Regulation of Cutaneous Malignancy by γδ T Cells

See allHide authors and affiliations

Science  19 Oct 2001:
Vol. 294, Issue 5542, pp. 605-609
DOI: 10.1126/science.1063916


The localization of γδ T cells within epithelia suggests that these cells may contribute to the down-regulation of epithelial malignancies. We report that mice lacking γδ cells are highly susceptible to multiple regimens of cutaneous carcinogenesis. After exposure to carcinogens, skin cells expressed Rae-1 and H60, major histocompatibility complex–related molecules structurally resembling human MICA. Each of these is a ligand for NKG2d, a receptor expressed by cytolytic T cells and natural killer (NK) cells. In vitro, skin-associated NKG2d+ γδ cells killed skin carcinoma cells by a mechanism that was sensitive to blocking NKG2d engagement. Thus, local T cells may use evolutionarily conserved proteins to negatively regulate malignancy.

A substantial fraction of the T cell pool is constitutively resident within epithelia. These intraepithelial lymphocytes (IELs) display limited T cell receptor (TCR) diversity and may recognize autologous proteins expressed on epithelial cells after infection or malignant transformation (1). Consistent with this, human bowel carcinomas show up-regulated expression of two major histocompatibility complex (MHC) class I–related molecules, MICA and MICB, and are targets for cytolysis by intestinal TCRγδ+ IELs expressing NKG2d, a receptor for MICA and MICB (2). Nonetheless, the capacity of either γδ cells or MICA to regulate malignancy in vivo is uncertain. Hence, we have studied three murine models of cutaneous malignancy. The skin was chosen because >90% of murine skin-associated IELs express TCRγδ (3) and because of the practicality of assessing tumor development in situ. Although the MICA/B locus is not conserved, mice express counterpart NKG2d ligands (4, 5), which might play a role in the detection of tumors.

Malignancy was induced either by inoculation of carcinoma cells or by chemical carcinogenesis. Thus, mice (i) were injected intradermally with the squamous cell carcinoma line PDV into two sites per mouse (6, 7); (ii) were injected intradermally with the carcinogen methylcholanthrene (MCA) (8, 9); or (iii) received skin applications of dimethylbenz[a]anthracene (DMBA) and phorbol ester (12-O-tetradecanoylphorbol; TPA), which induce and promote cutaneous malignancy, respectively (10,11). To directly test the role of γδ cells in regulating carcinogenesis, we compared tumor development in wild-type C57BL/6 mice with that in TCRδ / mice, which lack γδ cells. Parallel comparisons were made with tumor development in TCRβ / mice, which lack αβ T cells, and with TCRβ / δ / mice, which lack all T cells (12).

When wild-type and TCRδ / mice were challenged with PDV cells, the number of sites that had been inoculated and then developed into tumors was greater in the TCRδ / mice by a factor of 3 to 4. In total, 41 of 110 sites developed as tumors in TCRδ / mice, versus 13 of 134 sites in controls (P < 0.01). Consistent with this difference, 60% of TCRδ / mice developed at least one tumor, compared to <20% of controls. However, there was only a minor reduction in tumor latency (Fig. 1A), indicating that γδ cells reduced the number of events that develop as tumors, but not the time required for tumor development. In TCRβ / mice and TCRβ / δ / mice, ∼100% of sites developed as tumors, and latency was substantially reduced (Fig. 1A). These findings demonstrate that αβ T cells and γδ cells each regulate the growth of PDV-caused tumors, but in distinct fashions. The susceptible phenotype of TCRδ / mice demonstrates that the lack of γδ cells is not compensated for by the presence of αβ T cells and NK cells.

Figure 1

Increased cutaneous malignancy in TCRδ / mice. (A) Frequency of tumor formation, expressed as numbers of tumors per site of inoculation of 106 PDV cells and depicted as a percentage. Average latency for the development of palpable tumors is shown in weeks. (B) Kaplan-Meier plot of tumor-free FVB mice after single-hit MCA application. (C) Number of tumors per mouse developing with time after DMBA initiation and promotion with 5 nmol of TPA (*P < 0.002, **P < 0.001, ***P < 0.00001). (D) As in (C), but with 40 nmol of TPA (*P < 0.002, **P < 0.0005). (E) Frequency of irregular-shaped carcinomas per mouse after DMBA and 5 nmol of TPA (*P < 0.002, **P < 0.0005).

To directly assess the role of γδ cells in the development of MCA-induced fibrosarcomas and spindle cell carcinomas, we first backcrossed TCR-mutant mice (≥11 generations) to FVB mice, which are highly susceptible to chemical carcinogenesis. After MCA injection, greater numbers of TCRδ / mice and TCRβ / mice developed tumors relative to FVB mice (Fig. 1B). This finding agrees with previous studies showing T cell regulation of MCA-induced skin tumors (13). Again, the presence of either type of T cell failed to compensate for the absence of the other.

Naturally occurring human carcinomas often result from incremental insults that cause an accumulation of mutations (14). This type of etiology can be modeled by application of the tumor initiator DMBA followed by the tumor promoter TPA. Palpable local hyperplasias either regress or develop into regular-shaped papillomas, some of which evolve into irregular-shaped carcinomas (11). At 7 weeks, 67% of TCRδ / mice were tumor-bearing versus 16% of wild-type mice. The tumor burden was also increased in TCRδ / mice (Fig. 1C). By contrast, TCRβ / mice and wild-type mice were equally susceptible to DMBA- and TPA-induced carcinogenesis, and at higher doses of TPA, TCRβ−/−mice actually showed reduced susceptibility (Fig. 1D). This finding is consistent with other instances where components of the immune response promote rather than inhibit cutaneous malignancy (15). In addition to showing increased tumor burden, TCRδ−/−mice also revealed a higher incidence of progression of papillomas into carcinomas (Fig. 1E). These experiments provide additional evidence that γδ cells and αβ T cells make distinct contributions to the regulation of tumor growth.

In all three regimens, γδ cell deficiency reduced resistance to cutaneous malignancy. To determine whether the tumor cells might express a functional equivalent of human MICA/B that could act as a ligand for NKG2d on γδ cells, we constructed streptavidin beads displaying the ectodomains of recombinant murine NKG2d fused to a stalk provided by domains 3 and 4 (d3+4) of rat CD4 (16–18). These beads bound PDV cells (Fig. 2A). Mouse cells staining with NKG2d reagents have previously been shown to express the MHC class I–related proteins Rae-1 or H60 (4, 5, 19). Sequence analysis of RNA expressed by PDV cells identified Rae-1ɛ(Fig. 2B) (20–22), a novel fifth sequence encoded by the Rae-1 locus (Fig. 2C) (21). Rae-1 proteins are glycosylphosphatidylinositol (GPI)-linked (4, 5,21); consistent with this fact, PDV cell staining by NKG2d beads was sensitive to the presence of phospholipase C (Fig. 2A).

Figure 2

A cell surface ligand for murine NKG2d. (A) Left panel: PDV cells stained with NKG2d beads (dark shading) and unstained (light shading). Right panel: staining of PDV cells by NKG2d beads after 60-min pretreatment of cells with phospholipase C (dark shading) and unstained (light shading). (B) Rae-1 RT-PCR products detected by ethidium bromide staining in agarose. (C) Predicted amino acid alignment (40) of Rae-1 isoforms. The boxed residues are four charged residues that are conserved in corresponding positions in MICA where they contact homodimeric NKG2d (note that one of the four is not completely conserved in Rae-1); asterisks are primary sites of divergence between isoforms; putative N-linked glycosylation sites are underlined and in bold; italics in Rae-1ɛ denote primer sequences used to clone the cDNA; cysteines are in yellow; STP denotes a serine-threonine-proline–rich domain; α helix and β sheets, kinks, and loops are designated on the basis of the structural model (22).

To test whether murine NKG2d could directly interact with Rae-1ɛ, we immobilized Rae-1ɛ–CD4(d3+4) fusion protein on beads (23) that were incubated with recombinant NKG2d-CD4(d3+4). We observed that 110-kD homodimers of NKG2d-CD4(d3+4) were retained efficiently on the Rae-1–CD4 beads but inefficiently on control beads displaying CD4(d3+4) (Fig. 3B). A specific interaction between soluble Rae-1ɛ and immobilized NKG2d was confirmed by surface plasmon resonance (Fig. 3C) (24).

Figure 3

NKG2d is expressed on TCRγδ+cells and binds Rae-1ɛ. (A) NKG2d expression in TCRγδ+ DETC lines [7-17, B1, 30B4, and U10E1, maintained as resting cultures or stimulated with anti-CD3 (αCD3)]; in transformed keratinocyte cell lines (PAM2-12, PDV); and in epidermal preparations directly ex vivo (crude IEC) and in those enriched for primary DETC cells (enriched IEC). (B) Biochemical evidence for Rae-1ɛ binding to NKG2d was provided by use of magnetic streptavidin beads coated with either biotinylated recombinant CD4(d3+4) (CD4 bead) or a fusion of CD4(d3+4) with Rae-1ɛ (Rae-1ɛ bead). After a 5-hour incubation of beads with 100 nmol of NKG2d-CD4(d3+4), proteins were eluted from the beads with SDS or were sampled from the unbound supernatant and detected by Western blotting using OX68 (anti-CD4). The ∼110-kD NKG2d-CD4(d3+4) homodimer is detected most strongly in the supernatant of the CD4 beads and the bound fraction of the Rae-1ɛ beads. Eluted Rae-1ɛ–CD4(d3+4) is detected at ∼76 kD. (C) Surface plasmon resonance indicates specific binding of Rae-1ɛ–CD4 to NKG2d-CD4. Dark lines are increasing concentrations of Rae-1ɛ analyte: 4, 8, 16, and 32 μM; faint dotted line is CD4 analyte, 8 μM. (D) Phylogeny of MHC class I–related molecules generated with the topological algorithm of ClustalW (Rae-1ɛ and H60 are highlighted).

The hypothesis that the Rae-1–NKG2d interaction is homologous to that of human NKG2d and MICA was tested by molecular modeling. Fold recognition identified the best template for Rae-1 to be the MHC-like molecule ZAG (25, 26). When this model is compared with the crystal structure of MICA complexed to NKG2d (27), it is clear that the charge distribution and contours of surfaces of MICA and Rae-1 are similar and are quite distinct from those of conventional class I MHC (22). Likewise, the residues in human NKG2d that contact MICA are conserved and appropriately located in murine NKG2d (27, 28). Molecular phylogenetic analysis confirmed the relatedness of Rae-1 to MICA/B and to the recently described human ULBP proteins that also bind NKG2d (Fig. 3D) (29).

To determine whether cutaneous TCRγδ+ IELs [known as dendritic epidermal T cells (DETCs)] can respond to Rae-1 expression on PDV cells, we tested DETCs for NKG2d expression. By reverse transcription polymerase chain reaction (RT-PCR), NKG2d was detected at very low levels in a primary interface epidermal cell (IEC) preparation that contains 1% DETC (“crude IEC”), but was more clearly apparent in enriched IEC composed of ∼4% DETC (Fig. 3A). NKG2d was expressed in all DETC lines and was slightly enhanced by cell activation (5, 30). Two recently and independently derived DETC lines, 10-21 and 6-13, were tested for cytolytic effector function toward PDV cells (31, 32). Killing was consistently evident, even at an effector:target (E:T) ratio of 0.3:1 (Fig. 4A). As the E:T ratio was substantially increased, greater cytotoxicity (≤80%) was occasionally observed (33), although there was greater experimental variation, perhaps due to T cell inhibition by-products released from dying tumor cells.

Figure 4

Targeting of Rae-1+ cells by γδ cells and Rae-1 expression in vivo. (A) Left panel: Chromium release assay of 51Cr-loaded PDV cells after incubation with TCRγδ+ DETC, in the presence of the indicated reagents (sRAE-1ɛ = soluble recombinant Rae-1ɛ produced in 293T cells). Right panel: Killing at E:T 3:1, with anti-H2Db, anti-TCRγδ (GL3), anti-NKG2d, or normal rat serum. (B) RT-PCR products for HPRT, Rae-1, and H60 in individual samples of normal skin, skin treated with TPA and harvested several weeks later in the absence of overt lesion, skin treated with TPA and sampled 24 hours later, papillomas, and carcinomas. HPRT controls were performed in the same reaction and are present in each lane, irrespective of the expression of H60 or Rae-1; H60 gives rise to two transcripts detected as two fragments, as described (19). Detection was by ethidium bromide staining, shown in reverse image.

To investigate the molecules mediating the targeting of PDV cells by DETCs, we supplemented killing assays with soluble antibody to TCRγδ, soluble recombinant Rae-1ɛ, or an antiserum to NKG2d (34). Each reagent significantly inhibited killing (Fig. 4A). Moreover, the combination of anti-TCRγδ with either of the other two reagents reduced killing additively by 75 to 95% (Fig. 4A). These results are consistent with recent experiments showing that killing of virus-infected cells by NKG2d+ αβ T cells is contingent on the engagement of NKG2d and TCR (35).

To assess the general relevance of NKG2d-dependent killing of PDV to the immune surveillance of carcinomas, we examined the expression of Rae-1 by RT-PCR in normal FVB skin, in skin treated with DMBA and TPA, in papillomas, and in carcinomas (11, 36). Normal skin showed negligible levels of Rae-1, consistent with evidence that Rae-1 is primarily expressed in embryonic brain and limb buds (21). There was likewise no evidence of H60 expression (Fig. 4B). By contrast, Rae-1 and H60 expression were moderately increased in several areas of painted skin 24 hours after surface application of TPA. Where such areas did not develop any histological lesions, there was subsequently no evidence for Rae-1 or H60 expression. By contrast, Rae-1 and/or H60 were expressed in most freshly explanted papillomas and in all carcinomas.

The high susceptibility of immunosuppressed renal graft patients to squamous cell carcinomas is well established (37). By contrast, the individual immunological mechanisms that contribute to tumor surveillance are not fully defined. In this study, mice lacking γδ cells are shown to have increased susceptibility to three distinct regimens of induced cutaneous malignancy. Moreover, there are clear differences in the nature of the contributions that γδ cells and αβ T cells make to the regulation of malignancy induced by PDV cells and by application of DMBA and TPA.

In species as diverse as chickens, mice, and humans, local T cell subsets are commonly enriched in γδ cells. Here we have shown that TCRγδ+ DETCs can kill squamous carcinoma cells, contingent on the expression of Rae-1 by these cells. Consistent with the capacity of γδ cells to inhibit tumor development, we have shown that Rae-1 is up-regulated in vivo by chemical carcinogens. Because NKG2d is expressed on numerous cytolytic T cells and NK cells, transformed cells expressing NKG2d ligands such as Rae-1 and MICA may be vulnerable to several types of attack. It is likely that the nature of the cells that target tumor cells in vivo is determined by the anatomical accessibility of the tumor to the NKG2d+ cell and the presence of other ligands on the transformed cells that might either activate or inhibit particular types of cytolytic cell. In the case of cutaneous malignancy in the mouse, the nonredundant contribution of γδ cells may reflect the intimate juxtaposition of DETC with keratinocytes and/or the presence on keratinocytes of an as-yet-unidentified ligand for the γδ TCR. γδ T cells may perform similar roles in the human gut, where NKG2d+TCRγδ+ IELs are highly cytolytic (2), and in human skin, where a distinct subset of γδ T cells was recently identified (38).

DETC-mediated cytotoxicity may complement the interferon γ–mediated effects that were recently shown to contribute to immunosurveillance of MCA-induced tumors (13). Additionally, TCRγδ+ DETC can down-regulate inflammation provoked by systemic αβ T cells (39). Because αβ T cell responses can on occasion promote tumor growth (15), as shown for the DMBA-TPA regimen studied here, their down-regulation may be another means by which local γδ cells may reduce primary tumor development. These findings are clearly relevant to understanding the selective pressures on developing tumors and to considering the types of immune responses that would be useful in clinical intervention.

  • * These authors contributed equally to this report.

  • To whom correspondence should be addressed. E-mail: adrian.hayday{at}


View Abstract

Navigate This Article