The manuscript „Nintedanib overcomes imatinib induced KIT mutations and FGFR signaling upregulation mediated multiple drug resistances” by Liu and colleagues, describes the effects of the tyrosine kinase inhibitor nintedanib in models of gastrointestinal stromal tumor (GIST) in vitro and in vivo. First, the authors conducted a compound library screen in KIT-driven BaF3 cell line panel und thus identified nintedanib to have antiproloferative effects. The authors then evaluate the effects of this TKI in several in vitro assays and compare them to imatinib (IM) and other KIT-inhibitory drugs in non-GIST and GIST models. They emphasize the inhibitory effect of nintedanib in FGFR2 signaling. They show that effects comparable to the approved drugs IM and sunitinib (SU) can be achieved in some IM-sensitive and IM-resistant models, respectively in vivo. The authors conclude that nintedanib, as an approved drug should be further developed for IM-resistant GIST.

Although certainly in interesting study subject, the authors might overestimate the role of the T670I mutation and the role of FGFR signaling. Although they authors present a sufficient body of work to support their hypothesis, I feel the study has several serious flaws.
Choice of models:
- In the introduction, they state that “A variety of factors lead to the emergence of imatinib resistance, with the most common being secondary drug-resistance-conferring mutations in the KIT kinase domain, especially the T670I “gatekeeper” mutation [10, 11].”
o This implies that T670I would be the most prominent resistance mutation. This is neither supported by their references, nor by clinical reality. (E.g. compare DOI: 10.1200/JCO.2006.06.2265 ; doi: 10.1158/1535-7163.MCT-18-1224) . It may thus justify their most prominent model/mutation T670I
- In the first part of the result + Fig. 1, the authors use BaF3 models. However most of them do not reflect clinical reality in GIST: Resistance mostly arises from secondary mutations in the kinase domain or activation loop. However, there are only two models actually harboring double mutations. All primary muts. In kinase domain or activation loop are either very rare or nonexistent in GIST. Furthermore primary e17 mutations do display different sensitivity profile e.g. to IM than ex11+e17 mutated GIST.
o Therefore all conclusions drawn from these experiments may be false leads.
o Why where the Baf3 T670I-only cells used in vivo, why were only those (not being really representative (see above)) used for western blot?
- GIST models: Why were those specific models chosen?
o Why are there two T670I-mutated models? What’s the difference? The origin of “GIST-T1-T670I” is unclear as neither the cell line itself, nor its CRISPR/Cas9 origin is found in the given ref. 30, nor in the literature.
o As ex17 double mutants are available (e.g. 10.1158/1078-0432.CCR-14-1397), it would have been important to test these.  To draw conclusions about activation loop mutations.
- KIT –WT models:
o Why were KIT-WT BaF3, as well as a WT patient culture chosen as models? Generally KIT-WT inhibition would be considered as an unwanted off-target effect, which may cause hematologic and other toxicities. So called “wild type GISTs” are not actually driven by WT-KIT and therefore do not need to undergo KIT-inhibitor evaluation.

Rationale:
- One gets the impression the authors are exaggerating in some points:
o The importance of the T670I mutation (see above)
o About Avapritinib and ripretinib they state: “However, their clinical applications were limited by severe side effects such as intracranial hemorrhage, severe headache, and risk in secondary tumor induction [17, 18].” , while ava and rip are rather well tolerated drugs and rip is being evaluated to supplant sunitinib in second line.
Other:
- Why was regorafenib not used in this study?
- Why were only subsets of the drugs used in Fig. 1 used for consecutive experiments?
- The heatmap in Fig.1A/B uses arbitrary color coding: What does green actually mean? If green implies sensitivity, most mutations would be sensitive to most drugs. What are the thresholds, how were they chosen? Why do they differ between A and B?
- In fig. 4, much higher concentrations of nin and IM are used. Why? The combined (?) sequential(?) treatment is not explained well.
o One key question about FGFR-signalling induction is: If this effect is induced so rapidly, why is IM still effective for so long in vitro, in vivo and in patients?
o It would have been interesting to see whether the broader spectrum TKIs Su and regorafenib would be able to inhibit FGFR/ERK signling at comparable concentrations.
- What does the H2AX observation tell us? It is mentioned in the text and shown in a figure but not further commented or discussed.
- The statement “ Substitution mutations, such as V559A/D/G and L576P, are located in the
juxtamembrane domain where they can stabilize autoinhibited KIT and thus suppress its kinase activity” is not correct.
- The statement “Acquired on-target mutations of KIT kinase and alternative signaling pathway reactivation are two major resistance mechanisms.” Is somewhat imprecise or does not convey what the authors are trying to say. More in line with literature might be “Acquired on-target mutations of KIT kinase is the major resistance mechanism.” alternative signaling pathway activation, may play a role.
- The authors should consider that VEGFR inhibition by nintedanib might have additional impact on in vivo growth.

All my queries have been answered sufficiently.