Hypoxia-induced Changes in SUMO Conjugation Affect Transcriptional Regulation Under Low Oxygen*

HeLa cells grown under normoxia or hypoxia for 48h, were subjected to endogenous SUMO-immunoprecipitation in combination with quantitative mass spectrometry (SILAC) to gain insights into differences of the SUMO1 and SUMO2/3 proteome. Proteins whose SUMOylation changed without concomitant change in abundance were predominantly transcriptions factors. Particularly interesting was transcription factor TFAP2A (Activating enhancer binding Protein 2 alpha), whose sumoylation decreased on hypoxia. DeSUMOylation of TFAP2A enhanced transcriptional activity of HIF-1 under hypoxia contributing at the cellular response to low oxygen. Graphical Abstract Highlights Identification of endogenous SUMO1,SUMO2/3 conjugates and their response to hypoxia. TFAP2A is a target of sumoylation and its modification is inhibited under hypoxia. TFAP2A enhances the transcriptional activity of HIF-1 in a sumo-dependent manner. TFAP2A interacts physically with HIF-1 and HIF-2 in a sumo-independent manner. Hypoxia occurs in pathological conditions, such as cancer, as a result of the imbalance between oxygen supply and consumption by proliferating cells. HIFs are critical molecular mediators of the physiological response to hypoxia but also regulate multiple steps of carcinogenesis including tumor progression and metastasis. Recent data support that sumoylation, the covalent attachment of the Small Ubiquitin-related MOdifier (SUMO) to proteins, is involved in the activation of the hypoxic response and the ensuing signaling cascade. To gain insights into differences of the SUMO1 and SUMO2/3 proteome of HeLa cells under normoxia and cells grown for 48 h under hypoxic conditions, we employed endogenous SUMO-immunoprecipitation in combination with quantitative mass spectrometry (SILAC). The group of proteins whose abundance was increased both in the total proteome and in the SUMO IPs from hypoxic conditions was enriched in enzymes linked to the hypoxic response. In contrast, proteins whose SUMOylation status changed without concomitant change in abundance were predominantly transcriptions factors or transcription regulators. Particularly interesting was transcription factor TFAP2A (Activating enhancer binding Protein 2 alpha), whose sumoylation decreased on hypoxia. TFAP2A is known to interact with HIF-1 and we provide evidence that deSUMOylation of TFAP2A enhances the transcriptional activity of HIF-1 under hypoxic conditions. Overall, these results support the notion that SUMO-regulated signaling pathways contribute at many distinct levels to the cellular response to low oxygen.


In Brief
HeLa cells grown under normoxia or hypoxia for 48h, were subjected to endogenous SUMO-immunoprecipitation in combination with quantitative mass spectrometry (SILAC) to gain insights into differences of the SUMO1 and SUMO2/3 proteome. Proteins whose SUMOylation changed without concomitant change in abundance were predominantly transcriptions factors. Particularly interesting was transcription factor TFAP2A (Activating enhancer binding Protein 2 alpha), whose sumoylation decreased on hypoxia. DeSUMOylation of TFAP2A enhanced transcriptional activity of HIF-1 under hypoxia contributing at the cellular response to low oxygen.

Graphical Abstract
some segregation, nuclear transport and DNA damage (8, 10 -12) but requirements for sumoylation are often cell and tissue specific (13).
Several reports have shown that sumoylation fine-tunes the activity of critical mediators of the hypoxia-signaling cascade including HIF-1␣, HIF-2␣, HIF-1␤, and VHL (4), via multiple pathways including SUMO de-conjugation by the Sumo-Specific Isopeptidases (SENPS). For example, SENP1, which can be induced in hypoxia (14), was found to contribute to HIF-1␣ stability during hypoxia (15). On the other hand, enhanced SUMOylation of a subset of cellular proteins observed 24 h into hypoxic exposure (16), was linked to inactivation of SENP1 and SENP3. Interestingly, in addition to regulatory sumoylation of individual target proteins, hypoxia or ischemia also cause global changes in the SUMO proteome or "SUMOome" (17)(18)(19). The effects of these changes on the adaptation to low oxygen concentration are poorly characterized even though they are essential for cell survival (18,19).
To obtain further insights into the interplay between SUMOyaltion and hypoxia, we decided to compare the total, SUMO1 and SUMO2/3 proteomes of normoxic Hela cells with those incubated under hypoxia for 48 h, a time point at which cells have normally adapted transcriptionally to low oxygen conditions. For quantification, we used a SILAC-based approach. Although we did not observe a massive change in the SUMO-ome on hypoxia, we could identify a significant number of proteins with altered sumoylation levels, both for SUMO-1 and for SUMO-2/3, on hypoxia. For a subset of these proteins, modulation of their sumoylation level occurred with concomitant changes in their expression level. More interesting were proteins, SUMOylation of which was changed without concomitant changes in abundance. Those proteins may be targets and/or effectors in the hypoxic signal transduction cascade. One of these proteins is the transcription factor TFAP2A (Activating enhancer binding Protein 2 alpha), a general transcription factor involved in apoptosis, growth and differentiation (20). We demonstrate that sumoylation of TFAP2A is reduced by hypoxia and that undermodified TFAP2A contributes to the transcriptional activation of HIF-1 suggesting direct involvement of TFAP2A and its sumoylation in the cellular response to the hypoxic environment.
Cell Lines, Transfection and Luciferase Assays-Human HeLa cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Biosera, Nuaille, France) containing 10% fetal bovine serum (FBS) and 100 U/ml penicillin-streptomycin (Biosera). Transient transfections were carried out in 10-cm, 6 or 12-well plates by using a PEI (polyethylenimine) transfection reagent and were incubated for 24h. Briefly 15 g of DNA was added in a 10 cm plate in 1.5 ml DMEM without FBS and PS. After mixing 30 g PEI (1 g/l), was added and the mixture was incubated for 10 min. 4.5 ml of DMEM containing 5% FBS were added and the mixture was added on top of the cells for 4 h. Cells were incubated in fresh DMEM with 10% FBS with PS for 24 -48 h. When required, cells were treated for the indicated times under hypoxia, (1% O 2 , 94% N 2 and 5% CO 2 in an IN VIVO 2 200 hypoxia work station (Ruskinn Life Sciences, Bridgend, UK). Reporter gene assays were performed as described previously (25).
Immunoprecipitation-Two hundred milligrams of total mixed HeLa protein lysates from cells growing in normoxia and hypoxia (see above) were used for the SUMO-1 and SUMO-2/3 Immunoprecipitations (IPs). Endogenous SUMO-1 and SUMO-2/3 conjugates immunoprecipitation was described previously (27,28). In brief, HeLa adherent cells were lysed using a denaturing lysis buffer with 1%SDS. To immunoprecipitate SUMOylated proteins, the lysate was diluted 10-fold to achieve Ripa buffer conditions and was incubated with monoclonal anti-SUMO1 and anti-SUMO2/3 antibody (SUMO1 21C7 and SUMO2 8A2) coupled beads at 4°C overnight. Upon extensive washing and mock-elution for 30 min at 37 C without peptide, SUMO conjugates were eluted twice with an excess of epitope spanning peptides. Eluted proteins were precipitated with trichloroacetic acid and resuspended to 50 l before analysis by SDS-PAGE followed by Coomassie stain. (NuPAGE 10% Bis-tris, Invitrogen, Carlsbad, California). Thirty micrograms of the input protein lysate was also analyzed by SDS-PAGE, followed by Coomassie stain. The lanes containing SUMO-1 and SUMO-2/3-purified proteins and the input sample lanes were sliced into twenty-three sections before undergoing in-gel tryptic digestion.
For the anti-HA nondenaturing Immunoprecipitation the following procedure was followed: Briefly, HeLa cells were washed with cold PBS 24 h after transfection and lysed (20 min, 4°C) in buffer containing 25 mM Hepes, pH 7.6, 150 mM NaCl, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, EDTA-free protease inhibitor mixture (Roche, Basel, Switzerland), 50 mM glycerolphosphate, and 10 mM Na 3 VO 4 . After centrifugation, the sample volume was adjusted to 800 l with lysis buffer and samples were incubated for 3 h at 4°C with 1 l of anti-HA antibody. 20 l of Protein A-Sepharose (Amersham Biosciences, Little Chalfont, UK) bead slurry was added and incubation continued for 16 h at 4°C under gentle shaking. Beads were collected by centrifugation, washed 3 times with lysis buffer, and bound proteins were eluted by SDS sample buffer.
Mass Spectrometry-Proteins were separated by one-dimensional SDS-PAGE (4 -12% NuPAGE Bis-Tris Gel, Invitrogen) and the entire lane of the Coomassie blue-stained gel was cut into 23 slices. All slices were reduced with 10 mM DTT for 55 min at 56°C, alkylated with 55 mM iodoacetic acid (IAA) for 20 min at 26°C and digested with modified trypsin (Serva, Heidelberg, Germany) overnight at 37°C. Tryptic peptides were injected into a C18 precolumn (2.5 cm, 360 m o.d., 150 m i.d., Reprosil-Pur 120 Å, 5 m, C18-AQ, Dr Maisch GmbH, Ammerbuch, Germany) at a flow rate of 10 l/min. Bound peptides were eluted and separated on a C18 capillary column (15 cm, 360 m o.d., 75 m i.d., Reprosil-Pur 120 Å, 3 m, C18-AQ, Dr Maisch GmbH) at a flow rate of 300 nl/min, with a gradient from 7.5 to 37.5% Acetonitrile (ACN) in 0.1% formic acid for 50 min using an Agilent 1100 nano-flow LC system (Agilent Technologies, Santa Clara, California) coupled to an LTQ-Orbitrap XL hybrid mass spectrometer (Thermo Electron, Waltham, Massachusetts). The mass spectrometer was operated in the data-dependent mode to automatically switch between MS and MS/MS acquisition. Survey MS spectra were acquired in the Orbitrap (m/z 350 -1600) with the resolution set to 30,000 at m/z 400 and automatic gain control target at 5 ϫ 10 5 . The eight most intense ions were sequentially isolated for CID MS/MS fragmentation and detection in the linear ion trap. Ions with single and unrecognized charge states were excluded. The Raw Data was analyzed with Maxquant 1.3.0.5 and searched against Uniprot_ human_270812 database (2012_06 86725 sequences). Precursor and fragment ion tolerance was set to 20 ppm. Trypsin was allowed to cleave after Lysine and arginine with one missed cleavage. Fixed modification was Carbamidomethyl (C) and variable modification was Oxidation (M) and Acetyl (Protein N-Term). False discovery rate was 0.01 for precursor and fragments. Raw data and Maxquant tables have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository (29) with the data set identifier PXD010946.
The following statistics have been applied: Log 2 ratios (median) obtained from MaxQuant analyses (supplemental Table S1) of Hypoxia treated versus normoxic cells were tested to be different from zero (no changes) using limma package version 3.36.5 (30) within R statistical environment (R version 3.5.0). Specifically, a one-sample t test was approximated using a model of the form y ϭ a with a being a mean log2 ratio from the two biological experiments for a given protein in an experiment (Input, Sumo1 and Sumo2/3 IP). Only proteins having log2-ratios reported in both experiments were included in the analysis. Moderated t-statistics and p values resulting from empirical Bayes approach were used for the statistical interpretation of the intercept parameter a being nonzero. For multiple testing correction, an approach proposed by Storey et al. (31,32) was applied as implemented within an R package "qvalue" (version 2.12.0).
Immunofluoresence-HeLa cells were grown on coverslips incubated at normoxia or hypoxia for 8 -48 h and analyzed by immunofluoresence as previously described (23). Coverslips were incubated with a rabbit monoclonal anti-TFAP2A antibody (1:100 dilution), and with an Alexa 488-conjugated anti-rabbit secondary antibody (1: 1000, Jackson ImmunoResearch, Cambridgeshire, UK). Images were taken on a Zeiss Axioplan fluorescence microscope using an AxioCam MRm CCD sensor and 100ϫ objective with suitable filters.
Experimental Design and Statistical Rationale-Two biological experiments were performed to compare protein expression profile and the SUMO proteome in cells growing in normoxia versus cells growing for 48 h under 1% hypoxia. The Stable Isotope Labeling of Amino acids in Cell culture (SILAC) technique was used for quantitation of the proteomic results (26). The 2-state SILAC labeling was reversed between the two experiments (see above). Cell lysates from cells growing in normoxia and hypoxia in each experiment were pooled and subjected to SUMO-1 and SUMO2 immunoprecipitation. Input and SUMO-1 and SUMO-2/3 immunoprecipitated proteins were subjected to trypsin in-gel digestion, analysis by high-resolution LC-MS/MS and changes were estimated on precursor peptide intensities in MAxQuant (see above). Intensity ratios (log2) of Hypoxia treated cells/untreated (normoxic) cells for the proteins that detected in the two biological experiments was calculated and plotted in heat map and x/y scatter plots. (supplementary File S1 and Fig. 1).
Values shown for luciferase activity assays are derived from a minimum of three independent experiments performed in triplicates. For Statistical analysis the Graph Pad Instat Statistical package for Mac was used. Data are expressed as mean Ϯ s.e.m. Differences were examined by Student's t test (two-tailed) between two groups. p Ͻ 0.05 was considered statistically significant.

Identification of Endogenous SUMO-1 and SUMO-2/3 Conjugates and Their
Response to Hypoxia-Although a few studies report global changes to the SUMO proteome in ischemic models (oxygen and glucose deprivation) (33), specific targets and their physiological relevance are yet unclear. To identify specific cellular targets whose sumoylation is altered 48 h after initiation of hypoxia, we undertook a quantitative proteomic approach combined with a recently established protocol for denaturing immunoprecipitation of endogenous Sumo species (27,28). The Stable Isotope Labeling of Amino acids in Cell culture (SILAC) technique was used for quantitation of the proteomic results (26). Specifically, HeLa cells were grown either under normoxia in SILAC medium containing nonlabeled amino-acids (Lys 0 Arg 0 -Light) or incubated under 1% O 2 for 48 h in SILAC medium containing isotopically labeled amino-acids (Lys 4 Arg 6 -Heavy) (Fig. 1A). The experiment was repeated with reverse labeling (Fig. 1A) to identify and reject nonspecific contaminants of the immunoprecipitation.
Analysis of the cell lysates demonstrated the expected induction of HIF-1␣ expression under hypoxia (supplemental Analysis of the SUMO-1 and SUMO-2/3 IP eluates by Coomassie stain, followed by trypsin digestion (supplemental Fig.  S2) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) yielded quantitative data for several hundred sumoylated proteins in each IP (592 proteins were identified in the 1st SUMO1-IP and 430 in the 2nd IP, 554 were identified in the 1st SUMO2/3-IP and 332 in the 2nd IP, supplementary Files S1, S2 and S3). From our total proteins identified, 350 were recovered from both SUMO-1 IPs and 272 proteins from both SUMO-2/3 IPs (supplementary File S1). The SUMO conjugates identified by our analysis showed significant overlap with the ones initially identified by Becker et al. (28) (16), who employed a similar procedure to investigate the SUMO1 proteome of HeLa cells 24 h after initiation of hypoxia.
As shown in Fig. 1B and 1C, the number of proteins modified by SUMO-1 or SUMO-2/3 was not massively affected by exposure to hypoxia (marked as unchanged in Fig. 1C). Interestingly, close to 10% of our total proteins identified as SUMO conjugates were found to modulate their sumoylation after hypoxia (34 proteins for SUMO-1 and 28 for SUMO-2/3, Fig.  1C, marked in the light red squares, see also Table I). There were several proteins (17 modified by SUMO-1 and 11 by SUMO-2/3) that were enriched in the SUMO IP eluates from hypoxic cells but were also expressed at higher levels in the same cell lysates (Table II, see also supplementary Files S4), suggesting that the increased sumoylation level of the corresponding polypeptides correlated to the increased abundance of the proteins on hypoxia. These proteins included glycolytic enzymes, such as ALDOA, ENO1 and PGK1, and LDHA, all of which are known targets of hypoxia and HIF-1 (40), and up-regulation of their gene expression by hypoxia could explain their enrichment in our SUMO IPs. There has FIG. 1. Identification of endogenous SUMO-1 and Sumo 2/3 conjugates under hypoxia. A, Overview of the work flow of protein identification and quantification. HeLa cells were either untreated (normoxia) or exposed to 1% O 2 for 48h (hypoxia). Labeling of cells is indicated for both experiments (Light SILAC medium Lys 0 Arg 0 and heavy SILAC medium Lys 4 Arg 6 ). After denaturing lysis, the cell lysates were pooled and subjected to SUMO-1 and SUMO2 immunoprecipitation. Input and SUMO-1 and SUMO-2/3 immunoprecipitates were analyzed by SDS-PAGE electrophoresis and Coomassie staining. Proteins were further subjected to trypsin in-gel digestion, analysis by high-resolution LC-MS/MS and data processing (see Materials and Methods). B, Heat map representing the average of hypoxia treated/untreated (normoxia) log2 intensity ratios for the proteins that were detected in two biological experiments. First column includes proteins found in the INPUT analysis, the second and third columns includes proteins found in the SUMO-1 and SUMO-2/3 immunoprecipitates, respectively. Increased or decreased abundance of proteins on hypoxia is represented with the red to blue color scale. White indicates no change in protein abundance between hypoxia and normoxia or lack of data (protein not detected in any experiment). C, x/y scatter plots, representing comparison of hypoxia treated/untreated (normoxia) log2 intensity ratios for proteins common to both SILAC-SUMO-2/3 IPs (left) and SILAC-SUMO-1 IPs (right). Each protein is represented by one single point with coordinates coming from both IP experiments (x axis: log2 ratio heavy (hypoxia)/Light (normoxia) intensity ratio, y axis: log2 ratio Light (hypoxia)/heavy (normoxia) intensity ratio). Ratio cut-offs were determined graphically at 0.5 (dotted lines). Proteins that were only found more abundant in the unlabeled (Light) form in both experiments (upper left square), were rejected as external contaminants (in a red circle). verified in western blot 6 6 been substantial evidence for sumoylation of several glycolytic enzymes in the literature. ALDOA, ENO1, LDH and PGK1 have been found in several SUMO-IP experiements in S. cerevisiae (41) and human cells (16,37,42,43) both under normoxia and hypoxia (16,41). On the other hand, our analysis also identified proteins with altered sumoylation levels (Table I and Table III, see also supplementary Files S4), whose expression levels did not seem to significantly differ between normoxia and hypoxia. This group was enriched for factors linked to transcription and RNA stability. To validate their hypoxia-dependent changes in the sumoylation status, we turned to immunoprecipitation followed by immunoblotting, using specific antibodies for each target. As shown in Figs. 2A, supplemental Fig. S3 and Fig. S4, we could validate both, stable expression and changes in SUMOylation for several proteins involved in transcriptional regulation such as TFAP2A ( Fig. 2A), KCTD15, NFRKB, ZBTB38 (supplemental Fig. S3), ATRX, IRFBP2 (supplemental Fig. S4), two proteins with unknown function, ZNF687 (supplemental Fig. S3) and ZFP106 (supplemental Fig. S4), and for Exosc10 (Exosome subunit 10, known as Rrp6), a subunit of the RNA exosome (supplemental Fig. S4). Most of the sumoylated proteins migrated ϳ20 kDa higher than the unmodified proteins as shown in the immunoblot analysis and by the apparent MW at which they were identified in the MS analysis (supplemental Fig. S2), indicative of mono-SUMOylation. Some proteins such as KCTD15, ZNF687, and NFRKB gave rise to multiple higher migrating bands in the SUMO-2/3 IPs, indicating poly-sumoylation (supplemental Fig. S3, marked with asterisks).
TFAP2A Is a Target of Sumoylation and Its Modification is Inhibited Under Hypoxia-Among the nine SUMO targets for which we could validate modification changes in response to hypoxia that are independent of expression levels, we were most interested in TFAP2A, as it had previously been linked to hypoxia. TFAP2A was reported to interact with HIF-1␣, HIF-2␣ and CITED2 (21,44,45), and although it was known that TFAP2A could be sumoylated (46), nothing was known

and SUMO-2/3 identified in two IPs that modulate both their sumoylation by SUMO1 or SUMO-2/3 and their protein expression levels upon hypoxia (increase in SUMO conjugation and protein expression is marked with ϩ and decrease with Ϫ)
Genes 40S ribosomal protein S16 Ϫ STRAP Serine-threonine kinase receptor-associated protein Ϫ about the involvement of its sumoylation in these interactions or the hypoxic response. As shown in Fig. 2A, analysis of input samples and SUMO-2/3 immunoprecipitates with anti-TFAP2A antibodies confirmed a) that levels of endogenous TFAP2A do not change and b) that TFAP2A SUMOylation is lost on hypoxia. Moreover, the apparent MW of SUMOylated TFAP2A in the IP, which is 20 kDa larger than the calculated MW of unmodified TFAP2A, is indicative of mono-SUMOylated TFAP2A species. Independent experiments confirmed that the protein expression levels of TFAP2A did not significantly change after incubation of cells for 24 or 48h under hypoxia (Fig. 2B). Further, hypoxic treatment did not affect the subcellular localization of TFAP2A, which remained nuclear under all conditions (Fig.  2C). To begin to address the molecular mechanisms that may account for the loss of TFAP2A SUMOylation under hypoxia, we tested for transcriptional induction of SUMO isopeptidases by Q-PCR. As shown in supplemental Fig. S5, we did not detect significant changes in SENP mRNA levels. Taken together, our data confirmed that endogenous TFAP2A is a target for post-translational modification by SUMO-2/3 and revealed that its sumoylation is specifically inhibited under hypoxic conditions. The underlying molecular mechanism will have to await further investigations (see Discussion).
SUMOylation of TFAP2A on Lysine 10 Is Lost Under Hypoxia-TFAP2A is a member of the developmentally regulated family of AP-2 transcription factors. Individual members have both overlapping and distinct roles in development and in tumor biology. For example, although TFAP2A and TFAP2C sit on the same luminal gene cluster in breast cancer epithelial cells, only TFAP2C can induce the expression of luminal breast cancer genes. As shown by Bogachek et al. (46), this difference can be attributed to SUMOylation of TFAP2A on lysine 10, which has a repressive function. Our finding that TFAP2A is desumoylated in response to hypoxia suggests that this may revert TFAP2A from a repressor to an activator, at least on a subset of genes.
To follow up on this idea, we mutated the only consensus sumoylation site at Lys10 (46) (highlighted in Fig. 3A) into Arg. Wild-type HA-tagged TFAP2A and its SUMO-deficient K10R mutant were then overexpressed in HeLa cells and analyzed by immunoblotting. Both wild-type HA-TFAP2A and the K10R mutant were detected in HeLa cell lysates as 55 kDa protein bands by an anti-HA antibody (Fig. 3B). Further, an additional band with higher MW was recognized by the anti-HA antibody only in the case of wild-type HA-TFAP2A but not in the K10R TFAP2A mutant (marked with asterisk, Fig. 3B). To confirm the identity of this band as the sumoylated form of TFAP2A, we coexpressed a His-tagged form of SUMO2 and performed immunoprecipitation of the HA-TFAP2A forms with an anti-HA antibody. On analysis of the immunoprecipitates with an anti-HA antibody (Fig. 3B right upper panel, full blot can be found in supplemental Fig. S6A) we detected the same higher MW band (marked with asterisk), which was also recognized in another blot by an anti-His antibody (Fig. 3B, right bottom panel, full blot can be found in supplemental Fig. S6A) confirming that it indeed corresponded to the sumoylated form of TFAP2A. The absence of this form from the K10R mutant   (36,47), sumoylation at the K10 residue appears, according to our data, to be the predominant one in our system. We next compared the expression of wild-type TFAP2A and its K10R mutant form under normoxia or after 24-hour exposure to hypoxia (Fig. 3D). No apparent changes could be observed in the expression levels of the wild-type HA-TFAP2A under normoxia or hypoxia (Fig. 3C left panels, full blot can be found in supplemental Fig. S6B) but following immunoprecipitation, the sumoylated form of HA-TFAP2A wt, as detected by both anti-HA and anti-SUMO-2 antibodies, was clearly reduced under hypoxia (Fig. 3C right panels, marked with asterisk, full blot can be found in supplemental Fig. S6B), showing that sumoylation of the tagged over-expressed form of TFAP2A is regulated by hypoxia in a similar manner as endogenous TFAP2A. Of note, consistent with the observation that endogenous TFAP2A was only identified in SUMO2/3but not in SUMO1 IPs, analysis of immunoprecipitated tagged TFAP2A did not yield specific signals with SUMO1 antibodies (supplemental Fig. S6C).

and SUMO-2/3 identified in two IPs that modulate their sumoylation by SUMO1 or SUMO-2/3. Protein expression levels upon hypoxia were not changed or not detected by MS analysis. Further verification of these targets by small scale IPs and WB analysis is marked with an x in a different column
SUMOylation Deficient TFAP2A Is as Transcriptionally Active as wt TFAP2A-To test whether sumoylation of TFAP2A plays a role in its transcriptional activity, we used a luciferase reporter assay with plasmid p3xAP2-Bluc (see Materials and Methods). We could show that both wild-type and K10R mutant forms of TFAP2A were transcriptionally active, under both normoxia and hypoxia (Fig. 3D). In both cases the K10R mutant form of TFAP2A exhibited slightly higher activity than the wild-type form but the difference did not reach statistical significance, suggesting that neither the K10R mutation per se nor the lack of sumoylation at Lys10 affected significantly the transcriptional activity of TFAP2A. Although we can conclude from these experiments that SUMOylation is not required for TFAP2A activity, it remains possible that SUMOylation inhibits TFAP2A because the small fraction of sumoylated TFAP2A may not be enough to cause a significant decrease in the activity of the reporter constructs (Fig. 3B).
TFAP2A Enhances the Transcriptional Activity of HIF-1 in a Sumoylation-dependent Manner-As described above, TFAP2A is known to interact with other transcription factors including FIG. 2. TFAP2A sumoylation is inhibited under hypoxia. A, Soluble extracts (Input) and SUMO-2/3 immunoprecipitates, from Hela cells incubated under normoxia or hypoxia for 24 h, were analyzed by immunoblotting using a rabbit monoclonal anti-TFAP2A and a rabbit anti-SUMO2 antibody for verification of endogenous SUMO species enrichment. The asterisk shows the sumoylated version of TFAP2A and the arrowhead shows free Sumo2. Beta-actin was used as loading control. B, Hela cells were incubated in normoxia or hypoxia for the indicated times and expression levels of endogenous TFAP2A were analyzed by immunoblotting. Endogenous TFAP2A expression from 3 independent experiments was normalized against corresponding expression of ␤-actin and expressed as fold increase in relation to TFAP2A expression under normoxia (p values Ͼ0.05). C, Hela cells were incubated as in (B) and localization of endogenous TFAP2A was detected by indirect immunofluorescence microscopy. DAPI was used for nucleal visualization.
HIF-1. It is thus conceivable that SUMOylation of TFAP2A does not (only) serve to regulate all TFAP2A-dependent genes, but genes, expression of which expression depends on both HIF-1 and TFAP2A. To test this hypothesis, wild-type HA-TFAP2A and its K10R mutant sumo-deficient form were expressed in HeLa cells carrying an HRE-dependent luciferase reporter plasmid (pGL3-5HRE-VEGF). Overexpression of HA-TFAP2A stimulated (close to 3-fold) the HRE-dependent transcriptional activity observed under hypoxia (Fig. 4A) and this stimulation was similar for both wild-type and mutant forms of TFAP2A, as anticipated by the fact that sumoylation of TFAP2A is lost under hypoxia (Fig. 3D). To test whether the effect of TFAP2A on HRE-dependent transcription involves HIF-1, wild-type HA-TFAP2A and its K10R mutant sumodeficient form were coexpressed with Flag-HIF-1␣ in HeLa cells carrying the HRE-reporter and grown under normoxia. The transcriptional activity of HIF-1 in these cells was stimulated by both the wild-type and K10R mutant sumo-deficient forms of HA-TFAP2A but the effect of the K10R mutant form was significantly stronger (Fig. 4B), suggesting that lack of sumoylation facilitates a positive effect of TFAP2A on the activity of HIF-1. We could further show that the sumoylated form of TFAP2A was not affected in cells overexpressing HIF-1␣ under normoxia (Fig. 4C), suggesting that desumoylation of TFAP2A under hypoxia is not HIF-1 dependent.
TFAP2A Interacts Physically with HIF-1 and HIF-2 In a Sumoylation-independent Manner-Consistent with previous studies (45), the effect of TFAP2A on HIF-1␣ activity indicated that the two proteins may be interacting. To test whether SUMO contributes to this interaction, wild-type and K10R mutant sumo-deficient forms of HA-TFAP2A were immunoprecipitated from normoxic cells expressing also either Flag-HIF-1␣ or Flag-HIF-2␣. Analysis of the IPs showed association of TFAP2A with both ⌯⌱F-␣ isoforms ( Fig. 5A and 5B right panels) and this association was similar for both wildtype and K10R mutant sumo-deficient forms of HA-TFAP2A. ARNT (HIF-1␤) was also equally recovered in the IPs of both TFAP2A forms. Given that the HIF-1␣/ARNT (HIF-1) or HIF-2␣/ARNT (HIF-2) heterodimers are exclusively nuclear (48), these data suggest that TFAP2A can bind to both HIF-1 and HIF-2 inside the nucleus and may facilitate their transcriptional activity. Sumoylation of TFAP2A appears not to be required for the TFAP2A-HIF interaction although it enhances HIF-1 activity, suggesting that de-sumoylation of TFAP2A triggered by hypoxia may be required for the transcriptional activation of the TFAP2A-HIF complexes. However, we cannot exclude that SUMO to some extend could also inhibit the TFAP2A-HIF-1␣ interaction, but is not observed in our experiment because of low level of sumoylated TFAP2A in our IPs. DISCUSSION Sumoylation facilitates responses to physiological and pathophysiological events by either promoting global changes to the SUMO conjugation of proteins or by targeting specific and critical substrates of the ensuing signaling cascades. Increased global protein SUMO-conjugation under hypoxia was initially reported as a result of increased SUMO-1 expression (49), which was also confirmed by an in vivo study in hypoxic mouse brains and hearts (50). Global changes in the "SUMO proteome" were also reported in ischemic and OGD (oxygen glucose deprivation) models (17)(18)(19)51). A significant increase in the SUMO-2/3 (but not SUMO-1) conjugates was observed in the recovery period following harmful ischemia but the functional consequences of these global changes were poorly understood although they were required for survival (18,19).
In contrast to previous studies that focused on acute stress response, we investigated the SUMO proteome of cells that have already been kept for 48 h under hypoxic condition, and which have had significant transcriptional response to adapt to the stress (adaptive response included up-regulation of several SUMO targets; see below). At this late time point, we did not observe any significant increase in the amount of high molecular weight SUMO conjugates (supplemental Fig. S1), consistent with our quantitative proteomic results obtained from the 48 h time point, which did not show a massive increase in the number of SUMO-1 or SUMO-2/3 modified proteins. Previous studies have shown an increase in SUMO-1 conjugates in different cells of under early hours of hypoxia (16,49). We did however observe significant sumoylation-status changes in a small group (20 -30) of proteins. In line with this, down-regulation of Ubc9 acetylation via SIRT1 under hypoxia was previously shown to affect sumoylation/ desumoylation of a specific group of protein targets (52).
Our proteomic approach identified two categories of sumoylation targets, proteins whose expression and SUMOylation increased under hypoxia, and proteins whose expression remained constant while they lost or gained SUMO. The first group consisted mostly of glycolytic enzymes and known targets of HIF-1, the major transcription factor induced under hypoxia. This finding suggests that both SUMO-1 and SUMO-2/3 modification may be important for the function of enzymes like ALDOA, ENO1, PGK1 and LDHA that mediate metabolic adaptation to hypoxia. Sumoylation of several glycolytic enzymes like ALDOA, ENO1, LDH and PGK1 have been found in several SUMO-IP experiements in S. cerevisiae (41) and human cells (16,37,42,43) confirming once more the specificity of our method.
Intriguingly, the second group of proteins consisted mostly of transcription factors or regulators involved in activation or repression of transcription such as TFAP2A, ATRX, IRF2BP2, KCTD15, NFRKB, and ZBTB38. Some of these targets were also identified as hypoxia-affected SUMO1 targets in a recent study 24 h after hypoxia (16).
Focusing on the last group of proteins, we singled out transcription factor TFAP2A and could show that hypoxia reduced its sumoylation (by SUMO-2) without affecting its protein expression levels. TFAP2A coordinates a variety of cell processes, including cell differentiation during embryogenesis, growth, apoptosis and survival. The TFAP2A family of transcription factors is known to mediate both activation and repression of target genes (20). Changes in TFAP2A expression have been frequently detected in different types of cancer including melanoma, neuroglioma, and prostate or breast cancers (44,(53)(54)(55), indicating that TFAP2A plays a role in tumorigenesis, tumor invasion and metastasis. Sumoylation of TFAP2A has been previously shown to be involved in maintaining the basal cancer phenotype in breast cancer cells (46). Inhibiting sumoylation of TFAP2A by mutating the SU-MO-acceptor site enhanced the positive effect of TFAP2A on the transcriptional activity of HIF-1. In line with previous reports (44,45), we could also show that TFAP2A can physically interact with both HIF-1␣ and HIF-2␣. ⌯owever, sumoylation of TFAP2A was not critical for this interaction suggesting that the stimulatory role of TFAP2A de-sumoylation involves the interaction of the TFAP2A/HIF complex with chromatin or other transcriptional coactivators. TFAP2A as well as HIF-1␣ are known to bind to coactivators such as CITED-2 and p300/CBP. It has been suggested that TFAP2A may compete with HIF-1␣ for binding to CITED2 (56). However, our results suggest that these two factors rather cooperate than compete as part of the cellular response to hypoxia and such a cooperation has been documented in melanoma cells (44). Our data are also in line with a previous report showing that knockdown of TFAP2A inhibited expression of VEGF, a known HIF target, and impaired cell viability under hypoxia (45).
Another question raised by our findings is what triggers de-sumoylation of TFAP2A under hypoxia. One possibility is activation of a specific SUMO isopeptidase (SENP) responsible for the de-sumoylation of TFAP2A. This seems unlikely, considering that very few targets are deSUMOylated and that isopeptidases are not transcriptionally up-regulated in our experiments. Moreover, Kunz et al. (16) have reported that SENP1 is inhibited in hypoxic conditions. Alternatively, a specific E3 ligase for TFAP2A sumoylation is impaired under hypoxia, or TFAP2A aquires a modification that prevents or allows recruitment of unregulated E3 ligases or isopeptidases. Further investigation of the links between hypoxia and the sumoylation machinery is required to address this question. Our demonstration that lack of oxygen can either stimulate or inhibit the sumoylation of a specific set of proteins that, as shown for TFAP2A, can in turn modulate the transcriptional response to hypoxia underlines the significance of the connections between hypoxia and sumoylation and their implications for hypoxia related diseases.