The protein tyrosine phosphatase PTP-PEST mediates hypoxia-induced endothelial autophagy and angiogenesis via AMPK activation

Physiological hypoxia is a potent agonist for embryonic development and post-natal angiogenesis. Oxygen concentrations ranging from 1% to 5% are observed in the uterine environment between embryonic days 3.5 to 14.5 (E3.5–E14.5) to facilitate development of the placenta. Similarly, in response to hypoxia (<2% oxygen), endocardial and vascular endothelial cells mediate formation of foetal heart and vasculature, respectively, in mice between embryonic days E7.5 and E15 (Dunwoodie, 2009; Simon and Keith, 2008). Post-natal angiogenesis, as seen in the female reproductive tract during the menstrual cycle, as well as formation of collateral circulation to overcome coronary artery blocks, also depends on hypoxia-induced endothelial signalling.

Multiple studies in rodents have reported increased activity of cytosolic protein tyrosine phosphatases (PTPs) at sites of post-natal angiogenesis, including ischemic myocardium and skeletal muscles (Sugano et al., 2004; Yang et al., 2020). An increase in the cytosolic PTP activity during hypoxia is also seen in the cerebral cortex of newborn piglets (Ashraf et al., 2004). Paradoxically, others have shown that a non-selective PTP inhibitor, sodium orthovanadate, enhances VEGFR2 (also known as KDR) signalling and capillary morphogenesis (Montesano et al., 1988; Sugano et al., 2004). Although these studies allude to the involvement of different PTPs in hypoxia-induced angiogenic signalling, barring the involvement of a few, for instance VE-PTP (also known as PTPRB; a receptor PTP), the identity of hypoxia-responsive angiogenic cytosolic PTPs remains largely unknown.

Human genome encodes 38 classical PTPs, which are specific for tyrosine residues (Barr et al., 2009). Among these, PTP-PEST (also known as PTPN12) is a ubiquitously expressed cytosolic PTP with an N-terminal catalytic domain (1–300 amino acids) and a regulatory C-terminal PEST domain (301–780 amino acids), which consists of four PEST motifs. The latter plays a crucial role in protein–protein interaction, allowing the enzyme to interact with its known substrates [Cas, Paxillin, FAK (PTK2) or Pyk2 (PTK2B)], in addition to being a protein degradation signal (Davidson and Veillette, 2001; Lee and Rhee, 2019; Veillette et al., 2009). The catalytic domain harbours the conserved phosphatase ‘HC[X]₅R’ motif, surrounded by the ‘WPD loop’ and the ‘Q loop’ on one side to assist in catalysis, and by the ‘P-Tyr-loop’ on the other side to regulate substrate recognition and specificity (Dong et al., 2017). Both global and endothelial deficiency of PTP-PEST in mice leads to embryonic lethality between E9.5 and E10.5 due to defective heart formation and impaired endothelial network in the yolk sac (Sirois et al., 2006; Souza et al., 2012). It is worth noting that this stage in embryonic development coincides with hypoxia-induced morphogenesis (Dunwoodie, 2009; Simon and Keith, 2008). PTP-PEST appears to be crucial for vascular development even in humans, since a partial deletion of PTP-PEST is associated with disrupted aortic arch development (Duffy et al., 2015). Intriguingly, in a recent study, enhanced endothelial expression of PTP-PEST was observed in the vascularized core of glioblastoma tumours, with others demonstrating involvement of PTP-PEST in integrin-mediated endothelial cell adhesion and migration (Chen et al., 2018; Souza et al., 2012). Based on these observations, which illustrate an essential role of PTP-PEST in vascular development, and given the fact that hypoxia is indispensable for angiogenesis, in this study we set out to determine the functional role of PTP-PEST in hypoxia-induced endothelial responses. We assessed the effect of hypoxia on the expression, activity and the binding partners of PTP-PEST in primary human umbilical vein-derived endothelial cells (HUVECs).

To determine the effect of hypoxia (1% oxygen) on protein levels and enzyme activity of PTP-PEST in endothelial cells, HUVECs were cultured as monolayer and exposed to hypoxia for different time points. As can be seen in Fig. 1A and summarized in Fig. 1B, immunoblotting for PTP-PEST demonstrated a significant increase in protein levels from 3 h onwards that was sustained until 24 h. Increase in protein levels of HIF-1α (also known as HIF1A) confirmed induction of hypoxia in these cells. This experiment was performed on ten biological replicates, each time with a fresh batch of HUVECs. Furthermore, to determine whether this effect of hypoxia on PTP-PEST expression is an endothelial-specific phenomenon, or is also observed in other cell types, we checked for changes in PTP-PEST expression in response to hypoxia (1% oxygen) in other cell lines, including HEK293 (n=5), HA-VSMC (human aortic vascular smooth muscle cell; n=3), HeLa (human cervical epitheloid carcinoma cell line; n=4) and Huh7 (hepatocyte derived; n=3). As seen in Fig. 1C, hypoxia promotes PTP-PEST protein expression even in other cell lines, suggesting this to be a universal phenomenon. Equal loading and induction of hypoxia were confirmed for each of the cell lines via immunoblotting for β-actin and HIF-1α, respectively (Fig. 1C). We also checked for changes in subcellular localization of PTP-PEST in response to hypoxia (1% oxygen) by immunofluorescence imaging. Cytoplasmic localization of PTP-PEST was observed in normoxia and did not alter upon hypoxia (1% oxygen) treatment (Fig. S1A). This experiment was performed thrice, each time with a new batch of HUVECs.

Since endothelial cells, by virtue of their location, are one of the early responders to changes in oxygen tension and are resilient to hypoxia to promote adaptive angiogenesis (Filippi et al., 2018; Koziel and Jarmuszkiewicz, 2017), we employed primary endothelial cells (HUVECs) as model cell system to determine the role of PTP-PEST in hypoxia-mediated cellular responses. We set out to determine whether hypoxia (1% oxygen) influences the catalytic activity of PTP-PEST. Equal amounts of endogenous PTP-PEST were immunoprecipitated from endothelial lysates taken at different time points of hypoxia treatment (Fig. 1D; Fig. S1B) and were subjected to an immunophosphatase assay. As seen in the bar graph of Fig. 1D, hypoxia induced a significant increase in phosphatase activity of PTP-PEST, with the activity being maximal at 3 h compared with normoxia treatment. Thus, hypoxia increases both protein levels and enzyme activity of PTP-PEST.

To understand the functional relevance of hypoxia-mediated enhanced expression and activity of PTP-PEST, it was necessary to recognize its binding partners. Most of the reported binding partners of PTP-PEST are involved in cell adhesion, migration and cytoskeletal reorganization (Davidson and Veillette, 2001; Lee and Rhee, 2019; Veillette et al., 2009). We wanted to identify binding partners of PTP-PEST, specifically in endothelial cells exposed to hypoxia and normoxia. For this, HUVECs were transduced with N-terminal GST-tagged PTP-PEST lentiviral particles (as described in the Materials and Methods). Post-transduction, the same pool of cells was split and exposed to either normoxia or hypoxia (1% oxygen) for 24 h. Equal amounts of GST-tagged PTP-PEST were pulled down via immunoprecipitation using an anti-GST antibody, and the co-immunoprecipitated proteins were subjected to liquid chromatography-mass spectrometry (LC/MS/MS) analysis (see Materials and Methods).

The protein IDs obtained after mass spectrometry (MS) data were filtered for non-specific binding partners by removal of proteins appearing in MS of the isotype control sample. In addition, MS contaminants, as per the common repository of adventitious proteins (https://www.thegpm.org/crap/), as well as proteins that are a part of the Sepharose bead proteome were excluded from the analysis (Trinkle-Mulcahy et al., 2008). Proteins unique only to normoxia or hypoxia were subjected to analysis using the gene ontology program PANTHER (Thomas et al., 2003). The prominent binding partners observed solely in normoxia were signalling molecules, transcription factors and cytoskeletal proteins (Fig. 2A),whereas binding partners such as chaperones, hydrolases and oxidoreductases appeared exclusively in hypoxia (Fig. 2A). Some of the specific binding partners observed in normoxia and hypoxia are listed in Fig. 2B. In hypoxia, we found proteins including OTUB1, PGK1 and PKC-ε (PRKCE), which are known to play an important role in regulating autophagy, interacting with PTP-PEST. Surprisingly, proteomic studies revealed AMPK α₁ and α₂ (encoded by PRKAA1 and PRKAA2, respectively), the catalytic subunits of 5′-AMP-activated protein kinase (AMPK), as interacting partners of PTP-PEST in normoxia (Fig. 2B). Co-immunostaining of PTP-PEST and AMPK α subunits (referred to hereafter as AMPKα) followed by confocal immunofluorescence imaging demonstrated that both PTP-PEST and AMPKα indeed colocalize in the cytoplasm of HUVECs (Fig. 2C).

Intriguingly, our proteomics data revealed that interaction of PTP-PEST with AMPKα was lost upon hypoxia treatment. In order to validate the interaction of PTP-PEST with AMPKα, we performed co-immunoprecipitation experiments. Equal amounts of endogenous PTP-PEST were pulled down from normoxia- and hypoxia (1% oxygen, 24 h)-exposed HUVECs and immunoblotting was performed for AMPKα. It is worth noting that the anti-AMPKα antibody used for immunoblotting recognizes both α₁ and α₂ isoforms of the α subunit. In accordance with the proteomics data, we found interaction of PTP-PEST with AMPKα in normoxic conditions; however, this interaction was abrogated in HUVECs exposed to hypoxia (1% oxygen) for 24 h (Fig. 3A). Equal pulldown of PTP-PEST across the two conditions was confirmed by re-probing the blot for PTP-PEST. This experiment was performed four independent times, each time with a different batch of HUVECs.

PTP-PEST consists of an N-terminal catalytic domain and a long C-terminal non-catalytic domain, rich in PEST sequences. The C-terminal domain plays a significant role in regulating the enzyme activity of PTP-PEST by facilitating its interaction with either substrates and/or other adaptor proteins (Davidson and Veillette, 2001; Lee and Rhee, 2019; Veillette et al., 2009). Next, we checked whether the PEST domain is also essential for the interaction of PTP-PEST with AMPKα. We cloned the His-tagged wild type (WT) and enzymatically inactive (C231S) N-terminal catalytic domain (1–300 amino acids) lacking the C-terminal PEST sequences of human PTP-PEST into the pET28a(+) bacterial expression vector. The two proteins were overexpressed and purified from the E. coli BL21 codon plus RIL (DE3) strain (Fig. S2A–C). A total of 100 µg of purified WT and C231S mutant PTP-PEST protein lacking the PEST motifs were independently incubated with 500 µg of HUVEC lysate for 8 h at 4°C. This was followed by immunoprecipitation of PTP-PEST using an anti-His-tag antibody and immunoblotting for AMPKα (Fig. 3B). Equal pulldown of PTP-PEST was confirmed by re-probing the blot for PTP-PEST. We found co-immunoprecipitation of AMPKα with the N-terminal catalytic domain of PTP-PEST (Fig. 3B), which suggests that the interaction of AMPKα with PTP-PEST is not dependent on the C-terminal PEST domain. Moreover, co-immunoprecipitation of AMPKα with the C231S (catalytically inactive) mutant was greater in comparison to that with WT (Fig. 3B), despite equal pulldown of PTP-PEST protein across the two conditions, suggesting that AMPKα is a likely substrate of PTP-PEST. This experiment was performed thrice, each time with a fresh batch of HUVECs.

AMPK is a heterotrimeric stress responsive serine-threonine kinase known to play an important role in maintaining cellular energy homeostasis as well as autophagy (Jeon, 2016). Next, we wanted to understand the relevance of the interaction of AMPKα subunit with PTP-PEST, a tyrosine phosphatase. Since PTP-PEST interacts with AMPKα via its catalytic domain, we were interested in examining whether PTP-PEST can dephosphorylate AMPKα. First, we determined the effect of hypoxia on total tyrosine phosphorylation of AMPKα. We initially tried immunoprecipitating endogenous AMPKα but faced some difficulties. Hence, we resorted to an alternative approach that was employed by Yamada et al. (2016). For this, HUVECs were exposed to hypoxia (1% oxygen) for different time periods. Equal amounts of HUVEC lysate (300 µg) following experimental treatment were subjected to immunoprecipitation with 3 µg of anti-phospho-tyrosine antibody, followed by immunoblotting for AMPKα. Tyrosine phosphorylation of AMPKα in HUVECs exposed to hypoxia for 24 h was lower in comparison to that in HUVECs exposed to normoxia (Fig. 3C), indicating dephosphorylation of AMPKα. It is worth noting that the total AMPKα levels in endothelial cells did not change across treatment conditions (Fig. 3C). The experiment was performed in n=4 biological replicates, each time with a different batch of HUVECs.

Next, we knocked down PTP-PEST using shRNA lentivirus to check its effect on AMPKα tyrosine dephosphorylation in response to hypoxia (1% oxygen, 24 h). Scrambled shRNA lentivirus was used as negative control. We found that hypoxia-induced AMPKα dephosphorylation was indeed abrogated upon PTP-PEST knockdown (Fig. 3D,E). In fact, in the absence of PTP-PEST, the basal tyrosine phosphorylation of AMPKα itself was increased, whereas the total protein levels of AMPKα remained unchanged upon PTP-PEST knockdown. Successful knockdown of PTP-PEST was confirmed through western blotting, as seen in Fig. 3D. Thus, three independent experiments confirmed that PTP-PEST does regulate tyrosine dephosphorylation of AMPKα in response to hypoxia without influencing AMPKα protein levels.

AMPK can be regulated both allosterically and by means of posttranslational modifications. Phosphorylation at Thr¹⁷² residing in the kinase domain of the α subunit is an efficient mechanism of AMPK activation (Jeon, 2016). Therefore, next, the effect of hypoxia on AMPK activation in endothelial cells was examined. HUVECs were exposed to hypoxia for different time points, and immunoblotting was performed for phosphorylation of Thr¹⁷². Thr¹⁷² phosphorylation was seen in response to hypoxia as early as 15 min and was sustained up to 24 h (Fig. 3F; Fig. S2D) for five independent experiments. Since we observed that PTP-PEST mediates hypoxia-induced tyrosine dephosphorylation of AMPK, we set out to determine if absence of PTP-PEST also influences Thr¹⁷² phosphorylation of AMPK. For this purpose, HUVECs were transduced with PTP-PEST shRNA and scrambled shRNA lentivirus followed by hypoxia treatment (1% oxygen, 24 h) and immunoblotting for phospho-Thr¹⁷² AMPKα. Interestingly, we found that hypoxia-induced AMPK activation was attenuated in HUVECs upon PTP-PEST knockdown (Fig. 3G). Simultaneously, we also checked for phosphorylation of ACC at Ser⁷⁹ (acetyl-CoA carboxylase, a known substrate of AMPK) as a readout of AMPK activity. We found that hypoxia-induced phosphorylation of ACC at Ser⁷⁹ was also attenuated in PTP-PEST knockdown cells (Fig. 3G). These observations, in conjunction with loss of interaction between PTP-PEST and AMPKα under hypoxia, demonstrate that AMPKα is a substrate of PTP-PEST, wherein PTP-PEST mediates tyrosine dephosphorylation of AMPKα and regulates its catalytic activity.

AMPK is known to regulate autophagy via dual mechanisms involving inactivation of mTORC1 and direct phosphorylation of ULK1 at Ser³¹⁷ (Alers et al., 2012; Kim et al., 2011). Because we observed that PTP-PEST-dependent tyrosine dephosphorylation of AMPKα is associated with its activation in response to hypoxia, we next wanted to understand the role of PTP-PEST in hypoxia-induced endothelial cell autophagy. For this, we first tested the effect of hypoxia on endothelial autophagy followed by knockdown experiments. HUVECs exposed to hypoxia (1% oxygen) for different time intervals displayed an increase in LC3 (here referring to MAP1LC3A and MAP1LC3B) degradation (LC3II form) following hypoxia (Fig. S3A,B). An increase in abundance of the LC3II form indicates induction of autophagy, because it plays an indispensable role in autophagosome biogenesis. Along with an increase in LC3II levels, other autophagic markers like beclin-1 and phosphorylation of ULK1 at Ser³¹⁷ were also enhanced in response to hypoxia (Fig. S3A,C,D). In addition, a decrease in ULK1 Ser⁷⁵⁷ phosphorylation, another indicator of autophagy, was observed (Fig. S3A,E). It should be noted that Ser⁷⁵⁷ dephosphorylation of ULK1 reflects inactivation of mTORC1 (Kim et al., 2011). We also examined LC3 puncta formation in response to hypoxia using LC3 immunostaining followed by confocal imaging. As seen in Fig. S3F, high accumulation of LC3 puncta representing autophagosomes was observed in HUVECs treated with hypoxia (1% oxygen, 24 h). This led to a significant increase in the number of puncta per cell, as well as in the percentage of LC3-positive cells (Fig. S3G,H). Treatment with bafilomycin A1 (100 nM) further increased the number of LC3 puncta per cell, indicating that autophagy was in progress during hypoxia. Bafilomycin A1 treatment was given for the last 6 h of hypoxia treatment. These experiments were performed thrice, each time with a new batch of HUVECs.

Next, we wanted to determine whether hypoxia-induced autophagy was dependent on PTP-PEST-mediated AMPK activation. As seen in Fig. 4A,B, hypoxia (1% oxygen, 24 h)-induced LC3 degradation was significantly attenuated upon knockdown of PTP-PEST in five independent experiments. Further, we performed LC3 immunostaining in scrambled and PTP-PEST shRNA-treated endothelial cells in the absence and presence of the AMPK activator metformin (500 µM). Metformin treatment was given for 24 h along with hypoxia. As can be seen in Fig. 4C–E, the hypoxia-induced increase in the number of LC3 puncta per cell, as well as in the percentage of cells with puncta-like structures, was significantly attenuated upon PTP-PEST knockdown. This attenuation in autophagy due to the absence of PTP-PEST was, however, recovered upon metformin treatment, leading us to conclude that PTP-PEST is necessary for hypoxia-induced autophagy via AMPK activation. These experiments were repeated thrice, each time with a fresh batch of HUVECs.

Hypoxia, the principal physiological stimulus for angiogenesis, regulates multiple steps of sprouting angiogenesis including endothelial cell migration, tube formation and vessel branching (Forsythe et al., 1996; Krock et al., 2011; Rankin et al., 2008). Additionally, induction of autophagy in response to hypoxia is both a protective survival mechanism as well as an inducer of angiogenesis in endothelial cells (Du et al., 2012; Liang et al., 2018). We thus examined the effect of PTP-PEST knockdown on hypoxia-induced angiogenic responses such as migration and capillary tube formation. As seen in Fig. 5A and B, both basal and hypoxia (1% oxygen, 24 h)-induced endothelial cell migration was attenuated in PTP-PEST knockdown cells in wound healing experiments. Scrambled shRNA lentivirus was used as a mock transduction. The knockdown of endogenous PTP-PEST was confirmed through western blotting (Fig. 5C). This experiment was repeated four times, each time with a fresh batch of endothelial cells.

To examine the role of PTP-PEST in hypoxia-induced tube formation in vitro, HUVECs (15×10³ cells) were placed on growth factor-free Matrigel in low serum (1% FBS) MCDB 131-based endothelial growth medium after transduction with scrambled shRNA or PTP-PEST shRNA lentivirus, and tube formation was allowed under normoxic and hypoxic (1% oxygen) conditions. The tubular network (Fig. 6A) was quantified after 16 h of incubation using ImageJ software. We observed an increase in the number of segments (Fig. 6C), number of junctions (Fig. 6D) and the total length of tubes (Fig. 6E) in response to hypoxia (1% oxygen). Among these effects of hypoxia, PTP-PEST knockdown significantly attenuated the number of segments and junctions. PTP-PEST knockdown was confirmed through western blotting (Fig. 6B). Interestingly, use of rapamycin (200 nM), an mTOR inhibitor and autophagy inducer, reversed the effect of loss of PTP-PEST during hypoxia by increasing the number of segments (Fig. 6A,C) and junctions (Fig. 6A,D) of tubes for PTP-PEST-knockdown cells. These experiments were repeated four times, each time with a different batch of HUVECs. Taken together, these observations indicate that PTP-PEST promotes hypoxia-induced angiogenesis through the AMPK-dependent autophagy pathway.

The results of this study demonstrate that hypoxia increases the expression and catalytic activity of the cytosolic protein tyrosine phosphatase PTP-PEST to promote endothelial autophagy and consequent angiogenesis. These functional effects of hypoxia are dependent on PTP-PEST-mediated tyrosine dephosphorylation and activation of AMPK.

As the name suggests, AMPK is a physiological energy sensor that is activated in response to increased intracellular concentrations of AMP under stress conditions of hypoxia, calorie restriction or exercise. Upon activation, it inhibits energy-utilizing anabolic processes and instead promotes catabolic processes such as fatty acid oxidation, glycolysis and autophagy (Jeon, 2016; Long and Zierath, 2006). AMPK holoenzyme is a heterotrimeric complex composed of a catalytic α subunit in complex with two regulatory subunits, β and γ (Ross et al., 2016). Binding of AMP molecules to the γ subunits during stress mediates allosteric activation of the catalytic domain of the α subunits. The kinase domain (KD) of the α subunit harbours the Thr¹⁷² residue in the catalytic cleft formed by its N- and C-terminal lobes (Fig. 7) (Stein et al., 2000; Woods et al., 2003). In an inactive state, the back end of these lobes opposite to the catalytic cleft are held by the evolutionarily conserved auto-inhibitory domain (AID). The AID in turn is connected to a flexible regulatory motif termed the α-regulatory subunit-interacting motif (α-RIM). Multiple crystallography studies have demonstrated that conformational changes in the α-RIM promote pulling of the AID away from the kinase domain to relieve inhibition (Fig. 7) (Chen et al., 2009; Li et al., 2017). As a consequence of this pulling away, Thr¹⁷² at the catalytic site becomes accessible for phosphorylation by LKB1 (also known as STK11) or CAMKKβ (also known as CAMKK2) to enhance AMPK enzyme activity several hundredfold (Hawley et al., 2005; Woods et al., 2003). Two isoforms of α subunit (α₁ and α₂), which exhibit considerable sequence identity and conformational similarity, have been reported (Calabrese et al., 2014). It is worth noting that both these isoforms of α subunit interacted with PTP-PEST under normoxic conditions, with the interaction being lost in hypoxia. Interestingly, both these isoforms of α subunit are expressed in endothelial cells, where they participate in hypoxia-induced angiogenesis (Davis et al., 2006; Nagata et al., 2003).

In the current study, it was found that hypoxia-induced activation of AMPK was dependent on PTP-PEST. The catalytic domain of PTP-PEST (1–300 amino acids) was sufficient for interaction with AMPKα, thereby suggesting that AMPKα subunits are PTP-PEST substrates. The endogenous interaction between AMPKα and PTP-PEST was lost in hypoxia. This loss of interaction coincided with tyrosine dephosphorylation of AMPKα. Interestingly, knockdown of PTP-PEST not only enhanced the basal tyrosine phosphorylation of AMPK, but also prevented hypoxia-mediated activation of AMPK, as assessed through Thr¹⁷² phosphorylation. To the best of our knowledge, this is the first study to directly demonstrate regulation of AMPK activity by a cytosolic protein tyrosine phosphatase. Although not much is known about the regulation of AMPK by tyrosine phosphorylation, a recent study demonstrated that phosphorylation of Tyr⁴³⁶ in the α-subunit (Fig. 7) reduces AMPK catalytic activity by modulating the AID–αRIM interaction (Yamada et al., 2016). It is thus tempting to speculate that PTP-PEST may dephosphorylate this residue of the α-subunit under hypoxic conditions to activate AMPK by inducing a conformational change in the flexible α-RIM domain to relieve auto-inhibition imparted by AID (Fig. 7). This possibility, however, needs to be examined in future studies.

One of the major consequences of AMPK activation in endothelial cells is induction of autophagy to promote survival and angiogenesis (Filippi et al., 2018; Jeon, 2016; Long and Zierath, 2006; Stempien-Otero et al., 1999). This adaptive response imparts resilience to endothelial cells towards stress conditions such as heat shock, hypoxia, shear stress and calorie restriction (De Meyer et al., 2015; Jiang, 2016; Nussenzweig et al., 2015). In fact, defective endothelial autophagy is associated with vascular aging, thrombosis, atherosclerosis and even arterial stiffness (Jiang, 2016; Xie et al., 2011). AMPK activates autophagy either by phosphorylating ULK1 at Ser³¹⁷ or by inhibiting mTORC1 (Alers et al., 2012; Kim et al., 2011). Both these events lead to autophagosome formation. To couple the changes in PTP-PEST and AMPK activity with endothelial cell function, we assessed the consequence of PTP-PEST knockdown on hypoxia-induced autophagy. We found that despite induction of hypoxia, autophagy was attenuated in the absence of PTP-PEST. This effect was, however, rescued by the AMPK activator metformin, demonstrating that AMPK activation lies downstream of PTP-PEST in hypoxia-induced endothelial signalling. It is worth noting that metformin treatment also promotes endothelial migration and angiogenesis in experimental models of hypoxia such as stroke (Bakhashab et al., 2018; Jin et al., 2014; Nagata et al., 2003; Venna et al., 2014). Thus, AMPK is indispensable for both autophagy and angiogenesis. Likewise, we have identified an essential role of PTP-PEST in hypoxia-induced AMPK activation, autophagy and consequent angiogenesis. It is interesting to note that apart from activating AMPK, PTP-PEST also interacted with other proteins including ubiquitin thioesterase (OTUB1), PKC-ε, myotubularin-related protein 6 (MTMR6) and sarcolemmal membrane-associated protein (SLMAP), all of which are either directly involved in autophagy or its regulation (Lakshmanan et al., 2016; Mochizuki et al., 2013; Toton et al., 2018; Zhao et al., 2018). Thus, the current findings identify a hitherto unreported role of PTP-PEST in regulating endothelial autophagy via activation of AMPK.

PTP-PEST is an efficient enzyme with a K_cat/K_m greater than 7×10⁶ M⁻¹ s⁻¹ (Li et al., 2016; Selner et al., 2014). A key finding of the current study is that hypoxia increases both expression and activity of PTP-PEST. It is presently unclear how hypoxia brings about these effects, but a preliminary CONSITE-based in silico analysis of the putative promoter region of PTP-PEST indicates presence of multiple HIF-1α binding sites (data not shown), suggesting a possible increase in PTP-PEST transcription in response to hypoxia. Alternatively, hypoxia may enhance the protein stability of PTP-PEST. A great deal is known about the ability of the ‘PEST’ sequences to regulate protein stability through interactions with ubiquitin ligases (Rechsteiner and Rogers, 1996). Posttranslational modifications of proline-rich ‘PEST’ sequences through phosphorylation or prolyl hydroxylation assist in these interactions (García-Alai et al., 2006; Ravid and Hochstrasser, 2008). HIF-prolyl hydroxylases (PHDs), in the presence of oxygen, hydroxylate proline residues to recruit von Hippel–Lindau (vHL) ubiquitin ligases (Strowitzki et al., 2019). However, under hypoxic conditions they fail to hydroxylate prolines and thus fail to induce protein degradation. In fact, this mode of regulation is primarily responsible for increasing the stability of HIF-1α under hypoxia. PHD2 (also known as EGLN1), a major oxygen sensor in endothelial cells, is found to be higher in the nuclear fraction during hypoxia (Berchner-Pfannschmidt et al., 2008), with nuclear localization of PHD2 being necessary for induction of its hydroxylase activity (Pientka et al., 2012). Since we observed cytoplasmic retention of PTP-PEST during hypoxia, as seen in Fig. S1A, it is possible that under hypoxic conditions, nuclear PHD2 fails to hydroxylate cytoplasmic PTP-PEST, thereby enhancing the protein stability of this phosphatase during hypoxia. The fact that we observed an increased interaction of deubiquitinase OTUB1 with PTP-PEST during hypoxia also supports the notion that hypoxia may increase stability of the PTP-PEST protein.

Increased activity of PTP-PEST could be a result of several events, including changes in subcellular localization, protein–protein interactions or even posttranslational modifications in response to hypoxia. We did not observe any change in subcellular localization of PTP-PEST in response to hypoxia. However, it is possible that PTP-PEST may undergo serine, threonine or tyrosine phosphorylation upon induction of hypoxia. Indeed phosphorylation of Ser³⁹ is known to reduce the substrate specificity and activity of PTP-PEST (Lee and Rhee, 2019). Likewise, Tyr⁶⁴, an evolutionarily conserved residue in the ‘P-Tyr-loop’ of PTP-PEST is reported to be essential for its catalytic activity (Li et al., 2016). In contrast, Ser⁵⁷¹ phosphorylation enhances PTP-PEST substrate binding (Zheng et al., 2011). Whether hypoxia induces phosphorylation of PTP-PEST to modulate its stability and/or activity is the subject of ongoing investigation in our laboratory.

By virtue of its ability to regulate important cellular processes such as cell adhesion and migration of numerous cell types including embryonic fibroblasts, endothelial cells, dendritic cells, T cells and macrophages, PTP-PEST plays an essential role in cardiovascular development, tissue differentiation and immune function (Lee and Rhee, 2019; Souza et al., 2012; Veillette et al., 2009). Yet the role of PTP-PEST in cancer is controversial, with some regarding it as a tumour suppressor, whereas others associate it with metastasis and tumour vasculature (Chen et al., 2018; Lee and Rhee, 2019; Zheng et al., 2011). Our observations of PTP-PEST regulating hypoxia-induced autophagy in native endothelial cells identifies a new functional role of this phosphatase in endothelial physiology. Whether this role of endothelial PTP-PEST is of consequence in tumour angiogenesis under transformed settings needs to be elucidated in future studies. In conclusion, data presented here demonstrate that hypoxia increases expression and activity of cytosolic PTP-PEST, which in turn activates AMPK to promote endothelial autophagy and angiogenesis.

Experimental procedures involving isolation of endothelial cells from umbilical cords was approved by the IIT-Madras Institutional Ethics Committee as per the Indian Council of Medical Research (ICMR), Government of India guidelines. These 2017 ICMR guidelines are in accordance with the Declaration of Helsinki.

Tissue culture-grade plasticware was from Tarsons Products Pvt. Ltd., Kolkata, West Bengal, India. Antibodies against HIF-1α (CST #3716), phospho-AMPK (CST #2535), total AMPK (CST #2532), LC3 (CST #4108), phospho-ULK1 (Ser⁷⁵⁷) (CST #6888), phospho-ULK1 (Ser³¹⁷) (CST #12753), total ULK1 (CST #8054), phospho-ACC (CST #11818), total ACC (CST #3676), PTP-PEST (AG10) (CST #4864), GST (CST #2622), His-Tag (CST #2365) and phospho-tyrosine (P-Tyr-102) (CST #9416) were obtained from Cell Signaling Technology, Boston, MA, USA. All the antibodies were used at recommended dilutions, as per product datasheets. Dulbecco's modified Eagle's medium (DMEM) and MCDB131 medium were purchased from HiMedia, Mumbai, Maharashtra, India, and foetal bovine serum (FBS) was from Gibco, Thermo Fisher Scientific, Waltham, MA, USA. Endothelial cell growth medium SingleQuots supplements (CC-4133) required for culture of endothelial cells were from Lonza, Basel, Switzerland. Unless specified otherwise, the rest of the molecular biology and biochemistry reagents were from Sigma, St Louis, MO, USA.

HUVECs were isolated from freshly collected umbilical cords by collagenase digestion. They were cultured in fibronectin-coated T25 flasks until a monolayer was formed. MCDB131 medium with endothelial growth supplements was used for culturing HUVECs. All the experiments were performed in passage one. The human aortic vascular smooth muscle cell line (T/G HA-VSMC; ATCC CRL-1999) was purchased from American Type Culture Collection (ATCC). HeLa (human cervical adenocarcinoma epithelial cell line), HEK293 (human embryonic kidney 293) and Huh7 (a hepatocyte cell line) cell lines were procured from the National Centre for Cell Science (NCCS), Pune, Maharashtra, India. All the cell lines were confirmed for no contamination before starting experiments. For cell lines HEK293, HeLa, Huh7 and HA-VSMC, cells were cultured in DMEM with 10% FBS. For hypoxia exposure, cells were incubated in a hypoxia incubator (5% CO₂, 1% O₂ and 95% relative humidity) for the specified duration. After degassing, the cell culture media used for hypoxia experiments were pre-equilibrated to the hypoxic environment for 2 h prior to the start of the experiment.

All the buffers used for washing and lysing of cells were degassed before use. Cells were washed with 1× cold phosphate-buffered saline (PBS) and lysed in buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% TritonX-100, 0.1% SDS, 1% sodium deoxycholate, 1 mM PMSF and protease inhibitor cocktail from Sigma (St Louis, MO, USA), followed by sonication for 1 min (30% amplitude, 3 s on/off cycle; QSONICA Q700 sonicator). 50 µg of total protein was resolved on an 8% SDS–PAGE gel and transferred onto PVDF membrane. For western blotting involving detection of LC3 degradation, a 15% resolving gel was used. Blotted proteins were incubated with corresponding primary antibodies at 4°C overnight followed by incubation with horseradish peroxidase-conjugated secondary antibodies (Jackson ImmunoResearch, West Grove, PA, USA). Amersham ECL western blotting detection reagent (GE Healthcare Life Sciences) was used for detection. ImageJ version 1.45 (NIH, Bethesda, MD, USA) was used for densitometry analysis.

Cells were lysed in ice-cold buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton-X 100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM sodium orthovanadate, protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail 2 (Sigma) followed by 30 s of sonication. 300 µg of total protein was pre-cleared using protein A/G Sepharose beads for 1 h at 4°C. This was followed by overnight incubation with 3 µg PTP-PEST AG10 monoclonal antibody at 4°C. Protein–antibody complexes were pulled down using protein A/G Sepharose beads in 2× SDS sample buffer (10 mM Tris-HCl pH 6.8, 4% SDS, 0.2% Bromophenol Blue, 20% glycerol and 200 mM DTT). Specific pulldown of PTP-PEST was confirmed by employing a corresponding IgG isotype antibody as negative control. For the immunophosphatase assay, PTP-PEST elution was performed in Na-acetate phosphatase assay buffer (50 mM Na-acetate, 2 mM EDTA and 1 mM DTT, pH 5.0). The phosphatase assay was carried out in Na-acetate assay buffer at 30°C for 2 h, and 50 mM pNPP (para-nitrophenyl phosphate) was used as substrate. The reaction was stopped by addition of NaOH (final concentration of 0.45 M). The catalytic activity was determined spectrophotometrically by measuring the amount of pNP (para-nitrophenol) generated from pNPP. pNP released was quantified by measuring absorbance at 420 nm. For experiments involving pulldown of PTP-PEST from 500 µg of HUVEC lysate using a GST tag, 5 µg anti-GST antibody was used. Immunoprecipitation was carried out overnight at 4°C, followed by pulldown of protein–antibody complexes with protein A/G Sepharose beads for 2 h at 4°C.

Flag-PTP-PEST pcDNA3.1 plasmid construct was a kind gift from Dr Zhimin Lu, Department of Neuro-oncology, The University of Texas M. D. Anderson Cancer Center, Houston, USA (Zheng et al., 2011). PTP-PEST was subcloned in pENTR4-GST-6P1 (Addgene #17741) lentivirus entry vector backbone at BamH1 and XbaI restriction sites, followed by LR clonase-mediated gateway recombination into the third-generation lentivirus destination vector pLenti CMV Puro Dest (Addgene #17452). shRNA oligos (oligo 1, 5′-GATCCCACCAGAAGAATCCCAGAATCTCGAGATTCTGGGATTCTTCTGGTGGTTTTTG-3′; and oligo 2, 5′-AGCTCAAAAACCACCAGAAGAATCCCAGAATCTCGAGATTCTGGGATTCTTCTGGTGG-3′) were annealed in annealing buffer (10 mM Tris pH 8.0, 50 mM NaCl and 1 mM EDTA) and cloned in pENTR/pSUPER⁺ entry vector (Addgene #17338), which was recombined into pLenti X1 Puro DEST (Addgene #17297). These lentivirus entry and destination plasmids were a gift from Eric Campeau and Paul Kaufman, Program in Gene Function and Expression, University of Massachusetts Medical School, Worcester, USA (Campeau et al., 2009). For lentivirus production, recombined expression vectors were co-transfected with packaging plasmids pLP1, pLP2 and pVSVG (Invitrogen, Carlsbad, CA, USA) in 293T cells. Supernatant was collected 48 h and 72 h post-transfection and was concentrated through ultracentrifugation. Lentiviral transduction was performed in HUVECs (P₀) at a multiplicity of infection (MOI) of 15 in low serum (5%) endothelial growth medium at 70% confluency in the presence of polybrene (6 µg/ml) for 8 h. This was followed by a second round of transduction. 48 h post-transduction, puromycin (2 µg/ml) selection was performed for the next 3 days followed by splitting of cells for experimental treatment.

Following appropriate experimental treatments, HUVECs were washed with 1× PBS, followed by paraformaldehyde (4%) fixation for 20 min at room temperature. The cells were then made permeable with 0.25% Triton X-100 (10 min at room temperature). Cells were blocked with 5% serum in 1× PBS for an hour at room temperature, followed by overnight staining with primary antibodies at 4°C. Cells were then washed and incubated with fluorophore-conjugated secondary antibodies (Thermo Fisher Scientific, Waltham, MA, USA). Nuclei were counterstained with DAPI (1 µg/ml) for 5 min at room temperature, and images were captured using an LSM 700 Zeiss confocal microscope.

HUVECs were transduced with N-terminal GST-tagged PTP-PEST lentiviral particles as described above. Post-transduction, the same pool of cells was split and upon reaching confluency were exposed to either normoxia or hypoxia (1% oxygen) for 24 h. GST-tagged PTP-PEST was pulled down via immunoprecipitation using GST antibody (Cell Signaling Technology, USA). Immunoprecipitated proteins were loaded on to an SDS-PAGE gel. SDS-PAGE was done in such a way that the electrophoresis run was stopped the moment the entire sample entered the resolving gel (12%). Each band, ∼5 mm in size, visible after Coommassie Blue staining was excised and used for proteomics analysis. The proteomic profiling was performed by LC/MS/MS at the Mass Spectrometry and Proteomics Core facility of RGCB, Thiruvananthpuram, Kerala, India. Briefly, the excised gel pieces were subjected to in-gel trypsin digestion using sequence grade trypsin (Sigma), as per Shevchenko et al. (2006). The LC/MS/MS analyses of the extracted tryptic peptides were performed in a SYNAPT G2 High Definition mass spectrometer (Waters, Manchester, UK), which was connected to a nanoACQUITY UPLC chromatographic system (Waters) for the separation of peptides. The LC/MS/MS-acquired raw data was analysed by Progenesis QI for Proteomics V3.0 (NonLinear Dynamics, Waters) for protein identification using the Human protein database downloaded from UniProt.

HUVECs were cultured as a tight monolayer, and after 8 h of serum starvation a scratch was made with the aid of a 10 µl sterile micro-pipette tip, to create a wound. Immediately after wound creation, scraped cells were washed away and images were taken at the start as the ‘0 hour’ time point. HUVECs were then incubated in the presence of 5 mM hydroxyurea (anti-proliferative molecule) for the next 24 h, either in normoxia or hypoxia (1% oxygen) conditions, and images were retaken at the same locations following incubation to assess migration. The area of wound closure in 24 h was calculated using ImageJ software.

Wells in a 96-well plate were coated with 80 µl of growth factor free Matrigel and allowed to polymerize in a CO₂ cell culture incubator. Following appropriate experimental treatment (PTP-PEST knockdown), 15×10³ HUVECs (with or without PTP-PEST knockdown) were seeded onto each Matrigel-coated well in low serum (1% FBS) endothelial growth medium (MCDB131) and incubated for 16 h under normoxia or hypoxia (1% oxygen) followed by imaging. Tube networks were analysed using the Angioanalyser feature of ImageJ.

Sequence encoding the N-terminal 300 amino acids of wild-type PTP-PEST was PCR amplified (forward primer, 5′-CGCGGATCCATGGAGCAAGTGGAGATC-3′; reverse primer, 5′-CCGCTCGAGTCATAGTTGTAGCTGTTTTTC-3′) and cloned in Novagen-pET28a(+) bacterial expression vector (Merck Millipore Sigma, Burlington, MA, USA) between BamH1 and Xho1 restriction sites. Site directed mutagenesis was carried out to generate the C231S mutant using the following primers: forward primer, 5′-TGTATTCATTCCAGTGCAGGCTG-3′; and reverse primer, 5′-CAGCCTGCACTGGAATGAATACA-3′. PCR was performed using 10 ng of His tagged PTP-PEST (WT) plasmid as template, and this was followed by Dpn1 (New England Biolabs, UK) digestion for 2 h at 37°C. The Dpn1-digested PCR product was transformed into E. coli DH5α ultra-competent cells. Positive clones were selected by kanamycin resistance and grown overnight in Luria Bertani (LB) broth with 0.1 mg/ml kanamycin. These constructs encode the first 300 amino acids along with an additional hexa-histidine tag at the N-terminal end. The sequences of the constructs (WT and C231S) were confirmed using DNA sequencing.

E. coli strain BL21 codon plus RIL (DE3) harboring the relevant plasmid was grown overnight at 37°C in LB broth medium supplemented with kanamycin (0.1 mg/ml) and chloramphenicol (0.034 mg/ml). 1% of this overnight culture was transferred into fresh medium with kanamycin and chloramphenicol and grown until a cell density equivalent to an OD₆₀₀ of 0.6 was reached. Protein expression was induced with a final concentration of 1 mM IPTG for 6 h at 30°C. Cells were then harvested by centrifugation at 4500 g for 10 min at 4°C, washed with milli-Q water and frozen at −20°C. Overexpressed proteins were purified using immobilized metal affinity chromatography (IMAC) based on the affinity of the hexa-histidine tag for Ni²⁺ following cell lysis in lysis buffer (50 mM Tris-HCl, 200 mM NaCl, 1 mM PMSF, pH 8.0). All subsequent steps were performed at 4°C. Cell lysis was achieved through sonication performed thrice for 5 min with a pulse of 5 s ‘on’ and 5 s ‘off’, at 30% amplitude. The lysate was centrifuged at 4500 g for 30 min at 4°C. The supernatant was then applied to a Ni²⁺-NTA (nickel-nitrilotriacetic acid) Sepharose column (GE Healthcare, USA) pre-equilibrated with the lysis buffer at a flow rate of 0.2 ml/min. The protein was eluted in 50 mM Tris-HCl, 200 mM NaCl and 300 mM imidazole (pH 8.0). The fractions containing the eluted protein were pooled and immediately subjected to buffer exchange using a HiPrep 26/10 desalting column on an AKTA FPLC purification system (GE Healthcare, USA) into storage buffer (50 mM Tris-HCl, 200 mM NaCl, pH 8.0). The protein was concentrated using Amicon Ultra-15 centrifugal filter units (Millipore, Germany) to around 8 mg/ml. Size-exclusion chromatography was performed at 4°C using a HiLoad Superdex 200 pg 16/600 GE preparative column (GE Healthcare, USA). The column was equilibrated with 50 mM Tris-HCl, 200 mM NaCl, pH 8.0. Total protein (250 μg) was loaded into the column and eluted at a rate of 0.5 ml/min. The chromatograms were calibrated with the absorption of the following proteins at 280 nm: ferritin (440 kDa), aldolase (158 kDa), conalbumin (75 kDa), ovalbumin (44 kDa) and ribonuclease A (13.7 kDa). Overexpression and purification of PTP-PEST was confirmed through SDS-PAGE and size exclusion chromatography, respectively (Fig. S2A–C). Size exclusion chromatography confirmed that the purified PTP-PEST eluted as a monomeric protein. Protein concentration estimations were performed using Bradford's assay with bovine serum albumin serving as a standard.

All experimental data are represented as mean±s.e.m. for a minimum of three independent experiments. Statistical evaluation was performed using Student's t-test (two tailed), using GraphPad Prism version 6.0 software for Windows (GraphPad Prism Software Inc. San Diego, CA, USA). A P-value<0.05 was considered to be statistically significant.

Authors thank Dr S. Umadevi of the Translational Research Platform for Veterinary Biologicals, Tamil Nadu Veterinary and Animal Sciences University, Chennai for her technical assistance with confocal microscopy.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.250274.reviewer-comments.pdf