Dietary Cholesterol Causes Inflammatory Imbalance and Exacerbates Morbidity in Mice Infected with Influenza A Virus

This article is featured in Top Reads, p.

Respiratory viruses are a prevalent and persistent global health threat. The World Health Organization estimates that influenza virus is responsible for 290,000–650,000 deaths worldwide every year (1). Individuals with existing comorbidities, such as those with diabetes, cardiovascular disease, or who are immunocompromised are at elevated risk for disease complications, which can result in hospitalization and death (2, 3). Dyslipidemia is a common underlying metabolic condition of patients at risk for inflammatory diseases (4, 5). In fact, there is an association between serum lipid levels and prognosis after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection (6–8). Obesity, for which dyslipidemia is a hallmark feature (4), has long been associated with immunosuppression and increased susceptibility to infection (9, 10) and is an independent risk factor for severe disease caused by SARS-CoV-2 and influenza virus (11–13). Both circulating lipoproteins and cellular cholesterol metabolism can act to influence immune responses (14–16), but there is still conflicting evidence on the extent to which dietary cholesterol contributes to serum cholesterol levels in the absence of mediating confounders such as obesity and metabolic conditions (17, 18). Observational reports show that statins, cholesterol-lowering medications, are associated with improved survival during influenza pneumonia (19), SARS-CoV-2 (20), and sepsis (21). While dietary cholesterol appears to influence the pathogenesis and natural history of other infectious pathogens (22–26), further investigation is needed on whether dietary cholesterol can impact immunity to viral respiratory infection.

Increased cellular cholesterol may affect the pathogenesis of IAV infection by acting on the pathogen or the host. First, cholesterol is crucial for entry and replication of enveloped viruses, including coronaviruses (27–29) and influenza viruses (30–33). Cholesterol depletion studies indicate that cholesterol content of the viral envelope is essential for viral fusion and, furthermore, that infectivity can be partially restored by addition of exogenous cholesterol (30). Depletion of cholesterol content of the host cell also appears to disrupt viral infectivity (32). Second, cholesterol supply is crucial to immune cell proliferation and function during infection. Upon activation, T cells upregulate cholesterol biosynthesis and uptake, as well as downregulate cholesterol efflux processes during clonal expansion (34, 35). Inhibition of sterol regulatory element–binding proteins decreases T cell proliferation (35), and overloading cholesterol content of immune cell membranes results in a hyperresponsive phenotype, which may be a consequence of lipid raft modulation. Macrophages that lack ATP-binding cassette genes Abca1 or Abcg1, responsible for cellular cholesterol efflux, exhibit higher levels of free cholesterol, increased lipid raft formation, increased TLR-mediated signaling, and a shift toward an inflammatory phenotype (36–38). Lipid rafts also appear to be involved in either excluding or aggregating key signaling molecules during lymphocyte activation (39, 40). In both B and T cells, the Ag receptor is translocated to the lipid raft region upon Ag binding (41, 42). Moreover, ablation of cholesterol esterification enzyme acetyl-CoA acetyltransferase 1 resulted in cholesterol enrichment of the plasma membrane, which led to enhanced TCR clustering and signaling, potentiated effector function, and greater proliferation of CD8⁺ T cells (43).

Elucidating the effect of dietary cholesterol on the immune system during viral infection is especially relevant given the removal of any quantitative limit to dietary cholesterol intake from the Dietary Guidelines for Americans starting in 2015 (44). Thus, the purpose of this study was to determine the role of dietary cholesterol in modulating the immune response to experimental mouse-adapted human IAV infection. We demonstrate that dietary cholesterol increased morbidity in IAV-infected mice and induced an aberrant immune response to IAV infection characterized by an increased number of inflammatory CD4⁺ and CD8⁺ T cells, IL-10⁺CD8⁺ cells, and Ly6C⁺ conventional dendritic cells without affecting viral load.

Six-week-old C57BL/6J mice were obtained from The Jackson Laboratory (no. 000664). Animals were tail-marked and housed four per cage under constant 12-h light/12-h dark cycles (10 am–10 pm) and constant temperature. For experiments with 1 hemagglutination unit (HAU), male (n = 3 per diet per time point) and female (n = 3 per diet per time point) mice were used. Because there was no observed effect of sex, results from males and females were combined. For experiments with 0.7 HAU and 100 PFU, only male mice were used. At 11 wk of age, mice were anesthetized with 3% isoflurane and then intranasally inoculated with either sterile PBS, 1 HAU, or 0.7 HAU of mouse-adapted human influenza A virus (IAV) strain A/Puerto Rico/8/1934 (H1N1) (A/PR8) diluted in PBS. For cohorts 2 and 3 an inoculation titer of 100 PFU was used. The total inoculation volume was 30 µl. Animal weights and food disappearance were measured daily. Percent weight change was calculated from day 0 postinfection (p.i.). Mice were euthanized by CO₂ asphyxiation at days 4, 8, and 16 p.i. for follow-up experiments. Mice that lost 30% of body weight were euthanized immediately. All animal care protocols were in accordance with the National Institutes of Health’s Guidelines for Care and Use of Laboratory Animals and were approved by the University of Illinois Institutional Animal Care and Use Committee.

For 0.7 HAU and 100 PFU experiments, mice were fed either a standard rodent chow (Teklad 2016) or a matched diet containing 2% cholesterol (TD.07841) that differed only in cholesterol content. For 1 HAU experiments, mice were fed a standard rodent chow (Teklad 2018) or energy density–matched 2% cholesterol diet (TD.07841). Feeding of each diet began at 6 wk of age. Mice were maintained on their respective diets for 5 wk prior to inoculation and until termination.

Sickness behavior was assessed by the burrowing behavior test in which a mouse is presented with a polyvinyl chloride tube that is filled with a premeasured amount (∼200 g) of food pellets with the entrance of the tube elevated. Due to its natural inclination to burrow, the mouse digs out pellets when behaving normally. This test has been shown to be a sensitive measure of behavioral abnormalities in rodents after LPS or cytokine challenge (45).

Total serum cholesterol and triglyceride levels were spectrophotometrically measured on the Beckman Coulter AU680 clinical chemistry analyzer at the Veterinary Diagnostic Laboratory of the College of Veterinary Medicine at the University of Illinois at Urbana-Champaign. Serum samples for cholesterol and triglyceride quantitation were treated per the manufacturer’s instructions (Beckman Coulter, nos. OSR6516 and OSR60118, respectively). Briefly, for cholesterol measurement, cholesterol esters were hydrolyzed by cholesterol esterase to produce free cholesterol, which was oxidized by cholesterol oxidase causing simultaneous hydrogen peroxide production. Hydrogen peroxide reacted with 4-aminoantipyrine and phenol in the presence of peroxidase to produce a red chromophore that can be measured at 540/600 nm as an increase in absorbance. For triglyceride measurement, triglycerides were hydrolyzed to produce glycerol and fatty acids, then glycerol was phosphorylated to produce glycerol-3-phosphate, which was oxidized to produce hydrogen peroxide. Hydrogen peroxide reacted with 4-aminophenazone and N,N-bis(4-sulfobutyl)-3,5-dimethylaniline, disodium salt in the presence of peroxidase to produce a blue chromophore that can be measured at 660/800 nm as an increase in absorbance. Two samples of the same condition were pooled for each sample run resulting in n = 3 pooled samples per diet per time point.

Images of liver were taken immediately upon tissue collection. Liver color was scored by mean gray value measurements using ImageJ software (National Institutes of Health, public domain). Rectangular selection was held constant and areas of image distortion (e.g., glare) were avoided. For the crossover experiment (Fig. 9), upon tissue collection, each liver was blindly assigned a color score of 0 (deep red), 1 (intermediate), or 2 (pale).

Spleens from mice inoculated with PBS or 0.7 HAU of IAV, and lungs from mice inoculated with PBS or 100 PFU of IAV, were collected and immediately placed in RPMI 1640 media on ice. Single-cell suspensions were prepared and counted, then mixed spleen or lung cells were plated at a density of 5 × 10⁵ cells per well for spleen and 2 × 10⁵, 1 × 10⁵, and 5 × 10⁴ cells per well for lung in a round-bottom 96-well plates in RPMI 1640 supplemented with 5% FBS, l-glutamine, and 1% penicillin/streptomycin. Spleen cells were stimulated for 48 h and lung cells were stimulated for 72 h in a total volume of 200 µl with 1) media, 2) VP2_121–130, the immunodominant MHC class I–restricted peptide derived from Theiler’s murine encephalomyelitis virus (TMEV), 3) nucleoprotein (NP)_366–374, or 4) acid polymerase (PA)_224–233, with the latter two being immunodominant MHC class I–restricted peptides of A/PR8 IAV. Peptides were sourced from Anaspec (Fremont, CA). Each peptide was added at a concentration of 2 µM. Each stimulation condition was performed in duplicate. Following stimulation, supernatants were collected and IFN-γ levels were determined by ELISA (Thermo Fisher Scientific, no. 88-7314).

Neutralizing Ab levels were measured by a hemagglutination inhibition (HAI) assay as described by the World Health Organization’s Manual on Animal Influenza Diagnosis and Surveillance (46). Serum was collected from mice inoculated with PBS or 0.7 HAU of IAV and treated with receptor-destroying enzyme II (Denka Seiken, no. 370013) to inactivate nonspecific inhibitors of hemagglutination. Twenty-five–microliter vol of 2-fold serial dilutions of treated serum in PBS were prepared and incubated with 25 µl of 4 HAU/25 μl of A/PR8 IAV in a V-shaped 96-well plate for 15 min. Then, 50 µl of standardized turkey RBCs (Innovative Research, no. ITKRBC5P) was added at a concentration of 0.5% for a final reaction volume of 100 µl. Eight dilutions were prepared per sample and each dilution was performed in triplicate. Results were interpreted after 30 min of incubation or once the RBC control well was fully settled. The HAI titer is reported as the reciprocal of the highest dilution of serum that completely inhibited hemagglutination.

Whole blood was collected with EDTA-lined syringes by cardiac puncture, and RBCs were lysed by ammonium-chloride-potassium (ACK) lysis buffer. Lungs were enzymatically digested into single-cell suspensions and prepared for flow cytometry analysis as previously described (47). Briefly, the right lung (cohort 1) or both lungs (cohorts 2 and 3) were collected from mice inoculated with PBS or IAV. Tissues were minced and digested in a type II collagenase/DNase solution, then passed through a 40-µm filter and incubated with ACK lysis buffer. After washing, the total number of cells per lung was determined by an automated cell counter (Cellometer Auto T4, Nexcelom Bioscience).

Lung cells (1 × 10⁶ per reaction) were stimulated with either media (RPMI 1640 supplemented with 5% FBS, l-glutamine, and 1% penicillin/streptomycin) or a combination of PMA (81 nM) and ionomycin (1.34 µM) in the presence of brefeldin A (3 µg/ml) for 5 h at 37°C. The following staining steps were performed at 4°C. Cells were incubated with anti-CD16/32 Ab for 10 min to block Fc receptors, stained with surface-binding Abs for 20 min in the dark, fixed for 20 min with intracellular fixation buffer, permeabilized for 10 min with permeabilization buffer (Invitrogen, no. 00-8333-56), and then stained with Abs against cytokines for 20 min in the dark. For staining of myeloid cell surface markers (no intracellular staining), cells were not stimulated. Lung cells were labeled with the following Abs: CD45–eFluor 450 (30-F11; Thermo Fisher Scientific), CD45-BV421 (30-F11; BioLegend), TNF-BV605 (MP6-XT22; BioLegend), CD8a-SB645 (53-6.7; Thermo Fisher Scientific), CD8a-allophycocyanin (53-6.7; BioLegend), CD3-PerCP-Cy5.5 (17A2; BioLegend), CD19-BV711 (6D5; BioLegend), CD4-SB780 (RM4-5; Thermo Fisher Scientific), IL-10–AF488 (JES5-16E3; Thermo Fisher Scientific), IL-2–PE (JES6-5H4; Thermo Fisher Scientific), IFN-γ–PE/Dazzle594 (XMG1.2; BioLegend), IL-17A–AF700 (TC11-18H10.1; BioLegend), γδ TCR–PE-Cy7 (GL3; BioLegend), Siglec F–SB645 (1RNM44N; Thermo Fisher Scientific), CD11c-SB780 (N418; Thermo Fisher Scientific), Ly6C-PerCP-Cy5.5 (HK1.4; Thermo Fisher Scientific), Ly6C-FITC (HK1.4; BioLegend), Ly6G-PE-Cy7 (1A8; BioLegend), CD11b–PE–eFluor 610 (M1/70; Thermo Fisher Scientific), I-A/I-E (H-2^b)–AF700 (M5/114.15.2; BioLegend), CD192 (CCR2)–PE (SA203G11; BioLegend), and for viability eFluor 780 (Thermo Fisher Scientific, no. 65-0865-14). Blood cells were labeled with the following Abs: CD4–Pacific Blue (RM4-5; BioLegend), Ly6G-FITC (1A8; BioLegend), Ly6C-PerCp–Cy5.5 (HK1.4; BioLegend), B220-PE-Cy7 (RA3-6B2; BioLegend), CD11b-allophycoycanin (M1/70; BioLegend), and CD8a-allophycocyanin-Cy7 (53-6.7; BioLegend). Samples were run on an Attune NxT flow cytometer (Thermo Fisher Scientific), and data were analyzed using FlowJo software (v10.7.2). Compensation for all channels except viability dye was determined using single-stained compensation beads (Thermo Fisher Scientific, no. 01-2222-42). For viability dye compensation, lung cells were heated to 65°C for 15 min and then stained with viability dye. Gating of surface staining was facilitated by unstained samples. Gating of cytokine-stained samples was facilitated by media-stimulated controls as well as lung cells stained only with surface markers. The total number of cells per set of digested lungs was multiplied by the percentage of viable single cells and then by the appropriate subtype percentage to determine the number of cells of that subtype. For cohort 1, the numbers of cells were multiplied by 2 to account for both lungs.

Following euthanasia, the left lung was collected from mice inoculated with PBS or IAV, immediately flash frozen in liquid nitrogen, and stored at −80°C until they could be further processed. Frozen lungs were pulverized into a homogenized powder using a mortar and pestle (Cole-Parmer) on dry ice, and RNA was extracted by TRIzol (Thermo Fisher Scientific, no. 15596026) and then column purified using the GeneJET RNA purification kit (Thermo Fisher Scientific, no. K0732) as per the manufacturer’s instructions. Purified RNA quantity was determined by a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific), and quality was determined by running samples on a 1% agarose gel.

Total lung RNA was sent to the Roy J. Carver Biotechnology Center at the University of Illinois Urbana-Champaign for preparation, sequencing, and analysis. The RNA sequencing (RNA-seq) libraries were prepared with an Illumina TruSeq stranded mRNA-seq sample prep kit (Illumina), then pooled, quantitated by quantitative PCR, and sequenced on one S4 lane for 151 cycles from both ends of the fragment on a NovaSeq 6000. FastQ files were generated and demultiplexed with the bcl2fastq v2.20 conversion software (Illumina) and adaptors were trimmed. Salmon (48) (v1.4.0) was used to index the National Center for Biotechnology Information Mus musculus Annotation Release 109 transcriptome combined with the IAV (A/Puerto Rico/8/1934(H1N1)) transcriptome using the decoy-aware method with the entire GRCm39 + H1N1 genomes as the decoy sequence. Then quasi-mapping was performed to map reads to the combined transcriptomes with additional arguments –seqBias, –gcBias, –numBootstraps = 30, –validateMappings, and –recoverOrphans to help improve the accuracy of mappings.

The remaining statistical analyses were performed in R (49) (v4.1.0) using packages as indicated below. Gene-level counts were estimated from transcript-level counts using the “bias corrected counts without an offset” method from the tximport package (v1.20.0). This method provides more accurate gene-level count estimates and keeps multi-mapped reads in the analysis compared with the traditional alignment-based method (50). Genes without at least 0.25 counts per million in at least five samples were filtered out. The trimmed mean of M values normalized log₂ count per million values (51) was calculated and tested for differential expression using the limma-trend method in the limma package (52) (v3.48.0) using a final model that included six extra factors estimated from remove unwanted variation (RUV) sequencing analysis (53) of the 5000 least significant genes (first model with no factors) using the RUVs method to control for nuisance effects such as batch or library preparation. Ten pairwise comparisons were made: cholesterol versus standard diet within each time point and within each diet, with each of the later time points back to day 0. Multiple test correction was done using the false discovery rate (FDR) method (54) across all 10 comparisons together. Additionally, a weighted gene coexpression network analysis (WGCNA) (55) (package v1.70-3) was done on all genes for all samples using expression values that had the RUV factor effects removed using the blockwiseModules() function with default values except maxBlockSize = 22000, corType = “bicor”, power = 4, networkType = “signed hybrid”, and minModuleSize = 20. Because the day 8 time point overwhelmingly dominated the resulting modules of coexpressed genes, two additional WGCNAs were performed on just the day 0, 4, and 16 samples (power = 7) and the day 8 samples by themselves (power = 6). For each WGCNA, the eigengene values from each module were tested for differential expression across the 10 pairwise comparisons using regular limma modeling and global FDR correction. The RNA sequencing datasets associated with this paper have been deposited into the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/) and can be accessed using the accession number GSE197986.

Gene Ontology analyses on upregulated and downregulated gene sets were determined using DAVID bioinformatics resources 6.8 (56, 57). Lists of genes involved in specific immune pathways were determined using the KEGG database (58–60). A threshold FDR < 0.05 was used to determine differentially expressed genes (DEGs) from pairwise comparisons between cholesterol diet-fed and control diet-fed mice at each time point. Modules with cholesterol versus normal FDR < 0.05 were considered to have a significant diet effect. Individual genes within modules may not have a FDR < 0.05.

Madin–Darby canine kidney (MDCK) cells were purchased from American Type Tissue Culture (Manassas, VA; NBL2, ATCC CCL34); MDCK cells at passage 67 of type 1 morphology were received from the laboratory of Dr. J. Leibowitz (Texas A&M Health Science Center, Bryan, TX). American Type Tissue Culture–purchased cells were initially a heterogeneous pool and were enriched for type 1 morphology (61) by serial passages of light trypsinization–washed cultures to remove dysmorphic cells. Cells were propagated and maintained in 175-cm² Falcon-Corning tissue culture flasks with filter caps at 37°C and 5% CO₂. Complete growth medium for MDCK cells consisted of 10% FBS and 1% 10,000 U of penicillin/10,000 U of streptomycin (Lonza, no. 17-602E) in 1× DMEM with l-glutamine and sodium pyruvate (Corning Life Sciences, no. 10-017-CV).

Harvested mouse whole-lung tissues were flash frozen and stored at −80°C. Tissues were homogenized in a glass Dounce homogenizer (Corning Life Sciences VWR, no. 22877-282) while chilled on ice in sterile PBS at 500 μl/100 mg tissue. Lysate was clarified by centrifugation at 1000 × g for 10 min and stored at −80°C for subsequent plaque assay.

The plaque assay protocol was obtained from Dr. Luis Martinez-Sobrido and from published standard protocols (62, 63). MDCK cells were grown to 90% confluence in Nunclon six-well plates from VWR (no. 73520-906), washed with PBS (pH 7.2) twice, and inoculated in serial 10-fold dilutions of clarified virus. After a 1-h incubation at 37°C, virus inoculum was removed and overlaid with TPCK-trypsin, DEAE-dextran–modified complete 2× DMEM/F12 with Oxoid agar. The 2× DMEM/F12 complete media consisted of DMEM/F12 powder (Life Technologies, no. 12400-024), 2% 10,000 U of penicillin/10,000 U of streptomycin (Lonza, no. 17-602E), 2% HEPES buffer solution at 1 M (Life Technologies, no. 15630-080), 0.24% NaHCO₃ (from 5% stock), and 0.18% BSA fraction V (from 7.5% stock; Fisher Scientific, BP1605-100) in 500 ml of ddH₂O and filter sterilized. A 50% 2× DMEM/F12 complete media agar overlay was made with supplement of 0.01% DEAE-dextran (from 1% stock; Sigma-Aldrich, no. 93556-1G), 0.05% NaHCO₃ (from 5% stock), 0.6% Oxoid agar (from a 2% stock; Thermo Fisher Scientific, no. OXLLP0011B), and 2 μg/ml TPCK-trypsin (from a 2 mg/ml stock; Thermo Fisher Scientific, no. 20233) in ddH₂O. Infected cells were incubated inverted with overlay agar for 48 h at 37°C and 5% CO₂ and stained with 0.1% Crystal violet (Sigma-Aldrich, no. C0775-25G).

GraphPad Prism software (v9.1.0) was used to execute statistical analyses and to create graphs. For weight change, burrowing, and food disappearance data, repeated-measures mixed effects analysis with fixed effects of time, diet, and infection, matched values by time, and a Holm–Šídák correction for multiple comparisons were performed. Random effects of subject and residual were included in the model. For survival data, the log-rank (Mantel–Cox) test was performed. For liver scores from the crossover experiment, the Kruskal–Wallis nonparametric test with Dunn’s correction for multiple comparisons was performed. For spleen Ag recall assays, two-way ANOVA with fixed effects of time and diet and a Holm–Šídák correction for multiple comparisons were performed. For lung Ag recall assays, repeated-measures mixed effects analysis with fixed effects of diet, infection, and cell concentration, matched values by cell concentration, and a Holm–Šídák correction for multiple comparisons were performed. For polyfunctional flow cytometry data, a three-way ANOVA with fixed effects of diet, infection, and cell subtype was performed to determine any main effect of diet, or diet by infection interaction. Subsequently, multiple comparisons were determined per cell subtype by two-way ANOVA with fixed effects of diet and infection and a Holm–Šídák correction. All other flow cytometry data were analyzed by two-way ANOVA with fixed effects of diet and infection and Holm–Šídák correction. Correlations were analyzed by Pearson’s correlation test. For weight change and burrowing data from the crossover experiment, repeated-measures mixed effects analyses with fixed effects of time and diet, matched values by time, and a Holm–Šídák correction for multiple comparisons were performed. Significant results (p < 0.05) for all other measures were determined by a Student t test between control-fed and cholesterol-fed groups.

To determine whether dietary cholesterol influenced the pathogenesis of IAV infection, 6-wk-old C57BL/6J mice were fed either a standard rodent chow diet or a matched chow diet containing 2% added cholesterol. After 5 wk, mice were inoculated with either PBS or IAV, and then sickness behaviors and immune responses were assessed at days 0, 4, 8, and 16 p.i. (Fig. 1A). Prior to infection, mice fed a high-cholesterol diet did not exhibit differences in weight compared with control mice (Fig. 1B). In addition, food disappearance, a reflection of food intake, was comparable between mice fed a control diet and those fed a high-cholesterol diet prior to infection, but this was decreased in cholesterol-fed mice postinfection (Fig. 1C). As a result of infection, both groups of mice began losing weight at day 3 p.i. Weight loss in the cholesterol-fed group of mice mirrored that of controls until peak disease at day 8 p.i. After day 8 p.i., control mice began recovering weight. In contrast, cholesterol-fed mice did not begin to recover weight until day 12 p.i. (Fig. 1D). Likewise, both diet groups showed reduced burrowing behavior as a result of infection. However, this effect was exaggerated in cholesterol-fed mice compared with controls (Fig. 1E). Inoculation with a high dose of IAV resulted in early termination of a substantial number of mice due to the fact that they reached euthanasia criteria. There was no significant difference in the incidence of overall mortality between groups (Fig. 1F). In order to assess the effect of a high-cholesterol diet on disease exacerbation rather than survival, the experiment was repeated using a lower viral titer (0.7 HAU). Again, food disappearance was not altered by diet prior to infection (Fig. 1G). Although cholesterol-fed mice appeared to have reduced food disappearance postinfection, the effect did not reach statistical significance. Similar to the high-dose experiment, inoculation with a low dose of IAV resulted in exacerbated weight loss in the cholesterol diet group compared with the control diet group (Fig. 1H). Taken together, these data show that mice fed a high-cholesterol diet exhibited increased morbidity following IAV infection.

To determine whether cholesterol-fed mice developed dyslipidemia, circulating lipids were measured during the course of the experiment. Total free-cholesterol levels in the serum were higher in cholesterol-fed mice than those fed a control diet prior to infection. However, at days 4, 8, and 16 p.i., there was no difference in serum cholesterol levels between diet groups (Fig. 2A). In contrast, serum triglyceride levels were decreased in cholesterol-fed mice at days 4, 8, and 16 p.i. compared with controls (Fig. 2B). Finally, cholesterol-fed mice developed an altered liver appearance compared with controls, regardless of infection status, that resembled hepatic steatosis (Fig. 2C). These data show that cholesterol-fed mice developed dyslipidemia and confirm that the dietary treatment altered lipid metabolism.

Since cholesterol-fed mice were more susceptible to IAV, the effect of dietary cholesterol on the generation of virus-specific adaptive immune responses was assessed. The effect of diet on lymphocyte populations and effector responses was first assessed in the periphery. Infection did not alter the percentage of either CD4⁺ or CD8⁺ T cells in circulation (Supplemental Fig. 1A). However, in IAV-infected mice, the percentage of circulating B220⁺ B cells decreased at day 4 p.i. and returned to baseline levels by day 16 p.i. In contrast, the percentage of circulating granulocytes was increased at day 4 p.i., whereas the percentage of monocytes was not affected. With the exception of day 16 p.i., diet had no effect on circulating leukocytes. However, at this time point, dietary cholesterol decreased the percentage of circulating CD4⁺ T cells and increased the percentage of monocytes. To determine whether dietary cholesterol altered the generation of humoral immunity, serum HAI tests were used to determine levels of influenza-specific neutralizing Abs. Virus-specific Abs were increased at day 16 p.i. as a result of infection, but there was no significant difference between diet groups at any time point (Fig. 2D). Next, to test the effect of diet on virus-specific CD8⁺ T cell responses in the periphery, Ag recall assays were performed on cells isolated from spleen tissue (Fig. 2E). Mixed splenocyte cultures were stimulated with media or immunodominant H-2D^b–restricted peptides specific for either TMEV (VP2_121–130) (64) or IAV (NP_366–374, PA_224–223) (65, 66), and IFN-γ secretion was assessed. Little to no IFN-γ production was observed in response to stimulation with either media or the TMEV-specific control peptide VP2_121–130. In contrast, splenocytes from infected mice produced IFN-γ in response to stimulation with the IAV-specific NP_366–374 and PA_224–233 peptides at days 8 and 16 p.i. However, there was no effect of dietary treatment on IFN-γ production. Finally, Ag recall assays were performed on cells isolated from lung at day 8 p.i. to test the effect of diet on virus-specific CD8⁺ T cell responses in the target tissue at peak disease (Fig. 2F). Lung cells from PBS-inoculated mice produced no IFN-γ with media or peptide stimulation. For infected mice, low amounts of IFN-γ were detectable from lung cells stimulated with media and VP2_121–130 regardless of dietary group. In contrast, stimulation of lung cells from infected mice with IAV-specific NP_366–374 and PA_224–233 peptides increased IFN-γ levels 10- to 100-fold in a manner that was dependent on the number of cells in the culture. As in the periphery, there was no effect of dietary treatment on IFN-γ production in the lungs in response to NP_366–374 or PA_224–233 stimulation. These data indicate that dietary cholesterol did not inhibit the generation of virus-specific adaptive immunity and that immunosuppression was not the cause of increased morbidity observed in mice fed a high-cholesterol diet.

To assess transcriptomic changes to the lungs during influenza infection, RNA-seq was performed. A one-way ANOVA performed on all eight time-by-diet groups revealed that most transcriptomic changes occurred at day 8 p.i. compared with day 0 p.i. (FDR < 0.05), indicating that responses in the lung due to infection were most robust at day 8 p.i. Dietary cholesterol influenced the lung transcriptome such that 678 genes at day 0 p.i., 341 genes at day 4 p.i., 194 genes at day 8 p.i., and 233 genes at day 16 p.i. were differentially expressed between cholesterol- and control-fed mice (FDR < 0.05). The top 15 DEGs from each time point (Fig. 3A–D) as well as the significant Gene Ontology terms identified from all upregulated and downregulated genes are shown (Fig. 3E–H). This initial analysis indicated that noninfected cholesterol-fed mice had increased expression of genes involved in the cellular response to IFN-γ, but decreased expression of cholesterol-synthesizing genes compared with their control diet counterparts (Fig. 3E). At day 4 p.i., the lungs of cholesterol-fed mice exhibited increased expression of genes involved in the regulation of transcription and decreased expression of cholesterol-synthesizing genes (Fig. 3F). Fewer ontology terms were significant at day 8 p.i. and none exhibited an FDR of <0.05, indicating weak changes due to diet at this time point (Fig. 3G). Nevertheless, cholesterol-fed mice had increased expression of genes associated with innate immune responses, including Ccr1, Cxcl12, Csf3, Igf1, Mydgf, Plau, Lgals1, Saa4, and Serpina3n (Supplemental Fig. 2B). At day 16 p.i., cholesterol-fed mice had increased expression of genes involved in cell division compared with controls (Fig. 3H).

To garner further insight into how dietary cholesterol affected the biological processes occurring within the lung, a WGCNA on gene changes that occurred on day 8 p.i. alone was performed. This analysis resulted in 181 significant gene coexpression networks, or modules. None of these modules exhibited significant diet effects, but they reflected an overwhelming response to infection at this time point. Since these changes masked the effect of diet, we then performed WGCNA without day 8 p.i. gene changes. This analysis resulted in 84 total modules, 16 of which had a significant effect on diet (FDR < 0.05): 9 modules were significant at day 0 p.i., 4 at day 4 p.i., 2 at day 16 p.i., and 1 shared module was significant at all three time points analyzed (Fig. 4A, Supplemental Fig. 2A). Gene Ontology analysis performed on the shared module (module D0.D4.D16.A) identified genes involved in cholesterol and sterol biosynthesis pathways, again indicating that dietary cholesterol decreased expression of cholesterol biosynthesis genes at days 0, 4, and 16 p.i. but not at day 8 p.i. (Fig. 4B, 4C). Because genes involved in cholesterol biosynthesis in the lungs were downregulated in response to dietary cholesterol, we hypothesized that genes encoding cholesterol efflux proteins would be upregulated. Indeed, transcripts of the major cholesterol transporters involved in reverse cholesterol transport of Abca1 and Abcg1 were upregulated in the lungs of cholesterol-fed mice at days 0 and 16 p.i. and at days 4 and 16 p.i., respectively (Supplemental Fig. 2B). Neither Abca1 nor Abcg1 was upregulated at day 8 p.i. Taken together, these data indicate dietary cholesterol downregulated cholesterol biosynthesis and upregulated cholesterol efflux in the lungs prior to infection (day 0 p.i.), during the early stages of immune response (day 4 p.i.), and after resolution (day 16 p.i.), but not during the height of immune responses to IAV (day 8 p.i.).

Next, to assess the effect of diet on the pathogenesis of infection, modules associated with immune system pathways were examined. Gene Ontology analysis of all other nonimmune-related modules with significant diet effects are provided (Supplemental Table I). Gene Ontology analysis of module D0.H indicated that dietary cholesterol upregulated genes involved in Ag presentation, immune system processes, and response to IFNs even in the absence of infection at day 0 p.i. (Fig. 4D, 4F). Module D4.C contained sets of immune-related genes upregulated by dietary cholesterol at day 4 p.i. that were distinct from those upregulated at day 0 p.i. These genes included Tlr4, Tlr6, Tlr8, Ticam2, Itgb2, Itgb7, Itga4, Itgal, Itgax, Rhoa, Rac2, Ptk2b, and Plcg2 (Fig. 4G), which were associated with ontology terms such as leukocyte cell–cell adhesion, lymphocyte aggregation, and T cell activation (Fig. 4E). These results suggest that enhanced leukocyte trafficking and activation occurred in the lungs of cholesterol-fed mice compared with controls.

Transcriptomic analysis of gene changes between cholesterol-fed mice and those on control diet indicated that dietary cholesterol may promote leukocyte trafficking and activation in response to IAV infection. Therefore, the effects of diet on lung-infiltrating lymphocyte and leukocyte populations were assessed at the peak of disease (day 8 p.i.). Infection increased the number of immune subtypes examined at day 8 p.i. (Figs. 5, 6). The numbers of CD45⁺, CD4⁺, CD8⁺, and CD4⁻CD8⁻ cells per lung were not different between diet groups in the lungs of infected cholesterol-fed mice compared with infected control-fed mice (Fig. 5A–F). To determine whether diet influenced effector T cell function, polyfunctional T cell responses were measured, as these cells are critical for viral clearance (67–69). The number of TNF⁺IFN-γ⁺CD4⁺ T cells was increased in the lungs of infected cholesterol-fed mice compared with infected control-fed mice (Fig. 5K–M). In addition, dietary cholesterol increased numbers of IFN-γ⁺CD4⁺ T cells (Fig. 5K–M) and TNF⁺CD8⁺ T cells (Fig. 5N–P). The number of IL-10⁺CD4⁺ T cells was not altered (Fig. 5G, 5H), but the number of IL-10⁺CD8⁺ T cells was increased in the lungs of infected cholesterol-fed mice compared with infected controls (Fig. 5I, 5J).

Next, alterations to number of infiltrating CD4⁻CD8⁻ subtypes such as B cells, γδ T cells, dendritic cells, monocytes, and granulocytes were examined. Neither the number of CD19⁺ B cells, γδ T cells, nor CD11c⁺CD11b⁻Siglec F⁺Ly6C⁻ alveolar macrophages was affected by diet (Supplemental Fig. 3E–I). Moreover, the number of CD4⁻CD8⁻ cells in the lungs that produced IL-17A, which came almost exclusively from these cells and not from T cells, did not differ between diet groups (Supplemental Fig. 3B–D). Likewise, diet had no effect on numbers of MHC class II (MHC II)⁻CD11b⁺SSC^hi granulocytes, MHC II⁻CD11b⁺SSC^loLy6C⁺ monocytes, MHC II⁻CD11b⁺SSC^loLy6C⁻ monocytes, and MHC II⁺CD11c⁺CD11b⁻Ly6C⁺ plasmacytoid dendritic cells in the lungs during infection (Fig. 6A, 6C–K). Dietary cholesterol, however, increased the number of MHC II⁺CD11c⁺CD11b⁺Ly6C⁺ conventional dendritic cells (cDCs) during infection (Fig. 6A, 6B). Notably, numbers of IFN-γ⁺CD4⁺ T cells, IFN-γ⁺CD8⁺ T cells, TNF⁺CD4⁺ T cells, IL-10⁺CD4⁺ T cells, IL-10⁺CD8⁺ T cells, and especially MHC II⁻CD11b⁺SSC^hi granulocytes were highly correlated with percent weight loss of mice at peak disease, day 8 p.i. (Fig. 6L–T). In contrast to leukocyte numbers, percentages were not substantially altered by diet (Supplemental Fig. 1B). Taken together, these data show that dietary cholesterol increased the number and effector functions of infiltrating immune cells during infection and suggest that aberrant immune responsiveness contributed to the observed increase in morbidity.

Diet-induced potentiation of infiltrating leukocytes during infection is consistent with our RNA-seq data that showed transcriptional upregulation of chemokine signaling, leukocyte migration, and lymphocyte activation pathways at days 4 and 8 p.i. This effect was even more pronounced among infected mice with mild to moderate disease severity whose weight loss mirrored that of low-dose IAV-inoculated mice (Fig. 7). In contrast, infected mice displaying a more severe disease phenotype, mirroring weight loss of high-dose IAV-inoculated mice, had no differences in the numbers of leukocytes in the lungs at day 8 p.i. (Fig. 7A and data not shown). When analysis of IFN-γ production in the lungs at day 8 p.i. was restricted to mice exhibiting mild to moderate disease severity, infected cholesterol-fed mice produced greater amounts of IFN-γ compared with infected control-fed mice in response to media alone (Supplemental Fig. 2C). This demonstrates that with mild to moderate disease severity, infected cholesterol-fed mice not only had increased numbers of IFN-γ–producing cells (Fig. 7B) but also greater production of IFN-γ in the lungs compared with infected control-fed mice (Supplemental Fig. 2C). Taken together, these data suggest that the extent of diet-induced disease exacerbation at peak disease (day 8 p.i.) is dependent on severity of infection.

Cellular cholesterol status has been shown to influence infectivity of IAV (32). Therefore, it was possible that morbidity and heightened immune responsiveness within the lungs of infected cholesterol-fed mice was a consequence of increased viral load. To determine whether this was the case, we first mapped the reads generated from our RNA-seq experiment to the A/PR8 influenza genome. Viral genomic RNA was readily detectable in infected mice at days 4 and 8 p.i., but it did not differ between dietary groups (Fig. 8A). Moreover, dietary cholesterol did not influence lung viral titers at any time point examined, as determined by plaque assays (Fig. 8B). These data indicate that neither increased viral replication nor impaired viral clearance accounted for the increased morbidity that we observed in the infected cholesterol-fed mice. In contrast to correlations with the number of immune cells (Fig. 6L–T), weight loss was not correlated with viral load at peak disease (Fig. 8C).

Mice fed a diet supplemented with 2% cholesterol for 5 wk had increased morbidity following IAV infection that was characterized by an aberrant immune response. To determine whether this effect was temporally persistent, a random crossover experiment was performedfollowing a washout period of 5 wk (Fig. 9A). As before, dietary cholesterol did not affect weights prior to infection (Fig. 9B). Liver appearance was altered in mice that were fed a high-cholesterol diet for 10 wk (Chol/Chol) as well as in those that were switched to a high-cholesterol diet from a control diet (Control/Chol) when compared with mice that remained on a control diet (Control/Control). Liver appearance in Chol/Chol and Control/Chol were also different from mice that were initially fed a high-cholesterol diet for 5 wk but were switched to a control diet (Chol/Control). Liver appearance did not differ between Control/Control and Chol/Control groups (Fig. 9C). These data indicate that a 5-wk washout period was sufficient to restore changes to liver appearance. In contrast, diet reversal was not sufficient to alter degree of weight loss following infection. Specifically, mice that were ever fed dietary cholesterol (Chol/Chol, Chol/Control, Control/Chol), whether before or during infection, lost more weight as a result of infection than did mice fed a control diet for the duration of the experiment (Control/Control) (Fig. 9D). Finally, burrowing activity was decreased in the Chol/Chol group at day 6 p.i. compared with the Control/Control, Chol/Control, and Control/Chol groups, and at day 12 p.i. compared with the Control/Control group (Fig. 9E). These data suggest that a 5-wk diet washout period was not sufficient to fully reverse the effect of dietary cholesterol on IAV-induced morbidity.

In the current study we provide data showing that mice fed a high-cholesterol diet for 5 wk exhibited increased morbidity to a relevant viral respiratory pathogen, IAV. Regardless of whether mice received a low or high dose of virus inoculum, cholesterol-fed mice exhibited greater weight loss and decreased burrowing activity compared with those on a standard control diet. The exacerbation of virus-induced sickness behaviors resulting from dietary cholesterol was accompanied by an aberrant inflammatory response occurring in the lung. This response was characterized by increased expression of genes involved in leukocyte trafficking and activation. These transcriptomic changes were corroborated by our flow cytometry findings that demonstrated increased numbers of cytokine-producing CD4⁺ and CD8⁺ T cells as well as Ly6C⁺ cDCs in the lungs of infected cholesterol-fed mice compared with infected control-fed mice. Notably, this inflammatory response occurred in the absence of changes to viral load or clearance. Finally, we show that reversal of the diet for 5 wk was insufficient to completely overcome the adverse effects of dietary cholesterol. These results show that dietary cholesterol can influence the pathogenesis of respiratory viral infection and identify a modifiable risk factor for morbidity.

Five weeks of cholesterol diet treatment elevated total cholesterol levels in the serum prior to infection. This is consistent with previous studies showing that C57BL/6 mice have increased circulating cholesterol levels due to a high cholesterol diet (70). However, the effect of diet on circulating cholesterol dissipated with infection, likely due to the immense need for cholesterol to generate an immune response (34, 35). Likewise, cholesterol biosynthesis transcripts were downregulated at each time point examined except at day 8 p.i., when the host response to infection was greatest.

To interpret the effects of a high-cholesterol diet on infectious morbidity, it may be helpful to consider the effects of cholesterol-lowering drugs on viral infections. Statins, in particular, are cholesterol-lowering drugs that inhibit 3-hydroxy-3-methylglutaryl (HMG)–CoA reductase, the rate-limiting enzyme of cholesterol biosynthesis. Notably, statins have been shown to have antiviral effects against IAV, at least in vitro, whereby pretreatment of cells reduced viral titers and the expression of proinflammatory cytokines in IAV-infected cells (71–74). Studies on statin use in vivo, however, have produced varied results. One group proposed a combination of statin and caffeine as a potential prophylactic, as these drugs reduced lung viral titers and ameliorated lung damage in mice infected with various strains of influenza virus (75). In contrast, several studies found that statin treatment did not affect viral clearance, lung injury, or weight loss and did not improve survival in murine models of influenza infection (76–79). However, the protective effect of statins may require a pre-existing lipid metabolism disorder, as treatment was found to improve survival of infected obese mice by 40% but had no protective benefit for influenza-infected wild-type mice (80). In clinical observational studies, statin use has been associated with improved survival during severe influenza pneumonia (81) as well as SARS-CoV-2 (20). These variable results are likely due to confounding factors such as viral strain, infectious dose, mouse strain, and comorbidities, which underscores the need to further elucidate the role of cholesterol during infection if statins are to be considered as a therapeutic intervention.

Transcriptomic analysis of lung tissue revealed that dietary cholesterol alone was sufficient to increase expression of inflammatory genes in noninfected mice. Indeed, evidence suggests that cellular cholesterol accumulation can lead to inflammatory signaling under sterile conditions. For instance, macrophages that are deficient in ABCA1 and ABCG1, which mediate cellular cholesterol efflux, exhibit increased expression of chemokines CCL2 and CCL3 (37). Our dataset showed increased gene expression of several chemokines, including Ccl3, Ccl4, and Ccl5, in the lungs of cholesterol-fed mice compared with controls at day 0 p.i. (Supplemental Fig. 2B). Cholesterol may also activate inflammatory signaling pathways via recognition by the surface scavenger receptor CD36. Specifically, oxidized low-density lipoprotein has been shown to be a ligand for CD36 and can upregulate the production of chemokines and reactive oxygen species via formation of the CD36-TLR4-TLR6 heterotrimer complex under sterile conditions (82). Cholesterol accumulation can lead to formation of cholesterol crystals and lysosomal disruption, which in conjunction with TLR signaling results in activation of the NLRP3 inflammasome (83). As such, it is notable that cholesterol crystals have been found to activate the NLRP3 inflammasome in a complement-dependent manner, resulting in the production of cytokines such as TNF (84). We found that the C1q component gene C1qbp as well as Tnf were upregulated in the lungs of cholesterol-fed mice compared with controls at day 0 p.i., although Il18 and Il1b transcripts were not (Supplemental Fig. 2B).

Although cholesterol diet did not affect the number of CCR2⁺CD11b⁺ cells in the lungs of infected mice (Supplemental Fig. 3J, 3K), it increased the number of CCR2⁺CD11b⁺ cells and MHC II⁺CD11c⁺CD11b⁺ cDCs in the lungs of PBS-inoculated mice (Supplemental Fig. 3A). CCR2 is primarily considered a monocyte chemoattractant, but inflammatory cDCs have also been identified by expression of CCR2 (85). These data indicate that dietary cholesterol altered immune cell profiles in the lungs of PBS-inoculated mice. Given the effects of dietary cholesterol in the absence of infection shown in the present study, future studies may consider the contribution of CCR2⁺ myeloid cell subsets to the adverse effects of dietary cholesterol during sterile inflammatory conditions.

Polyfunctional T cells are thought to provide superior control of viral infections (67–69). However, during severe influenza disease, excessive production of proinflammatory cytokines, or a “cytokine storm,” is detrimental to host outcome (86). In the current study, dietary cholesterol increased the number of polyfunctional TNF⁺IFN-γ⁺CD4⁺ T cells but also of IFN-γ⁺CD4⁺ T cells, TNF⁺CD8⁺ T cells, and Ly6C⁺ cDCs in the lungs of infected mice. Thus, any protective benefits of polyfunctional T cell responses may have been potentially masked by the detrimental effects of excessive cytokine production. In fact, the number of several leukocyte subtypes in the lungs (granulocytes, in particular) were highly correlated with percent weight loss of mice at day 8 p.i., indicating a possible mechanism of exacerbated morbidity during respiratory viral infection to be investigated further.

Disease exacerbation by dietary cholesterol was even more pronounced among infected mice with mild to moderate disease severity as opposed to those with an acute disease phenotype. We suspect that in the case of acute morbidity, a ceiling of inflammation is reached at peak disease whereby the robust effect of infection saturates any observable effect of diet. For acute disease course, analysis at a later time point when cholesterol-fed mice exhibited delayed recovery may better demonstrate diet-induced differences in immune cell populations of the lung. These data highlight the highly dynamic effect of dietary cholesterol on exacerbating respiratory viral infection in contrast to the more prominent effects of high-fat diet and obesity models (87).

Despite overall exacerbation of disease, dietary cholesterol also increased the number of anti-inflammatory IL-10⁺CD8⁺ T cells at day 8 p.i. The cholesterol biosynthesis pathway regulates IL-10 expression in human CD4⁺ T cells (88), and thus our finding that the number of IL-10–producing T cells is altered by dietary cholesterol may reflect the link between sterol metabolism and IL-10 expression. This is also supported by the strong correlations between percent weight loss and the number of IL-10⁺CD4⁺ and IL-10⁺CD8⁺ T cells at day 8 p.i. Alternatively, increases in IL-10–producing cells may reflect the degree of inflammation and an attempt to return the tissue to a state of homeostasis.

Enhanced immune responses observed in infected cholesterol-fed mice may be considered a response to increased viral replication or persistence rather than a direct interaction with cholesterol. Cholesterol is essential for efficient influenza virus entry, replication, and budding (30, 32, 89), and thus cholesterol may affect viral infectivity. However, our RNA-seq results showed that no viral transcripts were increased in the lungs of infected cholesterol-fed mice compared with infected control-fed mice. Furthermore, infectious viral titer as determined by plaque assay clearly indicated no differences between dietary groups. As such, increased inflammatory responses observed as a result of dietary cholesterol were not likely attributable to increased lung viral titers.

To our knowledge, this is the first study to examine the effects of dietary cholesterol on IAV pathogenesis. Our data suggest that dietary cholesterol caused greater morbidity in IAV-infected mice due to an aberrant immune response in the lungs without an effect on viral load.

We thank Lisa Wetzel, Katiria Soto-Diaz, Faizah Rauther, Carl Raczka, and Isiah Ramos for technical support. We also thank Jenny Drnevich, HPCBio, and the sequencing facility within the UIUC Roy J. Carver Biotechnology Center.

The authors have no financial conflicts of interest.