Heterozygous transgenic humanized ACE2 (K18-hACE2) mice (strain: 2B6.Cg-Tg(K18-ACE2)2Prlmn/J) were obtained from the Jackson Laboratory (Bar Harbor, ME) and maintained in Tecniplast green line individually ventilated cages (Tecniplast, Buguggiate, Italy). Mice were maintained on a 12:12 light cycle at 30–70% humidity and provided ad libitum water and standard chow diets (LabDiet, St. Louis, MO, USA). Experimental procedures with animals were approved by the Boston University Biomedical Research, Institutional Biosafety Committee and Institutional Animal Care and Use Committee (IACUC).

The propagation of SARS-CoV-2 followed established procedures (20). The SARS-CoV-2 isolate 2019-nCoV/USA-WA1/2020 (NCBI accession number: MN985325; WA-1) was obtained from the CDC (Atlanta, GA, USA) and BEI Resources (Manassas, VA, USA). African green monkey kidney Vero E6 cells (ATCC^® CRL-1586™, American Type Culture Collection, Manassas, VA) were seeded at a concentration of 1x10⁷ cells in a T175 flask one day before virus generation. Then, the cells were infected with the virus diluted in 10 mL of Opti-MEM (ThermoFisher Scientific, Waltham, MA, USA) and incubated for 1 hour at 37°C for virus adsorption. Thereafter, 15 mL of DMEM containing 10% FBS and 1% penicillin/streptomycin was added, and the cells were incubated overnight. Subsequently, the media was removed, cells were rinsed with 1X PBS, pH 7.5 (ThermoFisher Scientific), and 25 mL of fresh DMEM containing 2% FBS was added. The cells were monitored for cytopathic effect, the media was harvested, filtered through a 0.22 μm filter, and concentrated using a sucrose gradient. The concentrated virus was suspended in sterile 1X PBS, pH 7.5, aliquoted, and stored at −80°C.

Viral stock titration was evaluated using plaque assay. Vero E6 cells were seeded into a 12-well plate at a concentration of 2x10⁵ cells/well. The next day, the cells were exposed to 10-fold serially diluted viral stock and incubated for 1 hour at 37°C. Afterwards, 1 mL of overlay media (comprising 1.2% Avicel (DuPont, Wilmington, DE, USA; RC-581) in DMEM with 2% FBS and 1% Pen/Strep) was added per well. Three days later, the overlay media was discarded, and the cells were fixed with 10% neutral buffered formalin (ThermoFisher Scientific) for 1 hour at room temperature. Subsequently, the formalin was removed, and the cells were stained with 0.1% crystal violet (Sigma-Aldrich) in 10% ethanol/water for 30 min at room temperature. The stain was rinsed off, cells were washed with water, and plaque-forming units (PFU) were counted to determine viral titers.

Male and female K18-hACE2 transgenic mice (12–15 and 72–112 weeks of age) were intranasally inoculated with 1x10⁴ (survival) or 1x10⁶ (endpoint studies) PFU of SARS-CoV-2 in 50 μL of sterile 1X PBS or sham inoculated. Inoculations were performed under 1–3% isoflurane anesthesia. Survival was assessed up to 12dpi. Mice were euthanized with ketamine/xylazine at predetermined time points for sample collection or earlier if they met euthanasia criteria (defined by an IACUC-approved clinical scoring system).

An IACUC-approved clinical scoring system was used to monitor disease progression and to establish humane endpoints of infected mice (21). The evaluated categories were body weight, general appearance, responsiveness, respiration, and neurological signs. Clinical signs and body temperature were recorded once daily for the full duration of the studies.

Microvascular lung ECs were isolated from K18-hACE2 mice infected with 1x10⁶ PFU of SARS-CoV-2 or mock-infected K18-hACE2 mice. Lungs were placed in DMEM containing 2% FBS prior to dissociation with a Miltenyi Biotec mouse lung dissociation kit (cat.# 130–095-927) following the manufacturer’s protocol. Tissues were minced using a Miltenyi GentleMACS, filtered through a 70-μm cell strainer, centrifuged at 300 x g at 4°C for 8 min, and cell pellets were suspended in MACS buffer prior to endothelial isolation.

ECs were isolated using a two-step sorting method. First, cells were negatively selected with a CD45 Microbead mouse kit (Miltenyi Biotec; cat.# 130–052-301) per manufacturer’s protocol. The collected negative fraction was washed thrice with MACS buffer prior to a positive selection for lung ECs using a CD31 Microbead mouse kit (cat.# 130–097-418) following the manufacturer’s protocol. The positive fraction was washed thrice with MACS buffer then resuspended in 500 μL of MACS buffer. All sorting protocols were performed on an AutoMACS Pro Cell separator using manufacturer’s recommended settings. After final isolation and washing, 50 μL of each sample was checked for purity by flow cytometry. The remaining 450 μL was centrifuged at 300 x g at 4°C for 8 min and the cell pellets were lysed in 600 μL QIAGEN RLT buffer containing beta-mercaptoethanol for RNA isolation (QIAGEN, Venlo, Netherlands).

Cell suspensions (50 µL) from the positive or the negative fraction were stained using surface staining protocol as previously described (22) with the PE-conjugated CD31 (PECAM-1) monoclonal antibody (clone 390, 1:100) and isotype control (clone eBR2a) (eBioscience, USA). Samples were fixed with 4% paraformaldehyde for 1 hour then washed twice with 1X PBS. Sample acquisition was performed on a LSRII instrument (BD Biosciences, USA), and data were analyzed using FlowJo v10.10.0 (FlowJo, Ashland, OR, USA).

The lungs were prepared as described before (21). Immunohistochemistry was performed using a Ventana Discovery Ultra (Roche, Basel, Switzerland). The following monoclonal antibodies were used: mouse SARS-CoV-2 Nucleocapsid protein (clone:1C7C7; Cell Signaling Technology, Danvers, MA, USA); rabbit mouse on mouse linking antibody (clone M204–3; Abcam, Waltham, MA, USA), HRP anti-rabbit IgG polymer (Vector Labs, Neward, CA, USA), developed with 3,3’-Diaminobenzidine, and counter stained with hematoxylin. Slides were imaged using a Vectra Polaris whole slide scanner (Akoya Biosciences, Marlborough, MA, USA) and analyzed utilizing the HALO™ image analysis platform (Indica labs, Albuquerque, NM, USA).

RNA was extracted using QIAGEN RNAeasy Plus micro kit. Ultra-low input RNA-seq was performed by Genewiz (Azenta US, Inc., NJ, USA). The FASTQ (150bp; paired-end) files from Genewiz were aligned to a custom combined reference (FASTQ and GTF) of mouse (GRCm39, Gencode v27) and SARS-CoV-2 (isolate Wuhan-Hu-1; NCBI accession ID: NC_045512.2) genomes using STAR aligner (v2.7.9a). The aligned reads were indexed and sorted using samtools (v1.10) and quantified using featureCounts from the subread (v1.6.2) (23) package. Further analyses were performed in RStudioServer/R (v4.1.1) framework with the DESeq2 (v1.34.0) (24) package. The principal components were visualized using pcaExplorer (v2.20.1) (25) after regularized log transformation and variance stabilizing transformation (vst) of counts. Differential expression (DE) analysis was based on Wald’s test with multiple test adjustment using Benjamini-Hochberg (BH) method in DESeq2 and shrinkage of log2-fold changes using the apeglm (v1.16.0) (26) package. The mouse DE genes (DEGs) were annotated from Ensembl (release 104) through the biomaRt (v2.50.1) (27) package; for the SARS-CoV-2 DEGs, gene symbols from the quantification step were retained. Statistical significance threshold was set at adjusted P<0.05. Venn data were estimated from the VennDetail (v1.10.0) (28) package. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment for significant DEGs was performed using clusterProfiler (v4.2.1) (29) with BH multiple testing, and pathways with adjusted P<0.05 considered statistically significant. QIAGEN Ingenuity Pathway Analysis (IPA, QIAGEN Inc., https://digitalinsights.qiagen.com/IPA) (30) was also performed with pathway significance threshold set at P<0.05 based on right-tailed Fisher’s exact test, and the enrichment was restricted to ECs through IPA’s tissue and cell filtering option. Heatmaps were constructed using the vst counts for DEGs with baseMean>50 with the ComplexHeatmap (v2.10.0) (31) package and all other figures were generated using ggplot2 (v3.3.5) (32). The heatmap visualizing z-scores for IPA enrichment was constructed using GraphPad Prism (v9.5.1; GraphPad Software, San Diego, CA, USA). Weighted gene co-expression network analysis (WGCNA) was performed using the WGCNA package (v1.72–5) (33) to identify co-regulated genes. The vst counts data were used as input after filtering out outlier genes as determined by the goodSamplesGenes function of WGCNA. The scale free topology was analyzed at different powers ranging from 10 to 30 for signed networks. However, as none of the powers reached an R² of 0.9, potentially due to biological heterogeneity between aged and young mice, we selected a soft-thresholding power of 18 as suggested by the WGCNA developers for signed networks for <20 samples (34). The network construction and module detection were performed using the blockwise approach with the parameters: maxBlockSize = 50000, TOMType = “signed”, power = 18, numericLabels = TRUE, randomSeed = 1234, saveTOMS = TRUE, minModuleSize = 30, mergeCutHeight = 0, pamRespectsDendro = FALSE (35). The correlation (Pearson) between module eigengenes and experimental groups and its statistical significance (Student asymptotic p-value) were calculated through WGCNA. A module eigengene represents the expression profiles of genes in a module (33). The experimental groups were coded in a binary format (0/1) through WGCNA’s binarizeCategoricalColumns function and included the groups aged mock, aged 2dpi, young mock, young 2dpi. In addition, correlation between eigengenes and infection status (binarized format: infected = 1, uninfected = 0) and age (aged = 1, young = 0) were also assessed. Gene ontology biological process (GO BP) enrichment of modules of biological relevance was performed using clusterProfiler with BH multiple testing, and processes with adjusted P<0.05 considered statistically significant. Genes that showed significant positive correlation with the module (correlation >0, correlation P<0.05) were used for the enrichment analysis. Correlation heatmap was visualized using pheatmap (v1.0.12) and module eigengene heatmaps were constructed using ComplexHeatmap. All other plots were generated using ggplot2. WGCNA heatmap and boxplot were based on code from (36).

Statistical analysis was performed with Prism v8 software (GraphPad). Sample sizes and number of technical and biological replicates are included in the figure legends. Data in the bar graphs represent mean ± standard error of the mean (s.e.m.). Comparison of two groups was performed based on the two-sided Student’s t test while multiple group testing was through one-way analysis of variance (ANOVA) with Tukey multiple comparison test. Statistical significance threshold was set at P<0.05.

To explore age-related infection outcomes, we investigated the pathogenicity of SARS-CoV-2 in aged (85–112 weeks) and young (12–15 weeks) mice. Due to high lethality rates at 1x10⁶ dose (21, 37), an inoculation dose of 10⁴ PFU of SARS-CoV-2 WA-1 was used for survival experiments. Survival studies revealed a significantly higher mortality rate in aged mice than young mice when challenged with SARS-CoV-2 ( Figure 1A ). Both young and aged mice lost 10% body weight between 8–10dpi and reached a nadir of 25% body weight loss between 10–12dpi ( Figure 1B ). However, no significant differences in body weight were observed. Similarly, there were no significant differences in body temperature between young and aged mice ( Figure 1C ). Although there was no difference in overall clinical score ( Figure 1D ), the aged mice displayed more ruffled fur, hunched postures, and labored breathing compared to young mice at 6dpi.

We next evaluated the histopathological changes in infected aged and young mice. Histological examination of the lung cross-sections revealed mild-to-moderate multifocal mononuclear immune cell infiltration at 2dpi in both young and aged mice ( Figure 1Ea-d ). In both age groups mononuclear inflammation was observed in peribronchiolar, perivascular, and interstitial compartments, consistent with an interstitial pneumonia. The endothelium was frequently hypertrophied in areas with perivascular mononuclear infiltrates suggestive of dysfunction. Next, we examined viral loads (N protein) in lung cross-sections at 2dpi by immunohistochemistry ( Figure 1F ). Both young and aged mice showed abundant N protein presence in alveolar epithelial cells (type I and II alveolar cells) although there were no significant differences between age groups ( Figure 1G ). Importantly, careful examination of arteries and veins of lung cross-sections revealed the absence of detectable SARS-CoV-2 N protein in ECs ( Figure 1Fc-d ).

In summary, these findings suggest that infection of lung epithelial cells is predominant, while EC infectivity of SARS-CoV-2 is an unlikely event in K18-hACE2 mice at the studied time point.

To investigate the age-dependent alterations during COVID-19, we performed bulk RNA-seq of lung ECs. ECs were MACS isolated and cell purity was evaluated by flow cytometry ( Supplementary Figures 1A–C ). RNA was isolated from highly enriched ECs for bulk RNA-seq. The principal component analysis plot illustrated clear differences between experimental groups ( Figure 2A ). SARS-CoV-2 induced significant changes in the abundance of many genes at 2dpi versus mock in aged mice (n=712; 229↓/483↑; P<0.05; Figure 2B ; Supplementary Table 1 ), young mice (n=2294; 1068↓/1226↑; P<0.05; Figure 2C ; Supplementary Table 2 ), and young versus aged mice (n=1440; 878↓/562↑; P<0.05; Figure 2D ; Supplementary Table 3 ). Interestingly, ~83% of the DE genes in young mice (n=1889; P<0.05) and ~43% in aged mice (n=307; P<0.05) were specific to each age group ( Figure 2E ; Supplementary Table 4 ). A subset of genes was dysregulated in both young and aged mice (n=422; 299↑/106↓; P<0.05; Figure 2E ; Supplementary Table 4 ). Viral transcripts represented <1–2% of all mapped reads at 2dpi.

KEGG enrichment analysis showed predominant upregulation of viral infection pathways (overlapping with Herpes simplex virus 1, influenza A, Measles, and Epstein-Barr virus infection), NOD-like receptor signaling pathway, virus sensing RIG-I-like receptor signaling pathway in both aged and young infected mice ( Figures 2F, G ). IPA enrichment analysis revealed that compared to their respective mock controls, TREM1 signaling, PI3/AKT signaling, and Role of RIG1-like Receptors in Antiviral Innate Immunity pathways were more highly activated in young mice than aged mice ( Supplementary Figure 2A ), while eNOS signaling, and coronavirus pathogenesis pathways were more repressed in aged mice than young mice ( Supplementary Figure 2A ). Similarly, IFN, Death receptor, and Pyroptosis signaling pathways were more activated in infected young mice versus aged mice ( Supplementary Figure 2B ). The heatmap of up- and downregulated genes highlights relevant genes connected to antiviral response, coagulation factors, and apoptosis signaling ( Figure 2H ). In summary, SARS-CoV-2 infection drastically dysregulated the endothelial transcriptome in young and aged mice compared to respective mock controls with young mice showing enhanced responses to infection.

Aging is accompanied by progressive biological changes in the immune system leading to a functional decline as evidenced by increased susceptibility to respiratory infections such as influenza and novel coronaviruses (38). Thus, we investigated changes in IFN signaling in lung ECs from young and aged mice at 2dpi. Young mice displayed a more pronounced IFN response compared to aged mice ( Figure 3A ). For instance, IFNβ (while only weakly expressed at the studied 2dpi time point) was upregulated in young mice compared to mock controls and aged mice, while aged mice demonstrated no difference compared with mock controls ( Figure 3B ). Stat1 (type I, II, III) and Stat2 (type I, III) are key transcription factors of IFN signaling, and are essential components of the cellular antiviral response and adaptive immunity (39). Stat1 and Stat2 expressions were ~2-fold higher in young mice compared to aged mice ( Figures 3C, D ). The IFIT gene family encodes defense proteins that are induced after viral infection or pathogen-associated molecular pattern recognition (40). ECs from young mice expressed ~20-fold higher Ifit1, Ifit3, and Isg15 compared to mock controls; aged murine ECs exhibited relatively less induction of Ifit1, Ifit3, and Isg15 than young mice ( Figures 3E–G ). Similarly, Ifitm3 induction was lower in old mice (~2-fold) during infection relative to young mice (~5-fold) ( Figure 3H ). A schematic diagram of the signaling pathways based on the expression profile in young and aged mice at 2dpi is presented in Figure 3I .

Pathway analysis revealed that Triggering Receptor Expressed on Myeloid cells-1 (TREM1) signaling was highly activated in young mice compared to aged mice ( Supplementary Figure 2A ). TREM1 is an immunoreceptor expressed on neutrophils, monocytes/macrophages, and ECs (41). It amplifies the inflammatory response driven by Toll-Like Receptor (TLR) engagement. Several TREM1-associated signaling molecules, receptors, and transcriptional factors (Tlr2, Tlr3, Cd40, Myd88, Jak2, Stat3, Akt3, Icam1) were significantly more upregulated in young mice than in aged mice ( Figure 3J ). Induction of TREM1-mediated chemokine ligand 2, Ccl2/MCP1 (42), was profoundly abrogated in aged mice relative to young mice ( Figure 3K ). Interestingly, the negative regulator, Sigirr (single immunoglobulin IL-1R-related molecule) (43), was significantly downregulated in young mice with comparatively less downregulation in aged mice versus the respective mock controls ( Figure 3L ). The schematic diagram shows the signaling pathways based on the expression profile in young and aged mice at 2dpi ( Figure 3M ).

The pathway analyses above highlight key inflammatory responses that are activated by the regulated genes in response to infection. However, biological responses are typically not driven by a single gene but a set of genes that are regulated together and share similar expression patterns. To identify whether coregulated genes in aged and young mice define specific responses during infection with SARS-CoV-2, we performed WGCNA (33, 35). In total, we identified 87 modules of co-expressed genes (module eigengenes). The correlation between the modules and experimental groups is presented in Supplementary Figure 3 . Infection status was significantly correlated with module ME1 ( Supplementary Figure 3 ). Genes in ME1 were highly induced in young mice 2dpi, while aged 2dpi mice showed diminished induction ( Supplementary Figures 4A, B ). Genes positively correlated with ME1 were enriched in host defense processes such as interferon response and cytokine production ( Supplementary Figure 4C ). ME1 genes included chemokines (e.g., Cxcl9, Cxcl10, Cxcl11), interferon-induced genes (Isg15, Rsad2) and genes involved in viral sensing (Oasl1) that were highly induced in young mice 2dpi compared to young mock controls. ME1 genes were also induced in infected aged mice, albeit at a lower level than infected young mice, and included immune-related genes such as Ifit1, Rsad2 and Oasl1 ( Supplementary Table 5 ). The complete gene-module and gene-group correlation data is presented in Supplementary Table 6 .

Overall, these findings suggest that senescent lung ECs mount an impaired immune response at 2dpi in humanized ACE2 mice, suggesting delayed pathogen defense mechanisms that might contribute to delayed viral clearance and reactive hyperinflammation, which in turn lead to high mortality.

As the aged mice demonstrated defective IFN response ( Figure 3A ), we examined age-dependent expression changes in cytokine, chemokine, and acute phase response signaling in lung ECs during SARS-CoV-2 infection at 2dpi. Young mice exhibited a more prominent cytokine/chemokine response in lung ECs than aged mice ( Figure 4A ). Consistent with the IFN response genes, IFN-regulated chemokines (44), such as Ccl2 (~10-fold), Cxcl9 (~50-fold), Cxcl10 (~180-fold), and Cxcl11 (~400-fold), were significantly upregulated in young mice compared to aged mice ( Figures 3K and 4B–D ). Notably, aged mice ECs manifested significantly higher Cxcl1 expression (~20-fold) compared with young mice ( Figure 4E ).

Acute-phase response genes are directly activated as a part of the adaptive stress response during infection and various other disease states. Acute-phase protein concentrations are altered with disease severity in COVID-19 patients, suggesting their potential utility in diagnosis and treatment (45). In our study, lung ECs after SARS-CoV-2 infection of young mice showed an increased activation of acute-phase response signaling in comparison to aged mice ( Figure 4F ). Various transcriptional factors were significantly upregulated (e.g., Ralb, Relb, and Nfkb2) or downregulated (e.g., Elp1, Map2k6) in young mice relative to old mice. Complement genes like C1ra, C2 and C4b were especially highly induced in young mice. Aging was associated with a diminished capacity to induce acute-phase genes like PAI-I (Serpine1) and C1-inh (Serping1) as well as transcriptional factors such as Map2k1 and Stat3 ( Figures 4G–J ).

In conclusion, aged lung ECs exhibited impaired cytokine/chemokine expression and inadequate acute phase response signaling in humanized ACE2 mice at 2dpi.