Circulating cell-free methylated DNA and lactate dehydrogenase release in colorectal cancer

Background Hypermethylation of DNA is an epigenetic alteration commonly found in colorectal cancer (CRC) and can also be detected in blood samples of cancer patients. Methylation of the genes helicase-like transcription factor (HLTF) and hyperplastic polyposis 1 (HPP1) have been proposed as prognostic, and neurogenin 1 (NEUROG1) as diagnostic biomarker. However the underlying mechanisms leading to the release of these genes are unclear. This study aimed at examining the possible correlation of the presence of methylated genes NEUROG1, HLTF and HPP1 in serum with tissue breakdown as a possible mechanism using serum lactate dehydrogenase (LDH) as a surrogate marker. Additionally the prognostic impact of these markers was examined. Methods Pretherapeutic serum samples from 259 patients from all cancer stages were analyzed. Presence of hypermethylation of the genes HLTF, HPP1, and NEUROG1 was examined using methylation-specific quantitative PCR (MethyLight). LDH was determined using an UV kinetic test. Results Hypermethylation of HLTF and HPP1 was detected significantly more often in patients with elevated LDH levels (32% vs. 12% [p = 0.0005], and 68% vs. 11% [p < 0.0001], respectively). Also, higher LDH values correlated with a higher percentage of a fully methylated reference in a linear fashion (Spearman correlation coefficient 0.18 for HLTF [p = 0.004]; 0.49 [p < .0001] for HPP1). No correlation between methylation of NEUROG1 and LDH was found in this study. Concerning the clinical characteristics, high levels of LDH as well as methylation of HLTF and HPP1 were significantly associated with larger and more advanced stages of CRC. Accordingly, these three markers were correlated with significantly shorter survival in the overall population. Moreover, all three identified patients with a worse prognosis in the subgroup of stage IV patients. Conclusions We were able to provide evidence that methylation of HLTF and especially HPP1 detected in serum is strongly correlated with cell death in CRC using LDH as surrogate marker. Additionally, we found that prognostic information is given by both HLTF and HPP1 as well as LDH. In sum, determining the methylation of HLTF and HPP1 in serum might be useful in order to identify patients with more aggressive tumors.


Background
Colorectal cancer (CRC) is the third most common cancer and the fourth most frequent cause of death from cancer worldwide with about 1.2 million cases and about 633,000 deaths in 2008 [1]. Despite significant advances in the last decades, especially patients with metastatic disease suffer from poor prognosis [2]. In addition to new therapeutic options, biomarkers are needed that allow the identification of different subgroups of patients potentially benefitting from different treatment regimens and intensity.
In many human cancers aberrant hypermethylation of CpG islands is a common epigenetic DNA modification leading to transcriptional silencing of genes that is already detectable in early stages of carcinogenesis [3]. Genes found hypermethylated in colorectal cancer have many functions, including mismatch repair, cell-cycle regulation and cell differentiation [4]. Methylated tumor DNA cannot only be found in primary colorectal cancer tissue, but can also be detected in remote media like serum or stool and potentially be used as biomarkers for various purposes [5][6][7]. We have previously described methylation of the genes neurogenin 1 (NEUROG1) in serum and HIC1 in stool as diagnostic markers [8,9] and helicase-like transcription factor (HLTF) and hyperplastic polyposis 1 (HPP1), also known as transmembrane protein with EGF-like and two follistatin-like domains 2 (TMEFF2), as prognostic serum markers [10,11].
NEUROG1 is a basic helix-loop-helix transcription factor which has been identified as one of the main players in neurosensory evolution and development, especially of the inner ear [12]. Moreover NEUROG1 has been described to be frequently hypermethylated in colorectal cancers and has been proposed as a marker to classify the CpGisland methylator phenotype in colorectal cancers [13,14].
HLTF is a transcription factor and a member of the SWI/SNF family of chromatin-remodeling factors [15]. The physiological function of HLTF has not yet been fully understood, but evidence for its association with genesis and progression of cancer exists [16]. Recently HLTF deficiency has been reported to significantly increase the formation of small intestinal adenocarcinoma and colon cancer in mice on a Apc min/+ mutant background and to be associated with chromosomal instability [15]. Hypermethylation of HLTF can commonly be found in all stages of CRC as well as in adenomas and is associated with tumor size, stage and poor prognosis [17][18][19][20]. Besides its occurrence in serum, methylated HLTF has also been detected in stool samples of CRC patients [21,22].
Lactate dehydrogenase (LDH) is essential for anaerobic glycolysis and reversably converts pyruvate to lactate. Its expression has been shown to be related to the hypoxia inducible factor HIF-1 [33][34][35][36]. Activation of the HIF pathway is a common finding in cancers [37,38]. LDH in serum is a frequently used parameter in clinical routine and is released upon cell membrane disintegration. Thus, it is an unspecific marker for tissue damage, e.g. caused by necrosis. Elevated LDH levels can be found in numerous diseases including myocardial infarction, hemolysis and malignancies [39]. Additionally LDH has been reported to be associated with more aggressive tumors and shorter survival [40][41][42][43] in CRC. In other cancer entities like testicular cancer [44,45] and aggressive non-hodgkin lymphoma [46] elevated LDH levels are used as prognostic biomarkers. Recently, LDH has been discussed as a predictive biomarker for anti-angiogenic therapies in colorectal cancer [43,47,48].
Cell death, especially necrosis, is considered to be the source of circulating cell-free DNA (cfDNA) in cancer patients [49,50]. However, the exact mechanisms leading to the release of the tumor markers discussed here with prognostic (HLTF and HPP1) or diagnostic (NEUROG1) information have not been examined so far. This study aimed at investigating a possible correlation of the presence of the methylated genes NEUROG1, HLTF and HPP1 in serum with tissue breakdown as a possible release mechanism using serum lactate dehydrogenase (LDH) as a surrogate marker. Additionally, the prognostic information given by these markers was examined.

Patients and serum samples
Pretherapeutic serum samples from 259 patients with colorectal cancer were included in the study. For these cases clinicopathologic and follow-up data as well as pretherapeutic lactate dehydrogenase values were available. Characteristics of the cohort are shown in Table 1. All measurements were performed blinded to patient data. Blood samples were obtained pretherapeutically and underwent the following standardized preanalytical procedure: All specimens were transported by a shock absorbed tube mailing system within 15 to 30 minutes after blood drawing to the central laboratory, followed by centrifugation at 2,000 g at 4°C for 10 minutes. The supernatant serum was transferred into polypropylene cryotubes and stored frozen at −80°C. In each case, DNA methylation and lactate dehydrogenase levels were determined in the same blood sample. The study was approved by the ethical committee of the Medical Faculty of the University of Munich.

DNA isolation and bisulfite conversion
The frozen serum samples were thawed at room temperature and homogenized by smoothly flipping the tube containing the serum. Genomic DNA from 200 μL of each serum sample was isolated using the High Pure Viral Nucleic Acid Kit (Roche Applied Science, Mannheim, Germany) according to the manufacturer's instructions and eluted in 50 μl of Elution Buffer. Bisulfite conversion was performed as described previously [11].

Analysis of DNA methylation
Bisulfite-treated DNA was analyzed by a fluorescencebased, real-time PCR assay, described previously as Methy-Light [51]. Dispersed Alu repeats were used to control for DNA amplification and to normalize for input DNA. Primer and probe sequences are listed in Additional file 1:

Clinicopathologic features and DNA methylation in serum
A total number of 259 serum samples were analyzed. An overview of the clinocopathologic characteristics is shown in Table 1. Methylation of HLTF was detected in 41 cases (16%), methylation of HPP1 in 57 cases (22%) and methylation of NEUROG1 in 66 cases (25%). The distribution of PMR values is demonstrated in Additional file 2: Table S2. HLTF methylation in the serum was significantly correlated with metastatic diseases (p = 0.013) and advanced tumor stages (p = 0.0489) as well as T4 tumors (T1-3 vs. T4, p = 0.046). A non-significant trend towards spread to lymph nodes was observed (N0 vs. N1-2, p = 0.050). HPP1 methylation in serum was significantly correlated with larger tumor size (p < 0.001), positive nodal status (p < 0.0001), metastatic disease (p < 0.0001), tumor stage (p < .0001) as well as higher tumor grades (p = 0.0002). No significant correlation between NEUROG1 methylation and clinicopathologic features existed. The complete distribution of the markers among the clinicopathologic features is presented in Table 2. LDH values ranged from 100 to 1730 U/l with a mean value of 238 U/l (standard deviation 202 U/l) and a median value of 185 U/l. A cutoff of 250 U/l, representing the upper limit of normal of the assay used, was chosen, resulting in 50 patients (19%) with elevated LDH levels. These patients suffered more frequently from T4 tumors (T1-3 vs. T4, p = 0.038), nodal and distant metastases (p = 0.0006 and p < 0.0001, respectively) as well as higher tumor stages (p < 0.0001). Additionally, a non-significant trend towards higher LDH levels in younger patients was found (p = 0.055).

Correlation between LDH and DNA methylation in serum
First we analyzed the correlation of methylation of HLTF, HPP1 and NEUROG1 with LDH in a binary way. For this purpose we used a cutoff of LDH at 250 U/l as mentioned above. For the methylation markers we considered a PMR > 0 as methylation positive which has been shown previously to be reasonable for serum methylation analysis by our and other groups [10,52,53]. In the 50 samples with elevated LDH levels, methylation of HLTF, HPP1, or NEUROG1 was detected in 16 (32%), 34 (68%), or 12 cases (24%), respectively, compared to 25 (12%), 23 (11%), or 54 (26%) in those 209 samples with normal LDH levels. Patients with elevated LDH levels revealed significantly more often methylation of HLTF or HPP1 (p = 0.0005 or p < 0.0001, respectively), whereas no correlation between NEUROG1 methylation and elevated LDH was found.
We also examined the relation of the methylation markers between each other. Methylation of HLTF was found significantly more often in HPP1 positive samples (51% vs. 17%, p < 0.0001). No significant difference in the frequency of either HLTF or HPP1 methylation was observed between NEUROG1 positive and NEUROG1 negative cases (32% vs. 24% and 26% vs. 25%, respectively). In a second step, correlations were analyzed using LDH as a continous variable without cutoff. In HPP1 positive samples significantly higher LDH levels were measured (median 298 U/l vs. 173 U/l, p < 0.0001). Patients with methylation of HLTF had slightly, but still significantly higher LDH levels (median 208 U/l vs. 180 U/l, p = 0.0050), while no difference was found in LDH levels between NEUROG1 positive and negative samples (median 187 U/L vs. 184 U/l, p = 0.95). Figure 1 provides a more detailed view on the distribution of LDH levels among the three methylation markers.
Additionally, we tested HLTF, HPP1 and NEUROG1 as continuous variables without cutoff using the PMR values and calculated univariate Spearman correlation coefficients. As in the analyses before, HLTF and HPP1 showed significant correlation with LDH, while NEUROG1 did not. All linear correlation coefficients and p-values are presented in Table 3.

Multivariate model
Next, a multivariate model was developed using logistic regression analysis with LDH values higher than 250 U/l as target variable. HPP1 and HLTF methylation as binary parameters, i.e. with a PMR > 0, as well as clinicopathological features were entered as independent variables. Only presence of distant metastases and HPP1 correlated significantly and independently with elevated LDH levels higher than 250 U/l. The odds ratios were 3.1 for metastatic disease (95% CI 1.3-7.2, p = 0.009) and 9.5 for HPP1 methylation (95% CI 4.2-21.9, p < 0.0001).

Survival analysis
We earlier reported methylation of HLTF and HPP1 to be independent prognostic markers in metastastatic colorectal cancer [11]. On the other hand, elevated LDH levels have been described to be linked to shorter survival [54]. Thus we compared methylation of HLTF and HPP1 with LDH as prognostic factors in our patient population.

Discussion
In this study we examined the correlation between cell damage using LDH as a surrogate marker and the methylation status of three genes which have previously been proposed as prognostic (HLTF, HPP1) [10,11] or diagnostic (NEUROG1) [8] biomarkers for patients with CRC.
Our data confirm our previous findings that methylation of HLTF or HPP1 in serum is found more often in Table 3 Linear Spearman correlation coefficients for the percentage of fully methylated reference (PMR) of HLTF, HPP1 and NEUROG1, and LDH levels among each other patients with advanced stages of colorectal cancer, especially in those with distant metastases, whereas no correlation between methylation of NEUROG1 and any clinicopathologic data was found. While methylation of HLTF was only correlated with metastastatic disease, methylation of HPP1 was also associated with local tumor extent and nodal status as well as tumor grade with high statistic significance. It is well known that patients with elevated serum levels of LDH tend to have more aggressive tumors and a shorter survival time [40][41][42][43]. Consistent with the literature high LDH levels in our data were significantly correlated with advanced tumor stages as well as nodal and distant metastases. Trends towards larger tumor size and younger age were observed but did not reach statistical significance.
Cell death associated mechanisms like apoptosis or, especially in cancer, necrosis have been suggested as main sources for cell-free DNA (cfDNA) in the blood, but other mechanisms like physiological active release have been described as well (for reviews see refs. [49,50]). In this study we found methylation of HLTF and, even to a higher degree, HPP1 to be correlated with elevated LDH levels. This finding was robust, as it was confirmed by different statistical methods. Given that elevated LDH indicates cell membrane damage, this observation might be a hint that methylated HLTF and HPP1 DNA is released by tumor cells undergoing cell death. The fact that necrosis tends to be found more often in larger, more aggressive tumours and advanced cancer stages [55,56], which was likewise the case for LDH as well as methylated HLTF and HPP1 in our data, also suggests an interrelation.
For NEUROG1, on the other hand, hypermethylation in serum was detectable independently of LDH levels and tumor stage. This is consistent with earlier analyses revealing methylation of NEUROG1 in primary tissue not to be associated with tumor stage (A.P. and F.K., data not published). Hence the observed correlation between DNA methylation in serum and LDH seems not to be linked to global methylation levels and cell death alone. Besides the methylation status of distinct genes, other parameters influencing this observation might include DNA integrity and stability of the respective segments as well as still unknown factors. Therefore it seems likely that tumor cell death might not be the only mechanism by which methylated tumor DNA is released to the blood.
In addition to the correlation analysis we examined the prognostic significance of the methylation markers HPP1 and HLTF as well as of LDH. All three markers were significantly associated with worse overall survival. This could be attributed to the fact that all three markers are found more frequently in advanced cancer stages. However, earlier analyses [11] as well as the survival data presented here furthermore divide patients with already metastasized disease into two subgroups with better or worse prognosis, respectively.

Conclusion
In conclusion we were able to provide evidence that methylation of HLTF and especially HPP1 detected in serum is strongly correlated with cell death in colorectal cancer using LDH as surrogate marker. However, this finding was specific for those two genes and did not occur for NEUROG1, suggesting that mechanisms other than release by membrane disintegration could be responsible for the occurrence of cell-free DNA in blood of CRC patients. Additionally, we found that prognostic information is given by both HLTF and HPP1 as well as LDH. In sum, determining the methylation of HLTF and HPP1 in serum might be useful in order to identify patients with more aggressive tumors. Future research needs to further clarify the underlying biological mechanisms and to validate methylated cell-free circulating DNA as a biomarker for colorectal cancer.

Competing interests
The authors declare that they have no competing interest.