Peroxiredoxin 1 Controls Prostate Cancer Growth through Toll-Like Receptor 4–Dependent Regulation of Tumor Vasculature

Prostate cancer (CaP) is a highly heterogeneous disease of unknown etiology (1, 2). In recent years a number of histopathologic and epidemiologic studies have implicated chronic inflammation in prostate carcinogenesis (3). Toll-like receptor 4 (TLR4) can contribute to CaP cell proliferation and invasion (4, 5) and to CaP risk (6). Preclinical studies have shown that TLR4 expression is required for CaP growth (4), suggesting the presence of a tumor-derived TLR4 ligand that contributes to the maintenance of a chronic inflammatory and protumorigenic environment and promotion of CaP growth.

Peroxiredoxin 1 (Prx1) is a multifunctional member of the 2-Cys subfamily of the evolutionarily conserved thiol-dependent antioxidant Prx family of enzymes (7) that is over-expressed by multiple cancers. Elevated Prx1 expression in lung and bladder cancer is associated with diminished overall survival and poor clinical outcome (8–10).

Prx1 is secreted in a nonclassical fashion by stressed, activated, and transformed cells, including prostate tumor cells (11–13). Extracellular Prx1 is a TLR4 ligand that stimulates the expression of proinflammatory cytokines from macrophages and dendritic cells (13). We hypothesize that secretion of Prx1 by prostate tumor cells leads to the generation of a protumorigenic microenvironment through its interaction with TLR4. Activation of TLR4 increases VEGF expression in both cancer and normal cells (14, 15). We further predict that Prx1 enhances prostate carcinogenesis and CaP growth through TLR4-dependent induction of VEGF and augmentation of tumor vasculature. The results presented here support these hypotheses. Reduction of Prx1 expression by prostate tumor cells with shRNA-inhibited tumor growth in s.c. CaP models. Delayed tumor growth in tumors with reduced Prx1 levels seems to be owing to reduced tumor vasculature. These findings suggest that Prx1 plays a pivotal role in CaP progression by orchestrating VEGF expression and vascular network formation.

Bovine serum albumin, insulin, and antibodies specific for β-actin were obtained from Sigma-Aldrich. Antibodies specific for SV40 large T antigen, CD31, and all isotype control antibodies were purchased from PharMingen. Antibodies against VEGF were purchased from Santa Cruz Biotechnology. Antibodies against NG2 were obtained from R&D Systems. Antibodies specific for Prx1 were obtained from Lab Frontier (13). Sunitinib malate was purchased from Selleck Chemicals LLC. The VEGF₁₂₁ expression construct was a gift from Douglas Fraker (University of Pennsylvania, Philadelphia, PA; ref. 16). The MyD88DN expression vector was a gift from Stuart Calderwood (Harvard Medical School, Boston, MA; ref. 17).

Tissue microarrays (TMA) were constructed by the Pathology Resource Network at Roswell Park Cancer Institute (RPCI; Buffalo, NY). Tissue cores (0.6 mm) were cut from 92 formalin-fixed paraffin-embedded donor blocks of prostatic adenocarcinoma, each representing a different case. Controls include cores of normal prostate tissue taken from each case of CaP. The TMA contains an unequal number of samples of each tumor grade and reflects the patient population at RPCI.

The TMA was stained with antibodies specific for β-cytokeratin, p63, racemase, and Prx1. Scoring of Prx1 expression was done in a noncontinuous (semiquantitative) scale and was determined for both the percentage of cells positive for Prx1 and intensity of epithelial staining. Overall Prx1 expression was grouped as follows: 1, 0%–5%; 2, 6%–25%; 3, 26%–50%; 4, 51%–75%; 5, 75%–100%. The epithelial staining intensity was ranked on a relative scale of 1–4 with 4 being the greatest staining intensity. To overcome the intraobserver variability the score was done twice and compared.

The murine CaP cell line C2H was maintained as described in the presence of 10⁻⁸ M dihydrotestosterone (Sigma Chemical Co.) at 37°C and 10% CO₂ (18). The human CaP cell line, PC-3M, was obtained from American Type Culture Collection and cultured according to the company's directions. Retroviral short hairpin RNA expression constructs specific for Prx1 expression (shPrx1) and nonspecific (scramble) were used to reduce Prx1 expression in PC-3M and C2H cell lines (9, 19). Stable scramble/shPrx1 cell lines were maintained in the presence of G418 (CellGro). The VEGF cDNA was obtained from Dr. D. Fraker (16) and subcloned into pcDNA3.1/hygro (Invitrogen). Stable shPrx1_VEGF PC-3M cell lines were maintained in G418 and hygromycin.

Human umbilical vein endothelial cells (HUVEC) were cultured in supplemented endothelial growth media (Promega) at 37°C and 5% CO₂. Murine endothelial cells, 2H-11, were cultured in DMEM supplemented with 10% defined FBS at 37°C and 10% CO_2.

Cells were transfected with vectors encoding a MyD88 dominant negative (DN) cDNA (20) using FuGENE6 (Invitrogen) according to the manufacturer's protocol.

C57BL/6J (TLR4^+/+) and C57BL/6ScN^tlr4lps-del/JthJ (TLR4^−/−) pathogen-free mice were purchased from The Jackson Laboratory. C57BL/6N (TLR4^+/+) were purchased from the National Cancer Institute; no difference in C2H scramble tumor growth was observed in B6 mice purchased from either source (data not shown). Severe combined immunodeficiency (SCID) pathogen-free mice were purchased through the Laboratory Animal Resource of RPCI. Transgenic adenocarcinoma of mouse prostate (TRAMP) mice (21) were bred, maintained, and housed at RPCI. The RPCI Institutional Animal Care and Use Committee approved both animal care and experiments.

Immunohistochemical analyses of murine tissue included staining tissues with antibodies specific for Prx1 (10 μg/mL, Lab Frontier) and SV40 T antigen (1.25 μg/mL, Pharminogen) expression in TRAMP prostatic tissues (22). Briefly, prostatic tissues (collected at necropsy) were fixed in 10% (vol/vol) buffered formalin and sections (5.0 μm) were cut from paraffin-embedded tissues. Slides/tissues were incubated overnight with primary antibodies. Secondary antibodies were incubated with tissues for 2 hours. Products were visualized with the chromogen 3′3′-diaminobenzidine tetrahydrochloride (DAB; Sigma). The cell nuclei were counterstained with hematoxylin.

PC-3M scramble/shPrx1 tumors (150 mm³) were harvested, fixed in zinc-formalin (R&D Systems), and paraffin embedded. CD31 and Prx1 expression in PC-3M scramble/shPrx1 tumors was carried out by the immunohistochemistry facility at RPCI as previously described (23).

Tumor growth was monitored by measuring the orthogonal diameters of tumors once every 3 days with calipers (24). The tumor volume, V, was calculated with the formula V = (lw²/2), where l is the longest axis of the tumor and w is the axis perpendicular to l.

Microvascular assessment of PC-3M scramble and shPrx1 tumors (150 mm³) was carried out using dynamic contrast, magnetic resonance imaging (MRI). A multislice, T2-weighted fast spin echo (TE/TR = 40/2,500 ms) scan was acquired in the axial orientation to provide an anatomic reference for tumor volume and localization. Three precontrast T1 relaxation rate measurements of the tumors were carried out using a saturation recovery, fast spin echo scan with variable repetition times ranging from 360 to 6,000 ms. Human serum albumin-(Gd-DTPA₃₅; Center for Pharmaceutical and Molecular Imaging, University of California, San Francisco) was then injected i.v. via the tail vein at a dose of 0.1 mmol/kg. Serial T1 measurements were then carried out for up to 45 minutes after injection.

Magnetic resonance data sets were converted into the Analyze 7.5 format and regions of interest (ROI) were defined for the tumor and kidneys using Analyze 7.0 (AnalyzeDirect). Mean signal intensity at each repetition time was determined within each ROI and T1 relaxation rates (R1) were calculated by nonlinear fitting of the equation: S_(TR) = S_MAX (1−e ^−(R1·TR)), using Matlab's Curve Fitting Toolbox (Matlab 2008a, MathWorks Inc.). The increase in R1 (ΔR1) was calculated by subtracting the average baseline R1 from each post-injection value. All ΔR1 values were corrected by dividing by the ΔR1 value of the kidney to account for injection variability. Normalized tumor ΔR1 values were plotted against time after injection, and a linear regression curve was calculated to yield the relative permeability (slope) and fractional blood volume (intercept; ref. 25).

Endothelial cell proliferation assays were carried out as described (26). In indicated experiments VEGF receptor (VEGFR) inhibitor (sunitinib malate) or blocking antibodies for VEGF (4 μg/mL) were added 2 hours prior to stimulation with rPrx1 or cultured supernatants from PC-3M scramble/shPrx1 cells. MyD88DN transfection was conducted 24 hours prior to stimulation. Proliferation was determined 24 hours after stimulation by MTT assay (27).

Endothelial cell migration assays were carried out by coating 24-well plates with 50 μL of collagen per well (Roche) mixed with rPrx1. Endothelial cell media was added to the bottom of the collagen coated well and HUVECs or 2H-11 cells (5 × 10⁵) were placed in the top of a 3.0 μm pore transwell. The number of migrated cells was determined after 24 hours by trypan blue staining.

To assess the role of Prx1 in the recruitment of vessels in vivo, Matrigel (BD Bioscience) containing rPrx1 was injected s.c. into SCID mice. Eight days later mice were euthanized and the hemoglobulin content present in the Matrigel plugs was determined (Sigma).

Endothelial differentiation assays were carried out according to Godoy and colleagues (26). Total tube length per image was quantified using Optimas 6.2 (Media Cybernetics, Inc.). Graphs represent a summation of the lengths of all individual tubular structures.

Mouse and human ELISA kits specific for mouse and human VEGF were purchased from R&D systems. The mouse/human Prx1 ELISA kit was purchased from Northwest Life Science Specialties, LLC. Assays were completed according to manufacturers' instructions.

Statistical analyses were done using a standardized Student's t-test with Welch's correction. For tumor growth analysis, hours-to-400 mm³ tumor volume were calculated for each animal by linearly interpolating between the times just before and after this volume was reached, using log (tumor volume) for the calculations (28). Tumor responses between groups were compared using the Kaplan–Meier analysis. Event curves were compared using the log-rank test, which calculated a 2-tailed P value testing the null hypothesis that the curves were identical. Differences were considered significant when P values were ≤ 0.05.

Prx1 expression in human CaP was determined utilizing a human CaP TMA. Tumor grade present in each core was determined by analysis of β-cytokeratin, p63, and racemace expression (Triple stain; Fig. 1A; ref. 29). Both overall Prx1 (Fig. 1B) and epithelial cell expression of Prx1 (Fig. 1C) increased in malignant prostatic cancer samples with respect to normal tissue. Prx1 expression was increased at the earliest tumor grade examined when compared to expression in normal tissue.

Prx1 expression levels also increase during CaP progression in the TRAMP murine model of CaP. Normal prostates from (C57BL/6 × FVB)F₁ WT mice express minimal levels of Prx1 (Fig. 1D). Prx1 expression is increased in foci of transformed cells and is further increased in intermediate and late-stage CaP tissues.

The role of Prx1 in CaP was evaluated in 2 s.c. models of CaP, human PC-3M and murine C2H. PC-3M and C2H cell lines were transfected with vectors engineered to express Prx1 specific shRNA (shPrx1), which resulted in a 50% reduction in Prx1 expression as compared to cells expressing nonspecific (scramble) shRNA (Fig. 2A, 2B). PC-3M shPrx1 tumor growth was delayed and growing tumors exhibited slower growth rate when compared with PC-3M scramble tumors (1.90 ± 0.16 mm³/day vs. 21.2 ± 2.02 mm³/day; P ≤ 0.0001; Fig. 2C). C2H shPrx1 tumors did not grow (Fig. 2D). To confirm that Prx1 control of CaP tumor growth was specific, Prx1 expression was restored in shPrx1 tumor cells by expression of shRNA resistant Prx1. Expression of shRNA resistant Prx1 (sRP) restored Prx1 expression and tumor growth, indicating that Prx1 specifically regulates CaP growth (Fig. 2A). Prx1 can regulate cell proliferation/apoptosis through its peroxidase/chaperone activities (7); however, PC-3M shPrx1 tumors did not show decreased numbers of proliferating cells or increases in apoptosis (data not shown).

To examine the role of TLR4 in mediating Prx1 control of tumor growth, C2H scramble cells were injected into TLR4 signaling deficient (TLR4^−/−) mice. Tumor growth was not observed. In TLR4^−/− mice (Fig. 2D), suggesting that Prx1 promotion of C2H tumor growth is TLR4 dependent. Cumulatively, these studies support the hypothesis that Prx1 regulation of CaP is an in vivo phenomenon that depends on host expression of TLR4.

PC-3M shPrx1 tumors have decreased numbers of vessels and overall vascular area as compared to control (scramble) tumors (Fig. 3A, 3C, 3D). The vascular area per vessel, or the average size of each vessel, is increased in shPrx1 tumors (Fig. 3E), which correlates with the presence of large vessels and minimal microvasculature (Fig. 3A).

Colocalization of pericytes and endothelial cells is a marker of mature vasculature and an indication of vessel function (30). A majority of endothelial cells (CD31⁺) and pericytes (NG2⁺) are colocalized in PC-3M scramble tumors (Fig. 4A). Endothelial cells and pericytes are also colocalized in the limited number of vessels present in PC-3M shPrx1 tumors, suggesting that although vasculature is limited in shPrx1 tumors, the vasculature that is present is mature. Reduction of Prx1 expression in PC-3M tumors leads to reduced vascular permeability, which was determined using dynamic contrast MRI (Fig. 4B, 4C). In summary, these results suggest that Prx1 expression contributes to tumor vasculature formation and function.

Analysis of angiogenic protein expression in PC-3M tumor lysates showed that shPrx1 tumors contained reduced levels of VEGF, TNFα, interleukin 6 (IL-6), TGFα, and TGFβ (data not shown). Further analysis showed that total host (mouse) and tumor derived (human) VEGF levels were decreased in shPrx1 tumors (Fig. 5A). Total VEGF consists of a number of isoforms, the predominant of which are VEGF₁₂₁ and VEGF₁₆₅. The ratio of VEGF₁₂₁ to VEGF₁₆₅ is a prognostic factor of malignancy in clinical and preclinical models of CaP (31). The ratio of VEGF₁₂₁ to VEGF₁₆₅ is decreased in shPrx1 tumors and cells (Fig. 5B). Restoration of VEGF expression by PC-3M shPrx1 cells restores tumor growth (Fig. 5C).

VEGF secretion is reduced in PC-3M scramble cells expressing MyD88DN (Fig. 5D), suggesting that Prx1 induction of VEGF is mediated by interaction with TLR4 and dependent upon MyD88. This finding is supported by results showing that extracellular addition of recombinant Prx1 (rPrx1) or restoration of Prx1 by expression of a shPrx1 resistant Prx1 (sRP) stimulated VEGF expression in shPrx1 cells. The induction of VEGF by rPrx1 or sRP was abrogated by the expression of MyD88DN (Fig. 5D). VEGF secretion from macrophages was also stimulated by incubation with rPrx1 and was abrogated in macrophages derived from TLR4^−/− mice, confirming the TLR4 dependence of Prx1 regulation of VEGF (Fig. 5E).

VEGF mediates angiogenesis partly through its effects on endothelial cell proliferation, migration, and differentiation (32). To determine if Prx1 influences endothelial cell proliferation, endothelial cells were cultured in the presence or absence of rPrx1. Addition of rPrx1 to endothelial cells increased their proliferation (Fig. 6A). CM from scramble PC-3M, but not CM from shPrx1 PC-3M cells also increased the proliferation of endothelial cells (Fig. 6A). Inhibition of TLR4 signaling by expression of MyD88DN in endothelial cells ablated the proliferative effects of Prx1 (Fig. 6A). Similar results were observed with C2H CM (data not shown).

To test whether Prx1 control of endothelial cell proliferation is dependent on its chaperone or peroxidase activity, endothelial cells were incubated with wild-type (Prx1), peroxidase-null (Prx1C52S), or chaperone-null (Prx1C83S) Prx1. Proliferative effects on endothelial cells were peroxidase independent and chaperone/structure dependent (Fig. 6B), supporting the conclusion that Prx1-TLR4 interaction promotes endothelial cell proliferation.

A Matrigel model was utilized to address the effects of Prx1 on endothelial cell recruitment or migration in an in vivo setting. Prx1-augmented recruitment of endothelial cells into Matrigel in vivo (Fig. 6C). Transwell assays were carried out with rPrx1 or PC-3M scramble/shPrx1 CM to further evaluate regulation of endothelial cell recruitment and migration. Addition of rPrx1 and CM from scramble PC-3M cells, but not CM from shPrx1 PC-3M cells, led to increased endothelial cell migration. Migration was inhibited by endothelial cell expression of MyD88DN (Fig. 6D). Incubation of endothelial cells with Prx1 led to a dose dependent increase in tubule formation that was abrogated by endothelial cell expression of MyD88DN (Fig. 6E).

To address if Prx1 regulation of endothelial cell proliferation was dependent on VEGF, endothelial cells were incubated with PC-3M scramble CM in the presence of blocking antibodies specific to VEGF or control antibodies. The proliferative effects of PC-3M scramble CM on HUVECs was reduced when VEGF activity was blocked (Fig. 7A). This finding was confirmed by preincubation of endothelial cells with suntinib malate, a tyrosine kinase inhibitor that inhibits VEGF signaling (33)(Fig. 7B).

PC-3M scramble CM contains tumor-derived VEGF, which could mediate the proliferative effects on endothelial cells independent of Prx1 interaction with TLR4. To confirm PC-3M secreted Prx1 was directly inducing endothelial cell proliferation, Prx1 was depleted from scramble CM. Prx1-depleted scramble CM failed to stimulate endothelial cell proliferation (Fig. 7C). Thus, tumor cell-derived VEGF does not seem to be sufficient to drive endothelial cell proliferation.

Our results show that Prx1 expression correlates with CaP progression and that Prx1 expression is critical for CaP growth. Reduction in tumor expression of Prx1 led to reduced tumor vasculature formation and function. Investigation into the mechanism of Prx1 control of CaP showed that Prx1 stimulates the expression of VEGF in both tumor and host cells in a TLR4-dependent manner. Tumor cell–secreted Prx1 stimulates TLR4 and VEGF-dependent endothelial cell proliferation, migration, and differentiation. These findings suggest that tumor-secreted Prx1 interacts with TLR4 to promote a proangiogenic tumor microenvironment that is required for tumor growth. TLR4 activation has previously been linked to angiogenesis (14, 15) and increased CaP progression (4, 5, 34), however, to our knowledge this is the first identification of a tumor-derived TLR4 ligand that promotes CaP growth by enhancing angiogenesis. Tumor-derived high mobility group B1 (HMGB1), which interacts with TLRs 2 and 4, can stimulate angiogenesis, but these effects are attributed to its interaction with the receptor for advanced glycation end products (RAGE) and not TLR (35).

PC-3M tumor cells with reduced Prx1 seem to undergo a period of adaption in vivo that results in tumor growth delay and slower growth kinetics. Overexpression of VEGF in these tumor cells reduces the tumor growth delay but does not completely eliminate the period of adaptation. These results suggest that reduction of VEGF is not the sole reason for the limited formation of tumor vasculature in the shPrx1 expressing tumors. Prx1 reduction also leads to lower expression of other angiogenic factors including IL-6 and TGF-β, which can also influence angiogenesis in CaP (36).

Numerous studies have implicated bacteria-associated immune responses as mediators of prostate carcinogenesis (3, 37). This hypothesis is supported by studies showing that TLR4 can contribute to CaP cell proliferation and invasion (4, 5) and to CaP risk (6). However, not all incidences of CaP are associated with a prior pathogen infection, suggesting the presence of a host-derived TLR4 ligand that contributes to the maintenance of a chronic inflammatory and protumorigenic environment, initiation of an angiogenic switch, and promotion of CaP growth. Our findings indicate that Prx1 expression is increased at the time of transformation, suggesting that Prx1 may be a link between inflammation and prostate carcinogenesis.

The role of Prx1-TLR4 interaction in CaP in our experiments was largely shown by inhibition of MyD88 signaling. All TLRs, except TLR3, signal through MyD88 (38) and genetic variations of TLR1, TLR6, and TLR10 resulting in increased activity are associated with increased prostate cancer risk (39). Therefore, the effects of Prx1 on CaP growth may also be owing to interactions with these receptors. However, VEGF secretion in response to Prx1 was lost in the absence of TLR4 signaling and our previous study showed that Prx1 seems to interact with solely TLR4 (13).

Prx1 was first identified as a tumor suppressor (40, 41) because Prx1 deficient mice have increased susceptibility to sarcomas and blood malignancies (42). However, Prx1 expression is increased in numerous cancers and is correlated with poor prognosis (8, 12, 43–49). Our findings indicate that Prx1 is elevated during CaP progression.

The tumor suppressive activity of Prx1 was attributed to its intracellular interactions with the known tumor suppressors, c-myc and PTEN (40, 50) and is associated with its intracellular peroxidase activity. Our studies indicate that promotion of CaP by Prx1 is owing to its extracellular chaperone activity and interaction with TLR4. Thus, the role of Prx1 in carcinogenesis may depend upon its location and activity. In this scenario, Prx1 acts as a tumor suppressor by controlling the activity of PTEN and c-myc prior to transformation. Upon transformation Prx1 is secreted and interacts with TLR4 to promote a protumorigenic microenvironment, stimulate angiogenesis, and tumor growth.

Prx1 may provide a new therapeutic target for treatment of CaP. Prx1 interaction with TLR4 and subsequent regulation of angiogenesis and VEGF production are dependent on its chaperone activity and/or decameric structure (13). Targeted inhibition of Prx1 decameric structure/chaperone activity with small molecule inhibitors may reduce chronic inflammation, inhibit VEGF expression, and limit angiogenesis in prostate carcinogenesis.

No potential conflicts of interest were disclosed.

Drs. James Mohler and Saroop Sing (Department of Urology) graciously provided the analysis program used for quantization of CD31 expression. Dr. David Gold, Department of Bioinformatics and Biostatistics, University at Buffalo assisted in the statistical analysis of the TMA.

This work was supported by NIH Grants CA109480 (SOG), CA98156 (SOG), CA095367 (BAF), CA111846 (BAF) and in part by the Roswell Park Cancer Center Support Grant CA16056.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.