Differential dynamics of peripheral immune responses to acute SARS-CoV-2 infection in the elderly

Abstract

Severe COVID-19 disproportionately impacts older individuals and those with comorbidities. Indeed, although the majority of COVID-19 cases have been reported in individuals under the age of 50, 95% of COVID-19 deaths have occurred in people who were 50 or older. However, the immunological underpinnings of severe COVID-19 in older patients have yet to be defined. This study investigated the longitudinal immune response to SARS-CoV-2 infection in a cohort of young and aged patients with varying disease severity. Phenotypic transcriptional and functional examination of peripheral blood mononuclear cells revealed age-, time, and disease severity-specific adaptations. Gene expression signatures within memory B cells and plasmablasts correlated with reduced frequency of antigen specific B cells and neutralizing antibody titers in aged patients with severe COVID-19. Moreover, older patients with severe disease exhibited exacerbated T cell lymphopenia, which correlated with lower levels of plasma IL-2, and diminished antigen specific T cell responses. Single cell RNA sequencing revealed augmented signatures of activation, exhaustion, cytotoxicity, and type-I interferon signaling in memory T cells and NK cells with age. Although hallmarks of a cytokine storm were evident in both age groups, older individuals exhibited elevated levels of chemokines that mobilize inflammatory myeloid cells, notably in those who succumbed to disease. Correspondingly, we observed a re-distribution of DC and monocytes with severe disease that was accompanied by a rewiring towards a more regulatory phenotype. Several of these critical changes, such as the reduction of surface HLA-DR on myeloid cells, were reversed in young but not aged patients over time. In summary, the data presented here provide novel insights into the impact of aging on the host response to SARS-CoV2 infection.

INTRODUCTION

The coronavirus disease-2019 (COVID-19) pandemic, caused by the respiratory virus severe acute respiratory syndrome coronavirus (SARS-CoV-2), resulted in over 32 million cases and 580,000 deaths in the U.S. alone. SARS-CoV-2 targets lung airway and alveolar epithelial cells, vascular endothelial cells, and macrophages in the lung. Infection triggers a vigorous innate and adaptive immune response that culminates in viral clearance or, in the case of severe infection, the development of cytokine storm, acute respiratory distress syndrome (ARDS), and multi-organ failure¹. The estimated mortality rate of COVID-19 in the US is 1–2%, while ~80% of patients are thought to be asymptomatic or experience mild disease symptoms^{2, 3}.

Although there are numerous questions that remain to be answered regarding the large discrepancies in disease outcomes, it is clear that age is a significant risk factor for severe COVID-19 and death^{4, 5, 6}. Indeed, case fatality rate worldwide has been shown to increase progressively with age (0.4% at age 55, 1.4% at age 65, 4.6% at age 75, and 15% at age 85)^{7, 8}. The highest proportion of patients that develop severe respiratory complications such as respiratory failure, dyspnea, pneumonia, ARDS, and death are the elderly^{9, 10, 11, 12, 13, 14}. Age-related changes in immunity, dubbed “immunosenescence”, result in a marked increase in susceptibility to respiratory viral infections and attenuated vaccine response^{15, 16}. Some of the hallmarks of immunosenescence include decreased frequency of naive T and B cells, increased prevalence of effector/terminally differentiated memory cells, as well as increased levels of circulating inflammatory mediators including IL-6 and C-reactive protein referred to as “inflammaging¹⁷. Despite the rich literature on COVID-19 pathogenesis, the specific immune mechanisms associated with increased severity of COVID-19 amongst the elderly and the immune correlates of mild and severe disease in the aged population remains elusive.

While age-associated immune changes in the lung undoubtedly contribute largely to a reduced ability to fight SARS-CoV-2 infection¹⁸, those studies are challenges to conduct. Consequently, most studies to date have relied on profiling circulating immune cells¹⁹. These studies have shown that severe COVID-19 is mediated by dysregulated innate and adaptive immune responses. Elevated levels of interleukin (IL)-1, IL-6, IL-8, and CXCL10 in the blood have been associated with more severe infection or death^{20, 21, 22, 23, 24}. Disruptions in distribution, activation as well as gene expression of circulating immune cells have been identified by single cell RNA sequencing. Notably, a pronounced decrease of HLA-DR on monocytes^{22, 25, 26}, an increase in non-classical CD16+ monocytes²², and a decrease in the frequency of total dendritic cells (DCs)²⁶ compared to patients with moderate/mild COVID-19 and healthy controls. Additionally, it has been shown that mild disease results in increased IFN-signaling gene expression, whereas severe disease leads to impaired type-I interferon responses²⁷ and reduced ability to respond to a secondary stimulation in monocytes²⁸. Studies examining NK cells have reported reduced numbers of total NK cells with disease severity, but no differences within subsets^{29, 30, 31}. Similarly, NK cells have been shown to upregulate interferon response genes but have an exhausted phenotype and reduced cytolytic activity^{29, 30}. Another hallmark of severe COVID-19 is lymphopenia³² that is accompanied by increased T cell activation and exhaustion^{30, 33, 34, 35, 36}.

Data from these studies have begun to paint a picture of significant immune dysregulation with COVID-19 that is potentially exacerbated in aged patients^{37, 38}. Regardless, gaps in our knowledge remain. Notably, comprehensive longitudinal studies that include both young and aged subjects with varying disease severity (from mild to fatal) as well age-matched controls and integrated transcriptional and functional analyses are needed. In this study, we carried out a longitudinal phenotypic, transcriptional, and functional analysis of peripheral blood mononuclear cells (PBMC) from young and aged COVID-19 patients compared to age-matched healthy controls. Analysis of circulating immune mediators revealed higher levels of cytokines important for T cell survival and IFNβ in young subjects, while older subjects exhibited higher levels of myeloid cell recruiting chemokines. Single cell RNA sequencing analysis revealed dampened activation signatures in plasmablasts and memory B cells isolated from aged subject that correlated with reduced neutralizing titers. Analysis of the T cell subsets revealed heightened activation and exhaustion states with severe disease and an age-associated exacerbation of interferon and cytotoxicity signaling. Functionally, this led to immune paralysis in the aged, with worsened T cell lymphopenia and diminished anti-SARS-CoV-2 T cell responses. Similarly, signatures of increased exhaustion and cytotoxicity markers were more evident in NK cells from aged subjects. Finally, severe COVID-19 resulted in a functional rewiring of monocytes and DC towards a more regulatory phenotype where NF-κB mediated responses were diminished while type I IFN responses were enhanced. The findings presented enhance our understanding of age-mediated disruptions in anti-viral responses and will aid the design of therapeutics and vaccines for this at-risk population.

MATERIALS AND METHODS

Ethics Statement

This study was approved by University of California Irvine Institutional Review Boards. Informed consent was obtained from all enrolled subjects.

Study Participants and Experimental Design

Blood samples from patients admitted to University of California Irvine Medical Center (UCIMC) and participating in the NIH ACT-1 trial were used in these studies. Participants gave written consent to have the remainder of their blood samples used for ancillary studies related to COVID-19. Samples were stratified by disease severity - healthy donors (HD), mild/moderate COVID-19, and severe COVID-19 and age (<60 categorized as young and ≥ 60 categorized as aged where median ages of the young and aged patients were 43 and 70, respectively). Samples from healthy Donors (n=49; 37 young and 12 aged) were obtained from previous unrelated studies collected prior to 2018 and blood from seronegative healthcare workers collected after March 2020. Individuals with asymptomatic/mild disease (n=25; 13 young and 12 aged) were identified as those that tested positive for SARS-CoV-2 during their visit to UCIMC for reasons unrelated to COVID-19 symptoms (e.g. heart attack, exacerbation of auto-immune disease, elective surgeries). These samples were obtained through the COVID-19 biospecimen bank at UCIMC. A total of 51 patients with severe COVID-19 (35 young and 16 aged) were profiled, including patients with severe illness requiring hospitalization (ward), critical illness requiring intensive care unit admission without/with intubation (ICU; ICU/I). For patients with severe disease, blood was collected longitudinally over several days post symptom (DPS) onset, with median DPS being 10 and 11 days for young and aged group respectively. Six patients (2 young and 4 aged) from our analysis succumbed to disease. Due to limited sample availability, only a subset of samples were utilized in each of the assays. Detailed characteristics of the cohorts and experimental breakdown by samples is provided in Supp Table 1.

Plasma and Peripheral Blood Mononuclear Cells (PBMC) Isolation

Whole blood samples were collected in EDTA vacutainer tubes. Peripheral blood mononuclear cells (PBMC) and blood plasma samples were isolated after whole blood centrifugation 1200 g for 10 minutes at room temperature in SepMate tubes (STEMCELL Technologies). Plasma was stored at −80°C until analysis. PBMC were cryo-preserved using 10% DMSO/FBS and Mr. Frosty Freezing containers (Thermo Fisher Scientific) at −80C then transferred to a cryogenic unit until analysis.

Measuring circulating immune mediators

Concentration of immune mediators in the plasma were measured using the Human XL Cytokine Discovery Premixed Kit (R&D Systems) that includes cytokines (IFNα, IFNβ, IFNγ, IL-1β, IL-10, IL-12p70, IL-13, IL-15, IL-17A, IL17E, IL-1RA, IL-2, IL-7, TNFα, TRAIL, IL-33), chemokines (CCL2/MCP-1, CCL3/MIP-1α, CCL4/MIP-1β, CCL20/MIP-3α, CCL5/RANTES, CCL11/Eotaxin, CXCL1/GROα, CXCL2/GROβ, CXCL10/IP-10, CX3CL1/Fractalkine), growth factors (GM-CSF, G-CSF, EGF, VEGF, PDGF-AA, TGFα) and effector molecules (Granzyme B, PD-L1, CD40L, Flt-3L). Samples were diluted per manufacturer’s instructions and run in duplicates on the Magpix Instrument (Luminex, Austin, TX). Data were fit using a 5P-logistic regression on xPONENT software (version 7.0c).

Antibody End Point Titer ELISA

Clear 96 well, high-binding polystyrene ELISA plates were coated with 100 uL/well of 500 ng/mL SARS-CoV-2 Spike-protein Receptor-Binding Domain (RBD) (GenScript) in PBS overnight at 4C. Plates were blocked using 5% milk in wash buffer (0.05% Tween in phosphate-buffered saline [PBS]) for 1 h at room temperature, washed three times with wash buffer, and incubated with heat-inactivated (55°C for 30 min) plasma samples (1:30 starting dilution) in 3-fold dilutions in duplicate for 1 h. Responses were visualized by adding HRP-anti-human IgG (BD Pharmingen) or IgA (Cell Sciences) to the wells another 90 min at RT, followed by 3 washes and the addition of o-Phenylenediamine dihydrochloride substrate (ThermoFisher Scientific) as substrate. Reaction was stopped using 1M HCl. ODs were read at 490 nm on a Victor3|^™ Multilabel plate reader (Perkin Elmer). Endpoint IgG and IgA titers were calculated using log-log transformation of the linear portion of the curve with 0.1 optical density (OD) units as the cutoff. Titers were standardized using a positive-control sample included with each assay.

Focus Reduction Neutralization Test (FRNT)

Heat inactivated plasma serially diluted (1:3) in HyClone Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10mM of HEPES buffer were combined with SARS-CoV-2 (100 PFU) at 37°C 5% CO2 for 1 hour. The antibody-virus inoculum was then transferred onto Vero E6 cells (ATCC C1008) that were seeded in a 96-well plate at 3.5×10⁴ cells/well the prior day. After 1 hour incubation at 37°C 5% CO2, 1% methylcellulose (Sigma Aldrich) in DMEM (5% FBS) was overlaid on the infected Vero cell layer. Plates were incubated at 37°C 5% CO2 for 24 hours. After 24 hours, the medium was carefully removed, and the plates were fixed with 100μl of 10% neutral buffered formalin for 1 hour at room temperature. Following fixation, plates were washed with deionized (DI) water and 50μl of ice-cold Methanol supplemented with 0.3% of hydrogen peroxide was added to each well. Plates were incubated at −20°C for 10 minutes followed by 20 minutes at room temperature. The plates were washed 3 times with DI water and blocked for 1 hour with 5% non-fat dry milk in PBS. Anti-SARS Nucleocapsid antibody (Novus Biologicals NB100–56576) at 1:1000 dilution in 5% non-fat dry milk/PBS was added. After an overnight incubation at 4°C, plates were washed 4 times with PBS, 40μl of HRP anti-rabbit IgG antibody (BioLegend) 1:1500 was added to each well, and plates were incubated 2 hours at room temperature. Plates were developed using True Blue HRP substrate and imaged on an ELISPOT reader. Each plate included a positive and a negative control. The half maximum inhibitory concentration (IC50) was calculated by non-linear regression analysis using normalized counted foci on Prism 7 (Graphpad Software). 100% of infectivity was obtained normalizing the number of foci counted in the wells derived from the cells infected with SARS-CoV-2 virus in the absence of plasma.

Phenotypic analysis of circulating lymphocytes and monocyte by flow cytometry

Frozen PBMCs were thawed, washed in FACS buffer (2% FBS, 1mM EDTA in PBS) and counted on TC20 (Biorad) before surface staining using two independent flow panels (Supp. Figure 8). For the innate panel, the following antibodies were used: CD3 (SP34, BD Pharmingen) and CD20 (2H7, BioLegend) for the exclusion of T & B lymphocytes, respectively. We further stained for CD56 (BV711, BioLegend), CD57 (HNK-1, BioLegend), KLRG1 (SA231A2, BioLegend) CD16 (3G8, BioLegend), CD14 (M5E2, BioLegend), HLA-DR (L243, BioLegend), CD11c (3.9, ThermoFisher Scientific), CD123 (6H6, BioLegend) and CD86 (IT2.2, BioLegend). For the adaptive panel, the following antibodies were used: CD4 (OKT4, BioLegend), CD8b (2ST8.5H7, Beckman Coulter), CD45RA (HI100, TONBO Biosciences), CCR7 (G043H7, BD Biosciences), CD19 (HIB19, BioLegend), IgD (IA6–2, BioLegend), CD27 (M-T271, BioLegend), KLRG1 (SA231A2, BioLegend) and PD-1 (Eh12.2h7, BioLegend). Cells were surface stained at 4°C for 30 minutes, permeabilized, fixed and stained intracellularly with Ki67 (clone B56) for adaptive panel and Granzyme B (QA16A02) for innate panel. Dead cells were excluded using the Ghost Dye viability dye (TONBO biosciences).

Monocytes from a subset of samples were phenotyped using an additional panel. Briefly cells were thawed, incubated with Fc blocking reagent (Human TruStain FcX, BioLegend) for 10 minutes at RT, and stained at 4°C for 30 minutes using the following panel: CD14 (M5E2, BioLegend), HLA-DR (L243, BioLegend), CD62L (DREG-56, BioLegend), CD163 (GHI/61, BioLegend), CCR5 (J418F1, BioLegend), CD40 (5C3, BioLegend), CCR2 (FAB151C, R&D Systems), CD64 (10.1, BioLegend), CX3CR1 (2A9–1, BioLegend), CD80 (2D10, BioLegend), CD11b (ICRF44, BioLegend), and Monocyte Blocker (Tru-Stain Monocyte Blocker, BioLegend). Samples acquired on the Attune NxT acoustic focusing cytometer (Life Technologies).

T cells from a subset of samples were phenotyped using two separate panels. PBMC were stained at 4°C for 30 minutes using either: 1) panel 1- CD4 (OKT4, BioLegend), CD8 (RPA-T8, BD Biosciences), CD3 (OKT3, BioLegend), CD38 (HIT2, Tonbo Biosciences), HLA-DR (LN3, BD Biosciences), PD-1 (EH12.2H7, BioLegend), CD25 (IL2R, BC96, BioLegend), CD127 (IL7R, RDR5, Invitrogen), CD95 (DX2, BioLegend), CTLA-4 (BNI3, BioLegend); 2) panel 2 - CD4 (OKT4, Tonbo Biosciences), CD3 (OKT3, BioLegend), CD134 (OX40, BER-ACT35, BioLegend), CD69 (FN50, BioLegend), CD154 (CD40L, TRAP1, BD Biosciences), CD137 (4–1BB, 4B4–1, BioLegend) separately. Samples were washed and analyzed on the Attune NxT (Life Technologies). All flow data were analyzed using FlowJo v10 (TreeStar, Ashland, OR, USA).

To detect antigen specific B cells, ~ 1×10e6 PBMC were stained with 100 ng of full-length biotinylated spike protein (Sino Biological) that was pre-incubated with Streptavidin-BV510 (BioLegend) at 2:1 ratio for 1 h at 4C and CD20-FITC (2H7, BioLegend) for 30 minutes at 4°C. Negative controls were PBMC stained with Streptavidin PE (BioLegend) and CD20-FITC. Samples were washed twice and resuspended in 200 uL FACS buffer before being analyzed on Attune NxT (Life Technologies)

Monocyte stimulation assay

Approximately 1×10e6 PBMC were cultured in the presence of RPMI (controls), bacterial and viral agonists at 37C. Bacterial agonist cocktail consists of a combination of 2ug/mL Pam3CSK4 (TLR1/2 agonist, InvivoGen), 1 ug/mL FSL-1 (TLR2/6 agonist, Sigma Aldrich), and 1 ug/mL LPS (TLR4 agonist from E. coli O111:B4, InvivoGen). Viral agonist cocktail consists of 5 ug/mL Imiquimod (R848, TLR7 agonist, InvivoGen), 1 ug/mL ssRNA (ssRNA40 LyoVec, TLR8 agonist, InvivoGen), and 5 ug/mL ODN2216 (CpG ODN, TLR9 agonist, InvivoGen). Samples were stimulated for 1 hour after which protein transport inhibitor (Brefeldin A) was added and incubated for an additional 7 hours at 37C. Cells were then washed twice in FACS buffer and surface stained using the following antibody cocktail - CD14 (M5E2, BioLegend), HLA-DR (L243, BioLegend), CD11b (ICRF44, BioLegend) for 30 minutes at 4C. Stained cells were then fixed and permeabilized using Fixation buffer (BioLegend) and incubated overnight with a cocktail of intracellular antibodies - IL-6 (MQ2–6A3, BioLegend), TNFɑ (MAb11, eBioScience), IFNɑ (LT27:295, Miltenyi Biotec). Samples were acquired on the Attune NxT acoustic focusing cytometer (Life Technologies). Data were analyzed using FlowJo v10 (TreeStar, Ashland, OR, USA).

T cell stimulation assay with SARS-CoV-2 overlapping peptide pools

T cells from healthy donors (young and aged) and severe patients (young and aged) were MACS purified from PBMC using anti-CD2 magnetic beads (Miltenyi Biotec). Purity of positive fractions was confirmed using flow cytometry and were >90%. 1×10e5 T cells/well were stimulated with 1 ug of a SARS-CoV-2 overlapping peptide pool covering Nsp14/E/M/ORF3a/6/7a/7b/8/10 (15 aa overlapping by 12 aa, Thermo Fischer Scientific, Rockford, IL), anti CD3/CD28 (positive control), or media (negative control) in 96 well plates for 24h at 37C and 5% CO₂. T cell responses in convalescent patients served as a positive control. Plates were spun, and supernatants collected to measure IFNγ, IL-2, IL-10, TNFα, and IL-4 using a custom Human HS High Sensitivity ProcartaPlex luminex kit (Invitrogen). Undiluted samples run in duplicates on the Magpix Instrument (Luminex, Austin, TX). Data were fit using a 5P-logistic regression on xPONENT software (version 7.0c).

Single cell RNA library preparation

Cryopreserved PBMC from each patient (n=4/group for HD and Mild; n=4/time point for severe) were thawed, washed, and stained with 1 ug/test cell-hashing antibody (TotalSeq B0251,B0254, B0256, B0260, clones LNH-95, 2M2, BioLegend) for 30 minutes at 4C. Samples were washed three times in ice cold PBS supplemented with 2% FBS and sorted on the FACSAria Fusion (BD Biosciences) with Ghost Dye Red 710 (Tonbo Biosciences) for dead cell exclusion. Live cells were counted in triplicates on a TC20 Automated Cell Counter (BioRad) and pooled in groups of 4 based on disease severity and time point. Pooled cells were resuspended in ice cold PBS with 0.04% BSA in a final concentration of 1200 cells/uL. Single cell suspensions were then immediately loaded on the 10X Genomics Chromium Controller with a loading target of 17,600 cells. Libraries were generated using the v3.1 chemistry per manufacturer’s instructions with additional steps for the amplification of HTO barcodes and library preparation using Chromium Single Cell 3’ Feature Barcoding Library Kit (10X Genomics, Pleasanton CA). Libraries were sequenced on Illumina NovaSeq with a sequencing target of 30,000 reads per cell RNA library and 2,000 reads per cell HTO barcode library.

Single cell RNA-Seq data analysis

Raw reads were aligned and quantified using the Cell Ranger Single-Cell Software Suite with Feature Barcode addition (version 4.0, 10X Genomics) against the GRCh38 human reference genome using the STAR aligner. Downstream processing of aligned reads was performed using Seurat (version 4.0). Droplets with ambient RNA (cells fewer than 200 detected genes) and dying cells (cells with more than 25% total mitochondrial gene expression) were excluded during initial QC. Data objects from all groups were integrated using Seurat³⁹ using the healthy, young donor samples as the reference. Data normalization and variance stabilization was performed on the integrated object using the SCTransform function where a regularized negative binomial regression corrected for differential effects of mitochondrial gene expression levels. The HTODemux function was then used to demultiplex donors and further to identify doublets, which were then removed from the analysis. Dimension reduction was performed using RunPCA function to obtain the first 30 principal components followed by clustering using the FindClusters function in Seurat. Clusters were visualized using the UMAP algorithm as implemented by Seurat’s runUMAP function. Cell types were assigned to individual clusters using FindMarkers function with a fold change cutoff of at least 0.4 and using a known catalog of well-characterized scRNA markers for human PBMC⁴⁰. Clusters from three major cell types (T and NK cells, myeloid cells, B cells) were further subsetted from the total cells and re-clustered to identify minor subsets within those groups. List of cluster specific markers identified from this study are cataloged in Supp Table 2.

Pseudo-temporal analysis

Pseudotime trajectory of CD4 and CD8 T cells and monocytes was reconstructed using Slingshot⁴¹. The analysis was performed on each of these subsets individually with Treg and MAIT cells excluded from T cell analysis due to unique developmental origins. For each set of cells, the UMAP dimensional reduction performed in Seurat was used as the input for Slingshot. For calculation of the lineages and pseudotime in the T cell subsets, the naive clusters were set as the root state. In the monocytes, MS1 was set as the root state.

Differential expression analyses

Differential expression analysis was performed using MAST using default settings in Seurat. All disease comparisons were performed relative to healthy donors from corresponding age groups. Only statistically significant genes (Fold change cutoff ≥ 1.5; adjusted p-value ≤ 0.05) were included in downstream analysis.

Module Scoring and functional enrichment

For gene scoring analysis, we compared gene signatures and pathways from KEGG (https://www.genome.jp/kegg/pathway.html) (Supp Table 3) in subpopulations using Seurat’s AddModuleScore function. CD8 T cell specific modules were derived from previously published work⁴². NK cell exhaustion scores were calculated using aggregate expression of exhaustion markers PDCD1 (PD-1), LAG3, HAVCR2 (TIM-3), and B3GAT1. Fas-Jnk signaling scores were calculated using aggregate expression of FAS, FASLG, DAXX, MAP3K5, MAPK8, MAPK9, and MAPK10. HLA Class II module scores were calculated using aggregate expression of HLA-DM (A, B), HLA-DP (A1, B1), HLA-DO (A, B), HLA-DQ (A1, A2, B1, B2), and HLA-DR (A, B1, B5). Over representative gene ontologies were identified using 1-way, 2-way, 4-way and 8-way enrichment of differential signatures using Metascape⁴³. Functional enrichment networks were edited and annotated using Cytoscape (version 3.6.1). All plots were generated using ggplot2 and Seurat. Differences in module scores were tested for statistical significance using one-way ANOVA.

Statistical analysis

Two group comparisons were tested using unpaired t-test followed with Welch’s correction after testing for normality using Shapiro Wilk test with alpha levels set at 0.05. Group differences were tested using ordinary one-way analysis of variance (ANOVA) test followed by Holm-Sidak’s multiple comparison test with alpha levels set at 0.05. Linear regression analysis compared significant shifts in curve over horizontal line, with spearman correlation coefficient reported for each age group. Group differences with regression were tested by fitting an analysis of covariance (ANCOVA) model on each parameter with age as a binary variable and days post symptom (DPS) as a continuous covariate. Alpha levels were set at 0.1.

RESULTS

Magnitude of cytokine storm in severe COVID-19 is modulated by age

To comprehensively assess the peripheral immune response to SARS-CoV-2 infection with age, we performed a combination of immunological, single cell transcriptomic, and functional assays (Figure 1A) using blood samples obtained from 49 healthy donors (37 young and 12 aged), 25 mild/ asymptomatic COVID-19 patients (13 young and 12 aged), and 51 severely infected COVID-19 patients (35 young and 16 aged) with longitudinal sampling based on days post symptom (DPS) onset (median ~10 DPS). Each group contained young (<60 years, median age 42.5 years) and aged patients (≥60 years, median age 69) including six patients (2 young and 4 aged) who succumbed to infection (Supp. Table 1).

Figure 1:

Age-dependent alterations in circulating factors and major immune cell subsets with COVID-19.

(A) Experimental design for the study. Blood was collected from healthy donors (n=49; 37 young and 12 aged), mild/asymptomatic patients (n=25; 13 young and 12 aged) and longitudinally in 35 young and 16 aged patients with severe COVID-19. Immune phenotypes of PBMC and concentration of soluble mediators in plasma were determined using flow cytometry and luminex respectively. Longitudinal serological responses to SARS-CoV-2 were measured using ELISA. A subset of PBMC samples (n=4 HD, 4 Mild, 4 Severe per age group) were profiled using scRNA-Seq to determine dynamic peripheral immune adaptations of COVID-19 in mild/severe disease in young and aged subjects. (B) Bubble plot representing linear regression analyses of the concentration (pg/mL) of select cytokines, chemokines, and growth factors across days post symptom onset (DPS). The size of the bubble represents spearman correlation coefficient (positive in red and negative in blue) and the color intensity represents the significance of that correlation (-log₁₀(p value)) in each of the three groups. Differences with age are denoted as * (between aged survived and aged deceased) or # (between young and aged) calculated using an ANCOVA test. (C) Heatmap of proportions of major immune cell subjects derived from live singlets quantified using flow cytometry. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups where */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

We first began by measuring circulating levels of key immune mediators using luminex with a focus on factors significantly altered with severity and age. While several analytes were elevated with severe disease (Supp. Figure 1A), levels of IP-10, GM-CSF, IL-10, and TRAIL were more elevated in aged subjects with severe disease compared to young subjects. On the other hand, Granzyme B and RANTES were markedly lower in aged subjects with severe disease relative to their young counterparts (Supp. Figure 1A).

We next tracked temporal changes in plasma cytokines, chemokines, and growth factors within young patients, aged patients that survived, and aged patients who succumbed to COVID-19. Linear regression analysis revealed significant increase in T cell (IL-2, IL-7, TNFα, MIP-1β, IL-4, IL-5, IL-13, CD40L) and myeloid cell associated factors (IL-1α, IL-1β, GRO-β, CX3CL1) and growth factors (PDGF-AA, TGFα, EGF) exclusively in young subjects with severe disease (Supp. Figure 1B). To test differences with age, we used an ANCOVA on individual factors treated independently with days post symptom (DPS) as a continuous covariate. Our analysis revealed sharp increase in circulating Th2 cytokine IL-5 and type-I interferon IFNβ in young subjects relative to aged subjects with severe disease (Figure 1B). On the other hand, levels of IFNα were reduced significantly in young and older subjects who survived over time (Figure 1B). Interestingly, within the aged group, a number of myeloid cell associated growth factors (IL-3, G-CSF, GM-CSF), chemokines (MIP-3β, MCP-1, Eotaxin), and regulatory cytokines (IL-1RA and IL-10) were significantly different between older subjects who survived versus those who succumbed to infection (Figure 1B). Furthermore, levels of IFNγ were significantly positively correlated while levels of GM-CSF were significantly negatively correlated with DPS only in aged patients who succumbed to infection (Figure 1B).

Finally, we used supervised random forest modeling to identify factors that could have early (1–5 DPS) predictive value of COVID-19 outcomes. (Supp. Figures 1C, D). Independent of age, plasma levels of IL-17A, IL-33, IP-10, IL-15, MIP-1β, and IFNα were the strongest predictors of disease severity. (Supp. Figures 1C, 1D). IL-17 levels were highly predictive of mild disease regardless of age, while MIP-1β levels were more predictive of severe disease in the elderly (Supp. Figure 1D).

Peripheral immune adaptations with disease severity and age.

To assess the impact of age and COVID-19 disease severity on peripheral immune adaptations, we profiled peripheral blood mononuclear cells (PBMC) from each patient by flow cytometry. Analysis of major populations in the blood including B cells, T cells, NK cells, dendritic cells (DC), and monocytes identified significant differences in immune cell frequencies, particularly with disease severity and age (Figure 1C). Specifically, decreased percentages of naive B cells and a concomitant increase in memory B cell fractions was observed in aged patients with severe disease compared to healthy donors. Furthermore, severe disease in aged patients was associated with significant decrease in central memory CD4 (TCM) and naïve CD8 relative to their young counterparts (Figure 1C). In contrast, mild disease was associated with an expansion of CD8 TCM in both young and aged patients, albeit to a greater extent in older subjects. Frequencies of DC subsets decreased significantly in both aged and young patients regardless of disease severity, whereas monocytosis was observed only in young patients with severe COVID-19 (Figure 1C).

To further characterize the unique aspects of peripheral immune response to acute COVID-19 with age, PBMC from 16 patients (Young Mild=4; Young Severe=4; Aged Mild=4; Aged Severe=4) and 8 healthy donors (Young HD=4; Aged HD=4) were profiled by scRNA-Seq (Figure 1A). For patients with severe disease, we profiled PBMC at three time points (ranging from DPS 2–22) over the course of acute infection (Supp. Figure 2A). An average of 2,082 cells were sequenced per patient and time point for a total of 92,648 cells (Supp. Figure 2B). Principal component analysis and dimensional reduction using Uniform Manifold Approximation and Projection (UMAP) revealed distinct clustering into the major peripheral innate and adaptive immune subsets (Supp. Figure 2C). Clusters were annotated based on highly expressed marker genes using Seurat’s FindMarkers function (Supp. Figure 2D and Supp. Table 2). This orthogonal immune profiling approach recapitulated salient features of COVID-19, notably a reduction of naïve CD8 T cells (especially in aged patients) and DCs with severe disease (Supp. Figure 2E). Frequency of B cells contracted in young patients but expanded in older patients with several COVID-19, while frequency of plasmablasts increased in patients with mild disease (Supp. Figure 2E). Finally, frequency of NKT cells increased in both young and aged subjects who experiences severe disease (Supp. Figure 2E).

Qualitative differences in B cell responses to acute SARS-CoV-2 infection in the aged.

To assess the impact of disease severity and age on humoral responses to SARS-CoV-2 infection, we re-clustered the B cells and plasmablasts and identified five subsets based on expression of key B cell markers (Figure 2A and Supp. Figure 3A). Modest increases in proportions of plasmablasts were observed with disease but were more prominent in younger patients with mild disease (Supp. Figure 2E). We next measured frequencies of proliferating B cells using flow cytometry. Independent of age, frequency of total and proliferating terminally differentiated B cells was increased in patients with severe disease, while that of total and proliferating marginal zone-like (MZ) B cells was increased only in patients with mild disease (Figure 2B).

Figure 2:

Qualitative differences in SARS-CoV-2 specific B cell responses with age.

(A) UMAP projection of 6,334 B cells and plasmablasts with major subsets annotated. (B) Dot plots showing frequency of total (top) and Ki67 expressing (bottom) terminally differentiated memory B cells (left) and marginal-zone like (MZ) B cells (right) determined from flow cytometry. All percentages are reflective of fraction of total live cells. (C) Bubble plot showing expression of select genes in the plasmablast subset. The color of the dot represents the average expression, and the size represents the percentage of cells expressing that gene. (D) Bubble plot showing expression of select genes in memory B cell subsets. The color of the dot represents the average expression, and the size represents the percentage of cells expressing that gene. (E) Bar plots of spike-specific IgG End Point Titers from young and aged patients with severe COVID-19. (F) Bar plots of antibody neutralization of SARS-CoV-2 virus (IC50) measured from the plasma of young and aged patients with severe COVID-19. (G) Bar plots showing percentage of spike-positive B cells in PBMC of young and aged patients with severe COVID-19. Two group differences were tested using unpaired t-test with Welch’s correction. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups. Error bars represent medians and inter-quartile range. P-values: */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Differential expression analysis in plasmablasts (Figure 2C) within each group relative to age-matched healthy donors revealed increased expression in immunoglobulin heavy chain genes IGHG3 and IGHM only in young patients with severe disease (Figure 2C). Additionally, expression of SSR4, important for B cell effector function, was increased with infection in young patients but decreased in aged patients (Figure 2C). Similarly, expression of heavy chain gene IGHG3 within memory B cells increased with infection in young patients but decreased in aged patients (Figure 2D). Finally, while severe disease was associated with induction of genes associated with B cell activation (CD40, TRAF4, NFKB1/2 and REL), these changes were less pronounced in aged subjects (Figure 2D).

Given significant age-differences in transcriptional landscapes of memory B cells and plasmablasts following severe COVID-19, we tested for differences in humoral responses with age. SARS-CoV-2 spike -specific IgG and IgA endpoint titers were measured using ELISA and neutralizing antibodies measured using focus reduction assay. While no differences were observed in IgG or IgA endpoint titers (Figure 2E and Supp. Figure 3B), advanced age was associated with significantly reduced SARS-CoV-2 neutralization titers (Figure 2F). Moreover, frequencies of spike protein-specific B cells were significantly reduced in aged patients with severe disease compared to their young counterpart (Supp. Figure 3C and Figure 2G).

Severe COVID-19 in aged patients is associated with increased severity of lymphopenia, exacerbated IFN signaling and dampened antiviral responses.

A closer examination of T cell frequencies revealed a linear drop in total and naïve CD4 and CD8 T cells across DPS in aged patients (Figure 3A, B). Severe disease was associated with increased frequencies of proliferating CD4 TCM (Figure 3C) and CD8 TEM (Figure 3D), though this induction was reduced in the aged. In line with the significant loss of naïve T cells and reduced proliferation, plasma levels of T cell maintenance factors IL-2 and IL-7 were only significantly increased with DPS in young patients with severe disease (Figure 3D). Furthermore, surface expression of IL2R (CD25) on CD8 T cells was significantly lower in aged subjects compared to young subjects with severe disease (Figure 3E).

Figure 3:

Heightened activation and cytotoxicity of T cells from aged patients with severe COVID-19

(A,B) Linear regression analysis of frequencies of (A) total and (B) naive T cell subsets with days post symptom onset (DPS) in young and aged subjects. (C) Dot plots comparing frequencies of (left) proliferating CD4 TCM and (right) proliferating CD8 TEM with mild and severe COVID-19 in young and aged subjects measured using flow cytometry. (D) Linear regression of plasma IL-2 and IL-7 levels and days post symptoms (DPS) within young (yellow) and aged subjects (red) with severe COVID-19. P-value on the graph refers to ANCOVA testing differences with age. Spearman correlation coefficient and corresponding significant p-value for over DPS is highlighted under each graph. (E) Bar plots showing median fluorescence intensities (MFI) of CD25 expression on CD4 and CD8 T cells from young and aged patients with severe COVID-19. (F) UMAP projection of 66,656 lymphocytes re-clustered from the main UMAP to identify T and Natural Killer (NK) cell subpopulations at a higher resolution. Major subsets of CD4 T cells, CD8 T cells, and NK cells are highlighted. (G) Bubble plot of gene ontology terms derived from COVID-19 induced DEG in CD4 TEM T cells from aged and young subjects. Color and size of the bubble represents statistical significance and number of genes, respectively. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups. Error bars represent medians and inter-quartile range. P-values: */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

We next re-clustered T and NK cells to identify subsets at higher resolution (Figure 3F). This allowed us to identify several subsets of memory CD4 and CD8 T cells as well as NK cells (Figure 3F and Supp. Figure 4A). Within CD4 TEM subsets, differentially expressed genes (DEGs) with severe COVID-19 enriched predominantly to pathways associated with lymphocyte activation (CD3E, IL2RA, IL7R), response to type-I interferons (IFI6, IRF1), antiviral immunity (ISG15, ISG20, MX1, TRIM8), and positive regulation of cell death (FAS), particularly in the aged (Figure 3G). Severe COVID-19 was also associated with an expansion of the CD4 TEMRA subset, most notably in young patients (Supp. Figure 4B). Moreover, this subset was transcriptionally dysregulated with severe disease only in aged patients compared to age-matched controls and aged patients with mild disease (Figure 4A). Specifically, expression of genes associated with ATP metabolism, calcium signaling and response to hypoxia (including cytotoxic molecules such as GZMA, GZMB, and KLRD1) was decreased in aged patients with mild disease and increased in aged patients with severe disease (Figure 4A, B). Genes that were uniquely downregulated with severe disease in aged patients enriched to gene ontology (GO) terms associated with antiviral immunity, leukocyte differentiation, and negative regulation of cell cycle (Supp. Figure 4C). A consistent theme across in both CD4 TEM and TEMRA clusters was up-regulation of genes that map to GO terms associated with apoptosis in patients who experienced severe disease, as evidenced by increased scoring of FAS signaling module in memory CD4 T cells with severe COVID-19, that was more pronounced in aged patients (Supp. Figure 4D and Supp. Table 3). Furthermore, a sharper decline in IL-2 signaling signatures was detected in memory CD4 T cells from aged subjects with severe disease only (Supp. Figure 4E).

Figure 4:

Age-associated disparities in T cell activation and antigen specific T cell responses in severe COVID-19

(A) Venn diagram of DEG detected in CD4 TEMRA clusters in aged subjects with mild and severe COVID-19. (B) Bar graph showing GO enrichment of DEGs up-regulated with severe COVID-19 but down-regulated in mild COVID-19 aged subjects within the CD4 TEMRA cluster. Gene numbers associated with each term are indicated next to each term. Size of the bar indicates the significance of the enrichment. (C) Bubble plot of a subset of genes that are differentially responsive in mild versus severe COVID-19 within memory CD8 T cell clusters. Size of the bubble is indicative of percentage of cells expressing the marker and color indicates average expression level ranging from low (blue) to high (red). (D) T cells from young and aged patients with severe COVID-19 were purified by MACS positive selection beads and stimulated with SARS-CoV-2 overlapping peptides. (E) Secretion of selected immune mediators were measured by Luminex where bar graphs show pg/mL concentrations of IFNγ and IL-2 in young severe, aged severe, and convalescent patients after stimulation with peptide pool 6 or CD3 positive control. (F) Percentages of CD69+CD40L+ (top) or OX40+CD40L+ (bottom) within total CD4 (left) or CD8 (right) from young and aged patients with severe COVID-19. (G) SARS-CoV-2 specific CD4 and CD8 T cells measured as percentage of AIM+ (OX40+CD137+) CD4 T cells (top), and AIM+ (CD69+CD137+) CD8 T cells (bottom) following T cell stimulation with peptide pool 6. Two group comparisons were tested using unpaired t-test with Welch’s correction. Error bars represent medians and inter-quartile range. P-values: *=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Single cell analysis of CD8 T cells identified 4 memory subset clusters in addition to naive CD8 T cells (Figure 3F and Supp. Figure 4A). These memory clusters were identified as CD8 IFN (expressing high levels of ISGs such as IFIT2, IFIT3), activated memory CD8 (CD69^high expressing CCL4, and IFNG), and two subsets of cytotoxic CD8 T cells (GZMH^high and GZMK^high) (Figure 3F and Supp. Figure 4A). Comparisons of COVID-19 associated differentially expressed genes in memory CD8 T cells revealed up-regulation of pro-survival (BCL2) and antiviral factors (OAS3, TXNIP) with mild disease and enrichment of factors associated with the tissue resident CD8 T cell development program (RUNX3, VIM, TGFB1) in severe patients (Figure 4C). More importantly, while CD8 T cells from both young and aged severe COVID-19 patients expressed elevated levels of chemokines (CCL4L2, XCL2) and cytotoxic molecules (PRF1, GZMB, KLRD1) (Figure 4C), the overall cytotoxicity scores in GZMK^high clusters were significantly higher in aged patients (Supp. Figure 4F). As observed in the CD4 TEM cluster, severe disease was associated with signatures of heightened CD8 T cell activation (CD3D and CD8B) (Figure 4C). Finally, our analyses revealed significant up-regulation of interferon signaling module scores in all CD8 T cell clusters with mild disease in both young and aged subjects (Supp. Figure 4G), albeit to a greater extent in young patients (p< 0.0001). In contrast, IFN signaling module scores were significantly higher in aged patients relative to young patients with severe disease (p<0.0001) (Supp. Figure 4G) peaking at DPS 18 in the CD8 IFN cluster (Supp. Figure 4H).

We next investigated if these transcriptional changes were associated with altered antigen-specific T cell responses against SARS-CoV-2 (Figure 4D). To quantify antigen specific responses, purified CD2+ cells were stimulated with overlapping peptide pools covering major SARS-CoV-2 ORF (E/M/N/Nsp14/ORF3a/6/7/8/10) for 24 hours and secreted cytokine levels were assessed using Luminex (Figure 4E). T cells from aged patients with severe COVID-19 secreted attenuated levels of IFNγ (p=0.09) and IL-2 (p=0.1) relative to young patients with severe disease while response to CD3 were comparable between the two groups (Figure 4E) suggesting impaired antigen-specific responses with age. Flow cytometry analysis of the unstimulated T cells after 24 hours in culture showed decreased frequencies of CD69+CD40L+ and OX40+CD40L+ T cells in aged subjects with severe COVID-19 (Figure 4F). Furthermore, following peptide stimulation, aged subjects had fewer antigen specific CD4 (OX40+CD137+) and CD8 (CD69+CD137+) T cells (Figure 4G).

Enrichment of terminally differentiated NK cells with an exhausted phenotype in aged subjects with severe COVID-19

Given their critical role in antiviral immunity during acute infection, we next examined the impact of COVID-19 disease severity and age on NK cell phenotypes. Two major populations of NK cells were identified by scRNA-Seq based on expression of FCGR3A (CD16) and NCAM1 (CD56) (Figure 3F, Supp. Figure 5A). We also identified a cluster of cytotoxic NKs expressing high levels of PLGC2 (NKT cells) which have been shown to be drive innate immune responses against virally infected cells⁴⁴. The frequency of this subset increased significantly in young patients and only modestly in aged patients with severe disease (Supp. Figure 5B). NK cell exhaustion module scores from the NCAM1^hi cluster were increased with severe disease and with advanced age (Figure 5A). In contrast, induction of a robust cytotoxic program was only evident in young patients with severe disease as indicated by a significant increase in CD56++ GranzymeB+ cells by flow cytometry (Figure 5B). This observation is in line with increased levels of granzyme B in plasma from young, but not aged patients, with severe disease (Supp. Figure 1A). On the other hand, the frequency of NK cells expressing activation marker KLRG1 and/or terminal differentiation marker CD57 was significantly increased in aged patients with severe disease compared to their young counterparts (Figure 5C).

Figure 5:

Increased frequency of terminally differentiated NK cells with an exhausted phenotype in aged subjects.

(A) Violin plots comparing exhaustion module scores for CD56^high (NCAM^high) NK cell subset. (B) Dot plots comparing granzyme B expressing CD56++ NK cells measured using flow cytometry. (C) Scatter plots comparing frequencies of KLRG1+ NK cells expressing CD57 using flow cytometry. (D) Clustered heatmap comparing normalized expression of genes enriching to GO term “lymphocyte activation” from CD16^high NK subset obtained from patients with mild and severe COVID-19. Colors represent scaled gene expression ranging from blue (low) to red (high). Select genes are highlighted. Error bars represent medians and inter-quartile range. (E) Violin plots comparing interferon signaling module scores in CD56^high and the CD16^high NK cell subsets. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups. P-values: */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Even though frequencies of FCGR3A^hi NK cell cluster did not change with disease severity/age, we observed significant changes in their transcriptional program with COVID-19. DEG from each disease group compared to their age-matched healthy donors enriched to GO processes associated with leukocyte activation, cell death, and response to virus (Supp. Figure 5C). Interestingly, the pathways that were upregulated with severe disease in the aged patients (lymphocyte activation and leukocyte activation in immune response) were strongly downregulated with mild infection in the young patients. Indeed, genes involved in cell migration (LGALS3 and CCL5) and activation (HLA-DRB1) were highly upregulated with severe disease in aged patients (Figure 5D). On the other hand, expression of numerous IFN stimulated genes (ex. ISG20, IFIT3, IFITM1) as well as B2M, IRF7, and STAT1 were upregulated with mild disease (Figure 5D). Indeed, IFN module score was highest in NK cells from patients with mild disease (Figure 5E). However, and as described for T cells, IFN module score was higher in NK cells from aged compared to young patients with severe disease in both CD56^hi (p<0.05) and CD16^hi (p<0.0001) clusters (Figure 5E). Finally, expression of genes associated with cytokine signaling (TNFSF4, TNFRSF18, STAT3, TNF, MAP3K8, RELA) and several cytokines and chemokines and their receptors (CXCR4, IFNGR1, IL21R, IL4R) were upregulated with severe disease but to a less extent in aged patients (Figure 5D).

Aging exacerbates regulatory skewing of monocytes and DCs with severe COVID-19

Young patients showed an increased in both classical and non-classical monocytes (Supp. Figure 6A), On the other hand, frequency of classical monocytes were significantly decreased while that of nonclassical monocytes were increased in aged patients (Supp. Figure 6A). Regardless of age, frequencies of pDCs and mDCs was reduced in COVID-19 patients in a severity-dependent manner (Figure 1C and Supp. Figure 6B). The most striking change, however, was the downregulation of surface HLA-DR and activation marker CD86 on both monocytes (Figure 6A and Supp. Figure 6C) and DC subsets (Figure 6B and Supp. Figure 6C). Furthermore, HLA-DR downregulation was more severe on cells from aged patients compared to the young patients (Figures 6A and 6B) and regression analysis of HLA-DR expression over DPS revealed a recovery of its expression on classical monocytes over the course of infection in young but not in aged subjects (p<0.0001) (Figure 6C). On the other hand, expression of CD86 was relatively higher in aged subjects with severe disease compared to young subjects (Figures 6A and 6B).

Figure 6:

Myeloid cell adaptations with age and severe COVID-19.

(A-B) Scatter plots comparing median fluorescence intensities of HLA-DR and co-stimulation marker CD86 in (A) classical and non-classical (NC) monocytes and (B) myeloid DCs (mDC) and plasmacytoid DCs (pDC). (C) Linear regression analysis of HLA-DR surface expression on classical monocytes as a function of days post symptom (DPS) in young and aged subjects. Spearman correlation coefficient and corresponding significant p-value for over DPS is highlighted. (D) Violin plot comparing changes in MHC class II RNA module scores within classical and non-classical monocytes. (E) UMAP projection of 15,922 monocytes (classical and non-classical) clustered from the main UMAP. The three trajectories identified by slingshot are highlighted with markers describing each trajectory. (F) Surface expression changes in CD163 and CD62L on total monocytes from healthy donors (n=5) and patients with severe COVID-19 (n=15). Error bars represent medians and inter-quartile range. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups. P-values: */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Re-clustering of the myeloid cell subsets from the scRNA-Seq data identified four clusters of classical monocytes (MS1-MS4), intermediate and non-classical monocytes, as well as mDCs and pDCs based on the expression of canonical markers (CD14, FCGR3A, FCER1A, LILRA4) (Supp. Figure 6D,E). Age dependent differences in surface HLA-DR on monocytes were validated at the transcriptional level as MHC Class II module scores were reduced with infection in a severity and age dependent manner (p<0.0001) (Figure 6D). Trajectory analysis revealed three unique lineages within the monocyte subsets (Figure 6E) starting from the classical MS1 subset ending either in non-classical monocytes (lineage 1), antiviral cluster expressing ISGs (lineage 3), or regulatory monocytes (lineage 2) expressing CD163 and F13A1 (Figure 6E). This regulatory lineage was more prominent in patients with severe disease and was confirmed by flow cytometry with increased surface expression of CD163 and CD62L (Figure 6F). Additionally, surface expression of CX3CR1 and CD11b was downregulated and remained reduced, whereas both CCR5 and CD64 increased over time with acute infection (Supp. Figure 6F).

Severe COVID-19 paralyzes innate immune responses to secondary stimulation

Differential gene expression analysis within classical monocytes revealed significant enrichment to GO terms “response to virus” and “response to IFNγ” in all patients, that was most significant in patients with mild disease. Interestingly, genes associated with negative regulation of immune response were differentially expressed only in the aged group (Supp. Figure 7A). We next examined the impact of aged and disease severity on signaling modules within myeloid cells (Figure 7A). Within classical monocytes expression of several ISG and other viral response markers was significantly increased in both young and aged patients with mild disease (Supp. Figure 7B). Expression of alarmin molecules (S100A8, S100A9, S100A12) was increased in aged patients with severe disease while that of antigen presentation was decreased (HLA-DRA) in classical monocytes with severe disease (Supp. Figure 7B). Finally, significant changes in expression of genes associated with wound healing, myeloid cell activation, and hematopoiesis were observed in non-classical monocytes with both mild and severe disease in the aged subjects (Supp. Figure 7C). DEG analysis in the mDC subset revealed significantly increased expression of IFN-responsive genes ISG20, MX1, STAT1, and IFI44L with disease in aged patients (Supp. Figure 7D). RIGI signaling was also increased in the mDC population from the aged patients with only a modest increase in young patients (Supp. Figure 7E).

Figure 7:

Dysregulation of Toll-like receptor responses in myeloid cells from patients with severe COVID-19

(A) Violin plots comparing IFN and NF-κB signaling module scores in monocyte and DC subsets with COVID-19 severity. (B) Experimental design to measure functional responses of monocytes and myeloid DCs from patients with severe COVID-19. PBMC collected longitudinally over the course of infection from patients and healthy donors were stimulated with bacterial agonists (Pam3CSK4, LPS, and FSL-1) and viral agonists (ssRNA, Imiquimod, ODN2216) separately for 8 hours. Cytokine responses (TNFɑ, IL-6, and IFNɑ) were measured using intracellular staining followed by flow cytometry. (C) Dot plots comparing frequencies of monocytes (CD14+, top) and myeloid DCs (HLA-DR+CD11c+, bottom) producing IL-6, TNFɑ, or both, following ex vivo stimulation with bacterial (left) and viral agonists (right). (D) Scatter plots with linear regression lines tracking mean fluorescence of IFNɑ in IFNɑ+ monocytes (top) and IFNɑ+ myeloid DCs (bottom) following stimulation with bacterial (left) and viral agonists (right). Spearman correlation and p-values are highlighted in each graph. Severe young patients are shown in filled circles whereas severe aged patients are shown in empty circles. X-axis denotes days post symptom onset. Error bars represent medians and inter-quartile range. Six-way comparisons were tested using one-way ANOVA with Holm-Sidak’s multiple test comparison where *=significant compared to healthy donors, += significant compared to mild patients, #=significant difference between Aged Mild/Severe and Young Mild/Severe groups. Two way comparisons were tested using unpaired t-test with Welch’s correction. P-values: */+/#=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

We next examined changes in pro-inflammatory and anti-viral signaling modules in myeloid cell subsets with age and disease severity. While NF-κB signaling was downregulated in monocytes with both mild and severe disease, its suppression was more prominent in the aged subjects with severe disease (p<0.0001) (Figure 7A). In contrast, type-I interferon signaling scores were elevated in monocytes and DCs with both mild and severe disease, with a higher magnitude in severe aged subjects (Figure 7A). To uncover the functional consequences of the transcriptional changes, PBMC (HD n=4; severe n=15) were stimulated with either a viral or bacterial agonist cocktail and the production of IL-6, TNF⍺, and IFN⍺ by monocytes and DC were measured using flow cytometry (Figure 7B). In response to either bacterial or viral agonists, a lower percentage of both monocytes and DC in severe patients produced IL-6 and TNF⍺, regardless of age (Figure 7C). We also detected a positive correlation between IFN⍺ interferon production and DPS in severe patients (Figure 7D). These data suggest that severe COVID-19 leads to a re-wiring of the inflammatory response pathways in myeloid cells with implications for response to secondary infections.

DISCUSSION

The clinical presentation of COVID-19 is highly heterogeneous, influenced by host factors such as age, sex, BMI, and underlying medical conditions, notably diabetes, hypertension, and cardiovascular disease. Advanced age is a major driver of severe disease with individuals above 50 accounting for 95% of the deaths and exhibiting exponential increase in hospitalization and intensive care unit (ICU) admissions⁸.This increased vulnerability of older individuals to respiratory viral infections is mediated by age-associated deterioration of immune function (immunosenescence) and heightened inflammation at baseline (inflammaging)¹⁹. While recent studies have shed light on host response to mild and severe COVID-19, the precise immune correlates of the disease in at-risk populations, particularly the elderly remain unclear. To address this question, we used a systems immunology approach to profile the circulating inflammatory environment using luminex, frequency and phenotype of key immune cells using flow cytometry, transcriptional landscape using and single cell RNA-sequencing in PBMC samples from young and aged patients with mild and severe COVID-19 and compared these measurements to those obtained in age-matched healthy donors.

Shifts in plasma inflammatory environment and the ensuing cytokine storm in severe COVID-19 have been well documented^{22, 24}. Our analysis, however, revealed that older subjects with severe disease showed significantly higher induction of GM-CSF, TRAIL, IP-10, and IL-10 but lower levels of Granzyme B and RANTES compared to their younger counterparts. Lower levels of RANTES coupled with higher levels of IP-10 indicate dysregulated recruitment of T cells with potentially greater adhesion to endothelial cells^{45, 46, 47}. Furthermore, linear regression analysis revealed a failure to up-regulated IFNβ in aged individuals, but a progressive increase in IFNγ levels in aged subjects who succumbed to the disease. On the other hand, young subjects exhibited more robust increases in both T cell factors (IL-2, IL-7, TNFα, IL-4, IL-5) and myeloid factors such as IL-1α and IL-1β. Levels of IL-1RA increased with DPS in young, but not aged patients while the converse was true for GM-CSF, a myelopoietic pro-inflammatory growth factor⁴⁸. These data suggest discordant inflammatory and regulatory responses in aged subjects with severe COVID-19, particularly in non-survivors. Interestingly, lower levels of IL-17 during the first 5 DPS was associated with poor prognosis (ICU admission/intubation and death). This observation differs from recent reports that high levels of IL-17²² and Th17 cells⁴⁹ are correlated with immunopathology of severe COVID-19 and the proposed use of IL-17 blockade as a therapeutic avenue to limit acute lung injury⁵⁰. This discrepancy highlights the need for careful evaluation of the timing for administering immune modulators in the context of COVID-19.

As previously reported, we observed substantial induction of plasma B cells with mild disease^{30, 51, 52, 53}, albeit less prominent in aged individuals. Expression of JCHAIN was higher in plasmablasts from older subjects at baseline, in line with reported age-mediated increased activation/differentiation in the B cell compartment⁵⁴. Furthermore, gene signatures associated with commitment to plasma cells were more robust with mild disease. Interestingly, while mild disease was associated with accumulation and proliferation of marginal zone B cells, severe disease was associated with accumulation of terminally differentiated memory B cells, which have been linked to adverse COVID-19 outcomes⁵¹. Finally, transcriptional signatures associated with B cell activation and NF-κB signaling were dampened in aged subjects with severe COVID-19. Coupled with lower levels of plasma CD40LG in aged subjects with severe disease, these observations suggest a disruption in T follicular helper cell responses in the aged. While levels of binding antibodies did not vary with age, those of neutralization antibodies was significantly diminished in older subjects. Furthermore, older patients exhibited fewer spike-specific B cells. This observation agrees with recent studies demonstrating enrichment of ORF8 specific B cells in the elderly, which are primarily non-neutralizing⁵⁵. Collectively, these observations point at reduced plasticity of B cells with age, or their increased reliance on CD4 T cell help, which have been shown to be impaired with severe COVID-19 in the aged^{56, 57}.

T cell lymphopenia and impaired T cell responses is a hallmark of acute viral infection, including severe COVID-19^{53 25 58 59}. Our analyses revealed a significant drop in both total and naive T cells with DPS in aged patients with severe COVID-19, notably within the CD8 T cell compartment. This drop could potentially be explained by increased trafficking to the site of infection or cytokine-storm induced apoptosis as has been reported for other viral diseases³². To that end, we observe a clear up-regulation of cell death (Fas signaling) pathway in memory T cells in aged patients. Moreover, severe COVID-19 associated metabolites in blood (such as lactic acid) can suppress T cell proliferation³². Indeed, our data highlights that while relative abundance of Ki67+ proliferating T cells are reduced with age. These observations are in line with progressive decrease in plasma levels of T cell proliferation factors IL-2 and IL-7, concomitant reduction in surface CD25 expression (IL2R) on CD8 T cells, and IL-2R signaling module in the aged. There is mounting evidence of T cell dysfunction/exhaustion with severe COVID-19^{35, 49, 59, 60}. Gene expression differences in memory CD4 and CD8 T cells indicate increased T cell activation signatures with severe COVID-19 in the aged subjects. This hyperactivation of T cells could also explain the depletion of T cells, which is more significant in aged subjects. Functional exhaustion of T cells with severe COVID-19 is further evident from lack of IFNγ and IL-2 production in response to SARS-CoV-2 peptide pools containing ORFs 7 and 8. The poor antigen specific responses coupled with reduced frequencies of activation markers (CD40L, CD69) support the conclusion that advanced age is associated with aberrant T cell activation potentially mediated by increased systemic inflammation and impaired T cell immunity in severe COVID-19⁶¹.

Gene expression changes reported here also suggest that CD8 T cells from patients with severe disease exhibit increased cytotoxicity (based on increased expression of GZMB, PRF1, XCL1, XCL2) and NK cell markers (KLRD1), as previously reported using flow cytometry²⁶. We also report an increase in frequency of terminally differentiated, adaptive CD57+ NK cells³¹, which were recently shown to be indicative of poor patient survival following severe disease⁶². Functional exhaustion of NK cells have recently been described^{31, 37, 62} in the context of severe COVID-19. Our data suggests that this state is further exacerbated in aged subjects, with lower frequencies of GranzymeB+ NK cells, circulating Granzyme B levels, and reduced expression of IFNG and PRF gene in the more cytotoxic CD56++ NK cells in aged subjects with severe disease.

Our data show an increase in non-classical monocytes in both young and aged patients with severe disease; however, an increased prevalence of classical monocytes is only evident in young subjects. In contrast to monocytes, a significant drop in both myeloid and plasmacytoid dendritic cells in blood is observed with severe disease regardless of age^{30, 51}. As recently described^{28, 30, 63}, we report a drop in MHC class II molecules (protein and RNA) on both monocytes and DC subsets independent of age, but HLA-DR expression recovers faster in young subjects. This downregulation of HLA-DR in monocytes coupled with significantly lower CD86 expression suggests skewing of cells into a state of immune tolerance²⁸. Indeed, our analysis within classical monocytes suggests a preferential trajectory of classical monocytes towards a regulatory cluster expressing high levels of CD163 in patients with severe disease. Furthermore, monocytes and DCs from patients with severe COVID-19 exhibit attenuated response to TLR stimulation providing a potential explanation for increased susceptibility to secondary bacterial infections as recently described^{64, 65}. In contrast to the dampened response following stimulation with bacterial antigens, we report elevated IFN signaling module with severe disease, particularly in myeloid DCs in aged subjects. Stimulation of both monocytes and DCs with viral agonists targeting TLRs 7/8/9 resulted in higher IFN production, suggesting innate immune signaling preferentially geared towards antiviral signaling. Indeed, we detected a robust up-regulation of interferon signaling module across several cell types. Moreover, aged individuals with severe disease exhibited stronger ISG signaling compared to young subjects, potentially due to delayed viral clearance. Exaggerated innate immunity in the lung, impaired antigen presentation, and lymphopenia are hallmarks of other respiratory viral infections such as influenza and RSV in the aged¹⁸. Our analysis extends these observations and confirms the impairment of innate and adaptive immune responses in blood following severe COVID-19 in aged patients.

We acknowledge several limitations in our study design and implementation. Firstly, we analyzed immune parameters by days post symptom onset, which being self-reported can be rather inaccurate and arbitrary. Secondly, we broadly defined patients with mild disease as ones with a positive PCR test with either no symptoms associated with COVID-19 or a mild disease not requiring extensive (>3 days) hospital stay. Lack of longitudinal samples from patients from this category prevented us from modeling disease dynamics with varied severity. Thirdly, given the nature of this pandemic, there are some biases in patient and healthy donor cohorts. For example, healthy donor subjects were predominantly female (68% in young; 58% in aged) whereas a significant number of patients with severe disease were Hispanic (69% in young; 75% in aged). Young patients with severe COVID-19 had elevated BMI with respect to young healthy donors, but there was no significance compared to the severe aged patients. Additionally, patients in severe categories presented with a wide array of underlying conditions that might have played a role in disease severity/outcome. A significant number of patients were treated with Remdesivir, however, there is limited evidence suggesting its role in either immune activation or suppression in blood. Due to limited statistical power, we pooled patients with severe disease at any DPS timepoint into one category for initial analysis before regressing the data with time. Future studies will stratify young and aged patients by clinical scores to identify innate immune correlates of disease severity and identify determinants of disease resolution and survival in patients with severe COVID-19. More importantly, addressing the impact of age/infection on qualitative differences in humoral responses and long-term durability of T cell responses to SARS-CoV-2 vaccine in the aged population.

Data availability

The datasets supporting the conclusions of this article are available on NCBI’s Sequence Read Archive under project ID PRJNA727424. R code for analysis and figures in the study have been deposited in Github https://github.com/MessaoudiLab/COVID-19-Aging-Paper/.