Aging‐related cell type‐specific pathophysiologic immune responses that exacerbate disease severity in aged COVID‐19 patients

Abstract Coronavirus disease 2019 (COVID‐19) is especially severe in aged patients, defined as 65 years or older, for reasons that are currently unknown. To investigate the underlying basis for this vulnerability, we performed multimodal data analyses on immunity, inflammation, and COVID‐19 incidence and severity as a function of age. Our analysis leveraged age‐specific COVID‐19 mortality and laboratory testing from a large COVID‐19 registry, along with epidemiological data of ~3.4 million individuals, large‐scale deep immune cell profiling data, and single‐cell RNA‐sequencing data from aged COVID‐19 patients across diverse populations. We found that decreased lymphocyte count and elevated inflammatory markers (C‐reactive protein, D‐dimer, and neutrophil–lymphocyte ratio) are significantly associated with age‐specific COVID‐19 severities. We identified the reduced abundance of naïve CD8 T cells with decreased expression of antiviral defense genes (i.e., IFITM3 and TRIM22) in aged severe COVID‐19 patients. Older individuals with severe COVID‐19 displayed type I and II interferon deficiencies, which is correlated with SARS‐CoV‐2 viral load. Elevated expression of SARS‐CoV‐2 entry factors and reduced expression of antiviral defense genes (LY6E and IFNAR1) in the secretory cells are associated with critical COVID‐19 in aged individuals. Mechanistically, we identified strong TGF‐beta‐mediated immune–epithelial cell interactions (i.e., secretory‐non‐resident macrophages) in aged individuals with critical COVID‐19. Taken together, our findings point to immuno‐inflammatory factors that could be targeted therapeutically to reduce morbidity and mortality in aged COVID‐19 patients.


| INTRODUC TI ON
Coronavirus disease 2019 , a global pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has been diagnosed in more than 284 million people globally, with 5.4 million deaths since December 2019 (data on December 30, 2021). Although a serious risk at any age, SARS-CoV-2 infection is particularly debilitating and deadly for aged patients, defined in this study as 65 years and older (Channappanavar & Perlman, 2020;Clay et al., 2014;Davies et al., 2020;O'Driscoll et al., 2021). The molecular basis of this aging-related vulnerability is an important area of investigation as it is currently poorly understood.
Impaired and dysregulated host immunities, including both innate and adaptive immunities, have been hypothesized as age-based factors in COVID-19 disease severity (Brodin, 2021;Channappanavar & Perlman, 2020). Compared to younger individuals with COVID-19, aged individuals show disrupted antigen-specific adaptive immunity to SARS-CoV-2, such as reduced coordination of CD4-CD8 T-cell responses (Rydyznski Moderbacher et al., 2020). In addition, aged individuals typically produce a less robust type I interferon (IFN) response to flu virus infections (Molony et al., 2017), indicating compromised cellular antiviral defense in innate immunity. Indeed, 13% of aged patients with life-threatening COVID-19 display inborn errors in autoantibodies against type I IFN immunity . In addition, aberrant immunosenescence and inflammation also play crucial roles in age-medicated COVID-19 morbidity and mortality (Domingues et al., 2020). For example, senescent cells become hyper-inflammatory in response to pathogen-associated molecular patterns, and senolytics reduce COVID-19 mortality in aged mice (Camell et al., 2021). Based on these findings, we sought to systematically identify whether there are specific immuno-inflammatory determinants that promote ageassociated COVID-19 severity.

| Severe outcomes in aged COVID-19 patients
To begin, we investigated the prevalence of COVID-19 disease among different age groups with 9 months of data collection.
Analysis of U.S. Centers for Disease Control (CDC) epidemiological data from March to December 2020 (Tables S1-S3) revealed that 80.5% of fatal cases occurred in aged patients. Strikingly, this rate was 4.1 times higher than in 18-64 years old (19.5%), and 1653 times higher than in 0-17 years old (0.05%, Figure 1a). Fatality prevalence was influenced by sex in both older and younger groups ( Figure 1b). Interestingly, we found that average fatal percentage in aged COVID-19 patients is 16% higher than that of influenza (Flu) ( Table S2), indicating that COVID-19 is more hazard for aged individuals than Flu.
To further account for disease comorbidities, we next computed OR across different age groups using a large COVID-19 registry database with 12,651 aged (≥65 years) and 32,426 younger individuals (20-55 years old) ( Figure 1c, Table S4, see Methods). Specifically, we tested the OR Model-2, which is adjusted for sex, race, smoking, and five common disease comorbidities (Guan et al., 2020; COVID-19 severities. We identified the reduced abundance of naïve CD8 T cells with decreased expression of antiviral defense genes (i.e., IFITM3 and TRIM22) in aged severe COVID-19 patients. Older individuals with severe COVID-19 displayed type I and II interferon deficiencies, which is correlated with SARS-CoV-2 viral load. Elevated expression of SARS-CoV-2 entry factors and reduced expression of antiviral defense genes (LY6E and IFNAR1) in the secretory cells are associated with critical COVID-19 in aged individuals. Mechanistically, we identified strong TGF-beta-mediated immuneepithelial cell interactions (i.e., secretory-non-resident macrophages) in aged individuals with critical COVID-19. Taken together, our findings point to immuno-inflammatory factors that could be targeted therapeutically to reduce morbidity and mortality in aged COVID-19 patients.

K E Y W O R D S
aging, cellular immunology, COVID-19, molecular biology of aging, SARS-CoV-2 F I G U R E 1 Epidemiological data analysis between aged and younger COVID-19 patients. (a) The percentage of fatal cases of COVID-19 and flu across three age groups. Data source from U.S. CDC. The upper panel shows the percentage of fatal cases of COVID-19 in the United States. Each dot in the boxplot represents one state. The lower panel shows the percentage of fatal cases of flu from 2010 to 2020. Each dot in the boxplot represents one flu season. Statistical p-value was computed by two-tailed paired t test. For details about CDC dataset, see Tables S1 and S2. (b) Sex differences in the percentage of fatal cases of COVID-19 across three age groups. (c) Odds ratio (OR) analysis of U.S. CDC and COVID-19 registry datasets. U.S. CDC dataset, "Younger" is defined as 20 to 49 years of age (n = 2,369,919), and 'aged' is defined as >60 years old (n = 1,048,011); COVID-19 registry dataset, "Younger" is defined as 18 to 55 years of age (n = 12,651), and 'aged' is defined as ≥65 years old (n = 32,426). OR >1 indicates aged COVID-19 patients with increased likelihood of hospitalization, ICU admission, and death. Two colors denote OR models with different adjusted confounders. Features of the COVID-19 registry dataset are shown in Table  S3 Figure S1a), including longer duration of hospitalization (average duration = 8.9 days; p = 1.4 × 10 −15 , Mann-Whitney U test; Figure S1b), in COVID-19 patients. Taken together, our findings confirm an elevated likelihood of severe outcomes in aged COVID-19 patients has compared with younger patients, even when adjusted for all possible confounding factors.

| Elevated inflammatory responses in aged COVID-19 patients
As severe COVID-19 patients have been reported to have lower lymphocyte count  and higher C-reactive protein (CRP) (Manson et al., 2020), we examined the Cleveland Clinic COVID-19 registry for differences in inflammatory biomarkers as a function of aging. Here, we found lower peripheral lympho- Benjamini-Hochberg multiple test correction; Figure 1d) and higher circulating neutrophils in hospitalized aged COVID-19 patients (q = 0.004; Figure 1d), compared with younger patients. We also found that the neutrophil-lymphocyte ratio (NLR), a marker of systemic inflammation , was elevated in aged COVID-19 patients (q < 2.0 × 10 −16 ; Figure 1d). In addition, the inflammatory markers D-dimer (q < 2.0 × 10 −16 ; Figure 1e) and C-reactive peptide (CRP) (q = 2.7 × 10 −10 ; Figure 1e) were also significantly increased in hospitalized aged patients compared with hospitalized young COVID-19 patients. Those findings motivate us to inspect heterogeneities of immune cells using large-scale immune cell phenotypic profiles and single-cell transcriptomics datasets under a multimodal genomic analytic framework.

| Elevated pro-inflammatory cytokine expression in aged COVID-19 patients
We next examined peripheral immune cell profiles (Takahashi et al., 2020) of hospitalized aged and younger COVID-19 patients by querying a publicly available dataset of 12 major immune cell types (% peripheral blood mononuclear cells [PBMCs]) and 32 T-cell subtypes (% CD3, Table S5, see Methods). All markers and cell type/ subtype definitions are provided in the original study (Takahashi et al., 2020). There was no difference in abundance of the major im- Naïve CD8 T-cell-mediated homeostasis is an important component of antiviral defense (Kaech & Cui, 2012), and the naïve CD8 T-cell receptor repertoire is negatively correlated with age in COVID-19 patients (Ren et al., 2021). Thus, reduced abundance of naïve CD8 T cells may be associated with COVID-19 severities in aged individuals.
We next turned to investigate the ratio of naïve vs. Yet, the ratio of CD4 naïve T cells with other CD4 T sub-cell type and NKT with NK has no significant difference between aged and younger patients in both ICU and non-ICU.

COVID-19 patients
Because we observed loss of CD8 naïve T cells and T effector memory cells in hospitalized aged COVID-19 patients (Figure 2 b, d), we examined a publicly available single-cell transcriptomic dataset of CD8 T cells from 25 severe/critical COVID-19 patients (aged n = 12; younger n = 13) (Stephenson et al., 2021) in order to search for aging-related molecular mechanisms in a cell type-specific manner.
Uniform Manifold Approximation and Projection (Becht et al., 2019) (UMAP) analysis revealed five distinct CD8 sub-clusters ( Figure 3a and Figure S3) based on biomarkers provided from the original literature (See Method, Stephenson et al., 2021), including naïve CD8, T central memory (Tcm), Tem, CD8 proliferation, and CD8 terminal effector T cell (also designated as TEMRA, Thome et al., 2014). We found that aged and younger patients with severe COVID-19 showed age-dependent immune pathway profiles across five CD8 subtypes.
For example, type I and II IFN signaling showed decreased effect in CD8 naïve T cells, CD8 Tem, and CD8 proliferation T cells isolated from PBMC in aged severe COVD-19 patients, not in younger patients ( Figure 3b). In addition, the antigen processing and presentation pathway showed decreased effect in CD8 Tem and CD8 TEMRA in aged patients as well. Our finding indicates that type I and II IFN signaling and antigen processing and presentation pathways are age-related immune pathways associated with COVID-19 disease severity. Yet, Th17 cell differentiation pathway of CD8 TEMRA and exhaustion consensus of CD8 T proliferation cells were activated in both aged and younger patients with severe COVID-19.
We next turned to investigate the molecular network in CD8 naïve T cells. Comparing to severe young COVID-19 patients, upregulated genes (q < 0.05, log-fold change >0.1) in CD8 naïve T cells from aged patients formed a network module (the largest connected component) in the human protein-protein interactome ( Figure 3c).

CoV-2 viral load in aged patients
To further investigate the relationship between viral load and COVID-19 disease severity, we analyzed bulk RNA-seq data from nasopharyngeal samples (Lieberman et al., 2020) (see Methods).

Consistent with our findings in naïve CD8 T cells, expression levels
of IFNα genes (IFNA1, IFNA5, IFNA7, and IFNA8) were significantly decreased in aged patients with high viral load ( Figure 4a). In addition, the expression of IFNG was decreased in aged patients with low viral load ( Figure S4a). Notably, we found that the IFN-stimulated antiviral genes (Sadler & Williams, 2008), including IFIT1 and OAS1 (2'-5'-oligoadenylate synthetase 1), were down-regulated in aged patients with a higher viral load ( Figure 4b). Next, we performed gene set enrichment analysis (GSEA, see Methods) for differentially ex- The single-cell transcriptomic dataset (25 of Severe\Critical COVID-19 patients, aged n = 12, younger n = 13) was collected from a recent study (Stephenson et al., 2021) (Table S1 and Method). (b) Pathway enrichment analysis across five CD8 T-cell subtypes. Black circle indicates q < 0.05. (c) A highlighted protein-protein interaction subnetwork for age-biased differentially expressed genes in CD8 naïve T cells from patients with critical COVID-19. The colors for nodes and edges represent different immune pathways SARS-CoV-2 has evolved several mechanisms to blunt IFN induction, including the direct targeting of MDA5 (melanoma differentiationassociated protein 5), a RIG-I-like receptor, by the viral papain-like protease (PLpro) . Furthermore, IFN potently inhibits IL-8 expression (Aman et al., 1993) in viral infection, and we also showed that aged COVID-19 patients with high viral load exhibit elevated plasma IL-8 (p = 0.005, Mann-Whitney U test; Figure 4c).
Notably, up-regulated genes in aged patients with high viral load were not enriched in immune pathways (Figure 4b and Figure S4b), indicating decreased immune ability in response to SARS-CoV-2 infection. Taken together, our data show that IFN deficiency is associated with elevated SARS-CoV-2 viral load in aged patients.

| Age-dependent increased expression of SARS-CoV-2 entry factors
We  Table S6); furthermore, BSG and CD147 showed elevated expression in Treg (regulatory T cell) and CD8 T cells (Figure 4d) as well. We also found that the S protein priming proteases TMPRSS2  and FURIN  were highly expressed in epithelial cells in critical and moderate COVID-19, with no differences between aged and young patients ( Figure 4d and Table S6). However, FURIN levels were in-

| Increased immune-epithelial cell interactions in aged COVID-19 patients
To further investigate the immunological mechanisms underlying age-associated COVID-19 outcomes, we performed Gene-set enrichment analysis (GSEA) to explore transcriptomic signatures on 22 immune pathways across 15 cell types derived from nasal tissue (see Methods). Here, we observed distinct immune responses between older and younger individuals with critical or moderate COVID-19 ( Figure S6) in epithelial and immune cell types. We further used CellphoneDB (Efremova et al., 2020) to quantify ligand-receptor interactions between epithelial and immune cells and found an elevated number of significant ligand-receptor interactions involved in immune-epithelial interactions (q < 0.05, permutation test with BH multiple testing correction (Benjamini & Hochberg, 1995), Table   S7) in aged patients with critical COVID-19 (Figure 5a). In addition, we also found a stronger immune-epithelial cell interaction network in aged patients. In particular, we noted that secretory-non-resident macrophages (nrMa) displayed the highest connection with other cell types in aged patients with critical COVID-19 ( Figure 5a).
We next analyzed ligand-receptor interactions of secretory/

| DISCUSS ION
This study provides a comprehensive analysis of immune profiles in aged and younger COVID-19 patients using large, electronic patient data from the CDC and the Cleveland Clinic Registry database.  Figure 2h). IL-6 is a potential therapeutic target since it is a critical mediator of cytokine storm in COVID-19 . However, a recent phase III clinical trial (NCT04320615) showed no reduced mortality in severe COVID-19 patients treated with the anti-IL-6R monoclonal antibody tocilizumab (Rosas et al., 2021). Younger COVID-19 patients in the ICU also showed significantly higher IL-10 ( Figure 2h). Our observations suggest that targeting IL-10 might reduce mortality in younger patients with severe COVID-19. Furthermore, an anti-IL-8

drug (BMS-986253) is under testing for COVID-19 patients in a
Phase 2 clinical trial (ClinicalTrials.gov Identifier: NCT04347226). Therefore, our findings suggested that age is an important biological variable in evaluation of clinical benefits of anti-IL-8 intervention trials.
We also found reduced lymphocytes in hospitalized aged COVID-19 patients (Figure 1d). In particular, the abundance of naïve CD8 T cells was decreased in aged patients with severe COVID-19 ( Figure 2d). Reduction of naïve CD8 T cell is a hallmark of immunosenescence in older individuals (Goronzy et al., 2015), and through scRNA-seq data analysis, we observed significant enrichment of upregulated apoptosis genes in CD8 naïve T cells from aged COVID-19 patients. Mechanistically, the apoptosis driver gene CTSD (Cocchiaro et al., 2016) is significantly elevated in naïve CD8 T cells from aged severe/critical COVID-19 patients compared with younger patients (q < 2.0 × 10 −16 ). Thus, modulation of CD8 naïve T-cell dysfunction, especially targeting apoptosis pathway (Chu et al., 2021), may provide a new treatment strategy for severe COVID-19 in aged patients.
IFN-mediated immunity provides initial rapid protection against viral infection (McNab et al., 2015), and about 3.5% of patients with life-threatening COVID-19 show genetic aberrations in the type I IFN pathway . A recent genetic study in European ancestry revealed that the cis-protein quantitative trait loci (pQTL, rs4767027) in OAS1 (an IFN-stimulated gene) were significantly associated with decreased likelihood of COVID-19 susceptibility and severity . Herein, we found that aged individuals with severe COVID-19 show reduced expression of type I IFN genes (Figures 3b,c, 4a, and 5b). Notably, aged patients with high SARS-CoV-2 viral load show reduced expression of OAS1 and IFNA1, IFNA5, and IFNA7 ( Figure 4a) compared with younger patients. On the contrary, aged patients with high SARS-CoV-2 viral load have elevated expression of the pro-inflammatory cytokine IL-8 and decreased lymphocyte cell counts in plasma (Figure 4c), demonstrating dysregulation of cytokine responses that has been well described for COVID-19 (Acharya et al., 2020). Of note, the dysregulated cytokine response is likely the effect of a variety of immunomodulatory strategies employed by SARS-CoV-2 that are used to manipulate specific signaling pathways that lead to cytokine induction such as the RIG-Ilike receptor pathway. Now, there are more than 40 ongoing clinical trials (https://clini caltr ials.gov/) to test interferon-related therapies for potential treatment of COVID-19. Our findings suggested that interferon-related therapies may provide more clinical benefits for older individuals with COVID-19.
Although aged adults show increased susceptibility to SARS-CoV-2 infection compared to children (Davies et al., 2020), we did not find differences in SARS-CoV-2 viral load in the upper airways between younger and aged patients ( Figure S8). Using large-scale scRNA-seq data analysis, we did find, however, that the SARS-CoV-2 entry genes (ACE2, BSG, TMPRSS2, FURIN, and NPR1) showed cell type-specific expression profiles in both aged and younger individuals. In aged patients with critical COVID-19, the expression of BSG was increased in secretory, nrMa and CD8 T cells, and elevated expression of FURIN was found in Treg and CD8 T cells. Thus, cell type-specific host factor expression may play an important role in age-mediated disease susceptibility and severity in COVID-19.
We also identified age-specific increases in immune-epithelial cell interactions. For example, we found strong TGFβ-mediated immune-epithelial cell interactions in aged severe COVID-19 patients (Figure 5b and Figure S7). TGFβ plays a crucial role in pulmonary fibrosis (Khalil et al., 1991;Lee et al., 2001), which is a common complication in severe COVID-19 patients (Leeming et al., 2021).
Lastly, we acknowledge the potential limitations of our study.
Although we inspected omics data from multiple tissues, including PBMCs, plasma, and nasal tissues, additional analysis of other COVID-19 and aging relevant tissues, such as lung and brain, should be investigated in the future. In addition, our COVID-19 database and omics data were generated from acute COVID-19 patients, and identification of the underlying genetic and molecular basis of aging differences for long-haul COVID-19 patients will be an important area of future investigation (Sudre et al., 2021). As the inconsistent correlation between RNA expression and protein expression (Buccitelli & Selbach, 2020), further investigation of differential protein expression of ACE2, BSG receptors, and the TGFβ using proteomics data is highly warranted in the future studies. Finally, investigation of COVID-19 vaccine responses between aged and young patients is also warranted in the future.

| E XPERIMENTAL PROCEDURE S
"Younger" was defined as 18 to 55 years of age, and "aged" was defined as ≥65 years old.

| U.S. CDC COVID-19 epidemiological data
Publically accessible COVID-19 death counts in 54 states and territories in United States were downloaded from the CDC Web site (https://data.cdc.gov/NCHS/Provi siona l-COVID -19-Death -Count s-by-Sex-Age-and-S/9bhg-hcku/data) on 23 December 2020 (Table   S1). Publically accessible statistics of influenza mortality across December 2020. This dataset includes age-stratified COVID-19 case counts in hospitalization, ICU admission, death, sex, and race. We extracted two age subgroups from all laboratory-confirmed cases using the following criteria: i) the age range of younger group from 20 to 49 years and the age range of older group over than 60 years (Table S2); ii) deletion of all cases in which sex and race information was missing. In total, the younger subgroup includes 2,369,919 cases, with 94,161 in hospitalization, 9138 in ICU admission, and 6469 death cases. The older subgroup has 1,048,011 cases in total, with 243,109 in hospitalization,29,671 in ICU admission,and 124,566 death cases. This dataset was used to determine OR analysis.

| COVID-19 registry database
We used institutional review board-approved COVID-19 registry data, including 45,077 individuals (12,651 aged patients and 32,426 younger patients; Table S3)  The data in COVID-19 registry include COVID-19 test results, baseline demographic information, and all recorded disease conditions (Table S3). We conducted a series of retrospective studies to test the association of aging with COVID-19 outcomes, including hospitalization, ICU admission, mechanical ventilation, and death.
Data were extracted from electronic health records (EPIC Systems), and patient data were managed using REDCap electronic data capture tools. To ensure data quality, a study team trained on uniform sources for the study variables manually checked all datasets.
Statistical analysis for smoking, hypertension, diabetes, coronary artery disease asthma, and emphysema and COPD was calculated after missing value deletion.

| Clinical outcome analysis
The OR was used to measure the association between COVID-19 outcomes and aging based on logistic regression. An OR >1 indicates that aged patients are associated with a higher likelihood of the outcome. To reduce the bias from confounding factors, we employed OR analysis in two datasets. For U.S. CDC datasets, the OR model was adjusted by sex and race, due to limited information of other confounding factors. However, in the COVID-19 registry, we adjusted for sex, race, smoking, hypertension, diabetes, coronary artery disease, asthma, emphysema, and COPD. The Kaplan-Meier method was used to estimate the cumulative hazard of hospitalization of COVID-19 patients across age groups. For hospitalization outcome, the time was calculated from the start date of COVID-19 symptoms to hospital admission date. Log-rank test was used for comparison across different age groups with Benjamini and Hochberg adjustment (Benjamini & Hochberg, 1995). Cumulative hazard analysis was performed using the Survival and Survminer packages in R 3.6.0 (https://www.r-proje ct.org).

| Public available COVID-19 multi-omics datasets used in this study
Detailed information of the list datasets shown in Table S1.

| Two single-cell sequencing datasets
In this study, we used two COVID-19 single-cell datasets (Table S1).
Based on our aging criteria, the critical/severe COVID-19 patients were grouped to aged (n = 12) and younger patients (n = 13). 2) A single-cell dataset from nasal tissues (Chua et al., 2020) (European Genome-phenome Archive repository: EGAS00001004481) was from COVID-19-positive patients (11 critically ill patients and 8 moderately ill patients). Based on our aging criteria, we extracted a subpopulation from the original cohort. The final COVID-19 cohort used in this study included 8 critically ill patients (5 younger and 3 older patients) and 7 moderately ill patients (4 younger and 3 older patients). As the original dataset supplied cell type information, additional analysis was based on cell type annotation. The dataset contained 115,895 cells across 15 cell types (B cell, Basal, Ciliated, Ciliated-diff, CD8 T cell, moDC, Neu, NKT, NKT-p, nrMa, rMa, Secretory, Secretory-diff, Squamous, and Treg).

| Bulk RNA-sequencing dataset in nasal tissue (Lieberman et al., 2020)
The dataset was publically available from NCBI GEO database (GSE152075). Based on original meta-information, we extracted COVID-19-positive sample data with high or low viral load, deleting samples in which sex and age information were missing. 147 bulk RNA-seq samples were used in this study, including 61 aged patients (high viral load n = 27, low viral load n =34) and 86 younger patients (high viral load n = 46, low viral load n = 40).

| SARS-CoV-2 viral load dataset (Fajnzylber et al., 2020)
We quantified SARS-CoV-2 RNA load from 5 specimen types, in-   (Table S5). It also detected the plasma concentration of 71 cytokines through cytokine array. Based on our age criteria, the dataset included 81 hospitalized patients, 40 with longitudinal data. When the second follow-up time of a patient was greater than 7 days, it was recorded as two samples. Hence, 114 samples were analyzed, which included 94 older samples (non-ICU n = 66, ICU = 26) and 50 younger samples (non-ICU n = 37, ICU = 13).

| Single-cell sequencing data analyses
All single-cell data analyses and visualizations were performed with the R package Seurat v3.1.4 40. The data quality filtering was strictly followed by the original literature (Chua et al., 2020;Ren et al., 2021). "NormalizeData" was used to normalize the data. with 15 principal components were used. A resolution of 0.5 was used in "FindClusters()" step. "FindAllMarkers" function with the MAST test was employed as the finding maker method for each cell type.

| Cell-cell interaction analysis
Cell-cell interaction analysis was based on normalized expression data of known ligand-receptor pairs in 15 cell types of nasal singlecell sample. The analysis was performed by CellPhoneDB (Efremova et al., 2020) v2.1.4 (https://github.com/Teich lab/cellp honedb) based on the python 3.7 platform. Statistical analysis mode was used to identify significant ligand-receptor pairs in each cell number. A permutation test (1000 randomizations) with BH multiple testing correction was used to evaluate the significance.

| Bulk RNA-sequencing data analysis
All bulk RNA-sequencing data analysis started from raw counts value. R package edgeR (Robinson et al., 2010) v3.12 was used to analyze differentially expressed genes in older vs. younger groups.
Correction for sex and batch effects was added into the formula of design model. Statistical significance p-values were adjusted by BH (q value) method (Benjamini & Hochberg, 1995). Differentially expressed genes were identified as adjusted p-value (q) <0.05 and log-fold change >0.5.

| Immune gene set enrichment analysis
To evaluate the immune pathway profiles in young and aged COVID-19 patients, GSEA was conducted as previously described (Subramanian et al., 2005). Immune gene profiles were retrieved from the KEGG database (Kanehisa et al., 2017). We selected 22 immune-related pathways and 1241 genes from KEGG belonging to the immune system subtype. For each cell type, we performed a GSEA on the list of differential expressed genes (DEGs) ranked by the log 2 FC. The normalized enrichment score (NES, Equation 1) was calculated for 22 immune pathways in young and aged specific gene sets (Figure 4b), in which ES (Subramanian et al., 2005) denotes enrichment score.
Normalization of the enrichment score reduced the effect of the differences in gene set size and in correlations between gene sets and the expression dataset. NES score >0 and q < 0.05 indicate that up-regulated DEGs in aged vs. young are significantly enriched in immune pathways, while NES score <0 and q < 0.05 indicate down-regulated DEGs in aged vs. young are significantly enriched in immune pathways. Permutation test (1000 times) was performed to evaluate the significance. All analyses were performed with the prerank function in GSEApy package (https:// gseapy.readt hedocs.io/en/maste r/index.html) on Python 3.7 platform.

| Statistical analysis
Statistical tests for assessing categorical data through chi-square test and the two-tailed Mann-Whitney U test were used to compare the difference in continuous variable by aged vs. younger.
Spearman's ρ was assessed for correlation between two variables.
Statistical significance level was set at q < 0.05 and corrected by Benjamini-Hochberg (false discovery rate) method. All statistical analysis was performed by SciPy Statistics (https://docs.scipy.org/ doc/scipy/ refer ence/stats.html#modul e-scipy.stats).

ACK N OWLED G M ENTS
This work was primarily supported by the National Institute of Aging (R01AG066707, U01AG073323, and R01AG066707-01S1) to F.C. This work was supported in part by NIH Research Grant

DATA AVA I L A B I L I T Y S TAT E M E N T
The clinical and transcriptomic datasets used in this study are publicly available; for details, see Table S1. The code for singlecell analysis can be found in https://github.com/Cheng F-Lab/ COVID -19_Map.