Durable responses to ATR inhibition with ceralasertib in tumors with genomic defects and high inflammation

BACKGROUND Phase 1 study of ATRinhibition alone or with radiation therapy (PATRIOT) was a first-in-human phase I study of the oral ATR (ataxia telangiectasia and Rad3-related) inhibitor ceralasertib (AZD6738) in advanced solid tumors. METHODS The primary objective was safety. Secondary objectives included assessment of antitumor responses and pharmacokinetic (PK) and pharmacodynamic (PD) studies. Sixty-seven patients received 20–240 mg ceralasertib BD continuously or intermittently (14 of a 28-day cycle). RESULTS Intermittent dosing was better tolerated than continuous, which was associated with dose-limiting hematological toxicity. The recommended phase 2 dose of ceralasertib was 160 mg twice daily for 2 weeks in a 4-weekly cycle. Modulation of target and increased DNA damage were identified in tumor and surrogate PD. There were 5 (8%) confirmed partial responses (PRs) (40–240 mg BD), 34 (52%) stable disease (SD), including 1 unconfirmed PR, and 27 (41%) progressive disease. Durable responses were seen in tumors with loss of AT-rich interactive domain-containing protein 1A (ARID1A) and DNA damage–response defects. Treatment-modulated tumor and systemic immune markers and responding tumors were more immune inflamed than nonresponding. CONCLUSION Ceralasertib monotherapy was tolerated at 160 mg BD intermittently and associated with antitumor activity. TRIAL REGISTRATION Clinicaltrials.gov: NCT02223923, EudraCT: 2013-003994-84. FUNDING Cancer Research UK, AstraZeneca, UK Department of Health (National Institute for Health Research), Rosetrees Trust, Experimental Cancer Medicine Centre.

constant estimated by linear regression of logarithmically-transformed concentration-versustime data.A minimum of three data points were used in calculating lz as per AstraZeneca standard operating procedures.Data from multiple dosing were used to derive accumulation ratios based on Cmax (Race Cmax) and AUC(0-8) (Rac AUC0-8) defined as the ratio between Cmax or AUC(0-8) after multiple dose and Cmax and AUC(0-8) at day 1.All PK parameters for ceralasertib were derived using non-compartmental analysis (NCA) method in Phoenix WinNonLin v8.3 software or higher where the 'linear up/log down trapezoidal rule' for AUC was applied.All PK concentrations and parameters were listed and summarized as per AstraZeneca standard operating procedures by dose level and by treatment (single or multiple dose).

Sequencing
Formalin-fixed samples were assessed by a pathologist for tumor-rich areas, which were marked for microdissection at extraction, and tumor content estimated.Tumor and normal tissue (DNA extracted from buffy coats) were analyzed where possible (and for all non-archival biopsies) to remove germline variants.
Formalin-fixed paraffin embedded (FFPE) tissue slides, fresh frozen material, and frozen buffy coat samples were extracted for this study.FFPE slides were reviewed for tumor content by a pathologist and tumor rich areas were marked for macrodissection at extraction.DNA was extracted using from five 10-micron sections of FFPE tumor samples using QIAamp DNA FFPE tissue kit (56404 QIAGEN).FF samples were extracted using QIAamp DNA Mini kit (51304 QIAGEN).
Frozen buffy coat samples were extracted on Qiagen QIAsymphony SP instrument, using QIAsymphony DNA_Blood_400_V6_DSP protocol, and QIAsymphony DNA Midi Kit (931255, QIAGEN).DNA was quantified using the Qubit dsDNA High Sensitivity Assay Kit with the Qubit 3.0 Fluorometer (Invitrogen, Carlsbad, CA).NGS libraries were prepared from 25-400ng DNA using the KAPA HyperPlus Kit (Kapa Biosystems, Wilmington, MA, USA) and IDT UDI 8bp adapters (Integrated DNA Technologies, Coralville, USA), following the manufacturer's protocol, including dual-SPRI size selection of the libraries (250-450 bp).To optimize enrichment and reduce off-target capture, pooled, multiplexed, amplified pre-capture libraries (up to 20 samples per hybridization) were hybridized overnight using 1 µg of total DNA to a custom design of DNA baits complementary to the genomic regions of interest (NimbleGen SeqCap EZ library, Roche, Madison, WI, USA).
Hybridized DNA was PCR amplified and products purified using AMPure XP beads (Beckman Coulter, Danvers, MA, USA) and quantified using Qubit dsDNA High Sensitivity Assay Kit with the Qubit 3.0 Fluorometer (Invitrogen, Carlsbad, CA), and High Sensitivity D1000 TapeStation (Agilent, Santa Clara, USA).Samples were captured using a targeted capture panel (DDR panel) consisting of 173 genes, including multiple potential sensitizers to ATRi, such as DDR genes, and oncogenes (Supplementary Table 3).
Sequencing was performed on a NextSeq (Illumina, San Diego, CA, USA) with 75 bp pairedend reads and v2 chemistry, or NovaSeq6000 with 100 or 150bp paired-end reads and v1 chemistry, according to the manufacturer's instructions.
Sequencing runs were analyzed using an in-house pipeline.For the demultiplexing, Illumina bcl2fastq was used to assign reads for each sample based on the sequencing of 8-bp unique dual indexes.The reads were aligned to the reference genome build GRCh37/Hg19 using Burrows-Wheeler Aligner (BWA-MEM), followed by the marking of PCR duplicates and calculation of various quality control (QC) metrics using Picard.Genome Analysis ToolKit (GATK) was used for realigning around known indels to improve indel calling and base quality score recalibration for adjusting systematic errors made by the sequencer when estimating quality scores of each base call.HaplotypeCaller is used for variant calling in germline sample (limit of detection ~10%) and Mutect2 is used for tumor-normal paired somatic analysis (limit of detection ~5%).Mutect2 tumor only mode is used for tumor only somatic analysis (limit of detection ~5%).VCF files were then annotated using oncotator (for samples pre-2019) and Personal Cancer Genome Reporter (for post 2019 samples).The potential mutations identified by in-house pipeline were further checked manually on IGV.Copy number variant was estimated by generalizing the coverage expected for a copy of any given targeted region (i.e.an exon), taking the average coverage across all captured regions to estimate the average coverage of one targeted region.Any ratio below 0.5-fold was defined as a potential deletion, whereas a ratio above 2.4 was flagged as a potential amplification if 80% of the target regions had exceeded the thresholds.Manta and Pindel was used for the detection of structural variants including large indels, potential fusions and ITDs.
For whole exome sequencing, genomic DNA was extracted from Buffy Coat using the Qiagen DNeasy Blood & Tissue Kit, from Frozen Solid Tumors using the Qiagen All Prep DNA/RNA Micro Kit, and from FFPE samples using the Covaris truXTRAC FFPE DNA Kit.
All genomic DNA (200-1000ng) was fragmented to 200bp using a Covaris E Series and the resultant libraries were subjected to DNA Capture using SureSelect XT Human All Exon v5 or v6 kit (Agilent) following the manufacturer's instructions.
Final libraries were quantified using qPCR and clustered at a molarity of 14.5 pM; sequencing was performed on an Illumina HiSeq 2500 using 2x101 cycles of version 2 RAPID SBS chemistry.Tumor samples were sequenced at 100-150x depth and germline samples at 40-60x.
Tumor mutational burden was defined as total number of somatic mutations, including synonymous mutations, divided by the library panel size.Tumor mutational burden was defined as high (≥20 mutations/Mb), intermediate (10-20) or low (<10).
In part B, sequencing of archival tumor material, or external sequencing reports were used to select some patients for the study.Out of 21 archival/external sequencing, confirmatory sequencing using fresh biopsies was performed in 20 tumors, and was concordant in 10 (in 2, the gene of interest was not included on the sequencing panel; in 3, confirmatory sequencing found other mutations of interest, in 5, confirmatory sequencing found no mutations of interest).

Tumor RNAseq
RNA samples were quality-controlled and sequenced by the ICR Genomics Facility.RNA polyA method was used for mRNA selection.Strand-specific libraries were generated using the NEB ultra II directional kit.Illumina paired-end libraries were sequenced on a NovaSeq (Illumina) using Novaseq chemistry acquiring 100bp paired-end (PE) reads.Bcl2fastq software (v2.2.20, Illumina) was used for converting the raw base calls to fastqs and to de-multiplex further the sequencing data.The PE fastq files were used for further analysis.The STAR alignment software (v.2.7.6a) was used to align reads to the reference genome (GRCh38).Once the reads were aligned, HTSeq-count (HTSeq v0.12.4) was used to count the number of reads mapping unambiguously to genomic features in each sample.

IHC
Formalin fixed paraffin-embedded tumor samples were used.Nuclear phospho-(S635) Rad50 (Cell Signaling 14223), using a previously published method (52).H-score was calculated using HALO image analysis (Indica Labs).Tumor samples were also stained for γH2AX (S139, Cell Signaling 9718) positivity by IHC.A positive nucleus was defined as one with at least 5 nuclear foci or pan-nuclear staining, resulting in a total percent positive score.Nuclear ATM (Abcam 32420) IHC staining was carried out on the Ventana autostainer, using DAB detection and assessed by H-score, but samples were only deemed acceptable if lymphocyte staining (internal control) was at least 2+ (moderate) intensity.Cyclin E1 (Invitrogen HE12) staining was carried out on the Ventana autostainer, using DAB detection and the H-score was quantified using HALO image analysis (Indica Labs).Ki67 (Mib-1, Dako M7240) was scored by percent nuclear positivity using a global unweighted method, scoring 4 areas of 100 cells.ARID1A IHC was performed as previously published (36), using EPR13501 antibody (Abcam) and scored using H-score for nuclear positivity.Quantification of tumor-infiltrating lymphocytes was done using a semi-automated method in QuPath (53).Scanned H&E sections were first segmented into tumor and stroma, using a random trees machine-learning classifier, for each tumor section.Nuclei were segmented using Stardist (54), then stromal nuclei were sorted by size (area <20 µm 2 ) and circularity (>0.85), using small circular nuclei to approximate lymphocyte count.

Immunofluorescence
PBMCs were isolated from blood using BD Vacutainer® CPT™ Cell preparation tubes with sodium citrate (BD Biosciences, Wokingham, UK) and fixed with 4% formalin containing 1%

Flow cytometry
Peripheral blood was drawn into 8 mL EDTA tubes (Vacutainer, BD) and analyzed within 24 hours.ACK-lysed whole blood (for myeloid panel) or PBMC from density gradient centrifugation (for lymphocyte panel) were used.Samples were surface-stained with antibodies for 30 mins at 4 O C. Samples were analyzed on a LSR Fortessa (BD Biosciences).FACS analyses were performed in FlowJo v10.Participants who had available sequencing data for their tumors were included.If sequencing was not done, they were not included.Mutation of interest defined in Supp.Table 3.

Supp Fig 4:
tumor kinetics for participants at 1 site (Royal Marsden).Medical records were reviewed to obtain prior tumor measurements (negative days) for the target lesions.These were taken and normalized to the baseline measurements for this study (day 0 measurement).
Triton™ X-100 (Sigma-Aldrich, St. Louis, MO).PBMC cells were cytospun onto microscope slides and stained with anti-phospho-(S345) Chk1 [133D3] (Cell Signaling Technology 2348), or anti-gH2A.X (S139) (Abcam ab11174) antibodies and AlexaFluor 488 goat anti-rabbit IgG antibody (Invitrogen, Carlsbad, CA).Nuclei of PBMCs were counterstained with TOPRO-3 (Invitrogen).A Carl Zeiss LSM 700 confocal laser scanning fluorescence microscope (Zeiss, Jena, Germany) was used to visualize and capture images of the PBMCs.Fluorescent nuclear intensity for phospho-(S345) and total Chk1 in individual PBMC cells were quantified using the IN Cell Investigator Developer Toolbox v1.9 software (GE Healthcare Biosciences, Piscataway, NJ).For the phospho (S345) -and total Chk1 assays: the raw data values (fluorescent nuclear intensity for PBMCs), the percentage change for phospho (S345) biomarker was calculated for each subject by comparing the levels measured at each post-dose time point to pre-dose levels measured at baseline.The number of fluorescent gH2AX foci within individual PBMC cells were quantified using the IN Cell Investigator Developer Toolbox v1.9 software and the percentage of cells at each time point with greater than 5 foci calculated.The assays were GCP compliant and utilized healthy volunteer PBMCs as quality controls in every analytical run.
Change in PBMC phospho-(S345)Chk1 fluorescence intensity after 2-week treatment (D15), normalized to baseline (BL) per patient, by dose level.Color of line indicates RECIST response.accumulation ratio of ceralasertib, calculated by geometric mean Cmax or AUC0-8 at day 1 vs. day 0, or day 15 vs. day 0 for the indicated dose levels.

C
(left) B (right) Heatmap of top 100 differentially expressed genes, in baseline biopsies Heatmap of top 100 differentially expressed genes with the lowest adjusted p-value, clustering according to response for (A) baseline and (B) on-treatment tumor biopsies.The data are rlog transformed.Supp Fig 9 Geneset enrichment analysis using the 'Hallmarks' gene set.Pathway analysis of differentially expressed genes.Gene set enrichment analysis was carried out using the 'hallmarks' gene set.Normalized expression data for each condition was used from DEseq2.Conditions: OT vs. BL: all samples, on-treatment vs. baseline biopsy; PR vs. SD (BL biopsy): differences between patients having RECIST partial response (PR) or stable disease (SD) i.e., responders vs non-responders, comparing baseline biopsies; PR vs. SD (OT biopsy) difference between responders and non-responders, comparing on-treatment biopsies.Color indicates condition, size indicates false discovery rate q-value, 'NES': normalized enrichment score.type deconvolution data, by response A: Baseline samples, split by RECIST response B: On-treatment samples, split by RECIST response C: Fold-change in cell type proportion (on-treatment divided by baseline) split by RECIST response A B Supp Fig 11: cell-type signatures.A: T-cell functions signature, baseline biopsies B: NK cell function signature, baseline biopsies C: cytokine signature, on-treatment biopsies D: T-cell functions signature, on-treatment biopsies Interferon-stimulated gene signature.Interferon-stimulated genes, and categorisation by alpha or gamma interferon stimulation were obtained from Liu et al., 2011, PNAS 109(11): 4239."NE" indicates a patient who was not evaluable for response assessment."BL" baseline biopsy, "OT" on-treatment biopsy (generally day 15)