Primary γδ Cell Clones Can Be Defined Phenotypically and Functionally as Th1/Th2 Cells and Illustrate the Association of CD4 with Th2 Differentiation¹

All mammalian and avian species examined to date harbor two distinct types of T cell that, on the basis of their heterodimeric TCR expression, are classified as αβ and γδ T cells (1, 2). These two lineages of T cells share some common characteristics, such as the association of the TCR chains with CD3 molecules (3), the expression of certain other surface molecules (4), and cell functions, such as cytotoxicity (5, 6, 7). However, they differ significantly in their anatomical distribution, ontogeny, and immunobiologic roles (3, 8, 9, 10, 11, 12, 13, 14, 15).

Over the past few years, a classification of CD4⁺ αβ T cells has been established, primarily according to the cells’ pattern of cytokine production and consequent physiologic function (16). Th1 cells produce IFN-γ and IL-2, which activate T cells and macrophages to attack intracellular pathogens and promote, through B cell help, the synthesis of particular Ig isotypes (e.g., murine IgG2a). Th2 cells, conversely, produce IL-4, IL-5, and IL-10, which help B cells synthesize other Ig isotypes (e.g., murine IgG1 and IgE) commonly associated with the attack on extracellular pathogens. The analysis of numerous infection systems and autoimmune diseases indicates that the skewing of the response to either Th1 or Th2 activation, respectively, has significant consequences for clearance of pathogen and/or characteristics of lymphoid infiltration (17, 18, 19, 20). In addition, Th2 cells may prove to be major physiologic down-regulators of Th1 responses (21).

How γδ T cells might fit into the Th1/Th2 pattern has not been clearly elucidated. First, most γδ T cells are not CD4⁺; thus, if CD4 expression were an important component of Th1/Th2 differentiation, one might expect that γδ T cells would not conform to this paradigm. Furthermore, several experiments using mice congenitally deficient in the synthesis of αβ T cells have demonstrated that γδ cells, unlike αβ T cells, are either incapable of or at best inefficient in providing Ag-specific responses of the kind responsible for pathogen clearance or for Ag-specific autoimmunity (14, 15, 22, 23, 24).

This notwithstanding, human and murine γδ cells have been demonstrated to provide B cell help (25, 26, 27, 28, 29, 30, 31) and, in association with this, were clearly shown to produce IL-4 (26, 29), a signatory Th2-type cytokine. Likewise, mice and humans infected with bacteria, viruses, or protozoa have demonstrated increases in lymphoid or intraepithelial γδ cells, suggesting an involvement of γδ cells in the nature of the host response (32, 33, 34, 35, 36, 37, 38, 39). Indeed, when the intracellular expression of cytokines by such responding γδ cells was examined, it revealed Th1- and Th2-type patterns that paralleled the prevailing Th1 and Th2 αβ T cell responses (40). Since then, additional studies have demonstrated the production of Th1 and Th2 cytokines by populations of γδ cells (41). These data have provoked the question of how closely the production of cytokines by γδ cells might conform at the clonal level to the Th1/Th2 paradigm defined for αβ cells.

In this study, cellular, molecular, and functional evidence is provided for the classification of γδ T cell clones as Th1 or Th2. We discuss the potential relevance of this to emerging bioassays for γδ cells (15, 24, 42, 43, 44, 45) and to the idea that Th1/Th2 determination results from the mode of presentation of peptide-MHC by a professional APC to a responding CD4⁺ T cell (46, 47).

Mice were bred and maintained in specific pathogen-free animal facilities at Yale University (New Haven, CT). TCRα^−/− mice (H-2^b/d) were generated by gene targeting (48). TCRβ×δ^−/− mice were generated locally by breeding TCRβ^−/− mice (H-2^b) (49) with TCRδ^−/− (H-2^b) mice (50). CB17.SCID mice (H-2^d) were obtained from The Jackson Laboratory (Bar Harbor, ME).

TCRα^−/− splenocytes (2 × 10⁶/ml) were cultured in Click’s medium plus 5 U/ml IL-2 (supernatant of EL4), 5% heat-inactivated FCS (HyClone, Logan, UT), and antibiotics (Life Technologies, Grand Island, NY). The cultures were replenished with medium every 3 to 4 days. Irradiated (3000 rad) feeder cells (BALB/c splenocytes, 10⁶/ml) were added 2 wk after the initial culture and weekly thereafter for 4 wk. Once a line was established, it was weaned off feeders.

Limiting dilution of γδ cell lines was performed on irradiated feeder cells in 96-well microtiter plates at n ≤ 1 cell/well. Medium was replenished every 3 to 4 days, and irradiated (3000 rad) feeder cells were provided weekly for the first few weeks and at 2-wk intervals subsequently until the establishment of clones, after which the supply of feeder cells was gradually stopped. All experiments presented in this study were performed with established clones (free of APC).

The following directly conjugated mAbs were purchased from PharMingen (San Diego, CA): PE-conjugated anti-CD3 (2C11), anti-TCRγδ (GL3), and anti-CD8 (53-6.7); FITC-conjugated anti-TCRαβ (H57) and anti-CD4 (RM4-5); and biotin-conjugated anti-Thy1.2 (53-2.1), anti-CD45R (B220, RA3-6B2), and anti-CD40 ligand (CD40L, MRL). Hybridoma culture supernatants were either maintained in this laboratory (2C11, H57, and GL3) or provided by Dr. C. Janeway, Jr. (Howard Hughes Medical Institute, Yale University: 2.4G2, anti-Fc receptor; 212A.1, anti-I-A^b/d), or Dr. Albert Bendelac (Princeton University, Princeton, NJ; anti-CD1).

Single cell suspensions (∼10⁶ cells/ml) were incubated with PE- or FITC-conjugated Abs at pretitrated dilutions on ice for 30 min, followed by washing three times with PBS-1% FCS and 0.02% sodium azide. Biotin-conjugated mAbs were further incubated with fluorescence-conjugated streptavidin. Stained cells were fixed in PBS-1% paraformaldehyde and analyzed on a FACScan (Becton Dickinson, Mountain View, CA). Dead cells and nonlymphoid cells were excluded by selective gating on forward and side scatter. For sorting, splenocytes (10⁸/ml) were stained with PE-conjugated anti-TCRγδ (GL3) and FITC-conjugated anti-CD4 (RM4-5). After washing, cells were resuspended at 2 × 10⁷/ml in PBS-2% FCS for sorting on a FACStar (Becton Dickinson), after which they were either used directly as a source of RNA (ex vivo sample) or activated in Click’s medium, 5% FCS, and 2.5 μg/ml Con A for 48 h before being used as a source of RNA (activated sample). In the latter case, the viability of cells was confirmed before and after harvest by trypan blue exclusion.

RNA was prepared from γδ cell clones (∼2.5 × 10⁶ cells) by RNAzol (Biotecx Laboratories, Inc., Houston, TX) and reverse transcribed into single strand cDNA using an oligo(dT) primer (Pharmacia, Piscataway, NJ) and Moloney murine leukemia virus reverse transcriptase (Life Technologies) at 37°C (neutral pH) for 60 min. Two microliters of the cDNA (100 μl) was amplified with primers specific for IL-4, IL-5, IL-10, IFN-γ, TGF-β, Fas, and FasL together with hypoxanthine phoshoribosyl transferase (HPRT; as a control) in the presence of 100 ng of the 5′ and 3′ primers, 1 μl of dNTPs (10 mM), 1.5 mM MgCl₂, and 1 U of Taq polymerase (Boehringer Mannheim, Indianapolis, IN). The PCR reactions were denatured at 94°C for 3 min followed by 35 cycles of 94°C for 20 s, 60°C for 20 s, and 72°C for 40 s and a final extension at 72°C for 7 min. PCR products were analyzed on 1.5% agarose gels. Primers were synthesized in the Keck Facility of Yale University, and the primer sequences for cytokines were adopted from the report by Reiner et al. (51) with corrections: IL-4, 5′-CATCGGCATTTTGAACGAGGTCA-3′ and 5′-CTTATCGATGAATCCAGGCATCG-3′; IL-5, 5′-GAAAGAGACCTTGACACAGCTG-3′ and 5′-GAACTCTTGCAGGTAATCCAGG-3′; IL-10, 5′-CCAGTTTTACCTGGTAGAAGTGATG-3′ and 5′-TGTCTAGGTCCTGGAGTCCAGCAGACTCAA-3′; IFN-γ, 5′-CATTGAAAGCCTAGAAAGTCTG-3′ and 5′-CTCATGAATGCATCCTTTTTCG-3′; HPRT, 5′-GTTGGATACAGGCCAGACTTTGTTG-3′ and 5′-GAGGGTAGGCTGGCCTATGGCT-3′; and additionally, Fas, 5′-ATCCGAGCTCTGAGGAGGCGGGTTCATGAAAC-3′ and 5′-GGAGGTTCTAGATTCAGGGTCATCCTG-3′; and FasL, 5′-CAGCTCTTCCACCTGCAGAAGG-3′ and 5′-AGATTCCTCAAAATTGATCAGAGAGAG-3′.

Cytokine mRNA was quantitated by competitive PCR (51), employing the simultaneous amplification by the same primers of known quantities of competitor DNA fragments. The competitor (provided by Dr. Richard Locksley, University of California, San Francisco, CA) differs from the cDNA of interest by an insert that allows the relative amounts of the two amplification products to be distinguished by gel electrophoresis.

To establish comparable amounts of cDNA template for subsequent analysis of cytokine gene expression, the cDNA was used as a template for the amplification of the HPRT housekeeping gene in the presence of varying amounts of HPRT competitor fragment (see Fig. 6,B and accompanying text in Results; numbers at the top of the lanes refer to fold dilutions of the competitor HPRT fragment used). At each dilution of competitor, the ratio of the product derived from the cellular cDNA to the product derived from the competitor was assessed on a Bio-Rad Laboratory Densitometer (Richmond, CA), using photographic negatives from ethidium-stained agarose gels. These ratios were plotted against competitor concentration, and the linear ranges established and compared by determination of regression (Fig. 6, C and D, and accompanying text in Results). This allowed us to compare the competitor concentrations required to obtain specific ratios of cell product to competitor product for each of the cDNAs, from which we could determine the relative operational concentrations of the cDNAs under study. Dilutions of those equivalent concentrations of cDNAs were then used as substrates for amplification of cytokine genes in the presence of a range of competitor concentrations. The ratios of cDNA product to competitor product were likewise plotted against the competitor concentration. When these plots were compared, it was possible to assess the relative abundance of cytokine cDNA in the different samples (see Fig. 6, E–I, in Results).

TCR γ and δ gene rearrangements of γδ clones were independently amplified by RT-PCR using “hot start” (denaturing at 94°C for 5 min) followed by 38 cycles of 94°C for 1 min, 58°C for 40 s, and 72°C for 1 min in a DNA thermal cycler 480 (Perkin-Elmer/Cetus, Emeryville, CA). The amplified products were purified (Qiagen, Chatsworth, CA), ligated, and transformed into Escherichia coli using the TA method (Invitrogen, San Diego, CA). Sequencing analysis was performed using Sequenase (52), [³⁵S]dATP, and SP6-, T7-, or TCR-specific oligonucleotides as primers, as previously described (28).

Splenocytes (10–15 × 10⁶) from TCRβ×δ^−/− mice (10–16 wk old) mixed with 3 to 5 × 10⁶ cloned γδ cells were injected i.v. into CB17.SCID mice (6–8 wk-old). Cloned γδ cells (3–5 × 10⁶) were also i.v. transferred to TCRβ×δ^−/− mice. For cell tracing, γδ clones and splenocytes from a TCRβ×δ^−/− mouse were labeled with the fluorescent dye DiI (Molecular Probes, Inc., Eugene, OR) at 37°C for 30 min before adoptive transfer. Transferred cells that were labeled could be observed in frozen spleen sections by fluorescence microscopy, as previously described (27).

For Ig quantitation, recipients were bled, and individual serum samples were collected every 2 wk postreconstitution. Total levels of serum IgM, IgG1, IgG2a, IgG2b, IgG3, and IgA were determined by ELISA as described previously (26, 27), using reagents from Southern Biotechnology Associates, Inc. (Birmingham, AL). Briefly, microtiter ELISA plates (Dynatech Laboratory, Inc., Chantilly, VA) were coated with goat anti-mouse IgH+L (5 μg/ml) in coating buffer (carbonate buffer, pH 9.6). After blocking the plates with 1% BSA (PBS containing 1% BSA), diluted serum samples (1/100 in blocking buffer) were added in duplicate. For standards, serial dilutions of mouse IgM, IgG1, IgG2a, IgG2b, IgG3, and IgA (Southern Biotechnology Associates) were also added in duplicate, starting at 1 μg/ml. After incubation and washing, alkaline phosphatase-conjugated goat anti-mouse IgM, IgG1, IgG2a, IgG2b, IgG3, and IgA were added individually. The enzymatic reaction was developed by adding substrate p-nitrophenyl phosphate and was stopped by adding 1 N NaOH. The plates were read at an absorbance of 405 nm on an microplate reader (Dynatech). The concentrations of Ig isotypes were determined by referring to standard curves performed in the same assay with known concentrations of various mouse Ig isotypes using the equation y = intercept + slope × log(x); the actual serum concentrations were obtained by y × the serum dilution (i.e., 100). Note that the concentrations quoted in the figures in this paper are for the sera and are calculated from the concentrations of serum dilutions that were experimentally determined and compared with concentrations of standards in the same ranges as the serum dilutions. Secreted cytokines (IL-4 and IFN-γ) in culture supernatants were also measured by ELISA using mAbs against murine IL-4 or IFN-γ (PharMingen) together with different concentrations of rIFN-γ and IL-4 (Life Technologies) as standards. The concentrations of IL-4 and IFN-γ in the culture supernatants were converted as described for Ig isotypes.

GC formation was examined in the spleens of reconstituted SCID mice and TCRβ×δ^−/− mice, respectively, using immunohistochemistry as reported previously (27, 28).

Liver, kidney, intestine, and lung from the reconstituted SCID or TCRβ×δ^−/− mice were fixed in 10% buffered formalin, paraffin embedded, and stained with hematoxylin and eosin. The sections were examined microscopically for lymphocytic infiltration to evaluate the presence of graft-vs-host-disease.

Two γδ cell lines were derived from splenocytes of two TCRα^−/− mice (H-2^b). They were CD4⁺ and CD4⁻CD8⁻ (double negative, DN), respectively. From these, a total of five γδ cell clones were obtained by limiting dilution. Two (G5 and H4) were CD4⁺; three were CD4⁻,CD8⁻ (A3, F6, and H2; Fig. 1,A). Also studied was the expression on the clones of adhesion molecules (ICAM-1, LFA-1, and lymphocyte Peyers patch adhesion molecule-1 (LPAM)), costimulatory molecules (CD28 and CD40 ligand), and CD1 that are either known or speculated to play important roles in T cell function. Examples of the range of expression levels are provided in Figure 1,B, and the data are summarized in Table I. All five γδ cell clones expressed LFA-1, ICAM-1, and CD28 similarly, while the expressions of CD40L and α₄ integrin (LPAM) were more heterogeneous. The data are consistent with previous studies of γδ cell populations (4, 53, 54), in that expression of CD28 is more variable than is generally the case for αβ cells. Interestingly, CD1 was clearly expressed by all the clones in which it was tested, consistent with the recent report that CD1 is expressed by various hemopoietic cells (55).

Cytokine mRNA expression by γδ clones was determined by RT-PCR. Both CD4⁺ γδ T cell clones (H4 and G5) conformed strikingly to a typical Th2 phenotype: high levels of IL-4, IL-5, and IL-10 and undetectable levels of IFN-γ (Fig. 2,A) or IL-2 (data not shown). By contrast, all DN γδ T cell clones conformed to a typical Th1 phenotype, displaying high levels of IFN-γ (Fig. 2,A) and IL-2, against undetectable IL-4, IL-5, and IL-10 (Fig. 2 A).

It has been suggested that Fas/FasL interaction defines a differential regulatory capacity of CD4⁺ T cells (56, 57), with Th1 CD4⁺ αβ T cells expressing high ratios of FasL:fas, and Th2 CD4⁺ αβ T cells expressing higher ratios of fas:FasL (Th0 cells are reported to express Fas and FasL approximately equivalently) (57). To test the degree to which the γδ clones conformed to this Th1/Th2 dichotomy, RT-PCR was again applied. D10, a CD4⁺ αβ⁺ Th2 clone, and B3B3, a CD4⁺, αβ⁺ Th0 clone were included as positive controls. The data (Fig. 2,B) show that the FasL:Fas expression ratios of the γδ cell clones conformed to their classification as Th2 and Th1 cells, respectively: G5 and H4 (Th2) expressed higher levels of Fas than of FasL, while A3, F6, and H2 (together with an additional IFN-γ-expressing γδ clone, A7) showed significant levels of FasL, but negligible expression of Fas (Fig. 2 B).

To examine whether gene expression patterns by the clones were representative of effector molecule production, the secretion of IL-4 and IFN-γ was examined by ELISA. Consistent with the RT-PCR data, secreted IL-4 was detected only in the supernatants of H4 and G5, while secreted IFN-γ was detected only in the supernatants of F6 and H2 (Fig. 2 C) and A3 (data not shown).

A series of experiments was undertaken to determine whether the functional capabilities of the clones likewise conformed to the Th1/Th2 classification. First, the γδ clones were activated in vitro with anti-CD3 and cocultured with naive, primary B cells derived from TCRβ×δ^−/− mice; such cells were uninfluenced by any prior exposure to T cells. Igs of different isotypes were measured in the culture supernatants (n = 3) harvested 7 days postincubation. All γδ⁺ clones elicited Ab secretion (Fig. 3), but in the absence of stimulation by anti-CD3, most of the Ig produced by B cells was IgM. Conversely, activation of all the γδ⁺ clones followed by coculture with naive B cells provoked IgG secretion. However, the putative Th2-γδ⁺ clones, G5 and H4, induced class switching primarily to IgG1, whereas the putative Th1 γδ⁺ clones induced class switching primarily to IgG2a. This conforms strikingly to the behavior of Th1 and Th2 αβ T cell clones (Fig. 3).

To test whether the Th1/Th2 classification applied to the γδ⁺ clones in vivo, CB17.SCID recipients (n = 3–4/group) were adoptively transferred with various γδ T cell clones, admixed with splenic B cells derived from TCRβ×δ^−/− mice. As controls, γδ clones alone or TCRβ×δ^−/− splenic B cells alone were also transferred to CB17.SCID recipients (n = 2–3/group). The engraftment of cells was confirmed in the short term by tracing, using fluorescent dye (DiI)-labeled γδ clones or B cells (as in our previous studies (27)) and in the longer term by FACS analysis at 4 wk postadoptive transfer (Fig. 4 A).

GC formation, a defining signature of T-B collaboration (28, 58, 59), and the follicular development of memory B cells, were also examined in SCID mice receiving γδ T cell clones together with splenic B cells from TCRβ×δ^−/− mice. As shown in Figure 4,B, the DN (Th1-like) γδ T cell clone, F6, induced GC formation by B cells from TCRβ×δ^−/− mice. GC reconstitution was previously shown using the Th2 clone, G5 (27). No GCs formed when splenocytes from TCRβ×δ^−/− mice were inoculated without γδ clones (Fig. 4 C) (27). To extend this finding, DN and CD4⁺ γδ clones were transferred directly to TCRβ×δ^−/− mice, in which, in the congenital absence of T cells, most Ab production is IgM and in which GCs do not develop (28). Again, GC formation was induced by both putative Th1 and putative Th2 γδ clones (data not shown). Consistent with our earlier findings (28), no GCs were observed in the TCRβ×δ^−/− recipients that received PBS only.

At various time points post-transfer, the CB17.SCID recipients were also assessed for serum Ig. The data (Fig. 5) reveal that all γδ⁺ clones sustained the production of IgG by transferred B cells, albeit at low levels. Conversely, Abs in SCID mice receiving B cells alone were almost exclusively IgM. Strikingly, IgG1 production was reproducibly higher in mice receiving putative Th2 clones G5 and H4, whereas IgG2a production was higher in mice receiving putative Th1 γδ T cell clones, e.g. F6 (Fig. 5).

The TCRγ/δ gene usage of the clones was defined by RT-PCR, cloning, and sequencing (Table II). Both CD4⁺ Th2-type γδ⁺ T cell clones used Vγ7-Jγ1 gene segments, whereas the DN Th1-type γδ⁺ T cell clones used Vγ1-Jγ4 gene segments. All the γδ⁺ T cell clones used in this study expressed Vδ6-Jδ1 gene segments (Table II). The two CD4⁺ Th2-type γδ T cell clones that were originally derived from the same line shared identical joining sequences for both Vγ and Vδ gene segments, indicating that they are in all likelihood sister subclones of a single progenitor. The junctional sequences in two DN Th1-type γδ T cell clones (A3 and H2) were likewise identical. Thus, the stable retention of a uniform Th1 or Th2 phenotype by sibling clones over time (the clones have been maintained for >2 yr), which is a further criterion of Th1/Th2 differentiation in αβ cells, is likewise a characteristic of the γδ⁺ clones. A third clone, F6 (DN, Th1 type), expressed different Vγ1-Jγ4/Vδ6-Jδ1 rearrangements (Table II). Strikingly, the Vγ1-Jγ4 rearrangement of this clone is identical with a monoclonal rearrangement (15.32) (28) microdissected from a GC of a different, infected TCRα^−/− mouse in which there was marked IgG expression.

The Th2 clones described here are CD4⁺. Compared with DN and CD8αα γδ⁺ cells, CD4⁺ γδ cells are rare, even in TCRα−/− mice (48, 60). It therefore seemed surprising that the exercise of cloning Th2 γδ⁺ clones yielded cells that were CD4⁺. In turn, this provoked the hypothesis that CD4 expression might be more strongly associated with Th2 differentiation than with Th1 differentiation. To examine this in vivo, an experiment was undertaken to compare the amount of IL-4 RNA expressed by CD4⁺ γδ cells and CD4⁻ γδ cells. CD3⁺ TCRγδ⁺ cells were isolated directly from TCRα^−/− spleens and sorted by FACS into CD4⁻ or CD4⁺ subsets (windows R2 and R3 in Fig. 6,A). In one experiment, the sorted CD4⁺ γδ⁺ cells and CD4⁻ γδ cells were used immediately ex vivo to prepare RNA. In a second experiment, the CD4⁺ γδ cells and the CD4⁻ γδ cells were each activated for 48 h in the presence of Con A (2.5 μg/ml) and then harvested for RNA. In each case, the extracted RNA was used as substrate for cDNA synthesis. To establish comparable levels of cDNA template for subsequent analysis of cytokine gene expression, the protocol described in Materials and Methods was applied. Briefly, cDNAs were compared for their capacity to act as templates for HPRT gene amplification in the presence of varying amounts of competitor HPRT fragment. Data for the activated CD4⁺ γδ⁺ and CD4⁻ γδ⁺ samples are shown in Figure 6,B (numbers at the top of the lanes refer to fold dilutions of the competitor HPRT fragment used). At each dilution of competitor, densitometry was used to determine the ratio of the product derived from the cellular cDNA to the product derived from the competitor; these ratios were then plotted against the competitor concentration, and the linear range was established (Fig. 6, C and D; equations for linear regression: activated CD4⁺ cells, y = −2.9583e⁻² + 29.305x, R² = 0.977; activated CD4⁻ cells, y = 0.19125 + 30.626x, R² = 0.983). From these plots, the relative operational concentrations of the cDNAs could be determined, allowing equal amounts of cell cDNA to be used for subsequent competitive amplification of cytokine genes. When this was attempted for IL-4, using a range of concentrations of competitor, it was immediately apparent that cDNA from the activated CD4⁺ cells competed much more effectively for the IL-4 primers than did cDNA from CD4⁻ cells (Fig. 6,E), in support of the stated hypothesis. The same was true for the nonactivated ex vivo samples (Fig. 6,G). To calculate more precisely the difference in relative IL-4 cDNA concentrations, a broad range of competitor was used with each cDNA sample (Fig. 6,F) to establish the range (in each case) over which the ratio of product derived from the cDNA compared with product derived from the competitor showed a linear relationship to input competitor (Fig. 6, H and I show the data for the activated CD4⁺ and CD4⁻ samples). Comparison of these plots revealed IL-4 expression in activated CD4⁺ γδ cells to be 17.4 times more abundant than IL-4 expression in activated CD4⁻ γδ cells (see Fig. 6). When the same approach was applied to the ex vivo samples, a similar excess of >10-fold IL-4 RNA was found in the CD4⁺ γδ⁺ sample (Fig. 6 G; quantitation data not shown). The expression of IL-10 showed a similar pattern. These data demonstrate that the expression of Th2 cytokines is at least an order of magnitude greater in peripheral CD4⁺ γδ⁺ cells than in DN γδ⁺ cells. This did not apply to IFN-γ, which was more highly expressed by CD4⁻ γδ cells (data not shown).

Given that γδ cells are commonly involved in host responses to pathogens (32, 33, 34, 35, 36, 37, 38, 39, 40), it was important to establish the degree to which γδ cells conform to the Th1/Th2 paradigm. In this regard, Wen et al. (26) first showed that γδ cells could functionally help B cells in vivo by production of IL-4, whereafter Ferrick et al. (40) showed the production of IFN-γ and IL-4, respectively, by peritoneal and spleen γδ T cells in response to Th1- and Th2-stimulating pathogens, respectively. Nonetheless, there was, to date, no demonstration that these results reflected the differentiated phenotype of distinct γδ⁺ clones, nor was there a demonstration of the degree to which various Th1 and Th2 phenotypic markers cosegregate in γδ clones. These issues are clarified by the findings presented here that cosegregation of Th1 and Th2 markers clearly occurs in at least some γδ clones, possibly to a greater degree than is the case in αβ T cell clones (61). This may reflect the fact that by comparison to αβ T cells a greater proportion of peripheral γδ cells may be preactivated, and hence no longer in a plastic differentiation state. This would be consistent with our capacity to readily measure Th1/Th2 cytokine expression by γδ cells without the need for prior activation (e.g., Fig. 6).

The generation and maintenance of αβ T cell clones is exclusively dependent on Ag and APCs. Once established, the growth of the γδ clones reported in this study did not require the sustained presence of APCs (see Materials and Methods). This has now been observed in several independent instances. The specificity of the γδ clones described here is under investigation, but we have ruled out a requirement for conventional, professional APCs. This is consistent with >50 characterizations of human and murine γδ TCR specificity that collectively failed to demonstrate any response to conventional class I/II MHC-processed peptide (reviewed in 62 . Hence, the data most strongly suggest that a Th1/Th2 classification can be established in primary T cells in the absence of specific peptide presentation by conventional class I/II MHC.

The role of CD4 in the differentiation of Th1 and Th2 αβ cells in CD4-expressing mice has been difficult to assess because CD4 has an important role during αβ T cell development that is epistatic to the differentiation of peripheral Th1/Th2 cells. This is not the case for γδ cells, which mostly develop as DN cells. Indeed, recent data from our laboratory indicates that the rare, but reproducible, numbers of CD4⁺ γδ cells that are present in the periphery of mice (60) and humans (4) develop from CD4⁻ CD8⁻ thymocytes, not from the CD4⁺CD8⁺ (double positive) pool (63). Two sets of data in this report suggest that CD4 expression may be more involved in the differentiation of Th2 cells rather than Th1 cells. First, although CD4⁺ γδ cells are rare in vivo (48, 60), the Th2 clones (albeit only two sibling clones) were both CD4⁺, while the Th1 clones were DN. Second, Th2 cytokine RNA expression was enriched in polyclonal CD4⁺ γδ cells examined directly ex vivo, whereas IFN-γ expression was not. The Ig isotype profiles of αβ T cell-deficient mice also implicate CD4 in Th2 responses; in TCRα^−/− mice, some of the B cell help is provided by CD4⁺, TCRαβ⁺ cells (60, 64). Such T cells are easily detected in the GCs of this strain (64), and the prevalent Ig isotypes (e.g., IgG1) are primarily Th2 associated (26, 31). By contrast, in TCRβ^−/− mice, all help is provided by γδ cells (24, 28), most of which are DN. DN CD3⁺ cells are readily discernible in the GCs of such mice (28), and the prevalent Ig isotypes (e.g., IgG2a) are of the Th1 type (24, 28). In summary, the association of CD4 with Th2 differentiation can be taken to suggest that systemic γδ cells are more likely to be Th1-type cells than Th2 cells, consistent with which, IFN-γ production is more commonly noted as a product of γδ cells. Nonetheless, there may be an important biologic role for Th2 γδ cells (see below).

The involvement of CD4 in Th2 differentiation/function may reflect engagement of an APC by CD4 as well as by TCR γδ, inducing higher levels of signaling in the responding T cell that are thought to favor Th2 differentiation (46). This would be consistent with several observations that CD4 expression is nonetheless not essential for αβ T cells to display a Th2 phenotype (65, 66, 67). Additionally, gut CD8⁺ γδ cells have been reported to show Th2-like activity. In all these instances, Th2 signaling may be induced by high dose Ag alone and/or by engagement of other molecules in addition to the TCR.

The only known ligand on APCs for CD4 is MHC class II. Resolving the specificities of the γδ clones described here will clarify whether CD4 engagement of MHC class II can augment signaling from a γδ TCR that is reactive to an MHC class II-independent ligand, or whether the augmentation only occurs when CD4 and TCRγδ coengage MHC class II, the latter most likely through a nonconventional mechanism, previously reviewed (2, 62).

We also note that all the clones tested expressed surface CD1. The role of CD1 in the immune system is not fully clarified, but there are data that the direction (Th1/Th2) of an αβ T cell response is in part influenced by the production of cytokines by CD1-reactive NK-T cells. The data provided here raise the intriguing possibility that γδ cells might themselves interact with T cells reactive to CD1. Consistent with this, CD1 was expressed on a subset(s) of γδ cells in vivo (data not shown), an issue currently under study.

Collectively, numerous reports have indicated that the number of γδ cells can greatly increase in humans and/or mice infected with bacteria, parasites, or viruses, and thus may contribute to the immune responses to these challenges (32, 33, 34, 35, 36, 37, 38, 39, 40). This may be particularly true during the neonatal period, when γδ cells are relatively abundant, and αβ T cell-APC interactions may not be fully established. The γδ clones described here may thus be representative of Th1/Th2 effector cells (40, 41). Indeed, the recent analysis of CD1−/− mice (68) indicates that significant levels of IL-4 are produced in the absence of CD1-reactive, NK1.1⁺ αβ T cells, previously considered as the T cells that skewed αβ T cell responses toward Th2. Given our original findings with γδ cells (26), γδ Th2 cells might under some circumstances be an important initiator of Th responses to infection. An important role for Th1/Th2 γδ cells in the establishment of Th cell responses would be consistent with the impairment in IgA synthesis seen in TCRδ^−/− mice (69).

At the same time, an increasingly noted phenotype of TCRδ^−/− mice is one of dysregulated, hyperactive immune function toward either foreign or self Ags (15, 24, 42, 44). Thus, it has been inferred from these and other data (43, 45) that γδ cells ordinarily down-regulate αβ T cells of either Th1 (15, 24, 42, 44, 45) or Th2 (43) function, either directly and/or indirectly. Indeed, γδ T cells have been shown to regulate the activation of macrophages (70), NK cells (71), and αβ T cells (72). It is quite conceivable that such regulation is mediated by Th1 and Th2 cytokines. Since exposure to IL-4 of professional APCs, such as macrophages, reduces their capacity to stimulate Th1 αβ cells (reviewed in 21 , it is possible that a major physiologic function of Th2 γδ cells is to attenuate the responses of Th1 αβ cells. In this regard it is notable that although they can be rare, CD4⁺ γδ cells appear to be conserved in all vertebrates in which they have been sought. A converse regulatory role (acting on αβ Th2 responses) may prove true for Th1 γδ cells. This hypothesis, that Th1 and Th2 γδ clones play important regulatory roles, would be entirely consistent with the nonredundant function of γδ cells and αβ cells that is evident from several independent, recently reported analyses of TCRδ^−/− mice (15, 24, 44).

During the preparation of this manuscript, two analyses primarily of the CD4⁻ mouse, indicated that CD4 is also a critical molecule in Th2 differentiation of αβ T cells (76, 77). Such studies would appear complementary to these studies of γδ cells in CD4-sufficient hosts.

We thank C. A. Janeway and J. Craft for review of the early forms of this manuscript, Adrian Smith for discussions, and C. A. Janeway and A. Bendelac for Abs.