Correlation of SHOX2 Gene Amplification and DNA Methylation in Lung Cancer Tumors

Lung cancer represents the second most common cancer in both men and women and accounts for about 15% of all cancer diagnoses [1]. In the absence of screening, lung cancer patients either exhibit symptoms or are accidentally diagnosed by clinical imaging performed for other indications. Patients suspected of having malignant lung disease usually undergo clinical investigation (workup) including CT-scanning of the thorax and bronchoscopy. The latter is preferentially applied to confirm a lung neoplasm by pathological assessment of tissue or a cytological specimen obtained during the procedure. A final diagnosis after the first bronchoscopy fails in about half of these patients [2], which requires additional invasive diagnostic procedures. Even when signs, symptoms and radiological findings are such that the clinical diagnosis of malignant lung disease appears obvious, it often takes considerable effort and invasive procedures to obtain tissue material suitable for definitively confirming the presence of malignant disease.

Ambiguous results following bronchoscopy are frequent (i.e. the presence of a malignancy cannot be confirmed), e.g. because the tumor is not visible endoscopically and cells obtained by brushing or aspiration do not allow the pathologist to confirm or exclude malignancy.

Biomarkers exhibit great potential for improving the management of cancer in clinical routine. So far, several biomarkers from genetic, proteomic and epigenetic approaches are available for clinical research purposes [3‐5]. The use of DNA methylation as a biomarker is an emerging field that provides potential for advancing the clinical process of cancer diagnosis [6‐10]. Methylation of DNA is an important epigenetic process involved in fundamental biological events such as development and cell differentiation [for review: [11]]. Aberrant DNA methylation has been reported to play a major role in carcinogenesis [for review: [12]], suggesting that DNA methylation analysis may be a valuable source for cancer biomarkers [13]. Recently, SHOX2 DNA methylation was shown to be a useful biomarker for detecting cancer patients at high specificity and sensitivity in a group of critical controls based on the analysis of bronchial lavage samples [14]. Interestingly, samples that were classified as 'cytologically negative' or 'inconclusive' due to no (or too few) visible tumor cell content could be identified as cancer-positive, based on their SHOX2 DNA methylation level.

SHOX2 has been identified as highly homologous to the short stature homeobox gene SHOX. Both homeodomain-transcription factors are involved in skeletogenesis, and the murine Shox2 gene has been shown to play a pivotal role during heart development [15‐18].

Genomic gain of chromosome 3q involving the SHOX2 gene has been recognized as one of the most prevalent and significant chromosomal rearrangements in lung cancer [19‐23]. The positive performance of SHOX2 DNA methylation as a biomarker in cytologically negative bronchial lavage samples might be due to the concerted effects of locus amplification and DNA methylation of SHOX2 in tumor cells. As a result, an increase of SHOX2 DNA copies in tumor cells compared to normal cells also increases the SHOX2 DNA methylation level in a mixed sample of tumor and normal cells as compared to the methylation level of normal cells alone. In this study the DNA methylation, gene expression and gene amplification of SHOX2 in matched tumor and normal adjacent tissues (NAT) from 55 lung cancer patients were investigated.

SHOX2 DNA methylation has previously been reported to be applicable for the diagnosis of lung cancer based on the analysis of bronchial lavage samples [14]. In this study, the SHOX2 DNA methylation was quantified in tumor tissues and morphologically normal adjacent tissues from 55 lung cancer patients. In brief, this assay represented a duplex real-time PCR consisting of a HeavyMethyl (HM) assay [28] for sensitive and quantitative detection of SHOX2 DNA methylation (promoter region of transcript variant b [14]) and a reference PCR for quantification of the total DNA using the ACTB locus. A response curve of this assay is given in figure 1 with different mixtures of bisulfite converted DNA from sperm (known to be unmethylated at many loci [25]) and bisulfite converted DNA from artificially methylated DNA. The assay allowed for highly accurate quantification of the SHOX2 DNA methylation over a broad range of relative methylation of the template DNA.

As indicated in figure 2, SHOX2 was methylated in the vast majority of tumors. Ninety-six per cent (53 out of 55) of matched pairs showed a higher methylation level in tumor tissues as compared to the matched NAT from the same patient. The analyzed SHOX2 was significantly higher methylated in squamous cell carcinomas as compared to adenocarcinomas (p = 0,0006 [Wilcoxon Rank-Sum test], Additional File 1). Low SHOX2 DNA methylation levels were also present in 55% (30 out of 55) of the NAT ranging from 0.01% to 0.3%. Interestingly, the SHOX2 methylation level of three patients' DNAs was above the theoretically expected maximum of 100% (patient #1: 291%, patient #2: 130%, patient #3: 122%). This observation is unlikely to be a measurement artefact since the response curve in figure 1 showed high accuracy of the detection assay.

DNA methylation within the promoter region of a gene is known to be part of a regulatory mechanism that is able to silence gene expression [for review: [29]]. The correlation between SHOX2 DNA methylation and gene expression was analyzed with three different quantitative real-time PCR assays. Two assays that were specific for the transcript variants SHOX2a and SHOX2b, respectively, which are suggested to be transcribed from alternative promoters [15] were designed. The third assay was designed to detect the total amount of SHOX2 RNA transcripts. The relative expression level of the two SHOX2a and b variants alone, as well as the total amount of SHOX2 RNA in tumor tissues and normal tissues, is shown in figure 3 and Additional File 1. A downregulation of SHOX2 expression due to a hypermethylation of the SHOX2 region in tumor tissues could not be observed. Contrarily, the SHOX2 expression seemed slightly elevated in tumor tissues, however, this was not statistically significant.

Genomic gain on chromosome 3q, where SHOX2 is located, has been recognized as one of the most prevalent and significant alterations in lung cancer [19‐23]. Therefore, the SHOX2 copy numbers were analyzed in order to investigate a potential correlation of SHOX2 DNA methylation and gene amplification. Quantities of SHOX2 copy numbers were characterized by the ratio of total SHOX2 DNA (to the total amount of DNA). The amount of total DNA was quantified by UV spectrophotometry. Quantities of SHOX2 copies were measured using a real-time PCR assay which specifically amplifies bisulfite converted copies of the SHOX2 gene (methylated and unmethylated). Even though the primers used for this assay did not include CpG islands, a PCR bias (i.e. preferred amplification of unmethylated or methylated DNA), had to be excluded [30‐32]. Otherwise, a more efficient amplification of methylated DNA would result in overestimated SHOX2 copy numbers in methylated specimens. Therefore, mixtures (0 to 100%) of bisulfite converted DNA from sperm and from artificially methylated DNA were analyzed with respect to a potential PCR bias. 20 ng DNA (according to UV) was subjected to a triplicate real-time PCR analysis. The resulting CT were 28.70 ± 0.33 (0% mixture), 28.55 ± 0.13 (25%), 28.54 ± 0.30 (50%), 28.37 ± 0.12 (75%), and 28.16 ± 0.11 (100%), respectively, indicating that no significant PCR bias exists for this assay. The determined SHOX2 copy numbers in relation to DNA methylation in the 55 tumor tissue samples is shown in figure 4 and Additional File 1. High SHOX2 DNA methylation was found to correlate strongly with gene amplification (p = 0.0002 [Kendall's tau rank correlation], p < 0.0001 [Spearman rank correlation]). Amplification of the SHOX2 gene locus did not differ significantly between adeno- and squamous cell carcinomas.

DNA methylation of the SHOX2 locus as shown in figure 4 was determined using a SHOX2/ACTB duplex real-time PCR. The SHOX2 assay specifically amplified methylated SHOX2 gene copies, while the ACTB primers amplified methylated and unmethylated copies of the ACTB gene. Thus, it cannot be excluded that the apparent higher relative SHOX2 methylation in tumors with SHOX2 gene amplification is simply due to the presence of higher numbers of total SHOX2 copies and therefore also a higher number of methylated copies. This would result in an overestimation of the relative methylation in samples with amplification of the SHOX2 locus. A sample where both SHOX2 gene copies are methylated would result in a methylation level of 100%. However, a methylation level of 200% would indicate that the SHOX2 locus is duplicated when a different locus, i.e. ACTB, is used for quantification of the total DNA. Thus, the apparent relative methylation of the SHOX2 locus correlates with the copy numbers due to the usage of a different locus as a reference assay. To circumvent this overestimation of relative SHOX2 methylation a reference assay which is located at the same locus was applied, where completely methylated SHOX2 gene is determined as 100%, independent of agene amplification. Thus, a duplex PCR was developed, where methylated SHOX2 copies were directly correlated to the total copy number of the SHOX2 gene itself. The correlation of the SHOX2 methylation as determined with an ACTB and a SHOX2 assay for quantifying total DNA is shown in figure 5 and Additional File 1. Using this duplex assay, none of the samples was determined as methylated above 100%. This confirmed that methylation determined by the ACTB reference assay was overestimated for samples showing gene amplification. However, the DNA methylation levels quantified by both reference assays showed strong correlation (p < 0.0001 [Kendall's tau and Spearman rank correlation]). In addition, DNA methylation quantified with the SHOX2 reference assay still showed a significant correlation with gene amplification (p = 0.002 [Kendall's tau and Spearman rank correlation]).

Tumor tissue derived from patient #1 showed a very high methylation rate of 291% (referred to ACTB) and 46% (referred to total SHOX2 copies), respectively, as well as a strong copy number amplification of the SHOX2 gene (4.9 fold). This tumor DNA was analyzed in more detail using array CGH technology. A summary of the CGH results are depicted in figure 6 showing that the complete q arm of chromosome 3 is overrepresented in the tumor. Taken together, this data clearly demonstrate that the SHOX2 locus is highly amplified in the tissue derived from patient #1. In summary, the presented results indicate that DNA methylation of the SHOX2 gene highly correlates with the amplification of its chromosomal location.

SHOX2 DNA methylation, operating as a biomarker at a high specificity and sensitivity in a group of lung cancer patients and critical controls based on the analysis of bronchial lavage samples, has recently been proven to be a suitable means for diagnosing lung cancer [14]. The biomarker also worked with samples that contained no visible tumor cell content. Since it has also been found that genomic gain at chromosome 3q, including SHOX2, is one of the most prevalent and significant alterations in lung cancer [19‐23], the objective of this study was to investigate the relationship between gene amplification, expression and DNA methylation of SHOX2. In this study, it was shown that in 96% of all cases of lung tumor patients, the SHOX2 gene is hypermethylated and frequently accompanied by increased copy numbers of the respective locus. Hence, the positive performance of the SHOX2 DNA methylation as a biomarker in cytologically negative bronchial lavage samples might be due to a correlation of locus amplification and DNA methylation in tumor cells. Genomic amplification did not lead to a statistically significant higher expression of SHOX2 a or b transcripts as compared to the respective NAT. However, it has to be considered that SHOX2 mRNA levels were measured in homogenized (lysed) tumor tissues. Therefore, the extracted mRNA from tumor tissues is susceptible to contamination by non-neoplastic cells. This contamination might mask tumor-specific alterations. Such contamination affects mRNA expression analyses much more dramatically as compared to DNA methylation and gene amplification since the expected ranges (fold changes) of mRNA expression between different cell types can be much higher. In cancer research, the development of laser microdissection (LMD) systems has addressed this dilemma and could be implemented in the future to study SHOX2 expression in distinct tumor cell types.

Interestingly, NAT also showed low levels of methylation (< 0.3%). This observation might arise from a potential field effect, leading to alterations of methylation in normal components adjacent to the tumor, or from the existence of a subpopulation of cells with similarities to the tumor cells on an epigenetic level. Alterations of methylation in NAT and the existence of a subpopulation of cells within the normal epithelia with epigenetic similarities to the tumor cells have previously been reported in several cancers [33, 34]. In addition, tumor development and progression occur as an interaction between tumor cells and their stromal environment [for review: [35, 36]], and the concept of epithelial to mesenchymal transition (EMT) as a driver of tumor progression has become more important in recent years [37, 38]. Thus, cells, which might not allow for a morphological cancer diagnosis do already harbour the cancerogenic information by their epigenetic state. However, in order to address the question of whether methylation in NAT comes from such a field effect or is also present in tissue from healthy individuals, normal components at larger distances from the tumor or in tissue from healthy controls would have to be analyzed.

Both, a potential field effect on an epigenetic level, as well as a concerted mechanism of gene amplification and DNA methylation, renders DNA methylation an attractive source of biomarkers for the diagnosis of cancer in body fluids where sensitive detection is required. The reinforced epigenetic signal, together with chromosomal rearrangements and thus locus amplifications, may be part of a cellular mechanism during tumorigenesis. Therefore, DNA methylation may be preferrentially employed as a diagnostic test compared to methods based on cell counting. Additionally, the detection and accurate quantification of DNA is easier as compared to protein and RNA, due to the higher stability of DNA in quantitative real-time PCR. This study also indicates that a combination of genome-wide DNA methylation analysis, i.e. MeDIP and DMH, and CGH is effective for the screening of candidate marker genes and therefore might also influence the discovery of novel biomarkers for clinical applications. A multiplexing of such markers, ideally from different chromosomes might further improve a diagnostic test.

However, the functional relationship between DNA methylation and amplification as well as the underlying pathomechanism remains to be elucidated. It is not yet clear whether DNA methylation is a response reaction to locus amplification or if DNA hypermethylation promotes destabilization of the genome and with that gene amplification. However, it is well acknowledged that a genome-wide hypomethylation induces chromosomal instability [for review: [11]] and that genome-wide hypomethylation in cancer goes hand-in-hand with gene-specific hypermethylation, e.g. of tumor suppressor genes [for review: [39]]. Thus, SHOX2 hypermethylation in lung cancer might be indicative of an overall hypomethylation of the corresponding chromosomal region and therefore of genetic instability which results in gene amplification.

The tumor-specific hypermethylation of SHOX2 and its frequent gene amplification in lung cancer renders SHOX2 DNA methylation is an attractive biomarker for lung cancer when sensitive detection is desired, i.e. in bronchial lavage specimens.