Transcriptomic Changes Triggered by Hypoxia: Evidence for HIF-1α -Independent, [Na+]i/[K+]i-Mediated, Excitation-Transcription Coupling

This study examines the relative impact of canonical hypoxia-inducible factor-1alpha- (HIF-1α and Na+ i/K+ i-mediated signaling on transcriptomic changes evoked by hypoxia and glucose deprivation. Incubation of RASMC in ischemic conditions resulted in ∼3-fold elevation of [Na+]i and 2-fold reduction of [K+]i. Using global gene expression profiling we found that Na+,K+-ATPase inhibition by ouabain or K+-free medium in rat aortic vascular smooth muscle cells (RASMC) led to the differential expression of dozens of genes whose altered expression was previously detected in cells subjected to hypoxia and ischemia/reperfusion. For further investigations, we selected Cyp1a1, Fos, Atf3, Klf10, Ptgs2, Nr4a1, Per2 and Hes1, i.e. genes possessing the highest increments of expression under sustained Na+,K+-ATPase inhibition and whose implication in the pathogenesis of hypoxia was proved in previous studies. In ouabain-treated RASMC, low-Na+, high-K+ medium abolished amplification of the [Na+]i/[K+]i ratio as well as the increased expression of all tested genes. In cells subjected to hypoxia and glucose deprivation, dissipation of the transmembrane gradient of Na+ and K+ completely eliminated increment of Fos, Atf3, Ptgs2 and Per2 mRNAs and sharply diminished augmentation expression of Klf10, Edn1, Nr4a1 and Hes1. In contrast to low-Na+, high-K+ medium, RASMC transfection with Hif-1a siRNA attenuated increments of Vegfa, Edn1, Klf10 and Nr4a1 mRNAs triggered by hypoxia but did not impact Fos, Atf3, Ptgs2 and Per2 expression. Thus, our investigation demonstrates, for the first time, that Na+ i/K+ i-mediated, Hif-1α- -independent excitation-transcription coupling contributes to transcriptomic changes evoked in RASMC by hypoxia and glucose deprivation.


Introduction
Hypoxia is characteristic of numerous pathologies, including inflammation [1], cancer [2], obesity [3], systemic and pulmonary hypertension [4;5], atherosclerosis [6] and kidney disease [7]. In 1986, Murry and colleagues reported that the size of myocardial infarcts, arising from 40-min occlusion of the circumplex artery, could be reduced by 75% if the myocardium had been subjected to so-called ischemic preconditioning, i.e., several short occlusions interspersed by periods of reperfusion [8]. Later on, the protective action of brief ischemia was documented in other tissues, including blood vessels [9]. Significantly, the prophylactic influence of ischemic preconditioning was at least partially blocked by inhibitors of RNA synthesis [10;11], suggesting a key role of profound transcriptomic changes documented in global gene expression profiling studies of ischemic tissues [12][13][14][15][16][17][18].

Effect of ouabain, K+-free medium and hypoxia on intracellular content of monovalent ions and ATP
Six-hr inhibition of Na + ,K + -ATPase in RASMC by ouabain increased [Na + ] i from 15-20 to 130 mM and decreased [K + ] i from ,150 to 25 mM (Fig. 1). Somewhat similar elevation of the [Na + ] i /[K + ] i ratio was detected with 6 hr of Na + ,K + -ATPase inhibition in K + -free medium. Dissipation of the transmembrane gradients of monovalent cations, triggered by ouabain and K + -free medium, was accompanied by elevation of [Cl -] i from ,40 to 80 and 60 mM, respectively (Fig. 1). Figure 2 shows that 24-hr incubation of RASMC in hypoxia and glucose starvation decreased intracellular ATP content by ,3 -fold whereas ouabain attenuated this parameter by less than 20%. The actions of hypoxia and ouabain on ATP content were preserved in low-Na + , high-K + medium. Treatment with ouabain resulted in almost 10-fold gain of [Na + ] i and virtually similar loss of [K + ] i . In hypoxic conditions, [Na + ] i and [K + ] i were increased and decreased by 3-and 2-fold, respectively. As predicted, dissipation of the transmembrane gradients of monovalent cations in low-Na + , high-K + medium almost completely abolished the actions of ouabain and hypoxia on the [Na + ] i /]K + ] i ratio (Fig. 2). Viewed collectively, these results allowed us to hypothesize that transcriptomics changes triggered by hypoxia are at least partially caused by Na + i /K + i -mediated excitation-transcription coupling discovered in our recent studies [26]. Data considered below support this hypothesis.
Identification and functional characterization of [Na + ] i /]K + ] i -sensitive transcriptome in RASMC Affymetrix data from 3 independent experiments were normalized and analyzed by PCA [27]. Each point on PCA represents the gene expression profile of an individual sample. Samples that are near each other in the resulting 3-dimensional plot have a similar transcriptome while those that are further apart have dissimilar transcriptional profiles. This approach identified ouabain and K + -free medium as major sources of variability within datasets (Fig. 3A). Figure 3B discloses that the number of differentially-expressed transcripts in RASMC treated for 6 hr with ouabain or K + -free medium totalled 8,266 and 8,264, respectively. Further analysis determined that the expression of 6,412 transcripts was affected by both stimuli (Fig. 3B). Significantly, we observed highly significant (p,4610 29 ) and positive (R 2 .0.80) correlations between levels of differentially-expressed transcripts identified in the presence of ouabain and K + -free medium (Fig. 4). Because the gain of [Na + ] i and loss of [K + ] i in cells treated with ouabain or K + -free medium are similar (Fig. 1), the results strongly suggest that the changes in gene expression evoked by both stimuli occur in response to elevation of the [Na + ] i /[K + ] i ratio rather than [Na + ] i /[K + ] iindependent events. Effect of ouabain and hypoxia on intracellular Na + , K + and ATP concentrations. RASMC were incubated for 24 hr under normal oxygen partial pressure (5% CO 2 /air -control) 63 mM ouabain or exposure to hypoxia (5% CO 2 /95% N 2 )/glucose deprivation in normal high-Na + , low-  . Effect of Na + ,K + -ATPase inhibition on the RASMC transcriptome. Cells were incubated for 6 hr in control DMEM, K + -free DMEM or DMEM containing 3 mM ouabain. All experiments are repeated 3 times. A. PCA of transcriptomic changes. Ellipsoids highlight portioning of samples based on type of treatment. The principal components in 3-dimensional graphs (PC#1, PC#2 and PC#3) represent the variability of gene expression level within datasets. B. Comparative analysis of the impact of Na + ,K + -ATPase inhibition by ouabain and K + -free medium on the RASMC transcriptome. The total number of genes whose expression is altered by ouabain and K + -free medium by more than 1.2-fold with p,0.05 is indicated; the number of genes affected by both stimuli appears in bold. doi:10.1371/journal.pone.0110597.g003 It should be noted that, together with increment of the [Na + ] i / [K + ] i ratio, ouabain and K + -free medium may affect cells independently of suppression of Na + ,K + -ATPase-mediated ion fluxes. Thus, recent studies revealed that ouabain triggered interaction of the Na + ,K + -ATPase a-subunit with the membrane-associated nonreceptor tyrosine kinase Src, activation of Ras/Raf/ERK1,2, phosphatidyl inositol 3-kinase (PI(3)K), PI(3)Kdependent protein kinase B, phospholipase C, [Ca 2+ ] i oscillations and augmented production of ROS (for review, see [28;29]). On the other hand, transfer of highly K + -permeable cells to K + -free medium results in transient membrane hyperpolarization that has a tissue-specific impact on the activity of diverse voltage-sensitive, membrane-bound proteins [30;31] and the distribution of other permeable ions, including Cl - (Fig. 1). Indeed, we noted that the expression of several genes detected, such as Cxcl2, Cxcl5, Tnfs5, Lif and Vcam1, is sharply increased in K + -free medium compared to ouabain-treated cells (Table 1). This considered, we focused our analysis on [Na + ] i /[K + ] i -sensitive genes whose expression in K +free medium and in the presence of ouabain was less than 2-fold different. [Na + ] i /[K + ] i -sensitive genes, whose expression was increased or decreased by more that 3-fold, are listed in Tables 2  and 3, respectively. Although functional characterization is somewhat artificialbecause genes are usually multifunctional and fall into several categories -we ascertained that both up-and down-regulated [Na + ] i /[K + ] i -sensitive transcriptomes were enriched with genes involved in transcription/translation, cell adhesion, migration, proliferation, differentiation and death (Tables 1 and 2, Fig. 5). We also noted that, among [Na + ] i /[K + ] i -sensitive genes, the relative content of transcription/translation regulators was ,3-4 -fold higher than in total mammalian genomes [32]. Keeping this in mind, we undertook an additional search for genes encoding HIF-1, AP-1, cyclic AMP response element-binding protein (CREB), nuclear factor kappa-B (NFkB), early growth response factors (EGR), i.e. major transcription factors involved in transcriptomic changes evoked by hypoxia (for review, see [33]). Table 3 demonstrates that Na + ,K + -ATPase inhibition resulted in augmented expression of genes encoding AP-1 and Egr1 and down-regulation of genes encoding regulators of the NFkB-and p53-mediated signalling pathways. We found less than 2-fold elevation of Sp1, Creb1 and Creb5 and lack of any impact of increment of the [Na + ] i /]K + ] i ratio on the transcription of other hypoxia-inducible transcription factors: Hif-1a, Hif-1b, Hif-2a, Hif-3a, p65, cRel, RelB, p50, p52, IkB, p53, Sp3, Gata2, Stat5, Gadd153.
In silico search for subset of genes whose expression is affected both elevation of the [Na + ] i /]K + ] i ratio and hypoxia To select candidate genes for the investigation of relative impact of HIF-1a-and [Na + ] i /[K + ] i -mediated signaling triggered by hypoxia, we performed a comparative analysis of our results on the identification of [Na + ] i /]K + ] i -sensitive in RASMC and PubMed database for genes affected by hypoxia/ischemia. In Tables 1 and  2, [Na + ] i /]K + ] i -sensitive transcripts found in PubMed database as genes affected by hypoxia and/or ischemia are shown in bold with number of publications given in parentheses. Then, we performed and additional search [Na + ] i /]K + ] i -and hypoxiasensitive genes in manuscripts investigated transcriptomic changes triggered by hypoxia/ischemia using the global gene expression profiling technology [12-14;16;17]. [Na + ] i /[K + ] i -sensitive genes found in these papers are italicized in Tables 1 and 2. These two approaches led us to conclusion that relative percentages of hypoxia-sensitive genes among up-and down-regulated [Na + ] i / [K + ] i -sensitive genes were ,40% and 12%, respectively, i.e. much higher than predicted from random distribution in the rat genome of 280 [Na + ] i /[K + ] i -sensitive genes and 60 hypoxia-sensitive genes annotated in Tables 1 and 2.

Role of HIF-1a-and [Na + ] i /[K + ] i -mediated signaling
For further investigations, we selected Cyp1a1, Fos, Atf3, Klf10, Ptgs2, Nr4a1, Per2 and Hes1, i.e. genes possessing the highest increments of expression under sustained Na + ,K + -ATPase inhibition and whose implication in the pathogenesis of hypoxia was proved in previous studies. Thus, FOS and ATF3, together with JUN, form dimeric transcription factor AP-1 whose augmented expression was documented in all types of cells subjected to hypoxia [33]. Ptgs2 encodes an inducible isoform of cyclooxygenase-2 (COX-2) whose role in the pathophysiology of hypoxia is well-documented [34]. Klf10 is a Kruppel-like zincfinger transcription factor family member involved in hypoxiadependent angiogenesis via COX-1 activation [35]. Nerve growth factor IB, also known as Nur77 or Nr4a1, is the nuclear receptor of transcription factors stabilizing HIF-1a, which increases its transcriptional activity [36]. Hes1 is the basic helix-loop-helix transcription factor whose expression is sharply augmented after ischemic renal failure [37]. The core circadian oscillator is composed of a transcription-translation feedback loop in which Clock and Bmal1 are positive regulators, and Per1, Per2, Cry1 and Cry2 act as negative regulators [38]. It has been shown that Per2 promotes circadian stabilization of HIF-1a activity that is critical for myocardial adaptation to ischemia [39;40]. Cyp1a1 encodes a cytochrome P450 family member and its expression is mediated by HIF-1b [44]. Vascular endothelial growth factor (Vegfa) and endothelin (Edn1) were chosen as positive controls for canonical HIF-1a-sensitive genes.
To examine the relative impact of HIF-1a-mediated and [Na + ] i /[K + ] i -dependent signaling, we compared the effects of hypoxia and ouabain on expression of the above-listed selected genes in control high-Na + , low-K + medium, after dissipation of the transmembrane gradients of monovalent cations in high-K + , low-Na + medium and in cells transfected with Hif-1a siRNA. As . Correlation analysis of transcripts whose expression is altered by ouabain and K + -free medium in RVSMC by more than by 1.2-fold with p,0.05. The total number of transcripts subjected to analysis is shown in Figure 2B. Transcript expression in control cells was taken as 1.00. The fold change was determined as log transformed treatment/control expression ratio. doi:10.1371/journal.pone.0110597.g004 Transcriptomic Changes Triggered by Hypoxia PLOS ONE | www.plosone.org Table 1. Genes whose expression was increased in RASMC subjected to Na + ,K + -ATPase inhibition.   (Table 4) and increased immunoreactive HIF-1a protein content by ,5-fold ( Fig. 6). RASMC transfection with Hif-1a siRNA but not with scrambled siRNA decreased Hif-1a expression by ,3-fold and sharply attenuated the rise in HIF-1a protein triggered by hypoxia (Fig. 6).
Ouabain increased baseline Hif-1a mRNA by ,50% (Table 4) and slightly curbed HIF-1a protein content (Fig. 6). Confirming previous observations [10], hypoxia increased Vegfa and Edn1 mRNA content by 12-and 4-fold, respectively (Table 5). Transfection with Hif-1a siRNA decreased hypoxia-dependent increments of Vegfa and Edn1 mRNA by ,4-and 2-fold, respectively ( Fig. 6). Ouabain did not significantly affect Vegfa and augmented Edn1 mRNA by 2.5-fold. Dissipation of the transmembrane gradients of monovalent cations in low-Na + , high-K + medium did not alter the expression of Vegfa triggered by hypoxia and decreased Edn1 mRNA by 2-fold. Viewed collectively, these data strongly support the efficacy of Hif1a-siRNA function. Dissipation of the transmembrane gradients of monovalent cations completely suppressed increments of Fos, Atf3, Ptgs2 and Per2 mRNA and sharply diminished elevations of Klf10, Edn1, Nr4a1 and Hes1 expression seen in hypoxic conditions (Fig. 7). Consistent with data obtained in other cell types, including human VSMC [42;43], hypoxia increased Fos, Atf3, Klf10, Ptgs2, Nr4a2, Per2 and Hes1 expression from 2-to 6-fold (Fig. 7). Transfection with Hif-1a siRNA decreased Klf10 and Nr4a mRNA increments evoked by hypoxia by ,2-fold but did not affect hypoxia-induced Fos, Atf3, Ptgs2 and Per2 expression. In contrast to the other genes listed in Table 4, hypoxia decreased Cyp1a1 mRNA by 2-fold in concordance with attenuated Cyp1a1 expression in the human microvasculature subjected to hypoxia [44]. The expression of all 8 tested genes was heightened from 3to 10-fold in the presence of ouabain. These increments were completely abolished under dissipation of the transmembrane gradients of monovalent cations in low-Na + , high-K + medium. In contrast to low-Na + , high-K + medium, transfection with Hif-1a siRNA did not affect the expression of these genes in ouabaintreated RASMC ( Table 4). Dissipation of the transmembrane gradient of monovalent cations completely inhibited increments of Fos, Atf3, Ptgs2 and Per2 mRNA and sharply diminished elevation of Klf10, Edn1, Nr4a1 and Hes1 expression seen in hypoxic conditions (Fig. 7).

Localization of (A/G)CGTG hypoxia response elements within 59-UTR
Several research teams reported that HIF-1a regulates gene expression in ischemic tissues via interaction of HIF-1a/HIF-1b heterodimer with HREs containing (A/G)CGTG consensus in promoter/enhancer regions of the target gene's DNA such as VEGFA [45] and EDN1 [46]. Considering this, we employed SCOPE service (Suite for Computational Identification of Promoter Elements): http://genie.dartmouth.edu/scope/ [47] for the search of (A/G)CGTG consensus within 59-untranslated regions (59-UTR) of [Na + ] i /[K + ] i -sensitive genes listed in Table 4. Using this approach we found numerous (A/G)CGTG sequences within 59-UTR encoding canonical HIF-1-sensitve genes (Edn1 and Vegfa) as well as all [Na + ] i /[K + ] i -sensitive genes listed in Table 4. Importantly, we failed to find any fixed position for this consensus within 10,000 bp 59-UTRs of HIF1a-sensitive vs HIF1a-resistant genes (Fig. 8). Moreover, we observed that in several [Na + ] i /[K + ] i -sensitive genes proximal 1,500 bp segments of 59-UTRs are more abundant with (A/G)CGTG sequence as compared to canonical HIF-sensitive transcripts (Fig. 9). Thus, 1,500 bp 59-UTRs of Atf3 and Edn1 contains 8 and 3 (A/ G)CGTG sequences. This observation is also confirmed by Sig Value parameter having a value of 28.4 for 1500 bp 59-UTRs of genes listed in Table 5. If the search is not restricted to positions of 1500 bp, Sig Value is negative indicating the absence of predictive capabilities for the consensus sequence.

Discussion
HIF-1a the sole known oxygen sensor, regulates gene expression in ischemic tissues via interaction of HIF-1a/HIF-1b heterodimer with HREs in promoter/enhancer regions of the target gene's DNA [19][20][21][22]. Our investigation demonstrates, for the first time, that [Na + ] i /[K + ] i -sensitive excitation-transcription coupling contributes to the transcriptomic changes triggered by hypoxia independently of HIF-1a-mediated signaling. Evidences supporting this conclusion are listed below.
First, elevation of the [Na + ] i /[K + ] i ratio evoked by 6-hr inhibition of Na + ,K + -ATPase by ouabain or K + -free medium resulted in differential expression of more than 6,000 transcripts (Fig. 3). The list of [Na + ] i /[K + ] i -sensitive genes whose expression changed by more than 3-fold (Tables 1 and 2) is abundant with genes whose differential expression was reported to be affected by hypoxia or ischemia/reperfusion. RASMC were treated with 3 mM ouabain or K + -free medium for 6 hr. Listed are assigned genes whose expression was increased in K + -free medium by more than 3-fold and was different by less than 2-fold in the presence of ouabain compared to K + -free medium. GeneChip expression analysis was performed as described in the Methods section. mRNA content in control cells was taken as 1.00. Genes whose expression is altered in ischemia/hypoxia are shown in bold. Appearing in parentheses are numbers of citations in PubMed. Listed in italics are genes whose differential expression was detected in ischemic tissue by whole genome microarray-based analysis [12][13][14][15][16][17][18]. Data on gene function from GeneCards database (www.genecards.org) were used for identification of gene function.  Table 2. Genes whose expression was decreased in RASMC subjected to Na + ,K + -ATPase inhibition.     (Fig. 2). Previously, it was shown that transient ischemia of cardiac myocytes increased [Na + ] i from 5-8 to 25-40 mM and decreased [K + ] i by 30% [48]. Augmentation of the [Na + ] i /[K + ] i ratio observed in hypoxic RASMC is probably caused by attenuation of intracellular ATP content (Fig. 2) that, in turn, leads to partial inhibition of Na + ,K + -ATPase.
Third, RASMC transfection with Hif-1a siRNA curbed the increment of HIF-1a protein as well as Vegfa, Edn1, Klf10 and RVSMC were treated with 3 mM ouabain or K + -free medium for 6 hr. Listed are assigned genes whose expression was decreased in K + -free medium by more than 3-fold and was different by less than 2-fold in the presence of ouabain compared to K + -free medium. GeneChip expression analysis was performed as described in the Methods section. mRNA content in control cells was taken as 1.00. Genes whose expression is altered in ischemia/hypoxia are shown in bold. Appearing in parentheses are numbers of citations in PubMed. Given in italics are genes whose differential expression was detected in ischemic tissue by whole genome microarray-based analysis [12][13][14][15][16][17][18]. Data on gene function from GeneCards database (www.genecards.org) were used for identification of gene function.  Table 3. Hypoxia-responsive transcription factors whose expression was increased in RASMC subjected to Na + ,K + -ATPase inhibition.
Viewed collectively our results demonstrate the dominated role of [Na + ] i /[K + ] i -mediated excitation-transcription coupling in overall transcriptomic changes triggered by ischemic conditions. It was shown that in ischemic tissues HIF-1a increases expression of Vegfa and Edn1 via interaction of HIF-1a/HIF-1b heterodimer with hypoxia response elements (HRE) encoding by 59-UTR (A/ G)CGTG consensus [45;46;49]. We noted that side-by-side with Hif-1a siRNA-sensitive Vegfa and Edn1, 59-UTRs of Hif-1a siRNA-resistant Atf3, Ptgs2, Fos and Per2 are also abundant with (A/G)CGTG sequences ( Fig. 8 and 9). Therefore, the presence of (A/G)CGTG consensus within 59-UTR is not sufficient alone to predict HIF-1-mediated mechanism of gene expression regulation in hypoxic conditions.
Side-by-side with HIF-1a protein accumulation, hypoxia triggers the expression of other transcription factors listed in Table 3 and reviewed by Cummins et al. [33]). Do these transcription factors contribute to [Na + ] i /[K + ] i -mediated transcriptomic changes evoked by hypoxia? We found a negligible impact of Na + ,K + -ATPase inhibition on Hif-1b expression and 2fold elevation of mRNA encoding aryl hydrocarbon receptors for dioxins, benzopyrenes and other environmental pollutions (AhR) ( Table 3). It was shown that, in addition to HIF-1a, HIF-1b forms a dimer with AhR [50] that leads to similar expression levels of the P450 isoforms CYP1A1 and CYP1B1 via binding of AhR/HIF-1b complex to the TNGCGTG consensus sequence in xenobioticresponsive elements [44;51]. However, the involvement of this regulatory pathway in the expression of [Na + ] i /[K + ] i -sensitive genes seems unlikely. Indeed, we observed very modest elevation of Cyp1b1 expression elicited by ouabain and K + -free medium (1.41-and 1.54-fold, respectively), in contrast to the ,15-fold increase of Cyp1a1 expression (Table 1). We did not find any changes in mRNAs encoding regulatory (p65, cRel, RelB, p50, p52) and inhibitory (IkB) subunits of NFkB. Ikbkg and Ikbke encode kinases that phosphorylate IkB, causing its dissociation and activation of NFkB-mediated transcription, whereas SIKE1 interacts with Ikbke and inhibits it [52]. Both Ikbkg/Ikbke and Sike1 expression was decreased up to 5-fold by ouabain and K + -free medium (Table 3). Thus, the final outcome of elevation of the [Na + ] i /[K + ] i ratio on the activity of this regulatory pathway remains unknown. MDM2 is a major negative regulator of p53. Elevation of the [Na + ] i /[K + ] i ratio increased Mdm2 expression by ,3-fold (Table 3), suggesting inhibition of p53 transcription rather than activation detected in hypoxia [53]. We observed that 6-hr inhibition of the Na + ,K + -ATPase resulted in 2-3-fold attenuation of mRNAs encoding AMP-activated protein kinase (AMPK) regulatory subunits Prkag1 and Prag2 (Table 2). These data suggest that elevation of the [Na + ] i /[K + ] i ratio attenuates rather than activates AMPK whose augmented activity was detected in hypoxic cells [53;54].
Previously, we demonstrated that 3-hr inhibition of Na + ,K + -ATPase in RASMC by ouabain and K + -free medium augmented Egr1 by ,5 and 7-fold, respectively [26]. Prolongation of incubation time up to 6 hr decreased the increments of Egr1 mRNA (Table 3), suggesting transient activation of this transcription factor. We also documented activation of the transcription factor AP-1, indicated by up to 8-fold augmentation of mRNAs encoding its major subunits, including Fos, Jun, Atf3, Maff and Mafk (Table 3). These data show that AP-1 and Egr1 are major  Table 4. Effect of high-K + , low-Na + medium and Hif-1a siRNA on gene expression triggered by hypoxia and ouabain.

4.09+ +58
Control non-transfected RASMC or RASMC transfected with scrambled or Hif-1a siRNA were exposed to hypoxia/glucose deprivation or 3 mM ouabain in control or low-Na + hypoxia-inducible transcriptions factors that are also activated by elevation of the [Na + ] i /[K + ] i ratio. Yan and co-workers reported that hypoxia triggered Egr1 in cultured hepatoma-derived cells deficient in HIF-1b [55]. We state here that Fos mRNA accumulation triggered by ouabain occurs in HIF-1a-deficient RASMC ( Table 4), indicating that that both the Egr1 and AP-1 pathways are initiated in response to oxygen deprivation independently of HIF-1. As an alternative hypothesis, we propose that activation of Egr1, AP-1 and other Hif-1a siRNA-resistant genes listed in Table 4 in hypoxic cells is mediated by ATP depletion, Na + ,K + -ATPase inhibition and dissipation of the transmembrane gradients of monovalent cations (Fig. 10). It was shown that gain of Na + i rather than loss of K + i sparks augmented Fos expression in RASMC [56]. Numerous studies have disclosed that [Na + ] i elevation heightens [Ca 2+ ] i via activation of Na + i /Ca 2+ o exchanger as well as via depolarization and activation of voltage-gated Ca 2+ channels (for review, see [57;58]). It has been well-documented that [Ca 2+ ] i elevation affects gene expression by activation of cAMP-response elements via CREB phosphorylation by (Ca 2+ +calmodulin)-dependent protein kinase and nuclear factor AT (NFAT) dephosphorylation by calcineurin [59]. In previous investigation, we found that addition of extracellular (EGTA and intracellular (BAPTA) Ca2+ chelators increased rather than decreased the number of [Na + ]i/[K + ]isensitive genes [26]. It should be noted, however, that these compounds may affects cellular functions independently of Ca 2+ depletion. Thus, we observed that addition of EGTA increases permeability of vascular smooth muscle cells for Na + [60]. Thus, additional experiments should be performed to clarify relative impact of Ca 2+ -mediated and -independent signaling shown in Fig. 10 in transcriptomics changes evoked by hypoxia.

Conclusion
We report here that elevation of the [Na + ] i /[K + ] i ratio contributes to transcriptomic changes triggered by hypoxia and glucose deprivation independently of HIF-1-mediated signalling. The molecular origin of the upstream sensor and the relative contribution of the gain of [Na + ] i and loss of [K + ] i in the triggering of this novel signalling pathway, including the augmented expression of hypoxia-inducible Egr1 and AP-1 transcription factors, remains unknown. Recent studies have demonstrated that modulation of histone methylation via an epigenetic mechanism is  a key device that cells use to adapt to hypoxia [61]. Increasing evidence indicates that side-by-side with regulation of the 59-UTR by transcription factors, gene activation or silencing is under the complex control of 3-dimensional positioning of genetic materials and chromatin in nuclear spaces [62;63]. The role of the [Na + ] i / [K + ] i ratio in the epigenetic regulation of 3-dimensional genome Figure 7. Effect of hypoxia and ouabain on gene expression in RASMC. Cells were exposed to normoxia, hypoxia/glucose deprivation or 3 mM ouabain for 24 hr in control high-Na + , low-K + medium (A, C), or high-K + , low-Na + medium (B). In some experiments, RASMC were transfected with Hif-1a siRNA (C). mRNA content in normoxia was taken as 1.00 and shown by broken lines. For more details, see figure 4 legend. doi:10.1371/journal.pone.0110597.g007 Figure 8. Position of (A/G)CGTG consensus within 10,000 bp 59-UTR of genes listed in Table 5. doi:10.1371/journal.pone.0110597.g008 Figure 9. Position of (A/G)CGTG consensus within 1,500 bp 59-UTR of genes listed in Table 5. doi:10.1371/journal.pone.0110597.g009 organization and its relationship to gene silencing and activation are currently being examined in our laboratory.

Cell culture
Rat aortic smooth muscle cells (RASMC), purchased from Lonza (Walkersville, MD, USA), were grown at 37uC in a CO 2 incubator in Dulbecco's modified Eagle medium (DMEM, Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 mg/ml streptomycin, and subjected to less than 10 passages. To establish quiescence, the cells were incubated for 24 hr in media in which FBS concentration was reduced to 0.2%.

Cell treatment
Quiescent cells were washed with K + -free DMEM (Sp-DMEM; Invitrogen) and incubated for 6 hr at 37uC in a humidified atmosphere with 5% CO 2 /balance air in control DMEM, K + -free DMEM and DMEM containing 3 mM ouabain. In some experiments, we used DMEM-like medium containing (in mM) NaCl 109. 4 [56]. To inhibit this signalling pathway, 0.1 mM nicardipine was added in certain experiments. To trigger hypoxia, RASMC were incubated in custom-designed, air-tight, flow-through cuvettes in media containing 0.5 mM glucose under substitution of 5% CO 2 /air by 5% CO 2 /N 2 . Eight hr after incubation in a CO 2 /N 2 environment, pO2 was ,30 mm Hg compared to ,150 mmHg in normoxia. We did not observe any impact of these treatment protocols on RASMC survival, estimated by lactate dehydrogenase release, caspase-3 activity and chromatin cleavage (data not included).

Intracellular content of monovalent ions
Intracellular K + , Na + and Clcontent was measured as the steady-state distribution of extra-and intracellular 86 Rb, 22 Na and 36 Cl, respectively. To establish isotope equilibrium, cells growing in 12-well plates were preincubated for 3 hr in control or K + -free medium (Sp-DMEM+Ca) containing 0.5 mCi/ml 86 RbCl, 4 mCi/ ml 22 NaCl or 3 H 36 Cl with ouabain added for the next 3 hr. To test the action of K + -free medium, the cells were washed twice with ice-cold Sp-DMEM+Ca. Then, cells loaded with 22 Na or 36 Cl were transferred to Sp-DMEM+Ca containing 22 NaCl and H 36 Cl, respectively, whereas cells loaded with 86 Rb were transferred to isotope-free Sp-DMEM+Ca. After 3 hr, they were transferred onto ice, washed 4 times with 2 ml of ice-cold medium W containing 100 mM MgCl 2 and 10 mM HEPES-tris buffer (pH 7.4). The washing medium was aspirated and the cells lysed with 1% SDS and 4 mM EDTA solution. Radioactivity of the incubation media and cell lysates was quantified, and intracellular cation content was calculated as A/am, where A was the radioactivity of the samples (cpm), a was the specific radioactivity of 86 Rb (K + ), 22 Na or 36 Cl in medium (cpm/nmol), and m was protein content (mg). For more details, see [64].

Intracellular ATP content
Intracellular ATP content was measured by assaying luciferasedependent luminescence with ATP bioluminescent assay kit (Sigma, St. Louis, MO, USA), as described in detail elsewhere [65].
Intracellula Na + , K + , Cland ATP concentrations Intracellular Na + , K + , Cland ATP concentrations were calculated on the basis of intracellular water volume in cells seeded in 12-well plates. The volume of intracellular water was measured as [ 14 C]-urea available space and calculated as V = A c / A m m, where A c was radioactivity of the cells after 30-min incubation with 2 mCi/ml [ 14 C]-urea (dpm), A m was radioactivity of the incubation medium (dmp/ml), and m was protein content in cell lysates (mg) [66].

RNA isolatio
Total RNA was extracted from cells grown in 6-well plates with TRIzol reagent (Invitrogen) and purified with RNeasy MinElute cleanup kit (Qiagen, Valencia, CA, USA), following the manufacturers' protocols. Only RNA samples that had more than 7.0 RNA integrity number and no detectable genomic DNA contamination were considered for subsequent gene array analyses. RNA quality was assessed by 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Microarray experiments were performed with GeneChip Human Gene 1.0 ST array (which detects 28,869 gene products) and GeneChip Rat Gene 1.0 ST array (which detects 27,342 gene products). On both arrays, each gene was represented by approximately 26 probes along the entire transcript's length (Affymetrix, Santa Clara, CA, USA). 100 ng of total RNA from each sample were processed with Ambion WT expression kit (Invitrogen), a reverse transcription method that specifically primes non-ribosomal RNA, including both poly(A) and non-poly(A) mRNA, and generates sense-strand cDNA as final product. 5.5 mg of single-stranded cDNA was fragmented and labeled by Affymetrix GeneChip WT terminal labeling kit, with 2.0 mg of the resulting cDNA hybridized on chips.

GeneChip expression analysis
RNA samples obtained from control cells and cells subjected to Na + ,K + -ATPase inhibition with ouabain and K + -free medium in 3 independent experiments were employed for GeneChip expression analysis. The controls and treatments were performed in parallel in each experiment independently from the other experiments. The entire hybridization procedure was conducted with the Affymetrix GeneChip system according to the manufacturer's recommended protocol. The hybridization results were evaluated with Affymetrix GeneChip Command Console Software. Chip quality was assessed by Affymetrix Expression Console. The data were analyzed by Partek Genomics Suite (Partek, St. Louis, MO, USA) and uploaded on the GEO repository with the accession number http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=G SE61131.
The normalized data then underwent principal component analysis (PCA) [27] to identify patterns in the dataset and highlight similarities and differences among samples. Major sources of variability found within the dataset by PCA served as grouping variabilities for analysis of variance with n = 4 for each group of samples. The ensuing data were filtered to identify transcripts with statistically significant variations of expression among groups that were modulated by at least 20%, with multiple testing corrections by the false discovery rate. Calculated p-values and geometric fold changes for each probe set identifier were imported into Ingenuity Pathway Analysis (Ingenuity Systems, http://www.ingenuity.com) to ascertain networks, biological functions and their pathophysiological implications. Functional information on regulated genes was also obtained from publications in PubMed.
Real-time quantitative reverse transcription-polymerase chain reaction (RT-PCR) RT-PCR was performed with Express SYBR GreenER qPCR Supermix kit (Invitrogen) according to the manufacturer's instructions. The reaction was carried out with a 7900 HT Fast RT -PCR system (Applied Biosystems, Foster City, CA, USA). The primers presented in Table 5 were designed with Primer3Plus online software from consensus sequences provided by Affymetrix for each gene of interest. All experiments were analyzed in duplicate. b 2 microglobulin mRNA expression served to normalize and compare the expression values of genes of interest. The results were quantified by the DDCt method in Microsoft Excel.

Western blotting
The probe sets were then filtered on flags (present, marginal, or absent), and expression levels were quantified. Statistically significant probe sets were identified by the false discovery rate followed by 2-way ANOVA with strain and age as major factors. Student's t-test was used for 2-group comparisons. When comparing more than 2 groups, 2-way ANOVA was employed with strain and age as the main factors, followed by Tukey's honest significant difference post-hoc test. Correlation analyses were performed with Pearson product-moment correlation coefficient (r). Null hypothesis was rejected whenever p,0.05.