Biomarker-directed targeted therapy plus durvalumab in advanced non-small-cell lung cancer: a phase 2 umbrella trial

Patients with advanced NSCLC who lack targetable oncogenic drivers are typically treated with anti–PD-(L)1 immunotherapy, alone or with platinum-based chemotherapy, but many exhibit primary or acquired resistance. Primary resistance is defined as progression or stable disease lasting less than 24 weeks on immunotherapy, while acquired resistance is progression after at least 24 weeks of disease control or response. Tumor-intrinsic mechanisms (for example, STK11/LKB1 or KEAP1 alterations, impaired antigen presentation, defects in DNA damage response pathways) and immunosuppressive features in the tumor microenvironment contribute to immune checkpoint blockade failure. The HUDSON phase 2 umbrella study was designed to test rational combinations intended to overcome resistance to prior PD-(L)1 therapy in advanced/metastatic NSCLC and to explore biomarkers of sensitivity and resistance, with a particular hypothesis that targeting ATR in ATM-altered tumors could enhance antitumor immunity in combination with PD-L1 blockade.

The background highlights that although PD-L1 expression and tumor mutational burden can enrich for benefit from checkpoint inhibitors, resistance is frequent and multifactorial. Prior work links STK11/LKB1 and KEAP1 alterations, low PD-L1 expression, and defective antigen presentation to poor immunotherapy outcomes. DDR defects can increase neoantigen load and immunogenicity, with clinical success of PARP inhibition in DDR-defective tumors and preclinical evidence that ATR inhibition may be effective in ATM-deficient settings. Additional strategies to reverse immunosuppression include targeting STAT3 and the adenosine axis via CD73 inhibition. However, in NSCLC after ICB, the efficacy of combining durvalumab with PARP inhibition, STAT3 inhibition, or CD73 inhibition had not been established, motivating the evaluation within HUDSON.

Design: HUDSON (NCT03334617) is an ongoing, open-label, multicenter, nonrandomized, modular phase 2 umbrella trial enrolling patients with locally advanced/metastatic NSCLC who progressed on prior anti–PD-1/PD-L1 therapy and had received platinum-doublet chemotherapy. Patients were assigned to biomarker-matched (Group A) or biomarker-nonmatched (Group B) cohorts using centralized molecular profiling (FoundationOne CDx on tissue or Guardant360 ctDNA; PD-L1 by IHC). Group B was stratified by primary or acquired resistance to prior ICB. Cohorts and regimens: Four modules are reported: durvalumab plus (1) olaparib (PARP inhibitor), (2) danvatirsen (STAT3 antisense oligonucleotide), (3) ceralasertib (ATR inhibitor), and (5) oleclumab (anti-CD73). Biomarker-matched assignments included: ATM alteration/low ATM expression → durvalumab-ceralasertib (Module 3); HRR mutation or STK11/LKB1 alteration → durvalumab-olaparib (Module 1); high CD73 expression → durvalumab-oleclumab (Module 5). Nonmatched patients were allocated to primary or acquired resistance cohorts for each regimen. Planned initial cohort size was 20 evaluable patients with possible expansion to 40 based on ORR signals. Eligibility: Adults ≥18 years with histologically/cytologically confirmed progressive advanced or recurrent NSCLC; ECOG 0–1; measurable disease per RECIST v1.1; prior anti–PD-1/PD-L1 therapy and exposure to platinum-doublet chemotherapy; suitable for biopsy (archival or new). Key exclusions included known targetable EGFR/ALK (and known ROS1, BRAF, MET, RET) alterations, unresolved significant immune-related AEs from prior ICB, active autoimmune disease requiring systemic therapy, uncontrolled illness, symptomatic brain metastases, active infections, and inadequate organ function. Treatments and dosing: Durvalumab 1,500 mg IV Q4W for all. Olaparib 300 mg PO BID (Module 1). Danvatirsen 200 mg IV every other day for 1-week lead-in, then weekly with durvalumab (Module 2). Ceralasertib 240 mg PO BID Days 1–7 as a lead-in, then Days 22–28 each 4-week cycle with durvalumab (Module 3). Oleclumab 3,000 mg IV Q2W in cycles 1–2, then Q4W from cycle 3 (Module 5). Assessments: Tumor imaging by contrast-enhanced CT (or MRI if needed) at baseline, every 6 weeks for 24 weeks, then every 8 weeks until progression confirmed per RECIST v1.1. Patients could continue therapy beyond radiographic progression if clinically benefiting. Primary endpoint: objective response rate (ORR) per RECIST v1.1. Secondary endpoints: disease control rates at 12/24 weeks, progression-free survival (PFS), overall survival (OS), and safety/tolerability (AEs per NCI CTCAE v4.03). Exploratory endpoints included multi-omic biomarker analyses (bulk RNA-seq TIS signature, NanoString immune profiling in blood, TCRβ repertoire sequencing dynamics) and outcomes across molecularly defined subgroups. Biomarker analyses: FoundationOne CDx profiled 324 genes; TMB categorized as low (<10 mut/Mb) vs high (≥10 mut/Mb). PD-L1 assessed locally; positivity defined as ≥1% tumor cells. Bulk RNA-seq was performed on prescreening tumor biopsies to compute the 18-gene Tumor Inflammation Signature (TIS) via ssGSEA. Longitudinal blood RNA (NanoString) and TCRβ sequencing (Adaptive immunoSEQ) were collected at baseline, after ceralasertib lead-in, and after durvalumab to assess immune modulation (e.g., IFN signatures, T cell clonality, clone expansion). Statistical analyses included Kaplan-Meier for PFS/OS, with 80% CIs reported; Wilcoxon tests for paired/unpaired comparisons in exploratory analyses; multiplicity controlled with Benjamini–Hochberg where applicable. Population: Between 26 Jan 2018 and 26 Apr 2022, 941 patients were screened; 324 for these modules; 268 treated: 79 durvalumab-ceralasertib; 87 durvalumab-olaparib; 45 durvalumab-danvatirsen; 57 durvalumab-oleclumab. Of treated, 88 were biomarker-matched; 77 and 103 were biomarker-nonmatched primary and acquired resistance cohorts, respectively.

- Primary endpoint ORR: Durvalumab-ceralasertib 13.9% (11/79) vs pooled other regimens 2.6% (5/189). All responses were confirmed partial responses. - Disease control: At 12 weeks 50.6% vs 32.3%; at 24 weeks 35.4% vs 15.9% (durvalumab-ceralasertib vs pooled others). - Time-to-event: Median PFS 5.8 months (80% CI 4.6–7.4) vs 2.7 months (1.8–2.8); Median OS 17.4 months (14.1–20.3) vs 9.4 months (7.5–10.6). - ATM-altered biomarker-matched (durvalumab-ceralasertib): ORR 26.1% (6/23); median PFS 8.4 months; median OS 22.8 months. In nonmatched cohorts with durvalumab-ceralasertib, ORR was 13.0% (primary) and 6.1% (acquired); PFS 4.6 and 4.6 months; OS 12.0 and 19.1 months, respectively. - Other combinations: Durvalumab-olaparib ORR 4.6% overall; 9.5% in HRRm-matched and 4.8% in STK11/LKB1-matched cohorts; 0% (primary) and 4.3% (acquired) in nonmatched cohorts. Durvalumab-danvatirsen: no objective responses observed. Durvalumab-oleclumab: 1 partial response (1.8%). - Subgroups (durvalumab-ceralasertib vs pooled others): • Primary resistance: PFS 6.0 vs 1.8 months; OS 12.6 vs 6.3 months. • Acquired resistance: PFS 5.8 vs 2.8 months; OS 19.1 vs 12.6 months. • PD-L1-negative: PFS 7.5 vs 2.1 months; OS 20.3 vs 10.7 months. • PD-L1-positive: PFS 4.6 vs 3.0 months; OS 14.2 vs 9.7 months. • Squamous histology: PFS 4.8 vs 2.6 months; OS 15.9 vs 6.0 months. • Nonsquamous histology: PFS 6.0 vs 2.7 months; OS 18.8 vs 10.8 months. • Consistent benefit also in patients with liver/bone metastases, low TMB, and with STK11 or KRAS variants. - Safety: Similar overall TEAE incidence between durvalumab-ceralasertib and other regimens: any TEAE 93.7% vs 89.9%; grade ≥3 TEAE 44.3% vs 51.3%; any TRAE 75.9% vs 71.4%; grade ≥3 TRAE 20.3% vs 30.2%; serious AEs 36.7% vs 34.4%. TRAE-related discontinuations 7.6% vs 10.1%. TEAE deaths 2.5% vs 2.6%, none considered treatment-related. With durvalumab-ceralasertib, grade ≥3 thrombocytopenia occurred in 5.1%; with durvalumab-olaparib, grade ≥3 anemia in 13.8%; with durvalumab-danvatirsen, grade ≥3 ALT increased 11.1% and thrombocytopenia 8.9%. - Peripheral immunomodulation (durvalumab-ceralasertib): After 7 days of ceralasertib lead-in, observed decreases in monocyte- and exhausted CD8 T-cell–associated signatures and increases in TNF-α, IFN-γ, and IFN-α response signatures. TCR repertoire showed reduced clonality after ceralasertib, followed by increased clonality, expansion of abundant T cell clones, and emergence of newly expanded clones after durvalumab addition. These dynamic changes were not observed with the olaparib or danvatirsen combinations.

The study addresses the critical need for effective therapies after failure of platinum chemotherapy and PD-(L)1 inhibitors in advanced NSCLC. Among four rationally selected combinations intended to overcome immunotherapy resistance mechanisms, durvalumab-ceralasertib demonstrated notably higher response rates and substantially longer PFS and OS compared with other combinations, with consistent benefit across subgroups typically recalcitrant to immunotherapy. The enhanced activity in ATM-altered tumors supports the biological rationale that ATR inhibition exploits ATR dependency in ATM-deficient contexts, potentially synergizing with PD-L1 blockade. Exploratory translational analyses indicate that ceralasertib induces systemic immunomodulation—dampening signatures of T cell exhaustion and augmenting interferon pathway activation—followed by increased T cell clonality and clonal expansion after durvalumab, consistent with reinvigoration of antitumor immunity. In contrast, combinations with olaparib, danvatirsen, or oleclumab did not produce compelling efficacy in this post-ICB setting, suggesting limited therapeutic leverage of those pathways under the tested conditions. Collectively, the findings support ATR inhibition plus PD-L1 blockade as a strategy to overcome immune resistance in NSCLC and motivate further confirmatory evaluation.

Durvalumab plus ceralasertib demonstrated a meaningful efficacy signal in advanced/metastatic NSCLC following prior PD-1/PD-L1 therapy and platinum chemotherapy, with favorable activity particularly in ATM-altered tumors and consistent benefits across multiple immunotherapy-refractory subgroups. The regimen showed a manageable safety profile and peripheral immune changes consistent with an enhanced antitumor response in combination with PD-L1 blockade. These results have prompted initiation of the phase 3 LAT-IFY trial (NCT05450692) comparing durvalumab-ceralasertib versus docetaxel in this population. HUDSON continues to accrue to durvalumab-ceralasertib (including monotherapy ceralasertib cohorts) and to additional durvalumab-based combinations; future work will further elucidate mechanistic biomarkers and optimize patient selection.

Key limitations include the open-label, nonrandomized platform design without a durvalumab monotherapy control, limiting formal comparative inferences about the contribution of each partner agent. Cohort sizes were modest, and accrual occurred over different timeframes, with limited follow-up for some patients in the durvalumab-ceralasertib module at data cutoff. Potential confounding from baseline imbalances across cohorts and use of different NGS and IHC assays (and archival versus fresh biopsies) may affect biomarker allocations and comparisons. The study was not powered for definitive subgroup analyses; exploratory findings should be interpreted cautiously.