Disruption of FOXO3a-miRNA feedback inhibition of IGF2/IGF-1R/IRS1 signaling confers Herceptin resistance in HER2-positive breast cancer

Resistance to Herceptin represents a significant challenge for successful treatment of HER2-positive breast cancer. Here, we show that in Herceptin-sensitive cells, FOXO3a regulates specific miRNAs to control IGF2 and IRS1 expression, retaining basic IGF2/IGF-1R/IRS1 signaling. The basic activity maintains expression of PPP3CB, a subunit of the serine/threonine-protein phosphatase 2B, to restrict FOXO3a phosphorylation (p-FOXO3a), inducing IGF2- and IRS1-targeting miRNAs. However, in Herceptin-resistant cells, p-FOXO3a levels are elevated due to transcriptional suppression of PPP3CB, disrupting the negative feedback inhibition loop formed by FOXO3a and the miRNAs, thereby upregulating IGF2 and IRS1. Moreover, we detect significantly increased IGF2 in blood and IRS1 in the tumors of breast cancer patients with poor response to Herceptin-containing regimens. Collectively, we demonstrate that the IGF2/IGF-1R/IRS1 signaling is aberrantly activated in Herceptin-resistant breast cancer via disruption of the FOXO3a-miRNA negative feedback inhibition. Such insights provide avenues to identify predictive biomarkers and effective strategies overcoming Herceptin resistance.

Reviewer #2: Remarks to the Author: In this manuscript the authors report that resistance to Herceptin treatment of HER2+ breast cancers could be due to activation of the IGF2/IRS1 signal because of disruption of a negative loop between FOXO3a-miR-128 and miR-30 There are several technical problems associated with this study General comments Supplementary figures all lack statistical tests, this issue should be addressed. It is unclear why main figures deserve statistic but not supplementary material.
Western blotting mostly show signals in saturation, therefore especially for loading controls such as b-actin, is difficult to assess whether the levels of the proteins loaded is truly equal amongst the various lanes. This is true for most of the western blottings shown. The authors should show western blotting images containing signal after lower exposures, when signals have not yet reached saturation. This is very important and would permit to better assess the quality of the data. Another issue is that the authors only show a single western blotting for each experiment in all cases. It would be more appropriate to show a representative western blotting as well as graphs of densitometric scanning ratios between target and loading controls derived from at least three independent experiments, thoroughly. In addition to this, statistical test should be performed to provide significance evidence of the changes shown.

Specific comments
Experiments in supplementary figure 1 lacks statistic. Also, in supplementary figure 1b and c it would be more appropriate to show data in combined box/whisker plot instead of barplot to have a better idea of the experimental variations.
The authors have used shRNA against IRS1 to test its role in Herceptin resistance and mTOR activation. It would be appropriate to also use CRISPR/CAS9 against IRS1 in these cell lines as an additional tool to test reproducibility of IRS1 resistance phenotypes It is generally unclear how many times the western blotting experiments have been done. It would be appropriate to also provide densitometric scan of the bands and to plot the average of these values divided by actin control accompanied to standard deviations and statistical tests. Supplementary figure 3a. The authors should also measure miRNAs that do not change as control in addition to miR-128-3p and miR-30a-5p. Figure 3a, b-actin signal is in saturation, difficult to assess whether inhibition of miR-128-3p and miR-30a-5p really increases IRS1 expression. Representative western blotting with lower exposure intensity should be shown and average of three independent experiment and p-values should be plotted. Also, how can the authors can explain that in figure 3b there is not increase in IRS1 protein levels upon miRNA inhibition? (For example in figure 3b comparing lane 2, with lane 5, 8 and 11 does not seem to be any difference in IRS1 expression). Additionally, what happens to IRS1 mRNA upon miRNA ectopic modification? What happen in cells where the genomic loci expressing miR-30 and miR-128-3p are removed by CRISPR/CAS9? Figure 4b. In addition to miR-128-3p and miR-30a-5p control miRNAs that do not change should be shown.
Supplementary figure 5d, why miR-126 expression is shown and why changes expression of this miRNA is similar to miR-128/30b and 193? Maybe a general process regulating general miRNA biogenesis could explain these effects here? Is there any miRNA that do not change in these conditions?
Reviewer #3: Remarks to the Author: In the present article Luo L et al, describes the role of FOXO3a-miRNA in the control of IGF2/IGFR1/IRS1 axis in relation to resistance to trastuzumab. The article is well developed, and mechanistically is well executed, including in vitro and in vivo models. However, the major limitation is the fact that the novelty of the findings described are not extremelly new. The role of the IGF2/IGFR1/IRS1 axis in resistance to trastuzumab is well described and documented.

Responses to Reviewer #1:
The manuscript discovered a novel mechanism of Herceptin resistance ……. Overall the findings are interesting and novel. The authors should address the following issues: Response: We thank the reviewer for his/her generous comments of research work.
1. In Fig 3e, they found that FOXO3a binds to the promoter region of specific microRNAs, then promotes the expression of these microRNAs. Since PPP3CB promotes transcriptional activity of FOXO3a, it would be interesting to examine whether knockdown of PPP3CB will change the binding of FOXO3a to the promoter of these microRNAs.

Response:
We appreciate the reviewer for this constructive criticism. According to the suggestion, we performed new experiments with specific shRNAs to downregulate PPP3CB expression. Our data revealed that specific knockdown of PPP3CB not only dramatically increased the levels of p-FOXO3a in both SKBR3 and BT474 cells treated with high concentration of rhIGF2 (80 ng/ml), and it also significantly abolished rhIGF2 (80 ng/ml)mediated enrichment of FOXO3a at the promoters of miR-128-3p and miR-30a-5p. These new findings are now shown in figure 4e and 4f, respectively, in the revised manuscript.
complex at transcriptional level in the resistant cells. Thus, our studies demonstrated that PPP3CB was translationally regulated by mTOR-mediated negative feedback regulation of IGF2/IRS1/mTOR signaling in sensitive cells. However, the STAT6/HDAC1 complex in the resistant cells transcriptionally suppressed PPP3CB, thereby conferring a constitutive activation of IGF2/IRS1/mTOR signaling. 5. The study showed that PPP3CB is an important regulator of FOXO3a. In Fig 6a, they detected higher expression levels of PPP3CB in sensitive cells than resistant cells, so some experiments can be designed to check whether sensitive cells expressing PPP3CB knockdown will become resistant to Herceptin and overexpression of PPP3CB will sensitize resistant cells to Herceptin.
Response: Thank the reviewer for this helpful suggestion. We performed additional experiments and found that ectopic expression of PPP3CB re-sensitized the resistant cells to Herceptin. In contrast, specific knockdown of PPP3CB elicited the sensitive cells becoming resistant to Herceptin. These data are now shown in Fig. 6b and supplementary figure 6a in the revision.
6. For mouse model experiment in Fig 7, expression level of p-FOXO3a, IRS1 and p-Akt should be tested by immunohistochemistry analysis.

Response:
We performed immunohistochemistry assays to examine the levels of p-FOXO3a, IRS1, and p-Akt in the tumor tissues obtained from our animal experiments. IRS1 knockdown or deletion markedly decreased the levels of p-Akt and p-FOXO3a in vivo (Supplementary figure  7c), confirming that p-Akt/p-FOXO3a acting as the downstream of IRS1 in resistant cells.

Responses to Reviewer #2:
In this manuscript the authors report that resistance to Herceptin treatment of HER2+ breast cancers could be due to activation of the IGF2/IRS1 signal because of disruption of a negative loop between FOXO3a-miR-128 and miR-30. There are several technical problems associated with this study.
General comments 1. Supplementary figures all lack statistical tests, this issue should be addressed. It is unclear why main figures deserve statistic but not supplementary material.

Response:
We appreciate the reviewer for his/her helpful comments. Statistical analyses on majority of the supplementary data have been performed and are now included in the revised Supplementary figures.
2. Western blotting mostly show signals in saturation, therefore especially for loading controls such as b-actin, is difficult to assess whether the levels of the proteins loaded is truly equal amongst the various lanes. This is true for most of the western blottings shown. The authors should show western blotting images containing signal after lower exposures, when signals have not yet reached saturation. This is very important and would permit to better assess the quality of the data.

Response:
Thank you for the kind suggestions. We repeated a number of our western blot assays, and took a shorter exposure time during film development. Some of new data with lower exposures are now included in the revised manuscript.
3. Another issue is that the authors only show a single western blotting for each experiment in all cases. It would be more appropriate to show a representative western blotting as well as graphs of densitometric scanning ratios between target and loading controls derived from at least three independent experiments, thoroughly. In addition to this, statistical test should be performed to provide significance evidence of the changes shown.
Response: All of the western blot assays were repeated at least three times, and sometimes by two independent lab people. We fully understand the reviewer's concern about the quantification issue of our western blotting. We have performed densitometric analysis on the western blot gels. Since we have a huge amount of data and each figure contains multiple panels, the space limitation becomes problematic. Thus, we have organized the data of our densitometric analysis on the key proteins into an Excel file, which is now shown as "Source Data" in the revised manuscript. I hope that this arrangement can satisfy the reviewer. 5. The authors have used shRNA against IRS1 to test its role in Herceptin resistance and mTOR activation. It would be appropriate to also use CRISPR/CAS9 against IRS1 in these cell lines as an additional tool to test reproducibility of IRS1 resistance phenotypes.
Response: It was greatly appreciated for this constructive suggestion. We utilized CRISPR/Cas9 gene editing technology to knockout IRS1 and performed additional assays, including cell viability examination upon Herceptin treatment, western blot detection, and in vivo animal experiments. We are extremely happy that our new data (Fig. 1d, Fig. 1e, and Fig. 7b in the revised manuscript) are able to confirm the reproducibility of IRS1-mediated resistance phenotypes.
6. It is generally unclear how many times the western blotting experiments have been done. It would be appropriate to also provide densitometric scan of the bands and to plot the average of these values divided by actin control accompanied to standard deviations and statistical tests.