Memory T Cells Specific for Murine Cytomegalovirus Re-Emerge after Multiple Challenges and Recapitulate Immunity in Various Adoptive Transfer Scenarios

Latent CMV is present within a large percentage of the population but is effectively controlled by the immune system (1–6). However, in transplant patients, immune suppression can allow CMV reactivations to progress to disease and increase mortality. Despite the advancements of antiviral medications, long-term prevention of CMV disease is dependent on the reconstitution of CMV-specific immunity, which can be achieved through adoptive immunotherapy (5–18).

In adoptive immunotherapy, healthy CMV-seropositive donors provide CMV-specific T cells to an immunosuppressed recipient. Because of the persistent nature of CMV infection, CMV-seropositive donors accumulate large numbers of CMV-specific CD8⁺ T cells (∼5–10% of the total CD8⁺ T cells), a process termed “memory inflation” (19–28). Studies in humans and the well-characterized murine CMV (MCMV) model have shown that the majority of inflationary populations are composed of terminally differentiated effector phenotype CD8⁺ T cells (T_EFF) that presumably develop as a result of repeated Ag stimulation and may not possess the proliferative or survival capacity necessary for long-term maintenance of CMV immunity (22, 27, 29–34). Interestingly, however, a fraction of these inflationary T cells retain a memory-like phenotype (memory phenotype CD8⁺ T cells [T_M]), despite sharing epitope specificity and TCR sequences with the T_EFF subset (23, 25, 35–37). Studies with other infection models have shown that such a memory phenotype can identify cells that have stem cell–like characteristics (38, 39). If this model holds true for CMV immunity, the CMV-specific T_M would be ideal to use in an adoptive immunotherapy setting. Recent evidence supports this hypothesis. In a nonhuman primate model, CMV-specific effector T cells that were expanded in vitro from sorted T_M had a superior ability to survive after adoptive transfer (40). Moreover, a human study showed a positive correlation between the presence of CMV-specific T_M in a donor transfer and the long-term maintenance of donor-derived cells (41).

The goal of our study was to use the mouse model (MCMV) to directly address the capacity of the CMV-specific T_M population to restore long-term CMV-specific immunity after transfer. Importantly, we found that the MCMV-specific T_M share a common genetic program with their human CMV (HCMV)–specific counterparts and that these cells could repeatedly restore long-term CMV-specific immunity under a spectrum of transfer scenarios. Our data suggest that adoptive immunotherapy with CMV-specific T_M will improve consistency and clinical outcomes in patients at risk for development of CMV disease.

Unless otherwise indicated, C57BL/6 (B6), CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ), Thy1.1 (B6.PL-Thy1^a/CyJ), and Rag^−/− mice (B6.129S7-Rag1 tm1Mom/J) were purchased from The Jackson Laboratory. OT-I transgenic mice (C57BL/6-Tg[TcraTcrb]1100Mjb/J), also purchased from The Jackson Laboratory, were bred with CD45.1 mice to produce double-positive (CD45.2⁺/CD45.1⁺) OT-I mice. All protocols were approved by the Thomas Jefferson University Institutional Animal Care and Use Committee.

Unless otherwise indicated, mice were infected i.p. with 2 × 10⁵ PFU MCMV strain MW97.01 (42). Mice were considered latently infected at 8 wk postinfection. Rag^−/− mice were infected with 5 × 10⁴ PFU MCMV-TK virus (43). OT-I T cell transfer recipients were challenged with 2 × 10⁵ PFU MCMV-SL8, which expresses the SIINFEKL peptide (44, 45).

MHC tetramers were provided by the National Institutes of Health Tetramer Core Facility (http://tetramer.yerkes.emory.edu) and have been described previously (27). Staining was performed as described previously (27) with tetramers and the following Abs: [CD8(53-6.7); CD44(IM7); CD27(LG.3A10); CD127(A7R34); KLRG1(2F1); CD62L(MEL-14); CD45.1(A20); CD45.1(104); Thy1.1(OX-7); Thy1.2(30-H12); IFN-γ(XMG1.2); TNF-α(MP6-XT22); CD107a(1D4B)]. In all cases, samples were collected on an LSR II and analyzed with FlowJo software (Tree Star). The gating strategy for phenotypic characterization of tetramer⁺ CD8⁺ T cells involved first gating lymphocytes and then singlets. CD8⁺ cells were gated as frequency of singlets. Tetramer⁺ cells were identified as a frequency of CD8⁺ cells. A B8R tetramer (specific for the B8R peptide from Vaccinia) was used as a negative tetramer control. Tetramer⁺ cells were phenotypically defined by their expression of KLRG1, CD27, CD127, or CD62L.

CD8^pos splenocytes from latently infected donors were enriched using EasySep Biotin selection kit (Stemcell Technologies) and biotinylated Abs against RBCs(Ter119), CD4(GK1.5), and CD19(6D5) according to the recommended protocol. Enriched cells were stained to determine the frequency of tetramer⁺ cells within the enriched fraction and then sorted on either a MoFlo (Dako Cytomation) or an ARIA II (BD Biosciences) cell sorter. Sorted cells were counted, and 5 × 10⁴ cells were transferred via the retro-orbital sinus. Sort purity was analyzed on an LSR II. The number of transferred tetramer-binding CD8⁺ T cells was estimated using the tetramer frequency within the enriched CD8⁺ population and the postsort purity analysis. Fold change was calculated as the number of tetramer-binding T cells in the spleen 7 d postchallenge over the total number of tetramer⁺ cells transferred (assuming 100% engraftment). The gating strategy for analyzing donor cells in the recipients was identical to that described earlier with Abs specific for the relevant congenic marker (CD45.1 or Thy1.2).

For OT-I adoptive transfers, splenocytes from naive mice containing 600 OT-I T cells were transferred. Recipients were challenged with MCMV-SL8. To establish secondary and tertiary populations, we FACS sorted and transferred OT-I T_M as described in the legend for Fig. 5 and in Supplemental Fig. 1. After challenge with MCMV-SL8, the frequencies of donor OT-Is were determined in the blood of recipients using the strategy described earlier except that singlets were not identified and OT-I donors were identified by expression of CD45.1 and Vα2.

Intracellular stimulation and staining were performed as previously described (27, 45), with minor modifications. Specifically, cells were incubated with 1 μg/ml peptide (Genemed Synthesis), 1 μg/ml brefeldin A (GolgiPlug; BD Biosciences), and CD107a-specific Ab for 3 h.

CD70 Ab blockade was performed as previously described (46), with minor modifications. In brief, mice received either 150 μg anti-CD70(FR70) or control rat IgG2b (both from BioXCell) via the i.p. route. Injections were administered at days −1, 0, and 3 postinfection.

Ab depletions were performed with Thy1.1(19E12), CD4(GK1.5), and NK1.1(PK136) Abs. A total of 300 μg of each Ab was administered i.p. in PBS. Three subsequent injections of 100 μg of each Ab were given at 7-d intervals.

Splenocytes from latently infected mice were costained with tetramers loaded with the antigenic peptides from M38, m139, and IE3 (25) and sorted on a MoFlo (Dako Cytomation) cell sorter. MCMV-specific T cells were identified as CD8⁺, CD44^hi, and tetramer binding, and then further segregated into T_M and T_EFF subsets by expression of KLRG1 and CD127. Naive CD8⁺ cells were CD44^lo. Total RNA was isolated using the Qiagen RNeasy Plus Kit (Qiagen), quantified on a NanoDrop 2000c Spectrophotometer (Thermo Scientific), and processed at the Microarray Core Facility at Thomas Jefferson University. In brief, 2.5 μg fragmented and biotinylated cDNA was hybridized to Mouse gene 1.0 ST array (Affymetrix). Chips were scanned on an Affymetrix Gene Chip Scanner 3000, and data were analyzed using the R programming language and various packages from Bioconductor (47). The oligo package (48) was used to extract expression data from the Affymetrix CEL files and perform background and RMA normalization (49). Annotation information was added using the mogene10sttranscriptcluster.db (50) package. Probes without valid annotations (7196 of 35,556 probes) were removed before differential expression analysis using the Limma package’s (51) linear modeling and Bayes methods (52). Genes showing upregulation or downregulation of at least 2-fold and p < 0.05 in each of three contrasts (T_EFF versus naive, T_M versus naive, and T_EFF versus T_M) were considered for gene set enrichment analysis (GSEA). Microarray data have been deposited in the Gene Expression Omnibus (GEO) database (53) under accession no. GSE61927.

Human data for series GSE24151 (54) were retrieved from National Center for Biotechnology Information’s GEO database (53), extracted using Partek Genomics Suite software, version 6.6 (Partek, St. Louis, MO) and curated for input into GSEA software (55) (http://www.broadinstitute.org/gsea). Because the data for GSE24151 have been deposited in GEO as log₁₀ ratios of the reference pool to sample, each value was inverted by multiplying by −1. Gene names in the six mouse gene lists (upregulated or downregulated in each of the three contrasts described earlier) were converted to human names using data from National Center for Biotechnology Information’s HomoloGene database, Release 68 (http://www.ncbi.nlm.nih.gov/homologene). The converted gene lists along with genes specific to the liver and the TCR pathway from the Molecular Signature Database (MSigDB) (55) were analyzed for enrichment in the human data using recommended settings for the GSEA command-line interface.

In the mouse model, MCMV infection of B6 mice results in inflation of select MCMV-specific CD8^pos T cells specific for peptides from the viral proteins M38, m139, and IE3 (Fig. 1A) (25, 27, 28). As in humans infected with HCMV, the majority of MCMV-specific inflationary T cells express a T_EFF phenotype (often defined as terminally differentiated CD8⁺ T cell phenotype in humans), whereas only a small fraction express a T_M-like phenotype, defined in this article as KLRG1^lo/CD27^hi and further subdivided into central memory CD8⁺ T cells (T_CM: CD127^hi/CD62L^hi) and effector memory CD8⁺ T cells (T_EM: CD127^hi/CD62L^lo) subsets (Fig. 1A) (22, 23, 27, 29–33, 56). In contrast, noninflationary MCMV-specific CD8^pos T cell responses, represented by the response against the viral protein M45, contract after acute infection and are thought to be maintained by homeostatic mechanisms thereafter (Fig. 1A) (25, 27, 57). As expected, noninflators express a predominately memory (T_M) phenotype, which also includes both T_CM and T_EM subsets (Fig. 1A) (23, 27).

It remains unknown whether the constant immune stimulation needed to maintain memory inflation causes a decline of the T_M subset within inflationary populations over time. Using infection-matched cohorts, we found that the numbers of T_M that were specific for inflationary Ags were stable over time and remarkably similar to the numbers of noninflationary T_M, despite great differences between the numbers of inflationary and noninflationary T_EFF (Fig. 1B, 1C). Thus, although continuous Ag stimulation maintains memory inflation, the inflationary T_M population remains stable.

The MCMV model is well characterized and the T cell responses clearly recapitulate those seen in HCMV-infected people. To determine whether MCMV-specific T_M and T_EFF share a common transcriptional program with their human counterparts, we sorted MCMV-specific T_M (CD44^hi/CD127^hi/KLRG1^lo) and T_EFF (CD44^hi/CD127^lo/KLRG1^hi) specific for the M38, m139, and IE3 Ags. Microarray analyses were performed on these cells. Genes that were significantly upregulated or downregulated in T_M and T_EFF subsets relative to each other or to naive (CD44^lo) T cells were mapped to the corresponding human genes and compared with the profiles of HCMV-specific T cells, previously defined by the van Lier group (54) as CD27^hi/CD45RA^lo (T_M) or CD27^lo/CD45RA^hi (T_EFF). The CD27 and CD127 (IL-7Rα) molecules both mark CMV-specific T cells with a memory phenotype in mice and humans (27, 29, 32, 58, 59), and nearly all MCMV-specific KLRG1^lo/CD27^hi cells (T_M) coexpressed CD127 (either T_CM or T_EM; Fig. 1A). Gene set enrichment analyses (GSEA) were used to measure the overall correlation between the mouse and human gene expression data. As shown in Fig. 2A, genes that distinguished mouse T_EFF and T_M from each other were highly enriched within the corresponding human data set; that is, genes upregulated specifically in mouse T_M relative to mouse T_EFF were highly enriched within the genes that distinguish human T_M from human T_EFF and vice versa. Moreover, relative to naive T cells, mouse genes that were upregulated and downregulated by T_EFF or T_M were highly enriched within genes that distinguished their human counterparts from human naive T cells (Fig. 2B). The analyzed mouse genes and the core enrichment profiles of each comparison are listed in Supplemental Table I. Importantly, several of these genes corresponded to our sorting parameters and the known phenotypes of T_M and T_EFF populations. As controls, identical analyses were performed with genes associated with the TCR signaling pathway or liver, and the data exhibited expected patterns (Fig. 2B).

Overall, these data show that MCMV- and HCMV-specific T cells share a common genetic program, validating the use of the MCMV model to investigate the function of HCMV-specific T cells. To our knowledge, this is the first direct comparison of MCMV-specific and HCMV-specific T cell gene expression profiles.

To test the proliferative capacity of the T_M and T_EFF, both populations were sorted from spleens of latently infected B6 mice (>3 mo postinfection) using their differential expression of KLRG1 and CD27. Sorted cells were transferred into naive congenic recipients and rechallenged. The M45- and M38-specific T_M proliferated robustly within 7 d after challenge, each expanding almost 1000-fold in the spleen alone, assuming 100% engraftment of the donor cells (Fig. 3A, 3B). In contrast, the M38-specific T_EFF population expanded <10-fold in the same period. Importantly, whereas the T_EFF donor cells remained exclusively KLRG1^hi, the T_M donor cells produced large numbers of both T_EFF and T_M progeny (Fig. 3C). In fact, donor M45- and M38-specific T_M were present in the spleen 7 d after challenge at numbers that were ∼50- to 100-fold higher than had been transferred (Fig. 3D, dotted line), indicating expansion of this subset without terminal differentiation. These data show that MCMV-specific T_M retain robust proliferative capacity and can produce phenotypically diverse progeny including new T_M.

Recent work has shown that interaction between CD27 and its ligand CD70 plays a functional role in the proliferation of MCMV-specific inflationary T cells (46). To test the contribution of this interaction specifically within the T_M population, we sorted and transferred T_M as described earlier and blocked the CD27–CD70 interaction as described in 2Materials and Methods. Blocking the CD27–CD70 interaction significantly decreased the expansion of the M38- and M45-specific T_M 7 d postchallenge by ∼4- to 6-fold (Fig. 3E), which is in line with the impact of CD70 blockade on unsorted (i.e., combined T_M and T_EFF populations) inflationary T cells (46). These data further suggest that the majority of proliferative potential of inflationary T cells is contained within the minor T_M subset. It should be noted that even in the presence of CD70 blockade, the T_M population retained a proliferative capacity that was greater than the T_EFF population, suggesting that additional pathways contribute to the total proliferative potential of these cells (Fig. 3B, 3E) (M. Quinn and C.M. Snyder, unpublished observations).

To determine the ability of the T_M donor cells to persist long term, we tracked the progeny from T_M donor cells in the blood after rechallenge. M38-specific T cells from T_M-sorted donors persisted at high frequencies in recipients, whereas the M45-specific donor cells contracted after their initial expansion in the same mice (Fig. 4A, 4B). Despite their initial T_M phenotype, the donor M38-specific T cells largely expressed a T_EFF phenotype after challenge (Fig. 4C, 4D), consistent with a typical inflationary population. The population as a whole retained its ability produce IFN-γ, TNF-α, and expose CD107a (Fig. 4E, 4F). Importantly, a small portion of donor T cells retained their T_M phenotype even after this secondary challenge (Fig. 4C, 4D).

To understand whether these persistent T_M phenotype donors continued to be functional, we turned to the OT-I transgenic system to facilitate sorting and avoid the possible selection of different T cell clones (Fig. 5A). As shown previously, transferred naive OT-Is undergo inflation and produce both T_M and T_EFF progeny after primary challenge with SIINFEKL-expressing MCMV-SL8 (45). We sorted the T_M phenotype OT-I cells that formed after primary challenge, transferred these cells, and challenged the recipients to establish secondary populations (Supplemental Fig. 1A). As with nontransgenic T cells (Fig. 4), secondary challenge of T_M OT-Is induced inflation and T_EFF formation, as well as a persistent KLRG1^lo population (Supplemental Fig. 1B). These secondary T_M were again sorted (Supplemental Fig. 1C), transferred into a third set of naive recipients, and rechallenged. Incredibly, the donor secondary T_M population inflated and produced both KLRG1^hi and KLRG1^lo progeny after this tertiary challenge (Fig. 5B–E).

Repeated acute viral challenges of small numbers of T cells in naive mice drive T_EFF differentiation (60–63), and indeed the overall frequency of tertiary inflationary cells that retained a T_M phenotype was reduced (Fig. 5E and Supplemental Fig. 1B). However, these tertiary stimulated OT-Is remained functional, producing both IFN-γ and TNF-α, as well as exposing CD107a (Fig. 5F, 5G). These data show that T_M specific for inflationary Ags can repeatedly recapitulate memory inflation upon viral challenge and produce functional T_EFF and T_M progeny.

To test the ability of transferred T_M to protect against a lethal MCMV challenge, we sorted T_M and T_EFF populations from latently infected B6 mice as described earlier and transferred them into Rag^−/− recipients. One day later, the Rag^−/− recipients were challenged with MCMV-TK, which lacks the m157 gene and is therefore resistant to NK-mediated control (43). Both transferred T_M and T_EFF expanded after the challenge and were sufficient to protect the recipients (Fig. 6A, 6B). In contrast, Rag^−/− mice that received no T cell therapy became moribund in 2–4 wk and had to be sacrificed (Fig. 6B). Notably, the T_EFF population, which proliferated very poorly in immune-replete mice (Fig. 3), expanded and persisted in immune-deficient hosts for at least 11 wk postchallenge (Fig. 6A, 6B). However, the T_EFF responses lacked M45-specific, noninflationary T cells (Fig. 6A). These data show that MCMV-specific T_M are capable of protecting immune-deficient mice and producing immune responses with broad specificities.

Patients undergoing hematopoietic stem cell transplantation (HSCT) are most susceptible to late-onset (>100 d) reactivating CMV, as opposed to an acute CMV infection (12, 64–66 and reviewed in Ref. 67). Furthermore, transferred CMV-specific T cells will need to compete with host immunity. Therefore, we developed a model to test whether T_M and T_EFF subsets could respond to viral reactivation after a long delay. To this end, T_M and T_EFF were sorted from latently infected mice (>3 mo postinfection) and transferred into immune-replete, infection-matched, or naive, congenic recipients differing at the Thy1 locus (Fig. 7A). After the transfer, the latently infected recipients were rested as described in the legend for Fig. 7. Donor T cells did not expand dramatically in any animal after transfer (Supplemental Fig. 2A), supporting our previous conclusion that competition between T cells dictates MCMV-specific T cell expansion (45). Recipient T cells and NK cells were then eliminated in all mice using a mixture of depletion Abs that targeted the host cells (Thy1.1⁺) but left the donor cells (Thy1.2⁺) intact (Fig. 7B and Supplemental Fig. 2B). This depletion protocol did not induce detectable viral transcription in any animal as assessed by nested RT-PCR (M. Quinn, T. Moghbeli, and C.M Snyder, unpublished observations), likely because of the presence of antiviral Abs (68). Despite the 9- to 12-wk rest period, MCMV-specific donor T_M responded robustly in all infected recipients after host depletion (Fig. 7C, 7D). Importantly, donor T_M did not expand to detectable levels in depleted naive recipients (Supplemental Fig. 2C). However, viral challenge of naive mice that received T_M donor cells 12 wk previously induced a robust donor response in three of the four animals, indicating that the T_M persisted in these mice, even without any Ag (Supplemental Fig. 2C). Thus, Ag rather than homeostatic mechanisms account for the donor T_M response in infected recipients.

In marked contrast, after depletion, donor T cells were only detectable in two animals that had received T_EFF, and then only at very low frequencies (Fig. 7C, 7D). Control experiments (Supplemental Fig. 3A–C) supported previous work (69), suggesting that the KLRG1-specific Ab did not induce depletion of the transferred T_EFF subset. Thus, the failure of T_EFF to expand in this setting is not a sorting artifact, but rather the inability to persist and/or expand in response to low amounts of viral Ag.

After expansion, all infected mice that received T_M had a donor population specific for multiple epitopes, and the progeny had differentiated to form new T_EFF populations (Fig. 7E) (M. Quinn and C.M. Snyder, unpublished observations). Furthermore, the four tetramers used stained only ∼60% of the total donor population in each animal (Fig. 7E), suggesting that the remaining 40% of each donor population contained cells specific for additional MCMV Ags. In contrast, in the two animals in which T_EFF donors expanded to detectable levels, each was skewed substantially toward a single inflationary epitope (Fig. 7E). Because these sorted T_EFF populations included large numbers of T cells specific for M38, m139, and IE3, this hit-or-miss expansion of donor T cells with select specificities implies that a very small number of non-T_EFF may have contaminated the transfer.

In the mice that received T_M donor cells, their diverse progeny persisted in recipients for >11 wk after termination of the depletion regimen, even though host immunity had returned (Fig. 7F). These data suggest that T_M with inflationary specificities are capable of surviving in an environment with very little or no Ag stimulation and then responding as needed during a period in which the host is immune compromised and viral Ag becomes available.

In total, these data show that protective MCMV-specific T_M persist throughout infection, retain superior proliferative function, and can respond to viral Ag as needed, in contrast with the numerically dominant T_EFF. Because MCMV-specific T_M share a transcriptional program with HCMV-specific T_M, our data suggest that T_M may be ideal candidates to restore functional immune surveillance in patients at risk for CMV reactivation.

Adoptive immunotherapy using CMV-specific CD8^pos T cells can be a successful therapeutic strategy for combating CMV reactivations (5–18). However, the majority of CMV-specific CD8⁺ T cells isolated from healthy donors will express an effector-differentiated phenotype (CD27^lo/CD127^lo/CD45RA^hi/KLRG-1^hi/CD57^hi) (reviewed in Ref. 70), and in vitro expansion of CMV-specific T cells drives their differentiation toward an effector phenotype (40). We used the MCMV model to show that the ability to restore MCMV immunity is contained almost entirely within the minor T_M subset that retains CD27. Although both T_M and T_EFF protected Rag^−/− mice (Fig. 6), humans are unlikely to remain completely immune depleted like Rag^−/− mice, and bolus CMV infections are of lesser concern than reactivation after transplantation. The inability of the T_EFF population to consistently expand after immune depletion in latently infected hosts suggests that these cells will only be protective under limited conditions. These data support a previous study in humans that correlated the transfer of CD27^hi CMV-specific T cells with an increased likelihood of T cell persistence and expansion (41).

To validate the use of the MCMV model, we compared human and mouse MCMV-specific T cells and show for the first time, to our knowledge, that T_M and T_EFF populations in mice and humans share a common transcriptional profile. The power of the GSEA analysis used for this comparison is that it identifies significant correlations across the entire transcriptional profile, rather than comparing individual genes. Nevertheless, we expect that future studies examining conserved and divergent genetic pathways will reveal significant and relevant information about CMV-specific immunity in mouse and human. These results highlight the usefulness of the MCMV model to: 1) perform CMV-specific CD8⁺ T cell functional studies that are difficult or impossible to perform in humans, and 2) provide translational insights into novel or improved therapeutic strategies.

Understanding how CMV-specific T cell immunity is maintained is critical for the improvement of CMV adoptive immunotherapy. Persistent Ag stimulation from CMV reactivations results in the majority of inflationary CD8⁺ T cells developing a T_EFF phenotype and function. However, our previous work showed that unsorted inflationary CD8⁺ T cells, containing primarily T_EFF, declined after transfer into congenic, latently infected recipients (27). These data suggest that MCMV-specific T_EFF are unable to sustain themselves in an immune-replete environment, even in the presence of Ag. Thus, we proposed that the accumulation of T_EFF is the result of continual Ag stimulation of the T_M population. Our data show that a small, stable MCMV-specific T_M population has strong functional similarities to classical memory T cells that develop after acute infections. For example, the ability to proliferate in response to Ag without terminal differentiation is a hallmark of functional memory T cells (71). In addition to producing differentiated progeny that accumulated after MCMV challenge (Fig. 4D), donor T_M also produced T_M phenotype progeny that outnumbered the cells transferred (Fig. 3C, 3D) and persisted throughout our observation period (Fig. 4C, 4D). These data suggest that MCMV-specific T_M have the ability to replace themselves even while producing differentiated progeny in response to Ag. Importantly, this was true through at least three rounds of stimulation using sorted splenic T_M (Fig. 5). Thus, MCMV-specific T_M have the capacity to respond repeatedly to viral Ag during this persistent infection and can recapitulate memory inflation.

It is interesting that transferred T_M failed to expand in immune-replete, latently infected hosts. Detectable numbers of donor T cells were only evident in one of six mice before immune depletion (Supplemental Fig. 2A). In this case, the donors were not positive for any of the tetramers used in the analyses and made up <1% of the total CD8⁺ population. However, loss of the host T cell populations led to rapid and robust expansion of donor T cells with diverse specificities and phenotypes in all T_M recipients (Fig. 7). The failure of transferred T_M to expand in the presence of host MCMV-specific immunity may reflect the relative lack of available Ag during the latent phase of MCMV infection. Indeed, viral reactivations occur in only a fraction of latently infected cells at any given time, and only rarely produce infectious viral particles (72, 73). Moreover, we have found that competition between T cells for access to this limited Ag regulates the expansion of individual T cell clones (45). Thus, the combination of low Ag and large numbers of MCMV-specific T cells in the recipients may have “shielded” the majority of the donor T_M from the ongoing infection, an idea we have proposed previously (45, 74). Importantly, MCMV Ag is not required for MCMV-specific T_M survival. We have previously shown that MCMV-specific T_M divide at a consistent rate with or without Ag (28), and our new data (Supplemental Fig. 2C) show that inflationary T_M can survive in naive mice without any Ag. Thus, homeostatic mechanisms can support the inflationary T_M population when it does not have access to Ag, which may partially explain the preservation of memory function within the T_M subset. Taken together, these data suggest that the highly functional T_M population, which can persist without access to Ag, proliferates robustly and produces new T_M, as well as more differentiated progeny, upon Ag stimulation.

Overall, our data further support the model that the burden of maintaining memory inflation falls on the functional T_M population, which can provide a stable and consistent source of new T_EFF progeny whenever needed, over prolonged periods. However, T cell competition for limited Ag appears to prevent the continuous stimulation of most T_M. Nonetheless, the T_M population is capable of robustly responding if T cell competition is lost, a conclusion with important clinical implications for adoptive immunotherapy. Variations in transplant protocols, patients, and antiviral therapy responses make it difficult to predict and standardize CMV prevention therapies. Our data suggest that the plasticity of the T_M population, transferred before any disease develops, may offer a personalized therapy, where the treatment adapts to the conditions of the patient and responds if and when Ag becomes available. Future studies will be needed to explore whether the addition of homeostatic cytokines (e.g., IL-15) or pharmacotherapeutics (e.g., rapamycin) (75) preserves the T_M phenotype either in vivo or during in vitro expansion.

We thank the Kimmel Cancer Center Flow Cytometry Facility and Animal Facility at Thomas Jefferson University.

The authors have no financial conflicts of interest.