The Neonatal CD8⁺ T Cell Repertoire Rapidly Diversifies during Persistent Viral Infection Free

Cytomegalovirus infection is the most common congenital viral infection worldwide (occurring in ∼1% of all live births) (1). Acquisition of CMV in utero can result in a lifetime of disability, including hearing loss and mental retardation. Interestingly, the severity of disease largely depends upon the gestational age of the individual at the time of infection. For example, neurologic outcomes tend to be most devastating when transmission to the fetus occurs in the first trimester and becomes less severe with progressing age and development (2–5). Also, viral shedding can persist for years in congenitally infected infants, whereas most adults are able to limit CMV replication in a matter of weeks or months and remain largely asymptomatic (6, 7). Although both lines of evidence demonstrate the importance of immune maturation in the control of infection, the major developmental-related immune defects remain undefined.

Previous reports have demonstrated that CD8⁺ T cells are required to control CMV (8–10). The production of new CD8⁺ T cells occurs in the thymus and involves the acquisition of a TCR that is functional but not self-reactive. It is well established that the ability of an individual to respond to specific Ags and mount protective CD8⁺ T cell responses is closely correlated with the level of diversification of TCRs (11, 12), which increases with progressing development (13, 14). Diversity of TCR usage is accomplished by somatic recombination of V, D, and J gene segments and random addition of N-nucleotides (in between germline segments) by the developmentally regulated enzyme TdT. Although the addition of N-nucleotides by TdT is responsible for ∼90% of TCR diversity in adulthood, its expression is restricted early in life (15–17). As a result, the number of unique TCRs initially produced in humans in the first trimester is severely limited and structurally distinct due to a lack of N-nucleotides. Throughout the gestational period, the diversity and length of the TCRs steadily increase with rising expression levels of TdT (14). A similar program of T cell development occurs in mice, and significant levels of TdT expression are not detected in the thymus until 4–8 d after birth (16). Whether individuals in early life have a sufficiently diverse repertoire of CD8⁺ T cells to mount protective immune responses against persistent viral infections is unknown.

To better understand developmental-related differences in the CD8⁺ T cell repertoire, we recently examined the clonal composition of the Ag-specific CD8⁺ T response in different aged mice following acute infection with a recombinant vaccinia virus expressing the immundominant HSV-1 peptide (gB-8p, derived from the gB glycoprotein) (13). We found that in this acute infection, the gB-8p–specific repertoire in neonatal mice was highly restricted and made up of a higher proportion of clonotypes lacking V-D-J junctional diversity. We recently extended this work to show that memory CD8⁺ T cells elicited by an acute neonatal infection respond poorly to a secondary infection later in adulthood (18). Our results demonstrated that neonatal clonotypes are partially locked into the memory pool in adult life. However, upon secondary infection, the restricted repertoire of neonatal memory CD8⁺ T cells is outcompeted by the more diverse adult clonotypes produced later in life. This previous work was limited to studying the effects of transient antigenic stimulation (from acute infection) in secondary lymphoid organs (the spleen).

The goal of this study was to determine the maintenance or turnover of the memory CD8⁺ T cell repertoire in the spleen and tissues of neonatal-infected mice during persistent virus (murine CMV [MCMV]) infection. In contrast to acute infections, we hypothesized that neonatal CD8⁺ T cells may be selectively maintained by continued priming from reactivating Ag, and that these cells may also be locked into nonlymphoid tissues (the brain) following neonatal infection. Indeed, recent studies have shown that memory CD8⁺ T cells can persist at sites of infection [e.g., lung (19), brain (20), gut (21), skin (22)] long after the infectious agents have been cleared. Tissue-resident memory T cells remain detached from the peripheral circulation and thus retain an imprint of the T cell compartment that was available at the time of infection (23–26). Maintaining a distinct pool of memory cells at the site of infection that can respond quickly to reinfection or reactivating Ag has many obvious advantages, but could be detrimental if these tissue-resident CD8⁺ T cells were derived from a less diverse pool of neonatal CD8⁺ T cells early in life.

In this report, we peripherally infected newborn mice with MCMV and examined the repertoire of CD8⁺ T cells in the tissue (brain) and periphery (spleen). Then we asked whether neonatal CD8⁺ T cells remain sequestered in the brain into adulthood or whether new T cells produced later in life, expressing more diverse and optimal TCRs, infiltrate the brain and gradually convert the neonatal repertoire to a more adult-like repertoire. We envisioned a number of possibilities, as follows: 1) the neonatal repertoire is locked into both the brain and the spleen; 2) the neonatal repertoire is locked into the brain, but is replaced by the adult repertoire in the spleen; and 3) the neonatal repertoire is replaced by the adult repertoire in both organs. The results described in this report provide critical insight into how the repertoire of CD8⁺ T cells evolves in nonlymphoid and lymphoid tissue following neonatal CMV infection. Collectively, these findings have important implications for the development of therapeutic strategies to promote enhanced control of CMV in early life.

Pregnant and adult (7–8 wk) C57BL/6 mice were purchased from the National Cancer Institute (Frederick, MD). Pregnant mice were individually housed and monitored daily for births. Only male mice were used for experiments, and all mice were maintained under pathogen-free conditions at Cornell University College of Veterinary Medicine, accredited by the American Association of Accreditation of Laboratory Animal Care. The experiments in this study were performed in strict 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 Institutional Animal Care and Use Committee at Cornell University.

Recombinant MCMV (Smith strain) expressing the MHC class I–restricted CTL epitope HSV gB_498–505 (SSIEFARL; denoted gB-8p in this text), designated MCMV-gB, was generously provided by Dr. L. Cicin-Sain (Helmholtz Centre for Infection Research). The gB-8p peptide was inserted into the MCMV genome at the 3′ end of the MCMV ie2 gene, as previously described (27). MCMV-gB viral stocks were propagated and quantified on M2-10B4 cells (CRL-1972; American Type Culture Collection). Infected cells were overlaid with 1% (w/v) carboxy-methylcellulose, and plaques were enumerated 6 d later. Newborn pups (6–18 h postpartum) and adult mice (2–4 mo) were infected (i.p.) with either 2 × 10² or 2 × 10⁵ PFU, respectively.

After extensive perfusion with PBS, spleens and brains were harvested and single-cell suspensions were prepared by passing through a mesh screen. Lymphocytes from the brain were further separated on Percoll gradient. Cells from both organs were stained for 1 h at 4°C with gB-8p tetramer and conjugated Abs to detect gB-8p–specific CD8⁺ T cells. The gB-8p:Kb tetramer was obtained from the National Institutes of Health Tetramer Core Facility (Emory University, Atlanta, GA). The following conjugated Abs were obtained from either eBioscience or Biolegend: anti-CD8a (clone 53-6.7), anti-CD4 (RM4-5), anti-Vβ10 (B21.5), anti-Vβ8 (F23.1), anti-CD103 (2E7), and anti-CD69 (H1.2F3). Flow cytometry data were acquired on a custom-made FACS LSRII instrument equipped with four lasers, using Diva software (BD Biosciences), and analysis was performed using FloJo software (Tree Star).

Our RT-PCR protocol was adapted from previous studies (28, 29). Single CD4⁻CD8⁺gB-8p⁺Vβ10⁻Vβ8⁺ cells were sorted directly into 96-well PCR plates containing 5 μl cDNA reaction mix using the FACSAria cell sorter system (BD Biosciences). Control wells without sorted cells were included on every plate to identify any possible contamination. The cDNA reaction mix contained 0.25 μl Sensiscript reverse transcriptase (Qiagen, Valencia, CA), 1× cDNA buffer (Qiagen), 0.5 mM 2′-deoxynucleoside 5′-triphosphate (Qiagen), 100 μg/ml tRNA (Invitrogen, Carlsbad, CA), 50 ng oligodeoxythymine_12–18 (Invitrogen), 20 U RNAse OUT (Invitrogen), and 0.1% Triton X-100 (Sigma-Aldrich, St. Louis, MO). cDNA synthesis was performed immediately after sorting by incubating plates at 37°C for 90 min, followed by 5 min at 95°C, and plates were stored at −80°C.

The Vβ8 transcripts were amplified by nested PCR, and the entire 5 μl cDNA reaction was used for the first PCR in a final 25-μl volume containing 1.5 U Taq polymerase in the manufacturer’s 1× Buffer A with 1.5 mM MgCl₂ (Fisher Scientific, Malvern, PA), 200 μM each 2′-deoxynucleoside 5′-triphosphate (Fisher Scientific), and 100 nM external sense Vβ8 primer (5′-GCTGCAGTCACCCAAAG-3′) and external antisense Vβ8 primer (5′- CCAGAAGGTAGCAGAGACCC-3′). The PCR conditions for the first PCR program began with 95°C for 2 min, followed by 35 cycles of 20 s at 95°C, 20 s at 56°C, and 45 s at 72°C, ending with 5 min at 72°C. A 1.5-μl aliquot of the first-round PCR was used for the second PCR with the internal sense Vβ8 primer (5′-GTACTGGTATCGGCAGGAC-3′) and internal antisense Vβ8 primer (5′-GGGTAGCCTTTTGTTTGTTTG-3′). The second PCR program began with 95°C for 2 min, followed by 35 cycles of 20 s at 95°C, 20 s at 56°C, and 45 s at 72°C, ending with 5 min at 72°C. PCR products were resolved on a 2% agarose gel, purified with the MinElute 96 UF PCR purification kit (Qiagen), and sequenced with 12 pmol internal Vβ8 sense primer, using an Applied Biosystems (Foster City, CA) 3730XL DNA Analyzer at the Genomics Facility (Cornell University, Ithaca, NY).

The gB-8p–specific CD8⁺ TCR-β repertoires were characterized by sequentially aligning each TCR-β sequence with the V regions for the Vβ8 (TRBV13 in IMGT nomenclature) genes, followed by the best-match J region and the best-match D region. This analysis was done using the IMGT reference alleles for the Mus musculus TRB genes (30). The CDR3β sequence was then identified between, and inclusive of, the conserved cysteine in the V region and the conserved phenylalanine in the J region. The minimum number of nucleotide additions required to produce a CDR3β sequence was determined by first identifying the nucleotides that could be attributed to the V region at the 5′ end of the CDR3β sequence and then the J region at the 3′ end of the CDR3β sequence. The D regions were subsequently aligned to the nucleotides in the junction between the identified V and J region–encoded portions of the CDR3β sequence, with no less than 2 nt attributed to a d region. Nucleotides in the junctions between the identified V, D, and J region–encoded segments were considered to be nucleotide additions.

The phenotypes and Vβ gene usage of the gB-8p–specific CD8⁺ T cell populations were compared between adult spleen, neonate spleen, and neonate brain using a Mann–Whitney U test for each pairwise comparison between the three groups, with the statistical significance for each pairwise comparison determined at p < 0.0167 (using Bonferroni correction for multiple pairwise comparisons). The features of the gB-8p–specific Vβ8⁺ CD8⁺ TCR repertoires were compared between neonatal-infected and adult-infected mice, or between acute and latent infection, using a Mann–Whitney U test. The features of the gB-8p–specific Vβ8⁺ CD8⁺ TCR repertoires were compared between paired brain and spleen samples using a Wilcoxon test. All statistical analyses were performed using GraphPad Prism software (GraphPad Software, San Diego, CA; Version 6).

The diversities of the gB-8p–specific Vβ8⁺ CD8⁺ TCR repertoires in each organ per mouse were evaluated using two different measures of diversity, the number of unique TCR clonotypes and the Simpson’s diversity index (31):

where n_i is the number of copies of the ith clonotype, c is the number of different clonotypes in the TCR sample, and n is the total number of TCR sequences sampled.

The similarities of the gB-8p–specific Vβ8⁺ CD8⁺ TCR repertoires between brain and spleen were assessed using the number of common clonotypes and the Morisita-Horn similarity index (32):

where f_i=n_1i / N₁ and g_i=n_2i / N_2,n_1i and n_2i are the number of copies of the ith clonotype in samples 1 and 2, and N₁ and N₂ are the total number of TCRs in samples 1 and 2, respectively. The summations in the numerator and the denominator are over the c unique clonotypes in both samples.

The Simpson’s diversity and Morisita-Horn similarity indices account for both the variety of TCR clonotypes and their frequencies. These relative diversity and similarity indices range in value from 0 (minimal diversity/similarity) to 1 (maximal diversity/similarity). To account for the differences in the sizes of the TCR repertoire samples, the diversity of a TCR repertoire or the similarity between TCR repertoires was estimated as the median value of 10,000 random draws of subsamples of 52 TCR sequences from the total TCR repertoires obtained for each sample (31, 32). The diversity and similarity analyses were performed using Matlab (The Mathworks, Natick, MA).

MCMV infection elicits a very broad CD8⁺ T cell response in adult C57BL/6 mice, comprised of at least 19 peptide epitopes (33, 34). To determine whether the epitope recognition patterns are altered in neonatal mice, we first systemically (i.p.) infected newborn and adult mice with wild-type MCMV (Smith strain) and compared the epitope immunodominance hierarchy during acute and latent stages of infection. When splenocytes were stimulated with a panel of 19 peptides early postinfection, we found that the overall epitope hierarchy between neonatal and adult mice was largely similar (Fig. 1A), although the relative dominance of M45- and M57-specific CD8⁺ T cells was significantly higher in adult mice (p < 0.0001 for M45 and p < 0.05 for M57). During latent stages of infection, M45- and M57-specific CD8⁺ T cells contracted, and the CD8⁺ T cell response became mostly comprised of ie3-, M38-, and m139-specific CD8⁺ T cells in both age groups (Fig. 1B). However, the relative abundance of ie3-specific CD8⁺ T cells was significantly greater in neonatal-infected mice (p < 0.0001). We also tracked the CD8⁺ T cell response over time in the blood of individual mice using tetramers to five epitopes and found the more rapid inflation of ie3-specific CD8⁺ T cells in mice infected at birth to be apparent by 2 mo postinfection (Fig. 1C–G). Thus, whereas neonatal- and adult-infected mice appear to target the same general MCMV peptides, the dynamics of the response vary throughout the course of infection.

To better understand how timing of infection alters the dynamics of the CD8⁺ T cell response, we next set out to determine how the TCR repertoire changes during MCMV infection. However, examining the fine specificity of the CD8⁺ T cell response to MCMV is challenging because of the many different epitope-specific CD8⁺ T cells that respond during various stages of infection. To circumvent this issue, we employed a recombinant strain of MCMV that expresses the HSV-1 gB-8p peptide (MCMV-gB). This virus elicits a strong immunodominant gB-8p–specific CD8⁺ T cell response during the acute and latent phases of infection (27).

Previous work by Britt and colleagues (8) elegantly showed that CD8⁺ T cells are critical for the resolution of neonatal MCMV brain infections, and the subsequent pathological abnormalities observed in the brain resemble those seen in human infants (35). Using this mouse model, we systemically (i.p.) infected mice at birth with 200 PFU MCMV-gB and enumerated the number of gB-8p–specific CD8⁺ T cells in the brain and spleen at various times postinfection using tetramers. We also infected a group of adult mice with MCMV-gB, allowing us to better understand how timing of infection may influence the TCR repertoire. However, as adults fail to mount a detectable gB-8p–specific CD8⁺ T cell response to the neonatal inoculum, the adult mice were infected with a higher dose of MCMV-gB (2 × 10⁵ PFU MCMV-gB, i.p.), similar to previous studies (36, 37). This difference in viral dose is also required because the adult dose of MCMV-gB is lethal in newborn pups.

We analyzed the populations of gB-8p–specific CD8⁺ T cells in the spleens and brains of neonate-infected mice. In mice infected as adults, we focused our analysis on the spleen, as very few CD8⁺ T cells infiltrate the brain when infection occurs in adult life. The numbers of splenic gB-8p–specific CD8⁺ T cells remained relatively constant over the course of infection following both neonatal infection (Fig. 2A) and adult infection (Fig. 2B). In the brains of neonatal-infected mice, viral replication peaked on day 14 (Fig. 2C), and the largest population of gB-8p–specific CD8⁺ T cells appeared on day 17, although significant numbers of cells could still be observed as late as day 58 (Fig. 2D, 2E). Thus, large numbers of gB-8p–specific CD8⁺ T cells are maintained in the brain and spleen after neonatal MCMV infection.

To further characterize memory gB-8p–specific CD8⁺ T cells from neonatally-infected animals, we next compared their phenotype to adult-infected animals during latent infection. CD8⁺ T cells that are dominant during latter stages of adult infection typically express the NK cell receptor (KLRG-1) (Fig. 3A). Indeed, the majority of gB-8p–specific CD8⁺ T cells in the brain (∼70%) and spleen (>80%) of neonatal-infected mice expressed KLRG-1, indicating these cells were also highly differentiated (Fig. 3A). In contrast, only the gB-8p–specific CD8⁺ T cells in the brain expressed the integrin CD103 and the early activation marker CD69 (Fig. 3B, 3C). Interestingly, the proportion of cells expressing CD103 in the brain increased from day 17 (∼35%) to day 58 (∼60%) (data not shown), which is consistent with their conversion into more of a tissue-resident memory CD8⁺ T cell. Together, these data demonstrate a phenotypic dichotomy between gB-8p–specific memory CD8⁺ T cells found in the brain and spleen following MCMV infection.

To gain insight into how the composition of CD8⁺ T cells differs in the brain and spleen over time, we compared their usage of different Vβ TCRs during the acute and latent phase of infection. Following infection, we analyzed Vβ usage of tetramer-positive cells using Abs against Vβ10 and Vβ8, because these TCRs have previously been shown to dominate the gB-8p–specific CD8⁺ T cell response (13, 29). Our data showed a slight bias toward Vβ8-expressing TCRs in both the spleens and brains of neonatal mice compared with adult mice during acute stages of infection (Fig. 4A, 4B). Interestingly, Vβ8 TCRs became even more pronounced in neonatal-infected mice during latency, particularly in the spleen (Fig. 4C, 4D). We also observed extensive variability in Vβ usage in neonatal-infected mice compared with adult-infected mice, which is substantially reduced in the spleen with progressing infection. For example, Vβ10 usage in the spleens of neonatal-infected mice ranged from 1 to 99% during acute stages of infection, but only varied from 2 to 20% during latent stages of infection. These results indicate that the repertoire of gB-8p–specific CD8⁺ T cells responding to infection in early life is uneven and highly variable.

To examine how the gB-8p–specific TCR repertoire changes during the course of MCMV infection requires both a sufficient depth of sampling of the TCR repertoire and a high level of confidence in the clonotype sequences, and their copy numbers, that comprise each TCR repertoire sample. We therefore focused on a specific portion of the TCR repertoire, using single-cell sorting and Sanger sequencing of Vβ-specific TCRs. We chose to focus on the Vβ8⁺ TCRs as these become dominant during the latent phase of infection. In order for us to establish a starting baseline for age-related differences in the TCR repertoire, we first compared the clonal composition of gB-8p–specific Vβ8⁺ TCRs found in the spleens of neonatal- and adult-infected mice during the peak of the primary response. Following infection of neonatal and adult mice with MCMV-gB, we sorted individual gB-8p–specific Vβ8⁺ CD8⁺ T cells from the spleens and sequenced their respective TCRs. Because most of the TCR diversity resides in the hypervariable CDR3 spanning the junction of V, D, and J gene segments, we defined individual clonotypes by their V and J gene usage and the CDR3 sequence. We analyzed >50 TCR-β sequences per mouse, in six to eight mice per age group, for a total of 1116 TCRβ sequences (Table I). Examples of the gB-8p–specific Vβ8⁺ TCR-β repertoires in neonatal- and adult-infected mice are provided in Table II.

To compare the levels of TCR diversity between mice infected at different ages, we used a previously published approach to standardizing the sample sizes (i.e., numbers of TCR sequences per mouse) and enumerated the number or diversity of unique gB-8p–specific Vβ8⁺ TCRs in neonatal- and adult-infected mice. As shown in Fig. 5A, we found significantly lower numbers of distinct clonotypes in the spleens from neonatal-infected mice compared with adult-infected mice (median: 3.5 versus 17 different clonotypes). Although the gB-8p–specific Vβ8⁺ TCR repertoire demonstrated an immunodominance hierarchy of different clonotypes in both adult- and neonatal-infected mice, the dominant clonotypes in neonatal-infected mice comprised an average of 70% of the gB-8p–specific repertoire. In contrast, the dominant clonotypes in adult-infected mice made up only an average of 26% of the repertoire. To account for these differences, we also evaluated the diversity of gB-8p–specific Vβ8⁺ TCR repertoires using Simpson’s diversity index, which takes into account both the number of distinct TCR clonotypes and their clonal dominance hierarchy. As with the number of TCR clonotypes, the Simpson’s diversity index was found to also be significantly lower in neonatal-infected mice (Fig. 5B). These data indicate that the gB-8p–specific Vβ8⁺ TCR repertoire in the spleens from neonate-infected mice is significantly less diverse than that in adult-infected mice, both in terms of the number of clonotypes and the evenness of clone size distribution.

We hypothesized that the limited number of distinct TCRs present in neonatal-infected mice could be due to the reduced amounts of TdT-mediated N-addition during early development. Thus, we also examined the extent to which the repertoire was comprised of germline-encoded clonotypes. A germline-encoded clonotype was considered to be a TCR-β sequence lacking nucleotides in the junctions between Vβ, Dβ, and Jβ gene segments. As shown in Fig. 5C, we found a significantly greater percentage of the neonatal gB-8p–specific Vβ8⁺ TCR clonotypes to be germline encoded compared with adults (median: 80.0% versus 22.9% clonotypes). We also assessed whether the lack of N-additions in neonatal mice resulted in shorter CDR3β lengths. In both age groups, the dominant CDR3β length was 13 aa. Although there was a trend toward shorter CDR3β lengths in neonatal mice, these differences depended on a small number of mice responding with CDR3β lengths of 12 aa (Supplemental Fig. 1H–K) and were not statistically significant. We also did not observe significant differences between the age groups in terms of usage of the three Vβ8 genes (Supplemental Fig. 1A–D). When we analyzed the diversity of Jβ gene usage in neonatal-infected mice, we found that this varied widely across different mice (Supplemental Fig. 1E–G), largely due to the small numbers of TCR clonotypes in the neonatal mice. By contrast, Jβ gene usage in the adult-infected mice was consistently diverse. Thus, the defining features of gB-8p–specific Vβ8⁺ CD8⁺ T cell repertoires in the spleens of neonatal-infected mice compared with adult-infected mice during early infection are that they are less diverse and comprised of more germline-encoded clonotypes.

Because we observed a significant difference in the diversity and specificity of the repertoire between adult and neonatally-infected mice during the acute stages of MCMV-gB infection, we next asked whether the characteristics associated with neonatal Vβ8⁺ TCRs persist in the spleen during latent infection. We suspected that neonatal Vβ8⁺ TCRs either became locked into the spleen for extended periods of time or were replaced by adult Vβ8⁺ TCRs that were later recruited into the response. To differentiate between these possibilities, we compared the clonal composition of gB-8p–specific Vβ8⁺ TCRs in the spleens of both age groups at 8 wk postinfection. The gB-8p–specific Vβ8⁺ TCR repertoire data used for this comparison are summarized in Table III, and a representative mouse/age group is depicted in Table IV.

Our analysis revealed a striking similarity in the features of the gB-8p–specific TCR-β repertoire in the spleens of neonatal-infected and adult-infected mice during latent infection. For example, the number of different clonotypes (Fig. 6A) and Simpson’s diversity index (Fig. 6B) were no longer significantly different between the age groups at 8 wk postinfection. This was largely due to the diversification of the repertoire in neonatal-infected mice. In addition, there was a reduction in the number of germline-encoded clonotypes present during early stages of neonatal infection (Fig. 6C). Overall, the general characteristics of the gB-8p–specific Vβ8⁺ TCR repertoires between neonatal- and adult-infected mice were similar (Supplemental Fig. 2). Thus, there was no evidence to suggest that neonatal MCMV-gB infection locks in a neonatal-like T cell repertoire in the spleen. Instead, our results indicate that neonatal clonotypes are replaced by adult clonotypes with progressing infection.

Tissue-resident memory cells in the brain have been shown to persist for long periods of time and are thought to be independent of the pool in the peripheral circulation (38). To investigate this, we compared the similarity of the TCR repertoires in the brain and spleen at different times postinfection. We hypothesized that the repertoires in the spleen and brain would be similar early after neonatal infection, because both populations were being formed at this time. However, if the resident memory cells in the brain were anatomically isolated, they would retain the features of the neonatal repertoire in latent infection despite the loss of these features in the splenic repertoire. As a consequence, we would expect the repertoires in the brain and spleen to diverge, as the spleen adopted more adult characteristics, and the brain maintained the neonatal character.

To investigate whether the neonatal gB-8p–specific Vβ8⁺ TCR repertoire was locked into the brain following neonatal MCMV infection, we first compared the repertoire characteristics of cells from each site. A total of 1111 Vβ8⁺ TCR sequences was obtained from the brains of infected mice for this analysis (Tables I, III). Because we were unable to sample brain and spleen repertoires longitudinally, we compared characteristics in the brain and spleen of individual animals both at the peak of the CD8⁺ T cell response (17 d postinfection) as well as at 8 wk postinfection. We found no significant differences between brain and spleen in the broader features of the total gB-8p–specific Vβ8⁺ TCR repertoires, including V gene usage, J gene usage, CDR3 length, extent of germline encoding, or clonotypic diversity in early infection (Fig. 7G, 7H, Supplemental Fig. 3D, 3G, 3K, 3O) or late infection (Fig. 7I, 7J, Supplemental Fig. 4D, 4G, 4K, 4O). However, at the level of the unique TCR clonotypes comprising the splenic and brain TCR repertoires, we observed several important differences.

First, although all the observed splenic clonotypes used either Vβ8.1 or Vβ8.2, we observed low-abundance clonotypes using Vβ8.3 that were unique to the brain in both early and late infection (Supplemental Figs. 3A, 4A, 7A, 7B). Secondly, in later infection, a significantly higher proportion of unique TCR clonotypes in the brain could be made with no N-additions (Fig. 7E). Although there was a substantial increase in the diversity of TCR clonotypes in both organs due to the recruitment of nongermline-encoded clonotypes in later infection, there was greater persistence of the neonatal-like germline-encoded clonotypes in the brain, as indicated by the proportion of germline-encoded TCR clonotypes unique to the brain in many of the mice in Fig. 7B (dark blue segments). However, these germline-encoded clonotypes that were present only in the brain were not always dominant in the repertoire.

To further investigate whether the TCR repertoire in the spleen had diverged from the TCR repertoire in the brain during chronic infection, we also performed a test of the similarity of the brain and splenic repertoires during early and late infection. We found that the repertoires in the brain and spleen were significantly more similar to each other in early infection than in later infection, as assessed by the Morisita-Horn similarity index (Fig. 8B). However, there were significantly more TCR clonotypes in common between the brain and spleen in late infection compared with early infection (Fig. 8A). Similar results were obtained when evaluating similarity at the level of the TCR nucleotide sequence (data not shown). In early infection, the common clonotypes tend to have more similar numbers of copies in both the brain and spleen. In later infection, it appears that the increased diversity of the repertoires leads to some of the same new clonotypes in the brain and spleen, but that these new common clonotypes are not necessarily dominant in both the organs.

Collectively, these results suggest that there is a high level of similarity between the CD8⁺ T cell pools in the brain and spleen in early infection, and that substantial similarity between these compartments is maintained as the TCR repertoire diversifies in late infection. However, it also appears that neonatal-like germline clonotypes are locked into the brain, but that they often do not dominate the repertoire.

That the immune responses to vaccines and infections in neonates are sometimes less effective than in adults suggests that priming in early life may recruit cells with lesser ability to respond and control infection. Previously, we showed that acute neonatal infections partially lock in a highly restricted repertoire of low-avidity memory CD8⁺ T cells that are incapable of competing with the more diverse repertoire that arises in adult life (18). In the current study, we investigated how chronic infection affects the persistence of the neonatal CD8⁺ T cell response, as well as whether a restricted anatomical location locks in the neonatal response. In particular, we focused on the neonatal CD8⁺ T cell response to the immunodominant HSV-gB-8p epitope in the context of MCMV infection in the spleen and brain.

One particular aspect of the neonatal gB-8p–specific CD8⁺ T cell response that we were interested to investigate was the TCR repertoire. However, to effectively assess the changes over time in a TCR repertoire requires both a sufficient depth of sampling of the TCR repertoire and also high-quality data for each sample. We therefore chose to use single-cell sorting and Sanger sequencing to focus on a narrower portion of the TCR repertoire, specifically the Vβ8⁺ TCRs that dominated the gB-8p–specific CD8⁺ T cell response in latent infection. To obtain a comparable depth of sampling of the broader TCR repertoire (i.e., analyzing all Vβ) using this high-quality experimental approach would not have been feasible. Although next-generation sequencing would have feasibly provided a sufficient depth of sampling of the broader TCR repertoire, the high error rates and potential biases associated with high-throughput approaches would not have allowed us the same high degree of confidence in the determination of the clonotype sequences and their frequencies as our current approach.

We first investigated whether the neonatal gB-8p–specific TCR repertoire is locked in the spleen during chronic MCMV infection by comparing the splenic TCR repertoires between neonate- and adult-infected mice during early and late infection. Apart from the TCR repertoire diversity and the extent of germline encoding, there was very little difference in the features of the gB-8p–specific Vβ8⁺ TCR repertoires between neonates and adults during early stages of infection. However, these differences were sufficient for us to determine that the neonatal-like TCR repertoire did not persist in the spleen during chronic infection. The clonal structure of gB-8p–specific Vβ8⁺ TCRs in both age groups was characterized by a prevalent Vβ8.1 gene usage, diverse array of J gene usage, and a CDR3 length of 13 aa. Also, a RGQ motif in CDR3 position 4–6 was found in the majority of clonotypes from neonatal- and adult-infected mice (see example repertoires in Tables II, IV), suggesting clonotypes in early life are capable of conforming to the necessary pMHC structural constraints and some of the same gB-8p–specific Vβ8⁺ clonotypes observed in the adults can also be found in neonatal mice. Thus, it is somewhat surprising that neonatal clonotypes do not remain prevalent during persistent infection with MCMV. Although the underlying basis for why new clonotypes need to be recruited into the CD8⁺ T cell response with progressing infection is unclear, several possibilities are worth mentioning.

First, it is possible that adult CD8⁺ T cells produced later in life may exhibit higher-avidity TCRs, which allow them to preferentially respond to reactivating Ag. Indeed, others have argued an affinity selection process exists during latent CMV infection, and those clonotypes exhibiting high avidity for pMHC are most competitive (39, 40). Previous reports have demonstrated that mice with broader CD8⁺ T cell repertoires are more easily able to mobilize protective T cells exhibiting higher TCR avidity (41–44). In this study, we infected neonatal mice with MCMV prior to the expression of TdT and diversification of the repertoire. Thus, it is possible that the initial pool of neonatal CD8⁺ T cells that are recruited into the response does not exhibit a sufficient amount of diversity to include high-avidity TCRs. Also, even if the same TCR-β clonotypes are observed in neonate and adult mice, they may confer higher avidity when paired with different TCR-α chains, of which there will be greater diversity in the adult mice. This explanation would be consistent with our previous findings showing neonatal memory CD8⁺ T cells elicited by acute infection are comprised of lower-avidity TCRs than adult memory CD8⁺ T cells (18).

Secondly, there may be substantial turnover of TCR clonotypes regardless of the timing of infection, although we are less able to detect these changes in the adult repertoire. That is, we are able to conclude that the gB-8p–specific repertoire rapidly turns over in neonatal-infected mice because we saw a reduced prevalence of germline-encoded clonotypes and an increase in TCR diversity. Adult-infected mice, in contrast, are comprised of clonotypes that appear to be structurally similar throughout the response, precluding us from drawing any conclusions about how dynamic the repertoire is during latency. However, we did observe a small increase in Vβ8⁺-expressing cells and a modest decrease in TCR repertoire diversity in adult-infected mice. These data suggest that the repertoire might be narrowing over time, which would be in line with previous reports describing longitudinal changes in the repertoire of CD8⁺ T cells responding to human CMV in healthy adults (40).

Third, the neonatal repertoire may convert to an adult repertoire because neonatal clonotypes rapidly become functionally exhausted. Earlier studies performed in adult-infected mice have demonstrated that the memory CD8⁺ T cell pool is extensively differentiated, but remains fully functional (45). However, if neonatal mice have an insufficient repertoire or number of precursors to control infection, then repeated stimulation during persistent infection could force these cells into a state of dysfunction. We therefore compared the expression levels of PD-1, a marker associated with exhaustion, but did not observe increased expression of PD-1 in neonatal cells at day 17 postinfection (data not shown). We also measured the proportion of tetramer-positive cells that are capable of producing IFN-γ during the acute stages of infection and again found no evidence of immune dysfunction in neonatal cells responding to MCMV (data not shown). Thus, we have yet to obtain any correlation between the culling of the neonatal repertoire and the loss of immune capabilities during persistent infection.

Finally, it is entirely possible that the rapid turnover of neonatal clonotypes is due to cell-intrinsic differences that are independent of developmentally related changes in TCR repertoire. In early life, the thymus is seeded by a different pool of hematopoietic stem cells and generates fewer distinct naive CD8⁺ T cells, which in turn undergo less postthymic maturation and more increased homeostatic proliferation. Thus, there are a number of factors that could confound age-related differences in the CD8⁺ T cell response. We used TCR sequencing in this study to track the CD8⁺ T cell repertoire during infection. The main advantage of this approach is that it accounts for age-related changes in the CD8⁺ T cell pool and allows us to monitor the endogenous pool of neonatal cells. However, additional studies are required to uncover the precise mechanism that underlies the loss of neonatal clonotypes during MCMV infection.

We next investigated whether a neonatal-like TCR repertoire of tissue-resident memory CD8⁺ T cells is locked in the brain following neonatal MCMV infection. Tissue-resident memory is believed to be a nonmigrating pool of memory CD8⁺ T cells that are derived from cells that were present during the initial primary infection. Thus, we were surprised to find that, although there was some evidence that neonate-like clonotypes were locked in the brain, there were substantial and similar changes to the TCR repertoires in the spleen and brain. These changes included a reduction in the proportion of germline-encoded clonotypes comprising the TCR repertoire and an increase in TCR clonotype diversity in both the spleen and brain during chronic infection. One possible explanation for these parallel changes is interaction between the brain and spleen. The brain is an immunoprivileged site and is not typically patrolled by lymphocytes. However, in response to viral infection, large numbers of Ag-specific CD8⁺ T cells migrate to the brain, which may facilitate an exchange of cells. However, we cannot exclude that the similarities of the repertoires in the brain and spleen arose due to seeding of both sites from a common source, rather than a direct exchange.

Once the Ag has been cleared, a small subset of cells, termed tissue-resident memory cells, persists in the brain for long periods of time. However, despite expressing the appropriate tissue-resident memory phenotype (CD103⁺, CD69⁺) and the greater persistence of neonate-like clonotypes in the brain, the overall composition of the gB-8p–specific CD8⁺ T cell repertoires in the brain during later stages of MCMV infection mirrored the adult-like repertoire of CD8⁺ T cells found in the spleen. Thus, it is interesting to speculate on why neonatal cells also fail to form long-lived tissue-resident memory cells in the brain.

One possibility is that immune surveillance of CMV in tissues, such as the brain, is largely confined to intravascular T cells. A recent report by Smith et al. (36) elegantly showed that memory T cell priming predominantly occurs in the circulation, rather than the parenchymal tissues. Although we extensively perfused the brain before cell isolation, it is possible that some blood-borne lymphocytes persist in the extensive circulatory system of this tissue. Indeed, there are examples in the literature of CD8⁺ T cells being confined to the vasculature in peripheral organs, despite perfusion (46). Another possibility is that the turnover of memory CD8⁺ T cells may vary among different pathogens. We initially chose to work with MCMV because it is a natural mouse pathogen, immune protection depends on CD8⁺ T cells, and large numbers of CD8⁺ T cells localize to the brain following neonatal infection. However, the behavior of tissue-resident memory cells has not been closely examined in the context of MCMV, which may exhibit different dynamics of Ag presentation and cell tropism than more commonly studied brain tropic pathogens (vesicular stomatitis virus, HSV). Thus, it will be important to determine whether the limited establishment of tissue-resident memory CD8⁺ T cells in the brain following neonatal MCMV infection relates to the timing of infection or the nature of the pathogen that is involved.

In addition to showing how the clonal composition of CD8⁺ T cells responding to an individual epitope changes over time, we also examined CD8⁺ T cell epitope recognition patterns at the bulk population level. Interestingly, we found that inflation of memory CD8⁺ T cells was more rapid in mice infected at birth compared with those infected later as adults. Although the immunodominance hierarchy in neonatal mice was similar to that observed in adult mice, we observed an increase in the accumulation of ie3-specific CD8⁺ T cells over the course of infection. These findings are significant, because persistent viral infections have been shown to disrupt the balance of naive and memory T cells and alter the homeostasis in older adults. Years and decades following initial viral exposure, memory CD8⁺ T cells accumulate or inflate, which can lead to a loss in the number and repertoire diversity of T cells that are available to respond to new pathogens. Our results suggest that these changes may be exaggerated in the neonate, when the naive T cell compartment is not yet fully formed. These data also warrant future investigation to determine whether congenital CMV infection is associated with more rapid T cell aging and impaired ability to mount new immune responses in later life.

Our investigations of the phenotypic and TCR repertoire characteristics in the spleen and brain during chronic MCMV infection suggest that there is some evidence for the locking in of neonatal Ag-specific CD8⁺ T cells in the brain. However, this is accompanied by substantial diversification of the Ag-specific CD8⁺ T cell populations in both the brain and spleen following neonatal infection, as well as possible interaction between the compartments. With the brain being a major target in congenital CMV infection, these findings provide important insights into the evolution of CD8⁺ T cell responses to neonatal CMV infection and also identify several key questions for future investigation.

We thank the University of Rochester Flow Cytometry Core Facility for expert sorting assistance and the National Institutes of Health Tetramer Facility at the Emory University for proficient tetramer production.

The authors have no financial conflicts of interest.