Immunovirological and environmental screening reveals actionable risk factors for fatal COVID-19 during post-vaccination nursing home outbreaks

Epidemiological profile of SARS-CoV-2 nursing home outbreaks

For the largest of the three outbreaks (nursing home A), the first infection was documented in the dementia ward on 17 May 2021, for an 89-year-old woman who developed COVID-19-related symptoms, who was subsequently hospitalized and who died after 2 weeks of hospitalization. A total of 102 cases were documented related to this outbreak between 18 May and 24 June, of which 75 were residents, 25 were staff members and two were family members of staff. All departments of the nursing home were involved, and consecutive screening moments were scheduled. Among 120 residents, 75 were SARS-CoV-2 positive by polymerase chain reaction (PCR) (62.5%; Table 1), whereas only 25 of 146 (17.1%) staff members tested positive (Supplementary Table 1). Timing of diagnosis by a positive PCR result and longitudinal follow-up is illustrated in Fig. 1a, which clearly shows late-onset PCR positivity for a large subset of residents who tested PCR negative at the start of the outbreak. This ‘second wave’ of delayed infections was corroborated by the continuous detection of SARS-CoV-2 by quantitative PCR (qPCR) in aerosol samples taken from the common areas of both staff and residents (Fig. 1b). For 58 of 102 (56.9%) positive cases, whole-genome sequencing (WGS) information was available, identifying the Delta variant (Pangolin lineage B.1.617.2) for all of them. Phylogenetic analysis indicates that all samples from the nursing home cluster were within the same clade, hence suggesting a single introduction event (Fig. 1c). Among the 75 PCR-positive residents, 15 died (case fatality ratio of 20%). Considering all individuals for whom vaccination status was known (Table 1), 96% of residents, but only 66% of staff members, were fully vaccinated. One resident and five staff members were partially vaccinated at the time of the outbreak, whereas one resident and 28.7% of staff members were not vaccinated.

Table 1 Demographic and clinical characteristics of nursing home residents involved in the three post-vaccination outbreaks
Fig. 1: Overview of the outbreak in nursing home A (Delta/B.1.617.2).
figure 1

a,b, We report the evolution through time of Ct values measured in both infected residents and staff members (a) and aerosols analyzed in various sections within the nursing home (b). Gray dots refer to negative PCR results. c, Time-scaled phylogenetic analysis involving Delta (B.1.617.2) genomes sampled and sequenced from this outbreak reveals that all 58 full genomes originating from nursing home A are clearly clustered within the overall phylogenetic tree (orange dots), suggesting a single introduction event. The phylogenetic tree is time calibrated, meaning that branch lengths are in units of time (year).

The first documented PCR-positive case for nursing home B dates from 20 May 2021, and the presumed index case developed COVID-19 symptoms the day before. Overall, 19 of 29 residents (65.5%) tested positive for SARS-CoV-2, but none of the 17 staff members tested positive on the repetitive screening moments organized between 20 May and 24 June. Despite high cycle threshold (Ct) values for this outbreak (Extended Data Fig. 1), WGS was successful for 19 of 19 (100%) PCR-positive cases, all classified as Gamma variant (Pangolin lineage P.1). Our phylogenetic analysis highlights that all samples clustered together within the more global Gamma phylogeny inferred in our study, again pointing toward the hypothesis of a single introduction event (Extended Data Fig. 1). Overall, seven fatal cases were reported in this outbreak, of which one resident tested negative by PCR. Although this death was classified as COVID-19 related according to World Health Organization (WHO) criteria16, due to severe respiratory symptoms and recent close contact with positive residents, we conservatively used only PCR-positive residents to calculate the case fatality ratio (6/19, 32%). For this nursing home, the vaccination rate was high among residents (86.2%), whereas only 52.9% of the staff members were fully vaccinated at the time of the outbreak. Nevertheless, none of the latter tested positive for SARS-CoV-2.

The post-vaccination outbreak in nursing home C was initially alerted by two cases (related resident and staff) infected with the Delta variant a few days before the large testing initiative for the other residents and staff members (20 July 2021). Twenty-five additional SARS-CoV-2-positive cases were identified during the outbreak. WGS determined the presence of the variant of interest, Mu (Pangolin lineage B.1.621), complemented with the mutation K417N in the spike protein, and, for one isolated staff member without resident contact, an additional Delta infection was identified. The single Delta-infected resident was, therefore, not included for further analysis of the outbreak (Table 1; 27/27 PCR-positive cases (100%) were confirmed by WGS: three Delta and 24 Mu). The Mu variant saw relatively limited circulation in Belgium, resulting in a restricted sampling of related genomic sequences in the local community. Our phylogenetic analysis, however, indicates that PCR-positive cases in this nursing home related to that variant clearly clustered within the overall phylogeny inferred for that variant (Extended Data Fig. 2), again advocating for a single introduction event. Moreover, all 24 PCR-positive cases infected with variant Mu (20 residents and four staff members) were linked to the dementia unit of the nursing home. Overall, seven infected residents died of COVID-19 (7/20, case fatality ratio 35%), and one additional resident died of a COVID-19-unrelated cause. Considering the 229 residents and staff members with known vaccination status, the overall vaccination rate was 98.3%. For the group of PCR-positive residents, 100% were fully vaccinated.

Demographic and clinical profile of SARS-CoV-2 outbreaks

Demographic and clinical risk factors for fatal COVID-19 among residents were identified by multivariable logistic regression models (Table 2), with the best model including age, male sex, non-Delta SARS-CoV-2 variants (Gamma and Mu) and later onset of infection (PCR positivity >7 d after the start of the outbreak). In the sensitivity analysis, only fully vaccinated and PCR-positive residents (n = 107) were included. The results remained statistically significant, with a similar effect size (Supplementary Table 2). The importance of these four factors as predictors of mortality was confirmed by Kaplan–Meier survival estimates (Extended Data Fig. 3) and time-to-event analysis (Cox proportional hazard regression; Supplementary Table 3). Of interest, dementia or peak viral load (nadir cycle quantification (Cq) value) were not predictive of fatal cases in the joint analysis of the three outbreaks (Table 2) but were significant predictors in single nursing homes (Supplementary Table 3). Because nursing home size was found to be a major risk factor for COVID-19 mortality in several countries, including Belgium17,18, we included this as an additional parameter in both logistic and Cox regression models. As shown in Supplementary Table 6, nursing home size was not an independent predictor (in addition to age, sex and late PCR positive) of fatal COVID-19, whereas the preferred model (corrected Akaike information criterion (cAIC)) contained age, sex, late PCR positive and variants/outbreaks as independent predictors.

Table 2 Multivariate logistic regression of demographic and clinical characteristics of residents with COVID-19 (all PCR-positive residents, n = 114)

Digital transcriptomic analysis of SARS-CoV-2 outbreaks

In search of candidate biomarkers for post-vaccine fatal COVID-19, as well as possible therapeutic targets, we opted for nCounter digital transcriptomics for immunovirological profiling of the nasal mucosa, encouraged by previous results19,20,21. For 20 of 28 fatal cases, a sufficient volume of diagnostic nasopharyngeal swabs was available for nCounter analysis, to explore immunological (600 genes representative of the major immune cell types) and virological (SARS-CoV-2 transcripts and ACE2/TMPRSS2 receptors) parameters as possible risk factors for fatal post-vaccine COVID-19. Thus, we carefully matched (age, sex and outbreak) 20 fatal cases (all those with available nasopharyngeal swabs) with 30 PCR-positive non-fatal cases, with similar timing of infection, as well as 10 PCR-negative but SARS-CoV-2-exposed residents. Because these samples were obtained at SARS-CoV-2 diagnosis, before hospitalization or treatment (oxygen and/or dexamethasone), the transcriptomic immune signatures are not modified by immunomodulatory treatment and can be used to predict fatal outcome. In addition, only four of 118 PCR-positive residents had received corticosteroids before their SARS-CoV-2 diagnosis. None of them was a fatal case, and they were not included for nCounter analysis.

As shown in Fig. 2 (volcano plot), a total of 193 human and seven viral gene transcripts were significantly upregulated or downregulated (P < 0.05) when comparing fatal versus non-fatal cases. In addition to the antiviral cytokines IL28A (also known as IFNL2 (interferon-λ2)) and IFNB1 (the gene encoding interferon-beta (IFN-β)), the most upregulated genes were predominantly expressed by innate immune cells: monocytes/macrophages (CX3CR1, TNFSF15, CLEC6A, ITLN1 and LILRB5), natural killer (NK) cells (THY1, CDH5, KIR3DL3, CD160, B3GAT1, NCAM1 and CCL3) and conventional dendritic cells (XCR1). Thus, the predominant immunopathogenic signature of fatal COVID-19 in vaccinated residents represents exacerbated innate immune activation rather than a failed adaptive (B cell and T cell) vaccine response. Likewise, a large subset of B cell genes (CD19, CR2, CD79A, CD79B, PAX5 and CD70), regulatory T cell (Treg) genes (FOXP3 and PTGER4) and cytotoxic CD8 T cell genes (EOMES and PTGER4) were also significantly upregulated in fatal cases, arguing against a curtailed B cell or T cell response or a failure of B cells or T cells to migrate to the nasal mucosa. On the other hand, a generalized downregulation of major histocompatibility complex (MHC) class I-mediated antigen presentation (B2M and HLA-C) was observed across all cell types, in agreement with previous reports demonstrating loss of MHC class I activity at the transcriptomic, epigenomic and functional level22,23,24,25,26,27.

Fig. 2: Differentially expressed genes in nasal mucosa of fatal COVID-19 outbreak cases as compared to matched PCR-positive residents from three nursing homes.
figure 2

Volcano plot of differentially expressed genes in nasal mucosa of fatal (n = 20) versus age-matched, sex-matched and outbreak-matched non-fatal PCR-positive cases (n = 30), quantified by nCounter digital transcriptomics (uncorrected P values from linear model, negative binomial distribution, dotted line showing P < 0.05, FDR q values provided in Source Data). Selected viral (red circles) and host immune transcripts (blue circles) significantly upregulated or downregulated in fatal versus non-fatal cases are highlighted with gene names. Details on immune genes are given in the Results section. PCR+, PCR positive.

Source data

Because the top downregulated genes were most representative of mucosal epithelial cells (PIGR, CD9 and MUC1), the observed exacerbated innate response might represent enhanced migration of innate immune cells but also virus-mediated destruction of the mucosal epithelial cells. In favor of the latter hypothesis, fatal cases were characterized by significantly higher viral transcript levels when measured by nCounter. Transcript levels for spike, envelope, nucleoprotein, ORF1ab, ORF3a and ORF7a genes (Fig. 3a and data not shown, all P < 0.05 with false discovery rate (FDR) correction) were higher in fatal cases compared to non-fatal PCR-positive residents. In addition, antisense SARS-CoV-2 was selectively increased in eight of 20 fatal cases (Fig. 3a) versus PCR-positive cases, indicating heightened intracellular viral replication. Of note, peak viral load (nadir Cq values) or viral load of the first PCR-positive sample, measured by qPCR, was not significantly different between fatal cases and PCR-positive controls (Fig. 3a), underscoring the sensitivity of nCounter digital transcriptomics. Exacerbated viral replication in fatal cases was paralleled by a marked eight-fold increase in viral receptor ACE2 transcript levels (P < 0.001) as well as an unexpected two-fold decrease (P < 0.01) in viral co-receptor TMPRSS2 expression (Fig. 3b).

Fig. 3: Immunological and virological risk factors identified in fatal COVID-19 outbreak cases among residents in three nursing homes.
figure 3

ac, Viral transcript levels for spike protein (left: fatal versus PCRpos P = 0.012, fatal versus PCRneg P = 0.000022, PCRneg versus PCRpos P = 0.0089) and ORF1ab antisense RNA (middle), measured by nCounter digital transcriptomics. Right panel shows peak viral load (nadir Cq values) as quantified by qPCR. Viral receptors (ACE2: fatal versus PCRpos P = 0.0009; TMPRSS2: fatal versus PCRpos P = 0.0036, fatal versus PCRneg P = 0.0005, PCRneg versus PCRpos P = 0.0422) (b) and antiviral cytokine IFNB1 (fatal versus PCRpos P = 0.0022, fatal versus PCRneg P = 0.0022) (c) were quantified by nCounter digital transcriptomics. Data are presented as median values ± s.d. d, Left: visualization of best predictive model (multivariate logistic regression, selected by cAIC), including age (not depicted) and ORF7a and ACE2 transcripts. Dashed gray lines indicate the detection limit of SARS-CoV-2 transcripts. Each circle represents a resident, and the size of the circle is proportional to ACE2 normalized expression. Right: comparison of ROC curves of predictive models by univariate (IFNB1) or multivariate (IFNB1/age/sex and age/ORF7a/ACE2) logistic regression. ROC curves showing significant prediction of fatal versus non-fatal COVID-19 according to IFNB1 transcript levels (right), with and without age and sex as additional factors (detailed in the Results section). For ac, statistical results are from Kruskal–Wallis test with FDR correction for multiple testing (PCRneg n = 10, PCRpos n = 30, fatal n = 20), ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, NS, not significant. PCRpos, PCR positive; PCRneg, PCR negative.

Source data

Among all immune genes, IFNB1 transcripts displayed the strongest negative correlation to survival time (starting from the date of PCR-positive diagnosis, Spearman’s ρ = −0.24, P = 0.0024). Corroborating our previous findings in a Belgian cohort of intensive care unit (ICU) patients19, we found that increased IFNB1 transcript levels significantly predicted a fatal outcome (Fig. 3c,d; area under the receiver operating characteristic (AUROC) curve 0.76 (95% confidence interval (CI) 0.63–0.89), P = 0.0013), which was slightly increased by adding age and sex to the model (Fig. 3d; AUROC 0.82 (95% CI 0.71–0.93), P = 0.000064). IFNB1 remained a significant predictor in multivariable logistic regression, independent of age, sex and peak viral load (nadir Cq value), which was also confirmed by time-to-event analysis (Cox proportional hazard models; Table 3).

Table 3 Multivariate Cox proportional hazard regression of immunological and virological parameters in fatal versus non-fatal post-vaccination COVID-19 in nursing home residents

Lastly, when combining all available demographic, immune and viral parameters, the best predictive model for mortality, according to the cAIC, included age (odds ratio (OR) 1.07, 95% CI 0.98–1.19), increased viral ORF7a (OR 1.67, 95% CI 0.98–3.46) and viral receptor ACE2 (15.43, 95% CI 2.54–165.9) transcript levels, resulting in correct classification of 18 of 20 (90%) fatal cases (AUROC 0.88, 95% CI 078–0.98, P = 0.000002), as visualized in Fig. 3d.

A unique immune signature in post-vaccination fatal COVID-19

To our knowledge, no well-powered study of immune signatures in post-vaccination fatal COVID-19 in the older adult population have been published at present. Thus, no public datasets are currently available for independent validation of our ‘post-vaccine fatal COVID-19’ immune signature in a comparable epidemiological setting. Therefore, we compared published transcriptomic and genomic signatures of pre-vaccination fatal and/or life-threatening COVID-19.

Upon cross-comparison of our ‘post-vaccine fatal COVID-19’ transcriptomic signature with previously described IEI (inborn errors in type I IFN immunity) genes linked to life-threatening COVID-19 (ref. 28), we identified a clear dichotomy between upregulated (IRF7) versus downregulated (IRF3) IEI genes in fatal cases (Fig. 4a). Therefore, we quantified type I IFN signaling score based on nCounter gene expression data (Supplementary Table 5). Of note, type I IFN signaling score was not significantly different between fatal cases and matched controls (Fig. 4b). However, lymphocyte activation, Th17 and Treg differentiation pathways were significantly increased in fatal cases (Fig. 4b), in agreement with our finding of upregulated EOMES, SRC, THY1, RORC, IL6R, FOXP3 and PTGER4 genes (Fig. 2). As shown in Fig. 4c, type I IFN signaling score was highly correlated to IRF7 (ρ = 0.84, P = 7 × 10−14) as well as STAT2 (ref. 29) (ρ = 0.91, P = 6 × 10−20) but not to IRF3 or IFNA2 (both P > 0.05) transcripts. In contrast to IFNA2, but in agreement with our multivariable logistic regression models for mortality, IFNB1 levels were most strongly correlated to IRF7 and TLR7 plasmacytoid dendritic cell (pDC)-specific type I IFN drivers as well as inflammatory targets, such as IL6R (Fig. 4c, lower panel). Similar to IFNB1, we found that IRF3 transcript levels were also able to predict mortality in residents (Fig. 4d; Kaplan–Meier curve, P = 0.0030). In addition, classification of nursing home residents according to IFNB1 levels demonstrated a significant link with lower IRF3 expression as well as higher viral replication and apoptosis, providing a possible molecular and cellular mechanism of action of IFN-β.

Fig. 4: IRF3/IRF7 dichotomy in type I IFN signaling underlies IFN-β link to inflammation, apoptosis and mortality in nursing home residents during post-vaccine COVID-19 outbreaks.
figure 4

a, Venn diagram shows overlap between gene transcripts upregulated (‘up Fatal’) or downregulated (‘down Fatal) in fatal cases versus PCR-positive controls (quantified by nCounter digital transcriptomics) and the gene mutations (IEI) identified in life-threatening COVID-19 (ref. 28) (pre-vaccine era). The five IEI genes not differentially expressed between cases and controls are TICAM1, TBK1, UNC93B1, IFNAR1 and TLR3. b, Pathway scores (calculated by nSolver from gene expression profiling by nCounter) for lymphocyte activation (P = 0.043), Th17 (P = 0.028) and Treg differentiation (P = 0.022) were increased in fatal cases versus PCR-positive controls, whereas type I IFN signaling was not (t-test with Welch’s correction). No pathways were significant after stringent Bonferroni correction for multiple testing. Data are presented as median values ± s.d. Red circles: fatal cases; green circles: PCR-positive controls. c, Spearmanʼs correlation of type I signaling score (upper panel) and IFNB1 expression (lower panel) with drivers of IFN signaling (STAT2, IRF7, IRF3, IFNA2 and TLR7) and inflammation (IL6R), across all 50 residents (20 fatal cases and 30 PCR-positive controls). d, Kaplan–Meier curve demonstrating significantly lower (log-rank test) survival in nursing home residents with ‘IRF3 low’ status (nCounter normalized expression below the median). e, Classification of nursing home residents into ‘IFNB1 high’ versus ‘IFNB1 low’ (below or above 100 normalized counts) reveals a significant link with IRF3 expression (Mann–Whitney test P = 0.000011), intracellular viral replication (measured as SARS-CoV-2 antisense RNA, Mann–Whitney test P = 0.000072) and apoptosis score (calculated by nSolver, Mann–Whitney test P = 0.044) in upper airway mucosa. Data are presented as median values ± s.d. ****P < 0.0001, **P < 0.01, *P < 0.05, NS, not significant. PCR+, PCR positive.

Source data

In addition to the genetic link to type I IFN signaling, anti-type I IFN neutralizing antibodies have been shown by several groups to be an additional risk factor for life-threatening COVID-19 (refs. 30,31,32,33). Because no serum samples were available from the fatal cases, we cross-examined our fatal COVID-19 immune gene signature with the LAIR1 biomarker recently described by van der Wijst et al.33 as strongly correlated to anti-IFN auto-antibodies. Confirming its antagonistic role in IFN/antiviral signaling, LAIR1 level was positively correlated with peak viral load (nadir Cq value, P = 7.0 × 10−5, ρ = −0.49, n = 50; Extended Data Fig. 4). Similar to peak viral load (Fig. 2 and Tables 2 and 3), LAIR1 transcript levels were not able to predict survival in this cohort (data not shown). However, we found that LAIR1 transcript level was significantly and negatively correlated with IRF3 (the major upstream driver of type I IFN production in epithelial cells) in fatal cases (ρ = −0.59, P = 0.0067, n = 20), whereas no significant correlation was observed for matched PCR-positive controls (ρ = −0.037, P = 0.84, n = 30; Extended Data Fig. 4). This demonstrated a major difference in the type I IFN pathway regulation between fatal cases and controls, probably more pronounced in the epithelial cells of the upper airway mucosa, in agreement with Zhang et al.34.

Taken together, cross-examination of published transcriptomic and genomic pre-vaccine fatal COVID-19 signatures highlights the unique innate and adaptive immune signature observed in post-vaccination fatal COVID-19 in nursing home residents.

2023-05-22 15:15:24