Structure-based Discovery of a Series of NSD2-PWWP1 Inhibitor

Overexpression, point mutations or translocations of protein lysine methyltransferase NSD2 was occurred in many types of cancer cells. Therefore, it was recognized as oncoprotein and considered as a promising anticancer drug target. NSD2 consists of a SET catalytic domain and two PWWP domains binding to methylated histone proteins. Here, we reported our efforts to develop a series of NSD2 PWWP1 inhibitors, and further structure-based optimization resulted a potent inhibitor 38, which has the high selectivity towards NSD2 PWWP1 domain. The detailed biological evaluation revealed that compound 38 can bind to NSD2-PWWP1 and then affect the expression of genes regulated by NSD2. We believe that the current discovery will provide a useful chemical probe to the future research in understanding the specific regulation mode of NSD2 by PWWP1 recognition.


INTRODUCTION
The nuclear receptor-binding SET domain (NSD) family of protein lysine methyltransferases consists of three members: NSD1, NSD2 and NSD3，and which predominantly catalyzes the mono-and di-methylation of histone 3 lysine 36 (H3K36). 1 High expression, point mutations or translocations of NSD family frequently occurred in multiple types of human cancers. NSD2, also known as MMSET (multiple myeloma SET domain) or WHSC1 (Wolf-Hirschhorn syndrome candidate 1) is involved in several cellular processes, including DNA damage repair, epithelial-mesenchymal transformation (EMT) and cell cycle regulation, and was identified as an important driver in oncogenesis. 2 Notably, NSD2 is closely related to accelerated disease progression and rapid relapse in 15-20% of multiple myeloma (MM) harboring the t(4;14) chromosomal translocation with increased levels of H3K36me2, 3 and among the most frequently mutated genes in pediatric cancer genomes, E1099K mutant of NSD2 is hyperactivated in acute lymphoblastic leukemia. 4 Recent studies have revealed that overexpression of NSD2 occurred in several solid tumors including bladder, prostate, glioblastoma and gastrointestinal. 5,6 NSD2 is a multi-domain protein, consists of the SET domain performing the catalytic methylation function and two PWWP (proline-tryptophan-tryptophan-proline) domains. The N-terminal PWWP domain (PWWP1) acts as the reader domain that preferentially binds to the nucleosomes containing H3K36me2while PWWP2 domain can bind to histone H3K36/79me3 and dsDNA. Though PWWP1 and PWWP2 are different in affinity to histone substrates, they have been reported to cooperate in maintaining appropriate nuclear localization. This process relies on the conserved aromatic cage consists of PWWP1 Y233, W236 and F266, which are orthogonally positioned and interact with the methylated side chain of H3K36 residue. The NSD2-PWWP1:H3K36me2 interaction plays a critical role in stabilizing NSD2 at chromatin, and then facilitates epigenetic spreading and propagation of H3K36me2which is required for the recruitment of DNMT3A and maintenance of DNA methylation at intergenic regions for gene transcription. 7 In carcinoma cells, H3K36me2 signal alteration underlies epithelial-to-mesenchymal transition and mesenchymal-toepithelial transition that is critical for metastatic. 8 Since NSD2 is a promising target for targeted cancer therapy, several groups have reported their efforts in developing selective small-molecule inhibitors targeting the catalytic domain or the PWWP domains of NSD proteins (Figure 1). Huang   Since there is only one chemotype of NSD2 PWWP1 inhibitors that merely showed moderate binding activity, therefore, more potent NSD2 PWWP1 inhibitors need to be developed to further assess the pharmacological potential of NSD2. Herein, we reported our effort to conduct a structure-based optimization of NSD2 PWWP1 ligands, leading to a selective NSD2-PWWP1 inhibitor 38 with significantly improved potency. Compound 38 can clearly bind to NSD2-PWWP1 and further affect the expression of genes regulated by NSD2. We believe that the discovery and related biological evaluation will contribute to the further research of the specific regulation mode of NSD2 by PWWP1 recognition, and furthermore, our discovery also provided a certain material basis for the functional study of NSD2.

Rational Design of Selective NSD2-PWWP1 Inhibitors.
At the initial stage of this project, we analyzed the binding model of reported small-molecule compounds to understand the binding basis between inhibitors and proteins, which enabled us to rationally design a new class of NSD2 inhibitors.
Currently, the crystal structure of NSD2 PWWP1 domain bound with MR837 was reported (PDB ID: 6UE6) 12 , which revealed two important binding features: 1) Ncyclopropyl amide of MR837 occupied the aromatic cage sub-pocket forming by three conservative aromatic residues, Tyr233, Trp236 and Phe266; 2) an essential hydrogen bonding interaction between Ala270 residue and the N atom of benzonitrile. As mentioned above, BI-9321, the first selective and potent PWWP inhibitor in NSD family was reported by Bottcher et al. and demonstrated in the cocrystal structure (PDB code: 6G2O) 10 , binds to NSD3-PWWP1 domain by the following patterns: the aromatic cage was occupied by N-methylimidazole ring; the N atom in the quinoline ring interacted with Ser314 through hydrogen bonding; and the NH2 of benzylamine extending out the binding pocket and forming a strong electrostatic interaction with nearby negative charged residue Glu318.  and NSD3-PWWP1 (BI-9321, PDB code 6G2O).
By superposition of these two cocrystal structures (Figure 2A) Therefore, we rationally designed the new chemotype of NSD2 PWWP1 inhibitors by incorporating the benzonitrile into the BI-9321 to replace the quinoline ring, which was thought that could achieve the binding selectivity of NSD2 over the NSD3 protein.
Besides, we also noticed that the methyl group on the benzene ring of BI-9321 was at the similar position of the O atom of the amide group of MR837, and the distance between the methyl C atom on the benzene ring of BI-9321 and Tyr233 was 3.6 Å. We hypothesized that if the methyl on the benzene ring was replaced with methoxy group, the O atom of methoxyl group may be able to form a hydrogen bond with Tyr233.
Taking together, we designed and synthesized compound 5 in Figure 2B for the biological test to verify the design strategy.

Structure-Activity Relationship (SAR) Exploration and Optimization of Ligands.
As shown in Table 1, compound 5 with two methoxy groups displayed a good binding affinity to NSD2 PWWP1 domain, with an IC50 value of 4.44 μM，which is much better than MR837 (4). This encouraged us to synthesize analogs 6-12. According to the binding assay, removal of one or two methoxy groups led to compounds 6 and 7 showing slightly reduced potency with IC50 values of 11.97 μM and 13.95 μM, respectively. When mono-halogen groups such as chlorine and fluorine were introduced (compound 8 and 9), they also slightly reduced potency if comparison with compound 5. However, single trifluoromethyl-substituent and methyl-substituent compounds (10 and 11) showed equal potency to compound 5. When another methyl group was introduced to yield compound 12, the potency was significantly improved to an IC50 value of 0.57 μM, about 8-fold more potent than of compound 5. This encouraged us to further explore the effect of other parts of this chemotype inhibitors on NSD2 PWWP1 binding.  To verify the design, we performed the crystallization experiment and solved the cocrystal structure of NSD2 PWWP1 domain with compound 5. As illustrated in Figure   3, compound 5 bound to the NSD2 in a mode very similar to BI-9321 bound to the NSD3: the N-methylimidazole situated into the aromatic cage and the benzylamine group interacted with the nearby Glu277; while the benzonitrile motif represents as an essential pharmacophore for PWWP1 of NSD2, which, as predicted, interacts with residues Asp269 and Ala270. Clearly, the nitrile group would be collided with the serine residue in NSD3.
Based on this solved structure, we next conducted a series of structural modifications to get compounds 13 -20 to further explore the interactions with this benzonitrile binding subpocket. As shown in Table 2  respectively. These results indicated that benzylamine group was necessary for activity against NSD2 and minor modifications led to much reduced potency. During the exploration of SARs, we found that tert-butoxycarbonyl-protected compound 13, compound 28 still showed moderate potency with an IC50 value of 1.67 μM. This indicated that there is certain space around this binding subsite to conduct structural modifications. Therefore, we further explored steric and hydrophilic tolerance by introducing bulky alkyl-or water-soluable substituents. As shown in Table   4, Notably, among these compounds, compound 38 showed the best potency against NSD2, 5-fold more potent than compound 13.

Selectivity profile with Protein Thermal shift
We carried out protein thermal shift assay to directly confirm the binding between NSD2-PWWP protein and compound 38. As shown in

Target validation in molecular and cellular level
We proved in the previous section that compound 38 could bind to PWWP domains.
To further detect whether 38 can affect HK36me2, we adopted AlphaLisa which was designed for screening NSD2 enzymatic inhibitors. It was shown that 38 did not display inhibition effect on H3K36me2 in vitro( Figure 5A). Consistent with Alphalisa results, 38 caused little effect on general H3K36me2 in cellular level( Figure 5B). In conclusion, 38 had no impact on SET domain or enzyme activity catalytic function, which further validated that 38 could specifically binds to PWWP domain.   showed little response to 9c. 14 In this study, NSD2-PWWP1 inhibitor 38 showed inhibitory effect in both NSD2 mutant and wild type cells. Therefore, we speculated that NSD2-PWWP1 inhibitor inhibited cell proliferation independent of NSD2 genotype, which was different from NSD2-SET inhibitor.