Delayed viral clearance and altered inflammatory responses affect severity of SARS-CoV-2 infection in aged mice

Abstract

Epidemiological investigations consistently demonstrate an overrepresentation of the elderly in COVID-19 hospitalizations and fatalities, making the advanced age as a major predictor of disease severity. Despite this, a comprehensive understanding of the cellular and molecular mechanisms explaining how old age represents a major risk factor remain elusive. To investigate this, we compared SARS-CoV-2 infection outcomes in young adults (2 months) and geriatric (15–22 months) mice. Both groups of K18-ACE2 mice were intranasally infected with 500 TCID₅₀ of SARS-CoV-2 Delta variant with analyses performed on days 3, 5, and 7 post-infection (DPI). Analyses included pulmonary cytokines, lung RNA-seq, viral loads, lipidomic profiles, and histological assessments, with a concurrent evaluation of the percentage of mice reaching humane endpoints. The findings unveiled notable differences, with aged mice exhibiting impaired viral clearance, reduced survival, and failure to recover weight loss due to infection. RNA-seq data suggested greater lung damage and reduced respiratory function in infected aged mice. Additionally, elderly-infected mice exhibited a deficient antiviral response characterized by reduced Th1-associated mediators (IFNγ, CCL2, CCL3, CXCL9) and diminished number of macrophages, NK cells, and T cells. Furthermore, mass-spectrometry analysis of the lung lipidome indicated altered expression of several lipids with immunomodulatory and pro-resolution effects in aged mice such as Resolvin, HOTrEs, and NeuroP, but also DiHOMEs-related ARDS. These findings indicate that aging affects antiviral immunity, leading to prolonged infection, greater lung damage, and poorer clinical outcomes. This underscores the potential efficacy of immunomodulatory treatments for elderly subjects experiencing symptoms of severe COVID-19.

Introduction

In May 2023, the World Health Organization (WHO) declared that COVID-19 is no longer a Public Health Emergency of International Concern (PHEIC) as the disease switched from pandemic to endemic. WHO data indicate that SARS-CoV-2 was responsible for approximately 1000–4000 monthly deaths between Jully 2023 and January 2025 despite the broad vaccination campaigns. The diseases associated with SARS-CoV-2 infection therefore remain an important medical concern.

The etiologic agent causing COVID-19 is SARS-CoV-2 belonging to the Betacoronaviruses genus [1, 2]. The SARS-CoV-2 use the Angiotensin-Converting Enzyme 2 (ACE2) as a cellular receptor [3]. The presence of transmembrane serine protease 2 (TMPRSS-2), cathepsin L (CTSL), and Furin proteases that cleave the virus’ spike protein have been shown to be important for the entry in the host cell [3,4,5]. Induction of cellular death via the pyroptosis pathway by the virus leads to the release of damage-associated molecular patterns (DAMPs) and pathogen-associated molecular patterns (PAMPs)[6, 7]. Recognition of those molecules via pattern recognition receptors (PRRs) such as retinoic acid-inducible gene I -like receptors (RLRs), Toll-like receptors (TLRs), nucleotide-binding oligomerization domain-like receptors -like receptors (NLRs), or Absent In Melanoma 2-like receptors (ALRs) leads to the production of several pro-inflammatory and antiviral mediators [8,9,10,11,12,13,14].

The activation of inflammatory and antiviral pathways through PRRs leads to the production of several cytokines, chemokines, and interferons (IFNs) as well as bioactive lipids [15,16,17]. These mediators contribute to the early leucocyte recruitment involving monocytes/macrophages, dendritic cells, and Natural Killer (NK) cells which fuel the inflammatory environment and contribute to manage the viral infection [18, 19]. This environment also favors the build-up of an efficient T cell response ensuring viral clearance and the resolution of inflammation. On the other hand, inability to control the inflammatory responses despite the elimination of the virus can lead to the development of severe COVID-19 [20].

Rapidly after the COVID-19 pandemic declaration by the World Health Organization, it became clear that not everyone was equally susceptible to the disease. The main risk factor identified was old age. Several epidemiologic studies revealed that the COVID-19 fatality rates increase with age going from 1–3% for patients below 50 years to 10–15% for 70–79 years and 15–25% for the above 80 years [21, 22]. Patients over 60–65 years of age represent about 70–80% of COVID-19 hospitalizations and deaths. A need for a more precise understanding of the disease course in the elderly population is required.

In the current study, we compared the susceptibility of young adults and elderly K18-ACE2 mice cohorts to pathogenic SARS-CoV-2 infection. Our results show that aged mice were more severely affected by COVID-19 than younger mice. Old mice experienced a prolonged viral persistence with deep profiling of the mouse lung lipidome revealing a dysregulation in the timing and balance of the pro-inflammatory and immune-modulatory signals. In line with this finding, older mice exhibited impaired Th1-like chemokines and type II IFN response and fewer antiviral leucocytes in response to infection. In summary, our results suggest that the age-related severe COVID-19 was caused by a weaker early immune response that led to delayed viral clearance and increased lung inflammation and morbidity.

Materials and methods

Cell culture and virus

Vero cells were purchased from the American Type Culture Collection (Manassas, VA, USA). This cell line was passaged twice a week and cultured in Medium 199 (Multicell Wisent Inc.) supplemented with 10% fetal bovine serum (FBS) (Corning Cellgro, Manassas, VA, USA), 10mM HEPES pH 7.2, 1mM of sodium pyruvate (Corning Cellgro) and 5μg/mL of Plasmocin® (Invivogen, San Diego, CA, USA) to prevent mycoplasma contamination. Cells were grown at 37°C with 5% CO2. The SARS-CoV-2 Delta strain was obtained from the BC Centre for Disease Control ([BCCDC] Vancouver, BC, Canada). SARS-CoV-2 was propagated on Vero cells. The infectious titer of viral preparations was 1.04 × 10⁶ Tissue Culture Infectious Dose 50_/mL (TCID₅₀/mL). Viral titration was performed as described [23]. Experiments involving infectious SARS-CoV-2 viruses were performed under biosafety level 3 conditions.

Mice

B6.Cg-Tg(K18-hACE2)2Prlmn/J mice were purchased from the Jackson Laboratories (Bar Harbor, ME, USA) and then bred in a colony at the Centre de Recherche du Centre Hospitalier Universitaire de Québec-Université Laval animal facility. Male and female mice with a mean age of seven weeks or eighty weeks were infected with 25µL of M-199 media containing 5 × 10² (TCID₅₀) of SARS-CoV-2 or 25µL of M-199 media for mock-infected mice. Mouse weight was recorded every day until euthanasia. Mice were sacrificed on days 3, 5, and 7 post-infections. A small lobe (approximatively 4mg) of the right lung was used for RNA extraction, the whole left lung for histological analysis and remaining right lung lobes (approximatively 80mg) for tissue homogenisation. Briefly, lung tissue was homoginized using the Bead Ruptor Bead Mill homogenizer (Omni, Kennesaw, GA, USA) then centrifuge at 3000xg for 20min at 4°C. Homogenate supernatant was use for cytokine analysis, infectious titer (TCID50_/mL) analysis and lipidomic analysis.

Reverse transcription digital PCR and quantitative real-time PCR analysis

RNA from mouse lungs was extracted using the Bead Mill Tissue RNA Purification Kit and the Bead Ruptor Bead Mill homogenizer (Omni). Following extraction, residual DNA was removed by treating the samples with DNAse I (Roche, Mississauga, ON, Canada). RNA was reverse transcribed to cDNA using SuperScript™ IV VILO™ master mix (ThermoFisher Scientific, Waltham, MA, USA)). SARS-CoV-2 viral RNA loads were determined using Droplet Digital PCR (ddPCR) supermix for probes without dUTP (Bio-Rad Laboratories Ltd, Hercules, CA, USA) and the QX200 Droplet Digital PCR System Workflow (Bio-Rad Laboratories Ltd). Quantitative real-time PCRs (qPCR) were performed on cDNA produced for RT-ddPCR analysis using the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories Ltd) on the Rotor-Gene Q 5plex (Qiagen). Primers and probes are listed in our previous work[23].

Histological analysis

Mouse lungs were fixed in formalin and embedded in parafilm blocks following standard method [24]. Once sliced, the lung section was deparaffined and rehydrated. Lung inflammation was evaluated on sections stained using the Carstair method (Electron Microscopy Sciences, Hatfield, PA, USA). Stained sections were digitalized using an AxioScan Z1 (Carl Zeiss, Oberkochen, Germany) at 10X magnitude. Each lung image was split into six sub-sections where inflammation in the perivascular, peribronchial, and parenchymal areas was scored. Scores of 1 to 5 were attributed for each area by sub-section and the mean score from the sub-sections presented.

Histological immunofluorescence

Lung sections were processed as described above. Following deparaffination and rehydration, antigen retrieval was performed on tissue following the protocol described in our previous work [25]. Then sections were permeabilized with TBS with 0.025% Triton X100 twice times 5 min and blocked with TBS containing 1% BSA and 5–10% goat serum (Multicell Wisent Inc.) for two hours. Tissues were labeled with 10 µg/mL of rat anti-mouse-GR1 clone RB6-8C5 (Thermo Fisher Scientific), 5 µg/mL of rat anti-mouse-CD4 clone 4SM95 (ThermoFisher Scientific), 1.14 µg/mL of rabbit anti-mouseCD8a clone EPR20305 (Abcam Inc., Toronto, ON, Canada), 1 µg/mL of rabbit anti-mouse-NK1.1 clone 39,197 (Cell Signaling Technology, Danvers, MA, USA), 1.5 µg/mL of rat anti-mouse-F4/80 clone BM8 (BioLegend, San Diego, CA, USA), 25 µg/mL of rabbit anti-SARS-CoV-2-N (Rockland Immunochemicals Inc., Limerick, PA, USA) in TBS with 1% BSA and 2.5–5% goat serum for 16 h at 4°C. Primary antibodies were stained with eider 4 μg/mL of goat anti-rabbit-Alexa-488 (Thermo Fisher Scientific), 1.2 µg/mL of goat anti-rabbit-Alexa-647 (Jackson ImmunoResearch Labs, Baltimore Pike, PA, USA), or 0.7 µg/mL goat anti-rat-Alexa-647 + (Thermo Fisher Scientific) for 1 h at room temperature in TBS with 1% BSA and 2.5–5% goat serum. Sections were incubated into TBS with 1.67 μg/mL of DAPI (Thermo Fisher Scientific) then washed for 5 min in TBS and mounted with ProLong Glass Antifade reagent (Thermo Fisher Scientific). Immunostained sections were scanned using an AxioScan Z1 (Carl Zeiss) at 20X magnitude. Images were exported from Zen Blue (Carl Zeiss) to 8-bit TIFF. CD4, CD8 and GR1 stained cells were quantified using macro generated using Celeste Image Analysis Software (Thermo Fisher Scientific) and SARS-CoV-2 N positive areas were quantified using Fiji [26] and then normalized with the tissue area analyzed. Cell densities for NK1.1 and F4/80 staining were computed with the help of the high throughput image analysis platform at CRCHUQ-UL. An algorithm processed AxioScan.Z1 raw images using Mathworks Matlab 2018b software to detect, measure and sort each cell according to experiment factors and tissue areas.

Cytokines, chemokines, and interferon quantification

The protein inflammatory mediators were quantified using ProcartaPlexTM Mouse Mix & Match Panels kit (Thermo Fisher Scientific) in lung homogenates.

Lipidomic analysis

Mouse lungs were collected and homogenized as described above. Lung homogenates were heated at 65°C for 30 min to inactivate infectious viruses before being taken out of the BSL3 facility. Oxylipin extractions and analysis were carried out as described elsewhere [27]. Briefly, 10 µL of a deuterated standard was combined with 200 µL of lung homogenates prior to freeze-drying. A volume of 289 µL of an ethanolic solution (41%) was then added and mixed thoroughly to reconstitute the samples. The specimens were incubated with acetonitrile (2:1) at room temperature and then at −20°C to precipitate the proteins. After centrifugation, supernatants were removed and combined with 600 µL of an ammonium hydroxide solution (0.01 M) before being loaded on solid phase extraction columns. A portion of the unwanted compounds were washed from the cartridges with the solution of ammonium hydroxide and then a mixture of acetonitrile and methanol (8:2). Compounds were eluted using an acidified version of the latter organic solution, nitrogen dried and reconstituted in 60uL (30% acetonitrile with 0.01% acetic acid). To enhance the precision of the quantification, a spiked calibration curve was generated by using 10 µL of each final sample to form a pool, which was employed to reconstitute pure standards extracted with the same steps as the samples. Analysis was performed with a High-performance liquid chromatography-tandem mass spectrometry system operated in negative mode and using specific transitions in scheduled multiple reaction monitoring (Supplementary Table 1) previously described [28]. The statistical analyses were performed using lipid concentrations normalized to total protein quantification. Then data were imported into MetaboAnalysis 6.0 online tools where they were log10 transformed and range scaled before any statistical analysis. The One-factor statistical analysis module was used to compare each mouse group for each time point using an unpaired t-test assuming an equal group variance. The metadata table statistical analysis module with covariates design was used to highlight the interactions between age or infection and expression of different lipids. A Two-way ANOVA was used to highlight the interaction between lipid expression and covariates (Age or infection time point). Linear models with adjustment for infection covariate were used to characterize the relation between age and lipid expression. Bubble plot was generated using ggplot2 3.5.1 R package [29].

RNAseq analysis

Library preparation and sequencing runs were performed by sequencing core facility at the CHU de Quebec Research Center-Université Laval. Briefly, ribosomal RNA-depleted RNA from lungs of mock and infected-mice (5DPI) was used to prepare sequencing libraries. The library was sequenced using the the NovaSeq 6000 apparatus (Illumina, San Diego, CA, USA) that generated 20–25 million reads per sample. Data preprocessing was performed on the Digital Research Alliance of Canada Narval server (Montréal, QC, Canada). Sequenced reads were trimmed using fastp 0.24.0[30] then relative gene expression was evaluated using kallisto 0.46.1[31]. Principal components analysis and graphs were made using FactoMineR 2.8 and factoextra 1.0.7 R packages [32]. Due to high divergence of one mouse from young mock and one mouse from infected old mouse group within their group, these were removed from further transcriptomic analysis (Supplementary Fig. 1). Gene counts were exported on RStudio using tximport 1.22.0 R package[33] and statistical analyses were performed with DESeq2 1.34.0 R package[34]. Gene set enrichment analyses were performed using the msigbr 2022.1.1 and clusterProfiler 4.2.2 R packages [35,36,37] using Hallmark gene set collection[38] and Reactome Pathway knowledgebase[39]. ChatGPT 4 (OpenAI, San Francisco, CA, USA) was used to build and improve the computational analysis pipeline. Volcano plots and GSEA Dot plots were generated using ggplot2 3.5.1 and enrichplot 1.14.2 R packages [29, 37].

Statistical analysis and graphs

R packages were run on R 4.1.2 using RStudio 2023.09.1 + 494 and other statistical analysis and graphs were performed using Prism 10.4.1 (Boston, MA, USA).

Results:

a. COVID-19 course in K18-hACE2 elderly and young adult mice

As shown in Fig. 1A, mice with a mean age of 1.7 months (young group) or 18.5 months (old group) were infected intranasally with 500 TCID₅₀ of SARS-CoV-2 and monitored for weight loss and overall appearance/behavior for 14 days. Mice that reached ethical checkpoint limits (> 20% weight loss plus one other sign of discomfort) were euthanized. hACE2 transgene expression was evaluated using RNAseq and showed no difference between age groups nor between mock and infected mice (Fig. 1B). As shown in Fig. 1C, infected young mice experienced greater weight losses, especially on days 6–7 post-infection (DPI) relative to aged mice. However, while young mice started to regain weight on day 7, older mice continued losing weight, reaching endpoint limits. As a result, none of the aged mice survived the infection beyond day 10 while 30% of young mice did (Fig. 1D). Likely due to the insufficient number of animals used (n = 10/group), no statistical differences in survival between the two groups were observed.

Fig. 1

SARS-CoV-2 pathogenesis and viral replication in aged and younger K18-hACE2 mice. A Mice age distribution across young and old groups in months. Error bars display minimum and maximum age within groups, while bars in the boxplots represent the median age. B hACE2 transgene transcript expression in mock and 5 DPI. Value represent Transcript per kilobase million (TPM) calculated by Kallisto Quant (Mean ± SD, n = 4–5/group). Mixed-effect analysis was used to determine the impact or age and infection time-point on transcript expression. C Mouse weight variations after infection. Values represent mean ± SD expressed in percent weight relative to day 0 (n = 30 at Day 0 for infected groups and 5 for mock groups). The weight curves of both SARS-CoV-2 infected groups were compared using Two-way ANOVA. Uncorrected Fisher's LSD tests were used to evaluate weight loss recovery by comparing weight at 7 to 9 DPI with the weights from 6 DPI that correspond to the weight loss peak for young mice. D Mice survival curves following infection with SARS-CoV-2. Mice were sacrificed when reaching the ethical checkpoints and those that reached the 14 DPI were considered as survivors. The difference in the survival was assay using a Log-rank (Mantel-Cox) test (n = 10 at Day 0). E Lung homogenate infectious viral loads at 3 DPI to 7 DPI expressed in Tissue Culture Infectious Dose 50 by mg of lung protein (TCID_50/mg) (Mean ± SD, n = 3–5/group) and LOD line show the limit of detection of the assay. F Viral RNA quantitation in the lungs using RT-ddPCR expressed in copies of SARS-CoV-2 E gene relative to copies of mRPP30 mRNA housekeeping gene (Mean ± SD, n = 3–5/group). G SARS-CoV-2 N antigen quantification on lung section. Staining scores represent the proportion of tissue area with a positive signal (Mean ± SD, n = 3–5/group). For D to F, comparisons between groups were performed using an Unpaired T-test or Mann Whitney according to the data distribution. P value: < 0.05 (*), < 0.0021 (**), < 0.0002(***), non-significant (ns)

b. Deficient lung viral clearance in elderly mice

Following virus inoculation, up to five mice per group were sacrificed on day 3, 5, and 7 post-infection and their lungs harvested for viral loads determination. Infectious virus titers are displayed in Fig. 1E. No significant viral load differences were detected between groups during the acute phase of infection (3–5 DPI). However, a major difference was observed on day 7. While all young mice had viral loads that were below the detection limits of the assay, none of the elderly mice efficiently cleared the virus. Reduced viral clearance by aged mice was confirmed using viral RNA (Fig. 1F) and N antigen (Fig. 1G) measurements.

Young mice produce chemokines in greater quantities at early times post infection relatively to aged mice

From the mouse lung homogenates at 3, 5 and 7 DPI, several inflammatory mediators were measured. As shown in Fig. 2A, the total inflammatory burden was significantly higher at 3–5 DPI in young mice relative to aged mice. On day 7, aged mice produced higher levels of cytokines in response to infection. This result should be interpreted with caution as one of the three remaining mice over produced IL-6 (Supplementary Fig. 2). When analyzed separately, IFNγ, CCL2 and CXCL9 were produced at greater levels by young mice at early time points (3 and 5 DPI) (Fig. 2B-E). In response to SARS-CoV-2 infection, CCL3 was produced at significantly higher levels in aged mice on day 7 post-infection relative to young mice (Fig. 2F).

Fig. 2

Chemokines, cytokines, and interferons (IFN) production in the lung following SARS-CoV-2 infection. Mediators were quantified from lung homogenates using Luminex technology (ProcartaPlex kit). A Total protein mediator burst. The subsections of each bar graph represent the mean concentration value for one mediator expressed in pg of mediator by mg of lung protein. Groups were compared using an Uncorrected Fisher's LSD on the mean concentration by bare. IFNγ (B), CXCL1 (C), CXCL9 (D), CCL2 (E), and CCL3 (F) production expressed in pg of mediator by mg of lung protein (Mean ± SD, n = 3–5/group). For each groups infection time points were compared with an Uncorrected Fisher's LSD or Uncorrected Dunn’s test with their mock condition according to the data distribution. Time points between groups were compared with the Unpaired T-test or Mann Whitney test according to the data distribution. P value: < 0.05 (*), < 0.0021 (**), < 0.0002(***) and < 0.00001(****)

No significant differences in the induction of proinflammatory cytokines and Type I-III IFN production or ISG expression

IL-6 production was measured in lung homogenates of the different mouse groups, but no significative differences were observed (Supplementary Fig. 2A). Similarly, CCL4, CXCL10, IFNα(2,4), IFNβ were induced by the infection without being significantly differentially expressed between groups (Supplementary Fig. 2 B to E). No increase in Il-1β gene expression was noted (Fig. 3A). Il18, Ifnl2-3 and ISGs (IRF7, Isg15 and Isg56) genes expression was also induced by infection but did not differ between young and old mice (Fig. 3 B to F). In contrast, Tnf gene expression was reduced in aged mice at 3DPI relative to young mice (Fig. 3G).

Fig. 3

Inflammatory and antiviral genes expression levels in the lungs following SARS-CoV-2 infection. Gene expression was quantified using RT-qPCR from lung total RNA. Il1b(A), Il18(B), Ifnl2-3(C), Irf7(D), Isg15(E), Isg56(F), Tnf(G) gene expression were reported as fold of change in comparison with the respective mice group mock controls (Mean ± SD, n = 3–5/group). For each group infection time points were compared with an Uncorrected Fisher's LSD or Uncorrected Dunn’s test with their mock condition according to the data distribution. Time points between groups were compared with the Unpaired T-test or Mann Whitney test. P value: < 0.05 (*), < 0.0021 (**) and < 0.0002(***)

Relative to young mice, elderly mice show reduced signs of inflammation at early time points after infection

Using the lung section stained using the Carstair method, the lung was evaluated for perivascular, peribronchial, and parenchymal signs of inflammation. As shown in Fig. 4A, when the whole lung inflammation scores were tabulated, increased inflammation was observed in the lungs of young mice at 3 DPI and 5 DPI in comparison with aged mice. Each lung section of old mice at 3 DPI evaluated showed lower inflammation, but on day 5 this was restricted to the peribranchial and parenchymal zones (Fig. 4 B to D).

Fig. 4

Lung histological inflammation in K18-mice in response against SARS-CoV-2 infection. Carstair stained sections were scanned and scored for inflammation (0 to 5) in the three different lung regions (perivascular, peribronchial, and parenchymal). Global inflammation score (A) (Mean, n = 3–5/group). Each subsection of the stacked bar graph represented the respective score for each lung region. Perivascular (B), peribronchial (C), and parenchymal (D) individual inflammation scores scaled from 0 to 5 (Mean ± SD, n = 3–5/group). Mouse groups and infection time points were compared using Uncorrected Fisher's LSD. P value: < 0.05 (*), < 0.0021 (**), < 0.0002(***) and < 0.00001(****). Image representative of a Carstair stained lung section for each mice group was shown in (E)

Younger mice exhibit higher lung leucocyte infiltration relative to old mice in response to SARS-CoV-2 infection

Lung sections were also analyzed for leukocyte subpopulations following infection. CD4 marker was used to quantify T helper cells, CD8α was used to quantify cytotoxic T cells, NK1.1 marker was used to label NK cells, GR1 antigen (anti-Ly-6C/Ly-6G) was used as markers of granulocytes (Ly-6G) and monocytes (Ly-6C) and F4/80 marker was used to identify macrophages. As shown in Fig. 5A, the number of CD4 cells in mock infected animals was higher in young versus aged mice. Upon infection, a reduction in the number of CD4-positive cells in the lungs of young mice but not in older mice was observed. Similar to CD4 + T cells, young mice had higher absolute numbers of CD8 + T cells than aged mice, although only statistically significant at 5 DPI (Fig. 5B). In contrast, mock-infected aged mice had lower numbers of resident CD8 + T cells than young mice. Infection led to a significant increase in CD8 + T cell infiltration in young mice but not in old mice. For NK cells, lower counts were observed 5 DPI aged mice than 5DPI young mice and a significant increase in NK cells was observed between mock and young mice at 7 DPI (Fig. 5C). Alike NK cells, significant recruitment of macrophages was detected at 7 DPI in younger mice only (Fig. 5D). No differences in NK and macrophage cell counts were observed at the 3DPI time point (data not shown). Noticeably, macrophage numbers in each group were slightly lower in elderly mice, although only statistically significant observed at 5 DPI. Regarding granulocytes/monocytes, relative to mock-infected mice, an increase of GR1 positive cell counts was observed in aged mice at 7 DPI but not in younger mice (Fig. 5E). Mixed-effect analysis highlight significant effects of age and infection time points on CD4 and F4/80 counts, age only for NK1.1 and infection time points only for CD8 and GR1 count.

Fig. 5

Leucocyte recruitment in mouse lungs following SARS-CoV-2 infection. Lung sections were stained with antibodies against mouse CD4 (A), CD8a(B), NK1.1(C), F4/80 (D) or GR1(E). Stained tissues were scanned, and positive cells were quantified on the whole section. Results were expressed as number of positive cells by mm.² of tissue section (Mean ± SD, n = 3–5/group). Images are representative of each immunostaining and shown beside each graph. For each group, infection time points were compared with an Uncorrected Fisher's LSD or Uncorrected Dunn’s test with their mock condition according to the data distribution. Time points between groups were compared with the Unpaired T-test or Mann Whitney test. Effects of the age and infection time point factors were evaluated using a mixed-effect model. P value: > 0.05 (ns), < 0.05 (*), < 0.0021 (**) and < 0.0002(***)

Aged mice showed an altered bioactive lipid basal profile and differential regulation than young mice in response to SARS-CoV-2 infection

We have extracted lipids from the mouse lungs to measure the lipid mediators of inflammation (LMI) produced in response to infection. The lipidome was quantified using a panel containing a hundred lipids (listed in Supplementary Table 1). Those lipids are produced from arachidonic acid (AA), eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA), linoleic Acid (LA) or α-linolenic acid (ALA) via cyclooxygenases (COX), lipoxygenase (LOX), cytochrome P450 (CYP) or non-enzymatic reactions (Fig. 6). Of quantified lipids, approximately 40 differed in elderly mice comparatively to the juveniles in at least one infection time-point and there relative levels were presented in a bubble plot (Fig. 7A). When comparing the elderly and juvenile mock groups, ± 8-/11-HEPE, ± 8(9)DiHETE, 8(9)EpETE, 9(S)-HOTrE, prostaglandin E2 (PGE₂), 8-iso-PGE2 and thromboxane B2 (TBX₂) were detected at significantly higher levels in the elderly than young mice. On day 3 post-infection, ± 11-/15-HEPE, and ± 14(15)-DiHETE derived from EPA were observed at significantly higher concentrations in older than young animals. Concomitantly LA/ALA derivatives like 13S-/9S-HOTrE or DIHOMEs and, DHA derivatives like resolvin D3 (RvD3), protectin Dx (PDX), ± 16(17)-DiHDPA, 4-series neuroprostanes (4-F_4t-NeuroP), 5-series isoprostanes (5-(RS)−5-F_2c-IsoP) were also more prevalent in older than young animals. In opposition to other time points, at day 5 post-infection a cluster of decreased lipids concentration was observed in aged mice. This cluster contains lipids belonging to the HETEs, HEPEs, DiHDPAs, EpDPAs, HDPAs, and OxoETEs together with RvD2 and PGF_3α For the day 7 post-infection due to the small sample size and high variability, only 17(18)-DiHETE and 8(S)-HETE levels were differentially changed between the two mouse groups.

Fig. 6

Production pathways of differentially expressed lipids (DELs) between elderly and young mice infected of not (mock) with SARS-CoV-2. These lipids are derived from Arachidonic Acid (AA), Eicosapentaenoic acid (EPA), Docosahexaenoic acid (DHA), Docosapentaenoic acid (DPA), Linoleic Acid (LA) alpha-Linolenic acid (ALA), gamma-Linolenic acid (GLA), Dihomo-gamma-Linolenic acid (DGLA) using enzyme Cyclooxygenases (COX), lipoxygenase (LOX), cytochrome P450 (CYP), Soluble epoxide hydrolase (sEH) or non-enzymatic reaction. Fatty Acid Desaturase (FADS), Elongase, Glutathione peroxidase 4 (GPX4), Leukotriene-A4 hydrolase (LTAH), Prostaglandin E synthase (PTGES), Aldose reductase (AKR1B3) enzymes were also needed to produce some of those metabolites from Leukotriene A4 (LTA4), Prostaglandin H (PGH), Prostaglandin G (PGG), Isoleukotoxin/Leukotoxin (EpOME), Hydroperoxy Docosahexaenoic Acid (HpDHA). DELs identified belong Hydroxy Eicosatetraenoic Acids (HETE), Oxo Eicosatetraenoic Acid (oxoETE/KETE), Prostaglandin E (PGE), Prostaglandin F (PGF), Thromboxane B2 (TXB₂), Leukotriene B4 (LTB₄), Isoprostane (IsoP), Hydroxy Octadecatrienoic Acid (HOTre), Isoleukotoxin/Leukotoxin diol (DiHOME), Hydroxy Eicosapentaenoic Acid (HEPE), Dihydroxy Eicosatetraenoic Acid (DiHETE), Epoxy Eicosatetraenoic Acid (EpETE), Hydroxy Docosahexaenoic Acid (HDPA/HDoHE), Epoxy Docosapentaenoic Acid (EpDPA/EpDPE), Dihydroxy Docosapentaenoic Acid (DiHDPA/DiHDPE), Neuroprostane (NeuroP), Resolvin D (RvD), Protectin DX (PDX) sub-class of lipids. Production pathways were built using the KEGG and WikiPathways (LIPID MAPS) database and work from Coras & al., Calder, Wen [28, 103,104,105]. The figure was created with BioRender.com

Fig. 7

Lipidomic profiling of lung homogenates from the elderly and young mice infected or not (mock) with SARS-CoV-2. A Differentially expressed lipids (DELs) between the two mouse groups for the same infection time-point (mock, 3 DPI, 5 DPI, 7 DPI). Color blue to red indicates log₂ of the fold change (FC) in the aged mice in comparison with the young mice and the dot size indicates the P value calculated with uncorrected unpaired T-test assuming equal group variance (Mean n = 3–5/group). B Levels of Lipids significantly different according to age when adjusted (marked with #) or not with the infection time-point. Lipids with an unadjusted P value below 0.05 in the bar graph. P values were calculated with a linear model based on the limma method [106] (n = 3–5/group). C Specific Lipids showed a significant interaction with age and infection time-point. Lipids that display a raw P value below 0.05 at the Two-way ANOVA tests are shown in the Venn diagram (n = 3–5/group). The 27 lipids that show no only interaction and differed only with the stage of infection were ± 19(20)-DiHDPA, ± 8-HEPE, ± 16-HDPA, ± 17(18)-DiHETE, ± 20-HDPA, Prostaglandin E3, RvD2 n−3 DPA, ± 5-HEPE, ± 7(8)-DiHDPA, ± 13-HDPA, 13(S)-HODE, ± 8-HDHA, 9(S)-HODE, 9(R)-HETE, ± 14(15)-DiHETE, 11(S)-HETE, ± 5(6)-DiHETE, Thromboxane B2, ± (18)-HEPE, 17(S)-HDHA, ± 4-HDPA, ± 7-HDPA, ± 11-HDPA, Protectin DX, 8-isoPG E2 and 5-F_2t-IsoP

Following a linear model analysis, our results show that 19 lipid mediators display a significant difference with the advanced age (old vs young) when adjusted or not for infection time point (Fig. 7B). Among these lipids, 9(S)/13S-HOTrE, ± 11-HEPE, and ± 9(10)-DiHOME have shown significant difference only when adjusted with the covariate infection time point. Six lipids globally reduced with old age, and they were produced from DHA or EPA via CYP or AA using LOX15/5. The 13 other mediators showed a significant positive association, and they were mainly produced from EPA, DHA, AA, or LA/ALA through the actions of CYP, LOX, or via non-enzymatic reaction. When analyzed in an ANOVA 2 for interaction between age (old/young) and infection time point, six mediators were mainly influenced by age (Fig. 7C). Most of them displayed were negatively impacted except 13(S)-HOTrE and LTB₄ which showed a significant positive association (Fig. 7B). The level of nine lipids was significantly impacted by both, infection and age and 27 lipids only by the infection. A summary of the differential lipid metabolite functions is shown in Supplementary Table 2.

Young and elderly mice exhibit distinct transcriptomic profiles

Whole RNA sequencing was performed on lung RNA of mock and 5 DPI mouse groups. Sequenced transcripts were pseudoaligned and quantitated using mouse and SARS-CoV-2 reference transcripts. Principal components analysis using gene counts show that each group (mock and 5 DPI for aged and young mice) form a distinct population, despite variance between individuals (supplementary Fig. 2B). Corresponding infection time point tends to cluster in Dim2(PCA2) while different age groups transcripts per kilobase million (TPM) to cluster in Dim1(PCA1). Those findings suggest that these two factors seem to explain most of the transcriptomic changes between mouse groups.

Prior statistical analysis, low expressed transcript with a TPM value below 1 in all mice were filtered out. Gene expression changes between conditions were evaluated (Fig. 8A to 8D). As shown in Figs. 8A and 8B, absolute number of totally differentially expressed genes (DEGs) upon infection are relatively similar between both age groups. However, at 5 days post SARS-CoV-2 infection, an increase in the number of up-regulated genes in young mice and increase in down-regulated gene in elderly mice is observed. As indicated by the distribution of the P adjusted values, old mice DEG show a weaker confidence level probably caused by the higher intra-individual variability described above. When comparing non-infected young and old mice groups, a predominance of up-regulated DEGs is observed in aged mice relative to the younger mice, with a tendency to switch at 5 DPI with a balanced proportion of up and down regulated DEGs (Fig. 8 C and D). Notably, ORF1a/ab transcripts were the only SARS-CoV-2 transcript differentially expressed between the two infected groups and were expressed at higher levels in elderly mice (not shown).

Fig. 8

Transcriptomic modulation in mouse lungs of aged and young mice at 5 days post SARS-CoV-2 infection. Differential gene expression between 5 DPI and mock infected young mice (A) and aged mice (B). Differential gene expression between the old and young mice for mock (C) and 5 DPI groups (D). Transcripts with a P ajusted value below 0.05 and with a fold of change bellow 0.6 or above 1.5 were highlighted in red or blue respectively (n = 4–5/group)

Following the gene set enrichment analysis (GSEA) on the log₂(Fold of change) of each genes using Hallmark gene sets, pathways related to proto-oncogenes and tumor invasion such as phosphoinositide 3-kinase(PI3K)/ protein kinase B (AKT)/ Serine/threonine-protein kinase (mTOR) signaling, epithelial mesenchymal transition and androgen response were induced by infection in young mice but not in older mice (Fig. 9 A and B). When comparing both mock groups, only genes related to allograft rejection were significatively up-regulated in elderly mice (Fig. 9C). For the 5 DPI comparison, in aged mice a down-regulation of genes associated with pathways linked to tissue development/reorganisation such as epithelia mesenchymal transition and apical junction was observed. This down-regulation also included genes associated with cell division involved in the G2M Checkpoint and mitotic spindle formation (Fig. 9D).

Fig. 9

Gene Set Enrichment Analysis on transcriptomic profiles of mock and 5 DPI young and elderly mice against the mouse hallmark database. Pathways showing significant enrichment at 5 DPI in comparison with their correspondent mock control group for young mice (A) and old mice (B). Pathways showing significant enrichment or depletion in mock (C) or 5 DPI (D) aged mice when compared with young mice group. Count values represent the number genes identified in a specific pathway and enrichment score represent the ratio between gene identified and the total number of gene in the pathway. Only pathways showing P adjusted value below 0.05 (Benjamini–Hochberg method) were display in the different dotplot (n = 4–5/group)

GSEAs were also performed using Reactome Pathways. As more pathways were highlighted, gene set similarly modulated in both linked comparisons were removed (Fig. 10 A and B). As shown in the Fig. 10A, pathways which were activated at 5 DPI only in young mice were mainly involved in DNA replication/reparation, transcription regulation, translation regulation and cell cycle. On the other hand, gene sets up-regulated at 5 DPI in old mice were involved in innate and adaptative immunity such as TLR, NLR and RLR signaling and regulation together with TCR and antigen presentation signaling (Fig. 10B). Pathways overexpressed in mock elderly mice also included Fc receptor and complement signaling (Fig. 10C). Comparison of 5 DPI groups highlighted the suppression in elderly mice of gene sets involved in the organisation and maintenance of the cytoskeleton and also an activation of PD1 signaling linked with T cell death and exhausting (Fig. 10D).

Fig. 10

Gene Set Enrichment Analysis on transcriptomic profile of mock and 5 DPI young and elderly mice against the mice Reactome Pathway database. Pathways showing significant enrichment or depletion in 5 DPI in comparison with their correspondent mock control group only in young mice (A) and only in old mice (B). Pathways showing significant enrichment or depletion in mock aged mice when compared with young mice (C). Pathways showing significant enrichment or depletion in 5 DPI old mice in comparison with young mice but not in mock group comparison (D). Pathways Count value represent number gene identified in a specific pathway and enrichment score represent the ratio between gene identified and the total number of gene in the pathway. All the pathways showing P adjusted value below 0.05 (Benjamini–Hochberg method) were displayed in panel B to D and the 25 pathways with lower P adjusted value were displayed in panel A (n = 4–5/group)

Discussion

Following SARS-CoV-2 infection, both young and aged mice demonstrated signs of infection including ruffled hair and weight losses. However, elderly mice failed to recover their weight by 9 DPI and experienced a tendency for a worse survival rate. Differences in the severity of COVID-19 are less age-dependent in the K18-hACE2 model than what was observed in human patients [21]. This discordance between this model and humans is likely attributable, at least partially, to the wider tissue distribution and overexpression of the hACE2 receptor in K18 mice that lead to very high susceptibility of these mice to SARS-CoV-2 [40]. Study from Onodera and collaborator highlighted greater weight loss and lung injury in male K18-mice in comparison with female [41]. Due to the small size of our mice group and the complexity in obtaining matched sex cohorts for eighteen-month-old mice, we could not detect any differences according to the sex of the animal.

While young mice successfully cleared the infection by day 7, significant infectious viral loads were present in the lungs of aged mice. A similar viral persistence in elderly human patients was suggested previously by Longtin and collaborators [42]. Considering viral loads and type I and III IFN responses were similar in young and aged mice at early time points (days 3 and 5), this suggests that the viral persistence results from improper viral clearance rather than increased viral replication in aged mice.

When CCL2, CCL3, CXCL9, and IFNγ were measured in lung homogenates, older mice produced less such mediators in response to infection relative to young mice. Those mediators are well known as recruiters and activators of several leucocyte populations involved in the T cell response build-up [43,44,45]. More precisely CXCL9 and IFNγ are associated with a T helper (Th) 1 response that favors the elimination of intracellular pathogen-like virus [46]. Despite not being directly involved in the Th1 polarisation, CCL2 and CCL3 are associated with responses against viral infections and could favor the build-up of a potent antiviral response through the recruitment of NK cells and monocytes/macrophages [47]. Tnf gene expression was also reduced in infected old mice supporting the hypothesis of an impairment of the Th1 response in this group. Weaker production of those mediators could explain the viral persistence observed at 7 DPI in elderly mice. A reduced level of CXCL1 was also measured, but it did not impact negatively the granulocyte recruitment in the lung, which was known as the main target of this chemokine [48, 49].

As highlighted above, the global level of protein mediators (Cytokine, Chemokine, IFN) was higher in younger mice at 3–5 DPI relative to older mice. This finding correlates with higher global inflammatory scores for the same time points in the young mouse group. For this group, inflammation scores increased with the same magnitude from mock to 3 DPI in the three lung compartments (blood vessels, bronchus, and parenchyma). For the same time points in the aged mouse group, a significant increase in inflammatory scores was observed only outside of the blood vessels region at 5 DPI. This suggests that in elderly mice, leucocytes recruitment was reduced and failed to efficiently migrate through the lung tissu to reach the infection site. This finding is also supported by the global immunosenescent observed in lungs from aged patients, where inaccurate migration and impaired anti-microbial function of lung leucocytes were observed [50].

As was observed in peripheral blood of severe COVID-19 patients, CD4⁺ cell count in mouse lungs were decreased during SARS-CoV-2 infection [51]. In our model, the SARS-CoV-2 infection increased the CD8⁺ cell count in K18-mice lungs contrary to what was reported in peripheral blood from humans. A decrease in CD4⁺ cells in aged mock-infected mice was observed which is in accordance with T cell age-dependent depletion in COVID-19 patient [50, 52, 53]. The lower CD4⁺ and CD8⁺ at 5 DPI in elderly mice is consequent to the lower production of CCL2, CCL3, CXCL9, and IFNγ and could explain the impaired viral clearance in this mouse group. In accordance with this finding, a global reduction of F4/80⁺ (macrophages) and NK1.1⁺ (NK cells) was observed in aged mice groups. Indeed, no NK1.1⁺ recruitment and a weak F4/80⁺ recruitment were observed in old mice infected with SAR-CoV-2. NK1.1⁺ recruitment patterns were in accordance with late NK cell recruitment and NK cell dysfunction in aged patient [53, 54]. F4/80⁺ also co-ordinate with impairment in lung macrophages function in SARS-CoV-2 infected hamster model[55]. Winkler et al. observed significant recruitment of neutrophils 7 DPI in lung of 9-week K18-hACE2 mice[56]. The difference in this cell recruitment could be caused by the use a SARS-CoV-2 dose about fifty times higher (of 2.5 × 10⁴ plaque forming unit) than the 5 × 10² TCID₅₀ used in the present study.

In the absence of infection, elderly mice express higher levels of PGE₂ and TXB₂, two lipids associated with pro-inflammatory activities [57, 58]. LTB_4, also a pro-inflammatory lipid mediator, is also expressed at higher levels in aged mice. In contrast, 15-OxoETE identified as an NF-κB response inhibitor, is expressed at reduced levels in aged mice [59, 60]. During infection, some HOTrEs and 11-HEPE were globally present at higher levels in the elderly than in young mice 3 days post-infection. These mediators are mainly connected with anti-inflammatory and pro-resolution responses [61,62,63,64]. Together, these results suggest that aged mice exhibit a higher basal level of inflammatory lipid mediators in comparison to young mice. This finding is in accordance with the high expression of stress-related genes observed in elderly human subjects [52].

Following SARS-CoV-2 infection, elderly mice also displayed a different lung lipidome than younger mice. Notably, 12(13)-DiHOME and 9(10)-DiHOME, also known as Leukotoxin diol, showed a positive relation with age and were present at significantly higher levels in aged mice day 3 post-infection. As these mediators are closely related to lung inflammation and potential contributors to ARDS, our results suggest a possible explanation for the worse outcome observed in elderly mice [65,66,67,68,69,70]. Elevated levels of DiHOMEs in plasma have been linked with the severity of COVID-19 [71]. In contrast, a major part of upregulated and differentially expressed lipids (DELs) at 3 DPI in elderly mice are known to have protective or pro-resolution effects, such as RvD3, PDx, 4-F_4t-NeuroP, HOTrEs, ± 14(15)-DiHETE, ± 15-HEPE, and ± 11-HEPE [61,62,63,64, 72,73,74,75,76,77,78]. This protective inflammatory response could mitigate the inflammation burst linked with DiHOMEs production. In addition to the potent inhibitory effect of ± 14(15)-DiHETE on NK cells, these mediators could contribute to the impairment of viral clearance [79]. In accordance with the high production of immunomodulator lipids in aged mice at 3 DPI, most of the DELs at 5 DPI were lower in old than young mice. Not surprisingly, part of them have moderate pro-inflammatory roles like PGF_3α [80], 11(S)-HETE, and 8(S)-HETE [81, 82]. However, most of them have also immunomodulatory functions, notably, RvD2, 19(20)-DiHDPA, HDPAs, 15-OxoETE, ± 17(18)-DiHETE, and ± 18-HEPE [72, 77, 78, 83,84,85,86,87,88]. Notably, the level of 15(S)-HETE was lower in 5 DPI aged mice than in their younger counterparts. The urinary 15-HETE expression in mild COVID-19 patients has been correlated with higher CD4 and CD8 T cell activation [71].

The differential regulation of anti-inflammatory mediators at 3 DPI and 5 DPI may suggest that the pro-resolution response takes a longer time to build up in younger mice allowing the establishment of a more efficient antiviral response and better infection management.

Specific isomers of 5-series F₂-isoPs and 4-series F₄-NeuroPs, markers of oxidative stress, were much higher in old than younger mice 3 days post-infection as reported here in the lung. This is in accordance with plasma levels of 5-series and 14-series F₂-isoPs that were previously associated with the severity of COVID-19 infection in human patients [71, 89]. Increase in 5-series and 15-series F₂-isoPs in plasma had also been associated with COVID-19 ICU admission[90]. Moreover, patients who survived or died from COVID-19 showed higher levels of 4-hydroxy-nonenal (4-HNE) a marker of lipid peroxidation in plasma than healthy individuals [91]. Our results are in line with the literature for increased oxidative stress in more severe infections found in old animals.

The difference in the lipidome profiles between aged and young mice did not match with the lipid mediator increased in hospitalized COVID-19 patient BALF or exhaled breath condensates. Indeed, increased levels of ± 19(20)-EpDPA, 15(S)-HETE, ± 15-HEPE, and ± 5-HEPE in COVID-19 patients have been reported while our results suggest a lower level of these lipids in the lungs of older mice [15, 92]. These results confirm the difficulty of comparing mouse lung homogenates with COVID-19 patients BALF, as previously stated [25]. Severe COVID-19 in human patients tends to progress more slowly than in acute infection animal models such as K18-hACE2 mice. The median time from onset to hospitalization was 7 to 11 days and admission to an intensive care unit was 9 days [93,94,95]. In our previous work, a higher level of LTB4, PDX, and 13-HOTrE was detected in BALF from intubated COVID-19 patients which is in line with what was observed in aged mice. On the other hand, 5/15-HETE/OXOETE and RvD2 were detected at lower levels in elderly mice while highly produced in human patients [15]. Lower level of 5/15-HETE in the plasma had been associated with COVID-19 ICU admission which was in accordance with the decreased level of those compounds in aged mice [90]. Moreover, we detected an increased level of RvD3 and PGF_3α that was not observed in the Severe/ICU patient BALF or plasma[15, 90]. In our experiment setup, the mice began to show severe respiratory disease signs from day 5 post-infection, it is predictable that the lipidomic profile didn’t reflect what has been observed in hospitalized COVID-19 patients. Work from Wong & al. highlights the implication of the PGD₂ pathways on the viral clearance and disease outcome in middle-aged mice infected with adapted SARS-CoV-2 strain [96]. Our work did not detect a change in the PGD₂ expression in elderly infected mice in comparison with young ones. This may be explained by the different disease courses observed in wild-type C57BL/6 used in Wong's studies versus the K18-hACE2 mice used in the present studies. As the K18 model is more susceptible to SARS-CoV-2, we observed a decrease in survival in young mice which is not the case with young wild-type C57BL/6 [96].

GSEA indicate that most of the significantly suppressed pathways in 5 DPI old mice were associated with tissue and epithelium morphogenesis. This finding suggests that a potential disorganization and destruction of lung tissue leading to impaired lung functions in aged mice. The down regulation of genes associated with tissue/epithelium morphogenesis were also observed in aged COVID-19 patients and were correlated with lung senescence and lung-term lung sequelae[97]. This finding is also in agreement with the upregulation of necrosis pathways observed only in infected aged mice.

Moreover, an upregulation of PI3K/AKT signaling, mTOR signaling, MYC targets, mTOR signaling, P53 pathways and KRAS signaling targets following infection in 5 DPI young mice suggests a differential metabolic switch following SARS-CoV-2 infection. All those factors are known as modulators and effectors of Warburg effect or aerobic glycolysis [98,99,100]. As aerobic glycolysis supports the activation of pro-inflammatory macrophage and cytotoxic immune cells, this finding is coherent poorer viral clearance observed in aged mice [101, 102]. In accordance with other signs of impaired antiviral responses, negative regulation of RLR signaling through DDX58, impairment in TLR4 signaling, PD1 signaling were also observed in 5 DPI aged mice but not in younger mice.

In summary, the present study highlights that SARS-CoV-2 infection of K18-hACE2 mice triggers a robust early immune response in younger mice characterized by a production of Th1-related mediators and antiviral chemokines. This response was significantly reduced in aged mice. Consequently, more severe clinical symptoms and deficient viral elimination were observed in the aged mouse cohort. Moreover, we observed a deep modulation of the bioactive lipid profile between the two mouse groups following SARS-CoV-2 infection. Indeed, lipids that were involved in anti-inflammatory and protective activity were detected at a higher level in aged mouse lungs at an early infection time-point. This finding suggests that the balance between immune activator and repressor signals given plays a crucial role in the build-up of an efficient antiviral response against SARS-CoV-2. Those results highlight the potential of targeting the immune response instead of the viral replication as a treatment against severe COVID-19 experienced by elderly people.