CD4⁺ T cell effector commitment coupled to self-renewal by asymmetric cell divisions

Upon infection, CD4⁺ T cells differentiate into short-lived effector cells as well as long-lived memory cells that mount responses upon re-exposure to the microbe. CD4⁺ T cells have the ability to differentiate into multiple effector subsets including T helper 1 cells (Th1 cells). The Th1 subset is defined by expression of the lineage-determining transcription factor T-bet and the capacity to secrete the effector molecule IFN-γ (Zhu et al., 2010). Th1 cells migrate to the site of microbial entry to exert their function (Swain et al., 2012). After the contraction phase, wherein the vast majority of effector cells undergo apoptosis, a small portion of cells persist in the host as memory T cells to combat future infections. Memory T cells can be divided into different subsets based on distinct effector function and homing capacity (Sallusto et al., 1999; Masopust et al., 2001; Reinhardt et al., 2001). One population of memory cells called central memory cells share similar features with naive T cells. They are characterized by expression of the chemokine receptor CCR7 and L-selectin (CD62L), circulate through secondary lymphoid organs, and have a less differentiated phenotype than bona fide effector cells. Upon rechallenge, they have the ability to regenerate differentiated effector cells in addition to self-renewing the central memory pool (Sallusto et al., 1999; Reinhardt et al., 2001; Zaph et al., 2004). In contrast, effector memory cells do not express CCR7 or CD62L and produce effector cytokines.

Using a variety of approaches, it has been suggested that a single T or B lymphocyte can generate progeny with intraclonal effector subclass diversity and memory cell renewal (Stemberger et al., 2007; Gerlach et al., 2010, 2013; Buchholz et al., 2013; Plumlee et al., 2013; Tubo et al., 2013, 2016; Graef et al., 2014; Becattini et al., 2015; Taylor et al., 2015). Whether cell-intrinsic, cell-extrinsic, stochastic, or deterministic mechanisms are responsible for the generation of intraclonal cell fate diversity of lymphocytes is an unresolved issue (Reiner and Adams, 2014). In this study, we have identified discrete stages of CD4⁺ T cell clonal selection distinguished by cell division, TCF1 expression, and anatomical localization. TCF1^hi cells had a less differentiated phenotype, showed increased expression of CD62L, and homed to noninflamed secondary lymphoid organs within the initial cell divisions. TCF1^hi cells from later divisions in the draining LNs (DLNs) had the capacity to asymmetrically self-renew while also generating PI3K-driven, TCF1^lo Th1 effector cells. The Th1 cell–like, TCF1^lo cells appeared incapable of reverting to central memory–phenotype cells and instead migrated to the site of infection. Some of the TCF1^hi cells in the DLNs also appeared to be T follicular helper cell (Tfh cell)–like and noncirculating. These findings offer a potential mechanistic explanation for the seemingly hard-wired regeneration and functional diversity of CD4⁺ T cell clonal selection (Tubo et al., 2013, 2016; Becattini et al., 2015).

TCF1 is a key regulator of T cell development in the thymus (Germar et al., 2011; Weber et al., 2011). In the periphery, TCF1 has been shown to be a negative regulator of effector cell and a positive regulator of memory cell CD8⁺ responses (Jeannet et al., 2010; Zhao et al., 2010; Zhou et al., 2010; Thaventhiran et al., 2013; Tiemessen et al., 2014). To examine the expression of TCF1 in CD4⁺ T cell responses, we used influenza viral infection, in which the primary activation of responding CD4⁺ T cells (mediastinal LNs) is anatomically distinct from the site of Th1 effector function (lung tissue). OTII T cells were labeled with a cell proliferation dye (CPD) and adoptively transferred into Thy1 disparate recipient mice, which were subsequently infected intranasally with a recombinant strain of PR8 influenza virus expressing a peptide epitope of OVA recognized by the OTII TCR (hereafter referred to as PR8-OVA). TCF1 expression was examined in the dividing antigen-specific OTII T cells in the LNs draining the site of infection (mediastinal LNs and DLNs), the nondraining LNs (NDLNs), and the lungs on day 4 after infection (Fig. 1 A).

After approximately four to five divisions in the DLNs, a population of cells that had repressed TCF1 emerged alongside a population of cells that had retained TCF1 expression (Fig. 1 A). In contrast to DLNs, the lung, site of effector T cell function, and the NDLNs, site of central memory cell recirculation (Zaph et al., 2004; Colpitts and Scott, 2010), were each enriched in an exclusive subset (Fig. 1 A). Lungs were enriched for cells that had undergone greater division (five to seven rounds) and down-regulated TCF1, whereas NDLNs were enriched with cells that divided less (one to four divisions) and maintained high levels of TCF1.

Both TCF1^hi and TCF1^lo CD4⁺ T cell subsets showed similar levels of CD44 induction, as well as CD127 (IL-7rα) and CCR7 repression (Fig. 1 B). In contrast, TCF1^hi cells had detectably less expression of the markers associated with Th1 cell differentiation, T-bet, CD25, and signaling lymphocyte activation molecule (SLAM; Fig. 1 B). Compared with TCF1^lo cells, lower Th1 cell–associated markers (despite comparable CD44 expression) were evident in TCF1^hi cells from both NDLNs and DLNs (Fig. 1 B) and across early and later TCF1^hi cell divisions (Fig. 1 C). Compared with the TCF1^lo subset, the TCF1^hi population also showed an increased percentage of cells that expressed CD62L (Fig. 1 D), which may explain the enrichment of TCF1^hi cells in lymphoid sites (Fig. 1 A).

In addition to central memory cells, other TCF1^hi progeny that reside in DLNs would include Tfh cells (Choi et al., 2015; Wu et al., 2015; Xu et al., 2015). We found a detectable portion of the TCF1^hi cells in the DLNs expressed substantial levels of the chemokine receptor CXCR5 and the transcription factor Bcl6, markers used to identify the Tfh cell lineage (Fig. 1 E). However, unlike the DLNs, the NDLNs and lung contained few if any Tfh-phenotype cells (Fig. 1 F). Within the first few days of infection, descendants of naive CD4⁺ T cells appeared to exhibit regenerative and functional diversity that could be delineated by division, TCF1 expression, and anatomical localization.

Among the later cell divisions in the DLNs, repression of TCF1 appears to mark Th1-like cells (with higher expression of CD25, SLAM, and T-bet) that are programmed to populate the site of infection. In contrast, TCF1^hi cells appear to be less differentiated and more likely to retain lymphoid homing capacity. We therefore wished to determine the flexibility and interrelatedness of the two populations. TCF1 is encoded by the Tcf7 locus, and we therefore used OTII⁺ Tcf7-GFP/+ reporter mice (Choi et al., 2015) to trace the fate of expressing and nonexpressing cells from PR8-OVA–infected recipient mice. We first determined that the pattern of Tcf7-GFP reporter expression in the DLNs (Fig. 2 A) was similar to TCF1 protein expression (Fig. 1 A), which established this reporter line as a suitable tool for investigating lineage relationships between Tcf7-GFP–high and Tcf7-GFP–low subsets.

We next sorted Tcf7-GFP–high and Tcf7-GFP–low OTII⁺ CD4⁺ T cells by flow cytometry from the DLNs of PR8-OVA–infected recipient mice. Equal numbers of sorted Tcf7-GFP–high and Tcf7-GFP–low cells were individually restimulated with anti-CD3 and anti-CD28 in vitro (Fig. 2 B). We found that the dividing Tcf7-GFP–high cells maintained the Tcf7-GFP–high pool while also abundantly giving rise to Tcf7-GFP–low cells (Fig. 2 B). However, the Tcf7-GFP–low cells simply generated more Tcf7-GFP–low cells (Fig. 2 B). To rule out the possibility that the difference in the ability to self-renew between the Tcf7-GFP–high and Tcf7-GFP–low cells was caused by a difference in the rounds of division they had undergone, we compared Tcf7-GFP–high and Tcf7-GFP–low cells that had divided the same number of times (five to eight rounds). After relabeling with division dye and restimulation, we found that the Tcf7-GFP–high cells divided further and self-renewed, producing more Tcf7-GFP–high and making Tcf7-GFP–low progeny. Tcf7-GFP–low cells also divided more but only produced additional Tcf7-GFP–low progeny (Fig. 2 C).

We obtained similar results from a system of modeled effector differentiation. Stimulation of naive T cells in Th1 cell–inducing conditions in vitro recapitulated the TCF1 protein expression heterogeneity in the progeny of activated CD4⁺ T cells. Naive CD4⁺ T cells were purified and stimulated in vitro with plate-bound anti-CD3 and anti-CD28 in the presence of IL-2 and IL-12 and were analyzed on days 3–6 of culture for TCF1 protein expression (Fig. 3 A). After about four cell divisions, a population of cells that showed low TCF1 protein abundance emerged, whereas some cells remained high for TCF1 protein content. Similarly, CD4⁺ Tcf7-GFP T cells stimulated in Th1 cell–inducing conditions for 5 d also exhibited heterogeneity of GFP expression after a few cell divisions (Fig. 3 B). Then, we sorted the Tcf7-GFP–high and Tcf7-GFP–low cells, relabeled them with division dye, and restimulated them with plate-bound anti-CD3 and anti-CD28 for an additional 5 d. Similar to observations with ex vivo cells, we found that in vitro–derived Tcf7-GFP–high cells remained bipotent, able to generate Tcf7-GFP–low and Tcf7-GFP–high cells, whereas Tcf7-GFP–low cells were unable to revert to production of Tcf7-GFP–high cells (Fig. 3 C).

We also cultured Tcf7-GFP/+ cells in Th1 cell–inducing conditions for a longer period of time, 7 d, which allowed the Tcf7-GFP–high cells to undergo more rounds of division (Fig. 3 D). Tcf7-GFP–high and Tcf7-GFP–low cells from matched later divisions were sorted by flow cytometry, relabeled with CPD, and restimulated with anti-CD3 and anti-CD28 for 5 d. Tcf7-GFP–high cells divided further and remained bipotent, able to generate Tcf7-GFP–low and Tcf7-GFP–high cells, whereas Tcf7-GFP–low cells divided further but were unable to revert to production of Tcf7-GFP–high cells (Fig. 3 D). Therefore, Tcf7 expression appears to mark a subset of CD4⁺ T cells with the ability to self-renew. Silencing of Tcf7 during commitment to effector differentiation, in contrast, appears to mark the inability to revert to a state of self-renewal.

It has been suggested that IL-2 drives the down-regulation of TCF1 in CD4⁺ T cells (Wu et al., 2015). We found that activating CD4⁺ T cells with plate-bound anti-CD3 and anti-CD28 in the presence of IL-2 or IL-12 repressed TCF1 expression (Fig. 4 A). PI3K signaling is important for driving repression of TCF1 in CD8⁺ T cells and Pax5 in B cells (Lin et al., 2015). The effect of the cytokines on TCF1 repression in CD4⁺ T cells was indeed PI3K dependent insofar as addition of the small molecule inhibitor of PI3K, LY294002, reversed the repression of TCF1 (Fig. 4, A and B). Cells that were generated in the presence of the PI3K inhibitor also underwent fewer rounds of division, failed to produce IFN-γ, granzyme B, and TNF, and were unable to induce expression of Blimp-1, which is known to promote terminal differentiation (Fig. 4, B and C). We also noted that activated CD4⁺ T cells were heterogeneous for the activation of mTOR, as assessed by phospho-S6 ribosomal protein (Fig. 4 D). The subpopulation with high mTOR activity appeared to be the cells undergoing TCF1 silencing and terminal maturation insofar as PI3K inhibitor treatment coordinately suppressed the generation of each of those attributes. These results demonstrate that PI3K signaling is also critical for the activation of mTOR, repression of TCF1, and effector commitment in CD4⁺ T cells.

How heterogeneity is generated in the CD4⁺ T cell compartment is poorly understood. The PI3K-dependent bimodality in expression of TCF1 (Fig. 4 B) prompted us to test whether unequal signaling to TCF1 repression could bifurcate CD4⁺ T cell fates. OTII⁺ CD4⁺ T cells sorted from recipient mice 4 d after PR8-OVA infection were examined for expression of TCF1 by confocal microscopy. TCF1 was localized to the nucleus in interphase cells. In metaphase cells, TCF1 was excluded from condensed chromatin and symmetrically localized to the cytoplasm. In contrast to the invariant symmetry of TCF1 in metaphase, ∼74% of conjoined sibling pairs, identified by the presence of an intercellular cytoplasmic bridge containing tubulin, contained asymmetric abundance of TCF1 (Fig. 5, A and B). Similar results were obtained from in vitro differentiation, with 56% of cytokinetic sibling pairs having asymmetric TCF1 abundance (Fig. 5, C and D). The asymmetric expression of TCF1 in cytokinetic cells does not appear to be caused by unequal partitioning of preformed TCF1 protein during metaphase but rather a signaling event occurring after telophase, which is consistent with recent findings in B cells and CD8⁺ T cells (Lin et al., 2015).

Our findings suggest that maintenance of expression of TCF1 among the progeny of a selected CD4⁺ T cell marks a subset of cells that have undergone few divisions and retained capacity for lymphoid recirculation, resembling canonical central memory cells. TCF1 expression also marks cells that have divided more, remained in the DLNs, and are capable of producing TCF1^lo, Th1 effector progeny, also renewing the TCF1^hi fate during active infection. Whether those TCF1^hi cells that are producing Th1 cell progeny and self-renewing during antigen encounter will die or, instead, revert to quiescence after resolution of infection will require further investigation. In contrast to the bipotency of TCF1^hi cells, TCF1 repression in Th1 cells appears to mark an inability to revert to the TCF1^hi fate under the physiological conditions tested here.

Among the TCF1^hi cells in the DLNs, we also found a subset of cells with Tfh cell characteristics, consistent with other recent studies revealing a role for TCF1 in Tfh cell development (Choi et al., 2015; Wu et al., 2015; Xu et al., 2015). Insofar as Tfh cells possess characteristics of memory-like renewal (Choi et al., 2013; Hale et al., 2013; Suan et al., 2015), it is notable that self-renewing CD4⁺ and CD8⁺ T cells can share phenotypic markers with Tfh cells, such as expression of CXCR5 (Pepper et al., 2011; He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016). The present findings suggest that TCF1^hi cell self-renewal occurs in the daughter cells with weaker anabolic (PI3K/mTOR) signaling, whereas TCF1^lo Th1 effector cells emerge from the daughter cells with stronger PI3K/mTOR signaling. Tfh cells also appear to diverge from Th1 cells owing to a weaker versus stronger PI3K/AKT/mTOR signaling divergence (Ray et al., 2015; Wu et al., 2015). The nature and extent of the phenotypic, functional, and metabolic similarities between central memory and Tfh cells, as well as the corresponding differences between those two subsets and Th1 cells, will also require further investigation.

An enigmatic finding in the current study, which is analogous to recent observations of TCF1 and Pax5 heterogeneity in CD8⁺ T cells and B cells, respectively (Lin et al., 2015), is why there exists a lag of three or four divisions before silencing TCF1 expression. In CD4⁺ T cells undergoing Th2 cell differentiation, there is an analogous lag in induction of IL-4 expression (Bird et al., 1998). Insofar as strong but unequal PI3K/mTOR signaling appears to commence from the very first T cell divisions (Lin et al., 2015; Pollizzi et al., 2016; Verbist et al., 2016), there would seem to exist some other rate-limiting feature, resolved by cell division, that delays gene silencing or induction. In B and T cells, silencing of genes associated with terminal differentiation appears to result from de novo establishment of epigenetic silencing, which may require more than one cell division (Scharer et al., 2013; Caron et al., 2015; Barwick et al., 2016; Ladle et al., 2016). Synthesis of the posttranslational modifications (such as acetyl and methyl groups) that induce or silence gene activity may also require some time- or division-dependent metabolic remodeling (Lu and Thompson, 2012; Londoño Gentile et al., 2013). Future studies examining signaling, epigenetic, and metabolic changes during the first three cell divisions may help resolve the nature of this division-linked lag in differentiation. In summary, the present findings suggest that metabolic inequality during cell division might ensure determinism for memory cell self-renewal as well as diversification of effector subclasses during the clonal selection of a CD4⁺ T cell.

C57BL/6 mice and C57BL/6 OTII mice (with transgenic expression of an I-A^b–restricted TCR specific for OVA amino acids 323–339) were either bred at Columbia University or purchased from The Jackson Laboratory. C57BL/6 Blimp-1–YFP mice (Fooksman et al., 2010) and C57BL/6 Tcf7-GFP/+ mice (Choi et al., 2015) were previously described. Both male and female mice were used in the study. Mice were housed in specific pathogen–free conditions. All animal studies were performed according to the Institutional Animal Care and Use Guidelines of Columbia University.

OTII wild-type CD4⁺ T cells were purified by magnetic cell separation (Miltenyi Biotec) and labeled with cell trace violet (CTV) proliferation dye. A total of 10⁶ cells were adoptively transferred by intravenous injection into Thy1 disparate C57BL/6 recipient mice. The next day, mice were anesthetized by isoflurane and inoculated intranasally with 250 50% tissue culture infective doses of PR8-OVA influenza virus. Expression of different markers was examined on day 4 or 5 after infection. In some experiments, a total of 10⁶ OTII Tcf7-GFP/+ CD4⁺ T cells were adoptively transferred into recipient mice that were subsequently infected with PR8-OVA. Tcf7-GFP/+–high and –low cells were sorted on day 5 after infection, relabeled with a CPD (either CTV or eFluor 670), and restimulated to examine responses.

Naive WT CD4⁺ T cells purified by magnetic cell separation were labeled with CTV, and 5 × 10⁵ cells were cultured in 48-well tissue culture plates that were precoated with 1 µg/ml anti-CD3 and anti-CD28 in the presence of 25 U/ml IL-2 and/or 10 ng/ml IL-12. In some experiments, a small molecule inhibitor to PI3K (5 µM), LY294002 (Cell Signaling Technology), was included at the start of the culture. Cells were harvested at the indicated time points. Tcf7-GFP/+–high and –low cells were also sorted, relabeled with CTV or eFluor 670 CPD, restimulated, and examined for expression of GFP.

Single-cell suspensions were prepared, and cells were stained according to standard staining protocol. Surface staining was done with the following antibodies: CD4 (RM4-5; BD), CD127 (A7R34; eBioscience), CCR7 (4B12; eBioscience), CD44 (IM7; BioLegend), SLAM (TC15-12F12.2; BioLegend), CD62L (MEL-14; BD), CXCR5 (2G8; BD), and CD25 (PC61; BD). Intracellular staining was done with the following antibodies: T-bet (4B10; BioLegend), TCF1 (C63D9; Cell Signaling Technology), Bcl6 (K112-91; BD), and phospho-S6 (S235/236; Cell Signaling Technology). Flow cytometry was acquired on LSRII and Fortessa machines (BD), and analysis was performed with FlowJo software (Tree Star).

Immunofluorescence was performed as previously described (Chang et al., 2007; Barnett et al., 2012; Ciocca et al., 2012; Lin et al., 2015). In brief, FACS-purified cells were transferred to poly-l-lysine–coated coverslips (no. 1.5) and incubated at 37°C for adherence. Cells were fixed with 3% paraformaldehyde, and after rehydration in 50 mM NH₄Cl in PBS, cells were permeabilized with 0.3% Triton X-100 in PBS and then incubated in blocking buffer (0.01% saponin and 0.25% fish skin gelatin in PBS). Cells were stained with primary and secondary antibodies diluted in blocking buffer. Images were acquired on a laser-scanning inverted confocal microscope (LSM710; ZEISS), and total fluorescence in cells was quantified using ImageJ software (National Institutes of Health). The following primary antibodies were used: rat anti–α-tubulin (YOL1/34; Abcam) and rabbit anti-TCF1 (C63D9; Cell Signaling Technology). The following secondary antibodies were used: goat α-rat AF488 and goat α-rabbit AF568. ProlongGold with DAPI (Invitrogen) was used to stain DNA and mount coverslips on glass slides. TCF1 protein was considered asymmetric in conjoined cells if the ratio of the fluorescence was greater than the mean of DNA plus two standard deviations. P-values were calculated with two-tailed Fisher’s exact test.

This study was supported by National Institutes of Health grants (AI061699, AI113365, and AI076458 to S.L. Reiner and AI112579, AI115149, AI119160, and AI121080 to H.-H. Xue) and the U.S. Department of Veterans Affairs Merit Review Award (I01 BX002903 to H.-H. Xue).

The authors declare no competing financial interests.

Author contributions: S.A. Nish and S.L. Reiner conceived and designed the study. S.A. Nish, R. Kratchmarov, K.D. Zens, W.-H.W. Lin, W.C. Adams, Y.-H. Chen, and B. Yen performed experiments. S.A. Nish and S.L. Reiner analyzed data and wrote the manuscript. N.J. Rothman assisted with animal experiments. A. Bhandoola and H.-H. Xue provided reagents and expertise.