Circulating Cell-Free SHOX2 DNA Methylation Is a Predictive, Prognostic, and Monitoring Biomarker in Adjuvant and Palliative Anti-PD-1-Treated Melanoma

Abstract

Background

The majority of metastatic melanoma patients initially do not respond or acquire resistance to anti-programmed cell death 1 (PD-1) immunotherapy. Liquid biopsy biomarkers might provide useful early response information and allow for personalized treatment decisions.

Methods

We prospectively assessed circulating cell-free SHOX2 DNA methylation (SHOX2 ccfDNAm) levels and their dynamic changes in blood plasma of melanoma patients by quantitative methylation-specific polymerase chain reaction. Patients were treated with either palliative (n = 42) or adjuvant (n = 55) anti-PD-1 immunotherapy. Moreover, we included n = 126 control patients without evidence of malignant disease. We analyzed SHOX2 ccfDNAm status prior to and 4 weeks after palliative treatment initiation with regard to outcome [objective response, progression-free survival (PFS), and overall survival (OS)]. In the adjuvant setting, we associated longitudinal SHOX2 ccfDNAm status with disease recurrence.

Results

Sensitivity was 60% with 25/42 melanoma patients showing increased SHOX2 ccfDNAm levels, whereas specificity was 98% with 123/126 (P < 0.001) control patients having SHOX2 ccfDNAm levels below cut-off. Pretreatment SHOX2 ccfDNAm status did not correlate with outcome; however, SHOX2 ccfDNAm negativity 4 weeks after palliative treatment initiation was strongly associated with improved survival [PFS: hazard ratio (HR) = 0.25, P = 0.002; OS: HR = 0.12, P = 0.007]. Pretreatment positive patients who reached SHOX2 ccfDNAm clearance after 4 weeks of immunotherapy showed an exceptionally beneficial outcome. SHOX2 ccfDNAm testing allowed for an early detection of distant metastases in adjuvant-treated melanoma patients.

Conclusions

Our study suggests SHOX2 ccfDNAm to be an early predictor of outcome in anti-PD-1 treated melanoma patients. SHOX2 ccfDNAm testing may aid individualized treatment decision-making.

Introduction

Currently, monoclonal antibodies targeting the immune checkpoints cytotoxic T lymphocyte-associated protein 4 (CTLA-4), programmed cell death 1 (PD-1), and lymphocyte-activation gene 3 are approved for the treatment of unresectable or metastatic melanoma (1). Moreover, the anti-PD-1 antibodies pembrolizumab and nivolumab have gained approval for adjuvant treatment of resected metastatic melanoma (1). Combination therapies with anti-PD-1 and anti-CTLA-4 inhibitors show objective response rates of approximately 61% (1). However, a large portion of patients initially do not respond or acquire resistance in the course of disease. Melanoma patients lacking response to anti-PD-1 immunotherapy may benefit from switching to immune checkpoint combination therapy or targeted therapy (2, 3). Treatment decisions are mainly based on clinical and radiological assessments. Besides other limitations, cross-sectional imaging is only valuable if applied regularly over a relatively long period of at least 3 to 6 months. Moreover, in patients undergoing immune checkpoint blockade (ICB), interpreting imaging can be ambiguous due to pseudoprogression (4). Therefore, complementary parameters that allow us to predict and monitor response prior to and early during therapy are needed. Currently, there are no standardized predictive or monitoring biomarkers for patients receiving ICB available that can assist treatment decisions (5). Ideally, such biomarkers would be testable in liquid biopsies, e.g., blood plasma.

Blood plasma contains a significant amount of circulating cell-free DNA (ccfDNA), mainly derived from hematopoietic cells, in both healthy individuals and cancer patients. Tumor cells release DNA into the circulation via apoptosis, necrosis, and active secretion (6). Hence, circulating tumor DNA (ctDNA) represents a fraction of the total ccfDNA. In recent years, ctDNA came into the focus as an analyte for diagnostic applications, including cancer screening, diagnosis, treatment response prediction, staging, risk stratification, prognosis, minimal residual disease detection, and response and recurrence monitoring (5, 7). In melanoma, ctDNA was suggested to have prognostic and predictive value in patients undergoing immunotherapy, targeted therapy, and T-cell transfer (8–10). Moreover, it has shown utility in disease monitoring and accurate classification of pseudoprogression (10, 11). The most common way to detect ctDNA in melanoma is polymerase chain reaction-based analysis aiming at the driver mutations BRAF (B-Raf proto-oncogene, serine/threonine kinase) codon 600 (BRAF^V600) and NRAS (NRAS proto-oncogene, GTPase) codon 61 (NRAS^Q61) (5, 7). However, approximately one-third of all melanomas are BRAF/NRAS wild-type, necessitating the application of multigene panels or next-generation sequencing for ctDNA detection (5). Almost 20 years ago, Hoon and colleagues suggested the analysis of aberrant DNA methylation in blood from melanoma patients (12, 13). DNA methylation testing harbors the potential to allow for a mutation-independent determination of ctDNA in melanoma.

Aberrant DNA methylation is a hallmark of cancer and is considered to play an essential role in carcinogenesis and tumor progression (14). Aberrant methylation patterns can be detected in ccfDNA of cancer patients and have already been implemented into clinical routine, e.g., as a U.S. Food and Drug Administration-approved colorectal cancer screening test (15). The clinical use of methylated ccfDNA biomarkers in melanoma has only been investigated in a few studies. Hypermethylation of the tumor suppressor gene Ras association domain family member 1 (RASSF1) was shown to enable the discrimination between melanoma patients and healthy controls and to be associated with negative response and worse overall survival (OS) in patients receiving biochemotherapy (12, 13, 16). So far, there is no evidence for a prognostic or predictive value of aberrant ccfDNA methylation (ccfDNAm) in melanoma patients undergoing immunotherapy or targeted therapy. Aberrantly methylated short-stature homeobox 2 (SHOX2) is a feature of several malignancies, including melanoma (17). SHOX2 is a homeobox gene, which encodes transcriptional regulators crucial for the development of multiple organs (18). Moreover, SHOX2 hypermethylation has been found in blood plasma from patients with various cancers, including lung cancer, renal cell carcinoma, colorectal cancer, biliary tract cancer, and head and neck squamous cell carcinoma (14, 19–23). In melanoma tissue, SHOX2 is hypermethylated in 93% of tumors at 90% specificity [area under the curve = 0.95 (95% confidence interval [CI], 0.93–0.97) of the receiver operating characteristic] (17), thereby representing a promising plasma biomarker candidate for melanoma. In advanced lung cancer and head and neck squamous cell carcinoma, SHOX2 ccfDNAm has been shown to have prognostic and predictive value and allowed for early detection of treatment failure/disease recurrence (21, 22).

In this study, we prospectively assessed the value of SHOX2 ccfDNAm in blood plasma of melanoma patients to predict and monitor response to anti-PD-1-based immunotherapy. Moreover, we evaluated the clinical performance with regard to recurrence monitoring in patients with locally advanced or metastasized melanoma who received adjuvant anti-PD-1 immunotherapy after surgical tumor resection.

Materials and Methods

Patient Cohorts and Ethics

Palliative treatment cohort

We prospectively analyzed plasma samples of n = 42 patients with metastatic (stage IIIC–IV) melanoma who were treated with palliative anti-PD-1-based immunotherapy at the University Hospital Bonn between 2017 and 2020. Blood samples were collected prior to (pretreatment) and during treatment (on-treatment). A study flow chart is shown in Supplemental Data 1. Pretreatment samples were obtained at the day of treatment initiation. Median time until first staging was 2.7 months [interquartile range (IQR): 2.4–3.6 months]; median follow-up time was 18.3 months (IQR: 11.0–34.4). Clinical and molecular baseline characteristics are listed in Table 1. Primary endpoints were progression-free survival (PFS) and OS. Survival time was defined as the time between treatment initiation and death/progress or censored at the last date of patient contact. Secondary endpoints included best overall response, objective response rate, and lead time gain regarding diagnosis of recurrence or progression, respectively. The response to therapy was assessed radiologically with computed tomography (CT) scans of the trunk and magnetic resonance imaging of the brain at 2- to 3-month intervals. In accordance with response evaluation criteria in solid tumors v.1.1 (RECIST v.1.1), response was described as either complete response, partial response, stable disease (SD), or progressive disease (PD).

Table 1.

Clinical and molecular baseline characteristics of the palliative treatment cohort.

Characteristic	Number [n] (%), median (IQR)	Median SHOX2 ccfDNAm [%] (IQR)	P value
All patients (%)	42 (100)	1.62 (0.00–17.31)	NA
Age (yrs.)	67 (IQR: 55–70)	NA	0.56
Sex (%)			0.010
Male	22 (52)	9.46 (0.47–27.59)
Female	20 (48)	0.00 (0.00–3.74)
AJCC stage (%)			NA
IIIc	1 (2)	0.00 (NA)
IV	41 (98)	1.71 (0.00–17.38)
M category (%)			0.12
M0	1 (2)	0.00 (NA)
M1a	3 (7)	2.69 (NA)
M1b	15 (36)	0.00 (0.00–14.64)
M1c	9 (21)	17.24 (2.10–29.90)
M1d	14 (33)	2.81 (0.30–16.70)
Number of metastatic sites (%)			0.002
<3	18 (43)	0.00 (0.00–2.72)
>2	24 (57)	9.46 (0.81–20.59)
Brain metastases (%)			0.64
No	27 (64)	1.71 (0.00–17.52)
Yes	15 (36)	1.53 (0.00–12.61)
Mutational status (%)			0.61
BRAF mutation	16 (38)	2.81 (0.00–18.02)
NRAS mutation	9 (21)	4.11 (0.00–57.01)
BRAF and NRAS wild-type	17 (40)	1.34 (0.00–10.97)
Disease origin (%)			0.40
Cutaneous	23 (55)	1.17 (0.00–11.57)
Acral	3 (7)	0.40 (NA)
Melanoma of unknown primary	11 (26)	8.07 (0.00–52.65)
Mucosal	4 (10)	1.35 (0.00–3.81)
Uveal	1 (2)	30.30 (NA)
First-line therapy (%)			0.40
Anti-PD-1 mAb pembrolizumab	19 (45)	0.63 (0.00–14.64)
Anti-PD-1 mAb nivolumab	3 (7)	0.00 (NA)
Anti-PD-1 + anti-CTLA-4 mAbs (nivolumab + ipilimumab)	20 (48)	6.62 (0.00–20.60)
Best overall response (%)			0.43
Progressive disease	13 (31)	2.69 (0.00–16.90)
Stable disease	12 (29)	1.41 (0.00–15.59)
Partial response	14 (33)	6.85 (0.00–36.91)
Complete response	3 (7)	0.00 (NA)

AJCC, American Joint Committee on Cancer; mAb, monoclonal antibody; NA, not applicable.

P values refer to Spearman correlation, Mann–Whitney U, and Kruskal–Wallis tests.

Adjuvant treatment cohort

The second cohort included n = 55 patients with locally advanced or metastatic (stage IIIA–IV) melanoma that received adjuvant ICB after surgical tumor resection at the University Hospital Bonn between 2017 and 2020 (Supplemental Data 1). Median follow-up time was 15.6 months (IQR: 11.7–29.6). Clinical and molecular characteristics are shown in Supplemental Table 1. The primary endpoint was tumor recurrence. The occurrence of lymph node or (sub)cutaneous metastases and distant metastases was determined by experienced dermatologists and radiologists. All patients underwent clinical examination and lymph node sonography every 3 months. CT scans of the trunk and magnetic resonance imaging of the brain were conducted every 3 months under adjuvant treatment and every 6 months after completed adjuvant treatment. Patients without tumor recurrence were censored 6 months after last blood sampling.

Controls

Plasma samples of n = 126 control patients with no evidence of malignant disease were analyzed. All control patients were treated at the University Hospital Bonn between 2014 and 2016 and matched to the palliative treatment cohort with respect to age and sex.

Plasma Preparation and SHOX2 CcfDNAm Analysis

Blood samples were collected in EDTA-containing collection tubes, and bisulfite-converted DNA was prepared as previously described (22–24). We assessed SHOX2 ccfDNAm performing a SHOX2 quantitative methylation-specific real-time polymerase chain reaction (20, 23). Actin β (ACTB) was used as a reference gene for the quantification of the total amount of ccfDNA in the plasma sample (20, 22–24). Our assay targets chromosome 3:158 103 550–158 103 661 (SHOX2) and chromosome 7:5 532 100–5 532 228 (ACTB; Genome Reference Consortium Human Build 38 patch release 13 [GRCh38.p13]).

We calculated relative SHOX2 ccfDNAm using the ΔΔ cycle threshold (ΔΔC_T) method as previously described and evaluated (20, 22, 23, 25, 26). Test results above the previously validated cutoff (0.25%) were considered ccfDNAm-positive (17, 22, 23). Biomarker levels below this value were defined as ccfDNAm-negative.

Statistical Analysis

Statistical analyses were performed using SPSS software version 28 (SPSS Inc.). Associations between nominal variables were tested using the Pearson χ²-test. Mean value comparisons of paired and unpaired not normally distributed variables were performed using the Wilcoxon test (2 groups) and Mann–Whitney U (2 groups) or Kruskal–Wallis (>2 groups) tests, respectively. Correlations were calculated using Spearman ρ. All variables were tested for normality. Survival analyses were performed using the Kaplan–Meier method and Cox proportional hazards analysis. Relative risk of death or progression was indicated as hazard ratio, including 95% CIs. P values refer to log-rank and Wald tests, respectively. P < 0.05 was considered statistically significant.

Results

SHOX2 CcfDNA Is Hypermethylated in Blood of Patients wIth Metastatic Melanoma

We previously reported SHOX2 hypermethylation in 93% of melanoma tissues at 90% specificity (17). In order to test the ability of SHOX2 ccfDNAm as a plasma biomarker, we compared SHOX2 ccfDNAm in patients with metastatic melanoma and in individuals without evidence of a malignant disease. We found significantly higher SHOX2 ccfDNAm in blood from melanoma patients compared to controls (P < 0.001, Fig. 1A). Applying a previously validated cutoff (0.25% SHOX2 ccfDNAm [17, 22, 23, 25, 26]), 25/42 (60% sensitivity) patients with metastatic melanoma showed SHOX2 ccfDNAm levels above the cutoff (SHOX2 ccfDNAm-positive); 123/126 (98% specificity) control individuals showed SHOX2 ccfDNAm levels below the cutoff (SHOX2 ccfDNAm-negative). Next, we tested the prognostic value of SHOX2 ccfDNAm at baseline prior treatment initiation (pretreatment). Melanoma patients (n = 42) receiving palliative anti-PD-1 mono- or anti-PD-1/CTLA-4 combination immunotherapy were categorized as pretreatment SHOX2 ccfDNAm-positive or -negative, respectively. We found no significant association between pretreatment SHOX2 ccfDNAm positivity and PFS (P = 0.31; Fig. 1B) or OS (P = 0.19; log-rank tests, Fig. 1C).

Fig. 1.

SHOX2 ccfDNAm and its association with survival in melanoma patients treated with palliative immunotherapy. SHOX2 ccfDNAm levels in blood plasma at baseline (pretreatment) of n = 42 patients with metastatic (stage IIIC–IV) melanoma treated with palliative anti-PD-1 immunotherapy compared to n = 126 individuals without evidence of malignant disease (A). Kaplan–Meier estimates of progression-free (B) and overall survival (C) in metastatic melanoma patients stratified according to pretreatment SHOX2 ccfDNA positivity (>0.25% cutoff). P values refer to Mann–Whitney U-test and log-rank test, respectively.

On-treatment SHOX2 CcfDNAm Negativity Correlates wIth Beneficial Survival

To test the association between early on-treatment SHOX2 ccfDNAm and survival, we stratified patients by positivity of SHOX2 ccfDNAm 4 weeks after treatment initiation. We found significantly prolonged PFS (P < 0.001; Fig. 2A) and OS (P = 0.002; log rank tests, Fig. 2B) associated with on-treatment SHOX2 ccfDNAm negativity. On-treatment SHOX2 ccfDNAm-positive patients had a median PFS of only 3.5 (95% CI, 0.0–7.0) months and a median OS of 26.4 (95% CI, 4.2–48.7) months, whereas the group of patients with negative SHOX2 ccfDNAm neither reached median PFS nor OS. Accordingly, patients with negative on-treatment SHOX2 ccfDNAm had a significant lower risk of death or progression when compared to patients with positive SHOX2 ccfDNAm [PFS: hazard ratio = 0.25 (95% CI, 0.10–0.60), P = 0.002; OS: hazard ratio = 0.12 (95% CI, 0.03–0.57), P = 0.007]. Six months after start of treatment, 14/19 (74%) patients with on-treatment SHOX2 ccfDNAm-positive plasma had a tumor progress versus 4/23 (17%) patients with negative on-treatment SHOX2 ccfDNAm. Five of 19 (26%) patients with positive on-treatment SHOX2 ccfDNAm died within 6 months, whereas no death in the on-treatment SHOX2 ccfDNAm-negative group was reported.

Fig. 2.

Progression-free survival and overall survival depending on early on-treatment SHOX2 ccfDNAm positivity. Kaplan–Meier estimates of progression-free (A) and overall survival (B) in n = 42 patients with metastatic (stage IIIC–IV) melanoma receiving palliative anti-PD-1 immunotherapy stratified by on-treatment SHOX2 ccfDNAm positivity 4 weeks after initiation of immunotherapy (log-rank test). SHOX2 ccfDNAm levels >0.25% were defined as SHOX2 ccfDNAm-positive.

Rapid On-Treatment SHOX2 CcfDNAm Clearance Predicts Response to Anti-PD-1 Immunotherapy and Survival

Next, we tested the association between rapid SHOX2 ccfDNAm clearance and therapy response. Comparing pre- to on-treatment SHOX2 ccfDNAm changes within each individual patient, objective response (complete response or partial response) was accompanied by a rapid SHOX2 ccfDNAm clearance within 4 weeks of treatment in 17/42 (40%) of patients (P = 0.005; Fig. 3A), whereas 13/42 (31%) of patients with PD had no significant change with undetected SHOX2 ccfDNAm during both pre- and on-treatment time points (P = 0.65, Mann–Whitney U-tests; Fig. 3A). Of note, we could also find significantly decreasing SHOX2 ccfDNAm levels in 12/42 (29%) of patients with SD (P = 0.036, Mann–Whitney U-test; Fig. 3A).

Fig. 3.

Pre- and on-treatment SHOX2 ccfDNAm positivity in association with response and lead time gain in diagnosis of disease progression. (A), SHOX2 ccfDNAm levels prior to (pretreatment) and 4 weeks after treatment initiation (on-treatment) in patients with SD, partial response and complete response, and PD (Wilcoxon test). Bars and whiskers indicate upper/lower quartiles and median values; (B), Lead time gain regarding the diagnosis of disease progression by SHOX2 ccfDNAm testing compared to current clinical practice in patients with PD (n = 13).

We categorized patients according to class switch regarding pre- to on-treatment SHOX2 ccfDNAm status (Fig. 4A). We built 3 groups of patients according to (A) SHOX2 ccfDNAm-positive pre- and on-treatment, (B) pre-treatment ccfDNAm-negative (irrespective of on-treatment ccfDNAm status), and (C) ccfDNAm-positive pre- and ccfDNAm–negative on-treatment plasma. We observed a significantly higher number of responses in patients showing rapid SHOX2 ccfDNAm clearance [group (C)] compared to the other groups (P = 0.030, Pearson χ²-test; Fig. 4A). Objective response was found in 7/8 (88%) patients from group (C) showing rapid SHOX2 ccfDNAm clearance within 4 weeks of ICB compared to 7/17 (41%) responses in group (B) and only 3/17 (18%) responses in patients from group (A). All 8 patients included into group (C) experienced at least disease control (SD/partial response/complete response), whereas 4/17 (24%) patients from group (B) and 9/17 (53%) patients with continuously positive SHOX2 ccfDNAm [group (A)] suffered PD.

Fig. 4.

Outcome in anti-PD-1-treated metastatic melanoma patients in association with pre- to on-treatment SHOX2 ccfDNAm changes. (A), Distribution of responses according to SHOX2 ccfDNAm pretreatment to 4 weeks on-treatment class switch (positive to positive, negative to positive/negative, and positive to negative; Pearson χ² test, n = 42). Created with BioRender.com. Kaplan–Meier estimates of progression-free (B) and overall survival (C) according to pretreatment to 4 weeks on-treatment SHOX2 ccfDNAm class switch (log-rank test, n = 42).

Moreover, we examined the association between SHOX2 ccfDNAm clearance and survival. As expected, patients with pretreatment SHOX2 ccfDNAm positivity to rapid on-treatment ccfDNAm negativity class switch [group (C)] showed the best, patients with pretreatment SHOX2 ccfDNAm negativity [group (B)] an intermediate, and persistent SHOX2 ccfDNAm positivity [group (A)] the poorest PFS (P < 0.001; Fig. 4B) and OS (P = 0.008; log-rank tests, Fig. 4C). While patients with persistently positive SHOX2 ccfDNAm [group (A)] only reached a median PFS of 2.8 (95% CI, 0.0–6.4) months and a median OS of 26.4 (95% CI, 7.5–45.4) months, the other groups did not reach median PFS or OS.

We further estimated a potential lead time gain regarding the accurate diagnosis of progression in patients with PD by means of SHOX2 ccfDNAm compared to imaging. Among the 10/13 (77%) on-treatment SHOX2 ccfDNAm-positive patients with PD, SHOX2 ccfDNAm positivity was found with a median lead time gain of 1.5 (IQR: 0.3–1.9) months compared to current clinical practice (Fig. 3B).

SHOX2 CcfDNAm Testing Aids in Recurrence Diagnostics Under Adjuvant Anti-PD-1 Immunotherapy

In order to test the utility of SHOX2 ccfDNAm testing during follow-up to detect recurrence in patients with adjuvant ICB, we analyzed plasma samples of n = 55 patients with locally advanced or metastatic melanoma receiving adjuvant anti-PD-1 monotherapy after surgical tumor resection. During follow-up, 11/55 (20%) patients developed lymph node or (sub)cutaneous metastases, 6/55 (11%) patients developed distant metastases, and 38/55 (69%) patients remained recurrence-free (Fig. 5). In 2/11 (18%) patients with lymph node or (sub)cutaneous metastases and in 5/6 (83%) patients with distant metastases, SHOX2 ccfDNAm turned positive prior to or simultaneously with recurrence diagnostics by standard of care follow-up examination (Fig. 5). Of note, the patient with distant metastasis and negative SHOX2 ccfDNAm developed a solitary brain metastasis, whereas the 5 patients with distant metastases and positive SHOX2 ccfDNAm developed metastases located in the lung, the liver, the bones, and at multiple sites, respectively. In all 38 recurrence-free patients, SHOX2 ccfDNAm remained negative during follow-up. Notably, SHOX2 ccfDNAm positivity was present at least 1 month prior to the clinical/radiographic detection of tumor recurrence in 2/11 (18%) patients with lymph node or (sub)cutaneous metastases and in 2/7 (29%) patients with distant metastases.

Fig. 5.

SHOX2 ccfDNAm positivity in relapsing and recurrence-free melanoma patients with adjuvant anti-PD-1 immunotherapy. Swimmer plot including SHOX2 ccfDNAm status during and after adjuvant anti-PD-1 therapy of n = 55 patients with locally advanced or metastatic melanoma. Patients were sorted by the localization of occurred metastases [lymph node or (sub)cutaneous metastases (n = 11) vs distant metastases (n = 6) and the absence of progression (n = 38), respectively]. SHOX2 ccfDNAm levels >0.25% were defined as SHOX2 ccfDNAm-positive.

Discussion

Besides its success in the first-line treatment of BRAF-mutated melanoma (1), targeted therapy has been successfully used as second-line treatment after tumor progression under anti-PD-1 immunotherapy (27), and efforts are increasingly focusing on alternative treatment regimens in case of resistance to immunotherapy. Adoptive T-cell therapy has shown efficacy in anti-PD-1 nonresponders (28). Ongoing clinical studies aim at several target structures of the tumor microenvironment and assess combination therapies ([29]; ClinicalTrials.gov Identifiers: NCT04303169, NCT03708328, NCT02652455, NCT03978611). Recently, the combination of pembrolizumab with the receptor tyrosine kinase inhibitor lenvatinib has shown promising results in overcoming resistance to anti-PD-1 immunotherapy alone or plus anti-CTLA-4 therapy (30). Early and reliable response monitoring is crucial for treatment planning. However, traditional cross-sectional imaging has several limitations including radiation exposure, availability, cost, and difficulties in interpretation, especially during the first months after treatment initiation (31). Early detection of an ineffective treatment can protect patients from unnecessary exposure to toxic agents. Dynamic noninvasive liquid biopsy biomarkers harbor the potential to give insights about response to treatment and survival.

In this study, we assessed the prognostic and predictive value of early SHOX2 ccfDNAm changes in blood plasma of patients with locally advanced or metastatic melanoma undergoing anti-PD-1 immunotherapy. We showed that on-treatment SHOX2 ccfDNAm testing allows for the discrimination of patients with therapy response and those with a high risk for early tumor progression. Accordingly, patients with positive on-treatment SHOX2 ccfDNAm showed shorter PFS and adverse OS. In a recent meta-analysis of 18 trials of ICB in advanced solid tumors, ctDNA was suggested to be an early response biomarker that may allow for de-escalation of cross-sectional imaging (32). Eight of 18 trials were conducted in patients with locally advanced or metastatic melanoma. Overall, the study revealed a 3 times lower hazard of progression and a 6 times lower risk of death in patients with >50% ctDNA reduction during treatment. In line with these findings, we found a 4 times lower hazard of progression and an 8 times lower risk of death in patients with negative on-treatment SHOX2 ccfDNAm. These hazards were even notably lower in case of positive pretreatment and negative on-treatment SHOX2 ccfDNAm. Of note, among the 18 trials analyzed, the second measurement of ctDNA levels was performed 6 to 16 weeks after treatment initiation. Whereas, in our study, on-treatment SHOX2 ccfDNAm was assessed only 4 weeks after starting treatment.

Little is known about ctDNA dynamics in patients with SD treated with ICB (32). Subgroup analyses of 2 studies with patients suffering from non-small cell lung cancer revealed a trend toward prolonged PFS in patients with SD upon ICB when on-treatment ctDNA decreased (33, 34). In accordance with these results, we found a significant reduction in SHOX2 ccfDNAm in melanoma patients with SD. Despite the small sample size in this subgroup, patients with negative on-treatment SHOX2 ccfDNAm showed significantly improved PFS compared to SHOX2 ccfDNAm-positive patients. These data support the applicability of SHOX2 ccfDNAm as an early response biomarker to anti-PD-1 immunotherapy, which may be used to guide treatment planning in metastatic melanoma.

Patients with metastatic melanoma frequently undergo CT scans, resulting in high exposure to radiation. Several studies have reported an increased incidence of cancer after CT scan exposure depending on radiation dose and the number of CT scan sites (35). SHOX2 ccfDNAm testing accurately identified patients who benefitted from ICB. This was particularly true for patients with positive pretreatment and negative on-treatment SHOX2 ccfDNAm who achieved disease control in any case. Therefore, SHOX2 ccfDNAm testing might allow for de-escalation of cross-sectional imaging, thereby reducing radiation exposure.

In the adjuvant setting, SHOX2 ccfDNAm testing showed a rather low clinical performance but was able to detect the relatively rare formation of distant metastases in relapsing patients undergoing adjuvant ICB. In most patients with newly detected lymph node or (sub)cutaneous metastases and in all recurrence-free patients, SHOX2 ccfDNAm remained negative. In view of the fact that the combination of physical examination and ultrasonography has the highest sensitivity and specificity in detecting lymph node and cutaneous metastases in melanoma (36), SHOX2 ccfDNAm testing might allow for an extension of staging intervals under adjuvant ICB.

Lactate dehydrogenase is an established prognostic factor in melanoma and part of the 8th revised staging guidelines of the American Joint Committee on Cancer (37). Currently, it is the only regularly used blood-based biomarker in melanoma (37), though it lacks sensitivity and specificity in predicting treatment response (37). Melanoma driver mutations (BRAF^V600 and NRAS^Q61) in ctDNA have emerged as a promising ctDNA biomarker test in BRAF- and NRAS-mutated melanomas (9, 10). However, BRAF and NRAS mutation testing is not applicable to BRAF/NRAS wild-type melanomas that constitute approximately a third of all melanomas. SHOX2 ccfDNAm testing could close the diagnostic gap of mutation-based ctDNA testing in wild-type melanoma, providing a mutation-independent way for ctDNA detection. A significant number of patients included in our study suffered from BRAF- or NRAS-mutated melanomas. Self-evidently, BRAF or NRAS ctDNA mutation testing is expected to show a high clinical performance in these patients. The lack of such mutation testing represents a major limitation of our study. Further studies comparing mutation and methylation testing are needed to allow for a direct comparison of both biomarker approaches and to evaluate potential synergistic effects of a combined analysis. The combination of SHOX2 ccfDNAm with additional methylation biomarkers (e.g., RASSF1) and mutation testing (e.g., BRAF and NRAS) might further increase the clinical sensitivity and specificity of a ctDNA biomarker test. In addition, a harmonized sequence of blood testing, physical examination and ultrasonography, and imaging, potentially including a retesting of positive samples in the adjuvant setting, could help to increase sensitivity and specificity of recurrence and response monitoring.

Recent studies suggested a predictive value for multiple other biomarkers in melanoma, including S100 calcium-binding protein B, composition of the gut microbiome, and several tissue-based biomarkers such as tumor mutational burden, T-cell infiltration pattern, interferon-γ signatures, CTLA4 promoter methylation, and PD-1 ligand 1 expression (38, 39). However, these biomarkers demonstrated only limited predictive performance, and large-scale validation studies demonstrating the clinical use are missing (38, 39). Forschner and colleagues tested the ability of tumor mutational burden to predict treatment response and survival alone or in combination with ctDNA levels in a cohort of n = 35 melanoma patients (8). This study revealed an additional prognostic and predictive value when tumor mutational burden and ctDNA levels were analyzed simultaneously. The combination of a tissue-based and an early blood-based biomarker could be beneficial: the tissue-based biomarker might allow for a pretreatment selection of suitable patients, while the blood-based biomarker could identify those patients shortly after therapy start who actually benefit from treatment. SHOX2 ccfDNAm might provide additive or synergistic information when assessed together with other promising biomarkers in melanoma.

Our study does not allow one to conclude that SHOX2 methylation has biological relevance in the context of melanoma immunobiology. Functional analyses are required to elucidate if SHOX2 methylation in melanoma tissue mechanistically associates with response to ICB or if SHOX2 methylation in the heterogeneous melanoma bulk tissue only represents a surrogate measure for tumor cell content.

In previous studies, we have already shown repeatability, precision, and accuracy of ccfDNA methylation testing (21, 40); however, larger reproducibility studies are warranted in order to determine precision between laboratories prior to clinical implementation. The quantitative methylation-specific polymerase chain reaction assay for SHOX2 ccfDNAm gives the quantitative SHOX2 ccfDNAm levels that we reported in this study in order to avoid the loss of information due to the result dichotomization based on a cut-off. However, as the analytical performance of such a quantitative test needs further rigorous validation, qualitative test results should be considered with regard to a timely implementation of a diagnostic test into clinical routine.

In conclusion, a decrease in SHOX2 ccfDNAm is associated with response to ICB and beneficial survival in patients with metastatic melanoma receiving anti-PD-1 immunotherapy. SHOX2 ccfDNAm is an early response liquid biopsy biomarker that harbors the potential to guide treatment planning in metastatic melanoma and supports treatment de-escalation strategies. Especially in patients with adjuvant immunotherapy, SHOX2 ccfDNAm testing might allow for de-escalation of cross-sectional imaging. Further studies are needed to assess the clinical applicability of SHOX2 ccfDNAm and whether it allows for aimproved clinical decision-making.

Ethics approval and consent to participate

The study was carried out in accordance with the Declaration of Helsinki of 1975, and all patients gave written informed consent. The study has been approved by the institutional review board of the University Hospital Bonn (vote number 301/19).

Supplemental Material

Supplemental material is available at Clinical Chemistry online.

Nonstandard Abbreviations

CTLA-4, cytotoxic T lymphocyte-associated protein 4; PD-1, programmed cell death 1; ICB, immune checkpoint blockade; ccfDNA, circulating cell-free DNA; ctDNA, circulating tumor DNA; OS, overall survival; ccfDNAm, circulating cell-free DNA methylation; CI, confidence interval; IQR, interquartile range; PFS, progression-free survival; CT, computed tomography; SD, stable disease; PD, progressive disease.

Human Genes

ACTB, actin β; BRAF, proto-oncogene B-Raf; CTLA4, cytotoxic T lymphocyte-associated protein 4; NRAS, neuroblastoma RAS viral oncogene homolog; RASSF1, Ras association domain family member 1; SHOX2, short-stature homeobox 2.

Author Contributions

The corresponding author takes full responsibility that all authors on this publication have met the following required criteria of eligibility for authorship: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved. Nobody who qualifies for authorship has been omitted from the list.

Simon Fietz (Conceptualization-Lead, Data curation-Lead, Formal analysis-Lead, Funding acquisition-Supporting, Investigation-Equal, Project administration-Equal, Resources-Supporting, Software-Equal, Validation-Equal, Visualization-Lead, Writing—original draft-Equal, Writing—review & editing-Equal), Eric Diekmann (Data curation-Supporting, Formal analysis-Supporting, Investigation-Equal, Writing—original draft-Equal), Luka de Vos (Investigation-Supporting, Writing—review & editing-Supporting), Romina Zarbl (Investigation-Supporting, Resources-Lead, Writing—review & editing-Equal), Alina Hunecke (Investigation-Supporting, Resources-Supporting, Writing—review & editing-Equal), Ann-Kathrin Glosch (Investigation-Supporting, Resources-Supporting, Writing—review & editing-Equal), Moritz Färber (Investigation-Supporting, Writing—review & editing-Equal), Judith Sirokay (Resources-Equal, Writing—review & editing-Supporting), Friederike Hoffmann (Resources-Supporting, Writing—review & editing-Supporting), Anne Fröhlich (Resources-Supporting, Writing—review & editing-Supporting), Alina Franzen (Resources-Supporting, Writing—review & editing-Supporting), Sebastian Strieth (Funding acquisition-Equal, Writing—review & editing-Supporting), Jennifer Landsberg (Conceptualization-Supporting, Funding acquisition-Supporting, Methodology-Supporting, Project administration-Supporting, Resources-Lead, Supervision-Supporting, Writing—original draft-Supporting, Writing—review & editing-Supporting), and Dimo Dietrich (Conceptualization-Lead, Data curation-Supporting, Formal analysis-Supporting, Funding acquisition-Lead, Methodology-Lead, Project administration-Equal, Resources-Equal, Software-Equal, Supervision-Lead, Validation-Equal, Writing—original draft-Supporting, Writing—review & editing-Equal)

Authors’ Disclosures or Potential Conflicts of Interest

Upon manuscript submission, all authors completed the author disclosure form.

Research Funding

This study was supported by the BioBank Bonn of the Bonn University Medical Faculty and the University Hospital Bonn. S. Fietz received funding from the University Hospital Bonn BONFOR program (O-105.0069). L. de Vos received funding from the University Hospital Bonn GEROK program (O-105.0072). F. Hoffmann was partly funded by the Deutsche Krebshilfe through a Mildred Scheel Foundation Grant (grant number 70113307).

Disclosures

D. Dietrich owns patents and patent applications on biomarker technologies (e.g., United States Patent Application 20210108255) that are licensed to Qiagen GmbH (Hilden, Germany). J. Sirokay is a consultant/advisory board member and received speaker’s honoraria from Novartis, BMS, MSD, and Roche. J. Landsberg is a consultant/advisory board member and received speaker’s honoraria from Novartis, BMS, MSD, and Roche. L. de Vos received travel support from Pierre Fabre.

Role of Sponsor

The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, preparation of manuscript, or final approval of manuscript.

Author notes

Simon Fietz and Eric Diekmann are joint first authors on this work.