Anti-CS-A mAb 2H6 was from Associates of Cape Cod/Seikagaku America (Falmouth, MA, USA). Fluorescence-conjugated, anti-mouse IgM and 5-aza-2′-deoxycytidine (5AzadC) were from Sigma (St. Louis, MO, USA). Primers were from Integrated DNA Technologies (IDT, Coralville, IA, USA). Real-time PCR reagents were from Applied Biosystems (Foster City, CA, USA). TRIzol reagent was from Invitrogen (Carlsbad, CA, USA), FailSafe PCR PreMix Selection kits and Enzyme Mix were from Epicentre Biotechnologies (Madison, WI, USA). Alkaline phosphatase (rAPid) was from Roche (Nutley, NJ, USA).

Publicly available Oncomine cancer microarray database (Compendia Biosciences; Ann Arbor, MI, USA; www.oncomine.com) was used to examine the expression of CHST11 in cancer tissue and determine association of CHST11 expression with breast cancer outcomes. Richardson et al (GEO accession GSE3744) (26), Finak et al (GEO accession GSE9014) (27), The Cancer Genome Atlas (TCGA, http://tcga-data.nci.nih.gov/tcga/) invasive breast carcinoma gene expression data and Gluck et al (GEO accession GSE22358) (28) dataset were used to compare CHST11 expression levels between cancer and normal tissues. Hoeflich et al (GEO accession GSE12777) (29) dataset was used to compare expression levels of CHST11 among a panel of cell lines that represent luminal, Her-2-amplified and basal-like molecular subtypes of breast cancer. Schuetz et al dataset (GEO accession GSE3893) (30) was used to compare CHST11 expression levels between matched ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC). Log-transformed, median-centered and normalized expression values (31) were extracted, analyzed and graphed accordingly.

Cell lines MCF7, MDA-MB-231, MDA-MB-468, T47-D, and ZR-75-1 were from ATCC (Manassas, VA, USA). MDA-MB-231, MDA-MB-468, T47-D, and ZR-75-1 cells were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies), 50 U/ml penicillin and 50 μg/ml streptomycin. MCF7 were grown as described before (14). Cells are checked every six months to be free from Mycoplasma contamination using the MycoAlert^® Mycoplasma Detection kit (Lonza Rockland Inc., Rockland, ME, USA).

De-identified paraffin-embedded specimens from 5 female breast cancer patients were provided by the Department of Pathology of the University of Arkansas for Medical Sciences (UAMS). For this study, an active human tissue use protocol approved by the UAMS Institutional Review Board was used.

RNA isolation and quantitative real-time was performed as described earlier (14). Briefly, total RNA was isolated from cultured cells using TRIzol reagent (Life Technologies, Grand Island, NY, USA) following the manufacturer’s instructions. The quantity and quality of the isolated RNA was determined using an Agilent 2100 Bioanalyzer (Palo Alto, CA, USA). Total RNA (1 μg) was reverse-transcribed using random-hexamer primers with TaqMan Reverse Transcription reagents (Applied Biosystems). Reverse-transcribed RNA was amplified with SYBR Green PCR Master Mix (Applied Biosystems) plus 0.3 μM of gene-specific upstream and downstream primers during 40 cycles on an Applied Biosystems 7900 HT Fast Real-time system. Data were analyzed by absolute and relative quantification. In absolute quantification, data were expressed in relation to 18S RNA, where the standard curves were generated using pooled RNA from the samples assayed. In relative quantification, the 2^−ΔΔC_T method was used to assess the target transcript in a treatment group to that of untreated control using expression of an internal control (reference gene) to normalize data (32). Expression of GAPDH was used as internal control. The primer sequences are shown in Table I.

5AzadC was dissolved in ice-cold phosphate-buffered saline, filter sterilized at 4°C and the resulting solution used to treat the MCF7 cell line. For dose–response experiments, cells were harvested 5 days after the initial treatment. Cell growth medium was refreshed every other day.

DNA from cells was extracted as described before (33). DNA from the paraffin-embedded tissues of breast cancer patients was extracted using Ex-Wax DNA extraction kit (Chemicon International, Temecula, CA, USA), following the manufacturer’s instructions with the additional steps of extracting recovered DNA with phenol (Amresco, Solon, OH, USA) and then 1-bromo-3-chloropropane (Molecular Research Center Inc., Cincinnati, OH, USA). DNA was bisulfite modified with an Epitect kit (Qiagen, Valencia, CA, USA) using 300 ng of DNA per reaction. PCR was performed using a FailSafe PCR PreMix Selection kit and FailSafe Enzyme Mix. Each 25-μl PCR reaction included 1.0 μM of each primer, 2.5 U of the FailSafe Enzyme Mix and 12.5 μl of the FailSafe PCR PreMixes A or C. Bisulfite-modified genomic DNA was amplified by semi-nested PCR using two sets of primers for part of intron 1 that is within a CpG island spanning exon 1 of the CHST11 gene (Genbank NM_000012 and exon 1 located with NM_018413). The same amplification profile was used for both reactions of the semi-nested PCR: 1 cycle at 80°C for 1 min, 1 cycle at 94°C for 1 min; 1 cycle at 95°C for 1 min, 54°C for 1 min, 72°C for 1 min; 1 cycle at 95°C for 1 min, 53°C for 1 min, 72°C for 1 min; 1 cycle at 95°C for 1 min, 52°C for 1 min, 72°C for 1 min; 1 cycle at 95°C for 1 min, 51°C for 1 min, 72°C for 1 min; 36 cycles at 95°C for 1 min, 50°C for 1 min, 72°C for 1 min; 72°C for 5 min and cooling to 4°C.

A forward, outside primer and reverse primer were used for CHST11 for the first reaction (Table I). A second, semi-nested, PCR was then performed on 1 μl of the amplificate (in a 25-μl PCR reaction) using a forward nested primer and the reverse primer from the first reaction (346-bp PCR product, Table I). The primers were designed using MethPrimer web software (34) (http://www.urogene.org/methprimer/). The CpG island was defined using CpG Island Searcher set on the default criteria for defining CpG islands (35) (http://cpgislands.usc.edu/cpg.aspx).

BGS was performed as described before (33) with the following minor modifications. We used rAPid alkaline phosphatase and PCR products were sequenced using the nested forward (upstream) CHST11 primer. We also analyzed DNA methylation in the region spanning >1.5 kb of the CHST11 CpG island with reduced representation bisulfite sequencing [RRBS (36,37)] data generated on breast cancer cell lines for the Illumina Idea Challenge (http://www.illumina.com/landing/idea) (25). RRBS selects DNA fragments ≤220 bp in the vicinity of MspI recognition sites (C.CGG) (36). We only considered CpG dinucleotides with coverage >10x. The percentage of methylation was inferred from the number of methylated reads divided by the sum of methylated and unmethylated reads. All reads were mapped to the hg18 release of the human genome (http://genome.ucsc.edu/cgi-bin/hgGateway?db=hg18). Second generation sequencing DNA methylation results were plotted in Matlab (http://www.mathworks.com).

One-way ANOVA with Tukey’s post hoc procedure test was used to compare gene expression between subtypes of cell lines. For comparison of gene expression data generated by real-time PCR, the raw amount for each mRNA was log transformed and normalized to the control mRNA (18S) amount, and analyzed via one-way ANOVA with Tukey’s post hoc procedure. The Mann-Whitney U test was performed to compare gene expression between DCIS and IDC. For 5AzadC induced fold change in gene expression, the mRNA levels of the non-zero doses for each transcript and experimental replication were normalized to that of the zero-dose control, transformed to their base-10 logarithms and analyzed for trend with dose via one-way ANOVA. Statistical analyses were performed using Excel (Microsoft, Seattle, WA, USA) or GraphPad Prism version 5.00 for Windows (GraphPad Software, San Diego, CA, USA). All P-values were 2-sided.

We have shown that the expression of CHST11 is elevated in tumor cells compared to normal cells of breast cancer patient tissue specimens (14). We used Oncomine database to confirm our original data. The data from several studies were analyzed to compare CHST11 expression in breast carcinoma versus normal breast tissue (Table II). The comparison showed average increases of 2–3.6-fold in CHST11 expression in breast cancer compared to normal tissue.

Our previously published data analyzing CHST11 expression in a set of commonly used human breast cancer cell lines suggest that CHST11 expression is low in luminal and high in basal-like cell lines (14). The data relate CHST11 expression to basal-like cancer cells. To further investigate such an association, we analyzed an Oncomine dataset that screened 50 breast cancer cell lines representative of the molecular subtypes luminal, Her2-amplified, and basal-like (29). We observed that the expression of CHST11 in basal-like cell lines was significantly higher than in luminal cell types (Fig. 1a). CHST11 was also significantly overexpressed in Her-2-amplified cell lines. Considering cells with above average expression as positive, 75% of basal-like and 56% Her-2-amplified cell lines were positive; while only 15% of luminal cell lines were positive for CHST11 expression.

To investigate the association of CHST11 overexpression with tumor progression, we interrogated another dataset in the Oncomine database, comparing the expression of CHST11 between ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) (30). DCIS and IDC samples were isolated by laser capture microdissection (LCM) and matched for each individual specimen (30). We found a significant increase in the expression of CHST11 in IDC compared to DCIS samples (Fig. 1b). Notably, both groups displayed a similar distribution pattern for ER status (5 ER-positive specimens in each group) and HER2-amplification pattern (3 HER2-amplified specimens in each group) with histological grades 2 and 3. The data link the elevated expression of CHST11 to aggressive subtypes and progression of the disease.

Our current and previously published data suggest that CHST11 expression plays a role in tumor progression and metastasis. Therefore understanding the mechanisms controlling CHST11 gene expression will help formulate strategies for development of biomarkers and drug targets. We demonstrated previously that CHST11 and CS-A expression was high in highly metastatic MDA-MB-231 cells, while it was significantly lower in MCF7 cells as assayed by qRT-PCR and flow cytometry (14). To examine whether DNA methylation controls variation and regulation of CHST11 in breast cancer cell lines, we examined DNA methylation of the CHST11 sequence in a section of its CpG island covering part of its promoter, its first exon, and a portion of its first intron (Fig. 2a). Using BGS and the Mquant algorithm (33) we determined methylation levels of ten CpG sites for MCF7 and MDA-MB-231 cells (Fig. 2b). This revealed very high methylation levels (91%) averaged >10 CpGs in the CHST11 sequence in MCF7 cells, and very low levels (5%) over the same 10 CpGs in MDA-MB-231 cells (Fig. 2c). These observations suggest that low expression of CHST11 in MCF7 cells is due to DNA hypermethylation.

To further validate our data and confirm that low expression of CHST11 in the MCF7 cell line is due to hypermethylation status of the CHST11 CpG island, MCF7 cells were treated with 5AzadC and CHST11 gene expression was determined. Upon 5AzadC treatment, we observed a significant increase in the expression of CHST11 mRNA (Fig. 3a) that paralleled surface expression of CS-A as assayed by flow cytometry using anti-CS-A mAb 2H6 (Fig. 3b). Together, the data suggest that low expression of CHST11 in the MCF7 cells is due to promoter hypermethylation, and that DNA hypomethylation in the more aggressive mesenchymal-like MDA-MB-231 cell line is permissive for overt CHST11 expression.

To confirm a role for methylation of the CHST11 CpG island in the expression control of CHST11, we further analyzed second generation DNA methylation sequencing data on breast cancer cell lines (Illumina Idea Challenge and ref. 25). This method provided information on DNA methylation at single nucleotide resolution across 162 out of 186 CpGs in the ~2.5 kb long CHST11 CpG island. Methylation analysis of Illumina Idea Challenge data of these cell lines indicates a close relationship between expression levels and methylation status. A very hypomethylated state was observed for MDA-MB-231 (Fig. 4a). The methylation status of two other triple-negative cell lines MDA-MB-468 and BT-20, was similar to MDA-MB-231 (Fig 4b and c). Similar to MCF7 cells, the CpG island of the CHST11 gene in the T47-D and ZR-75-1 cell lines, was hypermethylated (Fig. 4d–f). The expression of CHST11 in T47-D and ZR-75-1 cell lines was further examined by qRT-PCR (Fig. 4g) and flow cytometry (Table III), and compared with CHST11 expression levels in MCF7 and MDA-MB-231. Consistent with the methylation data, the expression of the CHST11 gene and CS-A in both of these cell lines was significantly lower than in MDA-MB-231 and comparable with that of MCF7 cells. MDA-MB-468 displayed a very hypomethylated state but the expression of CHST11 in these cells was moderate and less than what we observed for MDA-MB-231 and MDA-MET (4,14). The expression of CHST11 in tumor tissue may also be controlled by DNA methylation similar to that observed in cell lines.

To determine whether the same CpG island can be methylated differently in actual breast cancer, we examined the methylation status of the CpG island of CHST11 gene in 5 de-identified clinical breast cancer specimens. We observed that in 4 triple negative (TN, i.e., negative for ER, Her2/neu and progesterone receptor) specimens the CpG island was highly hypomethylated, very similar to what was observed for ER-negative basal-like cell lines (Fig. 4h). The CpG island was hypermethylated in one ER-positive specimen tested. Therefore, methylation status of CHST11 CpG might be a useful marker to differentiate between luminal and basal-like breast cancer subtypes, however, a larger sample size is required to confirm this.