Phenotypic and Clonal Stability of Antigen-Inexperienced Memory-like T Cells across the Genetic Background, Hygienic Status, and Aging

Antigen-inexperienced memory-like T (AIMT) cells are divided into two subsets based on the site of their origin. Innate T cells originate in the thymus, whereas virtual memory T cells are formed in the peripheral lymphoid organs (1, 2). In BALB/c (BALB) mice and several gene knockouts on C57BL/6 (B6) background, innate memory cells are formed in the thymus via IL-4 produced by invariant NKT (iNKT) cells (3–5). Wild-type (WT) B6 mice have low numbers of IL-4–producing iNKT cells in the thymus and form virtual memory T cells in a process dependent of IL-15 presentation by peripheral CD8α dendritic cells (2, 5–7). Recently, it has been observed that mature CD8⁺ single-positive thymic precursors of virtual memory T cells upregulate Eomes (8), a transcription factor required for their existence (7), suggesting that the differentiation process of virtual memory cells might be initiated already in the thymus and accomplished in the periphery (9). This scenario is consistent with the observation that thymocytes developing in a neonatal thymus are more predisposed toward AIMT cell formation than thymocytes developing in an adult thymus of B6 mice (10).

In addition to these major routes, lymphopenia induces homeostatic proliferation coupled with the differentiation of peripheral naive CD8⁺ T cells into AIMT cells, probably via IL-7 and/or IL-15 in B6 mice (11, 12). Moreover, an excess of IL-4, IL-7, or IL-15 induces the formation and/or expansion of AIMT cells (6, 13, 14). Given the fact that these three cytokines signal through the common γ-chain (15), it is not surprising that all of them induce AIMT cell formation. Although it has been proposed that some genes, such as Tbx21/T-bet or Klrk1/NKG2D, are differentially expressed in thymic innate memory and virtual memory AIMT cells (2), a direct comparison of the gene expression between virtual memory and innate memory T cells has not been carried out.

Despite their abundance and functional uniqueness, AIMT cells are still understudied. The potential reasons for the underrepresentation of AIMT cells in the immunological research could be the confusing nomenclature (1) and unclear developmental and functional relationship between different types of AIMT cells. Moreover, only a minority of AIMT cell-related studies used germ-free mice (8, 16, 17), which might obscure the interpretation of the observations. It is unclear whether the very existence of AIMT cells is physiological because AIMT cells have been characterized exclusively in inbred mouse strains housed in clean conditions and because it has been proposed that the CD8⁺ T cell compartment in laboratory mice substantially differs from feral mice (18).

In this study, we show in side-by-side controlled experiments for the first time, to our knowledge, that the routes of AIMT cell formation are conserved in mice with different genetic backgrounds and hygienic conditions, including germ-free and feral mice.

The Absthat were used for flow cytometry were purchased from BD Pharmingen and included the following: CD4 PE-Cy7 (RM4-5), catalog no. 552775; CD4 allophycocyanin-Cy7 (GK1.5), catalog no. 552051; CD4 Alexa Fluor 700 (RM4-5), catalog no. 557956; CD4 FITC (H129.19), catalog no. 553651; CD8α Pacific Blue (53-6.7), catalog no. 558106; CD49d PE (R1-2), catalog no. 553157; Ki67 Alexa Fluor 647 (B56), catalog no. 558615; KLRG1 allophycocyanin (2F1/KLRG1), catalog no. 553157; Ly-6c FITC (MR5-2), catalog no. 553186; TCR Vα2 PE (B20.1), catalog no. 553289; TCR Vα8.3 FITC (B21.14), catalog no. 553376; TCR Vβ8.1/8.2 FITC (MR5-2), catalog no. 553185; TCRβ allophycocyanin (H57-597), catalog no. 553174; and TCRβ PE (H57-597), catalog no. 553171. The following Abs were from BioLegend: CD4 Brilliant Violet 650 (RM4-5), catalog no. 100545; CD5 Brilliant Violet 510 (53-7-3), catalog no. 100627; CD8α Brilliant Violet 421 (53-6.7), catalog no. 558106; CD8α Brilliant Violet 510 (53-6.7), catalog no. 558106; CD24 FITC (M1/69), catalog no. 101806; CD44 Brilliant Violet 650 (IM7), catalog no. 103049; CD44 PerCP-Cy5.5 (IM7), catalog no. 103032; CD49d PE-Cy7 (R1-2), catalog no. 103618; CD62L Alexa Fluor 700 (MEL14), catalog no. 104426; CD122 FITC (TM-β1), catalog no. 123207; CD127 PE-Cy7 (A7R34), catalog no. 135015; CD127 PE (A7R34), catalog no. L34975; CD218a Alexa Fluor 647 (BG/IL18RA), catalog no. 132903; CX3CR1 PE-Cy7 (SA011F11), catalog no. 149016; IFN-γ Alexa Fluor 700 (XMG1.2), catalog no. 505824; TCR Vα8.3 PE (B21.14), catalog no. 127708; and TCR γ/δ PerCP-Cy5.5 (GL3), catalog no. 118117. Abs from eBioscience included the following: CD122 PerCP–eFluor 710 (TM-β1), catalog no. 46-1222-80; and EOMES Alexa Fluor 488 (Dan11mag), catalog no. 53-4875-80.

Common laboratory mouse strains C57BL/6J and BALB and transgenic mouse strains Vβ5 (19) and C.Cg-Tg(DO11.10)10Dlo/J (DO11.10) (20) were used in this study.

Mice were bred in our facilities (specific pathogen–free [SPF] mice at the Institute of Molecular Genetics of the Czech Academy of Sciences, Prague, Czech Republic and the Department of Biomedicine, University Hospital Basel, Basel, Switzerland; germ-free mice at the Laboratory of Gnotobiology, Institute of Microbiology of the Czech Academy of Sciences, Novy Hradek, Czech Republic) in accordance with laws of the Czech Republic and Switzerland, respectively. Animal protocols were approved by the Cantonal Veterinary Office of Baselstadt, Switzerland, by the Czech Academy of Sciences, the Czech Republic, and by the respective Committees for the Protection and Use of Experimental Animals of the Institute of Microbiology (identification no. 5/2019) and of the Institute of Molecular Genetics (identification no. 8/2017, 15/2019) of the Czech Academy of Sciences.

Mice were kept in the animal facility with a 12 h of light–dark cycle with food and water ad libitum. Males and females were used for experiments. Age- and sex-matched pairs of animals were used in the experimental groups. If possible, littermates were equally divided into the experimental groups.

For the surface staining, cells were incubated with diluted Abs in PBS/2% FCS on ice. LIVE/DEAD Near-IR dye (Life Technologies, L34975) was used for discrimination of live and dead cells.

For the staining of feral mice, cohoused mice, and control animals, cells were fixed with 2% formaldehyde at room temperature for 3 min prior to staining. All monoclonal Abs used for this staining were tested for their compatibility with the fixation procedure.

For intracellular staining, cells were fixed and permeabilized using Foxp3/Transcription Factor Staining Buffer Set (00-5523-00; eBioscience). For some experiments, enrichment of CD8⁺ T cells was performed using EasySep Magnetic bead separation kits (STEMCELL Technologies) or Dynabeads (Thermo Fisher Scientific), according to manufacturers' instructions, prior to the analysis or sorting by flow cytometry.

Cells were counted using a Z2 Particle Counter (Beckman Coulter) or using AccuCheck Counting beads (Thermo Fisher Scientific) and a flow cytometer. Flow cytometry was carried out with BD LSR II, BD FACSCanto II, or Cytek Aurora cytometers. Cell sorting was performed using BD Influx (BD Biosciences). Data were analyzed using FlowJo software (Tree Star).

A total of 400,000 cells/ml freshly isolated lymph node cells in IMDM/10% FCS supplemented with IL-2 were stimulated with PMA (10 ng/ml) and ionomycin (1 μM) or IL-12 (10 ng/ml) and IL-18 (10 ng/ml) or left unstimulated for 8 h. A total of 5 μg/ml brefeldin A (420601; BioLegend) was added to the culture for the last four hours of the incubation. The cells were stained with LIVE/DEAD Near-IR dye and fluorescently labeled Abs to CD8α, CD44, and CD49d. After washing, the cells were fixed and permeabilized and labeled with anti–IFN-γ Ab and analyzed by flow cytometry.

C57BL/6J and BALB mice age 5–8 wk were sublethally irradiated with 4 Gy on a T-200 x-ray machine (WOLF-Medizintechnik). The mice were sacrificed at day 16 postirradiation, and the lymphocytes from the peripheral lymph nodes and the spleen were stained for CD122, CD8α, viability, CD44, CD62L, CD49d, CD4, TCRβ, Ly-6c, and CD5 and analyzed by flow cytometry. The control group consisted of nonirradiated littermates.

Lymphocytes from peripheral lymph nodes and spleen from 8- to 12-wk-old BALB mice were harvested and stained with 5 µM CellTrace Violet proliferation dye (C34557; Thermo Fisher Scientific) in PBS for 10 min at 37°C. A total of 1 × 10⁶ naive CD8⁺ CD44⁻ cells were sorted and i.v. injected into recipient DO11.10 mouse. The phenotype of the transferred cells was analyzed after 7 d.

Ten- to twelve-week-old BALB or C57BL/6J SPF mice were irradiated with 4 Gy and sacrificed 16 d postirradiation. Nonirradiated mice were young SPF (7–12 wk) or aged (12–19 mo) BALB or C57BL/6J mice. Lymphocytes from the peripheral lymph nodes and the spleen were stained with TCRβ, CD8, CD44, CD49d, CD122, and viability dye and sorted by Influx. Each sample consisted of a pool of lymphocytes originating from 2–5 mice.

RNA was isolated using TRIzol LS (Thermo Fisher Scientific) followed by in-column DNase digestion using RNA a Clean & Concentrator Kit (Zymo Research), according to manufacturers' instructions.

Samples were prepared with SMARTer Stranded Total RNA-Seq v2–Pico Input Mammalian library preparation kit (Takara Bio). The protocol included cDNA synthesis from mRNA, ligation of sequencing adaptors and indexes, ribosomal cDNA depletion, final PCR amplification, and product purification. Library size distribution was evaluated on the Agilent 2100 Bioanalyzer using the High Sensitivity DNA Kit (Agilent Technologies). Libraries were sequenced using an Illumina NextSeq 500 instrument using 76-bp single-end high-output configuration.

Read quality was assessed by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). For subsequent read processing, a bioinformatics pipeline nf-core/rnaseq version 1.1 (https://doi.org/10.5281/zenodo.1447331), was used. Individual steps included removing sequencing adaptors with Trim Galore! (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), mapping to reference genomes GRCm38 or BALB _cJ_v1 (21) (Ensembl annotation version 94) respective to the specimen strain with HISAT2 (22), and quantifying expression on gene level with featureCounts (23). Resulting per gene read counts for individual strains were merged together using a gene ortholog map obtained from Ensembl. Merged counts served as input for differential expression analysis using DESeq2 R Bioconductor package (24). Prior to the analysis, genes not expressed in at least two samples were discarded. Shrunken log₂-fold changes using the adaptive shrinkage estimator (25) were used for differential expression analysis. Genes exhibiting minimal absolute log₂-fold change value of 1 and statistical significance of adjusted p value <0.05 between compared groups of samples were considered as differentially expressed for subsequent interpretation and visualization.

To show sample expression relatedness with the generic effect of mouse strain erased, we employed a method of removing unwanted source of variation by fitting a linear model using the limma R Bioconductor package (26) (function removeBatchEffect) after normalizing the data using variance-stabilizing transformation.

Raw fastq files have been deposited in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) at the European Bioinformatics Institute under accession number E-MTAB-9091.

Lymphocytes from Vβ5 mice were sorted as viable CD8⁺ CD4⁻ TCRVα2⁺ CD62L⁺ CD44^- (naive) or CD44⁺ (AIMT). RNA was isolated using TRIzol (Thermo Fisher Scientific) followed by in-column DNase treatment using the RNA Clean & Concentrator Kit (Zymo Research). cDNA synthesis was performed using a TCR-specific primer (5′-TTGGCACCCGAGAATTCCACTGHHHHHACAHHHHHACAHHHHHCACATTGATTTGGGAGTC-3′) by using Maxima Reverse Transcriptase (Thermo Fisher Scientific) according to the manufacturer's instructions. The first round of PCR amplification was performed using primers 5′-CGTTCAGAGTTCTACAGTCCGACGATCHHHHACHHHHACHHHNGCAGCCCATGGTACCA GCAGTTCC-3′ and 5′-ACTGGAGTTCCTTGGCACCCGAGAATTC-3′, using KAPA HiFi HotStart ReadyMix (Kapa Biosystems) and the following conditions: 95°C for (5 min); 40 cycles at 98°C (20 s), 61°C (15 s), 72°C (15 s), and 72°C for (5 min). The resulting ∼440-bp-long PCR product was subjected to electrophoresis using an agarose gel and extracted using a Gel Extraction Kit (QIAGEN).

Following the first round of PCR, the TCRα copy number was quantified via quantitative PCR (LightCycler 480 II; Roche) and compared with a dilution series of OT-I TCRα cloned in MSCV vector. A total of 7.5 × 10⁴ copies of the PCR product were then used for the second PCR amplification round using primers 5′-AATGATACGGCGACCACCGAGATCTACACGTTCAGAGTTCTACAGTCCGACGAT*C-3′ and 5′-CAAGCAGAAGACGGCATACGAGATCGTGATGTGACTGGAGTTCCTTGGCACCC*G-3′. The polymerase was KAPA HiFi HotStart ReadyMix (2×) and the following conditions: 95°C (5 min); 40 cycles at 98°C (20 s), 62°C for (15 s), 72°C for (15 s), and 72°C for (5 min). The resulting ∼500-bp-long PCR product was subjected to electrophoresis in agarose gels and purified using Gel Extraction Kit (QIAGEN). The samples were mixed in equimolar ratios to obtain 30 µl of a final concentration of 10 nM and sequenced by Illumina MiSeq (v3 kit,; 300-bp paired-end reads) at the BIOCEV, Charles University, Vestec, Czech Republic.

FastQC was used to assess read quality. Next, MIGEC version 1.2.9 (27) was used for quantification of clonotypes, and the resulting CDR3 sequence abundances in all samples were summarized to a clonotype abundance matrix. From this matrix, relative frequencies of CDR3 usage were computed. Clonotypes present in less than three samples were disregarded, and the clonotype abundance matrix was treated similar to RNA-sequencing count matrix,: DESeq2 was used to normalize for sequencing depth and RNA composition, and to test for differential clonotype abundance between AIMT and naive sample groups. Normalized abundances were used for principal component analysis (PCA).

For comparison of the diversity of TCR repertoires (CDR3 clonotypes), a Simpson Diversity Index was calculated using the function “diversity” from package “vegan” version 2.5-6 on R 4.0.2.

The sequence data has been deposited in the Sequence Read Archive and BioProject databases at National Center for Biotechnology Information. (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA633838).

We analyzed feral mice Mus musculus musculus that were trapped in human houses or garden shelters in the region of Central Bohemia, Czech Republic. The mice used in this study were feral mice caught up to 1 mo prior to the analysis and 8- to 24-mo-old first generation (F1) offspring of the feral mice born in captivity. These mice were housed at the Department of Zoology, Faculty of Science, Charles University in Prague, Czech Republic. Mice were provided with food ad libitum and were kept under stable conditions in open cages (13/11 h, day:night, 23°C).

All animal procedures on feral mice were carried out in accordance with the law of the Czech Republic (paragraph 17, number 246/1992) and the local ethics committee of the Faculty of Science, Charles University in Prague, Czech Republic, and they were specifically approved for this study by accreditation number 13060/2014-MZE-17214.

Four SPF females were housed for 8 wk with one feral male per cage. The open cohousing cages were divided by a wire mesh with square openings (diameter was 1cm), allowing contacts of cohoused animals. Cohousing cages were supplied with fresh bedding at the beginning of the experiment and provided with water and food ad libitum.

The age of SPF females was 4- to 6-wk at the beginning of the cohousing experiment. The cohoused SPF females were matched with their littermates, which remained in the SPF facility and had the same food and the same bedding as the cohoused animals. We performed two experiments with slightly modified setups. For experiment A, during the 6-wk period of cohousing, the bedding was changed every 2 wk and the feral males were transferred into another cage. For experiment B, during the 6-wk period of cohousing, the bedding was not changed. Every 2 wk, the mice were transferred to the other side of the same cage to be exposed to the feces of the feral mice.

A reference sample of saliva was collected from each mouse participating in the cohousing experiment before the start of experiment B. Throughout both experiments A and B, the saliva was collected from cohoused animals prior to the cohousing and after every second week of cohousing by gentle flushing with a pipette using 50 μl of the 0.9% NaCl solution. At the end of the experiment, the cohoused animals, as well as the non-cohoused controls, were sacrificed, and samples from duodenum, jejunum, ileum, cecum, and colon were collected for high-throughput 16S rRNA sequencing of the intestinal microbiota. At the end of each experiment, two cohoused Vβ5 animals and one cohoused feral male were sent for a thorough health status examination at the National Veterinary Institute in Prague. The results of this examination are available in health monitoring data submitted to Zenodo (https://zenodo.org/record/4298697#.YGxxu2RKhts).

Metagenomic DNA from gut and saliva samples was extracted using the PowerSoil DNA Kit (MO BIO Laboratories, San Diego, CA). Barcoding was performed with previously published 16S rRNA primers Bact-0341-b-S-17 and S-D-Bact-0785-a-A-21 (28). Dual-indexed Nextera sequencing libraries were prepared using a previously published two-step PCR design (29). The first PCR was performed in 10 μl and consisted of 1× KAPA HiFi HotStart ReadyMix (Roche), each 16S rRNA primer at 0.2 μM and 3 μl DNA. PCR conditions were as follows: initial denaturation at 95°C for 3 min followed by 28 (for colon and cecum), 33 (for saliva), or 35 (ileum, duodenum, jejunum) cycles each of 95°C (30 s), 55°C (30 s), and 72°C (30 s) and a final extension at 72°C (5 min). The second PCR differed from the first PCR in the following parameters: it was performed in 20-μl volume; the concentration of each indexed primer was 1 μM; 1 μl (gut samples) or 6 μl (saliva) of the first PCR product were used as a template; and the number of PCR cycles was 12. Technical duplicates were prepared to account for noise due to PCR and sequencing stochasticity. PCR products were pooled equimolarly, size-selected with Pippin Prep (Sage Science, Beverly, MA) at 520–750 bp and sequenced using MiSeq (Illumina, San Diego, CA) and v3 chemistry (i.e., 2 × 300-bp paired-end reads). Sequencing data were archived in European Nucleotide Archive under project accession PRJEB38301.

Samples were demultiplexed and primers were trimmed by skewer software (30). Using DADA2 (31), we filtered out low-quality sequences (expected number of errors per read >1), denoised the quality-filtered fastq files, and constructed an abundance matrix representing reads counts for individual haplotypes in each sample. Using UCHIME (32) and the gold.fna database, we identified chimeric sequences and removed them from the abundance matrix. Taxonomic assignation of haplotypes was conducted by the Ribosomal Database Project classifier (80% confidence threshold) (33) and the SILVA reference database v132 (34).

The effect of cohousing on α-diversity (based on Shannon diversity indices) was tested using a linear mixed-effect model, considering individual identity as a random effect and the effect of treatment level (cohoused versus non-cohoused), gut section, and their interactions as explanatory variables. The effect of cohousing on microbiota composition was assessed by permutation tests, in which the observed difference in composition (i.e., Bray–Curtis similarities) between cohoused laboratory versus wild individuals and non-cohoused laboratory versus wild individuals was compared with permuted average differences, where belonging to the cohoused versus non-cohoused laboratory group was randomized. Individual haplotypes whose relative abundances were affected by cohousing were identified by the DESeq2 package (24).

It has been shown that the percentage of AIMT cells increases upon aging (35–39). However, these studies have used conventional SPF mice, and thus the role of microbial Ags in this process cannot be excluded. Moreover, previous studies analyzed only B6 mice but not BALB mice that show substantial differences in the site of origin and cytokine dependency of AIMT cells (5, 3). For this reason, we compared the CD8⁺ T cell compartment in young and aged germ-free B6 and BALB mice. We observed that the frequency of peripheral CD44⁺ CD49d⁻ AIMT cells increases in both strains upon aging (Fig. 1A, 1B; Supplemental Fig. 1A, 1B), suggesting a common homeostatic mechanism of age-dependent induction of AIMT cells.

AIMT cells from B6 mice depend on IL-15 and express high levels of CD122 (IL-2RB), a subunit of the IL-15 receptor (5, 7, 16). We observed that AIMT cells from young germ-free B6 mice express higher levels of CD122 than those from young BALB mice (Fig. 1C, 1D; Supplemental Fig. 1C, 1D). Interestingly, a population of CD122^HIGH AIMT cells appears in aged germ-free BALB mice, whereas no difference in CD122 expression was found between AIMT cells from germ-free young and aged B6 mice (Fig. 1C, 1D). We observed that CD122^HIGH AIMT cells show slightly stronger homeostatic proliferation than CD122^LOW AIMT cells in aged but not in young BALB mice (Fig. 1E, 1F), which probably contributes to their accumulation in aged mice.

CD122^HIGH AIMT cells in B6 mice are generated in the periphery via IL-15 and homeostatic TCR signaling (6, 17). These cells are more frequent in aged mice, probably because the thymic output is low and the lymphopenic environment augments AIMT cells formation. In contrast, CD122^LOWAIMT cells are generated in the thymus of young BALB mice via IL-4 (3). We hypothesized that, during aging of BALB mice, the thymic regress impairs de novo formation of CD122^LOW AIMT cells, leading to the lymphopenia-induced generation of proliferative CD122^HIGH AIMT cells in the periphery. To elucidate whether CD122 expression could discriminate between thymic (innate memory) and peripheral AIMT cells (virtual memory) in BALB mice, we employed two models for lymphopenia-induced AIMT cells formation.

The first lymphopenic model was sublethal irradiation of mice. Irradiation reduces the number of T cells, which leads to homeostatic proliferation and AIMT formation of the surviving T cells in the periphery (1). As expected, irradiation increased the frequency of AIMT cells in B6 and BALB mice (Fig. 2A, 2B). Irradiation-induced AIMT cells showed significantly higher CD122 expression than control AIMT cells in BALB but not in B6 mice (Fig. 2C, 2D). This result is in close agreement with the hypothesis that CD122 expression is a marker of peripherally induced AIMT cells in both strains.

The second model was TCR transgenic DO11.10 Rag⁺ BALB mice. Most of T cells in these mice are monoclonal CD4⁺ T cells specific to I-A^d-OVA_323-339 (40). However, rare T cells expressing endogenous TCRα differentiate into CD8⁺ T cells. Thymic AIMT cells should be substantially reduced in these mice because of the absence of IL-4–producing thymic iNKT cells that are crucial for the formation of thymic AIMT cells (3). In addition, this mouse is lymphopenic in the CD8⁺ T cell compartment, which should promote the formation of AIMT cells in the periphery.

DO11.10 mice showed comparable frequency of AIMT cells with polyclonal BALB mice. The frequency of AIMT CD8⁺ T cells during aging increased (Fig. 2E, 2F), indicating that the age-dependent increase in the AIMT cell compartment does not depend on the presence of iNKT cells. The expression of CD122 was higher in AIMT cells from DO11.10 mice than in AIMT cells from polyclonal BALB mice (Fig. 2G), which is consistent with the hypothesis that CD122^HIGH AIMT cells are formed in the periphery. When we transferred naive polyclonal CD8⁺ T cells into the DO11.10 mice, we observed the formation of AIMT cells, which was coupled with their homeostatic proliferation (Fig. 2H, 2I).

Overall, these data indicate that CD122 expression differs between peripherally induced virtual memory AIMT cells (both in B6 and BALB mice) and thymus-derived innate memory AIMT cells of BALB mice.

The AIMT cell compartment is heterogeneous. As mentioned above, AIMT cells are formed in different organs (thymus or peripheral lymphoid organs) and have distinct cytokine dependencies (IL-15 versus IL-4) in different strains. Moreover, aging might change the gene expression programs of AIMT cells (36–38). We addressed differences between particular AIMT cell populations by high-throughput RNA sequencing using sorted naive and AIMT cells from young B6 and BALB mice, irradiation-induced AIMT cells from B6 and BALB mice, AIMT cells from aged B6 mice, and CD122^LOW and CD122^HIGH AIMT cells from aged BALB mice (Supplemental Fig. 1E–I).

As expected, the major source of the variability among these samples was the different genetic background of B6 and BALB mice (Supplemental Fig. 2A). We normalized the gene expression data for the differences between the strains by fitting the data with a linear model using the strain identity as one of the explanatory variables (Fig. 3A). The expression of Il2rb (CD122) and Itga4 (CD49d) in the sorted populations corresponded to the sorting strategy (Fig. 3B) and the previous characterization of AIMT cells from both mouse strains (Fig. 1A–D).

We assumed that naive and irradiation-induced AIMT cells represent two relatively uniform cell types present in both strains, and we used them as controls for evaluating our data normalization. The PCA of the adjusted data revealed that naive cells and irradiation-induced AIMT cells from both strains clustered together, confirming the reliability of our adjusted data (Fig. 3A). We obtained very similar results when performing the PCA using data without strain-specific normalization but with the exclusion of significantly differentially expressed genes between the strains (Supplemental Fig. 2B).

The AIMT cells from aged mice showed moderate differences in their gene expression patterns in comparison with AIMT cells from young animals (Fig. 3A). We identified 36 genes upregulated in AIMT cells from aged mice versus young mice in both strains (Supplemental Fig. 2C). Twenty-four of these transcripts were at least 4-fold enriched in AIMT cells from aged B6 mice in comparison with young mice in the previous study by Quinn et al. (36). Interestingly, we did not identify a single gene significantly upregulated in AIMT cells from young mice versus aged mice in both strains.

Young AIMT cells from both strains showed comparable global gene expression signature. The gene expression similarity between these populations was revealed as the proximity of young AIMT cells from both strains in the PCA plots (Fig. 3A, Supplemental Fig. 2B) as well as a correlation of AIMT/naive cell fold changes for individual genes plotted as B6 versus BALB without using any correction for the strain-dependent differences (Fig. 3C). Despite the similarity of the gene expression patterns, we could identify one gene significantly upregulated, and 25 genes significantly downregulated in young BALB AIMT cells versus B6 AIMT cells (Fig. 3D), suggesting that these genes might distinguish thymic innate memory and peripheral virtual memory subsets of AIMT cells.

We observed that the genes upregulated in AIMT cells from young B6 versus young BALB mice are significantly enriched in aged CD122^HIGH AIMT cells versus aged CD122^LOW AIMT cells from aged BALB mice (Fig. 3D, 3E), supporting the hypothesis that high CD122 levels predict the peripheral origin of AIMT cells, irrespective of the strain.

It has been proposed that T-bet (Tbx21) and NKG2D (Klrk1) are expressed in peripheral but not in thymic AIMT cells (2). Moreover, a small group of genes potentially involved in bystander cytotoxicity (Gzmb, Il12rb2, Il18r1, Il18rap, Stat4) were shown to be upregulated in peripheral AIMT cells (2, 6). All of these genes but Gzmb were enriched in young AIMT cells over naive T cells in our data (Fig. 3F). Four genes (Tbet, Klrk1, Il18r1, Il18rap) showed higher expression in AIMT cells of the proposed peripheral origin (AIMT cells from young and aged B6 and CD122^HIGH AIMT cells from aged BALB mice) than in the corresponding AIMT cells of the proposed thymic origin (AIMT cells from young BALB and CD122^LOW AIMT cells from aged BALB mice) (Fig. 3F).

Flow cytometry confirmed higher expression of IL-18R in AIMT cells from B6 than from BALB mice (Fig. 3G, Supplemental Fig. 2D). Moreover, IL-18R was preferentially expressed in CD122^HIGH AIMT cells (Supplemental Fig. 2E). Accordingly, AIMT cells from B6 mice were more efficient in producing IFN-γ after stimulation with IL-12/IL-18 than AIMT cells from BALB mice, whereas the responses of these cells to PMA and ionomycin were comparable (Fig. 3H, Supplemental Fig. 2F).

Overall, these data revealed a stable gene expression program in AIMT cells that is fine-tuned by the age and genetic background. The most notable variation among AIMT cells is the differential expression of genes encoding proinflammatory receptors based on the site of their origin. Peripheral AIMT cells express higher levels of CD122 (a receptor for IL-2 and IL-15), IL-18R, and activating NKG2D receptor than thymic AIMT cells. This expression pattern predisposes peripheral AIMT cells for efficient bystander immune protection.

The frequency of murine AIMT cells changes over time. It peaks around 3 wk of age (41), and it increases again as the mouse ages (35). For this reason, it is unclear whether AIMT cells in young and aged animals have a different clonal origin (42). To address this question, we analyzed TCR sequences of naive and AIMT cells in young and aged mice. To overcome the problem with TCRα and TCRβ pairing, we analyzed TCR α-chains in a Vβ5 mouse that has a fixed TCRβ from the OT-I TCR.

Because central memory CD8⁺ CD44⁺ CD62L⁺ T cells in Vβ5 mice do not express CD49d (Supplemental Fig. 3A), we gated AIMT cells in Vβ5 mice as CD44⁺ CD62L⁺ double-positive. The frequency of AIMT cells in Vβ5 mice recapitulates WT mice, including the increase in the frequency of AIMT cells upon aging (Supplemental Fig. 3B).

First, we quantified the usage of TRAV14 (corresponding to Vα2 epitope) and TRAV12 (corresponding to Vα8.3) by staining with specific Abs. In line with our previous data (17), we observed a higher Vα2/Vα8.3 ratio in AIMT cells than in naive T cells of young mice (Fig. 4A, Supplemental Fig. 3C). Interestingly, the Vα2/Vα8.3 ratio in naive cells was higher in aged mice than in young mice (Supplemental Fig. 3C). The highest Vα2/Vα8.3 ratio was observed in the AIMT compartment in aged mice. Because Vα2⁺ T cells are, on average, more self-reactive than Vα8.3⁺ T cells in Vβ5 mice (17), our data suggest that the CD8⁺ T cell compartment in aged mice is enriched for self-reactive T cells and that AIMT cells are generally more self-reactive than naive cells in young as well as in aged mice.

In the next step, we analyzed TCRs from Vα2⁺ T cells in greater detail using high-throughput sequencing. Naive and AIMT cells showed differential usage of TRAJ segments, indicating that they are formed from different T cell clonotypes (Fig. 4B, Supplemental Fig. 3D), which is in line with previous results (8, 17). Interestingly, we did not see any differences in the TRAJ usage between young and aged mice (Fig. 4B, Supplemental Fig. 3D). The second step of our analysis was the comparison of individual clonotypes using CDR3 amino acid sequences. The PCA analysis separated a cluster of AIMT cells (both young and aged mice) from naive T cells (Fig. 4C). The comparison of the usage of individual clonotypes by naive and AIMT cells showed that some clonotypes are enriched in the naive compartment and some clonotypes are significantly enriched in AIMT cells (Fig. 4D). The AIMT cell repertoire in young and aged mice is constituted predominantly from a small number of AIMT cell–enriched clonotypes. In contrast, the naive repertoire is more diverse and contains naive-enriched as well as AIMT-enriched clones (Fig. 4E). The usage of the dominant clonotypes differ in naive and AIMT subsets but not so much between the corresponding subsets from young and aged mice (Fig. 4F). Accordingly, the diversity of the TCR repertoire is relatively high in naive T cells and relatively low in AIMT T cells, irrespective of the age of the animals (Supplemental Fig. 3E). Overall, these data showed that AIMT cells from young and aged mice have the same clonal origin, although it is possible that some minor clonotypes are lost from the AIMT repertoire during aging, as indicated by the reduction of TCR Vα8.3 clonotypes.

It has been suggested that the exposure to environmental Ags in feral and “pet shop” mice has a substantial role in shaping the CD8⁺ T cell compartment (18). To address the possibility that the occurrence of AIMT cells might be an artifact of specific hygienic conditions in germ-free and SPF facilities, we analyzed mice with natural microbial colonization. First, bred feral mice in the animal facility and analyzed F1 offspring born in captivity. Interestingly, we observed that AIMT cells are relatively frequent in these mice (Fig. 5A, 5C). The expression of CD122 was variable between individual animals, suggesting variable peripheral or thymic origin of these cells, likely based on the environmental or genetic factors (Fig 5A).

In the next step, we analyzed feral mice trapped in garden shelters and human houses. These mice are supposed to have even richer antigenic exposure history that the F1 feral mice born in captivity. We observed that these mice contain a large population of CD8⁺ CD44⁺ CD49d⁻ T cells, corresponding to AIMT cells (Fig. 5B, 5C). The expression of CD122 in these AIMT cells largely varied among individual animals (Fig. 5B). Overall, these data suggested that the population of AIMT cells exists in feral mice. Interestingly, the percentage of AIMT cells was higher in feral mice than in F1 offspring of feral mice kept in captivity (Fig. 5C). This indicated that exposure to environmental Ags might increase the size of the AIMT cell compartment, as was recently shown for helminth infections (43, 44). The relative size of the AIMT cell compartment correlated with the frequency of CD122^HIGH cells among them in feral mice (Fig. 5D). This indicates that environmental factors inducing AIMT cells preferentially induce CD122^HIGH peripheral virtual memory T cells. This correlation was not observed in F1 offspring of feral mice, indicating the major role of genetic background (Fig. 5E).

To observe how the AIMT cell compartment and other CD8⁺ T cell subsets change upon exposure to environmental Ags in laboratory mice, we cohoused B6 WT mice and transgenic Vβ5 mice with feral mice or F1 offspring of feral mice for 6 wk. Surprisingly, the cohousing had no significant effect on the proportions of the major CD8⁺ T cell subsets (i.e., naive, AIMT, and Ag-experienced [AgE] cells) in B6 and Vβ5 mice (Fig. 5F–H). The only observed difference between cohoused and control laboratory strains was a slightly higher percentage of CX3CR1⁺ CD44⁺ CD49d⁺ AgE cells in the spleen and mesenteric lymph nodes of the cohoused animals of the SPF origin (Fig. 5G).

We observed that the enrichment of the AIMT cell compartment in Vα2⁺ T cells was preserved in the cohoused animals (Fig. 5I), suggesting that the cohousing did not substantially alter the composition of the repertoire of the AIMT cell compartment. Overall, our data indicate that AIMT cells are a naturally occurring CD8⁺ T cell subset in laboratory mouse strains as well as in feral mice.

Because we observed a surprisingly small effect of the cohousing of laboratory strains with feral mice or F1 offspring of feral mice on their CD8⁺ T cell compartment, we examined the colonization of the feral and cohoused animals with microbes and viruses. A detailed health examination of a limited number of animals detected Helicobacter, Pasteurella, murine CMV, and endoparasites but no other pathogenic agents in the feral mice (see https://zenodo.org/record/4298697#.YGxxu2RKhts). Of these pathogens, only Helicobacter and endoparasites were transmitted to the cohoused animals in at least one case out of four tested cohoused mice. This was in line with the fact that none of the cohoused mice died before the end of the experiment, including Vβ5 mice, which are most likely immunocompromised because of the limited TCR repertoire.

Subsequently, we addressed the effects of cohousing on bacterial colonization. We performed 16S rRNA sequence analysis of the composition of the microbial community in the gut and salivary glands of cohoused animals. The overall microbial diversity in the intestine of cohoused laboratory strains was not affected, except for the ileum, where the diversity increased (Supplemental Fig. 4A). Interestingly, microbial diversity decreased in the colon of feral mice upon cohousing (Supplemental Fig. 4A). As expected, the microbial composition of the intestine of the laboratory strains converged to the feral mice upon cohousing (Fig. 6A, Supplemental Fig. 4B). The most notable effect of the cohousing was the abundance of Campylobacteria (a taxon including Helicobacter) in the colon and cecum of the cohoused mice, whereas this group of bacteria was not present in non-cohoused laboratory mice (Fig. 6B, Supplemental Fig. 4C). Two different Helicobacter taxons (Campylobacterales) and Alistipes (Bacteroidia) were detected in the colon and cecum of some of the cohoused but not SPF mice (Fig. 6C). In contrast, three taxonomic units from Clostridiales, Roseburia, and Ruminococcus taxons were depleted in colon, cecum, and duodenum of the cohoused mice (Supplemental Fig. 4D).

The salivary microbiome of the laboratory strains did not converge to feral mice, but rather the microbiome of the feral mice converged to the microbiome of SPF laboratory strains during the cohousing (Supplemental Fig. 4E). However, cohousing led to a dramatic increase in Bacteriodia taxon in saliva of laboratory strains (Fig. 6D, Supplemental Fig. 4F). Moreover, the abundance of Clostridia and Deltaproteobacteria was detected in some individuals during but not before cohousing (Fig. 6D, Supplemental Fig. 4F). Overall, these data document that cohousing with feral mice changed the microbial colonization of laboratory mice without a substantial impact on the CD8⁺ T cell compartment.

AIMT cells represent one of the two most numerous populations of CD8⁺ T cells in mice. Although it has been shown that these cells are functionally different from naive and AgE memory T cells (6, 17, 39, 43–46), our understanding of the biology of these cells is incomplete. There are multiple potential reasons why AIMT cells are an underrepresented area of focus. First, multiple ways of AIMT cells formation were proposed involving different cytokines (IL-4, IL-7, and IL-15), different anatomical sites (thymus, secondary lymphoid organs), and different physiological states (experimentally induced lymphopenia, newborn lymphopenia, aging) (1, 2, 4). Second, the phenotypic and functional similarities between particular AIMT cell subsets and their differences are not well understood (1, 2, 4). Third, only a minority of studies used germ-free mice to exclude microbial Ags (16, 17), and at the same time, no previous study addressed the presence of AIMT cells in feral mice, which were reported to have a substantially different CD8⁺ T cell compartment than laboratory mice (18).

In this study, we performed comprehensive phenotyping of AIMT cells to address multiple unresolved questions concerning their origin and heterogeneity. We concluded that the formation of AIMT cells represents a robust genetic differentiation program across the genetic background, aging, and hygienic status. However, these three conditions regulate AIMT cell abundance and internal variability.

We identified CD122 as a marker discriminating CD122^LOW thymic (innate memory) and peripheral CD122^HIGH (virtual memory) subsets of CD8⁺ AIMT cells. CD122^LOW AIMT cells prevail in WT young BALB mice and some feral mice. On the contrary, CD122^HIGH AIMT cells are dominant in B6, in some feral mice, and BALB mice with the impaired formation of CD122^LOW cells because of aging, expression of a transgenic TCR, disabling the formation of IL-4–producing iNKT cells, or irradiation. Collectively, our data indicate that the absence of thymic innate memory AIMT cells allows the conversion of naive T cells into AIMT cells in the periphery, presumably via IL-4, IL-7, and/or IL-15 (1, 4, 7, 11, 12).

AIMT cells represent the largest population of CD8⁺ T cells in aged B6 mice (35). In this study, we confirmed this observation in germ-free conditions and extended it by the analysis of aged BALB mice. The mechanism behind the high frequency of AIMT cells in aged animals was unclear. It has been proposed that the origin of these cells might differ from the AIMT cells in young animals (37). We addressed this question by analyzing gene expression profiles and TCR profiling of young and aged AIMT cells using high-throughput sequencing. We observed that AIMT cells show a common gene expression pattern, largely irrespective of their site of origin, genetic background, and the age of the host. However, our analysis revealed that the site of origin and aging fine-tune the gene expression among AIMT cells. In addition to confirming differences between AIMT cells from young and aged B6 mice as reported previously (36), we identified similarities between CD122^HIGH AIMT cells from aged BALB mice and AIMT cells from B6 mice. These similarities support the concept that CD122^HIGH AIMT cells are formed at the periphery in both B6 and BALB strains. Expression of key effector molecules involved in bystander cytotoxicity as well as IL-12/IL-18 production of IFN-γ was higher in peripheral AIMT cells than in thymic AIMT cells, suggesting that these two subsets differ in their ability to elicit bystander protective immunity, as proposed previously (2). Despite these relatively subtle differences, virtual memory T cells, mostly present in B6 mice, and innate memory T cells, mostly present in young BALB mice, exhibit a very similar gene expression pattern and likely represent two flavors of the AIMT cell lineage.

It has been proposed that aged AIMT cells are formed from T cell clones with a high level of self-reactivity (47). Recently, we have shown that AIMT cells are formed from highly self-reactive T cell clones in young adult mice (17), which was subsequently confirmed by another study (8). In line with these previous reports, we observed that CD8⁺ naive and AIMT cells use different TCR repertoires. Clones typically present in the naive population are virtually absent in AIMT cells, whereas typical AIMT clones can still be identified in the naive compartment, albeit with a much lower frequency. This partial overlap is not surprising, as AIMT cells are formed from naive T cells in the periphery. However, a recent study by Miller et al. (8) showed a clear separation of naive and AIMT cell repertoires. This discrepancy might be caused by the difference between the employed mouse models. We analyzed TCR Vα2⁺ repertoire in transgenic mice expressing TCR Vβ5 adopted from H2-K^b–restricted OT-I TCR. Importantly, the Vβ5 mouse has a similar frequency of AIMT cells as WT B6 mice. Miller et al. (8) analyzed TCRα repertoire from mice expressing TCR Vβ3-chain. This particular TCRβ chain is abundant among prostate cancer–infiltrating regulatory CD4⁺ T cells (48) and might produce a T cell repertoire biased toward MHC class II restriction, subsequently affecting the frequency, repertoire, or phenotype of CD8⁺ AIMT cells.

We observed that young and aged AIMT cells have comparable TCRVα2 repertoire with identical dominant T cell clones and similar level of diversity. Moreover, we observed that AIMT cells preferentially use Vα2 over Vα8.3 TCR chains, which are more pronounced in aged mice. Thus, it is likely that some minor, relatively less self-reactive clonotypes are removed from the AIMT cell repertoire during aging, which is consistent with a previous report (49).

Overall, the probable causes of the high frequency of AIMT cells in aged B6 and BALB mice are as follows: 1) low thymic egress of fresh naive T cells, 2) irreversible AIMT cells formation from highly self-reactive clones augmented by age-related lymphopenia, and 3) the survival and proliferative advantage of AIMT cells over naive T cells.

For the first time, to our knowledge, we studied AIMT cells in feral mice and F1 offspring of feral mice born in captivity, suggesting that AIMT cells are not an artifact of inbred strains and/or SPF conditions. Feral mice showed an even higher frequency of AIMT cells than laboratory strains. This could be caused by the enhancement of the AIMT cell compartment via IL-4 produced in response to helminth (43, 44) or other infections. Indeed, we observed a positive correlation between the frequency of AIMT cells and the percentage of CD122^HIGH cells among AIMT cells in feral mice but not in feral F1 offspring born in captivity. This suggests a role of environmentally transient bacterial or helminth infections in feral mice and highlight the role of genetic background in housed F1 offspring.

Although it has been shown that cohousing of laboratory strains with pet shop mice leads to dramatic changes of their CD8⁺ T cell compartment (18), we could detect only a minor increase of CX3CR1⁺ effector cells in cohoused B6 mice. Otherwise, the CD8⁺ T cell compartment in the spleen, peripheral lymph nodes, and mesenteric lymph nodes of cohoused mice was unaffected. A possible explanation is that our feral mice showed much lower pathogenic burden than the pet shop mice used by Beura et al. (18), which seems logical, given the high animal density and the lack of natural predation at pet shop farms. Although different populations of feral mice might have a different pathogenic burden, our data document that cohousing with feral mice does not always induce dramatic changes in the CD8⁺ T cell compartment. We observed that cohousing with feral mice substantially changes the microbial composition of the intestine and saliva of laboratory strains, including opportunistic pathogens such as Pasteurella and Helicobacter.

Altogether, our data document that AIMT cells represent a large CD8⁺ T cell compartment in mice that is present in germ-free as well as in feral conditions. The frequency of peripherally induced self-reactive CD122^HIGH AIMT cells increases during aging via homeostatic mechanisms in BALB and B6 mice.

We thank Ladislav Cupak for technical assistance and genotyping of mice, Sarka Kocourkova for preparation of RNA-sequencing libraries, Prof. Sai Reddy and Dr. Simon Friedensohn (both from ETH Zürich) for valuable help with designing the assays for TCR repertoire analysis, and Dr. Dominik Filipp (Institute of Molecular Genetics, Prague, Czech Republic) for providing DO11.10 transgenic mice.

The authors have no financial conflicts of interest.