The phase I study included patients with CM stages IIB, IIC, III, and IV (9). For the analysis here reported, patients at stage IV have been excluded. The Phase II CASVAC-0401 study included patients with CM stages IIB, IIC, and III (10). Both studies were carried out at the Instituto Alexander Fleming and were approved by the Ethics Committee of that institution. All patients signed informed consent following Helsinki protocols, and patient #5 (Pt#5) consented publication of her anonymized data.

ITM biopsies from Pt#5 obtained before and after VACCIMEL immunization, and following nivolumab treatment, were formalin-fixed and paraffin embedded for histological analyses. From the latter ITM, a fresh punch biopsy was preserved at −80°C in RNAlater (Ambion) for RNA-seq analysis. Five-micrometer tumor sections were stained with anti-human monoclonal Abs to HLA- class I (A, B, C) (clone EMR8-5, Abcam), PMEL (clone hmb45, Dako), TYR (clone T311, Santa Cruz), CD8 (cloneC8/144, Dako), CD68 (clone PG-M1, Abcam), CD11c (clone EP1347Y, Abcam), DC-LAMP (clone 1010E1.01, Dendritics), and CD163 (clone 10D6, Invitrogen). The Avidin-Biotin-Peroxidase (ABC) system (Vectastain, Vector Labs) was used. Sections were examined by optical microscopy (Olympus BX40 microscope, DP2-BSW software). Quantification of the tumor-infiltrating immune populations in the different ITM biopsies was performed on the 1-mm² hot spot area by duplicate (ImageJ software), distinguishing zones with peritumoral (PT) and intratumoral (IT) infiltrate. Immune cell counts in the different ITM biopsies were compared by Tukey’s multiple comparisons test; p < 0.05 was considered as statistically significant.

Total RNA from frozen tumor was isolated with RNAaqueous Micro kit (Life Technologies); after DNase I treatment, RNA was quantified by a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The RNA integrity number (RIN) algorithm was used to determine RNA quality on the Agilent 2100 Bioanalyzer using the Eukaryote Total RNA Nano kit (Agilent Technologies, Waldbronn, Germany). Finally, the generated libraries were sequenced on the Illumina HiSeq 2500 platform (Illumina, San Diego, CA, USA). In the bioinformatic analysis of the transcriptomic data, the FastQC software (18) was used to analyze the quality of raw reads. Sequencing reads were mapped on the human reference genome (GRCh38) using the TopHat2 software (19). The low-quality readings were removed with Picard Tools [Picard Tools By Broad Institute, available online at: http://broadinstitute.github.io/picard/ (accessed August28, 2020)] and the unmapped and non-properly paired reads were remapped using the BWA-MEM algorithm (20). Gene and isoform prediction were estimated using the Cufflinks method (21). The HTSeq software (v.0.6.0) (22) was used to calculate gene expression levels. The gene counts were normalized using the methods described in DESeq2 package (23). To explore the abundance of genes associated to the immune system, gene lists were collected and downloaded from two databases (24), the Immunology Database and Analysis Portal System (ImmPort) (https://immport.niaid.nih.gov) and InnateDB (http://www.innatedb.ca/). Manually curated gene subsets were also analyzed.

RNASeq data from the dermic metastasis was uploaded to the European Nucleotide Archive (ENA, EMBL EBI); the corresponding accession link is: https://www.ebi.ac.uk/ena/browser/view/PRJEB49671.

HLA haplotypes (class I and class II) were determined on PBMC isolated from pt#5 by Scisco Genetics 25, Nelson). Haplotype: HLA-A: 02:01:01/03:01:01; HLA-B 07:02:01/18:01:01; HLA-C: 07:01:01/07:02:01; DPA1: 01:03:01; DPB1: 04:01:01; DQA1: 01:02:01/05:05:01; DQB1: 03:01:01/06:02:01; DRB1: 11:04:01/15:01:01; DRB345: 02:02:01/01:01:01. We selected HLA-A*0201-restricted peptides corresponding to non-mutated PMEL, tyrosinase (TYR), MAGE-B2, and SOX-2 antigens, which were expressed in the vaccine cells, from the TANTIGEN DataBase (http://projects.met-hilab.org/tadb/). Selected peptides were either T-cell epitopes previously identified in functional assays (in vitro and/or in vivo) or HLA ligands as determined by physical detection (25). Tested peptides are detailed in Table 1 . PBMC samples were thawed and seeded (1 × 10⁶) in 1 ml of CTL medium consisting of RPMI 1640 supplemented with 10% heat-inactivated human AB sera, 2 mM glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 2.5 mM HEPES, and 50 U/ml IL-2, in 24-well plates. PBMCs were stimulated with peptides (10 μg/ml) and cultured at 37°C, in 5% CO₂ for 12 days (effector cells). Every 3 days, fresh CTL medium with IL-2 was added. At day 10, additional PBMC samples were thawed, percentages of CD20+ and CD14+ cells (Ag-presenting cells, APCs) were recorded by flow cytometry, and cells were pulsed with peptides for 48 h. At day 12, APCs were treated with mitomycin C (50 μg/ml) for 1 h, washed twice with PBS, and resuspended in RPMI 1640 supplemented with FBS. Effector cells (4 × 10⁴) were seeded in 96-well plates previously coated with 5 μg/ml mouse anti-human IFNγ, and APCs were added in a 3:1 ratio (1.3 × 10⁴ CD20+ plus CD14+ cells/well) and cultured O.N. As a positive control, PBMCs (4 × 10⁴) were seeded and stimulated with 30 ng/ml OKT3 plus 1/1,000 PHA (M form, Gibco Life Technologies). As a negative control, non-stimulated cells were co-cultured with non-pulsed APC. Each experimental condition was performed in triplicate. Background baseline was calculated as the average number of spots present in non-stimulated cells for each time point sample (PRE, POST1, POST2, and POST3 samples, obtained at 0, 12, 18, and 25 months throughout VACCIMEL immunization). ELISPOT plates were developed as previously described (26). Plates were scanned using an AIDiSPOT ELR088IFL analyzer to quantify the number of spots per well; 350 spots/well were set as the maximum quantification limit.

We compared the distribution of DMFS of VACCIMEL-treated patients with those obtained by nivolumab (5), pembrolizumab (6), ipilimumab (27), and dabrafenib plus trametinib (28) treatments. Guyot’s iterative algorithm was applied to digitized published Kaplan–Meier curves to reconstruct individual patient data (29, 30). The digitization of source Kaplan–Meier curves was performed using WebPlotDigitizer (https://automeris.io/WebPlotDigitizer/). Survival analysis, including Kaplan–Meier analysis and Cox model fitting, was performed using R software (R Core Team 2020) using “survival” and “survminer” packages. A forest plot of hazard ratio was used to compare DMFS of different treatment vs. VACCIMEL. p-value< 0.05 was considered statistically significant.

The results of the Phase II study CASVAC-0401 demonstrated that vaccination with VACCIMEL plus BCG and GM-CSF was superior than medium-dose IFNα2b in terms of DMFS, with a minimum and maximum follow-up of 23 and 84 months, respectively (10). The CASVAC-0401 study was updated as of October 5, 2021, with a minimum and maximum follow-up of 91 months and 150 months (mean = 121 months). As represented by Kaplan–Meier analysis, DMFS was still significantly longer in vaccinated patients than in IFNα2b-treated patients (p = 0.035), with a median DMFS of 96 months for VACCIMEL and 13 months for IFNα2b ( Figure 1A ). The risk–benefit ratio of VACCIMEL remained unchanged. In order to analyze more patients with longer follow-up, we made a combined analysis of DMFS of vaccinated patients at stages IIB, IIC, and III recruited in the Phase I (9) and Phase II (10) studies (n = 30, Table 2 ). The mean age of the patients was 42.8 years, including 16 women (53%) and 14 men (47%). Clinical stages of all patients were re-determined as per version 8, AJCC (31). Of the 30 CM patients analyzed, 16/30 were at stage IIIC (53%); 9/30 were at stage IIIB (30%); 3/30 patients (10%) were at stage IIB; 1/30 (3%) was at stage IIIA; and 1/30 (3%) was at stage IIC. The minimum follow-up was 91 months and the maximum follow-up was 223 months. Median DMFS of the assembled population was 169 months. Also, overall survival (OS) data are shown ( Table 2 ), but further analysis was not pursued (see under Discussion). We compared the DMFS of combined VACCIMEL data with other studies performed in similar populations of CM patients also treated in adjuvancy: ipilimumab versus placebo (4); pembrolizumab versus placebo (6); nivolumab versus ipilimumab (5); and dabrafenib/trametinib versus placebo (28) ( Figure 1B ). Analysis at 48 months follow-up between VACCIMEL and ipilimumab, nivolumab, pembrolizumab, and dabrafenib/trametinib, with Kaplan–Meier and Cox-regression models, did not attain statistically significant differences. Hazard ratio of the different treatments compared to VACCIMEL revealed no relevant statistically significant differences ( Figure 1C ). The only marginal significative difference (p = 0.052) was with ipilimumab presenting a hazard ratio of 1.81 versus VACCIMEL (VACCIMEL better). Therefore, a non-inferiority of VACCIMEL respect to the other treatments at 48 months follow up is suggested.

Based on studies performed in Macacus rhesus that suggested temporary reactivation of chronic SIV infections with anti-PD-1 treatment, clinical assays performed with anti-PD-1 mAbs precluded the use of live-attenuated vaccines during anti-PD-1 therapy (15, 16). The information about the effect of anti-PD-1 mAbs on tuberculosis evolution is conflicting, with evidence suggesting awakening of infection (32, 33), while other authors (34) suggest the opposite effect. Therefore, our hypothesis that vaccination expands lymphocytic clones that could be further amplified with anti-PD-1 treatment required previously answering the important question if treatment with VACCIMEL, which includes as adjuvants BCG and GM-CSF, induced BCG reactivation and infection after anti-PD-1 treatment. Three patients from our clinical assay CASVAC-0401 (#5, #7, and #19) ( Table 2 ) who after ending the CASVAC-0401 protocol progressed to metastatic disease and were treated with nivolumab achieved complete responses without any signs of toxicity due to BCG infection (JM, personal communication). As an example, Pt #5 shall be described in detail, since this patient received 16 i.d. vaccine doses during treatment and follow-up. Due to disease progression after VACCIMEL treatment, the patient underwent various treatments, including 4 years of anti-PD-1 treatment. Besides, due to her ITM disease, several biopsies could be obtained along her treatment.

The timeline of Pt#5 clinical evolution with most relevant events is shown in Figure 2A . She is a 44-year-old female who in 02/2008 presented in her right leg a superficial extensive melanoma, Breslow 0.89 mm, ulcerated, in vertical growth phase, and with brisk lymphocyte infiltration. Margins were enlarged and two sentinel node biopsies were negative for metastasis. In 03/2009, cutaneous satellites appeared, progressing into ITM, many of which were removed. On 07/2010, the patient entered the CASVAC-0401 Phase II protocol in the vaccine arm and completed the 2-year protocol in 08/2012, after which she received three more vaccinations. The cutaneous lesions, mostly located at the inner part of her right leg, which had remained stable until then, progressively began to increase in number and size from 06/2013 ( Figure 2B ). Biopsies documented melanoma metastases, positive for BRAF^V600E oncogene mutation, and PET/CT revealed no distant metastasis. Given the impossibility of complete tumor removal (Stage IIIB), she began treatment with vemurafenib 960 mg BID, with complete clinical response until 2015, when ITM relapsed without any sign of distant metastasis by PET/CT. Two cycles of hyperthermic perfusion with melphalan were performed, and flattening of the lesions without complete resolution was observed After recurrence of the lesions in 2016 ( Figure 2C ), mainly at the posterior part of the leg, nivolumab (240 mg every 14 days) was started with an excellent initial response, showing almost disappearance of the lesions and good tolerance ( Figure 2D ). Four years later, regional ITM reappeared in large numbers, as multiple 1- to 3-mm amelanotic tumors at the back of the leg, and nivolumab was interrupted ( Figure 2E ). As confirmed by CAT and PET-CAT scans, the disease continued confined to the right leg and no lung BCG nodules were observed. Two punch biopsies were obtained at 10/2020 for histology, immunohistochemistry, and RNA-seq analysis. Pt#5 started treatment with dabrafenib/trametinib (12/2020) that continues until the writing of this paper with stable disease. Histology and immunostaining comparative analysis were performed on ITM biopsies obtained before and after VACCIMEL immunization, along with biopsies after nivolumab treatment ( Figure 2A ), of which a fresh punch biopsy was obtained for RNA-seq analysis.

In order to assert if VACCIMEL plus BCG elicited toxicity after posterior anti-PD-1 treatment, immune proficiency of Pt #5 had to be previously demonstrated. The frequency of T-cell clones recognizing synthetic peptides of shared TAA was determined by ELISPOT in PBMC samples obtained throughout the CASVAC-0401 protocol (PRE, 6, 12, and 25 months). VACCIMEL induced expansion of a T-cell response to HLA-A0201-restricted peptides from shared non-mutated melanoma antigens ( Figures 3A, B ). It is important to note that reactivity was practically nil before vaccination (PRE sample), although PMEL and tyrosinase (TYR) were strongly expressed in the patient’s tumor ( Figure 4 ). For most of the assayed peptides, a clear increase in the immune response was observed during the 2-year vaccination protocol. It should be noted that the intensity of the immune response towards the different antigens appeared unrelated to their expression abundancy in the vaccine or in the tumor ( Figure 3C ). Thus, patient’s immune response to a low-expressing gene as SOX2 in the vaccine was equally strong as the response to highly expressed genes as PMEL and TYR.

Immunohistochemistry was performed on PRE- and POST-vaccination biopsies, and of that obtained after nivolumab recurrence, to determine if any changes in the expression of relevant TAA and immune cells occurred during the patient’s evolution ( Figure 4 ). HLA expression was maintained all along the disease evolution ( Figures 4A–C ), and the expression of TAA gp100/Pmel and tyrosinase/Tyr was also abundant ( Figures 4D–I ). The number of CD8⁺ lymphocytes, CD11c⁺ antigen-presenting cells, and CD68⁺ macrophages increased following successive treatments ( Figures 4J–R and Supplementary Figure 1 ). In addition, the biopsy obtained after nivolumab relapse was analyzed by RNASeq. We searched in this unique sample the expression of genes related to the immune system. The expression of genes involved in antigen processing and presentation was high, as well as TCR Signaling, IFN, TNF, and TGFβ receptors ( Figure 5A ). Expression of HLA class I and class II molecules as well as of β2M was abundant ( Figure 5B ). Also, several immunosuppressive molecules such as TGFB1, HLA-E, and galectins 1 and 3 were abundant; no IL-10 expression was detected ( Figure 5C ). Immune checkpoints (receptors and ligands) such as CTLA-4, PD-1 (PDCD1), and PD-L1 (CD274) had low expression levels in this post-treatment biopsy ( Figure 5D ). These results pointed to a coexistence in the tumor of immunostimulatory and immunosuppressive characteristics.