Hypoxia-inducible factor-2 stabilization is not sufficient to induce erythropoietin production in deeper medullary fibroblasts

Abstract

Under hypoxaemic conditions, cortical fibroblasts primarily produce erythropoietin (EPO). However, we have previously shown that most interstitial fibroblasts positive for platelet-derived growth factor receptor β (PDGFR-β) in all kidney zones are also able to produce EPO. Therefore, we wondered if either the physiological stimuli might not be sufficient to stabilize the hypoxia-inducible factor (HIF)-2 in medullary fibroblasts or if different expression patterns or functions of the HIF-regulating prolyl-4-hydroxylases (PHD) 2 and 3 might explain the restrictive EPO cell recruitment. This study shows that although HIF-2 can be clearly stabilized in deeper medullary fibroblasts by pharmacological PHD-inhibition, this is not sufficient to induce EPO in these cells. In contrast, genetic stabilization of HIF-2 by cell-specific deletion of either PHD2 or PHD2 and PHD3 in mice resulted in EPO production in all kidney zones. EPO induction in PHD2/3-KO mice was twice as high as in PHD2-KOs. PHD3 deletion slightly increased basal EPO expression. Accordingly, in contrast to PHD2, PHD3 expression was only detected in a subset of interstitial fibroblasts, without zonal accumulation or hypoxaemic upregulation. Exposure of PHD3-deficient mice to a hypoxaemic stimulus resulted in significantly higher EPO levels compared to controls, with EPO induction restricted to the cortex. Overall, our data suggest the existence of additional regulatory mechanisms beyond the HIF-2 signalling pathway that control EPO expression in deeper medullary fibroblasts. Furthermore, they identify PHD3 as an attenuating factor that delays EPO induction in a subset of cortical PDGFR-β⁺ cells, but its expression pattern is not the determining factor responsible for the cortical restriction of EPO. KEY POINTS: Pharmacological prolyl-4-hydroxylase (PHD) inhibition activates hypoxia-inducible factor (HIF)-2 signalling in interstitial fibroblasts from all renal zones. HIF-2 stabilization is not sufficient to induce erythropoietin (EPO) expression in deeper medullary fibroblasts, although they are in principle capable of producing EPO. There are two subsets of interstitial fibroblasts, PHD2⁺ and PHD2/PHD3⁺ fibroblasts, that are evenly distributed throughout the kidneys, thus also not determining the restrictive cortical induction of EPO under hypoxaemic and pharmacological conditions. Solely PHD2⁺ fibroblasts are the predominant EPO producers in the renal cortex, while PHD3 is an attenuating factor that delays EPO induction in PHD2/PHD3⁺ fibroblasts. HIF2-induced upregulation of PHD isoforms is not detectable in interstitial fibroblasts. Overall, our data suggest the existence of additional regulatory mechanisms, in addition to the HIF-2 signalling pathway, that control EPO expression in deeper medullary fibroblasts.

Keywords: PHD inhibition; erythropoietin; hypoxia signalling; interstitial fibroblasts; prolyl‐4‐hydroxylases; renal endocrine function.

Figure 1. Cellular HIF‐2α mRNA expression levels as well as renal HIF‐1α and HIF‐2α mRNA expression under different conditions

A, zonal details showing the co‐expression of HIF‐2α (red) and PDGFR‐β (green) using RNAscope. The upper row shows details from the cortex, outer medulla (OM), and inner medulla (IM) of a kidney section of a wild‐type (WT) mouse under normoxic conditions. The lower row shows the respective details from a wild‐type mouse treated 8 times with PHDi. Arrows highlight exemplarily some HIF‐2α/PDGFR‐β coexpressing fibroblasts. The arrowheads indicate some endothelial cells, and the dotted line indicates a tubular segment. Scoring the HIF‐2α expression level per PDGFR‐β⁺ cell revealed no differences between kidney zones or different conditions. Clear HIF‐2α signals were observed in endothelial cells, while only weak HIF‐2α expression was detectable in tubular cells. Nuclei were counterstained with DAPI (grey). Scale bars: 20 µm. B, automated co‐expression analysis revealed no significant differences in HIF‐2α/PDGFR‐β co‐expression among kidney zones within a given condition, nor between any given zone and its corresponding zone in the other analysed conditions. Statistical significance was determined using one‐way ANOVA with Tukey's multiple comparisons test. Respective P‐values are provided in the tables below the graphs. Values are means ± SD of n ≥ 5 per group. C, renal HIF‐1α and HIF‐2α mRNA abundances under normoxic, anaemic and PHDi (8×)‐treated conditions and in PDGFR‐β^CreERT2/+ Vhl^ff mice. There was no significant difference between the analysed conditions. Statistical significance was determined using one‐way ANOVA with Dunnett's multiple comparisons test. P‐values are stated above the lines. Values are means ± SD of n ≥ 7 per group.

Figure 2. HIF‐2 protein stabilization on kidney sections under different conditions

HIF‐2 stabilization was visualized using immunohistochemical staining for HIF‐2α (brown nuclear signal) on kidney sections of wild‐type mice under different conditions as well as PDGFR‐β^CreERT2/+ Vhl^ff mice. For each condition/genotype, details from the cortex, outer medulla and inner medulla are shown. The arrows exemplarily highlight some HIF‐2 positive interstitial fibroblasts. Under normoxic conditions only very few HIF‐2 positive nuclei could be detected in the cortex (upper row). Under anaemic conditions clusters of HIF‐2 positive fibroblasts could be detected in the cortex and in the outer zone of the outer medulla but not in the deeper medullary regions (second row). After 8× PHDi treatment most interstitial fibroblasts were positive for HIF‐2 throughout all kidney zones (third row). On sections of PDGFR‐β^CreERT2/+ Vhl^ff mice HIF‐2 positive interstitial fibroblasts could also be detected in all kidney zones (lower row). Scale bars: 50 µm.

Figure 3. Renal induction of different HIF‐2 target genes under different conditions

A, RNAscope for adrenomedullin (ADM, white) or RGS4 (white) on kidney sections of normoxic, anaemic, and 8× PHDi‐treated wild‐type mice compared to kidney sections of PDGFR‐β^CreERT2/+ Vhl^ff mice. Under normoxic conditions ADM could only be detected sporadically in interstitial fibroblasts (arrows) and at a low level in some tubular segments. RGS4 could be detected in some interstitial fibroblasts along the corticomedullary border (arrows) and along vasa recta. ADM and RGS4 were upregulated in interstitial fibroblasts along the corticomedullary border in anaemic mice. After 8× PHDi treatment and in PDGFR‐β^CreERT2/+ Vhl^ff mice, the expression patterns of ADM and RGS4 were quite similar. They could be detected in most fibroblasts across all kidney zones. In addition, ADM was upregulated in tubular segments of PHDi‐treated mice. In contrast, EPO (red) could only be detected in interstitial fibroblasts of the cortex and the outer zone of the outer medulla on kidney sections of PHDi‐treated mice, while EPO expression was observed in most fibroblasts across all kidney zones in PDGFR‐β^CreERT2/+ Vhl^ff mice. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm. B, renal mRNA abundances of ADM, RGS4 as well as EPO and plasma EPO concentrations of mice under the analysed conditions/genotypes. Statistical significance was determined using one‐way ANOVA with Dunnett's multiple comparisons test. P values are stated above the respective lines. Values are means ± SD of n ≥ 7 per group.

Figure 4. Recruitment pattern of renal EPO mRNA expression as well as the renal expression levels of EPO mRNA and other HIF target genes after serial treatment of wild‐type mice with the PHDi roxadustat

A–D, distribution of EPO‐producing cells on kidney sections of wild‐type mice treated with 1× (A), 3× (B), 6× (C), or 8× (D) PHDi roxadustat was analysed using RNAscope for EPO mRNA (white). While 1× PHDi led to EPO induction along the cortico‐medullary border (A), EPO induction was increased in the outer stripe of the outer medulla and spread outward towards the cortex surface after 3× PHDi (B). With 6× and 8× PHDi the number of EPO producing cells further increased throughout the cortex. Dotted turquoise lines indicate the zonal borders. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm. E, renal mRNA abundances of different HIF‐target genes – EPO, regulator of G protein signalling 4 (RGS4), adrenomedullin (ADM) and PHD3 – of wild‐type mice treated with an increasing number of PHDi doses. Serial treatment with 10× roxadustat did not result in a further significant increase in target induction compared to an 8× series, as determined using one‐way ANOVA with Tukey's multiple comparisons test. P‐values are stated above the lines or in the tables below graphs. Values are means ± SD of n ≥ 4 per genotype.

Figure 5. Recruitment pattern of renal ADM expression as well as HIF‐2 protein stabilization on kidney sections of wild‐type mice 3× treated with the PHDi roxadustat

A and B, distribution of ADM‐producing cells (red dots) on kidney sections of normoxic wild‐type mice and wild‐type mice 3× treated with the PHDi roxadustat. Under normoxic conditions, ADM was only detected in few cortical interstitial fibroblasts (A). After 3 PHDi treatments, ADM⁺ interstitial fibroblasts could be detected in a regular pattern throughout all kidney zones (B). Dotted white lines indicate the zonal borders. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm. C, HIF‐2 stabilization was visualized using immunohistochemical staining for HIF‐2α (brown nuclear signal) on kidney sections of wild‐type mice after 3 PHDi treatments. The arrows highlight HIF‐2 positive interstitial fibroblasts. After 3× PHDi treatment HIF‐2 positive interstitial fibroblasts could be detected throughout all kidney zones – cortex, outer medulla (OM), and inner medulla (IM). Scale bars: 20 µm.

Figure 6. Extent of renal EPO induction in mice with different PHD deletions or Vhl deletion in PDGFR‐β⁺ cells

EPO mRNA abundances (A) and plasma EPO concentrations (B). While PHD3 deletion alone had no effect on the EPO expression, PHD2 deletion led to a significant increase in renal EPO mRNA expression levels and in parallel elevated plasma EPO concentrations of about 54,000 pg/ml compared to about 210 pg/ml in control animals. Combined deletion of PHD2 and PHD3 further increased the renal EPO mRNA abundance and plasma EPO concentrations to levels similar to mice with PDGFR‐β‐cell‐specific Vhl deletion. Statistical significance was determined using one‐way ANOVA with Tukey's correction. P‐values are stated above the lines. Values are means ± SD of n ≥ 5 per genotype.

Figure 7. Extent and distribution of renal EPO induction in mice with different PHD deletions in PDGFR‐β⁺ cells

A and B, overviews of transverse kidney sections of control and PDGFR‐β^CreERT2/+ PHD3^ff mice showing the expression pattern of EPO producing cells (red dots). White dotted lines indicate zonal borders. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm. C and D, overviews of transverse kidney sections showing the expression pattern of EPO producing cells (red dots) after PHD2 (C) or PHD2/PHD3 (D) deletion. Deletion of both, PHD2 or PHD2/PHD3 resulted in a strong induction of EPO throughout the kidneys. However, a direct comparison showed that the combined deletion of PHD2 and PHD3 resulted in approximately twice as many EPO⁺ cells than PHD2 deletion alone. Dotted lines indicate zonal borders. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm. E and F, details from the kidney cortex of a co‐RNAscope for EPO (red) and PDGFR‐β (green). After PHD2 deletion, EPO expression was induced in about 47% of PDGFR‐β⁺ interstitial fibroblasts (E). In contrast, codeletion of PHD2/PHD3 led to EPO production in about 75% of interstitial fibroblasts (F). Circles indicate glomeruli. Nuclei were counterstained with DAPI (grey). Scale bars: 50 µm.

Figure 8. Extent and distribution of renal EPO induction in PHD3‐KO mice and controls under basal, anaemic and PHDi‐treated (3×) conditions

A–C, EPO mRNA abundance, plasma EPO concentrations as well as EPO cell numbers. The induction of EPO due to anaemia or 3× PHDi treatment was higher in PDGFR‐β^CreERT2/+ PHD3^ff mice compared to controls. Statistical significance between control and PHD3‐KO mice for each condition was determined using unpaired one‐tailed t test. P‐values are stated above the lines. Values are means ± SD of n ≥ 4 per genotype. D and E, overviews showing the expression pattern of EPO producing cells (red) on transverse kidney sections of control and PDGFR‐β^CreERT2/+ PHD3^ff mice under anaemic conditions. Both animals had haematocrit values of approximately 25%. Recruitment of additional EPO cells followed the classic pattern from the corticomedullary border into the cortex. However, the total number of EPO⁺ cells per kidney section was higher in the kidneys of PHD3‐KO animals. White dotted lines indicate zonal borders. Nuclei were counterstained with DAPI (grey). Scale bars: 500 µm.

Figure 9. Localization and regulation of PHD2 and PHD3 under basal, anaemic and PHDi‐treated conditions as well as in PDGFR‐β^CreERT2/+ Vhl^ff mice

A and B, details showing the colocalization of PHD2 (red) with PDGFR‐β (green) and the tubular marker cadherin 16 (light blue) under normoxic conditions (A) and after 8× PHDi treatment (B) using RNAscope. PHD2 could be detected in all PDGFR‐β⁺ fibroblasts (arrows) as well as tubular cells under normoxia. After treatment with the PHD‐inhibitor roxadustat, PHD2 expression was elevated in some tubular cells but not in interstitial fibroblasts (arrows). Asterisks highlight tubular cells with increased PHD2 expression compared to the controls. Nuclei were counterstained with DAPI (grey). Scale bars: 20 µm. C–F, RNAscope details showing the colocalization of PHD3 (red) with PDGFR‐β (green) and the tubular marker cadherin 16 (light blue) on kidney sections of wild‐type mice under normoxic (C), anaemic (D) or 8× PHDi‐treated (E) conditions and of PDGFR‐β^CreERT2/+ Vhl^ff (F) mice. Under normoxic conditions PHD3 was only expressed in about 35% of interstitial PDGFR‐β⁺ fibroblasts (arrows). In addition, PHD3 could be detected in some tubular segments. Under anaemic conditions and after PHDi treatment, PHD3 was only upregulated in some tubular segments, highlighted with asterisks. There was no upregulation of PHD3 in interstitial fibroblasts. Upregulation of PHD3 in interstitial fibroblasts could only be detected in Vhl‐KO mice (arrowheads). Nuclei were counterstained with DAPI (grey). Scale bars: 20 µm. G–I, RNAscope details showing the colocalization of PHD3 (red) with EPO (blue) and the tubular marker cadherin 16 (light blue) on kidney sections of wild‐type mice under anaemic (G) or 3× PHDi‐treated (H) conditions and of PDGFR‐β^CreERT2/+ Vhl^ff (I) mice. In anaemic mice and mice treated with three administrations of PHDi, EPO was almost exclusively detected in PHD3 negative fibroblasts. In Vhl‐KO mice a strong coexpression of PHD3 and EPO could be observed (arrowheads). Nuclei were counterstained with DAPI (grey). Scale bars: 20 µm. J, zonal PHD2/PDGFR‐β as well as PHD3/PDGFR‐β coexpression was determined on kidney sections of wild‐type mice under normoxic conditions with the Zeiss Intellesis software. Analysis showed that in each kidney zone about 90% of interstitial fibroblasts were positive for PHD2, while only about 35% of the interstitial PDGFR‐β⁺ fibroblasts coexpressed PHD3. Statistical significance was determined using one‐way ANOVA with Tukey's multiple comparisons test. P‐values are stated above the respective lines. Values are means ± SD of n = 5 animals. K, zonal PHD2 or PHD3 mRNA expression level per PDGFR‐β⁺ fibroblast was determined on kidney sections of wild‐type mice under normoxic conditions by manual scoring. All PHD2/PDGFR‐β⁺ or PHD3/PDGFR‐β⁺ fibroblasts per renal zone respectively constitute 100%. Score 1: 1–5 PHD2 or PHD3 signal dots per PDGFR‐β⁺ fibroblast; score 2: 5–10 signal dots; score 3: >10 dots without dot clusters. The PHD2 or PHD3 expression levels per PDGFR‐β⁺ fibroblast were quite similar between the different kidney zones. Values are means ± SD of n = 5 animals. L, renal mRNA abundances of PHD2 and PHD3 under different conditions. Statistical significance was determined using one‐way ANOVA with Dunnett's multiple comparisons test. P‐values are stated above the respective lines. Values are means ± SD of n ≥ 7 per group.

Figure 10. Renal mRNA coexpression of the transcription factors Tcf21 (red) or Cebpd (CCAAT/enhancer‐binding protein delta, red), EPO (blue) and PDGFR‐β (green) under different conditions

For each analysed condition a representative detail from the cortex and from the deeper medulla are shown. Nuclei were counterstained with DAPI (grey). Scale bars: 20 µm. A, under normoxic conditions Tcf21 could be detected in all interstitial fibroblasts. There was no difference in the expression level per cell between the different zones. Under anaemic conditions Tcf21 was significantly downregulated in EPO⁺ fibroblasts. Tcf21 expression in the deep medulla was unchanged compared to normoxic conditions. Under PHDi‐treated conditions, Tcf21 expression was strongly downregulated in interstitial (EPO⁺) fibroblasts of the cortex and outer zone of the outer medulla. In the deeper medulla Tcf21 expression in EPO negative fibroblasts was also downregulated to levels comparable to cortical fibroblasts. In PDGFR‐β^CreERT2/+ Vhl^ff animals Tcf21 was downregulated in all EPO⁺ fibroblasts across all kidney zones. Arrows highlight some of the interstitial (EPO⁺) fibroblasts, in which Tcf21 is downregulated. B, under normoxic conditions Cebpd was only sporadically detected in some interstitial fibroblasts along the cortico‐medullary border. Moreover, Cebpd expression could be detected in the S3 segment of proximal tubules. Under anaemic conditions Cebpd was detectable in EPO⁺ fibroblasts and some additional interstitial fibroblasts in the cortex and even in the deeper medulla. Under PHDi‐treated conditions, clear Cebpd expression was found in most interstitial (EPO⁺) fibroblasts of the cortex and outer zone of the outer medulla. In the deeper medulla Cebpd expression in EPO negative fibroblasts was also upregulated to levels comparable to cortical fibroblasts. PDGFR‐β^CreERT2/+ Vhl^ff animals showed similar Cebpd expression in interstitial fibroblasts across all kidney zones as PHDi‐treated mice. Arrows highlight some of the interstitial Cebpd⁺ (EPO⁺) fibroblasts.