Combined inhibition of de novo pyrimidine synthesis and ATR promotes ATF4‐mediated cell death in TP53‐deficient ALL

Abstract

Mutations or deletions affecting the TP53 gene predict a dismal outcome in relapsed acute lymphoblastic leukemia (ALL). Loss of p53 function compromises the response to many anti‐leukemic therapies, underscoring the need for agents that are effective in TP53‐deficient cells. Leukemic blasts depend on de novo pyrimidine synthesis to sustain proliferation, and the mitochondrial enzyme dihydroorotate dehydrogenase (DHODH) is essential for this pathway. While DHODH inhibitors (DHODHis) show preclinical anti‐leukemic activity, the efficacy in TP53‐deficient contexts has not been tested. Using both isogenic TP53‐wildtype and knock‐out cell line models and patient‐derived xenografts (PDXs), we show that dual inhibition of DHODH and Ataxia Telangiectasia and Rad3‐related (ATR) produces pronounced anti‐leukemic effects, irrespective of the TP53 status. Through integrated transcriptomic and metabolomic analyses, we show that combined inhibition of DHODH and ATR leads to a reduced flux of glucose into the TCA cycle, accompanied by an increase in oxidative stress. This metabolic phenotype triggers cell death through a mechanism converging on Activating Transcription Factor 4 (ATF4), a key integrator of cellular responses to metabolic and oxidative stress, which operates largely independently of p53. Taken together, our findings identify ATF4‐mediated cell death as a previously unrecognized vulnerability in TP53‐deficient ALL.

INTRODUCTION

TP53 is the most mutated gene across human cancers. However, while its mutation frequency exceeds 90% in some solid tumors, TP53 alterations are relatively rare in acute leukemias. ¹ , ² Particularly in primary B‐cell precursor acute lymphoblastic leukemia (BCP‐ALL), alterations in the TP53 gene are infrequent, with an incidence lower than 3% at diagnosis. ³ In relapsed BCP‐ALL, however, the frequency of TP53 mutations increases to about 12%, where it serves as a strong and independent predictor of treatment failure. ⁴ Low hypodiploid ALL represents an exception to the otherwise low frequency of TP53 mutations in BCP‐ALL, with approximately 90% of the cases exhibiting TP53 mutations or deletions. ⁵ Notably, more than half of these alterations are of germline origin and associated with Li‐Fraumeni syndrome. ⁵ , ⁶ Clinically, low hypodiploid BCP‐ALL is also characterized by poor treatment outcomes, with a 5‐year overall survival rate of less than 60%. ⁵ , ⁷

The p53 protein, encoded by the TP53 gene, acts as a transcription factor and is best known for controlling genes involved in cell cycle, apoptosis, and metabolism. ⁸ , ⁹ Since the pro‐apoptotic effects of most cytotoxic drugs primarily act via p53, loss of p53 function compromises the response to not only many existing anti‐leukemic agents, but also to newly introduced targeted therapies, such as the anti‐CD22 targeting antibody‐drug conjugate inotuzumab ozogamicin. ¹⁰ Moreover, although CAR‐T‐cell therapy has become an important salvage therapy in the treatment of (relapsed) BCP‐ALL, we and others have shown that TP53‐mutated tumors frequently escape from these engineered cytotoxic T cells. ¹¹ , ¹² , ¹³ Hence, there is a critical need to develop effective therapeutic strategies specifically tailored to this highly aggressive leukemia subtype.

To sustain their high proliferation rates, leukemic blasts are strongly reliant on nucleotide synthesis to support DNA replication, transcription, and ribosome synthesis. For this reason, antimetabolites targeting nucleotide biosynthesis, such as methotrexate, 6‐mercaptopurine, and cytarabine, have become indispensable components of current ALL treatment protocols. ¹⁴ However, small‐molecule inhibitors targeting rate‐limiting enzymes within nucleotide biosynthesis pathways are advancing as potentially less toxic alternatives to these conventional agents. A leading candidate target in this class is dihydroorotate dehydrogenase (DHODH), a key mitochondrial enzyme that catalyzes the fourth and rate‐limiting step of the de novo pyrimidine synthesis pathway. ¹⁵ , ¹⁶ DHODH inhibitors (DHODHis) were first approved as immunomodulatory agents for autoimmune indications, exemplified by their routine use in rheumatoid arthritis ¹⁷ and multiple sclerosis. ¹⁸ More recently, accumulating preclinical and early‐phase clinical data have prompted their evaluation as anti‐neoplastic agents, with several studies assessing their therapeutic activity in hematological malignancies such as acute myeloid leukemia (AML) ¹⁹ , ²⁰ , ²¹ , ²² and ALL. ²³ The effects of DHODHis on tumor cells are complex and most likely context‐dependent. Several studies attribute the anti‐tumor effects of these drugs to increased replication fork stalling and a subsequent proliferation arrest and apoptosis. ²⁴ , ²⁵ However, recent publications point to a requirement for DHODH in preventing ferroptosis, ²⁶ an iron‐dependent cell death mechanism involving lipid peroxidation, which, depending on the cellular context, is either enhanced or suppressed by p53. ²⁷ Here, we explored the response of isogenic homozygous TP53‐knock‐out and TP53‐wildtype BCP‐ALL cells to the DHODHi orludodstat (BAY‐2402234) and performed high‐throughput drug screens to identify drugs that effectively synergize with DHODHi in TP53‐deficient ALL.

METHODS

Detailed methods can be found in the Supporting Information S1: Supplemental Methods section.

Cell models

Generation of isogenic homozygous TP53 ^KO and TP53 ^WT Nalm6 (ACC 548) and RCH‐ACV (ACC 128) using CRISPR/Cas9 has been described before. ²⁸ BCP‐ALL cell lines were cultured for a maximum of 3 months in RPMI 1640 + GlutaMAX (Gibco, 72400‐054) supplemented with 10% fetal bovine serum (Bodinco, lot BDC‐16514) and 1% Pen/Strep (Gibco, 15140‐122) at a cell density between 0.2 and 3 × 10⁶ cells per milliliter. Cell line identity was confirmed by DNA fingerprinting, and cultures were assessed every 6 weeks for the presence of Mycoplasma. PDXs were previously generated by intrafemoral injection of 0.25 × 10⁵ to 1 × 10⁶ viable primary BCP‐ALL cells in NOD.Cg‐Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG)‐mice, and TP53 genotype was previously characterized ²⁸ , ²⁹ (Table 1).

Table 1.

Patient‐derived xenograft genotypes.

Patient‐derived xenografts
Sample	Status	Copy number	Mutation status
PDX 7—diagnosis	Wildtype	WT	WT
PDX 7—relapse	Aberrant	WT	R273C
PDX 8—diagnosis	Wildtype	WT	WT
PDX 8—relapse	Aberrant	Monoallelic loss	p.R280T
PDX 9—diagnosis	Wildtype	WT	WT
PDX 9—relapse	Aberrant	WT	R248Q

Note: Copy number and mutation status are shown.

Abbreviation: PDX, patient‐derived xenograft.

High‐throughput combination drug screens

For drug screenings, RCH‐ACV TP53 ^KO and Nalm6 TP53 ^KO cells were seeded at 0.36 × 10⁶ cells/mL in 384‐well plates using a Multidrop Combi Reagent Dispenser (Thermo Scientific). A total of 224 standard‐of‐care drugs as well as early‐phase clinical trial compounds, dissolved in DMSO or water, were added with or without 1 nM orludodstat shortly after seeding. Drugs were added with a fully automated system at the high‐throughput screening facility, using a Biomek i7 liquid handler and the Echo550 dispenser. After 72 h, MTT conversion was performed according to the manufacturer's instructions, and area under the curve (AUC) values were calculated. The AUC ratio was calculated by dividing the AUC values from combination treatments by the AUC values from orludodstat single treatment. Drug combinations yielding an AUC ratio < 0.9 were considered of interest, as they indicated enhanced reduction in cell viability compared to orludodstat alone.

In vivo studies

At the start of the experiments, mice were 6–20 weeks old (median: 6.4 weeks). For orludodstat monotherapy, starting ten days post‐engraftment, mice were randomly assigned to groups (n = 5 per group) and treated orally with either 4 mg/kg orludodstat (5 days on, 2 days off) or vehicle for up to 28 days or until a predefined endpoint was reached. To evaluate combination treatment, animals were randomly assigned to groups (n = 7 per group) and treated orally with either 4 mg/kg orludodstat and/or 60 mg/kg berzosertib, or vehicle for 28 days or until an endpoint was reached. Tumor load was monitored weekly. When the percentage of human lymphoblasts exceeded 50%, mice were humanely euthanized by CO₂ asphyxiation. For the event‐free survival analysis, we used interpolated growth curve fitting to calculate the time until 30% tumor load.

Liquid chromatography–mass spectrometry–based metabolomics

RCH‐ACV TP53 ^WT and TP53 ^KO cells (0.5 × 10⁶) were seeded in triplicate and pre‐treated for 24 h with either dimethyl sulfoxide (DMSO) vehicle or 4 nM orludodstat. Cells were then washed with phosphate‐buffered saline (PBS) and reseeded in glucose‐free RPMI 1640 medium (Gibco, 11879‐020) supplemented with 11.1 mM [U‐¹³C₆]‐glucose (Cambridge Isotopes, CLM‐1396‐PK), 10% fetal bovine serum (FBS), and 1% Pen/Strep. Treatments with DMSO vehicle, 4 nM orludodstat, 100 nM berzosertib, or the combination were initiated concurrently and maintained for 24 h during isotope labeling. Cells were washed with cold PBS, and pellets were lysed in 50–75 µL ice‐cold lysis solvent composed of methanol/acetonitrile/Milli‐Q (2:2:1) for 10 min in a 4°C Thermoshaker. After that, samples were cleared by centrifugation at 12.000 rcf for 15 min and supernatants were frozen at −80°C until Liquid chromatography–mass spectrometry (LC–MS) analysis.

Analysis of metabolites was performed on a Q Exactive HF mass spectrometer (Thermo Scientific) coupled to a Vanquish autosampler and pump (Thermo Scientific). The MS was operated in polarity‐switching mode with spray voltages of −3.5 and 4.5 kV. Sheath gas, auxiliary gas, and sweep gas flow rates were 35, 10, and 1 units, respectively. Separation of metabolites was done using a Sequant ZIC‐pHILIC column (2.1 × 150 mm, 5 μm; Merck) coupled to a ZIC‐pHILIC guard column (2.1 × 20 mm, 5 μm; Merck), using elution buffer A (acetonitrile), and B (20 mM (NH₄)₂CO₃, 0.1% NH₄OH in LC/MS grade water (Biosolve)). Flow rates were set at 100 μL/min, and a gradient was run from 80%A to 20%A.

Data were processed in TraceFinder (Thermo). Metabolites were identified by exact mass (±5 ppm), m/z, retention time, and peak shape, using reference standards. Intensities were normalized to total ion count, and isotope distributions were corrected for natural ¹³C abundance. Log₂ fold‐changes (log₂FC) in metabolite levels were calculated relative to DMSO controls and binned (range: −5 to 5) for visualization. Differences were assessed by one‐way analysis of variance (ANOVA) with Dunnett's post hoc test versus DMSO.

RESULTS

DHODH inhibition induces replication stress and synergizes with ATR inhibition

TP53 mutations frequently drive therapy resistance by altering cellular responses to stress. To determine whether TP53 status affects sensitivity to inhibition of the de novo pyrimidine synthesis enzyme DHODH, we exposed isogenic TP53‐wildtype (TP53 ^WT) and TP53‐knock‐out (TP53 ^KO) derivatives of the BCP‐ALL cell lines Nalm6 and RCH‐ACV ²⁸ to various concentrations of the DHODHi orludodstat (BAY‐2402234). In both genotypes, orludodstat treatment reduced cell viability. Although growth rates were identical between the isogenic models, ²⁸ TP53 ^WT cells were more responsive to single‐agent therapy, than their TP53 ^KO counterparts (Figure 1A).

Figure 1.

Dihydroorotate dehydrogenase (DHODH) inhibition induces replication stress and synergizes with Ataxia Telangiectasia and Rad3‐related (ATR) inhibition. (A) Dose–response curves showing cell viability of TP53 ^WT and TP53 ^KO Nalm6 and RCH‐ACV cells treated with the indicated concentrations of orludodstat for 72 h. Each data point represents a mean (±SD) of three independent experiments. The area under the curve (AUC) was determined, and indicated differences were tested for significance using an unpaired two‐tailed Student's t‐test. (B) Cell cycle analysis of cells treated with 8 nM orludodstat, 100 μM uridine, or their combination for 48 h. Cell cycle distribution was assessed by measuring DNA content using flow cytometry following Hoechst33342 staining. Data reflect the percentage of cells in each phase under the indicated treatment conditions, highlighting changes in cell cycle progression. Significance between TP53 ^WT and TP53 ^KO was tested using two‐way analysis of variance (ANOVA). Significance between treatments and dimethyl sulfoxide (DMSO) was determined using Dunnets multiple comparison. (C, D) Western blots visualizing P53, phospho‐Chk1 (Ser345), and Chk1 expression in orludodstat‐treated (C) and cytarabine‐treated (Ara‐C; D) TP53 ^WT and TP53 ^KO Nalm6 (top) and RCH‐ACV cells (bottom). Lamin‐B1 was used as a loading control. Western blot images present representative data replicated in two independent experiments. (E) Kaplan–Meier curve showing event‐free survival of mice engrafted with TP53‐deficient patient‐derived xenograft (PDX) and treated with 4 mg/kg orludodstat or vehicle (5 days on/2 days off). (F) Visual representation of the high‐throughput combination drug screen pipeline. In total, 224 compounds were tested as a single treatment or in combination with orludodstat in RCH‐ACV TP53 ^KO and Nalm6 TP53 ^KO. (G) Venn diagram showing the number of compounds with an AUC ratio of <0.9 (combination/single treatment) in RCH‐ACV TP53 ^KO and Nalm6 TP53 ^KO. (H,I) Dose–response curves (left) and zero interaction potency (ZIP) synergy landscape (right) of one B‐cell precursor acute lymphoblastic leukemia (BCP‐ALL) PDX sample seeded on hTERT‐immortalized mesenchymal stromal cells and treated with the indicated concentrations of orludodstat and/or berzosertib for 72 h. (J) Maximum ZIP synergy scores shown for three diagnosis–relapse pairs of PDX samples treated with orludodstat (0–1 μM) and/or berzosertib (0–20 μM) for 72 h. Diagnosis–relapse couples of the same patient are connected by lines. Each data point represents the synergy scores of one representative replicate.

Orludodstat inhibits the de novo pyrimidine biosynthesis pathway, leading to reduced intracellular pyrimidine levels. Supplementation with exogenous uridine is known to restore pyrimidine pools through the nucleotide salvage pathway, thereby bypassing the de novo pathway and mitigating the effects of DHODHi, such as orludodstat. ²⁵ , ²⁹ , ³⁰ To investigate whether the shortage in pyrimidines resulted in replication fork stalling and therefore cell cycle arrest, and whether this could be rescued by uridine supplementation, we performed DNA content analyses following orludodstat treatment. Consistent with previous reports, ²⁴ , ³¹ orludodstat treatment induced an early S‐phase cell cycle arrest (Figure 1B and Supporting Information S1: Figure 1A,B), indicative of incomplete replication, and increased the proportion of apoptotic sub‐G1 cells (Supporting Information S1: Figure 1A,C). Both effects were fully rescued by the addition of uridine, indicating that the orludodstat‐induced cell cycle effects are due to on‐target activity. A well‐known inducer of replication stress is cytarabine (Ara‐C), a cytidine analog, which is incorporated into the DNA and inhibits DNA polymerase, directly impairing DNA replication and inducing replication stress. ³² Replication stress is subsequently sensed by Ataxia Telangiectasia and Rad3‐related (ATR), a key kinase that coordinates cell cycle checkpoints, stabilizes stalled replication forks, and activates DNA repair pathways to maintain genome integrity. ³³ To determine whether DHODH inhibition triggers a replication stress response similar to that of Ara‐C, ³⁴ and whether this is modulated by TP53 status, we exposed TP53 ^WT and TP53 ^KO cells to either orludodstat or Ara‐C. Both treatments resulted in activation of ATR in all models but the RCH‐ACV TP53 ^KO, as indicated by the phosphorylation of its downstream target Chk1 (Figure 1C,D). Orludodstat increased the accumulation of chromatin‐bound proliferating cell nuclear antigen (PCNA), a widely used marker for replication stress, in all models. Ara‐C increased the accumulation of PCNA in all but the Nalm6 TP53 ^KO (Supporting Information S1: Figure 1D, Supporting Information S2: Supplementary Data 1). These data indicate that orludodstat indeed triggers a replication stress response.

We next evaluated the single‐agent efficacy of orludodstat in vivo using a patient‐derived BCP‐ALL xenograft model of a relapsed ALL with a complete loss of functional p53 (TP53 ^R280T/−). Treatment with orludodstat led to a moderate delay in leukemia development (Figure 1E) without any overt signs of toxicity, such as body weight loss (Supporting Information S1: Figure 1E). These findings indicate that therapeutic inhibition of DHODH is well tolerated and slows progression of TP53‐deficient BCP‐ALL in vivo.

To investigate how we could enhance the therapeutic efficacy of DHODHi in BCP‐ALL, we performed a high‐throughput combination drug screen using TP53‐deficient Nalm6 and RCH‐ACV cells. In this screen, we combined orludodstat with a custom pediatric cancer drug library comprising 224 standard‐of‐care drugs as well as early‐phase clinical trial compounds (Figure 1F and Supporting Information S1: Table 4). To identify compounds that enhance orludodstat's cytotoxic activity, we selected drug combinations with an AUC ratio (combination/orludodstat single treatment) of <0.9, indicating that the combination reduces overall cell viability more than orludodstat alone. We found only two drug combinations effective in both the Nalm6 and RCH‐ACV models: the ATR inhibitor berzosertib (VE‐822) and inhibitor of apoptosis protein (IAP) antagonist AZD5582 (Figure 1G and Supporting Information S1: Figure 1F). Given that berzosertib is currently in phase II clinical trials for various adult solid tumors (NCT02595892, NCT03641547, and NCT03896503) and has demonstrated efficacy in targeting cells undergoing replication stress, ³⁵ , ³⁶ we proceeded to further evaluate the potential of combining orludodstat with berzosertib.

Consistent with the results of the drug screen, we observed effective synergy between orludodstat and berzosertib, both in Nalm6 TP53 ^WT and Nalm6 TP53 ^KO cells, while in RCH‐ACV cells the effects were mostly additive (Supporting Information S1: Figure 1G). In addition, we observed effective synergy in the TP53 ^R248Q hotspot mutated BCP‐ALL cell line SEM ³⁷ (Supporting Information S1: Figures 1H), in TP53 ^R248Q mutated KOPN8, and TP53 ^R181C mutated REH cells (Supporting Information S1: Figure 1I,J). In TP53‐wildtype BCP‐ALL cell line RS4;11 we observe a strong single‐agent effect of orludodstat, resulting in an additive effect in the combination (Supporting Information S1: Figure 1K).

To validate our findings in patient‐derived xenografts (PDXs), we tested the combination of orludodstat and berzosertib in three unique diagnosis–relapse pairs. In each of these pairs, the relapse sample had acquired mutations/deletions in TP53, whereas the corresponding diagnosis sample was TP53‐wildtype (Table 1). While the combination induced cell death in all samples, synergy scores were higher in the inherently more resistant relapse samples relative to their diagnosis counterparts (Figure 1H–J and Supporting Information S1: Figure 1L). Based on these findings, we conclude that the orludodstat/berzosertib combination treatment is at least equally effective in TP53‐deficient ALL cells.

Dual inhibition of ATR and DHODH activates ATF4‐dependent gene expression signatures

To explore the mechanism responsible for the observed synergy between pyrimidine synthesis blockade and ATR inhibition, we performed bulk RNA sequencing comparing RCH‐ACV TP53 ^WT and TP53 ^KO cells after 24‐h treatment with berzosertib, orludodstat, or the combination thereof.

Subsequent pathway analysis on differentially expressed genes (DEGs) between the vehicle‐treated TP53 ^WT and TP53 ^KO samples mainly revealed genes involved in p53‐regulated pathways (Supporting Information S1: Figure 2A). Compared to vehicle‐treated control cells, treatment with berzosertib alone did not markedly alter gene expression, as only 23 DEGs in TP53 ^WT and 25 DEGs in TP53 ^KO were identified (Supporting Information S1: Figure 2B). In contrast, orludodstat monotherapy induced differential expression of 894 genes in TP53 ^WT and 1099 genes in TP53 ^KO (Supporting Information S1: Figure 2B). Pathway enrichment analyses revealed that many of the genes upregulated by orludodstat were associated with extracellular matrix organization, particularly collagen assembly, and Type I interferon (alpha/beta) signaling pathways (Supporting Information S1: Figure 2C).

To identify the main pathways responsible for the enhanced cytotoxicity of the combination treatment, we compared orludodstat monotherapy to the combination therapy (Figure 2A and Supporting Information S1: Figures 2D,E). We observed significant enrichment of pathways associated with the integrated stress response (ISR), a conserved adaptive mechanism that restores cellular homeostasis under various stress conditions. Key enriched pathways included oxidative stress signaling via EIF2AK1 (HRI), endoplasmic reticulum (ER) stress via PERK, and transcriptional activation of Activating Transcription Factor 4 (ATF4) target genes (Figure 2A). Supporting these findings, transcription factor analysis ³⁸ identified ATF4 as the primary activated transcription factor in both TP53 ^WT and TP53 ^KO cells following combination treatment (Figure 2B).

Figure 2.

Dual inhibition of Ataxia Telangiectasia and Rad3‐related (ATR) and dihydroorotate dehydrogenase (DHODH) activates ATF4‐dependent gene expression signatures in TP53 ^WT and TP53 ^KO B‐cell precursor acute lymphoblastic leukemia (BCP‐ALL). (A) Overlayed overrepresentation analysis of differentially expressed genes (DEGs) in RCH‐ACV TP53 ^WT and TP53 ^KO cells. DEGs between combination‐ and orludodstat‐treated RCH‐ACV TP53 ^WT (black) and TP53 ^KO (blue) were analyzed for pathway enrichment using the Reactome database. The top 11 significantly enriched pathways shared between the two genotypes are shown. Pathways are ranked by significance. Line lengths represent the gene ratio (number of DEGs in the pathway/total number of genes in that pathway), and dot sizes correspond to statistical significance (−log₁₀ adjusted P‐value). (B) Transcription factor activation analysis in RCH‐ACV TP53 ^WT and TP53 ^KO cells. Activation scores for predicted transcription factors are plotted with TP53 ^KO on the x‐axis and TP53 ^WT on the y‐axis. Transcription factors associated with ATF4 are highlighted in red and annotated. (C) Schematic overview of the integrated stress response (ISR). In response to various cellular stressors, eIF2α kinases—such as GCN2, HRI, and PERK—are activated and phosphorylate eIF2α. Phosphorylated eIF2α promotes the selective translation of ATF4, a transcription factor that regulates genes involved in adaptive stress responses or, under prolonged stress, the execution of cell death. (D) Heatmap of gene expression in the “Response of EIF2AK1 (HRI) to heme deficiency” Reactome pathway (R‐HSA‐9648895). Z scores representing expression levels of genes are shown for the designated RCH‐ACV cell lines. TP53 status is indicated by black (TP53 ^WT) and blue (TP53 ^KO). (E) mRNA expression of CHAC1, DDIT3, and TRIB3 in Nalm6 TP53 ^KO transduced with short‐hairpin RNAs (shRNAs) targeting ATF4 (shRNA1 and shRNA2) and a control shRNA (shScrambled). Bar graphs show the relative mRNA expression levels of CHAC1, DDIT3, and TRIB3 measured by real‐time quantitative PCR following treatment with dimethyl sulfoxide (DMSO), 100 nM berzosertib, 4 nM orludodstat, or the combination thereof. Expression was normalized to the TBP housekeeping gene and quantified using the 2^(−ΔΔCt) method. Data represent mean ± SD of n = 3 technical triplicates. Statistical significance was determined using Two‐way analysis of variance (ANOVA) with Dunnets multiple comparison. (F) viability (%) after 72 h incubation of the indicated concentrations of berzosertib and 4 nM of orludodstat in Nalm6 TP53 ^KO models transduced with shRNAs targeting ATF4 (shRNA1 and shRNA2) and a control shRNA (shScrambled). Data represent the mean ± SD of n = 3 technical triplicates.

ATF4 activation is a hallmark of the ISR, which is initiated through phosphorylation of eukaryotic translation initiation factor 2α (eIF2α) by one of four ISR kinases: EIF2AK1 (HRI), PERK, PKR, and GCN2 (Figure 2C, showing enriched pathways). This phosphorylation event suppresses global mRNA translation, while selectively enhancing translation of ATF4. Once activated, ATF4 regulates a broad transcriptional program that enables cells to manage stressors such as reactive oxygen species (ROS), protein misfolding, or amino acid deprivation. However, under prolonged stress conditions, ATF4 drives the expression of pro‐cell death genes, acting as a context‐dependent mediator of cell fate, much like p53 (Figure 2C). ³⁹ , ⁴⁰

Among the ATF4‐regulated pro‐death genes upregulated by combination therapy were DNA Damage‐Inducible Transcript 3 (DDIT3/CHOP) and Tribbles Pseudokinase 3 (TRIB3), both known to promote apoptosis. ⁴¹ , ⁴² In addition, CHAC1, which depletes intracellular glutathione and promotes oxidative stress, was also upregulated. Notably, expression of these genes was not significantly altered by orludodstat treatment, whereas berzosertib monotherapy only slightly increased CHAC1 expression. The combination treatment, however, robustly induced expression of all the aforementioned genes (Figure 2D and Supporting Information S1: Figure 2F,G).

To investigate the involvement of ATF4 in cell death, we used two different short‐hairpin RNAs (shRNA) to create ATF4 knockdowns (KD) in the Nalm6 TP53 ^KO model and quantified gene expression of ATF4 downstream targets after combination treatment with orludodstat and berzosertib. In the Nalm6 TP53 ^KO ATF4 ^KD models, we found a significant decrease in DDIT3 expression in both shATF4‐1 and shATF4‐2 knockdown models compared to the shScrambled control, as well as a decreasing trend in TRIB3 expression (Figure 2E). When looking at cell viability upon combination treatment, the Nalm6 TP53 ^KO ATF4 ^KD models showed a significant decrease in cell death compared to the shScrambled control (Figure 2F). Collectively, these results indicate that the orludodstat/berzosertib combination therapy exerts p53‐independent anti‐leukemic effects through activating the ISR, leading to ATF4‐driven expression of genes that promote oxidative stress and the induction of cell death.

Dual inhibition of DHODH and ATR rewires oxidative metabolism

DHODH inhibition has a profound effect on (mitochondrial) metabolism. ⁴¹ To explore the orludodstat/berzosertib‐induced metabolic responses, we performed [U‐¹³C₆]‐glucose stable isotope tracing in the BCP‐ALL TP53 ^WT and TP53 ^KO model RCH‐ACV. Analysis of total metabolite levels revealed that berzosertib treatment alone had a relatively modest impact on metabolic profiles (Supporting Information S1: Figure 3A), consistent with its limited effect on the transcriptome. As expected, orludodstat treatment led to significant changes in metabolites associated with pyrimidine biosynthesis, reducing levels of de novo pyrimidine synthesis products (e.g., UMP, UDP, UTP, CDP, and CTP) while accumulating upstream intermediates like N‐carbamoyl‐aspartate, due to the enzymatic blockade of DHODH (Figure 3A). Beyond pyrimidine biosynthesis, orludodstat with and without berzosertib reduced all detected TCA cycle intermediates, including acetyl‐CoA, citrate, α‐ketoglutarate, succinate, and malate (Figure 3B and Supporting Information S1: Figure 3A). To evaluate canonical TCA cycle activity more directly, we analyzed the incorporation of glucose‐derived ¹³C into TCA cycle intermediates. Specifically, we focused on the [M+2] isotopologues, which represent metabolites produced during the first turn of the TCA cycle (Figure 3C). As depicted in Figure 3D, treatment with orludodstat ± berzosertib markedly reduced the levels of [M+2]‐labeled TCA cycle intermediates (Supporting Information S1: Figure 3B), indicating a diminished flux of glucose into the TCA cycle suggestive of impaired TCA cycle activity. Consistently, we observed a decrease in the NADH/NAD⁺ ratio under these treatment conditions (Figure 3E), indicative of a disrupted redox balance. This effect was more pronounced in TP53 ^KO cells compared to their TP53 ^WT counterparts. An altered redox balance might result in mitochondrial dysfunction and increased production of ROS. We therefore tested whether ROS levels increased upon treatment. While monotherapy with either orludodstat or berzosertib led to a modest increase in intracellular ROS, combined treatment resulted in significantly elevated ROS levels compared to the DMSO control (Figure 3F).

Figure 3.

Dual inhibition of dihydroorotate dehydrogenase (DHODH) and Ataxia Telangiectasia and Rad3‐related (ATR) reduces oxidative metabolism and induces ATF4‐driven cell death. (A, B) Heatmap visualizing the log₂ fold‐change (log₂FC) in total metabolite levels (sum of unlabeled and labeled fractions) in berzosertib‐ (100 nM), orludodstat‐ (4 nM), and combination‐treated (combi, 100 nM berzosertib + 4 nM orludodstat) compared to dimethyl sulfoxide (DMSO) control in RCH‐ACV TP53 ^WT and TP53 ^KO models. Data are represented as the mean log₂FC of n = 3 technical replicates per condition. (C) Schematic overview of the incorporation of carbons from [U‐¹³C₆]‐glucose (yellow dots) into downstream metabolites, such as the glycolytic end‐product lactate ([M+3]) and TCA cycle intermediates like citrate ([M+2]). Colors of the dots depict: ¹²C (black), ¹³C (yellow), ¹³C derived from ¹³C‐labeled TCA intermediate oxaloacetate that enters the second turn (pink) of the TCA cycle. (D) Bar graph representing the fold‐change in [M + 2]‐labeled TCA cycle metabolites in DMSO‐, berzosertib‐ (100 nM), orludodstat‐ (4 nM), and combination‐treated (100 nM berzosertib + 4 nM orludodstat) RCH‐ACV TP53 ^WT and TP53 ^KO models compared to DMSO (treatment/DMSO). Data are represented as the mean fold‐change ± SD of n = 3 technical replicates per condition and are normalized to DMSO vehicle (DMSO = 1). Statistical analysis was performed by one‐way analysis of variance (ANOVA) followed by Dunnett's post hoc test, comparing each treatment group to the control (DMSO). (E) Bar graph depicting the ratio of fold‐changes in total NADH levels relative to fold‐changes in NAD⁺ levels in DMSO‐, berzosertib‐ (100 nM), orludodstat‐ (4 nM), and combination‐treated (100 nM berzosertib + 4 nM orludodstat) RCH‐ACV TP53 ^WT and TP53 ^KO models. Data are represented as the mean NADH/NAD⁺ ratio ± SD of n = 3 technical replicates per condition and are normalized to DMSO vehicle (DMSO = 1). Statistical analysis was performed by one‐way ANOVA followed by Dunnett's post hoc test, comparing each treatment group to the control (DMSO). (F) Bar graph representing fold‐change in ROS levels in DMSO‐, berzosertib‐ (100 nM), orludodstat‐ (4 nM), and combi‐treated (100 nM berzosertib + 4nM orludodstat). Data represent the mean fold‐change ± SD of n = 3 technical replicates per condition and are normalized to DMSO vehicle (DMSO = 1). (G) Schematic representation of ATF4‐mediated transcriptional activation of CHAC1, which encodes a glutathione (GSH)‐degrading enzyme. Under conditions of oxidative stress, reduced GSH is converted to glutathione disulfide (GSSG) by enzymes such as glutathione peroxidase 4 (GPX4), thereby mitigating lipid peroxidation and protecting against ferroptosis. Treatment with the lipid peroxidation inhibitor ferrostatin‐1 (Fer‐1) attenuates this cell death process. (H) Bar graph representing glutathione (GSH) concentrations in DMSO‐, berzosertib‐ (100 nM), orludodstat‐ (4 nM), and combination‐treated (100 nM berzosertib + 4 nM orludodstat). Data represent μM ± SD of n = 3 technical replicates per condition. (I, J) Viability (%) after 72 h of incubation, the indicated concentrations of berzosertib, 4 nM orludodstat, and 20 nM Ferrostatin‐1 in RCH‐ACV TP53 ^KO (I) and Nalm6 TP53 ^KO (J). Data represent the mean ± SD of n = 3 independent experiments. (K, L) Bars representing the population percentage containing peroxidated lipids in Nalm6 TP53 ^KO cells (K) and RCH‐ACV TP53 ^KO cells (L) based on BODIPY‐C11 581/591 staining. Data represent the mean ± SD of n = 3 independent experiments. Statistical analysis was performed by one‐way ANOVA followed by Dunnett's post hoc test, comparing each treatment group to the control (DMSO).

To combat ROS and maintain redox homeostasis, cells rely on antioxidant systems, particularly the glutathione (GSH) pathway. As previously mentioned, a well‐known transcriptional target of ATF4 is the CHAC1 gene, encoding a protein that catalyzes the cleavage of GSH ⁴² (Figure 3G). This enzymatic activity reduces the intracellular GSH pool, impairing a cell's capacity to mitigate oxidative stress. Consistent with the transcriptional activation of CHAC1 observed in the ATF4‐associated gene signature (Figure 2E) and the increased ROS levels (Figure 3F), treatment with the combination led to significantly diminished intracellular GSH levels (Figure 3H).

Elevated oxidative stress coupled with reduced availability of GSH can trigger ferroptosis, a form of iron‐dependent cell death characterized by the accumulation of peroxidized lipids. ⁴³ In specific contexts, ATF4 can promote ferroptosis by modulating the expression of genes, such as DDIT3 and CHAC1, resulting in failure of GSH‐dependent antioxidant defenses, via GPX4. This failure results in unchecked lipid peroxidation and eventually cell death ⁴⁴ (Figure 3G). To investigate the extent to which the combination of orludodstat and berzosertib induces BCP‐ALL cell death through the ferroptosis pathway, we combined this combination treatment with ferroptosis inhibitor Ferrostatin‐1 (Fer‐1) (Figure 3I,J), Ferroptosis activator RSL3 (Supporting Information S1: Figure 3D–I), and checked lipid peroxidation levels using BODIPY‐C11 (Figure 3K,L).

In Nalm6 TP53 ^KO cells, the addition of Fer‐1 led to a near‐complete rescue of cell viability (Figure 3I), whereas in the RCH‐ACV TP53 ^KO model, only partial rescue was observed following combination treatment with orludodstat and berzosertib (Figure 3J). The combination of ferroptosis activator RSL3 and berzosertib resulted in a synergistic effect in the Nalm6 TP53 ^KO and the RCH‐ACV TP53 ^KO cell lines. The addition of orludodstat did not enhance the synergy but increased the cell death in an additive manner (Supporting Information S1: Figure 3D,E). Additionally, combination treatment resulted in a significant increase in peroxidized lipids both in Nalm6 TP53 ^KO (Figure 3K) and the RCH‐ACV TP53 ^KO model (Figure 3L), although the amount of peroxidized lipids detected was significantly lower in Nalm6 (2%) than in the RCH‐ACV cells (15%). These results suggest that ferroptosis plays a major role in the therapeutic effects observed, although other cell death mechanisms may also contribute.

Taken together, these findings support a model in which combined orludodstat and berzosertib treatment activates the ATF4 pathway, driving ROS‐mediated and TP53‐independent cell death.

ATR pathway inhibition enhances orludodstat responses in vivo

Since the combination of berzosertib and orludodstat was effective in primary patient samples in vitro (Figure 1H–J and Supporting Information S1: Figure 1I), we tested the in vivo efficacy of this drug combination in immunocompromised NRGS mice. These mice were transplanted with a patient‐derived BCP‐ALL xenograft in which one TP53 allele was affected by a nonsense mutation and the other allele was deleted (TP53 ^P142Fs/del). Treatment with orludodstat, berzosertib, or a combination of the two drugs was started 28 days after the injection of BCP‐ALL TP53‐deficient cells and maintained for a total of 4 weeks (on a 5 days on, 2 days off schedule) (Figure 4A). This treatment regimen was well tolerated, as evidenced by a lack of weight loss during treatment (Supporting Information S1: Figure 4A). Leukemia burden was assessed by flow cytometric analysis of human lymphoblasts in the blood. Mice treated with berzosertib as a single agent did not benefit from treatment, whereas orludodstat was sufficient to delay leukemia development (Figure 4B–D), consistent with our earlier observations (Figure 1C). Importantly, the combination of orludodstat and berzosertib significantly suppressed leukemia growth and improved event‐free survival compared to both single treatments.

Figure 4.

Ataxia Telangiectasia and Rad3‐related (ATR) pathway inhibition enhances orludodstat responses in vivo. (A) Schematic representation of the in vivo experiment design. TP53 ^P142Fs/del patient‐derived B‐cell precursor acute lymphoblastic leukemia (BCP‐ALL) xenograft cells were injected into mice and treated for 4 weeks with either orludodstat, berzosertib, or the combination. Leukemia development was followed via weekly blood collection, using flow cytometric measurement of human CD19‐expressing (hCD19⁺) cells in the blood. Image created with BioRender.com. (B, C) Percentage of hCD19⁺ cells relative to the total lymphocyte count cells is visualized over the course of the experiment (B) and on Day 47 post‐injection (C). A two‐way analysis of variance (ANOVA) was performed to assess significant differences in the percentage of hCD19⁺/mCD45⁺ cells between treatment groups, followed by Tukey's post hoc test to correct for multiple comparisons. (D) Kaplan–Meier curve showing event‐free survival, with events defined as tumor burden exceeding 50% hCD19⁺/mCD45⁺ cells in peripheral blood. Group differences were assessed using pairwise log‐rank tests, and P‐values were adjusted for multiple comparisons using the Bonferroni correction. (E, F) Fold‐change of CD3⁺ cells (E) and CD19⁺ cells (F) compared to baseline (Day −7). Statistical analysis was performed by one‐way ANOVA followed by Dunnett's post hoc test, comparing each treatment group to baseline. Only the statistical significance of the combination treatment group is shown. Complete statistical significance of all treatment groups is visible in Supporting Information S1: Figure 4C,D.

To further investigate potential treatment‐induced toxicities, we treated BALB/c mice with orludodstat, berzosertib, or the combination of the two for 3 weeks. Again, we observed no decrease in mouse weights over the course of the treatment (Supporting Information S1: Figure 4B), and a macroscopic post‐mortem examination of the organs revealed no gross abnormalities. We measured the blood composition using an automated hematology analyzer and flow cytometry before, during, and after treatment. We observed a roughly 25% decrease in CD19⁺ B‐cells and a 50% decrease in CD3⁺ T cells with the combination treatment (Figure 4E,F, Supporting Information S1: Figure 4C,D). Platelet counts were stable in untreated controls and showed no systematic treatment‐related trend in any group (Supporting Information S1: Figure 4E). Although we measure statistically significant differences at isolated timepoints, unlike the effects measured on lymphocytes, we detect no progressive or sustained effect on platelet counts following treatment. We observed a decreasing trend of red blood cells during treatment (Supporting Information S1: Figure 4F). After the treatment stop, both lymphocyte and red blood cell counts recovered. Granulocyte and monocyte counts were not affected by the therapy (Supporting Information S1: Figure 4G,H). These data suggest that this 3‐week dosing schedule is well tolerated and only mildly affects the immune cell composition.

Together, our results provide important proof‐of‐concept for the potential of treating TP53‐deficient leukemias with a combination of DHODH and ATR inhibitors. More generally, our findings suggest that pharmaceutical strategies promoting ATF4‐mediated cell death represent an attractive therapy for leukemias without functional p53.

DISCUSSION

Loss of functional p53 is a major determinant of treatment failure and clonal evolution in relapsed ALL, yet direct pharmacologic restoration of p53 function remains challenging. TP53 mutations confer broad resistance to cytotoxic chemotherapy, targeted small molecules, and even immunotherapies, highlighting the need for therapies that circumvent p53 dependency. Here, we identify a previously unrecognized metabolic vulnerability of TP53‐deficient BCP‐ALL. We exploit this vulnerability by combining inhibition of de novo pyrimidine synthesis and ATR signaling. We show that the DHODHi orludodstat synergizes with the ATR inhibitor berzosertib to induce robust anti‐leukemic effects, irrespective of p53 status. Mechanistically, this synthetic lethality is mediated through activation of the ISR and subsequent ATF4‐driven cell death (Figure 5). These findings provide a mechanistic rationale for therapeutic exploitation of ATF4‐mediated stress responses as a p53‐independent vulnerability in ALL.

Figure 5.

Therapeutic mechanism of combination therapy with berzosertib and orludodstat. Schematic representation of the activation of ATF4 translation through phosphorylation of eIF2a. Orludodstat impacts mitochondrial respiration, resulting in increased reactive oxygen species (ROS) levels. Increased ROS levels contribute to ferroptosis and activation of eIF2A, a key kinase in the integrated stress response. Berzosertib inhibits Activating Transcription Factor 4 (ATR) and causes replication stress, also activating eIF2a. eIF2a subsequently promotes translation of ATF4, which regulates expression of its downstream targets leading to TP53‐independent cell death via different mechanisms, including ferroptosis.

DHODH inhibition alone exerts a modest anti‐leukemic activity in both TP53 ^WT and particularly in TP53 ^KO cells, and delays leukemia progression in vivo, consistent with prior reports. ¹⁹ , ²⁰ , ²¹ , ²² , ²³ This highlights the need for rational combination strategies to maximize therapeutic efficacy. We show that dual inhibition of DHODH and the DNA damage checkpoint kinase ATR enhances cytotoxicity in pediatric BCP‐ALL cells, irrespective of their TP53 status. Importantly, this combination therapy was well tolerated in immunocompetent mice, with only a transient reduction in circulating B and T cells, and a non‐significant decreasing trend in red blood cell counts. Granulocytes and monocyte counts were unaffected. The transient lymphopenia is consistent with the known sensitivity of proliferating lymphocytes to nucleotide depletion and replication stress, and aligns with the established immunomodulatory activity of DHODHis in autoimmune indications. ¹⁷ , ¹⁸ Together, the recovery of all cell populations following treatment stop, and the absence of effects on the myeloid compartment, compare favorably to the myelosuppression typically associated with conventional chemotherapy, supporting further preclinical and clinical development of this combination.

Several studies attribute the anti‐tumor effects of DHODHi to increased replication fork stalling and a subsequent proliferation arrest and apoptosis. ²⁴ , ³¹ As TP53‐deficient cells lack a functional G1 checkpoint, they may become more reliant on ATR–CHK1‐mediated S‐phase control. ⁴⁵ , ⁴⁶ , ⁴⁷ However, our data point towards a previously unrecognized vulnerability: the activation of the ISR and its central effector ATF4. Multiple converging stress signals contribute to ATF4 activation in this context. DHODH inhibition depletes pyrimidine nucleotide pools, causing replication fork stalling and replication stress, ²³ , ²⁵ , ⁴⁸ and activates ATR. ²⁵ Persistent DNA damage in turn has been shown to activate the PERK–eIF2α–ATF4 arm of the ISR. ⁴⁹ Moreover, combined targeting of DHODH and ATR impaired TCA cycle function and altered the NADH/NAD⁺ ratio. This alteration in NADH/NAD⁺ ratio disrupts redox balance and elevates ROS, which further stimulates PERK‐dependent ATF4 activation. ⁵⁰ , ⁵¹

Suppression of TCA cycle flux also reduces the availability of various non‐essential amino acids that are precursors for other amino acids, including glutamate. Depletion of amino acids, such as glutamate, ⁵² activates GCN2, another arm of the ISR that activates ATF4. ⁵³ , ⁵⁴ Another drug that activates the GCN2‐ATF4 axis is the amino acid‐depleting agent asparaginase. ⁵⁵ Recently, we showed that in ALL, the loss of TP53 has minimal impact on the response to amino acid‐depleting agent asparaginase. ²⁸ Therefore, we hypothesize that (combination) therapies leading to sustained activation of ATF4 could provide a much‐sought Achilles’ heel for TP53‐deficient ALL.

The observation that ATF4‐driven death can bypass p53 underscores the translational potential of this strategy for TP53‐deficient leukemias. While prior studies have described ATF4 as a survival factor during transient stress, our data support an alternative role as a context‐dependent executioner of cell death under sustained metabolic and oxidative stress. This alternative role of ATF4 aligns with literature linking ATF4 to pathways that function independently of p53, including ferroptosis, ⁴⁴ , ⁵⁶ necroptosis, ⁵⁷ and apoptosis. ³⁹ , ⁴⁰ It is interesting to note that, in addition to berzosertib, we identified IAP inhibitor AZD5582 as the only drug synergizing with orludodstat in both model cell lines. Similar to our observations, IAP inhibitors were also shown to activate p53‐independent cell‐death mechanisms. ⁵⁸ , ⁵⁹

In summary, our study uncovers ATF4‐mediated cell death as a therapeutically exploitable liability in TP53‐deficient ALL and establishes combined DHODH and ATR inhibition as a promising, p53‐independent treatment strategy. Given the dismal prognosis associated with TP53‐mutated ALL, our findings support the clinical development of this rational combination approach to improve outcomes for this high‐risk patient population.

AUTHOR CONTRIBUTIONS

Kari J. T. Grünewald: Writing—original draft; methodology; software; data curation; formal analysis; investigation; validation. Marjolein M. G. Kes: Methodology; data curation; investigation; validation; formal analysis; writing—original draft. Gawin R. Stoll: Investigation; methodology. Willem P. J. Cox: Methodology; writing—review and editing; investigation; validation; formal analysis. Dorette S. van Ingen Schenau: Investigation; methodology. Enzo M. Scutigliani: Methodology; investigation. Emmely M. Wagener: Investigation. Ziqin Tang: Investigation; methodology. Senna Visser: Investigation. Jolanda Kooiman: Investigation. Jessica Buijs‐Gladdines: Investigation. Maria Rodionova: Investigation. Esther A. Zaal: Methodology; data curation; software; investigation. Jeroen W. A. Jansen: Investigation. Jeroen F. Velzen: Formal analysis; investigation; methodology. Jarno Drost: Supervision; writing—review and editing. Laurens T. van der Meer: Conceptualization; methodology; supervision; writing—review and editing; funding acquisition. Celia R. Berkers: Methodology; supervision; writing—review and editing; funding acquisition. Frank N. van Leeuwen: Conceptualization; writing—review and editing; supervision; funding acquisition.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

In vivo studies: Animal experiments were approved by the Animal Experimental Committee of Radboud University (RU‐DEC‐2019‐0036) and the Animal Experimental Committee of the Princess Maxima Center (AVD39900202216167).

FUNDING

This work was supported by KWF Kankerbestrijding (#14659) and KiKa (#377).

Supporting information

Supporting Information.

Supporting Information.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in GEO at https://www.ncbi.nlm.nih.gov/geo/, reference number GSE308646.