Left to itself, cutaneous melanoma (CM) is a fast-growing, devastating disease. There is consensus that the outcome of CM is a balance between the speed of tumor growth, its invasiveness and the efficiency of the immune system (IS) to raise defenses against it. Fortunately, the finding that 50% of CM patients have a mutation in the BRAF oncogene led to the synthesis of inhibitors targeting this mutation. The use of inhibitors against BRAF^V600 and MEK produced great advances in CM treatment (1). However, the treatment with inhibitors of the MAP Kinase pathway usually generates resistance (2). Another remarkable advance was the discovery of the immunosuppressive axes CTLA-4/CD28 (3) and PD-1/PD-L1 (4), which led to the synthesis of monoclonal antibodies (mAbs) against such pathways (Immune Checkpoint Inhibitors, ICI). The use of such mAbs, especially anti-PD-1, has led to major advances in the therapy of various tumors, among them CM. However, only a part of the CM population treated with ICI, either in metastatic disease or adjuvancy, responds to treatment (5, 6). Therefore, it is important to investigate the reasons for such disparate evolution to immunotherapy (IT). Metastatic tumor-draining lymph nodes (mTDLN) are the most common encounter site between the metastatic CM cells and an organized immune structure. Therefore, their study may reveal interaction mechanisms that could influence patients´ outcomes. Several attempts have been made to establish prognostic factors to differentiate progressor from non-progressor CM patients. Sivendran et al. analyzed mRNA from primary tumors from stages II-III CM patients and established a 53–gene panel that predicted spontaneously progressing from non-progressing patients, CD2 being the most expressed gene in the panel (7). Erdag et al. analyzed immune infiltrates in 183 metastases from 147 CM patients, 83 of whom (44%) were in lymph nodes (LN) (8). The authors defined histological immunotypes A (no immune infiltrate), B (immune infiltrate limited to surroundings of blood vessels) and C (immune infiltrate within tumors), which correlated with disease progression (A>B>C). Bogunovic et al. analyzed by gene chip technology the mRNA expression in LN from 38 CM patients and found 266 genes related to patients´ survival; the genes that positively correlated with survival were associated with the immune response and those that negatively correlated were involved in cell replication (9). More recently, Yaddanapudi et al. performed a single-cell analysis of immune cells in sentinel LN from 41 stage III CM patients and concluded that an immune-tolerant microenvironment was predominant in mTDLN (10).

Under the premise that challenging the IS with multiple antigens (Ags) would enhance the immune responsive repertoire of T-cell clones, we have developed the therapeutic vaccine VACCIMEL, composed of allogeneic irradiated CM cells adjuvanted with BCG and GM-CSF. VACCIMEL has been studied in CM patients in Phase I (11, 12) and Phase II studies (13, 14). The recently updated CASVAC 0401 Phase II study was an open-label, adjuvant, randomized study on stages IIB, IIC and III CM patients, in which VACCIMEL was compared in a 2:1 ratio to medium-dose IFN-α2b (15). In this study, after a minimum and maximum follow-up of 91 and 150 months, respectively, DMFS was significantly longer in vaccinated patients than in IFN-α2b treated patients with median DMFS of 96 months and 13 months (p=0.035). In these studies, some patients had quite a long DMFS, while others recurred rapidly. These results strongly suggest that some patients´ intrinsic characteristics, both tumor- and/or immune-related, present even before treatment, may influence the course of the disease. To investigate this possibility, two cohorts of stage III CM patients with extremely different outcomes after adjuvant IT were selected: a Good Outcome (GO) cohort with long DMFS and a Bad Outcome (BO) cohort with short DMFS. We analyzed mTDLN to understand in this environment which are the characteristics of tumor and immune cells that may influence the course of the disease.

The human genome contains about 7.5% of its genes coding for proteins related to the immune system (24). It is therefore not surprising that the immune system functionality may vary between individuals in many diseases, and this also applies to the immune response to cancer. This issue has become particularly important since manipulation of the IS has led to major advances in cancer treatment, and cutaneous melanoma (CM) has been a paradigmatic disease for the development of immunotherapies (25). The variety of clinical responses to diverse IT in CM has long been known. For example, high-dose IL-2 produced meaningful clinical responses in only 17% of the CM-treated patients (26). Adjuvant therapy with high-dose IFN-α2b prolonged disease-free intervals in only a small subset of CM patients (27). Also, a clinical trial in adjuvancy with anti-PD-1 pembrolizumab or placebo (EORTC 1325-MG/KEYNOTE-054) in stage III CM patients demonstrated, after a median follow-up of 42.3 months, that only 15% of the patients derived benefit from the treatment, since about 50% of the placebo arm patients did not progress, and around 35% of the patients did not increase DMFS (6).

Therefore, we need a better knowledge of both actors of the disease, the tumor and the IS, to select more appropriate treatments. For this purpose, we centered our studies on patients who had entered clinical trials dedicated to studying the effect of VACCIMEL, an allogeneic vaccine adjuvanted with BCG and GM-CSF (14). From this population, we selected two sets of patients: those with long DMFS (GO patients) and those with very short DMFS (BO patients). We analyzed mTDLN since they are the first site of encounter between metastatic cells from the primary tumor and an organized lymphoid structure. Besides, lymphadenectomy was mandatory prior to inclusion in the vaccine trial. To analyze the FFPE biopsies, we performed IHC and multiplex IF. When both cohorts were compared, significant differences were found. An important finding was that GO patients were heavily infiltrated by CD11c⁺ cells, whereas BO patients were practically devoid of them. GO patients also had more abundant PNAd⁺ HEVs and, most importantly, CD11c⁺ cells and DC-LAMP⁺ cells surrounding HEVs.

We describe to our knowledge for the first time a triad structure that could contribute to a stronger immune reactivity against tumors: the “Favoring Antigen-Presenting Structures (FAPS)”. This triad is formed by Ag-loaded DC, HEV and naïve lymphocytes establishing synaptic contacts with DC. Of note, HEV and FAPS were hardly found in BO patients. As to the analysis of the three components of the triad, DCs are the most efficient APC whose three main capabilities are to capture, process and present exogenous and endogenous antigens to cognate TCR CD4⁺ or CD8⁺ lymphocytes (28). We detected DC mainly by CD11c integrin and DC-LAMP expression. CD11c plays a fundamental role in DC physiology. Wang et al. found that blocking CD11c diminishes acute graft-versus-host disease in mice allogeneic bone marrow transplantation (29). Besides, CD11c would also intervene in the capture of CD47-deficient cells, CD47 being a “don´t eat me” molecule (30). It is important to decipher CD11c functional state, since Probst et al. demonstrated that resting CD11c⁺ cells induce tolerance whereas activated CD11c⁺ cells are capable of inducing effector CD8⁺ and CD4⁺ T lymphocytes (31). DC-LAMP⁺ cells are generally assumed to represent mature DCs (32). The question that should be asked is whether the CD11c⁺ cells found in mTDLN of GO patients are immature DCs, waiting for maturation stimuli to act as fully developed APCs, or whether they are already able to present Ag. To support the latter possibility, Pizzurro et al. demonstrated that CD11c⁺ cells, which phagocytosed tumor Ag, were able to stimulate IFN-gamma production of a specific T-cell clone (19). Besides, Barrio et al. demonstrated that DC or activated TAM are equally capable of cross-presenting MART-1 Ag to CD8⁺ T cells (33). It has also been reported by van de Hout et al. a sequential suppression of different DC subsets in CM sentinel LN (34). Another component of the triad, HEVs, are highly specialized post-capillary venules, endowed with a cuboid epithelium that expresses an ensemble of glycoproteins such as Peripheral Node Addressin (PNAd), a putative ligand of L-selectin (CD62L). Moussion and Girard demonstrated in mice that CD11c⁺ cells are essential for the formation of HEVs (35). Therefore, we suggest that, in GO patients, a higher number of CD11c⁺ cells could trigger the formation of more HEVs than in BO patients. Concerning the possible role of HEVs in cancer, Martinet et al. showed in primary CM that the number of HEVs was related to a good prognosis (36). On the contrary, in BO patients HEV would have been impacted by tumor-induced LN remodeling. The third component of the triad, the naïve lymphocytes, express L-selectin (CD62L), which binds to various sulfated syalyl LeX - containing molecules present in HEVs (37). Thus, we propose that FAPS are structures in which emerging naïve lymphocytes would make synaptic contact with Ag-loaded DC, thus being primed as soon as they leave HEVs; those naïve lymphocytes that fail priming at this instance would be later primed inside the tumor after encountering intratumoral DCs.

Emerging data support a model where LN architecture is progressively disrupted by tumors, leading to profound changes in immune cell function that may contribute to metastatic potential (38). Jones et al. revealed LN remodeling by tumor growth characterized by HEV dilation and dedifferentiation, with loss of PNAd expression and function (39). Our results partially contradict the findings of Yaddanapudi et al., who concluded that an immunosuppressive TME is predominant in sentinel LN (10). Although immunosuppression may be more frequent and permanent in BO patients, we suggest that, in selected GO patients, the Ag-presentation machinery was present and ready to be triggered by subsequent vaccination. The number and quality of DCs in mTDLN could also modulate the number of antitumor reactive T cell clones and the subsequent effect of immunotherapies. However, what could be the relationship between what is observed in an already excised mTDLN and the future evolution of the patients to adjuvant immunotherapy? Vaccination after mTDLN |resection may boost immune responses in other uncompromised LN. In a sequence of studies on VACCIMEL, we found that, in most patients, there was no pre-treatment basal reactivity against melanoma Ags or neoantigens, but such reactivity was triggered after vaccination, demonstrating a competent IS (40, 41).

Another finding of this work was that GO patients had a significantly higher CD8⁺ lymphocytic infiltration, confirming previous results (23). PD-1 expression was restricted to a small number of CD8⁺ lymphocytes, suggesting that only a few CD8⁺ cells had been primed by their cognate antigens (42), thus making infiltration and priming two different processes. Combining the findings of CD8⁺ lymphocytes and the abundance of CD11c⁺ cells in GO patients, it could also be hypothesized that CD8⁺ cells lack the “effector stimulus” provided by fully mature DC. Along this line, Makedonas et al. demonstrated in a model of influenza infection in mice, that the effector phase of CD8⁺ lymphocytes needs Ag presentation by DC and CD28 co-stimulation for sustaining activity (43). Another possibility is that the tumor immunosuppression established in the mTDLN, as previously reported (44), inhibits primed CD8⁺ lymphocytes to become fully active and later exhausted, even in GO patients. Indeed, the dynamics of CD8⁺ T cell responses to cancer immunotherapy was found to be disrupted in human mTDLN (45). Also, we found more Foxp3⁺ T regulatory cells (Treg) in BO patients. Beyond the negative regulatory function of Treg on effector T lymphocytes, it has recently been reported that depletion of Treg can induce HEV neogenesis through an alternative mechanism involving CD8⁺ cells activation, different from HEV ontogenic development (46). Thus, in a context of strong immune stimulation with a vaccine and a Th1 polarizing adjuvant (i.e. VACCIMEL plus BCG), some CM patients who are capable of developing FAPS in their LN could generate a potent immune response associated with GO. Other patients may have an impaired capacity to develop such structures and thus could be less efficient in mounting effective priming upon vaccination.

Regarding CD20⁺ B lymphocytes, they were mostly located in proliferating aggregates (as shown by the double positive staining with CD20 and Ki-67) and close to the tumor or marginally within it (Figures 1, 5–7). CD20⁺ aggregates were preferentially found in GO patients. Strikingly, a lack of CD21⁺ FDC network as shown by the absence of CD21 and the presence of CD11c⁺ cells in GO and BO patients suggest a disruption in B-cell function in mTDLN, although the presence of CD21^- FDC cannot be excluded. Of note, the role of CD20⁺ aggregates remains unknown, although most of them are clearly different from the mature TLS described in several tumors such as lung carcinomas (47).

The role of TAM in cancer has been extensively studied and debated, and two extreme types of TAM have been described: M1, which is detrimental to tumors; and M2, which is immunosuppressive and, therefore, induces tolerance and is often associated with a bad prognosis (48). However, transcriptomic analyses have revealed that TAM have a mixed phenotype (49). In a first approach, we used CD68, a type I glycosylated lysosomal membrane protein, to determine the whole TAM population (50). CD163, a glucocorticoid-inducible member of the scavenger receptor cysteine-rich family, expressed in innate immune cell populations (51), was used to detect M2 TAM (52). Neither the percentages of CD68⁺ nor CD163⁺ TAM attained statistical significance between GO and BO patients. An exploratory multiplex study performed on four patients revealed the extreme plasticity of TAM. Thus, CD68⁺CD64⁺ cells were only found in GO patients. Since myeloid cells express CD64 (FcyRI), a high-affinity receptor, the phagocytic role of tumor cells opsonized by locally produced Abs should be studied. Of note, CD64⁺ TAM were highly expressed in BO patients, both in intra-and peri-tumoral areas. Recently, Martinek et al. analyzed CM metastasis and found that CD14⁺ TAM displayed different signatures in tumor nests from those in tumor stroma associated with long-term survival (53).

It should be emphasized that both cohorts received adjuvant IT, since 22/29 patients were treated with VACCIMEL and 6/29 patients received IFN-α2b.

We have previously shown that peripheral lymphocytes, which in most patients were non-reactive to tumor-associated antigens and/or neoantigens prior to vaccination, became reactive after treatment with BCG-containing VACCIMEL, thus pointing to an increased number or awakening of reactive T-cell clones (14, 41). We provide here preliminary evidence that most analyzed GO patients (6/7 tested) reacted to vaccination by producing a larger number of CD8 lymphocyte clones against melanoma antigens than BO patients. However, some BO patients, (2/5 tested) who also produced reactive clones against melanoma antigens, failed to control the disease, probably due to a highly proliferating tumor, one of the ominous characteristics of melanoma.

Although radical lymphadenectomy is no longer mandatory in stage III CM patients, sentinel TDLN exeresis is still performed in patients with thick primary melanomas (54) and radical LN resection may be performed in more advanced stage III patients. It will be of interest to study prospectively whether the analysis here described on mTDLN may predict the evolution and response to different treatments.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

The studies involving humans were approved by Ethics Committee of the Alexander Fleming Institute (Buenos Aires, Argentina). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

JM: conception and design, collection and assembly of data, and manuscript writing. MA, MB and MP: collection and assembly of data, data analysis and interpretation, and manuscript writing. J-LT and M-CD-N: data interpretation and manuscript writing. AB and MPo: collection and assembly of data, data analysis and interpretation. AB: medical pathologist (Argentina); MA: Associate Researcher from CONICET (Argentina); MPa: PhD student (France); MPo: graduate student (Argentina); M-CD-N: Research Director at Inserm and Head of laboratory (France); J-LT: Research Director emeritus at Inserm (France); MB: Independent Researcher from CONICET and sub-Head of laboratory (Argentina); JM: Superior Researcher emeritus from CONICET and Head of laboratory (Argentina). All authors contributed to the article and approved the submitted version.