST study, heterologous mRNA vaccines appeared to provide a superior boost to third homologous doses.1, 13 In addition to the boost to humoral immunity, there was also a boost in broad cellular responses after a fourth vaccine dose. Due to the small number of samples available for analysis, it is difficult to quantify the size of the booster effect or make direct comparisons across all the schedules tested. A higher number of samples will be tested at the day 84 timepoint to investigate any differences.

Our study has several limitations. The number of participants within each subgroup is relatively small as we recruited only existing COV-BOOST participants who had received BNT162b2 as their third dose within the study. An even smaller number of samples were available for our analysis of cellular immunity, meaning low levels of precision to quantify T-cell responses. There were not enough samples to investigate any potential benefit of heterologous schedules on cellular responses. The timepoints after the third and fourth doses were different, but humoral responses in previous studies were at similar levels between day 7 and day 28 after vaccination.1, 6 Due to laboratory capacity, data for neutralising antibodies against variants of concern, including omicron, were not available when this Article was developed. Given that a strong correlation has been observed between anti-spike protein IgG titres and neutralising antibody titres against SARS-CoV-2 variants of concern,1 it is expected that the titres of neutralising antibodies after a fourth dose are similar to those observed following a third dose. Furthermore, only mRNA vaccines, which are currently difficult to obtain or are unavailable in many low-income or middle-income countries, were analysed as fourth-dose vaccines in this study.

The strengths of this study include it being the first to report on mixed-schedule fourth-dose data from a randomised trial and on populations who had received vaccines other than BNT162b2 as their first, second, or third dose. This study provides important data to help guide policy makers in decisions on how to use fourth doses of COVID-19 vaccines.

Contributors

SNF, MDS, XL and JSN-V-T conceived the trial and SNF is the chief investigator. SNF, APSM, MDS, and XL contributed to the protocol and design of the study. APSM, GB, and SS led the implementation of the study. XL, SF, LJ, and VC designed and did the statistical analysis and have accessed and verified the underlying data. XL, APSM, SF, and SNF drafted the manuscript. All other authors contributed to the implementation of the study and data collection. All authors reviewed and approved the final manuscript. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Data sharing

The study protocol is provided as appendix 2 and the statistical analysis plan is provided as appendix 3. Individual participant data will be made available when the study is complete upon reasonable requests made to the corresponding author ( s.faust@soton.ac.uk ); data can be shared through secure online platforms after proposals are approved. All the sequence datasets used in the T-cell analysis are available in the public GISAID database (https://www.gisaid.org).

FDA authorizes COVID vaccines for the littlest kids: what the data say

Nature NEWS 17 June 2022

The Moderna and Pfizer shots are hard to compare, so researchers and parents have lingering questions.

By Cassandra Willyard

Kids under five years old are the largest group of people in the United States not yet eligible for COVID-19 vaccines.
Credit: John Tlumacki/The Boston Globe via Getty

The US Food and Drug Administration (FDA) has given emergency authorization to COVID-19 vaccines for children aged five and younger. Assuming the Centers for Disease Control and Prevention (CDC) also signs off the decision, an extra 18 million people in the United States will be eligible for inoculation — the last large group to be granted access.

Vaccine makers Moderna and Pfizer each presented clinical-trial data to an FDA advisory panel on 15 June, showing that their mRNA-based vaccines are safe for children, and trigger antibody levels similar to those that have provided protection for adults. But researchers and parents still have questions about the real-world benefits of the vaccines, and which will perform best.

Young children have the lowest risk of becoming seriously ill with COVID-19. But that doesn’t mean the disease is benign. Since the pandemic began, 442 children aged 4 and younger have died of the disease in the United States, and thousands have been hospitalized. The coronavirus variant Omicron hit kids especially hard this year. After its emergence, the hospitalization rate for children under five was five times what it was during the previous surge, caused by the Delta variant. The numbers might seem small, says Yvonne Maldonado, a paediatrician and infectious-disease specialist at Stanford University in California, but children “shouldn’t be dying of anything”. “If we have a way to prevent deaths, we should be preventing them.”

Head to head

If the CDC green-lights the vaccines — which looks likely — parents will be eager for information about which to give their children. The most notable difference is in the number and timing of the doses. Moderna’s vaccine will be administered as two doses one month apart, each one-quarter of the amount given to adults. Pfizer’s will be given as three doses, with three weeks between the first two, and eight weeks between the second and third. Each shot is one-tenth the amount given to adults.

Safety was a top concern among FDA panel members, and both vaccines met the mark (the panel recommended authorizing them in a 21–0 vote). Most side effects were mild, such as pain at the injection site and fatigue, and resolved quickly.

The firms disclosed that serious adverse reactions related to the vaccine had occurred, but were rare. Moderna, based in Cambridge, Massachusetts, reported that one child who received its vaccine had a seizure triggered by a high fever (see ‘Moderna paediatric-trial results, at a glance’), and Pfizer, based in New York City, reported one case of fever and calf pain that might have been linked to vaccination (see ‘Pfizer paediatric-trial results, at a glance’).

“Beyond the one febrile seizure, there wasn’t anything that was highly concerning,” says Andrew Janowski, a paediatric infectious-disease specialist at Washington University School of Medicine in St. Louis, Missouri, who tuned in to the meeting virtually. “That’s what was very reassuring to me.”

Efficacy against infection with the coronavirus SARS-CoV-2 was a bit harder to parse for each vaccine. Regulators allowed the vaccine makers to infer efficacy by demonstrating that the vaccines could elicit antibody levels similar to those that have been protective for teens and young adults, a concept known as immunobridging. That helped to speed up the trials.

But the companies did manage to collect some efficacy data. In the Moderna trial, 265 out of 5,476 kids contracted COVID-19, and the efficacy ranged from about 50% in infants and toddlers to less than 40% in children aged 2–5. The Pfizer vaccine seemed to do better, with an average efficacy of about 80% in children aged 6 months to 4 years. But these figures are based on a tiny number of cases — just seven infections in the placebo group and three in the vaccine group. Doran Fink, deputy director of vaccines and related products applications at the FDA in Silver Spring, Maryland, said at the panel meeting that he regards those estimates as “preliminary” and “imprecise”.

Concerns remain

Despite wide agreement among panellists that the benefits of both vaccines outweigh the risks, some concerns did bubble up. Paul Offit, a vaccine and infectious-disease specialist at Children’s Hospital of Philadelphia in Pennsylvania, worried about the apparent lack of efficacy against Omicron demonstrated by the first two doses of the Pfizer vaccine, which was developed in partnership with biotechnology firm BioNTech, based in Mainz, Germany. Offit told Nature after the meeting: “You didn’t see any evidence for protection.” In other age groups, he added, the Moderna and Pfizer vaccines “track side by side in terms of efficacy”. This age group is “the first time you see them separate”.

That leaves young children who get the Pfizer vaccine potentially vulnerable for longer. It also means that children must have three doses to get protection, which could present a logistical challenge. “I have a lot of concern that many of these kids will not get the third dose, as we know the struggle to get people in for two,” said Jeannette Lee, a biostatistician at the University of Arkansas for Medical Sciences in Little Rock and a member of the advisory panel, during the meeting. “We’ve already seen with the boosters for adults, lots of people don’t take them.”

Wayne Marasco, a cancer immunologist at the Dana-Farber Cancer Institute in Boston, Massachusetts, brought up another concern that is relevant to both vaccines. He said that the first strain of a virus that a person is exposed to can bias their immune response to new variants of that virus for life — a phenomenon known as immune imprinting. That can be a problem for both children and adults. If young kids are given a vaccine against an early version of SARS-CoV-2, the question is whether their immune systems will protect them against a heavily evolved variant such as Omicron.

In a study published this month in Science1, triple-vaccinated health-care workers who became infected with Omicron displayed a boost in their T-cell, B-cell and antibody responses, but only against variants of concern that evolved before Omicron.

Despite these worries, says Andy Pekosz, an immunologist at Johns Hopkins University in Baltimore, Maryland, “you’re still much better off getting a vaccine and getting that immunity, than really taking a risk and acquiring that immunity via infection”.

An agonizing wait

This decision comes more than seven months after the first vaccine was authorized for US children aged five and older, and after a series of delays. The wait has been agonizing for some parents, and their frustration was palpable during the public-comment segment of the panel meeting. “I cannot know the FDA internal workings, but I can say the lack of transparency as to why the Moderna under-five review has taken longer than any other age cohort has made me feel like vaccinating my kids was not a priority for the FDA,” said Lauren Dunnington, who works in global public health and has two children under five.

According to a survey published in May by KFF, a health-policy organization based in San Francisco, California, these frustrated parents represent a minority. In the poll, just 18% of parents of under-fives planned to get their kids vaccinated “right away”. Another 38% would “wait and see”. And more than one-quarter — 27% — would not get their young children vaccinated at all. Eleven per cent would do so only if required. That could be due in part to a lack of information. A little more than half of the parents polled said they didn’t have enough information about the safety and effectiveness of vaccines in this age group.

Given the expected low uptake, the vaccine isn’t likely to have much of an impact on the pandemic. But it could make a substantial difference in the lives of families that choose to get their children vaccinated — especially those that have been completely isolating their children socially to protect them. Vaccinated kids might also miss fewer days of school and childcare owing to illness or quarantine restrictions.

The two vaccines are likely to be authorized by the FDA in the coming days. But it is up to the CDC to decide how they should be used. That agency’s Advisory Committee on Immunization Practices is expected to make its recommendation within days, and then the agency’s director, Rochelle Walensky, must sign off on it. Offit predicts that the vaccines will receive a full recommendation from the committee, but he notes that there is some precedent for recommending one vaccine over another.

If all goes to plan, the first shots could go into arms as soon as 21 June, according to senior White House officials. When that happens, the United States will join just a handful of countries that are vaccinating children under five, including Argentina, Bahrain, China, Cuba and Venezuela. It is unclear whether other countries will follow the US decision to make vaccines available to the youngest kids.

doi: https://doi.org/10.1038/d41586-022-01689-w

References

1. Reynolds, C. J. et al. Science https://doi.org/10.1126/science.abq1841 (2022).

Does Omicron hit kids harder? Scientists are trying to find out

Nature NEWS 04 February 2022

Children are making up a larger proportion of patients hospitalized with COVID than in previous infection waves.

by Max Kozlov

Children might be more susceptible to COVID because many have not yet been vaccinated. Credit: Majority World/Universal Images Group via Getty

As the highly transmissible Omicron coronavirus variant has swept the globe in the past two months, millions of people have been hospitalized. Children have been no exception, and, in the United States, they have made up a larger proportion of COVID-19 hospitalizations than at any other time of the pandemic.

Such paediatric hospitalizations might seem concerning, but estimates show that the individual risk of a child with Omicron being hospitalized is, in fact, lower — by one-third to one-half — than it was when the Delta variant was dominant. And hospitalized children are not presenting with any more severe illness than they were with other variants, says Michael Absoud, a specialist in women and children’s health at King’s College London. Preliminary UK data show that although there has been an increase in the proportion of children hospitalized with COVID-19 has increased during the Omicron wave — especially those under the age of one — the children have required fewer medical interventions, such as ventilators and supplemental oxygen.

These findings mirror the trend in the general population: Omicron seems less likely than Delta to cause hospitalization or death, especially in immunized and younger populations. But scientists are still trying to work out why Omicron has led to disproportionately more hospitalizations in children. In the United States, for example, children make up about 5% of all COVID-19 hospitalizations — a proportion up to four times higher than that of previous coronavirus waves.

One potential explanation is that the variant’s extremely high transmissibility, when coupled with a lack of built-up immunity from vaccination or past infection, leaves children more vulnerable to Omicron, compared with adults who have had access to vaccines for months. Most countries have not yet authorized a COVID-19 vaccine for children under the age of 5, and some have not yet offered it to children under 12. Even in the United States, which has authorized COVID-19 vaccinations for 5–11-year-olds, less than one-third of children in that age group have received a jab.

Omicron is less likely to cause severe illness in all age groups. But another possible explanation for the data is that Omicron’s multitude of mutations has made the illness different and perhaps slightly more serious in younger children than in adult populations, says Andrew Pavia, head of the division of paediatric infectious diseases at University of Utah Health in Salt Lake City. As evidence for this theory, Pavia cites early reports hinting that Omicron might not infect lung cells as readily as cells in the upper airways. In general, the lungs are where the coronavirus does much of its damage, and so fewer infected lung cells could mean a less severe illness.

A different infection

But children have relatively small nasal passageways that can easily be blocked, so paediatric upper respiratory infections sometimes warrant extra attention compared with those in adults. Roberta DeBiasi, who heads the division of paediatric infectious diseases at the Children’s National Hospital in Washington DC, says that she and her colleagues have noticed an increase in the number of children with ‘COVID croup’, which is an inflammation of the upper airway that produces a characteristic ‘barking’ cough. That adds credence to the theory that Omicron might infect children differently from adults.
But Absoud says hospitals are well equipped to treat children for croup and other symptoms of upper respiratory infection, because viruses such as respiratory syncytial virus send children to hospital with the same symptoms every year.

Even if children generally recover from an acute infection with Omicron, clinicians still worry that they might develop long COVID, in which symptoms persist for months, or a rare but serious condition called multisystem inflammatory system in children (MIS-C). It’s too early to assess the effect of Omicron on long COVID symptoms in children, says Absoud, but MIS-C symptoms usually develop two to four weeks after infection.

“We would have started seeing the signal [for MIS-C] by now, and we haven’t seen it,” he says. That doesn’t mean we’re in the clear, Absoud adds, because the illness can take longer to develop. But it is an encouraging sign that there hasn’t yet been a wave of children hospitalized for the condition.

doi: https://doi.org/10.1038/d41586-022-00309-x

CEBR Vaccines and Related Biological Products Advisory Committee meeting, June 14th and 15th, 2022

The Food and Drug Administration (FDA) and the Center for Biologics Evaluation (CEBR) Vaccines and Related Biological Products Advisory Committee met on June 14th and 15th, 2022 to review the Emergency Use Authorization amendment requests for the Pfizer-BioNTech COVID-19 Vaccine formulated for use in children 6 months to age 5 AND Moderna COVID-19 Vaccine formulated for use in children 6 months to 6 years.

The 174th Meeting of the Vaccines and Related Biological Products Committee meeting on June 15th (open session of over 7 hours long) is available at this link.

While I encourage everyone to review all of the presentations and data, I know that it took me 3 days to find the time to do it. Below are a few highlights and I have added approximate time points so one can speed ahead to parts that they are most interested in learning about. For example the Open Public Hearing with comments from 19 individuals from the American public begins at 4 hours and 39 minutes. Even a member of Congress weighs in.

I look at getting our son vaccinated as similar to the process to prepare our son for school. The decision is a personal one, but the data is compelling. Just as children take time to learn the knowledge and memory needed for writing and recognizing their name, the adaptive immune system needs training for each new pathogen. Over time, recognizing letters and words (foreign RNA and novel spike protein sequences) in new patterns become easier. That said, Omicron is an especially tough lesson for children with no prior SARS-CoV-2 exposure or passive (maternal) immunity. Parents routinely vaccinate their children against influenza, varicella (chicken-pox), rubella, hepatitis A, and rotavirus. Imagine a disease where the deaths per year are higher than each of these combined in their respective pre-vaccine era. That is Omicron and the many Omicron subvariants.

The 3 dose Pfizer and 2 dose Moderna pediatric formulations require TIME to build an ample antibody and immune compartment. These studies were conducted during the Omicron period where a peak of 1. 4 million cases per day were reported in the US. Of the 45,000 children under 5 that have been hospitalized due to COVID-19 since the beginning of the pandemic, half of these were during the Omicron surge of winter 2022. 24% of hospitalizations required ICU-level care. 63% of children under 5 hospitalized due to SARS-CoV-2 had NO underlying conditions. With an Omicron-specific vaccine on the way (Moderna’s 1273.214 in trials now and bivalent vaccines will be discussed by this committee in 2 weeks) getting the first doses underway this summer will begin the immune system training against this quickly evolving coronavirus. Just as adults have had to be boosted when new strains come along, I expect additional doses to be added to these pediatric formulations.

I anticipate when vaccinations roll out there to be reports of febrile seizures after vaccination due to the common reaction of fever post vaccine doses reported in both trials. Febrile seizures happen in about 3% of children in this age population at baseline. There was one case of a 17 month old female that had a seizure and rash within days of her first injection of Moderna. Another viral infection was not ruled out. The care team advised and her parents permitted her to stay in the study and she had no additional neurological sequelae. There were no new safety signals seen in the youngest age group of patients enrolled in either study that were not seen in the adolescent or adult trials.

It is important to caution that these are relatively small studies (5000 to 6000 enrolled for each company) with a short follow-up time. New safety signals that are more rare may come up. More studies are needed and are ongoing. Co-administration with other pediatric vaccines has not been tested.

While not every family will make the choice to vaccinate, it is a great step forward to allow parents and caregivers the option to vaccinate in the hopes to prevent severe illness, Long Covid, hospitalizations, and even deaths due to COVID-19.

Meeting notes:

The committee comprised of the following VRBPAC members: Dr. Arnoldo Monto (Chair), Dr. Prabhakara Atreya, Dr. Adam Berger, Dr. Hayley Gans, Dr. Henry (Hank) Bernstein, Dr. Archana Chatterjee, Dr. Amanda Cohn, Dr. David Kim, Dr. Paul Offit, Dr. Steven (Steve) A. Pergam, Dr. Jay Portnoy, and Dr. Eric Rubin. Temporary voting members include Dr. Oveta Fuller, Dr. James Hildreth, Dr. Jeannette Lee, Dr. Ofer Levy, Dr. Wayne Marasco, Dr. Pamela McInnes, Dr. Cody Meissner, Dr. Michael (Mike) Nelson, Dr. Art Reingold, Dr. Mark Sawyer, and Dr. Melinda Wharton. Dr. Paula Annuziato is the non-voting industry representative. All members were found to be in compliance with Federal conflict of interest laws included in 18 U.S.C. § 208. Dr. James Hildreth has a conflict and was granted a waiver that is available here on the FDA website. (He is the President of the Meharry Medical college that will be involved in upcoming vaccine trials. Dr. Hildreth will not receive any remuneration from those upcoming trials.)

These physicians and scientists representing academic and non-profit research groups, industry, and hospitals were asked after hearing the data to answer two YES or NO questions:

“Based on the totality of scientific evidence available, do the benefits of the Moderna COVID-19 Vaccine when administered as a 2-dose series (25 mcg each dose) outweigh its risks for use in children 6 months through 5 years of age?”

“Based on the totality of scientific evidence available, do the benefits of the PfizerBioNTech COVID-19 Vaccine when administered as a 3-dose series (3 µg each dose) outweigh its risks for use in children 6 months to 4 years of age?”

The votes were

Moderna: 21 YES, with 0 NO votes, and no abstentions.
PfizerBioNTech: 21 YES, with 0 NO votes, and no abstentions.

Here are a few highlights and I have added approximate time points so one can speed ahead to parts that they are most interested in learning about.

26 minutes: Dr. Peter Marks highlighted the data from the MMWR March 18, 2022 report (Vol. 71. No. 11) remarking that the omicron wave brought significant pediatric hospitalizations in the youngest children.

42 minutes: Dr. Carla Vinals of Moderna reported that in terms of safety, immunogenicity, and efficacy results that no NEW safety concerns were identified and real-world efficacy was shown during the omicron period a few months ago.

46 minutes: Dr. Evan Anderson described the unmet medical need for this vaccine and debunked misperceptions about the burden of COVID-19 in children 0 to 4.

63% of children 0-4 that are hospitalized with COVID-19 have NO underlying medical conditions.

There have been 147 deaths due to COVID19 in the 0 to 4 year old population in the first 5 months of 2022 and educational pre-school and childcare settings have been extremely disrupted in this population due to the high burden of disease.

The take-home for me from his remarks are the number of deaths per year due to COVID-19 are higher than the pre-vaccination levels of death per year for influenza, varicella, rubella, hepatitis A, and rotavirus (ALMOST combined each year)!

53 minutes: Dr. Ritaparna (Rita) Das discussed Study 204, and the dose responses of 25 ug, or 50ug versus the saline placebo. This study period had a peak of 1.4 million cases per day in the United States, where the majority of patients were enrolled. Only one severe adverse event (SAE) was determined to be vaccine-related. A 17 month old female had a febrile seizure and had a rash shortly after vaccination. Upon follow-up it was determined that it was safe for the patient to remain in the study and she received a second dose with no adverse events after the second dose.

For local reactions, pain was the most common and was characterized as low grade (1 or 2 severity) but it was similar to placebo groups that received saline injections, especially in the 6 to 23 month age group. Systemic reactions were fatigue and fevers, but most less than 39 degrees C. Fever was also common in the placebo group as this study group was conducted during a peak viral illnesses in the winter months of 2022.

There were 4 infections over 40 degrees C or 104 degrees F. One fever and resulting febrile seizure was considered study-related and an SAE as I shared before. Two of the high fevers were attributed to viral illness and one fever was attributed to the patient’s history of an underlying fever-related syndrome. One child had uticaria (hives) after Dose 1 and was discontinued in the study as this is considered an Adverse Event and the child is not likely a good candidate for this vaccine.

The immunogenicity data compared Study 204 to Study 301, or the 18-25 year old age group receiving two 100 ug doses, 1 month apart. The 6 to 23 month group had a GMT of 1.28 and the 2-5 year old group had a GMT of 1.01 showing non-inferiority.

Vaccine efficacy (VE) had both a CDC definition and a 301 case definition.

CDC VE 2-5 year olds: 36.8%
Study 301 VE in 2-5 year olds: 46.4%

CDC VE 6 to 23 month olds: 50.6%
Study 301 VE in 6 to 23 month olds: 31.5%

Compared to real-world data generated through Kaiser Permanente, vaccination was 84.5% effective against HOSPITALIZATION in this population.

1 hour 31 minutes: Dr. Robin Wisch described the immunobridging data and efficacy results. 6% of the vaccine-arm and 7% of the children in the placebo group were seropositive for SARS-CoV-2 at baseline in the 6-23 month age group. 9% of the vaccine-arm and 8% of the placebo—arm were seropositive at baseline in the 2 to 5 year olds.

No events met the CDC criteria for probable or confirmed myocarditis or pericarditis through the data cutoff of February 21, 2022 (median 68 days follow-up post Dose 2) in the 6-23 month group or the 2 to 5 year olds.

2 hours 35 minutes: Dr. William (Bill) Gruber discussed the BNT162b2 vaccine data.

7% of study participants were SARS-CoV2 seropositive. There were no severe or Grade 4 reactions including:

NO vaccine-related anaphylaxis
NO myocarditis/pericardidis
NO Bell’s palsy
NO MIS-C

Immunogenicity results: Non-inferior compared to 16-25 year olds receiving 2 doses of 30 ug

GMR of 1.3 for 2-5 year age group and 1.19 for 6-23 month age group

Efficacy: (Taken with a grain of salt due to the large confidence intervals): 80.3% after the third dose
6-23 month age group: 75.5% and 2-5 year age group: 82.3%

3 hours 19 minutes: Dr. Susan Wollersheim, FDA CBER review of Effectiveness and Safety

Dose selection considerations were discussed.
Patients from the US, Finland, Poland, and Spain were enrolled.

6-23 months: n=1776
2-4 years: n=2750

The study was blinded with some cases of unblinding described, i.e. when a child turned 5 before the added Dose 3, a child would be unblinded to allow to leave the study to get vaccinated or not with the approved formulation.

Immunogenicity results: Non-inferior compared to 16-25 year old age group receiving 2 doses of 30 ug

GMR of 1.3 for 2-5 year age group and 1.19 for 6 to 23 month age group

Used: USA WA1/2020 (Reference), B.1.617.2 Delta variant, and B.1.1.529 Omicron variant ratios at 1 month post Dose 3 versus Dose 2:

Efficacy (Taken with a grain of salt due to the large confidence intervals):
Two weeks after Dose 3 (Roughly 4 months after Dose 1):
6 to 23 month age group: 75.6%
2-5 year old age group: 82.4%

LIMITED by small numbers and short follow-up time.

4 hours 39 minutes: Open Public Hearing

Speakers: Jasmine King, Dr. Ashley Serrano, Michael Baker, Fatima Khan, Nicholas Giglia, Lauren Dunnington, Kathlyn Kinesley, Melissa Braveman, Congressman Louie Gohmert, Dr. Harvey Klein, Dr. Kailey Soller, Shae Lynn, Kate Schenk, Tamara Thomson, Sam Dodson, Donna Treubig, Catharine Diehl, Jessica Nehring, Katarina Lindley, and Caroline Bishop.

5 hours 46 minutes: Q&A Moderna Panel

The panel responded to questions about the choice of 25 ug vs. 50 ug. While the response was dose-dependent, the safety advisors recommended the lower dose to balance the safety and still achieve non-inferior immunogenicity.

6 hours 4 minutes: Safety Data

6 hours 11 minutes: Moderna Vote Discussion

7 hours: Q&A Pfizer

7 hours 7 minutes: Pfizer Voting Discussion

Dr. Paul Offit asked about the data following Dose 2 and shared his concerns of low to no visible change in the curves. Namely, his fear is the 3ug may be under dosed and that parents should be aware that life does not go back to normal after Dose 1 or 2, but rather months later after Dose 3.

Others echoed his concerns.

7 hours 19 minutes: Pfizer/BioNTech Vote

7 hours 23 minutes: Vote explanations

Dr. Archana Chatterjee describes the evolution of her confidence in this technology and WHY she was a NO vote in December of 2020 for the 1st mRNA vaccine. She was not alone as 4 voted NO then and wanted more data before providing an EUA.

B.1.1.529 Reported in vitro Therapeutic Activity, Update for June 13, 2022

New data is available on the NIH website here that compiles in vitro data with SARS-CoV-2 variants with regard to vaccines, antibodies, antivirals, and convalescent plasma. This interactive tool allows any individual to look for the parameters of their interest, including what has been published in the last seven days.

New data downloaded today for the reported in vitro Therapeutic Activity of Omicron or B.1.1.529 and the Omicron subvariants.

Sudden rise of more transmissible form of Omicron

The journal Science has published an article by Dr. Meridith Wadman, a staff writer at Science who has reported on the intersection of biomedical sciences and politics in various capacities for over 20 years.

The title of her article is “Sudden rise of more transmissible form of Omicron catches scientists by surprise.”

Nothing about viruses should take us as surprise. While I disagree with her selected title, I think the content of her article speaks to the amazing success of the Omicron branches. BA.1 or 21K is the dominant strain of Omicron circulating in the United States and many other countries. The BA.2 or 21L variant is now the dominant strain in Denmark (82%) and India, now detected in 57 countries but makes up only 4% of the total of specimens sampled and sequenced worldwide. 21L has been detected in over half of the US states and is the variant in 8% of samples sequenced to date. As you can see from the phylogeny tree below, the 21L is as far from 21K as each of the Alpha, Beta, and Gamma variants. Some scientists are calling for the Omicron 21L to be given a new name to illustrate this genetic diversity. As you can also appreciate the Delta strains 21J, 21A, and 21I are also far from Omicron on the phylogenetic tree. Just as people infected with an Alpha variant and later came down with Delta, and those that had Delta last summer and fall were infected in December and January with Omicron 21K, let us NOT be surprised if people infected with Omicron 21K also become infected with the 21L variant in the months to come. Fortunately many of those with vaccine-induced and boosted immunity and those previously infected will hopefully have “mild” disease. As a reminder a “mild” case by definition means that one will not require an urgent care visit or a hospital stay and does not reflect how terrible you may feel!

As this depiction of the SARS-CoV-2 evolutionary tree suggests, the BA.1 and BA.2 strains of the Omicron variant are about as genetically distinct as earlier variants Alpha, Beta, and Gamma are from each other.
Diagram credit: https://nextstrain.org/ncov/gisaid/global

Published: 31 JAN 2022 BY MEREDITH WADMAN

On 7 December 2021, as the Omicron variant of the pandemic coronavirus began to pummel the world, scientists officially identified a related strain. BA.2 differed by about 40 mutations from the original Omicron lineage, BA.1, but it was causing so few cases of ­COVID-19 that it seemed a sideshow to its rampaging counterpart.

“I was thinking: ‘BA.1 has the upper hand. We’ll never hear again from BA.2,’” recalls Mark Zeller, a genomic epidemiologist at the Scripps Research Institute. Eight weeks later, he says, “Clearly that’s not the case. … I’m pretty sure [BA.2] is going to be everywhere in the world, that it’s going to sweep and will be the dominant variant soon in most countries if not all.”

Zeller and other scientists are now trying to make sense of why BA.2 is exploding and what its emergence means for the Omicron surge and the pandemic overall. Already a U.K. report issued last week and a large household study from Denmark posted this week as a preprint make it clear BA.2 is inherently more transmissible than BA.1, leaving scientists to wonder which of its distinct mutations confer an advantage.

But so far, BA.2 does not appear to be making people sicker than BA.1, which itself poses less risk of severe disease than variants such as Delta and Beta. In Denmark, where by 21 January BA.2 accounted for 65% of new COVID-19 cases, “We see a continuous, steep decline in the number of intensive care unit patients and … now a decrease in the number of hospital admissions related to SARS-CoV-2,” says Tyra Grove Krause, an infectious disease epidemiologist at the country’s public health agency. In fact, the Danish government is so confident the variant won’t cause major upheaval that it lifted almost all pandemic restrictions on 1 February.

Still, some scientists predict BA.2 will extend Omicron’s impact. “I would guess we’ll see [BA.2] create a substantially longer tail of circulation of Omicron than would have existed with just [BA.1], but that it won’t drive the scale of epidemics we’ve experienced with Omicron in January,” computational biologist Trevor Bedford of the Fred Hutchinson Cancer Research Center tweeted on 28 January. In South Africa, BA.2 already may be stalling the rapid decline in new infections seen after the country’s Omicron wave peaked in December 2021.

Although BA.2 represented less than 4% of all Omicron sequences in the leading global virus database as of 30 January, it has been identified in 57 countries, with the earliest documented case dating to 17 November in South Africa. It likely now dominates in India, according to Bijaya Dhakal, a molecular biologist at the Sonic Reference Laboratory in Austin, Texas, who examined sequence data uploaded from eight large Indian states. In the United Kingdom, the proportion of likely BA.2 cases doubled from 2.2% to 4.4% in the 7 days that ended on 24 January.

In the United States, the Centers for Disease Control and Prevention is not yet tracking BA.2 separately. But Bedford estimates it accounted for 7% of new U.S. cases as of 30 January, up from 0.7% on 19 January. “In each country and across time, we see that the epidemic growth rate of Omicron BA.2 is greater than Omicron BA.1,” he says.

The report last week from the UK Health Security Agency (UKHSA) backs up that assessment in England, finding BA.2 was spreading faster than BA.1 in all regions where enough data were available to make an assessment. UKHSA data also show that in late December 2021 and early January, transmission was higher among household contacts of BA.2 cases, at 13.4%, than in contacts of other Omicron cases (10.3%).

The study from Denmark, which sequences the virus from virtually every person who gets COVID-19, paints a more dramatic picture. In households where the first case was BA.1, on average 29% of other people in the household became infected. When the first case was BA.2, 39% of household members were infected.

Omicron was already known to have mutations that help it evade antibodies, but the Danish researchers also found that BA.2 may be even better at dodging vaccine-induced immunity: Vaccinated and boosted people were three times as susceptible to being infected with BA.2 as with BA.1. Vaccinated but unboosted people were about 2.5 times as susceptible, and unvaccinated people 2.2 times as susceptible. Early U.K. data, however, showed vaccinated people, if boosted, had about the same level of protection against symptomatic infections with BA.1 or BA.2—63% and 70%, respectively.

In one hopeful and unexpected finding from Denmark, those who were vaccinated or vaccinated and boosted passed on BA.2 to household members less often, relative to BA.1. The same didn’t hold for unvaccinated people, who passed BA.2 to their household contacts at 2.6 times the rate they passed BA.1.

Much as scientists a few weeks ago wondered whether a previous infection with Delta or another variant would protect people from Omicron overall, some are now looking for data on whether Omicron’s first surge created a shield against BA.2. “To what extent does a BA.1 infection protect you against reinfection with BA.2?” Zeller asks. “From what I have seen in Denmark, it’s not going to be 100%.”

Scientists are also probing the variant’s ability to dodge vaccine-induced antibodies in lab dish studies. And drugmaker GlaxoSmithKline is testing its monoclonal antibody, sotrovimab, made with Vir Biotechnology, against BA.2 in lab studies. It’s the only widely authorized antibody that still thwarts BA.1.

Scientists note BA.1 and BA.2 are about as far apart on the evolutionary tree as earlier variants of concern—Alpha, Beta, and Gamma—are from each other (see graphic). Some even think BA.2 shouldn’t even be considered Omicron. “I hope in the near future that BA.2 gets its own variant of concern [label] because people assume it’s very similar which it’s not,” Zeller says.

BA.2 doesn’t have all of the mutations that help BA.1 avoid immune detection, but it has some its sibling doesn’t. Thomas Peacock, a virologist at Imperial College London, notes that most of the differences are in an area of the spike protein, called the N-terminal domain (NTD), that houses antibody targets. “What we don’t know is: Just because there are changes, are they changes that actually do something?” says Emma Hodcroft, a molecular epidemiologist at the University of Bern.

But one NTD difference—a deletion at amino acids 69 and 70 that is present in BA.1 and not in BA.2—could give researchers a tool for monitoring the spread of the up-and-coming Omicron strain. Certain SARS-CoV-2 polymerase chain reaction tests detect three genetic sequences of the virus, but the mutation in BA.1’s NTD gene eliminates one of those targets. Polymerase chain reaction tests pick up all three targets in BA.2, providing a proxy for distinguishing the Omicron strains if there is no full virus sequence. How the sibling strains were born is also preoccupying scientists. Viral evolution in a single immunocompromised patient is one theory, says Andrew ­Rambaut, an evolutionary biologist at the University of Edinburgh. “It’s possible that long-term infection could produce quite a lot of diversity within a single individual. It could be compartmentalized. So different variants living in different parts of the body.” Both Omicron strains could have also evolved in animals infected with human-adapted SARS-CoV-2, then spread back into people.

Why BA.2 is emerging only now is one more mystery, Hodcroft says. She speculates that a factor as simple as which Omicron caught an earlier flight out of South Africa, where both strains were first identified, may be the explanation. “BA.2 may have just been trapped for a little bit longer. But when it did finally get out and start spreading it started to show that it can edge out its big sister.”

With reporting by Kai Kupferschmidt.

Update, 1 February, 5:05 p.m.: This story has been updated.

doi: 10.1126/science.ada0810

B.1.1.529 Reported In Vitro Therapeutic Activity, Update For January 5, 2022

New data is available on the NIH website here that compiles in vitro data with SARS-CoV-2 variants with regard to vaccines, antibodies, antivirals, and convalescent plasma. This interactive tool allows any individual to look for the parameters of their interest, including what has been published in the last seven days.

Specifically for B.1.1.529, or the Omicron variant, the medical community has lost the majority of the therapeutic antibody treatments.

As the Nature article from December 21, 2021 reports:

But when virologists saw that Omicron has a multitude of mutations concentrated on its spike protein, they feared what it would mean for these treatments. The outcome was even worse than they anticipated, says Olivier Schwartz, a virologist at the Pasteur Institute in Paris and a co-author of one of the preprints 3. “We didn’t expect to see such a shift in the antibodies’ effectiveness,” he says.

”Sotrovimab is the best of the lot. Even so, the concentration required to halve viral replication was roughly three times higher for Omicron than for other coronavirus variants. Although sotrovimab’s drop in potency against the new variant is significant, says Stuart Turville, a virologist at the Kirby Institute in Sydney, Australia, and a co-author of one of the preprints2, “it’s nothing like what we saw for the others”. That might be because sotrovimab targets a part of the spike protein that is unchanged across many related coronaviruses.

While two recently Emergency Use Authorized oral antiviral therapies stop Omicron replication in vitro, don’t hang your hat on receiving them. There are very specific populations who are eligible to receive both the Merck product, molnapiravir, and the Pfizer combination therapy. These antivirals are most potent when used in the first days of infection, or the window of time that many people do not have fulminant symptoms and therefore do not seek testing or treatment.

https://opendata.ncats.nih.gov/variant/activity

A new tool to probe SARS-CoV-2 variants

Christmas came early for me this year. My exciting gift: A new tool in the scientific arsenal to study SARS-CoV-2 variants using Virus-Like-Particles (VLPS).

This novel technology and tool developed in the laboratory of Dr. Vineet D. Menachery allows scientists to study variant characteristics quickly and without studying actual viruses in a biosafety level 3 (BSL3) laboratory which require scientists to have specific facilities and training to conduct the studies as safely as possible. More about Dr. Vineet D. Menachery can be viewed here.

A description of the technology was published in the journal Science on December 23, 2021 and is available below.

BRYAN A. JOHNSON AND VINEET D. MENACHERY

SCIENCE • 23 Dec 2021 • Vol 374, Issue 6575 • pp. 1557-1558 • DOI: 10.1126/science.abn3781

Although efforts have been made to understand the biology of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a major focus has been on investigating genetic variation in the virus. However, progress is hampered by the need to perform experiments involving SARS-CoV-2 in biosafety level 3 (BSL3) laboratories, which require substantial training for safe operation. On page 1626 of this issue, Syed et al. (1) offer an alternative to using live virus, introducing a new SARS-CoV-2 virus-like particle (VLP) system. The authors innovate on previous VLP systems by incorporating a reporter construct to study infection (1, 2). Illustrating the system’s utility, they use VLPs to characterize mutations in SARS-CoV-2 variants of concern.

SARS-CoV-2 VLPs are created by expressing the four structural proteins—spike, membrane, envelope, and nucleocapsid—in a packaging cell line (2). Upon expression, VLPs consisting of these four proteins and a lipid membrane self-assemble and are released from the cell (3). Despite resembling SARS-CoV-2 morphologically, traditional VLPs cannot be used to study the effect of a mutation on fitness because they lack genetic material to deliver to target cells (2). Syed et al. introduced a key innovation. They first identified the SARS-CoV-2 packaging signal, a genetic marker used to identify full-length genomes for packaging into the virion (4). This packaging signal was incorporated into the 3’ untranslated region of a luciferase reporter plasmid, causing the resulting transcripts to be packaged within VLPs. Syed et al. show that VLPs deliver these luciferase reporters to target cells, allowing the resulting signal to be used as a proxy for SARS-CoV-2 infection. Thus, the effects of particular mutations on the strength of the luciferase signal can be used to determine modulation of SARS-CoV-2 infection (see the figure).

In the broader context of studying SARS-CoV-2 genetic variation, VLPs represent a middle ground between two commonly used methodologies: infectious clones and pseudovirus vectors. SARS-CoV-2 infectious clones are the gold standard because they create recombinant virus, incorporating mutations anywhere in the genome (5). However, using SARS-CoV-2 infectious clones is technically challenging and creates live SARS-CoV-2 that requires BSL3 laboratories for study. This limits the use of SARS-CoV-2 infectious clones to laboratories with access to such facilities and willingness to invest in developing a specialized skill set.

Pseudovirus systems are the leading alternative to using SARS-CoV-2 infectious clones. In these systems, SARS-CoV-2 spike protein is expressed in cells along with a noncoronavirus packaging system and a reporter gene, with the most common being lentivirus-based (6). Like the VLPs developed by Syed et al., pseudoviruses self-assemble, incorporating spike proteins on their surface and packaging reporter messenger RNA (mRNA) (6). The primary advantage of pseudovirus systems is their ease of use, allowing rapid analysis of spike mutations. Pseudoviruses can be generated in the widely available 293T cell line by simply expressing a small number of proteins (6). Additionally, because pseudoviruses replace replication genes, they do not undergo continued amplification in target cell lines (6). This makes them safe to use in BSL2 laboratories, which are available to most researchers. However, the only SARS-CoV-2 protein incorporated into pseudoviruses is spike. Because substantial genetic variation occurs outside of spike, the pseduovirus systems have limited applicability to study SARS-CoV-2 variants.

The SARS-CoV-2 VLPs used by Syed et al. offer researchers several advantages over pseudoviruses. Rather than relying on the packaging machinery of another virus, VLPs use SARS-CoV-2 proteins and recapitulate packaging, assembly, and release, as occurs in genuine virus infection (3). In principle, this allows the effects of variant mutations on these processes to be studied. Similarly, because all four structural proteins are incorporated into SARS-CoV-2 VLPs, additional genetic variation can be captured. Like pseudoviruses, VLPs do not undergo subsequent rounds of replication, allowing them to be used safely in BSL2 laboratories.

Illustrating the utility of SARS-CoV-2 VLPs, Syed et al. characterized several nucleocapsid mutations. SARS-CoV-2 nucleocapsid is a hotspot for coding mutations, particularly within its serine-rich (SR) motif (7, 8). Although its exact function is unclear, the SR motif has many phosphorylated amino acids and is located within a region of intrinsic structural disorder (7, 9, 10). Using their SARS-CoV-2 VLP system, Syed et al. analyzed the effects of several common nucleocapsid mutations and found that several enhanced infection, including those present in the Alpha, Gamma, and Delta variants (7). These data are consistent with findings using SARS-CoV-2 infectious clones (1, 11).

The finding that nucleocapsid mutations enhance SARS-CoV-2 infection has important implications. To date, most studies of SARS-CoV-2 genetic variation have focused on spike (12). This is understandable, because spike binds to the host cell receptor angiotensin-converting enzyme 2 (ACE2), and is thus the primary determinant of infection (13). Additionally, because spike is the target of available vaccines, determining if mutations affect protection is a pressing question (12). However, recent studies suggest that nucleocapsid mutations lead to enhanced virulence and fitness, highlighting the need to characterize genetic variation elsewhere in the viral genome (11). Because SARS-CoV-2 VLPs recapitulate enhancement of infection by these nucleocapsid mutations, they can be used to characterize mutations in emerging variants, such as deletion of amino acids 31 to 33 in the nucleocapsid protein of the Omicron variant.

Although a promising platform, there are limitations of this SARS-CoV-2 VLP system. Only the four structural proteins are present. Thus, like pseudoviruses, the scope of variation that can be captured is limited. For example, variant mutations in the viral replication machinery cannot be examined with VLPs (14). Additionally, while allowing for safe use in BSL2 laboratories, the inability of VLPs to undergo continued replication makes them unsuitable to study virulence or transmission. Furthermore, although data presented by Syed et al. suggest that enhancement of infection by VLPs and live SARS-CoV-2 are correlated, additional work is needed to determine how closely VLPs model infection. As SARS-CoV-2 evolves, it is critical that the effects of new mutations are characterized.

Acknowledgments

V.D.M. is funded by the National Institutes of Health and National Institute of Allergy and Infectious Diseases (grants AI153602, 1R21AI145400, and R24AI120942). V.D.M. has filed a patent on the reverse genetic system and reporter SARS-CoV-2.

References and Notes

1. A. M. Syed et al., Science 374, 1626 (2021).

CROSSREF=

2. H. Swann et al., Sci. Rep. 10, 21877 (2020).

CROSSREF

PUBMED

GOOGLE SCHOLAR

3. P. S. Masters, Adv. Virus Res. 66, 193 (2006).

CROSSREF

PUBMED

GOOGLE SCHOLAR

4. P.-K. Hsieh et al., J. Virol. 79, 13848 (2005).

GO TO REFERENCECROSSREFPUBMEDGOOGLE SCHOLAR

5. X. Xie et al., Cell Host Microbe 27, 841 (2020).

GO TO REFERENCECROSSREFPUBMEDGOOGLE SCHOLAR

6. M. Chen, X.-E. Zhang, Int. J. Biol. Sci. 17, 1574 (2021).

CROSSREF

PUBMED

GOOGLE SCHOLAR

7

Q. Ye et al., Protein Sci. 29, 1890 (2020).

CROSSREF

PUBMED

GOOGLE SCHOLAR

8. J. A. Plante et al., Cell Host Microbe 29, 508 (2021).

GO TO REFERENCECROSSREFPUBMEDGOOGLE SCHOLAR

9. M. Bouhaddou et al., Cell 182, 685 (2020).

GO TO REFERENCECROSSREFPUBMEDGOOGLE SCHOLAR

10. A. D. Davidson et al., Genome Med. 12, 68 (2020).

GO TO REFERENCECROSSREFPUBMEDGOOGLE SCHOLAR

11. B. A. Johnson et al., bioRxiv2021).

CROSSREF

GOOGLE SCHOLAR

12. W. T. Harvey et al., Nat. Rev. Microbiol. 19, 409 (2021).

CROSSREF

PUBMED

GOOGLE SCHOLAR

13. C. B. Jackson et al., Nat. Rev. Mol. Cell Biol. (2021).

GO TO REFERENCE

14. J. L. Mullen et al., https://outbreak.info/ (2020).

GO TO REFERENCE

COMIRNATY Booster Doses

The FDA shared the briefing document application for licensure of a booster dose for COMIRNATY (COVID-19 Vaccine, mRNA), available here. As of today, only 17% of Americans have raised their sleeve for a booster. COMIRNATY is the only FDA approved COVID-19 Vaccine. It has been widely studied and the safety profile proves it to be the most researched and safest of all COVID-19 vaccines used in the world and more studied and scrutinized than the EUA monoclonal antibody treatment, Regeneron. COMIRNATY is the booster my parents received and the booster Brad and I received. Speak with your doctor about your safety concerns. Here is some more to read to make the most informed choice.

As of today the FDA now recommends COVID-19 booster shots to ALL adults to mitigate the waning of antibody titers against the spike protein in some twice- mRNA vaccinated individuals and the sad reality of rising coronavirus cases in many parts of the world.

The 23-page document includes Section 6.3. for the proposed use of a COMIRNATY booster dose with Section 6.4 sharing the FDA review of clinical data from Study C4591001.

Study C4591001 is described below:

“Design Study C4591001 is ongoing. The study was initially designed to evaluate two vaccine candidates and several dosages in healthy adults in the United States (Phase 1), of which 24 participants (n=12 per age group: 18-55 years and 65-85 years) received a 2-dose primary series of BNT162b2 (30 µg); the study population included healthy men and women and excluded participants at high risk of SARS-CoV-2 infection or with serological evidence of prior or current SARS-CoV-2 infection.

The Phase 2/3 portion of the study is being conducted in the United States, Argentina, Brazil, Germany, South Africa and Turkey. Please see the Summary Basis for Regulatory Action for the approval of a 2-dose primary series of COMIRNATY for study design details. 1 Enrolled Phase 2/3 participants were initially stratified by age (18-55 years and >55 years), with the goal of older adults (>55 years of age) comprising 40% of the total study population. The protocol was later amended to include adolescents 16 and 17 years of age. The study population included participants at higher risk for acquiring COVID-19 and at higher risk of severe COVID19, such as health care workers, participants with autoimmune disease, and participants with chronic but stable medical conditions such as hypertension, asthma, diabetes, and infection with HIV, hepatitis B or hepatitis C. Participants were randomized 1:1 to receive two doses of either BNT162b2 or saline placebo 3 weeks apart. Per protocol, since December 14, 2020, following issuance of the Emergency Use Authorization for the Pfizer-BioNTech COVID-19 Vaccine, Phase 2/3 participants ≥16 years of age in the vaccine and placebo groups were progressively unblinded to their treatment assignment (when eligible for vaccination per local recommendations), and participants originally randomized to placebo were offered vaccination with BNT162b2 under the study protocol with continuing follow-up for safety and COVID-19- related outcomes. In February and March 2021, the protocol was amended to evaluate the safety and immunogenicity of booster dose of BNT162b2 in Phase 1 participants (N=12 per age cohort: 18- 55 years and 65-85 years) and a subset of Phase 2/3 adults (N=300, ages 18-55 years), who completed the 2-dose primary vaccination series with 30 µg BNT162b2. A booster dose of 30 µg BNT162b2 was administered approximately 7 to 9 months after a primary series for Phase 1 participants and approximately 6 months after a primary series for Phase 2/3 participants.

Immunogenicity evaluation

The effectiveness of the booster dose is based on an immunobridging analysis from the Phase 2/3 booster participants comparing 50% neutralizing antibody titers against the reference strain (recombinant USA-WA1/2020) at 1 month after the booster dose to those observed at 1 month post-primary series among subjects without evidence of prior SARS-CoV-2 infection. Immunobridging analyses included hypothesis testing for:

GMTs of SARS-CoV-2 neutralizing antibodies at 1 month after the booster dose vs. those values 1 month after a primary series, using a 1.5-fold non-inferiority margin as the success criterion for the lower bound of the confidence interval around the geometric mean ratio (GMR), and percentage of participants with seroresponse (≥4-fold rise from baseline) at 1 month after the booster dose vs. 1 month after a primary series, using a -10% non-inferiority margin as the success criterion for the lower bound of the confidence interval around the difference between seroresponse rates. In the protocol-specified analysis of seroresponse, the baseline neutralizing antibody titer for determining seroresponse to the booster dose was the pre-Dose 1 titer (same baseline titer as used for determining seroresponse to the primary series). However, FDA also asked Pfizer to conduct a post hoc seroresponse analysis using the pre-booster dose titer as the baseline for determining the booster dose seroresponse (defined as ≥4-fold increase from the pre-booster dose baseline titer). Exploratory analyses of neutralizing antibody titers elicited by the BNT162b2 primary series and a 30 µg BNT162b2 booster dose against the reference strain (Wuhan) of SARS-CoV-2 and the

Beta and Delta variants were performed using samples from the Phase 1 study population. Discussion of these exploratory analyses in this briefing document is focused on the Delta variant, since it is currently the predominant circulating variant in the US. Safety evaluation Phase 1 participants and Phase 2/3 participants recorded reactogenicity assessments and antipyretic/pain medication use from Day 1 through Day 7 after booster in an e-diary. Reactogenicity assessments included solicited injection site reactions (pain, redness, swelling) and systemic AEs (fever, fatigue, headache, chills, vomiting, diarrhea, new or worsened muscle pain, and new or worsened joint pain). Other safety assessments included: AEs occurring within 30 minutes after each dose, non-serious unsolicited AEs from Dose 1 through 1 month after the booster dose, and serious AEs (SAEs) from the booster dose to the data cut-off date of June 17, 2021 (Phase 2/3) or May 13, 2021 (Phase 1). Analysis populations pertaining to the 30 µg BNT162b2 booster dose

Safety: All randomized participants who received a booster dose of 30 µg BNT162b2. Analyses of reactogenicity endpoints were based on a subset of the safety population that included participants with any e-diary data reported after vaccination.

All-available immunogenicity: All participants who received a primary series of 30 µg BNT162b2 at initial randomization, received a booster dose of 30 µg BNT162b2, and had at least 1 valid and determinate immunogenicity result after the booster dose.

Evaluable immunogenicity: All eligible participants who received a primary series of 30 µg BNT162b2 as initially randomized, with Dose 2 received within 19-42 days after Dose 1, received a booster dose of 30 µg BNT162b2, had at least 1 valid and determinate immunogenicity result after the booster dose from a blood collection within 28-42 days after the booster dose, and had no other important protocol deviations as determined by the clinician.

6.4.2. Demographics and disposition

Demographic characteristics of the Phase 1 and Phase 2/3 study participants who received a BNT162b2 (30 µg) booster dose are summarized in Table 2 below. Booster recipients were predominantly White. Phase 1 excluded individuals with comorbidities that confer risk for severe COVID-19 (i.e., obesity, diabetes with or without complications, chronic pulmonary disease, cardiovascular conditions such as hypertension, congestive heart failure, ischemic heart disease, HIV). Approximately 20% of booster recipients in Phase 2/3 had such comorbidities.

Among the 24 Phase 1 study participants randomized to the BNT162b2 primary series, 23 participants received a BNT162b2 (30 µg) booster dose (11 adults ages 18-55 years and 12 adults ages 65-85 years). One participant in the 18-55 year-old cohort declined to receive a BNT162b2 booster dose. All 23 Phase 1 participants who received the booster dose were included in the safety analyses, and the booster dose evaluable immunogenicity population. The disposition of Phase 2/3 study participants who received a BNT162b2 (30 µg) booster dose is summarized in Table 3 below.

Vaccines and Related Biological Products Advisory Committee Meeting October 26, 2021 Briefing Document

The Food and Drug Administration (FDA) and the Center for Biologics Evaluation (CEBR) Vaccines and Related Biological Products Advisory Committee met on October 26, 2021 to, in part, review the Emergency Use Authorization amendment request for Pfizer-BioNTech COVID-19 Vaccine for use in children 5 through <12 years of age.

The 170th Meeting of the Vaccines and Related Biological Products Committee full meeting (open session of over 8 hours long) is available at this link.

The committee comprised of Dr. Arnoldo Monto, Dr. Amanda Cohn, Dr. Hayley Gans, Dr. Michael Kurilla, Dr. Cody Meissner, Dr. Paul Offit, Dr. Steve A. Pergam, Dr. Oveta Fuller, Dr. James Hildreth, Dr. Jeannette Lee, Dr. Patrick (Pat) Moore, Dr. Michael (Mike) Nelson, Dr. Stanley Perlman, Dr. Jay Portnoy, Dr. Eric Ruben, Dr. Mark Sawyer, Dr. Melinda Wharton and the non-voting industry representative Dr. Paula Annuziato were in compliance with Federal conflict of interest laws included in 18 U.S.C. § 208. These physicians and scientists representing academic and non-profit research groups, industry, and hospitals were asked after hearing the data to answer one question:

“Based on the totality of scientific evidence available, do the benefits of the PfizerBioNTech COVID-19 Vaccine when administered as a 2-dose series (10 µg each dose, 3 weeks apart) outweigh its risks for use in children 5-11 years of age?”

The vote was 17 YES, with 0 NO votes, and 1 abstention.

While I encourage everyone to review all of the presentations and data, here are a few highlights:

Dr. Doran Fink introduced the latest data of the US burden of disease due to COVID-19 including 1.9 million reported cases in children 5 to 11. The most recent data (week ending October 10th) shows that while children age 5 to 11 make up 8.7% of the population, they were accounting for 10.6% of all COVID-19 cases, the highest of any age group.

Dr. Ram Naik describes the change from the PBS buffer in the formulation to a TRIS buffer. This allows for a 10 week storage at 2 to 8 degree C storage versus using -20 to -80 degree storage.

Dr. Fiona Havers describes the epidemiology and risk factors in the pediatric population, noting that COVID-19 is the now the 8th leading cause of death in children. 39% of MIS-C cases have occurred in children 6 to 11 years old as well as the high levels of seroprevalence in their surveillance efforts of 50,000 samples in children every two weeks. Of note, 42% of samples from 5 to 11 year old-children in the May-June 2021 time period were positive for antibodies to the nucleoprotein. Perhaps most disturbing was her data on Post-COVID conditions in children that have a myriad of symptoms weeks and months after the infection. In addition, Dr. Havers described the closures affecting 2074 schools from August 2, 2021 to October 8, 2021 due to large scale outbreaks mostly in Texas, Georgia, and Kentucky. This has affected over 1 million children and their families in a two month period.

Dr. Matthew Oster (begins at 1 hour and 28 minutes) shares the prevalence data of myocarditis in the pre-COVID era, prevalence of myocarditis seen in individuals after SARS-CoV-2 infection, and review of myocarditis cases after COVID-19 vaccination. Dr. Oster is careful to describe how myocarditis differs in “classical” myocarditis, MIS-C myocarditis, post-COVID-19 myocarditis, and post- mRNA vaccine myocarditis in terms of measures of troponin, lymphocytes, C-reactive protein, platelets, and Echocardiogram ejection fraction recovery time. The data presented includes a review of 169,740,953 mRNA vaccine doses in males and 193,215,313 doses in females. Dr. Oster describes the 877 cases that met case definitions that result in an estimated 54 excess cases of myocarditis per one million adults fully vaccinated. Dr. Oster emphasized that children with acute myocarditis do well over time, while adults have more difficulty recovering than children.

The full data for the EUA amendment is available for download here.

The study had 2 cohorts. Cohort 1 was evaluated in terms of safety and immune response (in vitro immune assay responses) had 1,528 BNT162b2 10 µg participants and 757 placebo participants. Safety data included solicited ARs, unsolicited AEs, SAEs and AEs of clinical interest were assessed in a total of 2,268 (1,518 10 µg BNT162b2, 750 placebo) participants 5-11 years of age; 95% of participants in each study group completed at least 2 months of safety follow-up after Dose 2. Comparator group for immunogenicity: The comparator group for immunobridging analyses consisted of 300 evaluable participants 16-25 years of age who received both doses of BNT162b2 30 µg and were randomly selected from study C4591001 Phase 2/3.

In the Phase 2/3 safety expansion, a second cohort of 1,598 participants were randomized to receive BNT162b2 and 796 were randomized to placebo. At the time of the October 8, 2021 cutoff, most participants (98.7%) had received both Dose 1 and Dose 2. Seven participants in the BNT162b2 group did not receive vaccine, for a Safety Population of 1,591. One participant in the BNT162b2 group discontinued from the vaccination period due to AEs of pyrexia and neutropenia that worsened from baseline (see Section 7.6.7, AEs leading to withdrawal). Two participants (0.1%) in the BNT162b2 group withdrew from the study before the 1 month period. Neither withdrawal was due to an AE.

The following AEs were considered Grade 3 in severity: 1 tic, 1 rash (bilateral pleomorphic light eruption on arms). No Grade 4 (life-threatening AEs) were observed in the study. In Cohort 2, lymphadenopathy was reported in 6 (0.4%) vaccine recipients and 3 placebo recipients (0.4%).

Table 6 shows the neutralizing antibody levels comparing 264 samples from vaccinated Age 5-11 to 253 samples of the vaccinated Age 16-25 from study C4591001. The GMT values were an average of 1197.6 and 1146.5, respectively for a ratio of 1.04, showing non-inferiority.

Table 7 shows the same rate of neutralization in those two cohorts. Table 8 shows the difference of the Age 5-11 samples in plaque reduction assays comparing the delta variant strain to the reference strain, recombinant USA_WA1/2020. The GMT for the delta variant was 294, versus 365.3 for the reference strain.

I have included the risk analysis scenarios posed below for the potential of myocarditis/pericarditis post two-doses of 10 micrograms of mRNA vaccine, 3 weeks apart, per 1 million fully-vaccinated 5 to 11 year old children compared to the respective benefit of COVID-19 cases averted taking into account various outbreak scenarios.

For example, using this model in Scenario 2, if children age 5 to 11 become vaccinated and the United States experiences transmission levels like the delta wave we had in late August, at least 54,345 COVID cases in the vaccinated children, and 250 hospitalizations (average of 5 day hospital stay) will be avoided but with the potential of 92 mRNA-vaccine induced myocarditis cases that require a 1 day hospital stay during the same time period.

Myocarditis risks by age and sex.

Scenario 1: COVID-19 incidence as of September 11, 2021, VE 70% vs. COVID-19 cases and 80% vs. COVID-19 hospitalization.

Scenario 2: COVID-19 incidence at peak of U.S. Delta variant surge at end of August 2021, VE 70% vs. COVID-19 cases and 80% vs. COVID-19 hospitalization.

Scenario 3: COVID-19 incidence as of nadir in June 2021, VE 70% vs. COVID-19 cases and 80% vs. COVID-19 hospitalization.

Scenario 4: COVID-19 incidence as of September 11, 2021, VE 90% vs. COVID-19 cases and 100% vs. COVID-19 hospitalization.

Scenario 5: COVID-19 case incidence as of September 11, 2021, VE 70% vs. COVID-19 cases and 80% vs. COVID-19. hospitalization, COVID-19 death rate 300% that of Scenario 1.

Scenario 6: COVID-19 incidence as of September 11, 2021, VE 70% vs. COVID-19 cases and 80% vs. COVID-19 hospitalization, excess myocarditis cases 50% of Scenario 1.

6-month neurological and psychiatric outcomes in 236 379 survivors of COVID-19: a retrospective cohort study using electronic health records

The Lancet Psychiatry
Published: April 06, 2021

DOI: https://doi.org/10.1016/S2215-0366(21)00084-5

Raw Data: https://osf.io/7tzvy/wiki/home/

Figure 1 Kaplan-Meier estimates for the incidence of major outcomes after COVID-19 compared with other RTIs Shaded areas are 95% CIs. For incidences of first diagnoses, the number in brackets corresponds to all patients who did not have the outcome before the follow-up period. For diagnostic subcategories, see appendix (pp 8–10). RTI=respiratory tract infection.

AUTHORS: Maxime Taquet, PhD, Prof John R Geddes, MD, Prof Masud Husain, FRCP, Sierra Luciano, BA, Prof Paul J Harrison, FRCPsych

Summary

Background

Neurological and psychiatric sequelae of COVID-19 have been reported, but more data are needed to adequately assess the effects of COVID-19 on brain health. We aimed to provide robust estimates of incidence rates and relative risks of neurological and psychiatric diagnoses in patients in the 6 months following a COVID-19 diagnosis.

Methods

For this retrospective cohort study and time-to-event analysis, we used data obtained from the TriNetX electronic health records network (with over 81 million patients). Our primary cohort comprised patients who had a COVID-19 diagnosis; one matched control cohort included patients diagnosed with influenza, and the other matched control cohort included patients diagnosed with any respiratory tract infection including influenza in the same period. Patients with a diagnosis of COVID-19 or a positive test for SARS-CoV-2 were excluded from the control cohorts. All cohorts included patients older than 10 years who had an index event on or after Jan 20, 2020, and who were still alive on Dec 13, 2020. We estimated the incidence of 14 neurological and psychiatric outcomes in the 6 months after a confirmed diagnosis of COVID-19: intracranial haemorrhage; ischaemic stroke; parkinsonism; Guillain-Barré syndrome; nerve, nerve root, and plexus disorders; myoneural junction and muscle disease; encephalitis; dementia; psychotic, mood, and anxiety disorders (grouped and separately); substance use disorder; and insomnia. Using a Cox model, we compared incidences with those in propensity score-matched cohorts of patients with influenza or other respiratory tract infections. We investigated how these estimates were affected by COVID-19 severity, as proxied by hospitalisation, intensive therapy unit (ITU) admission, and encephalopathy (delirium and related disorders). We assessed the robustness of the differences in outcomes between cohorts by repeating the analysis in different scenarios. To provide benchmarking for the incidence and risk of neurological and psychiatric sequelae, we compared our primary cohort with four cohorts of patients diagnosed in the same period with additional index events: skin infection, urolithiasis, fracture of a large bone, and pulmonary embolism.

Findings

Among 236 379 patients diagnosed with COVID-19, the estimated incidence of a neurological or psychiatric diagnosis in the following 6 months was 33·62% (95% CI 33·17–34·07), with 12·84% (12·36–13·33) receiving their first such diagnosis. For patients who had been admitted to an ITU, the estimated incidence of a diagnosis was 46·42% (44·78–48·09) and for a first diagnosis was 25·79% (23·50–28·25). Regarding individual diagnoses of the study outcomes, the whole COVID-19 cohort had estimated incidences of 0·56% (0·50–0·63) for intracranial haemorrhage, 2·10% (1·97–2·23) for ischaemic stroke, 0·11% (0·08–0·14) for parkinsonism, 0·67% (0·59–0·75) for dementia, 17·39% (17·04–17·74) for anxiety disorder, and 1·40% (1·30–1·51) for psychotic disorder, among others. In the group with ITU admission, estimated incidences were 2·66% (2·24–3·16) for intracranial haemorrhage, 6·92% (6·17–7·76) for ischaemic stroke, 0·26% (0·15–0·45) for parkinsonism, 1·74% (1·31–2·30) for dementia, 19·15% (17·90–20·48) for anxiety disorder, and 2·77% (2·31–3·33) for psychotic disorder. Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40–1·47, for any diagnosis; 1·78, 1·68–1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14–1·17, for any diagnosis; 1·32, 1·27–1·36, for any first diagnosis). As with incidences, HRs were higher in patients who had more severe COVID-19 (eg, those admitted to ITU compared with those who were not: 1·58, 1·50–1·67, for any diagnosis; 2·87, 2·45–3·35, for any first diagnosis). Results were robust to various sensitivity analyses and benchmarking against the four additional index health events.

Interpretation

Our study provides evidence for substantial neurological and psychiatric morbidity in the 6 months after COVID-19 infection. Risks were greatest in, but not limited to, patients who had severe COVID-19. This information could help in service planning and identification of research priorities. Complementary study designs, including prospective cohorts, are needed to corroborate and explain these findings.

Funding

National Institute for Health Research (NIHR) Oxford Health Biomedical Research Centre.

• View related content for this article

Evidence before this study We searched Web of Science and Medline on Aug 1 and Dec 31, 2020, for studies in English, with the terms “(COVID-19 OR SARS-CoV2 OR SARS-CoV-2) AND (psychiatri* or neurologi*) AND (incidence OR epidemiologi* OR ‘systematic review’ or ‘meta-analysis’)”. We found case series and reviews of series reporting neurological and neuropsychiatric disorders during acute COVID-19 illness. We found one large electronic health records study of the psychiatric sequelae in the 3 months after a COVID-19 diagnosis. It reported an increased risk for anxiety and mood disorders and dementia after COVID-19 compared with a range of other health events; the study also reported the incidence of each disorder. We are not aware of any large-scale data regarding the incidence or relative risks of neurological diagnoses in patients who had recovered from COVID-19.

Added value of this study To our knowledge, we provide the first meaningful estimates of the risks of major neurological and psychiatric conditions in the 6 months after a COVID-19 diagnosis, using the electronic health records of over 236 000 patients with COVID-19. We report their incidence and hazard ratios compared with patients who had had influenza or other respiratory tract infections. We show that both incidence and hazard ratios were greater in patients who required hospitalisation or admission to the intensive therapy unit (ITU), and in those who had encephalopathy (delirium and other altered mental states) during the illness compared with those who did not.

Implications of all the available evidence COVID-19 was robustly associated with an increased risk of neurological and psychiatric disorders in the 6 months after a diagnosis. Given the size of the pandemic and the chronicity of many of the diagnoses and their consequences (eg, dementia, stroke, and intracranial haemorrhage), substantial effects on health and social care systems are likely to occur. Our data provide important evidence indicating the scale and nature of services that might be required. The findings also highlight the need for enhanced neurological follow-up of patients who were admitted to ITU or had encephalopathy during their COVID-19 illness.

Introduction

Since the COVID-19 pandemic began on March 11, 2020, there has been concern that survivors might be at an increased risk of neurological disorders. This concern, initially based on findings from other coronaviruses,1 was followed rapidly by case series,2, 3, 4 emerging evidence of COVID-19 CNS involvement,5, 6, 7 and the identification of mechanisms by which this could occur.8, 9, 10, 11 Similar concerns have been raised regarding psychiatric sequelae of COVID-19,12, 13 with evidence showing that survivors are indeed at increased risk of mood and anxiety disorders in the 3 months after infection.14 However, we need large scale, robust, and longer term data to properly identify and quantify the consequences of the COVID-19 pandemic on brain health. Such information is required both to plan services and identify research priorities.

In this study, we used an electronic health records network to investigate the incidence of neurological and psychiatric diagnoses in survivors in the 6 months after documented clinical COVID-19 infection, and we compared the associated risks with those following other health conditions. We explored whether the severity of COVID-19 infection, as proxied by hospitalisation, intensive therapy unit (ITU) admission, and encephalopathy, affects these risks. We also assessed the trajectory of hazard ratios (HRs) across the 6-month period.

Methods

 Study design and data collection

For this retrospective cohort study, we used The TriNetX Analytics Network, a federated network recording anonymised data from electronic health records in 62 health-care organisations, primarily in the USA, comprising 81 million patients. Available data include demographics, diagnoses (using codes from ICD-10), medications, procedures, and measurements (eg, blood pressure and body-mass index). The health-care organisations are a mixture of hospitals, primary care, and specialist providers, contributing data from uninsured and insured patients. These organisations warrant that they have all necessary rights, consents, approvals, and authority to provide the data to TriNetX, so long as their name remains anonymous as a data source and their data are used for research purposes. By use of the TriNetX user interface, cohorts can be created on the basis of inclusion and exclusion criteria, matched for confounding variables with a built-in propensity score-matching algorithm, and compared for outcomes of interest over specified time periods. Additional details about TriNetX, its data, provenance, and functionalities, are presented in the appendix (pp 1–2).

 Cohorts

The primary cohort was defined as all patients who had a confirmed diagnosis of COVID-19 (ICD-10 code U07.1). We also constructed two matched control cohorts: patients diagnosed with influenza (ICD-10 codes J09–11) and patients diagnosed with any respiratory tract infection including influenza (ICD-10 codes J00–06, J09–18, or J20–22). We excluded patients with a diagnosis of COVID-19 or a positive test for SARS-CoV-2 from the control cohorts. We refer to the diagnosis of COVID-19 (in the primary cohort) and influenza or other respiratory tract infections (in the control cohorts) as index events. The cohorts included all patients older than 10 years who had an index event on or after Jan 20, 2020 (the date of the first recorded COVID-19 case in the USA), and who were still alive at the time of the main analysis (Dec 13, 2020). Additional details on cohorts are provided in the appendix (pp 2–3).

 Covariates

We used a set of established and suspected risk factors for COVID-19 and for more severe COVID-19 illness:15, 16 age, sex, race, ethnicity, obesity, hypertension, diabetes, chronic kidney disease, asthma, chronic lower respiratory diseases, nicotine dependence, substance use disorder, ischaemic heart disease and other forms of heart disease, socioeconomic deprivation, cancer (and haematological cancer in particular), chronic liver disease, stroke, dementia, organ transplant, rheumatoid arthritis, lupus, psoriasis, and disorders involving an immune mechanism. To capture these risk factors in patients' health records, we used 55 variables. More details, including ICD-10 codes, are provided in the appendix (pp 3–4). Cohorts were matched for all these variables, as described in the following subsections.

We estimated the diagnostic incidence of the neurological and psychiatric outcomes of the primary cohort in the 6 months after a COVID-19 diagnosis. In the whole cohort, 33·62% (95% CI 33·17–34·07) of patients received a diagnosis (table 2). For the cohort subgroups, these estimates were 38·73% (37·87–39·60) for patients who were hospitalised, 46·42% (44·78–48·09) for those admitted to ITU, and 62·34% (60·14–64·55) for those diagnosed with encephalopathy. A similar, but more marked, increasing trend was observed for patients receiving their first recorded neurological or psychiatric diagnosis (table 2). Results according to sex, race, and age are shown in the appendix (p 28). The baseline characteristics of the COVID-19 cohort divided into those who did versus those who did not have a neurological or psychiatric outcome are also shown in the appendix (p 7).

Various adverse neurological and psychiatric outcomes occurring after COVID-19 have been predicted and reported.1, 2, 3, 4, 5, 14 The data presented in this study, from a large electronic health records network, support these predictions and provide estimates of the incidence and risk of these outcomes in patients who had COVID-19 compared with matched cohorts of patients with other health conditions occurring contemporaneously with the COVID-19 pandemic (Table 2, Table 3, figure 1).

The severity of COVID-19 had a clear effect on subsequent neurological diagnoses (Table 4, Table 5, figure 2). Overall, COVID-19 was associated with increased risk of neurological and psychiatric outcomes, but the incidences and HRs of these were greater in patients who had required hospitalisation, and markedly so in those who had required ITU admission or had developed encephalopathy, even after extensive propensity score matching for other factors (eg, age or previous cerebrovascular disease). Potential mechanisms for this association include viral invasion of the CNS,10, 11 hypercoagulable states,22 and neural effects of the immune response.9 However, the incidence and relative risk of neurological and psychiatric diagnoses were also increased even in patients with COVID-19 who did not require hospitalisation.

Some specific neurological diagnoses merit individual mention. Consistent with several other reports,23, 24 the risk of cerebrovascular events (ischaemic stroke and intracranial haemorrhage) was elevated after COVID-19, with the incidence of ischaemic stroke rising to almost one in ten (or three in 100 for a first stroke) in patients with encephalopathy. A similarly increased risk of stroke in patients who had COVID-19 compared with those who had influenza has been reported.25 Our previous study reported preliminary evidence for an association between COVID-19 and dementia.14 The data in this study support this association. Although the estimated incidence was modest in the whole COVID-19 cohort (table 2), 2·66% of patients older than 65 years (appendix p 28) and 4·72% who had encephalopathy (table 2), received a first diagnosis of dementia within 6 months of having COVID-19. The associations between COVID-19 and cerebrovascular and neurodegenerative diagnoses are concerning, and information about the severity and subsequent course of these diseases is required.

Whether COVID-19 is associated with Guillain-Barré syndrome remains unclear;26 our data were also equivocal, with HRs increased with COVID-19 compared with other respiratory tract infections but not with influenza (table 3), and increased compared with three of the four other index health events (appendix p 34). Concerns have also been raised about post-COVID-19 parkinsonian syndromes, driven by the encephalitis lethargica epidemic that followed the 1918 influenza pandemic.27 Our data provide some support for this possibility, although the incidence was low and not all HRs were significant. Parkinsonism might be a delayed outcome, in which case a clearer signal might emerge with a longer follow-up.

The findings regarding anxiety and mood disorders were broadly consistent with 3-month outcome data from a study done in a smaller number of cases than our cohort, using the same network,14 and showed that the HR remained elevated, although decreasing, at the 6-month period. Unlike the earlier study, and in line with previous suggestions,28 we also observed a significantly increased risk of psychotic disorders, probably reflecting the larger sample size and longer duration of follow-up reported here. Substance use disorders and insomnia were also more common in COVID-19 survivors than in those who had influenza or other respiratory tract infections (except for the incidence of a first diagnosis of substance use disorder after COVID-19 compared with other respiratory tract infections). Therefore, as with the neurological outcomes, the psychiatric sequelae of COVID-19 appear widespread and to persist up to, and probably beyond, 6 months. Compared with neurological disorders, common psychiatric disorders (mood and anxiety disorders) showed a weaker relationship with the markers of COVID-19 severity in terms of incidence (table 2) or HRs (table 5). This might indicate that their occurrence reflects, at least partly, the psychological and other implications of a COVID-19 diagnosis rather than being a direct manifestation of the illness.

HRs for most neurological outcomes were constant, and hence the risks associated with COVID-19 persisted up to the 6-month timepoint. Longer-term studies are needed to ascertain the duration of risk and the trajectory for individual diagnoses.

Rs for most neurological outcomes were constant, and hence the risks associated with COVID-19 persisted up to the 6-month timepoint. Longer-term studies are needed to ascertain the duration of risk and the trajectory for individual diagnoses.

Our findings are robust given the sample size, the propensity score matching, and the results of the sensitivity and secondary analyses. Nevertheless, they have weaknesses inherent to an electronic health records study,29 such as the unknown completeness of records, no validation of diagnoses, and sparse information on socioeconomic and lifestyle factors. These issues primarily affect the incidence estimates, but the choice of cohorts against which to compare COVID-19 outcomes influenced the magnitude of the HRs (table 3, appendix p 34). The analyses regarding encephalopathy (delirium and related conditions) deserve a note of caution. Even among patients who were hospitalised, only about 11% received this diagnosis, whereas much higher rates would be expected.18, 30 Under-recording of delirium during acute illness is well known and probably means that the diagnosed cases had prominent or sustained features; as such, results for this group should not be generalised to all patients with COVID-19 who experience delirium. We also note that encephalopathy is not just a severity marker but a diagnosis in itself, which might predispose to, or be an early sign of, other neuropsychiatric or neurodegenerative outcomes observed during follow-up. The timing of index events was such that most infections with influenza and many of the other respiratory tract infections occurred earlier on during the pandemic, whereas the incidence of COVID-19 diagnoses increased over time (appendix p 24). The effect of these timing differences on observed rates of sequelae is unclear but, if anything, they are likely to make the HRs an underestimate because COVID-19 cases were diagnosed at a time when all other diagnoses were made at a lower rate in the population (appendix p 24). Some patients in the comparison cohorts are likely to have had undiagnosed COVID-19; this would also tend to make our HRs an underestimate. Finally, a study of this kind can only show associations; efforts to identify mechanisms and assess causality will require prospective cohort studies and additional study designs.

In summary, the present data show that COVID-19 is followed by significant rates of neurological and psychiatric diagnoses over the subsequent 6 months. Services need to be configured, and resourced, to deal with this anticipated need.

Contributors

PJH and MT were granted unrestricted access to the TriNetX Analytics network for the purposes of research, and with no constraints on the analyses done or the decision to publish; they designed the study and directly accessed the TriNetX Analytics web interface to do it. MT, JRG, MH, and PJH defined cohort inclusion and exclusion criteria, and the outcome criteria and analytical approaches. MT did the data analyses, assisted by SL and PJH. All authors contributed to data interpretation. MT and PJH wrote the paper with input from JRG, MH, and SL. MT and PJH verified the data. PJH is the guarantor. PJH and MT had full access to all the data in the study, and the corresponding author had final responsibility for the decision to submit for publication.

Data sharing

The TriNetX system returned the results of these analyses as .csv files, which were downloaded and archived. Data presented in this paper can be freely accessed online. Additionally, TriNetX will grant access to researchers if they have a specific concern (through a third-party agreement option).

Declaration of interests

SL is an employee of TriNetX. All other authors declare no competing interests.

Acknowledgments

This work was supported by the NIHR Oxford Health Biomedical Research Centre ( grant BRC-1215–20005 ). MT is an NIHR Academic Clinical Fellow. MH is supported by a Wellcome Trust Principal Research Fellowship and the NIHR Oxford Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the UK National Health Service, NIHR, or the UK Department of Health.