Phosphorylation of seryl-tRNA synthetase by ATM/ATR is essential for hypoxia-induced angiogenesis

Hypoxia-induced angiogenesis maintains tissue oxygen supply and protects against ischemia but also enhances tumor progression and malignancy. This is mediated through activation of transcription factors like hypoxia-inducible factor 1 (HIF-1) and c-Myc, yet the impact of hypoxia on negative regulators of angiogenesis is unknown. During vascular development, seryl-tRNA synthetase (SerRS) regulates angiogenesis through a novel mechanism by counteracting c-Myc and transcriptionally repressing vascular endothelial growth factor A (VEGFA) expression. Here, we reveal that the transcriptional repressor role of SerRS is inactivated under hypoxia through phosphorylation by ataxia telangiectasia mutated (ATM) and ataxia telangiectasia mutated and RAD3-related (ATR) at Ser101 and Ser241 to attenuate its DNA binding capacity. In zebrafish, SerRSS101D/S241D, a phosphorylation-mimicry mutant, cannot suppress VEGFA expression to support normal vascular development. Moreover, expression of SerRSS101A/S241A, a phosphorylation-deficient and constitutively active mutant, prevents hypoxia-induced binding of c-Myc and HIF-1 to the VEGFA promoter, and activation of VEGFA expression. Consistently, SerRSS101A/S241A strongly inhibits normal and tumor-derived angiogenesis in mice. Therefore, we reveal a key step regulating hypoxic angiogenesis and highlight the importance of nuclear SerRS in post-developmental angiogenesis regulation in addition to vascular development. The role of nuclear SerRS in inhibiting both c-Myc and HIF-1 may provide therapeutic opportunities to correct dysregulation of angiogenesis in pathological settings.


Introduction
Oxygen is critical for the development and growth of most multicellular organisms. Sophisticated molecular mechanisms have been evolved to sense and respond to changes in oxygen levels in order to maintain cell and tissue homeostasis. Therefore, the hypoxia response is protective against ischemic tissue injury. On the other hand, pathophysiological hypoxia signaling also contributes to the development of diseases, such as cardiovascular disorders and solid tumors [1][2][3][4].
One of the many aspects of the hypoxia response is the stimulation of angiogenesis through the production of angiogenic factors, especially vascular endothelial growth factor A (VEGFA) [5][6][7]. The response is mediated mainly through a transcription factor called hypoxia-inducible factor (HIF)-1 [8]. HIF-1 is a heterodimer consisting of an oxygen-regulated HIF-1α subunit and a constitutively expressed HIF-1β subunit, also known as aryl hydrocarbon receptor nuclear translocator (ARNT). In the presence of oxygen, the HIF-1α subunit is rapidly hydroxylated by prolyl hydroxylases (PHDs) and subsequently ubiquitylated through von Hippel-Lindau protein (VHL) and degraded by the 26S proteasome [9,10]. However, in the absence of oxygen, HIF-1α subunit is stabilized; the HIF-1 heterodimer is formed and, in turn, translocated to the nucleus to bind to the hypoxia-response element (HRE) in the promoter region of VEGFA to activate gene expression [6,8,[10][11][12].
In addition to HIF-1, other transcription factors such as c-Myc have been reported to mediate hypoxia responses and stimulate VEGFA expression [12][13][14][15][16]. In fact, hypoxia-induced angiogenesis is preserved in HIF-1-deficient colon cancer xenograft [14,15,17]. In a recent clinical trial of EZN-2968, the antisense oligonucleotide inhibitor of HIF-1α successfully down-regulated both mRNA and protein levels of HIF-1α but failed to reduce the expression of VEGFA in tumor biopsies [18]. These preclinical and clinical observations highlight the significant role of HIF-1-independent regulation of VEGFA and angiogenesis in tumor progression.
In addition to the contribution of transcription factors that positively regulate angiogenesis, it is conceivable that a complete hypoxia response would require the collaboration between positive and negative transcriptional regulators. However, in contrast to the extensive understanding of how transcriptional activators respond to hypoxia, little is known on the response of negative regulators during hypoxia. We have recently identified seryl-tRNA synthetase (SerRS) as a key transcriptional repressor of VEGFA [19,20]. SerRS is well known as a member of the aminoacyl-tRNA synthetase family responsible for charging serine onto its cognate tRNA to generate substrates for protein biosynthesis in the cytosol. Surprisingly, forward genetic studies in zebrafish suggested a critical role of SerRS in vascular development independent of its aminoacylation activity [21][22][23]. Interestingly, this role of SerRS occurs in the nucleus, and a nucleus-targeting domain is found in all vertebrate SerRS [20]. Nuclear SerRS regulates the vasculature by transcriptionally repressing VEGFA expression and does so through binding to a 27-bp DNA on the VEGFA promoter and recruiting histone deacetylase SIRT2 for epigenetic gene silencing [19]. This role of SerRS directly counteracts that of c-Myc, the major transcriptional factor regulating VEGFA during development [19,24,25]. Because c-Myc is involved in hypoxia-stimulated VEGFA production [16], we postulated that the anti-c-Myc activity of SerRS would need to be attenuated in order to achieve VEGFA stimulation under hypoxia.
Through a series of studies in human cells, zebrafish, and mice, we found that this is indeed the case. The transcriptional repressor role of SerRS is inactivated during hypoxia through phosphorylation of SerRS by ataxia telangiectasia mutated (ATM) and the related kinase ATR to attenuate its DNA-binding capacity. Importantly, both c-Myc and HIF-1 cannot bind to the VEGFA promoter without SerRS being phosphorylated and removed from the DNA. Therefore, our findings reveal a missing piece in the regulation of angiogenesis during hypoxia and demonstrate the importance of inactivating negative regulators in this process. Such understanding may provide new opportunities in treating ischemic cardiovascular diseases and cancer.

SerRS inhibition of VEGFA transcription is blocked under hypoxia
To investigate whether SerRS is involved in the hypoxia response to regulate VEGFA expression, we knocked down SerRS in HEK293 cells with a short hairpin RNA (shRNA) targeting the 3 0 untranslated region (3 0 UTR) of the SerRS gene. Under normal levels of oxygen (normoxia), as we observed previously [19], VEGFA expression is up-regulated upon knocking down SerRS, compared with control cells transfected with a nonspecific control shRNA (shcontrol) or an shRNA targeting a different aminoacyl-tRNA synthetases (i.e., glycyl-tRNA synthetase (GlyRS); sh-GlyRS) (Fig 1A and 1B, S1A Fig). Under hypoxia, VEGFA expression is significantly increased in control cells as expected (Fig 1A and 1B). However, the stimulation is largely lost in SerRS knockdown cells (Fig 1A and 1B), suggesting under hypoxia VEGFA level is no longer dependent on SerRS. We hypothesize that this observation indicates a hypoxia-mediated inactivation of the angiostatic function of SerRS.

SerRS is phosphorylated at serine 101 and serine 241 by ATM/ATR during hypoxia
To test the hypothesis, we first examined the effect of hypoxia on the expression of SerRS and did not observe an obvious change (S1B Fig). We then investigated the possibility that SerRS is inactivated under hypoxia through posttranslational modifications. In a large-scale mass spectrometry study, SerRS was found to be phosphorylated at serine 241 (S241) by ATM/ATR kinases, which are activated under DNA damage [26]. In PhosphoSitePlus database [27], we found another possible SerRS phosphorylation site serine 101 (S101). Both sites have the conserved ATM/ATR substrate motif with a serine or threonine followed by a glutamine and preceded by 2 hydrophobic residues (at −1 and −3 positions relative to the serine/threonine) ( Fig  1C) [28]. Multiple sequence alignment revealed strict conservation of S/T101 and S/T241 and flanking ATM/ATR substrate motif residues in vertebrate SerRS (Fig 1C), concurrent with the emergence of the role of SerRS in regulating vascular development and angiogenesis [20].
DNA fragment-induced SerRS phosphorylation was confirmed in vitro by 32 P-labeling ( Fig  1D). Double-stranded DNA oligonucleotides were added to the nuclear extract of HEK293 cells to mimic DNA damage in order to activate ATM/ATR [29]. The "activated" nuclear extract specifically induced robust phosphorylation of the purified recombinant SerRS but not of GlyRS ( Fig 1D). SerRS phosphorylation was further verified by using specific phosphor-ATM/ATR substrate (p-S � Q) antibody ( Fig 1E). To confirm the phosphorylation sites on SerRS, we substituted S101 and S241 with alanine separately (S101A, S241A) and simultaneously (SerRS S101A/S241A or in abbreviation SerRS AA ). SerRS S101A showed a decreased level of phosphorylation, whereas SerRS S241A and SerRS AA had little phosphorylation in vitro as revealed by both p-SQ antibody detection ( Fig 1E) and 32 P-labeling (S1C Fig), suggesting that SerRS can be phosphorylated by ATM/ATR kinases at both S101 and S241 and that S241 is the major phosphorylation site on SerRS.
Next, we investigated whether SerRS is phosphorylated in hypoxic conditions, under which ATM/ATR is known to be activated [30][31][32]. Indeed, phosphorylation of endogenous SerRS was detected in HEK293 and human umbilical vein endothelial (HUVEC) cells under hypoxia within 12 h (Fig 1F and 1G). Interestingly, there is an apparent temporal delay between the activation of ATM/ATR (approximately 3 h after hypoxia exposure) and the phosphorylation of SerRS in HEK293 cells (at least 6 h after hypoxia exposure) ( Fig 1F). However, this delay was not observed in HUVEC cells, and phosphorylation of SerRS was detected 3 h after hypoxia exposure (Fig 1G). Possibly, an unknown inhibitor exists in HEK293 cells, but not in SerRS is involved in the hypoxia response to induce VEGFA expression through phosphorylation by ATM and ATR kinases at S101 and S241 residues. (A, B) qRT-PCR analysis of VEGFA expression in HEK293 cells transfected with sh-SerRS, sh-GlyRS, or nonspecific sh-Control under hypoxia or normoxia conditions. VEGFA levels (A) and relative induction of VEGFA under hypoxia (B) were plotted as means ± SEM (n = 4, biological replicates, Student t test, �� p < 0.01, ���� p < 0.0001). (C) Sequence alignment of SerRS proteins from vertebrates and non-vertebrates flanking serine 101 and serine 241 (highlighted in red) sites numbered according to the human sequence. The conserved ATM/ATR substrate motif residues are underlined. (D) 32 Plabeling to confirm phosphorylation of SerRS, but not GlyRS, in vitro using recombinant tRNA synthetases. Double-stranded DNA oligonucleotides are used to mimic DNA damage to activate ATM/ATR from the nuclear extract of HEK293 cells. (E) Western blot analysis to confirm recombinant SerRS, but not GlyRS and TyrRS, is phosphorylated by ATM/ATR by using a specific antibody against phosphor-ATM/ATR substrate (p-S � Q). The purified His 6 -tagged aaRS proteins are recognized by anti-His 6 -tag antibody. Activation of ATM and ATR was confirmed by autophosphorylation of ATM (at S1981) and phosphorylation of RPA32 (at S33), respectively. (F, G) IP and western blot analysis to confirm hypoxia-induced SerRS phosphorylation in HEK293 (F) and HUVEC (G) cells. IP was performed with anti-SerRS antibody from mouse and rabbit for HEK293 (F) and HUVEC (G) HUVEC cells, that needs to be removed prior to the phosphorylation of SerRS by ATM/ATR. Interestingly, SerRS AA , when exogenously expressed in hypoxic HEK293 cells, was phosphorylated to a lesser extent than the exogenously expressed wild-type SerRS (SerRS WT ) (Fig 1H), confirming that S241 and/or S101 are the phosphorylation sites.
To further confirm that ATM and ATR are responsible for SerRS phosphorylation under hypoxia, we knocked down ATM and ATR either separately or simultaneously by RNA interference. Hypoxia-induced phosphorylation of SerRS was greatly inhibited when either ATM or ATR was knocked down, and completely blocked when both kinases were knocked down simultaneously ( Fig 1I). Consistent with these results, SerRS phosphorylation under hypoxia could also be blocked by the specific ATM inhibitor KU-55933 [33] and ATR inhibitor VE-821 [34] (S1D Fig).
It is worth noting that we used a 2-enzyme approach to induce hypoxia in our cell-based experiments [35]. Glucose oxidase to deplete oxygen, whereas catalase was used to convert H 2 O 2 (produced by the prior oxidation reaction) to water to prevent DNA damage in cells. Because ATM and ATR are well known to be activated by DNA damage, we measured H2AX phosphorylation (γ-H2AX) to ensure lack of DNA damage during the 2-enzyme treatment (S2 Fig). Indeed, no obvious increase in gamma-H2AX level was observed in hypoxic cells, confirming that the activation of ATM/ATR is not an artifact of the 2-enzyme treatment.

Phosphorylation inactivates SerRS regulating VEGFA in cells and in vivo
To understand if phosphorylation of SerRS affects its role as a transcriptional repressor, we created a mutant form of SerRS bearing double substitutions of S101 and S241 with aspartic acid residues (SerRS S101D/S241D or in abbreviation SerRS DD ) to mimic the phosphorylated SerRS. In contrast to SerRS WT and SerRS AA , SerRS DD can no longer repress VEGFA transcription in HEK293 cells (Fig 2A), suggesting that phosphorylation of SerRS inhibits its transcriptional repressor activity.
To investigate the effect of SerRS phosphorylation in vivo, we took advantage of a previously well-established zebrafish system [19,20,22]. In zebrafish embryos, endogenous SerRS was knocked down by an antisense morpholino (SerRS-MO), which led to a 4-fold increase in the mRNA level of vegfaa ( Fig 2B). This effect can be completely and partially rescued by coinjection of human SerRS WT mRNA and SerRS AA mRNA with SerRS-MO, respectively. In fact, SerRS WT and SerRS AA should be functionally equivalent in this experiment because the zebrafish are not under hypoxia; the partial activity of SerRS AA is likely due to the lower expression level of SerRS AA compared with SerRS WT (Fig 2B). However, co-injection of SerRS DD mRNA, which was expressed at a similar level as SerRS AA , had no rescue effect ( Fig  2B), confirming that phosphorylations at S101 and S241 block the transcriptional repressor activity of SerRS on VEGFA in vivo.

Phosphorylation inactivates SerRS by attenuating its DNA binding capacity
To explore the molecular mechanism of how SerRS phosphorylation inactivates its function as a transcriptional repressor, we first examined the effect of hypoxia on SerRS nuclear localization in HEK293 cells. Hypoxia does not affect the nuclear localization of SerRS (S4A Fig). Consistently, we also found similar cytoplasmic versus nuclear distributions of exogenously expressed SerRS WT , SerRS DD , and SerRS AA proteins in HEK293 cells (S4B Fig). Given that SIRT2 is a necessary cofactor for SerRS to epigenetically silence VEGFA expression [19], we next examined the impact of hypoxia on the interaction between SerRS and SIRT2. Exposure to hypoxia did not obviously affect the amount of SIRT2 co-immunoprecipitated with SerRS in HEK293 cells (S4C Fig). Consistently, SIRT2 interacts with SerRS AA and SerRS DD as strongly as with SerRS WT (S4D Fig), indicating a lack of effect of hypoxia on the SerRS-SIRT2 interaction. SerRS-MO and with co-injection of SerRS-MO and SerRS WT , SerRS AA , or SerRS DD mRNA. Data plotted as means ± SEM (n = 3-4, biological replicates, �� p < 0.01, ��� p < 0.001, Student t test). The Coomassie blue staining was used to show equal total protein loading. See S1 Data for quantitative data and statistical analysis. See S2 Data for original, uncropped images supporting blots and gel results. qRT-PCR, real-time quantitative reverse transcription PCR; SerRS, seryl-tRNA synthetase; SerRS AA , S101A/S241A; SerRS DD , S101D/S241D; SerRS-MO, antisense morpholino against SerRS; SerRS WT , wild-type SerRS; VEGFA, vascular endothelial growth factor A. Finally, we explored the interaction of SerRS with the VEGFA promoter. As detected by electrophoresis mobility shift assay (EMSA), the direct binding between recombinant SerRS protein and a 32 P-labeled 27-bp DNA fragment previously identified as the SerRS binding site from the VEGFA promoter [19] was strongly attenuated with the phosphor-mimicking mutant SerRS DD but not the phosphor-deficient mutant SerRS AA (Fig 3A). Importantly, this decreased DNA binding of SerRS DD was not caused by mutation-induced protein misfolding, because both SerRS DD and SerRS AA proteins have largely maintained the enzymatic activity in tRNA aminoacylation (S5 Fig).
Consistently, in HEK293 cells, ectopically expressed SerRS DD mutant also showed a decreased binding to VEGFA promoter as determined by a chromatin immunoprecipitation assay ( Fig 3B). Thus, we concluded that hypoxia-induced phosphorylation blocks the transcriptional repressor activity of SerRS by weakening its DNA binding capacity. Interestingly, the S101 and S241 phosphorylation sites are located near the proposed DNA binding sites of SerRS [19] (Fig 3C). Phosphorylation of these serine residues would introduce negative charges to create electrostatic repulsion between SerRS and DNA. This could explain why phosphorylation of SerRS weakens its DNA binding capacity.

SerRS phosphorylation is required for hypoxia-induced c-Myc and HIF-1 binding to the VEGFA promoter
In previous work, we have established that SerRS and c-Myc compete with each other for binding to the VEGFA promoter [19]. Therefore, as SerRS is phosphorylated and loses its DNAbinding capacity, we would expect an enhanced binding of c-Myc to the VEGFA promoter. Indeed, during hypoxia, the amount of endogenous SerRS bound to the VEGFA promoter in both HUVEC and HEK293 cells gradually decreases, concurrent with a gradual increase in the level of c-Myc bound to the VEGFA promoter (Fig 3D and 3E). Remarkably, the amount of HIF-1α bound to the VEGFA promoter also increases in a trajectory parallel to that of c-Myc ( Fig 3D and 3E), raising the possibility that the removal of SerRS from the VEGFA promoter is also necessary for HIF-1 to bind to the DNA. It is worth noting that the release of SerRS from the VEGFA promoter starts earlier in HUVEC than in HEK293 cells (Fig 3D and 3E), which is consistent with the distinct time course of SerRS phosphorylation in these 2 cell types under hypoxia (Fig 1F and 1G). Nevertheless, the binding of HIF and Myc to the VEGFA promoter happens around the same time, i.e., 6 h after hypoxia exposure, in both HEK293 and HUVEC cells (Fig 3D and 3E).
To confirm that SerRS phosphorylation is essential for hypoxia-induced binding of c-Myc and HIF-1α to the VEGFA promoter, we compared DNA binding of c-Myc and HIF-1α in HEK293 cells that express SerRS WT versus phosphor-deficient SerRS AA and phosphor-mimicry SerRS DD . As expected, while the binding of SerRS WT to the VEGFA promoter decreases under hypoxia, binding of SerRS AA and SerRS DD no longer respond to change in the oxygen level ( Fig 3F). The amount of c-Myc and of HIF-1α bound to the VEGFA promoter significantly increases during hypoxia in control cells expressing SerRS WT (Fig 3F). However, the increase is blocked when SerRS AA is expressed (Fig 3F). These results strongly indicate that the SerRS phosphorylation event is a prerequisite for hypoxia-induced binding of c-Myc and HIF-1 to VEGFA. Consistently, hypoxia-induced VEGFA expression is inhibited in HEK293 cells expressing SerRS AA but not SerRS WT (Fig 3G).
Interestingly, we found no difference in HIF-1α binding to the VEGFA promoter between cells expressing SerRS WT and SerRS DD , whereas SerRS DD expression, compared with SerRS WT , stimulates c-Myc binding to the VEGFA promoter with or without hypoxia ( Fig 3F). This is likely due to the fact that c-Myc is constitutively present in the cell, whereas HIF-1α protein is

ATM/ATR-SerRS is a key pathway that regulates hypoxia-induced VEGFA expression
The fact that expressing the phosphorylation-deficient SerRS AA to bypass the ATM/ATR regulation can significantly inhibit hypoxia-induced VEGFA expression (Fig 3G) suggests that the ATM/ATR-SerRS pathway is critical for cellular induction of VEGFA during hypoxia. To confirm the importance of this pathway, we knocked down ATM or ATR or both in HEK293 cells (S8A Fig). Either ATM or ATR knockdown inhibited VEGFA induction under hypoxia, and it is unclear whether the dual knockdown has a stronger effect (S8B Fig), suggesting that both ATR and ATM are major players in stimulating VEGFA expression during hypoxia. We also blocked ATM and ATR individually in HEK293 cells by specific inhibitors. ATR inhibitor VE-821 dramatically inhibited VEGFA induction under hypoxia, whereas the effect of ATM inhibitor KU-55933 was reduced but still statistically significant ( Fig 3H). Consistently, ATR inhibitor VE-821 could restore the binding of SerRS to VEGFA promoter whereas it had no effect on SerRS AA and SerRS DD (Fig 3I), strongly suggesting that hypoxia activated ATR plays a key role in promoting the release of SerRS from VEGFA promoter by direct phosphorylation of SerRS at S101 and S241.

SerRS phosphorylation is required for activating VEGFA-mediated tube formation under hypoxia
To understand if the effect of SerRS on VEGFA transcription can be manifested at the protein level and because VEGFA mainly exists as an extracellular protein, we detected VEGFA in the culture media of both HEK293 and HUVEC cells by ELISA. As shown in Fig 4A and 4B, expression of SerRS WT and SerRS AA , but not the phosphor-mimicking SerRS DD , reduced the level of VEGFA protein in both cell types cultured in normoxia, although the observed effect is measure the effects of specific ATM inhibitor KU-55933 and specific ATR inhibitor VE-821 on hypoxia-induced VEGFA expression in HEK293 cells (means ± SEM, n = 2, biological replicates, � p < 0.05, ���� p < 0.0001, Student t test). (I) ChIP-qPCR to measure the binding of SerRS to VEGFA promoter in HEK293 cells stably transfected with SerRS WT , SerRS AA , or SerRS DD under normoxia or hypoxia, in the presence or absence of VE-821 (means ± SEM, n = 3, biological replicates, �� p < 0.01, ��� p < 0.001, Student t test). See S1 Data for quantitative data and statistical analysis. See S2 Data for original, uncropped images supporting blots and gel results. ATM, ataxia telangiectasia mutated; ATR, ataxia telangiectasia mutated and RAD3-related; ChIP-qPCR, chromatin immunoprecipitation-quantitative real-time PCR; EMSA, electrophoresis mobility shift assay; HIF-1; hypoxia-inducible factor 1; HUVEC, human umbilical vein endothelial cell; n.s., not significant; qRT-PCR, real-time quantitative reverse transcription PCR; SerRS, seryl-tRNA synthetase; SerRS AA , S101A/S241A; SerRS DD , S101D/S241D; SerRS WT , wild-type SerRS; VEGFA, vascular endothelial growth factor A.
https://doi.org/10.1371/journal.pbio.3000991.g004 PLOS BIOLOGY less obvious than that at the mRNA level (Fig 2A). Based on the ELISA analysis, the repressor activity of SerRS WT and SerRS AA on VEGFA protein level is similar at normal oxygen environment (Fig 4A and 4B). However, under hypoxia condition, the repressor activity of SerRS AA is significantly higher than that of SerRS WT (Fig 4A and 4B), in line with that SerRS WT but not SerRS AA can be phosphorylated and inactivated under hypoxia (Fig 1E and 1H, Fig 3G and 3I).
To further demonstrate the functional consequence of SerRS phosphorylation, we performed an endothelial cell tube formation assay with HUVEC cells cultured in different conditional media, which were obtained from culturing donor HUVEC cells stably transfected with various SerRS constructs (or an empty vector as the control) under normoxia or hypoxia. In the experiment where control donor cells (stably transfected with the empty vector) were involved, the conditional media from the hypoxia treatment group have much stronger activity in promoting HUVEC cell tube formation than the conditional media from the normoxia group, and the stimulation effect can be blocked by a VEGFA-neutralizing antibody, suggesting that the enhanced tube formation is mediated through hypoxia-induced VEGFA production from donor cells (Fig 4C and 4D). As expected, a similar result was found with the culture media from donor cells stably expressing the phosphor-mimicking SerRS DD , because SerRS DD is no longer active in suppressing VEGFA. However, the stimulation effect is compromised with the culture medium of donor cells stably expressing SerRS WT and abolished with that of donor cells stably expressing the phosphorylation-deficient SerRS AA (Fig 4C and 4D). These cell-based experiments demonstrate that phosphorylation and inactivation of SerRS is critical for stimulating angiogenesis under hypoxia.

SerRS AA bypasses hypoxia response and strongly inhibits angiogenesis in mice
Because ectopic expression of SerRS AA can block, to a large extent, the hypoxia-induced VEGFA expression in both HEK293 and HUVEC cells (Fig 3G and Fig 4A and 4B) and endothelial cell tube formation (Fig 4C and 4D), we next explored the activity of SerRS AA and the effect of SerRS phosphorylation in hypoxia-induced angiogenesis in mammals by using a matrigel plug angiogenesis assay. Mouse endothelial 3B11 cells were stably transfected with mouse SerRS WT , SerRS AA , or SerRS DD constructs to reach an expression level similar to that of the endogenous mouse SerRS (Fig 5A). The engineered 3B11 cells were mixed with matrigel in vitro at low temperature. The mixtures were injected subcutaneously into the mice and solidified into plugs, where a hypoxic environment would form prior to the induction of the vasculature. Two weeks after the injection, we confirmed, by the elevated HIF-1α protein level, that a hypoxia environment occurred inside the matrigel plugs (S9A Fig). At the same time, we evaluated the microvasculature in the plugs by VEGFA (Fig 5B and 5C) and CD31 (S9B and S9C Fig) immunostaining. As expected, expression of the phosphor-mimicking SerRS DD did not suppress microvascular formation, whereas the phosphor-deficient SerRS AA strongly suppress it, demonstrating that SerRS phosphorylation and inactivation is essential for hypoxiainduced angiogenesis in vivo. Importantly, the antiangiogenic activity of SerRS WT lies between that of SerRS AA and SerRS DD , indicating that SerRS WT was partially inactivated under hypoxia by phosphorylation (Fig 5B and 5C).

SerRS AA strongly inhibits tumor angiogenesis and tumor growth in mice
Hypoxia is a common phenomenon in solid tumors and leads to advanced and abnormal vascularization [2,3]. To evaluate whether hypoxia-induced SerRS phosphorylation and inactivation is also important for tumor angiogenesis and tumor growth, we stably transfected human triple negative breast cancer MDA-MB-231 cells with constructs expressing human SerRS WT , SerRS AA , or SerRS DD . Considering that angiogenesis-related transcription factors, such as c-Myc and HIF, are likely to be overexpressed in cancer cells [3,36], we selected MDA-MB-231 cells showing high levels of SerRS overexpression (approximately 7-fold) compared to that of the endogenous protein (S10A Fig). The engineered MDA-MB-231 cells were implanted subcutaneously into the mammary glands of T cell-deficient homozygote NU/J athymic nude (Nude) mice. Strong inhibitory effects of SerRS WT and SerRS AA , but not of SerRS DD , on tumor growth were observed during a 3-week period after the implantation (Fig 6A). Remarkably,  Student t test). See S1 Data for quantitative data and statistical analysis. See S2 Data for original, uncropped images supporting blots and gel results. n.s., not significant; SerRS, seryl-tRNA synthetase; SerRS AA , S101A/S241A; SerRS DD , S101D/S241D; SerRS WT , wild-type SerRS; Vegfa, vascular endothelial growth factor A https://doi.org/10.1371/journal.pbio.3000991.g005  (Fig 6A-6C). At the 3-week end point, the vasculature in tumor xenografts was examined by western blot analysis (Fig 6D-6F) and immunostaining (Fig 6G-6J) to detect VEGFA and endothelial cell marker VE-Cadherin (Cdh5). Consistent with tumor growth, strong inhibitory effects of SerRS WT and SerRS AA , but not of SerRS DD , were observed. And the effect of SerRS AA is significantly stronger than that of SerRS WT (Fig 6D-6J). We also detected the level of ATM/ ATR-mediated SerRS phosphorylation in tumor tissues. As expected, the level of phosphorylated SerRS was significantly reduced in tumors expressing either SerRS AA or SerRS DD , compared to SerRS WT (Fig 6K and 6L), confirming that SerRS phosphorylation at S101 and/or S241 indeed occurs in hypoxic tumor microenvironment (S10B Fig) to inactivate the antiangiogenic function of SerRS to allow sustained tumor growth.
Consistent with the result from the matrigel plug angiogenesis assay (Fig 5), SerRS WT exhibited partial, but nevertheless strong, activity in inhibiting tumor angiogenesis and tumor growth (Fig 6). These in vivo observations contrast to the more complete inactivation of SerRS WT by hypoxia in cell-based systems (Fig 3G and Fig 4A and 4B). Possibly, the hypoxic environment inside a matrigel plug or a tumor is more heterogenous than in cultured cells, rendering more SerRS WT to remain unphosphorylated and able to inhibit VEGFA expression.

Discussion
Until now, our understanding of angiogenesis regulation during hypoxia has focused on the activation of positive regulators. By studying SerRS, a novel inhibitor of angiogenesis, we revealed the importance of silencing negative regulators in this process. SerRS is inactivated through posttranslational modification by phosphoinositide-3 kinase (PI3K)-like kinase family members ATM and ATR, which are central kinases in DNA damage responses, important for maintaining genome stability under hypoxic stress in mammals [32,37]. ATR is activated with the accumulation of single-stranded DNA generated from stalled DNA replication forks under hypoxia [32]. On the other hand, ATM is activated during reoxygenation, when reactive oxygen species (ROS) are generated and lead to DNA double-stranded breaks [32,37] (Fig 7). Activated ATM/ATR kinases phosphorylate many effector substrates that regulate cell cycle, DNA damage repair, and apoptosis [26,38]. In addition to their role in maintaining genome stability, ATM and ATR, especially ATR, was reported to regulate pathological hypoxia-driven angiogenesis through an unknown mechanism [39]. This report is consistent with our observation that ATM and ATR inhibitors can block hypoxia-induced VEGFA expression (Fig 3H). Moreover, our study identifies SerRS as an ATM/ATR substrate and a novel effector to mediate hypoxia-driven angiogenesis. The facts that SerRS is a negative regulator of angiogenesis and that the activity of SerRS is inactivated by ATM/ATR reveal a general principle that hypoxia-driven angiogenesis requires not only the activation of positive regulators, such as HIF and c-Myc, but also the silencing of negative regulators, such as SerRS (Fig 7). 231 xenografts was analyzed by immunofluorescent staining of VEGFA (G) and Cdh5 (I) and quantified by measuring the fluorescent signal intensities (H, J). (K, L) IP and western blot analysis to confirm SerRS phosphorylation in tumor xenografts. IP was performed with anti-SerRS antibody from rabbit before phosphorylated SerRS was detected with a specific antibody against p-S � Q (K) and quantified by measuring the band intensities (L). Data are shown as means ± SEM, n = 5 mice, � p < 0.05, �� p < 0.01, ��� p < 0.001, n.s., Student t test). See S1 Data for quantitative data and statistical analysis. See S2 Data for original, uncropped images supporting blots and gel results. HIF-1α; hypoxia-inducible factor 1α; IP, immunoprecipitation; n.s., not significant; SerRS, seryl-tRNA synthetase; SerRS AA , S101A/S241A; SerRS DD , S101D/S241D; SerRS WT , wild-type SerRS; VEGFA, vascular endothelial growth factor A. https://doi.org/10.1371/journal.pbio.3000991.g006 Tumor cells can adapt to the hypoxic environment through orchestrating the expression of a large number of genes that functions in apoptosis, cell survival, energy metabolism, and angiogenesis [3]. Although HIF is considered as the most important transcription factor mediating the hypoxia response and has been targeted in many preclinical and clinical anticancer programs, success has been limited. In fact, HIF is not always associated with increased cancer mortality, and overexpression of HIF does not always occur in solid tumors [40]. In some cases, inhibition of HIF only delays tumors growth, but tumors growth eventually resumes in an HIF-independent manner [1].
For example, Kirsten rat sarcoma viral oncogene homolog (K-RAS) and hypoxia tension can synergistically increase the level of guanosine 5 0 -triphosphate (GTP)-bound Rho via PI3K and lead to the phosphorylation of c-Myc, which promotes VEGFA expression in the absence of HIF-1 [16] (Fig 7). Another HIF-independent angiogenesis pathway has been observed in HIF-1-deficient colon cancer cells, where increased production of ROS under hypoxia activates nuclear factor kappa B (NF-κB), which strongly induces angiogenic factors such as interleukin 8 (IL-8) [15,17] (Fig 7).
Our previous studies have revealed that SerRS strongly competes with c-Myc for binding to the VEGFA promoter and represses VEGFA expression [19]. Consistently, here we show that phosphorylation-mediated loss of DNA binding of SerRS under hypoxia is required for c-Myc to bind to the VEGFA promoter (Fig 3D and 3E). Remarkably, SerRS can also compete with HIF-1 for binding to the DNA (Fig 3D and 3E), indicating that SerRS inhibits both HIFdependent and HIF-independent activation of VEGFA. Interestingly, competitive binding between SerRS and an NF-κB family protein on VEGFA was also reported [41]. Moreover, we have observed a stimulation of IL-8 expression (among 215 angiogenesis related genes) in human endothelial cells when SerRS is knocked down (S11 Fig). Although it is unclear whether IL-8 is a direct target of SerRS, the apparent effect suggests that, in addition to inhibiting VEGFA expression through blocking c-Myc and HIF-1, SerRS may also inhibit IL-8 expression and potentially antagonize the ROS/NF-κB/IL-8 hypoxia response pathway (Fig 7). Therefore, we propose that inactivation of SerRS through ATM/ATR is required for the activation of several HIF-dependent and HIF-independent pathways for hypoxia-driven angiogenesis. Hence, blocking the effect of ATM/ATR on SerRS, for example, through the expression of a phosphorylation-deficient form of SerRS, can act as a master inhibitor of multiple hypoxia response pathways to effectively inhibit angiogenesis. Our results with the endothelial cell tube formation assay (Fig 4C and 4D) and in mice showing potent antiangiogenesis and antitumor effect with SerRS AA expression (Figs 5 and 6) are supportive of this concept.
In conclusion, our study identified a key regulation step in driving angiogenesis under hypoxia, which involves the inactivation of a seemingly broad functioning negative regulator of angiogenesis. Notably, this regulator is a nuclear localized tRNA synthetase, highlighting the importance of the non-translation function of a translation factor in angiogenesis regulation. The coding sequences for wild-type (WT) SerRS and its mutants (AA and DD) were subcloned into lentiviral vector pLV-EF1α-MCS-IRES-Bsd (Biosettia, San Diego, California, USA), and the lentivirus were prepared to infect HUVEC cells. The stably transfected HUVEC cells were selected by 2.5 μg/mL of blasticidin. Hypoxia was induced using an enzymatic hypoxia system [35]. Hypoxia medium was prepared by rapidly diluting glucose oxidase and catalase at a constant 1:10 ratio in cell culture medium (Sigma-Aldrich, #C3155 and #G0543), the concentration of the 2 enzymes are as follows: glucose oxidase (2 U/ml) and catalase (120 U/ml). For incubation periods longer than 24 h, the medium was replaced by preequilibrated hypoxic medium to maintain nutrients and substrates such as glucose.

Plasmid constructs
Human and mouse full-length SerRS genes were cloned into the pFlag-CMV2 vector (Sigma-Aldrich) and pBabe-puro vector (Addgene) and human SIRT2 genes into the pcDNA6c-V5/ His vector (Thermo Fisher Scientific). For mutations in SerRS, we performed site-directed mutagenesis PCR to obtain the SerRS AA  For protein purification, human SerRS and its mutant genes were subcloned into pET-20b (+) plasmid (Novagen, Darmstadt, Germany) and overexpressed in E. coli. The recombinant carboxyl-terminal His 6 -tagged proteins were purified using Ni-NTA beads (Qiagen, Valencia, California, USA). The purities of the recombinant proteins were assessed by Coomassie blue staining following 4% to 12% Mini Gel (ThermoFisher Scientific) electrophoresis. Protein concentrations were determined using Bradford protein assay (BioRad, Hercules, California, USA).

Real-time PCR assay
Total RNA was isolated from cells by TRIzol Reagent (ThermoFisher Scientific). One microgram (μg) of the total RNA from each sample was reversely transcribed to cDNA by M-MLV reverse transcriptase (Promega, Madison, Wisconsin, USA). All real-time PCR reactions were performed using the StepOnePlus Real-Time PCR system (ThermoFisher Scientific) with SYBR Select Master Mix (Applied Biosystems). The primer pairs for the PCR reactions were: The PCR reaction program started at 95˚C for 10 min, followed by 45 cycles of 95˚C for 20 s and 60˚C for 1 min. Each experiment was carried out in triplicate. The VEGFA gene expression was normalized to that of β-ACTIN. Statistical analyses were performed with the software Sig-maPlot (version 10.0). Student t test was used to analyze the changes between different groups.

P-labeling assay using nuclear extract
The in vitro phosphorylation of SerRS by ATM/ATR was performed based on the methods described by Schiotani and colleagues [29]. Briefly, to prepare the nuclear extract, 5 × 10 7 HEK293 cells were treated with swelling buffer [10 mM HEPES-NaOH (pH 7.9), 10 mM KCl, 1.5 mM MgCl 2 , 0.5 mM DTT, 0.5 mM PMSF, and protease inhibitor cocktail (added before use)] and homogenized to remove the cytoplasmic fraction. The nuclear pellets were lysed in lysis buffer [20 mM HEPES-NaOH (pH 7.9), 600 mM KCl, 1.

In vivo studies in zebrafish
Transgenic Tg (fli1a: EGFP) fish were maintained as described before [20]. The fish embryos were kept at 28.5˚C before and after microinjection.

Matrigel plug angiogenesis assay
A total of 10 6 of stably transfected 3B11 cells were resuspended in 100 μL of DMEM medium and then mixed with 200 μL of ice-cold matrigel (BD Biosciences, San Jose, California, USA) liquid on ice. The 300 μL of cell and matrigel mixture was injected subcutaneously into the flank of C3H/HeJ mice (2 injection sites per mouse and 5 to 6 mice in each group) (Jackson Laboratory). Fourteen days after inoculation, the matrigel plugs were excised and frozen in Tissue-Tek OCT compound for cryostat sectioning. All the mouse experiments were conducted according to the guidelines established by the IACUC (protocol number 13-0003) at The Scripps Research Institute.

Xenograft tumor model
A total of 10 6 of MDA-MB-231 cells stably overexpressing WT human SerRS, SerRS AA , or SerRS DD were subcutaneously injected into the mammal glands of 6-to 8-week-old female NU/J mice (5 mice in each group) (Jackson Laboratory, stock #002019). Twenty-two days after injection, the tumor xenografts were separated from mice and frozen in Tissue-Tek OCT compound for cryostat sectioning.

Immunohistochemistry and immunofluorescence
Sections of 5 μm in thickness from freshly frozen tumors xenograft and matrigel plugs were treated with acetone and 3% H 2 O 2 to block endogenous peroxidase. After 3 to 5 times wash and goat serum block, the sections were incubated with an anti-CD31 antibody (diluted at 1:3000; Cell Signaling Technology, #3528) overnight at 4˚C. The blood vessels were counted in 5 to 10 random viable fields (magnification, 400×) in tumor xenograft samples, and microvessel density in matrigel plugs was quantified by measuring the CD31 immunostaining density using Image J software. To detect the hypoxic regions, we incubated the slides with anti-HIF-1α antibody (1:100; Novus Biologicals, #NB100-105). The immunofluorescence staining was performed with anti-VEGFA antibody (diluted at 1:500; Abcam, #ab46154) and anti VE-cadherin (Cdh5) antibody (diluted at 1:500, Abcam, #ab33168).

Cell fractionation
Cytoplasmic and nuclear fractions were separated and extracted using NE-PER Nuclear and Cytoplasmic Extraction Kit (ThermoFisher Scientific). Exogenously or endogenously expressed SerRS proteins were detected by western blot analysis using anti-flag polyclonal antibody (Sigma-Aldrich) or polyclonal anti-SerRS antibody.

ELISA
After the stably transfected HUVEC or HEK293 cells were cultured under normoxia or hypoxia for 12 h, the level of secreted VEGFA proteins in cell culture medium were detected by Human VEGFA ELISA Kit (Wuxi Donglin Sci&Tech Development, Jiangsu, China).
Aliquots of 5 μL were applied into 100 μL quench solution (300 mM NaOAc (pH 3.0), 1 mg/ mL DNA, and 100 mM EDTA) in the Multi Screen 96-well filter plate (0.45 μm pore size hydrophobic, low-protein-binding membrane; Merck Millipore) at different time points of the reaction. After that, 100 μL of 20% (w/v) trichloroacetic acid (TCA) was added to precipitate the nucleic acids in each well. Plate was then washed 4 times with 200 μL/well of 5% TCA containing 100 mM cold L-serine, followed by 1-time wash with 95% ethanol. Air-dried plate was then treated with 70 μL/well of 100 mM NaOH to elute tRNAs. The eluted tRNAs were centrifuged into a 96-well flexible PET microplate (PerkinElmer, Santa Clara, California, USA) with 150 μL/well of Supermix scintillation mixture. After mixing, the radioactivity in each well of the plate was measured in a PerkinElmer 1450 Liquid Scintillation Counter and Luminescence Counter.

Tube formation assay
HUVEC cells (5 × 10 5 ) were suspended in fresh EGM medium mixed with same volume of conditional medium from empty vector, or SerRS WT , or SerRS AA , or SerRS DD stably transfected HUVEC cells, which were cultured in normoxia or hypoxia conditions for 12 h, and seeded on matrigel basement membrane matrix (BD Biosciences) in 48-well plates. Thirty-six hours later, the images of the endothelia cell tubular network were taken with a Q-IMAGING MicroPublisher 5.0 RTV CCD camera attached to an inverted Olympus IX71 microscope. The length of the tubes was measured by Image J software. To test the contribution of VEGFA, a VEGFA neutralizing antibody (Sino Biological, Beijing, China, #11066-R010-500) was added to the medium at the final concentration of 100 ng/mL.