CD8⁺ T Lymphocyte Self-renewal during Effector Cell Determination

SUMMARY

Selected CD8+ T cells must divide, produce differentiated effector cells, and self-renew, often repeatedly. We now show that silencing expression of the transcription factor TCF1 marks loss of self-renewal by determined effector cells, and that this requires cell division. In acute infections, the first three CD8+ T cell divisions produce daughter cells with unequal proliferative signaling but uniform maintenance of TCF1 expression. The more quiescent initial daughter cells resemble canonical central memory cells. The more proliferative, effector-prone cells from initial divisions can subsequently undergo division-dependent production of a TCF1-negative effector daughter cell along with a self-renewing TCF1-positive daughter cell, the latter also contributing to the memory cell pool upon resolution of infection. Self-renewal in the face of effector cell determination may promote clonal amplification and memory cell formation in acute infections, sustain effector regeneration during persistent subclinical infections, and be rate-limiting, but remediable, in chronic active infections and cancer.

Graphical abstract

INTRODUCTION

A single, activated CD8⁺ T lymphocyte appears to invariably give rise to effector cell and memory cell descendants (Buchholz et al., 2013; Gerlach et al., 2013; Gerlach et al., 2010; Plumlee et al., 2013; Stemberger et al., 2007). The mechanisms responsible for the generation of intraclonal diversity, however, remain controversial. Stochastic mechanisms have been proposed as a driving force behind diversification (Buchholz et al., 2013). Alternatively, it has been suggested that deterministic processes such as asymmetric cell division could assure the opposing outcomes of differentiation and self-renewal (Chang et al., 2011; Chang et al., 2007; Ciocca et al., 2012; Lin et al., 2015; Pollizzi et al., 2016; Verbist et al., 2016). Whether memory cells precede or follow the generation of effector cells has also been controversial (Restifo and Gattinoni, 2013).

Asymmetric inheritance of fate-determining proteins was originally described for the first T cell division of primary and secondary immune responses (Arsenio et al., 2014; Chang et al., 2011; Chang et al., 2007; Ciocca et al., 2012). The first asymmetric T cell division appeared to give rise to a more activated, effector-prone and a more quiescent, memory-prone pair of daughter cells. It was recently suggested that, after the third or fourth division, the more activated, effector-prone daughter cells underwent further asymmetric divisions characterized by sharp disparity in the expression of a key regulator of T cell memory (TCF1) between daughter cells (Lin et al., 2015).

The paradoxical finding of further asymmetric divisions subsequent to initial effector specification prompted us to explore the lineage relationship of TCF1-expressing and non-expressing subsets using a reporter mouse to track TCF1 expression in living cells (Choi et al., 2015). Our findings lead to a substantial revision of the original, two-pronged model of asymmetric T cell division. We conclude that the quiescent, memory-prone daughter cells are indeed less activated and differentiated, presumably serving to provide long-term self-renewal of the originally selected T cell clone. Despite their rapid division and heightened state of activation and differentiation, we now show that the initial effector-prone daughter cells actually retain the key memory-like property of progenitor cell self-renewal while producing their determined effector cell progeny. Production of the opposing outcomes of differentiation and self-renewal by effector-prone progenitors may explain why memory cells could have appeared to be derived from effector cells (Restifo and Gattinoni, 2013) and may provide a unifying framework for classifying antigen-activated T cell fates during successful and unsuccessful settings of long-term clonal T cell regeneration (Chu et al., 2016; He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016).

RESULTS

T cell clonal selection yielding progeny that retain and lose TCF1 expression

TCF1, encoded by the Tcf7 locus, is an essential transcription factor for T lymphocyte lineage specification during development (Germar et al., 2011; Weber et al., 2011). Following antigen activation, TCF1 limits CD8⁺ effector T cell differentiation and promotes central memory cell homeostasis (Jeannet et al., 2010; Tiemessen et al., 2014; Zhao et al., 2010; Zhou and Xue, 2012; Zhou et al., 2010). To examine the pattern of TCF1 expression in CD8⁺ T cells during an evolving infection, we transferred proliferation dye-labeled TCR transgenic P14 CD8⁺ T cells to naïve recipient mice followed by infection of recipients with Listeria monocytogenes (LMgp33) or lymphocytic choriomeningitis virus (LCMV). As previously suggested (Lin et al., 2015), we found TCF1 expression, using intracellular anti-TCF1 staining, was maintained in the first few divisions, and that after approximately three or four divisions, some cells underwent loss of TCF1 expression while some cells retained expression (Figure 1A). The pattern of TCF1 protein expression mirrored transcriptional activity as assessed using P14 CD8⁺ T cells expressing a Tcf7^GFP/+ reporter (Choi et al., 2015) (Figure S1A). Despite the dominance of TCF1^lo cells at the peak of clonal expansion, we consistently observed an unambiguous population of TCF1^hi cells in both transferred P14 cells as well as expanded endogenous D^b-gp33-specific CD8⁺ T cells from mice that did not receive transplants of P14 T cells (Figure1A and S1B).

Figure 1. Persistence of some TCF1-expressing T cells during clonal expansion.

(A) Cell division versus TCF1 protein expression of P14 CD8⁺ T cells at indicated times after LCMV infection, compared to cells transferred into uninfected mice for 4 days (far left plot).

(B) Left, expression of granzyme B and TCF1 in transferred P14 CD8⁺ T cells from uninfected and LCMV-infected mice on day 5 pi. Right, median fluorescence intensity (MFI) of granzyme B staining in TCF1^hi and TCF1^lo cells.

(C) Expression of granzyme B and KLRG1 versus TCF1 protein in splenic gp33-specific CD8⁺ T cells identified by tetramer staining 8d after LCMV infection (dot plots). Histogram plots comparing granzyme B and KLRG1 in naïve (gray fill), TCF1^hi (red line) and TCF1^lo (black line) cells.

(D) Left, expression of TCF1 protein in P14 CD8⁺ T cells from spleen and liver 3d after LMgp33 infection. Right, ratio of TCF1^hi to TCF1^lo cells in each organ.

(E) Left, Tcf7-GFP expression versus cell death in donor P14 CD8⁺ T cells from spleens of LCMV-infected recipient mice 5d pi. Binding of amine-reactive aqua dye to cytosolic targets signifies cell death. Right, frequency of cell death in Tcf7-GFP^hi and Tcf7-GFP^lo cells.

(F) Left, TCF1 versus Bcl2 expression of P14 CD8⁺ T cells 5d after LCMV infection. Right, MFI of Bcl2 in TCF1^hi and TCF1^lo cells.

t-test was performed to determine the significance *P <0.05, **P <0.01.

See also Figure S1.

As previously suggested (Lin et al., 2015), TCF1^lo P14 cells were more effector-like than the TCF1^hi cells as indicated by enrichment for lectin-like receptor KLRG1 expression in TCF1^lo cells (Figure S1C). We also found that TCF1^lo cells contain more granzyme B on a per cell basis than TCF1^hi cells (Figure 1B). Higher granzyme B and KLRG1 expression among TCF1^lo cells was also observed in polyclonal CD8⁺ T cells identified by gp33 tetramers at the peak of clonal expansion (Figure 1C). In addition to enrichment for effector markers, TCF1^lo cells preferentially localized to non-lymphoid anatomic sites associated with terminal differentiation, such as the liver of Listeria-infected mice (Figure 1D). Utilizing transferred Tcf7^GFP/+ reporter and non-reporter P14 CD8⁺ T cells, we found Tcf7-GFP^lo and TCF1-protein^lo cells freshly isolated from the spleen 5d after LCMV infection were more apoptotic as assessed by increased vital dye inclusion (Figure 1E), Annexin V binding (Figure S1D), and reduced expression of Bcl2 (Figure 1F).

Determined effector differentiation coupled to self-renewal

Despite the possibility that some daughter cells might be specified as effector-prone from the initial cell divisions (Arsenio et al., 2014; Chang et al., 2011; Chang et al., 2007), the foregoing results suggest they might not be irreversibly determined effector cells until they repress TCF1 in subsequent divisions. To formally address the flexibility of these populations, we first used an in vitro model that recapitulates in vivo patterns of TCF1 expression in the progeny of activated CD8⁺ T cells (Lin et al., 2015). Naïve P14 CD8⁺ T cells were purified and stimulated in vitro with gp33 peptide/antigen presenting cells (APCs) plus recombinant interleukin-2 (rIL-2). Repression of TCF1 became evident in a fraction of cells after approximately 3 cell divisions (Figure S2A). To track the plasticity and self-renewal capacity of TCF1^hi and TCF1^lo cells after the third cell division, we used P14 cells expressing a heterozygous knock-in allele Tcf7^GFP/+ that faithfully reports transcription from the Tcf7 locus with GFP expression (Choi et al., 2015). After initial activation by gp33/APCs, Tcf7-GFP^hi and Tcf7-GFP^lo cells from the fourth cell generation were sorted and re-cultured for 3.5d in the presence of gp33/APCs. Despite enrichment for cells with the earliest evidence of transcriptional repression of Tcf7-GFP expression, the Tcf7-GFP^lo cells gave rise largely to Tcf7-GFP^lo cells at the expense of Tcf7-GFP^hi cells. By contrast, Tcf7-GFP^hi cells were bipotent, capable of generating more Tcf7-GFP^hi cells as well as producing new Tcf7-GFP^lo cells (Figure 2A). The onset of TCF1 repression, thus, appears indicative of its heritable silencing in determined effector cells (Ladle et al., 2016; Scharer et al., 2013).

Figure 2. Self-renewal during effector differentiation marked by TCF1 expression.

(A) Left, cell division and Tcf7-GFP expression of bulk, pre-sorted P14 CD8⁺ T cells 2 days after stimulation with gp33 peptide/APCs, compared to cells that were not stimulated. Middle, post-sort purities of Tcf7-GFP^hi and Tcf7-GFP^lo cells sorted from the 4th generation. Right, cell division versus Tcf7-GFP expression 40h after re-stimulation with gp33/APCs. Results representative of 3 identical experiments.

(B–D) in vivo cell fate potential of TCF1-expressing and non-expressing donor cells from later cell generations.

(B) Top, pre-sort cell division, Tcf7-GFP and KLRG1 expression of donor P14 CD8⁺ T cells from spleens of LCMV-infected recipient mice d5 pi. KLRG1-negative, Tcf7-GFP^hi or Tcf7-GFP^lo cells were sorted from the cell generations following the initial appearance Tcf7-GFP^lo cells and post-sort purity of Tcf7-GFP^hi and Tcf7-GFP^lo cells is shown as cell division or KLRG1 versus Tcf7-GFP. Bottom, equal numbers of later generation Tcf7-GFP^hi and Tcf7-GFP^lo cells, which had been re-labeled with cell division dye, were transferred into day 5 LCMV infection-matched recipients. Four days after secondary transfer, i.e.- 9 days after primary infection (d4 pt; d9 pi), subsequent cell division, Tcf7-GFP and KLRG1 expression in donor cells from blood are shown. Vertical dashed line indicates MFI of cells re-labeled with division dye prior to secondary transfer (see histogram Figure S2B). Scatter plot indicates frequency of Tcf7-GFP^hi cells recovered from recipients of donor Tcf7-GFP^hi or Tcf7-GFP^lo cells. t-test was performed to determine the significance. **P <0.01, ***P <0.001.

(C) Representative FACS of Tcf7-GFP, CD127, and CD62L expression from primary recipients of naive, unsorted Tcf7^gfp/+ P14 CD8⁺ T cells at d13 pi (left) and secondary recipients of donor later generation Tcf7-GFP^hi and Tcf7-GFP^lo cells at d8 pt (d13 post primary infection, middle and right). Scatter plots indicate frequency of Tcf7-GFP^hi cells recovered from recipients of donor Tcf7-GFP^hi or Tcf7-GFP^lo cells (top) and frequency of central memory-like (CD62L⁺ CD127^hi) CD8⁺ T cells recovered from recipients of indicated donors (lower).

(D) Representative FACS of Tcf7-GFP, CD127, and CD62L expression from primary recipients of naive, unsorted Tcf7^gfp/+ P14 CD8⁺ T cells at d27 pi (left) and secondary recipients of donor Tcf7-GFP^hi and Tcf7-GFP^lo cells at d22 pt (d27 post primary infection, middle and right). (E) Top, spleens harvested at 30⁺ days post LCMV infection from mice that received Thy1.1⁺ P14 CD8⁺ T cell were labeled with cell division dye and transferred to secondary naïve recipients, which were then infected with LMgp33. Cell division versus TCF1 expression of P14 CD8⁺ T cells 3d after re-challenge (right), compared to cells donated into uninfected recipients (left). Middle, representative microscopy of a sibling pair of CD8⁺ T cells derived from rechallenged memory cells 3d after heterologous challenge, and stained for TCF1 (green), DNA (gray), and α-tubulin (red; superimposed on transmitted light, TL). Scale bar = 3 µm. Bottom, ratios of total TCF1 and DNA fluorescence between sibling cells (n=26, sibling pairs imaged) are depicted in scatter plot. Unequal TCF1 expression was found in ~50% of cytokinetic cell pairs when standardized to the ratio of DNA (pie chart). Results representative of 3 identical experiments.

See also Figure S2.

Similar results were obtained in vivo. From LCMV-infected recipient mice 5d post infection, we sorted donor P14 KLRG1-negative Tcf7-GFP^hi and Tcf7-GFP^lo cells from the later cell generations at which TCF1 repression was initially evident (Figure 2B). Equal numbers of Tcf7-GFP^hi and Tcf7-GFP^lo cells, which had been re-labeled with cell division dye, were transferred into separate infection-matched recipients that were also 5d post infection. After 4 days in infection-matched recipients (9 days post initial challenge), we found that Tcf7-GFP^hi donor cells underwent further division and self-renewed the Tcf7-GFP^hi pool, while also giving rise to de novo Tcf7-GFP^lo progeny (some of which progressed to KLRG1⁺) (Figure 2B). Tcf7-GFP^lo donor cells also divided further but simply generated more Tcf7-GFP^lo cells, which largely became KLRG1⁺. The foregoing result suggests silencing of Tcf7 marks determined effector cell differentiation in vivo. Retention of Tcf7 activity, by contrast, appears to signify the preservation of the bipotent capacity for self-renewal and differentiation during clonal expansion.

As predicted by prior studies (Jeannet et al., 2010; Ladle et al., 2016; Lin et al., 2015; Scharer et al., 2013; Tiemessen et al., 2014; Zhao et al., 2010; Zhou and Xue, 2012; Zhou et al., 2010), we found naïve and central memory (Tcm) CD8⁺ T cells express substantial Tcf7-GFP; effector memory cells that are CD127⁺ express detectable Tcf7-GFP, and bona fide effector cells silence Tcf7-GFP expression (Figure S2C). To examine the contribution of the later generation Tcf7-GFP^hi and Tcf7-GFP^lo cell from 5 days post infection to the memory pool, we analyzed the fate of transferred cells during and after the contraction phase that follows the peak of clonal expansion. At 13 days following initial infection (8 days post transfer into d5-infection-matched recipients), recipients of donor Tcf7-GFP^hi from the later generations continued to harbor Tcf7-GFP^hi and Tcf7-GFP^lo populations, as found in recipients of unsorted cells (Figure 2C). Many of the Tcf7-GFP^hi cells were Tcm phenotype (CD127^hi CD62L⁺). Recipients of donor Tcf7-GFP^hi from the later generations were also biased toward more Tcf7-GFP^hi and Tcm phenotype cells than recipients of unsorted cells (Figure 2C). Recipients of donor Tcf7-GFP^lo cells, by contrast, only harbored Tcf7-GFP^lo cells, none of which were Tcm.

At 27 days following initial infection, recipients of donor Tcf7-GFP^hi from the later generations were harboring fewer Tcf7-GFP^lo cells, presumably because the antigenic stimulus for differentiation had waned. The recipients of donor Tcf7-GFP^hi remained biased toward more Tcf7-GFP^hi and Tcm phenotype cells than recipients of unsorted cells (Figure 2D). By this time, recipients of donor Tcf7-GFP^lo cells contained few detectable donor cells at all, and those remaining did not bear a Tcm phenotype. The pattern of Tcf7-GFP expression after clonal contraction was consistent with predictions made from acute restimulation in vitro and in vivo (Figure 2A and 2B). Tcf7-GFP^hi cells seem to self-renew while producing more differentiated progeny, but Tcf7-GFP^lo cells are determined and unable to revert to Tcf7-GFP^hi cells under normal conditions. These findings suggest that TCF1-expressing cells from the later generations are not simply on an inevitable path toward terminal differentiation. Instead the self-renewal capacity of the later generation TCF1-expressing cells may ultimately contribute to the memory cell pool once the stimulus for continued effector generation has waned.

As previously suggested (Lin et al., 2015), re-challenge of gp33-specific memory CD8⁺ T cells also generates both TCF1^hi and TCF1^lo cells after approximately 3 divisions (Figure 2E). We, therefore, examined the clonal relationship between TCF1-expressing and -non-expressing cells derived from re-challenged memory cells using confocal microscopy. As predicted from findings in primary infection (Lin et al., 2015), we found ~50% of conjoined sibling memory cell pairs contained unequal expression of TCF1 between sister cells (Figure 2E). Self-renewal of a TCF1-expressing daughter cell, therefore, appears to be a hallmark of the production of a committed, TCF-silenced effector cell in both primary and secondary responses.

TCF1 repression occurs during cell division

Consistent with a model wherein production of a committed, TCF1-silenced effector cell is clonally coupled to the production of a self-renewing TCF1-expressing sibling, we found that the production of Tcf7-GFP^lo cells from Tcf7-GFP^hi is confined to cells undergoing a cell division. Tcf7-GFP^hi high cells from the fourth cell generation were sorted 2 days after primary stimulation with gp33/APCs, and then re-stimulated in the absence or presence of cell cycle inhibitors that arrest in G1, S, or G2/M phase. Many of the freely cycling cells became Tcf7-GFP^lo or TCF1-protein^lo but seemingly only after a further division (Figure 3A and S3A). Some of the Tcf7-GFP^hi cells, however, remained Tcf7-GFP^hi and TCF1-protein^hi despite dividing (Figure 3A and S3A). Cells that were arrested at any stage of the cell cycle, however, remained Tcf7-GFP^hi and contained abundant TCF1 protein (Figure 3A and S3A). Similarly, cell cycle inhibition abrogated the ability of in vivo generated Tcf7-GFP^hi to become Tcf7-GFP^lo cells, despite clear evidence of adequate activation (CD25 induction and increased cell size) during the cell cycle arrest (Figure 3B). Cell cycle inhibition itself was not an artificial stimulus of TCF1 expression insofar as Tcf7-GFP^lo cells remained Tcf7-GFP^lo in the presence of the drugs (Figure S3B).

Figure 3. TCF1 silencing during effector determination requiring cell division.

(A) Tcf7-GFP^hi cells from the 4th cell generation were sorted from Tcf7^GFP/+ P14 CD8⁺ T cells initially stimulated with gp33/APCs for 2 days as diagrammed in Figure 2A. Sorted, Tcf7-GFP^hi 4th generation cells were re-stimulated with gp33/APCs and rIL-2 in the absence or presence of agents that arrest at the indicated stage of the cell cycle for 24h and analyzed for cell division and expression of Tcf7-GFP. Scatter plot shows quantification of Tcf7-GFP MFI from individual experiments in indicated conditions. Similar results were obtained analyzing native TCF1 protein expression (Figure S3A).

(B) Tcf7-GFP^hi P14 donor CD8⁺ T cells sorted from recipient mice 5d post LCMV infection were re-stimulated ex vivo for 40 hours with gp33/APCs and rIL-2 in the absence or presence of indicated cell cycle arrest. Representative Tcf7-GFP, CD25, and cell size (forward light scatter, FSC) are shown as histograms and quantified for cumulative individual experiments in scatter plots as MFI (mean ± SEM). t-test was performed to determine the significance. *P < 0.05, **P < 0.01, ***P < 0.001, ns= no significant difference.

(C) TCF1 not silenced by homeostatic divisions. Proliferation dye-labeled P14 CD8⁺ T cells were transferred to Rag1^−/− mice. Three days post transfer, TCF1 expression was analyzed by FACS (far left) and by confocal microscopy (remaining panels). Representative sibling pair micrograph of TCF1 (green), DNA (gray), and α-tubulin (red; superimposed on transmitted light, TL) is shown. Scale bar = 3 µm. Ratios between sibling cells of total TCF1 abundance and DNA fluorescence (n=12, sibling pairs imaged) are depicted in scatter plot. Unequal TCF1 expression was not detected during homeostatic division (pie chart). Results representative of 3 identical experiments.

See also Figure S3.

Despite the apparent requirement for cell division to silence TCF1, its repression is not an inevitable consequence of any T cell division. Somewhat akin to the first three cell divisions of the immune response (Lin et al., 2015) and (Figure 1A), we found that P14 CD8⁺ T cells that underwent acute homeostatic proliferation in Rag1^−/− recipients generated progeny that were uniformly TCF1^hi (Figure 3C). Morphological examination of cytokinetic cells showed similar expression of TCF1 in conjoined sibling cells, akin to prior analyses of the first 3 cell divisions following infection (Lin et al., 2015). We conclude that immune activation-induced TCF1 repression is constrained to a cell division, which may be the enabling mechanism to couple differentiation to self-renewal.

Quiescent versus amplifying populations of self-renewing, TCF1^hi cells

We next examined whether there are functional and phenotypic differences between TCF1^hi cells in the initial divisions versus TCF1^hi cell from subsequent divisions. Three days after LMgp33 infection, we found that TCF1^hi cells expressed IL-2Rα (CD25) at two levels (Figure 4A). TCF1^hiCD25^lo cells were enriched for less-divided cells (first four generations) while TCF1^hiCD25^hi cells were enriched for cells beyond the fourth generation (Figure 4A). Comparison of TCF1^hi cells from first three generations to those beyond the 4th generation at day 3 post-infection revealed that those undergoing greater division also appeared more metabolically active, with higher expression of the PI3K-sensitive transcription factors, IRF4 and c-Myc, and higher mTORC1 activity, assessed by phosphorylation of the downstream target ribosomal S6 (p-S6) (Figure 4B). Cells undergoing greater division also expressed higher amounts of T-bet and interferon-γ (IFN-γ).

Figure 4. Distinct activation states in quiescent versus rapid-amplifying TCF1^hi cells.

(A) CD25 levels (left) are heterogeneous among TCF1^hi donor P14 CD8⁺ T cells (gray fill) in recipients infected with LMgp33 for 3 days (compared to naive CD8⁺ T cells, black line). Right, extent of proliferation of CD25^hi versus CD25^lo subsets of TCF1^hi cells. Cell generation numbers indicated. Results representative of 5 identical experiments.

(B, C) Cell division versus IRF4, c-Myc, p-S6, IFN-γ, T-bet, KLRG1, Ki-67, CD62L and Bcl2 in TCF1^hi P14 donor CD8⁺ T cells on day 3 (B) and/or day 6 (C) post LMgp33 infection. Histograms next to dot plots depict indicated markers in naïve P14 T cells (dashed line) and TCF1^hi P14 donor T cells that underwent less (red line) or more (gray fill) division. Gates defining lesser and greater division at top. Graphs next to histograms show quantitation of fold change in MFI (over naive levels) for less divided and more divided TCF1^hi cells from individual animals. Where indicated quantitation is percent positive for expression of marker instead of fold change over naive.

(D) Cell division and expression of Ki-67, IRF4, c-Myc, CD25, Tbet, and IFN-γ in TCF1^hi cells in the BM (magenta line) and the spleen (gray fill) 3 days post LMgp33 infection or in naïve P14 CD8⁺ T cells recovered from the BM of uninfected mice (black line).

(E) Quiescent versus rapid-amplifying TCF1^hi cells also arising in re-challenge responses. Priming (LCMV) and re-challenge (LMgp33) were performed as described in Figure 2E. Left, CD25 expression in re-challenged P14 TCF1^hi memory cells from infected secondary recipients 3d after re-challenge (gray fill) is compared to resting memory cells (black line). Right, cell division of CD25^hi and CD25^lo subpopulations of TCF1^hi cells. Cell generation numbers indicated. Results representative of 3 identical experiments.

(F) Cell division versus IRF4, c-Myc, T-bet and CD62L expression in in re-challenged TCF1^hi P14 memory cells from infected in secondary recipients 3d after re-challenge. Histograms next to dot plots depict indicated markers in resting memory P14 CD8⁺ T cells (dashed line) and TCF1^hi P14 donor memory cells that underwent less (red line) or more (gray fill) division.

Paired t-test was performed to determine the significance. *P < 0.05, **P < 0.01, ***P < 0.001.

At day 6 post infection, cells undergoing greater division continued to exhibit effector-like phenotypes, expressing higher T-bet and IFN-γ. A fraction of these cells began to express KLRG1 (Figure 4C). Cells that underwent lesser division became more quiescent (loss of proliferation antigen Ki-67 expression) and more central memory-like (higher levels of CD62L and Bcl2 and negligible IFN-γ (Figure 4C). The detection of more quiescent, central memory-like cells was not limited to the spleen. As early as 3 days after infection, cells that underwent less proliferation, with a quiescent phenotype (lower Ki-67, IRF4, c-Myc, CD25, T-bet, and IFN-γ), had localized to the bone marrow (Figure 4D), the preferred site of homeostatic memory cell renewal (Mazo et al., 2005).

Akin to TCF1^hi cells in the primary response, TCF1^hi cells generated after re-challenge also consisted of populations with distinct IL-2 sensitivity and proliferative potentials (Figure 4E). Cells that underwent less proliferation expressed lower T-bet, IRF4, c-Myc and more CD62L than the rapid-amplifying cells, and were phenotypically indistinguishable from the quiescent, Tcm-like cells generated after primary infection (Figure 4B and 4F). In both primary and re-challenge responses, the earliest generations of TCF1-expressing cells appear prone to quiescence even prior to the peak of clonal expansion, and without acquiring effector characteristics. The more activated, effector-prone cells from the earliest generations appear to progress to further differentiation and clonal amplification while maintaining a progenitor-like capacity for self-renewal.

DISCUSSION

The present results support the existence of an extended capacity for regenerative or conservative cell divisions of antigen-specific CD8⁺ T cells during acute immune responses. In the first three divisions, there appears to be differential activation leading to a more quiescent set of progeny as well as a more actively dividing population that will eventually give rise to effector cells. Both daughter cells in the initial divisions express TCF1, although we can extrapolate from prior studies that they differ in their expression of IRF4, cMyc, and activated mTOR components, factors that likely contribute to their differential proliferation (Lin et al., 2015; Pollizzi et al., 2016; Verbist et al., 2016).

The effector-prone TCF1^hi descendants from the initial divisions appear specified to become lineage-committed progenitors or transit-amplifying cells. After further, activation-associated divisions, the effector-prone TCF1^hi cells directly give rise to effector-determined, TCF1^lo descendants during the act of cell division. However, the divisions that yield TCF1^lo cells appear inherently conservative, yielding another daughter cell that remains TCF1^hi and self-renewable [Figure 2E and (Lin et al., 2015)]. It is notable that a B cell's differentiation to plasma cell is also coupled to cell division (Barwick et al., 2016; Caron et al., 2015; Lin et al., 2015). Confining determined differentiation to the act of cell division, rather than simply allowing a non-duplicating, interphase cell to transform into a new fate may be a highly conserved mechanism, together with polarized subcellular compartmentalization, for allowing a progenitor to meet the opposing demands of differentiation and self-renewal. Once a CD8⁺ T cell becomes TCF1^lo, it appears capable of some further divisions (Figure 2A and 2B). However, the silencing of TCF1 appears rapid and heritable, restricting the TCF1^lo progeny to move forward in a path to terminal differentiation.

TCF1^hi cells in the rapid-amplifying state possess a hybrid phenotype of memory and effector T cells. Their ability to generate both TCF1^hi and TCF1^lo progeny is memory-like, while their sensitivity to IL-2 and heightened expression of T-bet and IFN-γ are more effector-like. During an acute infection, the amplifying population could support the demands of rapid clonal expansion. Upon resolution of infection and withdrawal of antigenic stimulus, however, the remaining rapid-amplifying TCF1^hi cells might conceivably revert to quiescence, with diminished expression of effector-like traits, thereby contributing to the memory cell pool (Figure 2C and 2D). If repurposing of the rapid-amplifying TCF1^hi cells for entry into the central memory pool is possible following resolution of infection, it is still conceivable that TCF1^hi cells derived from earlier versus later cell generations could differ in their repopulating potentials (Gattinoni et al., 2011).

The ability to set aside memory cells both before and during the production of determined TCF1^lo cells may explain the apparent paradox of having features of both determinism and plasticity in clonal selection. Almost invariably, a recruited CD8⁺ T cell avoids clonal deletion and leaves behind memory cell progeny (Buchholz et al., 2013; Gerlach et al., 2013; Gerlach et al., 2010; Plumlee et al., 2013; Stemberger et al., 2007). Nonetheless, many variables alter the sizes of peak effector cell expansion as well the memory cell pool (Marsden et al., 2006). We speculate that, in acute infections, the deterministic production of a TCF1^hi sibling cell during the production of a TCF1^lo effector cell may represent an insurance policy against clonal deletion as well as a sensing mechanism to titer production of effector and memory cells according to the size and kinetics of the antigenic threat.

In the setting of controlled persistent infection, a proliferative intermediate population appears to enable the continuous production of effector cells (Chu et al., 2016). In chronic active infections, a similar self-renewal property has been identified among a TCF1^hi subpopulation of cells, although the natural history of those diseases appears to outstrip the normal limits of T cell regenerative capacity (He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016). Approaches that can expand the pool of amplifying intermediates may, therefore, extend T cell regenerative capacity in geriatric populations or prevent its expenditure during chronic active challenges and cancer elimination.

EXPERIMENTAL PROCEDURES

Mice

All animal work was done in accordance with Institutional Animal Care and Use Guidelines of Columbia University. C57BL/6 (wild type), Rag1^−/−, P14 TCR transgenic mice recognizing LCMV peptide gp33-41/H-2D^b, Tcf7^GFP/+ (Choi et al., 2015), and GFP-c-Myc (Huang et al., 2008) were housed in specific pathogen-free conditions prior to infectious challenges.

Adoptive transfers and infectious challenges

Resting CD8⁺ T cells were purified with the CD8⁺ T Cell Isolation Kit (Miltenyi Biotec) from P14 mice. 1–3 × 10⁶ CD8⁺ T cells were labeled with CFSE dye (Molecular Probes), a violet cell proliferation dye (Invitrogen), or an eFluor 670 cell proliferation dye (eBioscience) and adoptively transferred intravenously (iv) into wild-type congenic C57BL/6 mice. To generate acutely resolved systemic infections, mice were infected with either 2 × 10⁵ PFU of lymphocytic choriomenigitis virus Armstrong strain (LCMV) by intraperitoneal (ip) injection or 5 × 10³ Listeria monocytogenes expressing gp33-41 (LMgp33) by iv injection. For experiments requiring secondary adoptive transfers into infection-matched recipients, later division Tcf7-GFP^hi KLRG1⁻ and Tcf7-GFP^lo KLRG1⁻ P14 CD8⁺ T cells were sorted from LCMV-infected primary recipients on day 5 pi. Equal numbers of division dye-relabeled, sorted cells were transferred iv into day 5 LCMV-infected secondary recipients.

Cell culture

To activate naïve CD8⁺ T cells in vitro, P14⁺ CD8⁺ T cells (with or without Tcf7^GFP/+ reporter) were purified with the CD8⁺ T Cell Isolation Kit (Miltenyi Biotec), labeled with one of the cell proliferation dyes and activated with gp33 peptide (1µg/ml) (Anaspec) and recombinant IL-2 (100 IU) in the presence of congenic naïve splenocytes. All cells were cultured in complete lymphocyte media (Iscove's DMEM, 10% FBS, Pen-Strep, L-Glutamine, 55µM 2-mercaptoethanol). In some experiments, pre-activated P14 CD8⁺ T cells of designated phenotypes were sorted from infected mice or in vitro cultures and re-stimulated with or without gp33 peptides (1µg/ml) and in the presence of rIL-2 and congenic naïve splenocytes. Pharmacological inhibitors of G1 phase (Mimosine 250µM, Sigma), S phase (Hydroxyurea 200µM, Sigma) or G2/M phase (Nocodazole 1µg/ml, Sigma) and were added to the cells at the time of re-stimulation where indicated.

Flow cytometry

Cells were harvested from the blood, spleen, liver and bone marrow. Single-cell suspension from the spleen and bone marrow was prepared by filtering cells with a 70µm cell strainer and lysed by an ACK Lysing Buffer (Lonza). Mononuclear cells were isolated from heparinized blood by density gradient centrifugation in Lymphoprep (Stemcell technology). After LMgp33 infection, lymphocytes from PBS-perfused liver were isolated after filtering cells with a 70µm cell strainer and density gradient centrifugation. Surface staining was carried out at 4°C for 30 minutes in PBS⁺ containing 2% fetal bovine serum and 2 mM EDTA. Cells were then fixed and permeabilized and then stained for transcription factors and intracellular proteins. Dead cells were eliminated from analysis by using green or aqua amine reactive dyes (Invitrogen). Apoptotic cells were detected by Annexin V (BD Biosciences) following a standard staining protocol. Intracellular phospo-S6^235/236 in transferred P14 CD8⁺ T cells from the spleen after LMgp33 infection was assessed directly ex vivo by treating freshly isolated cells with 3% paraformaldehyde, ice-cold methanol, prior to intracellular antibody staining. H-2D^b-gp33 tetramers were used to detect endogenous LCMV-specific CD8⁺ T cells.

Flow cytometry antibodies against CD45.1 (A20), CD127 (A7R34), IRF4 (3E4) and KLRG1 (2F1) were from eBioscience; CD44 (IM7), CD45.2 (104), Granzyme B (GB11), T-bet (4B10) and Thy1.1 (OX-7), were from Biolegend; Bcl2 (3F11), CD25 (PC61), CD62L (MEL14), IFN-γ (XMG1.2) and Ki-67 (B56) were from BD Biosciences; c-Myc (D84C12), phospho-S6^235/236 (D57.2.2E), TCF1 (C63D9) were from Cell Signaling; CD4 (RM-45) and CD8 (5H10) were from Invitrogen. LSRII, LSRFortessa and FACSAria II flow cytometers (all from BD Biosciences) with BD FACSDiva software were used to analyze and/or purify cells. Data were analyzed using FlowJo v.8.8.7 (FlowJo, LLC).

Confocal microscopy

To analyze dividing memory cell during rechallenge, spleens harvested at 30⁺ days post LCMV infection from mice that received Thy1.1⁺ P14 CD8⁺ T cell were labeled with cell division dye and transferred to secondary naïve recipients, which were then infected with LMgp33 for 3 days prior to sorting cells for microscopy. To analyze naïve CD8⁺ T cells undergoing lymphopenia-induced homeostatic proliferation, Thy1.1⁺ P14 CD8⁺ T cells transferred into C57BL/6 Rag1^−/− were sorted from the recipient spleen on day 3 post-transfer. Immunofluorescence of dividing T cells was performed as previously described (Lin et al., 2015). The following antibodies were used: rat anti-α-tubulin (Abcam; YOL1/34), mouse anti-β-tubulin (Sigma; AA2), rabbit anti-TCF1 (Cell Signaling; C63D9). Cells undergoing cytokinesis were identified by cytoplasmic cleft in brightfield, tubulin bridge by antibody staining, plus the presence of dual nuclei using DAPI DNA stain. 15–20 Z stack sections were acquired using a Zeiss LSM710 laser scanning confocal microscope and ZEN software (Zeiss). 3-dimensional Z stacks were converted into 2-dimensional images. Total fluorescence of the TCF1 protein within each sister cell was calculated in ImageJ (NIH, USA) using the integrated density (IntDen) function. TCF1 protein was determined to be asymmetric in each conjoined cells if the ratio of the IntDen of either side of the cells was greater than that of the mean of DNA plus 2 standard deviations (SD). p values were calculated using chi-square analysis when comparing the asymmetry between designated protein to tubulin.

Statistical analyses

Where indicated, p-values were determined using a two-tailed t-test. p-values <0.05 were considered significant.