Given reported species-specific differences in immune responses^1–8 we wished to compare memory CD8 T cell distribution and differentiation between mice and humans. CD8 T cells are critical for adaptive immune control of intracellular infections and cancer and their distribution and differentiation relate directly to their function. We assessed nonlymphoid distribution by examining available specimens of normal cervical tissue from premenopausal adult women. We found that CD8 T cells were integrated within the mucosa, exhibited a tissue resident memory T cell (T_RM) phenotype⁹, and comprised ~15,000 of every million cells (Fig. 1a&b and data not shown). This agrees with previous reports that adult (unlike neonatal) human nonlymphoid tissues are abundantly populated with T_RM^10–12. In contrast, CD8 T cells were almost completely absent from cervical sections from adult inbred laboratory mice (C57BL/6 strain) housed under SPF conditions (Fig. 1a&b). We then compared major CD8 T cell lineages in blood, focusing on species-specific markers that define functionally homologous populations of naïve, central memory (T_CM), and terminally differentiated effector memory CD8 T cells (T_EM or T_EMRA) (Fig. 1c). Relative to adult humans, memory CD8 T cells were much scarcer in laboratory mice, and almost entirely comprised of T_CM rather than T_EM or T_EMRA. Also in contrast to humans, laboratory mice lacked CD27lo/Granzyme B+ effector differentiated memory CD8 T cells that are thought to respond most immediately to infection (Fig. 1c)^13,14. Thus, memory CD8 T cells in laboratory mice were less frequent and strikingly different from those in adult humans, and in fact, appeared much more similar to neonates (Fig. 1c&d and extended data Fig. 1).

Open in a separate window

Figure 1

Laboratory mice lack differentiated memory CD8 T cell subsets, which is more similar to neonates than adult humans

a,b) CD8 T cell density within adult laboratory mouse (n=5) and human cervix (n=3). Representative immunofluorescence staining of frozen sections (scale bars, 50 μm) is shown. CD8β (red), DAPI (nuclei, blue). c) CD8 T cell phenotype was compared between adult human blood (n=13), adult laboratory mouse blood (n=10), and human cord blood (n=8) among 2 independent experiments by fluorescent flow cytometry (representative plots shown). Top panels gated on CD8+/CD3+ cells. Bottom panels gated on antigen experienced CD8+/CD3+ cells, as defined by conventional lineage markers in each species (red and green quadrants). d) Enumeration of granzyme B+ CD8 T cell frequencies amongst antigen experienced subsets. Significance was determined using unpaired two-sided Mann-Whitney U-test (in b) or Kruskal-wallis (ANOVA) test (in d). * p<0.05, ** p<0.01, **** p<0.0001, bars indicate mean ± S.E.M.

Mice are amongst the closest relatives of primates, but diverged from human ancestors 65–75 million years ago¹⁵. We wished to test whether the differences we observed between mice and adult humans were intrinsic to all mice (divergent immune system evolution), or unique to laboratory mice. To this end, we trapped free-living barn populations of feral mice and also procured mice from commercial pet stores. Compared to inbred SPF laboratory mice, feral and petstore mice were enriched in antigen-experienced CD8 T cells, particularly those that displayed a differentiated effector memory phenotype (Fig. 2a). More extensive phenotypic characterization of differentiation markers previously defined among CD8 T cells in laboratory mice revealed that naïve (CD44lo) cells were phenotypically identical between SPF and petstore mice (Fig. 2b). These data validated the consistency of markers used to discriminate between naïve and antigen experienced CD8 T cell populations in the different mouse cohorts. However, examination of antigen-experienced (CD44hi) CD8 T cells indicated enrichment of granzyme B+ and CD27lo cells in petstore mice, as well as other signatures of terminal effector differentiation including CD62Llo, CXCR3lo, CD127lo, and 1B11hi phenotypes. However, cell size (Fsc), proliferation (Ki67), and other measures of recent antigen exposure (CD69 and PD-1) were equivalent between both cohorts of mice, indicating that differences in CD8 T cell differentiation state were not due to an ongoing acute infection in petstore mice. Moreover, the relative density of CD8 T cells in petstore mouse nonlymphoid tissues was up to 50-fold greater than that observed in SPF laboratory mice (Fig. 2c,d).

Open in a separate window

Figure 2

CD8 T cell subsets vary between feral, petstore, and laboratory mice

a) CD8 T cell subsets were compared in PBMC between laboratory mice (n=9), feral mice that were trapped in the wild (n=10), and mice obtained from a pet store (n=6), among 2 independent experiments. b) Phenotype of CD44lo/CD62Lhi (naïve) and CD44hi (antigen-experienced) CD8+ PBMC was compared by fluorescent flow cytometry. c,d) CD8 T cell density in non-lymphoid tissues was compared between laboratory and pet store mice by quantitative immunofluorescence microscopy (QIM). Immunofluorescence staining of frozen sections of indicated tissues (n=8 animals per group, scale bars = 50 μm). CD8β (red), DAPI (nuclei, blue). Significance was determined using unpaired two-sided Mann-Whitney U-test. * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001, bars indicate mean ± S.E.M.

Petstore mice recapitulated aspects of human CD8 T cell differentiation and distribution that were absent in laboratory mice. These differences could be solely due to genetics, or dependent on environment. Inbred mice are genetically homogenous, which permits refined approaches and reductionist comparisons not possible in outbred populations, but may be accompanied by mutations that impair immune system development or function. Laboratory mice live in filtered microisolator housing that provides extreme hygiene¹⁶. Conditions are so clean, mice harboring fatal immunodeficiency diseases often thrive in these environments due to the absence of pathogen exposure. Thus, modern husbandry has evolved to the point in which laboratory mice, the standard model for biomedical research, exhibit far less infectious experience than free-living (i.e., dirty) humans or mice naturally acquire^17–20. To test whether environmental conditions might impact the differentiation state and distribution of CD8 T cells, we co-housed inbred laboratory mice with petstore mice, which were not raised in ultra hygienic barrier facilities. Adult petstore mice were introduced into cages containing adult laboratory (C57BL/6 strain) inbred mice. Within four weeks of co-housing, the CD44hi cells increased from 15% to ~70% of CD8+ peripheral blood mononuclear cells (PBMC), and plateaued at ~50% after eight weeks for the duration of the study (Fig. 3a,b). 22% of laboratory mice died during the first eight weeks of co-housing, after which no further mortality was observed (Fig. 3c).

Open in a separate window

Figure 3

Co-housing with petstore mice changes the immune system of laboratory mice

a) Proportion of CD44hi (antigen-experienced) CD8+ PBMC in laboratory mice housed in either SPF conditions (blue, n=10 for each time point) or co-housed with pet store mice (green, n=9 for each time point) among 2 independent experiments. b) CD8 T cell phenotype in blood of laboratory mice after co-housing with pet store mice. Representative flow plots (n=15) are shown. c) Survival of laboratory mice after co-housing with pet store mice (n=65 mice sampled). d) Representative phenotypes of CD44hi CD8+ PBMC isolated from laboratory mice co-housed with pet store mice for 100 days (n=15) and non-cohoused age-matched controls (n=10) was compared by fluorescent flow cytometry. e) Enumeration of CD8 T cells in non-lymphoid tissues between laboratory and cohoused mice by QIM (n=8 animals per group). f) Ratio of indicated cell types isolated and enumerated by flow cytometry from cohoused (n=6) and laboratory mice (n=6) among 2 independent experiments. ILC = innate lymphoid cells, Th = CD4 T helper cells, Treg = Regulatory CD4 T cells, GC = Germinal center. Single positive staining for T-bet, Gata3, and Rorγt were used to determine Th1, Th2, and Th17 lineages, respectively. g) Serum antibody concentrations (n=7 per group). h) Principal component analysis of gene expression data from PBMC of laboratory (n=8), cohoused (n=7) and pet store mice (n=8). Significance was determined using unpaired two-sided Mann-Whitney U-test (in e) or one-way ANOVA with Bonferroni post-hoc analysis (in g). *p<0.05, ** p<0.01, ***p<0.001. Bars indicate mean ± S.E.M.

Serology was performed for common mouse pathogens, and revealed exposure to viral, bacterial, and helminth pathogens (although not murine cytomegalovirus, see extended data table 1). Co-housing resulted in a constitutive increase in highly differentiated effector memory cells that matched signatures in outbred mice and humans, including the accrual of granzyme B+ and CD27lo cells (Fig. 3d). Moreover, nonlymphoid tissues became populated by CD8 T cells expressing a T_RM phenotype (Fig. 3e and extended data Fig. 2). Expanding the cellular analysis beyond CD8 T cells revealed extensive and profound changes to many innate and adaptive immune cell lineages throughout the host and increased levels of serum antibody (Fig. 3f&g and extended data Fig. 2). Furthermore, Principal Components Analysis of PBMC gene expression data in the space of all detected genes (~18000) revealed a spatial shift of cohoused samples away from laboratory samples and towards pet store along PC1 (Fig. 3h). Together, these data demonstrate that co-housing profoundly impacted the status of the immune system.

Our data suggested that laboratory mice have features in common with neonatal humans, and that altering the environment reproduced phenotypic signatures of adult humans. To test this hypothesis more broadly, we queried expression-profiling data of maternal and neonatal cord PBMC from an unaffiliated study²¹ with petstore vs. laboratory and cohoused vs. laboratory signatures via gene set enrichment analysis (GSEA) (Fig. 4a and 4b). The top 400 genes up-regulated in petstore mice showed highly significant enrichment in human adult expression data, while the top 400 down-regulated genes showed enrichment in neonates. Laboratory mice acquired this gene expression program after co-housing.

Open in a separate window

Figure 4

Microbial experience matures mouse immune transcriptome from neonatal to adult human and impacts immune system function

a) Experimental design and b) GSEA plots showing the enrichment of gene signatures among the indicated mouse group comparisons relative to human adult vs. neonatal comparison. Signatures consist of top 400 significantly differentially expressed genes of laboratory (n=8), cohoused (n=7) and petstore mice (n=8). c) Pairwise overlaps of metagenes identified through leading-edge metagene analysis and corresponding GO terms. d) Bacterial load in the liver 3 days post-challenge with 8.5 × 10⁴ CFU of Listeria monocytogenes (LM) among indicated groups (n=9 for all except laboratory mice where n=8). e) Parasitic load in peripheral blood 5 days post Plasmodium berghei ANKA parasitized RBC challenge among the indicated groups of laboratory (n=15), cohoused (n=19), and petstore mice (n=15). f) 28 days after LCMV Armstrong infection, the proportion of H-2D^b/gp33-specific CD8 T cell MPECs (KLRG1-, CD127+) and SLECs (KLRG1+, CD127−) in PBMC of cohoused (n=8) and laboratory (n=9). Data are cumulative of 2 independent experiments. d–f) Data points represent individual mice. Significance was determined using Kruskal-wallis (ANOVA) test (in d), one-way ANOVA test (in e) and unpaired two-sided t-test (in f) * p<0.05, ** p<0.01, **** p<0.0001. Bars indicate mean ± S.E.M.

To more deeply interrogate transcriptional concordances between petstore mice, cohoused mice, and adult humans relative to laboratory mice and neonatal humans, we applied GSEA with the ImmuneSigDB database of immunological signatures²². We then used leading edge metagene analysis of GSEA results to identify modules of coregulated genes upregulated in human adult vs. cord PBMC, and those found in pairwise comparisons of laboratory, petstore, and co-housed mice. Overlap between the resulting metagenes was used to identify global similarities between each dataset. We observed highly significant overlaps between metagenes upregulated in adult vs. cord human blood and metagenes upregulated in petstore or co-housed vs. laboratory mouse blood. These included numerous pathways related to innate and adaptive immune functions (Fig. 4c, extended data Fig. 3 and supplemental data Tables 1&2). Conversely, metagenes upregulated in human cord blood overlapped with those upregulated in laboratory mice. Thus, these functional modules represent a major axis of similarity in immune status between petstore mice and adult humans relative to laboratory mice and neonatal humans, and can be conferred to laboratory mice via cohousing with petstore mice.

We next questioned whether mouse husbandry affected immune responses. We performed challenges with the intracellular pathogen Listeria monocytogenes (LM), which is a bacterial infection often used to gauge immune function in laboratory mice. Compared to laboratory mice, both petstore and co-housed mice exhibited a >10,000 fold reduction in bacterial burden three days after challenge, which actually matched bacterial control in laboratory mice that were previously vaccinated against LM (Fig. 4d and extended data Fig. 4a). Hence, physiological exposure to environmental microbes revealed that C57BL/6 mice exhibit considerably more innate resistance to LM infection than indicated from studies using SPF animals. The impact of mouse husbandry on infection control extended to a cerebral malaria model (Plasmodium berghei ANKA, Fig. 4e, extended data Fig. 4b). Lymphocytic choriomeningitis virus infection is an oft-used model to examine critical questions of adaptive T cell differentiation, including the regulation of memory precursor vs. short lived effector cell (MPEC vs. SLEC) development^23,24. In this case, we observed that the proportion of MPECs and SLECs were significantly impacted by mouse husbandry (Fig. 4f and extended data Fig. 4c).

Experiments in mice have informed much of our understanding of immune regulation, and have directly informed the development of life-saving clinical therapies. However, our study reveals an unanticipated impact of SPF husbandry on the immune system. Our study does not advocate for an end to SPF studies. But for studies of the immune system it is ironic that such an immunologically inexperienced organism has become de rigueur, as our data show that this compromises development of an adult-like human immune system. For maximizing opportunities for bench to bedside translational medicine, it may be opportune to add dirty mice to our repertoire of investigative tools. Much like the analysis of truly sterile “germ-free” mice has revealed how influential the commensal flora is on “normal” physiology and immune system function, our study suggests that the immune system in mice may not be fully “normalized” without more complete microbial exposure¹⁰. Indeed, just as many autoimmune diseases do not manifest in genetically predisposed mice in the absence of commensal flora, certain infectious experiences have been shown to induce heterologous and innate immune memory, trigger autoimmune disease, and affect transplantation tolerance ^25–30. Forward genetic screens are ongoing in mice to reveal the function of immunological genes, and it might be beneficial to conduct these screens in a “dirty” mouse model.

More generally, dirty mice might be valuable for interrogating aspects of the hygiene hypothesis, immune function and treatment from disease in the settings of transplantation, allergy, autoimmunity, and vaccination, and perhaps in disparate diseases that may have an immune system or inflammatory component (such as cardiovascular disease and cancer)⁴. Such mice could supplement current models to either increase translation potential for human disease or to better inform efficacy of preclinical prophylactic and therapeutic modalities, without sacrificing powerful experimental tools and approaches that are unavailable to human studies.