In the current study, we show that fractionation of DNA is highly essential for reliable detection of LOH in serum samples of ovarian cancer patients. Applying this new technique, we found more LOH in the LMWF than in the HMWF indicating that tumor-associated DNA is rather short-stranded than long-stranded. This observation is also supported by the fact that after chemotherapy, only the LMWF-DNA levels decreased, whereas the HMWF-DNA levels were similar prior to surgery and after chemotherapy. Moreover, our key findings showed that LOH at D6S1581 before surgery is predictive for reduced OS (combined fraction analysis) and LOH at D10S1765 after chemotherapy was significantly associated with tumor cell spread to the BM, respectively.
It has been reported that tumor-associated DNA, released into circulation, is rather of fragmented and low-molecular-weight character
. In this context, it was proposed that tumor cells, which proliferate at abnormal rates, frequently undergo apoptosis
[21–23] and cumulatively deliver small and fragmented DNA into the circulation. Therefore, it seems reasonable that the observed decline of cirDNA in our patients, selectively observed in the LMWF, can be ascribed to a systematic clearance of tumor cells upon chemotherapy, which leads to a global reduction of low molecular weight DNA. However, we did not see any significant change of HMWF-DNA prior to surgery and after chemotherapy. Long-stranded HMWF-DNA could contain an excess of normal DNA in the circulation and derives from normal blood lymphocytes
[24, 25], or necrotic stromal and inflammatory cells within the tumor microenvironment
. Accordingly, the observed steady-state HMWF-DNA content before surgery and after chemotherapy might indicate that chemotherapy does not sustainably affect HMWF-DNA in blood of ovarian cancer patients.
The majority of allelic losses were detected in the LMWF and also significantly correlated with common clinicopathological parameters of the patients whereas in the HMWF, LOH constituted a comparatively rare event. So far, previous studies on several tumor entities exclusively analyzed non-fractionated cirDNA and thus, there is a broad variety of LOH detection rates with partly contradictory results
[12, 24]. In cooperation with the laboratory of Schwarzenbach et al., we engaged its newly developed fractionation technique and could show that our findings were in accordance with data on prostate cancer, which showed a higher LOH detection rate in the LMWF compared to the HMWF
. To date, with the exception of this study, there is no further investigation of LOH on fractionated cell-free DNA. However, Wang et al. used a column-based, modified Guanidine/Promega Resin (G/R) method to selectively isolate fragmented DNA and showed for colorectal cancer patients that preferential isolation of fragmented DNA enhances the detection of circulating mutated K-RAS DNA sequences
. In this context, Wang proposed that in order to enhance the detection rate of somatic mutations or epigenetic modifications in circulating blood DNA, a method selectively extracting small DNA fragments should be used. Given these considerations and comprising our data, we can conclude for ovarian cancer that tumor DNA is rather reflected in the LMWF.
In comparison to preoperative sera, LOH incidence only moderately decreased after chemotherapy. Tracing microsatellite alterations in follow-up sera of patients with different cancer entities and investigating their clinical significance has previously been described in several studies
[27, 28]. In a previous investigation on breast cancer, overall LOH pattern of serial serum DNA samples was largely stable but it remained unclear if the remaining non concordant results were due to changes in clinical status
. In another study, the persistence of microsatellite alterations in plasma DNA of post-mastectomy patients has been associated with bad-prognosis histological parameters and indicated micrometastatic disease
. The detection of tumor associated microsatellite alteration after accomplished chemotherapy could indicate the presence of residual primary tumor or occult micrometastatic cells, such as circulating tumor cells or DTC, being resistant to treatment.
Apart from patients who’s LOH vanished upon chemotherapeutic intervention, a subset of patients displayed LOH events after chemotherapy which were not detectable before surgery. This observation has also been described in small cell lung cancer patients (SCLC). In case of a successful therapy, molecular alterations on cell-free DNA concomitantly vanished and new LOH events after chemotherapy coincided with disease recurrence
. A statistically substantiated conclusion about the prognostical impact of “de novo” LOH occurrence after chemotherapy in our patient cohort cannot be performed based on the limited number of this patient group. However, one can conceptually speculate that a certain subset of cancer cells with stem like properties is not affected by cytoreductive treatment, might evolve by clonal evolution, might systematically acquire de novo genetic alterations and might become capable of effecting recurrence. Accordingly, it could be observed previously that a side population of tumor cells with stem cell features, overexpressing ABC drug transporters, sustained the growth of drug-resistant ovarian tumors and induced recurrence
Considering LOH at the different loci, we observed that LOH at D10S1765 in LMWF after therapy was predictive for tumor cell dissemination to BM after therapy, FIGO stage and tumor grading. Significant associations of this marker with FIGO stage and grading coincide with our previous LOH investigation in primary ovarian tumor tissue
. D10S1765 is located at chromosomal band 10q23.3. This region encodes for PTEN, a dual-specificity protein phosphatase antagonizing the PI3K-Akt signaling pathway and regulating cellular proliferation, DNA repair, stem cell self-renewal, genomic instability and metastasis
. Particularly, in ovarian cancer, PTEN down-regulation has been shown to be functionally involved in platinum resistance
. It seems rational that highly advanced tumors are more dependent on the accumulation of genetic alterations than early-stage neoplastic lesions. Furthermore, it can be hypothesized that PTEN locus disruption by LOH, in combination with other inactivation mechanisms such as epigenetic events
, might functionally be involved in tumor progression, persistence of tumor cells after chemotherapy and their capacity to disseminate to the BM.
We revealed that LOH in cirDNA of ovarian cancer patients is predictive for FIGO stage and tumor grading. Moreover, one of our key finding was that LOH at marker D6S1581 in a combined analysis of both fractions, detected at primary diagnosis, was predictive for reduced OS. Microsatellite marker D6S1581 is located at 6q25.1, 25 kb proximal to the mannose 6-phosphate/insulin-like growth factor II receptor (M6P/IGF2R) locus
. This receptor is well described in breast cancer
[37–39] and is supposed to act as a tumor suppressor gene by negatively regulating cell survival, tumor invasion and metastasis
. Moreover, LOH at M6P/IGF2R is well known to be a frequent event in primary ovarian cancers
 and a low D6S1581 copy number was associated with platinum-resistant condition in ovarian cancer
. However, in our patient cohort, LOH occurrence did not significantly associate with platinum resistance. Generally, advanced ovarian cancer is a relatively chemo-sensitive tumor with overall clinical response rates of 70 – 80%
. Therefore, the observed number of resistant cases in our study (14.3%) might be too limited to allow a reliable and statistical substantiated conclusion.
The ascertained prognostic impact of marker D6S1581 in cirDNA of ovarian cancer patients in our current study is of particular interest, as our present serum data corroborate our previous findings on ovarian tumor tissue showing that allelic loss at D6S1581 is a biomarker for tumor cell spread to the BM
. Thus, evidence is accumulating, that genomic region proximal to M6P/IGF2R locus seems to be functionally implicated for ovarian tumorigenesis and may serve as a blood-based biomarker for prognosis. These data clearly sustain the reliability of cell-free cirDNA as “real-time liquid biopsy”, providing valuable prognostic information and being accessible non-invasively. Moreover, most of the significant associations with the clinicopathological parameters were found at the time of primary diagnosis before surgery. Thus, preoperative serum seems to constitute most informative reading point for patients with primary ovarian cancer.
The occurrence of artificial LOH in the LMWF, which is supposed to be random allelic loss due to low DNA concentration or impaired DNA integrity, constitutes a possible draw-back in the context of low-molecular-weight DNA amplification. However, random allelic loss due to a suboptimal DNA integrity would logically be accompanied by a statistically more frequent loss of the “shorter” allele of a polymorphic microsatellite maker, which should be more difficult to amplify than the “longer” allele. Despite from the addition of TMAC in order to prevent artificial LOH and to improve amplification condition, we found the number of losses of “short” and “long” alleles being well balanced throughout the entire marker panel and throughout all patients. Thus, it is very unlikely that observed LOH in the LMWF reflect a random process.
We previously tried to characterize fragment size distribution of HMWF and LMWF
, but, however, due to our experimental conditions and due to a partly overlapping performances of the different filter-column systems, so far, it was not possible to narrow down an exact cut-off for cirDNA fractionation. In this context, further studies are planned to better characterize size distribution of cirDNA fractions.
Conclusively, in the present study, we demonstrate the applicability and necessity of DNA fractionation in order to selectively monitor tumor associated microsatellite alterations in blood serum of ovarian cancer patients. Given our present findings, tumor DNA seems to be mostly reflected in the LMWF. It is of note that tumor specific circulating LOH with diagnostic relevance is not exclusively limited to the LMWF. Based on our findings, we have to consider that the HMWF might also contain a certain proportion of tumor derived DNA and should additionally be analyzed for blood-based LOH studies on ovarian cancer. Moreover, we newly identified LOH at D10S1765 and D6S1581 as novel blood-based biomarkers for tumor cell spread to the BM and prognosis, respectively, which might be of clinical relevance for monitoring studies. Genomic region proximal to M6P/IGF2R, a promising target, and its role in ovarian cancer is currently under investigation.