Tumor-infiltrating lymphocytes: apparently good for melanoma patients. But why?

Abstract

Tumor-infiltrating T lymphocytes (TILs) are observed in a number of human primary or metastatic tumors. Recently, gene expression profiling experiments suggested that the presence of T cells in metastatic melanomas before vaccinating the patients with tumor antigens could be a biomarker for clinical benefit from the vaccines. In this context, we review results pertaining to TILs in human melanomas, their prognostic value, and some possible reasons why their presence could help in selecting melanoma patients for vaccination against tumor-specific antigens.

Introduction

The presence of inflammatory cells within tumors was first noted by Rudolf Virchow in 1863 and led to work linking inflammation and cancer [1]. Later, T cells were shown to play a major role in rodent anti-tumor immune responses, and tumor-specific cytolytic T lymphocytes (CTL) were found in the blood [2] and among the leukocytes infiltrating the tumors [3] of melanoma patients. Then, Rosenberg and colleagues showed that melanoma regressions could be obtained in patients treated with a combination of interleukin-2 (IL-2), cyclophosphamide, and the adoptive transfer of autologous tumor-infiltrating lymphocytes (TILs) amplified in vitro with IL-2 [4, 5]. The identification of the antigens that were recognized by human tumor-specific CTL [6] provided a strong rationale for pursuing the development of active and passive T-cell immunotherapeutics.

There is a renewed interest in studying TILs, mostly in the context of their value as a biomarker in cancer immunotherapy [7]. Three clinical studies of active immunization of metastatic melanoma and lung cancer patients with defined tumor-specific antigens have included gene expression profiling of tumor samples taken before vaccination, and clinical benefit from the treatment appeared to correlate with a gene signature that included T-cell markers [7–9]. Though preliminary, these results are important for two reasons. They could lead to a significant improvement of the clinical efficacy of therapeutic anti-cancer vaccines, by selecting the responder patients. And they might help to identify the main factors that make vaccines ineffective in the other patients and to focus research towards rational immunomodulation, complementing vaccination or adoptive transfer.

Here, we will discuss in the context of active immunotherapy a few results about TILs in human cutaneous melanoma. Other aspects of TILs in melanoma can be found in a recent review by Oble et al. [10].

Prognostic significance of TILs in melanoma

If the presence of TILs in melanomas is predictive of a better clinical outcome following active immunotherapy [7], it is relevant to ask whether the presence of TILs itself has prognostic value. In primary melanomas, most studies point to the favorable effect of the presence of TILs, even though there is a lack of consensus (Table 1). The first large studies leading to this conclusion were conducted by Larsen and Grude [11] and Johnson et al. [12]. These were followed by the seminal work by Clark et al. [13] who defined the “brisk infiltration”, i.e., “TIL presence throughout the substance of the vertical growth phase or present and infiltrating across the entire base of the vertical growth phase”. Brisk T-cell infiltration was an independent predictive factor of patient survival. It is important to note that from then on, the presence of TILs in primary melanoma received a precise definition. The definition includes a melanoma in its vertical growth phase (the presence of lymphocytes during the earlier, radial growth phase has no prognostic value) and lymphocytes that infiltrate the tumor (an abundance of peritumoral lymphocytes had no influence on survival) [13]. Two other studies supported the prognostic significance of brisk TILs [14, 15], whereas four did not [16–19] (Table 1). In three of these [16, 18, 19], the patients underwent a sentinel lymph node biopsy procedure, and in each study, the absence of TILs predicted the presence of a metastasis in this node, which is itself an important predictor of shorter survival [20]. Thus, in primary melanoma, brisk TILs are associated with a better prognosis, although they are not an independent predictor of patient survival when all the well-established clinicopathologic factors are included in a multivariate analysis.

Table 1.

Tumor-infiltrating lymphocytes (TILs) in primary melanomas and patient survival

Studya	Staging	Groups	5-year overall survival (%)	P value	SLNe	Reference
Larsen and Grude (1978)	Stage I	TIL + (n = 210)h	78	“Significant”	No	[11]
		TIL +++ (n = 151)	91
Johnson et al. (1985)	Stage I	TIL + (n = 172)i	48	<0.05b	No	[12]
		TIL +++ (n = 90)	60
Clark et al. (1989)	Primary local disease	TIL absent (n = 108)	59d	0.0015b	No	[13]
		TIL brisk (n = 52)	89
		TIL non-brisk (n = 104)	75
Barnhill et al. (1996)	Stage I	TIL absent (n = 353)	86	n.s.b	No	[17]
		TIL brisk and non-brisk (n = 188)	90
Clemente et al. (1996)	Stages I/IIf	TIL absent (n = 105)	37	0.0003b	Yes	[14]
		TIL brisk (n = 47)	77
		TIL non-brisk (n = 133)	53
Tuthill et al. (2002)	Primary local disease	TIL absent (n = 122)	71	0.005c	No	[15]
		TIL brisk (n = 29)	100
		TIL non-brisk (n = 89)	71
Taylor et al. (2007)	Primary local disease	TIL absent (n = 195)	75	n.s.b	Yes	[18]
		TIL brisk and non-brisk (n = 692)	76
Mandala et al. (2009)	Stages I/II	TIL absent (n = 701)	90	n.s.c	Yes	[16]
		TIL brisk and non-brisk (n = 550)	95
Burton et al. (2011)	Stages I/II	TIL absent and non-brisk (n = 415)	84	n.s.g	Yes	[19]
		TIL brisk (n = 100)	95

^a Only studies including >200 patients were considered

^b Logrank or univariate statistics

^c Cox proportional hazards model

^d Proportion of patients with 8-year overall survival

^e If the study did or did not include a sentinel lymph node (SLN) biopsy procedure

^f Patients with clinically negative but histologically positive lymph nodes (N0+) were excluded from this study

^g Multivariate analysis

^h Slight lymphocyte infiltration limited to one or both sides of the invading tumor (+) or intense infiltration completely surrounding the invasive part of the tumor (+++)

ⁱ Lymphocyte infiltration recorded as absent to mild (+) or moderate to marked (+++)

In metastatic melanoma, the prognostic value of TILs is less clear because no results from a large study are published and because most reports include lymph node metastases in which it is not easy to distinguish between inflammatory and resident lymphocytes. Two studies are worth mentioning. In 1996, about 100 lymph node metastases were evaluated for TILs with the histologic criteria applied to the vertical growth phase of primary melanomas, and a more favorable prognosis was reported for patients with brisk TILs in their metastases [21]. Recently, about 30 metastases were analyzed using histology and gene expression profiles; more TILs correlated with longer survival, and it was confirmed in a validation set including 25 stage IIIc patients, but not in stage IIIb patients [22].

The prognostic value of TILs is not unique to melanomas. In colorectal tumors, multivariate analysis showed TILs to be an independent prognostic factor of survival [23]. More recent studies extended these results using high-throughput quantitative measurements of surface markers and of gene expression and showed a correlation between a low incidence of tumor recurrence after surgery and markers of Th1 polarization, cytotoxic, and memory T cells [24, 25]. In ovarian carcinomas, the presence of TILs was associated with a median duration of survival that was 2.8 times as long as that in patients whose tumors contained no TILs [26].

Specificity and functions of TILs in melanoma

The prognostic value of TILs, the identification of melanoma antigens recognized by T cells, and the presence of tumor-specific T cells in the blood and TILs from melanoma patients led to the inferences that TILs were tumor-specific and that they were anergic because of their coexistence with the melanoma cells. Which results support the conclusion that TILs are tumor specific? They are of two types as follows: evidences obtained with CTL clones or with tetramers, and suggestions with analyses of TCR repertoires or activation markers.

In the first group and on primary melanoma, there is only the work by Hercend and colleagues. In a primary tumor with histological signs of ongoing regression, they observed that some TCR Vβ sequences were enriched compared with autologous PBMCs [27]. One sequence belonged to a tumor-specific CTL clone [28] that recognized a neoantigen resulting from a somatic point mutation [29]. Another belonged to a CTL that recognized a MAGE-A6 antigenic peptide [30]. Altogether, these results demonstrated the presence of tumor-specific TILs in a primary melanoma. For metastases, several groups used tetramers to identify TILs specific for tyrosinase [31], Melan-A/MART-1 [32, 33], or LAGE/NY-ESO-1 [34]. The frequency of tumor-specific cells in these experiments was not very informative because the TILs were cultured with IL-2 over several weeks prior to tetramer labeling. In one study of 16 metastases analyzed for their ex vivo content of TILs labeled with four tetramers corresponding to melanocyte differentiation antigens, all the labeled T cells corresponded to only 0.2–8% of the TILs or 0.04–1.1% of living cells in the tissue [35]. In our own work, we found in a cutaneous metastasis that about 1.5% of the TILs were tumor specific. They consisted in six CTL clones recognizing five different antigens [36]. Finally, the work regarding the adoptive transfer of TIL-derived T-cell clones provided ample evidence for the presence of tumor-specific T cells among melanoma TILs, even though their proportions were not studied [37, 38].

TCR repertoire analyses belong to the works suggesting tumor specificity of at least a subpopulation of the TILs. In primary melanoma, several studies indicated the presence of oligoclonal T-cell populations [39–41]. Different clones were present in different regions of the same primary tumor [41, 42], suggesting bursts of T-cell activation that are too transient to spread throughout the entire tumor. In metastases, several groups also observed clonally expanded T cells [33, 43].

Other suggestions that TILs are tumor specific come from the analysis of their status of activation while contacting the melanoma cells. Ladanyi et al. used histochemistry to detect expression of the T-cell-activation markers CD25 and CD134 (OX40) on TILs from 76 primary melanomas [44]. CD25⁺ and CD134⁺ T cells were present in all tumors, and patients with higher densities of these cells around the tumor had superior survival. Anichini et al. used FOXP3 as an early activation marker on CD8⁺ TILs and showed the presence of CD8⁺ FOXP3⁺ T cells in advanced primary and metastatic melanomas [45]. A combination of surface markers categorized these cells as ‘early effectors’. After 6 h of coculture with autologous melanoma cells, an increase in the proportions of CD8⁺ FOXP3⁺ cells expressing the CTL degranulation marker CD107a strongly suggested recognition of tumor antigens.

In conclusion, there is no doubt that melanoma TILs do comprise tumor-specific T cells. But we are still far from being able to conclude that all TILs are tumor specific.

The hard road of anti-tumoral T cells towards tumor destruction

CTL-mediated rejection of melanoma is expected to occur if all the following conditions are met: (1) the tumor expresses tumor antigens recognized by the CTL; (2) the patient mounts a CTL response against one or more of these antigens; (3) anti-tumoral CTL generated in lymphoid organs reach the tumor sites and move into the tumor environment to make contact with tumor cells; (4) the CTL become activated; (5) the CTL effectively kill the tumor cells and (6), last but not least, this CTL response persists until all antigenic cells disappear.

It is almost certain that all these steps can take place in melanoma patients. It is common to see areas of tumor regression in primary melanomas associated with a T-cell infiltrate. It is not exceptional to diagnose a melanoma from a resected metastatic lymph node, without any trace of the primary tumor, suggesting that the latter regressed without being noticed. Spontaneous regressions of melanoma metastases, although very rare, have been reported. Finally, tumor regressions were observed in metastatic melanoma patients vaccinated with tumor-specific antigens. In all these cases, however, a direct causative role of anti-tumoral CTL in the regression process remains to be demonstrated.

In patients with progressing primary or metastatic melanoma, which steps fail and for what reason is currently unknown. We will consider each step individually and discuss its potential role in the predictive value of TILs for a clinical response to immunotherapy.

The limiting steps

The presence of tumor antigens on the melanoma cells

For 20 years, it has been known that tumor-specific CD8⁺ or CD4⁺ T lymphocytes recognize antigenic peptides presented by HLA molecules on melanoma cells. These peptides derive from endogenous proteins that are modified or aberrantly expressed as a result of genetic and epigenetic changes in the tumor cells. Peptides encoded by ‘cancer-germline’ genes such as MAGE, by mutated genes, by melanocyte differentiation genes such as TYR or MLANA, and possibly by genes that are overexpressed in melanomas when compared with normal cells are relevant for immunotherapy. All melanomas that carry HLA molecules display some level of these antigens. Cancer-germline genes, such as MAGE, are expressed more frequently in metastases than in primary melanomas, and about 50% of the latter do not express any MAGE gene. ‘Antigenic’ point mutations are already present at the primary tumor stage, as we observed for several of those identified with tumor-specific CTLs. Antigen processing defects can be present in melanoma cell lines, but are usually corrected by IFNγ which is likely to be produced by activated TILs [46, 47]. Only rare melanomas cannot be stained by anti-HLA class I antibodies, and a few melanoma cell lines are devoid of surface HLA class I. If TIL quiescence results from immune escape through antigen loss, one would expect melanoma lines derived from metastases with TILs to bear less antigens than lines derived from metastases without TILs. But we have not observed such a difference. In our view, antigenicity is not the main limiting factor, and defects in antigen presentation by melanoma cells are more a consequence of successful immunosurveillance by T cells than a cause of its initial failure.

The ability to mount a tumor-specific T-cell response

It is unlikely that some melanoma patients are unable to mount a spontaneous T-cell response against their tumor, and even if it was the case, this limitation would be bypassed by vaccination. We found a high frequency of circulating tumor-specific CTL in the blood of most but not all of the metastatic melanoma patients that we analyzed [48, 49]. The stage of the disease at which these responses did occur is unknown, but tumor-specific CTL can be present in primary melanomas [28].

The accessibility of the tumor cells to the T cells

It is plausible that some tumors are not permissive to lymphocyte homing. The vascular adhesion receptors, ICAM-1 and E- and P-selectins, were absent from vessels within melanoma metastases, but strongly expressed on the vessels of immediately adjacent tissues, and this result was linked with the observation that T lymphocytes tend to accumulate around rather than within melanoma metastases [50]. We also noted this peritumoral localization, even in tumors from a patient who displayed regression of metastases following vaccination [36]. To us, the obligatory role for melanoma rejection of lymphocyte entry within or migration towards the middle of the tumor is unclear. A complete regression could result from waves of centripetal attack by T cells remaining at a distance from tumor-derived immunosuppressive factors. Another difference between melanomas with or without TILs could be the production of chemokines, which attract T cells in melanomas with TILs. Gajewski and colleagues identified a subset of melanoma cell lines secreting these chemokines and found that these chemokine genes were expressed, together with T-cell genes, in the metastases of vaccinated patients who had a favorable clinical outcome [51]. It would be interesting to determine how many metastatic melanoma patients have tumor-specific T cells in their blood but no TILs in their tumors, as an indication of the proportion of patients for which lymphocytes entry into tumors could be the main limiting factor.

Activation of T cells within the tumor microenvironment

Once in the tumor, tumor-specific T cells could fail to be properly activated because of an immunosuppressive microenvironment. It seems that they proliferate, resulting in TIL oligoclonality and a higher proportion of tumor-specific T cells that we observed in TILs when compared with blood [36, 52]. As discussed earlier, there are also other signs of activation, and in a survey of cutaneous metastases, we found highly variable proportions of TILs, but they always had signs of activation (van Baren et al., in preparation). Thus, it seems that when tumor-specific T cells have access to tumors, they can be activated. What is not known is whether an absence of TILs could result from a strongly immunosuppressive environment, with blood tumor-specific T cells that could enter the tumor but exit quickly because they fail to be activated. Also, we do not know whether the activated T cells in tumors are directed at tumor antigens.

Lysis of the tumor cells

There is a scarcity of information about the in situ lytic activity of tumor-specific TILs in melanomas. In melanoma metastases that contain TIL, it would be very interesting to know whether there are at least some apoptotic tumor cells next to the CD8 T cells. Romero and colleagues analyzed anti-Melan-A/MART-1 CD8 T cells present in melanoma metastases and showed that upon in vitro restimulation they produced much less IFNγ than the anti-Melan-A blood T cells; however, their lytic activity could not be explored [53]. In mice, Frey and colleagues demonstrated an inhibition of the effector functions of TILs, including lysis, and their biochemical analysis pointed toward activation of p56^lck but not of ZAP70, and the involvement of the phosphatase, SHP-1 [54, 55]. Interestingly, the functional defects of TILs in these two studies were reversible after a brief culture in vitro. Thus, it is possible that tumor-specific T cells function normally upon arrival in a melanoma, including the lysis of tumor cells, and become anergic within a few hours or days. It is not known how long they stay anergic in the tumor, but upon leaving the tumor they would regain their full competence and be ready for a new cycle. This could help explain the paradoxical observations of activated TILs and no detectable tumor regression.

Persistence of the T-cell attack

Several mechanisms can account for a decrease and eventual loss of T-cell function inside tumors. These have been reviewed [56] and include soluble factors such as galectins [57] or immunosuppressive cytokines, inhibitory receptors such as CTLA-4 or PD-1 that participate in normal T-cell regulation, local nutrient shortages such as for tryptophan, the ‘classical’ anergy resulting from a lack of appropriate T-cell costimulation, or immunosuppressive cells such as regulatory T lymphocytes (Treg) or myeloid-derived suppressor cells. Which mechanism operates in a given melanoma is not known. However, the use of selective inhibitors should focus the image; anti-CTLA-4 antibodies appear to improve survival in metastatic melanoma patients [58], anti-PD-1 antibodies are promising [59], and inhibitors of the tryptophan degrading enzyme indoleamine 2,3-dioxygenase, which can be present in melanoma cells [60], are actively developed. Tregs, usually defined as CD4⁺ CD25⁺ FOXP3⁺ T lymphocytes, are present in all tissues including tumors [61]. They have been found also in melanomas, and Treg clones derived from melanoma patients were shown to recognize tumor-specific antigens [62–65]. It is not yet known whether a higher proportion of Tregs in melanoma TILs bears bad prognosis. An important methodological difficulty, now recognized by most authors, is the absence of an unambiguous Treg marker, except for the demethylation of regions of gene FOXP3 [66].

If local immunosuppression is the main limiting factor of TIL action, why would there be a correlation between the presence of TILs and a clinical benefit from anti-cancer vaccines? The vaccines could have an anti-immunosuppressive activity, either through the adjuvants that they contain or through the tumor-specific T cells that they prime or restimulate. In a detailed analysis of two melanoma patients who displayed regression of metastases following vaccination with MAGE antigens, we made several observations that are in line with the results presented in this review; metastases contained TILs prior to vaccination; tumor-specific CD8 T cells were part of these TILs; tumor-specific CTL clones could be derived from blood either before or after vaccination, suggesting that blood tumor-specific T cells were not anergic; a melanoma cell line derived from a metastasis expressed multiple antigens, and CTL recognizing these antigens and able to kill the cell line were present in tumors both before and after vaccination [36, 49]. Analyzing the situation following vaccination, when metastases were regressing, we found three differences compared with the situation before vaccination: (1) anti-vaccine T cells appeared in blood and metastases, but at unexpectedly low frequencies; (2) the proportions of several clones of tumor-specific CTL among TILs increased 10–1,000-fold in regressing metastases; and (3) new CTL clones appeared after vaccination, either recognizing antigens that were already targeted before vaccination, a process of clonal spreading, or recognizing tumor antigens that were not targeted before vaccination, a process of epitope spreading [52]. On the basis of these observations, we proposed a scenario in which an important function of anti-vaccine T cells would be, upon arrival in a tumor, to ‘spark’ processes of clonal and antigen spreading by locally and transiently decreasing immunosuppressive mechanisms [67]. This could lead to tumor rejection by all these T cells together, and not only by those elicited by the vaccine. In this scenario, the presence of TILs would favor clinical response to active immunotherapy.

Explaining the seemingly pacific coexistence of tumor cells and T lymphocytes inside melanoma tumors is challenging. Although we understand the antigenicity of melanomas much better than that of many other tumors, this review illustrates that the study of melanoma TILs is just beginning. Yet, understanding the biology of T lymphocytes inside human tumors is a prerequisite to the rational improvement of cancer immunotherapies.