Research ArticleCancer

An innate-like Vδ1+ γδ T cell compartment in the human breast is associated with remission in triple-negative breast cancer

See allHide authors and affiliations

Science Translational Medicine  09 Oct 2019:
Vol. 11, Issue 513, eaax9364
DOI: 10.1126/scitranslmed.aax9364

γδ T cells: Guarding and defending tissues

Specific γδ T cell subsets are known to populate tissues such as the skin and gut. To ascertain γδ T cell participation in breast cancer surveillance, Wu et al. isolated cells from grid cultures of human breast tissue. They observed cytolytic innate-like Vδ1+ T cells in healthy human breast tissue as well as tumor tissue. Analysis of a small cohort of women with triple-negative breast cancer revealed that these Vδ1+ T cells were associated with remission and overall survival. These results demonstrate that human tissue-resident γδ T cells may affect breast cancer progression; these cells could be key to existing or future immunotherapy interventions.

Abstract

Innate-like tissue-resident γδ T cell compartments capable of protecting against carcinogenesis are well established in mice. Conversely, the degree to which they exist in humans, their potential properties, and their contributions to host benefit are mostly unresolved. Here, we demonstrate that healthy human breast harbors a distinct γδ T cell compartment, primarily expressing T cell receptor (TCR) Vδ1 chains, by comparison to Vδ2 chains that predominate in peripheral blood. Breast-resident Vδ1+ cells were functionally skewed toward cytolysis and IFN-γ production, but not IL-17, which has been linked with inflammatory pathologies. Breast-resident Vδ1+ cells could be activated innately via the NKG2D receptor, whereas neighboring CD8+ αβ T cells required TCR signaling. A comparable population of Vδ1+ cells was found in human breast tumors, and when paired tumor and nonmalignant samples from 11 patients with triple-negative breast cancer were analyzed, progression-free and overall survival correlated with Vδ1+ cell representation, but not with either total γδ T cells or Vδ2+ T cells. As expected, progression-free survival also correlated with αβ TCRs. However, whereas in most cases TCRαβ repertoires focused, typical of antigen-specific responses, this was not observed for Vδ1+ cells, consistent with their innate-like responsiveness. Thus, maximal patient benefit may accrue from the collaboration of innate-like responses mounted by tissue-resident Vδ1+ compartments and adaptive responses mounted by αβ T cells.

INTRODUCTION

γδ T cells comprise a highly conserved third lineage of lymphocytes that uses somatic gene rearrangement to encode the defining antigen receptor (1, 2). Although this is a hallmark of adaptive immunity, subsets of murine γδ T cells also display innate-like activity, manifest in rapid responses to self-encoded “stress antigens” such as ligands for the NKG2D receptor (36). This is known as lymphoid stress surveillance (7).

Given that NKG2D ligands are up-regulated by overactivity of epidermal growth factor receptor (EGFR) signaling and DNA damage (8, 9), it is natural that lymphoid stress surveillance might contribute to cancer immunosurveillance (10). γδ T cell–deficient mice show greatly increased susceptibility to cancer in several systems (4, 1113), and many attempts are ongoing to exploit their activities clinically (14). Such approaches may enhance the efficacy of current immunotherapies such as checkpoint blockade and, in particular, chimeric antigen receptor (CAR) T cells, which have shown limited success in treating solid tumors. Moreover, the capacity of some γδ T cell subsets to secrete chemokines and cytokines and/or to present antigen argues strongly for their potential to promote the therapeutic potentials of other cell types (12, 1517).

In mice, signature γδ T cell compartments are associated with discrete tissues such as epidermis, dermis, lung, uterus, and intestinal epithelium (1825), seemingly offering optimal capacity to detect and respond to malignant transformation of neighboring cells. Accordingly, γδ T cell–deficient mice have increased susceptibility to skin carcinogens owing to the lack of dendritic epidermal γδ T cells (5). Whether local γδ T cell compartments populate all tissues is unresolved. Nonetheless, the prospect of a mouse breast-associated compartment was supported by the fact that the representation, albeit variable, of γδ T cells in lactating mammary glands was at least fourfold higher than in draining lymph nodes. Moreover, those cells used a variety of γδ T cell receptors (TCRγδs), distinguishing them from skin- and gut-resident γδ T cell compartments (26).

There has been long-standing interest in the degree to which tissue-associated γδ T cell compartments might be conserved in humans and whether or not they contribute to cancer immunosurveillance. On the one hand, humans harbor no obvious counterparts of dendritic epidermal γδ T cells; on the other hand, jawless vertebrates have skin-resident and gut intraepithelial cells with many parallels to γδ T cells, suggesting that such compartments have been conserved for over half a billion years (27). We therefore hypothesized that suboptimal methods for the detection and/or extraction of T cells from human tissues might have confounded attempts to identify and characterize conserved extralymphoid γδ T cell compartments. This hypothesis is consistent with inefficiencies and biases reported both for extracting TCRαβ+ tissue-resident memory T (TRM) cells (28) and for visualizing tumor-infiltrating lymphocytes (TILs) in situ (29) and derives support from our recent characterization of a large intraepithelial γδ T cell compartment in the human gut (30).

In this regard, the care of women in a large breast cancer risk surveillance and treatment practice offered a rare opportunity to analyze the status of γδ T cells in healthy tissue obtained from: reduction mammoplasty or risk-reducing mastectomy; malignant tissue from wide local resection; and paired malignant and nonmalignant tissues from therapeutic mastectomies. Additionally, the importance of investigating the possible existence of local γδ T cells was underlined by evidence that TIL densities were positive prognostic indicators in some types of breast cancer (31, 32). Despite this, the potential of immunotherapy in breast cancer remains unclear with disappointing response rates to current immunotherapies, such as checkpoint blockade (33, 34).

RESULTS

Vδ1 T cells compose a major human breast-resident γδ subset and are skewed toward cytolysis

Our initial goal was to assess the status of T cells resident within healthy human breast. However, the obtainment and characterization of lymphocytes from healthy human tissues has commonly been confounded by poor and irreproducible yields and low cell viability. To redress this problem, Clark and colleagues (3538) developed a “grid” explant culture system that permitted the recovery and characterization of large numbers of healthy skin-resident and lung-resident T cells without substantively changing their phenotype.

We therefore applied grids to disaggregated breast tissue from 29 healthy subjects undergoing reduction mammoplasties or risk-reducing mastectomies. Compared to the limited and variable recovery and poor viability of lymphocytes examined directly ex vivo, grids facilitated the recovery and maintenance of CD3 natural killer (NK) cells/innate lymphoid cells (ILCs), γδ T cells, and CD4+ and CD8+ αβ T cells in every case, albeit there was some enrichment of NK/ILC and γδ T cells (table S1 and Fig. 1A).

Fig. 1 Healthy human breast tissue harbors tissue-resident Vδ1+, Tc1-skewed γδ T cells.

(A) Representative flow cytometry plots showing the gating strategy to identify lymphocytes including γδ T cell subsets isolated from human breast tissue and following grid culture. Lymphocytes were gated on size and scatter (1), followed by live-dead exclusion (2), a singlet gate (3), and CD45 (4) before subsetting (5 to 8). (B) Summary dot plots showing TCRγδ+ cells isolated from healthy human breast tissue, expressed as a percentage of recovered CD3+ cells (n = 29) (median indicated). In a subset of these, Vδ chain usage was quantified and expressed as percentage of pan-TCRγδ+ (n = 18) (medians indicated). (C) Expression of cell surface markers NKG2D, CD28, PD-1, CD103, and CD69 on Vδ1+ T cells (n = 9 to 11) (medians indicated). (D) Functional phenotype of tissue-resident Vδ1+ T cells. Dot plots showing intracellular cytokine staining for IFN-γ (n = 12), IL-13 (n = 8), IL-17A (n = 9), TNF (n = 10), and cell surface CD107a (n = 5) expression, after in vitro stimulation of bulk CD3+ cultures with PMA and ionomycin (4 hours) (medians indicated). (E) Summary data showing the percentages of breast-resident Vδ1+ or CD4 αβ T cells stained intracellularly for IL-17A. Cells were isolated by explant culture and then grown in two forms of IL-17–skewing media, followed by in vitro activation with PMA and ionomycin (4 hours) (n = 3, except for Vδ1+ cells grown in TGF-β–containing medium, where n = 2) (mean with SEM indicated).

As is common for peripheral blood (PB) γδ T cells (39), γδ T cell representation in breast showed considerable interindividual variation (Fig. 1B). However, breast-extractable cells were clearly distinct from PB γδ T cells by TCRVδ chain usage. Whereas most PB cells express Vδ2 paired to Vγ9, most donors’ breast γδ T cells were predominantly Vδ1+ (median, 60.5% of γδ T cells), as are most human skin-resident and gut-resident γδ T cells (Fig. 1B) (40). In all cases, there was some representation of Vδ2+ T cells (median, 13.6% of γδ T cells) and of cells expressing neither Vδ1 nor Vδ2 (median, 20.45% of γδ T cells), which, in some cases, were almost exclusively Vδ3+ (fig. S1A).

The phenotypic consistency of γδ T cells in grid cultures and counterpart cells examined directly ex vivo was apparent from the expression patterns of several biologically important surface markers, albeit CD69 was expressed by more cells in grid culture (fig. S1, B and C, top two rows). By multiparameter analysis of a subset of donors, we could deduce a consensus Vδ1+ T cell phenotype that resembled that of extralymphoid γδ T cell subsets in other tissues of mice and humans (4, 4144), namely, uniform positivity for the activating NK cell receptor NKG2D, and for CD69, and largely lacking the lymphoid T cell costimulator, CD28 (Fig. 1C and fig. S1C, bottom row). On average, ~20% of breast Vδ1+ T cells expressed PD-1, whereas slightly more expressed the epithelial interaction integrin, CD103 (αEβ7), albeit with high interindividual variation (Fig. 1C and fig. S1C, bottom row).

To assess the cells’ functional potential, they were incubated with phorbol 12-myristate 13-acetate (PMA) and ionomycin, which jointly mimic TCR signaling, and analyzed for intracellular cytokine production and for surface expression of CD107a, a marker of degranulation and exocytosis of cytolytic mediators such as granzymes and perforin. Breast-associated Vδ1+ T cells combined CD107a expression with tumor necrosis factor (TNF) and interferon-γ (IFN-γ) production (Fig. 1D), a T-cytolytic type 1 (Tc1) phenotype that among CD8+ TCRαβ+ TILs is considered highly patient beneficial (4548). Cells from some donors expressed interleukin-13 (IL-13) (median, 5.2% of γδ T cells), which was recently linked to T cell tumor surveillance (49), but there was no production of IL-17A, an effector cytokine commonly produced by mouse γδ T cells, in which species it has been associated with tumor promotion (5052).

Because IL-17 production by human γδ T cells is reportedly difficult to observe (53), we tested whether breast-resident γδ T cells would respond in culture to IL-17–skewing conditions, namely, IL-1β, IL-6, IL-23, IL-2 ± transforming growth factor–β (TGF-β) (54). However, whereas breast-resident CD4+ αβ T cells extracted and maintained together with Vδ1+ T cells in the identical breast explant cultures (table S1) produced IL-17A [median, 16.5% (using IL-2) and 12.6% (using IL-2 + IL-15) of CD4+ T cells] and markedly increased production in IL-17–skewing conditions (medians, 41.5 and 35.5%, respectively) (Fig. 1E), breast Vδ1+ T cells produced negligible IL-17A under all circumstances (Fig. 1E).

To place the phenotypes of breast Vδ1+ T cells into context, similar analyses were performed on coextracted Vδ2+ and Vδ1Vδ2 γδ T cells, on CD4+ and CD8+ αβ T cells, and on CD3 lymphocytes that will include NK and ILC. The greatest similarities to Vδ1+ T cells were shown by Vδ2+ and Vδ1Vδ2 γδ cells and by CD8+ αβ T cells, although some such cells expressed CD28, and CD8+ αβ T cells were more uniformly CD103+ (fig. S1D). NK and ILC were also similar to Vδ1+ T cells except that they uniformly lacked PD-1. Last, and as anticipated, CD4+ αβ T cells lacked NKG2D expression and were mostly CD28+ and CD103 (fig. S1D), consistent with a recently described tissue-resident CD4+ phenotype (55).

Upon activation, breast-associated CD8+ αβ T cells were functionally most similar to Vδ1+ T cells in their Tc1 phenotype, showing an even greater frequency of IFN-γ–producing cells (fig. S1E). Likewise, breast-explanted CD4+ αβ T cells included more IL-13 producers than did Vδ1+ T cells. In sum, the healthy breast harbored a complex lymphoid ecosystem of multiple cell types with related but distinct phenotypes.

Human breast Vδ1+ T cells are innate-like

Murine skin-resident γδ T cells can respond in vivo to NKG2D ligand up-regulation without overt TCR stimulation (5). In relation to cancer, this is potentially important because NKG2D ligands are up-regulated by DNA damage (8) and EGFR overactivity (9). We therefore investigated whether breast-explanted Vδ1+ T cells would respond to plate-bound recombinant MICA protein, an NKG2D ligand commonly expressed by tumors. MICA provoked a subset of Vδ1+ T cells to produce TNF and IFN-γ and to up-regulate CD107a in a response inhibited by anti-NKG2D (Fig. 2A). Conversely, CD8+ αβ T cells within the identical grid cultures did not make significant responses to MICA relative to controls (Fig. 2A and fig. S2A), whereas both cell types showed increased responses when MICA was provided as a costimulator to anti-CD3 (Fig. 2B) (56, 57). The innate-like responsiveness of other breast γδ T cells was challenging to examine because even in cases where they composed a greater proportion of tissue γδ T cells, non-Vδ1+ cells were usually too few to assay reliably (fig. S2B). Despite this, we observed some NKG2D-dependent, innate-like responses among Vδ1 Vδ2 T cells but not among breast Vδ2+ T cells (fig. S2B).

Fig. 2 Breast tissue–resident Vδ1+ T cells are innate-like.

(A) Summary data showing intracellular staining for IFN-γ (n = 19 for Vδ1+ and n = 15 for CD8+), TNF (n = 17 for Vδ1+ and n = 13 for CD8+), and cell surface CD107a (n = 10 for Vδ1+ and n = 5 for CD8+) expression, after in vitro activation of breast-resident Vδ1+ cells or of Vδ1+ and CD8+ αβ T cells from within the same cultures, exposed to plate-bound recombinant MICA (10 μg/ml) in the presence of brefeldin A (BFA). Plotted as percentage of parent Vδ1+ or CD8+ gate. (B) Summary data showing intracellular staining for IFN-γ (n = 6 for Vδ1+ and n = 6 for CD8+) and TNF (n = 5 for Vδ1+ and n = 5 for CD8+) after in vitro activation of breast tissue–resident Vδ1+ and CD8+ αβ T cells with low-dose plate-bound anti-CD3 antibody (50 ng/ml) with or without plate-bound recombinant MICA (10 μg/ml) in the presence of BFA. Where indicated, MICA-stimulated cells were pretreated with anti-human NKG2D antibody (plotted as percentage of parent Vδ1+ or CD8+ gate). (C) Summary data for breast-resident Vδ1+ T cells, showing intracellular IFN-γ production after in vitro activation with IL-12 (n = 3) or IL-18 (n = 3) or IL-12 + IL-18 (n = 9) and with medium alone (n = 9). For all panels, mean with SEM is indicated. **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001, Kruskal-Wallis with post hoc Dunn’s test corrected for multiple testing.

It was also reported that innate-like γδ T cells make strong, TCR-independent responses to combinations of a STAT (signal transducer and activator of transcription)–signaling cytokine and an IL-1 family member (58). Breast-explanted Vδ1+ T cells produced IFN-γ in response to IL-12 + IL-18 but not to either alone, whereas CD4+ and CD8+ αβ T cells responded significantly less well to IL-12 + IL-18 (fig. S2C). In sum, the healthy breast harbored a mixture of innate-like Vδ1+ T cells and primarily adaptive αβ T cells.

Innate-like γδ T cells in human breast cancers

The identification of innate-like γδ T cells in healthy human breast formed a backdrop to examining the tissue-associated lymphoid compartment in breast cancer subjects. Although breast cancers vary in lymphoid infiltrates (32, 59), γδ T cells were invariably recovered and were largely comparable to those from healthy tissue in terms of TCR usage: Vδ1 was predominant, although the tumor samples included some examples where either Vδ2+ cells or Vδ1Vδ2 cells predominated (fig. S3A). As with γδ T cells from healthy breast, cells isolated from tumors using grids were phenotypically comparable with those examined directly ex vivo, albeit grid cells again showed higher expression of CD69 and, to some extent, NKG2D (fig. S3B). The similarities of TCR usage and surface marker expression for γδ T cells from tumor and nonmalignant tissue were particularly apparent in paired samples from 26 subjects (Fig. 3A and fig. S3C). Moreover, this comparability extended to other lymphocyte subsets simultaneously harvested from the paired samples (fig. S3D).

Fig. 3 Innate-like Vδ1+ Tc1 cells are a major subset of breast TILs.

(A) γδ T cells were consistently isolated from human breast tumor samples obtained from mastectomies after grid culture. Dot plots showing numbers of TCRγδ+ cells isolated from paired nonmalignant breast tissue (•) or tumor (ο) expressed as a percentage of CD3+ T cells. γδ T cells were further phenotyped for Vδ1+, Vδ2+, and Vδ1Vδ2 TCR usage expressed as a percentage of TCRγδ+ T cells (n = 25) (medians indicated). (B) Breast tumor–infiltrating Vδ1+ T cells are functionally skewed. Dot plots showing intracellular cytokine staining for IFN-γ (n = 12), IL-13 (n = 11), IL-17A (n = 4), TNF (n = 9), and cell surface CD107a expression (n = 4), after in vitro activation with PMA (10 ng/ml) and ionomycin (1 μg/ml) in the presence of BFA (medians indicated). (C) Summary data showing intracellular staining for IFN-γ (n = 9 to 12, depending on activating condition), TNF (n = 7), and cell surface CD107a (n = 4 to 7, depending on activating condition) expression after in vitro activation of Vδ1+ and CD8+ αβ T cells with plate-bound recombinant MICA (10 μg/ml) in the presence of BFA or in vitro activation with IL-12 (100 ng/ml) and IL-18 (100 ng/ml). Where indicated, MICA-stimulated cells were pretreated with blocking anti-human NKG2D antibody (mean with SEM indicated). (D) Tumor cell lines, MCF7 and HCC1954, were incubated with negatively sorted γδ T cells (γδ) derived from nonmalignant breast tissue or breast tumor at E:T of 5:1, in the presence (γδ + αNKG2D) or absence (γδ) of a blocking anti-NKG2D antibody for 48 hours. Cell lines without effector γδ T cells were used as negative controls (Control). Dot plots show concentrations of caspase-cleaved cytokeratin 18 (cCK18). Each data point represents the mean of two technical replicates; the median values for those data points are indicated by a horizontal line (note that there were only three donors for the killing assay of HCC1954 cells by tumor-derived γδ T cells in the presence of anti-NKG2D). △ and ○ are two donors for which there were paired nonmalignant breast tissue and breast tumor. *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001, Kruskal-Wallis with post hoc Dunn’s test corrected for multiple testing.

Likewise, the functional potential of Vδ1+ TILs was comparable to those of nonmalignant breast Vδ1+ T cells, in being Tc1 skewed and IL-17 deficient (Fig. 3B). Vδ1+ TILs were again responsive to NKG2D ligands and IL-12 + IL-18 in the absence of overt γδ TCR signaling, whereas co-isolated CD8+ αβ TILs did not show significant responses to MICA relative to controls, although they did respond better to IL-12 + IL-18 than did counterparts from healthy breast (Fig. 3C).

Given their strong cytolytic responsiveness, we tested the capacity of breast-derived γδ T cells to kill two breast tumor cell lines, MCF7 and HCC1954, for which tumor cell lysis was distinguished from lymphocyte death by quantitating cytokeratin 18 release (60). Note that because γδ T cells are not MHC (major histocompatibility complex)–restricted, it was possible to assess their functional responses to nonautologous tumor lines. γδ T cells from healthy breast donors (n = 4) reproducibly killed MCF7 cells at an effector:target (E:T) ratio of 5:1 (fig. S3E), and using this ratio, we found that cells from healthy breast and from tumor samples (n = 4) showed comparable capacity to kill MCF7 and HCC1954 (Fig. 3D). However, whereas NKG2D receptor blockade reduced killing by γδ T cells from healthy breast, killing by TCRγδ+ TILs was less affected (Fig. 3D). We also observed primary tumor cell killing by autologous TCRγδ+ TILs for the one patient from whom primary tumor cells could be grown and stably maintained (fig. S3F). In sum, primary γδ T cells obtained from breast cancers were functionally competent, could respond innately via NKG2D engagement, and could lyse breast tumor cells, albeit this was not overtly NKG2D dependent.

Vδ1+ TILs and durable remission

Given the functional Tc1 skew of γδ TILs, we wished to examine their status in relation to clinical outcome in an aggressive subset of breast cancer where time to events is relatively short. To this end, we sought patients with triple-negative breast cancer (TNBC) from whom we could access sufficient paired tumor and nonmalignant tissues and for whom accurate clinical follow-up data were available. Those criteria were met by 11 patients treated at Guy’s and St Thomas’ Hospitals, London (BTBC study REC no.: 13/LO/1248), for whom there were bio-banked, formalin-fixed paraffin-embedded (FFPE) samples. Patients were otherwise unselected. All patients had localized or locoregional TNBC [American Joint Committee on Cancer (AJCC) stage I to III] and had surgery with curative intent. Five of 11 remained in complete remission at last follow-up (range, 48 to 63 months; table S2; demarcated in blue in the figures that follow), whereas six had relapsed with distant metastatic disease within 18 months (range, 6 to 18 months; table S2; demarcated in red in the figures that follow).

FFPE blocks were needle-dissected to delineate tumor and nonmalignant tissue for genomic DNA extraction. Given the difficulty of immunohistochemical approaches in detecting γδ T cells in FFPE (61), we used quantitative genomic DNA sequencing of rearranged TCRα and TCRδ chain genes to infer absolute counts of αβ and γδ T cells. This approach has been shown to be more sensitive than immunohistochemistry for detecting TILs (29) and is used clinically, for example, to assess minimal residual disease.

We found that both αβ and γδ T cells were significantly more abundant per microgram of DNA extracted from tumor tissue versus paired nonmalignant tissue (Fig. 4A and fig. S4A). However, it was notable that in cases of remission, the numbers of TCRα, TCRγδ, TCRVδ1, and TCRVδ2 DNAs were invariably enriched in paired malignant versus nonmalignant tissue, whereas the pattern in relapsed cases was essentially random (fig. S4A). Moreover, in addition to enrichment relative to healthy tissue, there were conspicuously more TCRα+ and Vδ1+ TCRs per microgram of total tumor DNA in remission cases versus relapse (Fig. 4B). Thus, in indicating positive clinical outcome, the dynamics of small numbers of Vδ1+ T cells were as potent as the much larger-scale dynamics of αβ T cells, defined subsets of which have been shown to predict survival in TNBC (62). Conversely, this was not so for either total TCRγδ or Vδ2 TCRs (Fig. 4B), the latter possibly reflective of cells infiltrating from the PB.

Fig. 4 Vδ1+ T cells in TNBC are predictive of disease-free survival and OS.

(A) Overall landscape of T cell subsets in nonmalignant breast tissue (“Tissue”) and matched tumor tissue (“Tumor”), determined by quantitative sequencing of rearranged TCR genes from patients undergoing mastectomy for TNBC. Absolute TCR copies (a surrogate for T cell numbers) are plotted per microgram of input DNA. Individual patients plotted: red, patients with relapsed disease; blue, patients in remission (median bar plotted). *P < 0.05 and **P ≤ 0.01, Wilcoxon matched pairs signed rank test. (B) Intratumoral presence of αβ T cells and γδ T cells in patients with subsequent relapsed disease versus those who remained in remission. *P < 0.05, Kolmogorov-Smirnov test. (C) PFS (months from surgery) split on median T cell subsets found in 11 TNBC tumors. *P < 0.05 and **P ≤ 0.01, Gehan-Breslow-Wilcoxon test. (D) OS (months from surgery) split on median T cell subsets found in 11 TNBC tumors. *P < 0.05 and **P ≤ 0.01, Gehan-Breslow-Wilcoxon test.

Notable manifestations of the correlations were evident from Kaplan-Meier plots of progression-free survival (PFS), where limited representation (less than median values) of intratumoral TCRα and Vδ1 TCRs was predictive of poor PFS, whereas neither total TCRγδ nor Vδ2+ TCRs predicted outcome (Fig. 4C). Furthermore, there was a positive and significant correlation of intratumor TCRα with Vδ1 TCRs (Spearman r = 0.75, P ≤ 0.01) (fig. S4B), whereas neither total TCRγδ nor Vδ2 TCRs correlated with TCRα: Moreover, no TCRα-TCRδ correlation existed in nonmalignant tissue (fig. S4B). In addition, Vδ1 TCRs predicted overall survival (OS), although TCRα did not (Fig. 4D).

In situ evidence of innate-like Vδ1+ TILs

Having established significant positive correlates of TCRαβ+ T cells with PFS and of Vδ1+ T cells with PFS and OS, we assessed their TCR repertoires. When represented as circular tree plots, the Vδ1+ repertoires in tumors compared to paired nonmalignant tissue showed no clear overall focusing (examples shown in Fig. 5A), as quantitated by D50 (the smallest number of clones accounting for 50% of the total number of sequences observed) from paired nonmalignant and tumor tissues (table S3). Focusing would have suggested an adaptive TCR-driven response, as we observed for the TCRα repertoires of most tumors versus paired normal tissue (examples shown in Fig. 5B), consistent with previous reports (62, 63).

Fig. 5 Lack of tumor Vδ1+ TCR focusing relative to focusing of TCRα sequences.

(A) Examples of circular tree plots of the Vδ1+ repertoire in paired nonmalignant tissue (“tissue”) and tumor tissue (“tumor”), where each circle represents a unique TCR clonotype and the size of the circle is proportional to the representation of the specified clone. Plots were generated from total Vδ1+ TCRs. (B) Examples of circular tree plots of the TCRα repertoire in paired nonmalignant tissue (“tissue”) and tumor tissue (“tumor”). Plots were generated from total TCRα+ TCRs. (C) Vδ1+ and TCRα+ sequences in nonmalignant and tumor tissue were down-sampled (within each patient) to equivalent numbers to calculate diversity metrics. The degree of repertoire focusing was assessed by the delta of the Gini coefficient and the delta D50 of sequences of Vδ1 chains and the top 10% in abundance of TCRα sequences in tumor versus paired nonmalignant repertoires. To test whether the degree of repertoire focusing was different between Vδ1+ and TCRα+ compartments in individual patients, the Wilcoxon matched pairs signed-rank test was used to compare ΔGini and ΔD50. All sequences were analyzed on the basis of amino acid sequence. n = 11.

To further analyze the data, we also calculated the Gini coefficient (a statistical measure of distribution where 0 is fully polyclonal and 1 is monoclonal) for Vδ1 TCRs from paired nonmalignant and tumor tissues (table S3). Note that the tumor TCRs were down-sampled so that equivalent numbers of TCRs were compared within each patient. The same treatment was then applied to TCRα, although we considered only the most abundant 10% of TCRα TCRs, given recent evidence that the most relevant antigen-reactive αβ T cells commonly sit within this fraction (64). For each patient, we then calculated the delta of the Gini coefficient of paired tumor tissues and nonmalignant tissues for both Vδ1 and TCRα: Note that tumor focusing would be reflected by positive ΔGini coefficient (Fig. 5C). We then likewise calculated the delta of the D50 values for paired tumor and nonmalignant tissues for both Vδ1 and TCRαβ: Note that tumor focusing would be reflected by negative ΔD50 (Fig. 5C). These analyses confirmed quantitatively that TCRα repertoire focusing occurred in all but two tumor samples (KCL-059 and KCL-202), whereas Vδ1 repertoires showed no bias either toward focusing or toward diversification (Fig. 5C).

As a complementary approach, we also applied repertoire metrics to non–down-sampled (raw) TCR reads, using normalized measures of clonality (normalized Shannon entropy, Gini coefficient, and D50), as previously used by others (29, 63). These methods also suggested tumor repertoire focusing for TCRα (significance was reached for D50), whereas there was no such finding for Vδ1 (fig. S5 and table S4). Collectively, these data strongly suggest that the Vδ1+ cell responses were not driven by clonotypic antigens.

Given that, among TCR gene rearrangements, TCRδ harbors the highest potential for junctional diversity (65), it was not surprising that no public Vδ1 sequences were observed across different donors’ tumors (Fig. 6A). However, there was some Vδ1 sequence overlap between tumors and tissue from the same donor (fig. S6). Although the sample size was small, the lack of public sequences would also be consistent with TCR-agnostic, innate-like regulation of tissue-resident Vδ1+ T cells. By contrast, assessment of a comparably sized sample showed that some Vδ2 TCRs were shared across multiple donors (Fig. 6B). Most shared sequences reflected TCRs reactive to phospho-antigens that can be up-regulated in tumors (66), with a conserved hydrophobic residue in position 97 (table S5) specifically associated with phospho-antigen–mediated selection of the Vδ2+ repertoire (67, 68).

Fig. 6 No detectable public intratumoral Vδ1+ clonotypes.

(A) Intersections of Vδ1+ clonotypes between 11 patient tumor samples. Vertical bars represent the number of unique TCRs and the dot matrix represents sharing of TCRs across patients. A shared or public clonotype would be represented by at least two red dots (sharing between two patients) joined by a vertical red line. Private sequences are presented by an unconnected single red dot. (B) Intersections of Vδ2+ clonotypes between 11 patient tumor samples. All sequences were analyzed on the basis of amino acid sequence.

DISCUSSION

The past decade has witnessed a sea change in cancer immunology, with the realization that tumors are often antigenic and that tumor-reactive T cells can provide patient-beneficial responses, particularly if derepressed by checkpoint blockade (69, 70). Hence, there is considerable interest in the immune ecology of tumors. Over the same period, it became clear that several extralymphoid tissues in which tumors commonly form ordinarily harbor large myeloid and lymphocyte compartments, including γδ T cells that become tissue resident during the cells’ development and systemic αβ T cells that become TRM cells after priming in secondary lymphoid organs (71, 72).

Although positive clinical outcomes in TNBC have been associated with CD8+ TCRαβ+ TRM cells (62), there has been little investigation of whether a human breast-resident γδ T cell compartment exists that might influence breast cancer outcomes (73). This study addresses this point by first establishing a tissue-resident γδ T cell compartment in healthy human mammary tissue. This may be evolutionarily conserved because TCRγδ+ lymphocytes were reported in alveolar mammary epithelia of lactating cows (74) and were associated with lactating mammary glands in mice (26). To characterize human breast-resident γδ T cells in sufficient numbers, we used grid cultures previously used to elucidate key features of human skin and lung T cells (3538). Although this is a limitation, there was strong phenotypic consistency with breast γδ T cells examined directly ex vivo. This permitted our description of the compartment as mostly Vδ1+, NKG2D+, CD69+, partly CD103+, and with a Tc1 phenotype lacking IL-17 production. This is very distinct from PB γδ T cells but shares features with human intestinal epithelial γδ T cells (40).

In addition, human breast γδ T cells were innately responsive to NKG2D activators, whereas colocated αβ T cells required coincident TCR stimulation. Thus, healthy human breast Vδ1+ cells have an inherent potential to detect and respond rapidly to local cells en route to malignancy. Some of these signature properties may be shared with local Vδ1δ2 T cells, which are often Vδ3+, although these cells were most often present in very small numbers.

Within breast tumors, Vδ1+ T cells and αβ T cells were frequently more abundant than in paired healthy tissue, particularly in patients in remission. This likely reflects an inflexion point at which an activating immune response to the tumors occurred. Furthermore, when extracted from breast tumors, the comparatively expanded Vδ1+ and αβ T cell populations were functionally competent, retaining the innate responsiveness and Tc1 potential of cells from healthy breast. These observations evoked evidence that PD-1+ breast cancer TILs responded functionally to restimulation (75) and that their presence could be associated with favorable outcome (62). Although only few patients were available for in-depth analysis in this study, they were sufficient to show significant positive correlations of tumor-derived Vδ1+ and αβ T cells with clinical outcome, with Vδ1 TCRs correlating with both PFS and OS. It is therefore attractive to hypothesize that maximum patient benefit accrues from a collaboration of the innate responsiveness of local Vδ1+ cells with the antigen-specific modus operandi of αβ T cells, particularly CD8+ TCRαβ+ TRM cells (62).

At least two patient-beneficial facets of collaboration between Vδ1+ and TCRαβ+ T cells may be envisioned. First, by recognizing tumors via innate stimuli, Vδ1+ T cell responses may not be limited either by the number of neoantigen-generating somatic mutations or by immune-evasive suppression of peptide antigen presentation (47, 7679). The innate stimuli may include ligands for several NK receptors (43), including but not limited to NKG2D.

Second, the cytolytic, Tc1 phenotype of Vδ1+ cells may be augmented by the cells’ capacity to promote tissue immunogenicity via chemokine secretion and possibly via direct antigen presentation (16, 17, 80). A critical role of tissue-resident Vδ1+ cells may be to orchestrate multicomponent local immune responses to defined challenges while remaining tolerant to others. In this context, tissue/tumor immunogenicity might be effectively enhanced in the clinic by agents promoting the activities of local Vδ1+ T cells, or by the cells’ adoptive transfer, in concert with the activation/derepression of adaptive T cells. Because γδ T cells are not MHC-restricted, they might be adoptively transferred from heterologous donors, and because they are naturally tissue resident, local Vδ1+ cells may cope with hypoxic environments that prove hostile to systemic lymphocytes (81, 82).

Such considerations may pertain to other human tissues harboring local γδ T cell compartments, such as the gut, and may underpin reportedly strong correlations of γδ T cells with favorable clinical outcomes across a broad spectrum of human tumors (83). Nonetheless, in this study, neither total γδ TCRs nor Vδ2+ TCRs correlated with clinical outcome. This emphasizes the fact that γδ T cells comprise biologically distinct subsets, as is the case for αβ T cells or ILCs. Even among intratumoral CD8+ T cells, which are traditionally associated with patient benefit, most benefit in TNBC was attributable to a discrete subset of local CD8+ TRM cells (62). In mice, functionally different γδ T cell subsets have been reported to either mediate or repress tumor immunosurveillance (84, 85). Most often, IL-17 has been implicated in tumor promotion by γδ T cells (5052), whereas IL-17–producing cells are seemingly rare in humans, wherein the predominant phenotype is cytolytic and TNF/IFN-γ producing (86), as described here.

In mice, the innate responsiveness of γδ T cells and their suppression of IL-17 production were induced developmentally by subset-specific, tissue-specific selecting elements of the butyrophilin-like (Btnl/BTNL) family, members of which can also regulate human γδ T cells (30, 40, 58, 87). It is therefore possible that such elements act locally to select and regulate human breast-resident Vδ1+ T cells, in which regard the mammary gland is one of reportedly few tissues expressing BTNL9 (88).

Our study did not focus on interactions of breast Vδ1+ T cells with other breast-resident immune cells including B cells (32). Likewise, spatial relationships between breast γδ T cells and tumor-associated tertiary lymphoid structures were not determined (31, 32). In practical terms, clinical studies have suggested that human breast cancer, including TNBC, can be vulnerable to immune attack (33, 34), yet the efficacy of immunotherapies in this indication has been relatively poor. We strongly believe that this may be redressed by shifting therapies away from their unique focus on conventional, adaptive T cell responses and by learning from the natural ecology of the local breast T cell compartment. In particular, we believe that this may promote the immunogenicity of tumor tissues that drives and sustains patient-beneficial adaptive responses.

MATERIALS AND METHODS

Study design

The aim of this study was to ascertain whether the human breast might contain a tissue-resident γδ T cell compartment and to determine whether this might be protective in breast cancers. This study was undertaken first by demonstrating the reproducible presence of γδ T cells in healthy breast tissue with a particular focus on Vδ chain usage. Then, we determined their functional potential and whether they might be consistent with protective tumor immunosurveillance. These approaches were subsequently applied to γδ T cells in breast tumors. Colocated αβ T cells isolated using the same protocol and maintained in the same culture conditions were used as controls. Having established the presence of γδ T cells in human breast tissue and tumors and their innate-like immunosurveillance capacity in vitro, we examined the presence of these cells in situ (via their DNA rearrangements) and correlated this to prognosis in clinical samples. We also sought in situ evidence, particularly TCR repertoire clonality, for cells functioning in an innate-like immunosurveillance capacity, as was established ex vivo. Detailed study design, sample sizes, replicates, and inclusion/exclusion criteria are provided in the figure legends or in Materials and Methods. The sample sizes and experimental repetitions were sufficient to permit rigorous statistical analysis as described in the figure legends and Materials and Methods. All antibodies and key reagents are listed in table S6. Primary data are reported in data file S1.

Clinical material

Human breast samples were obtained from adult female patients undergoing breast reduction or risk-reducing mastectomy (29 patients) or breast tumor resection (90 patients) after informed consent as part of a noninterventional clinical trial (BTBC study REC no.: 13/LO/1248, IRAS ID 131133; principal investigator: A.T; study title: “Analysis of functional immune cell stroma and malignant cell interactions in breast cancer in order to discover and develop diagnostics and therapies in breast cancer subtypes”). This study had local research ethics committee approval and was conducted adhering to the principles of the Declaration of Helsinki. Specimens were collected from surgery into sterile saline and transported immediately to cut up. A histopathologist or pathology-trained technician identified and collected tumor material and ipsilateral nonadjacent normal breast tissue from surgical specimens from which lymphocytes were subsequently isolated as described below. Demographics of the patients are detailed in table S2. In addition to the patients above, for TCR sequencing experiments, tumor and paired nonmalignant tissue DNA was extracted from bio-banked FFPE blocks from 11 patients with AJCC stage I to III TNBC who had mastectomies for which we could access sufficient material and accurate clinical follow-up data. No other criteria were applied. The 11 cases were also part of the BTBC clinical trial described above.

Primary lymphocyte extraction and culture

For direct ex vivo isolation, fresh breast tumor or tissue was coarsely minced with scalpels and then dissociated using the MACS human tumor dissociation kit on a gentleMACS dissociator as per the manufacturer’s instructions (Miltenyi Biotec). Samples were washed twice with sterile RPMI 1640 and used immediately for downstream assays. Lymphocytes were also harvested using a grid explant system adapted from a protocol first described by Clark and colleagues (35). Briefly, fresh breast tumor or tissue was minced using scalpels and placed onto rat tail collagen (100 μg/ml; BD Biosciences)–coated Cellfoam grids (Cytomatrix Pty Ltd.). Each grid was placed into a separate well of a 24-well tissue culture plate and cultured in complete medium [Iscove’s modified Dulbecco’s medium (Life Technologies), 10% heat-inactivated fetal bovine serum (FBS; Life Technologies), l-glutamine (292 μg/ml; Life Technologies), penicillin (100 U/ml; Life Technologies), streptomycin (100 μg/ml; Life Technologies), and 2-mercaptoethanol (3.5 μl/liter; Life Technologies)] supplemented with recombinant human IL-2 (rhIL-2) (100 IU/ml; Proleukin; Novartis Pharmaceuticals UK Ltd.) and rhIL-15 (10 ng/ml; BioLegend). The grids were maintained for 3 weeks in culture at 37°C/5% CO2, and the lymphocytes were harvested by washing the wells/grids with 0.01 mM Hepes/Hanks’ balanced salt solution.

Flow cytometry and fluorescence-activated cell sorting

Cells were washed in sterile phosphate-buffered saline (PBS) to remove traces of serum and stained for 20 min at room temperature with LIVE/DEAD Fixable Aqua (Thermo Fisher Scientific) in PBS. Subsequent surface staining was carried out in fluorescence-activated cell sorting (FACS) buffer (PBS, 2% fetal calf serum, and 1 mM EDTA) for 20 min at 4°C (see table S6) before washing twice with FACS buffer and fixing with CellFIX (BD) for 10 min at room temperature. For intracellular cytokine staining, fixed cells were washed twice with Intracellular Staining Permeabilization Wash Buffer (BioLegend) and stained for 20 min at room temperature before two further washes with Intracellular Staining Permeabilization Wash Buffer. Samples were acquired on a BD FACSCanto II or BD LSRFortessa and were analyzed using FlowJo software (FlowJo LLC). For FACS, cells were not fixed and sorted on a BD FACSAria II as detailed below. Antibodies are listed in table S6 and were used at 1:50 dilution unless otherwise specified.

In vitro lymphocyte activation assays

Directly isolated and grid explant–isolated lymphocytes were stimulated with PMA (10 ng/ml; Sigma) and ionomycin (1 μg/ml; Sigma) in the presence of brefeldin A (BFA; 20 μg/ml; Sigma) for 4 hours at 37°C/5% CO2 before surface marker and intracellular cytokine staining and acquisition on a BD FACSCanto or Fortessa. For plate-bound NKG2D ligand assays, lymphocytes were harvested from explant cultures 24 hours before activation and rested in complete media without cytokine supplementation. After resting, lymphocytes were transferred to 96-well flat-bottom cell culture plates (Corning) coated with rhMICA (10 μg/ml; R&D Systems), anti-CD3 (50 ng/ml; BioLegend), anti-CD3 (50 ng/ml; BioLegend) and rhMICA (10 μg/ml; R&D Systems), or mouse immunoglobulin G2a (IgG2a) (50 ng/ml; BioLegend) at 100,000 cells per well in 100 μl of complete medium. Plates were incubated for 6 hours at 37°C/5% CO2 in the presence of IL-15 (10 ng/ml; BioLegend) and BFA (20 μg/ml; Sigma). Where CD107a was used as a functional readout, anti-human CD107a antibody (1:400 final concentration; BioLegend) was also added at the start of the assay along with monensin at 1× (BioLegend). For NKG2D-blocking conditions, anti-NKG2D antibody (10 μg/ml; clone 1D11; BioLegend) was added to lymphocytes just before plating.

For cytokine activation assays, lymphocytes were incubated with IL-12 (100 ng/ml; PeproTech) and/or IL-18 (100 ng/ml; Medical and Biological Laboratories) for a total of 24 hours at 37°C/5% CO2 with BFA (20 μg/ml; Sigma) added for the last 4 hours before surface and intracellular cytokine staining for flow cytometry. For IL-17–skewing assays, breast tissue explants were cultured in complete medium and in conditions as described above with the addition of rhIL-2 ± rhIL-15, rhIL-1β, rhIL-6, rhIL-23, and rhTGF-β for 3 days. These cells were then activated with PMA and ionomycin in the presence of BFA (20 μg/ml) for 4 hours at 37°C/5% CO2 before surface and intracellular cytokine staining for flow cytometry.

Cytotoxicity assays

Grid explant–derived γδ T cells from breast tissues and tumors were isolated by FACS via depletion of TCRαβ+ and NKp46+ cells. Target cells, MCF7 and HCC1954 [Cancer Research UK (CRUK)/Francis Crick Institute Cell Service], were seeded at 10,000 cells per well in 96-well flat-bottom plates (Corning) 24 hours prior. Fifty thousand negatively sorted γδ T cells were added to target cells in the presence or absence of blocking NKG2D antibody (10 μg/ml; BioLegend). Cells were incubated at 37°C for 48 hours, after which supernatants were collected and stored at −20°C until further analysis. Target cell apoptosis was measured using enzyme-linked immunosorbent assay (ELISA) for the epithelial cell–specific caspase-cleaved cytokeratin 18 (Diapharma), according to the manufacturer’s instructions.

Cell lines and culture conditions

Target MCF7 and HCC1954 cell lines were sourced from CRUK Cell Services (Clare Hall, London) and maintained in Dulbecco’s modified Eagle’s medium (Life Technologies), supplemented with 10% heat-inactivated FBS (Life Technologies), penicillin (100 U/ml; Life Technologies), and streptomycin (100 μg/ml; Life Technologies) at 37°C/5% CO2. HCC1954 cells were maintained in RPMI 1640 (Gibco) supplemented with 10% heat-inactivated FBS (Life Technologies), penicillin (100 U/ml; Life Technologies), and streptomycin (100 μg/ml; Life Technologies) at 37°C/5% CO2.

DNA extraction

DNA was extracted from FFPE paired tumor tissue and normal tissue blocks of 11 patients with TNBC treated with mastectomy as part of the BTBC study. Tumor tissue was needle microdissected after sectioning. The QIAamp DNA FFPE Tissue kit (Qiagen) was used per the manufacturer’s instructions to extract DNA. DNA was quantified using a Qubit fluorometer, and material from patients with >1 μg of DNA from both tumor and normal tissue was sent for quantitative TCRα/δ locus sequencing by Adaptive Biotechnologies.

TCR sequencing

TCR sequencing was performed by Adaptive Biotechnologies. The Adaptive Biotechnologies platform uses genomic DNA and can quantitate T cell numbers. Reads were aligned and annotated by Adaptive Biotechnologies, and data were output as .csv files (available from https://osf.io/d4eu6/) for downstream analysis from the immunoSEQ Analyzer (https://clients.adaptivebiotech.com/login). Output was filtered on in-frame CDR3s as well as TCRA to TCRA V-J family joins for TCRαβ and TCRD to TCRD V-J family joins for TCRγδ T cells. We normalized the absolute counts of TCRs to 1 μg of input DNA for each sample to enable normalized comparison of infiltrating T cell numbers across all samples. All analyses were carried out using CDR3 amino acid sequences as opposed to nucleotide sequences.

Clonal repertoire analysis

To compare clonality metrics within each patient between paired tumor and nonmalignant tissue, we down-sampled TCRs from each pair of samples. For Vδ1, TCRs from tumor and nonmalignant tissue were down-sampled to the number of clones in the smaller sample with probability of drawing a clonotype equal to its frequency in the full sample. Down-sampling with replacement was performed 200 times. For TCRαβ, clonotypes were ordered in decreasing frequency, and the top 10% of total TCRs in each pair of tumor and nonmalignant tissue were used for down-sampling as described above. Clonality metrics were then applied to the down-sampled data, and the median values were plotted. As an alternative, we also applied normalized measures of clonality (normalized Shannon entropy, Gini coefficient, and D50) to the raw data. TreeMaps were generated using the Macrofocus TreeMap program (www.treemap.com). We visualized shared clonotype patients using the UpSet R package (https://ieeexplore.ieee.org/abstract/document/6876017 and https://vcg.github.io/upset).

Statistical analysis

Statistical significance was determined by Kolmogorov-Smirnov test, log-rank test, Wilcoxon matched pairs signed-rank test, or Mann-Whitney test, as indicated in the figure legends using Prism 7 software (GraphPad). All findings were considered significant at a P value threshold of <0.05. Where results of statistical test are shown, *P < 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001 unless otherwise indicated.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/11/513/eaax9364/DC1

Fig. S1. Explant culture permitted the isolation of substantial numbers of human tissue-resident lymphocytes.

Fig. S2. Vδ1+ T cells display innate-like responsiveness.

Fig. S3. αβ and γδ T cells could be isolated from breast tumors and phenotypically resemble those from healthy tissue.

Fig. S4. Both αβ and γδ T cells are enriched in tumors compared with paired nonmalignant tissue.

Fig. S5. Vδ1+ T cells show no evidence of tumoral clonal focusing in contrast to αβ T cells.

Fig. S6. There is limited Vδ1 repertoire overlap between tumor and paired nonmalignant tissue within patients.

Table S1. Lymphocyte subtypes observed ex vivo after enzymatic digestion and in grid explant cultures.

Table S2. Clinical features of KCL TNBC cohort.

Table S3. Clonality metrics of down-sampled TCRs.

Table S4. Clonality metrics of raw TCRs.

Table S5. Public intratumoral phospho-antigen reactive Vδ2 CDR3 sequences and samples in which they were shared.

Table S6. Antibodies and key reagents table.

Data file S1. Primary data.

REFERENCES AND NOTES

Acknowledgments: We thank A. Clifford, S. Irshad, and T. Alaguthurai for consenting and recruiting patients and current and former colleagues for helpful advice and discussions. We thank the flow cytometry service units of the Peter Gorer Department of Immunobiology and the Guy’s and St Thomas’ Hospital Trust Biomedical Research Centre (BRC) for outstanding support. Funding: The work was supported by a Wellcome Trust (WT) Investigator Award (106292/Z/14/Z) to A.H. and by the Francis Crick Institute, which receives its core funding from CRUK (FC001093), the UK Medical Research Council (FC001093), and the WT (FC001093). F.K.-C. was funded by a BRC fellowship. Y.W. was funded by a National Institute for Health Research fellowship. R.T.W. was funded by a Medical Research Council (UK) fellowship and a National Institute for Health Research lectureship. A.G. and A.T. were funded by Breast Cancer Now funding at King’s College London. The work was conducted as part of the CRUK Cancer Immune Therapy Accelerator (CITA) and the CRUK City of London Major Centre. Author contributions: F.K.-C., Y.W., R.T.W., and A.H. conceived the study and designed experiments. F.K.-C. and Y.W. performed and analyzed experiments. C.N.-L., J.O., and P.G. processed patient samples and provided technical assistance. A.G., A.L., D.B., and P.V. assisted with data analysis. A.T. recruited patients. F.K.-C., Y.W., and A.H. wrote the manuscript. A.T. and A.H. supervised the study. Competing interests: A.H. is a cofounder and equity holder in GammaDelta Therapeutics PLC. Data and materials availability: All data associated with this study are present in the paper or Supplementary Materials. TCR sequences are deposited at https://osf.io/d4eu6/.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article