Phase I/II Trial of Exportin 1 Inhibitor Selinexor plus Docetaxel in Previously Treated, Advanced KRAS-Mutant Non–Small Cell Lung Cancer

The past decade has seen truly meaningful progress in the treatment of advanced non–small cell lung cancer (NSCLC), with immune checkpoint inhibitors and molecularly targeted therapies leading these advances (1, 2). Nevertheless, particularly for tumors without actionable genomic alterations, there are few effective treatments once disease progression occurs after first-line therapy. Docetaxel, for decades the standard second-line treatment for advanced NSCLC regardless of Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation status, yields only modest efficacy, with time to progression of less than 3 months (3).

KRAS mutations are one of the most common oncogenic driver alterations, occurring in substantial proportions of lung, pancreatic, colon, and other cancers (4). Approximately 20% to 30% of advanced NSCLC cases harbor KRAS mutations, which occur predominantly in adenocarcinoma histology and—in contrast to most oncogenic driver mutations—in current or former smokers (5, 6). Due to its unusual protein structure and lack of classic drug-binding sites, mutant KRAS has historically been difficult to target therapeutically (7). Recently, the KRAS small-molecule inhibitors sotorasib and adagrasib received FDA approval, but their use is limited to KRAS^G12C mutations (approximately one third of KRAS-mutant NSCLC; refs. 6, 8). Compared with other targeted therapies, they provide relatively limited benefit, with a median progression-free survival (PFS) of approximately 6 months (9, 10). Accordingly, the treatment of KRAS-mutant NSCLC, both G12C and non-G12C variants, remains a major unmet clinical need.

Selective inhibitors of nuclear export (SINE) may play a role in treating KRAS-mutant NSCLC. SINEs bind and inactivate exportin 1, resulting in nuclear retention of key tumor-suppressor proteins and nucleic acids such as p53 (11). Selinexor is a SINE approved for relapsed/refractory multiple myeloma and diffuse large B-cell lymphoma (12, 13). In myeloma cells treated with SINE, nuclear retention of tumor-suppressor proteins activates cell-cycle checkpoint and genome surveying actions, leading to apoptosis (14).

In vitro and in vivo studies show potent selinexor activity against KRAS-mutant but not KRAS wild-type NSCLC (15). We therefore conducted a phase I/II study of selinexor plus docetaxel in previously treated, advanced KRAS-mutant NSCLC. We evaluated this combination because (i) the activity of docetaxel monotherapy has been extensively characterized in advanced NSCLC, including KRAS-mutant cases, and (ii) preclinical data suggest synergy between selinexor and taxane chemotherapy (16).

This phase I/II single-arm clinical trial (NCT03095612) was approved by the University of Texas Southwestern Medical Center Institutional Review Board (STU 032017-003) and by the institutional review boards of all participating centers. The study was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent prior to enrollment and undergoing any study procedures. The primary outcomes were safety and MTD of selinexor in combination with docetaxel. The secondary outcome was the best radiographic response according to RECIST v1.1 (17). Exploratory outcomes included PFS, overall survival (OS), and predictive and pharmacodynamic biomarkers.

Eligible patients had histologically or cytologically confirmed advanced NSCLC harboring an activating KRAS mutation (codons 12, 13, 61, and 117). Patients must have received at least one and up to two previous lines of systemic cytotoxic therapy (i.e., conventional chemotherapy) for advanced NSCLC, one of which must have been a platinum-based doublet. Patients may have received up to four total previous lines of systemic therapy [including immunotherapy and molecularly targeted therapy (including KRAS^G12C inhibitors)] for advanced NSCLC. Patients with stable brain metastases, either treated or untreated, were eligible. Additional inclusion criteria included measurable disease, Eastern Cooperative Oncology Group performance status 0 to 1, and adequate hematologic, renal, and hepatic function. Exclusion criteria included concurrent active malignancy that would interfere with treatment administration or assessment, unstable cardiovascular function, and prior exposure to docetaxel, selinexor, or another SINE. Overall representativeness of this study population to the general KRAS-mutant NSCLC population is shown in Supplementary Table S1.

Selinexor was administered once weekly orally starting 1 week prior to docetaxel initiation, which allowed assessment of selinexor pharmacodynamic effects alone and in combination with docetaxel. The dose of selinexor started at 60 mg and was escalated to 80 mg once weekly in a 3+3 schema. Docetaxel was started at 75 mg/m² intravenously once every 3 weeks with standard premedications. Prophylactic antiemetics were administered with each selinexor dose. Based on substantial rates of nausea and vomiting in selinexor monotherapy studies (18, 19), the weekly antiemetic regimen included 5-hydroxytryptamine 3 (5-HT3) antagonists and olanzapine. Study therapy continued until intolerable toxicity or disease progression.

Selinexor could be dose-reduced by two levels (60 mg weekly and 40 mg weekly) for patients receiving 80 mg weekly and one level (40 mg weekly) for patients receiving 60 mg weekly. Docetaxel could be dose-reduced to 60 mg/m² or 35 mg/m². For patients with toxicities clearly related to one but not both agents (e.g., peripheral neuropathy attributed to docetaxel), dose modification was applied to only one agent. Patients who discontinued one agent because of toxicity could continue the other agent. Patients could receive supportive care such as antiemetics, antidiarrheals, and growth factor support as needed.

Toxicities were graded according to the NCI Common Terminology Criteria for Adverse Events, version 4.0. Patients were assessed for radiographic response to therapy every two cycles. Dose-limiting toxicities (DLT) were assessed during the first cycle of treatment (28 days: 7-day selinexor lead-in plus 21 days of combination therapy). DLTs included the following when considered at least possibly related to study therapy: more than one missed dose of four selinexor doses during cycle 1 or discontinuation of study therapy before the completion of cycle 1 due to treatment-related toxicity; grade ≥3 nausea, vomiting, dehydration, diarrhea, or fatigue lasting >3 days despite optimal supportive care; other grade ≥3 nonhematologic toxicity (with exceptions of reversible and asymptomatic electrolyte abnormalities; alopecia; and grade 3 alanine aminotransferase, aspartate aminotransferase, or alkaline phosphatase elevations); delay in initiating cycle 2 >21 days; any other grade 4 nonhematologic toxicity; febrile neutropenia; grade 4 neutropenia lasting >7 days; and grade ≥3 thrombocytopenia with clinically significant bleeding, bruising, or requiring transfusion. To be evaluable for DLT, patients must have received at least one dose of docetaxel and at least three of four doses of selinexor (unless withheld or discontinued because of toxicity). Nonevaluable patients were replaced.

We obtained and manually reviewed available clinical next-generation sequencing (NGS) results for enrolled patients. If NGS had not been performed previously, we sent available pretreatment biopsy samples for exploratory analysis at Tempus. RNA sequencing (RNA-seq) was conducted using the TempusRS.v2 pipeline (RRID: SCR_025832). DNA sequencing was performed using the Tempus xT (v.4) pipeline (RRID: SCR_025832), covering 648 genes spanning ∼3.6 Mb of the genomic space (see Supplementary Methods S1). Single-sample gene set enrichment analysis was performed to obtain sample-level enrichment scores for the p53 pathway for each patient. These scores were compared between TP53 wild-type and TP53-altered cases using an unpaired t test (data were determined to be normally distributed using the Shapiro–Wilk test).

Because yes-associated protein 1 (YAP1), a transcription co-regulator, was associated with the efficacy of selinexor in preclinical studies (15), we determined YAP1 expression in archival tumor tissue by IHC. YAP1 antibody staining was performed using a Dako Autostainer Link 48 instrument in a research core facility (1:2,000, ab205270, Abcam, RRID: AB_2813833) with high-pH (10.0) Tris/EDTA solution (Agilent). Briefly, the slides were baked for 20 minutes at 60°C, then deparaffinized, and hydrated before the antigen retrieval step. Heat-induced antigen retrieval was performed at pH 9.0 for 20 minutes in Dako PT Link. Slides were incubated with a peroxidase block followed by antibody incubation (1:2,000 dilution) for 20 minutes.

Staining was visualized using the EnVision FLEX visualization system. YAP1 immunoreactivity was scored by a pathologist blinded to case clinical outcomes (U. Nadeem). The intensity of YAP1 staining was scored as follows: 0 (no staining), 1 (weak staining, faint yellow), 2 (moderate staining, light brown), and 3 (strong staining, brown). As designated in prior studies, a score of 3 was considered positive (20).

In preclinical KRAS-mutant NSCLC models, the primary mechanism underlying the effect of exportin 1 inhibition was accumulation of nuclear IκBα, with resulting NF-κB inhibition (15). Because serial tumor biopsies were not feasible, we analyzed baseline and posttreatment plasma cytokines as surrogates to characterize activity of this pathway (21). Blood samples were collected on cycle 1 day −7 (pre-selinexor, pre-docetaxel) and cycle 1 day 1 (after selinexor initiation, pre-docetaxel). Blood was centrifuged at 1,200 g at 4°C for 15 minutes to obtain plasma. Plasma cytokine levels were measured using the Bio-Plex Pro Human Chemokine 40-plex Panel (Bio-Rad Laboratories) on the Luminex 200 System (RRID: SCR_018025) as previously described (22).

Raw sequencing data for this study were generated at Tempus and processed by Tempus proprietary pipelines. Differential expression analysis and count normalization were performed using the DESeq2 package (RRID: SCR_015687; ref. 23). Gene set enrichment analyses were performed using the fgsea (RRID: SCR_020938) and hacksig (RRID: SCR_025833) packages for multisample and single-sample pathway analyses, respectively (24). Korotkevich and colleagues used fast gene set enrichment analysis (bioRxiv 2021.060012). Gene signature compilations were obtained from the Molecular Signatures database (RRID: SCR_016863; ref. 25). Oncoplots were generated using the maftools package (RRID: SCR_024519) and other plots generated using the ggplot2 package (RRID: SCR_014601).

The phase I component of the trial followed a 3+3 dose-escalation schema. For the phase II component, enrollment of 35 patients was determined to provide 85% power to detect an improvement in the primary endpoint of radiographic response rate from 10% (historical control) to 35%, with a two-sided α of 0.05. The sample size estimate was based on the exact one-sample binomial test for proportions. All patients receiving at least one dose of selinexor were considered evaluable for safety analyses. All patients who completed at least one cycle of therapy or discontinued therapy for reasons of toxicity or progression of disease were evaluable for efficacy. The rate of disease control [partial response (PR) or stable disease (SD)] and the exact 95% confidence intervals (CI) were computed. PFS and OS were estimated by the Kaplan–Meier method.

For cytokine analysis, the log₂ transformation was conducted on cytokine levels to mitigate deviation from normality. Comparisons between groups were performed using t tests for continuous variables, the Fisher exact test for categorical variables, and the log-rank test for event-time outcomes. Due to the exploratory nature of this study, no adjustment for multiple comparisons was performed.

The data generated in this study are not publicly available, as there may be information that could compromise patient privacy or consent. However, de-identified data may be available upon reasonable request to the corresponding author after review by the study sponsor, Karyopharm Therapeutics.

The trial enrolled 40 patients (7 in dose escalation and 33 in dose expansion) from April 2018 through October 2022. The median age was 66 years, 55% were female, 85% were White, and 80% were current or former smokers. The median number of prior total lines of therapy was two, 11 patients (28%) received prior taxane therapy (paclitaxel in all cases), and 2 patients (5%) received prior KRAS^G12C-targeted therapy. Among KRAS mutations, 11 (28%) were G12C.

Clinical NGS reports were available for 10 patients. For another 16 patients, we performed Tempus research NGS. Among the 26 total cases with NGS results, 11 (42%) had TP53 co-mutations and 7 (27%) had STK11 mutations. Additional patient and tumor characteristics are shown in Table 1.

Treatment-emergent adverse events (AE) by category and grade are shown in Table 2. The principal AEs of any grade were nausea (73%), fatigue (70%), neutropenia (65%), diarrhea (58%), anorexia (50%), vomiting (45%), alopecia (45%), and anemia (40%). The most common grade ≥3 AEs were neutropenia (63%), hyponatremia (15%), dyspnea/respiratory failure (15%), anemia (10%), and diarrhea (10%). Treatment-emergent AEs according to the dose level are displayed in Supplementary Table S2. Notably, all patients (n = 4) treated at the selinexor 80 mg orally weekly dosage had at least one grade ≥3 AE.

Seven patients (18%) had selinexor dose reductions, 25 (63%) had selinexor dose delays/interruptions, and 5 (13%) had selinexor discontinued because of toxicity. Twelve patients (30%) had docetaxel dose reductions, 16 (40%) had docetaxel dose delays/interruptions, and 11 (28%) had docetaxel discontinued because of toxicity. A total of 13 (33%) patients received myeloid growth factor support because of neutropenia, including 7 (18%) who received growth factor support as prophylaxis in subsequent cycles.

In the dose-escalation phase of the trial, three patients were enrolled to dose level 1 (selinexor 60 mg orally weekly plus docetaxel 75 mg/m² i.v. every 21 days), of which none had a DLT. Among the first three patients enrolled to dose level 2 (selinexor 80 mg orally weekly plus docetaxel 75 mg/m² i.v. every 21 days), one experienced a DLT of grade 4 neutropenia, and cohort expansion was planned. However, after a total of four patients were treated at that dose level, it became apparent that the regimen was not clinically feasible because of toxicities such as fatigue, nausea/vomiting, and neutropenia that did not meet formal criteria for DLT because of grade and/or timing. Accordingly, selinexor 60 mg orally weekly plus docetaxel 75 mg/m² i.v. every 21 days was deemed the MTD and recommended phase II dosage.

A total of 32 patients (80%) were evaluable for radiographic response as follows: PR(7, 22%), SD (18, 56%), and progressive disease (PD; 7, 22%). Because KRAS^G12C mutations now represent a distinct clinical subset because of the availability of FDA-approved direct KRAS^G12C inhibitors, we compared results in G12C patients (n = 11, 28%) with those in non-G12C patients (n = 29, 73%). Figure 1A and B shows radiographic response according to KRAS mutation type, with no significant difference observed (P = 0.30).

Of note, only two patients in the G12C group had previously received a KRAS^G12C inhibitor (sotorasib), which reflects the study’s enrollment period (April 2018–October 2022) and FDA approvals for KRAS^G12C inhibitors (June 2021 for sotorasib; December 2022 for adagrasib). One of the two patients previously treated with sotorasib achieved a near PR (sum of target lesion diameters decreased 28%) and then developed disease progression after six cycles of study therapy. The other received only one dose of selinexor before experiencing clinical deterioration attributed to the underlying cancer and dying approximately 2 weeks later.

Because TP53 alterations occur commonly as co-mutations in KRAS-mutant NSCLC and are known to influence the efficacy of nuclear export inhibitors and other treatments such as KRAS^G12C inhibitors (26–28), we also analyzed results according to TP53 status. Of the 26 patients from whom we obtained NGS results, 11 (42%) had loss-of-function TP53 alterations. These cases had significantly worse outcomes across all efficacy endpoints, including RECIST parameters. For TP53-altered cases, the response rate was 9% and the disease control rate was 27%, compared with 27% and 80%, respectively, for TP53 wild-type cases (P = 0.03; Fig. 2A and B).

PFS and OS for the entire cohort are shown in Fig. 3A. The median PFS was 4.1 months (95% CI, 1.1–5.3 months) for KRAS^G12C versus 4.0 months (95% CI, 1.9–4.9 months) for KRAS non-G12C (HR, 0.88; 95% CI, 0.43–1.83; P = 0.73). The median OS was 6.5 months (95% CI, 1.1 months–not reached) for KRAS^G12C versus 7.5 months (95% CI, 4.5–13.7 months) for KRAS non-G12C (HR, 0.95; 95% CI, 0.43–2.1; P = 0.91; Fig. 3B).

Figure 3C shows PFS and OS according to TP53 status. The median PFS was 1.8 months (95% CI, 1.0–4.1 months) for TP53-altered versus 7.4 months (95% CI, 1.5–12.0 months) for TP53 wild-type cases (HR, 0.22; 95% CI, 0.07–0.67; P = 0.003). The median OS was 5.8 months (95% CI, 1.2–12.0 months) for TP53-altered versus 15.7 months (95% CI, 1.8–29.5 months) for TP53 wild-type cases (HR, 0.36; 95% CI, 0.15–0.88; P = 0.02).

Among the 11 patients who previously received paclitaxel, eight were evaluable for RECIST best response (three PRs, two SDs, and three PDs) and nine were evaluable for PFS (median 4.2 months). The median OS was 5.9 months. TP53 status was available for eight patients (two wild-type cases; six altered cases).

For the 16 patients evaluated, Tempus research NGS mutation analysis identified 342 mutations in 212 genes. In addition to TP53, NOTCH4 and LRP1B were the most frequently mutated genes (n ≥8; Supplementary Fig. S1A and S1B). Through RNA-seq, we identified differential gene expression between cases with disease control (PR + SD) and those with PD (Supplementary Fig. S2A). These analyses performed on archival tumor specimens identified statistically significant (adjusted P < 0.05) enrichment of several hallmark pathways associated with response (29). Those most strongly associated with disease control were NF-κB, WNT, and p53 signaling pathways. Pathways most strongly associated with disease progression involved protein secretion, oxidative phosphorylation, and MYC signaling. Cases identified as enriched for the p53 pathway correlated significantly with TP53 wild-type status (P = 0.04; Supplementary Fig. S2B).

Pretreatment tissue from 12 patients was available for IHC assessment of YAP1, of which 6 had positive (IHC 3) and 6 had negative (IHC ≤2) YAP1 expression. Radiographic response was similar between the two groups: response rate, 17%; disease control rate, 66% in YAP1 strong and response rate, 17%; disease control rate, 83% in YAP1 weak–moderate (P = 1). For YAP1-strong patients, the median PFS was 6.0 months (95% CI, 1.8–22.6 months) versus 4.3 months (95% CI, 1.0–9.7 months) in YAP1 weak–moderate cases (HR, 0.40; 95% CI, 0.10–1.67; P = 0.20). The median OS was 19.1 months in YAP1 strong versus 5.9 months in YAP1 weak–moderate (HR, 0.41; 95% CI, 0.10–1.75; P = 0.20).

A total of 37 patients had baseline cytokine data and 32 had post-selinexor cytokine data. Among the 40 cytokines in the multiplex panel, four (IL-6, CXCL5, CCL1, and CX3CL1) had levels below the limit of detection and were excluded from the analysis. Supplementary Figure S3A–S3C displays those cytokines for which baseline, pretreatment values were significantly associated with outcomes (OS, PFS, and/or radiographic response). These include MIP-3a/CCL20, GCP-2/CXCL6, and IL-8/CXCL8. We also evaluated associations between pre- and post-selinexor (both pre-docetaxel) changes and clinical outcomes (Supplementary Fig. S4A–S4C). Greater decreases in MIP-1a/CCL3, eotaxin-2/CCL24, and GCP-2/CXCL6 were significantly correlated with improved outcomes. Among these, CCL20, CCL3, and CXCL8 are known to be induced by NF-κB (30). For all four of these cytokines, lower baseline levels and/or greater decreases after selinexor administration were associated with more favorable outcomes.

Notably, some patients reported improved symptoms (e.g., less pain) after the first dose of selinexor, prior to docetaxel administration. For some cases, we also noted early and profound radiographic responses that seemed atypical for docetaxel (Supplementary Fig. S5A–S5C). In some instances, clinical and radiographic findings seemed to correlate. For instance, a patient with substantial right back pain from tumor pleural involvement described a clear decrease in pain after one dose of selinexor (without change in the analgesic regimen or prior radiotherapy) when she presented 1 week later for docetaxel initiation. Her first postbaseline imaging study (after cycle 2) demonstrated clear reduction in the primary tumor, a right-sided pleural-based lesion (Supplementary Fig. S5B). Notably, all three cases featured here featured wild-type TP53.

Following these clinical observations, to estimate the effect of single-agent selinexor across the study population, we analyzed changes in serum lactate dehydrogenase (LDH) levels between cycle 1 day −7 (pretreatment baseline) and cycle 1 day 1 (after selinexor initiation and prior to docetaxel administration), as decreases after treatment initiation have been associated with subsequent favorable radiographic and clinical outcomes in lung cancer (31), according to TP53 status (Supplementary Fig. S6). Serum LDH changes had a near significant association with TP53 status, increasing a mean of 51 U/L in TP53-mutant cases and decreasing 48 U/L in TP53 wild-type cases (P = 0.06).

Given the compelling preclinical evidence for a patient cohort with a high unmet need, we conducted a single-arm clinical trial of the SINE selinexor plus docetaxel in previously treated, advanced KRAS-mutant NSCLC. Weekly patterns of fatigue, nausea, and vomiting indicated that these events were attributable to selinexor as well as docetaxel. With weekly multiagent antiemetic prophylaxis, the MTD as well as the recommended phase II dosage was selinexor 60 mg orally weekly plus docetaxel 75 mg/m² i.v. every 3 weeks.

KRAS-mutant NSCLC is widely recognized as a therapeutic challenge, with worse outcomes compared with KRAS wild-type disease (32). In the present study, across the entire study cohort, the response rate was approximately 20%, the disease control rate was about 80%, and the median PFS was around 4 months. Although no conclusions can be drawn from cross-study comparisons, these outcomes seem comparable or possibly favorable with historical controls. Previous studies of single-agent docetaxel in KRAS-mutant NSCLC showed response rates of 0% to 14%, disease control rates of 37% to 60%, and PFS ranging 2 to 4 months (10, 33–35). Consistent with studies of docetaxel monotherapy, prior receipt of paclitaxel did not seem to affect the efficacy of docetaxel plus selinexor (36).

In general, the efficacy of selinexor plus docetaxel seemed comparable across KRAS mutation types, a clinically meaningful observation because oncogene-directed targeted therapies are currently only available for one third of KRAS-mutant lung cancers that feature G12C mutations (6). Because direct KRAS^G12C inhibitors were not FDA approved until the last year of trial enrollment, only two patients received KRAS^G12C inhibitors prior to study treatment. Accordingly, we cannot make meaningful observations with regard to the sequencing of these therapies.

TP53 alterations—which occurred in almost half of treated cases, comparable with reported rates in broader KRAS-mutant NSCLC populations (37)—were strongly and significantly associated with detrimental clinical outcomes. TP53-altered cases had response and disease control rates less than half of those observed in TP53 wild-type tumors. Similarly, the median PFS and OS were less than half of those in TP53 wild-type cases. The extent to which this phenomenon may reflect tumor sensitivity to docetaxel versus selinexor cannot be determined in this single-arm trial. As a tumor suppressor, TP53 predicts poor prognosis and resistance to chemotherapy when mutated or absent (38, 39). In prostate cancer cell lines, TP53 mutants and knockouts have reduced docetaxel sensitivity (40). However, in a phase III trial of more than 1,800 patients with breast cancer, TP53 status did not predict sensitivity to docetaxel (41). In subset analysis of the CodeBreaK 200 trial of sotorasib versus docetaxel for KRAS^G12C-mutant NSCLC, TP53 alterations seemed to have a far greater detrimental effect on sotorasib than on docetaxel (42).

Mechanistically, TP53 status has clear relevance to the antitumor efficacy of nuclear export inhibition. By blocking nuclear export of p53, selinexor results in retention of p53 in the nucleus, potentially promoting and prolonging tumor-suppressive effects of wild-type p53 but conveying no such benefit if p53 is absent or mutated. Consistent with this mechanism, TP53-mutant cell lines are resistant to selinexor (43, 44). In an exploratory analysis from a trial of maintenance selinexor monotherapy in endometrial cancer, promising efficacy was observed in patients with TP53 wild-type tumors but not in patients with TP53-mutant tumors (45). These observations suggest the possibility that nuclear retention of mutant TP53 might have tumor-promoting properties, potentially by outcompeting other functional tumor suppressors and cell-cycle regulators (46, 47).

Whether our observations primarily reflect the influence of TP53 on docetaxel, selinexor, or both cannot be definitively determined in this single-arm study. Additionally, it is important to consider whether our observations support TP53 status as truly predictive rather than prognostic. Indeed, TP53 alterations convey worse prognosis across NSCLC stages and histologies (48). Nevertheless, the substantially greater rates of tumor shrinkage (including RECIST PRs) after treatment initiation suggest that beyond any prognostic implications, TP53 status conveys predictive effects in the present trial.

The trial’s staggered initiation design, in which selinexor was started 1 week prior to docetaxel, may provide hypothesis-generating insights to distinguish the effects of each therapy. For these observations, we compared dynamics in serum LDH during the window of selinexor monotherapy. In clinical oncology, LDH—a glycolytic enzyme found throughout normal tissues that has a half-life of approximately 24 hours—is most commonly used as a nonspecific tumor marker for monitoring germ cell tumors. LDH levels are also prognostic in NSCLC (31), and decreases in serum LDH after chemotherapy initiation are associated with improved survival (49). Given these considerations, the early LDH decrease after selinexor initiation observed only in TP53 wild-type tumors may suggest TP53-dependent single-agent activity of selinexor in KRAS-mutant NSCLC.

The relevance of TP53 status to treatment efficacy was also supported by tumor RNA-seq, which identified p53 signaling as one of the pathways most strongly associated with favorable outcomes. Based on preclinical models, we also hypothesized that correlative studies would demonstrate selinexor effects on NF-κB function. Indeed, in sequencing analyses, the NF-κB pathway was strongly associated with response, and reductions in NF-κB–induced cytokines after selinexor initiation were more pronounced in patients deriving greater benefit. Although tumor YAP1 expression predicted selinexor activity in preclinical studies, we did not observe this association in the present trial, which could reflect the small number of cases for which YAP1 IHC was performed.

Other limitations of this study include the single-arm design and the relatively small sample size, which precludes meaningful observations about treatment effects in subpopulations such as individuals previously exposed to KRAS^G12C inhibitors. Additionally, for the patients with clinical NGS results, we did not have access to the raw sequencing data and were therefore unable to include these cases in pathway analysis. Strengths of the study include a lead-in period to evaluate selinexor monotherapy and inclusion of detailed biomarker correlates that may inform further development of this regimen.

With multiagent supportive care, combination selinexor plus docetaxel is a feasible regimen in previously treated, advanced KRAS-mutant NSCLC. Although efficacy seemed comparable across all KRAS mutation subtypes, TP53 alterations were strongly associated with multiple clinical outcomes. Acknowledging the limitations of cross-trial comparisons, in TP53 wild-type tumors, efficacy exceeded that expected from docetaxel monotherapy. Moreover, rapid symptom improvement in some patients and early decreases in serum LDH suggest that selinexor may have single-agent activity in these cases. Further clinical evaluation of this treatment strategy in KRAS-mutant TP53 wild-type NSCLC is warranted.

T.F. Burns reports grants from Novartis and personal fees from Eli Lilly and Company, Advarra, Inc., Janssen Scientific Affairs, LLC, Jazz Pharmaceuticals Inc., Amgen, AstraZeneca, Takeda Pharmaceuticals U.S.A., Inc., Genentech, and Pfizer outside the submitted work. J.E. Dowell reports grants from Karyopharm Therapeutics during the conduct of the study, as well as personal fees from Takeda, Regeneron, Catalyst, Jazz Pharmaceuticals, AstraZeneca, Janssen, and BeiGene and other support from Puma Biotechnology outside the submitted work. L. Horn reports grants from Karyopharm Therapeutics Inc. during the conduct of the study, as well as other support from AstraZeneca outside the submitted work. D.R. Camidge reports serving in an advisory role (ad hoc advisory boards/consultations) for AbbVie, Anheart Therapeutics, Apollomics, AstraZeneca/Daiichi Sankyo, BeiGene, Betta Pharmaceuticals, Bristol Myers Squibb, Eli Lilly and Company, Ellipses, Gallapagos, GENESIS Pharma, Gilead, Imagene, InduPro, Janssen, Kestrel, Pfizer, Roche, Sutro, Takeda, and Triana. H. Mu-Mosley reports a patent for US patent applications 63/386387 and 63/382972 pending. D.E. Gerber reports grants from Karyopharm Therapeutics during the conduct of the study, as well as grants from AstraZeneca and Novocure, personal fees from Catalyst Pharmaceuticals, AstraZeneca, Elevation Oncology, Janssen Scientific Affairs, Jazz Pharmaceuticals, Regeneron Pharmaceuticals, Sanofi, and Daiichi Sankyo, and other support from Gilead and OncoSeer Diagnostics, Inc., outside the submitted work; in addition, D.E. Gerber reports a patent 11747345 issued to the University of Texas Southwestern Medical Center and patents for 17/045482, 18/504868, 63/386387, 63/382972, and 63/382257 pending. No disclosures were reported by the other authors.

M.S. von Itzstein: Resources, data curation, investigation, writing–original draft, writing–review and editing. T.F. Burns: Resources, investigation, writing–review and editing. J.E. Dowell: Resources, investigation, writing–review and editing. L. Horn: Resources, investigation, writing–review and editing. D.R. Camidge: Resources, investigation, writing–review and editing. S.J. York: Resources, investigation, writing–review and editing. K.D. Eaton: Resources, investigation, writing–review and editing. K. Kyle: Project administration, writing–review and editing. F.J. Fattah: Resources, supervision, investigation, project administration, writing–review and editing. J. Liu: Data curation, formal analysis, visualization, writing–review and editing. H. Mu-Mosley: Data curation, formal analysis, visualization, writing–review and editing. A. Gupta: Resources, methodology, writing–review and editing. U. Nadeem: Resources, investigation, writing–review and editing. A. Gao: Data curation, formal analysis, visualization, writing–review and editing. S. Zhang: Data curation, formal analysis, visualization, writing–review and editing. D.E. Gerber: Conceptualization, data curation, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

This research was supported by funding from Karyopharm Therapeutics, Newton, Massachusetts (D.E. Gerber); the University of Texas Lung Specialized Program of Research Excellence (P50CA070907-21; D.E. Gerber); and the Clinical Research Office, Biostatistics Shared Resource, Data Sciences Shared Resource, and Biomarker Research Core of the Harold C. Simmons Comprehensive Cancer Center (1P30 CA 142543-03; D.E. Gerber, F.J. Fattah, H. Mu-Mosley, J. Liu, S. Zhang, and A. Gao). The authors thank Dru Gray for assistance with manuscript preparation. The authors also thank Christopher J. Walker and Andrea A. Ellero of Karyopharm Therapeutics for assistance with mutation analyses.