An HRE-Binding Py-Im Polyamide Impairs Hypoxic Signaling in Tumors

This article is featured in Highlights of This Issue, p. 531

Oxygen sensing is involved in a range of natural physiologic processes such as embryonal development, wound healing, and immune response (1). However, its dysregulation can contribute to pathogenesis of multiple diseases including fibrosis, erythrocythemia, heart disease, and cancer—a group of diseases leading to nearly 600,000 deaths every year, in the United States alone (2). Although new cancer treatments are being developed, many of the cancer therapies are hampered by presence of low levels of oxygen in tumors (3), known as tumor hypoxia, regulated mainly by hypoxia-inducible factors (HIF; ref. 4). Among them, HIF1 (5), is a transcription factor that is often is associated with poorer patient survival (3). HIF1 orchestrates numerous aspects of cancer progression, tumor angiogenesis, cell survival in hypoxia, and metastasis (3, 6, 7). Most inhibitors affect HIF1 signaling indirectly, by targeting other proteins, including Topoisomerase I, mTOR, or microtubules (7, 8). Established therapeutic strategies focus on modulating HIF1 signaling by altering the HIF1 protein levels (8), its dimerization and protein interactions (9–11), or its DNA binding and transcriptional activity (12–14).

One method to inhibit transcriptional activity of HIF1 is to displace it from its DNA-binding site, HIF1-responsive elements (HRE), for example, by a molecule shown in this report (compound 1; Fig. 1A; refs. 12, 13). Compound 1 is a member of a class of DNA-binding small molecules, pyrrole-imidazole (Py-Im) polyamides, which can be programmed to recognize a broad repertoire of sequences with affinities and specificities comparable with those of transcription factors (15). We reported that polyamides can modulate gene expression driven by many transcription factors in tissue culture (12, 13, 16–18).

Experiments focused on HIF1–DNA binding inhibition demonstrated the ability of 1 to displace the HIF1 complex from DNA in vitro (12), and by reducing HIF1α occupancy at selected HREs in a common tissue culture model of hypoxic response—U251 cells as represented in Fig. 1B (13). When U251 cells were treated with deferoxamine (DFO), an iron chelator and HIF1 activator (19), 1 was capable of inhibiting 23% of the induced genes, establishing polyamide 1 as a partial inhibitor of HIF1 transcriptional activity (13). The concentration used was 1 μmol/L, well below what was measured in the tumor tissues investigated in this study (2–3.5 μmol/L). The overall gene expression changes included many important proangiogenic factors, such as VEGFA, FLT1, or Endothelin-1, but also other factors. In light of the recent developments in cancer systems biology (20, 21), the broad effects of Py-Im polyamides on gene expression 1 (13, 16–18) led us to posit that their effects may arise due to affecting multiple targets. Recent discoveries show that drugs commonly act upon multiple molecular entities (20, 21) and this feature could be beneficial in cancer treatment, thanks to the decreased likelihood of developing drug resistance (22). However, the potential for toxicity due to off-target effects is a point of concern. Our mechanistic investigations expanded beyond the transcription factor:DNA interface and found that Py-Im polyamides can induce degradation of the large subunit of RNA polymerase II, activation of the P53 stress response without concurrent DNA damage (23), and induction of DNA replication stress (24).

We decided to establish whether the pattern of gene expression caused by administration of a Py-Im polyamide binding DNA sequence found in HREs (5′-WTWCG-3′) could result in measurable and useful antitumor effects in vivo. Our experiments focused on suppression of the phenotypes typical for tumor hypoxia, such as tumor vascularization, hypoxic gene expression, or cancer cell survival (1, 8). We chose a subcutaneous tumor model, due to its hypoxic nature (3) and reliability of these measurements. The cell lines chosen for engraftment have been evaluated extensively in xenograft models of cancer (25, 26) and hypoxic gene expression in one of them (U251) can be regulated by 1 (13), making it a good choice for evaluating in vivo mechanism of action of 1. The second cell line, GBM39 (27), was derived from the same site as U251 cells (brain), but was maintained as subcutaneous xenografts and expanded in serum-free conditions as spheroids. The serum-free treatment maintains genetic and histologic variability of human tumors and was thus our choice to test generality of mechanism of action of polyamide 1 (28). Our previous in vivo investigations demonstrated the bioavailability of various Py-Im polyamides upon intravenous (29), intraperitoneal, and subcutaneous administration (30). Subsequent in vivo xenograft studies established that polyamides could accumulate in subcutaneously grafted tumors (18, 30, 31), modulate tumor gene expression (18, 30), and inhibit growth of prostate cancer xenografts (23, 32). The distinct profile of gene expression modulation, favorable pharmacokinetics, and ability to modulate all of the tested aspects of hypoxic response by 1, could establish it as an interesting candidate for treatment of cancer and other hypoxia-related diseases. In this study, we established the ability of Py-Im polyamide 1 to interfere with hypoxic response, regulate gene expression in tumors, and to inhibit tumor growth and angiogenesis.

U251 cell line was a kind gift of Dr. Giovanni Mellilo (NCI, Bethesda, MD) in 2005. GBM39 cells were a gift of Dr. David Akhavan (University of California, Los Angeles, Los Angeles, CA; UCLA) in 2013, originally obtained from Prof. David James (University of California, San Francisco, San Francisco, CA). The cell lines were authenticated using StemElite ID assay (Laragen). U251 cells were maintained in RPMI1640 medium supplemented with 10% FBS. GBM39 cells were cultured in suspension in F12/DMEM medium supplemented with b27, EGF (20 ng/mL), FGF2 (20 ng/mL), and Glutamax, heparin (50 μg/mL), with growth factors replaced every 2 days. After 25 passages, GBM39 were reestablished as subcutaneous xenografts (NSG mice). The tumor was disintegrated, digested in 10% accutase (3 hours), cells resuspended by pipetting, and then cultured as described above. Human umbilical vein endothelial cells (HUVEC) were cultured in 200 PRF medium (Gibco) supplemented with low serum growth supplement (LSGS; Invitrogen). Media were exchanged every 48 hours and cells passaged weekly.

HUVECs were plated at 2 × 10⁵ cells (75-cm flask). After 36 hours, compound 1 (5 μmol/L) was added for 72 hours. Twelve-well plates were coated with 100 μL/well Geltrex (Invitrogen) and allowed to solidify at 37°C for 60 minutes. Cells were trypsinized and resuspended in 200 PRF medium at 2 × 10⁵/mL, and plated in the 12-well plates at 400 μL/well. After 6 to 12 hours, the wells were imaged on an inverted microscope by selecting four random fields of view.

Animal experiments were carried out according to approved Institutional Animal Care and Use Committee protocols at the California Institute of Technology (Pasadena, CA). Pharmacokinetic studies were done as described previously (30). Serum for analysis of toxicity markers (IDEXX) was obtained by RO collection and centrifugation (2,000 × g, 15 minutes) in Serum Separator Tubes (BD Biosciences). The radioquantitation of 2 was performed as described elsewhere (33).

Tissue processing was conducted at the Tissue Processing Core Laboratory (TPCL) at UCLA, using 5-μm paraffin sections. Tumor microvessels were visualized using anti-mouse CD31 staining using SC-1506 antibody (SCBT), apoptosis was assayed for using anti-human cleaved caspase-3 (CC3) antibody (9664, CellSignal), and HIF1α using CME 349A (Biocare Medical). At least two tissue sections per slide were used for each experiment. Slides were scanned using an Aperio ScanScope AT (Aperio) and positive pixel measurements were done using Imagescope software.

Perinecrotic area was defined as a field of view at 20× magnification that contained 50% necrosis and 50% adjacent, nonnecrotic area, with edge of necrosis in the middle of the screen. Numbers of histologic measurements for each tissue section were chosen to limit variability of the final average value to <10%, per tumor, or until all available data in the section were counted. Ki-67 measurements were done using Immunoratio software at 20× magnification, with n = 24 per tumor section. Nuclear density (and by proxy–necrosis) was assessed by automatic nuclei counting using Image-based Tool for Counting Nuclei (ITCN), 16 fields of view were analyzed per tumor section.

At least 6 (CC3), 9 (CD31) or 3 (TUNEL, GBM39), 5 (TUNEL, U251), or 8 (HIF) fields of view per tumor were chosen randomly prior to analysis, at either 20× (CC3 staining, CD31 staining for U251, GBM39, schedules A and B, TUNEL, HIF) or 40× magnification (CD31 staining for GBM39, schedule C, due to higher CD31-positive vessel counts). For analysis of frequency of apoptotic blood vessels (CC3/CD31 double staining), we counted enough high-power fields (20×) containing CD31-positive vessels, to assess at least 600 microvessels per condition.

Perinecrotic MVD was measured by first delineating nuclei on the edge of necrosis, and then offsetting the obtained path by an equivalent of 50 to 350 μm in Illustrator CS5 (Adobe). The MVD was then counted manually for each 50-μm zone and area integrated in Photoshop CS5 (Adobe). At least 40 measurements per tumor were obtained. Measurements of the nearest blood vessel-necrosis distance were done manually using Imagescope at 40× magnification.

Experiments were carried out in 8- to 12-week-old male NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, JAX). Cells were injected subcutaneously into a flank area (2.5 × 10⁶ of U251 and 7.5 × 10⁵ of GBM39 cells). U251 cells were administered in RPMI1640 medium and GBM39 cells in 50% Matrigel/culture medium. Before treatment, each vehicle-treated control xenograft was paired with a xenograft of the same initial volume to be treated with 1. Compound was administered subcutaneously interscapularly in 20% DMSO/PBS. Mice were weighted weekly and euthanized if 15% weight loss or more was observed.

Primers (Supplementary Table S1) were designed using Harvard Primer bank (34). The total mRNA in tumors was processed as in our previous studies (30).

Cells for gene expression analysis were plated in 6-well plates at 5–10 × 10⁴ cells/well and incubated in RPMI1640 medium with 10% FBS until approximately 50% confluence (24–48 hours). Afterward, medium was exchanged with the growth medium supplemented with 1 (1 μmol/L). After 48 hours, either 300 μmol/L of DFO or PBS was supplemented, and the cells were incubated for an additional 16 hours. The RNA was then harvested using an RNEasy Kit (Qiagen) and treated with a Riboguard RNAse inhibitor and TurboDNA Free DNAse (Ambion), according to manufacturers' instructions. RNA-Seq libraries were prepared using standard Illumina reagents and protocols. Single-read sequencing with the read length of 50 nucleotides was conducted on the Illumina HiSeq2000 sequencer, producing 35 to 40 million reads per library, with a total of three biologic replicates per sample. Sequencing data were mapped against the combined human (hg19) transcriptome, using the Tophat2 program package 2.0.13 (35) and the GRCh37 annotation. Differential gene expression was calculated with the DEseq2 module (36). Clustering analysis was performed using Cluster 3.0 software (37).

Each sample was analyzed using a two-tailed, Student t test, assuming normality and unequal variance. P value of 0.05 or less was considered statistically significant.

Our previous results showed that 1 can act as an inhibitor of HIF1-driven gene expression (13) in U251 cells. However, we did not determine activity of 1 in cells without HIF1α induction. Tumors contain both hypoxic and normoxic regions and thus it is important to establish how 1 acts in cells with and without induced HIF1α. To address this question, U251 cells were first treated with 1 (1 μmol/L, 48 hours) or PBS (vehicle). Half of the samples were then dosed with a HIF1α inducer DFO (300 μmol/L, 16 hours), whereas the other half with PBS (Supplementary Fig. S1A). Their mRNA was then analyzed by RNA sequencing. Gene expression differences for each sample were obtained by normalizing a polyamide-treated sample with the appropriate nontreated control—cells treated with PBS and 1 were normalized to cells treated with PBS only (PBS vs. 1), and cells treated with both 1 and DFO to cells treated with DFO (DFO vs. DFO+1). When 1 was dosed into cells treated with DFO, we observed it downregulated 744 genes. When 1 was dosed into cells dosed with PBS, a different set of genes was regulated. In fact only 101 of 723 regulated genes were identical for both conditions (14%; Fig. 1C and D). The upregulated genes also varied between the two groups. Cells treated with 1 and DFO showed upregulation of 316 genes, whereas the same cells treated with 1 and PBS 1,514 genes. Few genes were in common between the two groups (6% for DFO, 1.3% for PBS; Fig. 1C and D). Overall, 17% DFO-induced genes were downregulated by 1 at least 2-fold (P < 0.05). Gene clustering analysis suggested 1 acted to reverse effects of DFO (Supplementary Fig. S1B and S1C). Fewer genes were affected by compound 3 at the same dose (Supplementary Fig. S1D). Interestingly, we found that among the genes that were changed significantly both by 1 (in DFO-induced cells) and by DFO, the Py-Im polyamide 1 inhibited DFO-induced gene expression in 96% of the genes (2,557/2,661 genes; Supplementary Fig. S1E). This indicates that gene expression changes caused by DFO are either unaffected, or reversed by 1, but rarely, if ever, exacerbated by treatment with 1.

To evaluate bioavailability of 1 in vivo, we injected it subcutaneously at 6.8 mg/kg into C57BL6 mice. Compound 1 reached a serum concentration of 11.3 ± 1.6 μmol/L within 1.5 hours and 7.8 ± 0.4 μmol/L at 4 hours after injection (n = 4; Fig. 2A). The radiolabeled analogue of 1, Py-Im polyamide 2, was injected intraperitoneally (i.p.) to quantitate whole-organ compound concentrations (n = 6; Supplementary Fig. S2A and S2B) into immunocompromised NSG mice bearing GBM39 tumors. After 24 hours, concentrations of 2 were measured in GBM39 xenografts (2.0 μmol/L), host's kidney (6.3 μmol/L), liver (4.2 μmol/L), and lung (3.1 μmol/L; Supplementary Fig. S2A). Following three injections (6.8 mg/kg s.c., every other day × 3, harvested 24-hour after last injection), all tested tissues showed compound accumulation (Supplementary Fig. S2B). The same administration of 3 (FITC-conjugate of 1) in NSG mice resulted in readily detectable nuclear staining in tumor cells and tested tissue sections (Supplementary Fig. S2C–S2E). The nuclear staining was also present when tested in unfixed cells in tissue culture setting (Fig. 2B). A single-injection toxicity test showed that 1 did not affect animal weight or levels of serum toxicity markers (ALT, AST, TBIL, BUN, creatinine) at concentrations up to 10 mg/kg (Supplementary Fig. S3A and S3B). Taken together, these results supported evaluation of antitumor effects of 1 in vivo.

To test whether previously established partial inhibition of HIF1-driven gene expression by 1 (13) would result in decrease in tumor growth, we engrafted U251 and GBM39 cells subcutaneously into NOD/SCID-γ (NSG) mice. We established three dosing regiments (Supplementary Fig. S4A) for 1—schedule A, consisting of three subcutaneous injections every other day at maximum tolerated dose (6.8 mg/kg, 8 days) and schedules B and C, with lower dose (4.5 mg/kg, 3 injections/week) designed for prolonged treatment with low weight loss (<10%) over 4 weeks (U251 tumors) or 6 weeks (GBM39 tumors). The schedule A was designed to elicit maximum response over a short period of time at a dose tolerable to animals. The schedules B and C were introduced to assess if the compound 1 could retain its activity at a dose that could be tolerated over longer periods of time. In all cases, the tumors were harvested 72 hours after last injection, due to our previous research, indicating that Py-Im polyamide–induced apoptosis occurs between 48 and 72 hours after compound administration in tissue culture studies (24).

Mice bearing U251 tumors were subjected to treatment with 1 (Schedule A), which resulted in a reduction in median tumor mass (1.8-fold, P < 0.013; n = 7, 8, for vehicle and treated groups; Fig. 2C) compared with vehicle. Similarly, prolonged treatment (schedule B) resulted in 1.9-fold lower median U251 tumor mass (P < 0.0125; n = 10 per condition; Fig. 2D). Consistent results were obtained with a primary glioma cell line (GBM39) xenograft. Treatment according to schedule A resulted in 1.8-fold lower median mass (P < 0.016; n = 11 per condition; Supplementary Fig. S4B) and prolonged treatment (schedule C) in 1.6-fold reduction (P < 0.041; n = 6 per condition; Supplementary Fig. S4C). The treatments resulted in minor mouse weight loss; the effects were less severe (<5% average weight loss throughout the experiment) for schedules B and C (Supplementary Fig. S5).

Hypoxic signaling is a major driver for angiogenesis and its inhibition leads to decrease in microvessel density (MVD; refs. 1, 8). We measured MVD using anti-mouse CD31 immunostaining of tumor sections. We observed a significant reduction of MVD in all tested scenarios. For U251 tumors, with mice subjected to schedule A, we observed a 1.4-fold median reduction of MVD (P < 0.014; n = 7, 8, for vehicle and treated groups; Fig. 2E) and 1.7-fold median decrease for schedule B (P < 0.05; n = 10 per condition; Fig. 2F). For GBM39 tumors, both treatment schedules (A and C) led to 1.4-fold reduction in MVD (P < 0.01; n = 11 and 6, for schedules A and C; Supplementary Fig. S6A and S6B).

The decrease in MVD could be caused by direct effect of 1 on the endothelium, for example, rendering it either apoptotic, or unable to form blood vessels. We evaluated endothelial apoptosis by double-staining of mouse-specific CD31 and CC3 in GBM39 tumors (schedule C; n = 5 tumors, approximately 600 microvessels per condition; Fig. 2G). We found no significant endothelial apoptosis in either of the treatment groups. To measure angiogenic functionality of endothelium, we used in vitro Matrigel tube formation assay with HUVECs, revealing 1 (5 μmol/L, 48 hours) had no effect on tube formation (n = 3; Fig. 2H).

Decrease in MVD, without endothelial apoptosis or dysfunction, suggests 1 could interfere with the hypoxic response (5). To further test this hypothesis, we evaluated other aspects affected by hypoxic response, such as tumor cell proliferation (38), apoptosis in HIF1-positive, perinecrotic areas (38), nuclear HIF1α protein accumulation, and cell survival in areas distant from blood vessels (39, 40). To analyze this, we divided the tumor sections into three areas: necrotic, nonnecrotic, and perinecrotic (Fig. 3A). The non-necrotic areas contain nucleated cells with intact cell membranes, whereas necrotic areas are show loss of nuclei and fragmentation of cells. The perinecrotic area represents field of view (20× magnification) that contains 50% of cells containing nuclei and 50% of necrotic areas.

Antiangiogenic therapy increases hypoxia in tumors and often leads to decreased proliferation (39, 41). Upon treatment with 1 (schedule B), we observed a 1.2-fold decrease (P < 0.05; n = 10 per condition; Fig. 3B) in proliferation marker (Ki-67) in nonnecrotic areas of the U251 tumors.

Another effect of induction of hypoxic response is accumulation of HIF1α (1). However, treatment with 1 did not lead to increase in levels of HIF1α in either U251 (Fig. 3C and D) or GBM39 (Supplementary Fig. S7A and S7B) tumors. Lack of HIF1α accumulation is unlikely to be caused by its increased degradation, as 1 does not affect HIF1α levels in tissue culture (Supplementary Fig. S7C). Overall, we observed a spatial distribution of HIF1α in U251 tumor sections—perinecrotic areas showed higher HIF1α levels compared with nonnecrotic areas (1.8-fold average increase for 1, P < 0.001; n = 10 per condition; Fig. 3C and D).

Induction of hypoxic signaling can lead to apoptosis resistance in cancer cells (39). However, treatment with 1 increased apoptosis significantly in perinecrotic, HIF1α-positive, areas (CC3 staining, 1.8-fold, P < 0.0012; n = 10 per condition; Fig. 3E). The apoptosis was absent in nonnecrotic regions of U251 tumors regardless of the treatment (Fig. 3F), suggesting 1 is toxic specifically to hypoxic cells. Similarly, TUNEL staining revealed convergent results—no increase in apoptotic cells in nonnecrotic regions, and 1.7-fold (P < 0.009, n = 10 per condition) increase in apoptosis in perinecrotic tumor regions upon treatment with 1 (Supplementary Fig. S7D and S7E). The GBM39 tumors harvested before they developed necrosis also revealed no induction of apoptosis by 1 (Supplementary Fig. S7F).

Hypoxic signaling mediates cell survival in areas with low oxygen pressures. With increasing distance to blood vessels, tumor necrosis appears as a result of the death of cells with inadequate oxygen supply (1, 42). The oxygen pressure decreases as the function of distance from blood vessels, and typically does not reach beyond approximately 200 μm. Cells adapt to low oxygen pressures by activating hypoxic signaling, and once oxygen pressures drop below a critical value, cell necrosis occurs (1, 42). The cells that succeed to survive at distances comparable with 200 μm from blood vessels can thus be deemed as having a functional hypoxic response. Impairment of hypoxic signaling will result in necrosis occurring at higher oxygen pressures, at shorter distances from blood vessels. Therefore, to measure whether 1 decreases ability of cells to adapt to low oxygen pressures, we measured MVD as a function of distance from the edge of necrotic areas. We found the expected lack of microvessels near necrotic edges (<100 μm). In tumors derived from animals treated with 1 (U251, schedule B), necrotic edge appeared closer to the blood vessels. The median distances between the edge of necrosis and the nearest blood vessels were significantly shorter for both U251 and GBM39 tumors treated with 1, as compared with vehicle-treated controls (P < 0.01; Supplementary Fig. S7G). Similarly, at short distances around necrosis, the mean MVD appeared higher for tumors treated with 1, suggesting treated tumor cells require higher level of nutrients and oxygen for survival (3.3-fold at 150 μm and 2.3-fold for 200 μm distance from the necrotic edge, P < 0.02 and <0.01; Fig. 4A and B).

Lower MVD and lower cell survival at a distance from blood vessels should lead to an increase in necrotic area. However, the complex pattern of necrosis (micronecroses) in U251 tumors made it difficult to quantify it by a simple delineation. Instead, we decided to automatically count nuclei in the whole tumor section and found lower nuclear density in tumors treated with 1 (10% fewer nuclei, P < 0.03; Supplementary Fig. S8A). The further evidence of necrosis induction by 1 was present after short-term treatment with 1 in GBM39 tumors (schedule A), where transient accumulation of HIF1α (Supplementary Fig. S8B–S8D) led to presence of necrosis (Supplementary Fig. S8E), specifically localized to HIF1α-positive areas that were distant from microvessels (Supplementary Fig. S8E and S8F).

Overall, compound 1 decreases proliferation and nuclear density, selectively induces apoptosis in perinecrotic, HIF1α-positive areas, and causes necrotic areas to appear in proximity of blood vessels despite the presence of a blood supply, but treatment does not result in long-term HIF1α accumulation.

A common adverse effect of antiangiogenic treatment is upregulation of proangiogenic and prometastatic gene expression which renders the antiangiogenic therapies less effective and is often a result of hypoxic signaling (43). We decided to test whether 1 affected mRNA expression of such factors in vivo. NSG mice bearing U251 tumors were treated with 1 according to schedule D (6.8 mg/kg, two injections, on days 1 and 3; tumors harvested on day 5; SI Supplementary Fig. S4A). Out of 10 tested proangiogenic factors, four transcripts had lower relative expression after treatment with 1 (Fig. 5A), and one (VEGFA) was upregulated (1.4-fold) in both mRNA (Fig. 5B) and serum protein levels (Fig. 5C). Downregulation of mRNA expression was also apparent in the panel of prometastatic genes. Overall, five of six tested transcripts were downregulated in the group treated with 1 (Fig. 5D). The factors significantly affected by 1, included: FLT1, NLGN1, HOXB9, NEDD9, MMP2, and MMP14. Interestingly, the mouse receptor of Angiopoietin, Tek, was also downregulated. Results for all tested genes can be found in Supplementary Table S2.

Regulation of hypoxic signaling is central to maintaining balance between health and disease. Its principal regulator, HIF1, is essential for tumor initiation and progression, for example, vascularization, cell survival, and metastasis (3, 44). Inhibition of HIF1 activity has suppressed tumor progression and reduced cancer resistance to available therapies (reviewed in refs. 1 and 3). Our group has previously reported on the function of an HRE-binding Py-Im polyamide 1 as a partial inhibitor of HIF1-dependent transcription in tissue culture (13). However, the in vivo activity and mechanism of 1 remained to be explored. We used Py-Im polyamide 1 (Fig. 1) to show its activity is dependent on HIF1α activation (Fig. 1), that it inhibits tumor growth, angiogenesis (Fig. 2), and the tested aspects of hypoxic response in vivo (Figs. 3 and 4), including inhibition of tumor-derived mRNA expression of proangiogenic and prometastatic factors that are often upregulated in hypoxic conditions (Fig. 5). We found that the effects of 1 are consistent with what would be expected in case of a partial inhibition of hypoxic response (Fig. 6).

Subcutaneous administration revealed 1 distributes into serum within 1.5 hours after injection, with 31% drop in concentration 2.5 hours later, indicating multihour long half-life. Even though 1 had favorable pharmacokinetics, its activity in vivo is likely dependent on the target tissue concentration. After three intraperitoneal injections (schedule A), the radioactive analogue 2 reached a concentration of 3.5 μmol/L in GBM39 tumor and higher levels in other tissues. Concentrations attained for 1 in all tested tissues were thus higher than used in our tissue culture studies (13).

In the current study, inhibition of tumor growth and reduction in MVD was observed in two different cell types in response to treatment with 1. The extent of these effects was comparable with ones exerted by bevacizumab (45) in xenografts derived from U251 (41) and GBM39 cells (26). Investigation of possible mechanisms of antiangiogenic action showed lack of apoptosis in blood vessel lining, and no measurable influence on tube formation on Matrigel. Lack of direct effects of 1 on endothelial cells suggests blood vessel recruitment might be impaired. One possible explanation is inhibition of expression of tumor-associated proangiogenic factors by 1. However, systemic concentration of those proteins could also be affected.

Compound 1 induces apoptosis selectively in HIF1α-positive perinecrotic areas. This suggests 1 sensitizes cancer cells to hypoxia, which could occur by hypoxic response inhibition (3). Our RNA-sequencing data show that activity of 1 depends on activation of HIF1α, which provides a possible explanation why activity of 1 was restricted to cells in HIF1α-positive areas. Further support of this hypothesis is the lack of increase of HIF1α-positive cells in the tumors treated according to schedules B and C. Because 1 did not affect HIF1α levels in U251 cells in tissue culture, it is likely that numbers of HIF1α-positive cells were reduced as they went through apoptosis. Finally, increased reliance of cancer cells on proximity to vasculature once again suggests their hindered ability to adapt to low partial oxygen pressures, which is the main function of the hypoxic response. It is worth noting that, despite the fact that more blood vessels exist in proximity of necrosis, the perinecrotic areas are not representative of the whole tumor. These areas contain only a small fraction of tumor area and inherently contain few blood vessels, which leads to necrosis. The transient induction of HIF1α levels in briefly treated GBM39 tumors indicates a likely increase in tumor hypoxia or a change in HIF1α protein levels by direct action of 1. The latter possibility is unlikely as 1 had no effect on HIF1α levels in tissue culture. The apparent disparity in HIF1α levels between the GBM39 tumors treated according to schedule A, and other regimens, could be due to two factors: concentration of 1 was insufficient to induce apoptosis of tumor cells in schedule A, or overall level of hypoxia in these smaller tumors (Supplementary Fig. S8C) was too low to induce sensitivity to 1. The latter hypothesis is supported by comparable HIF1α levels in nonnecrotic regions of GBM39 tumors from a group treated according to schedule C (Supplementary Fig. S7A) and by tendency of smaller tumors to be less hypoxic (46). Reduced proliferation upon treatment with 1 is likely a combination of both reduced MVD and a direct effect of 1 on cancer cells. The former hypothesis is supported by the dependence of proliferation on distance from necrosis, whereas the latter by a reduction in Ki-67 in nonnecrotic areas that are presumably sufficiently vascularized for unrestricted cell division.

Compound 1 reduced mRNA expression of a panel of genes involved in angiogenesis and metastasis. Many of the affected genes carry important functions in tumor adaptation to hypoxia, for example, platelet-derived growth factor subunit B (PDGFB) and angiopoietin-1 (ANGPT1) are involved in blood vessel maturation (43), whereas Lysyl Oxidase (LOX) or RHOC were involved in metastasis (47, 48). Other factors significantly affected by 1, included: FLT1, NLGN1, HOXB9, NEDD9, MMP2, and MMP14, and a mouse Angiopoietin receptor, Tek. We also observed a slight elevation (1.4-fold) in both mRNA expression and serum concentration of VEGFA, despite a visible downregulation in tissue culture (13). Reduction in microvessel density typically leads to decreased oxygen pressure and increased expression of proangiogenic factors, including VEGFA (41, 43). Interestingly, recent studies show that regulation of VEGFA activity can be HIF1 independent (49, 50). However, it is also possible that VEGFA expression would be higher had it not been for treatment with 1. Importantly, elevated expression of VEGFA did not prevent 1 from exerting antiangiogenic effect, suggesting its mechanism of action could be VEGF independent.

In summary, this study demonstrates that Py-Im polyamide 1 interferes with the tested endpoints of hypoxic response in tumors: it inhibits prometastatic and proangiogenic gene expression, induces tumor cell apoptosis selectively in HIF1α-positive areas and decreases nuclear density, proliferative index, and microvessel density of tumors. The RNA-sequencing data revealed that 1 regulates genes differentially in cells with induced and basal levels of HIF1α, providing a potential explanation for selective induction of apoptosis in HIF1α-positive cells. This compound could be useful in treatment of hypoxia-related disease as it distributes to tissues, has favorable pharmacokinetics, and antitumor and antiangiogenic activity in two different xenografts.

No potential conflicts of interest were disclosed.

Conception and design: J.O. Szablowski, J.A. Raskatov, P.B. Dervan

Development of methodology: J.O. Szablowski, J.A. Raskatov

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.O. Szablowski, J.A. Raskatov

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.O. Szablowski, J.A. Raskatov, P.B. Dervan

Writing, review, and/or revision of the manuscript: J.O. Szablowski, J.A. Raskatov, P.B. Dervan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.O. Szablowski, J.A. Raskatov

Study supervision: P.B. Dervan

The authors thank Caltech OLAR for technical assistance with animal experiments and TPCL (UCLA) for assisting with IHC. The authors also thank Drs. Nora Rozengurt and Dinesh Rao (UCLA) for helpful suggestions, discussion and preliminary pathologic evaluation of tumor sections and Drs. Nicholas Nickols and Bogdan Olenyuk for helpful discussions and suggestions.

This work was supported by the NIH grant GM-51747. J.O. Szablowski was supported by NIH GM-51747. J.A. Raskatov received postdoctoral support from the Alexander von Humboldt Foundation.

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.