A high-throughput and sensitive method to measure Global DNA Methylation: Application in Lung Cancer

To quantitate the amount of DNA methylation in any genome 100 ng of a sample was aliquoted nine times into a 96 well white Microfluor 2 plate (Thermo Electron, Waltham, MA). The genomic DNA in the first three wells was digested with 5 units of a methyl-sensitive restriction enzyme such as HpaII (New England BioLabs, Beverly, MA), the DNA in the second three wells was digested with 5 units of the methyl-insensitive restriction enzyme MspI (New England BioLabs, Beverly, MA) to normalize the data when calculating the Global DNA Methylation Index (GDMI), and the third set of triplicates was for buffer only (NEBuffer1 – New England BioLabs, Beverly, MA). All reactions were performed in a total volume of 30 μl. Prior to incubation the 96 well microtiter plate was sealed with Adhesive PCR Foil (ABgene Inc., Rochester, NY) using an ALPS 300™ (ABgene Inc., Rochester, NY), spun briefly and placed in an air incubator for 3 hours at 37°C. After incubation, the plate was spun briefly and the Adhesive PCR Foil removed. The digestion of DNA was followed by an end-fill reaction where 20 μl of biotinylation buffer containing Biotin-11-dCTP and Biotin-11-dGTP (Perkin Elmer, Boston, MA) and Sequenase (USB Corporation, Clevland, OH) in 40 mM Tris-HCl, pH 7.5, 20 mM Tris-HCl, 50 mM NaCl) was added. The final concentration of biotinylated dCTP and dGTP and Sequenase used per well was 0.1 μM and 0.1 units respectively. The plate was sealed, spun briefly and placed into an air incubator for 30 minutes at 37°C. After the incubation period the plate was spun briefly, the Adhesive PCR Foil removed and 100 μl of Reacti-Bind™ DNA Coating Solution (Pierce, Rockford, IL) was added to each well. After mixing, the plate was once again sealed and placed on an orbital platform shaker (Lab-Line Instruments Inc., Melrose Park, IL) and shaken at 150 RPM at room temperature overnight. The solution in the wells was removed and the wells were washed 4 times with TBS (10 mM Tris-HCL pH 8.0, 150 mM NaCl). The biotin was detected using HRP Neutravidin (Pierce, Rockford, IL) and the DNA Detector kit (KPL, Gaithersburg, MD). Briefly, 200 μl of the Detector Block Solution (KPL, Gaithersburg, MD) was added to each well and the plate was incubated for 30 minutes at room temperature. After removal of the Detector Block Solution, 150 μl of Detector Block Solution containing 0.5 μg/ml (1:2000 dilution) of HRP Neutravidin was added to each well and the plate was incubated at room temperature for 30 minutes. The Detector Block/HRP Neutravidin solution was removed and the wells washed 5 times with 1xBiotin wash solution (KPL, Gaithersburg, MD). After the final wash 150 μl of LumiGlo^® chemiluminescence substrate (KPL, Gaithersburg, MD) was added to each well. After 2 minutes the luminescence emitted from each well was quantitated by a Wallac Envision 2100 multilabel reader (Perkin Elmer, Boston, MA).

The linear range of the CpGlobal assay for human genomic DNA was determined by measuring global DNA methylation in human male genomic DNA isolated from whole blood (Novagen, San Diego, CA). A range of DNA concentrations (100 ng, 50 ng, 25 ng, 12.5 ng, 6.25 ng and 3.125 ng) were assayed to determine the linearity of CpGlobal. One set of triplicates were digested with HpaII (New England BioLabs, Beverely, MA), the second set with MspI (New England BioLabs, Beverely, MA) and the third set was treated with buffer only (NEBuffer1 – New England Biolabs, Beverely, MA). The assay and the quantitation of the chemiluminescence were performed as described above.

All cell lines were purchased from ATCC (Manassas, VA) and were grown according to the manufacturer's recommendations. The cell lines that were used to compare the CpGlobal with High Performance Capillary Electrophoresis (HPCE) were SW48 (CCL-231), LoVo (CCL-229), HT-29 (HTB-38) and NCI-H69 (HTB-119D). Assessment of the global DNA methylation was performed on Non-Small Cell Lung Cancer tumor cell lines from various stages of lung cancer. These included a normal lung NL20 (CRL-2503), Stage I NCI-H1703 (CRL-5889), Stage II NCI-H522 (CRL-5810), Stage IIIa NCI-H1993 (CRL-5909), Stage IIIb NCI-H1944 (CRL5907) and Stage IV metastatic liver NCI-H1755 (CRL-5892). All cells were grown in 75 cm² cell culture flasks (Corning Incorporated, Corning, NY) in 5% CO₂ at 37°C. Cells were harvested when they had reached 90% confluency. DNA was isolated using the Blood and Cell Culture DNA Midi Kit (Qiagen, Valencia, CA). The purity and concentration of the DNA was measured using a DU650 spectrophotometer (Beckman Coulter, Fullerton, CA). The quality of the DNA was determined by loading 200 ng on a 1% agarose gel to inspect for any degradation.

Twenty fresh frozen tissues samples, including lung tumors and areas of non involved lung, were obtained from a repository of biological specimens through the lung cancer Specialized Program of Research Excellence (SPORE) at the Vanderbilt Ingram Cancer Center. For each specimen, a quality control of the tissue specimen was obtained from hematoxylin and eosin stained tissue section on adjacent tissue and reviewed by a pathologist prior to analysis. De-identified clinical data elements were shared in compliance of the health insurance portability and accountability regulations. This study was approved by the local Institutional Review Board of both institutions.

Normal lung DNA from young male accident victims was purchased from BioChain Institute, Inc., Hayward, CA.

Approximately 0.1 gm lung tissue from biopsies of paired tumor and normal tissue from NSCLC patients was frozen in liquid nitrogen. The tissue was pulverized to a fine powder using a freezer mill 6750 (SPEX SamplePrep, Metuchen, NJ) according to the manufacturer's recommendations. DNA was prepared using the Blood and Cell Culture DNA midi kit (Qiagen, Chatsworth, CA). The DNA was checked for concentration and purity by reading the absorbance at 260 and 280 nM and was further tested for intactness by running 100 ng on a 1% agarose gel.

Linear regression analyss was performed using EXCEL Analysis ToolPak Regression Analysis. In each case presented single predictor variable was used. R square, Adjusted R square, intercept, coefficient, standard error and F significance were generated by standard Summary Report. Residual Plots and LineFit Plots were used to check for the linearity of the regression.

To determine the net luminescence, the average value for the no enzyme control was subtracted from the individual enzyme values. To calculate the global DNA methylation index (GDMI), the individual net luminescence values for the methylation sensitive enzyme were divided by the average Msp I net luminescence. All calculations were performed in Microsoft Excel.

Both means and medians were calculated when comparing group differences among Normal Non-Disease, Normal Disease and Tumor samples. Only medians were used when the distributions of continuous variables were significantly skewed by small sample sizes such as different histopathological groups. Due to the potential compromise of parametric assumption, non-parametric Wilcoxon/Kruskal-Wallis tests were applied to evaluate the statistical significances for all analysis. P-value <= 0.05 was considered as statistically significant and p-value > 0.05 and <= 0.10 was regarded as border-line significant. All analyses were conducted using JMP5.1 (SAS Institute, Inc., Cary, NC).

CpGlobal was designed to incorporate several key features: easy to use, did not require radioactivity or PCR, functioned in a 96-well microtiter plate and utilized equipment that most laboratories possessed. The design was adapted from a method that utilizes methyl-sensitive restriction enzymes [14], which when applied to genomic DNA would produce a set of digested fragments that could be quantitated and translated into an amount of DNA methylation present in the genome. We have modified the approach by performing all the steps in one 96 well microtiter plate. Therefore, once the genomic DNA from a sample was aliquoted into the well of the microtiter plate it remained in that location through the four stages of treatment: digestion, end-fill with biotinylated nucleotides, attachment to the surface of the well and chemiluminescence. The methyl-insensitive restriction enzyme MspI, the isoschizomer of HpaII, was used in the design to normalize the data collected from the methyl-sensitive restriction enzymes. This step aided in the removal of any intrinsic variations introduced through slight differences in DNA concentrations, and digestion and end-fill reactions. As such a global DNA methylation index (GDMI) was calculated. The arrangement of the DNA in the 96 well microtiter plate was designed so that up to 10 samples per plate could be analyzed (Figure 1). The final step in the assay was the measurement of luminescence through the use of a luminometer. The end-product was an assay designed for an operator to easily manage the measurement of global DNA methylation in multiple samples, without the need for radioactivity, PCR and expensive equipment.

Bild vergrößern

Based on the design of the assay we began with the application of CpGlobal on Lambda DNA with the aim to determine which methyl-sensitive restriction enzyme was the best indicator in measuring global DNA methylation. Two species of lambda DNA were created: one that was unmethylated and the other that was in vitro methylated using M.SssI, a CpG methylase. Five 4 base pair methyl-sensitive restriction enzymes, AciI (CCGC), BstUI (CGCG), HpaII (CCGG), HinP1I (GCGC) and HypCH4IV (ACGT), were chosen based on their frequency within the Lambda genome and the core recognition sequence site. These enzymes were used to digest a series of Lambda DNA mixtures, which consisted of 21 points that extended from 100% to 0% methylated in 5% intervals. GDMI was generated for each 5% increment, which was compared against the theoretical methylation level. A linear regression analysis was performed for each methyl-sensitive restriction enzyme (Table 1). We observed a highly significant correlation of GDMI with the theoretical methylation levels using HinP1I, HpaII and HypCH4IV. Thus, under these conditions the use of any one of these three methyl-sensitive restriction enzymes would be an effective indicator for global DNA methylation.

In order to quantitate global DNA methylation in the human genome it was essential that the assay operates within the linear range. In particular, it was important to determine if the performance of a methyl-sensitive and the methyl-insensitive restriction enzyme was linear with DNA concentration as the net luminescence generated from the assay would be used to calculate the GDMI. Subsequently, CpGlobal was applied to human genomic DNA isolated from whole blood. Using a two-fold serial dilution that ranged from 100 ng to 3.125 ng, the DNA for each concentration was digested in triplicate with HpaII, MspI and treated with buffer only. The net luminescence from each experiment was plotted against DNA concentration (Figure 2a). The results demonstrated a linear regression that correlated highly with DNA concentration and net luminescence (R² HpaII = 0.982, R² MspI = 0.999). The GDMI was calculated for human whole blood DNA and was plotted against DNA concentration. The GDMI was moderately consistent with DNA concentration (Figure 2b). We chose to use 100 ng of human genomic DNA in all the following experiments because the signal to background ratio was at least > 10 fold.

Bild vergrößern

HPCE is viewed as one of the gold standard technologies when measuring the global genomic content of 5-methylcytosine [8]. We applied CpGlobal to calculate the GDMI from a selection of cancer cell lines to determine whether a strong correlation existed between these two technologies. We measured the global DNA methylation in four cancer cell lines (SW48, LoVo, HT-29 and H69), which varied in methyl content from high to low as determined by HPCE. HpaII and MspI were utilized on 100 ng of cell line genomic DNA to calculate the GDMI. The HPCE results, extracted from the publication by Paz et al [20] were compared with the calculated GDMIs from the four cell lines. The CpGlobal results presented with a good linear fit when compared to the HPCE data. In addition, we observed that the number of unmethylated HpaII sites had an inverse linear relationship with the total number of methylated cytosines (data not shown).

The linear fit for this correlation was calculated:

HPCE = -6.2 × GDMI + 7.18

Adjusted R² = 0.77

Residual error = 0.36.

We directed the assay toward measuring the association of global DNA methylation with the natural history of lung cancer. Lung cancer is measured in 5 stages, which range from 0 to IV [21]. We measured the GDMI, using the HpaII methyl-sensitive restriction enzyme, in epithelial cell lines that represented the full extent of the disease: Normal lung, Stage I, Stage II, Stage IIIa, Stage IIIb and Stage IV metastatic liver. The results indicated that there was an increase in global DNA hypomethylation that was associated with the progressive development of the tumor (Figure 3). The greatest loss of methylation was observed in the Stage IIIb cell line and the least was measured in the Stage I. Compared with the normal epithelial cell line, this represented a loss of 5-methylcytosines that ranged from 10%, in an early stage tumor, to 50% in a late stage cancer. These data demonstrated the quantitative capacity of CpGlobal to measure changes in global DNA methylation in a cancer as it progresses from early to late stage and metastasis.

Bild vergrößern

We measured the GDMI using CpGlobal in lung DNA from 20 patients diagnosed with Non Small Cell Lung Cancer (NSCLC) (Table 2) and 12 normal individuals to ascertain whether global DNA hypomethylation is a prominent phenomenon in this disease. DNA from three types of lung tissues was analyzed: Normal Non-Disease from normal individuals, Normal Disease (pathological and histological defined normal) paired with Tumor from each cancer patient. The GDMI was measured in the lung DNAs using the methyl-sensitive restriction enzymes HinP1I, HpaII and HypCH4IV. A general trend was observed when the median GDMI was calculated for all three methyl-sensitive enzymes (Table 3a); an increase in the median GDMI was measured with lowest levels detected in the Normal Non-Diseased and the highest in the Tumors (median GDMI – Normal Non-Disease < Normal Disease < Tumor). Nonparametric Wilcoxon/Kruskal-Wallis test was performed on the median GDMI data to determine whether the observed trend was statistically significant (Table 3b). The greatest difference in GDMI was between the Normal Non-Disease and Tumor (p < 0.054–0.004), which was observed in all three methyl-sensitive enzymes. However, only HinP1I and HpaII were able to detect a GDMI difference between the paired Normal Disease and Tumor (p = 0.064 and p = 0.068 respectively). Out of the three methyl-sensitive enzymes only HpaII was observed to distinguish between the Normal Non-Disease and Normal Disease with statistical significance (p = 0.039). These results suggested that the histological and pathological defined normal part of the lung from a cancer patient had altered global DNA methylation levels.

To investigate further the changes in global DNA methylation in lung cancer we separated the major histopathologic characteristics assigned to the cancer patient samples into three groups (Stage – 1A and 1B, grade of differentiation – well and non-well and tumor size – T1 and T2) and searched for any association of the GDMI with these characteristics in the Normal Disease and Tumor samples. Nonparametric Wilcoxon/Kruskal-Wallis test was performed using the median GDMI data, which was measured in the Normal Disease and Tumor samples. No statistically significant changes were observed in global DNA methylation from the Normal Disease tissue samples for any methyl-sensitive restriction enzyme when compared against the three groups. However, the median GDMI from the Tumor tissue samples was significantly associated with all three histopathologic characteristics. In particular this was observed for HinP1I (Stage 1A versus 1B), HypCH4IV and HpaII (well versus non-well differentiated) and HinP1I and HpaII (T1 versus T2) (Table 4). When the analysis was performed using the ratio of median GDMI Tumor/median GDMI Normal Disease the statistical significance was improved for some methyl-sensitive restriction enzymes and made worse for others when compared against the three histopathologic characteristics (Table 4). These analyses indicated that global DNA hypomethylation remained a persistent observation in lung tissue from patients diagnosed with NSCLC. Furthermore, CpGlobal could clearly distinguish a difference in the Tumor samples between Stage 1A and 1B, tumor size T1 and T2, and well differentiated and non-well differentiated tissue.