The Transcription Factor Runx2 Is Required for Long-Term Persistence of Antiviral CD8⁺ Memory T Cells

The T cell response to acute viral infections has been well characterized at the cellular level. Following infection, a robust pathogen-specific CD8⁺ T cell response is observed and within 1–2 wk postinfection, the pathogen is cleared from the infected host. This early effector phase includes the proliferation and differentiation of cytotoxic effector T cells, a process that is dependent on inflammatory cytokines produced by innate immune cells and on the presentation of viral peptides on host APCs (1–3). After viral clearance, the majority of the effector CD8⁺ T cell population will undergo apoptosis, a process that continues for many weeks post–pathogen clearance (4). Ultimately, the host retains a small pool of pathogen-specific memory T cells that provide rapid protection upon secondary infection (5).

During an acute antiviral response, the pool of activated CD8⁺ T cells is not homogeneous. Based on differential expression of surface markers, such as KLRG1 and CD127, virus-specific CD8⁺ T cells can be classified as KLRG1^hi CD127^lo terminal effector cells (TECs) and KLRG1^lo CD127^hi memory precursor cells (MPCs) (6). TECs rapidly proliferate in response to infection, make up the majority of the CD8⁺ effector response, and undergo apoptosis after clearance of the infection. MPCs proliferate less than TECs but go on to survive and undergo homeostatic proliferation after the infection is eliminated (6, 7). Several transcription factors have been shown to play critical roles in the relative differentiation of TECs versus MPCs during acute viral infection. These include IRF4 (8–12), BATF (13–15), T-bet (16–19), Blimp-1 (20–22), and Id2 (23–26), which regulate TEC differentiation and effector cell function. In contrast, Eomesodermin (Eomes) (17, 19, 27), Tcf1 (28, 29), Id3 (24, 30), and Runx3 (31) are all required for CD8⁺ T cell memory formation and homeostasis.

In this study, we show that a member of the Runt-related transcription factor family (RUNX), Runx2, is also important for regulating the long-term persistence of CD8⁺ memory T cells following acute lymphocytic choriomeningitis virus (LCMV)–Armstrong infection. Runx2, like the other RUNX factors, contains a Runt DNA binding domain and pairs with CBFβ to bind to DNA (32). Runx2 functions primarily in bone development in which it is required for osteoblast generation (33) and bone formation (34).

Runx1 and Runx3 have well-characterized roles in T cells, including important functions during regulatory T cell development (35), T_H1 skewing (36), and CD8⁺ T cell differentiation (31, 37). In contrast, no clear function for Runx2 in T cells has been identified, although an earlier study showed that ectopic overexpression of Runx2 in thymocytes perturbed T cell development at the CD4⁻CD8⁻ stage (38). A genome-wide regulatory network generated by Hu and Chen (39) also suggested that Runx2 may play a role in CD8⁺ T cell memory. Using mice carrying floxed alleles of Runx2 crossed to CD4-cre, we find no apparent defects in T cell development or T cell homeostasis under steady-state conditions. However, following infection with LCMV–Armstrong, we identify a CD8⁺ T cell–intrinsic defect in the development and persistence of virus-specific MPCs. This correlates with our findings that Runx2 expression levels in activated CD8⁺ T cells are enhanced by TLR and memory cytokine stimulation but inhibited by IRF4 expression. Together, these data identify Runx2 as an important mediator of virus-specific memory T cells following resolution of infection by LCMV–Armstrong.

Mice were bred and housed in specific pathogen-free conditions at the University of Massachusetts Medical School (UMMS) in accordance with Institutional Animal Care and Use Committee guidelines. C57BL/6J mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred in house. OT-I TCR transgenic Rag2^−/− mice were purchased from Taconic Biosciences (Germantown, NY) and bred in house. Runx2^fl/fl mice were a gift from Dr. A. Javed (40) (University of Alabama at Birmingham). CD4-cre⁺ transgenic mice were a gift from Dr. J. Kang (UMMS). P14 TCR transgenic mice were a gift from S. Kaech (Yale University) and were crossed to Runx2^fl/fl CD4-cre⁺ transgenic mice. Runx2^+/+ and Runx2^+/+ CD4-cre⁺ mice were used as wild-type (WT) controls. Irf4^+/fl CD4-cre⁺ and Irf4^fl/fl CD4-cre⁺ mice have been described previously (8, 12, 41).

Adult male mice (7–11 wk) were infected with LCMV–Armstrong at 5 × 10⁴ PFU i.p. For rechallenge, mice were infected with LCMV–clone 13 at 2 × 10⁶ PFU i.v. LCMV–Armstrong and LCMV–clone 13 were graciously provided by Dr. R. Welsh (UMMS). For coadoptive transfers, splenocytes from P14 WT CD4-cre⁺ CD90.1⁺ CD90.2⁺ and P14 Runx2^fl/fl CD4-cre⁺ CD90.1⁺ were stained with Abs to CD8α and Vα2 to determine proportions of P14 cells, and equal numbers of WT and Runx2^fl/fl cells were mixed. Ten thousand P14 cells were transferred i.v. into CD90.2⁺ host mice 1 d prior to infection.

Spleens and fat pads were harvested 9 d after LCMV–Armstrong infection. For rechallenge, kidneys and livers were harvested 4 d after LCMV–clone 13 infection. Organs were homogenized in 1 ml RPMI 1640 medium and stored at −80°C. Plaque assays were performed as described previously (42).

Splenocytes from OT-I TCR transgenic mice were stimulated with OVA, T4, or G4 peptide with indicated doses for 72 h. Cells were harvested and analyzed for Runx2, Eomes, and CD44 expression by intracellular staining. For cytokine experiments, IFN-β and IL-12 were purchased from R&D Systems (Minneapolis, MN). IL-7 and IL-15 were purchased from PeproTech (Rocky Hill, NJ). OVA, T4, and G4 peptides were purchased from 21st Century Biochemicals (Marlborough, MA). Imiquimod was purchased from InvivoGen (San Diego, CA). LPS was purchased from Sigma-Aldrich (St. Louis, MO). Splenocytes from Irf4^+/+ CD4-cre⁺, Irf4^+/fl CD4-cre⁺, and Irf4^fl/fl CD4-cre⁺ mice were isolated and plated with plate-bound anti-CD3/CD28 for 72 h. Cells were harvested and analyzed for Runx2, Eomes, and CD44 expression by intracellular staining. For cytokine production, splenocytes from infected mice were stimulated with gp_33–41, gp_276–286, and nucleoprotein (np)_396–404 peptide for 4 h in the presence of 1 μg/ml GolgiStop and 1 μg/ml GolgiPlug and Abs to CD107a and CD107b. gp_33–41, gp_276–286, and np_396–404 peptides were generously provided by Dr. R. Welsh (UMMS) and generated by K. Daniels.

CD127 (FITC), Bcl2 (PE), Bcl6 (PE), IRF4 (PE), CD8α (PE–eFluor 610), TNF-α (PerCP–eFluor 710), Eomes (PerCP–eFluor 710), CD122 (PerCP–eFluor 710), T-bet (PE-Cy7), CD44 (PE-Cy7 and Alexa Fluor 700), KLRG1 (PE-Cy7 and eFluor 450), IFN-γ (eFluor 450), IL-2 (allophycocyanin), Vα2 (allophycocyanin), CD27 (allophycocyanin–eFluor 780), and CD90.2 (allophycocyanin–eFluor 780) Abs were purchased from eBioscience (San Diego, CA). CD107a (FITC), CD107b (FITC), CD62L (FITC), 7-AAD, annexin V (PE), and CD90.1 (V500) were purchased from BD Biosciences (Billerica, MA). Abs to granzyme B (PE), LIVE/DEAD Violet, LIVE/DEAD Aqua, and goat anti-rabbit (Alexa Fluor 647) were purchased from Life Technologies (Grand Island, NY). H2-D^b gp_33–41, H2-D^b gp_276–286, and H2-D^b np_396–404 monomers were obtained from the National Institutes of Health Tetramer Core Facility (Atlanta, GA). Runx2 and TCF-1 Abs were purchased from Cell Signaling Technology (Danvers, MA). Single-cell suspensions from spleens, bone marrow, lymph nodes, lung, and liver were prepared; RBCs were lysed; and Fc receptors were blocked using supernatant from 2.4G2 hybridomas. Lymphocytes were isolated from lung and liver using Lympholyte-M (Cedarlane Laboratories, Burlington, NC), and perfusion was performed on mice to prevent contamination of blood lymphocytes in these preparations. Cells were stained with H2-D^b gp_33–41, H2-D^b gp_276–286, and H2-D^b np_396–404 tetramers prior to staining with cell surface Abs. For cytokine staining, cells were stimulated ex vivo with 1 μg/ml gp_33–41, gp_276–286, or np_396–404 for 4 h at 37°C. Intracellular cytokine staining was performed using BD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit. Intracellular transcription factor staining was performed using eBioscience Foxp3/Transcription Factor Staining Buffer Set. Samples were analyzed on the LSR II flow cytometer (BD Bioscience), and data were analyzed on FlowJo (Tree Star).

All data are represented as mean ± SEM. Statistical significance is indicated by *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001 (NS, p > 0.05) based on unpaired Student t test.

To circumvent the neonatal lethality of a germline deficiency in Runx2, we generated mice that lacked Runx2 only in T cells. To this end, Runx2^fl/fl mice (40) were crossed to the CD4-cre⁺ transgenic line (8, 12) (hereafter referred to as Runx2^fl/fl mice). Runx2^fl/fl mice showed no apparent defect in thymic T cell development or in peripheral T cells compared with Runx2^+/+ CD4-cre⁺ controls (hereafter called WT mice) (Supplemental Fig. 1A–C). Additionally, Runx2^fl/fl mice exhibited no abnormalities in their numbers or proportions of CD8⁺CD44^hi and CD8⁺CD44^lo T cells within the spleen (Supplemental Fig. 1D).

To assess the function of Runx2 in antiviral T cell responses, we infected WT and Runx2^fl/fl mice with 5 × 10⁴ PFU LCMV–Armstrong i.p. and harvested spleens at day 9 (the peak of the CD8⁺ T cell response), day 14 (the attrition phase), and day 28 (the memory phase) postinfection (Fig. 1). Compared with WT mice, Runx2^fl/fl mice had no alterations in the numbers of H2-D^b gp_33–41 tetramer–specific cells (GP33) at days 9 and 14 postinfection but did show reduced numbers of virus-specific cells at day 28 postinfection (Fig. 1A, 1B). Similar results were observed for H2-D^b gp_276–286–specific cells (GP276) and H2-D^b np_396–404–specific cells (NP396), indicating the defect at day 28 postinfection in the CD8⁺ T cells was not an epitope-specific phenotype (Supplemental Fig. 1E).

These results suggested a potential defect in the virus-specific MPC population in LCMV–Armstrong-infected Runx2^fl/fl mice. To assess this possibility, we examined tetramer-positive CD8⁺ T cells for KLRG1 and CD127 expression at each time point postinfection. This analysis revealed a defect in the total number of GP33, GP276, and NP396 MPCs in Runx2^fl/fl mice compared with controls at all three time points tested; in contrast, no differences were found in the virus-specific TEC populations in this comparison (Fig. 1C–E, Supplemental Fig. 1F, 1G). Functional memory cells are able to produce IFN-γ, TNF-α, and IL-2, and this polyfunctionality is an indicator of a robust memory population (43). Comparisons of cytokine production by virus-specific WT and Runx2^fl/fl CD8⁺ T cells revealed a significant reduction in the total numbers of triple cytokine–producing cells when T cells lacked Runx2 expression (Fig. 1F, Supplemental Fig. 1K). These data indicated that the reduced numbers of tetramer-positive CD8⁺ T cells in LCMV–Armstrong-infected Runx2^fl/fl mice was due to a defect in the memory T cell population.

As mentioned above, Runx2^fl/fl mice showed no reduction in virus-specific TEC numbers compared with WT mice at days 9, 14, or 28 after LCMV infection. We also observed only modest defects in granzyme B expression and no defects in CD107a+b expression at day 9 postinfection in Runx2^fl/fl mice (Supplemental Fig. 1H, 1I). These results suggested there were no defects in antiviral effector cell functions when T cells lacked Runx2. Analysis of viral clearance in LCMV–Armstrong-infected WT versus Runx2^fl/fl mice confirmed this supposition (Supplemental Fig. 1J).

Based on these data, we considered whether virus-specific CD8⁺ T cell subsets might express different levels of Runx2. To test this, we infected WT C57/BL6 mice with LCMV–Armstrong and harvested spleens at day 9 postinfection. Runx2 protein levels were examined in CD8⁺ MPCs and TECs specific for three viral epitopes. Although Runx2 was upregulated in both subsets of virus-specific cells relative to the levels present in naive CD8⁺ T cells, we observed increased levels of Runx2 in the MPCs compared with the TECs for each of the epitopes tested (Fig. 1G). These results indicated that Runx2 is upregulated within all activated CD8⁺ T cells but is higher within the MPC population than in the TEC subset.

To further assess potential molecular differences between WT and Runx2^fl/fl antiviral CD8⁺ T cells, we examined the expression levels of several transcription factors that are important for TEC and MPC differentiation and function. This analysis revealed that at days 14 and 28 postinfection, Runx2^fl/fl CD8⁺ T cells expressed lower levels of Eomes and TCF-1 compared with WT T cells, consistent with the reduced number of MPCs in these mice. In contrast, expression of T-bet, a factor associated with CD8⁺ effector function, was not altered between WT and Runx2^fl/fl cells (Fig. 1H, Supplemental Fig. 1L, 1M), consistent with previous results indicating no defect in CD8⁺ effector function (Supplemental Fig. 1H, 1I).

We next considered whether the loss of MPCs in LCMV–Armstrong-infected Runx2^fl/fl mice were due to Runx2 promoting apoptosis in MPCs. However, examination of splenocytes from LCMV–Armstrong-infected WT or Runx2^fl/fl mice at day 10 postinfection failed to show a significant increase in apoptosis of virus-specific CD8⁺ T cells in the absence of Runx2 (Supplemental Fig. 2). Taken together, these results indicated that Runx2 is regulating an alternative aspect of memory CD8⁺ T cell persistence, such as homeostatic proliferation, rather than survival.

To determine whether the loss of LCMV-specific Runx2-deficient CD8⁺ memory T cells was caused by an intrinsic loss of Runx2 in the CD8⁺ T cells, we performed adoptive transfer experiments. WT and Runx2^fl/fl mice were crossed to the P14 TCR transgenic line that expresses a TCR specific for the GP33 epitope of LCMV. WT (CD90.1⁺ CD90.2⁺) and Runx2^fl/fl (CD90.1⁺) P14 cells were cotransferred into CD90.2⁺ host mice at a 1:1 mixture (Fig. 2A), and recipients were infected with LCMV–Armstrong the following day. Four weeks later, donor cells were examined in the spleens of infected mice. As shown in Fig. 2, Runx2^fl/fl P14 cells were significantly reduced in numbers relative to WT P14 cells (Fig. 2B). Assessment of TEC and MPC subsets showed that both populations of Runx2^fl/fl P14 cells were reduced in numbers compared with their WT counterparts, but a more dramatic reduction was observed in the Runx2^fl/fl MPC subset, in which WT P14 cells outnumbered Runx2^fl/fl P14 cells by a 23-fold margin (Fig. 2C). Examination of several proteins associated with MPC differentiation and survival, including Eomes, TCF-1, CD27, CD122, and Bcl2, showed reduced expression by Runx2^fl/fl P14 cells relative to WT P14 cells at this time point. Surprisingly, this was not the case for Bcl6, which was upregulated in Runx2^fl/fl cells compared with WT controls (Fig. 2D). We also assessed cytokine production by WT and Runx2^fl/fl P14 cells at day 28 postinfection and observed markedly fewer triple cytokine–producing cells among the Runx2^fl/fl P14 cells relative to WT P14 cells (Fig. 2E). Overall, these data demonstrate that the defect in memory CD8⁺ T cell persistence in Runx2^fl/fl mice following LCMV–Armstrong infection is a CD8⁺ T cell–intrinsic phenotype.

To determine whether the defect in numbers of Runx2^fl/fl CD8⁺ memory T cells would impact the recall response, we rechallenged LCMV–Armstrong-infected WT and Runx2^fl/fl mice with LCMV–clone 13. Initially, WT and Runx2^fl/fl mice were infected with LCMV–Armstrong, and 90 d postinfection, mice were rechallenged with a high dose of LCMV–clone 13. Prior to rechallenge, we examined the surviving populations of virus-specific CD8⁺ T cells and found reduced numbers of tetramer-positive cells in Runx2^fl/fl mice compared with controls; furthermore, Runx2^fl/fl mice had a more substantial deficit in the CD62L^hi than in the CD62L^lo subset of each epitope-specific memory T cell population (Fig. 3A, Supplemental Fig. 3A–C). Four days after rechallenge, we observed substantial expansion of both the WT and Runx2^fl/fl virus-specific memory cells. This robust recall response was observed for all three viral epitopes tested (Fig. 3B, Supplemental Fig. 3D). A careful examination of TECs and MPCs showed no significant differences in either population when comparing WT to Runx2^fl/fl mice (Fig. 3C). We also found no differences in viral clearance between WT and Runx2^fl/fl at day 4 post-rechallenge (Fig. 3D). Phenotypic and functional analysis of the cells after rechallenge showed few significant differences between virus-specific WT and Runx2^fl/fl cells (Fig. 3E, Supplemental Fig. 3E–J). These results indicate that the loss of memory T cells seen in Runx2^fl/fl mice does not impair their protective response to secondary challenge.

Based on these data, we considered whether the reduced numbers of virus-specific memory CD8⁺ T cells seen in the spleens of Runx2^fl/fl mice might be due to altered migration of these cells to other tissues rather than to an absolute reduction in the total memory cell population. This hypothesis was suggested by a previous study that showed loss of Runx2 in the plasmacytoid dendritic cell compartment led to a retention of the cells in the bone marrow (44). To test this, we examined several organs for pathogen-specific CD8⁺ T cells 14 d after primary infection with LCMV–Armstrong. As shown, we found significantly fewer LCMV-specific memory CD8⁺ T cells in nearly all organs examined and for each of the viral epitopes assessed (Fig. 3F, Supplemental Fig. 3K, 3L). These data indicate that the reduced numbers of splenic Runx2^fl/fl LCMV-specific memory CD8⁺ T cells is not due to enhanced homing or retention of these cells to or in other organs.

Previous work has shown a role for TLR signaling in CD8⁺ T cell memory formation (45, 46). We next wanted to determine if Runx2 expression in activated CD8⁺ T cells was regulated by TLR signaling in vitro. To determine the optimal day poststimulation to look at Runx2 expression, we isolated splenocytes from P14 TCR transgenic mice and looked for Runx2 expression up to 5 d after stimulation with GP33 peptide. We found that Runx2 was upregulated by day 3 poststimulation and remained elevated through day 5 (Supplemental Fig. 4A); however, cell death increased dramatically starting at day 4 postinfection (Supplemental Fig. 4B). We therefore chose 72 h poststimulation as the optimal time point for assessing Runx2 expression accompanied by minimal cell death in vitro. To test the role of TLR agonists on Runx2 expression, we isolated splenocytes from OT-I TCR transgenic Rag2^−/− mice, stimulated OT-I T cells with high or low concentrations of the lower-affinity OVA peptide variant SIITFEKL (T4), and incubated with or without TLR agonists LPS (Fig. 4A) or imiquimod (Fig. 4B). We found that Runx2 was upregulated in cells treated with either TLR agonist at both high and low T4 peptide doses, but this upregulation was more dramatic when cells were stimulated with the low peptide dose.

Previous work from M. Mescher and colleagues (47, 48) has shown that type I IFN and IL-12 regulate memory CD8⁺ T cell development in several viral and intracellular bacterial infection systems. These findings prompted us to assess whether type I IFN or IL-12 was contributing to the high expression of Runx2 in CD8⁺ T cells stimulated in the presence of TLR agonists. We stimulated OT-I T cells with either a high or low concentration of T4 peptide and incubated cells with or without IFN-β (Fig. 4C) or IL-12 (Fig. 4D). We found treatment with IFN-β had no significant effect on Runx2 expression, whereas treatment with IL-12 led to increased Runx2 expression at the lower peptide dose. These results suggest a direct effect of IL-12 on T cells. Furthermore, the fact that IFN-β had no direct effect on T cells suggests that the enhanced Runx2 upregulation observed when stimulation cultures were treated with TLR agonists might be acting via upregulation of costimulatory molecules on splenic APCs.

We also tested additional cytokines known to be important in homeostatic proliferation and survival of memory cells for their effects on Runx2 expression. Splenocytes from OT-I TCR transgenic mice were isolated and stimulated with high or low doses of the T4 peptide. Cultures were supplemented with or without IL-7 (Fig. 4E) or IL-15 (Fig. 4F) and analyzed at day 3 poststimulation. These studies demonstrated that both IL-7 and IL-15 promoted increased expression of Runx2 when T cells were activated with weak TCR stimulation.

Together, these data show that TLR signaling, most likely in APCs, and stimulation of T cells with cytokines known to be important in CD8⁺ T cell memory development and homeostasis enhance Runx2 upregulation; interestingly, these cytokines and TLR agonist signals have the greatest impact in T cells activated with weaker strength of TCR signaling.

From our TLR and cytokine data, we found that OT-I cells stimulated with low concentrations of T4 peptide expressed higher levels of Runx2 than cells stimulated with a higher concentration of peptides (Fig. 4). These data suggested an important role for TCR signal strength in controlling Runx2 expression levels. Previous studies have shown an important inverse correlation between TCR signal strength and the development of CD8⁺ MPCs postinfection (3, 49). To assess more comprehensively a role for TCR signal strength in regulating Runx2 expression, we performed experiments using extensive dose ranges of three peptides recognized by the OT-I TCR. For these studies, OT-I cells were stimulated with APCs plus various concentrations of the highly potent SIINFEKL (OVA) peptide, the medium potency (T4) peptide, or the low potency SIIGFEKL (G4) peptide. These experiments clearly revealed an inverse correlation between peptide dose and Runx2 expression (Fig. 5A). Similar results were also observed for Eomes expression (Fig. 5B). As expected, CD44 upregulation showed a positive correlation with peptide dose (Fig. 5C).

These findings suggested that the transcription factor IRF4 might regulate Runx2 expression. This possibility was suggested by our previous studies showing that IRF4 is upregulated to different levels based on TCR signal strength and, further, that IRF4 is a negative regulator of Eomes expression (8, 12). To test this possibility, we stimulated WT, Irf4^+/fl CD4-cre⁺, and Irf4^fl/fl CD4-cre⁺ (hereby referred to as Irf4^+/fl and Irf4^fl/fl, respectively) naive CD8⁺ T cells with anti-CD3/CD28 for 72 h and examined Runx2 expression. As shown, we observed a significant increase in Runx2 expression in stimulated Irf4^fl/fl CD8⁺ T cells compared with WT or Irf4^+/fl cells (Fig. 5D). Consistent with previously published data, similar results were observed for Eomes (Fig. 5E). Together, these results indicate that Runx2 is more highly expressed when CD8⁺ T cells are stimulated with low levels of Ag and that IRF4 is a negative regulator of Runx2 and Eomes expression following CD8⁺ T cell activation.

A key aspect of the adaptive immune response is the generation of long-term memory cells that provide rapid and robust protection upon secondary exposure to an infecting pathogen. Improving memory CD8⁺ T cell responses is currently a focus of many vaccine efforts, whereas inhibition of memory T cell responses is an ongoing challenge in the development of therapies to treat autoimmune diseases. Thus, there is a need to understand in more detail the molecular mechanisms regulating memory T cell development and persistence.

Our studies identified Runx2 as an important factor in the maintenance of long-term memory CD8⁺ T cells. However, we found that Runx2 was not necessary for antiviral effector function or for robust recall responses to secondary challenge. These features of the CD8⁺ T cell response to infection in Runx2^fl/fl mice show a striking resemblance to previous studies examining the consequences of a deficiency in IL-15 or IL-15Ra (50–52). In these earlier studies, IL-15 signaling was found to be essential for long-term memory CD8⁺ T cell homeostasis and self-renewal. We observed that IL-15 stimulation of CD8⁺ T cells in vitro promoted enhanced Runx2 upregulation only under conditions of weak TCR stimulation. These findings suggest the possibility that in vivo tonic TCR signals plus IL-15 are functioning to maintain memory cells in part through the upregulation of Runx2. Consistent with this possibility, IL-15 is thought to promote memory T cell homeostatic proliferation rather than memory cell survival (53–57), a function that aligns well with data demonstrating that Runx2 overexpression in thymocytes promotes uncontrolled cellular proliferation leading to lymphomagenesis (58).

We also found that Runx2 expression levels are enhanced by weak rather than strong TCR signal intensity, likely because of negative regulation by the transcription factor IRF4. This pattern mirrors that of Eomes, another transcription factor associated with long-term memory CD8⁺ T cell maintenance (27). Along with weak TCR signaling, cytokines that promote memory T cell formation and signals that activate APCs also promote enhanced Runx2 upregulation. These data further strengthen an association of Runx2 expression with optimal formation of long-lived memory CD8⁺ T cells (5, 49).

Determining the precise mechanism underlying the loss of memory CD8⁺ T cells in the absence of Runx2 has remained. Early on, studies using microarrays were performed to identify differentially expressed genes between WT and Runx2^fl/fl CD8⁺ T cells. Cells isolated from LCMV–Armstrong-infected mice and analyzed directly ex vivo failed to yield informative gene targets of Runx2 regulation; similar uninformative results were obtained when comparing WT and Runx2^fl/fl CD8⁺ T cells stimulated in vitro (data not shown). This prompted us to consider alternative mechanisms for Runx2 function, such as a change in rRNA transcription and processing, based on the known role for Runx2 in regulating this pathway in osteoblasts (59); these studies also failed to identify a function for Runx2 in CD8⁺ T cells (data not shown). One additional possibility is that Runx2 is affecting memory formation through regulation of chromatin accessibility, a known function of Runx3 in CD8⁺ T cells (31). RUNX family members have been shown to have redundant roles in CD4⁺ regulatory T cells; thus, the most robust alterations are observed in the absence of Cbfβ, the binding partner for all three Runx proteins (60). Future studies examining CD8⁺ T cell memory using Cbfβ-deficient T cells may be informative in this regard. Alternatively, Runx2 may function intermittently in memory CD8⁺ T cells during short time windows when memory cells associate with stromal cells in lymphoid organs and receive homeostatic TCR and IL-15/IL-7 signals. Capturing the cells during these brief interactions may be required to identify the specific genes and/or pathways regulated by Runx2 in memory CD8⁺ T cells.

We thank Dr. Raymond Welsh and Dr. Lucio Castilla for insightful discussions. We thank Dr. Amjad Javed for the Runx^fl/fl mouse line, Regina Whitehead and Sharlene Hubbard for technical assistance, and the National Institutes of Health Tetramer Core Facility for LCMV-specific H2-D^b gp_33–41, gp_276–286, and np_396–404 monomers.

The authors have no financial conflicts of interest.