Ex vivo enrichment of circulating anti-tumor T cells from both cutaneous and ocular melanoma patients: clinical implications for adoptive cell transfer therapy

To isolate in vitro anti-melanoma T cells, mixed lymphocyte tumor cell cultures (MLTCs) were set up by stimulating PBMCs from melanoma patients with irradiated autologous melanoma cells. Initially, the efficacy of different cytokines, alone or in combination, on the in vitro growth of T lymphocytes was compared. To this aim, independent MLTCs from the cutaneous melanoma patient #2710 were set up in the presence of the following cytokines: (1) rhIL-2; (2) rhIL-2 and rhIL-15; (3) rhIL-15; (4) rh-IL-7; (5) rhIL-2 and rhIL-7; (6) rhIL-2 and rhIL-21; and (7) rhIL-21. At day 21, the highest expansion rate of lymphocytes (105-fold increase of the T-cell number) was obtained with IL-2 alone (Fig. 1). Efficient expansion in vitro of T cells was also obtained by culturing the MLTCs with IL-2 and IL-15 (87.5-fold increase) or with IL-2 + IL-7 (68.4-fold increase) (Fig. 1). Interestingly, similar increases in the number of T cells were observed when performing the MLTC with IL-2 and IL-15, followed for the second week of culture with IL-2 plus IL-7 (data not shown). The MLTC grown either with IL-2 plus IL-21 or with IL-15 alone led to a lower expansion of T cells (35.88 and 13.5-fold increase, respectively) (Fig. 1). No proliferation of T cells was observed in the presence of IL-21 alone (Fig. 1).

To assess the reactivity of these MLTCs against tumor cells, IFN-γ release was determined by ELISPOT. As shown by the representative data in Fig. 2, T lymphocytes maintained in culture with IL-2 and IL-15 released the highest amount of IFN-γ (n. 337 spots/5,000 cells) after the incubation with the autologous tumor cells (#2710 mel) (Fig. 2b) compared to the other in vitro culture conditions (n. 38.5–199 spots/5,000 cells) of MLTCs (Fig. 2a, c, d). Specific inhibition of IFN-γ release was observed after pre-incubation of the autologous tumor cells with the anti-HLA class I mAb (W6/32), not with the anti-HLA class II (L243) mAb (Fig. 2b). Similarly, the recognition of the HLA-matched (HLA-A*0201) allogeneic cutaneous melanoma line 501 mel was superior (n. 371 spots/5,000 cells; Fig. 2b) by culturing T cells with IL-2 plus IL-15 compared to the others cytokines or their combinations (20–205 n. spot/5,000 cells; Fig. 2a, c, d). Moreover, all the MLTCs (Fig. 2) failed to recognize the HLA-mismatched allogeneic cutaneous melanoma 15392 mel. No significant reactivity against the K562 line was detected for the MLTCs cultured with IL-2 plus IL-15 or IL-2 plus IL-21 (Fig. 2b, d), while IFN-γ release, though at low levels (22 and 42 spots/5,000 cells, respectively), occurred following the incubation of the MLTCs cultured with IL-2 alone or IL-2 plus Il-7 with this cell line (Fig. 2a, c). Higher reactivity (195 spots/5,000 cells) against the K562 cell line as compared with the autologous melanoma (62 spots/5,000 cells) was observed by lymphocytes expanded in vitro with IL-15 alone, indicating that an enrichment of NK-type immune responses occurred in this MLTC (data not shown). Thus, IL-2 plus IL-15 led to the most efficient isolation and expansion in vitro of melanoma-specific T lymphocytes, and therefore, this cytokine combination was utilized for the subsequent ex vivo selection of anti-tumor T lymphocytes.

The MLTC-based protocol has been applied to activate and expand in vitro circulating T cells from cutaneous melanoma patients.

Independent MLTC cultures were set up from PBMCs isolated from 6 metastatic cutaneous melanoma patients (#2710, 4478D, JOFR-IA, DAJU, 0342 and 7). T cells were stimulated weekly with irradiated autologous tumor cells in the presence of IL-2 and IL-15. Following the two rounds of in vitro stimulation (day 15), the tumor specificity of these MLTCs was assessed by IFN-γ release (ELISPOT assay). Representative data of patients #2710 are shown in Fig. 3a. In addition, results of patient #4478D are shown in Fig. 1S Panel A of the supplementary online data. These T lymphocytes exhibit specific recognition of the autologous tumor lines (285 spots/5,000 cells for #2710 Fig. 3a) with significant inhibition (53%; P ≤ 0.001) of cytokine release after pre-treatment of target cells with the W6/32 mAb. Instead, low inhibition (26% for #2710) of IFN-γ release was found in the presence of the anti-MHC class II (L243) mAb (Fig. 3a).

Notably, the MLTC from the HLA-A*0201⁺ melanoma patient #2710 specifically (92% of inhibition in the presence of the W6/32 mAb) recognized the HLA class I-matched allogeneic 501 mel line and the HLA-A*0201-restricted Melan-A/MART-1-derived epitope (Fig. 3a). The evidence that the recognition of the Melan-A/MART-1-derived epitope by this MLTC was weaker than the autologous and the allogeneic 501 mel tumor lines (121 vs. 285 and 272 N. spots/5,000 cells, respectively; Fig. 3a) suggests that these T lymphocytes can exhibit specific recognition of a broad panel of TAAs expressed by tumor cells. On the contrary, no recognition of a panel of HLA-A1-restricted peptides (MAG3-A1, MAGE-A3 and COA-1) was observed for the MLTC of patient #4478D (data not shown). In fact, only for patient #0342, the MLTC-derived T lymphocytes recognized the HLA-A1-restricted MAGE-A3 antigen (data not shown). Thus, this protocol allowed to isolate in vitro MLTCs from 6 out of 6 cutaneous melanoma patients with specific reactivity against the autologous tumor cells and, in some cases, against allogeneic HLA-matched tumor lines (e.g., 2710, 0342 and DAJU mel). Most of these T lymphocyte cultures recognized a broad array of TAAs, probably including also unknown antigens, since only in limited cases (#2710, 0342), the recognition of molecularly defined epitopes was observed. The specificity of anti-tumor reactivity by the isolated MLTCs was corroborated by the lack of the reactivity of HLA-mismatched melanoma lines (49318 mel in Fig. 3a) and of the NK target line K562.

We were able to establish in vitro primary uveal melanoma lines with efficiencies approaching 57% (N = 8 tumor cell lines out 12 primary and 2 metastatic ocular tumor samples). Albeit lower than those obtainable for cutaneous melanoma (72%), it is still higher than expected. We found that the immune profile of these ocular melanoma cell lines was similar to that of cutaneous melanoma counterparts as described in the supplements and Tables 1S and 2S (available online). Thus, we investigated whether our MLTC protocol could also be applied to ocular melanoma patients.

As shown by the representative data for MLTCs 2 in Fig. 3b, these T-cell cultures exerted specific recognition of the autologous 15765 mel line (236 spots/5,000 cells) with 53% of inhibition with the anti-MHC class I mAbs (W6/32) (Fig. 3b). We could also isolate CD4⁺ T cells specifically directed against the autologous tumor (#15765) as shown by the inhibition of IFN-γ release by L243 mAb in MLTC1 (Figure 1S Panel B of supplementary results). These data indicate a preferential CD4⁺ and CD8⁺ T-cell enrichment in MLTC 1 and 2, respectively, as confirmed by the phenotype analysis (61 and 82% of positive cells, respectively; data not shown). The inhibition of autologous tumor reactivity by CD8⁺ T lymphocytes (MLTC 2, Fig. 3b) was increased in the presence of both anti-MHC class I and anti-NKG2D mAbs (53 vs. 69% of inhibition) with W6/32 and W6/32 + anti-NKG2D mAbs, respectively), suggesting that the specific recognition of the ocular melanoma cells occurred by the engagement of both TCR and NKG2D as observed for cutaneous melanoma (Fig. 3b). No reactivity directed to known TAAs (Gp100, MAGE-A1 and COA-1) associated with the HLA-A3 molecules expressed by the autologous cells from patient 15765 was detected.

Interestingly, we also observed that these T lymphocytes exhibited cytotoxic activity, measured by CD107a mobilization and production of perforin, which correlated with that of IFN-γ production directed to the autologous melanoma line (Fig. 4).

These results thus indicate that our MLTC protocol can successfully isolate circulating T cells inducing specific recognition of autologous tumor cells from both cutaneous (N = 6) and ocular melanoma patients (N = 1).

The phenotype analysis of T lymphocytes following MLTCs from the 6 cutaneous melanoma patients showed a common enrichment of CD3⁺CD8⁺CD45RO⁺ (44–95% of positive cells) with CD4⁺ T cells also detectable in the range of 25–51% of positive cells (4478D, JOFR, 0342, 7 mel) (representative results from #2710 patient are shown Fig. 5). These findings are in agreement with the evidence that most of tumor recognition activity is HLA class I-restricted (Figs. 3 and 1S). These T-cell cultures expressed, though to a variable extent (20–40% of positive cells), some co-stimulatory receptors such as CD28, NKG2D, OX40 (CD134), 4-1BB (CD137), while CD27 was found positive only in 3/6 patients (4478D, JOFR-IA and DAJU; representative data are shown in Fig. 5). In addition, CCR7 and CD62L were mostly expressed at low levels, and CD57 was not found on these MLTCs (data not shown). Therefore, the MLTC-derived T cells we have isolated are in agreement with the phenotype of non-terminally differentiated effector memory (T_EM) (CD57-negative) T lymphocytes [33, 34]. Interestingly, T lymphocytes expressing high levels of co-stimulatory molecules have previously manifested potentially efficient melanoma-specific reactivity and in vivo persistence [24, 25].

No T regulatory (Tregs) cells as identified by specific staining for CD4⁺CD25^highCD127^dim were found in MLTCs (data not shown).

Of note, a similar phenotype profile was detected for both the MLTC 1 and the MLTC 2 T cells of the ocular melanoma patient 15765 (data not shown).

Thus, non-terminally differentiated anti-tumor T_EM cells could be ex vivo isolated by applying our protocol to circulating lymphocytes.

The most encouraging results of ACT protocols for metastatic melanoma patients have been obtained by the use of TILs [20, 21]. Thus, we have carried out a comparison of the phenotype and functional activity between MLTCs and TILs, from both cutaneous and ocular melanoma patients.

Initially, we isolated TILs from 5 metastatic cutaneous melanoma (3 subcutaneous lesions #4478D, 9476, 25368 and 2 lymph node lesions #3681 and 4931, respectively) and from 4 ocular melanoma (3 primary #3470, 1141, 4022 and one metastatic #15765) patients. The phenotype analysis of the T lymphocytes is shown in Fig. 6 (cutaneous melanoma patients # 3681, 4931, 9476 and 25368, panels A, B and C and primary ocular melanoma patients #3470, 1141 and 4022, panels D, E and F). TILs were cultured in vitro for 5–7 days with 600 IU/ml of rh-IL-2 in order to enrich T cells and to remove tumor cell contaminations. These TILs from melanoma were CD8⁺ T cells (46–64% of positive cells), with 18–60% of CD4⁺ T cells (Fig. 6a, d). Only TILs 9476 were enriched for CD4⁺ T cells (94% of positive T cells; panel A). Higher levels of CD8⁺ T cells (57–80% of positive cells), with the exception of patient #1141 (98% of CD4⁺ T cells; Fig. 6d), were found in TILs deriving from ocular melanoma lesions. All T cells expressed homogeneously CD45RO (data not shown) and high levels of CD28 (40–94% of positive cells; Panels B and E), while CD27 was detected only in TILs from two cutaneous melanoma patients (#25368 and #9476, 39 and 46% of CD8⁺ T cells, respectively; panel B) and in CD8⁺ lymphocytes from one ocular melanoma patient (# 3470, 24% of CD8⁺ T cells; panel E). CD137 was detected at high levels in T cells from 2 out of 4 metastatic cutaneous melanoma patients (the 2 lymph node metastatic lesions #3681 and 4931), in the primary uveal melanoma #4022 (30% of CD8⁺ cells) (Fig. 6c, f, respectively) and homogeneously in the metastatic ocular melanoma patient #15765 (CD137 was also present in 100% of both CD4⁺ and CD8⁺ cells; data not shown). Of note, 10–22% of CD134⁺ T cells were detected in ocular melanoma TILs (both CD4 and CD8 T cells) and, to a lower extent (10–12% of positive cells), in cutaneous melanoma deriving T cells (Fig. 6c, f). CD134 was homogeneously associated with both CD4⁺ and CD8⁺ T cells from TILs of #9476 and #1141 cutaneous and ocular melanoma patients, respectively (Fig. 6c, f). In addition, for all the patients, CD8⁺ T cells expressed homogenously NKG2D (50–87% of positive cells) (data not shown). CD57 was detected to a variable extent (11–23% of positive cells) in subpopulations of both CD4⁺ and CD8⁺ T cells, indicating that most of TILs were not terminally differentiated. No Tregs were found in short-term cultured TILs of all the analyzed cutaneous and ocular melanoma patients (data not shown). Therefore, TILs displayed an effector memory non-terminally differentiated phenotype, and in most cases, lower levels of the expression of co-stimulatory molecules were detected as compared with MLTC-derived T cells (see Figs. 5, 6).

Notably, an immune infiltrate was detected also by IHC analysis in 10 primary ocular melanoma lesions (representative results of patients #50306324 and 50316250 are shown in the Fig. 6g–n and Figure 2S of Supplementary results). In 6/10 tissues, CD4⁺ T cells were detected (21–70 cells/1 mm²); Fig. 6h), while CD8⁺ enrichment in ocular melanoma tissues was found only in three patients (40–80 cells/1 mm²; Fig. 6m). In addition, heterogeneous levels of Tbet and GATA3 (transcription factors associated with TH1 and TH2 type immune responses, respectively) were observed (Figure 2S Panel B, C and F and G). Interestingly, the presence of Tregs (evaluated as CD25⁺FoxP3⁺) was found in association with the enrichment of CD4⁺ T cells and with low infiltration of CD8⁺ lymphocytes (Fig. 6i, n). High numbers of macrophages with suppressive activity were also observed as shown by the staining with anti-CD163 mAb (Figure 2S Panels D and H).

We found that TILs can be detected in primary ocular melanoma lesions, though mostly CD4⁺ and in association with the presence of Tregs and M2 macrophage, envisioning an immune infiltration possibly dominated by TH2 and/or immune suppressive activity.

The short-term in vitro culture of TILs can, in most cases, rescue their TH1 activity, as suggested by the phenotype analysis (Fig. 6a–f and 2S Panels A–H). To prove this hypothesis, we assessed—when the autologous tumor line was available—the anti-melanoma activity by measuring the IFN-γ release (ELISPOT) of TILs from ocular melanoma and compared to that of cutaneous melanoma. Figure 7 shows representative results of TILs from the cutaneous melanoma #4478D (Panel A) and from the ocular melanoma #15765 (Panel B) patients. High levels of IFN-γ release (355 and 334 spots/5,000 cells for patients #4478D and #15765, respectively) were observed following incubation of TILs from both patients with the autologous tumor cell lines. Moreover, the tumor recognition was HLA-restricted as the cytokine secretion was inhibited (65 and 58% for patients #4478D and #15765, respectively) by the incubation of tumor cells with the anti-HLA-class I (W6/32) mAb, but not by the anti-HLA class II mAb (L243). In addition, both TIL cultures failed to recognize the allogeneic HLA-mismatched melanoma lines (1067 and 15392 lines; Fig. 7). Furthermore, lack of recognition of HLA-A3-restricted TAA-derived epitopes (such as Gp100, MAGE-A1 and COA-1) was observed for TILs as well as for MLTC isolated from the 15765 patients (data not shown), thus indicating that these T cells can recognize TAAs specifically associated with ocular melanoma.

These results demonstrate that the short-term in vitro culture of TILs from ocular melanoma can rescue their TH1 activity and that anti-tumor TILs can be isolated from both cutaneous and ocular melanoma patients.

The REP protocol was applied to both MLTC and TILs isolated from 10 cutaneous and ocular melanoma patients in order to determine whether comparable expansion in vitro of effector T cells could be obtained. The results shown in Fig. 8 indicate that efficient expansion (68–148 × 10⁶ T cells) of TILs (#15765 and #4478D) and MLTCs (#4478D, #2710 and #0342) could be achieved with an increase value of 240–592 times (Fig. 8, insert). Notably, similar expansion of T lymphocytes was observed in TILs vs MLTCs of the same patient (#4478D) (240 and 272 times number, respectively) (insert of Fig. 8). Moreover, an efficient expansion of TILs from the ocular melanoma patient #15765 was achieved (8.4 × 10⁷ cells starting from 1.5 × 10⁵ cells with an increase of 560 times) as well, thus indicating that the possible anergic state of T cells isolated from ocular suppressive tumor milieu can be recovered ex vivo and such lymphocytes efficiently expanded.

The specific anti-tumor reactivity by MLTC-derived T lymphocytes expanded in vitro in large scale with the use of the REP protocol [31] was obtained for all of the 7 cutaneous or ocular melanoma patients (# 7, 2710, 0342, 4478D, DAJU-1A, JOFR and 15765). Representative results of patients 4478D and DAJU are shown in Figure 3S Panels A and B, respectively, of supplementary online results). Similar results were obtained for the TILs isolated from three patients (# 4478D, 25368 and 15765) following the application of the REP as shown by the representative results in Figure 4S of Supplementary (available online). In addition, both MLTC- and TIL-derived T cells following the application of REP maintained their phenotype observed prior to the large scale expansion (data not shown).

In conclusion, our results indicate that, similarly to TILs that have been thus far largely used for ACT studies [20], the protocol we have identified based on MLTC followed by REP can lead to the isolation of large numbers of effector T cells. This protocol appears then to be suitable for ACT clinical applications for cutaneous and, for the first time, for ocular melanoma patients.