Environmental cues dictate the fate of individual CD8⁺ T cells responding to infection

Summary

Many studies have examined pathways controlling effector T cell differentiation, but less is known about the fate of individual CD8+ T cells during infection. Here, we examined the antiviral and anti-bacterial responses of single CD8+ T cells from the polyclonal repertoire. The progeny of naïve clonal CD8+ T cells displayed unique profiles of differentiation based on extrinsic pathogen-induced environmental cues, with some clones demonstrating extreme bias towards a single developmental pathway. Moreover, even within the same animal, a single naïve CD8⁺+ T cell exhibited distinct fates that were controlled by tissue-specific events. However, memory CD8+ T cells relied on intrinsic factors to control differentiation upon challenge. Our results demonstrate that stochastic and instructive events differentially contribute to shaping the primary and secondary CD8⁺ T cell response. and provide insight into the underlying forces that drive effector differentiation and protective memory formation.

INTRODUCTION

CD8⁺. T cells are a vital component of the adaptive immune system, important for eliminating intracellular pathogens and cancerous cells(Alexander-Miller, 2005; Zhang and Bevan, 2011). When CD8⁺ T cells are activated through their T cell receptor (TCR) by peptide presentation on major histocompatibility complex (MHC) class I molecules, they gain the ability to secrete cytokines and additional effector functions such as cytotoxicity. Small numbers of antigen-specific naive CD8⁺ T cells (approximately 80-1,200 cells per specificity/mouse(Obar et al., 2008)) expand after infection or peptide stimulation to form a large effector population, generally peaking around a week after infection. This large effector population then undergoes contraction, leaving behind a smaller long-lived memory population. Remaining memory CD8⁺ T cells protect the host from subsequent re-infection with the same pathogen by quickly expanding and rapidly expressing lytic activity and effector cytokines(Lefrançois and Obar, 2010).

At the peak of the CD8⁺ T cell response, the large effector cell population is made up of multiple subsets that can be distinguished phenotypically(Joshi et al., 2007; Sarkar et al., 2008; Obar and Lefrançois, 2010c). The expression of several cell surface markers, including killer cell lectin-like receptor subfamily G, member 1 (KLRG1) and interleukin-7 receptor (CD127) have been used to identify different effector subpopulations(Joshi et al., 2007; Sarkar et al., 2008). Upon antigenic stimulation, CD8⁺ T cells upregulate activation markers, including CD11a and CD44, and downregulate CD127, which is expressed by all naïve CD8⁺ T cells(Schluns et al., 2000). The earliest effector cells observed lack both CD127 and KLRG1 and are termed early effector cells (EEC)(Obar et al., 2011). EEC are capable of differentiating into the two other major effector populations, short lived effector cells (SLEC) and memory precursor effector cells (MPEC)(Obar et al., 2011). SLEC express KLRG1 but not CD127 and as their name suggests gradually die off and do not remain at memory. MPEC express CD127 but not KLRG1 and go on to form long-lived memory cells which continue to express CD127. A fourth population that expresses both CD127 and KLRG1, named double positive effector cells (DPEC) can be found after infection, although little is known about their origin or function. All four subsets described can secrete cytokines and express granzyme B, and, thus, are true effector CD8⁺ T cells. Although CD127 expression identifies memory precursors, forced CD127 expression does not result in increased memory generation(Hand et al., 2007; Haring et al., 2008). In addition, although KLRG1 expression marks senescent CD8⁺ T cells in mice and humans, its function is unknown(Voehringer et al., 2002; Grundemann et al., 2010). Nevertheless, the MPEC versus SLEC paradigm holds true in most primary CD8⁺ T cell responses to infection.

Our studies and those of others have revealed that the heterogeneity among effector CD8⁺ T cells is dependent on the type of infection and is controlled by a number of cytokines and transcription factors(Joshi et al., 2007; Cui et al., 2009; Kaech and Wherry, 2007; Harty and Badovinac, 2008; Obar and Lefrançois, 2010a; Obar and Lefrançois, 2010c). For example, while a Listeria monocytogenes (LM) infection drives robust development of SLEC, a vesicular stomatitis virus (VSV) infection results in a smaller fraction of SLEC and larger percentages of EEC and MPEC. Although it has been shown that IL-12 promotes SLEC development(Cui et al., 2009), the overall composition of the environmental milieu that leads to a particular pattern of effector development is not entirely clear. Thus, how deterministic versus stochastic events control the outcome of the response has yet to be determined. Furthermore, whether the initial APC-naïve CD8⁺ T cell interaction fixes the resulting fate of a cell or whether the progeny of a responding cell continues to be malleable downstream is also unknown. A number of theories of CD8⁺ T cell effector and memory development continue to be discussed(Ahmed et al., 2009). For example, asymmetric division has been suggested to result in distinct lineage designations with one daughter cell becoming an effector and the other developing along the memory lineage(Chang et al., 2007). Yet, little is known about the clonal composition of effector cell populations leading to memory development or how the progeny of individual cells contribute to the shaping of the overall response. Earlier and more recent reports employed either single cell transfer of TCR transgenic OT-I cells after sorting with anti-CD8⁺ and anti-CD44 mAbs or used genetic barcoding of OT-I cells to track individual fates after infection with LM(Stemberger et al., 2007; Gerlach et al., 2010; Buchholz et al., 2013; Gerlach et al., 2013). These studies showed that the response of single OT-I cells to infection with LM, generated all effector and memory subsets (Stemberger et al., 2007) based on CD27, KLRG1 and CD62L expression. However, expression of the canonical memory marker CD127 was not examined and whether these results will extend to the single cells from the polyclonal repertoire is not known. In addition, those reports examined in elegant detail the numerical contribution of clonally-derived families to the overall response, but again whether single cells expressing the high-avidity OT-I TCR (Turner et al., 2008) leads to recapitulation of the polyclonal response needs to be determined. Furthermore, how distinct infections and tissue environments impinge upon CD8⁺ T cell fate was not a feature of these reports.

Here, we have utilized a single cell transfer system to dissect the polyclonal response to infection. Our results identified multiple potential pathways of effector CD8⁺ T cell development, with some clones generating homogeneous effector subsets and others differentiating toward mixed effector populations. Our findings also demonstrated unexpected effects on clonal burst size unrelated to precursor frequency. Finally, these experiments showed dramatic effects of the environmental milieu based on the type of infection as well as tissue locale on the development of effector CD8⁺ T cell subpopulations. Unlike naïve CD8⁺ T cells, memory T cells differentiated independently of their environment, relying on intrinsic factors for fate decisions. Thus, our results provide new insight into the developmental pathways in single cells that contribute to shape the overall CD8⁺ T cell response.

RESULTS

Developing and validating a single cell transfer system for polyclonal antigen-specific CD8⁺ T cells

To better understand how individual clones of antigen-specific CD8⁺ T cells contribute to the response against infection, we employed a limiting dilution strategy to transfer single antigen-specific cells to recipient mice, which could be distinguished from the responding host cells based on expression of CD45 congenic markers. We based the number of cells transferred on the known frequencies of naïve antigen-specific CD8⁺ T cells specific for OVA in the context of H-2K^b (Obar et al., 2008; Jenkins and Moon, 2012). We transferred 2×10⁵ splenic CD8⁺ T cells, which contained ~1.3 OVA-specific cells, from both CD45.1⁺ CD45.2⁻ and CD45.1⁺ CD45.2⁺ mice, into 20 CD45.1⁻ 2⁺ mice that were then infected with VSV-OVA one day later (Fig 1a). Seven days after infection, splenocytes were stained for CD8⁺ and CD45.1/2 expression and for reactivity with OVA-K^b tetramer. Three outcomes are possible within each individual recipient mouse: no clonal response (top row, Fig. 1b), one clone responding (middle row, a CD45.1/2⁺ population) and two clones responding (bottom row, populations detected from both the CD45.1/2 and CD45.1 transferred cells).

Figure 1.

Detection and phenotype of the progeny of single antigen-specific CD8⁺ T cells in response to viral infection. OVA-specific CD8⁺ T cells from the spleen were identified by OVAK^b tetramer staining, and donor populations were identified by CD45.1 and CD45.2 staining within the OVA-K^b+ population. Expression of KLRG1 and CD127 was then analyzed on gated donor populations. This experiment was repeated 3 times, with 60 mice total.

Our transfers were based on the Poisson distribution which states that if no more than 63% of the recipient mice contained a population derived from the transferred cells, it can be assumed that single antigen-specific cells were transferred(Taswell, 1981). This distribution held true for all experiments in this paper. To verify the clonal origin of the populations analyzed, we sequenced the expressed TCRα and TCRβ genes from sorted, congenically marked, donor populations. Nine TCRα and five TCRβ sequences were obtained (Table 1). Only one in-frame sequence was obtained from each population and no two were identical (Table 1, Supplementary Fig. 1) indicating that each responding population was derived from a single cell.

Table 1.

TCRα and TCRβ sequencing to prove clonality.

TCRα sequences
	Alpha-V	Alpha-J	CDR3
1-OVA	TRAV 5-4	TRAJ49	AATSNTGYQNY
2-OVA	TRAV6-1	TRAJ12	VLGGTGGYKVV
3-OVA	TRAV14-2	TRAJ22	AASRPSGSWQLI
4-N	TRAV7-3	TRAJ15	AVTPQGGRALI
5-N	TRAV9	TRAJ40	VLSPNTGNYKYV
6-OVA	TRAV4	TRAJ12	AADTGGYKVV
7-N	TRAV7-4	TRAJ26	AASEEANYAQGLT
8-N	TRAV6-5	TRAJ22	ALGASSGSWQLI
9-N	TRAV6-3	TRAJ27	AMSDLPNTGKLT

TCRβ sequences
	Beta-V	Beta-J	Beta-D	CDR3
1-OVA	TRBV14	TRBJ1-5	TRBD1	ASSLGWNNNQAPL
2-OVA	TRBV31	TRBJ2-2	TRBD2	AWGLGLANTGQLY
3-N	TRBV14	TRBJ1-2	TRBD1	ASSFGANSDYT
4-N	TRBV13-1	TRBJ1-2	TRBD1	ASSQGRSSDYT
5-N	TRBV29	TRBJ1-1	TRBD1	XSSLSTAPLEVF

Progeny of single CD8 T cells following VSV-OVA infection were sorted and the expressed TCR V-regions were sequenced. TCRα sequences for 9 clones and TCRβ sequences of 5 clones are shown. OVA or N specificities are indicated. See also Figure S1.

For each of the populations arising from a single cell, KLRG1 and CD127 expression was evaluated to determine the differentiation status of the population. Extreme phenotypic variation in the progeny of individual clones was observed (Fig. 1b). In some cases, two clones from the same mouse (Fig. 1b, bottom row) exhibited distinct differentiation profiles indicating independent differentiation despite identical environments.

Clonal responses are differentially regulated by the inflammatory environment but develop independently of TCR avidity

Our previous work shows that effector CD8⁺ T cell differentiation and the inflammatory environment is distinct between VSV and LM infections(Obar et al., 2011). To test whether the differences were the result of differential clonal contributions to each response, we compared the phenotype of OVA-specific clonal responses after VSV-OVA or LM-OVA infection (Fig. 2a,b). Responses derived from clones in each infection displayed substantial variability, with certain responses skewed heavily to the MPEC or SLEC phenotypes (e.g. clones 1 & 20 for VSV-OVA and clones 1 & 17 for LM-OVA). Comparing the clonal progeny between the two infections revealed that clonal responses to VSV infection were heavily skewed toward MPEC and EEC (Fig 2a) while the response to LM was biased toward SLEC and DPEC development (Fig. 2b). In both infections, while some clonal responses were heavily skewed to a particular phenotype, others were comprised of a mixture of effector subsets (Fig. 2). In order to classify the phenotype of clonal responses, clones with more than 2-fold greater MPEC than SLEC were classified as “MPEC”, while clones with 2-fold greater SLEC than MPEC were classified as “SLEC” while the remainder of the responders were considered “mixed”. The clonal responses following VSV-OVA infection showed a range of phenotypes, and 9 representative examples are shown in Fig. 2c.

Figure 2.

Pathogen-specific effector subset development from CD8⁺ T cell clones. a. Stacked graphs of the phenotypes of the progeny of 20 individual OVA-specific clones observed from one VSV-OVA experiment (20 mice, 40 possible clones). This experiment was repeated 4 times, with 80 mice total. b. Stacked graphs of the progeny of 17 individual OVA-specific clones from one LM-OVA experiment (20 mice, 40 possible clones). This experiment was repeated 3 times, with 60 mice total. c. 9 examples of distinct phenotypes of donor populations after VSV-OVA infection. Clones were classified as MPEC phenotype (2-fold greater MPEC than SLEC), SLEC phenotype (2-fold greater SLEC than MPEC), or MIXED. See also Figure S2.

To better understand if individual TCR expression and avidity influenced effector differentiation, we transferred single CD8⁺ TCR transgenic OT-I T cells (specific for OVA peptide) from naïve OT-I Rag−/− mice to congenic recipients infected with VSV-OVA the following day. Much like OVA-specific clones from the polyclonal repertoire (Fig. 2a), responding single OT-I cells exhibited a broad range of effector phenotypes, with certain clones displaying extremely biased differentiation (e.g clones 1, 2, 19, 20; Supplemental Fig. 2). The fact that individual OT-I CD8⁺ T cells could skew dramatically towards one phenotype or another was not observed in previously published single cell transfer studies(Stemberger et al., 2007; Gerlach et al., 2010). Thus, TCR avidity, although playing a role in clonal selection during the response(Busch et al., 1998; Turner et al., 2008), did not dramatically bias the fate decisions of responding clones.

Clonal heterogeneity contributes to the overall response

To examine the ability of single cells to contribute to the overall response, mice were again infected with either VSV-OVA or LM-OVA after transfer. VSV-OVA infection allowed us to examine responses to both OVA and to VSV nucleoprotein (N), thus allowing analysis of a larger set of clones, while LM-OVA infection allowed us examine OVA responses in a different infectious setting. In total, 80 mice were infected with VSV-OVA after single cell transfer, giving us a possible 160 responders for each specificity. We obtained 63 OVA-specific and 84 N-specific clones from these mice, (Fig. 3a), in keeping with our previous demonstration that N-specific versus OVA-specific frequencies among naïve CD8⁺ T cells are greater(Obar et al., 2008; Jenkins and Moon, 2012). 60 mice were infected with LM-OVA after single cell transfer and 39 OVA-specific clones were observed (Fig. 3a). The phenotypes of the progeny of all clones examined in this study are displayed in Supplemental Fig. 3. Remarkable heterogeneity in the phenotype of the clonal progeny was evident and the stark fate bias of certain clones was unexpected. Additionally, we compared the fate of clones that arose in mice in which two clones responded after VSV-OVA infection. There was no correlation between the phenotype of either OVA-specific clone (Supplemental Fig. 3) or N-specific clones (data not shown), that developed within a given mouse. These results further supported the potential for inherent as well as stochastic factors in controlling the heterogeneity of the response.

Figure 3.

Concatenation of responding clonal progeny recapitulates the host polyclonal CD8⁺ T cell response to infection. a. Phenotypes as classified in Fig. 2c., of all OVA- and N-specific clones from VSV-OVA and LM-OVA experiments. VSV-OVA and VSV-N clones are from 80 mice total (4 separate experiments) with 160 possible clones each. LM-OVA clones are from 60 mice (3 separate experiments) with 120 possible clones. b. The number of cells of MPEC, SLEC, DEPC or EEC phenotype for all clones was determined then mathematically combined to calculate total frequency of each effector phenotype in all VSV-OVA clones, VSV-N clones and LM-OVA clones. c. Breakdown of effector phenotypes within the bulk host OVA- and N-specific response in the same mice from part b. See also Figure S3.

To determine whether the individual clonal responses observed were representative of the bulk response, we calculated the total cell number for EEC, MPEC, SLEC and DPEC within each clonal response and compiled the data across all the mice. The phenotypic distribution of the response between N- and OVA-specific clones from VSV-OVA infection was similar to each other, while OVA clones from LM-OVA were skewed towards SLEC and DPEC (Fig. 3b). In addition we averaged the effector phenotype breakdown from the bulk host response of the same mice, and found that the compilation of the clonal responses recapitulated the bulk response (Fig. 3c). Thus, our sampling of a large number of individual clones was representative of the overall response.

Tissue location differentially regulates effector CD8⁺ T cell generation

Considering the role of the inflammatory environment in shaping the CD8⁺ T cell response, we tested whether the progeny of a single clone would adapt to local signals in distinct tissues. We first compared clonal progenies in two lymphoid tissues, the spleen and the lymph nodes (LN) following VSV-OVA infection. The bulk responses in both sites were highly similar (Fig. 4a, first two columns) and although some minor differences were noted among clonal progeny, these too were comparable (Fig. 4a). Given that CD8⁺ T cell responses in non-lymphoid tissues are often distinct from those in lymphoid tissues(Cauley and Lefrançois, 2013), we wished to test whether single naïve precursor cells would exhibit distinct response outcomes outside of lymphoid tissues. To do so, we compared responses in the spleen and intestine, because the inflammatory milieu of the intestinal immune system is substantially distinct from that of other tissues(Sheridan and Lefrançois, 2011; Honda and Littman, 2012). Again, using our knowledge of naïve precursor frequencies of OVA-specific CD8⁺ T cells, we transferred limited numbers of polyclonal CD8⁺ T cells to recipient mice followed by oral LM-OVA infection one day later. After ten days, the spleen and the intestinal intraepithelial lymphocyte (IEL) compartment were examined for the presence of clonal progeny. The bulk antigen-specific CD8⁺ T cell response in the IEL compartment was skewed away from SLEC and toward EEC and MPEC subsets compared to the splenic response following oral infection (Fig. 4b, first columns). Moreover, the clonal progeny in the two locations were strikingly distinct (clones 1-5), although some clones yielded similar phenotypes (clones 6-8). In some cases clonal progeny were detected in the spleen but not in the intestine, suggesting further heterogeneity perhaps at the level of homing molecule expression (data not shown). These results indicated that populations derived from a single precursor acquired distinct fates based on their tissue environment.

Figure 4.

The fate of a single clone is distinct between the spleen and the intestinal mucosa. a. Stacked graphs of the phenotype of the progeny of 9 transferred OVA-specific clones (present in 20 mice) seven days after VSV-OVA infection in both spleen (S) and peripheral LN (L). Peripheral lymph nodes include axial, brachial, and inguinal. b. Stacked graphs of the phenotype of the progeny of 8 transferred OVA-specific clones found 10 days after oral LM-OVA infection in both the spleen (S) and intestinal epithelium (I). The first set of stacked graphs represents the average distribution of effector populations in the bulk host response of all 20 mice in each experiment. This experiment was repeated twice with similar results.

Intrinsic programming of memory CD8⁺ T cells directs secondary effector cell differentiation

Memory CD8⁺ T cells raised by VSV infection are uniformly CD127⁺ KLRG1- (Fig. 5a). Upon subsequent challenge with VSV, the secondary effector cells were predominantly KLRG1⁺ with ~1/3 of the cells expressing both KLRG1 and IL-7R (DPEC) (Fig. 5b). This pattern was distinct from the primary effector phenotypes where MPEC and EEC predominated. Secondary LM challenge resulted in a very similar pattern of effector phenotypes as that obtained after VSV infection (data not shown). These results suggested that either certain clones were dominating the secondary response or that a default pathway for secondary effector cell differentiation was operating based on intrinsic regulation. As the response environment of a secondary infection may be distinct from that of the primary response, we wished to compare naïve and memory responses in the same host. To this end, we transferred single OVA-specific CD45.1 memory CD8⁺ T cells after VSV-OVA infection along with single OVA-specific CD45.1/2 naïve CD8⁺ T cells, as before, into recipient CD45.2 mice. One day later, recipient mice were infected with VSV-OVA, and at the peak of the response, the phenotype of resulting clonal progeny from both naïve and memory cells were analyzed. Fig. 5c shows an example of one recipient mouse in which both single cells responded and the resulting populations were dramatically different in phenotype. While the memory-derived population was exclusively KLRG1⁺, few cells of this phenotype were present in the naïve cell-derived responders. This pattern held true when all the clones from this experiment were examined (Fig. 5d). The progeny of single memory cells were largely SLEC and DPEC, while the populations derived from single naive CD8⁺ T cells were biased toward EEC and MPEC. Thus, memory CD8⁺ T cells involved in a secondary response appeared to be refractory to at least some external stimuli and followed an intrinsic default pathway toward SLEC and DPEC development.

Figure 5.

Biased development toward KLRG1⁺ secondary effector cells is intrinsic to memory CD8⁺ T cells. a. Phenotype of polyclonal OVA-specific memory CD8⁺ T cells after VSV-OVA infection (>60 days). b. One day after transfer of bulk VSV-OVA-specific memory cells into naïve recipients, mice were infected with VSV-OVA. Phenotype of transferred OVA-specific memory cells 6 days later is shown. c. A single CD45.1 OVA-specific memory CD8⁺ T cell and a single CD45.1/2 OVA-specific naïve CD8⁺ T cell were transferred to naïve mice that were then infected with VSV-OVA. Seven days later KLRG1 and CD127 expression was analyzed on the donor populations. d. Graphs show the phenotype of 13 clones derived from single memory cells and 10 clones derived from naïve cells from one VSV-OVA experiment (20 mice). This experiment was repeated 4 times with 80 mice total.

Clonal burst size analysis reveals additional regulatory mechanisms controlling the immune response

The amount of proliferation for each clone will also determine the degree of input of particular cells to the overall response. An examination of the burst sizes for all of the clones analyzed revealed a broad distribution with a range of ~1×10³-6×10⁵ cells for responders derived from the polyclonal repertoire (Fig. 6a). Surprisingly, the mean clonal burst size of VSV N-specific cells was nearly 2-fold greater than that of VSV-OVA specific clones, many of which developed in the same host. We had previously assumed that the reason that the bulk N-specific response peaked more rapidly than the OVA-specific response was the difference in precursor frequencies(Obar et al., 2008). However, our current result indicated that factors other than precursor frequency regulated the growth of each clone. While the LM-OVA driven responses were similar in size to those from VSV-OVA infected mice, single TCR transgenic OT-I cells expanded dramatically after VSV infection (~ 7-fold greater on average than polyclonal cells) (Fig. 6a), perhaps due to the high avidity of the OT-I TCR(Turner et al., 2008). Interestingly, clonal burst sizes for memory CD8⁺ T cells were similar to those of naïve T cells, although a recent report showed that bulk naïve OT-I CD8⁺ T cells exhibit greater expansion capacity than do memory cells(Martin et al., 2012).

Figure 6.

Clonal burst size is divergent between different specificities and endogenous compared to transgenic CD8⁺ T cells. a. Total number of cells present in the spleen derived from each transferred clone for all experimental conditions as indicated. Mean clone size values are shown with red bars. Below is shown a table of mean and median clone size for all conditions. Two stars indicate a p-value of less than 0.005 and three stars indicate a p-value of less than 0.0005 based on an unpaired T-test. Burst sizes of progeny of OVA-specific (b.) and N-specific clones (c.) from VSV-OVA infected recipients based on the percentage of KLRG1⁺ CD127- cells within each clone. Burst sizes of progeny of OVA-specific clones from LM-OVA infected recipients (d.) and OVA-specific clones from VSV-infected recipients following single OT-I transfer (e.). Spearman correlation (r) and p-values are displayed on each graph. These data are compiled from all of the experiments shown in Figures 2,3 and 5.

We also examined the clonal burst size of subsets based on the percentage of KLRG1⁺ CD127⁻ cells (SLEC) within each clone (Fig. 6b-e). The Spearman correlation (r) was positive for VSVOVA, VSV-N, LM-OVA and VSV-OVA OT-I clones but strongest and most significant for LMOVA and VSV-OVA OT-I clones (Fig. 6b-e). This result indicated that KLRG1⁺ CD127⁻ or SLEC clones had greater expansion capacity than KLRG1- clones. These findings also indicated that the overall composition of the response was a factor of not only the fate of each cell but also the expansion potential of each type of clone. Thus, as the VSV response is dominated by cells with an MPEC phenotype, but SLEC clones expand more robustly, the ratio of clones differentiating along a particular pathway is therefore critical to the overall outcome. The degree of cell death during the expansion phase would also play a role in this process. Of particular interest was the finding that in LM-OVA infection KLRG1⁻ clones expanded poorly compared to KLRG1⁺ clones (Fig 6d) or even KLRG1⁻ clones from VSV infection (Fig. 6b). These data showed that the predominance of SLEC in the anti-LM response was not only due to the greater number of SLEC clones generated but also due to poor expansion of MPEC clones.

DISCUSSION

Here, we utilized a limiting dilution strategy to transfer and track single CD8⁺ T cells from the polyclonal repertoire after infection with VSV or LM. In both infection settings, the progeny of individual clones showed dramatic heterogeneity in the generation of effector populations with some clones forming all of the subsets equally and others biased heavily towards a particular fate. Taking into account phenotype and burst size, the clonal compilation recapitulated the polyclonal response. Additionally, the different inflammatory environments associated with the infections resulted in disparate effector phenotypes, with LM infection inducing more SLEC and DPEC, and VSV infection inducing more MPEC and EEC.

Previous studies have suggested that each naïve OT-I CD8⁺ T cell is capable of differentiating into all effector subclasses(Stemberger et al., 2007; Gerlach et al., 2010). Although our results demonstrated that a majority of clones were able to generate multiple effector subsets, the ratios of each subset within the progeny of a given clone were often vastly different. This finding was evident for single CD8⁺ T cells from the polyclonal repertoire as well as for OT-I cells with a monoclonal TCR. Furthermore, we analyzed the fluorescence intensity of tetramer binding from individual endogenous clonal progenies as an approximate measure of TCR avidity, and while individual clones displayed a range of tetramer binding strengths, no correlation of tetramer intensity with phenotype was observed (data not shown). Thus, although TCR signal strength is clearly involved in CD8⁺ T cell activation and differentiation(Obar and Lefrançois, 2010b; Obar et al., 2010; Teixeiro et al., 2009; King et al., 2012b), these results suggested that TCR avidity is not an essential factor for determining clonal fate. So what is the nature of the signals that results in such heterogeneity? One possibility is that there are inherent differences in naïve CD8⁺ T cells, aside from the TCR. For example, the amount of tonic TCR signaling among individual naïve CD8⁺ T cells could be distinct(Moran et al., 2011; Palmer et al., 2011). In addition, virtual memory cells exist in the naïve CD8⁺ T cell pool and may exhibit distinct response characteristics(Haluszczak et al., 2009). Although these possibilities have not yet been examined with regard to effector subset or memory differentiation, it seems plausible that such factors will play deterministic roles in some way.

The data presented here speak more to the effect of downstream events rather than naïve CD8⁺ T cell heterogeneity. A number of studies have examined the role that the inflammatory environment plays in effector CD8⁺ T cell differentiation(Joshi et al., 2007; Rutishauser et al., 2009; Cui et al., 2009; Pham et al., 2009; Obar et al., 2011). Cytokines including IL-2, IL-12 and IL-27 act to drive robust expansion but promote terminal differentiation of KLRG1⁺ effectors(Cui et al., 2009; Obar et al., 2010; Kalia et al., 2010; Obar et al., 2011), at least in part through modulation of certain transcription factors including T-bet, eomesodermin and BLIMP1(Joshi et al., 2007; Rutishauser et al., 2009; Intlekofer et al., 2005; Takemoto et al., 2006). Moreover, the systemic cytokine milieu after VSV versus LM infection are vastly different in terms of identity and kinetics(Obar et al., 2011), which will also impinge upon differentiation. Additionally, the character of the inflammatory environment can serve to enhance the sensitivity of TCR signaling, and may therefore influence expansion and differentiation(Richer et al., 2013). Nevertheless, what remains unclear is how a single cell gives rise to multiple fates. In one possible scenario, daughter cells downstream of the initial activation event acquire a heterogeneous set of signals through encounters in multiple microenvironments. These environments may include sequential interactions with APC(Celli et al., 2005; Khanna et al., 2007), and we have shown that even when inflammation is unperturbed, antigen recognition late after infection continues to drive CD8⁺ T cell differentiation(Khanna et al., 2007; Obar and Lefrançois, 2010b; Blair et al., 2011). Thus, particular progeny would differentially integrate signals leading to distinct fates, thereby resulting in heterogeneity arising from a single naïve cell. This scenario changes dramatically for memory CD8⁺ T cells. Single memory CD8⁺ T cells developed primarily into DPEC and SLEC, even in an environment where naïve CD8⁺ T cells were driven to produce primarily MPEC and EEC. This result demonstrated cell-intrinsic changes, likely epigenetic, that occur during the generation of memory CD8⁺ T cells. These findings are supported by genetic analysis of repeatedly rechallenged bulk memory CD8⁺ T cells where a focused pattern of gene expression is evident(Wirth et al., 2010).

Our data also demonstrated that certain naive clones exhibit extreme fate bias towards a particular effector subclass. Thus, a subset of clones generated progeny that were nearly exclusively MPEC or SLEC. Understanding the cues that drive such focused development could provide considerable insight into the factors controlling memory development. These extreme examples also speak to the potential for asymmetric division in controlling fate decisions leading to effector versus memory development(Chang et al., 2007). In these cases, unless one daughter-derived lineage was extinguished early on, then all progeny of the initial naïve cell would necessarily have developed into cells of one particular subtype. Otherwise, it would be difficult to reconcile asymmetric division with these results. Alternatively, depending on the strength of TCR signaling, asymmetric division may not always occur(King et al., 2012a). In addition to the infection type influencing effector development, we also showed that tissue-specific factors modulated effector differentiation. Thus, the effector composition of clonal progeny derived from an oral LM infection, where CD8⁺ T cells are primarily primed in the mesenteric LN or Peyer’s patches, were largely distinct between the spleen and intestinal IEL compartment. If one considers that any given clone was initially primed for example in the mesenteric LN, then downstream effects within each tissue must have played a role in influencing the overall composition of the response. This finding once again supported a scenario in which cells were influenced by additional APC-T cell interactions and/or by cytokines downstream of the initial priming event.

In addition to phenotype, our results also emphasize the importance of clonal burst size in shaping the CD8⁺ T cell response to infection. For example, the VSV-N specific clones expanded ~2-fold more than did the OVA-specific clones. Our previous results had led us to hypothesize that the larger precursor frequency for N- versus OVA-specific CD8⁺ T cells explained the larger N-specific response that is a hallmark of the anti-VSV CD8⁺ T cell response (Obar et al., 2008). Our new results indicated that other factors were also influencing response magnitude which could potentially include the quantity and longevity of available antigen. Differences in the growth capacity of cells of each effector type also dramatically influenced the composition of each response. KLRG1⁺ clonal expansion was greater than that of KLRG1-clones and thus the biased composition of effectors in the VSV versus the LM infection was due in part to the ratio of MPEC versus SLEC clones and their relative burst sizes. That is, the dominance of MPEC in the VSV response was the result of a larger number of MPEC clones, despite the increased expansion of a smaller number of SLEC clones. Even more dramatic was the greatly reduced expansion of KLRG1- versus KLRG1⁺ clones after LM infection. These unexpected results suggested that while the inflammation surrounding the LM infection promoted SLEC development by inducing a larger number of SLEC clones, it also appeared to limit the growth of the MPEC population.

Two recent publications have examined single CD8⁺ T cells following LM infection, especially with regard to the contribution of clonal expansion to the response. Both reports utilized TCR transgenic OT-I cells in either single cell transfers or in bar-coding studies that allow tracking of progeny of individual cells (Buchholz et al., 2013; Gerlach et al., 2013). The results show a broad range of clonal expansions of individual T cells, as we also observed. However, we noted that single OT-I cells underwent substantially greater expansion compared to single CD8⁺ T cells from the polyclonal repertoire. These studies also noted a negative correlation between CD62L expression and robust expansion, in keeping with our previous findings where the bulk CD8⁺ T cell response was analyzed(Obar and Lefrançois, 2010b). Mathematical modeling was used to define a stochastic linear differentiation pathway for CD8⁺ T cell diversification, in which central memory (T_CM) precursors generate the other lineages(Buchholz et al., 2013). Such a model is difficult to reconcile with our current data. The regulation of CD62L expression is complex with proteolytic cleavage occurring early after T cell activation followed by transcriptional downregulation of gene expression(Chao et al., 1997) and thus may not provide the best indicator of early memory precursor development. Our previous findings as well as those of others do not support a model in which T_CM are antecedent to all other effector lineages in the primary response, but rather indicate a developmental pathway in which EEC develop into two main lineages of effector cells, SLEC and MPEC. Within the MPEC population two distinct lineages also develop, T_CM and effector memory cells based on CD62L and CCR7 expression. Our work also revealed additional information regarding factors controlling CD8⁺ T cell fate and highlighted the importance of pathogen-induced and tissue-specific inflammatory environments, in particular in the intestinal mucosa, for regulating the ultimate fate of individual clones.

Overall, our results proved the utility of the single cell transfer system using polyclonal T cells to identify the unique clonal response of individual CD8⁺ T cells. Our results also demonstrated that fate decisions after primary, but not secondary, CD8⁺ T cell activation were surprisingly diverse with evidence for extreme developmental biases in some cases. The demonstration that extrinsic factors were essential in shaping the primary but not the secondary response revealed new aspects regarding the importance of contributions of individual clones to the shaping of these responses. Further understanding of the factors that direct effector subset development at the clonal level could enable enhanced memory generation and enhance vaccine design.

Experimental Procedures

Mice

Female C57BL/6 mice (CD45.1 and CD45.2) were purchased from National Cancer Institute (Frederick, MD) and used between 8-12 weeks of age. CD45.1/2 C57BL/6 mice were generated in-house by crossing the single congenic NCI mice. TCR transgenic OT-I Rag1^−/− mice were bred and maintained in-house. The University of Connecticut Health Center Animal Care and Use Committee approved all experiments.

Infections

Infections were performed i.v. with 1×10⁵ pfu VSV-OVA Indiana(Kim et al., 1998) or 1×10³ cfu LM-OVA(Pope et al., 2001) per mouse. For oral LM-OVA infections, all mice were food and water deprived for ~4 hours prior to infection, housed individually with minimal bedding, and given an approximately 0.5 cm³ piece of bread inoculated with 2×10⁹ colony-forming units (CFU) of Listeria monocytogenes (LM) strain 10403s carrying a mutation in the internalin A protein and expressing a truncated form of ovalbumin, generated as described previously (Xayarath et al., 2009).

Flow Cytometry

OVA- and N-specific T cells were stained using the H-2K^b tetramer containing the OVA peptide (SIINFEKL) and the H-2K^b tetramer containing the N peptide (RGYVYQGL) generated in our lab as previously described(Altman et al., 1996; Masopust et al., 2001). Tetramer staining was performed for 1 hour at room-temperature. All other antibodies specific for the indicated molecules were stained for 20 min at 4°. Fluorescence intensities were measured on an LSR-II (BD Biosceinces, San Jose, CA) and data was analyzed using FlowJo software (Tree Star, Ashland, OR).

Single cell transfers

For transfer of single polyclonal CD8⁺ T cells specific for OVA or N, splenocytes from a naïve mouse were enriched for CD8⁺⁺ cells using a negative selection CD8⁺⁺ T cell enrichment kit (Miltenyi Biotec, Auburn, CA). After enrichment, CD8⁺ T cell purity was checked by flow cytometry and 2-4×10⁵ CD8⁺ T cells were transferred i.v. into each recipient mouse. One day following transfer, recipient mice were infected with either VSV-OVA or LMOVA. At the peak of the T cell response (day 7 for VSV, day 8 for i.v. LM and day 10 for oral LM) recipient mice were sacrificed and splenocytes, LN cells or IEL were stained with either APC or PE labeled tetramers. Tetramer positive cells were then bound with anti-PE or anti-APC magnetic beads (Miltenyi Biotec, Auburn, CA) and enriched using a Miltenyi autoMACS separator. Enriched tetramer positive cells were then stained with additional indicated antibodies. For transfer of single OT-I transgenic CD8⁺ T cells, splenocytes from a naïve OT-I Rag−/− mouse were stained with CD8⁺ and Vα2 TCR antibody to estimate frequency (normally >80% of splenocytes) and 1.3-3 OT-I cells were transferred to each recipient mouse. Because of the low number of OT-I cells transferred, 2×10⁵ CD45.2 naïve splenocytes were included per mouse.

For transfer of single memory cells, splenocytes from a VSV-OVA infected mouse (>60 days post infection) were stained with OVA-K^b tetramer to estimate OVA-specific memory cell frequency. 1.3-3 CD45.1 OVA memory cells (2,000-7,000 total splenocytes) and 2-4×10⁵ CD45.1/2 naïve CD8⁺ splenocytes (containing ~1.3 OVA-specific naïve CD8⁺ T cells) were transferred to each recipient mouse.

TCRα and TCRβ sequencing

Progeny of single cells at the peak of the response following VSV-OVA infection were sort purified using an ARIA II (BD Biosciences, San Jose, CA) based on tetramer reactivity and CD8⁺ expression. RNA was isolated using an RNeasy Micro Kit (Qiagen, Germantown, Maryland) and cDNA was made using iScript cDNA synthesis kit (Biorad, Hercules, CA). A multiplex nested RT-PCR protocol was used to amplify TCR sequences as recently described(Dash et al., 2011). After gel purifying PCR products and sequencing (Genewiz, South Plainfield, NJ), sequences were aligned with TCR genes using IMGT/V-QUEST (Giudicelli et al., 2011; Brochet et al., 2008). TCRα and TCRβ sequences are listed in IMGT nomenclature according to the IMGT website (Bosc and Lefranc, 2003).

Statistics

Statistical significance was determined with an unpaired, 2-tailed T-test or a Spearman (nonparametic) correlation test. Prism 5.0 (Graphpad, LaJolla, CA) was used for all statistical analyses.