In Vivo Hemin Conditioning Targets the Vascular and Immunologic Compartments and Restrains Prostate Tumor Development

This article is featured in Highlights of This Issue, p. 4945

Prostate cancer is the second most common cancer in men worldwide (1). Although most current therapies against this disease are designed to target the tumor cells, the surrounding microenvironment plays a leading role in enabling tumor development (2). Novel cancer therapies should consider the crosstalk between epithelial and stromal compartments, which has been reported to promote tumor progression by remodeling the extracellular matrix to enhance invasion and angiogenesis, releasing soluble factors and disarming the antitumor immune surveillance (3). Understanding the competing interactions between the several pro- and antitumorigenic components that shape the complex milieu of the tumor microenvironment, could lead to more integral approaches for cancer treatment (4).

Chronic inflammation has been associated with a high cancer incidence (5), providing clear evidence that a deregulated microenvironment affects tumorigenesis. An inflammatory setting fosters tumor progression through a wide range of mechanisms and represents a decisive factor in its evolution (6, 7). Of note, inflammation-driven anatomic expansion and increased activation of the remodeled microvascular bed promotes angiogenesis and further influx of immune cells, which become codependent processes (8).

Several molecular pathways have been linked to cancer and inflammation (9). In particular, the enzyme heme oxygenase-1 (HO-1) is part of an endogenous defense system implicated in the homeostatic response (10, 11). The intrinsic effect of HO-1 on tumor cells in different cancer models has been extensively addressed (12, 13). Furthermore, data are available from a wide spectrum of physiopathologic conditions that link HO-1 to modulation of angiogenesis and the immune function, 2 hallmarks of cancer (7).

In prostate cancer, we have demonstrated that HO-1–overexpressing human xenografts generated in nude mice show impaired growth (14) and angiogenesis (15) and that this protein modifies the bone microenvironment modulating prostate cancer bone metastasis (16). We recently reported that HO-1 shapes cell–cell interactions, favoring a less aggressive phenotype (17, 18). Moreover, HO-1 inhibited relevant pathways implicated in prostate tumorigenesis (16, 19). Other groups have also provided evidence showing that HO-1 finely tunes prostate cancer progression by exerting both protumor (20, 21) and antitumor roles (22). However, despite considerable evidence regarding the role of this enzyme in the epithelial tumor cell compartment, its role in the tumor microenvironment still remains elusive.

Conditioning strategies constitute a relatively unexplored and exciting opportunity to shape tumor fate by targeting the tumor microenvironment. In this study, we assessed whether conditioning with hemin, known to induce HO-1, affects prostate cancer development using an immunocompetent murine model. Hemin treatment prior to tumor challenge resulted in a significant increase in tumor latency by targeting both tumor vascularization and cytotoxic T-cell responses. Taken altogether, these data showcase a novel function of an already human-used drug as a treatment to boost the endogenous antitumor response.

TRAMP-C1 cells (T-C1; ATCC) were cultured in DMEM (Invitrogen), 10% FBS (Gibco), antibiotic–antimycotic (Gibco), and insulin (5 μg/mL). Cell morphology, androgen sensitivity, and mycoplasma contamination were routinely assessed. Human umbilical vein endothelial cells (HUVEC; Lonza) were maintained in EGM-2 (Lonza). Bovine aortic endothelial cells (BAECs) were provided by M.T. Elola (IQUIFIB, UBA-CONICET; UBA-FFyB-QB; Buenos Aires, Argentina) and cultured as described previously (23). Lymph node cell primary cultures were carried out in RPMI1640 (Invitrogen) containing 10% FBS (PAA), antibiotics, 2 mmol/L l-glutamine, and 2 × 10⁻⁵ mol/L β-mercaptoethanol. Hemin (Sigma-Aldrich) was dissolved in 1 mol/L Tris-HCl, pH 8, 0.5 mol/L NaOH, and PBS. This solution was 22-μm filtered and diluted in PBS or culture media.

Animal procedures complied with institutional guidelines. Six- to 8-week-old male C57BL/6 mice were housed in the animal facility of the FCEN-UBA. Foxn1^nu mice were acquired from the animal facility of UNLP. T-cell receptor transgenic mice specific for H-2Kb OVA_257-264 (male OT-1 mice) were raised and tested at FIL-IIBBA-CONICET.

Mice were subcutaneously injected with hemin (200 μL, 30 μmol/L) on days −8, −5, and −1 prior to challenge with 2 × 10⁶ T-C1 cells on the same flank. Alternatively, hemin was administered on the contralateral flank when specified. Control littermates were injected with PBS. T-C1 cells were subcutaneously injected in Matrigel (4–5 mg/mL; Corning). No changes in weight were detected. Euthanasia was practiced at the most when tumor volume reached 1,500 mm³. Tumor size was calculated as W² × L/2 (W = width, L = length). Tumor volume was normalized to that at the start day of exponential growth.

Transcriptional profile was analyzed in T-C1 tumor samples and lymph node samples, as described previously (23). Primers are listed in Supplementary Table S1. Human PPIA, murine Rplp0, and bovine GAPDH were used as internal reference genes.

Sections obtained from paraffin-embedded tissues were stained according to Masson trichrome technique or subjected to IHC, as described previously (14, 23), using anti-HO-1 (ab13243, Abcam), anti-Gal-1 (H-45, Santa Cruz Biotechnology, Inc.), and anti-CD31 (D8V9E, Cell Signaling Technology) antibodies. Blinded qualitative studies were carried out by a pathologist (R.P. Meiss).

Patient-derived xenograft (PDX) were generated at MD Anderson Cancer Center (Houston, TX; ref. 24) to prepare the TMA (n = 50). IHC was carried out and blinded semiquantitative studies were performed (R.P. Meiss): 0, no staining; 1, 2, and 3, low, mild, and high staining, respectively. Scores of individual blokes corresponding to the same PDX were averaged.

Eighty-percent confluent T-C1 cells were cultured for 24 hours in a 1:5 diluted growth media. Conditioned media (CM) was harvested and filtered. Sixty microliters of growth factor–reduced Matrigel was plated in a 96-well plate and incubated at 37°C for 15 minutes. A total of 12.5 × 10³ hemin-treated (50 μmol/L, 8 hours) or control HUVECs were plated on the Matrigel in the presence of control or T-C1–derived CM. Positive control wells were seeded in EGM-2 medium. Endothelial tube formation was evaluated after 18 hours. Five fields per well were photographed. AdobePhotoshop-processed photographs were evaluated using the NIH ImageJ Angiogenesis Analyzer Plug-in.

BAECs were treated with hemin (50 μmol/L, 8 hours), washed, and subsequently cultured for 16 hours in a 1:5 diluted growth media containing recombinant Gal-1 (ref. 25; 0, 2.25, 4.5, and 9 μmol/L). Detached and adherent cells were photographed, harvested (TripLE Express, Thermo Fisher), and stained with FITC-Annexin-V (Apoptosis Detection kit, BD Pharmingen) and propidium iodine (Sigma), followed by flow cytometry.

HUVECs were treated with hemin (50 μmol/L, 8 hours), washed, and subsequently cultured for 24 hours in a 1:5 diluted growth media. CM was harvested and filtered. Confluent T-C1 cells were washed and serum-starved for 8 hours. A 1-mm-wide scratch was made across the cell layer and, after washing with serum-free medium twice, hemin-treated or control HUVEC CM was added, and plates were photographed immediately and after incubation (12, 24, 48 hours) at the identical location of initial image. Wound area was quantified using NIH ImageJ.

Eighty-percent confluent HUVECs were treated with hemin (50 μmol/L, 8 hours) and subsequently washed. A total of 3 × 10⁴ CFSE-labeled T-C1 cells (2.5 μmol/L, 5 minutes, 1 × 10⁶ cells/mL in PBS 1% FBS; Sigma) were added. After incubating 2 hours at 37°C, unattached cells were washed away, and 3 different fields of each well were photographed. Cells were counted using the NIH ImageJ Cell Counter Plug-in.

A total of 2 × 10⁶ T-C1 cells in 500 μL of Matrigel (4–5 mg/mL) were subcutaneously injected into mice following hemin conditioning. Five days later, Matrigel plugs were harvested and photographed. Plugs were homogenized in H₂O and cleared by centrifugation. Hemoglobin and total protein content were determined using the Drabkin reagent (WienerLab) and the Pierce BCA Protein Assay kit (Thermo Scientific).

For T-cell proliferation assays, 5 × 10⁵ CFSE-stained murine lymph node cells (2.5 μmol/L, 5 minutes) were seeded in a 96-well U-bottomed plate. Hemin was added into culture (18.75, 37.5, and 75 μmol/L). A total of 1 × 10⁴ mitomycin-arrested T-C1 cells (3-hour treatment, 10 μg/mL; Sigma) were added into cultures to mimic a tumor microenvironment. For T-cell degranulation, cells were stimulated with coated anti-CD3 antibody (145-2C11 hybridoma; 1 μg/mL, 12 hours) and anti-CD107a was added (1 hour, 37°C; 1D4B, BD Pharmingen) followed by an additional 4-hour culture with monensin (3 μmol/L; BD Pharmingen). Spontaneous expression of CD107a was also evaluated. For T-cell proliferation, cells were stimulated with coated anti-CD3 antibody (1 μg/mL, 72 hours) and proliferation assessed by CFSE dilution. Cells were stained for CD8 (53-6.7, BD Pharmingen) in staining buffer (PBS 1% FBS, 0.01% sodium azide; 30 minutes on ice). HO-1 induction was confirmed by intracytoplasmatic staining (ab13248; Abcam). FACS was performed in a FACSAria (BD Biosciences) using the FlowJo software.

WT C57BL/6 mice were irradiated (1 Gy; ¹³⁷cesium source; Cebirsa), transferred with 3 × 10⁶ OT-1 lymph node cells, and subcutaneously injected with hemin on a daily basis for 3 days. On the next day, freshly isolated WT C57BL/6 spleen cells were obtained by mechanical disruption, and erythrocytes were lysed. Splenocytes were stained with CFSE (5 μmol/L or 0.5 μmol/L). CFSE_dim cells were incubated with OVA-derived SIINFEKL peptide (10 μg/mL, 30 minutes, 37°C; Invivogen); CFSE_bright cells were exposed to the vehicle alone. Mice were transferred with a 1:1 mix of CFSE_dim-OVA and CFSE_bright-ctrl cells (2 × 10⁶ cells, 100 μL). After 16 hours, spleens were processed for FACS analysis. Alternatively, hemin treatment was carried out ex vivo (75 μmol/L, 8 hours) prior to adoptive transfer. Mice with no previous transfer of OT-1 lymph node cells were used as a control to calculate specific cytotoxicity as follows: SC = [1 − (%CFSE_dim-OVA/%CFSE_bright-ctrl)_{transferred mice} × (%CFSE_bright-ctrl/%CFSE_dim-OVA)_{nontransferred mice}] × 100.

Tumor, spleen, tumor-draining lymph node (TDLN; axillary, brachial, and inguinal), and blood vessel–containing tissue samples were harvested and single-cell suspensions were obtained by mechanical disruption. Tumor samples were previously disaggregated with Collagenase type IV (1 mg/mL, 1 hour, 37°C; Sigma). Cells were stained and analyzed by FACS. Anti-CD3-FITC (17A2; BD Biosciences), anti-CD11b-PE (M 1/70; BD Biosciences), anti-Gr-1-FITC (RB6-8CS; BD Biosciences), anti-CD31 (390, BD Pharmingen), anti-HO-1, goat-anti-rat Ig-PE (550767, BD Biosciences) goat-anti-mouse IgG1-Alexa488 (A21121, Thermo Fisher Scientific) were used.

We searched Oncomine (26) to identify human expression microarray datasets that compare prostate adenocarcinoma versus normal gland. Supplementary Table S2 shows information about these datasets. Cited literature was reviewed to confirm that the analysis was as documented.

At least 3 biological replicates (i.e., independent experiments) were carried out. Data represent mean ± SD. GraphPad software was used. Two groups were compared with the Student t test for unpaired data. Two-way ANOVA tests were used for multiple comparisons (with Tukey and Bonferroni post hoc tests). Nonparametric Mantel–Cox tests were carried out when comparing tumor-free mice curves. P < 0.05 was considered statistically significant.

Given the scarce prostate cancer preclinical models available allowing assessment of the whole tumor microenvironment, we used a syngeneic model based on the subcutaneous injection of T-C1 cells into C57BL/6 mice, previously standardized in our laboratory (27). TRAMP cells constitute the only murine prostate cancer cells lines appropriate for implantation into syngeneic immunocompetent mice. To study whether hemin conditioning influences tumor development, mice (n = 5) were subcutaneously injected with this agent (200 μL; 30 μmol/L) on days −8, −5, and −1 prior to tumor challenge on the same flank (2 × 10⁶ T-C1 cells in Matrigel; Fig. 1A). Control littermates (n = 5) were injected with saline. Hemin dosage was not arbitrary chosen, it was adapted from the dose used for human porphyria. Pharmacokinetics, pharmacodynamics, and toxicologic studies have been performed in animals prior to hemin acceptance for use in humans. Although tumors developed in all the mice, pretreatment resulted in a marked increase in tumor latency (57 ± 3 days in hemin-treated mice vs. 47 ± 5 days in control mice; Fig. 1B, left). Furthermore, initial growth rate was significantly reduced in hemin-treated animals (Fig. 1B, right). Tumor histologic analysis at the time of euthanasia showed lower number of blood vessels in the experimental group, as revealed by both trichrome staining and CD31 IHC (Fig. 1C). Evaluation of mediators typically involved in angiogenesis [VEGF-A, CD142, uPA, FGFb, TIMP-1, TSP-1, and galectin-1 (Gal-1)] revealed that only the Gal-1-coding mRNA (Lgals1) expression was significantly reduced in hemin-treated animals (Fig. 1D). Remarkably, our group and others have previously reported an upregulated expression of this lectin in human prostate cancer and its correlation with the tumor angiogenic phenotype (23, 28).

Considering our previous findings that HO-1 and Gal-1 expression play a critical role in prostate cancer neovascularization (15, 23), we analyzed the effect of hemin on endothelial cells within a prostate tumor setting. HO-1 induction was confirmed by qRT-PCR (Hmox1 mRNA; Fig. 2A). An in vitro tube formation assay was performed using hemin pretreated HUVECs, seeded onto Matrigel-coated wells in the presence or absence of T-C1–derived CM. Hemin pretreatment dramatically inhibited tubulogenesis only in the presence of tumor CM (Fig. 2B; Supplementary Fig. S1A), as reflected by a significant reduction of total tube length (Fig. 2C) and the number of master segments and tube joints in this experimental condition (Supplementary Fig. S1B). Importantly, the tube network was not hindered both in complete growth media and in the absence of T-C1–derived CM, confirming HUVEC viability is not significantly affected by hemin pretreatment. To evaluate the role of Gal-1, we carried out an apoptosis assay in the presence of the recombinant lectin. While control cells did not show Gal-1–induced apoptosis, hemin-treated endothelial cells were significantly more susceptible to cell death in a Gal-1 dose-dependent fashion (Fig. 2D). Of note, hemin-treated cells did not show any significant differences in terms of basal cell death when compared with control cells. The increase in endothelial cell death was further supported by a significant decrease in CD146 mRNA levels (Supplementary Fig. S2), a coreceptor for VEGFR-2 and a decoy receptor for Gal-1 (29, 30). We next studied whether hemin conditioning influences the crosstalk between tumor and endothelial cells. T-C1 cell motility was significantly impaired when a wound-healing assay was performed in the presence of hemin pretreated HUVEC CM (Fig. 2E). Tumor cell adhesion to endothelial cells was severely reduced when HUVECs were pretreated with hemin (Fig. 2F). We ruled out there were hemin traces in the CM by assessing HMOX1 mRNA expression levels in T-C1 that had been exposed, and detected no significant differences (data not shown). Collectively, these findings suggest that hemin conditioning remodels the endothelial compartment and negatively regulates interactions between endothelial and tumor cells. To assess whether hemin pretreatment is implicated in tumor neovascularization in vivo, we performed a Matrigel plug assay. C57BL/6 mice (n = 5) were subjected to our hemin-conditioning protocol before being injected with T-C1 cells in Matrigel. Control mice (n = 5) were injected with saline-containing Matrigel. After 5 days, plugs were harvested and hemoglobin content was determined. In keeping with our in vitro findings, hemin pretreatment significantly inhibited vascularization in vivo (Fig. 2G). In summary, hemin conditioning impacts upon the angiogenic process, probably accounting for tumor growth inhibition (Fig. 1B). To verify that hemin conditioning induces HO-1 expression in endothelial cells, on the day that mice would have been challenged with tumor cells, we collected samples of blood vessel–containing tissue from the same animal flank that had been previously injected with either PBS (n = 6) or hemin (n = 5). We subsequently assessed HO-1 expression in endothelial cells using FACS and we confirmed that hemin treatment in vivo results in increased HO-1 expression in CD31⁺ cells (Fig. 2H).

Both HO-1 and Gal-1 have been reported to play major roles in shaping antitumor responses (31, 32). We thus analyzed hemin effect upon the immune compartment by performing in vitro cocultures of lymph node cells isolated from C57BL/6 mice with syngeneic T-C1 cells. Hemin induced HO-1 in a dose-dependent fashion in CD8⁺ cytotoxic T lymphocytes (CTLs), as determined by intracellular staining and flow cytometry (Fig. 3A). FACS analysis of cell division by CFSE staining demonstrated that hemin increases CD8⁺ T-cell proliferation in a dose-dependent manner in response to polyclonal activation (Fig. 3B, left), even when cocultured with tumor cells (Fig. 3B, right). We also examined hemin effects on the cytotoxic potential of CD8⁺ T cells, by measuring the expression of CD107a on the cell surface as a result of CTL degranulation. Hemin treatment resulted in enhanced CTL effector function both in normal and tumor microenvironments (Fig. 3C). To evaluate whether hemin also boosts the immune function in an established tumor immunosuppressive setting, lymph node cells and T-C1 cells were cocultured 24 hours prior to hemin addition. Hemin augmented CD8⁺ T-cell proliferation and degranulation even under these experimental conditions (Fig. 3D and E), reverting tumor immunosuppression. To verify the relevance of these findings in vivo, we tested cytotoxicity following hemin conditioning by performing a CTL assay. CD8⁺ T-cell effector function in vivo was determined by mixed transfer of OVA peptide–pulsed target cells with control cells into WT C57BL/6 mice previously transferred with OT-1–derived lymph node cells (n = 5; Fig. 3F). We confirmed that hemin treatment induced HO-1 expression at the transcriptional level in lymph node samples collected on the day that mice would have been transferred with the mix (Fig. 3G). The left panel of Fig. 3H shows representative results, where the relative frequencies between peptide-pulsed target cells and control cells were used as readouts of specific killing. The frequency of CFSE_dim-OVA cells was reduced to a greater extent in hemin-treated mice compared with control littermates. OVA-specific cytotoxicity was significantly enhanced when transferred mice were subcutaneously treated with hemin prior to challenge with OVA-loaded cells (Fig. 3H, right). We next sought to determine whether hemin boosts antigen-specific cytotoxicity in vivo if OT-1 lymph node cells were treated ex vivo prior to adoptive transfer into WT C57BL/6 mice. Hemin-treated lymphocytes led to augmented OVA-specific cytotoxicity (Fig. 3I). Our results indicate that the sole exposure of lymph node cells to hemin boosts CD8⁺-mediated cytotoxicity, which could account for tumor growth inhibition (Fig. 1B).

To investigate whether hemin conditioning influences vascular and immunologic compartments at earlier stages of tumor development, tumor-bearing C57BL/6 mice (n = 4) were sacrificed when control mice exhibited tumors in their exponential growth phase, while hemin pretreated mice showed tumors in their initial growth steps. The frequency of CD3⁺CD8⁺ cells in hemin-conditioned mice was systemically increased, as determined by immunophenotyping the tumor, draining lymph nodes, and spleen (Fig. 4A). Furthermore, the frequency of regulatory Gr-1⁺CD11b⁺ myeloid cells in the tumor-draining lymph nodes was reduced in hemin pretreated mice compared with control animals (Fig. 4B), probably allowing a more effective antitumor immune response. Histologic analysis revealed that control tumors, but not tumors derived from hemin-conditioned mice, showed dilated blood vessels mimicking an angiomatous image (Fig. 4C, top and centre). In addition, hemin pretreated animals displayed tumors with reduced atypical mitotic figures when compared with control mice (Fig. 4C, bottom). IHC revealed lower levels of Gal-1 expression in tumors generated under hemin conditioning (Fig. 4D). These results further emphasize the essential role of the tumor microenvironment in governing the fate of tumor development. Considering the systemic changes in conditioned mice (Fig. 4A and B), we next evaluated whether hemin exerts an antitumor effect when injected at a site other than the flank subsequently challenged with the tumor cells. Similar to hemin conditioning on the ipsilateral flank, its administration on the contralateral flank also led to a significant increase in tumor latency (45 ± 5 days in control mice, n = 8; 56 ± 3 days in ipsilateral conditioned animals, n = 10; and 55 ± 2 days in the contralateral-conditioned group, n = 5; Fig. 4E). No significant differences were found between both conditioning strategies in terms of tumor latency, suggesting both treatments had a similar antitumor effect. Furthermore, in an effort to elucidate the relative contribution of the vascular and immune compartments, we tested whether hemin conditioning was able to modulate tumor development in athymic nude mice. Tumor latency and growth did not significantly differ when comparing hemin-conditioned animals with control mice (Fig. 4F), confirming that the immune system is required to target prostate cancer tumors in our experimental model.

To address potential clinical implications of elevated HO-1 expression in human prostate tumor samples, we searched the microarray database Oncomine (26). We found 16 gene expression microarray datasets comparing prostate adenocarcinoma versus normal prostate tissue that met our eligibility criteria (Supplementary Table S2). Only 4 datasets presented increased expression for HMOX1 mRNA with P < 0.05 (Table 1), in which only 1 of these datasets showed fold induction greater than 1.5. These results support the idea that HO-1 expression is not associated to prostate carcinogenesis. Given the small P values obtained for these 4 datasets (n = 122, n = 112, n = 150, n = 19; refer to Supplementary Table S2), we extended our analysis to assess the expression levels for LGALS1. Table 1 sums up the results obtained with all the datasets, whereas Supplementary Fig. S3 depicts representative results obtained when assessing Grasso Prostate Statistics (33). Interestingly, the expression profile for LGALS1 showed significant downregulation for prostate adenocarcinoma versus normal gland in samples where HMOX1 was upregulated (Table 1; Supplementary Fig. S3). We assessed in these same datasets the expression profiles of different angiogenesis-related genes (CD34, FLT1, FLT4, KDR, VEGFA, and VEGFC). Accordingly, their expression levels were downregulated for the prostate adenocarcinoma versus normal gland comparisons in which HMOX1 was upregulated (Table 1; Supplementary Fig. S3). On the contrary, these datasets with increased expression of HMOX1 also exhibited VCAM1, CD28, CD80, and CD86 upregulation in prostate adenocarcinoma compared with the normal gland (Table 1; Supplementary Fig. S3). Interestingly, more intense lymphocyte infiltration was observed in tumors with elevated VCAM-1 (34). Furthermore, CD28, CD80, and CD86 constitute key costimulatory molecules required to mount an effective immune response, and their elevated expression levels are indicative of T-cell activation. These results support the aforementioned findings and point toward a subset of patients who might respond to antiangiogenic and/or immunostimulatory therapy.

To further understand potential clinical implications, we screened a TMA containing 50 PDXs, each represented by PDXs grown in 3 different mice and 3 cores per tumor. These xenografts were derived from primary prostate cancer and metastatic sites and encompass samples of advanced prostate cancer with various histopathologic patterns. PDXs with heterogeneous staining between different cores or tumors for either HO-1 or Gal-1 were excluded from the analysis. PDXs were classified according to HO-1 immunostaining and we studied in detail those with low-to-intermediate HO-1 staining. This subset of PDXs displayed 2 clear clusters with either low or high Gal-1 immunostaining (Table 2). Those PDXs with mild HO-1 and low Gal-1 expression could be potentially more susceptible to an antiangiogenic and/or immune therapy, and could even benefit from the antiangiogenic properties of HO-1 induction (15).

Carcinogenesis is a multistep process encompassing mechanisms in both the epithelial cell and the surrounding tissue. Numerous signaling pathways are involved in the dynamic crosstalk between these compartments, and their deregulation influences disease initiation, progression, and patient prognosis (2). Novel experimental strategies have been designed to reeducate the microenvironment to shift the balance toward an antitumorigenic profile (4). Here we designed an experimental protocol using a murine model to shape the niche where prostate tumor cells are subsequently implanted (Fig. 1). We demonstrate for the first time that hemin conditioning remodels the bidirectional interactions between the tumor and the surrounding tissue, affects tumor neovascularization and immune function, and impairs tumor growth. Our results indicate that hemin conditioning acts, at least in part, through the inhibition of angiogenesis and the potentiation of tumor immunity, thus restraining prostate cancer development. Our findings also demonstrate that hemin modulated tumor development when injected on the contralateral flank (Fig. 4E). This strategy led to an increase in tumor latency that was comparable with that corresponding to ipsilateral hemin conditioning, further supporting potential novel therapeutic avenues for prostate cancer. Of note, hemin is approved for use in treatments against acute intermittent porphyria.

Angiogenesis is essential for tumor development given that proliferation of cancer cells, as well as their metastatic spread, rely on an adequate supply of both oxygen and nutrients and the removal of waste products. We demonstrate that hemin conditioning inhibits tumor-associated neovascularization in vitro and in vivo (Figs. 2 and 4). Hemin pretreatment of HUVEC dramatically hindered tubulogenesis only in the presence of tumor CM, suggesting hemin-treated HUVECs are refractory to proangiogenic tumor-derived factors. We also found that hemin conditioning limits the tumor angiogenic process in vivo. We have previously documented the antiangiogenic properties of HO-1 induction in the prostate cancer epithelial compartment (15). Moreover, this antiangiogenic effect was also demonstrated in pancreatic cancer by exogenous administration of the HO-1 metabolic product CO (35). In line with these observations, Ahmad and colleagues (36) recently reported that CO exposure inhibits VEGF-induced endothelial cell proliferation, migration, and capillary-like tube formation. Furthermore, in our preclinical model we found that Gal-1 expression was significantly reduced in tumors generated in hemin-treated mice. Our previous findings have shown that this lectin is highly expressed and regulated in the prostate cancer microenvironment and plays a major role in prostate cancer neovascularization (23). Notably, Croci and colleagues (25, 37) elegantly demonstrated the existence of a glycosylation-based circuit in which Gal-1 interaction with VEGFR2 mimics VEGF-A function and induces angiogenesis. Here we confirmed that Gal-1 promotes the formation of tube-like structures in vitro. Strikingly, hemin treatment sensitized endothelial cells to apoptosis in a Gal-1–dependent dose (Fig. 2). It has been previously reported that endothelial cells may undergo Gal-1–induced apoptosis when a potential coreceptor for VEGFR-2, CD146, is downregulated (29, 30); in this regard, we have indeed confirmed that hemin treatment upon endothelial cells inhibits CD146 expression. Moreover, taking into consideration the prolonged inhibition of Gal-1 expression in tumors evidencing impaired growth, our data might indicate that hemin conditioning remodels the endothelial compartment, affecting its interaction with the tumor cell. However, other studies have reported proangiogenic roles for HO-1 and its metabolic products in different cancer models and other pathologies (reviewed in refs. 12, 38), providing evidence of the inconsistent function of HO-1 in tumorigenesis and unveiling the complexity of the angiogenic process. Thus, the role of this protein is far from clear, but the general consensus of opinion is that its effect is highly dependent on tumor type, the trigger of tissue-specific signaling pathways, and the relative contribution of the enzymatic products and the noncanonical roles.

Nevertheless, simultaneous targeting of multiple components of the tumor microenvironment will be required to achieve a durable and efficient antitumor response. Our results indicate that hemin conditioning can also boost the immune response (Figs. 3 and 4). Hemin treatment resulted in enhanced CD8⁺ T-cell proliferation and degranulation but, more importantly, we have provided strong evidence that subcutaneous administration of this agent augments in vivo antigen-specific cytotoxicity. Outstandingly, assays encompassing ex vivo lymph node hemin treatment prior to adoptive transfer supported hemin capacity to boost the CD8⁺ cytotoxic function, which might in turn account for T-C1 tumor growth impairment. Indeed, we also observed a significant systemic expansion of CD8⁺ T cells, suggesting that hemin conditioning may trigger long-term immunologic effects. Although the immunomodulatory function of HO-1 is well documented and its immunosuppressive role is widely accepted (31), relatively little attention has been paid to the effect of its substrate heme on the immune response. Remarkably, heme directly regulates various molecular and cellular processes, has been shown to act as a T-cell mitogen in vitro (39) and exerts mild proinflammatory responses encompassing the activation of macrophages and polymorphonuclear cells (40–42). Furthermore, recent findings have shown that oral administration of heme prior to induction of necrotizing enterocolitis decreased disease incidence and increased Treg/Teff ratios (43). In addition, Konrad and collaborators (44) demonstrated that topical administration of hemin was protective against pulmonary inflammation. Both studies have clearly associated the reported effects to HO-1 expression by using genetically modified animals. Two independent studies have shown that HO-1 expressed in macrophages may influence prostate cancer tumor fate and promote its growth (45, 46). Surprisingly, Jais and collaborators (47) claimed to have broken the HO-1 dogma by providing evidence of its proinflammatory role in hepatocytes and macrophages in the development of metabolic disease, using conditional genetically modified mice and patient datasets. In addition, Mashreghi and colleagues (48) showed that pharmacologic agents usually associated to HO-1 modulation have the capacity to shape the immune response independently of this protein. Thus, we may conclude that under our preclinical model hemin could trigger a complex immune response, characterized by augmented CD8⁺ T-cell function. Finally, Gal-1 downregulation in tumors grown in hemin-conditioned animals might reflect potentiation of a tumor-specific immune response, as several findings indicate that this lectin plays critical protumorigenic roles within the tumor microenvironment, in part by suppressing T-cell–mediated cytotoxic responses (reviewed in refs. 32, 49). The relevance of the immune system in our experimental model becomes plainly evident when immunodeficient mice were conditioned with hemin prior to tumor challenge, as no significant differences were found. Although immunodeficient mice provide valuable information in cancer research, our findings add to cumulative data in literature that urge for fully immunocompetent and syngeneic animal models.

Finally, to assess the potential clinical implications of both HO-1 and Gal-1 in prostate cancer, we searched the cancer database Oncomine. Interestingly, a subset of patients displays mild upregulation of HO-1 with significantly lower expression for Gal-1, when comparing prostate adenocarcinoma versus normal gland (Table 1; Supplementary Fig. S3). This was further evidenced by the analysis of PDXs, which also showcased a subgroup with mild HO-1 and low Gal-1 expression (Table 2). This subpopulation could potentially be more susceptible to hemin treatment as an adjuvant therapy, maintaining Gal-1 expression levels severely low and, in combination with other antiangiogenic and/or immune therapies, might interrupt tumor growth and/or spread.

To summarize, our work provides evidence of the effect of hemin conditioning in both vascular and immune compartments, and its implications on prostate cancer development. Our strategy was conceived to remodel the environment to counteract the effects of tumor escape mechanisms. We speculate that novel therapeutic interventions should be designed to specifically target the reactive stroma. As mentioned above, we have previously shown that hemin treatment in prostate cancer cell lines does not have toxic effects and it even impairs proliferation, migration, and invasion in vitro and HO-1 overexpression leads to smaller and less vascularized tumors in vivo (14, 15). It should be noted that there is a considerable number of reports showcasing the protumor role of HO-1 in different types of cancer, even in prostate cancer (20, 21). Consequently, further study should be carried out to elucidate the HO-1 function in cancer development. We are currently developing new experimental strategies to assess hemin effects on prostate tumor progression in vivo, including intratumor injections and other means of specifically targeting immune and endothelial cells infiltrating the tumor. The results from this article provide a solid proof of concept to move forward in this direction.

No potential conflicts of interest were disclosed.

Conception and design: L. Gentilini, A. Ahmed, D. Compagno, D.J. Laderach, E.S. Vazquez

Development of methodology: F.M. Jaworski, R.P. Meiss, N.M. Navone, D. Compagno, D.J. Laderach

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Gueron, R.P. Meiss, P.M. Berguer, D.J. Laderach, E.S. Vazquez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.M. Jaworski, G. Gueron, R.P. Meiss, E.G. Ortiz, A. Ahmed, N.M. Navone, D.J. Laderach, E.S. Vazquez

Writing, review, and/or revision of the manuscript: F.M. Jaworski, L. Gentilini, G. Gueron, P.M. Berguer, A. Ahmed, G.A. Rabinovich, D. Compagno, D.J. Laderach, E.S. Vazquez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.A. Rabinovich

Study supervision: R.P. Meiss, A. Ahmed, D.J. Laderach, E.S. Vazquez

We are grateful to the Prostate Cancer Foundation for the Young Investigator Award 2013 given to G. Gueron, and to the European Molecular Biology Organization for the short-term fellowship awarded to F.M. Jaworski.

This work was supported by grants ANPCyT PICTRAICES 2013-0996 (to E.S. Vazquez), ANPCyT PICT2012-1533 (to D. Compagno), and ANPCyT PICT2014-0983 (to D.J. Laderach). Argentine National Cancer Institute (to E.S. Vazquez).

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.