Common Heritable Immunological Variations Revealed in Genetically Diverse Inbred Mouse Strains of the Collaborative Cross

Interindividual variations for many traits, such as eye color, height, and body weight, are readily observed in healthy individuals. Similarly, the proportion, function, and phenotype of immune cell subsets also demonstrate interindividual variations (1–3). These variations are shaped both by heritable and nonheritable factors, with the degree of contribution varying for each immunological trait (3–7). For instance, some NK cell traits are highly heritable, whereas other T cell and dendritic cell (DC) traits are predominantly influenced by nonheritable factors (1, 5, 6, 8–10). Of interest, relevant parallels between mice and humans have been made regarding the genetic control of given traits and phenotypes, such as disease susceptibility (11–14). Indeed, despite notable differences between the two species (15), mice evidently share genetic similarities with humans and can be useful in defining the contribution of genetic elements shaping the immune system (16, 17).

Genetic studies in mice typically exploit two parental inbred strains, such that genetic diversity in these studies is limited. Circumventing limitations associated with a priori selection of two inbred mouse strains for genetic analyses, the Collaborative Cross (CC) has emerged as a powerful genetic resource (18, 19). The CC comprises dozens of new inbred strains descended from eight diverse founder strains, harnessing over 90% of common genetic diversity of the mouse species, which can be exploited to study biological traits (19, 20). The CC has already proved useful in identifying genes mediating different phenotypes, including defining the proportion of some immune cells and other hematological parameters (21–23), the response and susceptibility to infections (24–27), and a wide range variety of other traits, including metabolism, cancer development, and immunological traits such as contact hypersensitivity and IgG glycosylation (28–40). In this study, we expand on these studies by characterizing 18 immunological traits, including subsets of NK cells, DCs and T cells. We confirm previous findings of specific loci linked to T cell traits, including the CD4/CD8 T cell ratio, and we define new loci linked to other immune cell traits. More importantly, we find that, as for humans, some NK cell traits are highly influenced by heritable factors, whereas we did not observe significant genetic linkage for given T cell and DC traits that are predominantly influenced by nonheritable factors in humans. Our results support the use of inbred mouse strains to identify both heritable and nonheritable influences determining the variability of immunological traits.

All experimental and animal handling activities were performed in accordance with the guidelines of the Institutional Animal Ethics Committee and the Association for Research in Vision and Ophthalmology Statement for the Use of Animals. The principles, development, and initial characterization of the CC have been described (18, 38). Mice were bred at the Animal Resource Centre (Perth, WA, Australia) and were generously provided by Geniad (38). Male and female mice were used at 5–8 wk of age, housed under a 12-h light/dark cycle, and given a standard diet with free access to food and water. During the analysis of the CC strains, C57BL/6 (B6) mice were periodically included to ensure experimental reproducibility. The CC strains used in this study are listed in Table I. The phenotypes were assessed on two to eight mice per CC strain to account for intrastrain variability. Two strains were not phenotyped for DCs, and one strain was only phenotyped for DC traits. The eight founder strains were analyzed separately from the CC strains at the Maisonneuve-Rosemont Hospital (MRH) Research Center. Specifically, A/J (no. 646), B6 (no. 664), 129S (no. 2448), NOD (no. 1976), NZO (no. 2105), CAST (no. 928), PWK (no. 3715), and WSB (no. 1145) were ordered from The Jackson Laboratory and housed under specific-pathogen–free conditions at the MRH Research Center. B6 and NOD mice were maintained on-site by brother–sister mating, and the six other founder strains were sacrificed at least 1 wk after reception. Five- to nine-week-old mice were used for all phenotypic analyses. Two males and two females of each founder were included, except for B6 mice, of which one male and four females were included. No immunophenotypic differences between males and females were observed (data not shown). The MRH ethics committee, overseen by the Canadian Council for Animal Protection, approved the experimental procedures.

Spleens were treated with collagenase (1 mg/ml in PBS, Type V from Clostridium histolyticum; Sigma-Aldrich) for 15 min at 37°C and passed through a 70-μM cell strainer to yield single-cell suspensions prior to staining with Abs. Single-cell suspensions were stained with the following Abs: CD3 (145-2C11), CD4 (GK1.5), CD8 (53-6.7), CD11c (N418), CD19 (6D5), CD27 (LG.3A10), CD49b (DX5), B220 (RA3-6B2), and TCR β (H57-597) from BioLegend; CD11b (M1/70) from BD Pharmingen; mPDCA-1 (eBio927) from eBioscience; and mCD1d:PBS57 tetramer-PE from the National Institutes of Health tetramer core facility. Data were acquired using a BD FACSAria II or a BD LSRFortessa X-20 and analyzed using FlowJo software. The data provided in this article as Supplemental Table I have been deposited to the Mouse Phenome Database as project MPD:Lesage1 (Lesage, S., R. Collin, L. Balmer, and G. Morahan. Immunological variation in 77 Collaborative Cross strains of mice. MPD:Lesage1. Mouse Phenome Database web resource [RRID:SCR_003212], The Jackson Laboratory, Bar Harbor, ME; https://phenome.jax.org).

We used the previously described GeneMiner mapping tools (35) (http://130.95.9.22/Geniad2/) to obtain genome-wide logarithm of the odds (LOD) score results. Mean values of cell proportions for each CC strain and founder strain were entered. Genotypes of the latest-generation strains were used when available. Including the eight founder strains, a total of 71 strains for DC subsets and 72 strains for NK and T cell subsets were analyzed for genetic association using the MugaQTL (normalize) function. Between 242 and 251 mice were included in the phenotype analyses. To obtain empirical p values, sets of 1000 permutations were employed for each trait.

Broad-sense heritability (41) was calculated using the formula H² = V_G/V_P = V_G/(V_e + V_G), where V_P is the total phenotypic variance for a given trait, V_G is the variance explained by genetic components, and V_e is the variance explained by environmental components. The variance is the square of the SD. For each trait, V_G was calculated as the mean variance between the replicates of the same strain for the eight CC founder strains, and V_e + V_G was calculated as the variance between all strains, including the CC strains and the eight founders, for the mean value of the corresponding phenotypes. The maximal value for heritability is one.

A one-way ANOVA with Bonferroni post hoc test was performed to determine differences between founder strains, and p < 0.05 was considered significant. Founders with the highest and lowest proportion of each cell subset are summarized in Supplemental Table II. A heatmap of an unbiased hierarchical cluster analysis of the 18 immunophenotypes was plotted for the CC and founder strains. A principal component analysis (PCA) was performed on CC and founder strains using the log-transformed values of each immunophenotype. The biplot was generated to indicate the weight of the immunophenotypes driving the variability explained by the first two components. The biplot was generated using R 3.3.0.

We used three well established multiparameter flow cytometry panels to define 18 immunological phenotypes in the spleen (42–47). The gating strategy for these 18 phenotypes is depicted in Fig. 1, where each phenotype is numbered 1–18. Specifically, for NK cells, we included seven phenotypes: total mature NK (mNK) cells and the four functional maturation stages, as well as total pre-mNK cells and the proportion of pre-mNK cells among total NK cells (Fig. 1A). For DCs, we included plasmacytoid DCs, conventional DCs (cDCs), as well as three cDC subsets, namely cDC1, cDC2, and merocytic DCs, for a total of five phenotypes (Fig. 1B). Finally, the T cell panel included six phenotypes, namely NKT cells, total T cells, CD4⁻CD8⁻ TCRαβ⁺ T cells, CD4⁺ and CD8⁺ T cells, as well as the ratio of CD4⁺ to CD8⁺ T cells (Fig. 1C).

Prior to undertaking the analysis of the phenotypes in the CC strains, we characterized the eight founder strains: A/J, B6, 129S, NOD, NZO, CAST, PWK, and WSB. The 18 immunological phenotypes were successfully defined in all eight founder strains (Supplemental Fig. 1). Analysis of these traits revealed a wide range of variation for most NK cell and T cell traits but less so for cDC traits (Fig. 2, Supplemental Table I). In particular, the proportion of total NK cells was highly increased in NZO mice compared with the other seven founders. CAST and NOD mice exhibited low and high T cell numbers relative to the other strains, respectively. The strain-to-strain variation of the immune cell subsets was maintained when it was quantified in absolute numbers (Supplemental Fig. 2). These results suggest that genetic variations likely have major effects in regulating proportions of lymphocyte populations in the spleen.

To define such genetic variations, we quantified these traits in the CC strains listed in Table I. All 18 traits in the CC exceeded the range of variation observed in the founder strains, a phenomenon known as genetic transgression (Figs. 2, 3). For instance, the proportion of cDC1 in the founder strains ranged between 6.4% (PWK strain) and 24.1% (B6 strain), representing a variation of <4-fold, whereas a variation of 25-fold was observed in the CC strains, with cDC1 ranging from 1.8 to 45.4%. For many phenotypes, the majority of the CC strains were outside of the phenotypic range determined by the founder strains. Notably, among the founder strains, NZO mice exhibit both the highest and lowest proportion of CD27⁻CD11b⁻ and CD27⁺CD11b⁺ mNK cells, respectively, at 22 and 10% (Figs. 2, 3). Still, 70% of CC strains had a higher proportion of CD27⁻CD11b⁻ mNK cells than NZO mice, and 54% of CC strains have a lower proportion of CD27⁺CD11b⁺ mNK cells than NZO mice (Figs. 2, 3). Similarly, 70% of CC strains had a lower proportion of cDC2 than NZO mice, which, among founder strains, showed the lowest proportion of cDC2 (Figs. 2, 3). For some traits, most of the CC strains exhibited phenotypes in the range observed between the founder strains. For example, NK cells from founder strains ranged from 0.8% in CAST to 6.1% in NZO, with only three CC strains falling outside this range (Figs. 2, 3). These observations of transgressive segregation suggest that the traits result from interaction among multiple genetic variants inherited independently by the CC strains.

The data from the phenotypic analysis of the 18 traits were compared with the genotypes of each CC and founder strain. Genetic associations were plotted as LOD scores. LOD scores exceeding the genome-wide significant threshold of 0.001, corresponding to LOD scores above 7.0, were observed for eight traits, namely total NK cells, pre-mNK cells, CD27⁺CD11b⁻ mNK cells, total cDC, cDC1, CD4⁺ T cells, CD4⁺/CD8⁺ T cell ratio, and invariant NKT cells (Fig. 4). Analysis of founder coefficients at the loci associated with these phenotypes revealed which founder alleles were important in determining each trait (Fig. 5). For instance, CAST and NZO mice exhibit the lowest and highest proportion of NK cells, respectively (Fig. 2), and the genetic variation from both CAST and NZO founder strains importantly contributes to the genetic linkage on chromosome 16 (Fig. 5). Indeed, alleles from the founder strain that exhibit the highest and/or lowest proportion of a given immune cell type were linked to similar variations in the proportion of these cells in the CC strains for five of the traits, namely NK cells, pre-mNK cells, cDC1, CD4⁺ T cells, and invariant NKT cells. These five immunological traits are, thus, highly driven by specific allelic variations. In contrast, three traits (CD27⁺CD11b⁻ mNK cells, total cDC, and CD4⁺/CD8⁺ T cell ratio) are likely to be regulated by more complex genetic interactions.

Interestingly, the locus on chromosome six linked to cDC1 contains the Cd8 gene. As CD8α expression is used to identify cDC1, modulation of CD8α expression itself could lead to changes in the proportion of cDC1. To test this hypothesis, we compared the expression of CD8α on T cells and cDCs in all CC strains. Whereas some CC strains presented with very low levels of CD8α expression on cDCs, we did not observe a parallel decrease for CD8α expression in T cells (Fig. 6). This result suggests that either Cd8 is an unlikely candidate gene for defining cDC1 proportion or it is differentially regulated in T cells and cDC1.

In addition to genetic interactions, each immunological trait may vary as a consequence of nonheritable factors (41). Although mice held in laboratory facilities are subjected to fewer nonheritable traits than wild mice, there remains a possible significant influence of nonheritable factors in specific-pathogen–free facilities. Measuring heritability (41), we found high heritability for all traits, with eight having H² values exceeding 0.9. The lowest heritability score of 0.58 was shown by cDCs (Table II). These results suggest that variation in these immunological traits in the given experimental conditions was mostly driven by heritability.

Variations in immune cell populations are often thought to be reciprocal; that is, if the proportion of an immune cell population is increased, it is expected that the proportion of a different immune cell population must be decreased. If this hypothesis is true, one would expect some loci to influence more than one phenotype. However, we did not find this to be the case. For instance, loci on chromosomes 2 and 17 are linked to CD4⁺ T cell proportion, but no significant linkage was observed for CD8⁺ T cells (Figs. 4, 5). This can in part be explained by the fact that CD4⁺ and CD8⁺ T cells are not binary; there are multiple subsets of T cells, each regulated by different heritable and nonheritable factors. Therefore, our immunogenetic analyses suggest that immune cell traits are regulated independently. This is consistent with recent studies demonstrating that human immunophenotypes exhibit a continuous distribution rather than discrete values (1, 8). To determine whether the CC mouse strains also exhibited a continuous distribution of immunophenotypes, we performed a PCA for the CC and founder mouse strains with the 18 phenotypic traits. The PCA (Fig. 7A) did not reduce the 18 immunophenotypes to one or two components that would represent most of the variability. The first two components explained only 21 and 17% of the phenotypic variance, respectively. Moreover, there was no obvious clustering of the strains, illustrated by a relatively uniform distribution on the plot. This is confirmed by an unbiased hierarchical analysis, in which, again, specific clusters could not be delineated (Fig. 7B). Taken together, our findings support the view that the 18 immunological traits are independently genetically regulated and, as for humans, represent a continuous distribution.

The parameters regulating the proportion of various immune cell populations have been studied in humans and mice from both genetic (3, 5, 48–50) and nongenetic (environment, diet, age, etc.) perspectives (4, 6, 51–54). Although environmental factors can shape the immune system, intrinsic host factors, such as the genetic background, arguably explain a larger proportion of the immune variation observed in humans and mice. The CC strains are an invaluable tool for the identification of loci linked to different phenotypic traits. Previous studies and a proof-of-concept for the GeneMiner mapping software (35) successfully exploited the CC strains to map genes mediating a wide range of phenotypes. Additionally, the number of candidate genes could be greatly reduced based on the founder haplotype effects, and in many cases the causative genetic variant in the responsible gene could be identified (21–25, 30, 33, 39, 40). For instance, Phillippi et al. (22) used a nine-color flow cytometry panel to characterize various immune cells, with emphasis on B cells, and found a polymorphism in Fcer2a, encoding for CD23, linked to CD23 expression. Krištić et al. (39) also defined missense single-nucleotide polymorphisms affecting the nature of glycans that modify IgG glycosylation.

In this study, we successfully identified loci already known to be linked to the proportion of specific immune cells. Notably, a locus on chromosome 17 was linked to the CD4/CD8 T cell ratio (48, 55–57). Interestingly, this locus was not identified in a previous CC strain analysis by Phillippi et al. (22). The reason for this discrepancy is unclear and may be attributed to the differences in the mouse cohorts used in the study. Whereas they used F1 and pre-CC strains, we characterized inbred CC strains, which may affect the diversity of the tested genotypes. We also tested multiple mice per strain, whereas they tested only one mouse per pre-CC strain. Importantly, most of the loci we mapped regulating immunological traits have not been previously identified, including in previous studies using crosses between common inbred strains. For example, genetic control of the total NK cell population was previously investigated in F2 cohorts (crosses of B6.RAG^o/o × NOD.RAG^o/o and B6.H2^g7 × NOD), finding relevant loci on mouse chromosomes 8, 9, and 17 (46, 58). Linkage in the CC and founder strains identified a new locus on chromosome 16 for this same trait. Similarly, a previous study linked the distal region of chromosome 7 to the control of pre-mNK cells in B10.BR × NOD.H2^k F2 outcross (42), whereas the CC strains reveal a locus on chromosome 16 linked to this trait. These differences are due to the presence of a higher diversity of alleles in the CC strains that are absent in selected F2 outcrosses and that likely have a dominant genetic impact in determining the phenotypes (59).

Many loci reported to be linked to specific immune traits did not reach the significant LOD score threshold in the current study. This may be explained by the relatively small size of our study, which included only 63 or 64 CC strains per phenotype. Although it was initially intended that more than 1000 CC strains would be generated, infertility problems and extinction of many strains decreased the number of available CC strains (60). The current study of ∼64 CC strains plus the 8 founder strains has sufficient power to identify major effect genes (34) but not loci with minor impact on multigenic traits. Alternatively, the variations could be driven by nonheritable factors. Although mice are subjected to fewer nonheritable influences than humans because of a controlled diet and environment, they are still influenced by many nonheritable factors, including somatic mutations, epigenetic changes, age, seasonal changes, microbiota, etc. In this study, we controlled for age by exclusively testing 5–8-wk-old mice. We also controlled for sex, a heritable variable, by including an equal number of mice of both sexes in the analysis of the founder strains (see 2Materials and Methods). Although this does not preclude the impact of other nonheritable factors in modulating the immunological traits, we found that all immunological traits tested were highly heritable in our experimental setting and that nonheritable factors have little impact on variation of most traits we studied.

The genetic diversity of the CC strains and their multiparental origin is closer to what is seen in human populations than the two-founder approaches used in conventional mouse genetics. The range of variability observed in the phenotypes of the CC strains was also more comparable to variability among healthy humans, as has been shown for regulatory T cells (Tregs) (23). Although the conclusions vary between studies concerning the degree of genetic heritability of human immune system components, variations in some immune population frequencies and phenotypes seem to be more affected by the environment, whereas others are more affected by genetics (3–7). Interestingly, the highest genetic associations in the current study were linked to immune parameters that also have a strong genetic component in humans. For example, NKT cell frequencies vary considerably among healthy individuals (61) and have been suggested to have a strong genetic heritability in some studies (5, 6), although not in all studies (7). Similarly, HLA-DR⁺ NK cells and CD56-bright NK cells, which are similar to the mouse pre-mNK population (62), had a stronger heritability than other NK subsets in a twin study (6) and were shown to be associated with human loci in genome-wide association studies of healthy individuals or twin cohorts (3, 5, 10). Early NK maturation stages have also been associated with genetic components (3). This is in agreement with our finding that NK functional maturation is significantly associated with genetic parameters in the CC strains. Another population for which human studies showed strong genetic determinants are the proportion and subsets of Tregs (5, 7). Although we did not test for Treg proportion, a previous study using CC-RIX strains did show strong association of Treg subsets to chromosome X loci (23).

In conclusion, our results add to those of others and validate the ability of a relatively small cohort of CC strains to map loci affecting complex traits and reveal potential candidate genes. Follow-up studies investigating differential expression of the candidate genes in the founder strains and the impact of their deficiency in mouse models could help in the development of therapeutic strategies to manipulate the immune system. Beyond quantitative trait loci associations, the CC strains will also prove useful in the future for the study of correlations between different phenotypes, for example between immune cell frequencies and disease susceptibilities. The CC strains overall represent an important community resource that has only begun to be exploited and is likely to bring important new discoveries.

We thank Kevin Li for assistance running the Flow Cytometry Centre at Harry Perkins Institute of Medical Research (Perth, WA, Australia). We also thank Gabrielle Boucher for expert statistical advice and Geniad Pty. Ltd. for mice.

The authors have no financial conflicts of interest.