Structure-guided design of a selective inhibitor of the methyltransferase KMT9 with cellular activity

Inhibition of epigenetic regulators by small molecules is an attractive strategy for cancer treatment. Recently, we characterised the role of lysine methyltransferase 9 (KMT9) in prostate, lung, and colon cancer. Our observation that the enzymatic activity was required for tumour cell proliferation identified KMT9 as a potential therapeutic target. Here, we report the development of a potent and selective KMT9 inhibitor (compound 4, KMI169) with cellular activity through structure-based drug design. KMI169 functions as a bi-substrate inhibitor targeting the SAM and substrate binding pockets of KMT9 and exhibits high potency, selectivity, and cellular target engagement. KMT9 inhibition selectively downregulates target genes involved in cell cycle regulation and impairs proliferation of tumours cells including castration- and enzalutamide-resistant prostate cancer cells. KMI169 represents a valuable tool to probe cellular KMT9 functions and paves the way for the development of clinical candidate inhibitors as therapeutic options to treat malignancies such as therapy-resistant prostate cancer.

Editorial Note: This manuscript has been previously reviewed at another journal that is not operating a transparent peer review scheme.This document only contains reviewer comments and rebuttal letters for versions considered at Nature Communications.Mentions of prior reports have been redacted.
Reviewer #1 (Remarks to the Author): General: The authors have addressed most of the raised points.Please consider to remove "first" in the title as it is slightly controversial.
As suggested, we removed 'first' in the title (page 1, line 1) and throughout the text (page 4, line 12, page 10, line 24 and page 12, lines 4 and 11).
1.) The authors claim 14 additional ligand-bound KMT9 structures of close analogues supporting there modeling hypothesis.At least one example should be deposited, discussed, and at included in the supplementary section, it would be very helpfull to follow the argumentation.
We followed the reviewer's suggestion and deposited two representative and informative ligand-bound KMT9 structures (PDB ID: 8QDG and 8QDI), included them in the supplementary section (new Extended Data Fig. 2a-f) and describe the additional structures in the results (page 5, lines 21-28 and page 6, lines 1-3).Together, the ligand-bound KMT9 structures show an overall high rigidity of the pocket in the presence of analogues with different substrate moieties, which fully corroborates the modelling approach applied for our compound series 2 to 4.

2.) and 3)
Most of the issues have been addressed.However, the Ctrl compound should be profiled in the proliferation assays side by side (Figure 4e).The two insensitive cell lines (HepG2 cells, PANC-1 cells) should be included in figure 4e.
We profiled KMI169ctrl in the proliferation assay side by side and calculated the GI 50 of KMI169 in HepG2 and PANC-1 cells.We included the data in the revised Fig. 4e.In the revised manuscript (page 10, lines 20-23) we state: "Despite cellular target engagement in both cell lines (Extended Data Fig. 4f-i), KMI169 did not affect proliferation of HepG2 or PANC-1 cells at concentrations that were effective in responsive cell lines (Fig. 4e, and Extended Date Fig. 4 j, k)".
Reviewer #2 (Remarks to the Author): Wang et al. report in the manuscript "Structure-guided design of the first selective inhibitor of the methyltransferase KMT9 with cellular activity" the discovery of a novel inhibitor of the methyltransferase KMT9.and based on limited cancer cell viability and gene expression data, advertise it as the first cellular probe to dissect catalytic and non-catalytic functon(s) of KMT9.While the described experiments are well executed and have merit, the advancement is incremental and the cell-based data too preliminary to warrant publication in Nature Communications.
1.The conceptual novelty is limited given that bi-substrate methyltransferase inhibitors for KMT9 have been reported previously.
We acknowledge that previously a NTMT1 bi-substrate inhibitor was also found to bind KMT9.The major achievement of our study was to develop a potent and selective KMT9 inhibitor with cellular activity starting from an initial hit.In addition, we characterized this novel inhibitor in cellular assays allowing us to distinguish between catalytic and scaffolding functions of KMT9.In our opinion, these are novel and major advancements compared to the previously shown non-selective, non-cell permeable NTMT1 bi-substrate inhibitors, which are not suitable for the characterization of KMT9 biology.While individual experiments are certainly not sufficient to substantiate claims, the sum of all experiments and controls provides, in our opinion, strong evidence for our conclusions.Nevertheless, we acknowledged in the discussion "Finally, it cannot be excluded that gene expression changes upon KMT169 treatment are not solely caused by KMT9 inhibition" (page 13, lines 14-16).

In the revised manuscript the authors have strengthened data addressing
4. The questions around the limited overlap of KMT169-induced and KMT9 siRNAinduced gene expression changes remain.The authors use somewhat circular logic since they say the differences may by due to inhibition versus loss of protein by use of KMT169 and KMT9 specific siRNAs respectively.However, they use the concordance of KMT9 siRNA and KMT169 mediated growth phenotypes in PC-3M and the absence thereof in PANC1 and HepG2 as evidence for mechanism-based phenotypic effects.
As [REDACTED] added in the discussion (page 13, lines 9-15), partial overlaps for DEGs af ter inhibitor treatment and knockdown of the gene of interest appear to be commonly obser ved and have been documented in previous publications.(for example: Extended Data Fig. 5a-b, Yankova, E Cell Biol. 24, 384-399 (2022).The reasons for the limited overlap of KMI169-and KMT9 siRNA-mediated gene expression changes are multiple and might be due to differing experimental conditions optimized for the respective purpose (96 hours inhibitor treatment vs. 72 hours siRNA treatment).In fact, the response to inhibitor seems to be relatively slow since we only observed few differentially expressed genes after 72 hours of treatment.In addition, the efficiencies of knockdown and inhibition may differ.Assuming, for example, that KMT9 inhibition is very efficient, whereas the knockdown efficiency is 'only' in the range of e.g.80-90%, the remaining KMT9 protein may cause gene expression not observed upon (efficient) inhibition (and vice versa).Our entire chain of argumentation includes in vitro and in vivo selectivity tests, the use of a structurally related control compound (KMI169ctrl), assays in control cell lines, as well as comparisons of gene expression upon KMT9 inhibition and knockdown.While a single experiment would certainly not allow conclusions about KMI169 selectivity, the complete set of experiments substantiates our claims.We acknowledge, however, that we cannot exclude that a small percentage of differentially expressed genes might be the result of off-target activity of either siRNA or inhibitor.However, such potential 'side-effects' effects do not affect the overall conclusions of the manuscript.Accordingly, in the revised manuscript we now state "Finally, it cannot be excluded that gene expression changes upon KMT169 treatment are not solely caused by KMT9 inhibition" (page 13, lines 14-16).2E the interaction of the azetidine moiety to residue Y125 is highlighted as a key interaction, however mutation of which does not affect compound binding.

[REDACTED] in Fig
Fig. 2E shows key interactions of KMT9 with KMI169 around the azetidine moiety.The figure depicts three main contacts in this region, while in other parts of the ligand binding pocket there are multiple additional protein-inhibitor contacts.Since Y125A compromises only one out of multiple ligand contacts, the two-fold reduced inhibitor binding of the mutant (K d = 50 nM for KMI169 binding to Y125A vs. K d = 25 nM for KMI169 binding to wildtype KMT9) lies actually within the expected range.We did not expect Y125A to drastically or fully abolish inhibitor binding as possibly assumed by the referee.

Figure 1 :
Figure 1: Venn diagram showing the intersection of cell cycle genes downregulated upon treatment of PC-3M cells with KMI169 and RNAi-mediated KMT9a knockdown.