Novel Methionine Aminopeptidase 2 Inhibitor M8891 Synergizes with VEGF Receptor Inhibitors to Inhibit Tumor Growth of Renal Cell Carcinoma Models

Abstract

N-terminal processing by methionine aminopeptidases (MetAP) is a crucial step in the maturation of proteins during protein biosynthesis. Small-molecule inhibitors of MetAP2 have antiangiogenic and antitumoral activity. Herein, we characterize the structurally novel MetAP2 inhibitor M8891. M8891 is a potent, selective, reversible small-molecule inhibitor blocking the growth of human endothelial cells and differentially inhibiting cancer cell growth. A CRISPR genome-wide screen identified the tumor suppressor p53 and MetAP1/MetAP2 as determinants of resistance and sensitivity to pharmacologic MetAP2 inhibition. A newly identified substrate of MetAP2, translation elongation factor 1-alpha-1 (EF1a-1), served as a pharmacodynamic biomarker to follow target inhibition in cell and mouse studies. Robust angiogenesis and tumor growth inhibition was observed with M8891 monotherapy. In combination with VEGF receptor inhibitors, tumor stasis and regression occurred in patient-derived xenograft renal cell carcinoma models, particularly those that were p53 wild-type, had Von Hippel-Landau gene (VHL) loss-of-function mutations, and a mid/high MetAP1/2 expression score.

Introduction

The discovery of the antiangiogenic mycotoxin fumagillin (1) and its unique molecular target methionine aminopeptidase 2 (MetAP2) (2) in the 1990s led to the identification and development of MetAP2 inhibitors for cancer treatment. Initially, promising signals of antitumor activity were described for the fumagillin derivative TNP-470 (3). However, the drug showed rapid elimination and neurologic side effects, which were hypothesized to be compound rather than target related (4).

MetAPs are dimetallohydrolases that remove the amino-terminal methionine residue encoded by the start codon from nascent proteins. This processing step is important for the maturation of a subset of proteins, especially proteins that are N-terminally modified or whose structure/function/activity relies on N-terminal processing (5). Eukaryotic organisms possess two MetAPs, MetAP1 and MetAP2, which have overlapping substrate specificity but share low sequence similarity. They are functionally redundant, at least in some eukaryotic organisms such as Saccharomyces cerevisiae and Arabidopsis thaliana (6, 7). However, systemic knockout (KO) of MetAP2 in mice is embryonic lethal with early defects during gastrulation. Tissue-specific MetAP2 KO in the hemangioblast lineage indicated its functional importance for vascular development (8). Although disruption of the MetAP2 gene is lethal for mammalian development, in vivo pharmacologic studies with fumagillin-based MetAP2 inhibitors in cancer models demonstrated efficacy and tolerability, underlining its promise for further exploration (9).

Pharmacologic inhibition of MetAP2 is expected to have a broad and complex impact on the N-terminal processing of multiple proteins with diverse cellular functions, ultimately leading to inhibition of the proliferation of endothelial or tumor cells (10). The molecular mechanisms that causally affect cell proliferation are difficult to identify and may be cell-type specific. Activation of the tumor suppressor p53 and its downstream effectors, such as p21, may to some extent explain the antiproliferative mechanism of action in some cell types, whereas p53-independent mechanisms may exist in others (8, 11). Other studies suggest that MetAP2 activity and sensitivity of cells to MetAP2 inhibition may be modified by the redox state of the cell (12, 13).

Here, we describe the pharmacologic characteristics of M8891, a potent, selective, and reversible small-molecule inhibitor of MetAP2. The antiangiogenic and antitumoral activity of M8891 were characterized and predictive biomarker candidates were identified. M8891 is currently being advanced into phase I clinical trials in patients with solid tumors (NCT03138538).

Materials and Methods

Compound synthesis

M8891 and MSC2492281 were synthesized in the department of Medicinal Chemistry at Merck KGaA (14).

Biochemical MetAP1/2 assays

The enzyme reaction was conducted in a total assay volume of 50 μL (100 mmol/L HEPES pH 7, 50 mmol/L NaCl, 50 μmol/L MnCl₂, 140 nmol/L h-MetAP2, 1U POD, 0.02U AAO, 0.6 mmol/L dianisidine, 0.5 mmol/L MAS peptide). After preincubation of constituents for 15 minutes (25°C), the reaction was started by the addition of MAS tripeptide (Methionine-Alanine-Serine) and absorption (450 nm) was measured in an Envision plate reader. A second measurement was performed 45 minutes later. Enzyme activity was calculated based on the difference between measurements and served as the basis for the calculation of activity data and IC₅₀ values. For measurement of human MetAP1 activity, enzymes were replaced (human MetAP1: 174 nmol/L h-MetAP1). All other assay conditions were kept identical.

Human umbilical vein endothelial cell proliferation assay

Human umbilical vein endothelial cells (HUVEC; Promocell, #C-12203) were cultured in Endothelial Cell Growth Medium MV (Promocell, #C-22020) supplemented with 5% fetal bovine serum (FBS), 0.4% endothelial cell growth supplement, 10 ng/mL human recombinant epidermal growth factor, 90 μg/mL heparin, and 1 μg/mL hydrocortisone. Cells were plated at 500 cells/well in 70 μL complete medium (384-well plates) and incubated for 6–8 hours (37°C). Compounds were serially diluted in dimethyl sulfoxide (DMSO) and added to wells. Plates were incubated for 72 hours (37°C). CyQuant detection reagent (Invitrogen, #C35012) was added and incubation continued for 1 hour (37°C). Fluorescence was measured at excitation 480 nm/emission 535 nm, bottom read in an Envision plate reader.

Cell lines

Human and animal tumor cells used in this study were obtained from ATCC unless stated otherwise. IMR-5 cells were from the Wistar Institute (Philadelphia, PA, USA), PC-9 cells from the Max-Planck-Institute (Cologne, Germany), and KP-4 cells from Riken Biosource. Cells were maintained by the Merck KGaA Cell Bank (Merck KGaA) and cultured in basal medium + 10% FBS. Short tandem repeats were analyzed to confirm the identity of all cell lines by polymerase chain reaction (PCR) and electrophoretic fractionation. Mycoplasma infection was excluded by quantitative PCR-based testing performed quarterly.

Cancer cell line proliferation assays

Cells were plated at 1,000–2,500 cells/well (adjusted to the growth rate of the cell lines) in 175 μL/well complete medium and incubated overnight (37°C). Serial dilutions of compounds were prepared in DMSO, diluted 1:40 in complete medium, and 25 μL/well were added. Plates were incubated for 72 hours (37°C). BrdUrd stock solution supplied in the kit (Roche, #11 647229001) was added per well in complete medium at a final concentration of 10 μmol/L and incubation continued for 18 hours (37°C). According to the manufacturer's instructions cells were fixed and incorporated BrdUrd was detected with a peroxidase-labeled anti-BrdUrd antibody. Subsequently, a colorimetric reaction and measurement of absorbance (450 nm) allowed relative cell numbers/well to be determined.

Cell viability in an A549 mock clone and two A549 MetAP1 KO clones was assessed using CellTiter-Blue (Promega, #G8080). Plates seeded with 1,000 cells/96-well (Dulbecco's modified Eagle medium [DMEM] + 10% FBS) and treated the following day with M8891 in a 1:3 dilution series (3E-06 mol/L–4.57E-09 mol/L). Ninety-six hours post-treatment, 20 μL of CellTiter-Blue reagent was added per 96-well, and after a 4-hour incubation time (37°C) the fluorescence was measured using a plate reader (VarioskanFlash, Thermo Scientific). DMSO control was set to 100% viability, and the percentage of viable cells of each treatment condition was calculated accordingly.

To assess the colony-forming potential of A549 or A549 p53 KO cells, cells were seeded at 25 cell/cm² before being treated once (1E−05 mol/L, 2.5E−06 mol/L, 5E−07 mol/L or DMSO of MetAP2 inhibitors M8891 or TNP-470 or the MDM2 inhibitor nutlin-3A). Colony formation was monitored for 10–20 days, depending on size and density, before staining. Colonies were washed in Dulbecco's phosphate-buffered saline (Gibco, #14190136) and overlayed with staining solution (80% methanol, 0.5% crystal violet powder, Sigma-Aldrich, #C6158). Excess staining solution was removed after 5 minutes, and colonies were washed once with deionized water before being air-dried.

Simple western analysis, western blot

Simple western analysis was performed on the WES (ProteinSimple, #004-700) or Jess system (ProteinSimple, #004-650) using a 12–230 kDa separation module (ProteinSimple, #SMW004) according to the manufacturer's instructions. Based on primary antibodies [MetAP1 (1 μg/mL, Santa Cruz Biotechnology Inc, #sc-514653), γ-Tubulin (8 μg/mL, Sigma-Aldrich, #T6557), methionine-containing N-terminus of elongation factor 1-alpha 1 (MetEF1a-1; 2 μg/mL, MKV-3-165-11; see Supplementary methods S6)], the mouse or rabbit detection module was used (ProteinSimple, #DM-001, #DM-002). Data were analyzed using Compass software versions 3.1.7 or 5.0.0 (ProteinSimple). For classic western blot analysis, lysates were separated on a 4%–12% Bis-Tris poly-acrylamide gel (Thermo Fisher, #Waltham, #WG1403BOX) blotted onto polyvinylidene difluoride membranes. For both western blot and simple western analysis, cells were lysed using HGNT lysis buffer [20 mmol/L HEPES pH7.4, 10% (v/v) glycerol, 150 mmol/L NaCl, 1% (v/v) Triton-X-100, 2 mmol/L EDTA pH8, 25 mmol/L NaF, protease, and phosphatase inhibitors]. BCA (Thermo Fisher, #23225) was used for protein quantification. The primary antibodies used for western blot were p53 [1:1,000, Cell Signaling Technology (CST), #2524]; p-p53 S15 (1:1,000, CST, #9284); p21 (1:1,000, CST, #2947); p-MDM2 S166 (1:1,000, CST, #3521); MDM2 (1:1,000, CST, #86934); GAPDH (1:1,000, CST, #2,118); cyclin D (1:1,000, NeoMarkers, #DSC-6); cyclin B (1:1,000, Santa Cruz Biotechnology, #sc-245); p-Rb S807/811 (1:1,000, CST, #9308); p-Rb S780 (1:1,000, CST, #9,307); Rb (1:2000, CST #9309); and PARP (1:1,000, CST, #9532). The secondary antibodies used were: donkey anti-mouse IgG (H + L; 0.08 mg/mL, Jackson ImmunoResearch, #715-036-150); and donkey anti-rabbit IgG (H + L; 0.08 mg/mL, Jackson ImmunoResearch, #711-036-152).

MetEF1a-1 determination in tumor cell lines

Tumor cells were cultured overnight in the presence or absence of different concentrations of M8891. Cells were harvested, washed, and resuspended in phosphate-buffered saline (PBS) + 1% bovine serum albumin. Aliquots for compensation controls were transferred to separate tubes. Cells were stained with Viability Dye eFluor506, washed, and fixed in ice-cold 90% methanol for 30 minutes. The permeabilized cells were incubated for 45 minutes on ice with primary MetEF1a-1 antibody (MKV-3-165-11, 1 μg/mL). Cells were then washed and incubated with anti-rabbit IgG-AlexaFluor488 at 0.5 μg/mL for 30 minutes on ice. After the final wash, the cells were resuspended in PBS + 0.5% FBS, and 10,000 events/sample were collected on a FACSCanto II system using Diva software.

Generation of A549 knockout cell lines

A549 p53 KO cell generation was described previously (15). A549 MetAP1 KO cells were generated by nucleofection according to the manufacturer's indications. Briefly, 1×10E06 A549 cells were resuspended in the Nucleofection Solution T containing the CRISPR reagents [Cas9 protein (Alt-R S.p. Cas9 Nuclease 3NLS; IDT)], crRNAs [Synthetic RNA Sequence-specific for MetAP1 Exon1 (Sigma-Aldrich), crRNA 1: ACGCGGGTGTGCGAGACAGA, crRNA 2: TGTCCCACTTGCATCAAGCT], tracrRNA [SygRNA Cas9 Synthetic tracrRNA (Sigma-Aldrich)] and the Reporter (GFP mRNA, Eurofins). After nucleofection using Nucleofector 2D (Lonza, #VACA-1002), cells were transferred into a 6-well plate with fresh media. After 48 hours (37°C), the nucleofected pool was sorted for GFP-positive cells. These cells were again grown for several days and then seeded as single cells into 96-well plates. Three weeks later, the growing clones were consolidated, DNA was extracted with Quick Extract, and PCR was performed with MetAP1 primers (MetAP1 Fw: CCTCCGCCCGGCAGTTCCTC, MetAP1 Rw: CGTCCCGACTCGCCGCAAGG). The PCR products were sequenced. The mock clone used did not show any modifications. KO clone 1 harbors deletions of 12 bp (amino acids [aa] 26–29: DIKLG) and 48 bp (aa 12–27 deletion: DDGCSSEAKLQCPTCIK), respectively, whereas clone 2 harbors a 48 bp (aa 12–27 deletion: DDGCSSEAKLQCPTCIK) and 2 bp deletion, the latter leading to a frameshift in codon 11 and premature stop at codon 16 (MAAVETRVCERRLQQ*). Clone 1 exhibited residual MetAP1 expression whereas no expression was detectable for clone 2.

Genome-wide CRISPR screen

A549, HT1080, and U87MG cells were maintained in DMEM (Sigma-Aldrich, #11995-065) supplemented with 10% FBS (Life Technologies, #12483-020) and 1% penicillin/streptomycin (Life Technologies, #15140-122) at 37°C and 5% CO₂. Cells were infected with a genome-wide lentiviral TKOv3 library encompassing ∼18,000 genes (Addgene, #90294) and selected with puromycin (1 μg/mL HT1080; 2 μg/mL A549 and U87MG) for 48 hours. Triplicates of 15 million cells were seeded at T0 for each treatment (DMSO, M8891, TNP-470; final drug concentrations were based on IC₄₀ determination in 15-cm dishes) and maintained at 200-fold coverage. Cells were counted using the Z2 Coulter Counter (Beckman Coulter, #6605700). Genomic DNA was extracted from cell pellets (20 million cells) using the QIAamp Blood Maxi Kit (Qiagen, #51192). In addition to the protease treatment from the kit, cell pellets were also treated with RNase A (Qiagen, #19101). Genomic DNA was precipitated using ethanol and NaCl and resuspended in 10 mmol/L Tris-HCL pH7.5. Genomic RNA inserts were amplified via PCR using primers harboring Illumina TruSeq adaptors with i5 and i7 barcodes. PCR was performed using NEBNext Ultra II Q5 Master Mix (New England Biolabs, #M0544L) and the C1000 Touch Thermal Cycler (Bio-Rad, #1851196). Sequencing libraries were purified by gel extraction using PureLink Quick Gel Extraction Kit (Life Technologies, #K210012) and dsDNA concentration was measured using Qubit 2.0 Fluorometer (Life Technologies, #Q32866). The resulting libraries were sequenced on an Illumina HiSeq 2500, using single-read 50-cycle v4 chemistry. Demultiplexed FASTQ files were first trimmed by locating constant sequence anchors and extracting 20-bp guide sequences preceding the anchor sequence. Reads from each sample were mapped to the gRNA library reference and read counts for each gRNA sequence were tabulated. Samples were sequenced to a target depth of ∼400× for T0 samples or ∼200× for experimental time points. Read counts for each sequence were normalized to 10 million reads per sample, and the fold change of each gRNA was calculated based on a reference sample, i.e., T0 sample. Screen performance was evaluated using reference sets of essential and nonessential genes (16). All screens behaved appropriately based on the essential gene profile with a significant shift in fold-change distribution of essential gRNAs observed relative to nonessential genes. In addition, precision–recall curves were calculated for endpoint DMSO samples and all screens showed high performance with > 90% of gold-standard essential genes identified at a false discovery rate of 5% (QC plots can be found in the CRISPR database).

The BAGEL (Bayesian Analysis of Gene EssentiaLity) algorithm was used to calculate gene essentiality in each screen. The resulting Bayes factor (BF) score indicates the confidence of a gene's essentiality in a cell line under the conditions of the screen. More positive scores indicate a higher confidence that a given gene's knockout causes a decrease in fitness; however, the scores are not necessarily a measure of the severity of the phenotype. For information on how the BF score is calculated, please see Hart and colleagues, 2016 (16). For CRISPR data analyses the drugZ algorithm was used to calculate a normalized gene-level z-score (normZ) score (17), which allows the identification of both synergistic and suppressor chemogenetic interactions from CRISPR screens. The software is available at github.com/hart-lab/drugz. Identified hits for each cell line were consolidated and ranked based on the frequency observed in screened cell lines.

Cell-cycle analysis

A549 wild-type and A549 p53 KO cells were cultured in DMEM + 10% FBS (37°C, 10% CO₂). Cells were plated at 2E06 cells/dish (100 mm × 15 mm) per condition and incubated overnight. Cells were treated as indicated [0.03% (v/v) DMSO, 1E-06 mol/L M8891, 3E-06 mol/L nutlin-3A for 6 or 24 hours]. At 5 or 23 hours, respectively, cells were stained for 1 hour with 3.26 mmol/L BrdUrd using the Phase-Flow FITC BrdUrd Kit (BioLegend, #370704). Next, cells were fixed/stained according to the manufacturer's instructions using reagents supplied with the kit. Optional 7AAD staining was performed (30 minutes, 4°C) and subsequently, samples were analyzed using FACSCantoII. Data analysis was accomplished using FlowJo and GraphPad Prism.

In vitro PDX screen

Solid human tumor xenografts growing subcutaneously in serial passages in nude mice (NMRI nu/nu strain) were removed, mechanically disaggregated, and subsequently incubated with an enzyme cocktail consisting of collagenase type IV (41 U/mL), DNase I (125 U/mL), hyaluronidase (100 U/mL), and dispase II (1.0 U/mL) in RPMI-1640 medium for 45 minutes (37°C). Cells were passed through sieves of 100 μm and 40 μm mesh size (Cell Strainer, Becton Dickinson, Falcon), washed and viable cells were counted. Aliquots of the cells were frozen and stored in liquid nitrogen. The clonogenic assay was performed in a 96-well plate format according to a modified two-layer soft agar assay (18). Each test well contained three layers of equal volume, two layers of semisolid medium (bottom and top layer), and one layer of medium supernatant, with or without the test compound. The bottom layer consisted of 0.05 mL/well cell culture medium [IMDM or RPMI-1640 with or without pyruvate, supplemented with 20% (v/v) FBS, 0.01% (w/v) gentamicin and 0.75% (w/v) agar]. 1.5E03 to 1E04 cells were added to 0.05 mL of the same culture medium supplemented with 0.4% (w/v) agar and plated onto the bottom layer. The test compounds were added after serial dilution in DMSO, transferred in a cell culture medium, and left on the cells for the duration of the experiment (continuous exposure, 0.05 mL drug overlay). Each plate included six untreated control wells and drug-treated groups in duplicate at 10 concentrations. Cultures were incubated at 37°C and 7.5% CO₂ for 8–13 days and monitored closely for colony growth. Within this period, ex vivo tumor growth led to the formation of colonies with a diameter of >50 μm. At the time of maximum colony formation, counts were performed with an automatic image analysis system (BIOREADER 5000-Wα, Bio-Sys GmbH). Forty-eight hours prior to the evaluation, vital colonies were stained with a sterile aqueous solution of 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-phenyltetrazolium chloride (1 mg/mL, 100 μL/well; ref. 19). Drug effects were expressed as percentage of colony formation, obtained by comparison of the mean signal in the treated wells with the mean signal of the untreated controls [expressed by the test-versus-control value (T/C-value), %]. Sigmoidal concentration–response curves were fitted to the data points obtained for each tumor model using a 4-parameter nonlinear curve fit (Oncotest Warehouse Software). IC₅₀ values are reported as relative IC₅₀ values [ = the concentration of test compound that gives a response (inhibition of colony formation) halfway between the top and bottom plateau of the sigmoidal concentration–response curve (inflection point of the curve)]. The absolute IC₅₀ value is determined as the concentration at the intersection of the concentration effect curve with a T/C of 50%.

Matrigel plug assay

An angiogenesis model was developed in a bioluminescent imaging assay utilizing female VEGF receptor (VEGFR)2-luc transgenic mice (20). All mice in this study were used according to the guidelines approved by the EMD Serono Institutional Care and Animal Use Committee. On the day of subcutaneous GFR-Matrigel (Becton Dickinson Discovery Labware, #354248) injection, VEGFR2-luc mice (six animals per group) were stratified according to body weight. To promote blood vessel growth into the Matrigel plug, human growth factors VEGF and basic fibroblast growth factor (R&D Systems, #293-VE, #233-FB-025) were added to the GFR-Matrigel at 5 μg/100 μL each. Matrigel without growth factor supplementation served as a control (NGF Matrigel). After Matrigel plug injection, mice received either vehicle control, M8891 (5, 10, 20 mg/kg, vehicle: 0.25% Methocel) daily per os, or anti-murine VEGF antibody (B20-4.1-V2; vehicle: PBS) intravenously twice per week. Body weight was monitored twice weekly. On treatment day 21, luciferin (150 mg/kg) was injected subcutaneously into mice. The mice were sacrificed 10 minutes later, Matrigel plugs extracted and weighed, and bioluminescent images were taken using the Xenogen IVIS 200 (Caliper Life Sciences). To quantify the bioluminescent signal (photons/second) of the image, regions of interest for each plug image were measured, and data were expressed as photons/sec. Living Image 3.0 software (Caliper Life Sciences) was used for the analysis of bioluminescent signals. Statistical analysis of bioluminescent data was performed using GraphPad Prism and one-way analysis of variance (ANOVA) followed by Tukey post-hoc multiple pair-wise comparisons. Values ≤0.05 were considered statistically significant.

In vivo cell line–derived xenograft studies of M8891 monotherapy

Human Caki-1 RCC cells [HTB-46, 5 million in 100 μL PBS w/o Ca/Mg, Matrigel (1:1)] were subcutaneously inoculated into the right flank of female 5–6-week-old CD-1 nude mice provided by Charles River Laboratories. When the tumors had reached a volume of 100–150 mm³ animals were sorted into treatment groups. M8891 was administered at doses of 10, 25, and 50 mg/kg twice daily (6 hours apart, on weekends, the compound was administered daily at 20, 50, and 100 mg/kg). Tumor volumes and body weight were measured twice weekly using calipers (L × W²/2). Percentage T/C values were calculated as follows: [(end tumor volume treatment − start tumor volume treatment) / (end tumor volume control − start tumor volume control)] × 100. Statistical analysis was done at the end of the study in GraphPad Prism V5.04 using RM-two-way ANOVA. Values ≤ 0.05 were considered statistically significant.

Pharmacokinetic/pharmacodynamic analysis

When human Caki-1 xenograft tumors reached a tumor size of 300–500 mm³, mice were randomized into treatment groups (n = 5) and received either vehicle control or M8891 (vehicle: 0.25% Methocel) at different doses. Mice were treated with a single administration of M8891 per os at 10, 25, or 100 mg/kg. At designated time points (1, 7, 24, 48, 72, and 96 hours) following treatment, animals were euthanized, and plasma and tumor tissues were collected, snap frozen, and processed for pharmacokinetic (PK; ref. 14) and pharmacodynamic (PD) analysis. PD Biomarker analysis was performed at Indivumed GmbH, Hamburg, Germany. The Simple Western (SallySue and PeggySue, ProteinSimple) method for the detection of total EF1a-1 (Abcam, #ab186386, 0.2 mg/mL) and MetEF1a-1 (MKV-3-165-11) was used to analyze preclinical xenograft tumor tissues. Tissues were homogenized by Precellys-24 (Bertin Technologies) and total protein concentration was determined using the BCA Protein Assay Kit (Thermo Scientific, #23225). Total and MetEF1a-1 were quantified along a standard curve of recombinant total EF1a-1 protein (Abcam, #ab177675, 1 mg/mL) measured in each experiment and normalized to total protein concentrations.

In vivo monotherapy and combination studies in PDX models

Animal studies using PDXs were performed at Charles River Discovery Research Services GmbH. All experiments and protocols were approved by the animal welfare body at Charles River Discovery Research Services GmbH and the local authorities and were conducted according to all applicable international, national, and local laws and guidelines. Tumor fragments were obtained from xenografts in serial passage in female NMRI nu/nu mice (NMRI-Foxn1nu, Envigo RMS GmbH). Tumor fragments were prepared from donor mice, cut into fragments (3–4 mm edge length), and implanted subcutaneously in acceptor mice. Tumor xenografts with a take rate below 65% were implanted with one or two tumors per mouse and in case of a bilateral take, one of these tumors was explanted prior to randomization. Animals and tumor implants were monitored daily and distributed into experimental groups (n = 5) when a tumor volume of 50–250 mm³ was reached. In monotherapy studies, animals were treated with vehicle or M8891 at a dose of 150 mg/kg twice daily per os (treatments 12 hours apart). In combination studies, the VEGFR inhibitor was administered orally first (sunitinib in vehicle:1% HEC, 0.25% polysorbate 80, 0.05% antifoam; axitinib in vehicle: 0.5% carboxymethyl cellulose; cabozantinib in vehicle: 30% propylene glycol, 5% Tween 80, 65% 5% dextrose) followed by M8891 about 20–30 minutes later. Treatment was given for 5 days on and 2 days off in both the monotherapy and combination groups. Tumors and animal body weights were measured twice weekly. Animals were euthanized when the relative tumor volume (RTV) reached 500% (set as survival endpoint) or when the maximum treatment period of 90 days was reached. ‘Undefined’ was set for groups where median survival time could not be determined within the treatment period of 90 days.

Data analysis of PDX study

To evaluate the antitumor activity of M8891 as monotherapy and in combination, response to treatment was calculated based on the RTV area under the curve (AUC) derived for each mouse in all treatment groups. Continuous RTV AUC was derived based on the integration of the RTV curve between the first and the last tumor measurement performed in this study (maximally 94 days). If tumor volume was not measured after reaching an RTV limit of >500%, the last measurement was used for RTV calculation. For statistical analysis, RTV AUC values for all mice were log-transformed using a natural logarithmic scale (variable AUC_LOG). The linear mixed-effect model^a was fitted and the significance of the TREATMENT term was evaluated with a deviance analysis (type II ANOVA). The P values of the TREATMENT term were collected for each comparison (contrast). To identify potential predictive biomarkers for each biomarker type (such as gene mutations, expression, signatures expression, copy-number alterations) and each comparison (such as M8891 vs. vehicle, sunitinib + M8891 vs. sunitinib, etc.), two linear mixed-effect regression models^b,c were fitted using R package lme4. A deviance analysis was performed for each model, and the P values of the BM_CALL (biomarker positive/negative) and BM_CALL:TREATMENT were collected. P values were corrected for multiple testing for each comparison separately using the Benjamini–Hochberg procedure.

^aAUC_LOG ∼ TREATMENT + (1|MODEL)

^bAUC_LOG ∼ TREATMENT + BM_CALL + (1|MODEL)

^cAUC_LOG ∼ TREATMENT + BM_CALL + BM_CALL:TREATMENT + (1|MODEL)

Data availability

Any requests for data by qualified scientific and medical researchers for legitimate research purposes will be subject to the healthcare business of Merck KGaA's (CrossRef Funder ID: 10.13039/100009945) data sharing policy. All requests should be submitted in writing to the healthcare business of Merck KGaA's data sharing portal (https://www.emdgroup.com/en/research/our-approach-to-research-and-development/healthcare/clinical-trials/commitment-responsible-data-sharing.html). When the healthcare business of Merck KGaA has a co-research, co-development, or co-marketing or co-promotion agreement, or when the product has been out-licensed, the responsibility for disclosure might be dependent on the agreement between parties. Under these circumstances, the healthcare business of Merck KGaA will endeavor to gain agreement to share data in response to requests.

Results

Biochemical and cellular activity of MetAP2 inhibitor M8891

M8891 is a potent and selective inhibitor of MetAP2 (Fig. 1A; ref. 14). M8891 inhibited human and murine MetAP2 activity (IC₅₀ = 52 and 32 nmol/L, respectively) while sparing human type 1 MetAP1 (Fig. 1B; Supplementary Figs. S1-1 and S1-2). The stereochemical configuration at the cyclized tartronic acid was essential because the activity of the R-enantiomer dropped by approximately 150-fold (Fig. 1A and B).

Figure 1.

Biochemical and cellular activity of MetAP2 inhibitor M8891. A, Structure of MetAP2 inhibitor M8891. The active enantiomer (aka eutomer) M8891 and the less active enantiomer (aka distomer) MSC2492281, differ in conformation at the cyclic tartronic diamide ring (labeled with *). B, Concentration–response curves for M8891 and MSC2492281 in a MetAP2 biochemical assay. IC₅₀ values for both compounds are indicated. C, Concentration–response curves for M8891 and MSC2492281 in an HUVEC proliferation assay. IC₅₀ values for both compounds are indicated. D, Graphical representation of log (IC₅₀) values determined for M8891 in 16 different cancer cell lines. Tested models are categorized according to species origin and tumor indications. Dashed line at log (c) = −6 visualizes cell lines with increased sensitivity ≤1E−6 mol/L. E, Graphical presentation of log (IC₅₀) values for M8891 determined in patient-derived tumor lines tested in vitro in soft-agar assays. PDX models (125 tested) were categorized according to tumor indication. Indications with only one representative were omitted from this graph (see Supplementary Table S4-1). Dashed line at log (c) = −6 visualizes cell lines with increased sensitivity ≤1E−6 mol/L. Abbreviations: cpd, compound, HUVEC, human umbilical vein endothelial cell; IC₅₀, 50% inhibitory concentration; MetAP2, methionine aminopeptidase 2; PDX, patient-derived xenograft.

Proliferation of HUVEC, a well-established primary endothelial cell model isolated from human umbilical cords, was inhibited by M8891 with an IC₅₀ of 20 nmol/L (Fig. 1C). The distomer MSC2492281 was 750-fold less active, in line with the decrease in MetAP2 inhibition. HUVEC differentiation into endothelial tubes was only marginally affected by M8891 at high concentrations (1E−05 mol/L), as demonstrated by quantifying tube number, length, and foci formation of HUVECs grown in Matrigel, which serves as a basement membrane matrix in cell culture (see Supplementary Methods S2 and Supplementary Figs. S2-1 and S2-2).

M8891 inhibited the proliferation of cancer cells as measured in BrdUrd incorporation assays (Fig. 1D). Of 16 tested cancer cell lines, five (Caki-1, A549, SW982, IMR5, and C6) were potently inhibited with IC₅₀ values ≤1E−06 mol/L, but, notably, with varying efficacy (38%–98%; see Supplementary Table S3-1).

M8891 differentially inhibited colony formation of PDX models in a concentration-dependent and selective manner. Tumor cells from 125 PDX models were plated in soft agar, grown in 3D for 8–10 days in the presence of increasing concentrations of M8891, and inhibition of colony formation was quantified (Fig. 1E; see Supplementary Table S4-1). This cell screen confirmed the differential activity pattern observed in cancer cell lines. Of 125 models tested, 76 were inhibited with an IC₅₀ ≤1E−06 mol/L (61%), with efficacy ranging from 31% to 96%. Activity was observed across all tested tumor indications, including colon, lung, pancreas, and stomach as well as bladder, head and neck, liver, melanoma, ovarian, RCC, and sarcoma. The sensitivity of cancer cell lines, such as A549 and HCT116, in the 3D colony formation assay was in line with results from a 2D BrdUrd incorporation assay (see Supplementary Table S4-1).

Inhibition of N-terminal processing of the MetAP2 substrate EF1a-1 by M8891

Several substrates of MetAP2 have been described (14-3-3γ, GAPDH); however, we were not successful in detecting unprocessed substrates robustly and quantitatively in cell lysates using commercially available antibodies (21, 22). We identified EF1a-1 as a novel MetAP2 substrate by 2D gel electrophoresis and mass spectrometry (see Supplementary Figs. S5-1 to S5-4) and generated a rabbit monoclonal antibody specifically detecting MetEF1a-1 while sparing the processed form of the protein (see Supplementary Figs. S6-1 and S6-2 and Supplementary Table S6–1). A concentration-dependent increase in unprocessed MetEF1a-1 was observed in A549 cells treated with M8891 in line with its potency in proliferation assays (Fig. 2A). Similarly, three different cancer cell lines (A549, U87MG, and HT1080) treated with increasing concentrations of M8891 showed accumulation of MetEF1a-1 (Fig. 2B). With an EC₅₀ for MetEF1a-1 accumulation in the range of 14–37 nmol/L, the inhibition of processing of EF1a-1 correlated well with inhibition of cell proliferation, suggesting that intracellular inhibition of MetAP2 mediates the antiproliferative effect.

Figure 2.

Concentration-dependent inhibition of EF1a-1 processing in cancer cells. A, A549 lung cancer cells were treated for 24 hours with the indicated concentrations (mol/L) of M8891 and analyzed for the presence of MetEF1a-1 by simple western blotting. Tubulin served as loading control. A virtual blot-like image generated by the Protein Simple Compass software is shown. B, A549 lung cancer cell line, HT1080 fibrosarcoma cell line, and U87MG glioblastoma cancer cell line were treated with the indicated concentrations of M8891 for 24 hours and analyzed by flow cytometry using MetEF1a-1 antibody. C, Genome-wide CRISPR screen to identify modifiers of pharmacologic MetAP2 inhibition. CRISPR screens were performed with A549, HT1080, and U87MG in the presence of IC₄₀ concentrations of MetAP2 inhibitors, TNP-470 or M8891. normZ scores from both CRISPR screens were plotted (y axis: TNP-470, x axis: M8891) per cell line. Selected genes (MetAP1, MetAP2, CDKN1A, and TP53) are highlighted. Abbreviations: CRISPR, clustered regularly interspaced short palindromic repeats; EF1a-1, elongation factor 1-alpha 1; IC₄₀, 40% inhibitory concentration; MetAP2, methionine aminopeptidase 2; MetEF1a-1, methionine-containing N-terminus of elongation factor 1-alpha 1; normZ, normalized gene-level z-score.

Genome-wide sensitizer/resistance CRISPR screen with M8891 and TNP-470

With the goal of identifying genetic aberrations contributing to sensitivity or resistance to MetAP2 inhibitors, we performed genome-wide CRISPR screens in three MetAP2 inhibitor-sensitive cancer cell lines (A549, HT1080, and U87MG) using M8891 and TNP-470 as control (Fig. 2C). Results from both CRISPR screens per cell line showed a moderate to strong positive correlation (A549: r = 0.37; HT1080: r = 0.71; U87MG: r = 0.42). Remarkably, both MetAP1 and MetAP2 were identified as sensitizing genes in A549 cells, whereas only MetAP1 was identified in HT1080 and U87MG cells. Interestingly, gene essentiality analysis in the absence of treatment revealed MetAP1 and MetAP2 as essential in HT1080 and U87MG cells but not in A549 cells (see Supplementary Tables S7-1 and S7-2). Despite these differences, MetAP2 inhibitors further decreased cell fitness if the MetAP1 gene was also deleted, indicating a functional redundancy of MetAPs. Deletion of tumor suppressor p53 and the cyclin-dependent kinase inhibitor p21, which is transcriptionally regulated by p53, suppressed the effect of MetAP2 inhibition on tumor cell proliferation in A549 and HT1080 cells, while only a trend was noted in U87MG cells (Fig. 2C). TP53 and CDKN1A were both nonessential in all three cell lines, as expected. The variable effects in the three cancer cell lines suggest that other factors, such as genotype and cancer model origin, influence sensitivity/resistance to M8891.

Functional characterization of MetAP1 and p53 as M8891 response modifiers

To confirm the CRISPR screening results, we generated MetAP1 KO and TP53 KO in the A549 cell line, in which sensitizing and suppressing effects of MetAP1 and TP53, respectively, had been observed. Two MetAP1 KO cell lines with either strongly reduced or a complete lack of MetAP1 protein were evaluated in viability assays (Fig. 3A). Partial, but concentration-dependent inhibition of cell growth by M8891 in mock A549 cells was evident with ∼70% maximum inhibition and an IC₅₀ of 1.2E−07 mol/L. Notably, proliferation of both MetAP1 KO cell lines was inhibited by M8891 even more potently (IC₅₀: 2.7E−08 mol/L and 2.4E−08 mol/L; efficacy: 100%), confirming the observations from the CRISPR screen. The colony-forming potential of wild-type A549 and corresponding TP53 KO cells was evaluated in the presence of the MetAP2 inhibitors M8891 and TNP-470 and the MDM2 inhibitor nutlin-3A (Fig. 3B). Nutlin-3A leads to activation of p53 in cells with a functional TP53 gene. Reassuringly, treatment with nutlin-3A significantly reduced colony formation. Similarly, colony formation of wild-type A549 cells was blocked by M8891 and TNP-470, demonstrating the sensitivity of this cell line to MetAP2 inhibition. In contrast, the isogenic TP53 KO cell line was resistant to nutlin-3A and to both MetAP2 inhibitors. These data confirm that p53 is required for the antiproliferative activity of M8891 in A549 cells.

Figure 3.

p53 and MetAP1 are important determinants of MetAP2 inhibitor sensitivity. A, Effect of M8891 (mol/L) on cell viability of A549 mock and two MetAP1 KO clones. Clone 1 bears a heterozygous deletion, whereas clone 2 has a homozygous deletion of the MetAP1 gene. Residual expression of MetAP1 is clearly detectable in clone 1. The knockout of MetAP1 in both clones was verified by simple western analysis; a virtual blot-like image is shown and γ-tubulin served as loading control. M8891 was added in 1:3 dilution steps (3E–05 mol/L to 4.6E–09 mol/L) to cells and incubated for 96 hours, and viability was determined using CellTiter-Blue. Viability curves and IC₅₀ values for the cell lines are given. B, Colony formation assay with A549 wild-type and A549 p53 KO cell lines treated with different concentrations (from top to bottom: 1E−05 mol/L, 2.5E−06 mol/L, 5E−07 mol/L; vehicle control) of MetAP2 inhibitors M8891 or TNP-470, or the MDM2 inhibitor nutlin-3A. C, Cell-cycle distribution of A549 wild-type and A549 p53 KO cell lines treated with M8891 (1E−06 mol/L) or nutlin-3A (3E−06 mol/L) for the indicated time points analyzed by FACS. D, Western blot analysis of different cell-cycle marker proteins in A549 wild-type and A549 p53 KO cell lines treated with 1E−06 mol/L M8891 or 3E−06 mol/L nutlin-3A for the indicated time points. Abbreviations: DMSO, dimethyl sulfoxide; IC₅₀, 50% inhibitory concentration; KO, knockout; MetAP, methionine aminopeptidase; WT, wild-type.

Treatment of A549 parental cells with M8891 led to a moderate but consistent change in cell-cycle distribution (Fig. 3C). In comparison with vehicle control, cells treated for 24 hours with M8891 showed accumulation in G₁ and G₂–M accompanied by a reduction in the S phase. Although M8891 induced stronger G₁ arrest, nutlin-3A led to a more prominent arrest in G₂–M. In A549 TP53 KO cells, the shift in cell-cycle distribution was not observed for M8891 nor nutlin-3A, underscoring the relevance of p53 for the impact of M8891 on the cell-cycle. An increase in the level of p53 protein and p53-regulated proteins (p21, MDM2), and their phosphorylation (pSer15-p53, pSer166-MDM2), after 6 and 24 hours was also observed for M8891-treated cells, but to a lesser extent than in nutlin-3A–treated cells (Fig. 3D). Moreover, the concomitant reduction in Rb phosphorylation, as well as cyclin B, after 24-hour treatment with M8891 was also fully dependent on functional p53 and consistent with the arrest phenotype in the G₁ and G₂ phases of the cell-cycle. These data, together with the observed resistance of TP53 KO cells to M8891 and TNP-470, suggest that pharmacologic MetAP2 inhibition triggers activation of the tumor suppressor p53 and its downstream target p21, leading to an increase in G₁-phase cells and inhibition of cell proliferation.

Antiangiogenic and antitumoral activity of M8891 in vivo

The antiangiogenic activity of M8891 was evaluated in a Matrigel plug angiogenesis assay in vivo. The transgenic VEGFR2-luc mouse model allows bioluminescence imaging based on a VEGFR2 promoter-driven luciferase reporter gene as readout for antiangiogenic activity (Fig. 4A and B). M8891 significantly reduced the reporter signal to an extent comparable with that of the anti-muVEGF antibody used as a positive control. We conclude that M8891 has marked antiangiogenic activity in the VEGFR2-luc mouse model, which is in line with its antiproliferative effect on endothelial cells observed in vitro.

Figure 4.

Antiangiogenic, antitumoral activities, and PK/PD relationship of M8891. A, Treatment scheme to assess blood vessel formation in VEGFR2-luc mice implanted with Matrigel. One day after implantation, mice were treated with M8891 (5, 10, 20 mg/kg, daily, per os) or anti-VEGFR antibody B20 (20 mg/kg, twice weekly, intravenously). After 14 days of treatment mice were treated with luciferin and Matrigel plugs were removed, and bioluminescence was quantified. Representative images for Matrigel plugs from vehicle- and M8891-treated mice (10 mg/kg dose) are shown. B, Quantification of bioluminescence from different treatment groups from the study outlined in A displayed as a bar diagram ± standard error of the mean. Statistically significant reductions of bioluminescence in comparison with vehicle-treatment group are indicated by *. C, Tumor growth curves of ccRCC model Caki-1 treated with vehicle or different doses of M8891 (10, 25, 50 mg/kg, twice daily, per os). Statistically significant differences versus vehicle-treated control group are indicated (***, P < 0.0001; *, P < 0.05). D, Relationship of PK to PD effect on inhibition of N-terminal processing of EF1a-1 by M8891 administered to Caki-1 tumor-bearing mice. Plasma levels of M8891 and formation of unprocessed MetEF1a-1 in different treatment groups of M8891 (10, 25, or 100 mg/kg, single administration, per os) is shown as a function of time. In tumors from vehicle-treated animals sampled at 24 hours, MetEF1a1 was undetectable. E, Tumor growth curves of gastric cancer PDX model GXF-1172 treated with vehicle or M8891 (150 mg/kg, twice daily, per os). Mice in the vehicle-treated group were sacrificed on day 28 due to reaching a mean tumor volume >800 mm³. At treatment end with M8891 (day 80), 7 of 10 tumor-bearing mice were still alive, three mice were removed due to tumor progression on days 35 (two mice) and 49 (one mouse). F, Tumor growth curves of colon cancer PDX model CXF-1783 treated with vehicle or M8891 (150 mg/kg, twice daily, per os). Selected mice in the vehicle-treated group were sacrificed due to reaching the tumor growth limit on the following days: 24 (n = 2), 31 (n = 3), 35 (n = 2), 42 (n = 1), and 49 (n = 3). On the last treatment day with M8891 (day 80), 6 of 10 tumor-bearing mice were still alive and four were removed due to tumor progression on days 49 (n = 2), 56, and 59. Abbreviations: EF1a-1, elongation factor 1-alpha 1; NFG, Matrigel without growth factors; MetEF1a-1, methionine-containing N-terminus of elongation factor 1-alpha 1; PD, pharmacodynamics; PK, pharmacokinetics; SEM, standard error of the mean; VEGFR, VEGF receptor.

Antitumor activity of M8891 was investigated in the clear cell RCC (ccRCC) model Caki-1 (VHL⁺, TP53⁺), which was exquisitely sensitive to M8891 in vitro (Fig. 4C). Treatment of Caki-1 xenografts with 10, 25, and 50 mg/kg M8891 administered twice daily dose-dependently inhibited tumor growth, resulting in T/C values of 71%, 37%, and 16%, respectively, whereas tumors from mice treated with vehicle continued to grow faster. M8891 treatment was well tolerated at all doses (see Supplementary Figs. S8-1).

PK/PD of M8891 in tumor-bearing mice

The Caki-1 xenograft model was also used to evaluate the PK/PD relationship of M8891 (Fig. 4D). M8891 exposure increased dose proportionally in Caki-1 xenograft-bearing mice. Peak concentrations were reached at the earliest sampling time point of 1 hour and M8891 was almost eliminated over the course of 24 hours (lower limit of quantification = 0.4 ng/mL). MetEF1a-1 was detectable in all tumors from M8891-treated mice, but not in vehicle-treated animals. MetEF1a-1 signal increased up to 7 hours after oral gavage and then declined over the course of 48 to 96 hours. Both MetEF1a-1 rise (PD_max delayed as compared with C_max) and decline were offset in relation to the plasma PK of M8891, suggesting that PK and PD are not directly linked and that other factors contribute to the dynamics of MetEF1a-1 accumulation and removal. However, with increasing plasma exposure of M8891, the biomarker signal increased accordingly, suggesting that orally administered M8891 inhibits MetAP2-dependent processing of its substrate MetEF1a-1 in a dose- and time-dependent manner in tumor tissue.

Antitumor activity of M8891 in selected PDX models

Based on the screening of PDX models in vitro (Fig. 1E), we selected 10 patient-derived tumor models of different origin. In 6 of 10 PDX models, statistically significant antitumor activity was observed (see Supplementary Table S9-1); tumor volume graphs for gastric cancer model GXF1172 and colon cancer model CXF1783 are shown (Fig. 4E and F, Supplementary S8-2 and S8-3). Under treatment with 150 mg/kg M8891, statistically significant tumor growth inhibition relative to vehicle-treated tumors was observed in both models. Due to the growth kinetics of tumors, the mice in the vehicle group had to be sacrificed on day 28 for GXF1172 and on day 49 for CXF1783, respectively, while the M8891 treatment arm was continued. Remarkably, the tumor growth curves flattened further during continuous treatment with M8891, and tumors stopped growing in a fraction of mice (7/10 mice for GXF1172, 6/10 mice for CXF1783).

Combination of M8891 and VEGFR inhibitors in RCC PDX models

In subsequent studies, we focused on the evaluation of RCC for several reasons: first, we observed strong, dose-dependent antitumor activity in the RCC Caki-1 xenograft, a model that has functional p53 (23); second, ccRCC is known to have a very low prevalence of TP53 mutations (2.7% according to The Cancer Genome Atlas [TCGA]); and third, RCC is known to be sensitive to antiangiogenic agents like sunitinib and bevacizumab, which are approved treatments in this indication (24). We postulated that due to differences in the mechanism of action of antiangiogenic agents and MetAP2 inhibitors, combination therapy may result in increased antitumor activity, especially combinations with orally administered agents such as receptor tyrosine kinase inhibitors.

We selected 25 RCC PDX models and evaluated the antitumor activity of M8891 as monotherapy as well as in combination with sunitinib (Fig. 5; see Supplementary Table S10-1). An integrated survival analysis based on RTV changes (AUC) across all 25 PDX models (Fig. 5A) showed a clear survival benefit for sunitinib but not for M8891 monotherapy. Remarkably, the combination of M8891 with sunitinib demonstrated clear superiority versus all other arms, achieving tumor regression in three models (RXF-2264, RXF-2717, and RXF-2796; Fig. 5B–E; see Supplementary Table S10-1 and Supplementary Figs. S11-1, S11-3, and S11-4). As in the case of GXF-1172 and CXF-1783 (Fig. 4E and F), tumor growth of RXF-SMTCA75 ceased after ∼50 days of treatment, suggesting that M8891 alone or in combination with sunitinib may require time to unfold full antitumor efficacy. Of note, M8891 also demonstrated antitumor efficacy in combination with two other VEGFR inhibitors, cabozantinib and axitinib, combined with good tolerability (Fig. 5F–I, see Supplementary Fig. S11-1 to S11-8), indicating a benefit of M8891 with VEGFR inhibitors in general.

Figure 5.

TP53 mutation, VHL status, and MetAP1/MetAP2 expression score are associated with response in the RCC PDX study. A, Kaplan–Meier curve integrating response data for all 25 tested RCC models (five mice per model = 125 mice) to visualize the progression-free survival benefit (73% tumor growth increase was used as a cutoff to define progression) in models treated with combination sunitinib and MetAP2 inhibitor compared with single agent or vehicle. B, Tumor growth curves of RCC PDX model RXF-2264 treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, 5 days on, 2 days off, per os), or the combination of M8891 and sunitinib are shown. C, Tumor growth curve of RCC PDX model RXF-SMTCA75 treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, 5 days on, 2 days off, per os) or the combination of M8891 and sunitinib is shown. D, Tumor growth curves of RCC PDX model RXF-2717 treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, 5 days on, 2 days off, per os), or the combination of M8891 and sunitinib are shown. E, Tumor growth curve of RCC PDX model RXF-2796 treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, 5 days on, 2 days off, per os) or the combination of M8891 and sunitinib is shown. F, Tumor growth curves of RCC PDX model RXF-2717 treated with vehicle, M8891 (100 mg/kg, daily, per os), axitinib (4 mg/kg, 5 days on, 2 days off, bidaily, per os), or the combination of M8891 and axitinib are shown. G, Tumor growth curve of RCC PDX model RXF-2796 treated with vehicle, M8891 (100 mg/kg, daily, per os), axitinib (4 mg/kg, 5 days on, 2 days off, bidaily, per os) or the combination of M8891 and axitinib is shown. H, Tumor growth curves of RCC PDX model RXF-2717 treated with vehicle, M8891 (100 mg/kg, daily, per os), cabozantinib (10 mg/kg, 5 days on, 2 days off, daily, per os), or the combination of M8891 and cabozantinib are shown. Note that vehicle and M8891 treatment groups are identical to F. I, Tumor growth curve of RCC PDX model RXF-2796 treated with vehicle, M8891 (100 mg/kg, daily, per os), cabozantinib (10 mg/kg, 5 days on, 2 days off, bidaily, per os) or the combination of M8891 and cabozantinib is shown. Note that vehicle and M8891 treatment groups are identical to G. J, Waterfall plots visualizing median RTV AUC values (summary parameter of response) for all tested models for M8891, sunitinib, and the combination, complemented with heat maps of molecular markers TP53 and VHL mutation status (red denotes mutated and green wild-type) and MetAP1 and MetAP2 expression score (green denotes low/middle expression and red high/middle). Abbreviations: AUC, area under the curve; MetAP, methionine aminopeptidases; PDX, patient-derived xenograft; RTV, relative tumor volume; SEM, standard error of the mean; VHL, Von Hippel-Landau.

Identification of putative predictive biomarkers for response to M8891/sunitinib combination

All RCC models were subjected to whole-genome sequencing and gene-expression analyses with the objective of identifying mutation or expression markers correlating with the observed response pattern. RCC models were ranked according to the observed combination benefit (Fig. 5J) and analyzed. Mutations in the tumor suppressor TP53 were of low prevalence [4 of 25 models with loss-of-function (LOF) mutations], in agreement with the low frequency observed in TCGA data. Notably, mutations in TP53 were absent in the top 15 ranked models, in line with our hypothesis that p53 is a relevant mediator of the response to MetAP2 inhibition (Fig. 5J). In contrast, mutations in the VHL tumor suppressor gene, which is frequently inactivated in RCC, were found in 10 of 25 models, with no overlap with p53 LOF mutations. Interestingly, 9 of the 15 top-ranked models harbored VHL mutations, whereas only 1 of the remaining 10 less responsive models harbored a mutation. Both candidate biomarkers reached statistical significance compared with the sunitinib monotherapy treatment group [P value (TP53) = 0.013; P value (VHL) = 0.016].

Based on the CRISPR screen results and subsequent KO study, a potential effect of MetAP1 and MetAP2 on treatment outcomes would be expected. As no functionally relevant and frequent mutations in MetAP1/2 genes are known, we investigated the influence of gene expression on response. We used median split from RNA-seq vst-transformed values to categorize each expression biomarker gene into high/mid and mid/low groups. Although mRNA levels of either MetAP gene alone did not correlate with the observed response ranking, an integrated MetAP1 and MetAP2 expression score (the score was assigned to the high/mid category if at least 1 gene was defined as high/mid) showed a statistically significant association with response [P value (MetAP1/MetAP2 score) = 0.0064]. Overall, we observed a trend for an enrichment of low/mid METAP1/2-expressing models among responders to M8891/sunitinib combination therapy. As already suggested by the results in MetAP1 KO cells (Figs. 2C and 3A), functional redundancy could explain the relevance of MetAP expression for the antiproliferative activity of MetAP2 inhibitors. Further analysis of available genomic and expression data did not suggest additional associations with the observed PDX response to M8891–sunitinib combination therapy. Taken together, our results suggest that mutation status of both tumor suppressors TP53 and VHL, as well as expression of both MetAP1 and MetAP2, may be relevant parameters modifying the antitumor activity of M8891 in combination with VEGFR inhibitors in RCC; however, additional unknown factors may further influence the response to MetAP2 inhibition and combination therapies.

Discussion

M8891 is a novel, potent, and highly selective inhibitor of human and murine MetAP2 with a well-described binding mode in the catalytic pocket of the enzyme (14). Various MetAP2 inhibitors have been described to date and their common denominator is that they target angiogenesis as well as the tumor compartment (25). Here, we characterize the new MetAP2 inhibitor M8891 in vitro and in vivo. M8891 blocked the proliferation of endothelial and tumor cells. In a panel of patient-derived tumors of diverse tissue origin, approximately 60% of all PDX models were potently inhibited by M8891 (threshold 1E−06 mol/L). Proliferation inhibition varied between 40% and 100%, suggesting that tumor cell lines rely on MetAP2 to different extents.

To identify factors contributing to the response to MetAP2 inhibition, we performed CRISPR genome-wide screens with three cancer cell lines (A549, HT1080, and U87MG), which we had identified as sensitive to treatment with MetAP2 inhibitors. Most strikingly, guides targeting the tumor suppressors TP53 and cyclin-dependent kinase inhibitor CDKN1A were enriched in A549 and HT1080, enrichment was less evident in U87MG. In contrast, guides targeting MetAP1 or MetAP2 were reduced in A549, whereas HT1080 and U87MG showed a decrease for MetAP1 only. Although several other candidates modulating the response to MetAP2 inhibition were identified (see Supplementary Table S7-1), we focused on further characterizing TP53 and MetAP1 in this study. Consistent with the results from the CRISPR screens, p53 KO in A549 cells conferred resistance to MetAP2 inhibition, validating this tumor suppressor as an important determinant of response in this cell line. MetAP1 KO in A549 was generated to assess its impact on pharmacologic inhibition by M8891. In two independently generated KO clones, MetAP1 protein was strongly reduced (heterozygous deletion of MetAP1) or undetectable (homozygous KO). Both KO clones were clearly more sensitive to M8891, which was reflected in a 5-fold reduction of IC₅₀ value, as well as increased efficacy.

Our functional analyses of MetAP1 and TP53 helped to better understand the mechanism of action of MetAP2 inhibition. The influence of MetAP1 on antiproliferative activity by MetAP2 inhibitors observed in our studies suggests that some level of functional redundancy exists, and the total MetAP level/activity (MetAP1 and MetAP2) may be a key determinant of sensitivity. Previous studies in yeast and human cells demonstrating the interdependence between MetAP1 and MetAP2 further support this concept (6, 26). These data strongly suggest that dual inhibition of MetAP1 and MetAP2 may be more efficacious in inhibiting tumor cell proliferation; however, normal tissues may also be affected by dual inhibition.

Pharmacologic inhibition of MetAP2 potentially blocks the processing of a broad and diverse range of substrates; the functional impact on activity, localization, or stability of various proteins may be diverse. Understanding of MetAP1 and MetAP2 substrates may be incomplete and, therefore, additional research is necessary to fully reveal the detailed mechanism of the antiproliferative effects in endothelial and/or cancer cells. With M8891, we have generated a highly selective pharmacologic inhibitor that enables such studies. In A549 cells, the inactivation of TP53 and CDKN1A clearly blunts the response to M8891, suggesting that TP53 activation and transcriptional induction of its downstream genes like CDKN1A contribute to the antiproliferative activity of MetAP2 inhibition. Several independent studies using other MetAP2 inhibitors or mouse KO studies support this functional relationship (11, 27); however, the finding that some p53-mutant cancer cell lines are sensitive to MetAP2 inhibitors suggests that other pathways exist that can limit tumor growth/proliferation (28). It remains to be elucidated exactly how p53 gets activated upon pharmacologic inhibition, but it is conceivable that the accumulation of unprocessed protein substrates is a primary trigger for the activation of p53. Currently, there is no evidence that p53 is a direct substrate of MetAP2; in fact, acidic residues in the P2’–P5’ position in human p53 (P1’–P5’: MEEPQ) strongly disfavor cleavage by human MetAP enzymes (29).

We have identified EF1a-1 as a novel substrate for MetAP2 and generated specific antibodies detecting the methionine-containing, unprocessed EF1a-1 protein (MetEF1a-1). Processing of EF1a-1 served as a robust and quantifiable indicator for inhibition of MetAP2, whereas we were unable to reliably detect other proposed MetAP2 substrates such as GAPDH and 14-3-3γ (21, 30) with commercially available antibodies. The concentration–response relationship for MetEF1a-1 formation in cancer cell lines was consistent with other downstream effects, such as inhibition of cell proliferation, and, consequently, we selected this biomarker to assess target inhibition in mouse xenograft studies. MetEF1a-1 accumulation in tumor biopsies was clearly dose- and time-dependent but did not correlate with the concentration of M8891 in the circulation, as indicated by the delayed increase and decrease in comparison with the M8891 concentration in plasma. It is conceivable that the dynamics of MetEF1a-1 levels not only depend on protein translation rate but also on the kinetics of MetEF1a-1 protein degradation, resulting in a more complex PK/PD relationship. As the PK in a tumor is correlated with plasma PK, we can rule out tumor accumulation and retention as a factor influencing PD.

M8891 strongly inhibited the growth of new blood vessels in Matrigel plugs implanted in mice. Antiangiogenic activity of M8891 was comparable to a murine VEGF-targeting antibody and showed strong activity even at the lowest dose tested. Tumor growth was strongly and significantly delayed in the Caki-1 ccRCC xenograft, as well as in several PDXs, albeit at doses higher than in the angiogenesis assay in vivo. Of note, the tumor growth behavior in several tumor studies was biphasic, with an initial tumor growth delay switching to tumor stasis after some time of treatment.

To further explore the therapeutic potential of M8891, we reasoned that dual targeting of the vascular and tumor compartments would be a promising strategy and selected RCC as a tumor type in which antiangiogenic therapies are in use and the frequency of p53 LOF mutations is low. In a cohort of 25 RCC PDX models, we evaluated the activity of sunitinib, which is an approved standard-of-care agent for the treatment of advanced RCC (24), and M8891, as well as a combination of both agents, on tumor growth. Under the dose and schedule tested in this study, M8891 showed only minor antitumor activity, classified as weak progression in a few models, whereas the activity of sunitinib was more pronounced but still considered to be moderate in the tested tumor panel. Strikingly, the combination of sunitinib and M8891 demonstrated strong antitumor activity, with approximately one-third of the models showing regression/stasis, one-third growth inhibition, and the remaining models no benefit. Again, antitumor activity in the tested mouse models was long-lasting, even in progressing models such as RXF-SMTCA75, in which tumor growth stopped at later time points (after day 40). Combination efficacy studies with axitinib or cabozantinib also revealed enhancement of antitumor efficacy, suggesting a general combination benefit with VEGFR inhibition. Of the 25 tumor models, 21 were p53 wild-type and four were p53 mutated, confirming the low prevalence of functional p53 loss in RCC. A responder group analysis (combination versus sunitinib alone) showed that all four p53-mutated models clustered in the low responder group, a statistically significant finding.

Mutations in the tumor suppressor VHL were more prevalent in RCC and 10 of 25 models with functional VHL loss were identified. Interestingly, 9 of 10 VHL-mutated models clustered in the responder group, and the association was also found to be statistically significant. However, it should be noted that the Caki-1 model is sensitive to M8891 and has functional VHL. A functional relationship between MetAP2 and VHL has recently been investigated, but it is currently unclear how VHL LOF mechanistically sensitizes tumors to MetAP2 inhibition (31).

Although the expression of MetAP1 or MetAP2 alone did not correlate with response, the integrated expression score for both MetAP enzymes was significantly associated with response. A mid/low expression score of MetAP1 and MetAP2 was noted in six of nine top responding models, signifying that there is indeed a functional redundancy relationship in cells.

It is important to note that the studies performed focused on the impact of MetAP2 inhibition on the tumor and endothelial cell compartments but did not investigate modulation of the immune system. Immune modulatory effects of MetAP2 inhibition have been investigated anecdotally, but further, more systematic studies in immune-competent mouse tumor models are required to better characterize its function (32, 33). Further development of M8891 will be managed by the Oncoteq AG subsidiary of Cureteq AG (34), which licensed M8891 from Merck KGaA following the determination of the recommended phase II dose in a phase Ia trial (35) and the strategic portfolio decision by Merck KGaA. The preclinical data generated in this study as well as clinical phase I data on safety, tolerability, and pharmacokinetics/-dynamics (35) represent a solid foundation for further clinical exploration.

In summary, we have characterized a potent and selective MetAP2 inhibitor with strong antiangiogenic and antitumoral activity showing superior efficacy to monotherapy in combination with sunitinib in a panel of patient-derived RCC models. So far, M8891 is the first reversible, potent, and highly selective MetAP2 inhibitor that has progressed to a phase I clinical trial (NCT03138538). In conjunction with our identified candidate biomarkers, including TP53, VHL, and MetAP1/MetAP2 expression, MetAP2 inhibitor combinations represent an intriguing regimen for RCC and warrant in-depth clinical investigation.

Supplementary Material

Supplementary Figure S1-1

Increasing concentrations of M8891 (3E−09M to 1E−04M) were tested in an MetAP1 enzyme assay

Supplementary Figure S1-2

Increasing concentrations of M8891 (3E−10M to 1E−05M) were tested in a murine MetAP2 enzyme assay

Supplementary Figure S2-1

Representative photomicrographs of HUVEC tube formation after treatment with M8891

Supplementary Figure S2-2

HUVEC tube formation assay

Supplementary Figure S5-1

Treatment of three cell lines (bEND3, HT1080, HCT116) with MetAP2 inhibitors 1E−07 M TNP-470 or 1E−06 M A-832234 for 48 hours followed by IEF mobility shift separation of extracted protein lysates (lanes 1, 5, 9: protein marker).

Supplementary Figure S5-2

Protein identification: stained bands (upper and lower) were excised, trypsin in-gel digested, and analyzed by peptide mapping using MALDI MS

Supplementary Figure S5-3

Peptide sequencing by MS fragmentation: the amino acid sequence identified both peptides as N-terminal tryptic peptide of elongation factor 1 alpha-1 (EF1a-1; UniProtKB - P68104).

Supplementary Figure S5-4

Immuno-analytical verification of EF1a-1 as MetAP2 substrate: bEND3 cells were treated by DMSO (control) and 1E−07 M TNP-470.

Supplementary Figure S6-1

SDS-PAGE analysis of rabbit monoclonal antibody MKV-3-165-11 before and after purification by Protein A affinity chromatography

Supplementary Figure S6-2

EF1a-1 knockdown (96 hours) in A549 or HCT116 cancer cells, which were treated for 24 hours with DMSO or 1E−07 M of M8891 and analyzed for the presence of MetEF1a-1 by Simple Western using the generated rabbit monoclonal antibody MKV-3-165-11.

Supplementary Figure S8-1

Body weight curves of Caki-1 ccRCC xenograft-bearing mice treated with vehicle or M8891 (10, 25, 50 mg/kg twice daily, per os) as monotherapy

Supplementary Figure S8-2

Body weight curves of gastric cancer PDX model GXF-1172-bearing mice treated with vehicle or M8891 (150 mg/kg, twice daily, per os) as monotherapy

Supplementary methods S2

Description of effect of M8891 on HUVEC formation on 3D Matrigel methodology

Supplementary methods S6

Generation of rabbit monoclonal antibody against unprocessed human EF1a

Supplementary Figure S8-3

Body weight curves of colon cancer PDX model CXF-1783-bearing mice treated with vehicle or M8891 (150 mg/kg, twice daily, per os) monotherapy

Supplementary Figure S11-1

Body weight curves of RCC PDX model RXF-2264-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, daily, per os), or the combination are shown

Supplementary Figure S11-2

Body weight curves of RCC PDX model RXF-SMTCA75-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, daily, per os), or the combination are shown

Supplementary Figure S11-3

Body weight curves for RCC PDX model RXF-2717-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, daily, per os), or the combination are shown

Supplementary Figure S11-4

Body weight curves for RCC PDX model RXF-2796-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), sunitinib (40 mg/kg, daily, per os), or the combination are shown

Supplementary Figure S11-5

Body weight curves for renal cancer PDX model RXF-2717-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), axitinib (4 mg/kg, twice daily, per os), or the combination are shown

Supplementary Figure S11-6

Body weight curves for RCC PDX model RXF-2717-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), cabozantinib (10 mg/kg, daily, per os), or the combination are shown

Supplementary Figure S11-7

Body weight curves for RCC PDX model RXF-2796-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), axitinib (4 mg/kg, twice daily, per os), or the combination are shown

Supplementary Figure S11-8

Body weight curves for RCC PDX model RXF-2796-bearing mice treated with vehicle, M8891 (100 mg/kg, daily, per os), cabozantinib (10 mg/kg, daily, per os), or the combination are shown

Supplementary Table S3-1

IC50 and efficacy of M8891 in cancer cell lines

Supplementary Table S4-1

Tabulated IC50 and efficacy in PDX cancer cell (No. 1–125) and cancer cell line (No. 126–130) models in soft agar assays in vitro

Supplementary Table S6-1

ELISA data for increasing dilutions of unpurified, purified antibody MKV-3-165-11 and flow through fraction

Supplementary Table S7-1

Bayes Factor (BF) scores of MetAP1 and MetAP2 in the screens

Supplementary Table S7-2

Top CRISPR hits that score as suppressors or sensitizers

Supplementary Table S9-1

Antitumor activity of M8891 monotherapy evaluated in vivo in 10 PDX models (NSCLC, head and neck, colon, osteosarcoma, and gastric cancer) selected from the PDX in vitro evaluation (IC50 <1E−06 M)

Supplementary Table S10-1

Summary of area under the curve values for in vivo RCC PDX study

This article is featured in Selected Articles from This Issue, p. 125

Authors' Disclosures

M. Friese-Hamim reports Manja Friese-Hamim is an employee of Merck KGaA group function of Merck KGaA. Merck, KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors, including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. Manja Friese-Hamim is named inventor on the following patents: WO2013/149704, WO2016/020031,and WO2022/008469. O. Bogatyrova reports Olga Bogatyrova is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. Olga Bogatyrova is named inventor on the following patent: WO2022/008469. M. Keil reports Marina Keil is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Marina Keil not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. F. Rohdich reports Felix Rohdich is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patent applications are licensed to Oncoteq AG. Felix Rohdich is named inventor on the following patent applications: WO 2022/008469 and WO 2016/020031. B. Blume reports Beatrix Blume is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck Healthcare KGaA and/or its affiliates have certain rights in patents, patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Beatrix Blume not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. B. Leuthner reports Birgitta Leuthner is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck Healthcare KGaA and/or its affiliates have certain rights in patents, patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Birgitta Leuthner not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. F. Czauderna reports Frank Czauderna is an employee of EMD Serono Research and Development Institute, Inc., Billerica, MA, USA. EMD Serono Research and Development Institute, Inc. is an affiliate of Merck KGaA Darmstadt, Germany. Merck Healthcare KGaA and/or its affiliates have certain rights in patents, patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). Frank Czauderna is not listed as an inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. D. Hahn reports Diane D. Hahn has been an employee of Merck Healthcare KGaA, healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891. (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882) and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. Diane Hahn is named inventor on the following patent: WO2016/020031. J. Jabs reports Julia Jabs is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. F. Jaehrling reports Merck Healthcare KGaA and/or its affiliates have certain rights in patents, patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Frank Jaehrling not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. T. Heinrich reports Timo Heinrich is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. Timo Heinrich is named inventor on the following patents: WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031. D. Wienke reports Dirk Wienke is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck Healthcare KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Dirk Wienke not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. J. Moffat reports other support from Merck KGaA during the conduct of the study. A. Blaukat reports Merck Healthcare KGaA and/or its affiliates have certain rights in patents, patent applications pertaining to M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469) with Andree Blaukat not listed as inventor on the indicated patents. These patents are licensed to Oncoteq AG, Zug, Switzerland. F.T. Zenke reports Frank Zenke is an employee of Merck Healthcare KGaA, the healthcare business of Merck KGaA, Darmstadt, Germany. Merck KGaA and/or its affiliates have certain rights in patents and patent applications pertaining to MetAP2 inhibitors including M8891 (WO2013/149704, WO2012/048775, WO2021/001328, and WO2016/020031), EEF1A1 (WO2010/051882), and a combination (WO2022/008469). These patents are licensed to Oncoteq AG. Frank Zenke is named inventor on the following patents: WO2013/149704, WO2012/048775, WO2021/001328, WO2016/020031, WO2010/051882, and WO2022/008469. No disclosures were reported by the other authors.

Authors' Contributions

M. Friese-Hamim: Conceptualization, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. M.J. Ortiz Ruiz: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, writing–original draft, writing–review and editing. O. Bogatyrova: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, writing–original draft, writing–review and editing. M. Keil: Data curation, formal analysis, validation, investigation, visualization, writing–original draft, writing–review and editing. F. Rohdich: Conceptualization, formal analysis, validation, investigation, visualization, writing–original draft, writing–review and editing. B. Blume: Data curation, formal analysis, validation, investigation, visualization, writing–review and editing. B. Leuthner: Data curation, formal analysis, validation, investigation, visualization, writing–review and editing. F. Czauderna: Conceptualization, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. D. Hahn: Data curation, formal analysis, validation, investigation, visualization, writing–review and editing. J. Jabs: Writing–review and editing. F. Jaehrling: Data curation, formal analysis, validation, investigation, writing–review and editing. T. Heinrich: Conceptualization, resources, validation, writing–original draft, writing–review and editing. R. Kellner: Data curation, formal analysis, validation, investigation, visualization, writing–review and editing. K. Chan: Resources, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. A.H. Tong: Resources, data curation, formal analysis, validation, investigation, visualization, writing–review and editing. D. Wienke: Conceptualization, supervision, validation, writing–original draft, project administration, writing–review and editing. J. Moffat: Resources, data curation, formal analysis, validation, investigation, writing–review and editing. A. Blaukat: Conceptualization, supervision, writing–review and editing. F.T. Zenke: Conceptualization, supervision, validation, visualization, writing–original draft, project administration, writing–review and editing.