Stromal–epithelial interactions dictate prostate tumorigenesis and response to castration. Hydrogen peroxide-inducible clone 5 (Hic-5/ARA55) is a transforming growth factor-beta (TGF-β)-induced coactivator of androgen receptor (AR) expressed in the prostate stroma. Interestingly, following castration, we identified epithelial expression of Hic-5/ARA55 in mouse and human prostate tissues. To determine the role of epithelial Hic-5 in prostate cancer progression and castration responsiveness, we compared LNCaP cells having Hic-5 stably expressed with the parental LNCaP cells following tissue recombination xenografts with mouse prostate stromal cells. We previously identified knocking out prostate stromal TGF-β signaling potentiated castrate-resistant prostate tumors, in a Wnt-dependent manner. The LNCaP chimeric tumors containing prostate fibroblasts conditionally knocked out for the TGF-β type II receptor (Tgfbr2-KO) resulted in larger, more invasive, and castration-resistant tumors compared those with floxed (control) stromal cells. However, the LNCaP-Hic5 associated with Tgfbr2-KO fibroblasts generated chimeric tumors with reduced tumor volume, lack of invasion and restored castration dependence. Neutralization of canonical Wnt signaling is shown to reduce prostate tumor size and restore regression following castration. Thus, we hypothesized that epithelial Hic-5/ARA55 expression negatively regulated Wnt signaling. The mechanism of the Hic-5/ARA55 effects on castration was determined by analysis of the c-myc promoter. C-myc luciferase reporter activity suggested Hic-5/ARA55 expression inhibited c-myc activity by β-catenin. Sequential ChIP analysis indicated β-catenin and T-cell-specific 4 (TCF4) bound the endogenous c-myc promoter in the absence of Hic-5 expression. However, the formation of a TCF4/Hic-5 repressor complex inhibited c-myc promoter activity, by excluding β-catenin binding with TCF4 on the promoter. The data indicate Hic-5/ARA55 expression in response to castration-enabled epithelial regression through the repression of c-myc gene at the chromatin level.

DNA damage found in prostate cancer-associated fibroblasts (CAF) promotes tumor progression. In the absence of somatic mutations in CAF, epigenetic changes dictate how stromal co-evolution is mediated in tumors. Seventy percent of prostate cancer patients lose expression of transforming growth factor-beta type II receptor (TGFBR2) in the stromal compartment (n = 77, p value = 0.0001), similar to the rate of glutathione S-transferase P1 (GSTP1) silencing. Xenografting of human prostate cancer epithelia, LNCaP, resulted in the epigenetic Tgfbr2 silencing of host mouse prostatic fibroblasts. Stromal Tgfbr2 promoter hypermethylation initiated by LNCaP cells was found to be dependent on IL-6 expression, based on neutralizing antibody studies. We further found that pharmacologic and transgenic knockout of TGF-β responsiveness in prostatic fibroblasts induced Gstp1 promoter methylation. It is known that TGF-β promotes DNA stability, however the mechanism is not well understood. Both prostatic human CAF and mouse transgenic knockout of Tgbr2 had elevated DNA methyltransferase I (DNMT1) activity and histone H3 lysine 9 trimethylation (H3K9me3) to suggest greater promoter methylation. Interestingly, the conditional knockout of Tgfbr2 in mouse prostatic fibroblasts, in modeling epigenetic silencing of Tgfbr2, had greater epigenetic gene silencing of multiple DNA damage repair and oxidative stress response genes, based on promoter methylation array analysis. Homologous gene silencing was validated by RT-PCR in mouse and human prostatic CAF. Not surprisingly, DNA damage repair gene silencing in the prostatic stromal cells corresponded with the presence of DNA damage. Restoring the expression of the epigenetically silenced genes in wild type fibroblasts with radiation-induced DNA damage reduced tumor progression. Tumor progression was inhibited even when epigenetic silencing was reversed in the Tgfbr2 knockout prostatic fibroblasts. Thus, fibroblastic epigenetic changes causative of DNA damage, initiated by association with cancer epithelia, is a dominant mediator of tumor progression over TGF-β responsiveness.

The goal of this study was to determine if quantitative changes in dynamic contrast enhanced MRI (DCE-MRI) following a single cycle of chemotherapy can be used to separate pathologic responders (i.e., no residual tumor in the breast at surgery) from pathologic non-responders. 28 patients with Stage II/III breast cancer were enrolled in an IRB-approved clinical trial where breast MRI scans were acquired before (t1) and after one cycle of therapy (t2). Imaging was performed on a 3.0T MR scanner (Philips Healthcare, The Netherlands) and employed a 3D spoiled gradient echo sequence with a spatial resolution of 6.6 mm3 and a temporal resolution of 16 seconds collected at 25 time points before and after the intravenous injection of 0.1 mmol/kg of gadopentetate dimeglumine (Magnevist, Wayne, NJ). At surgery, 12 patients were responders while 16 patients were non-responders. Both semi-quantitative and quantitative analyses were used to summarize the DCE-MRI data. The semi-quantitative parameters were the signal enhancement rate (SER [1]) and tumor volume (TV). Three pharmacokinetic models, the Tofts-Kety (TK), the Extended Tofts-Kety (ETK), and the fast exchange regime (FXR), were used to estimate the following quantitative parameters: the volume transfer constant (Ktrans), efflux rate constant (kep), vascular volume (vp), and the extravascular extracellular volume fraction (ve) [2]. Each parameter was summarized in two ways: 1) the change in mean from t1 to t2, and 2) the mean at t2. Receiver operating characteristic (ROC) analysis was then performed to determine the ability of each parameter to predict treatment response. The table displays the areas under the ROC curves (AUC) for each parameter. For the early change in parameters, the AUC for TV and SER were 0.48 and 0.66, respectively. The best AUC of the quantitative parameters was 0.73 from kep estimated by the ETK model. The sensitivities/specificities for TV, SER, and kep for predicting pathologic response were 88%/33%, 64%/79%, and 56%/92%, respectively. For the mean parameter values at t2, the AUCs of TV, SER, and kep were 0.50, 0.56, and 0.79, with sensitivities/specificities for predicting pathologic response of 63%/50%, 93%/21%, and 81%/75%, respectively. Our results can be interpreted in light of the ACRIN 6657/I-SPY trial [3] which found that change in TV and SER at an early time point were the most predictive of response with AUCs of 0.72 and 0.71, respectively. Our preliminary results, especially our AUC of 0.79 for kep at t2, suggest that a more quantitative analysis of higher temporal resolution DCE-MRI data may achieve comparable or even superior results. Our ongoing efforts involve combining multiple parameters in a multivariate analysis with apparent diffusion coefficient data from diffusion weighted MRI. [1] Arasu et al. Acad Radiol 2011;18:716–721. [2] Yankeelov & Gore. CMIR 2009;3:91–107. [3] Hylton et al. Radiology. 2012;263:663–672. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-01-03.

Purpose: To develop a clinically‐relevant patient‐specific modeling framework for oncology that is amenable to readily available clinical imaging data and yet retains the most salient features of response prediction. We use a mechanically coupled mathematical model of tumor growth that is initialized and constrained by MRI data early in the course of therapy, to guide the determination of model parameters and predict the response of breast cancers to neoadjuvant chemotherapy (NAC). Methods: We adopt a patient‐scale spatiotemporal tumor growth modeling framework and apply patient‐specific predictive modeling, constrained by quantitative imaging data, to a group of 26 patients exhibiting a varying degree of response to NAC. Dynamic contrast enhanced MRI, diffusion weighted MRI, and anatomical T1‐weighted MRI volumes were acquired prior to beginning NAC, after one cycle of NAC, and at the conclusion of NAC. Tumor response is parameterized using data from before and after the first cycle of therapy, and the model is driven forward in time to predict tumor burden at the conclusion of therapy. Model reconstructed parameters and predictions are retrospectively assessed for prognostic value in predicting patients that eventually respond or do not respond to NAC. Results: Using our mechanics‐coupled modeling approach, we are able to discriminate, after the first cycle of therapy, breast cancer patients that would eventually achieve a complete pathological response and those who would not, with an area under the receiver operator characteristic curve of 0.81, sensitivity of 90%, and specificity of 56%. Conclusion: We show the potential for model‐predictions at the conclusion of therapy for use as a prognostic indicator of response to therapy. This work provides considerable promise for predictive modeling centered on integrating quantitative in vivo imaging data with biomechanical models of tumor growth. National Institutes of Health NCI 1U01CA142565, NCI U01CA174706, NCI R25CA092043, NCI 1P50 098131, NCI P30CA68485, NCI R01CA138599, NINDS R01NS049251. The Vanderbilt initiative in Surgery and Engineering Pilot Award Program and the Whitaker Foundation.