Co-receptor choice by Vα14i NKT cells is driven by Th-POK expression rather than avoidance of CD8-mediated negative selection

We previously performed surface plasmon resonance studies to identify an interaction between CD8 and the thymus leukemia antigen, a nonclassical class I molecule (Fisher et al., 1985; Leishman et al., 2001). Experiments performed in parallel failed to convincingly demonstrate binding of CD8 even to CD1d tetramers (unpublished data). However, because there could be biologically meaningful interactions close to or below the level of detection, we focused our efforts on in vitro and in vivo functional assays to determine if the expression of CD8 affects the antigen response of Vα14i NKT cells. Initially, retroviruses were used to direct the expression of CD8α and CD8β chains in three mouse Vα14i NKT cell hybridomas. The 1.2 and 3.C3 cells express a TCR with Vβ8.2 β chains having different CDR3 regions, whereas the 1.4 TCR utilizes an atypical Vβ10 polypeptide (Burdin et al., 1998). Cells expressing near physiological levels of CD8αβ were selected by magnetic and flow cytometric cell sorting (Fig. 1 A and not depicted). We also established control hybridoma sublines transduced with empty retrovirus vector, and we compared the CD8-expressing Vα14i NKT cell hybridomas with the controls for IL-2 production in response to thymocytes pulsed with the synthetic antigen αGalCer, a potent agonist related to the antigens found in Sphingomonas bacteria. Antigen-pulsed CD1d transfectants of the B cell lymphoma A20 and bone marrow–derived DCs were also tested as APCs. Furthermore, to allow for the possibility that CD8αβ expressed by the hybridomas might bind to classical class I molecules on the APC, providing an adhesion function separate from a true coreceptor function, we also tested the hybridomas in an APC-free system with αGalCer presented by plate-bound CD1d protein. If CD8 acted as a TCR coreceptor, we would expect a shift in the antigen dose response curve as a result of enhanced antigen sensitivity mediated by CD8. However, we did not find a consistent difference between the CD8-expressing hybridomas and empty vector control transductants in their antigen response (Fig. 1 B and not depicted), even at αGalCer concentrations as low as 2 ng/ml. Similar experiments using the Sphingomonas-derived glycolipid GalA-GSL, a less potent agonist, also failed to find any effect of CD8 expression on the antigen response (unpublished data).

We also measured the effect of CD8 expression on the avidity of the 1.2 and 1.4 hybridoma lines for αGalCer-loaded CD1d (αGalCer-CD1d) tetramers by staining the CD8-expressing and control lines with serial dilutions of tetramers and determining the fluorescence intensity via flow cytometry. The resulting tetramer isotherm analyses revealed that CD8 expression had little or no effect on αGalCer-CD1d tetramer avidity (Fig. S1). Similar experiments using CD1d tetramers loaded with the less potent αGalCer analogue α glucosyl ceramide also did not reveal any effect of CD8 expression on antigen avidity (unpublished data).

We examined if CD8 normally acts to modulate Vα14i NKT cell differentiation. CD8-null mice were previously reported to contain elevated fractions of Vβ7⁺ NKT cells, (Bendelac et al., 1994), suggesting that CD8 plays a role in shaping the Vβ repertoire of Vα14i NKT cells. Subsequent results from other studies suggested that Vα14i TCRs containing Vβ7 have a higher affinity for endogenous ligands presented by CD1d and, therefore, the increase in Vβ7-expressing cells in the absence of CD8 is consistent with a partial release from negative selection (Chun et al., 2003; Schümann et al., 2006; Wei et al., 2006). Precise identification of Vα14i NKT cells was difficult at the time these earlier studies were performed, however, because CD1d tetramers were not available.

We stained thymocytes from CD8-null and WT mice with αGalCer-CD1d tetramers together with antibodies against Vβ2, Vβ7, and Vβ8.1/8.2. We observed no significant difference between the Vβ repertoires of Vα14i NKT thymocytes from CD8-deficient and WT mice (Table I), suggesting that CD8 expression normally does not modulate the selection of Vβ segments by Vα14i NKT cells.

The observations in the previous sections led us to reexamine if the development of Vα14i NKT cells is affected by enforced expression of CD8 in the T cell lineage. We therefore obtained and analyzed the same CD8 transgenic line that was examined in earlier studies (Robey et al., 1991; Bendelac et al., 1994). In these mice, CD8α and β expression is controlled by enhancer and promoter sequences derived from the CD2 gene (Robey et al., 1991). Thymus, spleen, and liver cell suspensions prepared from CD8 transgenic and WT mice were stained with αGalCer-CD1d tetramers together with a mix of antibodies to define Vα14i NKT cells and other subsets. In agreement with the original study, we found that mice homozygous for the CD8 transgene (CD8 Tg/Tg) exhibited markedly reduced (>10-fold) numbers of Vα14i NKT cells in both the thymus and periphery (Fig. 2). Surprisingly, mice that were hemizygous for the CD8 transgene (CD8 Tg/+) had significant populations of Vα14i NKT cells, although the numbers were approximately two- to threefold lower than those in WT littermates (Fig. 2). Some of the Vα14i NKT cells that expressed CD8 also coexpressed CD4, although the percentage of DN Vα14i NKT cells was increased compared with WT mice (Fig. 2 A).

Vα14i NKT development in the thymus is characterized by the successive induction of high levels of CD44 followed by NK1.1 (Benlagha et al., 2002). Mutations that disrupt NKT development could cause a block in the transition from CD44^low NK1.1^low (stage 1) to CD44^high NK1.1^low (stage 2) or from stage 2 to CD44^high NK1.1^high (stage 3), and hence result in higher percentages of stage 1 or stage 2 NKT cells. To determine whether enforced CD8 expression arrested NKT development at either stage 1 or 2, we assessed the expression of CD44 and NK1.1 on Vα14i NKT thymocytes from CD8 transgenic and nontransgenic littermates. The fractions of NKT cells in stages 1, 2, and 3 in either CD8 hemizygous or homozygous transgenic mice were similar to those of nontransgenic littermates, suggesting that expression of CD8 did not affect the developmental transition from stage 1–3 (Fig. 2 A).

The influence of CD8 transgene copy number on Vα14i NKT cell differentiation suggested the possibility that there was a threshold level of CD8 beyond which Vα14i NKT cell development was nearly completely inhibited. Thus, it was important to compare the levels of CD8 expressed in CD8 Tg/+ mice to those expressed by WT CD8⁺ SP thymocytes and mature T cells. We measured the anti-CD8 median fluorescence intensity (MFI) observed in WT SP CD8⁺ thymocytes, mature CD8⁺ T cells, and Vα14i NKT cells and conventional T lymphocytes from CD8 transgenic mice. We determined the number of anti-CD8 binding sites by interpolation of the MFI from each population against a standard curve prepared by staining beads with defined numbers of antibody binding sites. An analogous approach was also used to measure the numbers of anti-CD4 binding sites in CD4⁺ Vα14i NKT cells and conventional CD4⁺ T lymphocytes. We found that the numbers of both anti-CD8α and anti-CD8β binding sites on mature (TCR^high) CD4⁺ thymocytes and CD4⁺ peripheral T cells in CD8 Tg/+ mice were similar to those observed in WT CD8⁺ SP thymocytes and T cells (Fig. S2 and not depicted). Vα14i NKT cells from CD8 Tg/+ mice were found to express levels of both CD8α and CD8β that were approximately half of those observed in WT CD8⁺ SP thymocytes and T cells (Fig. S2 and not depicted). However, a similar reduction in the amount of CD4 was also observed in CD4⁺ Vα14i NKT cells from either WT or CD8 transgenic animals as compared with WT CD4⁺ SP thymocytes and T cells (Fig. S2). These data demonstrate that the CD8 transgene construct drove expression levels in transgene hemizygous T cells that appeared physiologically appropriate. Furthermore, they suggest that the lower CD8 amounts observed in Vα14i NKT cells from CD8 Tg/+ mice might be attributable to general mechanisms that normally act to down-regulate CD4, and perhaps also expression of the TCR, in Vα14i NKT cells.

We examined the response of CD8 Tg/+ mice to αGalCer administered in vivo to determine if CD8 expression affected Vα14i NKT cell function. Spleen and liver cell suspensions, together with serum samples, were prepared 90 min after αGalCer injection. The cell suspensions were stained to identify Vα14i NKT cells, followed by intracellular staining for IFN-γ, IL-4, and TNF. We observed no difference in the response of CD8⁺ and WT Vα14i NKT cells to antigen administered in vivo, as measured by intracellular cytokine staining (Fig. 3 and not depicted). Thus the expression of CD8 had no apparent effect on the in vivo antigen response of Vα14i NKT cells.

If CD8 acted as a coreceptor when expressed by Vα14i NKT cells, it would be predicted to increase the avidity and to decrease the TCR off rate for cognate antigen (Garcia et al., 1996). We therefore examined if CD8αβ increased Vα14i NKT cell antigen avidity by comparing CD8 transgenic and WT Vα14i NKT cells in two different CD1d tetramer binding assays. An equilibrium binding or tetramer binding isotherm analysis was performed by mixing WT and CD8 Tg/+ thymocytes and by staining aliquots with serial dilutions of αGalCer-CD1d tetramers, followed by staining with antibodies against CD8 and other T cell surface markers that permit unambiguous identification of Vα14i NKT cells. We also measured the tetramer off rates from WT and CD8 transgenic Vα14i NKT cells by determining the decrease in αGalCer-CD1d tetramer mean fluorescence intensity after incubation of stained aliquots at 37°C. We could not detect any effect of CD8 expression on Vα14i NKT cell antigen avidity or the off rate (Fig. S3).

The modulation of CD1d expression by thymocytes can affect NKT cell Vβ repertoire, with changes in CD1d levels correlating inversely with the fraction of Vα14i NKT cells expressing Vβ7 (Chun et al., 2003; Schümann et al., 2006; Wei et al., 2006). These data suggest that Vβ7 usage is enhanced under conditions of reduced antigen avidity. We thus examined the fraction of cells expressing Vβ7 and Vβ8.1/8.2 in CD8⁺ Vα14i NKT cells from CD8 Tg/+ mice. The CD8⁺ Vα14i NKT cell populations in the thymus contained an increased fraction of cells expressing Vβ7 (Fig. S4), which would be the opposite of expectation if CD8 expression enhanced the avidity of Vα14i NKT cells for endogenous ligands. Breeding the CD8 transgene onto a K^b and D^b-null background had no influence on the increased Vβ7 expression by Vα14i NKT thymocytes from CD8 Tg mice (unpublished data; Vugmeyster et al., 1998). These findings indicate that the CD8 transgene could modulate Vα14i NKT cell selection through a mechanism independent of its ability to interact with classical MHC class I antigen-presenting molecules but in a manner apparently inconsistent with CD8 coreceptor function.

The effect of the CD8 transgene on the Vα14i NKT cell TCR Vβ repertoire was similar to that observed in mice with reduced levels of thymocyte CD1d expression (Schümann et al., 2006; Wei et al., 2006). These observations led us to examine CD1d expression by cells from CD8 Tg mice. We found no differences in the amount of CD1d between CD8 Tg/+ and WT mice in mononuclear cells obtained from the spleen or liver (not depicted) or in mature TCR^high thymocytes (Fig. S5 A). In contrast, we found that CD1d expression in TCR^low DP thymocytes from CD8 Tg hemizygotes was reduced, with the MFI obtained by anti-CD1d staining only ∼60% of that observed in nontransgenic littermates (Fig. S5). The CD1d staining intensity of the TCR^low DP thymocyte subset in CD8 Tg/Tg mice was reduced even further to ∼30% of the level of WT (Fig. S5 B and not depicted). The amount of CD1d expression by the CD8 Tg/+ mice was not much greater than that expressed by Cd1d^+/− mice, whereas the CD8 Tg homozygotes expressed less CD1d than either strain (unpublished data). These data implied that the effect of the CD8 Tg on the Vα14i NKT TCR Vβ repertoire was, at least in part, the result of an unanticipated reduction in CD1d expression by DP thymocytes in these mice, and they suggest that the effect of increased levels of CD8 in the Tg homozygous mice had an additive effect on CD1d expression.

The CD8 Tg mouse line used in this and the earlier study (Bendelac et al., 1994) is known to exhibit defects in thymocyte maturation (Wack et al., 2000). We therefore performed experiments to test if the deficit in Vα14i NKT cells in CD8 Tg homozygotes could be related to the more general defects in thymocyte development, as opposed to self-ligand–mediated negative selection. As previously reported (Wack et al., 2000), the transition from the DN to the DP stages is partially inhibited in thymocytes from the CD8 Tg line, which was reflected both by increases in the fractions of DN and immature SP thymocytes and by decreases in the fraction of DP thymocytes and overall thymocyte cellularity (Fig. 4, A and B; and not depicted). Furthermore, we found that the severity of these maturation defects was much more pronounced in mice that were homozygous for the CD8 transgene (Fig. 4, A and B; and not depicted).

Thymus hypocellularity can be attributed to defects either in thymocyte expansion or survival. To determine if thymocyte survival is perturbed in CD8 Tg mice, we used ex vivo cultures. DP thymocytes from CD8 Tg/Tg mice exhibited a markedly reduced viability in culture compared with either Tg/+ or WT littermates (Fig. S6 A). These data suggest that the lifespan of DP thymocytes in vivo in CD8 Tg/Tg mice was likely to be compromised, which could account for the more than fivefold decrease in total thymocyte number. Additionally, we observed that CD8 Tg/Tg DP thymocytes had lower TCR expression and higher forward light scatter than hemizygous and nontransgenic littermates (Fig. S6, B and C). These phenotypes suggest that CD8 Tg/Tg DP thymocytes had a higher fraction of relatively immature cells, and they are consistent with, although they do not formally demonstrate, a survival defect within the more mature DP subset.

Mutations that restrict DP thymocyte lifespan are known to cause severe reductions in the development of Vα14i NKT cells (Bezbradica et al., 2005; Egawa et al., 2005). This has been attributed to the fact that gene rearrangement within the TCR-α locus proceeds in an ordered fashion, with recombination beginning at the 5′ or Vα proximal end of the Jα cluster, and then progressing in a 3′ or distal direction via secondary α rearrangements that occur during thymocyte maturation (Yannoutsos et al., 2001). Because the canonical Vα14-Jα18 rearrangement utilizes a Jα element near the 3′ end of the Jα locus, Vα14i NKT cell development is highly dependent on these secondary rearrangements that are decreased when DP thymocyte survival is impaired. We therefore performed multiplex PCR analysis to assess the relative frequency of rearrangement between Vα14 and both proximal and distal Jα elements. Rearrangement events between Vα14 and Jα elements near the proximal end of the Jα locus were readily detectable in all of the thymus samples regardless of the presence of CD8 transgene (Fig. 4 C). The canonical Vα14-Jα18 rearrangement was dramatically reduced in CD8 Tg/Tg (Fig. 4 C), but not in CD8 Tg/+ thymocyte DNA, which fits with the low frequency of Vα14i NKT thymocytes in homozygotes and near normal numbers in CD8 Tg/+ mice. However, CD8 Tg thymocytes exhibited lower frequencies of rearrangement events between Vα14 and all distal Jα elements assessed. Although this was particularly evident in thymocytes from mice homozygous for the CD8 transgene, reductions in the very distal Jα gene segments were observed even in CD8 Tg/+ mice (Fig. 4 C). These data demonstrate that the CD8αβ transgene induced a general defect in rearrangement to the distal end of the Jα locus, and they imply that the inhibition of Vα14i NKT cell development in CD8 Tg/Tg mice is not a result of antigen receptor specificity but rather of a reduction in the frequency of rearrangements necessary for the expression of the Vα14i TCR.

To confirm that the decrease in Vα14i NKT cells in CD8 Tg/Tg mice is the result of a Jα rearrangement defect, we circumvented the putative gene rearrangement road block by generating mice that were homozygous for the CD8 transgene and also carried a transgene that enforced expression of a rearranged Vα14i TCR-α chain in all T cells (Bendelac et al., 1996). Analyses of these Vα14i Tg + CD8 Tg/Tg mice demonstrated that the introduction of the Vα14i transgene resulted in the generation of large numbers of αGalCer-CD1d tetramer-binding T cells and that a substantial portion of these cells exhibited a CD44^high NK1.1^high surface antigen phenotype that is characteristic of mature WT Vα14i NKT cells (Fig. 5 A). Furthermore, Vα14i NKT cells from Vα14i Tg + CD8 Tg/Tg double transgenic mice made a robust response to αGalCer stimulation in vivo, with equivalent IFN-γ synthesis and only slightly decreased IL-4 synthesis compared with Vα14i Tg mice hemizygous for the CD8 Tg (Fig. 5 B). These data confirm that thymocytes in CD8 Tg homozygotes expressing the Vα14i NKT TCR are not subject to negative selection and, furthermore, that the reduced amount of CD1d expressed by CD8 Tg/Tg thymocytes is sufficient to promote Vα14i NKT cell differentiation.

To determine why Vα14i NKT cells do not normally assume a CD8 SP fate, we explored the possibility that differentiation of the Vα14i NKT cell lineage might involve the expression of factors that inhibit CD8 expression. The Zbtb7b gene directs the expression of the Th-POK or c-krox protein, which is a member of the BTB-POZ family of Zn finger transcriptional repressors which is required for the development of conventional CD4 SP T cells (He et al., 2005). Mice that lack expression of a functional Th-POK protein lack mature CD4 SP T cells, and Th-POK–null thymocytes expressing TCRs that would normally promote positive selection to a CD4 SP lineage are instead redirected to a CD8 SP fate (Keefe et al., 1999; He et al., 2005). Th-POK has also been shown to repress CD8 expression by conventional CD4 peripheral T cells (Wang et al., 2008). Moreover, enforced expression of Th-POK changes the fate of thymocytes selected by classical MHC class I molecules from CD8 SP to CD4 SP (He et al., 2005; Sun et al., 2005). There is no evidence, however, that Th-POK has effects on the negative selection of T lymphocytes.

We examined whether a deficiency in Th-POK affected Vα14i NKT cells by phenotypic analysis of mice homozygous for the spontaneous helper deficient (hd) mutation in the Th-POK gene (Dave et al., 1998; He et al., 2005). We found that the numbers and percentages of Vα14i NKT cells in hd/hd mice were similar to those in WT or hd/+ littermates (Fig. 6, A and B). Interestingly, the mature Vα14i NKT cells from hd/hd mice were either CD8 SP or DN, with the ratio of CD8 SP to DN similar to the CD4/DN ratio observed in mice WT for Th-POK (Fig. 6 C). These data demonstrate that Th-POK is required for the differentiation of both CD4 SP Vα14i NKT and conventional T cells but not for the generation of DN Vα14i NKT cells, although there is evidence that these T lymphocytes are derived from a CD4⁺ precursor (Benlagha et al., 2005).

Although Vα14i NKT cells from hd/hd mice exhibited no defects in the expression of surface antigens such as CD44 and NK1.1, we did find that they expressed abnormally low levels of the activation marker CD69 (Fig. 7 A). These data suggested that hd/hd Vα14i NKT cells might have a maturation or activation defect. We therefore subjected hd/hd and WT control mice to in vivo challenge with αGalCer. As assessed by intracellular cytokine staining and measurement of serum IL-4 levels, we found that hd/hd mice exhibited a markedly reduced in vivo response to αGalCer (Fig. 7, B and C). The decrease was strongest for IL-4 production but also involved TNF synthesis and IFN-γ to a lesser extent. These findings demonstrate that Th-POK is required for full Vα14i NKT cell function, although normal numbers of Vα14i NKT cells are generated in the hd mutant mice. Furthermore, the reduced response was not a direct consequence of CD8 expression by the mature cells because DN Vα14i NKT cells from hd/hd mice exhibited intracellular cytokine levels that were the same as or lower than those of the CD8⁺ subset (Fig. S7). However, it remained possible that very low levels of CD8 expression were sufficient to render Vα14i NKT cells hyporesponsive or that prior expression of CD8 partially anergized DN Vα14i NKT cells. We thus crossed hd mutants with mice carrying a targeted deletion of the CD8α gene and measured the responses of hd/hd CD8α^−/− and hd/hd CD8α^+/− Vα14i NKT cells to in vivo αGalCer challenge by intracellular cytokine staining (Fung-Leung et al., 1991). The response of hd/hd NKT cells was not affected by their CD8α genotype (Fig. 7 D). These data indicate that the effects of Th-POK deficiency on the function of Vα14i NKT cells are independent of its role in repressing CD8 expression in these cells.

Our data suggested that Th-POK might play a role in the induction of cytokine transcription upon antigen recognition by Vα14i NKT cells. We therefore prepared liver mononuclear cell RNA from mice with or without WT Th-POK function that had been challenged by αGalCer and measured the relative amounts of Il4 and Ifng transcripts by quantitative real-time PCR analysis of cDNA generated from these preparations. In agreement with the intracellular cytokine staining data, the induction of Il4 transcripts by αGalCer was dramatically reduced in Th-POK mutants, whereas the effect of Th-POK deficiency on Ifng transcript accumulation was more modest (Fig. 8).

Vα14i NKT cells are distinguished by the constitutive expression of detectable amounts of both Il4 and Ifng RNA in the absence of exogenous antigen stimulation (Stetson et al., 2003). The presence of these transcripts is likely to be a reflection of the epigenetic structure of the Il4 and Ifng gene loci, which may contribute to the ability of these cells to produce cytokines rapidly upon exposure to antigen. To determine if Th-POK contributed to the expression of basal cytokine transcripts, we analyzed RNA from liver mononuclear cells from unstimulated mice that were highly enriched for Vα14i NKT cells. We found that Il4 transcripts were dramatically reduced in Th-POK mutant Vα14i NKT cells (Fig. 8). Basal Ifng transcripts, in contrast, were only modestly affected by Th-POK mutation (Fig. 8).

Our data suggest that Vα14i NKT cells normally express Th-POK, which might be required not only for the repression of CD8 expression but for their functional maturation. To examine Th-POK gene expression in Vα14i NKT cells, we analyzed Th-POK^GFP/+ mice, in which the coding region of one allele of the Th-POK gene was replaced with sequence encoding GFP (Fig. 9; Setoguchi et al., 2008). Analysis of both the thymus and periphery of Th-POK^GFP/+ mice revealed that the intensity of GFP expression in the vast majority of Vα14i NKT cells was similar to that of conventional CD4 SP T cells, regardless of whether the Vα14i NKT cells were CD4 SP or DN (Fig. 9). Qualitatively similar results were obtained when we analyzed CD4 SP and DN Vα14i NKT cells from F2F3 strain mice containing a Th-POK-GFP transgene (unpublished data; He et al., 2008). Altogether, these data indicate that Th-POK is expressed by essentially all Vα14i NKT cells and that it is necessary for normal Vα14i NKT cell basal cytokine gene transcription and secretion by antigen-stimulated cells.

The regulation of CD4 and CD8αβ expression during thymocyte maturation is largely dictated by the TCR coreceptor function of these molecules, as they participate in the recognition of self-ligands presented by MHC class I or class II molecules, respectively. The roles, if any, of these coreceptors in the function of those TCR-αβ⁺ T cell populations not selected by classical MHC class I or class II molecules are not as well understood. The initial characterizations of Vα14i NKT cells reported that they were absent from mice in which CD8 expression was enforced in T cells and, furthermore, that their Vβ repertoire was altered in CD8-deficient mice (Bendelac et al., 1994; Lantz and Bendelac, 1994). These data led to the conclusion that differentiating NKT cells in the thymus that expressed CD8 were negatively selected, and they implied that CD8 acts as a CD1d coreceptor to increase the avidity of these cells for targets expressing the selecting glycolipid antigens. According to this view, the thymic self-ligands for Vα14i NKT cells are of sufficient avidity that they are true TCR agonists or close to that level of affinity. Therefore, they can positively select Vα14i NKT cells in the absence of a coreceptor, with CD8 coreceptor expression in fact causing their negative selection.

In this manuscript, however, we report that Vα14i NKT cells can develop and function in the context of enforced CD8 expression, and we demonstrate that CD8 has no apparent effect on Vα14i NKT cell avidity for cognate antigen presented by CD1d. The negative effect of the CD8 Tg on the thymic differentiation of Vα14i NKT cells is not a result of negative selection, but is related to the effect of the transgene on the normal developmental progression of all thymocytes. The mouse line used in our studies was in fact previously shown to exhibit a reduced efficiency of early thymocyte maturation from the DN to DP stage (Wack et al., 2000), which we have also observed (unpublished data). The previous paper did not report any effects of the CD8αβ transgene on DP lifespan, however, and in fact demonstrated that a CD8 transgene construct that was not expressed until the DP stage had no effect on thymic cellularity. It is important to note, however, that all of the transgenic animals examined in the previous study were hemizygous for the CD8αβ transgenes. We also observed a much less dramatic phenotype in hemizygotes as compared with CD8 Tg homozygotes, in terms of the DP lifespan during in vitro culture, the accumulation of DP thymocytes in vivo and overall thymus cellularity, and the frequency of distal Jα rearrangements. Thus, our data collectively indicate that the CD8αβ transgene can affect at least two stages of thymocyte development but the effects of the transgene on DP thymocytes are much more apparent in homozygous animals. The mechanism for the developmental effect of the CD8 transgene on DP cell number is not known, and because the transgene overexpression is not physiological, we have not sought to determine the mechanism in this study of Vα14i NKT cell differentiation. We speculate, however, that the overexpression of CD8αβ could lead to the sequestration of critical signaling molecules, such as the src family tyrosine kinase lck, away from TCR signaling complexes. Aberrant TCR signaling in mice homozygous for the CD8 transgenes would then lead to a reduction in the DP thymocyte lifespan, with effects on rearrangement within the TCR-α locus that are the likely proximal cause of their Vα14i NKT cell deficiency.

The Vα14i NKT cells that are found in CD8 Tg/+ mice are functionally normal, as measured by intracellular cytokine expression after antigen challenge in vivo. In addition, we have found no differences in the expression of CD69, NK1.1, or other NK and activation markers between Vα14i NKT cells from CD8 Tg/+ mice and nontransgenic littermates (unpublished data). In contrast, the CD8 transgene does appear to affect the TCR repertoire of Vα14i NKT cells, in that Vα14i NKT thymocytes from CD8 Tg hemizygous mice have an elevated fraction of Vβ7⁺ cells. These data might suggest that the expression of CD8 directly affects the avidity of Vα14i NKT cells for selecting determinants, but contrary to the conclusions drawn from some previous studies, they imply that Vβ7-expressing Vα14i NKT cells are more dependent, rather than less dependent, on the added avidity putatively provided by CD8. However, a second unexpected finding from the CD8 Tg mice, in addition to a reduced DP lifespan, is decreased CD1d expression by thymocytes. We consider it likely that the changes in the Vβ repertoire of Vα14i NKT cells observed in these mice are the result of reduced CD1d expression rather than CD8 mediating a coreceptor function or enhanced cell–cell interactions. This interpretation is consistent with studies showing that the fraction of Vβ7⁺ Vα14i NKT cells correlated inversely with CD1d expression levels (Chun et al., 2003; Schümann et al., 2006; Wei et al., 2006). Although CD1d expression is further reduced in CD8 Tg homozygotes as compared with CD8 Tg hemizygotes or CD1d^+/− mice, the fact that introduction of a Vα14i transgene restores NKT development and function demonstrates that these low CD1d levels are nevertheless sufficient to drive the differentiation of Vα14i NKT cells, and they point to the rearrangement defect as the principal cause of the NKT deficit in the CD8 Tg/Tg background.

Vα14i NKT cell development is uniquely dependent on a variety of receptors, signaling molecules, and transcription factors (Matsuda and Gapin, 2005). Recent studies of mice deficient for the nuclear high mobility group protein Tox have shown that this factor is required for the development of both conventional CD4⁺ T cells and the Vα14i NKT cell subsets (Aliahmad and Kaye, 2008), demonstrating a common genetic requirement for these two T lymphocyte subtypes. Our data reveal that the BTB-POZ zinc finger transcription factor Th-POK also has important roles in directing the differentiation of both conventional CD4⁺ and Vα14i NKT cells. In Th-POK mutant mice, the differentiation of all CD4 SP cells, including conventional and Vα14i NKT cells, is redirected toward a CD8 SP fate (He et al., 2005; Keefe et al., 1999). Interestingly, in addition to CD8⁺ cells, a population of DN Vα14i NKT cells remains in these mice. Additionally, our data indicate that Th-POK is expressed in all mature Vα14i NKT cells, whether they are CD4 SP or DN. Th-POK–deficient Vα14i NKT cells are hyporesponsive to antigenic stimulation. In WT mice, DN Vα14i NKT cells are highly responsive to stimulation with αGalCer (Kronenberg and Gapin, 2002), and Vα14i NKT cells are also normal in CD4-deficient mice (Bendelac et al., 1994). In this manuscript, we demonstrate that DN Vα14i NKT cells in the Th-POK mutant mice are as hyporesponsive as their CD8-expressing counterparts and that deletion of the CD8α gene has no effect on the response of Th-POK–deficient Vα14i NKT cells to αGalCer. Collectively, these data suggest that the functional defect in immune responses by Vα14i NKT cells in the Th-POK mutant mice is not a direct consequence of either the absence of CD4 expression or the presence of CD8. Our data further suggest that factors such as Th-POK are expressed as a consequence of the activation of the Vα14i NKT cell developmental program. Because Th-POK is capable of repressing CD8, this implies that Vα14i NKT cells down-regulate CD8 at least in part as a consequence of Th-POK expression (Wang et al., 2008). Furthermore, unexpectedly Th-POK is also required for the full functional maturation of these cells. Interestingly, another BTB-POZ family member, PLZF, has been recently shown to play a major role in specifying the function of Vα14i NKT cells (Kovalovsky et al., 2008; Savage et al., 2008). However, PLZF-deficient Vα14i NKT cells exhibit changes in surface antigen expression, tissue distribution, and antigen response that are distinct from those in Th-POK mutants (Kovalovsky et al., 2008; Savage et al., 2008). Thus, it remains to be determined whether the structural similarities between Th-POK and PLZF bear any relation to the mechanisms underlying their diverse effects on Vα14i NKT development and function.

In conclusion, we provide strong evidence that the development of CD8⁺ Vα14i NKT cells is not prevented by negative selection and that when CD8 expression is either forced or removed by genetic manipulation there is little effect on Vα14i NKT cell repertoire or function. Furthermore, our data suggest that the transcription factor Th-POK, expressed essentially throughout the Vα14i NKT developmental program, is required for the normal function of these cells. As a by-product of this expression, they are CD8 negative and directed to either a CD4 SP or DN fate.

The CD8αβ Tg mouse line was obtained from cryopreserved storage at The Jackson Laboratory. Mice with targeted mutations in CD1d (Mendiratta et al., 1997) were provided by L. van Kaer (Vanderbilt University, Nashville, TN). CD8α-deficient mice were purchased from The Jackson Laboratory (Fung-Leung et al., 1991). The hd/hd and Th-POK^GFP mouse lines have been previously described (Keefe et al., 1999; Setoguchi et al., 2008). Mice with mutations in H-2K^b and H-2D^b (Vugmeyster et al., 1998) were gifts from A. Sette (La Jolla Institute for Allergy and Immunology, La Jolla, CA). Vα14i transgenic mice (Bendelac et al., 1996) were obtained from A. Bendelac (University of Chicago, Chicago, IL). All mice, except for the Th-POK^GFP line, had been backcrossed extensively onto a C57BL/6 background. Mice were housed under specific pathogen-free conditions and the experiments were in accordance with animal protocols approved by the Institutional Animal Care and Use Committees at the La Jolla Institute for Allergy and Immunology or the Fox Chase Cancer Center.

Full-length cDNAs encoding mouse CD8α and CD8β were subcloned into the retroviral vector MigRI (Pear et al., 1998). Transfection of the ϕNX ecotropic retroviral packaging cell line and transduction of iNKT cell hybridomas were performed as previously described (Kinsella and Nolan, 1996; Engel and Murre, 1999). Transduced lines were enriched for CD8β⁺ cells by labeling with magnetic µ beads and passage over LS columns (Miltenyi Biotec) and selected for expression of near-physiological levels of CD8 by cell sorting with a FACSVantage SE (BD). TCR expression on transduced hybridoma lines was examined by flow cytometry before each αGalCer stimulation assay, and magnetic enrichment or cell sorting was used when necessary to equalize TCR expression levels between CD8-expressing and empty vector–transduced lines.

Thymocytes (5 × 10⁶/ml) or A20.CD1d cells (10⁶/ml; Brossay et al., 1998b) were incubated with various concentrations of αGalCer for 4 h at 37°C and then washed, resuspended at the same cell concentration, and added to 96-well flat-bottom tissue culture plates in a volume of 0.1 ml/well. Hybridomas (5 × 10⁴/well) were added to wells in a volume of 0.1 ml, and cells were incubated overnight at 37°C. Stimulation with αGalCer-loaded purified CD1d and ELISA assays were performed as previously described (Sidobre et al., 2002). In vivo antigen challenge was performed by injecting mice i.p. or i.v. with 2 µg αGalCer dissolved in 0.2 ml of PBS. Mice were killed 90–120 min after injection, at which time blood was drawn for serum cytokine analysis and spleen and liver mononuclear cell suspensions were prepared.

R-phycoerythrin–conjugated and allophycocyanin streptavidin-CD1d tetramers loaded with αGalCer were produced as previously described (Sidobre et al., 2002). Allophycocyanin-conjugated streptavidin-CD1d tetramers loaded with the αGalCer analogue PBS57 were supplied by the National Institutes of Health Core Tetramer Facility. All antibodies were purchased from BD, eBioscience, or BioLegend. Flow cytometry was done using either a FACSCalibur or LSRII (BD), and data was analyzed using FlowJo software (Tree Star, Inc.). Intracellular cytokine staining was performed using a kit and protocol purchased from BD. Primary mouse cell suspensions analyzed were obtained from mice between 6 and 10 wk of age.

A multiplex PCR strategy to detect TCR-α rearrangements has already been described (Pasqual et al., 2002). Genomic DNA was isolated from thymocytes or RAG-deficient splenocytes using the DNeasy Blood and Tissue kit (QIAGEN) and was amplified using an upstream primer specific for the Vα14 gene segments paired in separate reactions with primers specific for Jα 2, 9, 16, 48, and 56. Each primer pair amplified the specific Vα14-Jα (x) rearrangement, as well as rearrangements to Jα gene segments within up to ∼5 kb upstream of the targeted Jα. DNA samples were also amplified with Cα primers as a positive control. PCR primers used in this study have been described previously (Mancini et al., 2001; Pasqual et al., 2002; Hager et al., 2007). Amplifications were performed with Expand High-Fidelity Polymerase (Roche Applied Science) or Platinum Taq DNA Polymerase High Fidelity (Invitrogen) kits. Thermocycling conditions used were as previously described (Hager et al., 2007), except that the extension temperature was lowered to 68°C when the enzyme and kit were used (Invitrogen). PCR products were resolved on 1.4% agarose gels and blotted onto positively charged nylon membranes under alkaline conditions. A Vα14-specific probe was generated by PCR amplification of plasmid DNA containing a Vα14i NKT TCR construct using the following primers: 5′-GGGAGAGAACTGCGTCCTTCAATGTA-3′ and 5′-AGATGTAGGTGGCAGTGTCATCCAG-3′. The resulting PCR product was gel purified and labeled with ³²P-dCTP by random oligonucleotide priming, according to the DNA labeling systems kit and protocol (GE Healthcare). The resulting radiolabeled probe was used for Southern blot hybridization as previously described (Budowle and Baechtel, 1990).

For Vα14i NKT cell preparation, liver mononuclear cells were first incubated using biotinylated antibodies against CD19, CD45R, CD62L, TER-119, CD11b, CD24, F4/80, and Ly6G/Gr-1 (2 µg/ml each), and antibody-bound cells were removed using the EasySep system (Stem Cell Technologies). The remaining cells were then incubated with PE-CD1d tetramers loaded with αGalCer (1:100 dilution) together with 1 µg/ml of free streptavidin, and tetramer-binding cells were isolated using anti-PE microbeads and LS columns (Miltenyi Biotec). The cells were kept at 4°C after tetramer incubation until final isolation, at which point they were stored at −80°C. Purity as determined by flow cytometry was routinely 70% or greater. RNA was prepared using the RNeasy kit and protocol (QIAGEN). cDNA was prepared using the iScript kit and protocol (Bio-Rad Laboratories), and PCR reactions were performed using the Light SYBR 480 Master Mix (Roche). Cytokine gene transcript levels were normalized either to L32 transcript levels and the percentage of Vα14i NKT cells in each cell preparation or to Vα14-Jα18 transcript levels. Similar results were observed using either method of normalization. PCR primer sequences were as follows: Il4 forward, 5′-ACAGGAGAAGGGACGCCAT-3′; Il4 reverse, 5′-GAAGCCCTACAGACGAGCTCA-3′; Ifng forward, 5′-GGATGCATTCATGAGTATTGC-3′; Ifng reverse, 5′-CCTTTTCCGCTTCCTGAGG-3′; L32 forward, 5′-GAAACTGGCGGAAACCCA-3′; L32 reverse, 5′-GGATCTGGCCCTTGAACCTT-3′; Vα14 forward, 5′-CTAAGCACAGCACGCTGCACA-3′; and Jα18 reverse, 5′-CAGGTATGACAATCAGCTGAGTCC-3′.

All calculations of statistical significance were determined using a paired Student's t test. When multiple samples of the same genotype were determined in the same experiment, the arithmetic mean of those samples was used for the paired t test.

Fig. S1 shows that CD8 expression has no effect on the avidity of Vα14i NKT hybridomas for CD1d tetramers. Fig. S2 shows CD8 and CD4 expression levels by Vα14i NKT cells and conventional T lymphocytes from the thymus and liver of CD8 Tg/+ and WT littermates. Fig. S3 shows CD1d tetramer binding by Vα14i NKT thymocytes from CD8 Tg and nontransgenic littermates. Fig. S4 depicts the effect of the CD8 transgene on Vα14i NKT cell Vβ usage. Fig. S5 shows that the CD8 transgene acts to lower CD1d expression in DP thymocytes. Fig. S6 shows that CD8 Tg thymocytes have reduced survival in vitro and an altered phenotype. Fig. S7 shows that the reduced response of hd/hd Vα14i NKT cells to αGalCer is not restricted to the CD8-expressing subset.

We thank Archana Khurana for preparation of CD1d tetramers, Laurent Gapin for advice on rearrangement analysis, and Jr-Wen Shui for help with Southern blotting and hybridization.

This work was supported by National Institutes of Health grants R37 AI71922 (M. Kronenberg) and R01 AI42915 (D. Kappes), and a NHMRC Australia CJ Martin Fellowship #237029 (K. Hammond).

The authors have no conflicting financial interests.