Transient IGF-1R inhibition combined with osimertinib eradicates AXL-low expressing EGFR mutated lung cancer

Drug tolerance is the basis for acquired resistance to epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKIs) including osimertinib, through mechanisms that still remain unclear. Here, we show that while AXL-low expressing EGFR mutated lung cancer (EGFRmut-LC) cells are more sensitive to osimertinib than AXL-high expressing EGFRmut-LC cells, a small population emerge osimertinib tolerance. The tolerance is mediated by the increased expression and phosphorylation of insulin-like growth factor-1 receptor (IGF-1R), caused by the induction of its transcription factor FOXA1. IGF-1R maintains association with EGFR and adaptor proteins, including Gab1 and IRS1, in the presence of osimertinib and restores the survival signal. In AXL-low-expressing EGFRmut-LC cell-derived xenograft and patient-derived xenograft models, transient IGF-1R inhibition combined with continuous osimertinib treatment could eradicate tumors and prevent regrowth even after the cessation of osimertinib. These results indicate that optimal inhibition of tolerant signals combined with osimertinib may dramatically improve the outcome of EGFRmut-LC.


Reviewers' comments:
Reviewer #1 (Remarks to the Author): This manuscript reports that transient inhibition of the RTK IGF-1R is an effective potential approach to eliminate EGFR mutant lung cancer cells that are AXL-low yet survive to seed osimertinib tumors. A mechanism for IGF-1R induction is provided, centering on FOXA1.
The manuscript is clear. The findings are interesting and could be clinically relevant. There are several issues that diminish enthusiasm, as described below.
1-Which other RTKs were activated upon osimertinib treatment in the EGFR mutant cell lines?

Answer:
As indicated by the reviewers, phosphorylation of MET (HGF-R, point C3, C4) and FGFR3 (point B13, B14) was increased in H3255 and HCC4006 cells, respectively, following osimertinib treatment, as shown in Fig 1E. We evaluated the phosphorylation of MET and FGFR3 by western blotting. However, we detected no discernible increase in phosphorylated MET or FGFR3 in H3255, HCC4006, or HCC827 cells treated with osimertinib for 72 h (A). We further assessed the effect of a MET inhibitor (crizotinib) and FGFR inhibitor (BGJ398) on the osimertinib sensitivity of H3255 and HCC4006 cells, respectively, using the MTT assay. Although crizotinib and BGJ398 inhibited the phosphorylation of MET and FGFR3 in H3255 and HCC4006 cells, respectively, neither crizotinib nor BGJ398 remarkably affected the osimertinib sensitivity of H3255 (B, C) and HCC4006 (D, E) cells, respectively. These results indicate that MET and FGFR3 are unlikely to play predominant roles in mediating osimertinib sensitivity in EGFR mutated lung cancer cells tested, at least, in our experimental conditions. These results were included in the Results section (p6, lines 11-21) and Supplementary Figure 3.
2-What is the mechanism by which osimertinib induces FOXA1? What is the pathway and factor that is responsible?

Answer:
We would like to thank the reviewer for this scientifically relevant question. We performed additional experiments and uncovered the mechanism by which FOXA1 could be upregulated following osimertinib exposure.
In contrast to the IGF-1R expression results shown in Supplementary Fig. 4A of the original version, FOXA1 induction following osimertinib exposure was not impacted by cycloheximide treatment, indicating that FOXA1 upregulation by osimertinib does not require de novo protein synthesis ( Figure 3G). We hypothesized that pre-existing signaling proteins or pathways might be responsible for the induction of FOXA1 mRNA by osimertinib. Accordingly, we observed that osimertinib-dependent FOXA1 induction was significantly inhibited in IGF-1R knockout HCC827 cell clones ( Figure 3H). These results suggested that endogenous IGF-1R protein was involved in the signal transduction activating FOXA1 mRNA expression following osimertinib exposure. These new data indicated that osimertinib exposure activated FOXA1 expression through the signaling pathway comprising endogenous IGF-1R protein. Then, FOXA1 induced the transcriptionally more active epigenetic status of the IGF-1R gene, resulting in the positive feedback activation of IGF-1R in HCC827 cells ( Figure 3J).
These findings are now stated in the revised manuscript (p8, lines 3-32). Additionally, these new findings were included in Figure 3. The schema was also revised based on these findings ( Figure   3J).

3-
The findings are very similar to those originally reported by in the initial manuscript describing drug tolerant persister cells in cancers under TKI treatment (PMID: 20371346).
How do the authors move beyond that prior study? Are there epigenetic therapies that can combat the osimertinib tolerance, similar to those employed in that published work?

Answer:
As highlighted by the reviewer, IGF-1R involvement in the emergence of tolerant cells to first-generation EGFR-TKIs has been previously reported. Sharma et al have shown that first-generation EGFR-TKIs stimulated the expression of KDM5A, a histone demethylase, resulting in increased IGF-1R phosphorylation, associated with IGF-BP3 overexpression, and thereby inducing drug-tolerant cells. Furthermore, they reported that the combined use of the IGF-1R inhibitor, AEW541, and gefitinib inhibited the emergence of drug-tolerant cells in vitro. However, these findings were mainly observed using PC-9 cells, which express a high level of AXL (please see In the present study, we demonstrated novel findings focused on AXL-low expressing EGFR mutated NSCLC, which are more sensitive to EGFR-TKIs when compared with AXL-high expressing EGFR mutated NSCLC. In this population of cell lines, tolerance to the clinically available third-generation EGFR-TKI osimertinib is induced via increased IGF-1R protein expression. We further uncovered the mechanism by which osimertinib increased IGF-1R expression. Moreover, we demonstrated the efficacy of a new therapeutic strategy, the transient combination of IGF-1R inhibition with continuous osimertinib treatment, revealing remarkable improvements in the outcome of AXL-low expressing EGFR mutated NSCLC. We believe that these novel findings are crucial in considering therapeutic strategies that cure or dramatically improve the prognosis of EGFR mutated lung cancer.
Regarding epigenetic therapy, the development of drugs that inhibit FOXA1 might be necessary.
We aim to develop FOXA1 inhibitors and report their efficacy on osimertinib tolerance in future reports.
4-What is the mechanism of constitutive adaptor protein binding to IGF-1R? This result seems counterintuitive based on current knowledge. Furthermore, what is the functional significance of the EGFR/IGF-1R interaction that appears to be stable? Does knockdown of EGFR impair the ability of IGF-1R to transduce signaling and promote drug tolerance?

Answer:
Thank you for this important question. As indicated by the reviewer, recent review articles have presented the importance of ligand binding for adaptor protein binding to IGF-1R as follows, "Ligand binding induces a conformational change that activates the kinase domain of IGF-1R resulting in autophosphorylation of specific tyrosine residues, which appears to be the critical step in receptor activation. This in turn leads to recruitment and phosphorylation of the docking proteins insulin receptor substrates (IRS-1/2) and Shc, ultimately resulting in the activation of multiple signaling pathways, of which the two most well-characterized are (PI3K-AKT) and RAS-MAPK." (Osher E and Macaulay VM, Cells 2019; 8, 895) Conversely, other review articles have presented the crosstalk between IGF-1R and other receptor tyrosine kinases (RTKs), including EGFR (Liu C et al, Curr Pharm Des 2014;20:2912-21). Therefore, even in the steady-state, crosstalk with RTKs containing EGFR may induce a certain activation of IGF-1R and bind scaffold proteins. We performed additional experiments to examine the interaction between IGF-1R and mutated EGFR protein, using exon19 deleted EGFR specific antibody (EGFR-del19). We observed that osimertinib increased the binding of IGF-1R to total EGFR protein, as well as mutated EGFR protein. These new findings indicate that IGF-1R is associated with mutated EGFR, irrespective of osimertinib exposure ( Figure 1A for reviewer only).
As suggested by the reviewer, we also performed additional experiments with EGFR knockdown to assess the ability of IGF-1R on promoting drug tolerance. We successfully knocked down EGFR expression in HCC4006 cells by siRNA for EGFR ( Figure 1B for reviewer only). The knockdown of EGFR expression resulted in decreasing cell viability, and linsitinib (IGF-1R inhibitor) further inhibited the viability ( Figure 1C for reviewer only). However, since HCC4006 cells treated with siRNA for EGFR still expressed a discernible level of EGFR, we could not directly evaluate the ability of IGF-1R in the absence of EGFR on promoting osimertinib tolerance.
5-Is IGF-1R expression and phosphorylation and FOXA1 expression associated with worse response to osimertinib in EGFR mutant lung cancer patients?

Answer:
We agree that the relevance of the target molecules needs to be demonstrated using clinical specimens. However, in this study, it is currently challenging to present the relevance of IGF-1R and FOXA1 expression in clinical specimens. Our in vitro data indicated that the expression of IGF-1R and FOXA1 was upregulated after osimertinib exposure ( Figure 1F and Supplementary Table 1).
For example, the level of constitutively expressed IGF-1R did not correlate with osimertinib sensitivity in EGFR mutated lung cancer cell lines ( Figure 1B). We collected clinical specimens obtained before osimertinib treatment ( Figure 1C). To demonstrate the induction of IGF-1R and FOXA1, clinical specimens obtained a few days after osimertinib treatment initiation are required, but could be impossible in clinical practice.
We stated these findings in the Results section (p6, lines 22-31). As indicated by the reviewer, data for phosphorylated EGFR and total EGFR were missing in Figure   3B and D. Hence, we added the data for total EGFR and phosphorylated EGFR. Figure 3D, in the original version, is now listed as Figure 3E because, as requested by Reviewer #2, we inserted new data as Figure 3D in the revised version.  Cortot et al have reported that IGF-1R activation as a result of the loss of IGFBP3 expression, via methylation of its promotor lesions, is involved in the emergence of tolerant or resistant cells to EGFR-TKIs, PF299804 and WZ4002. Furthermore, they have reported that the combined use of IGF-1R inhibitors, BMS536924 and OSI-906 (Linsitinib), restored sensitivity to EGFR-TKIs in vitro.
These findings were observed in resistant cells obtained from PC-9 cells, expressing a high level of AXL (please see Figure 1B in our paper). Moreover, the efficacy of combined treatment with EGFR-TKI and IGF-1R inhibitors in vivo has not been reported.
In the present study, we revealed novel findings focusing on AXL-low expressing EGFR mutated NSCLC, which are more sensitive to EGFR-TKIs when compared with AXL-high expressing EGFR We added these statements in the Discussion section (p11 line 27-p12 line 1, p12 lines 6-9, and p12 lines 32-33). .

Additional Comments
2. Fig 2 -Would be more convincing using multiple IGF1R CRISPR-CAS9 KOs given potential off target effects of IGF1R siRNAs.

Answer:
As suggested by the reviewer, we established IGF-1R knockout clones for HCC827 cells using CRISPR-CAS9 for two different sites of IGF-1R (KO1 and KO2) (Supplementary Figure 6).
These three clones were more sensitive to osimertinib when compared with parental HCC827 cells.
As these clones were markedly sensitive to osimertinib and protein was harvested from viable cells after 72 h osimertinib exposure, signal transduction in cells was examined after 2 h osimertinib exposure by western blotting. In the IGF-1R knockout clone (KO1-6), osimertinib inhibited the phosphorylation of GAB1, Shc, AKT, and ERK more remarkably when compared with HCC827 cells.
We added the data and statements for IGF-1R CRISPR-CAS9 in Figure 2 D, E, and F. and in the Results section (p7, lines 12-17), respectively. We also added the statements for methods of "generation of IGF-1R knockout HCC827 cells by CRISPR/Cas9 system" in the Methods section (p15, line 26 -p16, line 9). We added the reference for CRISPR/Cas9 system as reference 42.
Accordingly, we renumbered references 43 and 44 in the revised version of manuscript. Since we obtained CRISPR/Cas9 plasmid from Dr Ryo Imamura, we stated his name in the Acknowledgements.

Answer:
We agree with the reviewers' comments. The Associate Editor kindly mentioned that "if this is not possible please include additional shRNAs for validation." Therefore, we examined the effect of two additional shRNAs. Overall, three different shRNA for FOXA1 consistently suppressed the increased IGF-1R expression and its phosphorylation. Additionally, they inhibited colony formation of HCC827 cells in the presence of osimertinib. We added the data regarding the additional two FOXA1-shRNA in Figure 3 C and F. 4. To test whether AXL directly mediates this altered dependency on IGF1R, functional experiments examining the impact of over-expressing AXL in AXL-low cells would also significantly strengthen the manuscript, especially since this is the only novel point compared with prior work.

Answer:
As suggested, we attempted to overexpress AXL in AXL-low tumor cells. Although we obtained AXL overexpressing HCC827 cells, as well as AXL overexpressing PC-9 cells as a control, we were unable to obtain AXL overexpressing H3255 or HCC4006 cells, even on attempting overexpression using two different expression vectors (pIRESpuro and pEZ-Lv105). The transfection of AXL, but not the empty vector, was extremely toxic to H3255 cells. In HCC4006 cells, although tumor cells survived after transfection, the tumor cells failed to overexpress AXL. Therefore, we evaluated osimertinib sensitivity using HCC827 cells, as well as PC-9 cells as a control, transfected with or without AXL overexpression ( Figure 3A for reviewer only). In PC-9 cells, AXL overexpression discernibly increased sensitivity to osimertinib ( Figure 3B for reviewer only). In HCC827 cells, although AXL overexpression associated with AXL phosphorylation was successfully induced, it failed to explicitly affect osimertinib sensitivity ( Figure 3C for reviewer only). Interestingly, AXL overexpression reproducibly inhibited IGF-1R protein expression in HCC827 cells ( Figure 3A for reviewer only). Therefore, in HCC827 cells, AXL overexpression might suppress its effect on osimertinib sensitivity by decreasing IGF-1R protein expression, which may be induced by an unknown mechanism and potentially sensitize cells to osimertinib. These data are interesting, and the precise mechanisms should be intensively analyzed; however, we believe that this is beyond the scope of this paper. We would like to further analyze the mechanisms and report these results in future reports.