Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Mar 21.
Published in final edited form as: Nature. 2022 Sep 21;611(7935):346–351. doi: 10.1038/s41586-022-05344-2

Common human genetic variants of APOE impact murine COVID-19 mortality

Benjamin N Ostendorf 1,4,5,*, Mira A Patel 1,#, Jana Bilanovic 1,#, H-Heinrich Hoffmann 2, Sebastian E Carrasco 3, Charles M Rice 2, Sohail F Tavazoie 1,*
PMCID: PMC10957240  NIHMSID: NIHMS1913990  PMID: 36130725

Abstract

Clinical outcomes of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection are highly heterogeneous, ranging from asymptomatic infection to lethal coronavirus disease 2019 (COVID-19). The factors underlying this heterogeneity remain insufficiently understood. Genetic association studies have suggested that genetic variants contribute to the heterogeneity of COVID-19 outcomes, but the underlying potential causal mechanisms are insufficiently understood. Here we show that common variants of the Apolipoprotein E (APOE) gene, homozygous in approximately 3% of the world’s population1 and associated with Alzheimer’s disease, atherosclerosis and anti-tumor immunity2–5, impact COVID-19 outcome in a mouse model that recapitulates increased susceptibility conferred by male sex and advanced age. Mice bearing the APOE2 or APOE4 variant exhibited rapid disease progression and poor survival outcomes relative to mice bearing the most prevalent APOE3 allele. APOE2 and APOE4 mice exhibited increased viral loads as well as suppressed adaptive immune responses early after infection. In vitro assays demonstrated increased infection in the presence of APOE2 and APOE4 relative to APOE3, indicating that differential outcomes are mediated by differential effects of APOE variants on both viral infection and antiviral immunity. Consistent with these in vivo findings in mice, APOE genotype was associated with survival in SARS-CoV-2 infected patients in the UK Biobank (candidate variant analysis, P = 2.6×10−7). Our findings suggest APOE genotype to partially explain the heterogeneity of COVID-19 outcomes and warrant prospective studies to assess APOE genotyping as a means of identifying patients at high risk for adverse outcomes.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused the COVID-19 pandemic with more than 580 million confirmed infections and 6 million deaths to date worldwide. Clinical presentations of SARS-CoV-2 infection show pronounced variation, ranging from asymptomatic infection to lethal disease. Several epidemiological factors have been identified that associate with adverse outcome, including male sex, advanced age, select comorbidities, and genetic ancestry6. However, these factors only partially explain the wide inter-individual clinical spectrum of SARS-CoV-2 infection. There is thus a major need to identify the factors underlying susceptibility to poor outcome in COVID-19. Major efforts have shown germline genetics to correlate with disease severity in COVID-19 (reviewed in7). Amongst these, candidate gene approaches have revealed rare autosomal inborn errors of type I interferon (IFN) immunity to alter type I IFN signaling in vitro8,9. In addition, genome-wide association studies have identified several genomic loci to be significantly associated with critical COVID-1910–15,10,11,13,16. However, it remains unknown whether common germline variants causally impact the course of COVID-19.

APOE is a secreted protein with canonical roles in lipid metabolism. Importantly, APOE has also been shown to modulate immunity in different contexts, including infection and anti-tumor immunity2,17,18. Two single-nucleotide polymorphisms give rise to three highly prevalent variants of APOE, termed APOE2, APOE3, and APOE4. The proteins encoded by these alleles differ by one or two amino acids. Forty percent of the world’s population carry at least one copy of either the APOE2 or APOE4 allele, and approximately 3 % are homozygous for either APOE2 or APOE41. The APOE4 variant is the strongest monogenetic risk factor for Alzheimer’s disease19,20. APOE variants also modulate several immune-related processes, including atherosclerosis4 and anti-tumor immunity3, prompting us to determine whether APOE causally modulates SARS-CoV-2 infection. Using genetic mouse models of APOE human genetic variation as well as supportive clinical association studies, we found that the APOE2 and APOE4 variants confer adverse outcomes in SARS-CoV-2 infection in vivo including reduced survival.

APOE causally impacts murine COVID-19 outcomes

To assess the impact of APOE germline variation on SARS-CoV-2 infection, we infected 328 APOE knock-in mice across different ages and of both sexes with SARS-CoV-2 MA10, a mouse-adapted strain of SARS-CoV-221 (Extended Data Fig. 1a–c). In APOE knock-in mice, the murine Apoe gene is replaced with one of the three human APOE alleles. Multivariate analysis revealed that this murine model recapitulated the increased risk for poor survival conferred by male sex and those of advanced age as previously shown in humans (Fig. 1a–c). Remarkably, APOE genotype also significantly impacted survival, with both the APOE2 and APOE4 variants conferring poor survival outcomes relative to the APOE3 variant (Fig. 1a, d). In both male and female age-matched mice, APOE4 mice exhibited accelerated weight loss relative to the other variants (Fig. 1e–g, j–l; Extended Data Fig. 1d). While female mice showed higher survival overall, APOE2 and APOE4 conferred worse survival outcomes in both male and female mice (Fig. 1h–i, m–n). The impact was particularly pronounced in male mice, with 100% of APOE4 mice succumbing to COVID-19 in contrast to approximately 30% mortality in APOE3 mice. We observed a significant interaction between APOE genotype and age, with the impact of APOE on survival being more pronounced in younger mice. No significant interaction was observed for APOE genotype and sex (Extended Data Fig. 1e–h). Of note, no spontaneous deaths were detected in similarly aged and non-infected APOE knock-in mice over a comparable period, indicating that the known impact of APOE genotype on longevity does not confound these results (Extended Data Fig. 1i–k). Thus, APOE variants causally and dramatically impact the outcome of murine COVID-19.

Fig. 1 |. APOE variants modulate outcome of murine SARS-CoV-2 MA10 infection.

Fig. 1 |

Open in a new tab

a, Multivariate analysis of the impact of age, sex, and APOE genotype on survival of SARS-CoV-2 MA10-infected APOE knock-in mice (P values according to multivariable Cox proportional hazards model; n = 128, 82, and 118 for APOE2, APOE3, and APOE4, respectively; data pooled from 13 independent experiments). b-d, Survival of combined male and female SARS-CoV-2 MA10-infected APOE-knock-in mice stratified by age (cutoff: 30 weeks) (b), sex (c), and APOE genotype (d) (P values according to log-rank tests). e-n, Age distribution (e, j), weight course (f, k), weight on day 4 post infection (g, l), survival (h, m) and hazard ratios (i, n) of male (e-i) versus female (j-n) APOE-knock-in mice from (a) stratified by APOE genotype (P values according to Kruskal-Wallis-test (e, j), two-sided t-tests (g, l), log-rank test (h, m), and Cox proportional hazard models (i, n); note that (f) and (k) show group averages but some animals died or were censored for tissue harvest during the course of the experiment). Error bars in a, i, and n indicate 95% confidence intervals. Error bars in e, f, j, and k indicate standard error of the mean. Boxplot whiskers in g and l extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively. n, sample size; HR, hazard ratio for death.

APOE2 and APOE4 mice exhibit accelerated COVID-19 progression

To assess viral load, we performed TaqMan quantitative real-time PCR on lungs from APOE knock-in mice on day four post infection. Consistent with faster disease progression, APOE2 and APOE4 mice showed elevated viral loads relative to APOE3 mice (Fig. 2a). These differences were already evident on day two post infection (Extended Data Fig. 2a) and validated by SARS-CoV-2 nucleocapsid immunofluorescence staining (Fig. 2b). Histopathological analyses on day four post infection revealed pronounced lung injury in APOE4 mice with increased bronchiolar necrosis, alveolar damage, and fibrin deposition (Fig. 2c–f, Extended Data Fig. 2b–c). No differences were observed for inflammatory infiltrates in the pulmonary interstitium and vessels in APOE2 or APOE4 mice in comparison with APOE3 mice (Extended Data Fig. 2d–h). These results indicate accelerated COVID-19 progression in APOE2 and APOE4 mice relative to APOE3, with histopathologic features evident by day four primarily in APOE4 mice.

Fig. 2 |. APOE2 and APOE4 mice exhibit accelerated progression of COVID-19 relative to APOE3 mice.

Fig. 2 |

Open in a new tab

a, TaqMan qPCR for SARS-CoV-2 N1 in homogenized lungs from APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10 (data pooled from two experiments; P values according to two-sided Mann-Whitney test; n = 15, 20, 18 for APOE2, APOE3, and APOE4, respectively). b, Immunofluorescence staining for SARS-CoV-2 nucleocapsid in lungs of APOE knock-in mice on day 4 after infection with SARS-CoV-2 MA10 (P values according to two-sided Mann-Whitney tests; n = 10, 15, 10 for APOE2, APOE3, and APOE4, respectively). Images show representative sections; scale bar, 100 μm. c-f, Histopathologic scoring of bronchiolar necrosis (c), alveolar damage (d), and fibrin deposition (e) in lungs from APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10 (data pooled from two independent experiments,; n = 18, 22, 15 for APOE2, APOE3, and APOE4, respectively; P values according to two-sided Mann Whitney tests). f, Representative images for (c-e). Black arrowheads, bronchiolar epithelial necrosis; inset asterisks, fibrin; white arrowheads, interstitial and perivascular inflammation; inset arrows, endothelialitis. Scale bars, 1000μm (top row) and 400μm (middle and bottom rows). Boxplot whiskers in (a) and (b) extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively.

APOE genotype modulates COVID-19 outcome by impacting antiviral immunity and viral infection

We next performed transcriptional profiling of homogenized lungs of non-infected APOE knock-in mice and on days two and four post infection with SARS-CoV-2 MA10 (Fig. 3a). To identify clusters of highly correlated genes and relate their expression to genotype and timepoint relative to infection, we employed weighted gene co-expression network analysis (WGCNA; see methods)22. WGCNA revealed five modules of co-expressed genes that were significantly correlated with APOE genotype, and ten modules correlated with timepoint relative to infection (Fig. 3b). Assessment of the trajectories of the eigengene of these modules (a metric summarizing the weighted overall expression of a module) revealed modules that became specifically up- or downregulated in APOE3 relative to APOE2 and APOE4 mice during disease progression, most pronounced for modules greenyellow/midnightblue and black, respectively (Fig. 3c, Extended Data Fig. 3a), and validated in an independent cohort of mice (Extended Data Fig. 3b–c). Correlations of the greenyellow, midnightblue and pink modules with APOE genotype were also significant in an independent third cohort of young female mice (Extended Data Fig. 3d–e). Pathway analysis of the black module that exhibited higher expression in APOE2 and APOE4 relative to APOE3 mice on day four revealed enrichment of genes implicated in blood coagulation and hemostasis, abnormalities of which are frequent in severe COVID-1923 (Extended Data Fig. 3f–g). Strikingly, analysis of the modules that exhibited downregulation in APOE2 and APOE4 mice relative to APOE3 on day four (greenyellow, midnightblue, yellow) showed enrichment of genes implicated in T and B cell activation as well as positive immune response regulation (Extended Data Fig. 3h–o). Immunofluorescence staining indicated overall similar levels of CD45+ leukocytes in APOE2 and APOE4 relative to APOE3 mice (Extended Data Fig. 3p). These data are consistent with dampened adaptive immunity during early response to COVID-19 in APOE2 and APOE4 relative to APOE3 mice.

Fig. 3 |. APOE genotype impacts COVID-19 progression through immune modulation and altered viral infection.

Fig. 3 |

Open in a new tab

a, Schematic for transcriptional profiling of lungs from SARS-CoV-2 MA10-infected mice. b, Correlation of module eigengenes with time after infection and APOE genotype ordered by its impact on COVID-19 survival (E3 > E2 > E4); stars indicate significant correlations (Pearson correlation tests). c, Module eigengene trajectories for modules significantly correlating with APOE genotype (n = 4 (E2/d0; E3/d2), 3 (E4/d0), 6 (E2/d2), 5 (E4/d2; E3/d4; E4/d4), and 7 (E2/d4)). d, Flow cytometry for indicated cells in lungs of APOE knock-in mice on day 4 post infection (n = 21, 15, 20 for APOE2, APOE3, and APOE4, respectively; data pooled from two independent experiments; P values according to one-tailed t-tests). e-f, Density plots of 41,500 RNA-sequenced lung cells from APOE knock-in mice stratified by infection status (e) or APOE genotype in infected mice (f). g, Gene set enrichment analysis for grouped clusters from (f); grey boxes indicate no significant enrichment. h, Representative flow cytometry plots of tetramer+ CD8+ T cells on days 4 and 11 post infection (independent experiments). i, Proportion of tetramer+ CD8+ cells on day 11 post infection (n=14, 22, 17 for APOE2, APOE3, and APOE4, respectively; data pooled from two independent experiments, P values according to two-tailed t-tests; note that some mice died during the course of infection). j, Representative samples for (i). k, Fraction of infected cells after incubation with SARS-CoV-2 in the presence of the indicated proteins (n = 10 per group; representative of three independent experiments; P values according to two-tailed t tests). Boxplot whiskers in d, i and k extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively.

Consistent with these findings, flow cytometry on dissociated lungs on day four post infection confirmed an expansion of myeloid cells and relative depletion of lymphoid cells in the lungs of both APOE2 and APOE4 relative to APOE3 mice (Fig. 3d, Extended Data Fig. 4a–b). In humans with severe COVID-19, depletion of lymphoid subsets has also been observed in the peripheral blood24,25. To assess whether these changes were recapitulated by our animal model, we performed flow cytometry on peripheral blood of APOE knock-in mice. While total leukocyte numbers were not significantly different between APOE genotypes, both APOE2 and APOE4 mice showed expansion of myeloid cells mainly driven by Ly6G+ neutrophils with concomitant contraction of all major lymphoid populations (Extended Data Fig. 4c–g). These data are consistent with the reported elevation of myeloid/lymphoid ratios in patients with adverse COVID-19 outcomes24,25 and suggest that adaptive immune responses are blunted in APOE2 and APOE4 mice during early COVID-19 progression.

To further profile the immunological response in APOE knock-in mice during COVID-19, we performed single cell RNA-sequencing (scRNAseq) on a total of 41,500 cells (post-filtering) from 29 mice across all three genotypes with and without COVID-19 (Extended Data Fig. 5a–d). Infected mice showed a marked expansion of myeloid cells, which, consistent with our flow cytometry data, was more prominent in APOE2 and APOE4 relative to APOE3 mice (Fig. 3e–f, Extended Data Fig. 6). To assess changes in the functional status of cell clusters, we performed gene set enrichment analysis. Strikingly, APOE2 mice showed more pronounced enrichment of various immune-related pathways relative to APOE3 in comparison to APOE4 relative to APOE3 mice (Fig. 3g). In humans, hyperactivation of proinflammatory signaling has been implicated in adverse outcomes24,26. We therefore hypothesized that despite a similar change in immune subset abundances during early infection, antiviral immune responses might diverge between APOE2 and APOE4 mice over the course of infection. To test this, we assessed the generation of virus spike-specific CD8+ T cells during infection (Fig. 3h). The fraction of virus spike-specific CD8+ T cells as assessed by tetramer staining was significantly larger in APOE4 relative to APOE3 and APOE2 mice, consistent with APOE4 mice eventually mounting more effective adaptive antiviral immunity than APOE2 mice (Fig. 3i–j). These data indicate that while both APOE2 and APOE4 mice initially exhibited blunted adaptive immune responses, APOE4 mice generated more robust antiviral T cell responses in later stages of infection, which emerged after pathological tissue damage had occurred.

We next assessed whether APOE directly impacts viral infection, potentially explaining the emergence of differences in viral titer and immune responses early upon infection. Remarkably, recombinant APOE3, but not recombinant APOE2 or APOE4, significantly suppressed infection of Huh-7.5 cells in vitro (Fig. 3k). In sum, these data indicate that adverse outcomes in APOE2 and APOE4 mice are driven by both enhanced viral infection and dampened adaptive antiviral immunity.

APOE genotype associates with survival in COVID-19 patients

To assess the impact of APOE genotype on COVID-19 outcome in humans, we analyzed participants of the UK Biobank27. Overall distribution of APOE genotype in 402,763 UK Biobank participants was comparable to similarly aged individuals in the ARIC study28, with approximately 40% carrying at least one copy of the APOE2 or APOE4 allele (Extended Data Fig. 7a, b). Consistent with previous reports carried out at earlier times during the pandemic29,30, we observed a moderate enrichment of APOE4 homozygosity in participants with positive versus negative test results and in participants with positive test results versus the remaining participants (Extended Data Fig. 7c, d). There was no significant difference in APOE genotype distribution between patients with a positive test regarding the test origin (inpatient versus outpatient; Extended Data Fig. 7e).

We next performed survival analysis of patients with confirmed SARS-CoV-2 infection. Consistent with known epidemiological observations, multivariate analysis confirmed male sex and advanced age to confer adverse survival outcomes (Fig. 4a–c). Strikingly, homozygous APOE4 patients also exhibited poor survival with a more than two-fold increased hazard ratio for death relative to APOE3 homozygous patients (Fig. 4a, d). Homozygous APOE2 patients also experienced an increased hazard ratio for death that did not reach statistical significance (Fig. 4a, d). The association between APOE genotype and survival remained significant upon adjustment for the first ten principal components of genetic variation, indicating population structure to be unlikely to account for this association (Extended Data Fig. 8a–c). Consistently, the association of APOE with COVID-19 was maintained upon restriction of the analysis to individuals of European ancestry (Extended Data Fig. 8d–g). No significant association of APOE genotype with survival was detected over a similar period prior to the start of the COVID-19 pandemic, indicating that the known association of APOE genotype with longevity also does not confound these results (Extended Data Fig. 8h). Overall, these results are consistent with our animal studies that demonstrate a causal role of APOE genotype in modulating murine COVID-19 outcome. While the present work was in revision, an independent study validated the epidemiologic association of APOE4 with adverse outcomes in COVID-19 in the large FINNGEN cohort31.

Fig. 4 |. APOE germline variants are associated with survival in humans with SARS-CoV-2 infection.

Fig. 4 |

Open in a new tab

a, Multivariate analysis of the impact of age, sex, and APOE genotype on survival of patients with SARS-CoV-2 infection in the UK Biobank (P values according to multivariable Cox proportional hazards model, error bars indicate 95% confidence intervals, n = 13,207). b-d, Survival of patients from (a) stratified by age below or above median (b), sex (c), and APOE genotype (d). P values in (b-d) according to log-rank tests. n, sample size; HR, hazard ratio for death.

Discussion

The COVID-19 pandemic has had a devastating impact on public health, but individual outcomes are markedly heterogeneous. Comprehensive efforts have been made to uncover the genetic basis of COVID-19 outcome. These efforts were either carried out using genome-wide or candidate gene approaches and identified genetic variants and regions epidemiologically associated with COVID-19 outcome8,10–12,14,15,32. However, whether common germline variants could causally modulate COVID-19 outcomes in vivo is unknown. In this work, we undertook a reverse genetic approach and specifically focused on APOE variants given their previously established roles in modulating immunity. By employing genetic mouse models of human APOE germline variation, we established a causal link between APOE genotype and COVID-19 outcome in mice, supported by clinical association data in humans. Importantly, our focused genetic and biochemical studies of these APOE variants led us to assess their epidemiological associations with human outcomes. While previous genome-wide association studies for COVID-19 critical illness have not detected associations with variants in APOE that reached genome-wide significant threshold levels, our data on APOE variant association with survival in COVID-19 patients in a candidate analysis are supported by the reverse genetic approach in mice, suggesting a potential causal relationship between APOE4 genotype and COVID-19 outcome in human disease.

We uncovered two mechanisms underlying APOE-genotype dependent differences in murine COVID-19 outcomes: Both APOE2 and APOE4 mice showed impaired immune responses during early infection. Single cell transcriptional profiling indicated hyperactivation of proinflammatory signaling in APOE2 relative to APOE3 and APOE4 mice. In addition, APOE4 mice exhibited increased expansion of virus-specific CD8+ T cells during later stages of infection, indicating that antiviral T cell responses diverge between APOE2 and APOE4 during later infection stages. In addition to these effects on antiviral immunity, we found that recombinant APOE3, but not recombinant APOE2 or APOE4, inhibited viral infection in vitro. These findings are consistent with a prior study demonstrating increased infection of APOE4 relative to APOE3 neurons and astrocytes33. While this past study’s findings could be interpreted as APOE4 enhancing infection of neurons and astrocytes relative to APOE3, we interpret our findings as APOE3 repressing infection in contrast to APOE2 and APOE4. Our data indicate that adverse outcomes in APOE2 and APOE4 mice may be mediated by both enhanced viral infection and maladaptive immunity during early infection, with APOE4 mice ultimately generating more robust antiviral T cell immunity than APOE2 mice.

It will be important to further dissect the mechanistic basis of how these variants exert detrimental effects on COVID-19 outcome at a molecular level in future studies. APOE has been shown to directly modulate both innate and adaptive immune responses2,34,35, providing potential clues towards its molecular mechanism of action in immune modulation. In addition, a genetic screen identified cholesterol metabolism to impact SARS-CoV-2 infection36–38, and SARS-CoV-2 may bind directly to APOE39, providing starting points for additional mechanistic studies focused on how APOE impacts viral infection. It is important to note that the effects of APOE variants seem to be disease-context specific, with APOE2 and APOE4 conferring beneficial and/or detrimental outcomes depending on phenotype3,19,20,40–42. Moreover, the dual impact of APOE genetic variation on COVID-19 and Alzheimer’s outcomes has implications for understanding the neurocognitive changes imparted by both disorders.

Our findings have several potential clinical implications. Firstly, prospective clinical studies are warranted to determine if APOE genotyping could be used for risk stratification in SARS-CoV-2 and perhaps other virus infections. Such genotyping may allow future patients to benefit from more aggressive preventative and therapeutic approaches, including early booster vaccinations, anti-viral drugs and monoclonal antibody therapies. The impact of vaccination or prior infection history on APOE genotype dependence of COVID-19 outcomes will need to be determined. Additionally, it will be important to assess vaccination efficacy in individuals of distinct APOE genotypes. More generally, our work confirms that common genetic variation can give rise to heterogeneous outcomes of COVID-19.

Methods

Cell lines

VeroE6 cells (Chlorocebus sabaeus; sex: female, kidney epithelial) obtained from the ATCC (CRL-1586) and Ralph Baric (University of North Carolina at Chapel Hill), Caco-2 cells (H. sapiens, sex: male, colon epithelial) obtained from the ATCC (HTB-37), and Huh-7.5 hepatoma cells (Homo sapiens; sex: male, liver epithelial)43 were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 1 % nonessential amino acids (NEAA) and 10 % fetal bovine serum (FBS) at 37ºC and 5 % CO2. All cell lines were tested negative for contamination with mycoplasma.

Virus propagation and titration

The SARS-CoV-2 MA10 was generously provided by Ralph Baric (University of North Carolina at Chapel Hill). A P1 stock was amplified in VeroE6 cells obtained from the ATCC that were engineered to stably express TMPRSS2 (VeroE6TMPRSS2). To generate a P2 working stock, VeroE6TMPRSS2 cells were infected at a multiplicity of infection (MOI) of 0.1 plaque forming units (PFU)/cell and incubated at 37°C for four days. The virus-containing supernatant was subsequently harvested, clarified by centrifugation (3,000 g for 10 min), and filtered using a disposable vacuum filter system with a 0.22 μm membrane. Virus stock titers were measured by a standard plaque assay (PA) on Huh-7.5 cells that stably express ACE2 and TMPRSS2 (Huh-7.5ACE2/TMPRSS2) and on VeroE6 cells obtained from Ralph Baric (referred to as VeroE6UNC). Briefly, 500 μL of serial 10-fold virus dilutions in Opti-MEM were used to infect 4×105 cells seeded the day prior into wells of a 6-well plate. After 90 min adsorption, the virus inoculum was removed, and cells were overlayed with DMEM containing 10 % FBS with 1.2 % microcrystalline cellulose (Avicel). Cells were incubated for four days at 33°C, followed by fixation with 7 % formaldehyde and crystal violet staining for plaque enumeration. SARS-CoV-2, strain USA-WA1/2020, was obtained from BEI Resources and amplified in Caco-2 cells. Caco-2 cells were infected at a MOI = 0.05 PFU/cell and incubated for 6 days at 37°C. The virus-containing supernatant was subsequently harvested, clarified by centrifugation (3,000 g × 10 min) and stored at −80ºC. Viral titers were measured on Huh-7.5 cells by standard plaque assay as described above. All SARS-CoV-2 and SARS-CoV-2 MA10 experiments were performed in a biosafety level 3 (BSL-3) laboratory.

To confirm virus identity and evaluate for unwanted mutations that were acquired during the amplification process, RNA from virus stocks was purified using TRIzol Reagent (ThermoFisher Scientific, #15596026). In brief, 200 μL of each virus stock was added to 800 μL TRIzol Reagent, followed by 200 μL chloroform, which was then centrifuged at 12,000 g for 5 min. The upper aqueous phase was transferred to a new tube, mixed with an equal volume of isopropanol, and then added to RNeasy Mini Kit columns (Qiagen, #74014) to be further purified following the manufacturer’s instructions. Viral stocks were subsequently confirmed via next generation sequencing using libraries for Illumina MiSeq.

Animal studies

All animal experiments were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee at The Rockefeller University, including the use of SARS-CoV-2 MA10 virus under BSL-3 conditions. Human APOE2 (strain #1547), APOE3 (#1548), and APOE4 (#1549) targeted replacement (knock-in) mice on C57Bl/6 background were obtained from Taconic Biosciences.

SARS-CoV-2 MA10 in vivo infections

All infection experiments were performed in a dedicated BSL-3 facility at The Rockefeller University at negative pressure. Staff performing experiments were protected by wearing Tyvek suits connected to powered air purifying respirators. Mice were intranasally infected with 14,700 PFU (based on titration in VeroE6UNC cells) of SARS-CoV-2 MA10 in a volume of 30 μL under ketamine/xylazine anesthesia. APOE knock-in mice were infected between 7 and 45 weeks of age as indicated in the figures. Experimental cohorts were age matched. Mice were monitored daily for weight loss and general condition. Mice were recorded as dead when found dead in the cage or when meeting criteria for euthanasia as defined in the animal protocol, including when falling below 70 % initial body weight. All infected mice were included in survival and weight analyses. Some mice were selected before infection for tissue harvest on the days as indicated in the figure legends and censored for survival and weight analyses on the respective days. Mice were gently twirled before weighing to prevent measurement inaccuracies due to mouse movements. In addition, weight measurements were performed with investigators blinded for the genotype in two independent experiments which recapitulated the results of the overall large cohort.

RNA isolation from homogenized lungs

The right lung lobe was resected and homogenized in TRIzol (ThermoFisher Scientific, #15596026) in a gentleMACS dissociator (Miltenyi) according to the manufacturer’s instructions (program RNA_01). Debris was removed by centrifugation (2000g for one minute) and RNA was isolated using the Direct-zol RNA purification kit (Zymo Research, #R2050) including DNAse digestion according to the manufacturer’s instructions.

TaqMan quantitative real-time PCR

For quantification of SARS-CoV-2 MA10 titers from homogenized lungs, RNA was isolated as described above, reverse-transcribed and quantified using the TaqMan Fast Virus One Step Master Mix (ThermoFisher Scientific, #4444436) on a QuantStudio 5 system running QuantStudio Design & Analysis v1.4.3 (Thermo Fisher Scientific) according to the manufacturer’s instructions. Primers for viral nucleocapsid were as recommended in the US Centers for Disease Control and Prevention diagnostic N1 assay (IDT, #10006713), and 18S rRNA was used as housekeeping control (Thermo Scientific, #4319413E).

Bulk RNA-sequencing

For preparation of RNA-seq libraries, 250–500 ng of RNA isolated from homogenized lungs as outlined above was used as input for the Quantseq 3’ FWD library preparation kit (Lexogen, #015). For cohorts 1–2, age-matched 17–23 weeks old male mice were used. For cohort 3, seven weeks old female mice were used. Libraries were sequenced on an Illumina NovaSeq sequencer (single end, 100 bp read length), and polyA and adapter sequences were trimmed using the BBDuk utility (v38.9; options k=13, ktrim=r, forcetrimleft=11, useshortkmers=t, mink=5, qtrim=t, trimq=10, minlength=20). As genome reference, mouse (assembly GRCm38) and SARS-CoV-2 MA10 (Genbank accession number MT95260221) genomes were concatenated and trimmed reads were aligned using STAR aligner (v2.7.8a) with default settings, apart from “--outFilterMismatchNoverLmax 0.1” as recommended by Lexogen (personal communication). STAR was also used for counting reads mapping to genes. Further analysis was performed using R (v4.1.0). Two samples (out of 88 samples total) were removed from analysis because of their identification as outliers based on PCA and/or SARS-CoV-2 MA10 transcript abundance.

Weighted gene co-expression network analysis and pathway analysis

Weighted gene co-expression network analysis (WGCNA)44 was employed using the WGCNA R package (v1.70) to identify modules of co-expressed genes. WGCNA identifies clusters of genes whose expression correlates with each other and relates these clusters to traits, such as APOE genotype and timepoint relative to infection in our study. The module eigengene represents the first principal component of the expression matrix and can be used to summarize the (weighted average) expression of a module. Gene expression data were subjected to library size normalization and variant stabilizing transformation using DESeq2 (v1.32.0) and the top 30% genes in terms of variance of expression were used as input for WGCNA. To compute the adjacency matrix for a signed co-expression network, a soft threshold power of 10 was used. To calculate correlations between traits and module eigengenes, APOE genotype was assigned values based on its impact on survival as shown in Fig. 1a, and the timepoint trait was assigned values in terms of days relative to infection. Hub genes were identified as the genes exhibiting the highest connectivity within a given model.

To assess enrichment of gene sets listed in the Gene Ontology “biological processes”, the clusterProfiler package for R (v4.0.0) was used to perform over-representation analysis based on a hypergeometric model.

Single cell RNA-sequencing

For single cell RNA-sequencing (scRNA-seq) of lung-resident cells, mice were anesthetized, and the pulmonary circulation was flushed with 5–10 mL ice-cold PBS. The right lung lobe was dissociated using the lung dissociation kit (130–095-927, Miltenyi Biotec) with a gentleMACS dissociator according to the manufacturer’s instructions (program 37C_m_LDK_1). Cells were strained using a 70 μm filter, washed, pelleted, and red blood cells were lysed by incubation in ACK buffer (A10492, Gibco) for two minutes before neutralization with PBS. Cells were then strained again with a 40 μm filter and processed using the cell fixation (SB1001) and single cell whole transcriptome (SB2001) kits from Parse Biosciences according to the manufacturer’s instructions. This scRNA-seq approach is based on combinatorial barcoding, which enabled us to multiplex lungs from a total of 29 mice representing each of the three APOE genotypes and conditions in the absence and presence of SARS-CoV-2 MA10 infection (Extended Data Fig. 5a). One of the eight resulting sublibraries was sequenced on an Illumina Nextseq 500 sequencer and the other seven sublibraries were pooled and sequenced on an Illumina Novaseq sequencer (S2 flowcell) to an average depth of 65,256 reads/cell.

For data processing, the ParseBioscience processing pipeline (v0.9.6p) was employed at default settings to align sequencing reads to the GRCm38 mouse genome and to demultiplex samples. In brief, each of the eight sublibraries was first processed individually with the `split-pipe –mode all` command and the output of the eight sublibraries was combined with `split-pipe –mode combinè. Downstream processing was performed using the R package Seurat (v4.0.2) at default settings unless otherwise noted. Cells with fewer than 150 or more than 7,500 detected unique genes, more than 40,000 unique molecular identifiers, or more than 15% mitochondrial reads were excluded from analysis. The resulting gene-cell matrix was normalized and scaled using Seurat’s ǸormalizeDatà and `ScaleDatà functions and principal component analysis was performed with Seurat’s `RunPCÀ function; cells were clustered using the `FindNeighbors` (30 dimensions of reduction) and `FindClusters` (resolution = 1.4) functions; for visualizing clusters, `RunUMAP` (30 dimensions) was run. Wilcoxon rank-sum tests were performed to determine differentially expressed genes between clusters using the `FindAllMarkers` function (minimal fraction of 25% and log-transformed fold change threshold of 0.25). The identity of cell clusters was determined by cross-referencing top differentially expressed transcripts with previous studies reporting on single cell transcriptomes of the lung45–47. Ambiguous cells with expression of distinct lineage markers were deemed to be likely multiplets and were excluded. Three clusters expressing T cell markers were characterized further using Seurat’s `subset` function and reanalyzed similarly to the main dataset, including running the `RunPCÀ, `FindNeighbors` (20 dimensions), `FindClusters` (resolution = 0.5) and `RunUMAP` functions. Ambiguous cells from the subset were removed, and annotations for the remaining clusters were added to the main dataset. For summary analyses, clusters were grouped as follows: Alveolar mø A and B and proliferating alveolar mø as alveolar macrophages; monocytes A and B as monocytes; T cells naïve, T cells and T cells proliferating as T cells; myofibroblasts, lipofibroblasts, and Col14a1+ fibroblasts as fibroblasts; capillary ECs, vascular ECs A and B, other ECs, and Vcam1+ ECs A and B as endothelial cells; AT1, AT2, ciliated cells, airway epithelial A and B, and mesothelial cells as epithelial cells. In total, filtering low quality and ambiguous cells resulted in 41,500 cells for analysis (of 50,104 cells before filtering).

For gene set enrichment analysis (GSEA) of the samples from infected mice, differentially expressed genes between either APOE2 and APOE3 or APOE4 and APOE3 were calculated according to Wilcoxon rank-sum tests using Seurat’s `FindMarkers` function. Genes were ranked using the metric [-log10(P value)]/sign of log-fold change]. The ranked gene lists were used as input for the `GSEÀ function of the clusterProfiler R package (v4.0.0) to assess enrichment of selected immune-related pathways of the Hallmark gene set of the MSigDB database (http://www.gsea-msigdb.org).

Histological analysis and immunofluorescence staining

The left lung lobe was resected and fixed by submersion in 4% paraformaldehyde for 24 hours at room temperature. Fixed lungs were embedded in paraffin and sectioned in 5 μm thick slices. Sections were dewaxed and rehydrated by incubation with xylene and descending ethanol concentrations and then either stained with hematoxylin/eosin for histological analysis or processed for immunofluorescence staining.

For immunofluorescence staining, samples were permeabilized with 0.1% Triton-X for 15 min. Antigen retrieval was performed by microwaving samples in Tris-EDTA buffer (Abcam, #ab93684) for 20 minutes. Samples were blocked by incubation with 5 % goat serum in PBST (PBS with 0.1 % Tween-20) for one hour. Subsequently, sections were stained with anti-CD45 (polyclonal, Abcam, #ab10558; 1:750) or anti-SARS Nucleocapsid (polyclonal, Novus Biologicals, #56576, 1:1000) at 4°C overnight. All antibodies were diluted in PBST with 5% goat serum. Slides were washed three times with PBS and stained with AF555-conjugated anti-rabbit antibody (1:200 in PBST, ThermoFisher Scientific) for 45 minutes. Slides were washed with PBS and nuclei were counterstained with DAPI (1 μg/ml, Roche) before mounting with Prolong Gold (ThermoFisher Scientific). Images of lung sections were acquired using a Nikon A1R confocal microscope at 20× magnification using Nikon NIS elements software (v5.20.02). Images were quantified using CellProfiler (v4.2.1). Three to four randomly sampled fields of view per lung were analyzed and averaged.

For histological analysis, H&E-stained lung sections were evaluated and scored by a board-certified veterinary pathologist (S.E.C) using a semiquantitative histopathology scoring system used in mouse models of SARS-CoV-248,49. Briefly, five random fields of the lung lobe at 200× total magnification were chosen and scored in a blinded manner for histopathological changes. Ordinal scores for lesion parameters were assigned using the following tiers: 0 - within expected limits; 1 - uncommon, < 5%; 2 - detectable in 5–33%; 3 - detectable in 34–66% and 4 - detectable in > 66% of lung fields. Tissues were graded for the presence of edema, hemorrhage, fibrin, and/or necrotic debris in alveoli, bronchiolar epithelial necrosis, perivascular and interstitial inflammation, and mononuclear cell infiltrates. Endothelial inflammation (endothelialitis) was evaluated by the extent of the lesion using the following ordinal scoring: 0 – absent; 1 - minor, solitary to loose adhesion or aggregation of leukocytes to the vascular endothelium with or without infiltration of leukocytes in the vascular wall / up to five blood vessels affected; 2 – moderate, small to medium adhesion/aggregates and infiltration / six to ten blood vessels affected; and 3 – severe, robust leukocytic aggregates and infiltrates around pulmonary vessels / more than ten blood vessels affected. For scoring neutrophil cell infiltration (200–600× objective magnification): 0 - within normal limits; 1 - scattered neutrophils sequestered in septa and/or infiltrating blood vessels; 2 - #1 plus solitary neutrophils extravasated in alveolar spaces; 3 - #2 plus small aggregates in blood vessels, alveolar spaces, and perivascular and peribronchiolar interstitium. An Olympus BX45 light microscope was used to capture images with a DP26 camera using cellSens Dimension software (v1.16). Lungs in SARS-CoV-2 MA10-infected mice exhibited multifocal areas of airway epithelial damage in bronchioles. Bronchioles had focal to multifocal changes characterized by segmental attenuation of bronchiolar epithelium with an accumulation of necrotic cellular debris, fibrin, and sloughed epithelial cells, and occasional foamy macrophages in the airway lumens (Fig. 2f). Peribronchiolar interstitium was multifocally infiltrated by increased numbers of neutrophils and lymphocytes. The adjacent alveolar sacs and septae exhibited multifocal to coalescing areas of alveolar damage. Histological changes included hypercellular thickening of the alveolar septae caused by infiltrating leukocytes and congestion of alveolar capillaries, pneumocyte degeneration and necrosis, edema, fibrin strands and increased numbers of macrophages and scattered neutrophils and lymphocytes in alveolar spaces. Often, the vascular endothelium of pulmonary vessels was reactive with adherence and aggregation of leukocytes to the endothelium and transmigrating within vessel walls, indicative of endothelialitis.

Flow cytometry

All steps were performed on ice and under protection from light unless stated otherwise. Peripheral blood was obtained by submandibular bleedings and red blood cells were lysed by incubation in ACK buffer (A10492, Gibco) for three minutes at room temperature before addition of PBS for neutralization. For flow cytometry of dissociated lungs, mice were anesthetized, and the pulmonary circulation was flushed with 5–10 mL ice-cold PBS. The right lung lobe was then dissociated using the lung dissociation kit (130–095-927, Miltenyi Biotec) with a gentleMACS dissociator according to manufacturer’s instructions (program 37C_m_LDK_1). Cells were strained through a 70 μm filter, washed, pelleted and red blood cells were lysed by incubation in ACK buffer as indicated above before addition of PBS for neutralization. Cells were pelleted by centrifugation at 200×g for 5 minutes and resuspended in staining buffer (25 mM HEPES, 2 % FBS, 10 mM EDTA (351–027, Quality Biological), and 0.1 % sodium azide (7144.8–16, Ricca) in PBS). To block Fc receptors, cells were incubated with 2.5 μg/mL anti-CD16/32 antibody in staining buffer (clone 93; 101320, BioLegend) before incubation with antibodies diluted in staining buffer for 20 minutes. After washing with PBS, cells were incubated with Zombie NIR Fixable Live/Dead stain (423105, BioLegend; 1:10,000 in PBS) for 15 minutes at room temperature, washed with staining buffer, and fixed in 4% PFA. CountBright counting beads (C36950, Thermo Fisher) were added to the peripheral blood samples before analysis on an LSR Fortessa (BD Biosciences). For compensation, single color controls with UltraComp beads (01–2222-42, ThermoFisher) for antibodies and amine-reactive beads (A10628, ThermoFisher) for Zombie live-dead stain were used. The following anti-mouse fluorophore-conjugated antibodies were used: CD45-BV785 (clone: 30-F11, cat#: 103149, supplier: BioLegend, dilution: 1:3,000), CD11b-FITC (M1/70, 101206, BioLegend, 1:1,000), Ly6G-PerCP/Cy5.5 (1A8, 127616, Biolegend, 1:1,000), Ly6C-BV711 (HK1.4, 128037, BioLegend, 1:10,000), I-A/I-E-PE (M5/114.15.2, 107607, BioLegend, 1:10,000), CD19-PB (6D5, 115526, Biolegend, 1:500), CD19-BV421 (6D5, 115549, BioLegend, 1:500), NK1.1-APC (PK136, 17–5941-82, eBiosciences, 1:500), CD4-BV605 (GK1.5, 100451, BioLegend, 1:300), CD8α-AF700 (53–6.7, 100730, BioLegend, 1:1,000). For staining of SARS-CoV-2 spike-specific CD8+ T cells, BV421-labeled SARS-CoV-2 S 539–546 tetramer was used (NIH Tetramer Core Facility, 1:200).

SARS-CoV-2 in vitro infections

The day prior to infection, Huh-7.5 cells were seeded into 96-well plates at a density of 7.5×103 cells/well. The next day, recombinant APOE2, −3, −4 (21–9195, 21–9189, 21–9190, Tonbo Biosciences) or BSA (A9576, Sigma) as control were added to the wells at a concentration of 10 μg/mL, followed by infection with SARS-CoV-2 (WA1/2020) at an MOI of 0.01 PFU/cell. Cells were then incubated at 33°C for 48 h. Next, they were fixed by adding an equal volume of 7% formaldehyde to the wells and subsequently permeabilized with 0.1% Triton X-100 for 10 min. After extensive washing, SARS-CoV-2 infected cells were incubated for one hour at room temperature with blocking solution of 5% goat serum in PBS (005–000–121, Jackson ImmunoResearch). A rabbit polyclonal anti-SARS-CoV-2 nucleocapsid antibody (GTX135357, GeneTex) was added to the cells at 1:1,000 dilution in blocking solution and incubated at 4°C overnight. A goat anti-rabbit AlexaFluor 594 (A-11012, Life Technologies) was used as a secondary antibody at a 1:2,000 dilution. Nuclei were stained with Hoechst 33342 (62249, Thermo Scientific) at a 1 μg/mL dilution. Images were acquired with a fluorescence microscope and analyzed using ImageXpress Micro XLS (Molecular Devices, Sunnyvale, CA). All SARS-CoV-2 experiments were performed in a biosafety level 3 laboratory.

Analysis of the UK Biobank

APOE genotyping results as determined by the rs7412 and rs429358 single nucleotide polymorphisms were downloaded from the UK Biobank50. Clinical data, including SARS-CoV-2 test results and survival data were downloaded from the UK Biobank data portal on June 22, 2021. For survival analyses, in patients with multiple tests the earliest positive test result was used as day zero of infection and COVID-19-associated death was recorded if the death cause was ICD10-coded as U07.1 or U07.2. Out of 502,619 patients, APOE genotype could be determined in 413,219 patients. 77,221 participants had SARS-CoV-2 test results available, and 16,562 patients of these were tested positive at least once (Extended Data Fig. 7a). APOE2/APOE4 heterozygous patients (n = 10,456) were excluded from analyses except for summary statistics shown in Extended Data Fig. 7a. For visualization purposes, survival data were truncated at 40 days. To account for genetic ancestry, the first ten genetic principal components as provided by the UK Biobank were included in a multivariate analysis. To restrict the analyses to individuals of European genetic ancestry, field 22006 provided by UK Biobank was used.

Statistical analysis

R v 4.1.0 was used for data visualization and analyses. Statistical tests and sample sizes are listed in the respective figure legends. Unless otherwise noted, data are expressed as mean ± standard error of the mean. For boxplots, hinges represent the first and third quartiles, whiskers extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and points represent individual mice. Survival analyses were performed using the R packages ‘survival’ and ‘survminer’; summary tables were compiled using the ‘gtsummary’ package. Multivariate analyses were performed according to a Cox proportional hazards model using the ‘survival’ package and visualized with the ‘forestmodel’ package. A significant difference was concluded at P < 0.05 in all figures.

Data availability

Bulk RNA-seq and scRNA-seq data have been deposited at the Gene Expression Omnibus (GEO) under accession numbers GSE184289 and GSE199498, respectively. All data from the UK Biobank is publicly available under www.ukbiobank.ac.uk. MSigDB is publicly available under http://www.gsea-msigdb.org. Source data is provided with this paper.

Code availability

Code is publicly available at https://github.com/benostendorf/ostendorf_etal_2022.

Extended Data

Extended Data Fig. 1 |. Expanded characteristics of APOE knock-in mice infected with SARS-CoV-2 MA10.

Extended Data Fig. 1 |

Open in a new tab

a-c, Distribution of age at infection (a), sex (b), and APOE genotype (c) of APOE knock-in mice infected with SARS-CoV-2 MA10 (n = 328; data pooled from 13 independent experiments). d, Individual weight course of male and female APOE knock-in mice infected with SARS-CoV-2 MA10 from (a). e-f, Multivariate analysis of the impact of age, sex, APOE genotype, and the interaction of age/APOE and sex/APOE on survival of SARS-CoV-2 MA10-infected APOE knock-in mice from (a) (P values according to multivariable Cox proportional hazards model; error bars in (f) denote 95% confidence intervals; n = 128, 82, and 118 for APOE2, APOE3, and APOE4, respectively). g-h, Survival of young (< 30 weeks old) (g) and old (> 30 weeks old) (h) SARS-CoV-2 MA10-infected APOE-knock-in mice from (a) stratified by APOE genotype; P values according to log-rank tests. i-k, Age (i), sex distribution (j), and survival of non infected APOE knock-in mice over a two-week period (k) (n = 67, 55, 67 for APOE2, APOE3, and APOE4, respectively; P values according to Kruskal-Wallis (i) and logrank (k) tests). Boxplot whiskers in (a) extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively.

Extended Data Fig. 2 |. Viral load early post infection and extended histopathologic analysis of lungs from SARS-CoV-2 MA10-infected APOE knock-in mice.

Extended Data Fig. 2 |

Open in a new tab

a, TaqMan qPCR for SARS-CoV-2 N1 in homogenized lungs from APOE knock-in mice on day 2 post infection with SARS-CoV-2 MA10 (data pooled from two experiments; P values according to one-tailed Mann-Whitney test; n = 12, 11, 11 for APOE2, APOE3, and APOE4, respectively; boxplot whiskers extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively). b-h, Histopathologic scoring of lungs from APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10 for hemorrhage (b), edema (c), mononuclear cell infiltrates (d), neutrophilic cell infiltrates (e), interstitial infiltrates (f), perivascular infiltrates (g), and endothelialitis/vascular changes (h); P values according to two-sided Mann Whitney-tests, n = 18, 22, 15 for APOE2, APOE3, and APOE4, respectively.

Extended Data Fig. 3 |. Extended analysis of transcriptional profiles and immune cell infiltration in APOE knock-in mice during COVID-19.

Extended Data Fig. 3 |

Open in a new tab

a, Module eigengenes averaged per condition. b, Independent validation of correlation of specific gene modules with APOE genotype in lungs of male APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10. Red indicates positive correlation of module eigengenes with APOE genotype ordered by its impact on COVID-19 survival (E3 > E2 > E4); stars indicate significant correlations (one-sided Pearson correlation tests). c, Averaged module eigengenes in the validation experiment. d, Correlation of gene modules with APOE genotype in lungs of 7-weeks old female APOE knock-in mice on day 4 post infection. Red indicates positive correlation of module eigengenes with APOE genotype ordered by its impact on COVID-19 survival (E3 > E2 > E4); stars indicate significant correlations (one-sided Pearson correlation tests). e, Averaged module eigengenes of modules significantly associated with APOE genotype in mice from (d). f-g, Network plots of the top ten hubgenes (genes with highest intramodular connectivity) (f) and the top five GO pathways enriched in the 348 genes of the black module (g) (P values according to hypergeometric tests adjusted for FDR). h, Expression of genes constituting the midnightblue and greenyellow modules. Hubgenes are annotated by name. i-o, Network plots of the top ten hubgenes (I, k, m, o) and the top five GO pathways enriched in 45, 67, and 58 genes of the greenyellow, midnightblue, and yellow modules, respectively (j, l, n) (P values in j, l, and n according to hypergeometric tests adjusted for FDR). No pathways were enriched in the 24 genes making up the pink module. p, Immunofluorescence staining for CD45+ cells in lungs of APOE knock-in mice on day 4 post infection (n = 10, 15, 10 for APOE2, APOE3, and APOE4, respectively; P values according to two-sided Mann Whitney tests; boxplot whiskers extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively). Images on the right show representative sections; scale bar, 100 μm.

Extended Data Fig. 4 |. Immune cell profiling of lungs and peripheral blood of APOE knock-in mice with COVID-19.

Extended Data Fig. 4 |

Open in a new tab

a-b, Gating strategy to delineate leukocyte subsets (a) and assessment of the proportion of leukocyte subsets (b) in dissociated lungs of APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10 (n = 21, 15, 20 for APOE2, APOE3, and APOE4, respectively; data pooled from two independent experiments; P values according to one-tailed t-tests). c, Gating strategy to delineate leukocyte subsets in peripheral blood of APOE knock-in mice with COVID-19. d-g, Concentration of CD45+ leukocytes (d) and proportion of myeloid (e) and lymphoid (f) subsets in the peripheral blood of APOE knock-in mice on day 4 post infection with SARS-CoV-2 MA10 as assessed by flow cytometry (n = 10, 9, 7 for APOE2, APOE3, and APOE4, respectively; P values according to two-sided t tests). g, Representative flow cytometry plots for (e-f). Boxplot whiskers in b and d-f extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively.

Extended Data Fig. 5 |. Extended single cell RNA-sequencing data.

Extended Data Fig. 5 |

Open in a new tab

a, Number of samples per genotype and condition for single cell RNA-sequencing (scRNA-seq). b, Uniform manifold approximation and projection (UMAP) plot of 41,500 RNA-sequenced cells from APOE knock-in mice with or without COVID-19. c-d, Heatmaps of manually curated marker genes (c) and of top three differentially expressed genes per cluster (d) for cells from (b).

Extended Data Fig. 6 |. Cellular composition in lungs from APOE knock-in mice with or without COVID-19.

Extended Data Fig. 6 |

Open in a new tab

a, Fraction of grouped clusters of RNA-sequenced lung cells in APOE knock-in mice with or without COVID-19 (n = 9 and 20 for non-infected and infected, respectively; P values according to two-tailed t tests). b-c, Fraction of clusters in immune (b) and non-immune (c) cells from infected mice from (a) (n = 6, 6, and 8 for APOE2, APOE3, and APOE4, respectively; P values according to two-tailed t tests). Boxplot whiskers in a-c extend to the smallest and largest value within 1.5 × interquartile ranges of the hinges, and box centre and hinges indicate median and first and third quartiles, respectively.

Extended Data Fig. 7 |. APOE genotyping and SARS-CoV-2 test results in participants of the UK Biobank.

Extended Data Fig. 7 |

Open in a new tab

a, General characteristics of the UK Biobank population. b, Distribution of APOE genotype in participants of the UK Biobank versus the ARIC study (Blair et al., Neurology, 2015) (P = 0.2, Chi-squared test). c-e, Distribution of APOE genotype in UK Biobank patients with positive versus negative SARS-CoV-2 test (c), with positive test versus negative test or untested (d), and with SARS-CoV-2 test performed in- versus outpatient (e). Tables in (c-e) show odds ratios for testing positive versus negative, having a positive versus negative or no test, and having at least one inpatient versus only outpatient tests, respectively; P values are based on binomial general linearized models. Numbers on top of bars indicate sample sizes.

Extended Data Fig. 8 |. The impact of APOE genotype on COVID-19 outcome is not confounded by population structure or its impact on longevity.

Extended Data Fig. 8 |

Open in a new tab

a, Multivariate analysis of the impact of age, sex, the first ten genetic principal components, and APOE genotype on survival of patients with SARS-CoV-2 infection in the UK Biobank (P values according to multivariable Cox proportional hazards model, n = 13,207). b-c, Dot plot of the genetic principal components 1–2 (b) and 3–4 (c) colored by APOE genotype of SARS-CoV-2-positive patients of the UK Biobank. PC, principal component. d-e, Dot plot of the genetic principal components 1–2 (d) and 3–4 (e) colored by APOE genotype of SARS-CoV-2-positive patients with European ancestry in the UK Biobank. f, Multivariate analysis of the impact of age, sex, and APOE genotype on survival of patients with European ancestry and SARS-CoV-2 infection in the UK Biobank (P values according to multivariable Cox proportional hazards model, n = 10,333). g, Survival of patients from (f) stratified by APOE genotype (P value in according to log-rank test). h, Survival of UK biobank participants over a 30 day observation period in January 2019. The start of the observation period was Jan 1, 2019, and data were censored on Jan 31, 2019 (P value according to log-rank test; n = 384,106). Error bars in a and f indicate 95% confidence intervals.

Acknowledgments

We are grateful to members of the Tavazoie laboratory and to Daniel Mucida for comments on previous versions of the manuscript. We thank Rockefeller University resource centers for assistance: Vaughn Francis and other veterinary staff of the Comparative Bioscience Center for animal husbandry and care, Connie Zhao, Christine Lai and staff at the genomics resource center for assistance with RNA-seq, and Gaitree McNab for advice on biosafety measures. We thank the Laboratory of Comparative Pathology for histopathology support (with funding from NIH Core Grant P30 CA008748), the group of Ralph Baric for generating the mouse adapted SARS-CoV-2 MA10 strain and making it available to the scientific community, and Rachel Yamin for experimental advice. We thank the NIH Tetramer Core Facility for providing BV421-labeled SARS-CoV-2 S 539-546 tetramer [H-2K(b)]. This work was supported by the Pershing Square Innovation Fund. B.N.O. was supported by a Max Eder grant of the German Cancer Aid (reference 70114327), the David Rockefeller Graduate program in Bioscience, and is a fellow of the digital clinician scientist program at BIH-Charité. J.B. was supported by a scholarship from the German National Academic Foundation. M.A.P. was supported in part by the National Center for Advancing Translational Sciences, NIH, through Rockefeller University, Grant UL1 TR001866. S.F.T. was supported by the Black Family Foundation, The Rockefeller Center for Human Metastasis Research, and The Breast Cancer Research Foundation. C.M.R. was supported by the G. Harold and Leila Y. Mathers Charitable Foundation, the Bawd Foundation, and Fast Grants, a part of Emergent Ventures at the Mercatus Center, George Mason University. Human data analyses were conducted using the UK Biobank Resource under Application Number 62709.

Footnotes

Competing interests

The authors declare no competing financial interests.

References

  • 1.Farrer LA et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. JAMA 278, 1349–56 (1997). [PubMed] [Google Scholar]
  • 2.Tavazoie MF et al. LXR/ApoE Activation Restricts Innate Immune Suppression in Cancer. Cell 172, 825–840.e18 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Ostendorf BN et al. Common germline variants of the human APOE gene modulate melanoma progression and survival. Nature Medicine 26, 1048–1053 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Mahley RW Apolipoprotein E: from cardiovascular disease to neurodegenerative disorders. J Mol Med (Berl) 94, 739–46 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Shi Y. & Holtzman DM Interplay between innate immunity and Alzheimer disease: APOE and TREM2 in the spotlight. Nat Rev Immunol 18, 759–772 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Williamson EJ et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 584, 430–436 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Zhang Q, Bastard P, Cobat A. & Casanova J-L Human genetic and immunological determinants of critical COVID-19 pneumonia. Nature 1–15 (2022) doi: 10.1038/s41586-022-04447-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Zhang Q. et al. Inborn errors of type I IFN immunity in patients with life-threatening COVID-19. Science 370, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Asano T. et al. X-linked recessive TLR7 deficiency in ~1% of men under 60 years old with life-threatening COVID-19. Sci. Immunol. 6, eabl4348 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.The Severe Covid-19 GWAS Group. Genomewide Association Study of Severe Covid-19 with Respiratory Failure. New England Journal of Medicine 383, 1522–1534 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Pairo-Castineira E. et al. Genetic mechanisms of critical illness in COVID-19. Nature 591, 92–98 (2021). [DOI] [PubMed] [Google Scholar]
  • 12.COVID-19 Host Genetics Initiative et al. Mapping the human genetic architecture of COVID-19. Nature 600, 472–477 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Nakanishi T. et al. Age-dependent impact of the major common genetic risk factor for COVID-19 on severity and mortality. Journal of Clinical Investigation 131, e152386 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kousathanas A. et al. Whole-genome sequencing reveals host factors underlying critical COVID-19. Nature 607, 97–103 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Pathak GA et al. A first update on mapping the human genetic architecture of COVID-19. Nature 608, E1–E10 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.COVID-19 Host Genetics Initiative. Mapping the human genetic architecture of COVID-19. Nature 1–8 (2021) doi: 10.1038/s41586-021-03767-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ludewig B. et al. Hypercholesterolemia Exacerbates Virus-Induced Immunopathologic Liver Disease Via Suppression of Antiviral Cytotoxic T Cell Responses. The Journal of Immunology 166, 3369–3376 (2001). [DOI] [PubMed] [Google Scholar]
  • 18.Toledo A, Monzón JD, Coleman JL, Garcia-Monco JC & Benach JL Hypercholesterolemia and ApoE deficiency result in severe infection with Lyme disease and relapsing-fever Borrelia. Proceedings of the National Academy of Sciences 112, 5491–5496 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Strittmatter WJ et al. Apolipoprotein E: high-avidity binding to beta-amyloid and increased frequency of type 4 allele in late-onset familial Alzheimer disease. Proc Natl Acad Sci U S A 90, 1977–81 (1993). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Corder EH et al. Protective effect of apolipoprotein E type 2 allele for late onset Alzheimer disease. Nat Genet 7, 180–4 (1994). [DOI] [PubMed] [Google Scholar]
  • 21.Leist SR et al. A Mouse-Adapted SARS-CoV-2 Induces Acute Lung Injury and Mortality in Standard Laboratory Mice. Cell 183, 1070–1085.e12 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Langfelder P. & Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9, 559 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Perico L. et al. Immunity, endothelial injury and complement-induced coagulopathy in COVID-19. Nat Rev Nephrol 17, 46–64 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Lucas C. et al. Longitudinal analyses reveal immunological misfiring in severe COVID-19. Nature 584, 463–469 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Mathew D. et al. Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications. Science 369, eabc8511 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lee JS et al. Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19. Sci. Immunol. 5, eabd1554 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Sudlow C. et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLOS Medicine 12, e1001779 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Blair CK et al. APOE genotype and cognitive decline in a middle-aged cohort. Neurology 64, 268–76 (2005). [DOI] [PubMed] [Google Scholar]
  • 29.Ser T. del et al. Residence, Clinical Features, and Genetic Risk Factors Associated with Symptoms of COVID-19 in a Cohort of Older People in Madrid. GER 67, 281–289 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Kuo C-L et al. APOE e4 Genotype Predicts Severe COVID-19 in the UK Biobank Community Cohort. The Journals of Gerontology: Series A 75, 2231–2232 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Kurki SN et al. APOE ε4 associates with increased risk of severe COVID-19, cerebral microhaemorrhages and post-COVID mental fatigue: a Finnish biobank, autopsy and clinical study. Acta Neuropathologica Communications 9, 199 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Kosmicki JA et al. Pan-ancestry exome-wide association analyses of COVID-19 outcomes in 586,157 individuals. The American Journal of Human Genetics 108, 1350–1355 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Wang C. et al. ApoE-Isoform-Dependent SARS-CoV-2 Neurotropism and Cellular Response. Cell Stem Cell 28, 331–342.e5 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Bonacina F. et al. Myeloid apolipoprotein E controls dendritic cell antigen presentation and T cell activation. Nat Commun 9, 3083 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Laskowitz DT, Lee DM, Schmechel D. & Staats HF Altered immune responses in apolipoprotein E-deficient mice. J. Lipid Res. 41, 613–620 (2000). [PubMed] [Google Scholar]
  • 36.Wang R. et al. Genetic Screens Identify Host Factors for SARS-CoV-2 and Common Cold Coronaviruses. Cell 184, 106119.e14 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Schneider WM et al. Genome-Scale Identification of SARS-CoV-2 and Pan-coronavirus Host Factor Networks. Cell 184, 120–132.e14 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Hoffmann H-H et al. Functional interrogation of a SARS-CoV-2 host protein interactome identifies unique and shared coronavirus host factors. Cell Host & Microbe 29, 267–280.e5 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Yin Y, Sheng Y, Wang M, Ma Y. & Ding H. Interaction of serum proteins with SARS-CoV-2 RBD. Nanoscale (2021) doi: 10.1039/D1NR02687A. [DOI] [PubMed] [Google Scholar]
  • 40.Wozniak MA et al. Apolipoprotein E-ϵ4 protects against severe liver disease caused by hepatitis C virus. Hepatology 36, 456–463 (2002). [DOI] [PubMed] [Google Scholar]
  • 41.Exel E. van et al. Effect of APOE ε4 allele on survival and fertility in an adverse environment. PLOS ONE 12, e0179497 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Burt TD et al. Apolipoprotein (apo) E4 enhances HIV-1 cell entry in vitro, and the APOE ε4/ε4 genotype accelerates HIV disease progression. Proceedings of the National Academy of Sciences 105, 8718–8723 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Blight KJ, McKeating JA & Rice CM Highly Permissive Cell Lines for Subgenomic and Genomic Hepatitis C Virus RNA Replication. Journal of Virology 76, 13001–13014 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Zhang B. & Horvath S. A General Framework for Weighted Gene Co-Expression Network Analysis. Statistical Applications in Genetics and Molecular Biology 4, (2005). [DOI] [PubMed] [Google Scholar]
  • 45.Angelidis I. et al. An atlas of the aging lung mapped by single cell transcriptomics and deep tissue proteomics. Nat Commun 10, 963 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Hurskainen M. et al. Single cell transcriptomic analysis of murine lung development on hyperoxia-induced damage. Nat Commun 12, 1565 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Melms JC et al. A molecular single-cell lung atlas of lethal COVID-19. Nature 595, 114–119 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Zheng J. et al. COVID-19 treatments and pathogenesis including anosmia in K18-hACE2 mice. Nature 589, 603–607 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Sun J. et al. Generation of a Broadly Useful Model for COVID-19 Pathogenesis, Vaccination, and Treatment. Cell S0092867420307418 (2020) doi: 10.1016/j.cell.2020.06.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Bycroft C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Bulk RNA-seq and scRNA-seq data have been deposited at the Gene Expression Omnibus (GEO) under accession numbers GSE184289 and GSE199498, respectively. All data from the UK Biobank is publicly available under www.ukbiobank.ac.uk. MSigDB is publicly available under http://www.gsea-msigdb.org. Source data is provided with this paper.

RESOURCES