Profilin-1 Phosphorylation Directs Angiocrine Expression and Glioblastoma Progression through HIF-1α Accumulation

SUMMARY

The tumor vascular microenvironment supports tumorigenesis by supplying not only oxygen and diffusible nutrients but also by secreting soluble factors that promote tumorigenesis. Here we identify a feed-forward mechanism in which endothelial cells (EC), in response to tumor-derived mediators, release angiocrines driving aberrant vascularization and glioblastoma multiforme (GBM) progression through a hypoxia-independent induction of hypoxia-inducible factor (HIF)-1α. Phosphorylation of profilin-1 (Pfn-1) at Tyr¹²⁹ in EC induces binding to tumor suppressor protein von Hippel-Lindau (VHL), prevents VHL-mediated degradation of prolyl-hydroxylated HIF-1α, culminating in HIF-1α accumulation even in normoxia. Elevated HIF-1α induces expression of multiple angiogenic factors, leading to vascular abnormality and tumor progression. In a genetic model of GBM, mice with an EC-specific defect in Pfn-1 phosphorylation exhibit reduced tumor angiogenesis, normalized vasculature, and improved survival. Moreover, EC-specific Pfn-1 phosphorylation is associated with tumor aggressiveness in human glioma. These findings suggest that targeting Pfn-1 phosphorylation may offer a selective strategy for therapeutic intervention of malignant solid tumors.

Tumors are now recognized as organ-like tissues of extreme complexity¹. Stromal cells, extracellular matrix, and soluble factors constitute the microenvironment that promotes tumor progression and metastasis, and induces therapeutic resistance^1–3. Newly formed blood vessels deliver oxygen and nutrients to solid tumors and are crucial for their growth⁴. However, overgrown, topologically and structurally abnormal vasculature characterizes the microenvironment of most highly malignant tumors. These vessels are characterized by tortuous morphology and excessive sprouting, and are highly proliferative as well as leaky⁵. They create a host-hostile but tumor-friendly microenvironment that fuels tumor progression⁶. Therapies targeting vascular normalization recently have joined anti-angiogenesis as important strategies for reconditioning the tumor microenvironment and treating cancer^5,7.

Glioblastoma multiforme (GBM), the grade IV glioma, is among the most malignant and highly vascularized tumors, with a current median survival of about 14 months in the United States^8,9. Most GBM tumors are refractory to conventional cytotoxic therapies, and during the past half-century there has been only limited improvement in patient treatment and prognosis¹⁰. GBM is distinguished by microvascular hyperplasia and extraordinary vascular abnormality of unknown etiology⁸. Anti-angiogenesis and vasculature-normalizing therapies, primarily targeting vascular endothelial growth factor (VEGF)-A and its receptors, have been developed and exploited in recent years. However, the therapeutic benefits have been small and transient, possibly due to contributions of other angiogenic factors, acquired resistance to VEGF-antagonistic treatments, and harmful treatment side-effects^11–13. Development of strategies that target signaling molecules at points of convergence downstream of discrete angiogenic growth factor pathways may provide a solution for improving the clinical outcomes of patients with GBM.

We have recently shown that Tyr¹²⁹ phosphorylation of profilin-1 (Pfn-1), a ubiquitously expressed actin-binding protein¹⁴, promotes sprouting angiogenesis after ischemic injury through regulation of actin filament dynamics and increased vascular endothelial cell (EC) motility¹⁵. Here, we show an unexpected role of Pfn-1 phosphorylation in inducing expression of a plethora of angiogenic factors that drive vascular abnormality and GBM progression. Pfn-1 phosphorylation directs expression of endothelial angiocrines, significantly contributing to global expression of angiogenic factors in tumors. Inhibition of this EC-specific event reduces GBM progression in a genetic mouse model that recapitulates the major features of human disease, suggesting that Pfn-1 phosphorylation represents a selective therapeutic target for treatment of GBM and other malignant tumors.

RESULTS

EC Pfn-1 phosphorylation at Tyr¹²⁹ drives GBM progression and aberrant vascularization

Tissue array-based, multiplex analysis of human breast, liver, brain, skin, and prostate tumors revealed marked elevation of Tyr¹²⁹-phosphorylated Pfn-1 in multiple brain tumors types, with maximum expression in GBM (Supplementary Fig. 1). To investigate the potential role of Pfn-1 phosphorylation in tumor progression, we took advantage of our conditional, EC-specific, phos-phorylation-deficient Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) knock-in mice¹⁵, and a genetically-engineered, allogeneic mouse model of GBM that recapitulates the major features of human GBM¹⁶. A donor glioma was formed by stereotactic injection of DF-1 fibroblasts infected with retroviral RCAS-platelet-derived growth factor (PDGF)-B into the cortex of Nestin-tv-a Ink4a-Arf^−/− Gli-luc donor mice (Fig. 1a). The nestin promoter drives PDGF-B expression specifically in neural stem/progenitor cells. A single-cell suspension of the glioma was injected into the cortex of Pfn1^Y129F knock-in (i.e., Tie2-Cre; Pfn1^{flox/flox:Y129F}) or Pfn1^WT wild-type (i.e., Pfn1^{flox/flox:Y129F}) mice recipients. Consistent with previous studies, Pfn1^WT mice developed aggressive tumors exhibiting prominent vascularity with microvascular hyperplasia and extensive regions of necrosis (Fig. 1b). Pfn1^WT mice became moribund after about 30 days, and all were dead by 70 days. EC knock-in of Pfn1^Y129F increased median survival by about 7 days (p < 0.05) (Fig. 1c), nearly comparable to the 10- to 14-day increase in mouse survival by temozolomide, the current first-line standard of care for patient treatment of glioma^17,18. Interestingly, 3 of 14 Pfn1^Y129F mice survived for at least 90 days when the experiment was terminated (Fig. 1c), further suggesting the significance of this microenvironment-dependent mechanism of tumor progression.

Figure 1.

Pfn-1 phosphorylation at Tyr¹²⁹ in EC induces aberrant vascularization and GBM progression. (a) GBM was induced by orthotopic injection of GBM tumor cells, and brain sections subjected to pathological and immunohistochemical analyses. The primary GBM in donor mice is induced by overexpression of PDGF in glial cells through RCAS/tv-a-mediated somatic gene transfer. Recipient mice were had conditional knock-in of Pfn1^Y129F/Y129F driven by EC-specific Tie2 promoter. (b) Four weeks after tumor cell injection, brains of Pfn1^{flox/flox:Y129F} mice were excised and sections subjected to H&E stain. (c) Survival was monitored for 90 days post-injection (n = 16 and 14 for Pfn1^{flox/flox:Y129F} and Tie2-Cre; Pfn1^{flox/flox:Y129F} mice, respectively), and analyzed by two-tailed Logrank test. (d) Tumor sections were probed with anti-CD31 antibody. Vascular density was quantified by immunofluorescence (mean ± SEM, n = 9 mice, two-tailed unpaired t test). (e) Hematoxylin and eosin (H&E)-stained sections. (f, g) Tumor sections were stained with anti-CD31 and anti-NG-2 antibodies (n = 5 mice). (h) Mice were perfused intravenously with FITC-dextran, and EC in tumor sections detected with anti-CD31 antibody (n = 5 mice). (i) Mice were injected with Hypoxyprobe-1 (pimonidazole HCl), and tumor sections probed with monoclonal antibody against pimonidazole adducts and detected with Alexa Fluor 561-conjugated IgG (n = 5 mice).

Tumors from Pfn1^Y129F knock-in mice have markedly reduced microvascular density (Fig. 1d), and the vessels exhibit morphological features typical of the vascular abnormality common in human GBM, i.e., tortuous, granular, and dilated with multi-layer wall thickening accompanied by extensive hemorrhage (Fig. 1e, f). Remarkably, blood vessels in Pfn1^Y129F knock-in mice appear essentially “normalized” as evidenced by non-tortuous, thin capillary vessels with less hemorrhage. Abnormal tumor vasculature is also characterized by insufficient coverage of nascent endothelium by pericytes, which hampers vessel maturation. Defective pericyte support and structural deformity of blood vessels are associated with poor functionality of the vasculature, leading to plasma leakage and spatially heterogeneous hypoxia¹⁹. Defective EC Pfn-1 phosphorylation substantially increased the extent of pericyte coverage on tumor blood vessels from 56.1 ± 8.2% to 87.3 ± 6.9% (mean ± SEM, n = 5 mice, P < 0.05 by two-tailed unpaired t test, Fig. 1g), inhibited leakage of intravenous fluorescein isothiocyanate (FITC)-dextran (as shown by decreased fluorescence intensity from 473 ± 117 to 145 ± 21, mean ± SEM, n = 5 mice, P < 0.05 by two-tailed unpaired t test, Fig. 1h), and reduced intra-tumoral hypoxia as determined by Hydroxyprobe-1 Kit (as shown by decreased fluorescence intensity from 718 ± 104 to 329 ± 32, mean ± SEM, n = 5 mice, P < 0.05 by two-tailed unpaired t test, Fig. 1i), Together these results indicate an important contribution of Pfn-1 phosphorylation to vascular abnormality.

EC Pfn-1 phosphorylation at Tyr¹²⁹ contributes to aberrant tube formation, exuberant proliferation, and monolayer hyperpermeability in vitro

To further investigate the EC-specific function of Pfn-1 phosphorylation, EC isolated from Pfn1^WT and Pfn1^Y129F knock-in mice were incubated with human U251 glioma cell-conditioned medium (GCM) on matrigel-coated dishes. Consistent with in vivo results, defective Pfn-1 phosphorylation of EC substantially reduced vessel thickening and sprouting in vitro (Fig. 2a). Similarly, GCM-stimulated proliferation (Fig. 2b) and monolayer permeability (Fig. 2c) was markedly reduced in Pfn1^Y129F EC. These results suggest that tumor microenvironment-dependent Pfn-1 phosphorylation drives EC over-proliferation and vascular hyperpermeability, and contributes importantly to microvascular hyperplasia during GBM progression. We previously established a role of Pfn-1 phosphorylation at Tyr¹²⁹ in EC motility and vessel sprouting, consistent with its restricted localization at the cell leading edge during chemotaxis in response to VEGF-A¹⁵. Interestingly, GCM induced accumulation of Tyr¹²⁹-phosphorylated Pfn-1 to the perinuclear area as well as the cell leading edge in migrating EC, in contrast to its restricted localization at cell protrusion sites in the absence of GCM (Fig. 2d), possibly indicating an alternate function of Pfn-1 phosphorylation in the glioma microenvironment.

Figure 2.

Pfn-1 phosphorylation at Tyr¹²⁹ is critical for in vitro vascular abnormality. (a–c) EC were isolated from Pfn1^{flox/flox:Y129F} (Pfn1^WT) and Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) mice, and treated with U251 glioma cell-conditioned medium (GCM). (a) Tube formation was induced on Matrigel. (b) EC proliferation determined by MTT assay (mean ± SEM, n = 5 biologically independent samples per group, one representative experiment shown, and the experiment was repeated 3 times). (c) Endothelial monolayer permeability was analyzed by fluorescence of FITC-dextran in transwell plate (mean ± SEM, n = 5 biologically independent samples per group, one representative experiment shown, and the experiment was repeated 3 times). (d) Human microvascular EC were treated with GCM, and stained with anti-P-Pfn-1-Tyr¹²⁹ and anti-Pfn-1 antibodies.

Pfn-1 phosphorylation at Tyr¹²⁹ is critical for expression of angiogenic growth factors in vitro and in vivo

We investigated the molecular mechanism underlying Pfn-1 phosphorylation-mediated vascular abnormality. Mouse GBM with defective EC-specific phosphorylation of Pfn-1 exhibited markedly reduced expression of the angiogenic growth factors VEGF-A and basic fibroblast growth factor (bFGF) (Fig. 3a). The result suggests that expression of Pfn-1 phosphorylation-mediated EC angiocrines contribute importantly to the burden of angiogenic growth factors in GBM. The specific role of EC as growth factor source was investigated by GCM stimulation of EC isolated from mouse aorta. An angiogenesis-specific antibody array showed that GCM-induced EC expression of multiple growth factors and cytokines, including VEGF-A, VEGF-B, placenta growth factor (PlGF), heparin-binding epidermal growth factor-like growth factor (HB-EGF), and bFGF, was substantially higher in Tyr¹²⁹ Pfn-1 phosphorylation-competent EC (Fig. 3b and Supplementary Fig. 2). The requirement for Tyr¹²⁹ phosphorylation and GCM stimulation for enhanced expression of select targets was confirmed by immunoblot analysis (Fig. 3c). Real-time PCR revealed that regulation is primarily at the level of mRNA expression (Fig. 3d). These results suggest that the abundant vasculature in the GBM microenvironment expresses angiogenic factors at a high level in a Pfn-1 phosphorylation-dependent manner, contributing to vascular abnormality and tumor progression.

Figure 3.

EC Pfn-1 phosphorylation at Tyr¹²⁹ induces HIF-1α-dependent expression of angiogenic growth factors in glioma microenvironment. (a) Tissue lysates from GBM tumors in Pfn1^{flox/flox:Y129F} and Tie2-Cre; Pfn1^{flox/flox:Y129F} mice were subjected to immunoblot analysis with antibodies described. (b–d) EC were isolated from Pfn1^{flox/flox:Y129F} (Pfn1^WT) and Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) mice, and treated with U251 glioma cell-conditioned medium (GCM). (b) Cell lysates were analyzed with a Mouse Angiogenesis Proteome Profiler Array. (c) Cell lysates were subjected to immunoblot analysis. (d) mRNA was isolated and analyzed by real-time RT-PCR (mean, 5 technical replicates per group, one representative experiment shown, the experiment was repeated 3 times, two-tailed unpaired t test). (e) EC were isolated from Pfn1^{flox/flox:Y129F} (Pfn1^WT) and Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) mice, and treated with GCM. Cell lysates were analyzed by immunoblot with anti-HIF-1α and anti-HIF-2α antibodies. (f, g) EC were transfected with siRNA targeting HIF-1α (or control), and incubated with GCM. (f) Cell lysates were subjected to immunoblot analysis. (g) EC proliferation and permeability were determined by MTT and transwell assays (mean ± SEM, n = 5 biologically independent samples per group, one representative experiment shown, and the experiment was repeated 3 times, two-tailed unpaired t test), respectively. (h) EC were transfected with siRNA targeting HIF-2α (or control), and incubated with GCM. Cell lysates were subjected to immunoblot analysis as shown. Uncropped images of blots are shown in Supplementary Fig. 8.

Pfn-1 phosphorylation-mediated expression of angiogenic factors requires HIF-1α

The expression of several phospho-Pfn-1-dependent factors, including VEGF-A, HB-EGF, and bFGF, is induced by the hypoxia-inducible factor (HIF)-1α transcription factor in tumors^20,21. Consistent with these findings, GCM stimulated HIF-1α and -2α expression, the induction of both requiring Tyr¹²⁹ Pfn-1 phosphorylation in EC (Fig. 3e). Remarkably, HIF-1α induction was observed even under normoxia, which ordinarily induces rapid ubiquitinylation and degradation. siRNA-mediated knockdown showed that HIF-1α is essential for GCM-stimulated induction of growth factors (Fig. 3f), as well as for increased EC proliferation and permeability (Fig. 3g). In contrast, HIF-2α knockdown did not appreciably inhibit growth factor expression (Fig. 3h). Transcriptome-wide analysis of tumor-derived EC shows diminished expression of multiple HIF-1α-inducible genes in Pfn-1 phosphorylation-deficient mice (Supplementary Fig. 3). These results reveal an unexpected role of Pfn-1 phosphorylation in HIF-1α–mediated expression of angiogenic factors, which might contribute to EC proliferation and vascular hyperpermeability in the glioma microenvironment.

Tyr¹²⁹-phosphorylated Pfn-1 induces accumulation of Pro-hydroxylated HIF-1α by inhibiting VHL-mediated degradation

During normoxia, HIF-1α is subject to prolyl hydroxylation and consequent degradation mediated by tumor suppressor protein von Hippel-Lindau (VHL), thus driving low-level expression of HIF-1α in physiological conditions^22,23. The hypoxic tumor microenvironment suppresses prolyl hydroxylation of HIF-1α and VHL binding, thereby inducing HIF-1α accumulation^24,25. To explore the mechanism by which Pfn-1 phosphorylation induces HIF-1α in normoxia, we investigated the stimulus-dependent interactions of phospho-Pfn-1 with VHL and HIF-1α. Angiogenic growth factors expressed in human GBM, i.e., VEGF-A, PlGF, bFGF, and PDGF^26,27, stimulated phosphorylation of Pfn-1 and its interaction with VHL (Fig. 4a); particularly robust stimulation of both processes was observed with GCM. Pfn-1 bearing the Y129F mutation did not bind VHL in GCM-treated EC indicating a strict requirement for phosphorylation (Fig. 4b). Co-immunoprecipitation confirmed the GCM-stimulated interaction between endogenous Pfn-1 and VHL (Fig. 4c). Furthermore, co-localization of Tyr¹²⁹-phosphorylated Pfn-1 and VHL was detected, primarily in the perinuclear region of GBM-stimulated EC (Fig. 4d).

Figure 4.

Tyr¹²⁹-phosphorylated Pfn-1 increases HIF-1α expression by blocking VHL-mediated degradation. (a) Human microvascular EC were transfected with pCMV-HA-VHL plus pcDNA-Pfn-1-Flag constructs (or pcDNA vector control). EC were treated with angiogenic factors or GCM, and lysates immunoprecipitated with anti-HA antibody and subjected to immunoblot analysis. (b) EC were transfected with pcDNA (vector), or pCMV-HA-VHL and pcDNA-Pfn-1-Flag with or without Y129F mutation, and treated with GCM as in (a). (c) EC were treated with GCM. Cell lysates were immunoprecipitated with anti-VHL or anti-Pfn-1 antibody, and subjected to immunoblot analysis as shown. (d) EC were treated with GCM and subjected to immunofluorescence analysis. (e) Purified Pfn-1^WT-His and Pfn-1^Y129A-His were in vitro phosphorylated with lysate from GCM-treated EC and re-purified with Ni⁺-beads, and proteins resolved by SDS/PAGE and stained with GelCode blue (left). Pfn-1 interaction with chip-immobilized VHL was determined by SPR (right). (f) VHL interaction with immobilized HIF-1α in the presence of Pfn-1^WT or Pfn-1^Y129F was determined by SPR. (g, h) EC were isolated from Pfn1^{flox/flox:Y129F} (Pfn1^WT) and Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) mice. EC were (g) exposed to hypoxia or treated with GCM, or (h) treated with GCM plus 0.1% DMSO (vehicle) or 10 mM MG132, and subjected to immunoblot analysis. (i) GBM was induced in Pfn1^{flox/flox:Y129F} (Pfn1^WT) mice, and tumor cells were isolated and cultured. Fibroblasts and tumor cells were treated with glioma-conditioned medium (GCM), followed by immunoblot analysis. Uncropped images of blots are shown in Supplementary Fig. 8.

To investigate direct binding of Pfn-1 to VHL, the proteins were expressed and purified, and their interaction analyzed by surface plasmon resonance (SPR) spectroscopy. Phosphorylated wild-type Pfn-1 (generated by in vitro incubation with lysate from GCM-activated cells or with purified Src kinase) exhibited greater affinity for VHL than the Y129F mutant (Fig. 4e and Supplementary Fig. 4). Furthermore, pre-incubation of VHL with lysate-treated Pfn-1^WT abrogated VHL binding to HIF-1α, whereas pre-incubation with Pfn-1^Y129F was ineffective, indicating that phosphorylation-dependent Pfn-1 binding to VHL prevents VHL binding to HIF-1α (Fig. 4f). Pfn-1 also interacts with G-actin in a Tyr¹²⁹-phosphorylation-dependent manner¹⁵. We found that G-actin competes with VHL for Pfn-1 binding, confirming the interaction of both partners at or near the same site (Supplementary Fig. 5).

Phospho-Pfn-1 competition with HIF-1α for VHL binding might lead to constitutive expression of HIF-1α in its Pro-hydroxylated form. Indeed, GCM treatment of EC in normoxia induced HIF-1α expression to nearly the same extent as hypoxic treatment in the absence of GCM; however, only GCM induced accumulation of Pro⁵⁶⁴-hydroxylated HIF-1α (Fig. 4g). In the presence of non-phosphorylatable Pfn-1, HIF-1α was induced by hypoxia but not by GCM. Treatment of EC with the proteasome inhibitor MG-132 induced robust accumulation of HIF-1α and Pro⁵⁶⁴-hydroxylated HIF-1α in both Pfn-1^WT and Pfn-1^Y129F cells, suggesting that Pfn-1 phosphorylation state does not influence prolyl hydroxylation of HIF-1α (Fig. 4h). GCM did not stimulate Pfn-1 phosphorylation or HIF-1α expression in mouse fibroblasts or GBM cells, indicating EC specificity of the pathway (Fig. 4i). Furthermore, knockdown of VHL substantially increased vessel thickening and sprouting on Matrigel and cell proliferation, and abolished the difference between Pfn-1^WT and Pfn-1^Y129F EC, consistent with an important contribution of VHL to Pfn-1 phosphorylation-mediated vascular abnormality in glioma condition (Fig. 5a–d).

Figure 5.

VHL is critical for Pfn-1 phosphorylation-mediated vascular abnormality. EC were isolated from Pfn1^{flox/flox:Y129F} (Pfn1^WT) and Tie2-Cre; Pfn1^{flox/flox:Y129F} (Pfn1^Y129F) mice and transfected with control or VHL siRNA. (a) Cell lysate from Pfn1^WT EC was resolved by SDS-PAGE and subjected to immunoblot analysis with anti- VHL and anti-GAPDH antibodies. (b, c) Tube formation was induced on Matrigel for 24 hr in the presence of glioma- conditioned medium (GCM). (b) Representative images. (c) Total tube length, the number of vessel sprouts, and the thick- ness of vessel walls were quantified (Mean ± SEM, n = 5 biologically independent samples per group, one representative experiment shown, and the experiment was repeated 3 times, two-tailed unpaired t test). (d) Cells were treated with GCM, and cell proliferation deter- mined by MTT assay (Mean ± SEM, n = 5 biologically independent samples per group, one representative experiment shown, and the experiment was repeated 3 times). Uncropped images of blots are shown in Supplementary Fig. 8.

To determine the in vivo role of this regulatory mechanism during GBM progression, Pro⁵⁶⁴-hydroxylated HIF-1α was visualized in sections of mouse GBM tumors. Pro-hydroxylated HIF-1α was abundant in the tumor and co-localized almost exclusively with CD31⁺ blood vessels (Fig. 6a, b); tumors of Pfn-1^Y129F mice exhibited very low-amounts of Pro-hydroxylated HIF-1α and HIF-1α (Fig. 6a–c). Similarly, human GBM specimens showed preferential co-localization of Pro-hydroxylated HIF-1α with EC, whereas total HIF-1α was distributed throughout the tissue (Fig. 6d). Together, these results reveal a non-canonical, hypoxia-independent mechanism for HIF-1α induction, by which phospho-Pfn-1-Tyr¹²⁹ binds VHL and prevents VHL-mediated HIF-1α degradation, leading to accumulation of Pro-hydroxylated HIF-1α in EC in culture and in vivo, even in normoxia.

Figure 6.

Expression of HIF-1α and Pro⁵⁶⁴-hydroxylated HIF-1α in EC of GBM tumor. (a–c) GBM was induced in Pfn1^{flox/flox:Y129F} and Tie2-Cre; Pfn1^{flox/flox:Y129F} mice (n = 5 mice). Brain sections were probed with anti-CD31, and (a, b) anti-hydroxy-HIF-1α-Pro⁵⁶⁴ or (c) anti-HIF-1α antibodies, and visualized with Alexa Fluor 488- and Alexa Fluor 561- conjugated second antibodies. (b) Brain sections retrieved from GBM tumor in Pfn1^{flox/flox:Y129F} mice were probed with anti-CD31 and anti-hydroxy- Pro⁵⁶⁴ or control isotype IgGs. (c) Left, representative images. Right, mean fluorescence intensity was quantified in CD31⁺ EC and tumor area (Mean ± SEM, n = 5 mice, two-tailed unpaired t test). (d) Brain sections from tumor biopsy specimens of GBM patients (n = 3 samples) were probed with anti-CD31 and -hydroxy-HIF-1α-Pro⁵⁶⁴ or -HIF-1α antibodies.

Phosphorylation of Pfn-1 at Tyr¹²⁹ is elevated in blood vessels of human GBM tumors, and is associated with glioma grade

Pfn-1 phosphorylation was determined in human biopsy specimens during glioma progression to GBM. Elevated phospho-Pfn-1-Tyr¹²⁹ expression was observed in Grades III and IV glioma, but not in Grades I and II (Fig. 7a). Quantitative immunofluorescence showed that Pfn-1 phosphorylation correlated with glioma grade in tumor regions and in EC (Fig. 7b, c). In all grades, phospho-Pfn-1-Tyr¹²⁹ expression was observed almost exclusively co-localizing with CD31⁺ blood vessels highly abundant in Grade IV (GBM) specimens (Fig. 7a, d). Moreover, essentially all blood vessels in GBM tumor were associated with phospho-Pfn-1-Tyr¹²⁹. Co-localization with CD144 (VE-cadherin)⁺ vessels confirmed EC-specific Pfn-1 phosphorylation (Fig. 7e). In contrast, phospho-Akt-Ser⁴⁷³, the critical regulatory product of PI3K activation, was more uniformly expressed, and co-localization with CD144⁺ blood vessels was minimal. Moreover, the distribution of total Pfn-1 was uniformly expressed throughout the tissue, and in healthy brain tissue (Fig. 7f). These results further support EC-selective phosphorylation of Pfn-1 during GBM progression. Growth factor-stimulated induction of Pfn-1 phosphorylation in EC, in tandem with the consequent increase in angiocrine expression, might constitute a feed-forward loop, driving uncontrolled growth factor expression, aberrant vascularization, and consequent accelerated disease progression (Fig. 8). Knockdown of HIF-1α inhibited GCM-induced Pfn-1 Tyr¹²⁹ phosphorylation supporting the proposed feed-forward mechanism of angiocrine production (Supplementary Fig. 6).

Figure 7.

Increased Tyr¹²⁹ phosphorylation of Pfn-1 in blood vessels of human glioma tumors correlates with grade. (a–d) Sections from biopsy specimens from subjects with glioma (n = 57 samples, total) and from normal brains (n = 6 samples) were probed with anti-P-Pfn-1-Tyr¹²⁹ and -CD31 antibodies. (a) Representative images. (b) P-Pfn-1-Tyr¹²⁹ in brain sections from glioma patients or from healthy controls was quantified by immunofluorescence using anti-P-Pfn-1-Tyr¹²⁹ antibody (bars indicate means, total n = 63 samples, two-tailed unpaired t test). Background fluorescence detected with omission of primary antibody was subtracted. (c) After background subtraction, fluorescence intensity of P-Pfn-1-Tyr¹²⁹ was normalized by area of CD31⁺ EC, and expressed as percentage of control (mean ± SEM, total n = 63 samples, two-tailed unpaired t test). (d) Quantitative analysis of co-localization of CD31⁺ EC with P-Pfn-1-Tyr129 (mean ± SEM, total n = 63 samples, two-tailed unpaired t test). (e) Human GBM specimens probed with anti-P-Pfn-1-Tyr¹²⁹, -CD144 and -P-Akt-Ser⁴⁷³ antibodies (n = 5 samples). (f) Uniform expression of Pfn-1 in human glioma tumors. Sections from biopsy specimens of human subjects with glioma (n = 4 samples) and from normal brains (n = 5 samples) were probed with anti-Pfn-1 antibodies.

Figure 8.

Schematic illustrating phospho-Pfn-1-mediated induction of HIF-1α driving growth factor production, aberrant vascularization, and GBM progression. The model shows that phospho-Pfn-1 interacts with VHL, prevents VHL-mediated degradation of HIF-1α, and induces accumulation of Pro-hydroxylated HIF-1α in GBM EC with pathological consequences.

Pfn-1 mRNA expression correlates with poor glioma patient survival

Consistent with the role of Pfn-1 phosphorylation in glioma progression, database analysis of GBM gene expression reveals elevated mRNA encoding Pfn-1, but not other actin-binding proteins including cofilin-1, formin, fascin, and N-WASP (Supplementary Fig. 7a, b), suggesting a non-canonical function of Pfn-1 besides regulation of actin dynamics. Moreover, high Pfn-1 mRNA expression predicts poor glioma patient survival (Supplementary Fig. 7c) suggesting the possible role of Pfn-1 as a biomarker of glioma aggressiveness. However, the specific role of Pfn-1 phosphorylation in patient survival has not been investigated.

DISCUSSION

HIF-1α is a critically important transcription factor instrumental in regulating diverse cellular responses to hypoxia, inducing expression of glycolytic enzymes and multiple angiogenic growth factors that drive aberrant vascularization during tumorigenesis^6,24,25,28. HIF-1α is upregulated in most malignant tumors, primarily by hypoxia-induced protein stabilization²⁵. Here we show a non-canonical, hypoxia-independent mechanism of HIF-1α induction in EC, in which phosphorylation of Pfn-1 at Tyr¹²⁹ inhibits VHL-mediated HIF-1α degradation by direct, competitive inhibition of VHL binding to HIF-1α. In turn, elevated HIF-1α induces expression of multiple angiogenic factors, eventually leading to aberrant vascularization and GBM progression (Fig. 8).

Phospho-Pfn-1 directs accumulation of HIF-1α even under normoxic conditions. Normoxic induction of HIF-1α by reactive oxygen species²⁹, PLC/PKC³⁰, and succinate^31,32 has been previously reported, and these agents inhibit protein hydroxylation thereby preventing recognition and degradation of HIF-1α by VHL. Likewise, human epidermal growth factor-like receptor 2 (HER2) induces HIF-1α expression in normoxia; however, the mechanism depends on increased synthesis of HIF-1α, not VHL-dependent degradation^33,34. HER2 expression correlates with the breast cancer progression and is an important diagnostic biomarker and therapeutic target for breast cancer, but is overexpressed in only 10–15% of GBM tumors^35–37. Here we describe a distinct, phospho-Pfn-1-driven mechanism that prevents recognition of prolyl hydroxylated HIF-1α by VHL, thereby facilitating accumulation of the modified protein in tumor EC, even in a normoxic microenvironment.

The classic role of EC is to provide physical structure that lines blood vessels, however, secretion of soluble factors by EC robustly regulates post-injury tissue regeneration and tumorigenesis^38–41. Autocrine and paracrine effects of angiogenic growth factors, including VEGF and basic FGF, on EC contribute importantly to vascular homeostasis and angiogenesis^21,42–44. Here, we show that expression of endothelial angiocrines driven by Pfn-1 phosphorylation significantly contributes to vascular abnormality and GBM progression. Endothelial angiocrines are a major source of angiogenic factors in GBM, likely due to prominent vascularization of the tumor. Importantly, extensive and abnormal vasculature in malignant solid tumors has been linked to resistance to radiation and chemotherapy, and poor patient outcome⁶.

In more than half of the GBM patients treated with anti-VEGF therapy, a transient response, therapeutic ineffectiveness, and harmful side-effects have been observed, possibly implicating other angiogenic factors, e.g., placenta-derived, hepatocyte, and fibroblast growth factors, in therapeutic resistance^12,13,45. Therapeutic strategies specifically targeting signaling molecules at points of convergence downstream of these diverse angiogenic growth factor pathways might prove to be more effective with lesser adverse, off-target consequences. This concept is the basis of intensive investigation of inhibitors of the pro-angiogenic PI-3-kinase/Akt pathway as therapeutic targets for GBM¹⁰. We have previously shown that Tyr¹²⁹ phosphorylation of Pfn-1 is a principal component of an alternative pathway of sprouting angiogenesis that is independent of and complementary to the PI-3-kinase/Akt pathway¹⁵. Here, we show that Pfn-1 phosphorylation may present a critical regulatory node in EC expression of angiogenic factors, in addition to its direct stimulation of sprouting angiogenesis, both critical events in tumor vascularization. Morevoer, Pfn-1 phosphorylation is preferentially induced in the blood vessels of GBM tumors, consistent with a role in regulation of EC-derived angiocrines, in contrast to the more uniform activation of PI-3-kinase/Akt in multiple cell types. Thus, Inhibition of Pfn-1 phosphorylation may offer a selective therapeutic strategy for targeting expression of multiple angiogenic factors in tumors.

Pfn-1 is an actin-binding protein that regulates processes involved in cell propulsion including increasing G-actin availability and polymerization, as well as protease expression^46–48. Our findings reveal an unconventional function of Pfn-1 in autocrine or paracrine signaling through HIF-1α induction and growth factor expression during GBM progression. Pfn-1 phosphorylation may have a major effect on HIF-1 stability rather than actin dynamics in glioma. Although Tie2 robustly drives EC gene expression, the Tie2 promoter is also activated in the myeloid lineage and thus Tie2-Cre-mediated knockin of Pfn1^Y129F might reduce Pfn-1 phosphorylation in macrophages^49,50. Certainly tumor-associated macrophage have an important role in tumor vascularization and progression in glioma^50,51, but the contribution of macrophage Pfn-1 phosphorylation is not known.

In summary, Pfn-1 phosphorylation regulates endothelial angiocrine expression and drives aberrant vascularization and GBM progression through a hypoxia-independent HIF-1α induction mechanism. As a corollary, prevention of Pfn-1 phosphorylation might offer a selective target and strategy to inhibit endothelial angiocrine expression and normalize aberrant vascularture, recondition the microenvironment, and improve the therapeutic window for traditional chemotherapy, for example, treatment of GBM and possibly other solid malignant tumors.