Vaccine-associated enhanced respiratory pathology in COVID-19 hamsters after T_H2-biased immunization

Abstract

Vaccine-associated enhanced respiratory disease (VAERD) is a severe complication for some respiratory infections. To investigate the potential for VAERD induction in coronavirus disease 2019 (COVID-19), we evaluate two vaccine leads utilizing a severe hamster infection model: a T helper type 1 (T_H1)-biased measles vaccine-derived candidate and a T_H2-biased alum-adjuvanted, non-stabilized spike protein. The measles virus (MeV)-derived vaccine protects the animals, but the protein lead induces VAERD, which can be alleviated by dexamethasone treatment. Bulk transcriptomic analysis reveals that our protein vaccine prepares enhanced host gene dysregulation in the lung, exclusively up-regulating mRNAs encoding the eosinophil attractant CCL-11, T_H2-driving interleukin (IL)-19, or T_H2 cytokines IL-4, IL-5, and IL-13. Single-cell RNA sequencing (scRNA-seq) identifies lung macrophages or lymphoid cells as sources, respectively. Our findings imply that VAERD is caused by the concerted action of hyperstimulated macrophages and T_H2 cytokine-secreting lymphoid cells and potentially links VAERD to antibody-dependent enhancement (ADE). In summary, we identify the cytokine drivers and cellular contributors that mediate VAERD after T_H2-biased vaccination.

Graphical abstract

Ebenig et al. report vaccine-associated enhanced respiratory disease (VAERD) in SARS-CoV-2-infected Syrian hamsters previously immunized with a suboptimal vaccine consisting of misfolded SARS-CoV-2 spike protein and Alum adjuvants. VAERD is correlated with dysregulation of specific immune cell subsets in the lungs and could be alleviated by dexamethasone treatment.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the etiological agent of the current coronavirus disease 2019 (COVID-19) pandemic (Wu et al., 2020; Zhu et al., 2020), resulting in more than 513 million confirmed cases (as of May 5, 2022) with an overall case fatality rate of 1.21% (WHO, 2021). To combat this pandemic, the first vaccine candidates were developed, tested in animal models, and approved for human use within the first year after pathogen identification. To date, more than 10 billion vaccine doses have been administered of all authorized vaccines based on different vaccine technologies, namely mRNA vaccines, adenoviral vector vaccines, inactivated viruses, or proteins. Potential safety risks including the induction of a more severe or altered clinical pathology after breakthrough infection in vaccinated patients were considered, but not yet described, during their development.

One possible risk associated with immunization in the context of respiratory infections is an immunopathology termed vaccine-associated enhanced respiratory disease (VAERD) (Munoz et al., 2021). Such enhancement of disease has been well described in the context of formalin-inactivated respiratory syncytial virus (RSV) vaccines for infants and young children (Kim et al., 1969). Vaccinated children who were later exposed to circulating RSV developed an enhanced and atypical phenotype of clinical symptoms that resulted in higher hospitalization rates and even a small number of associated fatalities (Kim et al., 1969). Lung sections of the fatal cases displayed monocytic infiltration with massive excess of eosinophils (Kim et al., 1969). Induction of low-affinity and non-neutralizing antibody (Ab) responses provoked by the formalin-inactivated virus was found to be causative for enhanced disease (Polack et al., 2002). Such a pathologic eosinophil infiltration triggered by T_H2-biased T cell responses has also been demonstrated in animal models after immunization with formalin-inactivated virus prior to RSV infection (Ruckwardt et al., 2019).

Immunopathologies such as VAERD or Ab-dependent enhancement (ADE) of disease were also reported in CoV-infected vaccinated animals or animal models for highly pathogenic human CoV. Cats immunized with a recombinant S protein-expressing vaccinia virus developed more severe disease and early death syndrome following infection with feline infectious peritonitis virus, which was linked to a low amount of vaccine-induced neutralizing Abs in the affected animals (Vennema et al., 1990). Passive transfer of Abs directed against the virus before infection was sufficient for pathology, which was correlated to accelerated virus uptake by macrophages via Fc receptor (Olsen et al., 1992; Takano et al., 2008). Such processes are considered a marker and relevant for classical ADE as described for dengue virus (DENV) (Beltramello et al., 2010). Also, animal models for the first two highly pathogenic zoonotic beta-CoVs, SARS-CoV and Middle Eastern respiratory syndrome (MERS)-CoV, revealed potential for VAERD. Entry receptor-transgenic K18-ACE2 or hDPP4 mice that were immunized with whole inactivated virus vaccines developed severe immunopathology in lung tissue with infiltration of eosinophils after infection (Bolles et al., 2011; Tseng et al., 2012; Agrawal et al., 2016).

These scenarios prompted careful evaluation of the predisposition of COVID-19 infection for disease enhancement. Induction of T_H2-biased immunity or non-functional Ab responses by the different front-runner vaccine candidates was avoided (Corbett et al., 2020a, 2020b; Ramasamy et al., 2020; Sahin et al., 2020; Sadoff et al., 2021; van der Lubbe et al., 2021). Accordingly, no ERD or ADE has been reported in vaccinated individuals to date. However, two recent publications described evidence for VAERD in per se T_H2-biased mouse models using mouse-adapted recombinant SARS-CoV-2 after immunization with alum-adjuvanted, whole-inactivated SARS-CoV-2 or adjuvanted S protein. Following virus challenge, these animals developed considerable pneumonia with infiltration of eosinophils and neutrophils, a clear sign of VAERD (DiPiazza et al., 2021; Iwata-Yoshikawa et al., 2022). Again, the pathology was linked to induction of cytokines typical for T_H2-biased immune responses, but no detailed mechanism for the pathogenesis was described.

We have previously shown in the Syrian hamster (Mesocricetus auratus) COVID-19 model that alum-adjuvanted spike protein failed to protect against severe disease after infection, in contrast to a measles vaccine-derived prototypic COVID-19 vaccine (Hörner et al., 2020). Here, we report reproducible VAERD in outbred hamsters with no inbred helper cell bias. Those hamsters were vaccinated with alum-adjuvanted S protein and later infected with a low-passage human patient isolate. This model allowed us to identify the mechanisms that underlie the observed immunopathology. VAERD became evident in the respective hamsters’ lungs: ISH, immunohistochemistry, staining for eosinophils, and general histopathology revealed exaggeration of pneumonia and eosinophilic infiltration, while virus load was reduced. These effects could be correlated to a broadly enhanced dysregulation of gene expression: up-regulation of the major eosinophil attractant eotaxin-1/CCL-11 combined with induction of T_H2-marker cytokines interleukin (IL)-4, IL-5, and IL-13 supported by IL-19 was evident. Furthermore, using single-cell RNA sequencing (scRNA-seq), we attributed this dysregulation to specific immune cell subsets, suggesting that ADE via Fc-receptor-mediated skewing of virion uptake drives CCL-11 and IL-19 secretion by lung macrophages. These cytokines attract eosinophils and positively feedback on the T/NK cell compartment responsible for T_H2-marker cytokine secretion. Eosinophil infiltration and pathology observed in recombinant, non-stabilized SARS-CoV-2 S protein adjuvanted with aluminum hydroxide (Alum+S)-vaccinated animals were dramatically reduced by treatment of infected animals with dexamethasone, while the whole syndrome was absent in animals that received the T helper type 1 (T_H1)-biased measles virus (MeV)-derived vaccine candidate. These data allow us to propose a detailed model of vaccine-induced immunopathogenesis, involving immune cells of the lymphoid and myeloid compartment that release pro-inflammatory cytokines, in a gold-standard animal model of COVID-19. Our data also potentially link VAERD mechanistically to ADE, which will be the focus of follow-up studies.

Results

Histological analysis and pulmonary virus load

To study the efficacy of MeV-derived COVID-19 vaccines, we vaccinated Syrian hamsters with the experimental vaccine candidate MeV_vac2-SARS2-S(H) and a classic protein vaccine formulation of recombinant, non-stabilized SARS-CoV-2 S protein adjuvanted with aluminum hydroxide (Alum+S). The recombinant S revealed distinct changes in secondary structural elements (31.7% α-helices and 35.0% β-sheets versus 19.3% α-helices and 50.8% β-sheets) compared with a stabilized soluble S (Figure S1), highlighting deviation (i.e., de-folding) of the vaccine protein from the native pre-fusion structure. Vaccinated animals were challenged in parallel to naive or vector-control hamsters (MV_vac2-ATU(P)) using the low-passage SARS-CoV-2 patient isolate BavPat1 (Hörner et al., 2020) (Figure 1A). During histological analysis of lung necropsies, the expected pathology was observed for the control animals in hematoxylin and eosin (H&E)-stained tissue slices of the lungs (Figure 1B). In both naive and vector control hamsters, epithelia and endothelia in bronchii and vasculature, respectively, showed inflammation and hemorrhages. Between 20% and 80% (median 40%) of the lung area had become dense due to cellular infiltrates of macrophages and lymphocytes; in general, no or few eosinophils or granulocytes were observed. Single foci to moderate karyorrhexis were apparent (Table S1). This pathology was significantly ameliorated in MeV_vac2-SARS2-S(H)-vaccinated hamsters, with reduced inflammation (median of 27.5% dense area), very few hemorrhages, only a few eosinophils in perivascular regions, and no to minimal karyorrhexis. In contrast, animals vaccinated with Alum+S displayed a considerably intensified pathogenesis: while bronchial epithelia showed few inflammatory alterations, vascular endothelium revealed explicit inflammation. The area of dense infiltrates exceeded 50% of the slices, with massive infiltration of eosinophils (Figures 1B and 1D). Together, this histopathology suggests induction of VAERD by the immunization of hamsters with Alum+S.

Figure 1.

Pulmonary pathology in vaccinated Syrian hamsters upon challenge with SARS-CoV-2

(A) Schematic depiction of experiment. Syrian hamsters were bled and subsequently immunized with indicated vaccines on days 0 and 21 (n = 6/cohort). On day 35, immunized hamsters were bled and intranasally challenged with low-passage SARS-CoV-2 and observed for another 4 days before sacrifice for organ preparation.

(B) Depicted are representative sections of left lobes of vaccinated hamster lungs prepared 4 days post infection (dpi) with low-passage SARS-CoV-2 (n = 6). H&E staining (two top rows) reveals histopathological changes and immune cell infiltration. In situ hybridization (ISH) for SARS-CoV-2 viral RNA (third row) and immunohistochemistry (IHC) staining for SARS-CoV-2 nucleocapsid protein N (fourth row) depict extent of infection, while Sirius red staining (bottom row) reveals infiltration of eosinophils (scale bars: top row, 500 μm; second row, 20 μm; other rows, 40 μm). Yellow arrowheads and yellow circle depict individual eosinophils or a cluster of eosinophils in H&E-stained samples, respectively.

(C) Quantification of lung infection by determining the relative area of tissue staining positive by ISH by automated image evaluation of the whole slices.

(D) Eosinophil quantification for all animals is depicted. Eosinophil numbers in 10 randomly selected fields of view at 200× magnification were counted, and means were correlated against the area of view. Each data point reveals the mean number of eosinophils per mm² in individual animals. For statistical analysis, ordinary one-way ANOVA was applied with Tukey’s multiple comparisons test. ^∗∗∗∗p < 0.0001.

(C and D) Single dots represent individual animals, bars median of groups; black circles, MOCK immunized animals; open squares, MeV vector-immunized animals; blue diamonds, MeV_vac2-SARS2-S(H)-immunized animals; red triangles, animals vaccinated with Alum+S. ^∗∗∗∗p < 0.0001.

Active SARS-CoV-2 infection became evident by in situ hybridization (ISH) of SARS-CoV-2 RNA genomes (Figures 1B, third row, and 1C) and immunohistochemistry for SARS-CoV-2 nucleocapsid protein N (Figure 1B, fourth row). MeV_vac2-SARS2-S(H)-vaccinated animals revealed significantly reduced staining for SARS-CoV-2 in both ISH and immunohistochemistry. These data are consistent with the previously published protective efficacy of this T_H1-biased vaccine concept. Additionally, Alum+S-vaccinated animals showed somewhat reduced staining for SARS-CoV-2 genomes and N, despite enhanced pathology. While up to 20% of the lung tissue stained positive for SARS-CoV-2 genomes in control animals, a maximum of only 5% did in hamsters vaccinated with MeV_vac2-SARS2-S(H), while Alum+S-vaccinated animals showed a split behavior (Figure 1C). Sirius red staining for eosinophils (Figure 1B, bottom row) validated H&E findings for significantly enhanced infiltration, which became evident to this extent only in this vaccine cohort. Only in 3 out of 6 animals vaccinated with MeV_vac2-SARS2-S(H) were some eosinophils found in perivascular regions of inflamed tissue (Figure S2). Between 84 and 155 eosinophils/mm² were found in lung sections of hamsters immunized with Alum+S, whereas in hamsters inoculated with MeV_vac2-SARS2-S(H), or the MeV vector control, eosinophil numbers remained within the same range as uninfected control hamsters (3–24 eosinophils/mm²) (Figure 1D).

Thus, the COVID-19 hamster model revealed features suggestive of VAERD for animals vaccinated using the traditional T_H2-biased alum-adjuvanted protein vaccine approach but an absence of such an effect for the T_H1-biased measles-based COVID-19 vaccine candidate (Hörner et al., 2020).

Transcriptional profiling of SARS-CoV-2-infected hamster lungs

To identify determinants underlying the observed enhanced pathology, we performed RNA-seq profiling of hamster lung tissue. Lung transcriptomes from SARS-CoV-2-infected naive hamsters were compared with infected animals previously immunized with either T_H1- or T_H2-biased vaccines. Differentially expressed genes (DEGs) in lungs were determined by comparison with baseline expression levels, derived from uninfected, unvaccinated control animals. Principal component analysis (PCA) of individual transcriptomes revealed segregation of signals according to infection and vaccination status, confirming distinct lung transcriptional responses in the different groups (Figure 2A). Significant transcriptional dysregulation was evident in infected lung tissue, with ∼2,000 genes down-regulated and ∼1,500 genes up-regulated in vaccine-naïve, SARS-CoV-2-infected hamsters (Figure 2B, left panel). However, prior vaccination with MeV_vac2-SARS2-S(H) caused a 40% reduction in numbers of significantly dysregulated genes. In line with this, lung resident viral RNAs were also 20-fold reduced when compared with unvaccinated animals (Figure 2B, left panel). In contrast, challenge with SARS-CoV-2 after vaccination with Alum+S resulted in comparable numbers of dysregulated genes to those observed in unvaccinated animals (Figure 2B, left panel), with an increased median fold change in gene expression also apparent (Figure 2B, right panel). Only a 2-fold reduction in lung-resident viral RNAs was observed here (Figure 2B, left panel). Together, these data suggest that prior immunization with a T_H1-biased vaccine confers protection from disease, substantially reducing viral loads in hamster lungs, with a concomitant reduction in the magnitude of lung gene dysregulation. In contrast, the gene dysregulation and viral loads observed using the T_H2-biased vaccine more closely resembled patterns seen in unvaccinated animals. However, inspection of the PCA plot revealed separate clustering of unvaccinated and Alum+S groups (Figure 2A), indicating unique transcriptional signatures underlying Alum+S-associated pathology.

Figure 2.

Global host gene dysregulation revealed by global RNA-seq in lungs of infected, differently vaccinated hamsters

(A) Principal component analysis (PCA) of hamster lung transcriptomes. Transcriptomes from individual animals (n = 4/cohort) are represented by a single data point. Yellow, unvaccinated and not infected; blue, unvaccinated and infected; red, vaccinated with MeV_vac2-SARS2-S(H) and infected; green, vaccinated with alum-adjuvanted protein and infected.

(B) Viral RNA copies and host gene dysregulation in hamster lungs after infection and vaccination. Left panel: data plotted on the left y axis represent the number of differentially expressed genes (DEGs) (false discovery rate [FDR] p < 0.05) in the hamster lung under each condition, when compared with uninfected unvaccinated controls (n = 4). Blue: upregulated; orange: downregulated. Data plotted on the right y axis (gray circles) represents the number of lung-resident SARS-CoV-2 mapped reads per condition (mean ± SEM). Right panel: violin plots depict range of log₂ fold change in expression for all significantly dysregulated genes under each condition, when compared with uninfected unvaccinated controls. Horizontal lines, median values.

(C) Gene Ontology (GO) enrichment analysis of SARS-CoV-2-induced DEGs. GO categories are labeled on the y axes. Circle size represents the ratio of significantly dysregulated genes relative to the total gene number in a specific GO term. Circles are shaded relative to activation Z score, with significantly enriched categories highlighted. Left panel: selected GO categories related to immune response. Right panel: significant GO categories exclusive to Alum+S-vaccinated animals.

(D) Plot depicting log₂ fold change of 28 identified genes (left) exhibiting differential expression patterns between the different conditions, and their associated cellular function (right).

(E) Dysregulation of selected cytokine genes including eosinophil chemo-attractant Ccl11 and T_H2 cytokine mRNAs in lungs of differently vaccinated hamsters upon SARS-CoV-2 infection. Heatmap represents normalized mRNA expression (reads per kilobase per million bases mapped [RPKM]) of genes. Gray cells with an X represents no detectable mRNA expression (RPKM = 0).

To investigate this in more detail, we performed Gene Ontology (GO) enrichment analyses to determine the associated biological processes (Figure 2C). In all groups, infection was associated with significant enrichment of genes associated with defense against pathogens and vigorous induction of classical antiviral and inflammatory response genes (Figure 2C, left panel). Of note, the magnitude of these responses was similar in unvaccinated and Alum+S-immunized hamsters and reduced in MeV_vac2-SARS2-S(H)-immunized animals. Visualizing significantly enriched GO categories that were unique to Alum+S-treated animals revealed a profile of dysregulated gene classes that likely contributes to the observed vaccine-associated pathology after superinfection (Figure 2C, right panel).

We next visualized fold change in expression of differentially induced genes, considering their associated cellular functions (Figure 2D). These analyses led us to further explore normalized expression of a refined subset of genes that exhibited unique induction profiles and likely contribute to the vaccine-associated immunopathogenesis (Figure 2E). While the suite of genes involved in eosinophil chemotaxis were similarly upregulated under all conditions, uncontrolled induction of the major eosinophil chemotaxin Ccl11 (eotaxin-1) mRNA was unique to the Alum+S cohort (Figure 2E). Furthermore, T_H2 cytokine mRNAs Il4, Il5, Il13, and Il19 were potently induced in the majority of Alum+S-vaccinated animals but largely undetectable in the other groups. These analyses pinpoint likely transcriptional mediators underlying VAERD and hyperinflammation after T_H2-biased vaccination: the specific induction of the Il4/Il5/Il13/Il19 cytokine axis combined with the potent esosinophil chemotaxin Ccl11 potentially results in uncontrolled recruitment of eosinophils to the site of infection and enhanced pathogenesis.

Induction of T_H2-biased anti-S immunity by alum-adjuvanted protein

To further demonstrate induction of T_H2-biased immunity by vaccination as the trigger of VAERD, hamsters were just immunized with Alum+S, MeV_vac2-SARS2-S(H), or medium (MOCK) without SARS-CoV-2 challenge (Figure 3A). We then analyzed antigen-specific cytokine responses of these animals splenocytes’ in re-call experiments 14 days after the second immunization.

Figure 3.

Induction of T_H2-biased antigen-specific immune cells in protein-vaccinated Syrian hamsters

(A) Syrian hamsters were immunized with MeV_vac2-SARS2-S(H) (MeV-S) (intraperitoneally [i.p.]) or Alum+S (subcutaneously [s.c.]) on days 0 and 21 (n = 6 animals/cohort). Two weeks after immunization, hamsters were sacrificed, blood was collected, and splenocytes were isolated.

(B) IFN-γ ELISpot analysis using splenocytes of hamsters vaccinated on days 0 and 21, isolated 14 days after the boost immunization and after stimulation with recombinant S protein (rS protein) or MeV bulk antigens (MeV bulk). The reactivity of splenocytes was confirmed by Concanavalin A (ConA) (5 μg/mL) or flagellin treatment (20 μg/mL). Recombinant ovalbumin (rOva) or medium served as negative controls. The number of cells per 1×10⁶ splenocytes represent the number of cells expressing IFN-γ upon re-stimulation. Dots represents individual animals, horizontals bars median per group (n = 6). Samples above the upper limit of detection (ULOD) were displayed as such. For statistical analysis of grouped ELISpot data, paired t test was applied. ns, not significant (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01.

(C) Relative fold change expression of mRNAs encoding IL-4, IL-5, or IL-13 was determined using quantitative RT-PCR and the ΔΔct method. mRNA encoding RPL18 was used as housekeeping gene for normalization. Mean of samples from mock-treated hamsters served as reference. For statistical analysis, ordinary one-way ANOVA was applied with Tukey’s multiple comparisons test. ns, not significant (p > 0.05), ^∗p < 0.05; ^∗∗p < 0 .01; ^∗∗∗p < 0.001.

(D) Absolute mRNA copy numbers were determined by quantitative RT-PCR using a plasmid DNA standard for each gene. Mock-vaccinated hamsters, black circles; MeV_vac2-SARS2-S(H)-vaccinated hamsters, blue diamonds; protein-vaccinated hamsters, red triangles.

As a control for successful vaccination, interferon gamma (IFN-γ) secretion was determined. Splenocytes of all animals reacted with IFN-γ secretion after stimulation with the unspecific stimuli Concanavalin A (ConA) or flagellin, which triggered more than 600 spots/10⁶ cells in ELISpot, thus demonstrating general functionality. On the other hand, medium- or ovalbumin-stimulated splenocytes showed a background reactivity of 100–250 IFN-γ⁺ spots/10⁶ cells (Figure 3B). Upon antigen-specific stimulation with recombinant S protein, only animals vaccinated with MeV_vac2-SARS2-S(H) or alum-adjuvanted S revealed specific reactivity with approximately 400 spots/10⁶ cells or in the range of the upper limit of detection, respectively, thereby demonstrating successful induction of cellular immune responses against the SARS-CoV-2 S protein.

In the absence of suitable assays to quantify hamster IL-4, IL-5, or IL-13 protein, we determined mRNA copy numbers of these cytokines by quantitative RT-PCR in re-stimulated splenocytes and correlated the signals to the hamster housekeeping gene RPL18 mRNA copies. The total RNA from re-stimulated splenocytes was subjected to analysis (Figure 3C). For normalization, data are presented relative to the average signal in mock-immunized hamsters’ splenocytes after the respective stimulation (Figure 3D). S-specific induction of Il4 (5-fold), Il5 (>15,000-fold), or Il13 (70-fold) mRNA was only detected in splenocytes of animals immunized by alum-adjuvanted S and was significantly higher than in splenocytes of MeV-vaccinated animals or naive animals that did not show relevant up-regulation of these T_H2-cytokine genes after re-stimulation.

These data demonstrate that, in contrast to the MeV-derived vaccine, the alum-adjuvanted S protein induced S-specific T_H2-biased immune cells after vaccination, corroborating the hypothesis of VAERD induction after SARS-CoV-2 challenge specifically by Alum+S.

VAERD in hamsters responds to dexamethasone treatment

To further dissect the role of individual lung cell populations during VAERD by scRNA-seq analysis and to evaluate treatment options for VAERD, the initial challenge experiment of vaccinated hamsters was replicated (Figure 4A). Cohorts of 4–6 hamsters were vaccinated as before with alum-adjuvanted S, MeV_vac2-SARS2-S(H), or medium (MOCK). For the cohort receiving Alum+S, the number of animals was doubled to assess the impact of treatment with the clinically used immunosuppressant dexamethasone.

Figure 4.

Protection of vaccinated hamsters and pathology upon SARS-CoV-2 challenge in combination with dexamethasone treatment

(A) Schematic depiction of experimental setting. Syrian hamsters were bled and subsequently immunized with indicated vaccines on days 0 and 21 (n = 4 for mock, n = 5 for MeV_vac2-SARS2-S(H), n = 6 for both Alum+S-vaccinated cohorts). Ten days after the second immunization, hamsters were bled again. On day 35, immunized hamsters were intranasally challenged with low-passage SARS-CoV-2 and observed for another 4 days before sacrifice for organ preparation.

(B and C) Protective efficacy of vaccination was analyzed by (B) quantifying weight loss over time and (C) histopathology of lung sections prepared 4 dpi (scale bars: top row, 1 mm; second row, 20 μm; all other rows, 40 μm). For statistical analysis, two-way ANOVA was applied with Tukey’s multiple comparisons test. ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001.

(D) Blinded histopathological analysis of H&E-stained lung samples (C, second row) was performed, and findings were scored. For statistical analysis, ordinary one-way ANOVA was applied with Mann-Whitney test. ns, not significant; ^∗∗p < 0.01.

(E) Infected areas positive by ISH against SARS-CoV-2 viral RNA (C, third row) were quantified by determining the relative area of tissue staining positive for viral RNA by automated image evaluation of whole slice. Infection was confirmed by IHC against SARS-CoV-2 nucleocapsid protein (C, fourth row).

(F) Infiltration of eosinophils positive by Sirius red staining (C, bottom row) was quantified by determination of the mean number of eosinophils per mm² for each animal. Eosinophils of 10 randomly selected fields of view at 200× magnification were counted, and mean was correlated against the area of view. For statistical analysis, ordinary one-way ANOVA was applied with Tukey’s multiple comparisons test. ^∗p < 0.05; ^∗∗p < 0.01. Mock-immunized, black circles; MeV_vac2-SARS2-S(H), blue diamonds; Alum+S without (red triangles) or with (open circles) dexamethasone treatment.

To control immunization and to stratify the cohorts, sera of immunized animals were tested for binding Abs by ELISA and neutralizing Abs by titration of VNT (Figure S3). As expected, immunity against MeV was detected solely after vaccination with MeV_vac2-SARS2-S(H). Binding Abs to SARS-CoV-2 S were detected in sera from all animals that received either adjuvanted protein or the MeV-derived vaccine (Figure S3B). However, neutralizing activity inhibiting SARS-CoV-2 was evident only in hamsters immunized with recombinant MeV_vac2-SARS2-S(H) (VNT of 10–80), but not in the protein-vaccinated animals (Figure S3H). The binding Ab titers targeting SARS-CoV-2 S were used together with the animals’ sex to stratify both Alum+S cohorts for treatment with dexamethasone upon challenge or not. All animals were challenged 14 days after the second vaccination. During the next 4 days, one of both groups that had received Alum+S were treated twice daily with dexamethasone.

Upon infection, weight loss was initially observed in all groups (Figures 4B and S4A). Consistent with previous observations, hamsters immunized with MeV_vac2-SARS2-S(H) stopped losing weight by day 2 post infection (p.i.) and recovered thereafter, while naive animals revealed progressive weight loss. This was also evident for hamsters immunized with the protein vaccine that showed comparable decay. Remarkably, the protein-vaccinated, infected animals treated with dexamethasone revealed another phenotype: weight loss stopped on day 2.5, and the animals’ weights stabilized over the next 2 days. All animals were sacrificed on day 4 p.i. for subsequent necropsy.

Analysis of gross pathology replicated the phenotype observed previously (Table S2), with signs of pneumonia and inflamed areas on the lungs of naive animals, while MeV_vac2-SARS2-S(H)-immunized hamsters showed few lesions on the lungs’ surface (Figure S4B). Moreover, this experiment replicated VAERD as evident already by the overall pathology, with large areas of inflammation on the surface of the S-protein-vaccinated animals’ lungs and a swollen appearance of the entire lung. This effect was prevented in the dexamethasone-treated animals. Their lungs were comparable to the lung explants of the MeV_vac2-SARS2-S(H)-immunized cohort. Lungs were subjected to bronchoalveolar lavage (BAL) and subsequently sampled.

Histopathologic analyses of lung samples after H&E, ISH, immunohistochemistry, and Sirius red staining exactly replicated the VAERD phenotype for Alum+S-vaccinated animals and the absence thereof in MeV_vac2-SARS2-S(H)-vaccinated animals (Figure 4C). When Alum+S-immunized animals were treated with dexamethasone, the VAERD phenotype was lost. In these samples, the inflammatory phenotype was absent and lung tissue reflected the appearance of MeV_vac2-SARS2-S(H)-immunized hamsters, while the pathology score decreased significantly (Figure 4D). Despite this normalization of lung pathology, immunohistochemistry for SARS-CoV-2 N or ISH for SARS-CoV-2 genomes was not markedly different from naive challenged animals and was enhanced when compared with Alum+S-vaccinated, challenged hamsters without dexamethasone treatment (Figure 4C, last column).

Furthermore, live-virus titers and viral RNA copy numbers, as determined in lung tissue (Figures 5A and 5B) and BAL cells (Figure 5C), were in agreement with this phenotype. Virus RNA copy numbers in lung tissue were significantly reduced in hamsters vaccinated with MeV_vac2-SARS2-CoV (0.36–8.28 E-gene copies/RPL18 copy) compared with naive infected animals (110.8–235.3 E-gene copies/RPL18 copy), consistent with the absence of live virus in the lungs of all animals in the MeV group (Figure 5A). Compared with naive/unvaccinated infected hamsters, viral burden was lower in hamsters immunized with Alum+S but increased slightly when these hamsters were treated with dexamethasone during the challenge. Comparable virus RNA copy numbers were obtained for the BAL cells studied.

Figure 5.

Impact of vaccination or dexamethasone treatment on SARS-CoV-2 titers upon challenge of Syrian hamsters

Protection of vaccinated hamsters upon challenge with SARS-CoV-2 was analyzed 4 dpi by (A) titration of live-virus titers in lung tissue or (B and C) determination of relative SARS-CoV-2 E gene copy numbers in (B) lung tissue or (C) bronchoalveolar lavage (BAL) cells by quantitative RT-PCR. RPL18 housekeeping gene copies were used for normalization. Lower LoDs are indicated by dotted lines. Each data point represents an individual animal vaccinated with medium (black circles), MeV_vac2-SARS2-S(H) (blue diamonds), or alum-adjuvanted S protein without (red triangles) or with (open circles) dexamethasone treatment. For statistical analysis, ordinary one-way ANOVA was applied with Tukey’s multiple comparisons test. ns, not significant (p > 0.05), ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001.

Thus, viral loads and extent of tissue infection did not correlate with the inflammatory phenotype of pathology, consistent with an immunopathogenesis as the basis for VAERD, which again was not observed for the T_H1-biased MeV-COVID-19 model vaccine candidate. Moreover, the responsiveness of the pathology to dexamethasone treatment indicates a role for immune cells in this process.

Dysregulation of T_H2 cytokines in lungs and BAL cells

To confirm pulmonary up-regulation of Il4, Il5, Il13, and eotaxin-1 mRNAs in protein-vaccinated animals and to determine if this dysregulation was also observed in BAL cells after infection, total RNA from the respective cells was subjected to quantitative RT-PCR (qRT-PCR). These analyses revealed significant induction of all four cytokine genes in lung cells and revealed the same tendency in BAL cells after infection of protein-vaccinated animals. In contrast, vaccination with the T_H1-biased MeV_vac2-SARS2-S(H) down-regulated expression of these critical genes, while dexamethasone treatment did not alter the cytokine expression pattern (Figure S5). Therefore, induction of T_H2 cytokines in protein-vaccinated animals identified by RNA-seq in the first experiment was confirmed in this second cohort of hamsters.

Assignment of dysregulation to individual cells in lungs by scRNA-seq

To delineate the role of individual cell populations, scRNA-Seq data of infected lungs were generated and first analyzed to determine cellular subsets according to individual cellular gene expression profiles, as described previously (Nouailles et al., 2021). Accordingly, 25 individual cell clusters were defined, and their transcriptional profiles identified 13 different cell populations including immune cells (Figure 6B). Minor differences between vaccine cohorts became evident in the relative cell frequencies in the infected lungs. In naive animals, and even more so in Alum+S-vaccinated animals, lung macrophages, which revealed traits of an interstitial macrophage phenotype, were overrepresented compared with the MeV-vaccinated group. In contrast, alveolar macrophages were observed at higher frequencies in MeV-vaccinated animals’ lungs (Figure 6C).

Figure 6.

Annotation of cell populations in infected hamster lungs and respective gene regulation

(A–C) Definition (A), annotation (B), and proportion (C) of specific cell populations found in the lungs of differently vaccinated hamsters 4 dpi with SARS-CoV-2 (n = 4).

(D) Differentially regulated genes of interest in these cell populations are displayed in the presented heatmap.

(E–G) To further pinpoint dysregulation of T or NK cells, the respective cell subsets were (E) annotated according to maker genes, (F) quantified, and (G) dysregulation of genes was resolved in the respective heatmap as displayed.

(A, C, D, F, and G) Black, mock-vaccinated hamsters; red, hamsters vaccinated with Alum+S; blue, hamsters vaccinated with MeV_vac2-SARS2-S(H) (MeV-S).

The gene-expression profiles of these distinct cell populations showed specific patterns (Figure 6D). Exclusive up-regulation of Ccl11 expression in the Alum+S group could be assigned to the overrepresented lung macrophages (Figure 6D). Induction of Il4, Il5, and Il13 became evident exclusively in the population of T and natural killer (NK) cells (Figure 6D, three top panels). Zooming in on specific T/NK cell subsets (Figures 6E and 6F), up-regulation of Il4 was found in regulatory T cells, while Il5 and Il13 expression was assigned to cells with an innate lymphoid cell phenotype (Figure 6G) and activated T cells (Table S3). In contrast, MeV_vac2-SARS2-S(H)-vaccinated animals reflected a similar, but dampened, response when compared with naive infected animals, with few genes being differentially regulated. On the other hand, monitoring the distribution of SARS-CoV-2 RNA sequences indicating virus infection or uptake revealed a broad distribution of viral RNA in most cell types in naive/unvaccinated, infected animals (Figures 7A and 7B). This distribution was more focused in lung macrophages in Alum+S-vaccinated animals and correlated to some extent with the expression profiles of Fcγ receptors IIb and IV, which were found to be specifically up-regulated after infection in both naive and Alum+S-vaccinated samples (Figure 7C).

Figure 7.

Assignment of SARS-CoV-2 reads and FcR expression to cell populations in infected hamster lungs

(A and B) Assignment of SARS-CoV-2 genome reads to cell populations defined in Figure 6 (A) and quantification of respective positive cells of the different treatment groups as indicated (B). Single dots represent individual animals. Differential expression of SARS-CoV-2 RNA was analyzed using the Loupe Browser 6.0.0. Statistical significance was tested by comparison of individual cell populations with all other cells within the different treatment groups, p values were adjusted using Benjamini-Hochberg correction for multiple testings. ns, not significant; ^∗∗∗∗p < 0.0001.

(C) Assignment of mRNA reads encoding hamster FcγR IIb (top panel) or FcγR IV (bottom panel) to cell populations defined in Figure 6.

To further mine the scRNA-seq data, we performed both gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) using the REACTOME, KEGG, and GO databases on specific cell populations that contribute to the VAERD observed in Alum+S-vaccinated animals (Figures S6 and S7), focusing on alveolar macrophages, lung macrophages, monocytic macrophages, Treml4 ⁺ monocytes, and T/NK cells. For GSVA, normalized, by-cell-population gene-expression values were used as input, while DEGs between the unvaccinated and either Alum+S- or MeV_vac2-SARS2-S(H)-vaccinated conditions were utilized for GSEA. These analyses highlight pathways and biological processes, which are induced by SARS-CoV-2 infection and are differentially targeted or regulated across these selected cell types, dependent on prior immune status and vaccination type. These supportive analyses of the hamster lung scRNA-seq data provide insights into how the observed transcriptional differences mediate their downstream effects, exhibiting differential targeting of a broad range of cellular process or canonical pathways in a cell-type-specific manner.

Taken together, the scRNA-seq data assigned the up-regulation of IL-4, IL-5, and IL-13 to specific T cell subpopulations induced by vaccination with Alum+S, while CCL-11 and IL-19 expression was contributed by lung macrophages, which were overrepresented, revealed up-regulation of Fcγ receptors, and were the main target population containing an excess of SARS-CoV-2 RNA. These aberrant patterns of dysregulated gene expression were not observed in animals vaccinated with the prototypic T_H1-biased MeV-COVID-19 vaccine candidate.

Discussion

Based on these data, we propose the following immune-cell-mediated mechanism for induction of VAERD: vaccination with alum-adjuvanted S protein in de-folded conformation induces low levels of non-protective S-specific binding Abs lacking neutralizing activity. In parallel, T_H2-biased S-specific T cell responses were induced as evidenced by the significant up-regulation of IL-4, IL-5, and IL-13 after recall. After infection, these immune responses lowered the virus load to some extent, but the induction of T_H2 immunity promoted VAERD via the IL-4/IL-5/IL-13 chemokine axis secreted by regulatory T cells and innate lymphoid cells. These processes are likely reinforced by IL-19 secreted by hyperstimulated lung macrophages, which, in addition, express further chemo-attractants such as CCL-11, the major attractant of eosinophils. Indeed, massive infiltration of eosinophils completed the picture of VAERD. Macrophage activity correlated with SARS-CoV-2 virus genome enrichment. It is therefore tempting to hypothesize that the enrichment of SARS-CoV-2 RNAs in macrophages is caused by Fcγ-receptor-mediated uptake of opsonized viruses as previously proposed for ADE processes and recently demonstrated for enrichment of SARS-CoV-2 in monocytes of the peripheral blood of patients with severe COVID-19 (Junqueira et al., 2022).

While our data support such a model for induction of VAERD by sub-optimal T_H2-biased prototypic vaccines targeting SARS-CoV-2, the nature of these findings seems quite striking. VAERD had been clearly observed for both MERS-CoV and SARS-CoV in the respective transgenic mouse models after vaccination with T_H2-biased, whole-inactivated virus vaccines (Bolles et al., 2011; Tseng et al., 2012; Iwata-Yoshikawa et al., 2014; Honda-Okubo et al., 2015; Agrawal et al., 2016). Since no human vaccines had been tested against these two pathogens, the transferability of these findings to the occurrence of VAERD in humans remained unclear. Nevertheless, this potential risk was perceived also for vaccine-induced immune responses that target the closely related SARS-CoV-2 and triggered vaccine developers to aim for T_H1-biased immunity and to assess the potential of immunopathogenesis in the available animal models (Anderson et al., 2020; Corbett et al., 2020a, 2020b; Jackson et al., 2020; Polack et al., 2020; Ramasamy et al., 2020; Walsh et al., 2020; Sadoff et al., 2021; Stephenson et al., 2021; van der Lubbe et al., 2021). Despite all these efforts, only two studies have been published so far that revealed evidence for VAERD potential upon vaccination of T_H2-prone BALB/c mice with denatured antigen followed by a challenge with a mouse-adapted recombinant SARS-CoV-2 (DiPiazza et al., 2021; Iwata-Yoshikawa et al., 2022). While using an unbiased animal model and non-adapted SARS-CoV-2 and going beyond the mechanistic hypotheses expressed (Iwata-Yoshikawa et al., 2022), our study aligns well with these mouse data in the observed VAERD phenotype with T_H2 cytokine secretion and massive eosinophil attraction. However, no other studies so far have identified evidence for the risk of enhanced disease upon COVID-19 vaccination. How can this discrepancy be explained?

The sub-optimal immunization regime of vaccinating Syrian hamster subcutaneously with alum-adjuvanted, non-stabilized S protein was expected to give a rather mediocre (Nürnberger et al., 2019), but T_H2-biased, immune response (Ko et al., 2017) as a worst-case control in our earlier study describing a MeV-derived COVID-19 vaccine candidate (Hörner et al., 2020). Indeed, only binding Abs with no neutralizing activity were induced, and no antigen-specific CD8⁺ T cell killing activity was observed by us (Hörner et al., 2020) or others (DiPiazza et al., 2021; Iwata-Yoshikawa et al., 2022). Our observations thereby closely resemble VAERD induced by inactivated RSV (Johnson and Graham, 1999; De Swart et al., 2002; Johnson et al., 2003, 2004; Moghaddam et al., 2006) or MeV in the respective animal models (Polack et al., 2003), which are quite reminiscent of the situation observed in human VAERD after RSV (Openshaw, 2001) or measles vaccination (Nader and Warren, 1968) using whole-inactivated virus vaccines. Moreover, induction of protective immune responses, or in the worst-case scenario, immunopathogenesis, represents a complex interplay of innate immune processes that set the stage for humoral and cellular adaptive immunity. This can also be deduced from our data. While dexamethasone did not inhibit T_H2 cytokine expression, here, it clearly reduced pathology in the VAERD setting, in addition to naive infected hamsters (Wyler et al., 2022; Yuan et al., 2022). Thereby, a quite delicate process is revealed, balancing immune activities destructive against pathogen or host tissue, which may be difficult to trigger.

Limitations of the study

As already pointed out by DiPiazza et al., drawing conclusions from animal models and extending these observations to the immunological situation in humans may be a difficult task (DiPiazza et al., 2021). However, we were able to replicate VAERD induced by a protein-based COVID-19 vaccine candidate with a very similar phenotype in a second animal model, Syrian hamsters, using an unmodified low-passage virus isolate. Therefore, our data strongly support the idea of monitoring vaccinated human patients that experience a break-through infection closely. In any case, while our experimental vaccines mimic, but are not the same as, the authorized vaccines, these and previously published data point to few concerns for vaccines developed to trigger T_H1-biased responses such as viral vector platform-based vaccines, mRNA vaccines, or other vaccine concepts developed for such bias by, e.g., using respective adjuvants. Moreover, even if VAERD as observed in our model should occur in human patients, this immunopathology would be treatable by dexamethasone, which was confirmed to be an effective medication for severe courses of COVID-19 anyway (Tomazini et al., 2020; Horby et al., 2021). This would be good news also for putative VAERD being mistakenly diagnosed as a variant of the usual forms of severe COVID-19 in a naive patient.

A further limitation of our study is the restriction of our analyses to a single time point of disease. It would be interesting to analyze further progression of the animals with or without VAERD. However, the applied challenge model is rather harsh. The hamsters lose weight after infection with a highly reproducible kinetics that necessitates sacrifice of 50%–83% of naive animals due to humane endpoints 5 to 6 days after infection (Hörner et al., 2020), and statistically significant differences cannot be expected by realistic animal numbers. Therefore, day 4 after infections was chosen as the day of analyses since this is the time point during infection when most dramatic effects can be expected (measurable weight loss starts at day 2 p.i.), but all animals are still available.

In conclusion, our data support the idea that ADE-like processes, as reported by Wan et al. for MERS-CoV, could be relevant in the mechanism of VAERD. For MERS-CoV, a monoclonal Ab binding to the RBD of the S is able to cross-link MERS-CoV S and Fc receptors. When tested in a pseudovirus assay, this monoclonal Ab mediated virus entry into CD32A-expressing (i.e., FcγRIIA-positive) cells and macrophages (Wan et al., 2020). Such an uptake mechanism would explain the enrichment of SARS-CoV-2 genomes in macrophage populations correlating with the Fc-receptor distribution as observed in our study. Recent clinical data have demonstrated evidence that monocytes in the peripheral blood of patients with severe COVID-19 take up opsonized SARS-CoV-2 particles via this route and are impacted by this process (Junqueira et al., 2022). In any case, enrichment of SARS-CoV-2 in and stimulation of this immune cell population result in the secretion of the major eosinophil attractant CCL-11 and IL-19, which can drive regulatory T cells (T_regs) into T_H2 polarization. Together with the IL-4/IL-5/IL-13 chemokine axis of such S-specific T_H2-biased CD4 helper cell populations, CCL-11 can be expected to cause infiltration of eosinophils in a process mechanistically reminiscent of allergic asthma, resulting in the immunopathogenesis observed in our study.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
KPL HRP-conjugated goat anti-hamster IgG	Seracare	Cat#5220-0371
mAb against SARS-CoV nucleocapsid protein	Bussmann et al. (2006)	clone 4F3C4
secondary biotinylated goat anti-mouse Ab	Vector Laboratories	Cat#BA-9200-1.5
mAb against M protein of Influenza A virus	ATCC	clone HB-64
Bacterial and virus strains
MeVvac2-SARS2-S(H)	Hörner et al. (2020)	N/A
MVvac2-GFP(P)	Malczyk et al. (2015)	N/A
SARS-CoV-2 isolate MUC-IMB1	Böhmer et al. (2020)	GenBank# MT270101.1
Chemicals, peptides, and recombinant proteins
SARS-CoV-2 (2019-nCoV) Spike S1+S2 ECD-His Recombinant Protein	Sino Biological Europe	Cat#40589-V08B1
SARS-CoV-2 Trimeric Spike (stabilized Spike)	NIBSC	Cat#101007
Rec. Ovalbumin	Schülke et al. (2010)	RefSeq# NM205152
Rec. Flagellin A	Schülke et al. (2010)	RefSeq# X65624
Concanavalin A	Sigma-Aldrich	Cat#11028-71-0
MeV bulk antigens	Virion Serion	Cat#BA102VS-S
Dispase	Corning	Cat#354235
DNAse	AppliChem	Cat#A3778,0100
Collagenase B	Roche	Cat#11088815001
Actinomycin D	Sigma-Aldrich	Cat#A9415-5MG
Aluminum hydroxid (Allhydrogel adjuvant 2%)	InvivoGen	Cat#Vac-alu-250
3-Amino-9-ethyl-carbozole	Sigma-Aldrich	Cat#A5754-10G
Dexamethosone	Bela-pharm	Cat#6933074.00.00
Direktrot 80	Sigma-Aldrich	Cat#365548
Entellan	Sigma-Aldrich	Cat#107060
Papanicolaous solution 1b Hematoxylin S	Sigma-Aldrich	Cat#1.09254
RBC Lysis buffer	Santa Cruz Biotechnology	Cat#sc-296258
RNAscope® Pretreatment Reagents	Bio-Techne	Cat#322330 and 322000
RNAscope® Negative Control Probe	Bio-Techne	Cat#310043
TMB substrate	Invitrogen	Cat#00-4201-56
TRIzol Reagent	Ambion, Thermo Fisher Schientific	Cat#15596026
Critical commercial assays
Chromium Next Gem Chip G Single Cell Kit	10x genomics	Cat#1000121
Chromium Next Gem Single cell 3ʹ GEM, Library & Gel Bead Kit v3.1	10x genomics	Cat#1000120
Direct-zol RNA MiniPrep kit	Zymo research	Cat#R2052
Hamster IFN-γ ELISpotBASIC kit	MABTECH	Cat#3102-2H
QIAseq FastSelect –rRNA HMR Kit	Qiagen	Cat#334385
RNAscope® 2.5 HD Assay – RED Kit	Bio-Techne	Cat#322360
Superscript III one step RT-PCR system with Platinum Tag Polymerase	Invitrogen	Cat#12574-026
Vectastain Elite ABC Kit	Vector Laboratories	Cat#PK6100
Deposited data
Bulk RNA-seq	This Paper	GEO: GSE195939
scRNA seq data	This Paper	GEO: GSE196938
Experimental models: Cell lines
Vero (African green monkey kidney)	ATCC	CCL-81
Vero clone E6	ATCC	CRL-1586
Experimental models: Organisms/strains
Syrian golden hamster (Mesocricetus auratus); HsdHan:AURA	Envigo	Art#8903
Oligonucleotides
Table S4
Software and algorithms
ELI.Analyse V5.0	AE.L.VIS	http://www.aelvis.de
fastp algorithm	Chen et al. (2018)	https://github.com/OpenGene/fastp
CLC Genomics Workbench 21.0.4	QIAGEN	https://digitalinsights.qiagen.com
CellRanger v5.0	10x genomics	https://support.10xgenomics.com/single-cell-gene-expression/software
R packages Seurat v4.0	Hao et al. (2021)	https://satijalab.org/seurat/
DoRothEA v3.12	Holland et al. (2020)	https://saezlab.github.io/dorothea/
GSVA R package	Hänzelmann et al. (2013)	http://bioconductor.org/packages/release/bioc/html/GSVA.html
clusterProfiler R package	Wu et al. (2021)	https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html
Prism 9.2.0	GraphPad Software	http://www.graphpad.com
Original code used in the analysis of the single cell data	This Paper	https://github.com/Berlin-Hamster-Single-Cell-Consortium/Vaccine-associated-enhanced-respiratory-pathology-in-COVID-19-hamsters-after-TH2-biased-immunization
Other
Lysing Matrix M tubes	MP biosciences	Cat#6923100
J-810 Spectropolarimeter	JASCO	https://www.jasco.de/en
Eli.Scan ELISpot scanner	AE.L.VIS	http://www.aelvis.de
Qubit 4 Fluorimeter	Thermo Fisher	Cat#Q33238
Precellys24 tissue homogenizer	bertin TECHNOLOGIES	Cat#P000669-PR240-A
2100 Bioanalyzer Instrument	Agilent	Part#G2939BA
Novaseq 6000	Illumina	Cat#20013850
NextSeq 550	Illumina	Cat#SY-415-1002
Hamamatsu S60 scanner	Hamamatsu Photonics	Prod#C13210
Chromium Controller	10x Genomics	Prod# 1000204

Resource availability

Lead contact

Lead Contact for requests concerning ressources is Michael D. Mühlebach, email: Michael.Muehlebach@pei.de.

Materials availability

Materials will be available upon request; (recombinant) viruses or plasmids may require filing of materials transfer agreements (MTAs) and compensation for shipment.

Experimental model and subject details

Cells

Vero (African green monkey kidney; ATCC# CCL-81) and Vero clone E6 (ATCC# CRL-1586) cells were purchased from ATCC (Manassas, VA, USA) and cultured in Dulbecco’s modified Eagle’s medium (DMEM, Sigma Aldrich, Steinheim, Germany) supplemented with 10% fetal bovine serum (FBS; Sigma Aldrich) and 2 mM L-glutamine (L-Gln, Sigma Aldrich). Cell cultures were incubated at 37°C in a humidified atmosphere containing 6% CO₂ up to 30 passages after thawing of the initial stocks to ensure authentication of utilized cell lines.

Viruses

MeV_vac2-SARS2-S(H) (Hörner et al., 2020) and MV_vac2-GFP(P) (i.e. MV_vac2-ATU(P) (Del Valle et al., 2007) with GFP inserted in the ATU (Malczyk et al., 2015) have been described previously. Subsequent passages were generated after TCID₅₀ titration of infectious virus according to the method of Kaerber and Spaerman (Kärber, 1931). Stocks were generated by infection of Vero cells at an MOI = 0.03, and viruses in P3 or P4 were used for vaccination experiments. SARS-CoV-2 isolate MUC-IMB1 (Böhmer et al., 2020) was used in passage 3 on Vero-E6 cells after isolation from the patient as described before (Hörner et al., 2020).

Syrian golden hamster animal model

All animal experiments were carried out in compliance with the regulations of German animal protection laws and as authorized by the RP Darmstadt and reported according to the ARRIVE guidelines. Six to 12-weeks old Syrian hamsters (Envigo RMS, Venray, Netherlands) were randomized for age- and sex-matched groups. Male and female hamsters were used in a 1:1 ratio to ensure transferability of results to both sexes, cohort sizes (n = 6) were chosen to ensure statistical validation with an error probability of p < 0.05 with a power of 80% conservatively assuming an effect size of 0.8. Animals were vaccinated intraperitonally (i.p.) in a prime-boost schedule (days 0 and 21) with 5×10⁵ TCID₅₀ of recombinant MeV-derived vaccine virus in 200 μL volume or subcutaneously (s.c.) with 10 μg recombinant SARS-CoV-2 S protein (Cat-No. 40589-V08B1, Sino Biological Europe, Eschborn, Germany) adjuvanted with 500 μg aluminum hydroxide (Allhydrogel adjuvant 2%, vac-alu-250, InvivoGen, San Diego, CA, USA) in 100 μL volume. Blood was drawn on days 0 and 21 or 31. Splenocytes of vaccinated animals were isolated 14 days after second immunization or hamsters were challenge by intranasal application of 4×10³ TCID₅₀ SARS-CoV-2 (isolate MUC-IMB1) in passage 3 in 100 μL volume. Alum+S vaccinated hamsters were divided in two groups: i) a untreated group and ii) a dexamethasone treated one (1 mg/kg body weight twice daily orally or subcutaneously). Animals were euthanized 4 days after infection. Specific sections of each lung were prepared for histology (left lobe), analysis of total RNA (right middle lobe), titration of live virus (right apical lobe), and analysis of transcriptomics by scRNA-Seq (caudal lobe), where applicable.

Method details

CD spectroscopy

The secondary structure of SARS-CoV-2 Spike was analysed using circular dichroism spectroscopy and J-810 spectropolarimeter (JASCO) equipped with a quartz cuvette (0.1 cm). Spectra were recorded at room temperature from 255 nm–185 nm by accumulating 10 runs (bandwidth 1 nm, scanning speed 50 nm/min) using 0.16–0.58 μM protein solutions in 20 mM sodium phosphate buffer pH 8.0.

Bronchoalveolar lavage (BAL)

Lungs from sacrificed hamsters were inflated with 2.5 mL PBS via the trachea. The BAL fluid was collected and cells were harvested by centrifugation (1,200 rpm, 4°C, 5 min). The cell-free fluids were removed and the cells were resuspended in 350 μL TRIzol Reagent (Ambion, Thermo Fisher Scientific).

Virus neutralization test (VNT)

Virus neutralization tests (VNT) were performed as described previously (Hörner et al., 2020). In short, serum samples were diluted in 2-fold series in DMEM. 50 PFU MV_vac2-GFP(P) or 100 TCID₅₀ SARS-CoV-2 were mixed with diluted serum samples and incubated at 37°C for 1 h. Subsequently, the virus-serum mixture was added to 1×10⁴ Vero or Vero E6 cells seeded 3 h before in 96 well plates (Thermo Fisher Scientific, Ulm, Germany). Cells were incubated for 4 days at 37°C in a humidified atmosphere containing 6% CO₂. Virus neutralizing titers were determined as the reciprocal of the highest serum dilution that completely abrogated infectivity.

Total IgG quantification

0.5 μg MeV bulk antigens (Virion Serion, Würzburg) or 0.25 μg recombinant SARS-CoV-2 S protein were coated in 50 μL/well carbonate buffer (Na₂CO₃ 30 mM; NaHCO₃ 70 mM; pH 9.6) on Nunc Maxisorp® 96 well ELISA plates (eBioscience). After overnight incubation at 4°C, the plates were washed three times with 200 μL PBS containing 0.1% Tween 20 (w/v) and blocked with 100 μL PBS containing 5% BSA and 0.1% Tween 20 for at least 2 h at ambient temperature. Hamster sera were 5-fold serially diluted in PBS containing 1% BSA and 0.1% Tween 20. 50 μL/well of serum dilutions were used in the assay and the plates were incubated at 37°C for 2 h prior to washing the plates three times using PBS containing 0.1% Tween20. Subsequently, 50 μL/well of HRP-conjugated goat anti-hamster IgG (1:1,000 in PBS containing 1% BSA and 0.1% Tween20; Sera care KPL, Cat. 5220-0371) were added and the plates were incubated for 1 h at room temperature. After washing the plates three times, 100 μL/well TMB substrate (Invitrogen) was added. The reaction was stopped by addition of 50 μL/well 1 N H₂SO₄ and the absorbance at 450 nm (specific signal) and 630 nm (reference wavelength) was measured.

IFN-γ ELISpot analysis

Hamster interferon gamma (IFN-γ) enzyme-linked immunosorbent spot (ELISpot) analysis was performed using the Hamster IFN-γ ELISpot^BASIC kit (MABTECH, Nacka Strand, Sweden) in combination with multiscreen immunoprecipitation (IP) ELISpot polyvinylidene difluoride (PVDF) 96-well plates (Merck Millipore, Darmstadt, Germany) according to the manufacturer’s instructions. 5×10⁵ isolated splenocytes were co-cultured with different stimuli in 200 mL RPMI containing 10% FBS, 2 mM L-Gln, 10 mM HEPES pH 7.4, 50 mM 2-mecaptoethanol and 1% Penicillin-streptomycin. To re-stimulate SARS-CoV-2 specific T cells, isolated splenocytes were cultured with 10 mg/mL recombinant SARS-CoV-2 (2019-nCoV) Spike Protein (S1+S2 ECD, His tag) (Sino biological Europe). Recombinant Ovalbumin [10 mg/mL] served as negative protein control. In parallel, splenocytes were stimulated with 10 mg/ml MeV bulk antigen (Virion Serion, Würzburg, Germany). General stimulation of T cells was achieved using 5 μg/mL concanavalin A (ConA, Sigma-Aldrich) or recombinant 20 μg/mL Flagellin A produced in house (Schülke et al., 2011). Untreated splenocytes served as negative control. After 36 h of stimulation, cells were removed and plates were incubated with biotinylated detection antibodies and Streptavidin-HRP conjugate following the manufactures introductions using a 1 in 100 dilution for the streptavidin-HRP conjugate with 3-Amino-9-ethyl-carbazole (AEC; Sigma-Aldrich) dissolved in N,N-dimethylformamide (Merck Millipore) as substrate. Spots were counted using an Eli.Scan ELISpot scanner (AE.L.VIS, Hamburg, Germany) and analysis software ELI.Analyse V5.0 (AE.L.VIS).

Determination of infectious virus lung titers

The right apical lobe of the lungs of infected animals was snap-frozen in liquid nitrogen and homogenized in 1 mL ice-cold DMEM containing 2 mM L-Gln and 1% Penicillin/Streptomycin in Lysing Matrix M tubes (MP Bioscience, Hilton, UK) using the Precellys24 tissue homogenizer (bertin TECHNOLOGIES, Montigny-le-Bretonneux, France) for 2 × 10 s at 6,000 rpm. Samples were kept on ice at all times. Subsequently, organ debris was removed by centrifugation (13 min, 6,800 rpm, 4°C). Vero E6 cells were inoculated with the supernatants in a 10-fold dilution series for 7 d at 37°C. SARS-CoV-2 organ titers were calculated by the TCID₅₀ method of Kaerber and Spearman according to virus-induced CPE and adjusted for 1 g of tissue.

RNA preparation

The right middle lobe of the lungs of infected animals was homogenized in 1 mL TRIzol Reagent (Ambion, Thermo Fisher Scientific) in Lysing Matrix M tubes (MP Bioscience, Hilton, UK) using the Precellys24 tissue homogenizer (bertin TECHNOLOGIES) for 2 × 15 s at 6,000 rpm. Samples were kept on ice at all times. Organ debris was removed by subsequent centrifugation (13 min, 6,800 rpm, 4°C). Clear supernatant was used for RNA purification with Direct-zol RNA MiniPrep kit (Zymo research, Freiburg (Breisgau), Germany) according to the manufactures introduction.

Quantitative reverse-transcription PCR (qRT-PCR)

RNA samples were quantified by quantitative reverse transcription-PCR (qPT-PCR) using Superscript III one step RT-PCR system with Platinum Tag Polymerase (Invitrogen, Darmstadt, Germany). Primer and probe sequences for mRNA encoding the SARS-CoV-2 E gene (Corman et al., 2020), hamster RPL18 (Zivcec et al., 2011), IL-4, and IL-13 (Espitia et al., 2010) were used as described and are indicated in Table S4. Primers for detection of Eoatxin-1 (Stanelle-Bertram et al., 2020) and forward primer sequences for IL-5 (Mendlovic et al., 2015) were ordered as described. The reverse primer sequence for IL-5 was adapted according to RNA Seq results of hamster lungs as described in this manuscript. Probes for mRNA encoding Eotaxin and IL-5 were designed as shown in Table S3. Reactions were run in 96-well plates (Bio-Rad Laboratories, Hercules, CA) using CFX96 qPCR cycler (Bio-Rad Laboratories) and 5 μL RNA in a total reaction volume of 25 μL in triplicates. An internal Hamster reference (linear range, 4.5×10⁶ to 4.5×10² copies (Hörner et al., 2020)) was used for quantification of SARS-CoV-2 E gene copy numbers. This reference was validated for copy numbers of RPL18 housekeeping gene by utilization of a PCR product DNA reference generated as described (Osterrieder et al., 2020), and was used for quantification in subsequent runs (linear range, 1.8×10⁵ to 1.82×10² copies). The following cycling conditions were used for all analyses: reverse transcription for 10 min at 55°C, denaturation for 180 s at 94°C, followed by 45 cycles of 15 s at 94°C and 30 s at 58°C. Quantified sample copy numbers were normalized to copy numbers of the hamster housekeeping gene RPL18. If direct quantification was not possible, the ΔΔct method was used.

Total RNA-Seq

The isolated RNA samples were used for NNSR priming based RNA-Seq library preparation (Levin et al., 2010) (as described in (Brown et al., 2020), vRNA NGS section) with the following modifications. Total RNA, was subjected to rRNA removal using the QIAseq FastSelect –rRNA HMR Kit (Qiagen) in combination with reverse transcription as follows. A 35 μL reaction mixture containing 1 μg RNA, 100 pmol NNSR_RT primer (gctcttccgatctctNNNNNN), 8 μL of 5× SuperScript IV buffer (Invitrogen) 20 pmol dNTPs and 1 μL of FastSelect-rRNA mix was subjected to the following hybridization protocol: 75°C 2 min, 70°C 2 min, 65°C 2 min, 60°C 2 min, 55°C 2 min, 37°C 5 min, 25°C 5 min, store at 4°C. For cDNA synthesis the reaction above was supplemented with dithiothreitol (10 mM), 20 U of RiboLock ribonuclease inhibitor (Thermo Fisher Scientific) and 200 U of SuperScript IV reverse transcriptase in a final reaction volume of 40 μL and incubated 45°C 5 min, 70°C 15 min.

The smears of 200–500 base pairs of the final barcoded libraries were purified from a 1.5% agarose gel and sequenced on a NextSeq 550 Illumina instrument using a single-end 86 bp setting. The RNS-Seq library preparation method used results in reads that start with the same two initial nucleotides. Hence, these were removed when performing quality- and minimum length- read trimming with the fastp algorithm (Chen et al., 2018) with the default parameters. Read mapping to the Hamster genome scaffold, statistical analyses of DEGs between groups and GO analyses were performed using CLC Genomics Workbench (QIAGEN).

Isolation of single cells from lung tissue

The right caudal lobe of the lung was removed from the body and transferred on ice in PBS containing 1% BSA (w/v) and 2 mg/mL Actinomycin D (Sigma-Aldrich) for further processing. Lung tissue was incubated in 2 mL Dispase (Corning, Bedford, MA, USA) containing 2 mg DNAse (AppliChem, Darmstadt, Germany), 4.6 mg Collagenase B (Roche, Basel, Switzerland) and 2 mg/mL Actinomycin D (Sigma-Aldrich) and at 37°C for 30 min. Digest was stopped by the addition of cold PBS containing 0.5% (w/v) BSA and Actinomycin D. The tissue was disrupted by pipetting in a repeated pumping motion. Cell suspension was collected and filtered through 70 mm-filter to obtain a single-cell suspension. Red Blood cell were lysed by incubation with RBC lysis buffer (Santa Cruz Biotechnology, Dallas, Texas, USA) for 4 min at room-temperature. Lysis reaction as stopped by the addition of PBS containing 0.04% BSA, the supernatant was removed by centrifugation (1,200 rpm, 6 min, 4°C) and cells were resuspended in PBS containing 0.04% BSA. Barcoding of single cells and RNA isolation was performed using the Chromium controller and Chromium Next Gem Chip G Single Cell Kit and Chromium Next Gem Single cell 3′ GEM, Library & Gel Bead Kit v3.1 (10x Genomics B.V., Leiden, The Netherlands) according to the manufactures instructions.

Single-cell RNA-Seq

After enzymatic fragmentation and size selection, resulting double-stranded cDNA amplicons optimized for library construction were subjected to adaptor ligation and sample index PCRs needed for Illumina bridge amplification and sequencing according to the manufacturer’s instruction (10x genomics). Single cell libraries were quantified using Qubit (Thermo Fisher) and quality-controlled using the Bioanalyzer System (Agilent). Sequencing was performed on a Novaseq 6000 (Illumina), aiming for 200 Million reads per library (read1: 28, read2: 150 nucleotides). Data were analysed using CellRanger v5.0 (10x Genomics) using hamster and SARS-CoV-2 genome scaffolds, and the R packages Seurat v4.0 (Hao et al., 2021) and DoRothEA v3.12 (Holland et al., 2020) were used for cell clustering, annotation, and transcription factor activity analysis. Median gene number detected per cell ranged between 2000 and 4400, with 3800–18500 median UMI counts per cell. Gene set variation analysis (GSVA) was performed using the GSVA R package (Hänzelmann et al., 2013) and gene set enrichment analysis was performed using the clusterProfiler R package (Wu et al., 2021).

Histopathology

The left lung lobe was carefully removed and immersion-fixed in 10% neutral-buffered formalin for 7 days. The tissue was subsequently paraffin-embedded and sections of 4 μm were prepared.

Hematoxilin-eosin staining

Hematoxylin-eosin staining was carried out in accordance with standard procedures (Mulisch and Welsch, 2015). H&E stained slices were subjected to histopathologic analyses on blinded samples.

Sirius red staining

For Sirius Red staining of lung tissue sections we followed the protocol published by Llewellyn with some modifications (Llewellyn, 1970). Briefly, sections were placed in Papanicolaous solution 1b Hematoxylin S (Sigma Aldrich) for 2 min and rinsed afterwards with water followed by ethanol, 3% HCl in ethanol and 70% ethanol. Subsequently, sections were stained for 90 min in alkaline Sirius red (0.5 g Direktrot 80, Sigma in 50% ethanol containing 0.1‰ NaOH) before rinsing with water. Sections were dehydrated afterwards with increasing ethanol concentrations and xylene. Finally, sections were covered with Entellan (Merck KGaA, Darmstadt, Germany).

In situ hybridisation

To detect viral RNA in the lungs, fixed paraffin-embedded tissue sections were mounted on glass slides and analyzed by in situ hybridization as described previously (Halwe et al., 2021; Tscherne et al., 2021). For this, the RNAscope® 2.5 HD Assay – RED Kit (Bio-Techne, cat. no. 322360) was used according to the manufacturer’s instructions. Slides were incubated at 60°C, deparaffinized with xylene and 100% ethanol and pretreated with RNAscope® Pretreatment Reagents (cat. no. 322330 and 322000), to enable access to the target RNA. Subsequently, the RNA-specific probe, targeting the S protein of the SARS-CoV-2 virus (cat. no. 848561) was hybridized to the RNA. After the amplification steps, Fast Red substrate was administered to the samples for signal detection. Slides were counterstained with Gill’s Hematoxylin I and 0.02% ammonia water. A RNAscope® Negative Control Probe (cat. no. 310043) was used in parallel to control background staining.

Immunohistochemistry

For SARS-CoV-2 antigen detection, a monoclonal Ab against the nucleocapsid protein (clone 4F3C4 (Bussmann et al., 2006)) was used according to standardized procedures of avidin-biotin-peroxidase complex-method (ABC, Vectastain Elite ABC Kit, Burlingame, CA, USA). Briefly, 2–3 μm sections were mounted on adhesive glass slides, dewaxed in xylene, followed by rehydration in descending graded alcohols. Endogenous peroxidase was quenched with 3% hydrogen peroxide in distilled water for 10 min at room temperature. Antigen heat retrieval was performed in 10 mM citrate buffer (pH 6) for 20 min in a pressure cooker. Nonspecific Ab binding was blocked for 30 min at room temperature with goat normal serum, diluted in PBS (1:2). The primary Ab was applied overnight at 4°C (1:50, diluted in TRIS buffer), the secondary biotinylated goat anti-mouse Ab was applied for 30 min at room temperature (Vector Laboratories, Burlingame, CA, USA, 1:200). Color was developed by incubating the slides with freshly prepared avidin-biotin-peroxidase complex (ABC) solution (Vectastain Elite ABC Kit; Vector Laboratories), followed by exposure to 3-amino-9-ethylcarbazole substrate (AEC, Dako, Carpinteria, CA, USA). The sections were counterstained with Mayer’s haematoxylin and coverslipped. As negative control, consecutive sections were labelled with an irrelevant Ab (M protein of Influenza A virus, ATCC clone HB-64). A positive control slide was included in each run. Slides were scanned using a Hamamatsu S60 scanner (Hamamatsu Photonics, K.K. Japan).

Quantification and statistical analysis

Animal cohort or sample size per treatment was between n = 4 and n = 6 as defined in the individual figure legends. Statistical analysis of different experiments is indicated in detail in the respective figure legends. For statistical analysis of samples from animal experiments, ordinary one-way or two-way ANOVA was applied with Tukey’s multiple comparisons test using Prism. For statistical analysis of grouped ELISpot data, paired t test was applied using Prism. For statistical analysis of histopathology, ordinary one-way ANOVA was applied with Mann-Whitney test. For statistical analysis of qRT-PCR analyses, ordinary one-way ANOVA was applied with Tukey’s multiple comparisons test using Prism.

Bulk RNAseq libraries from different hamster lungs (n = 4) were generated for each condition (mock uninfected, infected, Alum+S vaccinated + infected, and MeV_vac2-SARS2-S(H) vaccinated + infected). An average of 13,400,912.5 (range 8,468,387 to 22,180,134) reads mapped to the hamster genome scaffold MesAur1.0.102, an average of 90.1% (range 87.95%–92.04%) of the total number of reads, with 98.7% (range 97.61%–99.27%) representing protein-coding segments. An average 62,455.9 (range 2,649 to 205,001) reads mapped to SARS-CoV-2 reference genome (NCBI accession #LR824570), about 0.41% (range 0.03%–1.25%) of the total number of reads. For statistical comparison, read counts of libraries of vaccinated animals were tested against mock animals using CLC Genomics Workbench 21.0.4 employing multi-factorial statistics based on a negative binomial Generalized Linear Model (GLM). p-values of differentially expressed genes (DEG) were corrected with FDR correction method.

Four scRNA-seq libraries from four different infected hamster lungs were generated for each condition (mock, Alum+S vaccinated, and MeV_vac2-SARS2-S(H) vaccinated) and an average number of 2,679 cells (range: 1,392 to 4,645) were retrieved per library with an average read count of 7,271 mapped reads per cell. Differential expression of selected cytokines in different celltype compartments between treatment groups was performed using the FindMarkers function from the Seurat R package with the default parameters (Wilcoxon Rank Sum test with Bonferroni p value correction).

Differential expression of SARS-CoV-2 RNA in scRNA-Seq experiments was analysed using the Loupe Browser 6.0.0 software. Statistical significance was tested by comparison of individual cell populations with all other cells within the different treatment groups, p values were adjusted using Benjamini-Hochberg correction for multiple testings.

We did not determine by statistical tests whether the data met assumptions of the statistical approach, also for the reason that tests were conservatively chosen for data requirements to have to meet as few pre-assumptions as possible.

Published: August 3, 2022

Data and code availability

Data and Code used in this study are fully available. Bulk RNA-seq and scRNA-seq data generated in this study were submitted to the NCBI GEO database and can be found under accession number NCBI GEO: GSE195939 and GSE196938, respectively. All original code used in the analysis of the single cell data has been deposited at https://github.com/Berlin-Hamster-Single-Cell-Consortium/Vaccine-associated-enhanced-respiratory-pathology-in-COVID-19-hamsters-after-TH2-biased-immunization. All other data are available in the main text or the supplemental information. Any additional information required to reanalyse the data reported in this paper is available from the lead contact upon request.