In an effort to characterize the response of colon cancer cells to HDACi, we analyzed the gene expression profile of two colon cancer cell lines following treatment with two HDACi, vorinostat and LBH589. Both HDACi resulted in significant inhibition of tumor cell proliferation, an accumulation of acetylated histones and the onset of apoptotic cell death. However, LBH589 exerted antiproliferative effects at significantly lower concentrations than vorinostat, consistent with previous reports utilizing these HDACi [25, 34, 41]. Specifically, the IC50 for LBH589 was in the single digit nanomolar range while vorinostat required concentrations in excess of 1 μM. The concentrations at which HDACi induce their antiproliferative effects is of importance particularly in the context of clinically achievable serum concentrations and the extrapolation of in vitro observations to clinical settings. Pharmacokinetic data from clinical trials following a twice-daily dose of vorinostat determined that the half-life was in the range of 1 - 3.5 h and maximal serum concentrations did not peak over 2 μM and rapidly diminished [27, 42, 43]. Of note, the half-life of LBH589 was determined to be in the order of 10-14 h and serum concentrations of 400 - 700 nM are achievable at doses which are well tolerated . Therefore, the concentration of LBH589 required to achieve 50% growth inhibition in our colon cancer cells was well within clinically achievable concentrations whereas the concentration of vorinostat was within, but approaching the upper limit of reported serum concentration ranges.
The cDNA microarray analysis demonstrated that in each cell line that the gene expression profile was significantly altered after a 24 h exposure to either HDAC inhibitor, vorinostat (2 μM) or LBH589 (50 nM). Considering the mechanism of action of HDACi including histone acetylation-induced chromatin remodeling and the acetylation of non-histone proteins including transcription factors, it is intriguing that only 5-7% of genes in the colon cancer cell lines analyzed were modulated by HDACi treatment. However, our results are consistent with other microarray profiling experiments which reported as few as 2% and as high as 10% modulated by HDACi. These reports and the data presented herein would indicate that HDACi do not induce global gene expression changes and may instead target specific sets of genes. An important observation in this study was that vorinostat and LBH589 induced very similar transcriptional profiles within each cell line. As both of these agents are hydroxamate-class HDACi, this observation is somewhat expected. Additional studies have identified very similar transcriptional changes produced by the two hydroxamic-acid based HDACi, TSA and vorinostat, while also demonstrating a different gene expression profile obtained with the benzamide class HDACi MS-275 .
The analysis of our data demonstrates that HDACi induce significant cell-line specific effects on genes involved in the regulation of a number of critical tumor processes including angiogenesis, mitosis, DNA replication, recombination and repair and apoptosis. More specifically, the potent anti-angiogenic matrix glycoprotein thrombospondin-1 (THBS1), was significantly upregulated 14-fold in HCT116 cells at 24 h (Table 1). HT29 cells however, showed no modulation until 24 h post-treatment where only a modest increase of 2-fold was observed by qPCR. HDACi are reported to be potent inhibitors of tumor angiogenesis and induction of THBS1 has previously been reported following HDAC inhibition . Similarly, in HCT116 the most heavily upregulated gene following treatment with either vorinostat or LBH589 was connective tissue growth factor (CTGF; Table 1). CTGF is a multifunctional secreted matricellular protein associated with fibrotic disorders, angiogenic regulation, and possibly tumor development . Human tumors overexpressing CTGF demonstrated reduced microvessel density indicative of potential antiangiogenic properties, and ovarian tumors overexpressing CTGF demonstrated enhanced tumor cell invasion . In HT29 cells, fibroblast growth factor 19 (FGF19) was significantly downregulated by both HDACi (Table 2). FGF19 binds to FGF receptor 4 and has been shown to mediate cell cycle progression, angiogenesis and promote tumor growth through the beta-catenin pathway. Knockdown of FGF19 in colon cancer cells decreased tumor growth in vitro and in vivo . It is possible that the antiangiogenic and antitumor action of HDACi are mediated, in part, through modulation of key angiogenic regulators such as these and would indicate that HDACi may potentiate the therapeutic efficacy when used in combination with inhibitors of tumor angiogenesis.
In HT29 cells, microarray analysis identified that both HDACi induce a potent downregulation of the anti-apoptotic caspase inhibitor protein AVEN. qPCR confirmed that AVEN is significantly downregulated in HT29 cells by vorinostat and LBH589 >5-fold at 24 h and only modestly regulated in HCT116 cells <2-fold at 24 h. AVEN is reported to inhibit caspase activation through inhibition of APAF-1 self-association . The downregulation of AVEN would suggest that HDACi-induced apoptosis in the HT29 cells may be regulated in part via the mitochondria, supporting the mechanism of oxidative stress injury as previously reported .
We also observed significant cell-line-specific alterations in genes involved in mitosis. Aurora kinase B was identified as downregulated by both vorinostat and LBH589 in HCT116 cells. The Aurora kinase family are critical regulators of mitotic cell division having roles in centrosome function, mitotic spindle formation, chromosome segregation and cytokinesis . Overexpression of Aurora kinases A and B have been linked to genetic instability and are frequently overexpressed in solid tumors such as colorectal cancer [50, 51] and inhibition of aurora kinases has become an attractive therapeutic strategy with multiple inhibitors in clinical development. Of note, recent studies have reported that LBH589 induces the degradation of aurora kinase A and B in renal and non-small cell lung cancer cells resulting in G2/M arrest and apoptosis [52, 53]. Interestingly, we observed downregulation of aurora kinase B with HDACi treatment only in the HCT116 cells where a potent G2/M arrest and significant apoptosis was observed (Table 1, Figure 5B).
Approximately 18% of the DEGs identified after HDACi treatment were modulated in a similar manner in both cell lines. This core set of genes encompass genes involved in cell cycle, nucleotide metabolism, nucleosome assembly and apoptosis. We identified a panel of 11 genes, 6 up- and 5 downregulated by both HDACi in both cell lines. Previously, Glaser et al. identified a core set of 13 genes regulated by three HDACis in bladder and breast cancer carcinoma cell line models. Upon comparison, one upregulated gene (histone H2B) and one down-regulated gene (thymidylate synthase) are consistent between our core gene set and those reported by Glaser et al. . One of the primary reasons for this is that out core gene set was defined solely from colon cancer cells which are physiologically distinct from both bladder and breast cancers and may employ different mechanisms of gene expression regulation. An additional study analyzed the effects of HDACi in renal cancer cells and identified consistent directional modulation of short-chain alcohol dehydrogenase, aldo-keto reductase and fibroblast growth factor gene families .
Two genes within our core set of HDACi-modulated genes are directly involved in nucleotide metabolism and DNA repair. Downregulation of both thymidylate synthase (TYMS) and UNG was observed in both cell lines following treatment with either HDACi. Thymidylate synthase is essential for the de novo synthesis of thymidylate, an essential precursor required for DNA replication and repair. UNG is the gene encoding uracil-DNA glycosylase, a base excision repair protein involved in uracil excision from DNA. Both these enzymes are reported to mediate response to the antimetabolite class of chemotherapeutic agents including inhibitors of TS such as 5-fluorouracil [55, 56]. A number of other studies have confirmed that downregulation of TS mRNA and protein is a common event in response to HDACi treatment [14, 25, 57]. We recently confirmed that downregulation of TS was a common event in an extended panel of colon cell lines and was driven primarily through a transcriptional mechanism in response to HDAC inhibition. This interaction resulted in synergistic antiproliferative effects between HDACi and 5-FU in colon cancer cells  supporting the concept that HDACi-mediated alterations in known drug targets may provide opportunity for new therapeutic combinations.
Short-chain alcohol dehydrogenase family member 2 (DHRS2) was identified as the most heavily induced gene by HDACi in our core set of genes. DHRS2 was originally identified following its upregulation by treatment with butyrate and was later confirmed to be involved in the differentiation of monocytes to dendritic cells [58, 59]. HDACi treatment is reported to induce cellular differentiation and induction of pro-differentiation genes such as DHRS2 is a plausible mechanism .
MT1X and MT1G were both heavily induced in both cell lines by HDACi treatment. These genes encode two highly inducible ubiquitous proteins belonging to a family of cysteine-rich metallothionein proteins. Metallothioneins can bind to both physiological and xenobiotic heavy metals . Previous studies have identified regulation of other metallothionein family members in response to HDACi . MT1G is reported to be a tumor suppressor gene and is frequently epigenetically silenced in a number of human malignancies [62, 63]. Although the mechanism that results in the induction of metallothionein proteins is unknown, both the MT1X and MT1G genes map to chromosome 16q13 and it is likely that HDACi-mediated events in this region such as chromatin relaxation result in the increased transcription of both of these genes.
NF-κB regulates the expression of a significant number of genes involved in immune response, angiogenesis, cell adhesion, proliferation, differentiation, and apoptosis [64, 65]. The NFKB1 gene encodes the predominant p50/p105 form and represents one of the core genes significantly downregulated by HDACi treatment in this study. As such, many different types of human tumors have dysregulated NF-κB, primarily via constitutive activation that mediates continued cell proliferation and averts the onset of apoptosis . Downregulation of NF-κB is a likely mechanism by which HDACi induce aspects of their apoptotic effects in colon cancer cells. We also identified the IL-1 receptor associated kinase (IRAK1) as consistently downregulated by HDACi in our core set of genes. IRAK1 encodes the interleukin-1 receptor-associated kinase 1 which is reported to be partially responsible for IL1-induced upregulation of NF-κB  and was one of ~100 genes identified as consistently upregulated in a microarray meta-comparison of genes upregulated in solid tumors of epithelial origin .
Our core set of genes includes the histone family member HIST1H2BD which encodes the histone H2B protein and was >3-fold induced by HDACi treatment. HIST1H2BD has previously been reported to be induced by HDACi treatment . While the mechanism of induction of this gene is unknown, it is located within the large histone gene cluster on chromosome 6p22-p21.3 and it is likely that the HDACi-induced alterations in this region, possibly through chromatin relaxation allowing transcriptional machinery access, results in this induction.
We have analyzed the gene expression profiles of two of the most clinically advanced hydroxamate class HDACi, vorinostat and LBH589, in two colon cancer cell line models. We identified significant overlap in differentially expressed gene profiles for vorinostat and LBH589 within each cell line indicating similar mechanism of action for these HDACi. Interestingly, we also identified a strong cell-line dependence of gene expression changes induced by these HDACi with only 18% commonality in HDACi-induced DEGs. Within this gene expression overlap, we identified a core set of 6 up- and 5 downregulated genes that are regulated by both of HDACi in both cell lines. Defining a core set of genes that represent markers of HDAC inhibition is an important first step in the identification and validation of clinical markers for evaluating HDACi target inhibition and efficacy. Currently, analysis of histone acetylation from tumor tissue and more frequently from isolated peripheral blood mononuclear cells is used as evidence of HDACi biological activity. However, histone acetylation following HDACi treatment has been shown to be highly reversible and often inconsistent. A panel of HDACi-regulated genes may provide a more sensitive and reliable means to determining the efficacy of HDACi treatment in the clinic. We also identified alterations in additional pathways which may enhance the therapeutic potential of both conventional and targeted therapeutics, including genes involved in angiogenesis, nucleotide metabolism and mitosis. As HDACi advance in clinical development, these agents are likely to be incorporated into combination treatment strategies with both conventional and novel chemotherapeutic agents. Therefore, the identification of pathways and drug targets modulated by HDAC inhibition could be critically important in elucidating their disease-specific mechanism of action and assisting in the identification of effective drug combination partners.