Proline rich 11 (PRR11) overexpression amplifies PI3K signaling and promotes antiestrogen resistance in breast cancer

The 17q23 amplicon is associated with poor outcome in ER+ breast cancers, but the causal genes to endocrine resistance in this amplicon are unclear. Here, we interrogate transcriptome data from primary breast tumors and find that among genes in 17q23, PRR11 is a key gene associated with a poor response to therapeutic estrogen suppression. PRR11 promotes estrogen-independent proliferation and confers endocrine resistance in ER+ breast cancers. Mechanistically, the proline-rich motif-mediated interaction of PRR11 with the p85α regulatory subunit of PI3K suppresses p85 homodimerization, thus enhancing insulin-stimulated binding of p110-p85α heterodimers to IRS1 and activation of PI3K. PRR11-amplified breast cancer cells rely on PIK3CA and are highly sensitive to PI3K inhibitors, suggesting that PRR11 amplification confers PI3K dependence. Finally, genetic and pharmacological inhibition of PI3K suppresses PRR11-mediated, estrogen-independent growth. These data suggest ER+/PRR11-amplified breast cancers as a novel subgroup of tumors that may benefit from treatment with PI3K inhibitors and antiestrogens.

one of genes (n=323), which was significantly upregulated by estrogen stimulation. PRR11 is included as an upregulated gene in Supplementary Table 3 of the article in Cancer Research.

What is the distance of the sequences encoding BRIP1, SMARCD2 and TACO1 genes to that encoding PRR11?
We have included a cytogenetic band of 17q showing the distance among those genes in Supplementary Fig. 2a of the revised manuscript (Rebuttal Fig. 2). Rebuttal Fig. 2. Location of PRR11, BRIP1, TACO1 and SMARCD2 in the chromosome 17q.

Fig. 2e. Where was the V5 epitope introduced? The authors should present controls to show that the epitope does not perturb the function of the protein. How do the levels of expression achieved upon overexpression relate to the normal levels of expression?
These are commercially available constructs tagged with V5 at their C terminus. We did not test the function of tagged protein because the functions of each protein are unclear or unknown. Based on our previous experiments with pLX302 or pLX304 vectors and other publications using these vectors, the V5 epitope is not expected to affect their own protein function (PMID: 28978427, 24356096). In both MDA-MB-134VI and MDA-MB-175 cells, overexpression of PRR11 achieved higher levels of PRR11 protein compared to endogenous PRR11 protein (Rebuttal Fig. 3). Rebuttal Fig. 3. Lysates from MDA-MB-134VI and MDA-MB-175VII cells stably transduced with pLX302-LacZ or -PRR11 were subjected to immunoblot analysis with PRR11 and actin antibodies.

For readers outside the field, what is the meaning of the ATARiS score?
ATARiS is a computational quantification method of gene suppression phenotypes from multi-sample RNAi screens (PMID: 23269662). We have included this description in the revised manuscript.

Fig. 3b. Please, describe the different visible bands. Is the slow migrating band endogenous PRR11? If so, why does ectopically expressed PRR11 migrate slower?
The upper band in the last column of Fig. 3b represents exogenous PRR11 protein which is tagged with V5 epitope. Technically, it has slightly higher molecular weight compared to endogenous PRR11 protein.
For clarity, we have added arrows indicating endogenous and exogenous PRR11 protein (Rebuttal Fig.  4).
Rebuttal Fig. 4. MCF7 LTED and HCC1428 LTED cells were transduced with shRNA targeting the 3' UTR of PRR11 and then retransduced with pLX304-GFP or pLX304-PRR11-V5. Cell lysates were subjected to immunoblot analysis with PRR11 and actin antibodies. Exo, exogenous; endo, endogenous. Fig. 3b. The knock-out of PRR11 should completely abrogate expression of PRR11. However, there are detectable levels of expression in the data shown in extended data Fig. 3b.

Have the authors analyzed a pool of cells, some not knocked-out?
Due to the essentiality and multiple copies of PRR11 in MCF7 and HCC1428 cells, we failed to get clones with complete knockout of PRR11. Since we used the pooled population of cells with CRISPR knockout, PRR11 protein coded by a non-edited allele was also detected. Hence, in this revised manuscript, we focused on PRR11 shRNA and siRNA to knock down PRR11 expression.

The status of amplification of MCF7 cell, an experimental model repeatedly used in the manuscript is confusing: "MCF7 and HCC1428 cells, both of which display high PRR11 copy number". How to the levels of PR11 amplification of parental MCF7 cells compare to those of MCF7 LTED (longterm estrogen deprived), (FulvR) MCF7 cells tamoxifen-resistant (TamR) MCF7.
We thank the reviewer for this comment and apologize for the inadequate explanation. Based on the public dataset from Cancer Cell Line Encyclopedia, both MCF7 and HCC1428 parental cells harbor PRR11 high copy number. We did not evaluate further genomic copy number increases in MCF7 and HCC1428 LTED cells, but we did observe an elevation of PRR11 protein levels in both LTED lines compared to their respective parental counterparts (Rebuttal Fig. 5).
Rebuttal Fig. 5. Lysates of MCF7 and HCC1428 LTED cells were subjected to immunoblot analysis with PRR11 and actin antibodies.

Fig. 4a. How do the authors define "tumors with high PRR11 mRNA"?
The median level of PRR11 mRNA was used to define PRR11-high or -low tumors as described in figure legend of the original manuscript. However, as we replied above, we have split PRR11-high vs. -low tumors with the FPKM cutoff adopted from the human protein atlas in the revised manuscript. For the Kaplan-Meier plotter analyses, the auto select best cut-off was used for splitting patients with high PRR11 expression vs. low.

Fig. 4g. Does the rescue shown here correspond to the rescue of cell proliferation?
We apologize for the typo in the figure. Cells shown in Fig. 4g correspond to the rescue models shown in Figs. 2b and 2c. Fig. 4g and corresponding legend have been corrected accordingly. Based on a quantification of immunoblot band intensity, the p85α levels induced by PRR11 overexpression were about 2-3 folds compared to respective controls (Rebuttal Fig. 6A). The intensity of p85α immunoblot bands was measured using the Image Lab software (ver. 6.0, BioRad).  Fig. 1 and Fig. 2a-d,

g) strengthen the case for an association of PRR11 with clinical outcome and tumour cell proliferation markers. Cell-based studies were then undertaken to functionally explore the impact of PRR11 over/under-expression on in vitro breast cancer cell proliferation and drug sensitivity.
Evidence is presented that PRR11 downregulation leads to some level of antiproliferation/enhanced sensitivity to estrogen suppression, with overexpression having the opposite effect (Fig. 3). In a next series of analyses, PRR11 gene expression is linked to a PI3K activation signature in cancer and to enhanced PI3K activation in cell-based models. This is mechanistically supported by data to indicate that PRR11 can bind the p85α regulatory subunit of PI3K and hereby suppress p85 homodimer formation, allowing p85 instead to associate with the

The strongest conclusion of this manuscript is that, largely based on bio-informatic data and some cell-based observations, PRR11 amplification in ER+ breast cancer is a candidate genetic marker to aid patient selection for ER-targeted therapies. However, the role of the PRR11-PI3K signalling link remains unclear, and the current conclusion of increased cellular sensitivity to PI3K inhibitors lacks sufficient evidence.
Major points:

It is critical that all bio-informatic data/code are made available alongside extensive expansion of the methods to detail the key steps undertaken -essentially as provided to the reviewer upon additional request.
Data and code availability are stated in the reporting summary which is enclosed with the revised manuscript.

The authors should test the impact of modulation of PRR11 expression, particularly downregulation, on the proliferation of a range of non-ER responsive breast cancer lines, especially triple negative and/or HER2+ cell lines. Based on their clinical observations on p. 4, lines 86-89, one would expect such cell lines to be insensitive to PRR11 downregulation and will thus strengthen the conclusions they make with regards to this protein's role in ER-positive tumours only.
We thank the Reviewer for these suggestions. Our aim was not to forge a link between estrogen sensitivity and PRR11 levels, but to determine whether high PRR11 affects the response to estrogen suppression. The 'low estrogen conditions in vivo' are our attempt to try to reproduce the clinical scenario in the original clinical cohort where, in the setting of therapeutic estrogen suppression with letrozole (e.g., low estrogen in vivo), tumors with high PRR11 expression exhibited resistance as suggested by both a high PEPI score and recurrence after 5 years of follow up. We employed MCF7 and HCC1428 LTED cells because 1) parental MCF7 and HCC1428 cells harbor high PRR11 copy number, 2) although initially sensitive to estrogen suppression, these cells always adapt to experimental estrogen suppression, and 3) PRR11 protein levels were elevated in LTED cells compared to parental cells (Rebuttal Fig. 5). We selected these cells, again, not to forge a link with estrogen sensitivity, but, because of their ability to adapt to estrogen suppression, to show that high PRR11 levels can mediate escape from estrogen suppression, in line with the thesis of the paper. In this revised manuscript, we have used ER+ cell lines (MDA-MB-134VII and MDA-MB-175VI) without high PRR11 copy number that are initially sensitive to antiestrogens and demonstrated that PRR11 overexpression promotes estrogen-independent growth and drives fulvestrant resistance, lending further support to our hypothesis that PRR11 amplification, within the 17q23 amplicon, mediates escape from the estrogen suppression.
As requested by the Reviewer, we have extended our studies to other breast cancer subtypes. We ablated PRR11 in two triple negative breast cancer (TNBC) cell lines, HCC38 and MDA-MB-231, and one HER2+ breast cancer cell line, BT474 (Rebuttal Fig. 7A; included as Supplementary Fig. 4g of the revised manuscript). HCC38 and BT474 cells harbor PRR11 high copy number and also express high level of PRR11 protein (Rebuttal Fig. 7B; included as Supplementary Fig. 2d of the revised manuscript). MDA-MB-231 cells express high levels of PRR11 protein (Rebuttal Fig. 7B). PRR11 knockdown with siRNA promoted cell proliferation in TNBC cells, which is discordant with the observation in ER+ breast cancer cells. In BT-474 cells, the anti-proliferative effect of PRR11 silencing was not significant (Rebuttal Fig. 7C; included as Supplementary Fig. 4h of the revised manuscript). PRR11 silencing modestly reduced p-AKT in MDA-MB-231 and BT474 cells (Rebuttal Fig. 7A). We have not been able to investigate the mechanism of the discrepancy in other subtypes of breast cancer as they would require a second study. Unfortunately, we are in no position to state that the mechanism linking PI3K signaling to cell viability is specific to ER+ breast cancers, as this would require many more cell lines and human tumors.
We would like to respectfully emphasize that the basis of this study was an observation in a clinical cohort of patients with ER+ breast cancer treated with estrogen suppression, not estrogen stimulation. Herein, we demonstrated that PRR11 promotes hyperactivation of the PI3K/AKT pathway, a signaling pathway that, when aberrant, can bypass hormone dependence in ER+ breast cancer (PMID: 20530877). Many drivers of endocrine resistance activate growth factor signaling pathways in order to bypass estrogen dependence. A recent study by Razavi and colleagues reported that mutations/alterations on ESR1, transcription factors, the MAPK pathway, and other unknown pathways account for 18%, 9%, 13% and 60%, respectively, of endocrine resistance in ER + breast tumors (PMID:30205045). Except for the mutations in ESR1 and possibly transcription factor alterations, many of these potential drivers contribute to the resistance via alternative growth promoting signaling pathways rather than modulation of ER signaling per se. For instance, acquired mutations in HER2 drive resistance to endocrine therapy via PI3K/mTOR activation (PMID:30314968). Thus, we recognize that PRR11 regulates cell proliferation independently of estrogen signaling per se. Further, the role of PRR11 in cell proliferation and poor outcome has been previously suggested in various cancer types that are not associated with hormone receptor signaling (lung, pancreatic and gastric cancer). (Fig. 7g,h)

The statements that 'PRR11-amplified cancers are highly sensitive to PI3K inhibitors and rely on PIK3CA' should be removed from both the abstract and text (lines 267-269). This also applies to the title of the paragraph 'PRR11-amplified breast cancer cells are dependent on the PI3K pathway' (line 253). The final conclusion of the abstract ('These data suggest PI3K inhibitors as a novel therapeutic strategy for PRR11-amplified ER+ breast cancers') should also be toned down. At best, it could be speculated that PRR11 amplification in ER+ breast cancer is a candidate genetic marker to aid patient selection for ERα targeted therapies.
We thank the reviewer for this insightful comment and agree that GR metrics assay will improve the accuracy of drug sensitivity assessment. To address this point, we reassessed the drug responses to fulvestrant, alpelisib and taselisib using the GR metrics calculator (Rebuttal Fig. 8). All assessments of drug sensitivity in Fig. 3g, 3h, 7e and 7f of the original manuscript were replaced to the GR metrics (GR50 and GR value curves) in the revised manuscript. We observed a trend similar to the previous analyses. Furthermore, we wish to emphasize that data in Fig. 7f-h of the revised manuscript clearly show that PRR11 overexpression promotes estrogen-independent growth, which is blocked by PI3K inhibitors. Finally, we did not mean to imply that PI3K inhibitors are a novel therapeutic strategy. We have now clarified that ER+/PRR11 amplified tumors represent a novel subgroup that may benefit from the combination of a PI3K inhibitor and an antiestrogen. Accordingly, we have amended 'These data suggest PI3K inhibitors as a novel therapeutic strategy for PRR11-amplified ER+ breast cancers' to 'These data suggest ER+/PRR11-amplified breast cancers as a novel subgroup that may benefit from treatment with PI3K inhibitors.' in the abstract.
Rebuttal Fig. 8. A-H. GR and GR50 values were calculated using the GR Metrics calculator (t-test). MDA-MB-134VI and MDA-MB-175VII cells stably transduced with pLX302-LacZ and -PRR11 were treated with multiple doses of fulvestrant for 6 days (A and B). MCF7 FulvR cells transfected with control or PRR11 siRNA were treated with multiple doses of fulvestrant for 6 days (C and D). MDA-MB-134VI and MCF10A cells stably transduced with pLX302-LacZ and -PRR11 were treated with multiple doses of alpelisib (E and F) or taselisib (G and H) for 6 days.

Figure 7a and extended Data Figure 8. It is not clear why MCF10A was chosen as a cell model for these critical experiments as it is not a breast cancer cell line (but an immortalized but not transformed cell line) and is also ER-negative. These experiments should instead be performed on genuine breast cancer cell line, both ER-positive and ER-negative.
We agree with the reviewer's concern. We utilized MCF10A cells because this is a cell line that relies on insulin for propagation and lacks a PIK3CA mutation. We now include data suggesting that estrogenindependent growth promoted by PRR11 overexpression requires PIK3CA in MDA-MB-134VI ER+ breast cancer cells (Fig. 7f of the revised manuscript). We have moved the results with MCF10A cells to Supplementary Data.

Other than the FISH data presented in Extended data Fig 2b, the Authors should aim to show evidence for PRR11 protein overexpression in human breast cancer tissue. This protein can only have a functional impact if it is indeed overexpressed at the protein level.
We conducted PRR11 immunohistochemistry (IHC) in 175 primary ER+ breast tumors treated with neoadjuvant letrozole (Rebuttal Fig. 9A; included as Fig. 1g of the revised manuscript). Based on the distribution of PRR11 protein levels, PRR11-positive and -high tumors were classified by PRR11 >1% and >15%, respectively (Rebuttal Fig. 9B; included as Supplementary Fig. 1b of the revised manuscript). As a result, 14.3 % and 4.6% of tumors were defined as PRR11-positive and PRR11-high, respectively. Of note, tumors with poor response to letrozole exhibited higher levels of PRR11 protein, supporting the functional impact of PRR11 on endocrine therapy response (Rebuttal Fig. 9C; included as Fig. 1h of the revised manuscript). Rebuttal Fig. 9. A. Representative PRR11 IHC images of ER+ primary breast tumors. B. Distribution of PRR11 positivity in ER+ primary breast tumors. C. Tumoral PRR11 protein levels were plotted by letrozole responses that were defined by on-treatment Ki67 levels as described in the previous report (PMID: 28794284). Fig. 7).

What is known about this protein? Even if nothing is known, this should be stated. I found it rather odd to have to read through the MS without having no information about this protein.
We apologize for the lack of clarity. We have included a brief description about PRR11 in the Introduction section of the revised manuscript as below.
"PRR11 has been implicated in poor outcome of various cancer types, but the molecular basis for this association is unclear."

Extended data Fig 2a: what statistical test was performed?
As we described in Method section, t-tests (Nonparametric tests) were used for the calculation of p value.

Extended data Fig 2b: a negative control is missing: the Authors should also show a FISH image of a cancer without PRR11 amplification.
We have included the FISH image of a primary breast tumor without PRR11 amplification (Rebuttal Fig.  10; included as Supplementary Fig. 2e of the revised manuscript).
Rebuttal Fig. 10. Representative FISH image from a breast tumor specimen with PRR11 amplification. Magnification = 100x.

Page 5: line 118: 'we listed 90 genes located in the 17g23 amplicon': it was not clear whether these 90 are all the genes found in this amplicon? If not, how were these genes selected?
We listed all genes located in 17q23 amplicon based on the Atlas of Genetics and Cytogentics (PMID: 23161685).

Extended Fig 3a, Legend: what does 'low density monolayers' mean here and in all other figure legends where it appears? The authors should be much more specific with regards to how their assays were performed when it comes to cell numbers, media changes and similar, as these are factors that heavily influence the reproducibility of drug testing assays and similar.
We apologize for the lack of clarity. Technically, clonogenic populations propagated from low number of cells (1,000 cells) seeded in 12-well plate (2-D) were considered as the 'low density monolayers'. The information of media change was described in the "Clonogenic assays" section of Methods.

The information on antibody and other reagents need additional detail, for example for western blots, proximity ligation assay etc.
We have included information of antibodies used in this study as supplementary material.

1) The authors suggest that there is a higher rate of PRR11 amplification in metastatic (data from the MBC project compared to METABRIC and TCGA) compared to ER+ primary breast tumors. Yet, their in vivo experiment only examined primary tumor growth. Does PRR11 promote tumor progression/metastasis in vivo? Either spontaneous or experimental metastasis assays could be done to address this question.
PRR11 has been associated with an invasive phenotype in hepatocellular and ovarian carcinoma (PMID: 30248355, 30165366), so investigating a role of PRR11 in breast cancer metastasis would be very interesting, but also a significant additional study. However, our study focused on the implication of PRR11 overexpression/amplification in endocrine resistance, another phenotype indicative of increased cancer virulence.

2) An in vivo study demonstrating the effects of PI3K inhibition on PRR11 amplified cells would greatly strengthen the clinical relevance of the manuscript. Additionally, would overexpression of PRR11 confer endocrine therapy resistance in vivo?
We thank to the Reviewer for these insightful suggestions and wholeheartedly share his/her wishes. Regarding the evaluation of endocrine therapy resistance promoted by PRR11 overexpression in vivo, unfortunately, we failed to find a feasible model due to limited tumorigenicity of ER+/PIK3CA WT /PRR11 non-amplified breast cancer cell lines (MDA-MB-134VI, MDA-MB-175VII, CAMA1 and HCC1500). We and others have previously shown that the estrogen-independent growth of PRR11-amplified MCF7 xenografts is blocked by PI3K inhibition (PMID: 22049316).

3) All of the studies were done with stable cell lines of endocrine resistance or LTED. Do PRR11 levels increase upon acute endocrine treatment or in parental cells? And What are the absolute protein levels of PRR11 across a panel of cell lines representing multiple subtypes of breast cancer?
PRR11 expression was not affected by estrogen deprivation or by tamoxifen treatment for 24 h (data not shown). Moreover, we had shown that E2 stimulation does not induce PRR11 mRNA expression (Supplementary Fig. 1e of the revised manuscript).
To address the reviewer's suggestion, we examined PRR11 protein levels in 11 breast cancer cell lines (Rebuttal Fig. 11A; included as Supplementary Fig. 2d of the revised manuscript). As we replied to Reviewer #1, PRR11 protein levels were higher in PRR11-amplified cell lines. We did not observe a breast cancer subtype-specific expression of PRR11 protein among these cell lines (Rebuttal Fig. 11A,  right). Further, we interrogated 56 breast cancer cell lines representing all subtypes in the CCLE dataset (PMID: 31068700), and found that PRR11 mRNA expression was similar across subtypes (Rebuttal Fig.  11B). Rebuttal Fig. 11. A. Lysates from breast cancer cell lines (n=11) were subjected to immunoblot analysis with PRR11 and actin antibodies. PRR11-amplified cell lines are highlighted in red. The intensity of PRR11 immunoblot bands was quantified using the Image Lab software (ver. 6.0, BioRad) and then plotted by molecular subtype in the right. B. PRR11 mRNA levels in breast cancer cell lines (n=56). The normalized expression data, log2 transcript per kilobase million (TPM), was obtained from the CCLE dataset.

4) Figure 4: The results presented show that knockdown of PRR11 results in a S phase entry defect of the cell cycle, demonstrating that PRR11 is needed for S phase entry. In addition, the authors show that there is decrease in CyclinD1 levels. Previous studies have shown that loss of PRR11 results in S phase arrest. This raises question whether PRR11 loss mediated changes in cell cycle is general to all cells and not specific to endocrine resistant variants? Does PRR11 in endocrine sensitive or TNBC cells also result in cell cycle arrest?
As we replied to Reviewer #2, PRR11 knockdown by siRNA did not suppress proliferation in triple negative and HER2+ breast cancer cells, although it suppressed the AKT phosphorylation (Rebuttal Fig.  7). These results imply that PRR11 does not regulate cell viability through the PIK3/AKT pathway in all cancer cells. Given that PRR11 has been associated with poor outcome in other various cancer types such as pancreatic, gastric and lung cancers, however, the role of PRR11 in cell cycle is unlikely to be limited to ER+ breast cancers. Nevertheless, we would like to emphasize that this is the first study demonstrating the role of PRR11 in anti-estrogen resistance.

5) Figure 5a: it is not clear why knockdown of PRR11 results in a decrease in p110a levels. Is the low pAKT levels simply due to decrease in p110a levels and not related to PRR11?
The interaction of the p85 subunit with the p110 catalytic subunit of PI3K stabilizes p110. Since PRR11 knockdown reduces the p110-p85 interaction as a consequence of interrupting p85 homodimer formation, the p110 protein would be destabilized upon PRR11 knockdown. Although we cannot rule out other mechanisms, the lower levels of the p110 catalytic subunit would result in reduced PI3K activation and p-AKT levels.

6) Figure 5b: It is not clear why overexpression of PRR11 fails increase the levels of p-AKT and p-GSK in MDA-MD-134VI cells. It is certainly not due to difference levels of PRR11 expression or due to endogenous AKT or GSK protein levels. Is this an indication that PRR1 has PI3K independent functions that needs to be addressed?
We respectively disagree that ectopic expression of PRR11 failed to increase the levels of p-AKT and p-GSK3β. To clarity this point, however, we reassessed and replaced the data with improved immunoblot images in Fig. 5B of the revised manuscript (Rebuttal Fig. 12).