Background
Proper glycosylation of the dense layer of glycoconjugates on the cell surface is necessary for effective communication between cells and the environment. Aberrant glycosylation in cancer is associated with disrupted cell signaling, evasion of growth suppressors, immunoresistance, increased angiogenesis, and induction of invasion and metastasis [
1‐
3]. Furthermore, both defective and increased levels of glycosylation have been shown to contribute to multidrug resistance in various cancers [
4,
5]. As such, a better understanding of the cancer glycome and glycoproteome will be instrumental to the identification of more definitive cancer biomarkers and development of more effective cancer therapeutics.
Ultimately, the versatility of glycan structures help cancer cells respond, adapt, migrate, and invade. Although there is no genetic template for glycan structure, the type of glycosylation added to any particular protein is unequivocally determined by consensus sequences in the gene encoding for it. The structure of the glycan added then depends on expression of the enzymes required to synthesize the carbohydrate glycan structures. Thus, glycome regulation is complex, and involves more than 600 so-called “glycogenes” which encode for the various proteoglycans, glycosyltransferases, glycosidases, sulfatases, and carbohydrate biosynthetic enzymes that determine glycome function. Evidence suggests that individual glycome composition is largely genetically predetermined, but recent advances in glycobiology have identified disease- and cancer-specific glycophenotypes which are under epigenetic control [
6‐
8]. This is intriguing because unlike genetic mutations, epigenetic modifications are transient and reversible, making them extremely desirable therapeutic targets.
HDAC inhibitors are powerful epigenetic regulators and promising cancer therapeutics, but in solid tumors, these drugs are not effective as single agents, and instead are often use in combination with a chemotherapeutic. Furthermore, the broad-spectrum and often pleiotropic effects off HDAC inhibitors make it difficult to harness their full potential. Indeed, while growth rate is slowed by HDAC inhibitor treatment in nearly all cancer cells, many also increase oncogenic properties such as increased motility and invasive capacity [
9]. Although other epigenetic regulators, such as the DNA methyltransferase inhibitor 5-Aza-dc, have been used extensively in multiple cancer types to effectively target glycogenes and glycosylation enzymes to slow tumor progression [
10‐
12], the effects of HDAC inhibitors on glycosylation, oncogenic phenotype, and chemotherapeutic response have not been extensively investigated. At a broader level, even though epigenetic regulation determines many aspects of oncogenesis and normal cellular behavior, these epigenetic processes and their regulation are not well understood.
In the study of epigenetic regulation of the glycome, the human adrenal cortical carcinoma SW13 cell line represents an invaluable research model as it exists as two distinct phenotypes which are epigenetically plastic [
13,
14]. In this study, we utilize the SW13 cell line as a model of epigenetic phenotype regulation to investigate how HDAC inhibitors can influence the acquisition of distinct oncogenic characteristics, such as rapid proliferation and metastatic capacity. Data presented here suggest that HDAC inhibition significantly impacts the SW13 glycome. Specifically, we demonstrate differential glycogene expression in multiple biosynthetic pathways and document changes in glycome composition. We then go on to characterize the glycosylation signatures and glycan binding preferences of tumorigenic versus metastatic SW13 cells. Finally, we demonstrate that the changes induced by HDAC inhibitor treatment observed in the SW13 glycome are associated with decreased chemosensitivity. This work furthers our understanding of how HDAC inhibition might influence malignant transformation in cancer cell types that respond negatively to HDAC inhibitor treatment. This study also identifies candidate biomarkers for determining responsiveness to HDAC inhibition and chemotherapeutic treatments.
Methods
SW13 cell culture and treatment
SW13 cells were obtained from American Type Culture Collection (ATCC; CCL-105) and maintained in high glucose DMEM supplemented with 10% fetal bovine serum and 10 U/ml penicillin/streptomycin at 37 °C in a humidity controlled incubator. For HDAC inhibition, cells were treated with 1 nM romidepsin (FK228) (Selleckchem) for 24 h.
Morphology, proliferation, and MMP activity assays
Morphology was assessed via immunofluorescence using standard techniques. Specifically, SW13 control cells, or SW13 cells that had been treated with 1 nM FK228 were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton-X, and blocked with 1% BSA before a 1 h incubation with 25 U/ml Alexafluor 488 phalloidin (Invitrogen). Samples were mounted with ProLong Gold anti-fade reagent with DAPI to stain nuclei and photographed using a Zeiss Axiovert apotome with a 40X objective lens and a uniform exposure at each wavelength.
SW13 cell proliferation with and without 1 nM FK228 treatment was measured using a Click-iT EdU assay (ThermoFisher) according to the manufacturer’s instructions. Cells were imaged using a Zeiss Axiovert Apotome with a 10X objective lens and uniform exposure at each wavelength. Percent proliferation was quantified using NIH ImageJ software to divide the number of cells which stained positive for EdU by the total number of cells in each groups.
MMP activity was measured by plating SW13 cells at 1 × 104 cells/well in 8-chamber slides. After 24 h of no treatment (control) or treatment with 1 nM FK228 cells were incubated overnight with a DQ gel substrate (Life Technologies) diluted to 40 μg/ml in MMP activity buffer (100 mM NaCl, 100 mM Tris-HCl, pH 7.5, 10 mM CaCl2, 20 μM ZnCl, 0.05% NP40, 0.2 mM sodium azide). Cells were then washed 2 times with 1X PBS before fixation and DAPI staining. Relative differences in MMP activity were quantified using NIH ImageJ software to divide the relative amount of green fluorescence by the number of cells in a given well. All experiments were performed at least in triplicate.
Gene expression analysis of SW13 cells following HDAC inhibitor treatment
Total RNA was isolated from cells using Trizol according to the manufacturer’s recommendations. Following quality control analysis (260/230 > 1.5, 260/280 > 1.8, RIN > 6), whole genome microarray analysis was performed by PhalanxBiotech using their OneArray platform. Standard selection criteria to identify differentially expressed genes were established as log2 fold change ≥1 and p < 0.05. For advanced data and pathway analysis, intensity data were pooled and calculated to identify differentially expressed genes based on the threshold of fold change and p-value. A gene set enrichment analysis of Gene Ontology (GO) terms was then performed using the differentially expressed gene lists as input. A small subset of differentially expressed genes identified in the array were selected for validation by qRT-PCR. Briefly, the same total RNA described above was reverse transcribed using SuperScript II, and qRT-PCR was performed using 50 ng cDNA per reaction on an ABI StepOne Plus thermocycler using SYBR Green chemistry. Expression changes were analyzed via the ΔΔCT Method following normalization to GAPDH expression, which was utilized as an invariant control.
Lectin and glycan array analysis
Lectin and glycan arrays offer a high-throughput approach for cellular glycoprofiling and the identification of preferential carbohydrate binding moieties, respectively. Thus, to assess the functional consequences of glycogene expression changes in response to HDAC inhibition, total protein from SW13 control cells and SW13 cells that had been treated with 1 nM FK224 for 24 h were subjected to lectin and glycan array analysis (
n = 4 per group). Arrays were performed and analyzed by RayBiotech. The Lectin 40 array is designed to detect glycan profiles using 40 unique lectins, while the Glycan 100 array assess carbohydrate binding preferences using the 100 most frequently identified structures showing important binding function in the literature. Lists of all lectins and glycans utilized in the arrays are available in Additional file
1: Tables S1 and S2.
Lectin binding analysis of carbohydrate expression
Binding of select lectins were validated by lectin blot analysis wherein 25 μg total protein from SW13 control or FK228 treated cells was electrophoresed through a pre-cast 4–20% polyacrylamide gel and then transferred to a PVDF membrane. Membranes were blocked with Pierce SuperBlock® buffer for 30 min, and then incubated with 20 μg/ml FITC labeled lectins (EY Laboratories) for ~ 2 h before being imaged using a ChemiDoc. Alternatively, treated and untreated cells were fixed as described above before incubation with 20 μg/ml FITC labeled lectins for ~ 2 h. Samples were mounted with ProLong Gold anti-fade reagent with DAPI to stain nuclei and photographed using a Zeiss Axiovert apotome with a 40X objective lens and a uniform exposure at each wavelength. To demonstrate specificity of lectin binding, wheat germ agglutinin (WGA) stained cells were incubated with WGA elution buffer for 30 min for mounting and visualization.
Measuring GAG-sulfation
Total GAG-sulfation was assayed using a 1,9-Dimethylmethylene blue (DMB) assay as previously described [
15]. DMB is a cationic dye that specifically binds to sulfated glycosaminoglycans with an absorbance at ~ 525 nm. Briefly, ~ 1 × 10
6 cells were collected in 40 μl PBS or RIPA buffer. For the cells in PBS, 125 μl DMB buffer (31 μM DMB in 55 nM formic acid, pH 3.5) was added and transferred to a clear-walled 96-well plate. DMB binding was quantified by measuring the optical density at 525 nm. Total protein levels were assessed by BCA assay from the cells in RIPA buffer. DMB binding levels were normalized to total protein levels.
Viability assays
Chemotherapeutic sensitivity was measured using an MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay. Cells were plated at 3 × 104 cells per well in 96-well plates and incubated with 1, 10, or 50 nM paclitaxel (Cayman Chemical) for 24 or 48 h. Following treatment, 10 μl of 12 mM MTT was added to each well and cells were placed back in the incubator at 37 °C for 4 h. The formazan was solubilized by adding 100 ul solubilization solution (40% DMSO, 16% SDS, 2% acetic acid) followed by another incubation at 37 °C for 1 h. Spectrometric absorbance at 570 nm was measured with a microplate reader. Each assay included four technical replicates and was repeated at least three times. Paclitaxel effects at each dose, and differences in paclitaxel response between cell lines were analyzed using a two factor mixed design ANOVA using SPSS version 19 software (Chicago, IL). A Bonferroni test was used to adjust for multiple pairwise comparisons. Mean differences were considered statistically significant at p < 0.05.
Discussion
Data presented here suggests that HDAC inhibitor treatment of SW13 cells leads to > 2 fold changes in the expression of 3250 genes, a substantial portion of the genome. In our effort to characterize some of the additional functional consequences of HDAC inhibitor treatment, we documented significant impacts on SW13 cell morphology, growth, invasiveness, and chemotherapeutic response after HDAC inhibition. Importantly, these findings are not unique to the SW13 cell line, as numerous cell types have been shown to exhibit similar dramatic changes in gene expression and phenotypic characteristics with alterations in the glycome [
9,
26,
27]. In addition, multiple HDAC inhibitors (including changes valporic acid, sodium butyrate, apicidin, MS-275, SAHA, and TSA) are able to illicit similar phenotypic changes in multiple cell lines [
9]. Thus, we were confident SW13 cells and their response to FK228 treatment were a relevant model to investigate how HDAC inhibitor treatment contributes to such dichotomous epigenetically and phenotypically distinct oncogenic states.
Though in this study we used an HDAC inhibitor which primarily targets HDAC1, it is notable that multiple HDAC inhibitors have been demonstrated to lead to the characteristic phenotypic switch observed in the SW13 cell type [
9,
13,
14,
16]. Thus, these phenotypic changes are likely not dependent on inhibition of a specific HDAC, but are likely mediated by widespread alterations in the acetylome that subsequently impacts global cellular gene expression. As inhibitors of lysine deacetylation however, HDAC inhibitor treatment affects the activity of many acetylated proteins, including many non-histone targets. Interestingly, only 3.9% of the acetylation sites detected after FK228 treatment of a colon cancer cell line were on core histones [
28]. Recent work has demonstrated that acetylation patterns are tissue specific and have distinct subcellular distribution [
29] and these patterns have been linked to changes metabolism and to regulate a wide variety of biological functions [
28,
30,
31] and more importantly, acetylation of non-histone proteins has been linked to alterations in gene expression [
32,
33]. Amongst the effects of alterations in the acetylome, we have chosen to address the effects of the glycome due to as recent data has suggested that the aberrant glycosylation seen in cancer is epigenetically regulated [
34,
35].
Bioinformatics analyses of the microarray suggests that many glycome-related genes are differentially expressed after HDAC inhibitor treatment and many glycosylation pathways appear to be affected by treatment. Among the affected pathways, we found the prevalence of differentially expressed HSPG-related genes in the microarray data notable due to the link of these gene products with oncogenesis. Based on the data presented here, we suggest that HDAC inhibitor- mediated epigenetic regulation of the glycome, and possibly the HPSGs and their biosynthetic machinery, is an important contributory factor to the prevention and progression of oncogenesis.
Given the broad spectrum of differential expression of genes involved in glycome biosynthesis, alterations in lectin binding might be expected and lectin array analysis does indeed demonstrate HDAC inhibition significantly impacts the glycosylation signature of SW13 cell. We found it interesting that the profile of glycosylated proteins recognized by several lectins was also altered. This suggests alterations both in the amount and type of glycosylation, and that these glycan structures are attached to proteins of different molecular weights. Of particular interest is the increase in labeling by the sialic acid binding lectin LPA as increase in this glycosylation is associated with more aggressive tumors [
26,
27]. In addition, the fact that this differential binding is not seen when cells are permeabilized underscores need to carefully consider sample preparation when comparing different lectin assays. Specifically, cellular glycosylation is most closely associated with membrane lipids and proteins and these glycans are susceptible to extraction with detergent permeabilization. In addition, it is clear that some cytosolic and nuclear proteins are glycosylated and can be bound by lectins (e.g. WGA and UEA) [
36,
37] and these proteins may be enriched relative to other proteins due to sample preparation methods.
Importantly, changes in lectin binding corresponded with changes in mRNA expression identified in the microarray, and demonstrate the functional impact of HDAC inhibitor-mediated glycogene expression. For example, expression of the sialylatransferases ST6Gal2 and ST3GAL5 were significantly increased in HDAC inhibitor treated SW13 cells. Sialyltransferases function by linking sialic acids to terminal GalNAc and GlcNAc sugar residues on glycoproteins, and play important roles in cancer progression [
38]. Decreased ECA and GS-II binding in HDAC inhibitor treated SW13 cells indicates increased masking of GlcNAc residues, consistent with increased sialyltransferase activity. In-line with our results here, this pattern of activity has also been associated with increased invasive capacity [
27].
Similarly, decreased binding to ECA, as well ConA, has also been observed in mammary gland epithelial cells going through TGFβ induced epithelial to mesenchymal transition, which is also associated with cancer metastasis [
39]. Moreover, these changes in lectin binding were also associated with similar changes in N-glycan-related gene expression changes observed in our study as well. Thus, it is interesting to note decreased binding to both of those lectins in response to HDAC inhibitor treatment. VFA binding was also decreased in the HDAC inhibitor treated cells. Intriguingly, treatment with VFA has been demonstrated to decrease the malignant phenotype in other cell lines [
40]. Consequently, the identification of different patterns glycogene expression and lectin binding properties of HDAC inhibitor treated cells may play an important role in the development of innovative combinatorial approaches to cancer therapies.
Biological activity of the glycome is also dependent on the recognition of glycans by glycan binding proteins. SW13 cells exhibited binding to a broad number of glycan binding proteins in each of the eight classes present in the array. FK228 treatment significantly impacted the binding affinity for a number of glycans, but of particular interest was the 2-fold increase in α-Rha binding. Human serum contains an abundance of anti-carbohydrate antibodies, and anti-Rha antibodies are among the most abundant [
41]. Further, rhamnose glycoconjugates have been demonstrated to target anti-carbohydrate antibodies specifically to tumor cells prompting rigorous investigations into their use for cancer immunotherapies, with very promising results [
41‐
43]. Our findings that HDAC inhibitor treatment further increases the binding affinity for α-Rha raises the exciting possibility to use HDAC inhibitors as enhancers of cancer immunotherapy efficacy.
As mentioned above, HDAC inhibitors have largely proved ineffective as monotherapies for solid tumors, but have had some limited demonstrated efficacy in combination with other agents such as carboplatin and paclitaxel [
44,
45]. In contrast, we observed decreased responsiveness to paclitaxel treatment following HDAC inhibition. It is worth noting that alterations in glycogene expression and cellular glycosylation patterns can significantly impact chemosensitivity [
2,
4,
5]. Indeed, increased expression of ST8SIA4 in human leukemia and decreased expression of B4GALT2 in breast cancer cells has been associated with multidrug resistance [
4,
5]. In SW13 cells, we noted both an increase in ST8SIA4 expression and a decrease in B4GALT2 expression in response to HDAC inhibition, which could help explain their decreased sensitivity to paclitaxel treatment. Findings from these studies underscore the importance of glycomic alterations in the progression of cancer and their utility in identification of effective chemotherapeutics. The utility of B4GALT2 and ST8SIA4 as prognostic indicators for HDAC inhibitor treatment effectiveness and chemotherapeutic sensitivity should be further investigated.