CD27 is a thymic determinant of the balance between interferon-γ- and interleukin 17–producing γδ T cell subsets

Abstract

The production of cytokines such as interferon-γ and interleukin 17 by αβ and γδ T cells influences the outcome of immune responses. Here we show that most γδ T lymphocytes expressed the tumor necrosis factor receptor family member CD27 and secreted interferon-γ, whereas interleukin 17 production was restricted to CD27- γδ T cells. In contrast to the apparent plasticity of αβ T cells, the cytokine profiles of these distinct γδ T cell subsets were essentially stable, even during infection. These phenotypes were established during thymic development, when CD27 functions as a regulator of the differentiation of γδ T cells at least in part by inducing expression of the lymphotoxin-β receptor and genes associated with trans-conditioning and interferon-γ production. Thus, the cytokine profiles of peripheral γδ T cells are predetermined mainly by a mechanism involving CD27.

The past three decades have witnessed intense investigation of T cell generation in the mouse thymus. As a result, insight has been gained into how bone marrow–derived progenitors seed the thymus, rearrange their T cell antigen receptor (TCR) loci, become committed to either the αβ or the γδ T cell lineage and proceed through TCR- mediated selection processes that ensure that ‘useful’ T cells are exported to exert immune functions in the periphery1–3. This notwithstanding, consensus has not been reached on the degree to which intrathymic differentiation events influence peripheral T cell function. For example, although it seems clear that events in the thymus promote the generation of Foxp3+ ‘natural’ regulatory T cells, there is debate over which molecular interactions are involved, with the TCR, CD28 and the lymphotoxin-β receptor (LTβR) all being linked to this¹. There is likewise no clear sense of the relative influence of thymic determination versus peripheral ‘plasticity’ and functional ‘conversion’. This is true not only for regulatory T cells but also for the differentiation of helper T cells leading to the production of important cytokines, including interferon-γ (IFN-γ), interleukin 4 (IL-4) and IL-17. In this context, it may also be argued that there has been a disproportionate focus on αβ T cells, when it is increasingly acknowledged that γδ T cells have pleiotropic functional potential and, in arenas such as the production of IFN-γ and IL-17, make critical contributions to host immune competence^4,5. Indeed, whereas TCRαβ+ IL-17-producing T helper cells (T_H-17 cells) are restricted mainly to the gut⁶, γδ T cells are a chief source of IL-17 in lymphoid organs and peripheral tissues^7–10. Consequently, there are ongoing attempts to harness the functional potential of γδ T cells in the clinic. Functional T cell subsets are apparently generated by the combined actions of transcription factors. Thus, the transcription factors T-bet and eomesodermin, GATA-3 and c-Maf, and RORγt and Runx1 determine T cell production of IFN-γ, IL-4 and IL-17, respectively^11,12. However, their actions can be critically modulated by other molecules, including Notch receptors¹³. Indeed, there is much biological and clinical interest in the identification of cell surface receptors whose engagement might skew T cell differentiation. In this context, it has been reported that the differentiation of γδ T cells toward IFN-γ production and away from IL-17 production is regulated by specific γδ TCR agonists in the thymus¹⁴. Possibly consistent with those findings, particular γδ TCRs are associated with distinct cellular functions¹⁵. Those observations notwithstanding, the differentiation of γδ T cells is also regulated by other intrathymic molecular interactions. For example, the differentiation of proliferating, IFN-γ- producing γδ T cells requires TCR-independent interactions between early γδ T cell progenitors and CD4⁺CD8⁺ (double-positive (DP)) αβ T cell progenitors. These events, collectively called ‘trans-conditioning’, are determined in part by signals received through the LTβR^16–18.

On the basis of such perspectives, this paper defines discrete intrathymic progenitors of peripheral IFN-γ- and IL-17-producing γδ T cell subsets, distinguished and regulated by the tumor necrosis factor (TNF) receptor ‘superfamily’ member CD27 (A000546), which engages CD70 (A000547), a TNF-related transmembrane glycoprotein¹⁹. CD27⁺ γδ thymocytes expressed LTβR and genes associated with a T helper type 1 (T_H1) phenotype, in contrast to CD27^– _{γδ thym}ocytes, which included the progenitors of IL-17-producing γδ cells. Whereas the relative proportions of the two γδ subsets in the thymus and the periphery were biased toward CD27⁺ IFN-γ producers, IL-17-producing CD27^– _{γδ T cell} populations rapidly expanded in response to acute infection such as that posed by malaria.

RESULTS

CD27 expression segregates γδ lymphocyte subsets

Attempts have been made to identify markers of distinct γδ T cell subsets, particularly after the realization that γδ cells are key sources of IL-17 and IFN-γ and contribute to host defense and to certain inflammatory and autoimmune diseases^14,20–22. Although γδ cells have been subcategorized by their expression of CD25 and CD122 into IL-17 and IFN-γ producers, respectively²³, as discussed below, this classification is not generally applicable beyond the peritoneal cavity, and there is no evidence that these markers actively regulate the outcome of T cell differentiation. Here we examine the validity of CD27 as a subset marker for γδ T cells, as CD27 has been linked to the functional segregation of subsets of αβ T cells^19,24 and natural killer (NK) cells²⁵.

CD27 expression distinguished two γδ cell subsets in the spleen and lymph nodes (axillary, inguinal, branchial and mesenteric) and in the peripheral tissues (lung or gut) of adult C57BL/6 mice. Overall, 10–30% of γδ cells were CD27^– (called ‘γδ^27–’ here) and 70– 90% were CD27⁺ (called ‘γδ^27+’ here; Fig. 1a). The two subsets had very different surface phenotypes, with γδ27- cells generally being CD44hiCD62Llo _and γδ27+ cells having lower expression of CD44 and higher expression of CD62L (Fig. 1b). They also showed conspicuously little interconversion, as the CD27⁺ and CD27^– phenotypes were maintained even after 3–5 d of continuous stimulation with antibody to CD3ε (anti-CD3ε) in vitro or 2 weeks of residence in lymphopenic mice deficient in the ^{product of recom} bination-activating gene 2 (RAG-2; Fig. 1c).

Figure 1.

CD27 expression defines distinct subsets of peripheral γδ T cells. (a) Flow cytometry of cells obtained from various tissues (above plots) of adult wild- type C57BL/6 mice, stained for CD3ε, CD27 and TCRγδ and gated on CD3ε⁺TCRγδ⁺ cells. a+i(LN), axillary and inguinal lymph nodes; mLN, mesenteric lymph nodes. Data are representative of three to eight independent experiments with three to five mice each. (b) Flow cytometry of splenocytes stained for CD3ε, CD27 and TCRγδ, plus CD62L or CD44 and gated on CD3ε⁺TCRγδ⁺CD27⁺ cells (γδ²⁷⁺) and CD3ε⁺TCRγδ⁺CD27⁻cells (γδ²⁷⁻). Data are representative of three independent experiments with three to five mice each. (c) Flow cytometry of γδ²⁷⁺ or γδ²⁷⁻ T cells sorted from pooled spleens and lymph nodes and cultured separately for 3 d with antigen-presenting cells and anti-CD3ε (1 mg/ml; left) or injected separately intravenously into RAG-2- deficient mice and analyzed after 2 weeks (right), then recovered, stained for CD3ε, CD27 and TCRδ and gated on CD3ε⁺TCRδ⁺ cells. Data are representative of three independent experiments. (d–f) Flow cytometry analyzing the proliferation of γδ²⁷⁺ or γδ²⁷⁻ T cells sorted from pooled spleens and lymph nodes, labeled with CFSE and cultured separately with antigen- presenting cells in presence of various doses of anti-CD3e (horizontal axis; d) or injected intravenously into RAG-2-deficient mice separately (e) or together at ratio of 1:1 (f), and assessed on the basis of CFSE-dilution kinetics at day 3. Data are representative of four independent experiments.

These experiments also showed that the proliferation of γδ²⁷⁺ cells after TCR stimulation in vitro exceeded that of γδ^27– cells (Fig. 1d) and that γδ²⁷⁺ cells had a considerable advantage over γδ^27– lymphocytes when these cells were injected together into lymphopenic mice (Figs. 1e,f). Thus, γδ²⁷⁺ and γδ^27– cells constitute distinct and essentially stable subsets of γδ T lymphocytes in naive mice.

Cytokine production segregates with CD27

When comparing the gene expression of splenic γδ²⁷⁺ and γδ^27– cells from naive C57BL/6 mice, we noted considerable polarization toward constitutive expression of Ifnγ and II17, respectively (Fig. 2a). After 4 h of activation in vitro with phorbol 12-myristate 13-acetate (PMA) and ionomycin, γδ²⁷⁺ cells had typical T_H1-like characteristics, including abundant intracellular IFN-γ and TNF but no IL-17; in contrast, γδ^27– populations had far fewer cells that expressed IFN-γ and TNF but had a conspicuous population of IL-17-producing cells (Fig. 2b and Supplementary Fig. 1a online). Neither γδ²⁷⁺ nor γδ^27– cells had substantial expression of IL-4, IL-13 or IL-10 in the conditions used, but they had similar cytolytic capacity (Supplementary Fig. 1a,b and data not shown). Independent reports have cited an association of CD122 expression with IFN-γ-producing γδ T cells^14,23 and of the scavenger receptor SCART2 with IL-17-secreting γδ T cells²⁶. Consistent with those findings, expression of CD122 and SCART2 segregated with γδ²⁷⁺ and γδ^27– cells, respectively (Fig. 2c,d). However, the subsets did not overlap perfectly. Thus, many γδ²⁷⁺ cells were CD122^–, and a substantial percentage of CD122^– cells also produced IFN-γ after short-term activation (Supplementary Fig. 1c). Indeed, as most γδ T cells are CD122^–, over 50% of IFN-γ-producing γδ T cells may have been CD122^– whereas less than 5% were CD27^– (Fig. 2b,d) Hence, CD27 expression seemed to segregate much more accurately with IFN-γ production by γδ T cells than did CD122. The degree to which SCART2-expressing cells overlap with γδ^27– cells cannot be assessed until reagents facilitating the detection of SCART2 by flow cytometry become generally available.

Figure 2.

CD27 expression segregates IFN-γ- versus IL-17-producing γδ cells in naive and malaria-infected mice. (a) Quantitative real-time PCR of the expression of Ifnγ and II17 by γδ²⁷⁺ and γδ²⁷⁻ T cells sorted from pooled spleens and lymph nodes of naive C57BL/6 mice, normalized to Efa1 expression and presented as the percent of maximum expression of the gene of interest. (b) Intracellular staining for IFN-γ or IL-17 in γδ²⁷⁺ and γδ²⁷⁻ T cells sorted from pooled spleens and lymph nodes of naive or Plasmodium berghei ANKA–infected C57BL/6 mice (at day 4 after infection), then stimulated for 4 h with PMA and ionomycin. FSC, forward scatter. Numbers adjacent to outlined areas indicate percent IFN-γ⁺ cells (top row) or IL-17⁺ cells (bottom row). (c) Quantitative real-time PCR of the expression of II2rb (encoding CD122) and the gene encoding SCART2 (National Center for Biotechnology Information identification number 5830411N06Rik; called ‘Scart2’ here) by γδ²⁷⁺ and γδ²⁷⁻ T cells, as described in a. (d) Flow cytometry of splenocytes from naive mice, stained for CD3ε, CD27, TCRγδ and CD122 and gated on CD3ε⁺TCRγδ⁺ cells. Numbers in quadrants indicate percent cells in each. (e) Absolute number of total γδ T cells (left) and percent CD69⁺ cells (middle) and CD44^hiCD62L^lo cells (right) among γδ²⁷⁺ and γδ²⁷⁻ T cells from splenocytes obtained from naive mice (0) and mice at days 3 and 6 after infection with P. berghei ANKA and stained for CD3ε, CD27, TCRγδ and CD69, plus CD62L or CD44. (f) Absolute number of total γδ cells (left), γδ²⁷⁺ T cells (middle) and γδ²⁷⁻ T cells (right) among splenocytes obtained from naive or naive mice and mice at day 4 after infection with ‘titrated’ doses of P. berghei ANKA (5 ×10⁴, 5 ×10⁵ or 5 ×10⁶ parasites per mouse; to achieve variable parasitemia) and stained for CD3ε, CD27 and TCRγδ, presented relative to parasitemia (percent P. berghei ANKA in blood). Each symbol represents an individual mouse (n=4 mice/group). Data are representative of three independent experiments (a,c,e; error bars, s.d.; n=3 samples of four to five mice, pooled) or four to five (b), three (d) or two (f) independent experiments.

To analyze how each γδ subset might respond to infection, we used Plasmodium berghei, because in mice and humans, γδ T cells make important contributions to the response to this pathogen and other apicomplexan parasites^27–30. We evaluated the spleen and peripheral lymphoid organs of C57BL/6 mice infected with P. berghei (strain ANKA), a model of experimental cerebral malaria that leads to death between 6 and 8 d after infection³¹. The γδ²⁷⁺ and γδ^27– subsets each proliferated after inoculation with a fixed number of P. berghei ANKA– infected erythrocytes, but with different kinetics. The γδ²⁷⁺ populations expanded more rapidly (by day 3 after infection); however, by day 6 after infection, γδ^27– cells had also increased almost fivefold in number and accounted for approximately 40% of cells in the splenic γδ compartment (Fig. 2e). These findings show that whereas γδ^27– cells proliferated relatively poorly in response to stimulation in vitro or during homeostatic expansion in vivo (Fig. 1), they were able to proliferate robustly in response to physiological stimuli in vivo. Moreover, as early as day 3 after infection, approximately 60% of γδ^27– cells had an activated CD69+ surface phenotype, compared with only about 20% of γδ²⁷⁺ cells (Fig. 2e). A higher inoculum of P. berghei ANKA–infected erythrocytes resulted in augmented parasitemia (Supplementary Fig. 2a online) and population expansion of spleen and lymph node γδ T cells that was almost exclusively attributable to γδ^27– cells, which at high parasitemia represented over 50% of all γδ T cells (Fig. 2f and Supplementary Fig. 2b). Notably, however, both γδ T cell subsets showed strong conservation of the cytokine (IFN-γ versus IL-17) and surface marker (CD44 and CD62L) phenotypes of naive mice (Fig. 2b,e).

Factors that influence γδ^– cell IL-17 production

By analyzing cytokine production versus dilution of the cytosolic dye CFSE by total γδ T cells, we confirmed that most IL-17 producers, like γδ^27– cells, proliferated poorly after activation in vitro, whereas IFN-γ producers, like γδ²⁷⁺ cells, proliferated robustly (Supplementary Fig. 3a,b online). Similarly, IL-17 was produced by CD44hiCD62L^lo γδ T cells but not by CD44loCD62L^hi γδ T cells (Supplementary Fig. 3c,d). Collectively, these data show that the properties of the γδ^27– subset fully resemble those of the IL-17-producing subset of the peripheral γδ T cell pool. Therefore, we considered it appropriate to use γδ^27– cells to study the requirements for IL-17 expression. Consistent with reports that IL-17 production by γδ T cells does not require stringent conditioning by transforming growth factor-β (TGF-β), IL-6 or other inflammatory cytokines^7,23 required for αβ T_H-17 differentiation^32–34, γδ^27– cells expressed IL-17 mRNA immediately after isolation and required only 4 h of treatment with PMA and ionomycin to express IL-17 protein (Fig. 3a,b). Neither TCRαβ⁺CD4⁺CD^27– effector T cells nor γδ²⁷⁺ cells produced IL-17 mRNA or protein (Fig. 3a,b). Given that elaboration of IL-17 by γδ^27– cells occurs in different conditions than those established for αβ T_H-17 cells, we also determined if the T_H-17-determining transcription factor RORγt was expressed in γδ^27– cells. For this we used a mouse in which green fluorescent protein (GFP) is expressed under the control of Rorc (which encodes RORγt transcriptional elements)^6,35. By the criterion of GFP expression, all IL-17-producing γδ^27– cells expressed Rorc (Fig. 3c). A small percentage of γδ²⁷⁺ cells weakly expressed Rorc but did not express IL-17, consistent with data above (Figs. 2b and 3b) and with the generally accepted view that RORγt is necessary but not sufficient for IL-17 induction in αβ T cells¹¹. Nonetheless, we investigated whether the constitutive expression of IL-17 in γδ^27– cells could be further upregulated by particular stimuli. Indeed, II17 transcription and IL-17 protein production was augmented in 3-day cultures by activation with anti-CD3e plus IL-2 (in the absence of any other cytokines); this augmentation was further enhanced by TGF-β treatment (Fig. 3a,b). Of note, these stimuli also induced IFN-γ production in a substantial fraction of γδ^27– cells (although this fraction was always lower than that of the γδ²⁷⁺ cell subset; Fig. 3b), which demonstrates some degree of functional plasticity of this population, at least in vitro. That observation notwithstanding, IL-17 expression was achieved without the addition of exogenous IL-6 or IL-21.

Figure 3.

Constitutive expression of IL-17 and RORγt by γδ²⁷⁻ cells. (a) Quantitative real-time PCR analysis of II17 expression by CD4⁺CD25⁻ αβ T cells (4⁺25⁻) and γδ²⁷⁺ or γδ²⁷⁻ T cells sorted from pooled spleens and lymph nodes of C57BL/6 mice and assessed without activation (Fresh) or after activation for 4 d with plate-bound anti-CD3ε plus IL-2, with (+) or without (−) TGF-b, presented as described in Figure 2a. Data are representative of three independent experiments (error bars, s.d.; n=3 mice). (b) Intracellular staining of IFN-γ and IL-17 in CD27⁺ or CD27⁻ γδ and CD4⁺CD25⁻αβ⁻ T cells sorted from pooled spleens and lymph nodes from C57BL/6 mice, assessed without preactivation (Fresh) or after preactivation for 4 d on plate-bound anti- CD3ε and IL-2 (PreAct) with or without TGF-β, followed by stimulation with PMA and ionomycin. Data are representative of four independent experiments. (c) Intracellular staining for IL-17 and GFP in CD4⁺CD25⁻, γδ²⁷⁺ and γδ²⁷⁻ T cells isolated from pooled lymph nodes of RORγt-GFP mice. Data are representative of triplicate experiments with five mice each. (d) Intracellular staining for IFN-γ and IL-17 in flow cytometry–sorted γδ²⁷⁻ and γδ²⁷⁺ splenocytes activated in vitro for 3 d with anti-CD3ε and IL-2 in the presence (+) of various cytokines (above plots). Data are representative of three independent experiments. Numbers in quadrants (b–d) indicate percent cells in each. (e) Staining for surface V_γ1 or V_γ4 on and intracellular IL-17 in γδ²⁷⁺ and γδ²⁷⁻ cells sorted from pooled spleens and lymph nodes of C57BL/6 mice, preactivated for 4 d on plate-bound anti-CD3ε plus IL-2 and TGF-β, then stimulated with PMA and ionomycin. Data are representative of three independent experiments.

In addition, the finding that IL-17- producing γδ^27– cells lacked transcripts for IL-6 or IL-21 (Supplementary Fig. 4 online) excluded the possibility that these cytokines might be provided in an autocrine way. Likewise, in contrast to the need for antibody blockade of IFN-γ for the optimal polarization of CD4⁺CD25^27– αβ T_H-17 cells in vitro^7,32 (data not shown), γδ^27– cells produced IL-17 even in the presence of IFN-γ-producing cells (Fig. 3b). Moreover, when exogenous IL-6, IL-21 and neutralizing anti-IFN-γ were added to cultures alone or in combination, they had essentially no effect on IL-17 production by TGF-β-treated cells; the addition of IL-23 had only a very small effect (Fig. 3d). Notably, in all conditions tested, most γδ²⁷⁺ cells failed to produce IL-17 (Fig. 3b,d), which reemphasizes the functional distinction between the two γδ T cell subsets. Although it has been proposed that IL-17 production is more common among cells positive for the variable region V_γ4 than among V_γ1⁺ cells, conflicting data have left this point contentious and unresolved^21,22. In our analyses, IL-17-expressing cells accounted for almost 50% of V_γ4⁺ γδ^27– cells and only about 11% of V_γ1⁺ γδ^27– cells (Fig. 3e). This finding supports the hypothesis that there is an association between the TCR and a bias in the functional differentiation of γδ T cells^4,14.

Thymic progenitors of γδ²⁷⁺ and γδ^27– cells

Given their essential stability in naive and infected mice, we assessed whether the γδ²⁷⁺ and γδ^27– subsets were developmentally established in the thymus. We found that TCRγδ⁺ thymocytes had the highest expression of CD27 of all the T cell–committed subsets (Fig. 4a). However, in contrast to the homogeneous expression of CD27 on mature TCRαβ⁺ thymocytes, approximately 10% of γδ thymocytes conspicuously lacked CD27 (Fig. 4a). As thymocyte precursors at double-negative stages 2 and 3, from which γδ cells mainly derive^1,36,37, express CD25, we used staining for CD25 and CD27 to segregate γδ thymocytes into subsets; 5–10% were CD27⁺CD25⁺ (called ‘γδ^27,25+’ here), whereas the remaining 90–95% cells were negative for CD25, including 10–15% CD27– cells (γδ^27–) and 75–80% CD27⁺ cells (γδ²⁷⁺; Fig. 4b). This representation of the subsets was completely similar in independent mouse colonies maintained in two different countries. Moreover, each subset was detectable in the thymus from the earliest embryonic stages tested (embryonic days 14–15 (E14–E15), although the γδ²⁷⁺ subset greatly expanded between E15 and E18, by which time the relative proportions found in the adult were established (Fig. 4c). The finding that peripheral γδ cells were essentially all CD^25– (Fig. 4b) underpins the concern that the proposed use of CD25 to identify IL-17-producing γδ T cells may be applicable only to the peritoneal cavity²¹ (Supplementary Fig. 5 online) and suggests that the thymic γδ^27,25+ subset might constitute a progenitor compartment. Consistent with that hypothesis, γδ^27,25+ cells were larger in size than were γδ²⁷⁺ or γδ^27– thymocytes (Fig. 4d), had a CD24^hiCD25^hiCD62L^hiCD44^hiCD2^loCD5^lo ‘immature’ phenotype very similar to that of progenitors at double-negative stages 2 and 3 (Fig. 4e and Supplementary Fig. 6 online), had lower surface expression of TCR than did the γδ²⁷⁺ and γδ^27– subsets (data not shown) and had high proliferative turnover (Fig. 4f,g). In contrast, the surface phenotype and proliferative activity of γδ²⁷⁺ and γδ^27– thymocytes closely mimicked those of their respective peripheral counterparts (Figs. 1b and 4e–g). Next we repopulated fetal thymic reaggregation organ cultures with highly purified γδ^27,25+ thymocytes. After 7 d in culture, these γδ^27,25+ thymocytes gave rise to both γδ²⁷⁺ and γδ^27– cells (Fig. 4h). This observation confirms that γδ^27,25+ thymocytes have the potential to generate γδ²⁷⁺ and γδ^27– cells. Another demonstration of the relationship between the thymic γδ²⁷⁺ and γδ^27– subsets and their peripheral counterparts was evident in their constitutive expression of Ifng and II17, respectively (Fig. 4i). Because this expression was present in E18 thymi, it seems implausible to attribute it to the recirculation of activated peripheral γδ T cells back to the thymus. Likewise, γδ^27– cells from E18 thymi already expressed Rorc and its T_H-17-driving cofactor Runx1 (ref. 11), whereas fetal γδ²⁷⁺ cells expressed Tbx21, which encodes T-bet, a chief determinant of T_H1 differentiation (Fig. 4j). The idea that such functional segregation was already established in γδ²⁷⁺ and γδ^27– thymocytes, respectively, was also demonstrated by their corresponding expression of IFN-g and IL-17 protein (Supplementary Fig. 7a online). Incubation for 3 d with anti-CD3e and IL-2 supplemented with a ‘cocktail’ of IL-6, IL-21 and IL-23 further increased IL-17 production by γδ^27– cells but did not induce IL-17 production by γδ²⁷⁺ thymocytes (Supplementary Fig. 7b). Thus, the functional potential and constraints of γδ²⁷⁺ and γδ^27– cells seem to be ‘imprinted’ as early as in the fetal thymus.

Figure 4.

Both γδ²⁷⁺ and γδ^27– cells originate from common CD27⁺CD25⁺ thymic γδ progenitors. (a) Flow cytometry of thymocytes of C57BL/6 mice (n=3) gated on single-positive stage 4 cells (SP4; CD4⁺CD8^–) or γδ cells (TCRγδ⁺CD3ε). Dashed line, isotype control. (b) Flow cytometry of thymocytes (left) and splenocytes (right) from C57BL/6 mice, stained for CD3ε, CD27, CD25 and TCRγδ, gated on CD3ε⁺TCRγδ⁺ cells. Numbers in plots indicate percent cells in each area. (c) Proportion of the subsets defined n b among total γδ thymocytes from embryonic (E14–E18), newborn (days 1–9 (D0–D9)) or adult (6-week-old (W6)) C57BL/6 mice. (d) Flow cytometry of the cell size of the subsets defined in b. (e) Staining of thymocytes from newborn mice (day 2) with antibodies specific for various markers (below plots), gated on the subsets defined in b. (f) Incorporation of the thymidine analog BrdU (5-bromodeoxyuridine) by gated subsets (defined in b) of thymocytes obtained from C57BL/6 mice given BrdU daily intraperitoneally and then stained for CD27, CD25, TCRγδ and BrdU. (g) Intracellular staining of thymocyte subsets (defined in b) with 7- amino- actinomycin D (7-AAD), an indicator of DNA content. 27,25+, γδ^27,25+; 27+, γδ²⁷⁺; 27–, γδ^27–. (h) Flow cytometry of γδ^25,27+ thymocytes after sorting (Post-sort; left) and after the sorted cells were mixed with thymic stromal cell samples from E15 C57BL/6 mice, depleted of thymocytes and incubated for 7 d (RTOC; right). Numbers adjacent to or in outlined areas indicate percent cells in each. (i) Quantitative RT-PCR analysis of the expression of Ifnγ and II17 in various cell subsets, presented as described in Figure 2a. (j) Semiquantitative RT-PCR analysis of the expression of Rorc, Runx1 and Tbx21 in various cell subsets, normalized to Actb expression (encoding b-actin). For Rorc and Runx1, cells were sorted from pooled organs of over four mice; for Tbx21, γδ thymocyte subsets were purified from pooled FTOCs at day 6 (n>5 cultures). 4⁺8⁺, CD4⁺CD8⁺. Data are representative of three (a,d,e,h,j) or over ten (b) independent experiments or three (c,g,i) or two (f) independent experiments with three mice (f,g,i) or over three mice (c; error bars (c,f,g,i), s.d.).

CD27 regulates γδ T cell functional potential

The number of γδ T cells, TCR V_γ segment use and profiles of developmental markers (such as CD24, CD25, CD44 and CD62L) were similar in the thymus and peripheral lymphoid organs of wild- type and Cd27^–/– mice³⁸ (Fig. 5a, Supplementary Fig. 8a,b online and data not shown). However, TCRγδ⁺ thymocytes from Cd27^–/– mice had lower expression of Ifnγ (Supplementary Fig. 8c), and peripheral γδ cells from Cd27^–/– mice had lower expression of IFN-γ and TNF mRNA and protein (Fig. 5b and Supplementary Fig. 8d). The most overt difference was that the proportion of IFN-γ^hi cells in Cd27^–/– mice was approximately 65% lower (Fig. 5b). Likewise, CD44^loCD62L^hi γδ T cells from Cd27^–/– mice (which in wild-type mice would also express CD27) had less IFN-γ production than their wild-type counterparts had (Supplementary Fig. 8e). Thus, CD27 is a regulator as well as a marker of the functional differentiation of γδ T cells. IL-17 expression was essentially unchanged in Cd27^–/– mice and, as in wild-type mice, IL-17 expression was limited to the

Figure 5.

CD27 controls the functional potential of γδ T cells. (a) Staining of TCRγδ and CD3 surface proteins on wild-type (WT) C57BL/6 and CD27-deficient (Cd27^−/−) thymocytes (left), and absolute number of TCRγδ⁺CD3⁺ thymocytes (right). Data are representative of five independent experiments (error bars, s.d.; n=3 mice). (b) Intracellular staining of various cytokines (left margin) in sorted total γδ T cells obtained from pooled spleens and lymph nodes of wild-type and CD27-deficient mice and stimulated with PMA and ionomycin. Right, percent of total γδ T cells. NS, not significant. *, P<0.05 (Student's t-test). Data are representative of four independent experiments with four to five mice each. (c,d) Absolute number of γδ T cells (c) or IFN-γ⁺ or IFN-γ^hi γδ T cells (d) among spleen cells of naive or P berghei ANKA–infected wild-type or Cd27^−/− mice. D4PI, day 4 after infection. *, P<0.05 (Student's t-test). Data are representative of two independent experiments (error bars, s.d.; n=5 mice).

CD44^hiCD62L^lo subset in Cd27^–/– mice (Supplementary Fig. 8f)

Infection with P. berghei ANKA elicited a similar expansion of total γδ splenocyte populations in wild-type and Cd27^–/– mice (Fig. 5c), but the latter had 60% lower absolute numbers of IFN-γ producers and 80% fewer IFN-g^hi cells than did wild-type mice (Fig. 5d). Notably, there was a delay in the progression of cerebral malaria, a severe inflammatory syndrome partly attributable to IFN-γ²⁸, in Cd27^–/– mice (Supplementary Fig. 9 online).

To investigate possible mechanisms by which CD27 alters the thymic and peripheral phenotypes of γδ T cells, we considered the selective impairment of IFN-γ production in γδ cells that occurs in TCRβ-deficient mice. Like CD27-deficient mice, TCRβ-deficient mice show no overt defect in γδ T cell numbers. In TCRβ-deficient mice, γδ T cell functional alterations have been attributed mainly to the absence in the thymus of trans-conditioning of γδ cell progenitors by TCRβ⁺ DP thymocytes^1,17. Both γδ T subsets defined here were present in wild-type proportions in TCRβ-deficient mice (data not shown), but γδ²⁷⁺ cells in TCRβ-deficient mice showed impaired expression of IFN-γ and TNF mRNA (Fig. 6a). We did not find this impairment in γδ²⁷⁺ cells from TCRα-deficient mice, which had wild-type numbers of DP thymocytes; this suggests that the defective γδ T cell function in CD27-deficient mice may at least in part reflect altered trans-conditioning.

Figure 6.

CD27 signals regulate the differentiation of γδ thymocytes. (a) Quantitative RT-PCR analysis of the expression of Ifng and Tnfa in γδ²⁷⁺ cells sorted from TCRα-deficient mice (a) and TCRβ-deficient mice (b), presented as described in Figure 2a. Data are representative of two independent experiments (error bars, s.d.; n=3 mice). (b) Gene expression and fluorescence intensity of microarray analysis of wild-type and CD27- deficient (KO) γδ T cells. Black, genes with differences in expression of over twofold (WT/KO); gray, all other genes analyzed. P<0.00003 (adjusted) for genes selected as having differences in expression. Data are representative of two independent experiments. (c) Semiquantitative RT-PCR analysis of the expression of various genes (left margin) in total γδ thymocytes from wild-type or CD27-deficient mice. Data are representative of two independent experiments with triplicate samples of cells purified from four to six mice. (d) Quantitative RT-PCR analysis of the expression of Ltbr in total γδ thymocytes sorted from wild-type and CD27-deficient mice, presented as described in Figure 2a. *, P<0.001 (Student's t-test). Data are representative of two independent experiments (n=3 mice). (e) Semiquantitative RT-PCR analysis of the expression of various genes (left margin) in γδ subsets from wild-type thymi. Data are representative of two independent experiments with triplicate samples of cells purified from four to six mice. (f) Quantitative RT-PCR analysis of the expression of Ifng and II17 in γδ T cell subsets sorted at day 7 from FTOCs treated with a soluble recombinant fusion protein of CD70 and immunoglobulin (10 mg/ml; sCD70 +) or control human immunoglobulin G (sCD70 -), presented as described in Figure 2a. Data are representative of two independent experiments (error bars, s.d.; n=3 mice). *, P<0.05 (Student's t-test). (g) Flow cytometry of FTOCs at day 7, treated with medium alone, anti-CD3ε (0.67 mg/ml), anti–keyhole limpet hemocyanin (0.67 mg/ml; isotype control) or cyclosporin A (0.5 mM). Data are representative of two independent experiments with a pool of three E14 thymic lobes each.

To test that possibility, we compared the transcriptomes of wild-type and CD27-deficient γδ T cells with whole-genome microarrays (ArrayExpress accession number E-NCMF-23; Fig. 6b). Conspicuous among CD27-dependent genes was Ltbr, which encodes LTβR, a mediator of trans-conditioning¹⁷, as well as four ‘signature’ genes of LTβR signaling in γδ thymocytes: Crem (also called Icer), Nr4a2 (Nurr1), Rgs2 and Rgs1 (Table 1). We confirmed differences in the expression of these genes in wild-type and CD27-deficient γδ thymocytes by RT-PCR; in contrast, II7ra, which encodes a factor critical for γδ thymocyte development¹, was not impaired in CD27-deficient γδ thymocytes (Fig. 6c,d).

Table 1. Transcription in wild-type C57BL/6 and CD27-deficient γδ T lymphocytes.

Expression of genes linked to T cell biology, presented as expression in wild-type cells relative to expression in CD27-deficient γδ T cells (‘fold’ change). Gene ID, identification number in the National Center for Biotechnology Information database. Full cDNA microarray data, ArrayExpress accession number E-NCMF-23. P=0.00003 (threshold for microarray data analysis). Data are representative of two independent experiments.

Gene ID	Symbol	Description	Function	Expression
21940	Tnfrsf7	CD27 (TNF-RSF7)	Activation	21.11
17000	Ltbr	Lymphotoxin-β receptor (TNF-RSF3)	Cytokine receptor	9.00
12477	Ctla4	CTLA-4 coreceptor	Activation	7.01
12916	Crem	CREM/ICER	Transcription	5.10
18227	Nr4a2	Nuclear orphan receptor	Transcription	4.66
19735	Rgs2	Regulator of G protein signaling	Signaling	4.17
50778	Rgs1	Regulator of G protein signaling	Signaling	3.86
21943	Tnfsf11	RANK ligand (TNF-SF11)	Cytokine	3.66
12487	Cd28	CD28 coreceptor	Activation	3.36
19734	Rgs16	Regulator of G protein signaling	Signaling	3.25
16364	Irf4	Interferon-regulatory factor	Transcription	2.30
16478	Jund1	JunD	Transcription	2.20
16172	Il17ra	IL-17 receptor α-chain	Cytokine receptor	0.14

Consistent with the effects of CD27, wild-type mice showed expression of Ltbr and Ltbr-dependent genes in γδ^27,25+ and γδ²⁷⁺ thymocytes but not in γδ^27– thymocytes (Fig. 6e). To derive further evidence that CD27 regulates the differentiation of γδ thymo- cytes, we added to E14 fetal thymic organ cultures (FTOCs) a fusion protein of CD70 and immunoglobulin that acts as a CD27 agonist³⁹. After 1 week of culture, the recovered γδ²⁷⁺ thymocytes showed significant upregulation of Ifng and γδ^27– thymocytes showed down-regulation of II17 (Fig. 6f). Thus, embryonic γδ progenitors were effectively skewed toward IFN-γ production simply through the manipulation of CD70-CD27 signals. To assess whether CD27 might be acting together with other determinants of γδ differentiation, we added to parallel organ cultures either anti-CD3e or cyclosporin A; the former enhances and the latter inhibits TCR signaling. Relative to their representation in control cultures, these treatments increased and decreased representation of the γδ²⁷⁺ subset, respectively, with reciprocal effects on the γδ^27– subset (Fig. 6g). We conclude that the developmental preconditioning of discrete peripheral γδ cell subsets may involve some form of ‘collaboration’ between TCR and CD27 signaling in the thymus.

Finally, to determine whether the classification that we have introduced might apply to other unconventional T cells, we examined invariant NKT cells, as it has been reported that IL-17 expression is restricted to the NK1.1^– subset of NKT cells and that this restriction is ‘imprinted’ in the thymus^40,41. After isolating NKT cells on the basis of their reactivity to tetramers of CD1d and α-galactosylceramide, we found that surface marker staining ‘assigned’ over 80% of NK1.1⁻ cells to the CD27⁻ compartment, whereas over 80% of CD27⁺ cells were NK1.1⁺ (Supplementary Fig. 10 online). Thus, it seems that in the NKT compartment, IL-17-producing cells are also found mainly in the CD27⁻ subset.

DISCUSSION

Here we have described mouse γδ cell subsets that had distinct functional capacities and could be defined by CD27 expression. The γδ²⁷⁺ subset was more abundant and included cells that produced IFN-γ but essentially no IL-17 immediately after isolation and in all activation conditions used. In contrast, the γδ^27– subset comprised essentially all IL-17-producing γδ cells. Both subsets actively responded to infection in vivo by proliferating and producing cytokines. Although γδ^27– cells may have produced IFN-γ after in vitro activation, very few did so ex vivo, whether from naive or infected mice. Moreover, their thymic counterparts did not produce IFN-γ, which raises the possibility that the IFN-γ and IL-17 ‘double-producing’ cells were an artifact of in vitro activation protocols, even though ‘double production’ of IL-17 and IFN-γ has been described for TCRαβ⁺ T_H-17 cells^6,42. Whereas TCRαβ⁺ T_H1 and T_H-17 cells can take up to 7 d to differentiate and expand clonally under the influence of specific combinations of cytokines⁴³, the ability of the γδ T cell subsets described here to express their ‘signature’ cytokines directly ex vivo was apparently acquired during their differentiation in the thymus, when the subsets also adopted their characteristic phenotypic markers and other properties that remained essentially stable, even during infection. These observations define a parallel with NKT cells, another ‘innate-like’ lymphocyte lineage that promptly secretes a panoply of cytokines, including IL-17 and IFN-γ^40,44. Having found that NK1.1^– NKT cell populations, which are known to include ‘spontaneous’ IL-17-producing cells^40,41, were mostly CD27^–, we propose that CD27 may influence the functional differentiation of several lineages of unconventional T cells. The thymic ‘imprinting’ of the CD27⁺ and CD^27– subsets perhaps reflects the fact that γδ T cells (and other unconventional T cells) may need to provide specific effector cytokines very rapidly, to counter microbial growth and dissemination before maturation of the adaptive response. Indeed, myriad analyses have shown that γδ cells are crucial providers of either IFN-γ or IL-17 in diverse physiological settings^{8–10,14,20–22}. This may be particularly true in early life, when there is a precipitous encounter with environmental antigens and when γδ T cells seem precociously mature compared with αβ T cells in both mice³⁵ and humans (D. Gibbons and A.C.H., unpublished observations). Rapid γδ T cell responses may also determine in part the complexion and magnitude of ensuing adaptive responses^14,45. Developmental ‘imprinting’ may more strongly influence γδ T cells than αβ T cells, as defined sets of cytokines seem able to skew bulk populations of peripheral αβ T cells toward widely different functional differentiation pathways (such as T_H1, T_H-17 and regulatory T cells). In contrast, the various conditions we used here failed to skew peripheral γδ²⁷⁺ cells toward IL-17 production. Such functional ‘rigidity’ and our notable failure to identify any substantial interconversion of γδ²⁷⁺ and γδ^27– cells in vitro or in vivo suggest that these two subsets, although heterogeneous, constitute discrete lineages. Because it regulates γδ T cell functional differentiation, we believe the TNF receptor family member CD27 is a more incisive marker of γδ T cell subsets than others described^14,23,26. CD27 controls the expression of LTβR, another TNF receptor family member and a component of γδ T cell trans-conditioning that is associated with IFN- γ-biased differentiation and high in vitro proliferative potential¹⁷. Thus, CD27 may induce or sustain expression of LTβR and its ‘downstream’ targets in γδ²⁷⁺ thymocytes and in their γδ^27,25+ progenitors. This would be consistent with the finding that genes induced by trans-conditioning were expressed in early CD4^–CD8^– progenitors and their expression was then selectively retained or lost in more mature cells. The retention of the expression of such genes in γδ²⁷⁺ cells would confer their ‘signatory’ proliferative, T_H1-like func- tional potential, which is impaired in CD27-deficient mice and LTβR- deficient mice¹⁷. Because the lymphotoxin that trans-conditions through LTβR is derived at least in part from TCRβ⁺ DP thymocytes, this model also readily explains the functional impairment of the γδ²⁷⁺ subset in TCRβ-deficient mice but not in TCRα-deficient mice. It might also explain the notable increase in the size of the thymic γδ²⁷⁺ subset between E15 and E18, as this period corresponds to the main expansion of TCRβ⁺ DP thymocyte populations.

Clearly, as the impairment in IFN-γ production by γδ T cells in mice lacking CD27, LTβR or TCRβ is incomplete, other factors probably influence γδ T cell function. Signaling by the γδ TCR, which has been reported to skew γδ T cells toward IFN-γ production and away from IL-17 production¹⁴ may be one such factor. Our findings of TCR bias in the γδ²⁷⁺ and γδ^27– subsets and of the influence of TCR signal manipulation on the relative abundance of γδ²⁷⁺ and γδ^27– subsets in FTOC are consistent with that idea. TCR signals in the thymus may upregulate CD27 (ref. 37), which may then act together with TCR signaling (and/or other differentiation factors) by providing a critical competence to γδ T cells (and possibly other unconventional T cells) so they can progress through T_H1 differentiation. Although the transcription factor RORγt characterizes all IL-17-producing γδ^27– cells, its method of operating would seem different from that in TCRαβ⁺ T_H-17 cells as, consistent with other data⁷, we found that IL-17 production by γδ T cells required neither ‘collaborative’ signaling from TGF-β, IL-6 and IL-21 receptors nor interference with IFN production. It is likewise notable that microbe-induced IL-17 production in the gut seems to derive from αβ T cells⁴⁶, whereas IL-17-producing γδ T cells are relatively rare in the gut compared with their presence in other tissues. Indeed, the relative over-representation of TCRαβ⁺ T_H-17 cells in the mouse gut⁶ may represent a ‘compensation’ for the relative scarcity of gut γδ^27– cells. Another contrast in the IL-17 production of αβ and γδ T cells lies in the augmented γδ^27– _cell production of IL-17 in response to activation by TGF-β and TCR in the presence of IL-2, a combination that upregulates the transcription factor Foxp3 and a bias toward regulatory T cell function in CD4+ αβ T cells7,47. Indeed, it was reproducibly difficult to detect Foxp3 expression in γδ T cells in wild-type mice18 (data not shown). Possibly, a key feature of IL-17 production by γδ cells may be the ability of RORγt to act without interference from Foxp3. Although γδ^27– cells were mostly unaffected in CD27-deficient mice, as might be expected, CD27 gain of function in FTOC did result in lower IL-17 production by these cells, which indicates that CD27 may diminish IL-17 production as well as enhance IFN-g production, at least in T cell progenitors. CD27 is expressed on a plethora of lymphoid cells and bone marrow precursors, where it has been assigned many pleiotropic functions^38,39,48,49. Despite its expression on mouse and human thymocytes and its responsiveness to pre-TCR and TCR signals³⁷, CD27 seems to be not critical in the development of conventional αβ T cells³⁸ (unpublished data). That is notable given the constitutive expression of CD70 on thymic epithelial cells of mice and humans⁵⁰. Thus, here we have reported a previously unknown function for CD27 expression on thymic T cell precursors. Future studies should deter- mine which cells direct key CD70-CD27 interactions in the thymus and whether the functional and phenotypic stability of γδ²⁷⁺ cells requires sustained interactions between CD27 and CD70 in the periphery. The resolution of such issues will provide a better under- standing of the balance of precommitment and the peripheral plasticity of γδ T cell subsets.

METHODS

Mice

All mice used were adults 4–10 weeks of age. Embryos were obtained by the setting up of timed pregnancies. C57BL/6 (B6), B6.TCRα-deficient, B6.TCRβ-deficient and B6.RAG-2-deficient mice were from Jackson Laboratories. B6.CD27-deficient mice have been described³⁸. B6.RORγt-GFP mice35 were a gift from D. Littman. Mice were bred and maintained in the specific pathogen–free animal facilities of the Instituto de Medicina Molecular (Lisbon), Queen Mary London University (London) and Yale University School of Medicine (New Haven). All experiments involving animals were done in compliance with the relevant laws and institutional guidelines and were approved by local ethics committees.

Cell preparation

Thymuses, lymph nodes and spleens were homogenized and washed in RPMI medium containing 10% (vol/vol) FCS. In some cases, thymocyte samples were depleted of CD8⁺ and CD4⁺ cells by treatment with anti-CD8 (31.M) and anti-CD4 (RL172; both from J. van Meerwijk) plus complement (HD Supplies), followed by gradient centrifugation in Lympholyte M (Cederlane Laboratories). Lungs were cut into pieces 5 mm² and were incubated for 2 h at 37 1C with stirring in complete RPMI medium supplemented with collagenase (0.05 mg/ml; Sigma). Supernatants were collected and live cells were isolated on a gradient of Lympholyte M (Cederlane Laboratories).

Monoclonal antibodies

Fluorescein isothiocyanate–labeled anti-V_γ1 (2.11) was a gift from P. Pereira. The following monoclonal antibodies were from BD Pharmingen: fluorescein isothiocyanate–labeled anti-TCRδ (GL3), anti-V_γ4 (UC3-10A6) and anti-CD62L (MEL14); phycoerythrin-labeled anti-TCRδ (GL3), anti-CD25 (PC61), anti-CD44 (IM7), anti-IFN-γ (XMG1.2), and anti-TNF (MP6-XT22); peridinine chlorophyll protein complex–cyanine 5.5–labeled anti-CD4 (RM4-5) and anti-CD8α (53.6.7); allophycocyanin- labeled anti-CD25 (PC61); biotinylated anti-TCRδ (GL3); peridinine chloro- phyll protein complex–cyanine 5.5–conjugated streptavidin; and purified anti-CD3e (145.2C11). The following monoclonal antibodies were from eBiosciences: phycoerythrin-labeled anti-CD27 (LG.7F9), phycoerythrin- indotricarbocyanine– labeled anti-CD3e (145.2C11), fluorescein isothiocya- nate– or allophycocyanin-labeled anti-IL-17 (eBio17B7), and allophycocya- nin-labeled anti-CD27 (LG.7F9). Phycoerythrin-labeled mCD1d tetramer loaded with the glycolipid PBS57 was from the National Institutes of Health Tetramer Core Facility.

Flow cytometry and cell sorting. Information on these procedures is available in the Supplementary Methods online

Cell culture

For preactivation regimens, cells were incubated for 4 d on plate-bound anti-CD3e (10 mg/ml) with IL-2 (100 U/ml; Sigma), with or without TGF-β (5 ng/ml; R&D Systems). Where indicated, IL-6 (20 ng/ml), IL-21 (100 ng/ml), IL-23 (20 ng/ml) or neutralizing anti-IFN−γ (10 mg/ml) was also added to the medium. For analysis of proliferation, flow cytometry–sorted cells were stained by incubation for 12 min at 37 1C with 5 mM CFSE (carboxy- fluorescein diacetate succinimidyl ester; Molecular Probes). Reactions were quenched by washing with ice-cold RPMI medium supplemented with 10% (vol/vol) FCS. CFSE-labeled responder cells (5 ×10⁴) were cultured in 96-well round-bottomed plates in the presence of antigen-presenting cells (1 ×10⁵ erythrocyte-depleted splenocyte samples treated with mitomycin C (Sigma)) and various doses of anti-CD3e. After 3 d, the proliferation of responder cells was assessed by flow cytometry.

FTOC

The medium for FTOC was RPMI-1640 medium with 10% (vol/vol) FCS, 50 mM β-mercaptoethanol, L-glutamine, nonessential amino acids, 10 mM HEPES, pH 7.2, penicillin and streptomycin. C57BL/6 fetal thymic lobes at E15 were cultured for various times in FTOC medium on nucleopore filters (Whatman) and then were analyzed by flow cytometry. Cultures were provided fresh medium every 3 d. For reaggregate thymic organ culture, E15 thymic lobes treated with 1.35 mM 2-deoxyguanosine (Sigma) were digested for 30 min at 37 1C to a single-cell suspension with 0.125% (vol/vol) trypsin (Invitrogen). Reaggregates were formed by the centrifugation of 7 ×10⁵ stromal cells with 1×10³ to 10 ×10³ sorted γδ^25,27+ cells and pipetting of the pellet onto a nucleopore filter in FTOC medium. Cultures were incubated 7–10 d. For CD27 stimulation, a fusion protein of recombinant CD70 and immunoglobulin (10 mg/ml; provided by A. Al-Shamkhani) was added to the cultures.

Adoptive transfer

Sorted cells were stained with CFSE as described above and were injected intravenously into RAG-2-deficient mice (3 ×10⁵ to 5 ×10⁵ cells per mouse). Splenocytes from injected mice were collected after 3 d (for analysis of short-term division and stability of CD27 expression) or 3 weeks (for analysis of long-term stability of CD27 expression).

Malaria infection. Mice were infected as described³¹

RT-PCR. RNA was extracted with TRIzol reagent (Invitrogen) and was treated with RNase-free RQ1 DNase (Promega). RNA concentrations were measured with a Nanodrop ND 1000 (Nucliber). Superscript II reverse transcriptase was used according to the manufacturer's protocol (Invitrogen) for the production of cDNA in a reaction volume of 20 ml. Primers for quantitative real-time RT- PCR or semiquantitative RT-PCR (Supplementary Table 1 online) were designed with the Primer3 program or with the Universal ProbeLibrary Assay Design Center (Roche Applied Science). For semiquantitative RT-PCR, cDNA was amplified as described¹⁸ in a MyCycler thermocycler (BioRad). For quantitative PCR, cDNA was amplified in a reaction volume of 50 ml, with half the volume (25 ml) being SYBR Green PCR master mix (Applied Biosystems), in a Rotor Gene 6000 (Corbett Research). Reactions consisted of 2 min at 50 1C and 10 min at 95 1C, followed by 50 cycles of 15 s at 95 1C and 1 min at 60 1C, followed by 50 s at 95 1C. Expression was normalized to that of Eef1a1 (encoding eukaryotic translation elongation factor 1a1; called ‘Efa1’ here) or Hprt1 (encoding hypoxanthine guanine phosphoribosyl trans- ferase) and is presented as percent of maximum expression.