Biomarkers are of great interest in applied medical research, specifically in oncology. Here, applications of biomarker analysis include the detection of primary disease processes as well as metastasis and tumor recurrence, prediction of therapy response, and prognosis of disease outcome. Most often, the analytical target of biomarker analysis in oncology is DNA. From the technical point of view, an essential advantage of this biomolecule is the availability of analytical methods for the detection of characteristics of DNA, such as sequence variation, methylation, and abundance, even if only a few molecules of the DNA of interest are present.

From the biological point of view, it is advantageous that these DNA characteristics are often tightly linked to the disease process or its root cause. Moreover, as DNA is present in many body liquids, this biomolecule may be available for analysis by taking samples of these fluids (liquid biopsies). Establishing a DNA-based biomarker analysis for some intended use is a multistep process that starts with identifying the specific DNA-characteristic that can serve as a biomarker. It ends with validating the complete examination procedure, i.e., confirming that it is apt for the intended use. The work presented here aimed to develop biomarker analyses for two applications, namely i) a diagnostic biomarker test for detecting second neoplasms in patients with heritable retinoblastoma and ii) a prognostic test for uveal melanoma.

Heritable retinoblastoma (Rb) is an autosomal-dominant tumor predisposition syndrome caused by pathogenic variants of the RB1 gene. Rb-survivors have a high risk to develop a second primary malignancy (SPM). Development of retinoblastoma is often initiated by genetic mechanisms that result in loss of heterozygosity (LOH). It is to be expected that LOH is frequent in SPMs as well. We examined if LOH can be used as a biomarker for non-invasive early detection of SPMs in Rb-survivors.

We analyzed DNA derived from extracellular vesicles (EV-DNA) and cell-free DNA (cfDNA) extracted from blood plasma. DNA released by tumor cells with LOH at the RB1 locus is expected to skew the ratio of RB1 alleles in EV-DNA and cfDNA. We determined the allelic ratio of multiple Single Nucleotide Polymorphisms (SNPs) by Next-Generation Sequencing of cf-DNA. The assay was established for six SNPs located in the RB1 gene on the same linkage group.

We evaluated the suitability of EV-DNA and cfDNA for our liquid biopsy examination, with the criterion for suitability being the number of effective genome equivalents (GEs) needed for the detection of at least 5 % of ctDNA. The number of GEs in EV-DNA was insufficient, but cfDNA was suitable to detect around 10 % ctDNA, which may still be clinically relevant as there are no other biomarker screening tests available for SPMs in Rb-survivors. For test validation, we generated reference samples with tumor DNA admixtures. Tumor fractions of as little as 1% were detectable.

We applied our method to cfDNA from an Rb-cohort to review if the analytical procedure is suitable for the intended use. Contrary to our expectation, the assay was applicable to only one-third of patients because the proportion of patients being constitutionally heterozygous for the selected SNPs was lower in the cohorts from the cooperating countries. Due to the number of GEs measured in cfDNA, we expected a limit of detection (LoD) of 10 % ctDNA . The weighted arithmetic mean variant allele fraction (VAF) of two Rb-survivors and three SPM patients was above this threshold. However, these results need to be considered false positives as none of these patients had an SPM at the time of blood draw or in an appropriate period of time around the blood collection. Four patient samples collected at the time of diagnosis of SPM showed a VAF below the LoD and might be false negative.

Conclusively, we established a highly sensitive assay that allows the detection of 1 % tumor DNA in genomic DNA, but it was not applicable to the cfDNA samples collected within the European NIRBTEST study because of low DNA levels due to low sample volumes. However, this could not be adjusted during the project duration due to international standardization of sample collection and local ethical restrictions. As a next step, the suitability of the assay for the intended use should be evaluated using cfDNA reference material or patient blood samples of higher volumes.

Uveal Melanoma (UM) is a rare cancer but still the most common primary intraocular tumor in adults. Most UMs fall in one of two classes that show distinct genetic features and are highly correlated with prognosis. Monosomy 3, a distinctive feature of UM class two, is strongly associated with a poor prognosis. Currently, testing this biomarker in smaller tumors requires a tissue biopsy. The sampling procedure of tumor tissue is invasive and may cause side effects like bleeding, retinal detachment, or tumor cell seeding. Therefore, less invasive methods to obtain tumor-derived DNA (ctDNA) are desirable. The source of ctDNA that can be obtained least invasively is blood, but it has been shown that at diagnosis of UM the proportion of ctDNA in cfDNA from blood is mostly too low for reliable biomarker analysis (Le Guin et al. 2021).

We performed a feasibility study to test if cfDNA from aqueous humor or vitreous body aspirate contains sufficient amounts of ctDNA for genetic analyses. cfDNA was isolated from both specimens, which were obtained from eyes after enucleation. The amount of tumor-derived DNA was determined by deep amplicon sequencing targeting oncogenic variants in GNAQ and GNA11, which are both highly specific for UM. For analysis of sequencing data, we developed a Snakemake pipeline that includes bwa-mem (alignment) and freebayes (variant calling) to determine the fraction of alleles specific for tumor variants. In seven of nine vitreous body samples and five of 14 samples from aqueous humor GNAQ or GNA11 variants were detected with VAFs ranging from 3.7 % to 37.5 %.

To determine chromosome 3 dosage alterations in cfDNA, we developed an assay which is based on sequence enrichment of around 100 SNPs distributed over the whole chromosome 3. A balanced allelic ratio of informative SNPs is expected in cfDNA of healthy individuals or UM tumors with disomy 3, while an unbalanced allele ratio was expected for monosomy 3 tumors. We evaluated if this characteristic can serve as an indicator for chromosome 3 loss and, hence, as a prognostic biomarker for UM patients.

Testing the assay on cfDNA from plasma of healthy donors and UM patients, we found the VAF to be highly scattered around the expected VAF of $0.5$. The LoD of the assay was highly above the targeted LoD of 5 % ctDNA and therefore needs to be optimized before the assay can be applied to cfDNA from AH and VB.

Taken together, cfDNA derived from AH or VB is a very promising analytical target for diagnostic and prognostic biomarker analysis as it contains high levels of ctDNA at the time of diagnosis. Future studies on cfDNA from AH and VB from small tumors collected from non-enucleated eyes will show if these liquid biopsies can be used to detect the chromosome 3 status as a prognostic biomarker.