Nemtabrutinib and pirtobrutinib were purchased from ChemScene LLC (Monmouth Junction, NJ, USA). Acalabrutinib, ibrutinib and zanubrutinib were purchased from MedChemExpress LLC (Monmouth Junction, NJ, USA), Axon Medchem B.V. (Groningen, the Netherlands), and Activate Scientific GmbH (Prien am Chiemsee, Germany), respectively. All inhibitors were stored as dry powders at 4 °C and were dissolved in dimethyl sulfoxide (DMSO) at 10 mmol/L concentration before testing.

Nemtabrutinib and the approved BTK inhibitors were profiled on a panel of 254 wild-type kinases in mobility shift assays (MSA) at Carna Biosciences, Inc. (Kobe, Japan). Kinase inhibition was measured at a compound concentration of 1 µmol/L and an ATP concentration of K _M,bin (17), which is an ATP concentration within 2-fold of the affinity of an individual kinase for ATP (K _M,ATP). Percentage inhibition values were mapped to a phylogenetic kinome tree using Coral (http://phanstiel-lab.med.unc.edu/CORAL/, accessed September 4^th, 2024) (18). Half-maximal inhibitory concentration (IC₅₀) values were determined using duplicate 10-point dilution series in MSA at Carna Biosciences, Inc. for 15 kinases, while for MEK1 and MEK2, inhibition of enzymatic activity was measured in-house in enzyme-linked immunosorbent assays (ELISA) (Carna Biosciences, Inc., cat. no. 07–41 and 07-42). Percentage inhibition of MEK1 and MEK2 at 1 µmol/L nemtabrutinib and K _M,bin ATP concentration was derived from the dose-response curve as well and mapped to the phylogenetic kinome tree, as described above. Inhibition of MLK1 and B-RAF at K _M,ATP was determined in a radiometric assay using ³²P-labeled ATP at Eurofins Cerep SA (Celle-Lévescault, France). Interaction with SIK3 was assessed in a competition binding assay at Eurofins DiscoverX LLC (San Diego, CA, USA). The binding of nemtabrutinib to biotinylated, inactive MEK1 (Carna Biosciences, Inc., cat. no. 07-441-10-20N) and activated B-RAF (Carna Biosciences, Inc., cat. no. 09-422-20N) was measured by surface plasmon resonance (SPR) using a Biacore 1S+ (Cytiva), as described previously (9).

A panel of 160 cancer cell lines was used in this study. Supplementary Table S1 provides an overview of the 160 cell lines used for cell viability assays, the subset of 102 cell lines used for comparative profiling and the subsets used for different bioinformatic analyses, which were dependent on the availability of public data. Cell lines were purchased from the American Type Culture Collection (ATCC) (Manassas, VA, USA), the German Collection of Microorganisms and Cell Cultures (DSMZ) (Braunschweig, Germany), the RIKEN BioResource Research Center (Tsukuba, Ibaraki, Japan) or the Japanese Collection of Research Bioresources (JCRB) (Ibaraki City, Osaka, Japan). All cell lines were propagated in the cell culture media as indicated in Supplementary Table S1 . Cell viability assays were carried out within ten passages of the original vials. The authenticity of the ATCC and DSMZ cell lines has been confirmed by short tandem repeat analysis at both institutions. In addition, the mutation status of several cancer genes has been confirmed in various cell lines, including RIKEN and JCRB cell lines, by next-generation sequencing of genomic DNA isolated from the same cell batches as used for the viability assays (19, 20).

Intracellular ATP content was used as an indirect readout of cell number using the ATPlite 1Step bioluminescence assay (Revvity, Groningen, the Netherlands). Cells were seeded in 384-well plates at an optimized density to ensure unrestricted growth and maximal signal at the end of the experiment. After 24 hours of incubation, the starting cell number was determined by adding ATPlite to each well of a control plate and recording luminescence on an Envision multimode reader (Revvity, Waltham, MA, USA) (4). Compounds were diluted in DMSO in √10-fold steps from 10 mmol/L stocks to obtain 9-point dilution series. After further 31.6-fold dilution in 20 mmol/L HEPES (pH 7.4), the dilution series were added in duplicate to the cells in the 384-well plates to determine the effect of the compounds on cell viability. Vehicle-treated controls were included to determine maximal cell proliferation. The final DMSO concentration was 0.4% (v/v) in all wells. After incubation for an additional 72 hours, the ATP content was measured in each well using ATPlite. The luminescent signal in inhibitor-treated wells was normalized to vehicle-treated control wells to determine the percentage viability at each concentration. Cell doublings were determined by relating the cell number of vehicle-treated controls to the starting cell number for each cell line. The viability assay of a cell line was repeated when the cell doubling deviated > 2-fold from the historic doubling as determined by multiple independent experiments. The quality of the complete assay was determined by a parallel test with doxorubicin on two cell lines. IC₅₀ values were calculated by fitting a four-parameter logistic model to the percentage viability values using IDBS XLfit5 (IDBS, Guildford, United Kingdom). All curves were visually inspected and submitted to an F-test as implemented in IDBS XLfit5. Curves with an F-value above 1.5 were invalidated and IC₅₀ values were maximized at the highest concentration. The highest initial test concentration in each dose range was 31.6 µmol/L and the lowest was 3.16 nmol/L. When this dose range was too high in a certain cell line to allow for a reliable determination of the IC₅₀, a new dilution series was prepared using a diluted stock solution and was used to retest the compound on the cell line.

Multiple datasets were used for the bioinformatic analyses. The gene mutation, fusion and amplification status of the cell lines were retrieved from the COSMIC Cell Lines Project available at the COSMIC database (https://cancer.sanger.ac.uk) (21), the DepMap database (release 24Q2) (7) and literature. Mutations were filtered for oncogenic relevance with the Cancer Hotspots database (https://www.cancerhotspots.org/#/home) (22) and the OncoKB database (https://www.oncokb.org/) (23, 24), as described previously (6). Gene mutation data are available for 156 of the 160 profiled cell lines ( Supplementary Table S1 ).

The in-house Oncolines^® kinase inhibitor dataset contains the IC₅₀ profiles of 135 kinase inhibitors that have been profiled on a subset of 66 or 102 of the 160 cancer cell lines. The kinase inhibitor set includes approved drugs, clinical and pre-clinical inhibitors, and tool compounds ( Supplementary Table S2 ).

Gene and protein expression data of cell lines as well as gene dependency scores were retrieved from the DepMap database (7). Basal mRNA gene expression levels (release 23Q4) were used and are available for 145 of the 160 profiled cell lines ( Supplementary Table S1 ). Protein expression levels, as determined by Reverse Phase Protein Array (release 22Q2), were used and are available for 128 of the 160 cancer cell lines ( Supplementary Table S1 ).

Gene dependency scores from a large individual CRISPR knock-out screen and large RNA interference (RNAi) knock-down screens (release 23Q4 and Demeter Data v6, respectively) were used (25–29). In these screens, the effect of single-gene knock-out or knock-down on cell viability has been determined. The more negative a dependency score, the more dependent a cell line is on a specific gene. The CRISPR dataset includes data for 120 of the 160 profiled cell lines, while the RNAi dataset covers data for 107 of the 160 cell lines ( Supplementary Table S1 ).

For all bioinformatic analyses of cell viability assay data, ¹⁰logIC₅₀ values (in nmol/L) were used as a measure for cell line sensitivity. All calculations were performed in R (version 4.3.1) (30).

For the gene mutation analysis, profiled cell lines were classified as having an alteration in a gene if the gene was mutated, fused, amplified (oncogenes) or deleted (tumor suppressor genes). Otherwise, cell lines were classified as ‘wild-type’. The analysis focused on a subset including known oncogenic kinase genes (31) of which only 23 genes that were altered in three or more cell lines were included to allow for proper statistics. A type II analysis of variance (ANOVA) was performed to identify significant associations between cell line sensitivity to nemtabrutinib and kinase gene alterations. Significance after multiple testing correction was determined with the Benjamini-Hochberg procedure (i.e., false discovery rate < 20%).

The IC₅₀ profile of nemtabrutinib was compared to the IC₅₀ profiles of the kinase inhibitors in the Oncolines^® kinase inhibitor dataset by calculating the Pearson correlation between compounds or by hierarchical clustering. The network tree of inhibitors and Pearson correlations was generated using the Fruchterman–Reingold algorithm, as implemented in the R package igraph (32). Hierarchical clustering of cell panel viability data from the kinase inhibitors was performed with the Ward method, using 1 – Pearson correlation (r) as clustering distance, as described previously (8).

The IC₅₀ values of nemtabrutinib were correlated to basal gene expression levels of 371 genes and protein expression levels of 55 genes that have been experimentally determined to be involved in cancer (33) using Pearson correlations. Additionally, Pearson correlations were determined between the IC₅₀ values of nemtabrutinib and CRISPR or RNAi dependency scores of 465 and 204 kinase genes, respectively (34). Significance of all correlations was determined by calculating p-values, which were subjected to the Benjamini-Hochberg procedure to correct for multiple testing. Correlations with a false discovery rate of < 20% were considered significant.

Docking of nemtabrutinib in the ATP- and allosteric binding pocket of MEK1 was performed with the structure of MEK1 with the ATP-competitive MEK1 inhibitor DS03090629 (PDB ID:7XLP (35)) and the structure of MEK1 with adenosine 5’-(β,γ-methylene)triphosphate (AMP-PCP) and allosteric MEK1 inhibitor cobimetinib (PDB ID: 4AN2 (36)), respectively. Prior to docking, the structures were processed in MOE v2024.06, hydrogen atoms were added, partial charges were calculated using the Amber10:EHT forcefield and energy minimization was performed. The ATP- and allosteric binding pockets were selected as amino acids within 4.5 Å distance of DS03090629 and cobimetinib, respectively. DS03090629, cobimetinib and co-factors not involved in ligand binding were removed from the structures. Docking in the allosteric binding pocket was done in the presence of AMP-PCP and magnesium.

Docking was performed using the Amber10:EHT forcefield with Triangle Matcher placement and Induced Fit receptor refinement. Redocking of DS03090629 and cobimetinib in their respective protein structures validated the docking procedure (RMSDs: 0.61 and 0.92; docking scores: -12.2 and -10.1). Docking scores for nemtabrutinib in the ATP-binding pocket ranged from -10.9 to -10.5. For docking of nemtabrutinib in the allosteric binding pocket, 100 docking poses were generated of which the 25 highest-scoring poses after refinement (-9.8 to -8.9) were reported. Docking poses were visually inspected and analyzed for ligand-protein interactions using interaction fingerprints (IFPs).

Nemtabrutinib inhibited the tyrosine kinase activity of wild-type BTK in biochemical assays with an IC₅₀ of 1.4 nmol/L, which is in range with the inhibition of BTK by the approved BTK inhibitors ( Figure 1a ). The acquired resistance-related C481S-mutant BTK is inhibited by nemtabrutinib with a similar IC₅₀ (1.6 nmol/L, Table 1 ). As previously noted (15), nemtabrutinib shows cross-reactivity with many other kinases in biochemical assays, including other TEC family, SRC family and growth factor receptor tyrosine kinases (RTKs) ( Figure 1b , Table 1 ). When profiled on a panel of 254 wild-type kinases at a single concentration of 1 µmol/L, nemtabrutinib inhibited the activity of 63 kinases by more than 75% and 45 kinases by more than 90% ( Figure 1b , Supplementary Table S3 ). This is a considerably higher number of cross-reactivities than found for the approved BTK inhibitors on the same kinase panel ( Figure 1c , Supplementary Table S3 ). For example, the other reversible BTK inhibitor pirtobrutinib inhibited only five kinases by more than 75% (i.e., Aurora B, BRK, CSK, EGFR and TEC) ( Supplementary Table S3 ). These results indicate that nemtabrutinib has a broader selectivity in biochemical assays compared to pirtobrutinib and the approved irreversible BTK inhibitors.

To study the activity and selectivity of nemtabrutinib in cancer cells, nemtabrutinib was profiled on a panel of 160 human cancer cell lines in cell viability assays. The cell panel represents a wide range of solid tumors and hematological malignancies ( Supplementary Table S1 ). Nemtabrutinib demonstrated potent inhibitory activity across cell lines from all tissues and disease types. The most sensitive cell line was the chronic eosinophilic leukemia cell line EoL-1 (IC₅₀ = 59 nmol/L), expressing a fusion gene of Factor Interacting with PAPOLA and CPSF1 (FIP1L1) and platelet-derived growth factor receptor α (PDGFRα) ( Figure 2a , Supplementary Table S4 ) (37). Other cell lines among the most sensitive responders were two T-cell acute lymphoblastic leukemia (T-ALL) cell lines (Jurkat E6.1 and CCRF-HSB-2) and several cell lines derived from solid tumors, including endometrial (HEC-251 and HEC-1-B) and colon cancers (HT-29). The germinal center B-cell like diffuse large B-cell lymphoma (DLBCL) cell line HT and the prostate carcinoma cell line LNCaP clone FGC were among the least sensitive cell lines in the viability assays (IC₅₀ > 31.6 µmol/L) ( Supplementary Table S4).

The profiled cancer cell line panel contains two cell lines derived from BTK-dependent cancers, i.e., the DLBCL cell line SU-DHL-6 and the mantle cell lymphoma cell line REC-1. The SU-DHL-6 cell line harbors a mutation in MYD88 resulting in the activation of BTK (38), while the REC-1 cell line has constitutively active B-cell receptor signaling (39). Nemtabrutinib inhibited the viability of these cell lines with an IC₅₀ of 0.6 µmol/L and 2.0 µmol/L, respectively. This contrasts the potent IC₅₀ values of the irreversible BTK inhibitors ibrutinib, acalabrutinib and zanubrutinib on these cell lines ( Figure 2b , Supplementary Table S5 ). However, the approved BTK inhibitors showed only partial effects, while nemtabrutinib reached almost complete inhibition of viability ( Figure 2b ).

To investigate which kinases could be responsible for the broad cellular activity of nemtabrutinib, a number of bioinformatic analyses were performed on the cell panel profiling data. Both genomic alterations and aberrant expression of genes or proteins can induce tumor cell proliferation (31, 33). Therefore, we correlated the IC₅₀ profile of nemtabrutinib with the mutation status of oncogenic kinases, the gene dependency scores of human kinase genes and with the gene or protein expression levels of known cancer genes in the cell lines.

Associations between cell line sensitivity and gene mutations were determined by ANOVA. Of the 23 kinase genes analyzed, only the mutation status of BRAF was found to be significantly associated with nemtabrutinib sensitivity ( Figure 3a ). Cancer cell lines harboring mutations in BRAF were on average three times more sensitive than cell lines expressing wild-type BRAF. Mutant BRAF is also a predictive biomarker of drug response for B-RAF and MEK1 inhibitors profiled in our cancer cell line panel (4, 6).

To further investigate the mechanism underlying the association of cell line sensitivity to nemtabrutinib with BRAF mutation status, we compared the IC₅₀ profile of nemtabrutinib with the profiles of the 135 kinase inhibitors in the Oncolines^® kinase inhibitor dataset in a one-to-one comparison analysis. Inhibitors are considered to have a similar inhibitory profile if the Pearson correlation is 0.5 or higher. If inhibitors show a similar profile, they are connected to each other in a network tree for visualization ( Figure 3b ). The analysis revealed that nemtabrutinib is not connected with the approved BTK inhibitors, but shows connections with MEK, ERK and pan-RAF inhibitors ( Figure 3b ). A comparative analysis based on an unsupervised hierarchical clustering showed similar results ( Supplementary Figure S1 ). The approved BTK inhibitors cluster at distant locations in the hierarchical clustering tree. For instance, and as noted before (8), ibrutinib clusters with EGFR inhibitors ( Supplementary Figure S1 ).

Basal gene expression levels of 371 cancer genes were correlated to the cancer cell line sensitivity of nemtabrutinib. This correlation analysis revealed increased expression of fibroblast growth factor receptor 3 (FGFR3) as a significant marker of drug response ( Figure 4a ). Cancer cell line sensitivity was also correlated to the protein expression levels of 55 cancer genes. This showed that high levels of phosphorylated MEK1 are significantly correlated with sensitivity to nemtabrutinib ( Figure 4b ).

Finally, we compared the drug sensitivity profile of nemtabrutinib with the genetic dependency scores of kinase genes available from large-scale CRISPR (465 genes) and RNAi screens (204 genes). Significant positive correlations were found between the IC₅₀ profile of nemtabrutinib and the CRISPR knock-out dependency scores of two genes: BRAF and SIK3 ( Figure 4c ). Knock-down of BRAF by RNAi was also significantly correlated with response to nemtabrutinib. Additionally, response to nemtabrutinib exhibits a strong positive correlation with knock-down dependency scores of MAP2K1, MAP3K9, and MAPK1 ( Figure 4c ), encoding the MAPK pathway components MEK1, MLK1 and ERK2. The correlation analysis was repeated with the dataset minus the 16 BRAF-mutant cell lines to determine whether the preferential targeting of BRAF-mutant cell lines by nemtabrutinib may have strongly influenced the results. The significant positive correlations with SIK3 and MAP3K9 in the CRISPR and RNAi screen dataset, respectively, were maintained in this analysis ( Figure 4d ), indicating that the dependency of cell lines on these kinases can predict nemtabrutinib response independent of BRAF mutation status.

To determine whether the kinases identified in the bioinformatic analyses of the cell panel profiling study were genuine biochemical targets of nemtabrutinib, kinase enzyme activity assays were performed. The IC₅₀ values of nemtabrutinib in biochemical assays with these kinases are given in Table 1 . Table 1 summarizes a subset of kinases that were included in the panel of 254 wild-type kinases for the single concentration profiling ( Supplementary Table S3 ), such as FGFR1-4, as well as kinases that were not included in this panel (MEK1, MLK1, ERK2, B-RAF and SIK3).

The most sensitive cell line EoL-1 has a constitutively active PDGFRα due to a FIP1L1-PDGFRα fusion (37). In a biochemical assay, PDGFRα was inhibited with an IC₅₀ of 12.4 nmol/L. As mentioned before, the gene expression analysis revealed a correlation between high expression of FGFR3 and cell line sensitivity to nemtabrutinib ( Figure 4a ). Biochemically, nemtabrutinib inhibits FGFR3 with an IC₅₀ of 70.1 nmol/L and FGFR2 is even more potently inhibited with an IC₅₀ of 34.0 nmol/L. For FGFR1, moderately potent inhibition was found (IC₅₀ = 134 nmol/L), while FGFR4 was not potently inhibited (IC₅₀ = 1.40 µmol/L). The cell lines AN3-CA and KATO III have an FGFR2 mutation or amplification, respectively (6). These cell lines were inhibited by nemtabrutinib with a potency within the range of inhibition of the BTK-driven cell lines (IC₅₀ = 1.1 µmol/L) ( Supplementary Table S4 ). The FGFR1 fusion-positive cell line KG-1 and FGFR3 fusion-positive cell line RT-4 (6) are also inhibited within that range, however with a slightly lower potency (IC₅₀ = 1.4 and 1.7 µmol/L, respectively). No binding to SIK3 by nemtabrutinib was found ( Table 1 ). These results suggest that platelet-derived and fibroblast growth factor receptors are potential drug response biomarkers for nemtabrutinib.

The sensitivity of BRAF-mutant cell lines to nemtabrutinib indicates activity of nemtabrutinib on the MAPK pathway. Nemtabrutinib inhibited MEK1 and MEK2 with IC₅₀ values of 8.5 and 9.5 nmol/L, respectively ( Table 1 ). The compound also showed inhibitory activity in the B-RAF assay with an IC₅₀ of 73 nmol/L. It should be noted that the B-RAF kinase assay is a cascade assay, in which activity is indirectly measured via MEK1 and ERK2. Nemtabrutinib did not inhibit MLK1, ERK1 or ERK2 ( Table 1 ).

To investigate the specific target of nemtabrutinib in the MAPK pathway, surface plasmon resonance binding experiments were performed on B-RAF and MEK1. Nemtabrutinib bound inactive MEK1 with an affinity (K _D) of 10 nmol/L ( Figure 5a ), whereas it bound activated B-RAF with a K _D of 1.2 µmol/L ( Figure 5b ), corresponding to a 120-fold difference in affinity ( Supplementary Table S6 ). This indicates that the inhibitory effect on MAPK signaling is most likely due to inhibition of MEK1 rather than B-RAF.

X-ray crystallographic studies on nemtabrutinib-bound BTK indicate that nemtabrutinib is an ATP-competitive BTK inhibitor ( Figure 5c , Supplementary Table S7 ) (15). To investigate the binding mode of nemtabrutinib in MEK1, molecular docking in both the ATP- and allosteric binding pocket of MEK1 was performed. Docking of nemtabrutinib in the ATP-binding pocket of MEK1 resulted in multiple docking poses, which were described by interaction fingerprints (IFPs). These in silico predicted IFPs were compared to the IFP derived from the X-ray crystallographic analysis of the nemtabrutinib-BTK complex (15). Interestingly, the highest-scoring docking pose of nemtabrutinib in MEK1 has a very similar IFP as the IFP found in the crystal structure of nemtabrutinib-bound BTK ( Figures 5c, d , Supplementary Table S7 ). This includes two hydrogen bond interactions between the pyrrolo-pyrimidine ring of nemtabrutinib and two hinge region residues in the ATP-binding pocket of BTK (Met⁴⁷⁷ and Glu⁴⁷⁵) and MEK1 (Met¹⁴⁶ and Glu¹⁴⁴). Additionally, the hydroxymethyl-oxane moiety and the carbonyl oxygen atom of nemtabrutinib forms a (water-mediated) hydrogen bond interaction with the residues Cys⁴⁸¹ and Lys⁴³⁰ (BTK) or Ser¹⁵⁰ and Lys⁹⁷ (MEK1), respectively.

In contrast, docking of nemtabrutinib in the allosteric pocket of MEK1 indicates that nemtabrutinib exhibits extensive sampling of the allosteric pocket, suggesting a lack of strong and highly complementary interactions. Oppositely, the crystal structure of cobimetinib in the allosteric pocket of MEK1 showed multiple polar contacts, such as hydrogen bond interactions with Lys⁹⁷, Asp¹⁹⁰ and Met¹⁴³ as well as halogen bond interactions with Val¹²⁷ and Ser²¹² ( Supplementary Figure S2A ). Indeed, the IFP analysis shows that nemtabrutinib is unable to form the contacts observed with cobimetinib ( Supplementary Figure S2B, C ). Altogether, the molecular docking experiments suggest that nemtabrutinib preferentially binds to the ATP-binding pocket and does not exhibit strong binding to the allosteric pocket.