Abstract
Sarbecovirus (CoV) infections, including Severe Acute Respiratory CoV (SARS-CoV) and SARS-CoV-2, are considerable human threats. Human GWAS studies have recently identified loci associated with variation in SARS-CoV-2 susceptibility. However, genetically tractable models that reproduce human CoV disease outcomes are needed to mechanistically evaluate genetic determinants of CoV susceptibility. We used the Collaborative Cross (CC) and human GWAS datasets to elucidate host susceptibility loci that regulate CoV infections and to identify host quantitative trait loci that modulate severe CoV and pan-CoV disease outcomes including a major disease regulating loci including CCR9. CCR9 ablation resulted in enhanced titer, weight loss, respiratory dysfunction, mortality, and inflammation, providing mechanistic support in mitigating protection from severe SARS-CoV-2 pathogenesis across species. This study represents a comprehensive analysis of susceptibility loci for an entire genus of human pathogens conducted, identifies a large collection of susceptibility loci and candidate genes that regulate multiple aspects type-specific and cross-CoV pathogenesis, and also validates the paradigm of using the CC platform to identify common cross-species susceptibility loci and genes for newly emerging and pre-epidemic viruses.
Main Text
Natural host genetic variation regulates disease severity following most viral infections, yet the specific genes and loci that regulate differential disease outcomes are largely unknown across susceptible species1,2. Coronaviruses (CoVs) are significant human and animal pathogens; six CoVs (three human, three swine) have emerged or expanded their geographic range in the 21st century3,4. The most impactful emergent human CoVs (SARS-CoV and SARS-CoV-2) are group 2B coronaviruses, which likely emerged from bats to cause worldwide human epidemic or pandemic respiratory infections, leading to substantial morbidity and mortality5,6. Moreover, many high-risk group 2B Sarbecoviruses (SARS-like viruses) and group 2C MERS-like bat CoVs are poised for future human emergence events7–9. The Sarbecovirus subgenus is clustered into four clades that include the clade I SARS-CoV and high-risk SARS-like Bat CoVs (BtCoV), clade II SARS-like BtCoVs like HKU3, and clade III SARS-CoV-2 (Figure 1a)10. Sarbecoviruses vary widely in their ability to cause human and animal disease11. SARS-CoV caused ~8,000 infections with a 10% mortality rate, while SARS-CoV-2 has infected ~132 million, leading to ~2.9 million deaths to date12,13. SARS-CoV-2 infections cause asymptomatic to life-threatening disease outcomes, supporting a role for inter-host genetic control of emerging viral disease outcomes in both humans and mice14–16. Thus, understanding the functions of natural host variants in genes that regulate susceptibility and disease severity after diverse Sarbecovirus infections may reveal common genetic loci that regulate wildtype and variant virus pathogenic outcomes across species, inform threat potential, and reveal novel targets for therapeutic intervention.
The angiotensin-converting enzyme 2 (ACE2) receptor interacts with the Spike protein (S) receptor binding domains of SARS-CoV, SARS-CoV-2 and many BtCoV Sarbecoviruses, but not the clade II HKU3 strain (Figure 1b)17 As many of these strains do not produce disease in mice, reverse genetics and serial passaging were used to select for SARS-CoV-MA15 (SARS-MA), HKU3-MA, and SARS-CoV-2 MA10 strains that replicate efficiently and produce severe disease in mice11,18,19,20. Mouse genetic reference populations (GRPs) have been employed as highly relevant models of human disease, and coupled with systems genetics data algorithms, to identify host susceptibility loci, genes, genetic networks and higher-level genetic interactions that regulate phenotypic variation and disease severity21–23. Among these mouse GRPs, the Collaborative Cross (CC) genetic reference population encodes over 44 million single nucleotide polymorphisms (SNPs), 4 million insertions and deletions (InDels), as well as several thousand novel variants (both SNPs as well as InDels) present in single strains24,25. Building from our prior work, genetically mapping quantitative trait loci in incipient mice from CC strains (Pre-CC26, Table 1), we designed a panel of 115 F1 crosses between different CC strains (CC-RIX; an outbred population, like humans, but reproducible (Figure 1c)) to identify loci controlling Sarbecovirus pathogenesis and adaptive immune responses. We used our mouse-adapted CoV models, including SARS-CoV MA15 (SARS-MA), HKU3-MA, and SARS-CoV-2 MA10, which replicate efficiently and produce severe disease in mice11,18,19,20, to overcome the host-specific angiotensin-converting enzyme 2 (ACE-2) interaction with CoV Spike protein.
Results
Phenotypic distributions, genomics scans, and allele effects maps for 3 traits across the CC-RIX
We infected the CC-RIX population with two genetically distinct Sarbecoviruses, which included the clade I SARS-MA and clade II HKU3-MA strains, respectively (Figure 1a-b)11,27. Groups of CC-RIX mice were inoculated with 5×103 plaque-forming units (PFU) of SARS-MA, and viral burden, clinical disease (e.g., weight loss, mortality, and respiratory function), antibody titers, and immune cell infiltrates were measured at multiple timepoints post-infection (ranging from 2-32 days) (Figure 1d, Figure 2). A matched cohort of CC-RIX was inoculated with 1×105 PFU of HKU3-MA and evaluated for mortality and weight loss through day 4 post infection (Figure 2a). In both studies, virus challenge elicited an array of disease phenotypes, ranging from clinically inapparent infection to lethal outcomes within the first 4 days of infection. We estimated genetic contributions for many of these traits and determined that heritability for these responses was 44.4%-80.9%, estimates that agree with previous CC studies26,28–31. Importantly, the various CC-RIX disease phenotypes measured in response to infection appeared relatively uncorrelated (Figure 2a-c), suggesting that there are many independent genetic factors driving these responses. We next conducted genetic mapping to identify both genomic regions and specific founder haplotypes driving various aspects of SARS-MA and HKU3-MA disease responses. Given the likely complex genetic architecture underlying these phenotypes, we identified those loci surpassing community-accepted significance thresholds (p<0.33), with distinct allele effects (Figure 2a-c, Table 1)32. We identified 11 distinct and high-confidence loci in our RIX population affecting weight loss, mortality, titer, antibody responses or respiratory function after SARS-MA infection or weight loss and mortality following HKU3-MA infection. The effect sizes of these loci varied from 3-23% of the total trait variation (that is, largely moderate effect sizes), the loci were located in different genomic regions with different causal founder alleles, and most loci primarily impacted one or a few traits in this population (Table 1). The number of independent loci and their trait-specific impacts are consistent with our earlier observations of little to no correlation between disease states across the RIX lines (Figure 2a-c). Together, this analysis highlights the genetic complexity driving CoV disease outcomes and immunity and also the inability of any single animal model of CoV disease to fully address all aspects of the disease response.
To investigate the possibility of pan-sarbecovirus susceptibility loci, we evaluated whether any of our SARS-MA phenotypes also were associated with the haplotypes driving HKU3-MA mortality (Figure 2d, Table 1). For these analyses, we binned CC-RIX lines based on their diplotypes at these HKU3-MA diseases associated loci, then determined whether these diplotype bins provided an improved fit to the relationship between SARS-MA phenotypes and the CC-RIX IDs themselves. For example, was there any genetic signal at quantitative trait loci (QTL) HrS10 or HrS11 that was associated with differential SARS-MA disease when simplifying the underlying causal model; a method we have used to find additional QTLs with small effect sizes in genetically complex populations29. HrS10 was associated with enhanced SARS-MA weight loss, disease and mortality at day 2 post-infection (2 dpi), whereas HrS11 had moderate associations with SARS-MA lung titers at both 2 and 4 dpi (Figure 2d). These results indicate that common susceptibility loci regulate disease severity across two genetically distinct Sarbecovirus clades, and that HrS10 and HrS11 are both associated with HKU3-MA-induced mortality, but likely contribute to virus-induced disease through different mechanisms.
Disease phenotypes in parental strains CC011 and CC074 and identification of a quantitative multitrait locus on chromosome 9 in the CC011xCC074-F2
Concurrent with our large CC-RIX screen, we identified a pair of inbred CC strains showing highly divergent susceptibilities to SARS-CoV: the disease resistant CC011/Unc and the highly susceptible CC074/Unc strain (CC011 and CC074 from here on, respectively). After challenge with 1×104 PFU SARS-MA, these strains exhibited marked differences in clinical and virological disease phenotypes (e.g., hemorrhage, weight loss, virus titer, mortality, circulating immune cells), and all CC074 mice developed lethal disease by 4 dpi post infection (Figure 3a, left panel). Relevant for the current pandemic, these parental strains showed similar severe infection phenotypes during SARS-CoV-2 MA10 infection (Figure 3a, right panel, Figure S1). We generated 403 F2 mice by intercrossing these strains (Figure S2) and inoculated them intranasally at 9-12 weeks with 1×104 PFU of SARS-MA. These F2 mice showed an expanded range of disease responses relative to their parent CC strains, including mortality, weight loss, titer, respiratory function, circulating immune cell and hemorrhage phenotypes (Figure S3, Figure 3b). We conducted QTL mapping in these F2 mice and identified five significant QTL segregating in this population (HrS24-28, Table 1), and supply information on other potential loci (p<0.33, Table S1). Most of these loci impacted multiple aspects of the SARS-MA disease response during this infection time course (for example analysis of HrS26 on chromosome 9 indicated the locus contributed to mortality at 4 dpi (Odds ratio (OR) of 3.15), as well as lung hemorrhage or congestion (10.2% of population variation), airflow restriction at 2 dpi (7.8% PenH), as well as peripheral neutrophil (11.8%) and lymphocyte (12.4%) levels) at 4 dpi (Figure 3c, Table 1). Recently, three Genome-wide Association Study (GWAS) in humans identified a locus associated with respiratory failure. This locus (encompassing genes such as SLC6A20, LZTFL1, FYCO1, CXCR6, XCR1, and CCR9) is syntenous with a more proximal region of our chromosome 9 locus33–35. CCR9 emerged as a strong candidate based on the integration of our data with these studies and the presence of nonsynonomous SNPs in CCR9 as well as synonymous mutations in regulatory flanking sequences.
Identification of CCR9 as a major susceptibility allele during SARS-CoV-2 infection
To better understand how our contrasting CC strains and this locus regulates SARS-CoV-2 disease, we inoculated CC011, CC074, C57BL/6NJ and CCR9-/- mice on a C57BL/6NJ background with SARS-CoV-2 MA1020. CC011 and CC074 mice infected with SARS-CoV-2 MA10 showed concordance in their SARS-MA disease response phenotypes, including lethality in the CC074 susceptible line. The CCR9-/- mice developed more severe clinical disease (Figure 4a, Figure S4a), exhibited increased virus titers (Figure 4b), weight loss, mortality and prolonged pulmonary dysfunction and severe lung pathology as measured by whole body plethysmography (Figure 4c), lung hemorrhage (Figure S4b), and lung damage (Figure S4c, d and Figure S4e) relative to their wild-type controls, supporting its important role in protection from disease. Analysis of the cytokine profile in lungs and serum by multiplex immune-assay showed increased subsets of cytokines and chemokines that are involved in promoting allergic airway inflammation, including IL9, IL13, Il17, CCL2, CCL3, CCL5, G-CSF, and Eotaxin either in the lung tissue, serum, or both (Figure 4f and Figure S5b). By flow cytometric analysis, the composition of infiltrating cells into the lung tissue (Figure 4e) and bronchoalveolar (BAL) fluid (Figure S5a) of CCR9-/- mice showed a significant increase of CD4+ T cells, CD8+ effector T cells, inflammation-promoting CD11+ DCs and eosinophils at 6 dpi, consistent with an allergic airway inflammatory response. Although originally found to play an important role in chronic gut inflammation, CCR9 is mainly expressed on lymphocytes, dendritic cells (DCs) and monocytes/macrophages36,37, and also participates in early allergic airway inflammation including the migration and proliferation of eosinophils and lymphocytes38. In addition CCR9+ DCs are implicated in regulating inflammation, alloimmunity, and autoimmunity36. CCR9-/- mice develop chronic inflammatory responses and CD11b+ inflammatory macrophages contribute to the pathogenesis of liver fibrosis via the CCR9/CCL25 axis39,40. As inflammatory macrophages contribute significantly to increased SARS-CoV pathogenesis in mice41, together, these data target the CCR9/CCL25 axis as a major driver of SARS-CoV-2 pathogenesis across species and validate CCR9 as a driver of the human Chr3 susceptibility loci and mouse Chr9 susceptibility locus.
A key goal in animal studies is to identify relevant models of human disease. Syntenic genome regions between humans and rodents often regulate a number of infectious and chronic diseases, and our analysis of HrS26 extends this important pattern42–44 to sarbecovirus infections. All told, the synteny between human Chr3 and mouse Chr9, the effect sizes of the loci identified in this F2 (between 5-15% of the population-wide phenotypic variation in this F2, Table 1), as well as the prevalence of loci impacting multiple aspects of SARS-MA associated disease, highlight how sorting of multiple susceptibility alleles into individual CC strains model unique aspects of the genetic control of disease responses. Furthermore, the presence of alleles of at least six of the CC founder strains segregating across these loci (and often in opposite directions: a C57BL/6J allele is protective at HrS26 and HrS27 but exacerbating at HrS28, Table 1) highlights the utility of using genetically complex but reproducible models of disease.
Identification of major effect locus on chromosome 4 and of Trim14 as a susceptibility gene during SARS-MA and SARS-CoV-2 MA10 infection
Next, we revisited a previous CC-F2 intercross, CC003/UncxCC053/Unc (named CC003 and CC053 from here on) conducted by our group45, and utilized our refined analysis pipelines once the original SARS-MA disease loci (HrS5-9) were statistically accounted for. This re-analysis allowed us to identify an additional locus (HrS23 on chr4) impacting both weight loss and hemorrhagic damage to the lungs as determined by gross pathology (Figure 5a, b), as well as several other suggestive QTLs (Table S1). The chr4 locus also overlapped with the mortality QTL HrS24 identified in CC011xCC074-F2 cross (Figure 5c). SNP variation between CC003 and CC053, as well as between CC011 and CC074 in this locus pointed to Trim14 as a likely candidate gene driving these differences in SARS-CoV disease. Previous work identified Trim14 as a key docking platform in the context of MAVS signaling46,47. We used CRISPR/Cas9 targeting to edit Trim14 in C57BL/6J mice, create a functional knockout (Figure S6), and evaluate its role following infection (Figure 5d, top panel). Trim14Δ47/Δ47 mice inoculated with 1×105 PFU of SARS-MA had a modest increase in pathogenesis relative to C57BL/6J control mice. At 3 and 4 dpi, Trim14-deficient mice had increased weight loss, which corresponded with increases in viral titer within the lung at 2 and 4 dpi. This result show that an absence of Trim14 affects viral clearance. Similarly, Trim14Δ47/Δ47 mice inoculated with SARS-CoV-2 MA10 also sustained modest increases in weight loss and a delayed recovery phenotype when compared to C57BL/6 mice (Figure 5d, bottom panel). However, the difference in viral titer seen at early times post SARS-MA infection was not observed with SARS-CoV-2 MA10. Together, these data suggest that Trim14 has a shared role in attenuating Sarbecovirus disease potential, but that this effect varies between viruses. Taken together, the data demonstrates the potential of identifying common and shared QTLs among group 2B coronaviruses.
Discussion
Across these studies, we describe several dozen loci impacting different disease responses, including several which show broad responses to all tested Sarbecoviruses. We demonstrate connections to human GWAS studies, and as such these data represent a resource for future comparative studies of Sarbecovirus pathogenesis between humans and animals. Our study highlights the power of using animal GRPs to understand the role of host genetic variation on infectious diseases, generate new models of differential disease, probe the role of individual genes in disease progression, and provide mechanistic insight into the role of specific host genes and viral strains in regulating pathogenesis across species. In appropriately selected large population screens, highly penetrant genetic variants can be identified easily, as can their impacts on specific aspects of disease outcome. In contrast, targeted mapping crosses between highly discordant strains can help to identify more complex genetic interaction networks such as variants that are penetrant only in the context of specific genetic backgrounds, or epistatic (gene-gene) interaction networks.
We leveraged large-scale population mapping as well as focused intercrosses to better characterize the genetic susceptibility landscape of Sarbecovirus infections in mouse models26,45 and demonstrated that a large number of polymorphic loci (Figure 6) regulate the host disease responses to this subgroup of coronaviruses. Moreover, in this study and others, we have identified specific genes (Trim5526, Ticam245, and here Trim14 and CCR9), which have naturally occurring polymorphisms driving aberrant SARS-CoV disease responses. Importantly, we demonstrate that HrS10 and HrS11 influence disease severity following both clade I SARS-MA and clade II HKU3-MA infection in the CC-RIX, supporting the hypothesis that intrinsic virulence properties encoded within the Sarbecoviruses are subject to similar susceptibility loci in mammals. Further, the concordant susceptibility profiles of CC011, CC074, Trim14 and CCR9 deficient mice with SARS-MA and SARS-CoV-2 MA10 highlight the utility of pre-emergence disease models. Such findings are consistent with the discovery that group I and II human norovirus infection and pathogenesis are heavily regulated by polymorphisms in fucosyltransferase 2 (FUT2)48. Rich and complex datasets like the ones described here enable comparisons with human GWAS studies mapping QTL after SARS-CoV-2 infection33–35. The CC platform can be used to evaluate the role of these loci in mouse models. One advantage of our approach is the use of different mapping platforms, which provide opportunities to combine datasets across projects to gain a greater understanding of the role of how host genetic variation modulates CoV disease responses in mammals (Figure 6, Table 1). In addition, the unexplained heritability and suggestive loci (Table S1) we have identified, suggests that CoV disease and immunity are complex polygenic traits, with the accumulation of variants across many loci driving final disease susceptibility. Collectively, these studies represent the most comprehensive analysis of susceptibility loci for an entire genus of human pathogens, identify a large collection of susceptibility loci and candidate genes that regulate multiple aspects type-specific and cross-CoV pathogenesis, validate a role for the CCR9-CCL25 axis in regulating SARS-CoV-2 disease severity and provide a resource for community wide studies.
Supplemental Figure Legends
Author contributions
A.S., L.E.G., F.P.M.V., S.K.M., M.T.H., V.D.M., M.T.F., R.S.B designed screens and biology experiments; A.S., L.E.G., S.R.L., V.D.M. conducted and characterized infections in vivo; A.S., L.E.G., S.R.L., E.S.W., K.L.J., D.T.S. performed in vitro studies and immune cell quantifications; A.S., L.E.G, S.R.L., R.L.G., S.A.M., V.D.M., M.T.F. processed and analyzed data, and generated figures; A.S., B.K.H, M.A.M., S.J., S.C., S.K.M., M.T.F. performed genetic mapping studies; L. A.V., L.B.T., M.S.D. designed and isolated Trim14-deficient mice; A.S., L.E.G., V.D.M. performed in vivo evaluation of Trim14-deficient infected animals; S.A.M. scored pathologic changes in the lungs of infected mice; R.L.G., S.A. isolated recombinant viruses; T.A.B., P.H., G.D.S., D.R.M. produced CC-RIX and F2 animals, and processed samples for genotyping; L.E.G., S.R.L., B.K.H, M.A.M., K.L.J., S.J., T.A.B., L.B.T., D.R.M., G.D.S., M.S.D., F.P.M.V., S.K.M., M. T.H. edited the manuscript; A.S., M.T.F., R.S.B. wrote the manuscript.
Competing interests
The authors have no competing interests.
Online Material and Methods
Cells and viruses
Recombinant mouse-adapted SARS-CoV MA15 (SARS-MA), HKU3-SRBD-MA (HKU3-MA), and SARS-CoV-2 MA10 (SARS-2-MA) virus were generated as described previously (HKU3-SRBD-MA: GenBank Accession Number XXX, SARS-CoV-2 MA10: GenBank Accession Number XXX)11,18,20. For virus titration, the caudal lobe of the right lung was homogenized in PBS, resulting homogenate was serial-diluted and inoculated onto confluent monolayers of Vero E6 cells (ATCC XXX), followed by agarose overlay. Plaques were visualized with overlay of Neutral Red dye on day 2 (SARS-MA) or day 3 (SARS-CoV-2 MA10) post infection.
Mouse studies and in vivo infections
All mouse studies were performed at the University of North Carolina (Animal Welfare Assurance #A3410-01) using protocols approved by the UNC Institutional Animal Care and Use Committee (IACUC). Animal studies at Washington University were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the IACUC at the Washington University School of Medicine (Assurance number A3381-01).
Mouse studies fall into three major classes: CC-RIX, F2 intercross mice, and inbred wild-type or gene-edited mice. The laboratory of Pardo Manuel de Villena (FPMV) purchased CC mice from the Systems Genetics Core Facility at UNC between 2012 and 2018. These CC mice were used to breed CC-RIXs in the FPMV laboratory, to ensure proper cohorts and batch sizes. CC-RIXs were generated in a ring design such that each CC-RIX had one copy of the MHC H2Bb allele, and that each CC strain was used as both dam and sire in equal proportion across all RIXs. Mice (~105 CC-RIX strains, 3 animals each) were transferred at 5-6 weeks of age to the Baric (RSB) laboratory for infection between 9-12 weeks of age. The details of the CC003 x CC053 F2 are published45. The Systems Genetics Core Facility was contracted to generate the F2 cross between CC011 and CC074. F1 mice between CC011 and CC074 were generated in both potential cross directions, and F2 mice were bred in all 4 possible F1 x F1 combinations, to ensure appropriately balanced sex chromosome and parent-of-origin effects. F2 mice (226 males, 177 females) were weaned such that littermates were randomized to different experimental cages to further reduce litter- or batch-effects on the study, and mice were transferred at 5-6 weeks of age to the RSB laboratory for infection between 9-12 weeks of age.
15- week old CCR9-/- mice (strain 027041) and 15-week old female C57BL/6NJ mice (strain 005304) were purchased from Jackson Laboratory. CC-RIX, CC-F2 mice, Trim14-deficient, and CCR9-/- mice were infected with 5×103 (CC-RIX with SARS-MA), 1×104 (CC-F2 with SARS-MA and SARS-CoV-2 MA10), and 1×105 (CC-RIX with HKU3-MA, Trim14Δ47/Δ47 and CCR9-/- mice with SARS-MA and SARS-CoV-2 MA10) plaque forming units (PFU) in 50 μl PBS intranasally at 9-12 (CC-RIX and CC-F2 mice) or 15 (CCR9-/- and C57BL/6NJ) weeks of age, respectively. Body weight, mortality, and pulmonary function by whole body plethysmography49 were monitored daily where indicated. At indicated timepoints, mice were euthanized and gross pathology (hemorrhage score) of the lung was assessed and scored on a scale from 0 (no hemorrhage) to 4 (severe hemorrhage affecting all lung lobes). Then lung tissue was harvested for titer and histopathology analysis; and blood samples were harvested to determine antibody composition and for analysis of peripheral immune cells. Samples were stored at −80°C until homogenized and titered by plaque assay as described above. Serum was prepared and SARS-CoV spike-specific antibody were quantified by ELISA as previously described33. Peripheral blood was diluted 1:5 in PBS/EDTA and analyzed with the VetScan HM5 as previously described 50. Histopathology samples were fixed in 10% phosphate buffered formalin for 7 days before paraffin embedding, sectioning stained with hematoxylin and eosin.
Generation of Trim14-deficient mice
Gene-edited Trim14-deficient mice were generated with support from the Genome Engineering and iPSC center and Department of Pathology Micro-Injection Core (Washington University School of Medicine). A sgRNA targeting exon 4 of Trim14 was selected based on minimal off-target effects in silico and targeting efficiency in vitro. The sgRNA (5’-ACCAATGGACACTCGCCTGANGG-3’) was synthesized, transcribed (HiScribe T7 In vitro Transcription Kit, New England BioLabs), and purified (MEGAclear Transcription Clean-Up Kit, Thermo Fisher). The sgRNA was mixed and co-injected with Cas9 RNA at 5ng/μl and 10ng/μl final concentrations into half-day-old C57BL/6J embryos (E0.5). After next-generation sequencing of founders and two generations of mice backcrossed to C57BL/6J mice, a mouse line with a 47-nucleotide deletion (5’-GCCTGAAGGAAAGTGAGTTGCCTAAGACCAACTCCAAGTCCTTGCTC-3’) encompassing the 3’ splice site of intron 4 and part of the coding region of exon 4 was generated. These Trim14Δ47/Δ47 mice were bred as homozygotes and used for experiments. Trim14Δ47/Δ47 mice were born in normal Mendelian frequencies and showed no apparent defects in development, growth, or fecundity. Lung tissue from Trim14Δ47/Δ47 were found to lack detectable Trim14 mRNA, likely due to nonsense-mediated decay, as measured by RT-qPCR using a predesigned primer/probe set for Trim14 (IDT, Assay ID Mm.PT.58.286730) and the housekeeping gene GAPDH (IDT, Assay ID Mm.PT.39a.1). Sanger sequencing of a polymerase chain reaction amplicon [5’ GGCACAGCTCAACCCATGG −3’ (forward) and 5’-ACCAGCGAGCTCGTGCTCC −3’ (reverse)] was used for genotyping.
Flow cytometry analysis of immune cell infiltrates
For analysis of BAL fluid, mice were sacrificed by ketamine overdose, followed by cannulation of the trachea with a 19-G canula. BAL was performed with three washes of 0.8 ml of sterile PBS. BAL fluid was centrifuged, and single cell suspensions were generated for staining. For analysis of lung tissues, mice were perfused with sterile PBS, and the right inferior lung lobes were digested at 37°C with 630 μg/mL collagenase D (Roche) and 75 U/mL of DNase I (Sigma–Aldrich) for 2 h. Single cell suspensions of BAL fluid and lung digest were preincubated with Fc Block antibody (BD PharMingen) in PBS + 2% heat-inactivated FBS for 10 min at room temperature before staining. Cells were incubated with antibodies against the following markers: efluor506 Viability Dye (Thermo Fisher, 65-0866-14), BUV395 anti-CD45 (Clone 30-F11, BD Biosciences), BV711 anti-CD11b (Clone M1/70, Biolegend), APC-Cy7 anti-CD11c (Clone HL3, BD Biosciences), BV650 anti-Ly6G (Clone 1A8, Biolegend), Pacific Blue anti-Ly6C (Clone HK1.4, Biolegend) FITC anti-CD24 (Clone M1/69, Biolegend), PE anti-Siglec F (Clone E50-2440, Biolegend), PE-Cy7 anti-CD64 (Clone X54-5/7.1, Biolegend), AF700 anti-MHCII (Clone M5/114.15.2, Biolegend), BV421 anti-CD3 (Clone 17A2, Biolegend), BV785 anti-CD4 (Clone GK1.5, Biolegend), APC anti-CD8a (Clone 53-6.7, Biolegend) BV421 anti-B220 (Clone RA3-6B2, Biolegend) APC-Cy7 anti-CD44 (Clone IM7, Biolegend) BV605 anti-CD62L (Clone MEL-14, Biolegend). All antibodies were used at a dilution of 1:200. Cells were stained for 20 min at 4°C, washed, fixed and permeabilized for intracellular staining with Foxp3/Transcription Factor Staining Buffer Set (eBioscience) according to manufacturer’s instructions. Cells were incubated overnight at 4°C with BV421 anti-Foxp3 (Clone MF-14, Biolegend) washed, re-fixed with 4% PFA (EMS) for 20 min and resuspended in permeabilization buffer. Absolute cell counts were determined using Trucount beads (BD). Flow cytometric data were acquired on a cytometer (BD-X20; BD Biosciences) and analyzed using FlowJo software (Tree Star) (Figure S7).
Cytokine and chemokine protein analysis
The small center lung lobe of each mouse was homogenized in 1 ml of PBS and briefly centrifuged to remove debris. Fifty microliters of homogenate were used to measure cytokine and chemokine protein abundance using a Bio-Plex Pro mouse cytokine 23-plex assay (Bio-Rad) according to the manufacturer’s instructions.
Lung pathology scoring
Two separate lung pathology scoring scales, Matute-Bello and Diffuse Alveolar Damage (DAD), were used to quantify acute lung injury (ALI)51.
For Matute-Bello scoring samples were blinded and three random fields of lung tissue were chosen and scored for the following: (A) neutrophils in alveolar space (none = 0, 1–5 cells = 1, > 5 cells = 2), (B) neutrophils in interstitial space (none = 0, 1–5 cells = 1, > 5 cells = 2), (C) hyaline membranes (none = 0, one membrane = 1, > 1 membrane = 2), (D) Proteinaceous debris in air spaces (none = 0, one instance = 1, > 1 instance = 2), (E) alveolar septal thickening (< 2Å~ mock thickness = 0, 2–4Å~ mock thickness = 1, > 4Å~ mock thickness = 2). Scores from A–E were put into the following formula score = [(20x A) + (14 x B) + (7 x C) + (7 x D) + (2 x E)]/100 to obtain a lung injury score per field and then averaged for the final score for that sample.
In a similar way, for DAD scoring, three random fields of lung tissue were scored for the in a blinded manner for: 1= absence of cellular sloughing and necrosis, 2= uncommon solitary cell sloughing and necrosis (1–2 foci/field), 3=multifocal (3+foci) cellular sloughing and necrosis with uncommon septal wall hyalinization, or 4=multifocal (>75% of field) cellular sloughing and necrosis with common and/or prominent hyaline membranes. To obtain the final DAD score per mouse, the scores for the three fields per mouse were averaged.
Genotyping
CC003, CC053, their F1 progeny, and the F2 cross were genotyped as previously described45. CC011, CC074, their F1 progeny, and the F2 cross were genotyped on the MiniMUGA genotyping array52. Genomic DNA was isolated from tail-clips of animals using the Qiagen (Hilden, Germany) DNeasy Blood & Tissue kit. 1.5 μg was sent to Neogen (Lincoln, Nebraska) for processing. We filtered the genotypes upon return for informativeness within this cross. To be considered informative, the marker had to have one homozygous allele in all CC011 mice genotyped, the alternate homozygous allele in all CC074 mice genotyped, and the appropriate call in all F1 animals (H calls on the autosomes, an H call in females on the X chromosomes, and the relevant homozygous call in male F1s). This filtering reduced the ~10,800 SNPs on the MiniMUGA array to 2821 informative markers.
QTL mapping and statistical analyses
For the CC-RIX, we used the same pipeline we previously described29. Briefly, each CC-RIX had their genome represented as an array of probabilities of each of the 8 CC founder haplotypes (Figure 1C). This array was used in the DOQTL R package53 to run an 8-allele regression at each of 77,000 markers for our CC-RIX phenotypes. At each marker, a LOD score is calculated describing the goodness of fit of our trait~genotype model relative to a null model. Significance was determined by running 1000 permutations scrambling the relationship between phenotypes and haplotypes. In this way, significance is independent of both population allele frequencies, as well as the phenotypic distribution. For the F2 crosses, instead of a regression on haplotype probabilities, the R/QTL package conducts a regression of the trait of interest on the exact genotypes at each locus54. As with the CC-RIX mapping, permutation testing is used to identify significance.
Acknowledgments
This study was supported by grants in aid from the National Institutes of Health, Allergy and Infectious Diseases (AI100625 and AI149644 to R.S.B., M.T.H., M.T.F., and F.P.M.V.), (R00AG049092 to V.D.M.), a contract from the NIH (HHSN272201700036I; Task Order #38 75N93020F00001 to R.S.B), and support from NCATS (UL1TR002369 to M.A.M. and S.K.M.). We would like to thank the Systems Genetics Core Facility (UNC) for maintaining and distributing Collaborative Cross mice. The research was also supported by a generous gift from the Chan-Zuckerberg foundation.
Footnotes
↵* Co-first authors
+ Co-senior authors