STAT3 regulates the expression of a specific subset of hypoxia-dependent genes
To analyze the involvement of STAT3 in hypoxia-dependent processes, we decided to perform RNAseq and compare the data of Stat3+/+ and Stat3−/− cells incubated for 24 hours either in normoxia or in hypoxia (1% O2 tension). As shown in Fig. 1A and in Fig. S2, the comparison between Stat3+/+ ESCs grown in normoxia or hypoxia revealed a relevant group of differentially expressed genes. In particular, as reported in the Volcano plot (Fig. 1A), the expression of about 1’500 genes is altered in hypoxia and a large part of these genes are upregulated. As expected, genes involved in glycolysis and angiogenesis belong to the subset of upregulated genes. Among them we found Vascular Epithelial Growth Factor A (Vegfa), Hexokinase 1 (Hk1), Hexokinase 2 (Hk2), Phosphofructokinase (Pfkp), Lactate dehydrogenase A (Ldha), Aldolase A (Aldoa), Phosphoinositide-dependent kinase 1 (Pdk1) (Fig. 1A). Interestingly, when comparing the number of genes affected by hypoxia in Stat3+/+ and in Stat3−/− ESCs, we saw that the effect of hypoxia in Stat3 mutant cells was significantly attenuated: as reported in the box plots of Fig. 1B and B’, only 70% of hypoxia-dependent genes are induced by hypoxia in Stat3 null cells (FC > 1 and p-value < 0.05). Hence, 30% of hypoxia-responsive genes are not significantly induced when Stat3 is knocked-out. Of note, the comparison between hypoxic and normoxic Stat3−/− cells revealed a significant difference in both down- and up-regulated transcripts, however, this response to low oxygen was weaker when compared to the one observed in hypoxic Stat3+/+ ESCs (Fig. 1B,B’). Indeed, the comparison between hypoxic Stat3+/+ and hypoxic Stat3−/− cells allowed us to identify a subset of hypoxia responsive genes whose induction is significantly dampened in Stat3−/− cells. This subset of genes is reported in the heatmap of Fig. 1C and contains canonical Hif1α targets, such as Vegfa, Hk1, Hk2, Pfkp and Hypoxia lipid droplet-associated (Hilpda) whose expressions were plotted in Fig. 1D and validated with RT-qPCR in Fig. 1E. Moreover, as shown in Fig. S1, the comparison between Stat3+/+ and Stat3−/− hypoxic ESCs shows that hypoxia does not affect the expression of STAT3 target genes: as reported in Fig. S1, volcano plot and bar plots reveal that the overall number of STAT3-dependent genes does not change when cells are incubated in normoxia or in hypoxia. Box-plots reported in Fig. S1B demonstrated that hypoxia does not significantly affect the expression of STAT3-related genes. Stat3 itself and some of its target genes, such as Socs3 and Klf4, are downregulated in Stat3 knock-out cells, but they are not differentially expressed when comparing normoxic and hypoxic Stat3+/+ cells (Fig. S1D,E,F). Only Tet2 is significantly downregulated by hypoxia in Stat3+/+ ESCs, but this result is not confirmed by RT-qPCR (Fig. S1G). We can conclude that STAT3 plays an important role in the induction of hypoxia-dependent genes, including known HIF1α-targets involved in the regulation of glycolytic metabolism and in vascular remodeling (Fig. S2B), while HIF1α does not regulate STAT3 activity, at least as regards mouse ESCs in which STAT3 activity is constitutively induced by LIF in the culture media (Ohtsuka, et al., 2015; Carbognin et al., 2016; Betto et al., 2021; Wulansari et al., 2021). It is worth mentioning that some HIF1α-dependent genes are affected by hypoxia either in Stat3+/+ and Stat3−/− cells. In Fig. S2C we show that the expression of Egln3, one of the main targets of HIF1α which encodes for PHD determining the elastic feedback loop of the oxygen deprivation response (Minamishima et al., 2009; Walmsley et al., 2011; Santhakumar et al., 2012), is not affected by Stat3 mutation, demonstrating that not all the hypoxia-dependent genes rely on STAT3 activity, but only a very specific subset of genes. Moreover, we measured the levels of Y705 and S727 phosphorylation in ESCs incubated with low oxygen tensions for 24 hours and compared them with normoxic cells. As shown in Fig. S3A, hypoxia do not significantly affect the total level of STAT3 protein nor pSTAT3 Y705 and pSTAT3 S727 in ESCs.
To understand how STAT3 regulates hypoxia-dependent gene expression, we decided to study the crosstalk between STAT3 and HIF1α. We first asked whether the expression levels of Hif1α mRNA are affected by the genetic ablation of Stat3. Interestingly, as reported in Fig. 2A, there are no significant differences in the expression of Hif1α in Stat3−/− cells compared to Stat3+/+. Therefore, we sought to assess whether STAT3 is involved in the stabilization of HIF1α in hypoxia. Interestingly, as reported in Fig. 2B, western blot analysis revealed that hypoxia can stabilize HIF1α both in Stat3+/+ and in Stat3−/− cells. Once demonstrated that in mouse ESCs STAT3 is neither involved in the expression of Hif1α mRNA nor in HIF1α protein stabilization, we decided to test whether STAT3 might interact directly with HIF1α. To do so, we performed Proximity Ligation Assay (PLA) and found that PLA positive dots could be detected only in murine Stat3+/+ ESCs grown at low oxygen tensions (Fig. 2C), revealing that STAT3 and HIF1α interact with each other in hypoxic conditions. Furthermore, the cellular localization of dots indicates that the interaction between the two transcription factors occurs in the nucleus.
Next, we tested whether the interaction between STAT3 and HIF1α can be observed also in pseudohypoxia. Pseudohypoxia is a condition that occurs whenever HIF1α is stabilized even if the organism is exposed to atmospheric oxygen tensions. This phenomenon can be chemically triggered by some compounds like cobalt chloride (CoCl2) which substitutes with Co2+ the Fe2+ ion necessary for the catalytic activity of PHD, hence inactivating the enzyme (Triantafyllou et al., 2006; Elks et al., 2015; Munos-Sanchez and Chanez-Cardenas, 2018). As expected, wild type ESCs incubated for 24 hours with CoCl2 were characterized by high levels of expression of HIF1α target genes Vegfa and Hk2 compared to untreated Stat3+/+ cells. In contrast, Stat3−/− cells incubated with CoCl2 did not show significant upregulation of these transcripts (Fig. S3B,C), while Hif1α mRNA levels were not affected by the treatment (Fig. S3D). On the other hand, Egln3, which is a pure target gene of HIF1α involved in the elastic feedback loop of the oxygen deprivation response (Minamishima et al., 2009), was not affected by the absence of Stat3 (Fig. S3E). Interestingly, PLA analysis performed on ESCs treated with CoCl2 confirms the interaction between STAT3 and HIF1α (Fig. S3F). These results suggest that the formation of a nuclear STAT3-HIF1α complex is associated with full activation of hypoxia responsive genes.
Hif1α transcriptional activity is not induced in vivo when Stat3 is knocked-out or inhibited.
Zebrafish is an in vivo model in which the pathophysiological roles of HIF1α and JAK/STAT signalling have been extensively studied (Santhakumar et al., 2012; Gerri et al., 2017; Gerri et al., 2018; Vettori et al., 2017; Marchi et al., 2020; Liu et al., 2017; Peron et al., 2020; Peron et al., 2021; Dinarello et al., 2022; Risato et al., 2022). Thus, it appeared as a valid platform to analyze the STAT3-HIF1α crosstalk and the physiological implications of this interplay between transcription factors. We used the Tg(4xHRE-TATA:mCherry,cmlc2:EGFP)ia22 hypoxia reporter zebrafish line (herein called HRE:mCherry), in which the mCherry red fluorescent protein is expressed in all tissues experiencing low oxygen tensions or pseudohypoxic conditions (Vettori et al., 2017). To investigate the requirement of Stat3 in the transcriptional response to hypoxia, we combined hypoxic and pseudohypoxic treatments with chemical or genetic inhibition of Stat3. To inhibit Stat3 signaling we used AG490, which blocks the Jak-mediated Y705 phosphorylation of Stat3 and abrogates its nuclear transcriptional activity (Park et al., 2014; Peron et al., 2020; Peron et al., 2021). Larvae treated with this compound were also characterized by a reduction of stat3 gene expression (Fig. S4A). To induce hypoxia in zebrafish larvae, we incubated the animals for 3 days with 5% oxygen tension, while pseudohypoxia was forced by using either dimethyloxalylglycine (DMOG), an inhibitor of Phd-dependent degradation of Hif1α (Mole et al., 2003); CoCl2 or dexamethasone (Dex), a synthetic glucocorticoid that contributes to the stabilization of Hif1α by degrading Vhl (Vettori et al., 2017). Notably, while hypoxia and pseudohypoxia induced the fluorescence of HRE:mCherry zebrafish larvae, such induction was abolished by AG490, suggesting that, also in zebrafish, activated Stat3 is necessary for the correct induction of hypoxia-dependent transcription (Fig. 3A). Moreover, we tested the expression of vegfa and hk1 at different times of exposure to low oxygen tension. We could observe that hypoxia determines a boost of expression of both vegfa and hk1 after 8 hours of incubation in hypoxia; after 24 hours of incubation in hypoxic conditions, both transcripts return to the levels detected in normoxic larvae. Interestingly, vegfa expression increased after 72 hours of hypoxia, while hk1 expression was upregulated after 48 hours of hypoxia (Fig. S4B,C). AG490 abrogated this hypoxia-dependent dynamical expression of vegfa and hk1 (Fig. S4B,C)., confirming the results obtained with the HRE:mCherry reporter line and with ESCs. STAT3 transcriptional activity has also been reported to be positively affected by the phosphorylation of Serine 727 (Levy et al., 2002; Zhang et al., 1995; Wen et al., 1995) and this post-translational modification is triggered by MEK/ERK pathway (Gough et al., 2013). As we recently demonstrated that the MEK/ERK inhibitor PD98059 blocks the pS727 activities of STAT3 (Peron et al., 2021), we sought to assess whether the inhibition of S727 phosphorylation can affect the responsiveness of larvae to low oxygen tensions. When HRE:mCherry reporter larvae are treated with PD98059 and incubated in low oxygen tensions, we could not observe significant differences in reporter fluorescence compared to hypoxic DMSO-treated larvae, suggesting that the chemical inhibition of S727 phosphorylation does not affect the transcriptional responsiveness of zebrafish to hypoxia (Fig. 3B). To further elucidate the role of Serine or Tyrosine phosphorylation in Stat3-dependent regulation of hypoxia transcription, we injected Stat3 mRNAs (that we recently used and validated in Peron et al., 2021) in double transgenic eggs obtained from the breeding between Tg(7xStat3-Hsv.Ul23:EGFP)ia28, herein called SBE:EGFP (Peron et al., 2020), and HRE:mCherry transgenic animals to monitor at the same time the Stat3- and the hypoxia-dependent transcription. Of note, wild type Stat3 mRNA injection upregulated mCherry fluorescence suggesting that the overexpression of Stat3 increases hypoxia-dependent transcription, while Stat3 Y705F mRNA (that encodes a STAT3 that cannot be phosphorylated in 705 position, as described in Minami et al., 1996) cannot induce the fluorescence of both reporters (Fig. S4D,E). On the other hand, surprisingly, the overexpression of Stat3 S727A (in which the substitution of the Serine with an Alanine blocks the 727 phosphorylation, as shown in Wen et al., 1997), induces the Stat3-dependent transcriptional activity as well as the hypoxia-related fluorescence (Fig. S4D,E). These results suggest that Tyrosine phosphorylation is necessary for the regulation of hypoxia by Stat3, while Serine is not.
Next, we crossed the stat3ia23 mutant zebrafish line (herein called stat3−/−) (Peron et al., 2020) with the HRE:mCherry reporter line. We treated the HRE:mCherry;stat3+/+, HRE:mCherry;stat3+/− and HRE:mCherry;stat3−/− sibling larvae with 5% oxygen tension and measured mCherry fluorescence. Consistent with previous results obtained in ESC cells and in embryos treated with AG490, the Hif1α-dependent reporter activity increased in hypoxic stat3+/+ larvae, while no significant increase of the reporter fluorescence was detected in hypoxic stat3+/−and stat3−/− larvae when compared with normoxic siblings (Fig. 3C). These results indicate that the transcriptional activation of HIF1α targets requires STAT3 also in vivo.
Nuclear crosstalk between Stat3 and Hif1α determines the regulation of hypoxia-dependent genes in zebrafish
Given that Stat3 is needed for the induction of normal Hif1α transcriptional activities and that hypoxia-induced mechanisms are impaired when Stat3 is inhibited or deleted, we wanted to determine by which mechanisms Stat3 regulates Hif1α transcriptional activity.
We first tested whether Stat3 inactivation would affect hif1α expression levels, its stabilization, or the levels of HIF1α regulators vhl and egln3; we observed no significant differences in their expression levels when comparing stat3+/+ and stat3−/− 6-dpf larvae (Fig. S5A-D). We conclude that, as already observed in ESCs (Fig. 2B), the lack of a functional Stat3 does not impair the expression and stabilization of Hif1α in zebrafish larvae.
Given that in ESCs STAT3 and HIF1α interact in the nucleus, we focussed our attention on the nuclear activity of HIF1α in vivo. For this purpose, we injected an mRNA that encodes a dominant active (DA) form of hif1αb bearing mutations at two prolines and one asparagine (P402A, P564G, N804A), hence preventing their hydroxylation by Phd and subsequent Vhl-dependent degradation (Elks et al., 2011; Elks et al., 2013). This construct activates by default nuclear targets, independently from any upstream degrading cue. We treated HRE:mCherry hypoxia reporter larvae injected with hif1αb DA mRNA at 3 dpf with AG490 and analysed the reporter fluorescence at 4 dpf. Notably, hif1αb DA mRNA determines an induction of the reporter activity that is slightly decreased (p = 0.05) in injected larvae treated with AG490 (Fig. 4A). We further investigated the endogenous genes induced by hif1αb DA mRNA by performing RT-qPCR analysis. The expression levels of Hif1α-dependent genes, such as vegfa and hk1 were induced by hif1αb DA mRNA, however, AG490 abrogated their upregulation (Fig. 4B,C). These results confirmed in vivo that activation of HIF1α nuclear targets rely on active STAT3, as observed in mouse ESCs.
An organ in which STAT3 plays a pivotal role is the intestine (Peron et al., 2020; Matthews et al., 2011; Pickert et al., 2009). As we recently demonstrated in zebrafish, Stat3 is fundamental for the correct folding of gut and is particularly expressed in intestinal stem cells (Peron et al., 2020). In particular, the Stat3 zebrafish reporter line Tg(7xStat3-Hsv.Ul23:EGFP)ia28 that we recently published is characterized by EGFP positive cells in the intestinal track which represent the intestinal stem cells of zebrafish.
We took advantage of this zebrafish model to ask whether HIF1α and STAT3 are active in the same cells (cell autonomously), or whether they co-operate non-cell-autonomously in distinct cells somehow interconnected. To do so, we sorted Stat3 responsive intestinal stem cells from adult intestines. Results of RT-qPCR revealed a strong and significant upregulation of Hif1α transcriptional activity in EGFP sorted cells. In particular, hif1αb transcript levels were 20-fold higher in EGFP-positive cells compared to EGFP-negative cells. The upregulation of the vegfa and hk1, involved in angiogenesis and glucose metabolism, as well as of the pure Hif1α target gene egln3, involved in the Hif1α degradation feedback loop (Minamishima et al., 2009), revealed a massive Hif1α transcriptional activity in the EGFP-positive Stat3-responsive cells (Fig. 4D). flna, a gene that is neither affected by Stat3 nor by Hif1α was used in this experiment as a negative control. We conclude that in the context of a complex tissue like the intestine, HIF1α direct targets are strongly expressed in cells with active STAT3, further indicating cooperative transcriptional activation by the two factors and the cell-autonomous regulation of the STAT3-HIF1α crosstalk.
Hypoxia-dependent processes are impaired in stat3 mutant zebrafish
Our results so far indicate a cooperative activation of STAT3 and HIF1α in the transcriptional response to hypoxia. We then asked whether such cooperation is important also for the physiological responses in an organism.
One of the most relevant processes induced by hypoxia is angiogenesis (Fraisl et al., 2009). To study how this process is affected by hypoxia in stat3 mutant zebrafish larvae, we decided to use the zebrafish endothelial cell reporter line Tg(Fli1:EGFP)y1 (Lawson and Weinstein, 2002). stat3+/+, stat3+/− and stat3−/− sibling larvae in Tg(Fli1:EGFP)y1 transgenic background were treated with 5% oxygen tension from 48 hpf to 54 hpf. After the treatment, we detected a significant increase of endothelial fluorescence in stat3+/+ hypoxic larvae when compared with untreated siblings (Fig. 5A), indicating that angiogenesis is induced by low oxygen tensions, hence confirming what already observed by Eyries et al. (2008). A similar induction of fluorescence was also detected in hypoxic heterozygous stat3+/− larvae when compared to normoxic heterozygous siblings, but no significant differences were detected between normoxic and hypoxic null stat3−/− larvae, indicating a pivotal role of Stat3 in this process (Fig. 5A).
Hypoxia plays a central role in erythropoiesis (Haase, 2010; Zhang et al., 2012; Solak et al., 2016; Kietzmann, 2020) and in zebrafish gata1 is a specific marker for erythropoiesis and erythrocytes (Lyons et al., 2002; Galloway et al., 2005; Bresciani et al., 2010; Quintana et al., 2014; Li et al., 2014; Lenard et al., 2016). Hence, we decided to use the Tg(gata1:dsRed)sd2 transgenic line (Galloway et al., 2005), in which erythroid cells display a strong red fluorescence, to see whether Stat3 has a role in hypoxia-induced erythropoiesis. Notably, 6-hour long hypoxia treatments determine a significant increase of fluorescence in stat3+/+. Similarly, an upregulation of fluorescence was detected also in hypoxic stat3+/− larvae. Of note, no significant differences were detected between normoxic and hypoxic stat3−/− larvae, demonstrating that Stat3 is involved in the hypoxia-induced erythropoiesis (Fig. 5B).
Since the hypoxia/HIF1 pathway has been linked to mobilization and polarization of macrophages (Lewis et al., 1999; Elks et al., 2013; Ke et al., 2019; Lewis and Elks, 2019; Sadiku and Walmsley, 2019), we used the Tg(LysC:dsRed)nz50 transgenic line (Hall et al., 2007), to focus our attention on them. First, to confirm that the migration of macrophages outside the aorta-gonad-mesonephros (AGM) region is triggered by hypoxia and relies on Hif1α, we injected 1-cell stage Tg(LysC:dsRed)nz50 transgenic embryos with a solution containing both morpholinos (MO) against hif1αa and hif1αb (as previously described in Gerri et al., 2017). 48-hpf controls and hif1α morphants were subsequently incubated in normoxia or hypoxia for 6 hours and the number of macrophages was counted. As reported in Fig. 5C and Fig. S6A, hypoxia determines an increase of the ratio between the number of cells observed in the trunk of larvae (named region A) and that in the AGM (named region B), suggesting that hypoxia determines the migration of macrophages away from AGM. Interestingly, we could not detect significant differences between normoxic and hypoxic morphants, demonstrating that Hif1α is required for this process. Subsequently, we sought to see whether Stat3 is also involved in this process. To do so, we treated 48-hpf stat3+/+, stat3+/− and stat3−/− larvae Tg(LysC:dsRed)nz50 transgenic background with 5% oxygen tension for 6 hours. Treated stat3+/+ and stat3+/− showed a significant increase of Area A/Area B ratio compared to untreated siblings, while hypoxia did not have any effect in mobilizing macrophages of stat3−/− larvae, suggesting that Stat3 is involved in this Hif1α-dependent process (Fig. 5D and Fig. S6B). Moreover, we analysed the level of expression of genes involved in macrophage activity, like mfap4, tek, and lcp1 (Walton et al., 2020; Melcher et al., 2008; Kell et al., 2018). mfap4 and tek appeared to be downregulated in their expression in stat3−/− when compared to stat3+/+ siblings (Fig. S6C), highlighting a role of Stat3 in the homeostasis of these cells. Importantly, the total number of dsRed-positive cells does not seem to be affected significantly by genetic ablation of stat3, indicating a specific effect on migration rather than survival or proliferation (Fig. S6B,C).