Release of Cell-Free Tumor DNA in the Plasma of Uveal Melanoma Patients Under Radiotherapy

Abstract

Purpose: Uveal melanoma (UM) is a tumor of the eye that metastasizes in approximately half of cases. Prognostic testing requires accessibility to tumor tissue, which is usually not available with eye-preserving therapies. Noninvasive approaches to prognostic testing that provide valuable information for patient care are therefore needed. The aim of this study was to evaluate the use of circulating cell-free plasma DNA analysis in UM patients undergoing brachytherapy.

Methods: The study recruited 26 uveal melanoma patients referred to the department between February and October 2020. Blood samples were collected at various time points before, during, and after treatment, and deep amplicon sequencing was used to identify oncogenic variant alleles of the GNAQ and GNA11 genes, which serve as indicators for the presence of circulating tumor DNA (ctDNA).

Results: The results showed that all patients were ctDNA negative before brachytherapy. In 31% of patients, ctDNA was detected during therapy. The variant allele fraction of GNAQ or GNA11 alleles in ctDNA positive samples ranged from 0.24% to 2% and correlates with the largest basal diameter and thickness of the tumor.

Conclusions: The findings suggest that brachytherapy increases the presence of tumor DNA in the plasma of UM patients. Thus ctDNA analysis may offer a noninvasive approach for prognostic testing. However, efforts are still required to lower the limit of detection for tumor-specific genetic alterations.

Uveal melanoma (UM) is the most common primary intraocular malignancy in adults.¹ Analyses of genetic alterations, epigenetic markers, and expression profiles revealed two main classes of UM that are associated with patients’ prognosis.^2–5 Loss of chromosome 3 (monosomy 3 [M3]) in the tumor is a well-established and commonly used marker for prognostic testing of UM patients.⁶ So far, therapeutic decisions do not depend on prognosis. However, most patients want to know their individual risk, and even if the prognosis is poor, the long-term psychological impact on these patients is positive.^7,8 In the case that adjuvant therapy options become available, prognostic testing may also have clinical significance for stratifying patients according to their metastatic risk. Currently, prognostic testing is only possible if tumor material is available, which is usually the case in patients treated by enucleation. Most patients are initially treated with eye-preserving methods, so tumor material must be obtained by biopsy, which can cause serious complications such as hemorrhages, endophthalmitis, retinal detachment, tumor seeding, or extraocular dissemination.⁹ In these patients analysis of circulating cell-free DNA (cfDNA) could provide an attractive, noninvasive alternative approach. The presence of circulating tumor DNA (ctDNA) in the blood of UM patients with metastatic disease but also several months before metastatic progression becomes clinically diagnosed has been reported previously.^10–14 However, ctDNA is only rarely detected in blood at time of diagnosis of the primary tumor. In two studies an elevated level of ctDNA was found in the blood of patients during or shortly after radiotherapy.^15,16 This suggests that DNA from dying tumor cells is released into the blood and prompted us to analyze the ctDNA content at different time points during and after radiotherapy. 

Because almost all UM, irrespective of UM class, carry activating somatic mutations in exon 4 or 5 in one of the paralog oncogenes GNAQ or GNA11, these variants can serve as a marker for the presence and amount of ctDNA circulating in the blood.^15,17–19 The aim of the study presented here was to determine whether tumor DNA released by radiation therapy–induced cell death is detectable in the blood of UM patients and to evaluate the dynamics of ctDNA levels during brachytherapy in the blood of patients. 

Methods

Patients’ Cohort and Blood Collection

Between February and October 2020 all patients referred to our Department of Ophthalmology with a diagnosis of primary UM of the choroid or ciliary body and an indication for brachytherapy were asked to participate in the study. Written informed consent to participate in the study was given by 26 patients. The protocols of the Declaration of Helsinki have been followed. This study has been approved by the local ethics committee. Accounting clinical criteria like orange pigment, exudation, tumor size (largest basal diameter/height), and reduction of visual acuity may be applied to estimate the risk for malignant transformation of a choroidal nevus (TFSOM-DIM Mnemonic).²⁰ 

Patients were treated by brachytherapy with either mono-nuclide plaque (ruthenium¹⁰⁶) or bi-nuclide plaque (ruthenium¹⁰⁶ + iodine¹²⁵). A bi-nuclide plaque is used in larger tumors >6 mm in height that cannot be irradiated with ruthenium¹⁰⁶ because of dosimetric limitations. A combination of iodine and ruthenium helps to maintain the tumor dosage while reducing the dosage outside the target volume compared with iodine alone.²¹ Blood samples were taken by venipuncture from the patients immediately before, after, and several times during the course of treatment. The time interval depended on the duration of the treatment: daily sampling for short treatment periods and every two or three days for longer treatment periods. Venous blood was collected using two 7.5 mL EDTA tubes either from the forearm or the back of the hand. The vacutainers were drawn slowly to exclude possible hemolysis and contamination of the sample. The blood was processed within 60 minutes upon collection and kept frozen at −20° until further cfDNA extraction. 

Isolation of cfDNA and Library Preparation

Cell-free DNA was isolated from 3 to 5 mL plasma (depending on obtained plasma volume) using the QiAmp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany). Library preparation for deep amplicon sequencing has been performed as described.¹³ DNA sequencing was performed on Illumina MiSeq Platform in “paired-end sequencing” mode (2 × 150 bp). Each sample was sequenced with at least 20,000-fold coverage. The presence of cell-free tumor DNA was determined in all cfDNA samples by deep amplicon sequencing of the exon 4 and exon 5 regions of GNAQ and GNA11, specifically covering the positions R183 and Q209. Samples with a variant allele fraction of the mutant allele (VAF) >0.1% detected in one of the target regions were considered to be ctDNA positive.¹⁵ 

Statistics

Data analysis was performed with IBM SPSS Statistics 27.0. Standard sample characteristics (mean, median, range, quantiles) were used, including standard deviation when normally distributed. We applied the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables. 

Results

We collected blood samples from 26 patients the day before therapy, after therapy, and at different time points during radiation therapy, depending on the duration of therapy every day/every second or every third day (for details see Fig. 1). CfDNA was isolated from a total of 128 blood samples. 

Figure 1.

 

Heatmap illustrating ctDNA negative and ctDNA positive blood samples in individual patients during radiotherapy. Column “0” shows ctDNA status at the day before treatment. ctDNA negative samples are shown as red boxes and ctDNA positive samples as green boxes. Gray box: no blood sample was taken. Patients that dropped out before the end of study are marked with an asterisk in the first column. End of brachytherapy is indicated by a vertical solid black line.

No ctDNA signal was detected in any of the 26 blood samples collected before therapy (day 0 in Fig. 1). In eight of the 26 patients (31%), we found a ctDNA signal in at least one of the samples taken during or directly after treatment (Fig. 2). In the positive samples, the VAF varied from 0.25% to 2.0%. Two (no. 17 and no. 23) of the eight patients dropped out of the study early, and therefore complete data are not available from these. Two other patients (no. 14 and no. 16) showed a positive ctDNA signal at all time points during the course of therapy and after therapy. Interestingly, these tumors were the second and third largest tumors in our cohort. In the remaining four patients only one blood sample was found to be ctDNA positive (Fig. 1). Clinical characteristics of all recruited patients grouped according to their ctDNA status are addressed in Table 1. 

Figure 2.

 

CtDNA levels of individual ctDNA-positive patients during radiotherapy. Proportion of VAF is given with the threshold value of 0.1 % shown as a horizontal black line. Days at which a blood sample was taken are indicated by a yellow dot.

Table 1.

 

Clinical Characteristic of Study Cohort

 

We compared the clinical features of the patients with the presence of ctDNA in the blood (Table 1). A statistically significant correlation was observed for tumor height and largest basal diameter with the mean of 7.91 mm and 14 mm in ctDNA-positive compared to 4.74 mm and 1.4 mm in ctDNA-negative patients, respectively (Mann-Whitney U test P value < 0.05, Fig. 3). The apex dose was significantly less with 105 Gy in the ctDNA-positive compared to the ctDNA-negative patients (Mann-Whitney U test P value < 0.05). This implies that larger tumors are more likely to shed detectable amounts of DNA into the blood under radiation therapy than smaller tumors. 

Figure 3.

 

Distribution of tumor thickness [mm] (a) and largest basal diameter (LBD) [mm] (b) in ctDNA-positive and ctDNA-negative samples.

Three of eight ctDNA-positive patients exhibited mutations in the GNA11 gene, and five had mutations in the GNAQ gene. Table 2 shows the oncogenic GNAQ/GNA11 mutations in the ctDNA-positive patients. 

Table 2.

 

Genetic Variants Identified

 

Discussion

Analysis of ctDNA is increasingly being used in tumor diagnostics, because it provides access to genomic tumor DNA without the need to obtain tumor tissue.²² Moreover, it has been shown that ctDNA contains the genetic information of the entire tumor genome,²³ making it a promising source for the analysis of genetic markers such as gene mutation pattern or chromosome 3 status. In this study, we investigated whether cfDNA as potential source for noninvasive prognostic testing in patients with UM can be isolated from blood collected under radiotherapy. 

Controversial reports exist concerning the presence of plasma ctDNA levels in nonmetastasized UM patients. Bustamante et al.²⁴ reported detectable ctDNA levels in 100% of patients with primary uveal melanoma. However they did not analyze probes prior to the treatment of the primary tumor. Beasley et al found ctDNA in only 26% of non-treated patients.¹⁰ In a previous study, Le Guin et al.¹⁵ found detectable levels of ctDNA in the plasma of patients with metastatic UM. In patients with localized tumors, however, ctDNA was only detectable in rare cases at the time of diagnosis, regardless of tumor size. Francis et al.¹⁶ reported detectable levels of ctDNA in almost 29% of the patients in the perioperative period (enucleation/brachytherapy). 

Results of our present study are consistent with the previous study by Le Guin et al.¹⁵ in that, before treatment, blood ctDNA levels were below level of detection in blood from all 26 patients with localized disease (Fig. 1). Blood samples obtained under irradiation contained detectable levels of ctDNA in eight of 26 (31%) study participants. In ctDNA-positive samples, the variant allele fraction of GNAQ or GNA11 alleles ranged from 0.25% to 2.0%. Both the VAF and the number of ctDNA-positive serial samples during therapy was highly variable between patients. Only two of the patients (no. 14 and no. 16) were ctDNA positive at all time points tested during therapy including the post treatment samples. In contrast, in six patients (no. 2, no. 5, no. 17, no. 20, no. 21, and no. 23) ctDNA was detected in only one blood sample taken during or directly after brachytherapy. In two of these six patients we could not obtain the full number of planned blood samples as both patients interrupted or discontinued their participation in the study (Fig. 1). 

Because the ctDNA-positive samples are evenly distributed over the treatment period, it was not possible for us to determine a time point during therapy when ctDNA levels peaked. Shortening blood collection intervals and introducing additional sampling after the end of radiotherapy may allow better resolution of the dynamics of ctDNA release in UM patients. Our findings are in agreement with those of the study of Lo et al.,²⁵ where the kinetics of plasma Epstein-Barr virus DNA during radiation therapy in patients with nasopharyngeal carcinoma was explored. Blood collection was carried out before intervention, every day during the first week of radiotherapy, and on the first day of each weekly course of radiotherapy. The group has shown the initial rise of EBV DNA concentration during the first week of radiotherapy with a subsequent decrease in all patients.²⁵ The same results were demonstrated by Walls et al.,²⁶ who investigated the kinetics of ctDNA during radical radiotherapy for non-small cell lung cancer. Patients donated blood prior to radiotherapy, 72 hours after radiotherapy and seven days after radiotherapy. The group has shown a transient increase in ctDNA concentration at 72 hours followed by its decrease. Francis et al.¹⁶ recently analyzed ctDNA in UM patients undergoing primary tumor treatment (brachytherapy/enucleation) and found an increase of ctDNA on day 2 and 3 of brachytherapy. They reported an apex dose of 85 Gy, suggestive for usage of iodine plaques. In our study, ruthenium plaques and bi-nuclide plaques (iodine and ruthenium) have been used. Since ruthenium is emitting high energetic (2.4-3.5 MeV) beta irradiation instead of lower energy (0.028-0.035 MeV) gamma irradiation emitted by iodine the differences in detection rates might be explained. 

Presence of detectable levels of ctDNA during or shortly after therapy indicates that DNA from primary UM is released into the patient's blood during radiation therapy. However, it remains unclear why ctDNA is detectable only in a subset of patients and why in many of these patients a ctDNA signal is obtained only at a single point in time during therapy. We may have missed the presence of ctDNA in some patients, as the GNAQ/GNA11 regions targeted by our assay are mutated in only 92% of cases. The kinetics of cfDNA must be also taken into consideration (i.e., the rate of shedding of ctDNA from the tumor and clearance of ctDNA from blood). It is possible that cfDNA turnover rates are higher in the ctDNA negative patients with high levels of both shedding and clearance.^27,28 It has been estimated that the half-life of ctDNA is up to two hours and is dependent on factors including cell turnover, tumor size, excretion in bodily fluids and degradation rate by circulating nucleases.²⁹ Khier and Lohan³⁰ suggested that blood-brain barrier may affect the distribution and the abundance of ctDNA throughout the organism, which may partly contribute to the negative samples in our study. However, because we observed a significant correlation between tumor size and the probability of detection of ctDNA we suggest that, as tumor size increases, more cells are likely to release their ctDNA into the blood as a result of radiation-induced cell death (P value for tumor height and largest basal diameter is < 0.05). In the ctDNA-negative patients, the concentration of ctDNA in the blood may not have reached the detection limit of our method. The sensitivity of the methodology is strictly limited by the amount of cfDNA available for measurement. Here we used 6 ng of cfDNA for each analysis which corresponds to 2000 haploid human genomes. However, even the presence of more than one complete tumor genome in cfDNA does no guarantee that a single mutant allele can be detected. Therefore the use of a larger amount of cfDNA would improve the sensitivity of the method and thus increase the ctDNA detection rate in patients with smaller tumors. This, however, would require the collection of more blood from patients. Pooling plasma samples from multiple sampling time points might help to solve this problem. For this reason, Streck Cell-Free DNA BCT blood collection tubes with stabilizing properties may be advantageous. It was shown that Streck tubes provide better stabilization of ctDNA at 48 hours after blood draw as opposed to conventional EDTA tubes.³¹ 

CtDNA is a potential source for genetic prognostic testing in patients with uveal melanoma. In principle, two different genetic markers are available for this purpose: loss of chromosome 3 or the mutation pattern of BAP1, SF3B1 and EIF1AX. 

Each of the two markers can be used to estimate patient's prognosis, although with different sensitivity. Monosomy 3 as marker for poor prognosis has been validated in numerous retrospective studies and is considered as the most reliable genetic marker.^32,33 However, evaluation of the mutation pattern, with BAP1 mutations predicting poor prognosis and SF3B1 and EIF1AX mutations predicting moderate or good prognosis, respectively, might fail to predict the correct prognosis in a small proportion of tumors because not all mutations are detectable by sequence analysis.³⁴ In recent years, NGS-based approaches for cfDNA analysis have been developed.³⁵ These techniques allow us to determine chromosomal copy number changes or single nucleotide analysis of ctDNA, even at low VAF > 1%.^36,37 Future studies are needed to find out which techniques for detecting these genetic markers are most suitable for us on ctDNA. It is a limitation of the study that the ctDNA cannot be matched to the mutations of the associated primary tumor, and no survival data are available. The lack of tumor material also prevents classification of the tumor according to metastatic risk. 

In summary, our study demonstrates the presence of ctDNA in patients with localized UM under radiation therapy. Possible clinical applications are the noninvasive prognostic testing for stratifying tumors into high and low risk. The method needs to be further developed to reliably determine prognostic markers on ctDNA.^20,35 

Acknowledgments

The authors thank Martina Fleuringer and Lars Masshöfer for their technical assistance. 

Disclosure: V. Kim, None; M. Guberina, None; N.E. Bechrakis, None; D.R. Lohmann, None; M. Zeschnigk, None; C.H.D. Le Guin, None