Constitutive CD8 expression drives innate CD8+ T-cell differentiation via induction of iNKT2 cells

Constitutive CD8 expression induces differentiation of innate memory-like CD8+ thymocytes in both cell-intrinsic and cell-extrinsic manner, the latter being accomplished by an increase in the iNKT2 subset.


Introduction
The thymus provides a specific microenvironment that supports development of several types of T cells, including innate-like T cells, such as invariant natural killer T (iNKT) cells and mucosal-associated invariant T (MAIT) cells, and conventional T cells. Signaling via the TCR plays a central role in driving differentiation of both innate-like and conventional T cells (Hogquist & Jameson, 2014), although the TCR diversity and the selecting MHC-presenting antigens are quite different between these two types of T cells; innate-like T cells such as iNKT cells and MAIT cells express invariant TCRs that recognize nonpeptide antigens presented on non-classical MHC, CD1d and MR1, respectively (Godfrey et al, 2015), whereas conventional T cells express diverse TCRs recognizing peptide antigens presented on classical MHC (Hogquist & Jameson, 2014).
Essential roles of CD4/CD8 co-receptors in TCR/MHC interaction during differentiation of the two major conventional T-cell subsets, CD4 + helper and CD8 + cytotoxic cells, from common precursors, CD4 + CD8 + double-positive (DP) thymocytes, are well characterized. Thymocytes positively selected by class II MHC molecules (MHC-II selected thymocytes) develop into CD4 + CD8 − single-positive (SP) thymocytes that are committed to the helper lineage, whereas MHC-I-selected thymocytes are directed to become CD4 − CD8 + SP thymocytes committed to the cytotoxic lineage (Ellmeier et al, 1999). It has been proposed that differences in the duration of the positive-selection signal instruct distinct fates in post-selection thymocytes (Singer et al, 2008). Thus, briefer TCR signals in MHC-I-selected thymocytes caused by temporal down-regulation of the CD8 co-receptor guide post-selection thymocytes to differentiate into CD4 − CD8 + SP thymocytes. On the other hand, persistent TCR signals in MHC-II-selected thymocytes supported by constitutive CD4 expression activate a developmental program toward the helper-lineage T cells via induction of the zing-finger transcription factor ThPOK (He et al, 2005;Sun et al, 2005) through antagonizing a transcriptional silencer in the Zbtb7b gene encoding ThPOK (He et al, 2008;Setoguchi et al, 2008). Therefore, in what is called the kinetic signaling model, distinct expression kinetics between CD4 and CD8 co-receptors have been proposed to play a key role in segregating helper and cytotoxic lineages (Singer et al, 2008). In line with this model, perturbation of positive-selection signaling duration in MHC-II-selected thymocytes re-directs them to become CD8 + cytotoxic-lineage cells (Sarafova et al, 2005;Singer et al, 2008;Adoro et al, 2012). On the other hand, constitutive transgenic CD8 expression guides about 30% of MHC-I-selected thymocytes to differentiate into CD4 + cells (Bosselut et al, 2001). One proposed explanation for the low efficiency of such redirected differentiation was down-regulation of the transgenic CD8α chain that heterodimerized with endogenous CD8β chain.
In addition to TCR signals, cytokines play important roles in controlling T-cell differentiation in the thymus. Signals by IL-7 are crucial for the differentiation of CD8 SP thymocytes . Recently, IL-4 has been shown to support differentiation of another type of CD8 SP thymocyte with the characteristics of both the memory and innate cells, which is referred to as innate memory-like CD8 T cells (Weinreich et al, 2010). The iNKT2 subset of iNKT cells produces IL-4 and has been shown to be a major source of IL-4 in the thymic environment. Accordingly, an increase in the numbers of iNKT2 cells, although they represent only a tiny subpopulation of total thymocytes, has a significant impact on the generation of innate memory-like CD8 T cells (Lee et al, 2013). In addition to iNKT2 cells, iNKT1 cells expressing IFN-γ and iNKT17 cells expressing IL-17 are also differentiated from iNKT precursors (Constantinides & Bendelac, 2013). However, little is known about how balanced differentiation of such iNKT-cell subsets is regulated.
In this study, we generated a novel transgenic mouse model expressing the CD8αβ heterodimer or the CD8αα homodimer in the absence of endogenous CD8α/CD8β chains and MHC-II molecules and observed that two-thirds of MHC-I-selected thymocytes differentiated into CD4 − CD8 + SP thymocytes, most of which acquired signatures of innate memory-like CD8 T cells in both cell-intrinsic and cell-extrinsic manner. The cell-extrinsic mechanism was linked to results from enhanced differentiation of the iNKT2-cell subset. Thus, our study sheds new light on the physiological relevance of down-regulation of the Cd8 gene to fine-tune the balance of iNKTcell subsets.

Results
Developmental pathway to CD4 + T cells from the CD8 SP stage in Zbtb7b ΔTE/ΔTE mice The initial activation of the Zbtb7b gene upon receiving positiveselection signals is achieved mainly by a thymic enhancer (TE) (He et al, 2008;Muroi et al, 2013). Therefore, removal of the TE from the Zbtb7b locus results in delayed and low-level expression of ThPOK in newly selected thymocytes (Muroi et al, 2013). However, sequential activation of a proximal enhancer restores ThPOK expression in a later developmental stage . The impaired ThPOK expression kinetics due to loss of the TE in Zbtb7b ΔTE/ΔTE mice results in redirected differentiation of a small proportion of MHC-II-selected thymocytes into CD4 − CD8 + SP thymocytes (Muroi et al, 2013). On the other hand, enforced ThPOK expression in CD8 + T cells was shown to activate some helperlineage signature genes (Jenkinson et al, 2007). Given the restored ThPOK expression in the later developmental stage of MHC-IIselected thymocytes in Zbtb7b ΔTE/ΔTE mice, we examined whether there are MHC-II-selected thymocytes from some developmental stages toward the CD8-lineage that acquire a CD4 + CD8 − phenotype. We accomplished this by a fate-mapping approach using an E8I-Cre transgene, whose expression is driven by the E8I enhancer known to be activated specifically in mature CD4 − CD8 + SP thymocytes (Ellmeier et al, 1997) and can be monitored by GFP expression from ires-gfp sequences present in the transgene . By crossing the E8I-Cre transgenic mice to a Rosa26 YFP reporter strain, about 70% of CD8 + splenic T cells were marked by YFP expression, which could be distinguished from the weaker GFP signal derived from the E8I-Cre transgene (Fig 1A), whereas only 0.02% of CD4 + splenic T cells were marked by YFP. On the other hand, the percentage of YFP-positive CD4 + T cells was significantly increased, up to 0.07% on average, in Zbtb7b ΔTE/ΔTE mice (Fig 1A and B). This result indicates that there exists a developmental pathway toward CD4 + T cells in the Zbtb7b ΔTE/ΔTE mice from a developmental stage where the E8I-Cre transgene is activated, which occurs after CD4 downregulation in control mice . Such reversible differentiation to CD4 + T cells after possible loss of CD4 expression was not compatible with the kinetic signaling model, prompting us to revisit the relevance of distinct co-receptor expression kinetics in the CD4 + helper versus CD8 + cytotoxic lineage choice.

Differentiation of innate memory-like CD8 T cells under constitutive CD8αβ expression conditions
The effect of constitutive CD8αβ co-receptor expression in vivo was examined previously in the presence of the endogenous CD8β chain. Possible down-regulation of the transgenic CD8α chain, which could dimerize with the endogenous CD8β chain, was discussed as a possible reason why most of the MHC-I-restricted cells were differentiated into CD4-negative cytotoxic T cells in this transgenic model (Bosselut et al, 2001). To analyze the effect of constitutive CD8αβ expression in the absence of both endogenous CD8α and CD8β chains, we generated a mutant Cd8 locus, by sequential targeting in ES cells, referred to as Cd8 Δab , that lacks coding regions for both CD8α and CD8β chains in the Cd8a and Cd8b1 genes, respectively (Figs 2A-C and S1). For transgenic expression of CD8α and CD8β chains, a cDNA encoding either chain was inserted into the Rosa26 locus, generating a Rosa26 8a and Rosa26 8b locus, respectively (Fig 2A). Induction of CD8α and CD8β chains from the Rosa26 8a and Rosa26 8b loci upon Cre-mediated excision of the stop cassette was confirmed by flow cytometry analyses ( Fig 2D). However, the expression level of transgenic CD8α, which is represented in CD4 + population at the lower left panel in Fig 2D was lower than that of endogenous CD8α (CD8 SP population at the upper left panel in Fig 2D), presumably because of the weak promoter activity in the Rosa26 locus.
We next examined expression kinetics of transgenic CD8α expression in Rosa26 8a/8a : Cd8 Δab/Δab : Cd4-Cre mice. As expected, expression level of transgenic CD8α from the Rosa26 locus was stable and did not shown down-regulation after positive selection ( Fig S2), confirming constitutive expression of CD8 co-receptor. To examine differentiation of MHC-I selected-thymocytes, we then generated I-Ab −/− : Rosa26 8a/8b : Cd8 Δab/Δab : Cd4-Cre mice, hereafter referred to as MHC-II o :CD8ab-Tg mice. Despite lower transgenic CD8αβ expression levels, the percentage of mature thymocytes, defined as the CD24 lo TCRβ hi population, was equivalent between Wt and MHC-II o :CD8ab-Tg mice ( Fig 3A). In the mature thymocytes population of MHC-II o :CD8ab-Tg mice, although around 30% cells acquired CD4 expression, about 70% cells emerged as CD4-negative cells ( Fig 3A). The balance of CD4 + and CD4 − cells was maintained after their egress from the thymus. In the peripheral lymphoid tissues, such as spleen, of MHC-II o :CD8ab-Tg mice, the αβT-cell population consisted of about 65% CD4 − cells and 35% CD4 + cells ( Fig S3A). This proportion of CD4 + to CD4 − cells in MHC-II o :CD8ab-Tg mice was similar to that reported in the previous transgenic model expressing CD8α and CD8β from mini-transgenes at equivalent levels to endogenous CD8αβ (Bosselut et al, 2001), indicating that, in spite of the lower level of transgenic CD8αβ expression, our CD8αβ transgenic model is likely to mimic developmental processes in the previous model. We next examined expression of Runx3, a central transcription factor for cytotoxic T-cell development, by using a Runx3 tdTomato reporter allele (Kojo et al, 2017). CD4 − mature thymocytes emerged in MHC-II o :CD8ab-Tg mice expressed Runx3-tdTomato, albeit at a lower level than that in control CD8 + T cells in control mice (Fig 3A), indicating that these cells are related to the cytotoxic lineage. To further characterize the CD4 − mature thymocytes in MHC-II o :CD8ab-Tg mice, we performed transcriptome analyses by RNA-seq. Principal component analyses indicated that CD4 − mature thymocytes that had differentiated in MHC-II o :CD8ab-Tg mice were different from control wild-type CD8 SP thymocytes ( Fig S3B). Analyses of differentially expressed genes revealed that Gzmk, Zbtb16, and Tbx21 genes were up-regulated in the CD4 − mature thymocytes ( Fig 3B). We also confirmed higher expression of Eomes, one of the signature genes for innate memory-like CD8 T cells (Weinreich et al, 2010), in CD8 SP thymocyte population of MHC-II o :CD8ab-Tg mice in subsequent qRT-PCR and flow cytometry analyses (Figs 3B and S3C). CD44 and CD122 are known surface markers for innate memory-like CD8 T cells. Higher CD44 expression was detected in about half of the CD4 − mature thymocytes in MHC-II o :CD8ab-Tg mice and CD122 expression was detected in a half of the CD44 hi cells (Figs 3C and S3D). An increase in CD44 hi CD122 hi cells was also observed in spleen CD4 − αβTcell population ( Fig S3E). These observations demonstrate that half of the CD4 − mature thymocytes that emerge in MHC-II o :CD8ab-Tg mice are innate memory-like CD8 T cells. An increase in the CD44 hi CD122 hi subset was also induced by transgenic CD8αβ expression in an MHCsufficient background ( Fig 3C). We also found that expression of just the CD8α chain that could function as a CD8αα homodimer resulted in enhanced differentiation of CD44 hi CD122 hi innate memory-like CD8 T cells ( Fig 3C). Thus, under lower but constitutive CD8αβ or CD8αα expression, more than half of the MHC-I-selected cells remain to be differentiated into CD4 − cytotoxic lineage-related cells with a skewed differentiation toward innate memory-like CD8 + cells.

Increase in the iNKT2 subset under constitutive CD8αβ expression
During the analysis of promyelocytic leukaemia zinc finger (PLZF) expression, we noticed that around 40% of CD4 + mature thymocytes expressed a higher level of PLZF than that in innate memory-like CD8 T cells ( Fig 4A). Innate-like CD4 + T cells, including CD1d-restricted iNKT cells (Kovalovsky et al, 2008) and MR1-restricted MAIT cells (Rahimpour et al, 2015), have been shown to express PLZF at high levels. Interestingly, differentiation of innate memory-like CD8 T cells in the thymus has been shown to be enhanced by IL-4 secreted by iNKT2 cells (Verykokakis et al, 2010;Weinreich et al, 2010;Lee et al, 2013). We, therefore, examined CD4 + PLZF hi cells in MHC-II o : CD8ab-Tg mice in terms of iNKT characteristics, such as expression of the invariant TCR, which can be specifically detected by CD1d dimers loaded with αGalCer (αGC) (Matsuda et al, 2000). About 30% of CD4 + mature thymocyte of MHC-II o :CD8ab-Tg mice were stained with CD1d-αGC and were positive for transgenic CD8 expression, whereas only 5% of those cells were stained in control mice (Figs 4B and S4A). Such iNKT cells were present at a similar ratio in CD4 + T-cell population in MHC-II o mice expressing only endogenous CD8αβ ( Fig 4B). On the contrary, in MHC-II-sufficient background, both frequency and numbers of iNKT cells were reduced in our CD8ab-Tg mice as was previously reported (Lantz & Bendelac, 1994), whereas these are restored in MHC-IIdeficient background ( Fig S4A). These observations indicate that iNKT cells are one of the major CD4 + subsets arising after elimination of MHC-II-selected thymocytes. Whereas PLZF expression was not detected in the CD4 + CD1d-αGC − population in wild-type control mice, there remained a small proportion of CD1d-αGC − PLZF + CD4 + cells in the thymus of MHC-II o :CD8ab-Tg mice as well as MHC-II o mice ( Fig 4B).
Although the frequency of iNKT cells was similar between MHC-II o and MHC-II o :CD8ab-Tg mice, an increase in innate memory-like CD8 + T cells occurred only in MHC-II o :CD8ab-Tg mice ( Fig 3C). Given the important role of IL-4 to support innate memory-like CD8 + T-cell differentiation (Verykokakis et al, 2010;Lee et al, 2013), we asked whether differentiation of iNKT-cell subsets is different between MHC-II o and MHC-II o :CD8ab-Tg mice. After CD24 down-regulation, the developmental stages of iNKT cells are divided into the CD44 − NK1.1 − stage 1, the CD44 + NK1.1 − stage 2, and the CD44 + NK1.1 + stage 3 (Benlagha et al, 2002). In the thymus of MHC-II o :CD8ab-Tg mice, proportions of CD44 − NK1.1 − stage 1 and CD44 + NK1.1 − stage 2 cells were significantly increased compared with control and MHC-II o mice and, this was accompanied by a relative decrease in CD44 + NK1.1 + stage 3 cells ( Fig  4C). The T-bet lo PLZF hi iNKT2-cell subset is included in the stage 1 and stage 2 iNKT-cell population (Constantinides & Bendelac, 2013;Lee et al, 2013). As expected, the proportion of the T-bet lo PLZF hi iNKT2-cell subset was significantly higher in MHC-II o :CD8ab-Tg mice than in wildtype and MHC-II o mice (Fig 4C), which was consistent with an increase in the number of iNKT2 cells in MHC-II o :CD8ab-Tg mice (Fig S4B). We further confirmed an increase in the iNKT2 subset by using other markers, Plzf and RORγt, that define iNKT2 subset as Plzf hi RORγt − population ( Fig S4C). In addition, after ex vivo PMA/ionomycin stimulation, thymic iNKT cells of MHC-II o :CD8ab-Tg mice produced more IL-4 than wild-type control mice (Fig 4D), confirming the increase in the iNKT-cell subset that functionally produces IL-4. These observations indicate that constitutive expression of the CD8αβ co-receptor results in a skewed differentiation toward the iNKT2-cell subset.

Cell-intrinsic and cell-extrinsic mechanisms for innate-like CD8 T-cell development
Our results showed that differentiation of both innate memory-like CD8 thymocytes and the iNKT2-cell subset is enhanced in MHC-II o : CD8ab-Tg mice. Given the known role of IL-4 secreted by iNKT2 cells in supporting innate memory-like CD8 thymocyte differentiation (Verykokakis et al, 2010;Lee et al, 2013), we next tested to what extents the increase in the iNKT2-cell subset has impacts on innate memory-like CD8 T-cell differentiation in MHC-II o :CD8ab-Tg mice. To accomplish this aim, we set up mixed bone marrow chimera experiments in which equal numbers of bone marrow progenitors from CD45.1 wild-type mice and CD45.2 wild-type or MHC-II o : CD8ab-Tg mice were co-injected together into sublethal irradiated MHC-II o host mice. In the host mice that received CD45.1 wildtype and CD45.2 wild-type bone marrow cells, the percentages of CD44 + CD122 + innate memory-like cells in mature CD4 − CD8 + SP thymocytes were similar between CD45.1 + and CD45.2 + populations  and were less than 1% in both population (Fig 5A). In striking contrast, the percentage of CD45.1 + CD44 + CD122 + CD8α + cells were increased to around 5% when progenitors from CD45.2 MHC-II o :CD8ab-Tg mice were included (Fig 5A). The degree of CD45.1 + CD44 + CD122 + CD8 SP thymocyte differentiation tended to be correlated with the degree of chimerism of CD45.2 + cells to CD45.1 + cells. These results are consistent with previous studies showing that iNKT2-mediated cell-extrinsic mechanisms, such as an IL-4-enriched thymic environment, support innate memory-like CD8 T-cell differentiation (Verykokakis et al, 2010;Lee et al, 2013). Although actual involvement of IL-4-producing iNKT2-cell subsets in enhanced differentiation of innate memory-like CD8 T cells in MHC-II o :CD8ab-Tg needs to get confirmed by genetics using mice deficient for IL-4 or CD1d, it is likely that skewed differentiation of iNKT2 cells could serve as a cell-extrinsic mechanism driving innate memory-like CD8 T differentiation.
Interestingly, in this setting, the percentage of CD44 + CD122 + cells differentiated from CD45.2 MHC-II o :CD8ab-Tg progenitors was always higher (threefold) than that from CD45.1 wild-type cells (Fig 5A). This higher percentage of CD44 + CD122 + cells in the CD45.2 + population suggests that besides cell-extrinsic mechanisms such as higher level of IL-4 produced by iNKT2 cells, constitutive CD8αβ expression activates uncharacterized cell-intrinsic mechanisms that makes MHC-I-restricted cells more prone to activate a developmental program toward the innate memory-like CD8 thymocyte.
We next examined development of the iNKT-cell subset in the bone marrow chimera host mice by assessing the ratio of CD45.2 + cells to CD45.1 + cells at the three stages of iNKT-cell development. This was performed individually for each host mouse because iNKTcell reconstitution efficiency varied in each host. In the mixed chimera setting of CD45.1 wild-type with CD45.2 wild-type progenitors, the ratio of CD45.2 + cells to CD45.1 + cells in each of the three iNKT-cell stages was around one in all host mice (Fig 5B). On the contrary, the proportion of CD45.2 + cells relative to CD45.1 + cells at the CD44 − NK1.1 − stage 1 and the CD44 + NK1.1 − stage 2 was significantly increased in host mice injected with CD45.1 wild-type and CD45.2 MHC-II o :CD8ab-Tg progenitors (Fig 5B). We also confirmed an increased proportion of the T-bet lo PLZF hi iNKT2-cell subset in this setting ( Fig 5B). These results clearly demonstrate that cell-intrinsic mechanisms are involved in the skewed differentiation toward iNKT1 and iNKT2-cell subsets in MHC-II o :CD8ab-Tg mice, although involvement of cell-extrinsic mechanisms is not formally excluded.

Intracellular signaling from the CD8α chain is required for iNKT2 skewing
Our results revealed that constitutive CD8αβ expression enhanced iNKT2-cell differentiation in at least a cell-intrinsic manner. Recently, TCR signal strength was shown to be involved in regulating the differentiation of iNKT-cell subsets. Thus, strong TCR signaling promotes iNKT2 and iNKT17 development (Tuttle et al, 2018;Zhao et al, 2018), suggesting that constitutive CD8αβ expression might enhance TCR signaling during iNKT-cell development, either by aiding in antigen recognition and/or modulating intracellular TCR signals. Consistent with this notion, expression level of CD5, an indicator of TCR signal strength, was higher on thymic iNKT cells that were developed under constitutive CD8 expression (Fig S4D). iNKT cells are selected by CD1d molecules on DP thymocytes (Constantinides & Bendelac, 2013). Contrary to the previous study showing CD1d down-regulation by transgenic CD8 expression (Engel et al, 2010), there was no significant change in CD1d expression levels on DP thymocytes upon transgenic CD8αα or CD8αβ expression (Fig S4E), suggesting that CD1d expression level is unlikely to be involved in enhancement of TCR signaling by transgenic CD8 expression. Two cysteine residues in the CD8α cytoplasmic tail are essential for co-receptor function in TCR stimulation by classical MHC-I via interacting with Lck kinase and recruitment of LAT (Turner et al, 1990;Hoeveler & Malissen, 1993;Bosselut et al, 1999). We, therefore, tested whether these cysteine residues are necessary to enhance iNKT2-cell differentiation by generating a Rosa26 8aCA allele that expresses a mutant CD8α chain in which two cysteine residues are replaced with alanine ( Fig 6A). Unlike in the MHC-II o : CD8ab-Tg mice, the frequency and numbers of mature CD8 SP thymocytes were significantly decreased in the MHC-II o mice expressing CD8α CA mutant protein (MHC-II o :CD8a CA b−Tg mice) (Fig 6B), indicating that co-receptor function supporting conventional MHC-I-restricted cell development was abrogated by these amino acid replacements.
When we examined the iNKT-cell population, defined as CD24 lo TCRβ + CD1d-αGC + , in MHC-II o :CD8a CA b−Tg mice, there was no increase in the percentage of the iNKT2-cell subset (Fig 6C). These observations indicate that intracellular signaling through the CD8α chain is necessary to drive enhanced iNKT2-cell development.
It remains unclear whether CD8αβ or CD8αα interacts directly with the non-classical MHC-I molecule CD1d. Nevertheless, our results clearly demonstrated that constitutive CD8α expression alone is sufficient to affect iNKT-cell subset differentiation. We then examined expression kinetics of the endogenous CD8α chain during iNKTcell development. At the CD24 hi CD44 − NK1.1 − stage 0, most of iNKT-cell precursors express CD8α and CD4 and more than half of these cells down-regulate CD8α expression (Fig S5). During the transition from CD24 hi stage 0 to the CD24 lo CD44 − NK1.1 − stage 1, CD8α expression was efficiently down-regulated. The ThPOK transcription factor is essential to terminate Cd8 gene expression in conventional MHC-Irestricted cytotoxic T cells (He et al, 2005;Rui et al, 2012) as well as in iNKT cells (Engel et al, 2010). Interestingly, CD8α down-regulation at stage 0 seemed to precede ThPOK induction (Fig S5). Thus, similar to conventional cytotoxic T cells , the initial CD8 downregulation during iNKT-cell development is independent of ThPOK.

Discussion
Results shown in this study clearly demonstrate that constitutive expression of CD8αβ or CD8αα influences the proportion of iNKTcell subsets, with skewing to the iNKT2 subset in a cell-intrinsic subpopulation. One representative result of at least four experiments. Graphs showing summary of at least four experiments (lower). Mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 (one-way ANOVA with Tukey's multiple comparison). (D) Histogram showing intracellular IL-4 staining in the thymic iNKT population after stimulation with PMA and ionomycin for 4 h. One representative of two experiments. manner, generating an IL-4-rich thymic microenvironment that fosters differentiation of innate memory-like CD8 + T cells. In addition to the cell-extrinsic mechanism, constitutive CD8αβ expression somehow conditions MHC-I-restricted cells to become more prone to acquire innate memory-like signatures in a cellintrinsic manner.
Recent studies reported that TCR signal strength influences iNKT-cell subset differentiation (Tuttle et al, 2018;Zhao et al, 2018). Attenuated function of ZAP70, which plays an essential role in early TCR signaling events via phosphorylation of LAT, impaired differentiation of iNKT2 and iNKT17, but not iNKT1 cells (Tuttle et al, 2018;Zhao et al, 2018). The expression level of CD5, which parallels TCR signaling intensity, and induction of Nur77, another marker reflecting TCR strength, is higher in iNKT2 cells (Moran et al, 2011;Lee et al, 2013;Tuttle et al, 2018). These findings highlight that strong TCR signaling promotes iNKT2-cell development, at least in part by sustaining Egr2 expression, which results in increases of chromatin accessibility of iNKT2-specific regulatory elements harboring Egr2-and NFAT-binding motifs (Tuttle et al, 2018). Therefore, it is conceivable that constitutive CD8αβ expression enhances TCR signaling in iNKT precursors. However, a previous study failed to detect direct binding of CD8αβ to the non-classical MHC-I molecule, CD1d (Engel et al, 2010), leaving open the question of how CD8αβ enhances TCR signaling engaged by CD1d. Our results showed that the CD8αα homodimer, which has a distinct affinity for conventional MHC-I and a thymus leukemia antigen (TL), another non-classical MHC-I molecule, from CD8αβ co-receptor (Leishman et al, 2001;Gangadharan & Cheroutre, 2004), also functions to enhance iNKT2 development. This suggests that CD8αβ is unlikely to help iNKT precursors recognize antigen on CD1d as it does for antigens on classical MHC-I as a co-receptor, although the possibility that CD8αβ increases binding affinity of invariant TCR with antigen/CD1d molecules is not formally excluded. Using a retrogenic transgenic mouse system, Cruz Tleugabulova et al showed that TCR half-life is more important to modulate iNKT subset differentiation than avidity (Cruz Tleugabulova et al, 2016). This finding is not only in line with the idea that CD8αβ does not increase affinity for antigen/ CD1d complexes but also raise the possibility that CD8αβ stabilizes the interaction between TCR and antigen/CD1d complexes. On the other hands, our results show that two cysteine residues within the intracellular domain of CD8α chain are required for skewed differentiation toward the iNKT2 subset. These cysteine residues function as a docking module for Lck and LAT (Turner et al, 1990;Hoeveler & Malissen, 1993). Such an intracellular association of CD8α chain with TCR/CD3 complexes and signaling molecules in the absence of antigen recognition suggests a possible role of CD8α chain to strengthen TCR signaling independently of extracellular binding of CD8α to CD1d. Given the involvement of ZAP70 in shaping the development of iNKT subsets, the capacity of the CD8α chain to interact with the molecular machinery in early TCR signaling events is likely to be essential to strengthen or prolong TCR signaling. At any rate, given that CD4 is constitutively expressed during iNKT-cell development and interacts with Lck with higher affinity than CD8α, it is possible that CD8α modulates intracellular signaling from invariant TCR engaged by antigen/CD1d complexes differently from CD4.
Pioneering studies proposed that CD8αβ expression during iNKT-cell development inhibits this development through a process of negative selection (Lantz & Bendelac, 1994). However, emergence of CD8 + iNKT cells in ThPOK-deficient mice indicated that CD8 expression does not provide full inhibitory signals (Engel et al, 2010). Recently, reduction of iNKT-cell numbers in SLAM family receptor-deficient mice was shown to stem from increased apoptosis because of strong TCR signals (Lu et al, 2019). Thus, the SLAM family receptor functions to attenuate TCR signals to foster iNKTcell development, implying that differentiation of iNKT cells is impaired once TCR signal strength exceeds some thresholds. Therefore, enhanced TCR signaling by constitutive CD8αβ expression inhibits differentiation of some iNKT precursors that receive excessive TCR signals, whereas other iNKT precursors that can survive under constitutive CD8αβ expression are directed toward the iNKT2-cell developmental pathway. It is possible that such iNKT2 cells that emerged under constitutive CD8αβ expression are exposed to prolonged TCR signals and produce an increased amount of IL-4, generating the IL-4-rich thymic microenvironment despite a small increase in their numbers.
A previous publication discussed the possibility that downregulation of endogenous CD8β chain could be what limits the redirection of MHC-I selected cells into CD4 + helper T cells (Bosselut et al, 2001). However, in the absence of endogenous CD8β, the proportion of CD4 + versus CD4 − cells was about one-third, which is similar to that observed in the previous study (Bosselut et al, 2001), implying that this is not the case. Moreover, one-third of the CD4 + T cells are CD1d-restricted iNKT cells, and the rest of the CD4 + T population contains a significant proportion of PLZF + innate-like cells with uncharacterized MHC-restriction. Therefore, at least in our experimental setting for transgenic CD8αβ expression, most MHC-I-selected cells differentiate into CD8 + cytotoxic-lineage cells. This finding demonstrates that a persistent positive-selection signal alone is not sufficient to instruct the helper-lineage fate to MHC-I restricted cells. Thus, our results challenge the model in which a distinct duration of positive-selection signals between MHC-I-and MHC-II-selected thymocytes generated by distinct kinetics of expressions of the CD4 versus CD8αβ co-receptor is central to mechanisms that govern the helper versus cytotoxic lineage choice. Our results also revealed that down-regulation of CD8 plays an essential role in preventing development of innate memory-like CD8 T cells, rather than in inducing conventional cytotoxic fate, at least in part by shaping differentiation of iNKT-cell subsets. By using several experimental systems, disruption of CD4/ MHC-II-mediated positive-selection signals had been previously shown to result in a redirected differentiation of MHC-II restricted cells into CD8 + cytotoxic-lineage cells (Sarafova et al, 2005;Adoro et al, 2012). This genetic evidence supports the concept that persistence of positive-selection signals is necessary to guide MHC-IIselected thymocytes to become the helper-lineage T cells. On the contrary, whether persistence of MHC-I-mediated TCR signaling is sufficient to instruct the helper-lineage fate has not yet been well established. Given the lower expression level of transgenic CD8αβ in this study, it remains possible that the signal strength was not sufficient to instruct the helper-lineage fate in this setting. However, the ratio of CD4 + versus CD4 − mature thymocytes in this study was similar to that observed in mice expressing transgenic CD8αβ at a level similar to endogenous CD8αβ (Bosselut et al, 2001). In addition, the lower expression level of transgenic CD8αβ or CD8αα was sufficient to influence TCR signals from invariant TCR on iNKT precursors. Therefore, it is conceivable that persistence of MHC-I-derived positive-selection signals alone is not sufficient to instruct the helper-lineage fate, although it is still important to test thymocyte fate decision by achieving constitutive CD8αβ expression at the correct level (Littman, 2016). Our results suggest that there must be uncharacterized differences in positive-selection signals between MHC-I-and MHC-II-restricted thymocytes beyond duration of signal length. Understanding of such differences in positive-selection signals engaged by distinct MHCs is extremely important and awaits experiments in future studies.
The unique kinetics of CD8 expression, known as co-receptor reversal, was discovered 20 yr ago (Brugnera et al, 2000;Cibotti et al, 2000). CD8 co-receptor reversal is controlled at the transcriptional level and is accomplished by temporal termination of Cd8 gene expression in all post-selection thymocytes and sequential Cd8 reactivation specifically in the cytotoxic-lineage cells. However, the molecular mechanisms underlying such CD8 expression with its unique kinetics remains uncharacterized. Our results revealed a novel relevance of Cd8 down-regulation to shaping iNKT-cell subset differentiation, raising the possibility that mechanisms causing Cd8 down-regulation might be evolved to fine-tune innate T-cell differentiation in species that lack conventional cytotoxic T cells or in the extant positive-selection process of cytotoxic T cells by classical MHC-I molecules. However, expression of the CD8αβ co-receptor on cytotoxic-lineage cells is beneficial to efficiently mount adaptive immune responses by assisting recognition of peptide antigen presented on MHC-I molecules in the periphery, leading us to speculate that mechanism(s) that reactivate Cd8 gene might have been acquired later. In mice, the E8I enhancer, one of six enhancers in the Cd8 locus, shows CD8 + cytotoxic lineage-specific activity in a reporter transgenic assay (Ellmeier et al, 1997). However, CD8αβ coreceptor expression by MHC-I-restricted cells is maintained in the absence of the E8I enhancer (Ellmeier et al, 1998), indicating that compensatory mechanisms operate to maintain CD8αβ co-receptor expression on MHC-I-restricted cytotoxic T cells (Ellmeier et al, 2002). Such mechanisms for CD8 expression are likely to have been established under evolutional pressures to secure MHC-I-dependent pathogen clearance.

Mice
Thpok ΔTE mice (Muroi et al, 2013), Rosa26-STOP-YFP mice (Srinivas et al, 2001), and Runx3-tdTomato reporter mice (Kojo et al, 2017) have been described. I-Aβ-deficient mice were from Taconic. To generate a Cd8 Δab allele, we first replaced exon 2 in the Cd8b1 gene with a neomycin resistant gene (neo r ) flanked with FRT sites by homologous recombination with the targeting vector in MI ES cells, establishing ES clone 67. We next removed the neo r gene from the ES clone 67 by transient transfection of an FLP recombinase expression vector and isolated an ES clone harboring the Cd8 Δb allele, into which the targeting vector used to generate the Cd8a h2PA allele (Wada et al, 2018) was transfected to replace exon 1 of the Cd8a gene with an hCD2 cDNA and the neo r gene. After isolation of ES clones that had undergone homologous recombination with the second targeting vector at the Cd8a gene, we screened them to determine which allele, Cd8 or Cd8 Δb , was targeted by analyzing the type of FLP recombinase-mediated recombination occurring after transduction of a retrovirus vector encoding FLP, and then isolated ES clones harboring the Cd8 +/ΔbaCD2N genotype. After removal of the neo r gene from the Cd8 +/ΔbaCD2N ES clone by transient transfection of the FLP recombinase expression vector, ES clones were used to generate chimera mice by aggregation. To generate the Rosa26 8a or Rosa26 8b allele, a cDNA fragment encoding CD8α or CD8β was PCR-amplified by RT-PCR using thymus mRNA, sequenced, and cloned into the AscI site of the CTV vector (#159212; Addgene). A cDNA encoding the mutant CD8α CA chain harboring alanine substitutions at two cysteine residues in the CD8α cytoplasmic tail was amplified by overlap PCR. All mice were maintained in the specific pathogen-free animal facility at the RIKEN IMS, and all animal procedures were in accordance with institutional guidelines for animal care and with the protocol (28-017) approved by the Institutional Animal Care and Use Committee of RIKEN Yokohama Branch.

Quantitative RT-PCR
Total cellular RNA was extracted from purified cell subsets using Trizol (Thermo Fisher Scientific) and treated with RNase-free DNase I (Thermo Fisher Scientific). cDNA was synthesized from total RNA using the SuperScriptII First Strand Synthesis System (Invitrogen). Quantitative RT-PCR was performed using the ABI/PRISM 7000 sequence detection system with an internal fluorescent TaqMan probe. Primers for Zbtb7b (He et al, 2005) and Eomes (Intlekofer et al, 2005) were previously described.

RNA-seq
RNA was extracted from sorted CD8 + and CD4 − mature thymocytes of Wt and MHC-II o :CD8αβ Tg mice by TRIzol (Thermo Fisher Scientific), followed by RNeasy Micro Kit (QIAGEN) according to the manufacturer's protocols. Sequencing libraries were prepared using a NEBNext RNA Library Prep Kit for Illumina (E7530; NEB) with Poly(A) mRNA Magnetic Isolation Module (E7490; NEB) according to the manufacturer's protocol. Single-end 50-bp reads were obtained by Illumina HiSeq 1500. The reads were mapped by Tophat v2.1.1 onto the mouse genome mm10 and counts were generated using HTSeq v.0.6.1. The DESeq2 R package was used to perform principle component analysis and to generate the heat map of differentially expressed gene with normalized counts of wild-type (GSE48138) and Tg (GSE132059) in duplicates in this study.

Accession codes
GEO: raw sequencing data and processed files, GSE48138 and GSE132059.