Valproic acid causes dose- and time-dependent changes in nuclear structure in prostate cancer cells in vitro and in vivo

Histone deacetylase inhibitors such as valproic acid (VPA) are promising anticancer agents that change the acetylation status of histones and loosen the chromatin structure. We assessed nuclear structure changes induced by VPA in prostate cancer LNCaP, CWR22R, DU145, and PC3 cell lines and xenografts and their potential use as a biomarker of treatment. In vitro tissue microarrays consisted of prostate cancer cell lines treated for 3, 7, or 14 days with 0, 0.6, or 1.2 mmol/L VPA. In vivo tissue microarrays consisted of cores from prostate cancer xenografts from nude mice treated for 30 days with 0.2% or 0.4% VPA in drinking water. Digital images of at least 200 Feulgen DNA-stained nuclei were captured using the Nikon CoolScope and nuclear alterations were measured. With a set of seven most frequently significant nuclear alterations (determined by univariate logistic regression analysis), control and VPA treatment nuclei were compared in vitro and in vivo. Depending on the cell line, area under the curve-receiver operating characteristics ranged between 0.6 and 0.9 and were dose- and time-dependent both in vitro and in vivo. Also, VPA treatment caused significant nuclear alterations in normal drug-filtering organs (liver and kidney tissue). In vitro and in vivo VPA treatment of prostate cancer cell lines results in significant dose- and time-dependent changes in nuclear structure. Further, VPA induces nuclear structural changes in normal liver and kidney tissue, which likely reflects a natural physiologic response. Therefore, nuclear structural alterations may serve as a biomarker for histone deacetylase inhibitor treatment. [Mol Cancer Ther 2009;8(4):802–8]

Acetylation of histones and other nuclear proteins plays an important role in cancer development and progression (1). Deregulation of the balance between histone acetyltransferases and histone deacetylase (HDAC) has been associated with oncogenesis in many tissues (2, 3). HDAC inhibitors (HDACI) are promising anticancer agents that have shown stable disease and partial responses in a wide range of solid malignancies, including prostate cancer (4, 5). Recently, the Food and Drug Administration-approved HDACI vorinostat for the treatment of advanced cutaneous T-cell lymphoma. Clinical trials treating solid tumors with HDACI valproic acid (VPA) alone or combination are ongoing (6–8). Recently, two phase I clinical trials were published using VPA in combination with methyltransferase inhibitor 5-azacytidine or topoisomerase II inhibitor epirubicin (9, 10). These trials showed pharmacokinetic properties and toxicity profiles of VPA in patients with solid malignancies. In addition, stable disease and partial responses were reported in up to 39% and 22% of patients, respectively.

HDACIs change the acetylation status of histones, thereby enhancing the accessibility of the transcription and DNA repair machinery to DNA (11, 12). In addition to changing DNA accessibility by histone acetylation, HDACIs affect transcription, acetylation status, and function of proteins that are important for DNA mobility, DNA structure, and regulation of DNA accessibility (1, 13). Among these proteins are high-mobility group proteins (14–17), nuclear matrix proteins (18), tubulin (19, 20), and members of SWI/SNF complex (21).

Diamond and associates (22) were the first to employ nuclear structure measurements (nuclear roundness factor) by digital imaging to predict prostate cancer outcome. Subsequently, several investigators (23–28) have used nuclear alterations to predict pathologic stage and prognosis for prostate cancer patients. Nuclear morphometric alterations measured by computer-assisted image analysis can detect abnormal DNA content representing large-scale chromosomal alterations (tetraploidy and aneuploidy) reflecting genetic instability in tumor cells (29, 30). In addition, alterations in nuclear size, shape, and texture can be measured. In the current study, we assessed the nuclear structure changes induced by VPA treatment in prostate cancer cell lines and xenografts and their potential use as a biomarker of treatment.

The LNCaP, PC3, DU145, and C4-2 cell lines were grown in RPMI 1640 with l-glutamine (Cellgro) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies), 5 μg/mL ciprofloxacin hydrochloride (U.S. Biological), and 50 μg/mL gentamicin (Quality Biological). Cells were allowed to grow until 80% to 90% confluent and harvested with 0.05% trypsin/0.53 mmol/L EDTA (Cellgro) before each subsequent passage. VPA (1 mol/L; VPA sodium salt; Sigma) stock was made in PBS and filters were sterilized through a 0.22 μm filter. The in vitro tissue microarray consisted of core cell plugs from formalin-fixed, paraffin-embedded human prostate cancer cell lines LNCaP, C4-2, DU145, and PC3. Cells were treated for 3, 7, or 14 days with 0, 0.6, or 1.2 mmol/L VPA and resuspended in agarose before harvesting and paraffin embedding.

Human prostate cancer cell lines LNCaP, PC3, DU145, and C4-2 were allowed to grow to 80% to 90% confluency and harvested as described above. Cells were resuspended in 1× PBS (pH 7.4; BioSource), mixed 1× with Matrigel (BD Biosciences), and injected (1 × 10⁶ per injection) s.c. into the lateral flanks of male athymic nu/nu mice. Once palpable tumors were established, animals were randomized into control and treatment arms with the latter receiving 0.2% or 0.4% VPA in drinking water for 30 days before formalin fixation and paraffin embedding.

Tissue microarrays were prepared using a Beecher MT1 manual arrayer (Beecher Instruments). From formalin-fixed, paraffin-embedded tissue blocks, a spot was punched and transferred to a recipient block. The in vitro tissue microarray was constructed using two or three serial 0.60 mm core tissue samples from each cell plug. The in vivo tissue microarray was constructed using two serial 0.60 mm core tissue samples from each of the 4 to 10 replicate xenografts and 6 μm slides were prepared from each tissue microarray.

Feulgen DNA staining was done per manufacturer’s instructions (TriPath Imaging). Feulgen specifically and quantitatively stains DNA in cellular material by uncovering the free aldehyde groups in the DNA during the acid hydrolysis process, which then reacts with the Feulgen reagent to form a stable, blue color that absorbs light at 560 nm (31, 32). Figure 1 shows Feulgen-stained in vitro and in vivo tissue microarray spots.

Next, a minimum of 200 nuclei was captured from each tissue microarray spot using Nikon CoolScope imaging system with pathology software suite designed by Bacus Laboratories. The software calculates a total of 39 nuclear structure alterations using a single-step pixel analysis of each nucleus to quantitatively characterize nuclear size, shape, DNA content, and chromatin texture features. Tables 1 and 2 shows the number of nuclei captured from each treatment arm.

All data were analyzed using Stata v10.0 statistical analysis software (Stata). Univariate logistic regression analysis was done to determine which independent variables significantly differentiate between VPA and control groups. Two sets of the most common univariately significant nuclear structure alterations, one for in vitro and one for in vivo experiments, were chosen to compare VPA response using area under the curve-receiver operating characteristic (AUC-ROC) across different doses and different durations for in vitro and in vivo experiments. Statistical significance in this study was set as P ≤ 0.05.

To assess nuclear structure changes induced by VPA treatment in LNCaP, C4-2, DU145, and PC3 cell lines, 39 features of nuclear size and shape, texture, and DNA content were measured using Nikon CoolScope imaging system (supplementary text contains definition of all features). Univariate logistic regression analysis with a variable selection stringency of P_z ≤ 0.05 was done to identify significant alterations in nuclear structure after treatment with VPA for each cell line at each given dose and duration of treatment (Supplementary Table S1).⁴

A set of the seven most frequently significant nuclear structural alterations was selected to compare VPA drug response across all prostate cancer cell lines tested at each given dose and duration of treatment. These included maximum diameter, peak, valley, difference entropy, information measure A, maximum correlation, and product moment. Figure 2 shows that alterations in nuclear structure were dose- and time-dependent and highest at 14 days of treatment with 1.2 mmol/L VPA. The increase in VPA response was minimal when 3 days of treatment was compared with 7 days of treatment. The PC3 nuclei showed the largest change in nuclear structure compared with LNCaP, C4-2, and DU145 at 14 days of treatment with 1.2 mmol/L VPA. We were unable to capture sufficient nuclear images at days 3 and 7 for the PC3 cell line with 0.6 and 1.2 mmol/L VPA and day 7 with 0.6 mmol/L VPA for DU145 cell line. These were therefore excluded from analysis.

To determine whether the changes seen in vitro also occur in vivo, LNCaP, CWR22R, DU145, and PC3 xenografts were analyzed. Univariate logistic regression analysis with a variable selection stringency level of P_z ≤ 0.05 was done to identify significant alterations in nuclear structure in xenografts after treatment of nude mice with VPA (Supplementary Table S2).⁴ A set of the seven most frequently significant nuclear structural alterations were selected to compare VPA response across all xenograft experiments. These included maximum diameter, minimum diameter, average absorbance, SD, peak, valley, and triangular symmetry. Figure 3 shows that the quantitative changes in nuclear structure seen in vivo are dose-dependent. The extent of structural changes differed across different xenografts. The C4-2 xenograft tumors treated with 0.4% VPA treatment for 30 days showed the most pronounced change of all tested xenografts.

To assess whether VPA treatment also affects nuclear structure in drug-metabolizing and drug-filtering organs that appear normal by H&E staining, we captured images of normal liver and kidney cell nuclei 30 days after injection of tumor cells and compared them with mice injected and treated for 30 days with 0.4% VPA. Supplementary Table 2⁴ shows significant nuclear alterations in liver and kidney after treatment. Using the abovementioned xenograft set of significant nuclear structure alterations (maximum diameter, minimum diameter, average absorbance, SD, peak, valley, and triangular symmetry), AUC-ROC for liver and kidney were 0.64 and 0.71, respectively.

Malignant diseases are exemplified by aberrant transcriptional regulation that may be triggered by a change in recruitment of HDACs to the site of transcriptional initiation (33). Screening of HDAC expression in human prostate cancer has revealed distinct class I HDAC profiles between stromal and epithelial cells (34) and in vitro experiments with prostate cancer cell lines have shown a marked increase in HDAC levels for most HDACs compared with normal prostate tissue (35). Although it is thought that acetylation of histones is required for gene expression, inhibiting HDACs does not result in whole-genome reexpression. In fact, about half of the genes that are differentially expressed after VPA treatment are down-regulated (36).

Nuclear organization and chromatin dynamics have regained interest because recent reports suggest that, in interphase cells, gene activation and replication is associated with repositioning of the genetic locus relative to the nuclear compartment and other genomic loci (37, 38). Although individual chromosomes occupy distinct positions in the nucleus called chromosome territories (39), locus-specific movement within and across these boundaries occurs throughout the interphase. The (extent of) mobility of individual loci depends on nuclear localization (40), phase in cell cycle (41), and location on the chromosome (42) and might be actin-, myosin-, and ATP-dependent (43, 44). Although in many cases repositioning of activated genes is oriented toward the nuclear interior (44–48), there is not always a strict correlation between movement away from the periphery and gene activation (4647, 49). The purpose of these movements and the means by which this is possible is under intense debate. It is suggested that chromatin mobility may facilitate access for enzymes involved in histone modifications, nucleosome remodeling, and the ensuing folding or unfolding of the domain (38). Recently, investigators have described the existence of regulatory cross-talk between “kissing” loci within and across chromosomes, suggesting the existence of spatial networks of gene regulation. They propose that interacting loci form a “poised chromatin hub,” which might recruit remodeling complexes or histone acetyltransferases to form a positive environment for gene expression (50, 51). As described above, it is known that chromatin remodeling proteins such as high-mobility group proteins and the SWI/SNF complex are (in part) regulated by acetylation. Furthermore, it has been shown that acetylation is deregulated in cancer cells (2, 3).

In the current study, we show that VPA treatment results in dose- and time-dependent changes in the nuclear structure of prostate cancer cell lines, reflecting change in chromatin remodeling dynamics in prostate cancer cells. Further, these nuclear morphometric alterations can be used as a marker to measure therapeutic response. With the automated CoolScope, results can be easily calibrated and laboratory-dependent variation can be minimized. Of note, VPA treatment also induces nuclear structural changes in the mouse normal liver and kidney cells where it is metabolized and its breakdown products are excreted (52). These nuclear structural changes in normal liver and kidney cells may reflect underlying cause for reported elevation of liver enzyme, hepatitis, renal tubular dysfunction in children (53), and Fanconi syndrome (54, 55) during VPA treatment.

Previously, our group has shown that VPA treatment of prostate cancer cells increases histone H3 acetylation, CK18, and p21 and p27 expression and decreases androgen receptor, cyclin D1, and Ki-67 expression in vitro and in vivo (56, 57). Further, chronic low-dose administration of VPA results in an increase in apoptosis (assessed by TUNEL assay) and a decrease in angiogenesis (assessed by CD34 microvessel density) in prostate cancer xenografts (56). The next step in our investigation, the analysis of pretreatment and post-treatment human prostate cancer tissue, is not feasible for now, as VPA treatment for solid tumors is only given in a clinical trial setting at the moment. Patients eligible for a clinical trial are in an end stage of their disease and typically have had a prostatectomy or chemotherapy before enrollment.

Supplementary Tables S1⁴ and S2⁴ illustrate that in vitro and in vivo prostate cancer-treated cells had different structural alterations profiles. There are several potential reasons that may explain these differences. Chromatin remodeling and actin filament organization are interacting processes (58). It is known that the nucleus forms part of a continuous physical network spanning the extracellular matrix, the cytoskeleton, and the nuclear envelope. Environmentally mediated forces can be transmitted to the nucleus and induce deformations of the chromatin (59). In our experiments, in vitro cells were grown in a single layer in a dish before harvesting using trypsin, whereas in vivo cells were grown three-dimensionally in mouse stroma and were paraffin-embedded as such. In addition, in vitro and in vivo cells have access to a different set of growth factors and other nutrients. Their microenvironment differs grossly, which might affect their type of response to VPA treatment.

The quantitative assessment of nuclear structure by digital image analysis has identified numerous potential clinical applications. Our group has previously shown that there are significant alterations in nuclear structure between and within Gleason grading patterns 3 to 5 (60). Further, we showed that nuclear structure analysis is of additive value with clinicopathologic parameters in the prediction of biochemical recurrence, distant metastasis, and death after radical prostatectomy (23, 24). Recently, Makarov et al. (61) used quantitative nuclear structure alterations to predict conversion to unfavorable biopsy pathology during watchful waiting surveillance. In this article, we discuss the use of nuclear structure analysis in therapeutic response measurements. Identification of proteins that modify nuclear chromatin organization is of paramount importance to improve our understanding of the biology behind these measurements. One of such proteins was histone acetyltransferase protein p300 has shown to predict prostate cancer biochemical recurrence in men with long-term follow-up and correlate with changes in epithelia nuclear size and shape (62–64). The use of quantitative nuclear structure alterations and the molecular mechanisms that cause such changes provide the foundation for continued research in this area that can eventually change the management of prostate cancer patients.

In conclusion, HDACI VPA significantly changes the epithelial nuclear structure in a time- and dose-dependent fashion in prostate cancer both in vitro and in vivo. VPA also affects nuclear structure in normal mouse liver and kidney tissue, which likely reflects a natural physiologic response.

No potential conflicts of interest were disclosed.

We thank Shabana Shabbeer for providing xenograft material.