Natural T cell autoreactivity to melanoma antigens: clonally expanded melanoma-antigen specific CD8 + memory T cells can be detected in healthy humans

The present study was designed to address current technical challenges for ex vivo assessment of T cells in general, and MA-specific T cells in particular. This initial study serves as a baseline assessment using healthy donors, and building upon these data, eventually has application to characterizing T cell reactivity in melanoma patients. In contrast to previous studies that predominantly utilized select MA peptides in the context of specific HLA alleles and the flow cytometry-based multimer approach, we opted to utilize extensive peptide pools that systematically covered the entire amino acid sequence of the melanoma antigens; tyrosinase, MAGE-A3, Melan-A/Mart-1, gp100 and NY-ESO-1. The first surprising finding was that 23 of 40 healthy donors (58%) displayed ex vivo reactivity to at least one of these MA antigens (Table 1). Eight of these HD responded to two or more MA. Specifically, two HD responded to two MA pools and another two HD responded to three MA pools. One HD responded to four MA, and the remaining three responded to all five MA peptide pools, respectively. NY-ESO-1 and gp100, while frequently recognized in combination with other MA, were not targeted in isolation. Overall, the MA-specific repertoire in HD appears to be diverse, and capable of recognizing each of five MA used in this study. Collectively, this observation justifies the use of all five of these peptide pools for comprehensive immune monitoring, and indicates that additional MA should be considered for assessment of T cell reactivity.

The present study also illustrates the feasibility of such a broad coverage of multiple antigens. As performed, using 250,000 PBMC per well and three replicates for each of the five MA plus the medium control, only 5.4 million PBMC were required to establish the breadth of the MA-specific repertoire from individual HD. Importantly, while this broad screening was performed using a single color IFN-γ assay, the same number of PBMC could have been used to define the respective MA-specific T cell subpopulations using a 4-color four-color ImmunoSpot® assay. Thus, even after accounting for the limiting PBMC numbers that can be obtained from melanoma patients (e.g. 20 million PBMC), the number of MA peptide pools that could be readily tested can be increased to 25 when tested in triplicate, or 80 when tested as single replicates; which might well be permissible as hardly anyone runs samples in flow cytometry in replicates.

Using the ELISPOT platform for measurement of T cell responses against “foreign antigens”, such as viruses, one can unambiguously identify donor cohorts that have, or have not, been exposed to a particular antigen. Such studies demonstrate that specific viral peptides, or peptide pools, elicit T cell-derived IFN-γ production in excess of the medium background using PBMC from virus-exposed cohorts but not in virus-naïve individuals [22]. Confirming this notion, when we recently tested similar peptide pools that cover HCMV antigens, we found that IFN-γ SFU were elicited only in CMV-exposed donors, but not in HCMV-negative individuals [23]. A similar experiment is detailed in Suppl. Fig. 1, which further illustrates this important point. In this study [23], and in the data presented in Suppl. Fig. 1, HCMV peptide pools that represent 11 distinct proteins encoded by of HCMV virus, each consisting of 15-mer peptides that span the entire length of the respective aa sequences in steps of 11 aa, were used for stimulation. These HCMV peptides were from the same manufacturer, JPT, and were synthesized and handled under identical experimental condition, including being used at an equivalent concentration as the MA peptides reported here. Each of the three HCMV-infected donors shown in Suppl. Fig. 1 (donors A, B and C) responded vigorously to most of the HCMV antigens/peptide pools tested. In contrast, none of the three HCMV negative HD (donors D, E, and F) responded to any of the 11 HCMV peptide pools with significant SFU numbers over the medium background. Thus, in a situation in which HCMV infection, and hence the development of an immune response, can be confirmed, the exquisite specificity of such peptide pools for detecting ex vivo-expanded memory T cells can easily be verified. However, since MA are ubiquitously expressed self-antigens, they may be more likely to induce immunologic tolerance than trigger clonal expansions and cytokine differentiation in healthy donors. Nevertheless, MA may be unique self-antigens since environmental factors such as sunburn, which promotes a pro-inflammatory environment, may enhance their immunogenicity and trigger initiation of an immune response.

The relatively high frequency of MA-specific CD8 + T cells detected amongst PBMC from HD was not anticipated. For example, under the 24-h recall conditions (Table 1), tyrosinase-reactive T cells in HD #7, #9, #17, and #36 reached or exceeded 30 SFU per 250,000 PBMC, which is a precursory frequency of 1:8300 or 0.008%. In three additional HD (#6, #24 and #31), MAGE-3A reactive cells were also present in this frequency range. While still at the detection limit of flow cytometry, the magnitude of these recall responses in an ELISPOT assay are typical of clonally expanded memory/effector T cell populations [24]. To this end, the frequency of MA-reactive CD8 + cells detected in these HD were in the range of clonally expanded memory cells, and were substantially greater than the < 1 in 250,000 PBMC frequency anticipated in the naïve T cell compartment.

Using HLA-A2*01 tetramers, Melan-A _(26−36) peptide-specific CD8 + T cells were detected at an atypically high frequency (0.07%) in HD, and these T cells were apparently naïve since they exhibited a CD45RA⁺ surface phenotype [25]. As such a high frequency of naïve T cells is unique, it is thought that these cells emerge through a particularly productive thymic selection process. As noted above, we have detected tyrosinase and Melan-A reactive CD8 + T cells at frequencies around 0.01% following 24 h antigen stimulation directly ex vivo. After 72 h of antigen stimulation, 8 of 40 HD responded to at least one of the MA at this relatively high, > 30 SFU per 250,000 PBMC or ~ 0.008%, frequency range. Altogether, 58% of HD displayed a significant response to at least one of the five MA antigens used for testing.

In general, peptide-specific naïve T cells occur at very low frequencies; < 1 per 250,000 PBMC. Additionally, naïve T cells do not secrete IFN-γ within the first 24 h of antigen stimulation [10]. While naïve T cells are capable of extensive proliferation following antigen recognition, the 24-h incubation period of an ELISPOT assay is sufficient for only a single cell division cycle. Thus, detection of MA-reactive T cells capable of IFN-γ production within a 24-h assay (Table 1; Fig. 1) signifies that these T cells have most likely undergone previous expansion and differentiation into Th1-type memory cells in vivo. Since the MA-specific T cells do not secrete GzB within the first 24 h of stimulation (Fig. 2b), these cells most likely represent resting (central) memory cells directly ex vivo. Following continued antigen stimulation, these cells go on to acquire expression of cytolytic granules and convert into GzB + effector CD8 + T cells capable of both cytolysis and IFN-γ secretion. The observed behavioral pattern of these MA-reactive T cells is typical of resting CD8 + T cells specific for virus antigens, including CEF peptide-reactive CD8 + T cells [21].

In this study we also identified MA-specific T cells from several HD that failed to secrete IFN-γ, along with GzB, after 24 h of antigen stimulation. However, these same donor cells yielded both IFN-γ and GzB producing cells after 72 h of stimulation with MA (Table 1; Fig. 1). Such cells likely represent progeny of stem cell-like (SC) CD8 + memory T cells [9], which can engage in rapid proliferation and then differentiate into effector cells capable of both IFN-γ and GzB expression following a longer 72 h antigen stimulation. Another possible explanation for the delayed cytokine secretion phenotype is that MA-specific T cells are naive at isolation. While self-antigen is likely to deplete the high-affinity end of the autoreactive T cell repertoire [16], T cells are positively selected on self-antigens and ongoing T cell receptor engagement is required for survival [26]. Therefore, the increased frequency of naïve MA-specific T cells could be attributed to an increased abundance of low-affinity, autoreactive T cells relative to those recognizing foreign antigens with high-affinity.

Data consistent with the notion that self-reactive T cells in HD can possess a unique phenotype has also been reported for cytokeratin-18-specific CD8 + T cells [27]. Such cells have undergone extensive clonal expansion in vivo, but are functionally unresponsive and do not secrete cytokine in response to antigen encounter directly ex vivo. However, after antigen stimulation these cells regain their functionality. The MA-reactive T cells from HD that failed to secrete cytokines at 24 h, but which became positive for cytokine secretion at 72 h, therefore, might be the progeny of such clonally expanded autoreactive T cells. Overall, these findings suggest that in the context of comprehensive immune monitoring it is not advisable to restrict assessment of T cell function(s) to a single set time point following antigen stimulation. Instead, it would be preferable to evaluate T cell recall responses over extended time frames [28]. Importantly, this approach is technically feasible using the ELISPOT platform in spite of limiting PBMC acquired from patients. Furthermore, since ELISPOT is a non-destructive technique and T cells survive the assay largely unharmed, PBMC can theoretically be transferred from one assay into the next, facilitating testing of the same PBMC [29].

Initial efforts reported in the literature that sought to identify melanoma-specific T cells required over a week of in vitro expansion [5]. To avoid this in vitro expansion step, and to enable determination of MA-specific T cell precursor frequencies as they exist in vivo, we performed standard 24-h recall assays. This approach is commonly utilized to identify in vivo primed memory T cells. In an ELISPOT assay that does not require signal enhancers, the 24-h antigen-stimulation period is optimal for detecting peptide-elicited cytokine production by CD8 + memory T cells [28]. Since resting memory CD8 + T cells do not possess cytolytic granules immediately following antigen stimulation, but instead acquire this effector function within 3–4 days after antigen stimulation [21], we extended our in vitro culture to 72 h to determine whether the resting memory cells detected at 24 h could differentiate into cytolytic effectors. To enable segregation of such effector T cells into distinct sub-lineages, we also performed 4-color ImmunoSpot® analysis at the 72-h time point. The closer characterization of cytokine expression profiles by tyrosinase-specific CD8 + T cells demonstrate that, in addition to terminal effector cells (TE: IL-2⁻, IFN-γ⁺, TNF-α⁺ and GzB⁺), the polyfunctional (PF: IL-2⁺, IFN-γ⁺, TNF-α⁺ and GzB⁺) and stem-cell like (SC: IL-2⁺, IFN-γ⁻, TNF-α⁻ and GzB⁻) lineages were also present within the MA-specific CD8 + T cell repertoires, with the latter stem cell-like subset prevailing (Fig. 4). The abundance of stem cell-like CD8 + T cells we observed serves as indirect evidence that tyrosinase-specific CD8 + T cells in HD have not engaged in prolonged immune encounters in vivo. In contrast, in cancer patients, due to T cell exhaustion resulting from persistent antigen stimulation, senescent and dysfunctional tumor antigen-specific CD8 + T cells seem to prevail, with a concomitant reduction in the stem cell-like memory cell pool [8, 30].

Titration of the antigenic peptides permits determination of the functional affinity of antigen-specific T cells [16]. We found the tyrosinase-specific CD8 + T cells in HD to be of relatively low-affinity (K₅₀ ~ 1 µg/ml) compared to CEF-specific CD8 + T cells (K₅₀ ~ 0.005 µg/ml). Unlike viral antigen-specific CD8 + T cells which recognized foreign peptide determinants, T cells recognizing self-antigens must somehow evade negative selection. To this end, our data suggest that the tyrosinase-specific CD8 + T cells we detected in HD escaped negative selection due to their low affinity for this self-antigen. In contrast, T cells bearing receptors that recognize tyrosinase peptides with high-affinity are likely to be purged from the repertoire through clonal deletion.

The data presented in this study also highlights the feasibility of evaluating the MA-specific T cell repertoire using a novel modification of the ImmunoSpot® technology. The sensitivity of the ELISPOT assay enables detection of low abundance T cells, far below the < 0.01% frequency range, which is not reliably achieved using a flow cytometry approach. The use of MA peptide pools also permitted parallel assessment of all potential antigenic determinants of the MA, which is not technically feasible using multimers. Studies of the activation kinetics also revealed an apparent naïve/anergic state of MA-specific T cells in vivo. Unlike flow cytometric assessment of cytokine production, which requires fixation and membrane permeabilization of the sample, cell viability is maintained during the ELISPOT and secondary assays can be performed without requiring additional cell material. Moreover, multi-color measurements of cytokine expression profiles using the 4-color ImmunoSpot® assay also permitted identification of the respective memory lineages and fractionation into stem cell-like, polyfunctional or terminal effectors cells. Finally, measurement of functional affinity was also deduced through peptide titration experiments, and was accomplished using fewer cells and requiring considerably less labor compared to flow cytometry. Importantly, all of the assays detailed in this report were accomplished using less than 20 ml of blood per donor, and with relatively minor investment of investigator effort and cost. Furthermore, all of the reported ELISPOT assay variants can be readily validated and adapted for generating data in a regulated environment [11]. Thus, we believe, these novel extensions of the ImmunoSpot® technology will largely facilitate efforts to better understand tumor antigen-specific T cells in health and disease, including their mobilization for treatment of tumors.

Melanoma become immunogenic in the metastatic stage, and the T cell response to the malignant cells has been quite well-characterized. The T cells in melanoma patients target non-synonymous mutations on various proteins and/or recognize overexpressed non-mutated melanoma antigens [2, 31, 32]. However, it was not apparent prior to our findings that benign melanocytes were capable of triggering endogenous T cell responses against melanocyte autoantigens which are neither overexpressed, nor mutated, and as such qualify as “self-antigens”. Our data draw attention to the existence of these melanocyte-specific CD8 + T memory cells in healthy human donors. Likely, this natural T cell autoreactivity gets primed during sunburns, when melanocyte activation occurs in the context of skin inflammation. At present, one can only speculate about the immunobiological significance of natural CD8 + T cell autoreactivity. One possible interpretation of this finding is that natural T cell autoreactivity to melanocyte antigens helps protects against development of melanoma. This notion is supported by ongoing studies, in which we found that 23 of 24 melanoma patients tested lacked MA reactivity that was demonstrated in this report using HD, while still exhibiting robust T cell reactivity against additional recall antigens (Anna Przybyla and Paul V. Lehmann—manuscript in preparation). A similar observation has been reported for breast cancer. Healthy women displayed high levels of spontaneous T cell autoreactivity to HER-2, while women with breast cancer were found to selectively lack such T cell responses [33]. One possible interpretation of such findings is that individuals lacking natural T cell autoreactivity to a given tumor are at increased risk of developing that cancer. However, there is an equally likely interpretation for the selective absence of natural T cell autoreactivity in cancer patients. Specifically, these pre-primed T cell populations may become “burned out” and undergo senescence in the face of ongoing antigen stimulation [34]. Such senescent T cells, while undetectable in an IFN-γ recall assay, could even promote the growth of the tumor [8, 30]. In either case, natural T cell autoreactivity targets autoantigens, and expression levels thereof, that are present on melanocytes. Once melanoma cells arise in the body, they offer a new set of target antigens for CD8 + T cell recognition, which can either be mutated or overexpressed melanoma antigens. The latter, though qualitatively unchanged relative to “self”, become targeted because they are presented at much higher MHC-peptide density on tumor cells. When expressed at a sufficiently high copy number on transformed melanocytes, low-affinity autoreactive T cells will likely receive sufficient TCR stimulation that their activation threshold is exceeded. By contrast, these same low-affinity autoreactive T cells would otherwise be ignorant of the same MHC-peptide combination when expressed at physiologic levels on normal melanocytes [35]. It is likely that checkpoint inhibitors may act on tumor-specific T cells that recognize mutated self-peptides, but it remains unclear whether such biologics also enhances the activity of pre-existing, natural T cell autoreactivity. Furthermore, the role of natural CD8 + T cell reactivity and prevention of melanoma in HD is presently unknown. Whether these cells are inactive in melanoma patients, or if they contribute to promoting the immunosuppressive tumor micro-environment after undergoing senescence are also key questions that will require further study.