Role for stromal heterogeneity in prostate tumorigenesis

Abstract

Prostate cancer develops through a stochastic mechanism whereby precancerous lesions on occasion progress to multifocal adenocarcinoma. Analysis of human benign and cancer prostate tissues revealed heterogeneous loss of TGF-β signaling in the cancer-associated stromal fibroblastic cell compartment. To test the hypothesis that prostate cancer progression is dependent on the heterogeneous TGF-β responsive microenvironment, a tissue recombination experiment was designed in which the ratio of TGF-β responsive and nonresponsive stromal cells was varied. Although 100% TGF-β responsive stromal cells supported benign prostate growth and 100% TGF-β nonresponsive stromal cells resulted in precancerous lesions, only the mixture of TGF-β responsive and nonresponsive stromal cells resulted in adenocarcinoma. A computational model was used to resolve a mechanism of tumorigenic progression in which proliferation and invasion occur in two independent steps mediated by distinct stromally derived paracrine signals produced by TGF-β nonresponsive and responsive stromal cells. Complex spatial relationships of stromal and epithelial cells were incorporated into the model on the basis of experimental data. Informed by incorporation of experimentally derived spatial parameters for complex stromal-epithelial relationships, the computational model indicated ranges for the relative production of paracrine factors by each cell type and provided bounds for the diffusive range of the molecules. Because SDF-1 satisfied model predictions for an invasion-promoting paracrine factor, a more focused computational model was subsequently used to investigate whether SDF-1 was the invasion signal. Simulations replicating SDF-1 expression data revealed the requirement for cooperative SDF-1 expression, a prediction supported biologically by heterotypic stromal interleukin-1β signaling between fibroblastic cell populations. The cancer stromal field effect supports a functional role for the unaltered fibroblasts as a cooperative mediator of cancer progression.

Figure 1. Stromal heterogeneity of human and mouse prostate tissues

A) Sample data of benign prostate hyperplasia (BPH) tissues suggest near homogeneous phosphorylated-Smad2 localization in the stroma (S). The lower and higher grade PCa tissues (left and right respectively) have similarly heterogeneous phosphorylated-Smad2 expression. B) Immunofluorescent staining demonstrating focal areas of increased stromal CD90 expression in human PCa (red, highlighted by dashed line) concurrent with loss of stromal phosphorylated-Smad2 expression (green). In comparison, there were greater uniformity in CD90 and phosphorylated-Smad2 in the BPH stromal tissue. C) There were heterogeneous phosphorylated-Smad2 localization in the prostate stroma of Tgfbr2^fspKO compared to Tgfbr2^{floxE2/floxE2} control mice.

Figure 2. Percentage of WT and Tgfbr2-KO stroma in prostate tissue recombinant allografts promoted cancer progression

A) H&E of grafted WT prostatic epithelial organoids recombined with 100% Tgfbr2-flox prostate stromal cells had normal glandular architecture. B) There was observed progression to adenocarcinoma in grafts of WT prostatic epithelial organoids recombined with a 50/50 mixture of Tgfbr2-flox and Tgfbr2-KO prostate stromal cells. C) PIN lesions developed in grafts of WT prostatic epithelial organoids recombined with 100% Tgfbr2-KO prostate stromal cells. Arrowheads (B & C) indicate transformed epithelia.

Figure 3. Establishment of a biologically informed computational model

A) H&E stained section of normal mouse prostate (left) was used to assign corresponding positions for simulated cells (right) assuming 50% Tgfbr2-KO stromal cells (black epithelia, blue WT stromal cells, and cyan Tgfbr2-KO stromal cells). B) Simulations indicated that stromal heterogeneity altered the proliferative and invasive potential of prostate epithelial cells, where greatest invasion occurred at heterogeneous mixtures of stromal cells and the extent of epithelial proliferation and invasion depended on the ratio of paracrine factor production and threshold response. The number of cells that became proliferative (Step 1, dashed line) and invasive (Step 2, solid line) was based on paracrine factor diffusion lengths (L_M1=200 and L_M2=300 µm), transformation response thresholds (T_M1=0.0453 and T_M2=0.3432 paracrine factor units), and fixed total paracrine factor abundance per cell source (A_M1=A_M2=10,000 paracrine factor units). Error bars indicate the standard error of 100 simulations. C) Phase diagram illustrating the final epithelial classifications as a function of Tgfbr2-KO cell M₁ abundance (y-axis) and WT cell M₂ abundance (x-axis) in a tissue at 50/50 mixture of WT and Tgfbr2-KO stromal cells. If the production rate of both paracrine factors was low relative to the transformation threshold, the cells remained normal. Paracrine factor diffusion lengths and thresholds for M₁ and M₂ were as in 3B.

Figure 4. Epigenetic regulation of Wnt and SDF-1 signaling by TGF-β in prostate stromal cells

Quantitative RT-PCR revealed reversal of methylation induced silencing of SFRP1, SFRP2, and SDF-1 by 5-aza-dC treatment of Tgfbr2-KO stroma, comparable to WT stroma. Data are presented as mean ± S.D. and normalized to GAPDH.

Figure 5. Computational modeling incorporating experimental data for putative paracrine factors predicted inter-stromal cell cooperativity for paracrine factor M₂ production

A) In simulations, WT stromal cells produced 20% of M₁ levels produced by Tgfbr2-KO stroma, modeling experimental data for Wnt-3a (M₁) production. The gap between the minimum (dotted line) and the maximum (solid line) thresholds illustrated that the model cannot yield experimental proliferation rates over this range of diffusion lengths. B) In simulations, Tgfbr2-KO stromal cells produced 25% of M₂ levels produced by WT stroma cells, modeling experimental data for SDF-1 (M₂) production. The area of overlap (shaded region) above the minimum (dotted line) and below the maximum (solid line) thresholds were consistent with experimental proliferation rates for diffusion lengths greater than L_M2=24 µm. Minima and maxima determination for A and B are described in supplemental information. C) SDF-1 (candidate M₂) expression by WT (white bar) and Tgfbr2-KO (KO, black bar) cell types was cooperatively elevated following co-culture as measured by qRT-PCR (normalized to β-actin, relative to KO expression). D) Simulation of M₂ production cooperativity assumed paracrine factors M₃ and M₄ expression by confluent WT and Tgfbr2-KO stromal cells. Best-fit parameters were found to match data (in C) assuming first M₃ secretion by the Tgfbr2-KO stroma that resulted in SDF-1 induction by WT cells (gray line) and the added assumption of M₄ secretion by WT stroma augmented Tgfbr2-KO response to M₃ (dashed black line). Parameters for simulations are detailed in supplemental information.

Figure 6. Paracrine signaling between stromal sub-types drives carcinogenesis

A) SDF-1 mRNA expression was measured in Tgfbr2-Flox and Tgfbr2-KO prostate stromal cells by qRT-PCR in response to treatment with the candidate M₃ heterotypic stromal signaling factor, IL-1β. B) A model of stromal TGF-β responsiveness driving prostate carcinogenesis.. Loss of stromal responsiveness to TGF-β, resulted in elevated production of TGF-β and Wnt-3a by the stroma. While increased Wnt paracrine signaling promoted epithelial proliferation, the increase in TGF-β in the microenvironment resulted in CXCR4 expression by the epithelium, subsequently increasing its sensitivity to SDF-1. WT and KO stroma cooperate to express elevated SDF-1, driving the progression of prostatic carcinogenesis.