Loss of Phd1 but not Phd2 or Phd3 selectively protects mice against chronic colitis. Several studies have shown that loss of Phd1 is protective against acute colitis (15, 16). However, the importance of PHD1–3 in colitis-associated colon carcinogenesis is not yet known. To assess the putative functions of PHD1–3 in CAC, we employed the AOM/DSS model (26) in Phd1-, Phd2-, or Phd3-deficient (Phd1–/–, Phd2+/–, and Phd3–/–) mice and WT controls; after AOM-induced epithelial mutagenesis, mice underwent repeated cycles of DSS exposure followed by a recovery period (Figure 1A). In this chronic colitis model, Phd1–/– mice showed significantly attenuated colitis activity compared with WT control animals as assessed by the BW change and disease activity index (DAI) (Figure 1B, and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.153337DS1). Phd2+/– mice showed a significantly higher BW compared with WT control mice. However, this change was less pronounced compared with Phd1–/– mice and not reflected by the more comprehensive DAI (Figure 1B, and Supplemental Figure 1A). Phd3–/– mice showed an unchanged BW and DAI compared with WT control mice (Figure 1B, and Supplemental Figure 1A). In keeping with this, colitis-induced shortening of the colon was reduced in Phd1–/– but not Phd2+/– or Phd3–/– mice compared with WT controls (Figure 1C). Histological assessment of colonic mucosa after chronic DSS-induced colitis by H&E staining (parameters outlined in Supplemental Table 1) revealed decreased histological injury of Phd1–/– mice compared with WT control mice (Figure 1D). To further characterize the extent of inflammation, we assessed the expression of several proinflammatory cytokine and chemokine transcripts in colonic mucosa samples by semiquantitative real-time PCR (qRT-PCR). However, expression of none of the proinflammatory cytokines and chemokines was significantly changed in Phd1–/–, Phd2+/–, or Phd3–/– mice compared with their WT counterparts (Supplemental Figure 1B).

Loss of Phd1 but not Phd2 or Phd3 selectively protects mice against chronic colitis. (A) Model of chronic colitis and colitis-associated tumorigenesis induced by AOM and repeated cycles of DSS in WT control, Phd1–/–, Phd2+/–, and Phd3–/– mice. (B) BW change relative to baseline from WT (n = 10), Phd1–/– (n = 6), Phd2+/– (n = 5), and Phd3–/– (n = 9) mice over the course of AOM/DSS treatment. BW was measured every other day. (C) Colon length of AOM/DSS-treated WT (n = 10), Phd1–/– (n = 7), Phd2+/– (n = 5), and Phd3–/– (n = 10) mice after termination of the experiment at day 84. (D) Histological scoring of mucosal damage (top) as previously described by Katakura et al. (62) and representative H&E staining (bottom) of colons from WT (n = 7), Phd1–/– (n = 6), Phd2+/– (n = 5), and Phd3–/– (n = 8) mice after termination of the experiment on day 84. Scale bar: 100 μm. Statistical significance was calculated using 2-way ANOVA (B) or 1-way ANOVA with Dunnett’s multiple comparisons test in C and D. *P < 0.05, ****P < 0.0001.

Taken together, these results demonstrate that loss of Phd1, but not Phd2 or Phd3, is protective against DSS-induced chronic colitis.

Loss of Phd1 diminishes, Phd2 haplodeficiency aggravates, and Phd3 deficiency does not affect colitis-associated tumor growth. To evaluate the effect of Phd1, 2, or 3 deficiency on colitis-associated tumorigenesis, we analyzed tumor formation and size after AOM/DSS-induced CAC. Consistent with the protective effects of Phd1 deficiency against chronic colitis, Phd1–/– mice displayed a significantly reduced tumor number and size compared with WT controls (Figure 2, A and B, and Supplemental Figure 2A). Strikingly, although the number of tumors in Phd2+/– mice was unchanged compared with control mice, the tumors in Phd2+/– animals were significantly larger (Figure 2, A and B, and Supplemental Figure 2A). In Phd3–/– mice, tumor number and size did not differ from WT mice (Figure 2, A and B, and Supplemental Figure 2A). This indicates that the PHD isoenzymes PHD1–3 each has a distinct impact on tumor formation and growth.

Loss of Phd1 diminishes, Phd2 haplodeficiency aggravates, and Phd3 deficiency does not alter colitis-associated tumor growth. (A) Representative macroscopic images (top) and macroscopic quantification of AOM/DSS-induced tumors (bottom). Number of tumors per mouse (left; WT: n = 9; Phd1-/-: n = 9; Phd2+/-: n = 11; and Phd3-/-: n = 11 mice) and size of individual tumors (right; WT: n = 80; Phd1-/-: n = 12; Phd2+/-: n = 54; and Phd3-/-: n = 67 tumors). (B) H&E staining of colons from WT control, Phd1–/–, Phd2+/–, and Phd3–/– mice. Arrows indicate colitis-associated tumors. Scale bar: 2 mm. (C) Quantification of epithelial PCNA immunostaining in AOM/DSS-induced tumors (WT: n = 17; Phd1-/-: n = 13; Phd2+/-: n = 14; and Phd3-/-: n = 15 tumors) and representative histological images (right). Scale bar: 25 μm. (D) Quantification of epithelial CC3 immunostaining in AOM/DSS-induced tumors (WT: n = 19; Phd1–/–: n = 13; Phd2+/–: n = 32; and Phd3–/–: n = 20 tumors) and representative histological images (right). Scale bar: 25 μm. (E) Model of sporadic colorectal carcinogenesis induced by repeated injections of AOM (6xAOM) in WT control, Phd1–/–, Phd2+/–, and Phd3–/– mice. (F) Macroscopic quantification of 6xAOM-induced tumors. Number of tumors per mouse (left; WT: n = 10; Phd1–/–: n = 9; Phd2+/–: n = 8; and Phd3–/–: n = 10 mice) and size of individual tumors (right; WT: n = 5; Phd1–/–: n = 4; Phd2+/–: n = 11; and Phd3–/–: n = 11 tumors). Statistical significance was calculated using 1-way ANOVA with Dunnett’s multiple comparisons test in A, C, D, and F. *P < 0.05, **P < 0.01, ****P < 0.0001.

To further investigate tumor cell proliferation and apoptosis in these tumors, we performed IHC of proliferating cell nuclear antigen (PCNA) and cleaved caspase-3 (CC3). Remarkably, this revealed that, while tumor proliferation was unchanged in Phd1–/– and Phd3–/– tumors compared with control animals, Phd2+/– tumors proliferated significantly more (Figure 2C). Cell apoptosis in the tumors was unchanged among all experimental groups (Figure 2D). Collectively, these results demonstrate that loss of Phd1 diminishes CAC growth, whereas Phd2 haplodeficiency increases tumor proliferation and, thus, colitis-associated tumor growth. Loss of Phd3 did not result in any changes in tumor burden in the AOM/DSS model.

To interrogate whether intestinal inflammation is a prerequisite for these effects, we used an inflammation-independent, sporadic CRC tumor model comprising weekly injections of AOM, a potent carcinogen (27), administered for 6 weeks (6xAOM) (Figure 2E). Intriguingly, after 140 days, tumor number and size were equal in Phd1–/–, Phd2+/–, Phd3–/–, and WT control mice (Figure 2F), indicating that in the absence of underlying intestinal inflammation, the loss of any of the PHD isoenzymes does not result in differences concerning tumor number or size of sporadic CRC. Taken together, this demonstrates that intestinal inflammation is required for PHD-dependent tumor growth in CAC.

The activity of the oncogenic STAT3 and ERK1/2 signaling pathways is increased in colitis-associated Phd2+/– tumors. We next sought a possible molecular mechanism promoting tumorigenesis in colitis-associated Phd2+/– tumors. Since STAT3, ERK1/2, and WNT/β-catenin are key oncogenic signaling pathways in CRC (28–30), we examined phosphorylation of STAT3 and ERK1/2 as well as nuclear localization of β-catenin in the tumor cell compartment of Phd2+/– and WT control tumors by IHC. Strikingly, Phd2 haplodeficiency significantly increased STAT3 (Figure 3A) and ERK1/2 phosphorylation (Figure 3B) in epithelial cells of AOM/DSS-induced colon tumors. In contrast, there was no difference in nuclear β-catenin expression in epithelial cells of Phd2+/– tumors compared with WT controls (Supplemental Figure 3A). The augmented STAT3 and ERK1/2 phosphorylation was also validated by IB (Figure 3C). Of note, there was no difference in STAT3 phosphorylation in epithelial cells of Phd1–/– tumors compared with WT control tumors as assessed by IHC (Supplemental Figure 3B).

The activity of the oncogenic STAT3 and ERK1/2 signaling pathways is increased in colitis-associated Phd2+/– tumors. (A) Quantification of epithelial nuclear phosphorylated (p-) p-STAT3Y705 immunostaining in WT (n = 19) and Phd2+/– (n = 20) tumors and representative histological images (right). Scale bar: 100 μm. (B) Quantification of epithelial p-ERK1/2 immunostaining in WT (n = 30) and Phd2+/– (n = 37) tumors and representative histological images (right). Scale bar: 100 μm. (C) IB of STAT3, p-STAT3Y705, ERK1/2, and p-ERK1/2 in size- and location-matched WT (n = 4) and Phd2+/– (n = 4) tumors. qRT-PCR analysis of EGFR ligand Ereg (D), STAT3 and ERK1/2 target genes (E), and IL-6 and IL-11 (F) in WT (n = 16) and Phd2+/– (n = 16) tumors. Statistical significance was calculated using 1-way ANOVA with Dunnett’s multiple comparisons test in A and B or 2-tailed Student’s t test in D–F. *P < 0.05, **P < 0.01, ***P < 0.001.

Since the STAT3 and ERK1/2 signaling pathways are downstream of EGFR, we quantified the transcript expression of Egfr and all 7 known EGFR ligands — Ereg, Areg, Egf, Hbegf, Tgfa, Epgn, and Btc — in colitis-associated tumors from Phd2+/– and WT control mice by qRT-PCR. Intriguingly, while Egfr transcript expression was not significantly altered in Phd2+/– tumors compared with WT control tumors (Supplemental Figure 3C), Ereg was the only EGFR ligand that was significantly upregulated in Phd2+/– tumors compared with control tumors (Figure 3D and Supplemental Figure 3D). Further validating this, we reanalyzed a publicly available high-density microarray data set that includes transcriptomes from size- and location-matched AOM/DSS-induced and sporadic ApcMin/+ tumors (31). This verified that — in contrast to inflammation-independent ApcMin/+ tumors — Ereg, but none of the other EGFR ligands, was significantly upregulated in inflammation-associated tumors compared with the normal control samples (Supplemental Figure 3E). Taken together, this suggests that EREG signaling at least in part contributes to the activation of the oncogenic STAT3 and ERK1/2 signaling pathways in Phd2+/– tumor cells in CAC.

To further assess activation of the STAT3 and ERK1/2 signaling pathways, we analyzed the mRNA transcript expression of the target genes Myc, baculoviral IAP repeat-containing 5 (Birc5), and BCL2-like 1 (Bcl2l1). Consistent with increased STAT3 and ERK1/2 phosphorylation, transcript expression of these genes was significantly augmented in Phd2+/– tumors compared with WT control tumors (Figure 3E). Moreover, expression of the proinflammatory cytokines IL-6 and IL-11, which are key protumorigenic mediators in the AOM/DSS model and can signal both via STAT3 and ERK1/2 (28, 32, 33), was increased in Phd2+/– tumors compared with WT control tumors (Figure 3F).

Collectively, these results suggest that the enhanced colitis-associated tumorigenesis caused by Phd2 haplodeficiency is mediated, at least in part, by activation of the STAT3 and ERK1/2 signaling pathways in tumor cells through EREG.

The number of TAMs in colitis-associated Phd2+/– tumors is increased. Since immune cells, and specifically myeloid cells such as macrophages, are key determinants of tumor growth in the AOM/DSS model (34), we comprehensively profiled the immune cell landscape of AOM/DSS-induced tumors of Phd2+/– and WT controls using flow cytometry (Supplemental Figure 4A, and Supplemental Figure 5A). This revealed a significantly increased number of TAMs in AOM/DSS-induced tumors of Phd2+/– mice compared with WT controls (Figure 4A). Importantly, no other differences in the immune cell composition of Phd2+/– and WT tumors were detectable (Figure 4, A and B). To further validate these results, we performed IHC and immunofluorescence staining of F4/80-positive macrophages, CD11c-positive DCs, and CD3-positive T cells in AOM/DSS tumors. This supported an increased number of macrophages in tumors of Phd2+/– mice compared with WT controls (Figure 4C). Consistent with the flow cytometry results, staining of CD11c-positive DCs and CD3-positive T cells showed no differences among the animals (Figure 4, D and E).

The number of TAMs in colitis-associated Phd2+/– tumors is increased. Flow cytometry analysis of (A) myeloid and (B) lymphoid cells in tumors from Phd2+/– (n = 6) and WT (n = 6) control mice. (C) Quantification of F4/80 immunostaining in WT (n = 30), Phd1–/– (n = 18), and Phd2+/– (n = 44) tumors and representative histological images (right). Scale bar: 50 μm. (D) Quantification of CD11c immunofluorescence staining in WT (n = 7), Phd1–/– (n = 7), and Phd2+/– (n = 7) tumors and representative histological images (right). Scale bar: 50 μm. (E) Quantification of CD3 immunostaining in WT (n = 19), Phd1–/– (n = 8), and Phd2+/– (n = 26) tumors and representative histological images (right). Scale bar: 50 μm. Statistical significance was calculated using 2-tailed Student’s t test in A and B or 1-way ANOVA with Dunnett’s multiple comparisons test in C–E. *P < 0.05. HPF, high-power field.

To characterize the functional activation state of the TAMs, we assessed the expression of M1 (CD80, CD86, and CCR7) and M2 (CD163 and CD206) polarization markers within the TAM population of Phd2+/– and WT tumors by flow cytometry. Interestingly, this did not reveal significant changes of macrophage polarization in Phd2-deficient macrophages (Supplemental Figure 5B).

Taken together, this indicates that Phd2 deficiency is associated with an increased number of TAMs in CAC tumors, which could contribute to the increased tumor burden observed in Phd2+/– mice compared with WT control animals.

Phd2-deficient BMDMs stimulate tumor proliferation and show increased Ereg expression in vitro. After demonstrating that the presence of TAMs is increased and the STAT3 and ERK1/2 pathways are activated in Phd2+/– tumors in CAC, we next set out to determine how Phd2 haplodeficiency affects the functionality of macrophages using an established ex vivo model (35). For this, we performed a qRT-PCR analysis of Phd2+/– and WT control BMDMs, both unstimulated and upon proinflammatory stimulation, with LPS, TNF-α, or IL-4. Strikingly, the transcript expression of Ereg was significantly increased 2-fold in Phd2+/– BMDMs stimulated with LPS or TNF-α compared with WT control BMDMs (Figure 5A), suggesting that TAMs are at least 1 source of the increased Ereg expression observed in Phd2+/– tumors. Consistent with this, interrogation of Ereg expression in 2 large-scale single-cell RNA-Seq (scRNA-Seq) data sets of human CRC and UC (Broad Institute) (36, 37) verified that Ereg was also expressed in human macrophages and monocytes (Figure 5, B and C), underscoring the importance of EREG in these cells for CAC.

Phd2-deficient BMDMs stimulate tumor proliferation and show increased Ereg expression in vitro. (A) qRT-PCR analysis of Ereg in WT and Phd2+/– BMDMs upon stimulation with control, LPS (100 ng/mL), TNF-α (20 ng/mL), or IL-4 (20 ng/mL) for 24 hours. Dots represent biological replicates. Data represent 4 independent experiments. (B) t-Distributed stochastic neighbor embedding (tSNE) plots showing analysis of Ereg expression in myeloid cells from human CRC samples. Analysis was performed with publicly available scRNA-Seq data (36). (C) tSNE plots showing analysis of Ereg expression in immune cells from human UC mucosa samples. Analysis was performed with publicly available scRNA-Seq data (37). (D) Crystal violet viability assay of murine CMT-93 rectal cancer cells after 48 hours of treatment with control (RPMI medium + 1% FCS) or conditioned media from BMDMs stimulated with control, LPS, TNF-α, or IL-4. Dots represent technical replicates. Data represent 2 independent experiments. Statistical significance was calculated using Student’s t test. *P < 0.05.

In subsequent studies, we assessed the impact of Phd2 deficiency of BMDMs on the viability of CRC tumor cells in vitro. For this, we treated murine CMT-93 rectal cancer cells with the supernatant of stimulated Phd2+/– or WT BMDMs and assessed their viability after 48 hours of treatment. Strikingly, supernatant of Phd2+/– BMDMs stimulated with LPS or IL-4 significantly increased tumor cell viability of CMT-93 cells compared with treatment with supernatant of WT BMDMs stimulated with LPS or IL-4, indicating that Phd2+/– macrophages can promote tumor growth in vitro (Figure 5D). Taken together, this suggests that, in addition to their increased presence in AOM/DSS tumors, Phd2-deficient macrophages display protumorigenic features and increased Ereg expression in vitro, which implies a mechanistic link to the observed increase in oncogenic STAT3 and ERK1/2 signaling in Phd2+/– tumors in vivo.

Lineage-specific deletion of Phd2 in the hematopoietic but not the epithelial cell compartment aggravates colitis-associated tumor growth. To test the hypothesis that TAMs are crucial for promoting Phd2-deficient tumor growth in CAC, we used transgenic Vav:Cre-Phd2fl/fl mice that harbor a homozygous deletion of Phd2 in all hematopoietic lineages (including macrophages) and subjected them to AOM/DSS treatment. Furthermore, to exclude a potential impact of PHD2 in epithelial cells on CAC growth, we also induced AOM/DSS tumors in Villin:Cre-Phd2fl/fl mice, which are Phd2 deficient in the intestinal epithelial cells (IECs) of the small intestine and colon (38, 39). In line with our previous results with Phd2+/– mice, Vav:Cre-Phd2fl/fl mice displayed significantly bigger tumors, while the tumor number was not changed compared with control animals (Figure 6, A and B). While tumors of Vav:Cre-Phd2fl/fl and control mice were equally apoptotic as assessed by CC3 IHC staining (Figure 6C), tumors from Vav:Cre-Phd2fl/fl mice showed aggravated tumor proliferation compared with their controls as assessed by IHC for PCNA (Figure 6D). Moreover, IHC staining revealed significantly increased phosphorylation of STAT3 and ERK1/2 (Figure 6, E and F) but unchanged nuclear β-catenin expression in Vav:Cre-Phd2fl/fl tumors compared with the control tumors (Supplemental Figure 6A). Strikingly, consistent with the phenotype observed in Phd2+/– animals, IHC staining for F4/80 suggested a significant increase in the number of TAMs in Vav:Cre-Phd2fl/fl tumors compared with the control tumors (Figure 6G). Moreover, upstream of STAT3 and ERK1/2 signaling, the mRNA transcript level of EGFR ligand Ereg, as quantified by qRT-PCR, was significantly augmented (Figure 6H), while the expression of Egfr and all other EGFR ligands — Areg, Egf, Hbegf, Tgfa, Epgn, and Btc — was not significantly altered in Vav:Cre-Phd2fl/fl tumors compared with control tumors (Supplemental Figure 6, B and C). Furthermore, the transcript expression of protumorigenic IL-6 was significantly increased, while expression of IL-11 was modestly (but not significantly) elevated in Vav:Cre-Phd2fl/fl tumors compared with WT control tumors (Figure 6I).

Lineage-specific deletion of Phd2 in the hematopoietic but not the epithelial cell compartment aggravates colitis-associated tumor growth. (A) Macroscopic quantification of AOM/DSS-induced tumors. Number of tumors per mouse (Phd2fl/fl control: n = 8; and Vav:Cre-Phd2fl/fl: n = 7 mice) and size of individual tumors (control: n = 127; and Vav:Cre-Phd2fl/fl: n = 100 tumors), and representative macroscopic images (right) of colons from control and Vav:Cre-Phd2fl/fl mice (B) H&E staining of colons from control and Vav:Cre-Phd2fl/fl mice. Arrows indicate colitis-associated tumors. Scale bar: 2 mm. (C) Quantification of CC3 immunostaining in control (n = 32) and Vav:Cre-Phd2fl/fl (n = 53) tumors (top) and representative histological images (bottom). Scale bar: 50 μm. (D) Quantification of epithelial PCNA immunostaining in control (n = 9) and Vav:Cre-Phd2fl/fl (n = 11) tumors (top) and representative histological images (bottom). Scale bar: 25 μm. (E) Quantification of epithelial nuclear p-STAT3Y705 immunostaining in control (n = 14) and Vav:Cre-Phd2fl/fl (n = 14) tumors (top) and representative histological images (bottom). Scale bar: 25 μm. (F) Quantification of p-ERK1/2 immunostaining in control (n = 31) and Vav:Cre-Phd2fl/fl (n = 30) tumors (top) and representative histological images (bottom). Scale bar: 100 μm. (G) Quantification of F4/80 immunostaining in control (n = 38) and Vav:Cre-Phd2fl/fl (n = 45) tumors (top) and representative histological images (bottom). Scale bar: 100 μm. (H and I) qRT-PCR analysis of EGFR ligand Ereg in H and IL-6 and IL-11 in I in WT (n = 14) and Phd2+/– (n = 14) tumors. Statistical significance was calculated using 2-tailed Student’s t test. *P < 0.05, ***P < 0.001, ****P < 0.0001.

In contrast, Villin:Cre-Phd2fl/fl mice did not display significantly altered CAC tumor growth compared with control animals (Supplemental Figure 7, A and B). Consistently, tumor proliferation, apoptosis, and STAT3 phosphorylation (Supplemental Figure 7, C–E), as well as gene expression of Ereg (Supplemental Figure 7F), IL-6, and IL-11 (Supplemental Figure 7G), were not significantly altered in Villin:Cre-Phd2fl/fl mice as compared with the controls. Together, Phd2 deficiency in hematopoietic cells (including TAMs), but importantly not in IECs, promotes CAC tumor growth at least in part by activation of the STAT3 and ERK1/2 signaling pathways mediated by EREG.

In conclusion, each of the 3 different HIF-PHD isoenzymes PHD1–3 has a distinct impact on CAC but importantly not on inflammation-independent, sporadic CRC tumor growth. This effect is tumor promoting (PHD1), tumor inhibiting (PHD2), or neutral (PHD3) (Figure 7). PHD2 expression (i) reduces the number of TAMs in AOM/DSS tumors, (ii) impairs the protumorigenic properties of macrophages at least in part through decreased Ereg expression, and (iii) diminishes STAT3 and ERK1/2 signaling in colitis-associated tumors.

The HIF-prolyl hydroxylases have distinct and nonredundant roles in colitis-associated cancer. PHD1–3 have distinct effects on CAC growth, which is tumor promoting (PHD1), tumor inhibiting (PHD2), or neutral (PHD3). Intestinal inflammation is diminished in Phd1-deficient but unaltered in Phd2- and Phd3-deficient mice. In CAC, PHD2 deficiency (i) increases the number of TAMs, (ii) promotes Ereg expression in macrophages, and (iii) augments STAT3 and ERK1/2 signaling, which at least in part contributes to aggravated tumor cell proliferation in colitis-associated tumors.