Experimental cancer therapy using restoration of NAD⁺-linked 15-hydroxyprostaglandin dehydrogenase expression

Preclinical and clinical evidence shows that prostaglandins (PG) play an important role in the growth of tumors. Prostaglandin endoperoxide synthase 2 (cyclooxygenase-2; Cox-2) is the rate-limiting enzyme involved in the oxidative transformation of arachidonic acid into PGH₂, which represents the precursor of several bioactive molecules, including PGE₂, PGD₂, PGF₂α, PGI₂ (prostacyclin), and TXA₂ (thromboxane A₂; refs. 1, 2). Cox-2 is overexpressed in the majority of epithelial tumors, including breast and colon cancers. Aberrant upregulation of Cox-2 expression is correlated with prognostic markers that reflect a poor chance for survival, which includes tumor size, axillary node metastases, tumor grade, ductal histology, receptor negative disease, and overexpression of HER-2 (3, 4). Furthermore, elevated Cox-2 expression was significantly associated with reduced survival and recognized as an independent prognostic factor in colorectal cancer patients (5). Cox-2 is involved in activation of tumor-associated angiogenesis and in the inhibition of cancer cell apoptosis (6, 7). Moreover, elevated Cox-2 expression has recently been shown to correlate with distant breast cancer metastases (8). Inhibition of the Cox-2 pathway decreased colorectal metastasis growth in the liver (9–11). It was shown that activation of the Cox-2 signaling pathway leads to intestinal epithelial and colon tumor cell proliferation (12).

Growth, progression, invasion, and metastasis of cancers are accompanied by the formation of new blood vessels. Therefore, tumor-associated angiogenesis is an acquired capability of malignant tumors. High levels of PGE₂ expression in tumors are associated with increased angiogenesis (13). It has been shown that PGE₂ induces the expression of several proangiogenesis factors including vascular endothelial growth factor (VEGF; ref. 14) and basic fibroblast growth factor (bFGF; ref. 15) in blood vessel endothelial cells (16). VEGF and bFGF also induce Cox-2–mediated PGE₂ expression in endothelial cells, which raises the possibility of amplification of PGE₂ regulation of angiogenesis through a positive feedback loop (17). The addition of the anti-VEGF monoclonal antibody (mAb) bevacizumab (Avastin) to chemotherapy in the treatment of metastatic colorectal cancer has resulted in an increase of progression-free survival and overall survival (18).

NAD⁺-linked 15-hydroxyprostaglandin dehydrogenase (15-PGDH) catalyzes the oxidation of 15(S)-hydroxyl group of PGE₂ resulting in the formation of 15-keto metabolites with greatly reduced biological activities. Expression of 15-PGDH, a key enzyme responsible for PGE₂ inactivation, is suppressed in the majority of cancers, including breast and colorectal cancers (19–21). Thus, overexpression of Cox-2 and repression of 15-PGDH provide complementary pathways to increase levels of PGE₂ in the tumor microenvironment.

We have developed adenoviral vectors (Ad) encoding the human 15-PGDH gene under control of the VEGF receptor 1 (VEGFR-1/flt-1) and Cox-2 promoter element Adflt-PGDH and Adcox-PGDH, respectively. We investigated whether Ad-targeted 15-PGDH therapy produces cytotoxicity against cancer cells in vitro and in s.c. human breast and colon xenograft models. The results show that Ad-mediated overexpression of 15-PGDH significantly inhibited growth of the human breast and colon xenografts in athymic nude mice. Also, the antitumor effect 15-PGDH gene therapy was enhanced by coadministration of bevacizumab (Avastin).

The 2LMP and LCC6 breast cancer cell lines (both were obtained from Dr. M. Lippman, Georgetown University, Washington, DC), MDA-MB-231 breast cancer cells, colon adenocarcinoma cells LS174T (American Type Culture Collection), and embryonic kidney cells HEK293 (Microbix Biosystems) were cultured in DMEM/F12 (Mediatech) containing 10% fetal bovine serum (Summit Biotechnology). Normal established breast epithelial cells MCF 10A (American Type Culture Collection) were cultured in mammary epithelial growth medium (Clontech) containing 100 ng/mL cholera toxin (Calbiochem). Human umbilical vein endothelial cells (HUVEC) were purchased from Clonetics and were grown in endothelial growth medium-2 growth media (Clonetics). Avastin anti-VEGF recombinant IgG1 mAb was obtained from Genentech.

Replication-deficient E1- and E3-deleted Ad vectors were constructed using standard molecular cloning methods (Supplementary materials).

To determine VEGFR1/flt-1, EP receptor, Cox-2, and 15-PGDH mRNA expression, cells were lysed in TRIzol reagent (Invitrogen), and total RNA was extracted using RNeasy Mini Kit (Qiagen), following standard protocol, and quantified specrophotometrically using a MBA 2000 spectrophotometer (Perkin Elmer). cDNA was synthesized using random hexamer primers and an Omniscript RT kit (Qiagen). The first-strand cDNA was used as the template for PCR. For amplification of cDNA encoding the target genes the following forward (f) and reverse (r) primers were used: flt-1 f: GAG GAT TCT GAC GGT TTC; flt-1 r: CCT GTC AGT ATG GCA TTG; 15-PGDH f: GTA AAG CTG CCC TGG ATG AG; 15-PDGH r: AAC AAA GCC TGG ACA AAT GG; EP1 f: TCT ACC TCC CTG CAG CGG CCA CT; EP1 r: GAA GTG GCT GAG GCC GCT GTG CCG GGA; EP2 f: ATG GGC AAT GCC TCC AAT GAC TCC CAG; EP2 r: CTC CAG GGA ACA ATT TCA AAA T; EP3 f: GAG CAC TGC AAG ACA CAC ACG GAG; EP3 r: GAT CTC CCA TGG TAT TAC TGA CAA; EP4 f: CCT CCT GAG AAA GAC AGT GCT; EP4 r: AAG ACA CTC TCT GAG TCC T; COX-2 f: TGA AAC CCA CTC CAA ACA CAG; and COX-2 r: TCA TCA GGC ACA GGA GGA AG. The human glyceraldehyde 3-phosphate dehydrogenese (GAPDH) cDNA was used as an internal standard for template loading of PCR by using primers: GAPDH f: TCC CAT CAC CAT CTT CCA; and GAPDH r: CAT CAC GCC ACA GTT TCC. PCR was done in the Eppendorf MasterCycler (Eppendorf). PCR product was analyzed by 1% agarose electrophoresis with 0.01 μg/mL ethidium bromide (Bio-Rad) staining.

Cells were infected with 50% tissue culture infectious dose (TCID₅₀)/cell of Adflt-PGDH, Adcox-PGDH, or Adflt-Luc. Cell culture supernatants were collected at 48 h after treatment, centrifuged to remove floating cells, and stored at −20°C. The amount of PGE₂ released in the medium was measured with enzyme immunoassay kit (Cayman Chemical) following the manufacturer's instructions. The raw PGE₂ data were normalized according to the number of viable cells in each sample, determined by trypan blue staining. Cells were counted using a hemacytometer.

Cells were infected with 50 TCID₅₀/cell of Adflt-PGDH, Adcox-PGDH, or Adflt-Luc. For generation of conditioned medium, cells were incubated for 48 h after treatment. Cell culture supernatants were collected and VEGF concentrations were assessed using a quantitative ELISA (R&D systems) following the manufacturer's instructions. The raw VEGF data were normalized according to the number of viable cells in each sample.

To determine VEGFR1/flt-1, Cox-2, and cytomegalovirus promoter activity, cells were infected with 5 TCID₅₀/cell of Adflt-Luc, Adcox-Luc, or Adcmv-Luc, respectively. After 48 h the cells were treated with reporter lysis buffer (Promega), freeze-thawed once, and then these samples were mixed with luciferase assay reagent (Promega) and Luc activity was measured using Berthold Lumat LB9501. Luc activities were normalized for the number of viable cells in each sample and expressed as relative light units.

To investigate the levels of recombinant protein expression, 2LMP cells were infected with 50 TCID₅₀/cell of Adcox-PGDH, Adflt-PGDH, or Adflt-Luc. Cells and conditioned media were collected at 48 h after infection and subjected to Western blot analysis using anti–15-PGDH rabbit polyclonal antibody (Novus Biological). Cells were washed in TBS and homogenized in lysis buffer. Equal amounts of sample were loaded in each lane and separated on SDS-PAGE followed by transfer to a polyvinylidene difluoride membrane. The membrane was blocked with 5% nonfat milk in TBS. The membrane was incubated with specific and secondary antibodies, and then processed and treated with ECL plus Western Blotting Detection System (Amersham Biosciences).

At 3 h after infection with 50 TCID₅₀/cell of Adflt-PGDH, Adcox-PGDH, or Adflt-Luc, cells were diluted to an appropriate cell density, placed into 6-well culture plates, and cultured an additional 15 d. Cells were then fixed and stained with 2% (w/v) crystal violet in 70% ethanol. Colonies comprising ≥50 cells were counted. The plating efficiencies were calculated as the number of colonies divided by the number of test cells plated for each data point. Plating efficiencies were referenced back to the Adflt-Luc–treated control plating efficiency to determine the surviving fraction for Adflt-PGDH and Adcox-PGDH.

Migration assay was done using tissue culture plate inserts with 12.0 μm tissue culture–treated Millcell polycarbonate membranes (Millipore). Cells were infected with 50 TCID₅₀/cell of Adflt-PGDH, Adcox-PGDH, or Adflt-Luc. After overnight incubation the cells were harvested and added to the upper chambers at concentrations of 5 × 10⁴ cells/well. After incubation for 72 h nonmigrated cells on the upper side of the tissue culture plate insert were removed with a cotton swab, and the migrated cells attached to the lower side of the membrane were released using trypsin digest and counted using a hemacytometer.

Female athymic nude mice were purchased from the Frederick Cancer Research Facility and housed under aseptic conditions in microisolator cages, and experiments were carried out according to the protocols approved by the Institutional Animal Care and Use Committee. For tumor cell inoculation the mice were anesthetized by xylazine/ketamine (90/10 mg/kg), and 5 × 10⁶ 2LMP breast cancer cells were injected s.c. into the flanks of 8-week-old mice. When tumors reached 4 to 6 mm in diameter (day 0), groups of 8 to 9 mice were injected intratumorally with PBS, Adflt-Luc, Adflt-PGDH, or Adcox-PGDH at 1 × 10⁸ TCID₅₀/tumor on days 0, 5, 10, and 15.

To investigate the therapeutic efficacy of combination 15-PGDH–mediated therapy and anti-VEGF therapy, 5 × 10⁶ LS174T colon cancer cells were injected s.c. into the flanks of 8-week-old mice. When tumors reached 4 to 6 mm in diameter (day 0), groups of 8 to 9 mice were injected intratumorally with PBS, Adflt-Luc, or Adflt-PGDH at 1 × 10⁸ TCID₅₀/tumor on days 0, 4, and 8. Avastin was administrated i.p. at 5 mg/mouse on days 2 and 6. Tumor size was monitored twice a week using Vernier calipers. Tumor volumes (mm³) were calculated as width² × length × 0.5. Percent change from baseline at day 0 was computed by comparing the baseline value to the tumor size on each day of measurement. The mathematical model was fitted to tumor growth data to estimate tumor doubling time. The end point criterion for the experimental therapy study was a doubling in tumor volume from the initiation of treatment (day 0). Percent change from baseline at day 0 was computed by comparing the baseline value to the tumor volume on each day of measurement.

The formalin-fixed and paraffin-embedded 5-μm-thick sections of tumor samples were incubated overnight with rat anti-mouse CD31/PECAM1 polyclonal antibody (BD Biosciences Pharmingen). The sections were then incubated with donkey anti-rat antibody secondary antibody (Jackson ImmunoResearch Laboratories). Antigen-antibody complexes were visualized by incubation with 3,3-diaminobenzidine substrate and counterstained with diluted Harris hematoxylin. Tumor vessels containing CD31-positive (brown) cells were quantified by microscopy (original magnification, ×200) of at least 5 random fields/tumor, and were calculated as relative microvessel density.

All error terms are expressed as the SD of the mean. Significance levels for comparison of differences between groups in the in vitro experiments were analyzed by Student's t test. The differences were considered significant when P < 0.05. All reported P values are two-sided. In the animal model tumor therapy studies, the treatment groups were compared with respect to tumor size and percent of original tumor size over time. To test for significant differences in tumor size between treatment groups, ANOVA test was conducted. When the ANOVA indicated that a significant difference existed (P < 0.05), multiple comparison procedures were used to determine where the differences lay.

Recent studies have shown a strong interaction between the Cox-2 and the VEGF signaling pathways. In this study, 2LMP breast cancer cells and MCF 10A normal mammary gland epithelial cells were tested for PGE₂ EP1-4 receptor, Cox-2, and VEGFR1 mRNA expression, as well as VEGF and PGE₂ protein secretion. Expression of PGE₂ receptors EP1-4 in 2LMP and MCF 10A cells was investigated by reverse transcriptase-PCR (RT-PCR). As shown in Fig. 1A, 2LMP cancer cells expressed all four types of PGE₂ receptors, whereas MCF 10A cells showed undetectable levels of EP2 and EP3 as well as lower levels of EP1 and EP4 receptor mRNA expression compared with 2LMP cells.

Importantly, MCF 10A cells showed undetectable levels of VEGFR1 expression, and showed lower-level Cox-2 mRNA expression in comparison with 2LMP breast cancer cells as measured by RT-PCR (Fig. 1B). However, RT-PCR analysis revealed a higher level of 15-PGDH expression in MCF 10A cells in comparison with 2LMP cells (Fig. 1B).

Also, quantitative analysis of EP receptors, and Cox-2 and VEGFR1/flt-1 mRNA expression in an expanded panel of breast cancer cells was done using real-time RT-PCR. The results of real time RT-PCR analysis, summarized in Supplementary Table S1 (supplementary materials), are confirmed higher levels of EP receptor, Cox-2, and VEGFR1 mRNA expression in all tested cancer cells in comparison with MCF 10A cells. In contrast, the levels of 15-PGDH expression were significantly higher in normal breast cells in comparison with cancer cells (Supplementary Table S1).

The levels of PGE₂ were significantly higher in conditioned medium of 2LMP breast cancer cells in comparison with MCF-10A cells (30.4 ± 5.1 pg/mL and 11.3 ± 3.5 pg/mL, respectively) measured by ELISA. Similarly, the levels of VEGF evaluated using ELISA were significantly higher in conditioned medium of 2LMP cells (19.7 ± 4.2 ng/mL) in comparison with MCF 10A cells (7.1 ± 2.3 ng/mL).

Initially, cells were infected with Adcmv-Luc, Adflt-Luc, or Adcox-Luc, and the cytomegalovirus, VEGFR1, and Cox-2 promoter activity was evaluated using luciferase (Luc) assay. The levels of VEGFR1- and Cox-2–driven Ad-mediated Luc expression were higher in 2LMP breast cancer cells as well as LS174T colon cancer cells compared with MCF 10A normal mammary gland epithelial cells (Fig. 1C).

To identify Ad-mediated 15-PGDH expression, 2LMP breast cancer cells were infected with Adflt-PGDH, Adcox-PGDH, or Adflt-Luc. The preliminary study showed that infection of 2LMP and LS174T cancer cells with 50 TCID₅₀/cell of Adflt-Luc has a negligible effect on levels of 15-PGDH and Cox-2 expression as well as on the levels of PGE₂ and VEGF secretion in comparison with uninfected control. Thus, for in vitro study Adflt-Luc was used as control. The infected cells were collected at 48 hours after infection and then analyzed by RT-PCR as well as protein expression by SDS-PAGE followed by immunoblotting analysis. The results (Fig. 1D) show higher levels of 15-PGDH mRNA and protein expression in Adflt-PGDH– and Adcox-PGDH–infected cells as compared with Adflt-Luc–infected control cells. Similar results were obtained with LS174T cells (data not shown).

Next, we investigated whether Ad-mediated 15-PGDH overexpression affects PGE₂ and VEGF expression in cancer cells. 2LMP and LS174T cells were infected with Adcox-PGDH, Adflt-PGDH, or Adflt-Luc, and then the levels of PGE₂ and VEGF were measured by ELISA. As shown in Fig. 2A, infection of 2LMP cells with Adcox-PGDH and Adflt-PGDH resulted in decreased PGE₂ concentration by 2.0-fold and 3.2-fold (P < 0.05), respectively, compared with conditioned media from Adflt-Luc–infected control cells. PGE₂ secretion in LS174T cells following Adcox-PGDH and Adflt-PGDH infection was reduced by 1.5-fold and 1.6-fold (P > 0.05), respectively, in comparison with Adflt-Luc–infected cells.

In comparison with Adflt-Luc–infected cells, the levels of VEGF were significantly reduced in conditioned media from Adcox-PGDH– and Adflt-PGDH–infected 2LMP cells, by 2.9-fold and 1.9-fold, respectively (P < 0.05), and by 4.4-fold and 5.7-fold for Adcox-PGDH– and Adflt-PGDH–infected LS174T cells, respectively (P < 0.01; Fig. 2B).

To determine 15-PGDH–mediated cytotoxicity, 2LMP and LS174T cancer cells were infected with Adcox-PGDH, Adflt-PGDH, or Adflt-Luc, and subjected to a clonogenic survival assay. As shown in Fig. 2C, there was 3.0-fold and 1.9-fold (P < 0.01) reduction in the number of colonies for Adcox-PGDH–infected 2LMP and LS174T cells, respectively, as compared with Adflt-Luc–infected control cells. Adflt-PGDH infection of 2LMP and LS174T cells significantly reduced the number of colonies 18.2-fold and 8.1-fold (P < 0.001), respectively, in comparison with Adflt-Luc–infected cells.

To further evaluate whether Ad-mediated 15-PGDH expression inhibits cancer cell migration, the conditioned media obtained from 2LMP and LS174T cancer cells infected with Adflt-Luc, Adcox-PGDH, or Adflt-PGDH were applied to the lower chambers of culture plate transwell inserts, and uninfected cells were seeded to the upper chambers of transwell inserts. Conditioned media from Adcox-PGDH– and Adflt-PGDH–infected cancer cells inhibited 2LMP cell migration by 2.4-fold and 3.4-fold (P < 0.05), respectively, and migration of LS174T cells by 2.1-fold and 2.6-fold (P < 0.05), respectively, compared with conditioned media from Adflt-Luc infected cells (Fig. 3A).

Also, we determined human blood vessel endothelial cell migration using conditioned media obtained from cancer cells infected with Adflt-Luc, Adcox-PGDH, or Adflt-PGDH. The conditioned media were applied to the lower chambers of culture plate transwell inserts, and HUVECs were seeded to the upper chambers of transwell inserts. As shown in Fig. 3B, conditioned media from Adcox-PGDH– and Adflt-PGDH–infected 2LMP cancer cells inhibited HUVEC migration by 2.6-fold and 3.9-fold (P < 0.05), respectively, whereas conditioned media from Adcox-PGDH– and Adflt-PGDH–infected LS174T cells decreased migration of endothelial cells by 3.7-fold and 3.1-fold (P < 0.05), respectively, compared with conditioned media from Adflt-Luc–infected cancer cells.

Next, we investigated the PGE₂ receptor expression in cancer cells following Ad-mediated 15-PGDH overexpression. 2LMP and LS174T cells were infected with Adflt-PGDH, Adcox-PGDH, or Adflt-Luc, and expression of PGE₂ receptor and Cox-2 mRNA was investigated by RT-PCR. The results presented in Fig. 3C show that Adflt-PGDH and Adcox-PGDH decreased levels of Cox-2 mRNA in both cancer cell lines and EP4 mRNA expression in LS174T cells in comparison with Adflt-Luc–infected cells. At the same time 15-PGDH overexpression produced no alteration in levels of expression of VEGFR1 and EP1-3 receptor mRNA in 2LMP and LS174T cancer cells (data not shown).

To investigate the antitumor effect of the 15-PGDH–mediated therapy of breast tumor xenografts, 2LMP cells were s.c. injected into the flank of athymic nude mice. Before treatment, the mean tumor sizes in groups at baseline were not significantly different between treatment groups (P > 0.05) and the within-treatment variances (P > 0.05). The preliminary dose escalation study showed that multiple intratumoral injections of 2LMP and LS174T tumor xenografts with 1 × 10⁸ TCID₅₀ Adflt-Luc had a negligible effect on tumor growth. Consistency of delivery of Adflt-Luc vector into s.c. xenografts was confirmed using luciferase assay (3 tumors/group) at 3 days after the last injection (data not shown). In vivo tumor therapy with 1 × 10⁸ TCID₅₀ of each Ad occurred on days 0, 5, 10, and 15, beginning 5 days after tumor cell injection. The baseline mean and SD for tumor volumes on day 0 was 86.0 ± 9.1 mm³. The mean time to tumor doubling for the PBS-, Adflt-Luc–, Adcox-PGDH–, and Adflt-PGDH–treated groups were 11, 12, 28, and 39 days, respectively (Fig. 4A). Comparisons of mean time to tumor doubling of the group treated with Adcox-PGDH versus Adflt-Luc and Adflt-PGDH versus Adflt-Luc showed significant differences between the groups (P < 0.05).

To determine whether the inhibitory effect of Ad-mediated 15-PGDH therapy on tumor growth was linked to the suppression of tumor angiogenesis, we examined the distribution of the endothelial cell–specific antigen CD31/PECAM1. 2LMP s.c. xenograft-bearing mice received treatment with PBS, Adflt-Luc, Adcox-PGDH, and Adflt-PGDH, as described above, and 3 days after the last injection, the tumors with surrounded tissue (3 tumors/group) were excised, fixed in 10% neutral buffered formalin, and stained with anti-CD31 antibody. Microscopic examination of CD31-stained tumor sections showed that compared with PBS-injected control mice, Adcox-PGDH and Adflt-PGDH injection decreased the number of positive cells (Supplementary Fig. S1; supplementary materials). The quantitative evaluation of tumor-associated vascularity showed that the number of CD31-positive cells in 2LMP tumor xenografts treated with Adcox-PGDH and Adflt-PGDH decreased 1.9 (12.1 ± 3.7) and 2.7-fold (8.4 ± 3.1; P < 0.05), respectively, over that of Adflt-Luc–treated control tumors (22.5 ± 4.2), with PBS-treated xenografts at 21.8 ± 5.0.

Recently the anti-VEGF mAb bevacizumab (Avastin) in combination with chemotherapy was approved for first- and second-line treatments of metastatic colorectal cancer (18). We tested whether flt-1–driven overexpression of 15-PGDH alone and in combination with Avastin could delay growth of LS174T colon tumor xenografts in athymic nude mice. Tumor therapy was initiated on day 0, which corresponded to 7 days after tumor cell injection. The baseline mean and SD for tumor volumes on day 0 was 96.7 ± 12.0 mm³. The mean times to tumor doubling for the PBS, Adflt-Luc, Adflt-PGDH, and Adflt-Luc in combination with Avastin-treated groups were 5, 8, 21, and 25 days, respectively, in comparison with Adflt-PGDH plus Avastin, which produced prolonged delay of LS174T tumor growth (approximately 130% of original tumor size at day 32, the last day of the study; Fig. 4B). Moreover, in the group treated with Adflt-PGDH in combination with Avastin, two tumors underwent complete regression, in contrast to no regressions with PBS-, Adflt-Luc–, Adflt-PGDH–, or Adflt-Luc plus Avastin–treated groups. Comparisons of mean time to tumor doubling of the groups treated with Adflt-PGDH versus Adflt-Luc, Adflt-Luc in combination with Avastin versus Adflt-Luc, and Adflt-PGDH in combination with Avastin versus Adflt-PGDH or Adflt-Luc plus Avastin showed significant differences between the groups (P < 0.05).

Immunohistochemical analysis of LS174T tumors stained with anti-CD31 antibody showed fewer brown-stained CD31-positive endothelial cells (Supplementary Fig. S2; supplementary materials) and reduced microvessel density 1.7-, 3.5-, and 6.7-fold (P < 0.05) in a cross-section of tumor from mice treated with Adflt-PGDH (19.7 ± 5.3), Adflt-Luc plus Avastin (9.2 ± 2.7), and Adf-PGDH plus Avastin (4.9 ± 2.1), respectively, compared with Adflt-Luc–treated tumors (32.8 ± 5.9), with PBS-treated tumors at 36.1 ± 6.7. These results show the antiangiogenesis efficacy of 15-PGDH therapy against tumor-associated angiogenic vasculature in vivo.

The inducible Cox-2 is strongly implicated in both breast and colon neoplasia. Clinical trials are being conducted to study the use of Cox-2 inhibitors alone and in combination with other agents in the neoadjuvant, adjuvant, and metastatic treatment settings (22, 23). Despite tremendous promise of anti–Cox-2 therapy using selective Cox-2 inhibitors, which were developed to provide an alternative with reduced gastrointestinal risk to the traditional nonsteroidal anti-inflammatory drugs, these have been associated with an increased incidence of major cardiovascular toxicity due to inhibition of Cox-2–mediated PGI₂ expression (24). Thus, their utilization for cancer prevention and therapy is currently questionable, and suggest that further development of novel selective agents with lower toxicity is needed. It is reasonable that inhibition of only PGE₂, rather than global PG biosynthesis, would avoid some of the potential gastrointestinal and cardiovascular side effects.

Thus, clearly other novel therapy strategies are desperately needed to achieve adequate cancer therapy when the ineffectiveness of the existing treatment regimens is well recognized. The application of gene therapy to neoplastic diseases offers a promising approach to achieve targeted antitumor intervention. Further to this end, gene therapies for cancer may be combined with standard modalities to accomplish synergistic anticancer effects. The Adcox-PGDH and Adflt-PGDH vectors have been developed to function as therapeutic gene delivery vehicles by incorporating an expression cassette containing the human 15-PGDH transgene. The employment of tumor-specific promoters for 15-PGDH gene expression would increase therapeutic efficacy and avoid adverse side effects of anticancer treatment.

In a comparative study, MCF 10A, normal established breast epithelial cells, and 2LMP breast cancer cells, derived by passage of MDA-MB-231 tumor cells in ascites of nude mice (reflecting the metastatic state; refs. 25–27), were tested for PGE₂ and VEGF protein expression, as well as EP1-4, Cox-2, and VEGFR1 mRNA expression. The results of in vitro studies show that the levels of elevated PGE₂ and VEGF secretion in 2LMP breast cancer cell culture media were associated with high levels of expression of PGE₂ receptors, Cox-2, and VEGFR1 in comparison with MCF 10A normal cells. The high levels of EP1-4, Cox-2, and VEGFR1 mRNA were detected using quantitative RT-PCR in 2LMP, MDA-MB-231, and LCC6 breast cancer cells in contrast to relatively lower 15-PGDH mRNA expression.

Although the underlying molecular mechanisms of regulation of 15-PGDH expression remain to be characterized in more detail, it was shown that the loss of 15-PGDH expression in colon cancer is most likely due to the upregulation and binding of the transcriptional repressor Snail to E-boxes located in the 15-PGDH promoter (28). Also, ZEB1 and Slug are downstream targets of the epidermal growth factor receptor signaling pathway through activated ERK, and these transcriptional repressors are important for the regulation of 15-PGDH expression in a subset of non–small cell lung cancer cells (29). It was shown that upregulation of 15-PGDH induced by nonsteroidal anti-inflammatory drugs has the potential to inhibit growth of human glioma, in part, by upregulation of p21 possibly independent from Cox-2 enzymatic function (30). Recent studies identified a tumor suppressor activity of 15-PGDH in colon and bladder cancers, and suggested epigenetic silencing of the enzyme by DNA methylation and histone modification (19, 31). Low 15-PGDH levels were shown in the poorly differentiated, Cox-2–expressing MDA-MB-231 breast cancer cells isolated from pleural effusions as well as in primary breast cancers. Decreased 15-PGDH levels were correlated with unfavorable prognostic factors (32). Restoration of 15-PGDH expression produced a delay in breast tumor growth (33).

The results of in vitro studies showed that Ad-meditated 15-PGDH overexpression produced increased cytotoxicity and inhibited cancer cell migration in comparison with Adflt-Luc. It is becoming more obvious from preclinical and clinical studies that overexpression of Cox-2 and repression of 15-PGDH provide complementary pathways to increase PGE₂ levels in tumor. Ad-mediated restoration of 15-PGDH expression was associated with reduction of PGE₂ and VEGF concentration in cultured media and decreased levels of Cox-2 and EP4 mRNA expression in colon cancer cells. Although the role of PGE₂ receptors in cancer biology remains complex, it was shown that expression of EP4 plays an important role in colon and breast cancer development (34–38). However, additional studies are required to investigate molecular mechanisms of the therapeutic effect of 15-PGDH overexpression in cancer cells.

Future studies also need to establish precise mechanisms of regulation of tumor angiogenesis via Ad-mediated 15-PGDH overexpression. The in vitro study showed that cell conditioned medium from Adcox-PGDH– or Adflt-PGDH–infected cancer cells inhibited migration of HUVEC and decreased proliferation of endothelial cells (data not shown) in comparison with Adflt-Luc. Importantly, it was shown that PGE₂ induces the expression of VEGF (14) and bFGF (15) in blood vessel endothelial cells (16). On the other hand, both VEGF and bFGF can induce PGE₂ expression in endothelial cells, which raises the possibility of amplification of PGE₂ regulation of angiogenesis through a positive feedback loop (17). Thus, decreased levels of PGE₂ and VEGF in cultured media from 15-PGDH–overexpressed cancer cells can, in part, produce inhibition of tumor-associated microvasculature.

The animal therapy studies showed that Ad-mediated 15-PGDH therapy of breast and colon tumor xenografts produced growth inhibition. Importantly, a significant delay in tumor growth was associated with a decreased number of CD31-positive cells in Adcox-PGDH– and Adflt-PGDH–injected tumor xenografts in comparison with Adflt-Luc. Data from experimental and clinical studies indicate that growth of most cancers is strongly angiogenesis-dependent. High levels of PGE₂ expression in breast and colon tumors are associated with increased angiogenesis. Overexpression of PGE₂ induces VEGF production in tumor cells, which is associated with robust expression of its cognate VEGFR1/flt-1 receptor and results in activation of tumor-associated angiogenesis and in development of self-sufficiency in growth signaling in highly aggressive metastatic breast tumors (39). The antiangiogenic drug Avastin is widely being used for treatment of patients with colon cancer. The antitumor effect of Adflt-PGDH–mediated therapy of colon tumor xenografts was significantly improved by coadministration of Avastin.

Notably, the most effective growth inhibition of both breast and colon tumor xenografts occurred when expression of the 15-PGDH gene was driven by the flt-1 promoter. VEGF and its cognate receptor VEGFR1/flt-1 are frequently overexpressed in human malignancies, including breast and colon carcinomas. Recent studies in breast cancer show that high levels of VEGF production in tumor cells is often associated with robust expression of its cognate VEGFR1/flt-1 receptor in tumor-associated blood microvessels and tumor cells (40). The precise molecular mechanisms of regulation of carcinogenesis by VEGFR1 remains uncertain; it is obvious, however, that flt-1 overexpression plays a dual role through activation of tumor-associated angiogenesis and in development of self-sufficiency in growth signaling, which is involved in development of highly aggressive breast cancer phenotypes (41).

It was shown that high levels of VEGF/VEGFR1 were associated with reduced survival in breast cancer patients (42). Expression of flt-1 in colon cancer cells has been recently associated with induction of cell motility and invasion (43). Recently, we have shown that flt-1–driven expression of the tumor necrosis factor–related apoptosis-inducing ligand suppressed in vivo growth of human prostate tumors (44). Importantly, expression of flt-1 is inducible and significantly enhanced under hypoxic conditions or under exposure to oxidative stress (45). Both hypoxia and oxidative stress are hallmarks of the tumor microenvironment, and this may explain tumor specificity of the flt-1 promoter. Expression of flt-1 has been observed in tumor cells, endothelial cells, and in tumor-infiltrated CD11b myeloid cells. Inflammatory tumor-infiltrated CD11b myeloid cells are involved in both regulation of antitumor immune response and tumor-associated neovasculogenesis. These cells have been shown to support angiogenesis via secretion of MMP-9 and VEGF (46). Interestingly, flt-1–driven expression of the 15-PGDH gene in the tumor microenvironment promotes significant reduction of VEGF secretion by tumor-infiltrated CD11b myeloid cells and also stimulates the antitumor immune response in immunocompetent mice bearing established syngeneic murine colon tumors (47).

Collectively, our data show that Ad-driven expression of 15-PGDH promotes inhibition of breast and colon cancer cell growth. The 15-PGDH–mediated anticancer effect is associated with reduction of PGE₂ and VEGF levels and decreased Cox-2 and PGE₂ receptor EP4 mRNA expression. The results of experimental therapy studies show that administration of Adcox-PGDH and Adflt-PGDH into mice with established tumors resulted in delay of tumor growth. The results of in vivo studies reveal that the inhibitory effect of 15-PGDH therapy on tumor growth was associated with the suppression of tumor-associated angiogenesis. Our studies show that the antitumor effect can be further enhanced when 15-PGDH therapy was combined with anti-VEGF mAb. These results suggest that 15-PGDH–mediated regulation of PGE₂ catabolism in the tumor microenvironment represents a novel approach for therapy of breast and colon cancer.

No potential conflicts of interest were disclosed.

We thank Sally B. Lagan for assistance in preparing the manuscript.