This was a monocentric prospective observational study that included plasma samples from an existing biobank ( LiquiMel ). For the biobank, citrate blood samples from routinely drawn blood were collected with the patient's informed consent, and centrifuged for 10 min at 1800 x g. 2 ml Plasma was aliquoted and frozen at -80°C until miRNA multiplexing. Blood values and other clinically relevant parameters were known from patient records. Patients` plasma samples were included in different cohorts of melanoma stages, with and without metastasis, before or during therapy. Cohort 1: Melanoma patients in stages I - III with no or only locoregional metastatic sites (lymph nodes) (without any form of therapy) compared to treatment-naive stage IV melanoma patients who actively metastasize (Table S2). Cohort 2: Treatment-naive stage IV melanoma patients who actively metastasize at therapy-start (baseline) compared to an independent group of stage IV melanoma patients with the first occurrence of progressive disease after the start of immune therapy (progress) (Table S4). Cohort 3: Treatment-naive, actively metastasizing stage IV melanoma patients with negative, compared to low and high PD-L1 immunostaining of melanoma tissue (Table S6). Because the time point of sampling is important in this study, the baseline samples were taken directly on the day of the first therapy before infusion. Progress samples were collected on average within 16 days before or after the staging in which the progressive disease was first diagnosed. Progressive disease is defined as at least a 20 % growth in the size of the metastasis or as the appearance of a new metastasis, which means increasing tumor mass 31.

We got the ethical approval for this study from the IRB Ethical Review Board of Ham-burg and the Medical Association of Lower Saxony for analysis of human materials and all patients gave their written informed consent for this study.

The transcription of miRNAs was measured via flow cytometric quantification of barcode-labeled fluorescent miRNA hydrogel micro-particle ( FirePlex Particle Technology , Abcam) according to the manufacturer s protocol as previously described 32. Briefly, 25 µl of protease-digested plasma was added to customized fireplex particles (~35 µL) and incubated under shaking (1125 rpm) at 37 °C for 60 min. After binding of miRNAs, the particles, which contain a complementary sequence, were rinsed twice with rinse buffer A followed by a labeling reaction (RT, 60 min, 1125 rpm). During labelling each miRNA is ligated to two linkers. After washing of miRNA-linker-bound particles with rinse buffers B and A, the miRNA-linkers were eluted with H2O from the particles at 55 °C. The miRNA linkers were amplified by PCR and the PCR products were transferred back to the initial particles (shaking at 1125 rpm, 37 °C, 60 min). A fluorescent reporter was added (RT, 15 min, 1125 rpm) that binds to the miRNA linker complex. Fluorescence of the particles was then measured by flow cytometry (with Guava easycyte 8HT, Millipore). The raw data obtained from flow cytometry were then processed with the FirePlex Analysis Workbench software (Abcam). For the normalization, the geometric mean of the five overall most stably expressed miRNAs was determined (regardless of stage, therapy, or outcome) (Table S1).

PD-L1 staining in melanoma tumor tissue and determination of the tumor proportion score (TPS) was done as part of the ongoing diagnosis. The immunohistochemical staining was carried out using monoclonal rabbit anti-PD-L1 (clone 28-8) according to the manufacturer's protocol with PD-L1 IHC 28-8 pharmDx Kit (Dako). TPS was scored by a pathologist according to the manufacturer s interpretation manual 33.

Statistical analysis was performed using R 4.1.1 (https://www.r-project.org/). miRNA expression was log2-transformed after setting negative values to 0 and adding 1 to increase data symmetry and approximate a normal distribution. miRNAs were considered differentially expressed if significantly altered and the calculated transcriptional fold change was ≥ 1.5-fold (log2-FC: ≥ 0.585). For each hypothesis, we conducted a two-way analysis of variance (ANOVA) including the different miRNAs and the respective grouping variable for each cohort with an interaction term. ANOVA was performed with heteroscedasticity-corrected covariance matrices (hccm) to adjust for unequal variances. Pair-wise differences in each subgroup were then compared by the Games-Howell post-hoc test. Results were considered statistically significant if p < 0.05.

Statistical differences for baseline demographics were tested with the chi^² test for categorical data and ANOVA (> 2 groups) or Welch s t-test (2 groups) for numerical data.

Pathway analysis was performed with the web tool miRTargetLink 2.0 (https://ccb-compute.cs. uni-saarland.de/mirtargetlink2) using a unidirectional search with a set of miRNAs and performing enrichment analysis on this miRNA set in the database of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.

Survival analyses were conducted with the R packages survival and survminer 3435. OS is presented in Kaplan-Meier plots; hazard ratios were calculated with multi-variable Cox regression as a post-hoc analysis.

For this study, we included a total of 153 patients in different stages of cutaneous malignant melanoma with a median age of 64 (range: 17.0, 96.0) years. For 101 (66.1%) of these patients the v-Raf murine sarcoma viral oncogene homolog B (BRAF) status was known. 59 patients (38.6%) had a wildtype BRAF and 42 patients (27.5%) had a BRAF V600 mutation. Ten patients (6.5%) were negative in PD-L1 immunostaining of the tumor tissue, 20 were positive (13.1%) and 40 (26.1%) had elevated serum LDH concentrations.

104 patients were in stages I to III and had no distant metastases. In total 17 patients were in stage I (11.1%), 30 in stage II (19.6%), and 57 in stage III (37.3%). All these patients were therapy naïve at blood sampling. 49 patients were in stage IV with active distant metastases and one or more metastatic sites (Table 1).

To answer the question of whether there is a difference in miRNA expression pattern between patients without active distant metastases in stages I to III and actively metastasizing patients in stage IV, we compared the expression of 63 miRNAs (Table S1) based on hypothesis-driven literature search, meaning the miRNAs were selected based on their reported association with cancer development as oncomirs or tumor suppressors. Our selection did not exclusively focus on miRNAs specific to malignant melanoma but rather encompassed those involved in general carcinogenesis. In cohort one (Table S2) blood plasma of 104 patients in lower stages (< IV) and 35 actively metastasizing patients in stage IV were examined. The correlation with anatomical sites of metastases is not discussed in this report. To exclude possible influences of any kind of melanoma therapy we only included therapy naïve patients that did not get immunotherapy or targeted therapy at blood sampling. The group < IV consisted of 17 patients (12.2%) in stage I, 30 patients (21.6%) in stage II, and 57 patients (41.0%) in stage III. The overall median age of patients in this cohort is 66.0 ^{17.0 96.0} years. The gender distribution in stages < IV was n = 51 female (49.04%) and n = 53 male (50.96%), but in stage IV there was a higher male proportion with 62.9 % (n = 22) in contrast to females with 37.1% (n = 13). In stage IV 12 patients had one distant metastatic site (34.3%), 13 patients had two (37.1%) and 10 patients had three or more distant metastatic sites (28.6%).

Two-way ANOVA (analysis of variance) was conducted to examine the effects of clinical stage and type of miRNA on miRNA expression. There was no significant interaction between the type of miRNA and certain clinical stages (Figure 1). All of these miRNAs were downregulated in stages < IV and were upregulated in stage IV, which may be indicative of an increase in tumor mass by metastases and a correspondingly higher amount of secreted miRNAs into patients’ blood. This would correlate with patients’ data for tumor burden, measured during staging by radiologists (progressive disease is defined as at least a 20 percent growth in the size of the metastasis or appearance of a new metastasis 31).

Overrepresentation analysis on this set of nine miRNAs showed several enriched pathways involved in carcinogenesis, including microRNAs in cancer (p-value = 1,55e-15), proteoglycans in cancer (p-value = 2.02e-12), or pathways in cancer (p-value = 2.59e-11) (Table S3). Moreover, receiver operating characteristic (ROC) analyses showed that expression levels of this set of miRNAs are useful better classifiers for discriminating stage IV patients from patients with lower stages than of the single miRNAs alone (AUC = 0.739, p = 1.485513e-05) (Figure 2, Figure S1).

Therapy naïve stage IV melanoma patients show different miRNA expression patterns than melanoma patients at disease progress under therapy

Data from cohort 2 were used (Table S4) to compare miRNA expression between therapy naïve patients with stage IV melanoma (baseline, n = 35) and stage IV melanoma patients at first disease progression after the start of immune therapy (progress, n = 14).

The median age was higher in the baseline subgroup with 67.0 (20.0, 96.0) years compared to 57.5 (35.0, 76.0) years in the progress subgroup. The proportion of patients with BRAF mutated melanoma was higher in the progress subgroup (64.3% vs. 28.6%), however, LDH decreased during therapy with 62.9% of patients showing elevated LDH at baseline and 35.7% of patients at progress. Other covariates showed similar distribution patterns between the two groups (Table S4).

A two-way ANOVA was conducted to examine the effects of baseline-/progress-patients and type of miRNA on miRNA expression. No interaction effect could be observed, but a significant main effect for the two groups of melanoma patients (baseline vs. progress) was present (ANOVA, F1,2952 = 61.993, p = 4.80e-15). A post-hoc Games-Howell test showed that on the individual miRNA level, five miRNAs ‒ namely miR-146b-5p, miR-185-5p, miR-197-3p, miR-199a-5p, and miR-338-3p ‒ were downregulated in patients at progress (Figure 3). A subsequent pathway analysis yields significantly enriched pathways for this 5-miRNA set including pathways in cancer (p-value = 3.12335e-10), proteoglycans in cancer (p-value = 3.12335e-10), or microRNAs in cancer (p-value = 2.68618e-09) (Table S5).

It might, however, be interesting to note that there is no overlap in the miRNA expression patterns between cohort 1 and 2 (compare Figure.1 and Figure.3), although the first cohort discriminates between patients in stage < IV and IV, where stage IV group already represents metastasized melanoma patients which underwent some further progression compared to the stage < IV group patients. Also, the trend in miRNA expression is different between the two cohorts (in cohort 1 we detected an upregulation of the miRNAs involved (Figure 1), whereas in cohort 2 we found a downregulation of miRNAs with further progression). Apparently, in stage IV melanoma patients (cohort 2) further progression after the start of therapy might occur and is “controlled” by another set of miRNAs, with possibly different targets.

For 20 therapy naïve baseline patients, the tumor proportion score (TPS) of PD-L1 positive tumor cells was known by histological staining. These patients (cohort 3) were grouped by their TPS into negative (<1%), low (1-20%), and high (>20%) PD-L1 expression in tumor tissue. The median age of this cohort was 63.5 (20.0, 96.0) years with eight (40.0%) female and 12 (60.0%) male patients. Six patients (30.0%) showed a BRAF mutation, and 16 patients (80.0%) had elevated LDH levels (Table S6).

Two-way ANOVA revealed a significant difference in miRNA expression between the TPS groups (ANOVA, F2,1071 = 18.578, p = 1.17e-08). Two miRNAs (miR-150-5p and miR-199a-5p) were found to be able to distinguish between the PD-L1 negative, low, and high patients (Figure 4). In these two miRNAs, there was a significant upregulation from negative to highly positive PD-L1 tumors.

Examining these preliminary results, we speculated whether the miRNA expression pattern could be correlated with patients’ survival characteristics in our basic cohort (Table 1). Therefore, we investigated the OS of n=152 patients, grouped by “low” or “high”. In this setting, “low” corresponds to the expression of miRNA-150-5p and miRNA-199a-5p lower than the median of the corresponding miRNA for all members of the cohort (n=152), whereas “high” corresponds to higher expression than the median, respectively (Figure 5). miRNA-199a-5p did not show significant differences between the groups.

However, miR-150-5p exhibited significant differences in favor of high expression in a Kaplan-Meier analysis (p = 0.0022). To adjust for baseline differences in patients with lower and higher miR-150-5p expression than the median (Table S8), respectively, we performed a successive multivariable Cox regression (Figure 6). The previously observed differences remained significant with a more favorable OS for patients with higher miR-150-5p expression (HR (95%): 0.39 (0.16-0.92, p = 0.031). Furthermore, as expected, a higher stage and age were associated with a worse OS (HR (95%): 8.71 (2,68-28.36, p < 0,001; HR (95%): 1.04 (1.01-1.06, p = 0.017, respectively), whereas other covariates like sex, BRAF mutation status, LDH or melanoma type were not significantly altered.

These findings indicate that overexpression of only one miRNA (miR-150-5p) might be predictive not only for high PD-L1 expression in melanoma patients but also for OS of melanoma patients.

Melanoma of the skin is the deadliest form of skin cancer with increasing incidences. Melanoma has been recognized for a long time as a therapy-resistant cancer with no (or little) success of therapies in the past. However, ICI has revolutionized therapy outcomes in the last decade, which has been demonstrated by an increase in OS, PFS, and other surrogate survival parameters in many studies up to now. However, positive response to ICI seems to be restricted to only a subgroup of melanoma patients, and many patients do not respond and/or develop resistance to therapy. The reasons for this are still not understood very well. Thus, there is an urgent need for diagnostic, progression, and predictive biomarkers. It is still necessary to have early diagnostic tools to differentiate between healthy and melanoma-suffering individuals, and, in case of a melanoma diagnosis, to predict the progression of disease. In times of modern, individualized terms of precision medicine (targeted therapies as well as ICI) it is, furthermore, important to find predictive biomarkers, to characterize patients in their ability to respond to a certain therapy (ideally before the start of therapy) and how these patients might respond (or non-respond) during therapy.

A variety of markers has been described in the past including gene expression, ctDNA (circulating tumor DNA), circulating tumor cells, epigenetic markers like DNA methylation and histone modifications, BRAF (and other gene) mutational status, as well as protein expression (e.g. PD-L1-expression in tissue and liquid biopsies 1529. However, novel circulating biomarkers are still lacking, and, to date, the only ( validated ) marker that has been incorporated in clinical routine use seems to be LDH, which appears to be an independent, significant predictor of survival (when detected at high serum levels), although only in advanced melanoma stages 29.

Search for new, reliable biomarkers for diagnostics, progression, and prediction is, therefore, still ongoing and urgently needs to be investigated and to be validated for clinical use.

It turned out in the last years that circulating miRNAs (in liquid biopsies, e.g. blood (-plasma)) might be good candidates for the future to be used as biomarkers for disease progression, therapy resistance, and therapy response 1636.

In this paper, we used a clinically applicable, fast, and reliable assay (Abcam FirePlex) to measure miRNA expression patterns in the plasma of melanoma patients in different stages of disease and therapy. We have been able to show, that

- A plasma circulating miRNA expression pattern can discriminate between stage < IV and stage IV melanoma patients.

Several candidates for this kind of discrimination have been discussed in the literature 37383940. However, their use in a clinical setting (fast, reliable, validated) still awaits application. Using the Abcam FirePlex assay we have been able to show here, that on liquid biopsy (melanoma plasma) level, miRNAs are good candidates to discriminate between stages of melanoma progression (< IV vs. IV). The identified nine miRNAs (miR-132-3p, -193b-3p, -200c-3p, -204-5p, -205-5p, -211-5p, -221-3p, -34b-3p and -494-3p) are all known to be involved in melanoma development and progression and might belong to a group of outstanding, pathway determining biomarker candidates for the future of precision therapy in the field of malignant melanoma 414243444546.

- Therapy naive stage IV melanoma patients show a different miRNA expression pattern than melanoma patients at further disease progress under therapy.

This might be an important finding, indicating that further progression in stage IV melanoma patients is controlled by another miRNA pattern and their associated targets (genes) compared to simple differentiation between low stages (< IV) vs. advanced stages (IV) (see cohort1, Figure 1). There is no overlap between the significant miRNA expression patterns that discriminate between stage < IV and IV patients and those who suffered further progress in stage IV (results from cohorts 1 and 2, see Figs. 1 and These results might be indicative of the participation of different miRNA pathways involved in melanoma development and should be considered, after further validation, in next-generation therapy regimes. Further pathway analyses of gene targets associated with the miRNA changes have to show, specifically, which regulatory pathways are involved and how their knowledge can be used for better clinical prognosis and prediction.

- Expression of liquid biopsy miR-150-5p and miR-199a-5p can discriminate between PD-L1 negative and high expressing patients.

In a small cohort (n=20, cohort 3) we have been able to show, that two miRNAs (miR-150-5p and miR-199a-5p) can discriminate between negative expression (< 1%), low expression (1 - 20 %), and high expression (> 20%) of PD-L1 in liquid biopsies (plasma) of melanoma patients according to the TPS for immune-histologically detection of PD-L1 in melanoma tissue (Figure 4). There was a significant upregulation of both miRNAs from negative to high expression. This finding might indicate that plasma detection of this duo of miRNAs might reveal PD-L1 expression, which is seen as an important parameter in predicting therapy response for ICI in melanoma patients. The advantage of this liquid biopsy-based detection would be the use of a non-invasive assay, which might facilitate its clinical application.

- These miRNAs have recently been shown to be involved in an indirect regulation of PD-L1 expression. MiR-150 negatively regulates STAT1 and EIF4B, whereas miR-199a negatively regulates mTOR expression 47. A subsequent KEGG enrichment analysis for these two miRNAs, furthermore, indicates the involvement of different cancer-associated pathways like regulation of TWIST1 (p-value = 8.73 e-05) or immune response (p = 1.39e-04; Table S7). TWIST1 knockdown in breast cancer cells was reported to downregulate their metastatic abilities 48. In Melanoma TWIST1 is upregulated 49. Its overexpression triggers EMT 50 and is likely to affect progression. Hence, the less the expression of these miRNAs, the worse it could be for the patient's overall survival.

-Expression of miR-150-5p is correlated with overall survival.

Besides being indicative of the level of PD-L1 expression in melanoma patients, the expression of miR-150-5p was also correlated to OS in a cohort of n=152 patients. Especially, its overexpression was correlated with better OS, according to Kaplan-Meier analysis (p = 0.0022) (Figure 5). This could be confirmed by a successive multivariable Cox regression, yielding HR of 0.39 (95% CI:0.16 - 0.92) for high expression of miRNA-150-5p compared to low expression of this miRNA (Figure 6).