Early Changes in Circulating Cell-Free KRAS G12C Predict Response to Adagrasib in KRAS Mutant Non–Small Cell Lung Cancer Patients

Plasma genotyping is an emerging approach for the detection of circulating tumor DNA (ctDNA) within the broader pool of cell-free DNA (cfDNA) in patients with cancer (1–4). Initially, ctDNA was investigated as a biomarker because of its non-invasive nature to overcome the imperfect yield of solid organ biopsy. For example, in patients with non–small cell lung cancer (NSCLC), the presence of an EGFR driver mutation (Exon19 deletion or L858R, among others) or resistance mechanisms (e.g., EGFR T790M) predicts variable sensitivity to EGFR tyrosine kinase inhibitors (TKI; refs. 5–7). The EGFR paradigm has proven so successful that molecular testing by liquid biopsies has grown increasingly important and available in NSCLC care; now guiding the use of multiple oral-targeted therapies that are FDA approved (8–10).

Assessing for response and progression via serial CT scans is standard of care in patients with metastatic cancer undergoing systemic therapy. In advanced NSCLC, restaging scans of the chest, abdomen, pelvis, and the brain, if known metastatic lesions symptoms concerning for central nervous system involvement are present, are routinely performed every 6–8 weeks; this matches RECIST1.1 criteria, used for the assessment of treatment response and progression in clinical trials (11, 12). Although early applications of plasma genotyping assessed for the presence or absence of a clinically actionable mutation at a clinically relevant time point, recent data suggest that serial analyses may have clinical value as clearance of ctDNA among patients undergoing systemic chemo- or immunotherapy correlates with progression-free survival (PFS) and overall survival (OS; refs. 13–17). Collaborative efforts are ongoing to investigate and validate early ctDNA kinetics as a biomarker of response (https://www.focr.org/news/friends-cancer-research-launches-ctmonitr).

Adagrasib is a selective and irreversible oral small-molecule inhibitor of the mutant KRAS G12C protein (18) and was recently approved by the FDA for treatment of NSCLC who have received at least 1 prior systemic therapy (19). Here, we report on early ctDNA changes of KRAS G12C ctDNA in a Phase 2 trial of adagrasib in patients with advanced KRAS G12C-mutant NSCLC.

We analyzed that 60 patients enrolled in Cohort A of KRYSTAL-1 (NCT03785249). KRYSTAL-1 is a phase 1/2 single-arm clinical trial of registrational intent to evaluate adagrasib for advanced solid tumors with KRAS G12C mutations. Cohort A studies adagrasib given at 600-mg orally BID in patients with histopathology-confirmed NSCLC and measurable radiographic disease amenable to RECIST1.1 that were previously treated with a platinum-based chemotherapy (20). All patients were required to have a KRAS G12C driver mutation in tumor tissue. Adagrasib is administered over continuous 21 days cycles and CT imaging to assess clinical response was conducted every two cycles. Data cutoff value was 10/15/2021. All patients provided written informed consent to the institutional review board–approved protocol that allowed for ctDNA analysis of collected plasma. The study was conducted in accordance with the Declaration of Helsinki.

Plasma for ctDNA analysis was collected before adagrasib treatment on C1D1, C2D1, and C4D1. Blood (20 mL) was collected in Streck's cfDNA Blood Collection vacutainers, centrifuged for 10 minutes at 1,200 × g, followed by a second centrifugation of the plasma supernatant at 3,000 × g for 10 minutes. Cleared plasma was stored at −80°C until analysis. cfDNA was extracted using the DSP Circulating DNA Kit on a QIASymphonhy. Quantitative plasma assessment of KRAS G12C for C1D1 specimens was performed by droplet digital PCR (ddPCR; ref. 21) and by targeted hybrid capture next-generation sequencing using the ctDX First assay (Agilent) as previously described (22). The Agilent Resolution ctDx FIRST Assay is a qualitative, targeted hybrid capture, NGS-based in vitro diagnostic test that detects sequence variations in 112 genes using circulating cfDNA isolated from plasma of peripheral blood. Pre-treatment specimens were first assayed by NGS (ctDX First) at Resolution Biosciences, now Agilent, and if 10 ng of DNA was left remaining by ddPCR at the Belfer Center for Applied Cancer Science at the Dana Farber Cancer Institute. C2D1 and C4D1 specimens were assayed for KRAS G12C by ddPCR at the Belfer Center for Applied Science only. All laboratory-personal were blinded to sample identity. Results are presented as allelic fraction (AF%) of KRAS G12C to wild-type KRAS cfDNA.

Objective response rate (ORR), maximum change in target lesion size, and PFS were centrally assessed by RECIST 1.1. For ORR, the Fisher's exact tests (or χ² test) are conducted to compare the biomarker subgroups (complete clearance vs. incomplete clearance). For PFS and OS, median and 95% confidence interval within each biomarker group are estimated on the basis of the Kaplan–Meier method, and the comparison between the biomarker subgroups is based on the log-rank test.

The data generated for this study are available upon request. Requests should be directed to Cloud Paweletz at [email protected].

To study ctDNA plasma response of adagrasib in KRAS G12C patients with NSCLC, we performed plasma-targeted hybrid capture NGS and serial ddPCR on 60 patients enrolled in Cohort A of KRYSTAL-1. Fifty-three patients had their C1D1 pre-treatment baseline draw analyzed by plasma NGS, of these 24 were also analyzed by ddPCR, and 7 patients had missing C1D1 specimens. To confirm the quantitative concordance between NGS and ddPCR, we compared the KRAS G12C AF% results for 24 C1D1 specimens that had both ddPCR and NGS results. Concordance for ddPCR and NGS results was excellent across all AF% levels with an r² > 0.95 (Supplementary Fig. S1). Thus, NGS and ddPCR results on C1D1 were used interchangeably for analysis of baseline KRAS^G12C AF% concordance with clinical outcome measures and for analysis on the longitudinal effects of adagrasib on KRAS^G12C AF%. All available cycles 2 and 4 specimens were analyzed by ddPCR.

Of 60 patients with at least two evaluable plasma samples, 7 patients were excluded because of missing C1D1 draws, 13 patients were excluded for lack of detectable KRAS G12C levels at C1D1 and 1 patient was excluded because of a missing C2D1 draw resulting in 39 evaluable patients for early plasma response assessment between pre-treatment baseline and cycle 2 (Fig. 1). The baseline clinical characteristics and outcomes with adagrasib treatment were similar for the 39 patients to the overall patient population in the study (Supplementary Table S1). For the pre-treatment baseline to cycle 4 analysis; 4 additional patients were excluded because of missing C4D1 draws resulting in 36 patients evaluable for subsequent analysis (Fig. 1). Sensitivity for KRAS G12C detection at baseline for Cohort A patients with a confirmed KRAS G12C mutation by tumor tissue test was 75.5% (40/53), which is in the range of ctDNA oncogene driver detection in similar patient population cohorts (Fig. 2A; ref. 22). However, the correlation between disease burden (sum of diameters) to KRAS G12C AF% at baseline was weak (Supplementary Fig. S2). Evaluation of the correlation of baseline KRAS^G12C ctDNA levels and clinical response using RECIST 1.1 indicated that KRAS^G12C baseline ctDNA levels were not a predictive marker for clinical response (Fig. 2B).

We then analyzed the change in ctDNA KRAS G12C at 2 specific intervals, between pre-treatment baseline and cycle 2 and between baseline and cycle 4. We found that the maximal response in KRAS G12C ctDNA levels in most complete sample sets was observed between baseline and cycle 2 with 84.6% (33/39) of patients achieving complete KRAS^G12C plasma clearance by C2D1. By cycle 2, 34 patients had a >95% clearance of their plasma KRAS^G12C and a decrease of any magnitude was observed in 97.4% (38/39) of evaluable patients (Fig. 2C and D). Five patients initially showed a decrease in their plasma KRAS G12C at cycle 2 but a subsequent increase in ctDNA levels at cycle 4. In contrast, two patients that exhibited incomplete clearance of KRAS^G12C ctDNA at cycle 2 continued to deepen to achieve complete or near complete (>95%) clearance at cycle 4. All remaining patients evaluable at cycle 4 continued to have non-detectable ctDNA levels in their plasma and 80.6% (29/36) achieved a complete plasma response at cycle 4. Five patients never achieved complete clearance of plasma ctDNA at either cycle 2 or 4. Of note, patients with incomplete plasma clearance had higher AF% at baseline (Supplementary Fig. S3) and among those 4 patients who discontinued treatment before cycle 4, 3 had dose reductions or interruptions that may affect ctDNA clearance.

We next explored the clinical outcomes in patients that either cleared KRAS G12C in their plasma compared with those that did not clear KRAS G12C at either cycle 2 or 4 to investigate whether early ctDNA clearance can predict for clinical benefit of adagrasib. Patients with complete ctDNA clearance at cycle 2 showed improved ORR [cleared: 60.6% (20/33) vs. non-cleared: 33.3% (2/6)] and survival (cleared: 14.1 months vs. non-cleared: 8.7 months; P = 0.04; HR, 0.3). Patients with complete ctDNA clearance at cycle 2 also exhibit a nonsignificant trend toward shorter mPFS compared with patients with incomplete ctDNA clearance, likely a result of the small sample size in the “not-cleared” cohort. Complete ctDNA clearance at cycle 4 was associated with both improved survival (14.7 vs. 5.4 months; P <0.001; HR, 0.1) and mPFS (9.8 vs. 4.3 months; P value of 0.01; HR, 0.3; Fig. 3).

We then compared early plasma response with radiographic response in 39 patients for which baseline and cycle 2 results were available (Fig. 4). The median tumor diameter reduction of 33 patients that cleared their ctDNA completely at cycle 2 was 46% (range, −100% to 0%), whereas the median reduction for the 6 patients with incomplete ctDNA clearance at cycle 2 was 19.5% (range, 8% to 53%). For the 5 patients that initially showed a decrease in ctDNA levels but a subsequent rise at cycle 4, only one patient had a radiographic increase from baseline (−11% to 13%), whereas the other patients either were either clinically stable (−26% to −26%) or showed continued benefit of treatment 43.5% median decrease (range, 27 to 60%). These results are in line with our previous reports that best plasma response is not necessarily an indicator of best radiographic response (23). Of 7 patients that did not exhibit a complete plasma response at cycle 4, two patients had an increase in their tumor measurements from baseline (range, 13% to 16%;, median 14.5%) whereas 2 patients reached their radiographic nadir at week 12, one patient reached their radiographic nadir at week 6, one patient reached their radiographic nadir at week 21, and one patient had non-measurable disease at baseline (Supplementary Fig. S4).

Though the use of ctDNA is most often thought of as an alternative to tumor genotyping in patients with acquired resistance to targeted therapy, the field is increasingly focused on new applications such as detection of minimal residual disease, early detection, and monitoring of response and resistance (24, 25). In this prospective study of KRAS G12C-mutant patients with lung cancer participating in cohort A of KRYSTAL 1, complete clearance of ctDNA at cycle 2 was associated with improved ORR and OS whereas complete ctDNA clearance at cycle 4 was associated with improved PFS and OS. We previously accessed plasma samples collected at baseline and C2D1 for 64 patients with colorectal cancer treated on adagrasib with and without cetuximab. 55.2% (16/29) of evaluable patients in the monotherapy and 87.5% (14/16) in the cetuximab cohort achieved a greater than 95% reduction in KRASG12C-mutant allele fractions (26). Of interest, the response rate for the adagrasib monotherapy cohort was 18.6% and 46.6% for the adagrasib/cetuximab combination cohort. Together these studies support that early plasma response has potential as a marker of treatment outcome.

Several studies have shown that ctDNA is directly related to biologic factors like extent of metastatic spread and the aggressiveness of the cancer and that quantitative nature of ctDNA analysis has untapped potential as a monitoring tool for evaluating the effectiveness of a therapy and the aggressiveness of a cancer. The DYNAMIC study (27), for instance, assessed ctDNA dynamics of patients with lung cancer after surgery and showed that ctDNA quickly decreases after surgery and that detection of ctDNA may have a role in the postoperative lung cancer monitoring setting. Conclusions that were also supported by Chaudhuri and colleagues (28) in where the authors showed that serial ctDNA analyses could be useful for posttreatment surveillance after surgery or radiotherapy in stage I–III localized lung cancer.

Our data presented herein support previous studies that ctDNA response has the potential to be a marker of treatment outcome in patients with advanced NSCLC, and as such can be useful in certain clinical and drug development settings. In a phase I trial of ASP8273, a third-generation EGFR TKI, we detected a 10-fold decrease of mutant EGFR in plasma after just 1 cycle of therapy in EGFR-mutant patients, an effect seen starting at 100-mg daily and seen throughout dose escalation until 500-mg daily. Although 300-mg daily was selected as the phase 2 dose, this plasma response increased confidence in the activity of lower doses too (29). Similar findings were reported in serial plasma NGS in phase I trials of LOXO-292, a novel RET TKI, and BLU945, a fourth-generation EGFR TKI (30–32). Consistent with our results, several studies have already shown that clearance of ctDNA after treatment is prognostic of clinical benefit in scenarios when the recommended phase 2 dose has been established. Yu and colleagues (29) demonstrated that detection of mutant EGFR in plasma after 6 weeks in patients treated with osimertinib and bevacizumab in EGFR-mutant patients with NSCLC is associated with shorter PFS compared with patients with undetectable mutant EGFR. In addition, we have shown that ctDNA clearance in plasma after 6 weeks was associated with an improved ORR (74% vs. 41%) and PFS (HR, 2.6) in EGFR-mutant patients with NSCLC treated with osimertinib (33). Results that are also encouragingly observed in patients treated with anti-PD1 or anti-PDL1 regimens (15, 16, 24). In a retrospective study of 86 patients with NSCLC that were treated with anti–PD-1 antibodies, Guibert and colleagues (34) have shown that changes in ctDNA predict response to immune checkpoint inhibitors. And more recently, our group demonstrated that plasma response, defined as ≥50% decrease in max AF in ctDNA by cycle 3 using a commercial 36 gene NGS assay, is predictive of outcome in patients who received first-line pembrolizumab-based therapies. However, these results are still based on retrospective study designs and independent validation cohorts are still needed (15).

Our study also urges caution. This subset of patients that were evaluable for longitudinal KRAS^G12C ctDNA response in Cohort A exhibited a slightly higher ORR (56.4%, 22/39 cycle ctDNA response evaluable) compared with Cohort A overall (50.5%, 48/95 RECIST 1.1 evaluable patients; ref. 19; Supplementary Table S1) and the number of patients that had incomplete ctDNA clearance at cycle 2 or 4 was small. Of the 5 patients that showed a decrease in KRAS^G12C ctDNA by cycle 2 and an increase by cycle 4 only 1 discontinued treatment due to disease progression. The other 4 patients discontinued treatment due to either deterioration of health (n = 2), adverse events (n = 1) or death (n = 1). This is in line with reported lead times of plasma progression versus radiographic progression, which showed that the median lead time from ctDNA increase to radiographic progression was just under one month for EGFR TKIs, but greater than 25% of patients had a lead time of over 3 months (35). However, our study was limited to analysis of baseline, cycles 2 and 4 specimens only. Although some end of treatment specimens were collected on study and analyzed as part of Awad and colleagues (23), our study was not designed to longitudinally monitor until clinical progression or to characterize acquired resistance mutations. Thus, further research is needed whether initiating treatment changes at plasma progression leads to improved outcomes. Several studies are now underway to investigate the use ctDNA to either intensify or de-intensify treatment regimens in NSCLC and other cancers (NCT04166487; ref. 36).

Several protein biomarkers have been validated for use in patients undergoing systemic therapy for malignancy, including for patients with advanced prostate cancer (PSA), colorectal cancer (CEA), medullary thyroid cancer (CEA), and ovarian/fallopian tube cancer (CA-125). ctDNA may offer clinician's additional insight into therapeutic response with the potential for the dual benefit of providing data for regarding acquired resistance mechanisms (23).

C.P. Paweletz reports grants from National Institutes of Health and other support from Mirati Therapeutics during the conduct of the study. T. Kheoh reports other support from Mirati Therapeutics during the conduct of the study; as well as other support from Mirati Therapeutics outside the submitted work. R.C. Chao reports other support from Mirati Therapeutics outside the submitted work; and reports employment with Mirati Therapeutics. A.I. Spira reports Leadership at NEXT Oncology Virginia, Stock and Other; ownership interests at Eli Lilly; Honoraria from CytomX Therapeutics, AstraZeneca/MedImmune, Merck, Takeda, Amgen, Janssen Oncology, Novartis, Bristol-Myers Squibb, Bayer, Incyte, Mirati Therapeutics, Gritstone Oncology, Jazz Pharmaceuticals, Janssen Research & Development, Mersana, Gritstone Bio, Daiichi Sankyo/Astra Zeneca, Regeneron, Lilly, Black Diamond Therapeutics, Array BioPharma, and Blueprint Medicines; and research funding from LAM Therapeutics, Regeneron, Roche, AstraZeneca, Boehringer Ingelheim, Astellas Pharma, MedImmune, Novartis, Newlink Genetics, Incyte, Abbvie, Ignyta, Trovagene, Takeda, Macrogenics, CytomX Therapeutics, Astex Pharmaceuticals, Bristol-Myers Squibb, Loxo, Arch Therapeutics, Gritstone, Plexxikon, Amgen, Daiichi Sankyo, ADCT, Janssen Oncology, Mirati Therapeutics, Rubius, Synthekine, Mersana, Blueprint Medicines, Alkermes, and Revolution Medicines. K. Leventakos reports other support from Boehringer Ingelheim Pharmaceuticals, Amgen, AstraZeneca, Targeted Oncology, Takeda, Jazz Pharmaceuticals, Mirati Therapeutics, Janssen, Regeneron, OncLive State of the Summit, and MJH Life Sciences, as well as grants from AstraZeneca and Mirati Therapeutics outside the submitted work. M.L. Johnson reports grants and other support from Abbvie, Amgen, Arcus Biosciences, AstraZeneca, Black Diamond, Calithera Biosciences, Daiichi Sankyo, Genentech/Roche, Genmab, Genocea Biosciences, GlaxoSmithKline, Gritstone Oncology, IDEAYA Biosciences, Immunocore, Janssen, Merck, Mirati Therapeutics, Novartis, Regeneron Pharmaceuticals, Revolution Medicines, Sanofi, Takeda Pharmaceuticals, and Turning Point Therapeutics, and grants from Acerta, Adaptimmune, Apexigen, Array BioPharma, Artios Pharma, Atreca, BeiGene, BerGenBio, BioAtla, Boehringer Ingelheim, Bristol-Myers Squibb, Carisma Therapeutics, Checkpoint Therapeutics, City of Hope National Medical Center, Corvus Pharmaceuticals, Curis, CytomX, Dracen Pharmaceuticals, Dynavax, Lilly, Elicio Therapeutics, EMD Serono, EQRx, Erasca, Exelixis, Fate Therapeutics, Guardant Health, Harpoon, Helsinn Healthcare SA, Hengrui Therapeutics, Hutchinson MediPharma, IGM Biosciences, Immunitas Therapeutics, Incyte, Jounce Therapeutics, Kadmon Pharmaceuticals, Kartos Therapeutics, Loxo Oncology, Lycera, Memorial Sloan-Kettering, Merus, Mythic Therapeutics, NeoImmune Tech, Neovia Oncology, Numab Therapeutics, Nuvalent, OncoMed Pharmaceuticals, Palleon Pharmaceuticals, Pfizer, PMV Pharmaceuticals, Rain Therapeutics, RasCal Therapeutics, Relay Therapeutics, Ribon Therapeutics, Rubius Therapeutics, Seven and Eight Biopharmaceuticals/Birdie Biopharmaceuticals, Shattuck Labs, Silicon Therapeutics, Stem CentRx, Syndax Pharmaceuticals, Tarveda, TCR2 Therapeutics, Tempest Therapeutics, Tizona Therapeutics, TMUNITY Therapeutics, University of Michigan, Vyriad, WindMIL Therapeutics, and Y-mAbs Therapeutics, and other support from Arrivent, Astellas, Axelia Oncology, EcoR1, iTeos, Jazz Pharmaceuticals, Molecular Axiom, Oncorus, Pyramid Biosciences, SeaGen, Synthekine, and VBL Therapeutics outside the submitted work. S.-H.I. Ou reports grants from Mirati during the conduct of the study; as well as personal fees from Pfizer, JNJ/Janssen, AnHeart Therapeutics, BeiGene, Lilly, Elevation Oncology, DAVA Oncology LLP, and Turning Point Therapeutics outside the submitted work. G.J. Riely reports grants from Mirati during the conduct of the study; grants from Mirati, Lilly, Takeda, Merck, Roche, Pfizer, Rain Therapeutics, and Novartis outside the submitted work. W. Yang reports other support from Mirati Therapeutics Inc. during the conduct of the study. J.G. Christensen reports personal fees and other support from Mirati Therapeutics during the conduct of the study; personal fees from Mirati Therapeutics outside the submitted work; and reports patents 10,633,381, 10,689,377, and 10,125,134 issued. P.A. Janne reports personal fees from Mirati Therapeutics during the conduct of the study. P.A. Janne also reports grants and personal fees from AstraZeneca, Boehringer Ingelheim, Daiichi Sankyo, and Takeda Oncology; personal fees from Pfizer, Roche/Genentech, Chugai, Biocartis, Novartis, Sanofi Oncology, Transcenta, Silicon Therapeuticsons, Syndax, Nuvalent, Bayer, Eisai, Allorion Therapeutics, Accutar Biotech, Abbvie, Monte Rosa, Scorpion Therapeutics, Merus, Frontier Medicines, Hongyun Biotechnology, and Duality Biologics; and grants from Eli Lilly, SFJ Pharmaceuticals, Voronoi, Revolution Medicines, and PUMA outside the submitted work; and reports a patent for EGFR mutations issued and licensed to LabCorp. No disclosures were reported by the other authors.

C.P. Paweletz: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing. G.A. Heavey: Data curation, formal analysis, writing–review and editing. Y. Kuang: Data curation, formal analysis, supervision, methodology, writing–review and editing. E. Durlacher: Investigation, writing–review and editing. T. Kheoh: Data curation, formal analysis, visualization, writing–review and editing. R.C. Chao: Data curation, formal analysis, writing–review and editing. A.I. Spira: Investigation, writing–review and editing. K. Leventakos: Investigation, writing–review and editing. M.L. Johnson: Investigation, writing–review and editing. S.-H.I. Ou: Investigation, writing–review and editing. G.J. Riely: Investigation, writing–review and editing. K. Anderes: Formal analysis, supervision, writing–original draft, writing–review and editing. W. Yang: Data curation, formal analysis, methodology, writing–original draft, writing–review and editing. J.G. Christensen: Conceptualization, resources, supervision, writing–original draft, writing–review and editing. P.A. Janne: Conceptualization, resources, investigation, writing–original draft, writing–review and editing.

This study received funding from R01 CA240592–02 and Mirati Therapeutics. We like to thank all patients who participated in the trial as well as their families, their caregivers, and the study investigators. We have tried where possible to cite the primary publication in this field of early ctDNA response but relied on review articles to summarize the state of the art. We apologize to the authors whose work was summarized in our cited review articles.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.