Collagen prolyl 4-hydroxylase 1 is essential for HIF-1α stabilization and TNBC chemoresistance

Collagen prolyl 4-hydroxylase (P4H) expression and collagen hydroxylation in cancer cells are necessary for breast cancer progression. Here, we show that P4H alpha 1 subunit (P4HA1) protein expression is induced in triple-negative breast cancer (TNBC) and HER2 positive breast cancer. By modulating alpha ketoglutarate (α-KG) and succinate levels P4HA1 expression reduces proline hydroxylation on hypoxia-inducible factor (HIF) 1α, enhancing its stability in cancer cells. Activation of the P4HA/HIF-1 axis enhances cancer cell stemness, accompanied by decreased oxidative phosphorylation and reactive oxygen species (ROS) levels. Inhibition of P4HA1 sensitizes TNBC to the chemotherapeutic agent docetaxel and doxorubicin in xenografts and patient-derived models. We also show that increased P4HA1 expression correlates with short relapse-free survival in TNBC patients who received chemotherapy. These results suggest that P4HA1 promotes chemoresistance by modulating HIF-1-dependent cancer cell stemness. Targeting collagen P4H is a promising strategy to inhibit tumor progression and sensitize TNBC to chemotherapeutic agents.

• Since it is key for the mechanism proposed that PHD levels are not directly affected by the silencing of P4HA1, and even though this is shown in the figure S3 for PHD1 (known not to be upregulated by HIF1), the reviewer would like to see the levels of the three PHDs upon P4HA1 silencing (as PHD2 and PHD3 are HIF targets).
• C ould the authors provide the expression level of P4HA1 as a control of the effectiveness of the overexpression and silencing for each concerning western blot?
• C ould the authors specify which inhibitor and at which concentration it was used in each figure? Since PHDs could also be inhibited, could the authors comment on the specificity of such inhibitors and how it affects their conclusions? C ould the authors explain their choice regarding the inhibitor used in vivo?
• In response to chemotherapy the authors attribute the increase in resistance of the C D44+/C D24-/low (clones 3 and 4) to the increased levels of P4HA1. They proved that silencing P4HA1 rescues the sensitivity to docetaxel and doxorubicin. What is the effect of the silencing in the sensitive clones (clones 1 and 3)? Are they further sensitized? Why do the authors choose the HMLE (breast epithelial) cell line for these experiments instead of the TNBC cell lines (MDA MB231 or Hs578T)? C ould they explain in detail how do they obtain the C D44+/C D24-/low and the C D44low/C D24+ clones?
• Interestingly upon short term treatment with docetaxel or doxorubicin the levels of P4HA1 are up-regulated (figure S5). How do the authors explain it? Is it a selection (increased proliferation or decreased apoptosis) of the high expressing cells? Or is it because of a direct effect on the expression of P4HA1 in response to the drugs? What is the time line of these experiments?
• In the in vivo experiments (figure 7 a and b) there are differences on the effect in the primary tumor growth when looking at the effect of the KD vs the P4H inhibitor. How do the authors explain no effects upon silencing but significant reduction in tumor growth in response to the single treatment with the P4H inhibitor?
• The combo (KD+docetaxel) improves the effect of the single treatments but eventually the tumors grow exponentially. Moreover, it seems that the combination of P4Hi+docetaxel is not so effective as the combo KD+docetaxel. C ould the authors discuss this and the relevance of the eventual tumor escape?
Minor comments: • General: minor typing mistakes. The reviewer would like to know whether at the start of each treatment animals were randomized in groups with the same average. From the text and the figures, it seems that it is not the case. This is especially relevant in figure 7b. o 7b: arrow indicating the beginning of P4Hi treatment should be provided, since in the M&M is not clear whether the treatment starts right after tumor injection and for 2 weeks or after two weeks, for the rest of the experiment. o In 7b n=8, whereas in the IHC analysis + met analysis n is lower, suggesting that the analysis hasn't been performed in all tumors. The authors should clarify this aspect. o 7c: could the authors provide the quantification of HIF-1a and collagen? • Figure 8: o 8d: it should be specified which treatment course did the patients follow. Reviewer #2: Remarks to the Author: The manuscript by Xiong et al. shows the role of collagen prolyl hydroxylase P4HA1 in HIF1α pathway in triple-negative breast cancer (TNBC ). The authors propose that P4HA1 controls HIF1α stability by regulating the intracellular levels of succinate and α-ketoglutarate, which in turn control HIF-prolyl hydroxylase (PHD) activity. According to their results, this axis P4HA1/HIF1α plays an important role in the response of TNBC patients to chemotherapy, highlighting the strategy of targeting P4HA1 to improve TNBC treatment. The story is clearly presented, and robustly sustained by evidence, particularly the in vivo experiments. However, the manuscript presents some inconsistencies and technical issues. Specific comments are as follows. 1. All over the manuscript, the authors used inconsistently different cell lines for different groups of experiment. Sometimes, the specific cell line used in the experiment is not even indicated. The authors should be more consistent regarding the cellular model used. 2. In the manuscript, PHDs have been proposed by the authors to play a key role in the connection between P4HA1 and HIF1α. However, no experiment of lack-of-function or gain-of-function of PHDs has been included in the manuscript. Those experiments will help to elucidate the mechanism of HIF-1α stabilization by P4HA1. 3. Figure 1D. The authors claimed an increase of collagen secretion but this figure only shows the difference in protein level of P4HA1 between TNBC cell lines and luminal cancer cells. 4. Figure S1. The authors claimed to use shRNA of P4HA1 or P4H inhibitors to check tumor progression in 3D culture systems, but only the result with shRNA against P4HA1 was shown. Data showing pharmacological P4HA1 inhibition (as claimed in the text) should be included in this figure. 5. The authors used gene co-expression and gene ontology analysis but the protocol is missing in the materials and methods part. Please include in methods section the procedure for gene coexpression and gene ontology analysis, as well as details of the statistical analysis of these results. 6. Figure 2B. The authors should include a positive control, namely HIF1α levels in hypoxia. Moreover, the authors might consider to include collagen secretion upon overexpression of P4HA1 and knockdown of P4HA1. 7. Figure 2D. Why are the levels of HIF1 in these panels so high in comparison to Figure 2E (normoxia)? Have these experiments been performed in normoxic or hypoxic conditions? 8. Figure 2G, 2H and 2I. In experiments showing HRE-luciferase activity and expression of HIF1αtarget genes, positive control of cells incubated in hypoxia should be included. 9. Figure 2F and Figure S2A. The concentration of P4H inhibitors is not consistent. 10. Figure 3E. In comparison with Figure 2D, why a decrease in HIF1α level was not seen upon shP4HA1 in MDA231 cells? For comparative purposes, the authors should consider to do the overexpression of P4HA1 and shP4HA1 in the same cell line. 11. Figure 3D, 3E, 4F and 4H lack the control immunoblot for P4HA1. 12. Figure 3F and 3G lack a positive control in hypoxia. Again, it would have been convenient to perform these experiments in the same cell line. 13. Figure 3H should include an immunoblot for hydroxylated-HIF1α. 14. In page 7, the authors claimed that "We have shown that PHD is active in hydroxylating HIF-1α in these cells when P4HA1 is knocked down, but P4HA1 does not influence the protein level of the PHD ( Figure S3)". However, Figure S3 does not show any result regarding PHD activity. 15. Figure S3 only shows the immunoblot against PHD1. Immunoblots against PHD2 and PHD3 should be included as well. Indeed, among all the 3 members of the PHD family, PHD2 is the main isoform regulating HIF1α stability in cells (although all 3 enzymes can hydroxylate HIF1α in vitro). 16. In page 8, the sentence "We therefore determined whether P4HA1 enhances proline hydroxylation on HIF1α" would have a more logical meaning if expressed as "We therefore determined whether P4HA1 reduces proline hydroxylation on HIF1α". 17. Figure 4F and 4G. The concentration of octyl-α-ketoglutarate and dimethyl-succinate were not indicated. 18. Figure S4A and S4B. The authors should check if octyl-α-ketoglutarate treatment prevents the effect of P4HA1 overexpression (as they did in Figures 4H and 5F).

Reviewer #3:
Remarks to the Author: Xiong. G et demonstrated the novel role of collagen P4HA1 in promoting chemoresistance by modulating HIF-1 dependent cancer cell stemness in TNBC models, patient derived organoids and patient cohort. Additionally, they demonstrated that the P4HA1 inhibitor or inhibition of P4HA1 in combination with decetaxel treatment induces regression of tumor growth and colonization in TNBC breast cancer in vivo models. This is interesting, new and an important observation. However, authors need to address if high or low expression of P4HA1 correlates to the chemotherapy response in a HIF-1 alpha dependent manner in TNBC patients using publically available multiple datasets to address if P4HA1 is a predictive/prognostic marker in response to TNBC chemotherapy treatment.
Major Specific comments: 1. In fig. 1c, authors need to include P4HA1 staining of metastatic TNBC tumor sections ideally with matched primary tumors to address if P4HA1 expression is associate with TNBC progression. Additionally, they need to ideally compare the P4HA1 expression in metastatic ER+ tumor sections with matched primary tumor sections to address if high P4HA is specific to TNBC progression not ER+ progression. If matched samples cannot be found, then unmatched analysis would be acceptable.
2. In all P4HA1 knockdown experiments performed in TNBC cell lines, lack rescue controls. In all these experiments they need to reconstitute P4HA1 in TNBC knockdown cells and show the effect or phenotype is rescued and is specific to P4HA1. (Fig, 2d, e; 3d, e; 4b-e, g; 5e, g-k; Fig 6a, b, d, f, g and 7a-c).
3. The role of P4HA1 in breast cancer metastasis is well established in the literature so the key contribution here seems to be linking P4HA1 to HIF1 alpha for inducing chemotherapy resistance. They need to address if HIF1 alpha is critical for P4HA1 mediated chemotherapy resistance, cell stemness, and tumor growth using TNBC models by performing HIF1 alpha knockdown experiments. 7. In result section: "P4HA1 expression is up-regulated in TNBC and is associated with collagen production", expression comparison of P4HA1 in TC GA/C urtis -TNBC Vs Other subtypes + across TNBC stages is missing. Showing levels of P4HA1 in actual patient data will provide a more realistic perspective of high levels both across and within subtypes. 8. Similarly, lack of association between levels of P4HA1 and HIF-1α at mRNA level should be tested in patient samples as well.
9. The study shows survival curves from kmplot, which consolidates different datasets together. Performing survival analysis on individual dataset like C urtis et al (METABRIC ) by using the therapy annotations for the categorization needs to be done as a confirmation of the finding.
10. Studies in the past have shown importance of STAT3 along with HIF-1α in chemoresistance in TNBC . The status of STAT3 in P4H41-high and HIF-1α-stable TNBC cell-lines and samples needs to be examined and discussed.

Minor comments:
I thank reviewers for their very thoughtful and constructive comments, which are most helpful in refining and strengthening this study.

Reviewer #1, Expertise: HIF, metabolism (Remarks to the Author):
In the present study Xiong et al. describe a new mechanism contributing to HIF-1α stability in TNBC. The authors demonstrate that expression of the collagen prolyl 4-hydroxylase subunit 1 (P4HA1) in cancer cells is essential for HIF-1α stabilization in an oxygen independent manner. The role of HIF-1α in breast cancer progression and metastasis has been described before, however the authors shed light on a new mechanism that further contributes to HIF-1α mediated malignancy in TNBC. Furthermore, an interesting role of the P4HA1-HIF-1α axis in the resistance to chemotherapy supports the interest of the work. Together with the correlation with patient data and in vivo (xenografts and PDX) experiments the authors proposed that targeting P4HA1-HIF-1α axis could contribute to improve TNBC treatment.
We find the manuscript suitable for publication in Nature Communications after addressing the following comments.

Response:
We thank the reviewer for this and the following comments.
Major comments:

• How is P4HA1 induced in TNBC cells? Is it HIF1a dependent? Are TNBCs more HIF1a dependent than other breast cancer cell types?
Response: It has been shown that P4HA1 transcription is induced by HIF-1α and hypoxia in mammary epithelial cells (7,61). Our data and previous report suggest a positive feedback loop between collagen hydroxylation and the HIF-1 pathway. Our data also suggest that P4HA1 in TNBC is not only regulated by HIF-1α. We have unpublished data showing that P4HA1 expression was induced during the epithelial-mesenchymal transition (EMT). EMT inducer Twist, Snail, and TGF-β can stimulate P4HA1 expression in mammary epithelial cells. It has been reported that TNBC/basal-like breast cancer is associated with activation of the EMT program. Thus, P4HA1 expression in TNBC may be induced by EMT activators. This discussion has been included in the manuscript on page 17.
HIF-1 activation is associated with the development and progression of ER positive and HER2 positive breast cancer. The activation also contributes to ER positive and HER2 cancer progression. Gene expression analysis showed that the HIF-1 pathway is hyperactivated in basallike breast cancer (20). It is not clear if TNBC are more HIF-1 dependent than other breast cancer subtypes. We respectfully suggest this question to be addressed in future. [Redacted]

• The results in figure 1a, b suggest that Her2+ tumors express similar levels of P4HA1 than TNBC tumors. Could the authors comment on this?
Response: Thank the reviewer for bringing this to our attention. We also noticed that P4HA1 expression was upregulated in HER2+ tumor. In addition, it has been shown that ERBB2 overexpression in mammary epithelial is associated with P4HA2 expression (Mackay A et al, Oncogene, 2003, 1; 22 (17)). Gilkes DM, et al. 's study showed that P4HA1 mRNA levels were increased in HER2 breast cancer (7). We have analyzed P4HA1 mRNA expression in published microarray dataset (supplemental Figure S9a) and confirmed their findings. We now include the following sentences in the abstract "we showed that P4H alpha 1 subunit (P4HA1) protein expression was induced in triple-negative breast cancer (TNBC) and HER2 positive breast cancer " and in the discussion on page 14: 'P4HA1 expression is also upregulated in HER2 positive tumor. However, roles of P4HA1 in HER2 cancer progression still needs further clarification.' • Since it is key for the mechanism proposed that PHD levels are not directly affected by the silencing of P4HA1, and even though this is shown in the figure S3 for PHD1 (known not to be upregulated by HIF1), the reviewer would like to see the levels of the three PHDs upon P4HA1 silencing (as PHD2 and PHD3 are HIF targets).
Response: This point is well taken. We have examined PHD2 and PHD3 protein levels in control and P4HA1-silenced MDA-MB-231 cells (supplemental Figure S3). We found that PHD3 was expressed at low level in MDA-MB-231 cells. Silencing P4HA1 did not significantly change PHD1 and PHD3 protein expression (as showed in figure S3). PHD2 protein levels are slightly down-regulated in P4HA1-silenced MDA-MB-231 cell.
• Could the authors provide the expression level of P4HA1 as a control of the effectiveness of the overexpression and silencing for each concerning western blot?
Response: As the reviewer suggested, we now include the P4HA1 western blot data to confirm the overexpression and silencing efficiency (See revised Figure 2, 3, 4).

• Could the authors specify which inhibitor and at which concentration it was used in each figure? Since PHDs could also be inhibited, could the authors comment on the specificity of such inhibitors and how it affects their conclusions? Could the authors explain their choice regarding the inhibitor used in vivo?
Response: We have specified which inhibitor was used in each figure and also provided the concentration information in figure legends. It has been shown that 1,4-DPCA inhibits P4HA1 activity with an IC50 value of 3. Characterization of the human prolyl 4-hydroxylases that modify the hypoxia-inducible factor.). In the tissue culture experiment, we showed that 1,4-DPCA (5 µM) or DHB (10 µM) treatment or reduced HIF-1α protein level and HRE luciferase activity. It is not likely that PHDs activity is inhibited by those inhibitors at these concentrations.
DHB has been safely used in vivo to inhibit collagen deposition at 40 mg/kg weight per day (7). It also inhibits cancer metastasis at this concentration (7). Thus, we chose this inhibitor for the xenograft experiments. This rationale has been included in the supplemental material and methods on page 22. All the CD44+/CD24-/low and the CD44low/CD24+ clones were derived from HMLE cells. In brief, we passaged about 1000 HMLE cells to a 100 mm culture dish. After the mono-cell clones growth, we picked up the clones and expanded them. The CD44/CD24 levels in these clones were examined by FACS analysis. Similar method has also been used by (Mani SA et al,Cell. 2008 16;133(4):704-15). We now include this information in supplemental experimental procedure (page 21) and the reference in the manuscript.  Figure 5a, b, c). Therefore, short term treatment with docetaxel or doxorubicin may enrich cancer stem cells with high levels of P4HA1. We have included this discussion in the manuscript on page 11.
• In the in vivo experiments ( figure 7 a and b) there are differences on the effect in the primary tumor growth when looking at the effect of the KD vs the P4H inhibitor. How do the authors explain no effects upon silencing but significant reduction in tumor growth in response to the single treatment with the P4H inhibitor?
Response: Thank the reviewer for bringing this to our attention. We also noticed that the P4H inhibitor is more potent to inhibit primary tumor growth. P4HA family includes P4HA1, P4HA2 and P4HA3. Inhibitory activity of DHB is not specific to P4HA1 (Vasta JD, ACS Chem Biol. 2016 15; 11(1):193-9). It is possible that P4H inhibitor treatment suppresses tumor growth by inhibiting other P4HA family members. We now include this discussion on page 12.
• The combo (KD+docetaxel) improves the effect of the single treatments but eventually the tumors grow exponentially. Moreover, it seems that the combination of P4Hi+docetaxel is not so effective as the combo KD+docetaxel. Could the authors discuss this and the relevance of the eventual tumor escape?

Response:
The reviewer is right that the silencing P4HA1 was more effective to sensitize MDA-MB-231 cells to docetaxel treatment compared with P4Hi. One difference between shRNA experiment and DHB treatment is that P4HA1 was consistently knockdown by shRNA, while DHB treatment was stopped at day20. It is plausible that continually silencing P4HA1 delayed the relapse after the docetaxel treatment. Another possibility is that silencing P4HA1 may be more potent in reducing P4HA1 activity. Therefore, it is crucial to identify more potent and specific P4HA1 inhibitors in the future.

Response:
We apologize for these oversights. The typing mistakes have been corrected.  Response: Protein levels of P4HA1 expression in human breast cancer tissues were determined using tissue microarray (TMA). Each case in the TMA is represented by three separate tissue cores. IHC scoring results from each of these three tissue cores were averaged to produce a final score for each case. P4HA1 immunohistochemical staining was scored according to a standard semi-quantitative scale as follows: negative (0), weak (1), moderate (2), strong (3). Scoring was performed by a pathologist who was blinded to clinical and pathologic variables. This information is now included in the Supplemental Material and Methods on page 18. Response: The SIRM experiment has been used to identify metabolite changes at the cellular level as an unbiased approach, and n value is not big enough for the statistic significant analysis. We hypothesize that collagen hydroxylation mainly consumes the cytoplasmic a-KG, thus we specifically measured cytoplasmic a-KG level in 4c. Response: It has been reported that tumor-initiating cells are more resistant to the chemotherapeutic agent treatment (26, 37). It is most likely that Dox or Doc treatment selected the tumor-initiating cell population in shControl MDA-MB-231 cells, and consequently enhances tumor sphere formation. We now include this discussion on page 11.

o In 7b n=8, whereas in the IHC analysis + met analysis n is lower, suggesting that the analysis hasn't been performed in all tumors. The authors should clarify this aspect.
Response: We are sorry that we did not clarity it clearly in the manuscript. These are two experiments. For the experiment in 7b, the tumors were kept in mice after 21 days to examine the tumor regrowth without drug treatment. For the experiment in 7c-f, to examine the effects of drug treatments on cell proliferation and apoptosis, we performed surgery to remove tumors at Day 21 for H&E and IHC analyses. At this time point, tumor did not start to regrown and the drug may still be effective. The mice were kept for another month after the surgery to detect metastasis to lung. We now add this information in the Supplemental Material and Methods on page 22.

The authors claimed an increase of collagen secretion but this figure only shows the difference in protein level of P4HA1 between TNBC cell lines and luminal cancer cells.
Response: We are sorry for this oversight. The collagen secretion data are now included in Figure 1d.
4. Figure S1. The authors claimed to use shRNA of P4HA1 or P4H inhibitors to check tumor progression in 3D culture systems, but only the result with shRNA against P4HA1 was shown.

Data showing pharmacological P4HA1 inhibition (as claimed in the text) should be included in this figure.
Response: We have included these data in Figure S1d-f.

The authors used gene co-expression and gene ontology analysis but the protocol is missing in the materials and methods part. Please include in methods section the procedure for gene coexpression and gene ontology analysis, as well as details of the statistical analysis of these results.
Response: The protocol of gene co-expression and gene ontology analysis is now included in the supplemental table 1 on page 16.  7. Figure 2D. Why are the levels of HIF1 in these panels so high in comparison to Figure 2E (normoxia)? Have these experiments been performed in normoxic or hypoxic conditions? Response: These blots were from two different experiments. The HIF-1α protein level was very high in hypoxia condition, and thus the bot in Figure 2f (Original 2e) was exposed in much short time compared with the blot in Figure 2e (original 2d). Therefore, we cannot compare these two blots because they were generated under different exposure conditions.

Figure 2G, 2H and 2I. In experiments showing HRE-luciferase activity and expression of HIF1α-target genes, positive control of cells incubated in hypoxia should be included.
Response: This point is well taken. We have performed the HRE-luciferase report assay and examined HIF1α-target genes expression under hypoxia condition. The results are now included in Figure 2h, i, j. Figure 2F and Figure S2A. The concentration of P4H inhibitors is not consistent.

Response:
We apologize for the wrong labeling in Figure S2d. The concentration is at μM.
10. Figure 3E. In comparison with Figure 2D, why a decrease in HIF1α level was not seen upon shP4HA1 in MDA231 cells? For comparative purposes, the authors should consider to do the overexpression of P4HA1 and shP4HA1 in the same cell line.

Response:
We tried to compare HIF-1 hydroxylation at similar amount of total HIF-1α protein, thus we loaded more protein lysate for shP4HA1 MDA-MB-231 cells, as shown by the loading control Lamin A/C. We now provide a rationale for performing gene silencing and overexpression in different cell lines. See the response to comment 1. In general, we silenced P4HA1 in cell lines with high P4HA1 expression level (MDA-MB-231), and ectopically expressing P4HA1 in cells lines with low P4HA1 protein levels (MCF10A and HMLE). Figure 3D, 3E, 4F and 4H lack the control immunoblot for P4HA1.

11.
Response: As reviewer requested, we now include the immunoblot for P4HA1 in Figure 3e, 3f (original 3d), 3g (original 3e), 3j, 4f, 4g, and 4j. Figure 3F and 3G lack a positive control in hypoxia. Again, it would have been convenient to perform these experiments in the same cell line.

Response:
We have performed these experiments under both normoxia and hypoxia conditions, and the results are now included in Figure 3h (original 3f) and 3i (original 3g).
13. Figure 3H should include an immunoblot for hydroxylated-HIF1α. , and it has been shown that HIF-1α PPAA mutant is more resistant to hydroxylationdependent protein degradation (13). We respectfully suggest not repeating these experiments.
14. In page 7, the authors claimed that "We have shown that PHD is active in hydroxylating HIF-1α in these cells when P4HA1 is knocked down, but P4HA1 does not influence the protein level of the PHD (Figure S3)". However, Figure S3 does not show any result regarding PHD activity.

Response:
We are sorry for the inaccurate description. We now change the sentence to "We showed that silencing P4HA1 had little effect on the protein level of the PHD1 and PHD3, and slightly reduced PHD2 protein level ( Figure S3a), suggesting that reduced HIF-1 hydroxylation is not due to increased PHDs expression." (See page 7).

Figure S3 only shows the immunoblot against PHD1. Immunoblots against PHD2 and PHD3 should be included as well. Indeed, among all the 3 members of the PHD family, PHD2 is the main isoform regulating HIF1α stability in cells (although all 3 enzymes can hydroxylate HIF1α in vitro).
Response: We appreciate that this part of the study need more attention. We have examined all three PHDs protein levels in P4HA silencing MDA-MB-231 cell (Supplemental Figure S3a).
16. In page 8, the sentence "We therefore determined whether P4HA1 enhances proline hydroxylation on HIF1α" would have a more logical meaning if expressed as "We therefore determined whether P4HA1 reduces proline hydroxylation on HIF1α". Response: Thank the reviewer for identifying this mistake. We have corrected the sentence in the revised manuscript. Figure 4F and 4G. The concentration of octyl-α-ketoglutarate and dimethyl-succinate were not indicated.

Response:
We now indicate concentration of octyl-α-ketoglutarate and dimethyl-succinate in Figure 4f and 4g.
18. Figure S4A and S4B. The authors should check if octyl-α-ketoglutarate treatment prevents the effect of P4HA1 overexpression (as they did in Figures 4H and 5F).

Response:
We have performed the seahorse assay in P4HA1-expressing MCF10A cells in the presence of octyl-α-ketoglutarate (See FigureS5b), and results showed that octyl-α-ketoglutarate treatment rescued oxygen consumption in P4HA1 overexpression cells. Response: To address this comment, we have analyzed the published microarray dataset generated from 20 primary breast cancer tissues and corresponding metastatic lymph node tumors (Suzuki M, Mol Oncol. 2007 1(2):172-80.). We did not detect a significant difference in P4HA1 expression between the primary tumors and metastasis lesions; in fact, P4HA1 expression is slightly downregulated in the lymph node tumors. It has been reported that P4HA1 expression is essential for breast cancer metastasis (7). These results suggest that the gene promoting cancer metastasis is not necessarily highly expressed at the metastasis lesions. This is especially the case for the EMT-related gene. It is well-established that EMT promotes cancer metastasis by enhancing cancer cell invasion. However, the mesenchymal to epithelial transition . We have some unpublished data showing that P4HA1 expression is induced by Twist during EMT (see response to reviewer 1's comment). It is mostly likely that increased P4HA1 expression in primary tumor promotes cancer cell invasion (7) and stemness ( Figure. 5), which facilitates the initiation of cancer cell colonization at the distance organs. However, P4HA1 expression may be repressed during the mesenchymal-epithelial transition (MET) process that is required for the macrometastasis formation. It would be interesting to compare the P4HA1 protein levels between the matched primary tumors and metastatic lesions in specific subtypes; unfortunately, we don't have the matched samples collected in our institute to address this question.

Primary
Metastasis P4HA1 mRNA level 8 10 12 14 2. In all P4HA1 knockdown experiments performed in TNBC cell lines, lack rescue controls. In all these experiments they need to reconstitute P4HA1 in TNBC knockdown cells and show the effect or phenotype is rescued and is specific to P4HA1. (Fig, 2d, e; 3d, e; 4b-e, g; 5e, g-k; Fig  6a, b, d, f, g and 7a-c).
Response: Thank the reviewer for the suggestion. We have reconstituted P4HA1 expression in shP4HA1-expressing MDA-MB-231 cells (see supplemental information on page 17). We found that restoring P4HA1 expression rescued HIF-1α protein level in shP4HA1-expressing MDA-MB-231 cells ( Figure S2b). Increased α-KG level and reduced succinate levels in shP4HA1expressing cells were also rescued by P4HA1 restoration (Figure S4a, b). We also analyzed cancer cell stemness (tumorsphere formation assay, FACS analysis of aldehyde dehydrogenase activity) and ROS level after P4HA1 expression was restored in shP4HA1 expressing cells ( Figure S5a, b, e). We found that cancer cell stemness and ROS levels were rescued by P4HA1 expression. These results indicate that the effects or phenotypes in shP4HA1-expressing cells are specific to P4HA1 silence.
3. The role of P4HA1 in breast cancer metastasis is well established in the literature so the key contribution here seems to be linking P4HA1 to HIF1 alpha for inducing chemotherapy resistance. They need to address if HIF1 alpha is critical for P4HA1 mediated chemotherapy resistance, cell stemness, and tumor growth using TNBC models by performing HIF1 alpha knockdown experiments.

Response:
As the reviewer pointed out, P4HA1 roles in cancer metastasis is well established; thus the present study focus on function of P4HA1 in regulating the HIF-1 pathway and chemoresistance. To address if HIF-1α is critical for P4HA1 mediated chemotherapy resistance and cell stemness, we have performed a number of rescue experiments by restoring HIF-1α protein expression in P4HA1-silenced cells. We found that expression of the stable HIF-1α (PPAA) could at least partially rescue cancer cell stemness and reduced ROS levels and oxidative phosphorylation in P4HA1-silenced cells (See Figure. 5h, 5i, 5j, 5k, 5l). To determine whether P4HA1 promotes chemoresistance through HIF-1, we have knocked down P4HA1 in HIF-1α (PPAA)-expressing MDA-MB-231 cells, and then examined the response of those cells to docetaxel treatment in vivo. We found that silence of P4HA1 fails to sensitize PPAAexpressing cells to docetaxel (See Figure S7b). There results indicate that HIF-1α is critical for P4HA1-mediated cell stemness and chemotherapy resistance. It has been reported that inhibition of the HIF-1 pathway sensitizes TNBC cells to chemotherapy resistance in TNBC and inhibits cancer cell stemness (Zhang C et al, Proc Natl Acad Sci U S A. 2016 5;113(14):E2047-56.), thus we did not perform HIF-1α knockdown experiments to repeat previous studies. We hope that the rescue experiments in the revised manuscript can address the reviewer's concern.
4. Fig 3a- Response: We agree. We have performed the HIF-1α protein degradation experiments in MDA -MB-231 cells, and the results also show that knockdown of P4HA1 significantly increases the HIF-1 degradation at the early time points (Figure 3a). To analyze HIF-1α ubiquitination, we need to co-transfect three vectors expressing shP4HA1, HIF-1α-flag, and Ubiquitin-HA into cells (Figure 3e). We encountered a technic issue when performing the HIF-1α ubiquitination experiments in MDA-MB-231 cells. The transfection efficiency is relatively low in MDA-MB-231 cells compared with 293FT cells, and it is hard to pull down enough protein to analyze HIF-1α ubiquitination.  Figure 7a. Results show that silencing P4HA1 reduced ALDH positive staining (supplemental Figure S7a), suggesting that cancer cell stemness is enhanced by P4HA1 in vivo. 7. In result section: "P4HA1 expression is up-regulated in TNBC and is associated with collagen production", expression comparison of P4HA1 in TCGA/Curtis -TNBC Vs Other subtypes + across TNBC stages is missing. Showing levels of P4HA1 in actual patient data will provide a more realistic perspective of high levels both across and within subtypes.

Response:
We appreciate that this part needs more attention. Our human breast cancer TMA data (Figure 1a) indicate that P4HA1 is highly expressed in both TNBC and HER2 breast cancer tissue. Therefore, we have changed the sentence to "The upregulation of P4HA1 in breast cancer cell lines is associated with increased secretion of collagen". As reviewer suggested, we have analyzed the P4HA1 mRNA levels in TCGA dataset. Results showed that P4HA1 is also highly expression in highly proliferative ER positive breast cancer, and that Her2 positive cancer has highest expression of P4HA1 (Supplemental Figure S9). One potential explanation for the discrepancy between the microarray and IHC data is that we only quantified P4HA1 protein levels in cancer cell for the IHC analysis, while P4HA1 mRNA expression in TCGA data may derived from both stromal and cancer cells.

Similarly, lack of association between levels of P4HA1 and HIF-1α
at mRNA level should be tested in patient samples as well.
Response: As reviewer suggested, we have analyzed the association between P4HA1 and HIF-1α mRNA levels in the published microarray dataset (Chin K, 2007, Cancer Cell, 10 (6), 529-41). We identified a weak correlation (r=0.218) between P4HA1 and HIF-1α mRNA expression (see below). Our study mainly focuses on the regulation of HIF-1α at the protein level. Expression of the HIF-1 gene signature reflects HIF-1 activity in cancer tissue. Therefore, we determined the association between P4HA1 expression and HIF-1 gene signature, and results showed that expression of HIF-1 gene signature is associated with P4HA1 levels in human breast cancer tissue (Figure 8f). Response: Thank the reviewer for the suggestion and providing the detail information for the analysis. We have performed the survival analysis in the METABRIC dataset, and results confirm the finding that increased P4HA1 expression is associated with poor prognosis after chemotherapy. The data is now included in supplemental Figure S8g.
10. Studies in the past have shown importance of STAT3 along with HIF-1α in chemoresistance in TNBC. The status of STAT3 in P4H41-high and HIF-1α-stable TNBC cell-lines and samples needs to be examined and discussed.
Response: Roles of STAT3 in chemoresistance are well established. Currently we don't have strong data to support that P4HA1 promotes chemoresistance through STAT3. Our unbiased gene co-expression analysis did not detect the association between P4HA1 and the STAT3 pathway or gene signature. In addition, the western blot data show silencing P4HA1 only slightly reduced STAT3 phosphorylation (See below). We now include this discussion on page 16 and 17.
Minor comments: 1. In Figure 8D and e patient number in respective groups should be shown.

Response:
The patient number is now included in Figure 8d.

Quantification of fig S7 should be done to address the differences between different groups.
Response: As reviewer suggested, we have quantified IHC staining, and data is included in Figure S8f.
3. Fig 1A, ki67 staining should be included in the panel. Additionally, they need to include P4HA1 staining high stage HER2 positive tumor sections with matched primary tumors to address if P4HA1 expression is associate with only TNBC progression not HER2 positive breast cancer progression.
Response: As reviewer requested, we now include Ki67 staining and the Her2 panel in Figure  1A. As we mentioned in the response to comment 1, we don't have the access to the matched HER2 positive breast cancer tissue samples. However, we have analyzed association between P4HA1 expression and prognosis of HER2 breast cancer patients. We found that increased P4HA1 expression correlates with poor prognosis (see below). Therefore, it is most likely that P4HA1 expression is also associated with progression of HER2 positive breast cancer.