Gemcitabine plus selinexor in selective advanced sarcomas: a phase I of the Spanish group for research on sarcoma study

Abstract

Exportin-1 (XPO-1) is related to drug resistance and poor prognosis in solid tumors. Selinexor, an XPO-1 inhibitor, has shown preclinical and clinical activity in sarcomas. This Phase I study explores the combination of gemcitabine and selinexor in a classic 3 + 3 design. Adult patients with selected advanced sarcomas receive gemcitabine and weekly selinexor in 21-day cycles. The main endpoint is to determine the recommended phase 2 dose (RP2D). Secondary end-points include safety, overall response rate (ORR), overall survival (OS), and quality of life. Seventeen patients are included in this study. One dose-limiting toxicity (grade 4 thrombocytopenia) is detected in dose-level +3, but the R2PD is established at dose-level +2 (gemcitabine at 1200 mg/m² at 10 mg/m²/min followed by 60 mg weekly selinexor) based on its better tolerability. The most frequent adverse events are neutropenia (82.4%) and thrombocytopenia (76.5%). The ORR is 31.25 %, and the median OS (mOS) is 39.5 months (95% CI, 12.4-67) with a 36-month OS rate of 50.2%. A phase II is currently exploring this combination in leiomyosarcoma and malignant peripheral nerve sheath tumors. Trial registration: NCT04595994.

Exportin-1 (XPO-1) inhibitor selinexor has been reported as a therapeutic option in solid cancer. Here, the authors present a phase 1 clinical trial combining selinexor with chemotherapy by gemcitabine in patients with advanced sarcoma.

Introduction

Patients with advanced sarcomas represent a vulnerable population with a poor prognosis. The median overall survival (OS) has increased in the last decade due, at least partially, to the emergence of new treatment options in second lines. Specifically, in advanced soft tissue sarcomas (STS) the survival expectancy is now above one and a half years if we consider the recent randomized phase III trials^1,2. In the case of osteosarcoma patients, the figures are even worse, with life expectancy below one year in metastatic relapsed disease, except for surgically rescued cases^3–5.

Therefore, new strategies or new drugs with different mechanisms of action need to be incorporated into the therapeutic arsenal of advanced sarcoma patients.

Overexpression of XPO-1 has been related to therapy resistance and poor survival in solid tumors⁶. Selinexor is a first-in-class, highly specific small-molecule inhibitor of exportin-1 (XPO-1, also known as CMR1). This protein is the main mediator of nuclear export in many cell types, and it mediates the leucine-rich nuclear export signal. Inhibiting this function hinders the escape of various tumor suppressor proteins, thereby strengthening tumor suppression⁷. Selinexor as a monotherapy has received FDA approval for the treatment of relapsed or refractory diffuse large B-cell lymphoma and triple-class refractory myeloma, based on clinical trials that demonstrated overall response rates of 28⁸ and 26%⁹, respectively.

In preclinical studies, Selinexor’s efficacy was evaluated across multiple sarcoma subtypes through in vitro and in vivo models, encompassing 17 cell lines and 9 xenograft models. Most sarcoma cell lines showed sensitivity to selinexor with IC₅₀ values ranging from 28.8 nM to 218.2 nM. Selinexor suppressed sarcoma xenograft growth, including an alveolar soft part sarcoma model that was resistant in vitro¹⁰.

Selinexor, in combination with gemcitabine and other agents, exhibited increased efficacy in mouse xenograft models of pancreatic carcinoma¹¹. The synergistic action may derive from the biological effect of the selinexor reducing mRNA and the protein expression of DNA damage-repairing gene products. Thus, selinexor prevents different repair signaling, providing rationale for a combination with DNA damage agents, such as gemcitabine¹². Importantly, the order of administration proved to be important, with cytotoxicity being greater when the DNA-damaging agent was given prior to selinexor. Theoretically, selinexor could induce apoptosis not only through the blockade of DNA repair gene products but also by nuclear retention of p27, induction of Bax protein, and depletion of survivin¹¹. The Fas pathway can represent another potential synergistic mechanism between gemcitabine, a Fas sensitizer¹³, and selinexor, a Fas activator¹⁴.

Selinexor has been tested in a phase I trial including patients with solid tumors. In this trial, among the 157 patients evaluated for efficacy, 1 patient had a complete response (CR), 6 patients had partial responses (PR), and 67 patients had stable disease (SD), with 27 (17%) patients reaching stabilization at least for 4 months. Interestingly, a patient with STS had stable disease for almost two years¹⁵. The toxicity profile, predominantly characterized by mild gastrointestinal symptoms such as nausea, vomiting, and anorexia, as well as general effects including weight loss and fatigue, is manageable with supportive medications. This enhances tolerability, as evidenced in patients undergoing long-term treatment with selinexor, and supports its use in combination with other therapies. In addition, the metabolism of selinexor is independent of the cytochrome P450 isoenzymes, making its combination with other drugs more feasible. Additionally, selinexor was tested in a phase Ib trial involving 54 patients with STS and bone tumors that had progressed after at least one prior line of treatment. SD was reported in 30 out of 52 evaluable patients (58%), with 17 (33%) remaining stable for at least 4 months. The median progression-free survival (PFS) was 4.2 and 3.7 months for liposarcoma and leiomyosarcoma patients, respectively¹⁶. In the correlative studies associated with the previously mentioned phase I trials, selinexor was shown to decrease cellular proliferation and enhance apoptosis, attributable to the nuclear accumulation of tumor suppressor proteins, including p53 and FOXO3^15,16.

Based on previous considerations, a phase I trial was designed to explore the combination of gemcitabine plus selinexor in patients with sarcomas. Here we present the results of this trial, defining the recommended phase II dose of the combination and preliminary data of activity.

Results

Preclinical results

Leiomyosarcoma cell lines exhibited IC₅₀ values for selinexor ranging from 0.050 µM (AA) to 0.256 µM (CP0024), whereas IC₅₀ values for gemcitabine ranged from 0.342 nM (SK-UT-1) to 9.70 nM (IEC005) within the same histologic subtype. In osteosarcoma cells, selinexor IC₅₀ values ranged from 0.130 µM (SAOS-2) to 0.146 µM (U2OS), while gemcitabine IC₅₀ values ranged from 0.399 nM (SAOS-2) to 0.914 nM (U2OS). For MPNST cells, selinexor IC₅₀ values ranged from 0.102 µM (ICP060) to 1.60 µM (S462), and gemcitabine IC₅₀ values ranged from 14.5 nM (ICP060) to 35.1 nM (S462) in this histology (Supplementary Table S1). Three leiomyosarcoma cell lines exhibited synergistic CI values, with CP0024 and SK-UT-1 both at 0.79 and AA at 0.60. In contrast, another leiomyosarcoma cell line, IEC005, showed slight antagonism with a CI of 1.19. Similarly, among MPNST cell lines, ICP060 (0.76), S462 (0.88), and sNF96.2 (0.78) displayed synergistic interactions, whereas 88-14 showed antagonism with a CI of 1.38. In osteosarcoma cell lines, all combinations were antagonistic, with CI values of 1.67 for MG-63, 1.123 for U2OS, and 1.54 for SAOS-2 (Fig. 1A).

Fig. 1. Selinexor and gemcitabine preclinical synergy evaluation.

A Isobologram displays the IC₅₀ concentration of selinexor on the X-axis and the IC₅₀ concentration of gemcitabine on the Y-axis. The combination point represents the IC₅₀ values for the combined treatment: the X value indicates the IC₅₀ for selinexor in the combination, and the Y value indicates the IC₅₀ for gemcitabine in the combination. A combination-point below the line connecting the single-agent IC₅₀s indicates synergy, on the line indicates an additive effect, and above the line indicates antagonism (experimental replicas on different days: n = 3 with 3 technical replicas). B Apoptosis was assessed by Annexin V-FITC/propidium iodide (PI) flow cytometry assay. Quantification of necrotic and apoptotic cell populations is presented as percentages of total LMS (CP0024, SK-UT-1, AA and IEC005), osteosarcoma (MG-63, U2OS and SAOS-2), and MPNST (ICP060, S462, sNF96.2 and 88-14) cells after 72-h treatment with either 1 µM selinexor, 1.5 nM gemcitabine, or the combination of both (experimental replicas on different days: n = 3). Bar plots represent mean values with standard deviations; dots indicate individual samples. For statistical analysis, two-tailed t-tests were conducted. *p values: CP0024 p: 0.049, SK-UT-1 p: 0.016, ICP060 p: 0.017, S462 p: 0.011.

The evaluation of apoptotic populations following single-agent or combination treatments was consistent with these results. Specifically, annexin V-positive cells were significantly higher in the combination treatment compared to selinexor monotherapy for CP0024 (34.0% ± 2.9 vs. 25.1% ± 5.4, p < 0.05) and SK-UT-1 (59.9% ± 11.3 vs. 32.2% ± 3.5, p < 0.05), but not for IEC005 (29.3% ± 3.7 vs. 31.7% ± 4.8, p ≥ 0.05), which aligned with the additive effect observed in viability assays. In MPNST, apoptotic populations were also significantly higher in the combination treatment compared to selinexor: ICP060 (25.2% ± 1.7 vs. 16.5% ± 3.5, p < 0.05), S462 (70.7% ± 0.4 vs. 38.1% ± 4.9, p < 0.05), and 88–14 (69.45% ± 8.9 vs. 52.8% ± 7.5, p < 0.05), although the latter did not reach statistical significance and also corresponded with a lack of synergy in cell viability. In contrast, osteosarcoma cell lines showed no significant differences between combination and single-agent treatments, with apoptotic populations of 62.0% ± 2.2 vs. 58.1% ± 5.5, p ≥ 0.05 for MG-63, 55.2% ± 2.1 vs. 56.1% ± 8.2, p ≥ 0.05 for SAOS-2, and 48.3% ± 13.2 vs. 44.2% ± 10.5, p ≥ 0.05 for U2OS (Fig. 1B).

In the CP0024 cell line, we observed elevated DNA damage levels with the combination of selinexor and gemcitabine compared to only gemcitabine, as measured by γH2A.X accumulation in three separate experiments (Supplementary Fig. 1A and Source Data file). In contrast, in three cell lines where the combination exhibited antagonistic effects, γH2A.X levels were normally not increased in the combination-treated groups compared to gemcitabine monotherapy (Supplementary Fig. 1B–D).

Notably, survivin protein expression was completely abolished in CP0024 cells following selinexor treatment (Supplementary Fig. S1A). In the other three cell lines, survivin levels were partially reduced but not fully depleted (Supplementary Fig. S1B–D). Immunofluorescence analysis revealed selinexor-induced nuclear accumulation of IκBα in the CP0024 cell line (Supplementary Fig. S1E), which was absent in IEC005, MG-63, and SAOS-2 cells (Supplementary Fig. S1F–H), suggesting that IκBα nuclear localization may be critical for achieving complete survivin depletion and mediating the synergistic effect of selinexor with gemcitabine.

Clinical results

From November 2020 to September 2022, 17 patients diagnosed with advanced and progressing sarcoma were assessed for eligibility in 5 hospitals affiliated with the Spanish Sarcoma Group (GEIS) and with expertise in sarcoma care. These 17 enrolled patients underwent the designed treatment comprising the intent-to-treat and safety populations (Fig. 2). Two patients in the +3-dose level followed a noncompliant treatment regimen and were excluded from the DLT analysis. The first patient received the gemcitabine infusion in 30 min instead of 2 h on day 1 of cycle 1. The second patient had taken 20 mg instead of 80 mg of selinexor on day 1 of cycle 1. Both patients remained in the trial after correcting the dosages for subsequent administrations. The clinical cutoff for the data analysis was April 28, 2024. At that time, 1 patient was still receiving treatment in the clinical trial, while 16 patients had discontinued the treatment, 14 (87%) because of progression, 1 (6%) due to patient refusal, and 1 (6%) because of the investigator’s decision.

Fig. 2.

Consort diagram.

In total, 168 21-day cycles were administered. On average, participants received 4 cycles, with a range from 1.5 to 47 cycles. The median number of selinexor doses given was 9, ranging from 4 to 111 doses, while for gemcitabine, the median was 8 doses, with a range from 3 to 84 doses.

There were selinexor and gemcitabine administration delays in 82 and 76% of patients, respectively. Dose reduction of selinexor or gemcitabine occurred in 23 and 41% of patients, respectively. The median relative dose-intensity for gemcitabine was 87% (30–100), whereas for selinexor it was 83% (47–100).

The median age of patients was 50 years (22–71), with a male/female ratio of 53%/47%. In terms of baseline performance, 14 (82%) individuals were classified as ECOG 0, and 3 (18%) as ECOG 1. The histological subtypes were distributed as follows: leiomyosarcoma 9 (53%), osteosarcoma 6 (35%), alveolar soft part sarcoma 1 (6%), and synovial sarcoma 1 (6%). At the time of inclusion, 15 (88%) of the patients presented metastasis, while 2 (12%) had locally advanced, non-resectable tumors (Table 1).

Table 1.

Patient demographics

Characteristics	N (%)
Median age (range)	50 (22–71)
Sex (M/F)	9 (53%)/8 (47%)
Extension at diagnosis:
- Localized	9 (53%)
- Locally advanced	2 (12%)
- Metastatic	6 (35%)
ECOG baselinea:
- 0	14 (82%)
- 1	3 (18%)
Extension at baselinea:
- Locally advanced	2 (12%)
- Metastatic	15 (88%)
Median metastasis-free intervalb (range) (all patients)	11.9 (0–118.5)
Median metastasis-free intervalb (range) (metastatic only)	11.2 (0–118.5)
Median previous lines (range)	1 (1–2)
Histology:
- Leiomyosarcoma	9 (53%)
- Osteosarcoma	6 (35%)
- Alveolar soft part sarcoma	1 (6%)
- Synovial sarcoma	1 (6%)

^aBaseline refers as measurements and observations at the time of patient enrollment, before any intervention of the study.

^bThe metastasis-free interval is the time elapsed from initial localized diagnosis until the disease becomes advanced (metastatic or locally inoperable). If advanced at diagnosis, this interval is zero. M male; F female.

Safety and tolerability

No DLTs were detected from 0 to +2 dose-levels, with 3 patients treated at each level, while one DLT (grade 4 thrombocytopenia) was detected in a patient accrued at +3 dose-level. This latter dose-level was therefore expanded accordingly up to 6 patients; however, given that two patients were deemed non-evaluable for DLT due to dose-nonadherence, this cohort was expanded to include a total of eight patients. In this expanded cohort, no further DLTs were observed.

Although the recommended phase II dose (RP2D) was technically at the +3 dose-level (with gemcitabine at 1200 mg/m² administered at 10 mg/m²/min followed by selinexor at 80 mg weekly), the overall tolerance was noticeably better with selinexor at 60 mg weekly. Therefore, the RP2D was determined to be the +2 dose-level (gemcitabine at 1200 mg/m² at 10 mg/m²/min followed by selinexor at 60 mg weekly). When examining the highest toxicity events experienced by each patient, treatment-related hematologic side effects were the most common and occurred in the following distribution: neutropenia 82.4%, thrombocytopenia 76.5%, and anemia 70.6%. Among the grade 3 or 4 toxicities, the hematologic ones were still the most common, distributed as neutropenia (64.7%) and thrombocytopenia (47.1%). Regarding non-hematologic toxicities, nausea (70.6%), vomiting (64.7%), asthenia (64.7%), and ALT increase (47.1%) were the most frequent, and in most patients, they were grade 1-2 (Table 2). Granulocyte-colony stimulating factors (G-CSF) were used throughout the study, beyond the first cycle, to prevent delays and infections in 11 of 17 (65%) accrued patients. The most frequent schedules were 3 doses of G-CSF from day 2 and 5–7 doses from day 9 of each cycle. Only 2 (11%) patients experienced febrile neutropenia. No toxic death was reported in this trial. Notably, no patients left the study because of toxicity. This shows that managing bone marrow and gastrointestinal side effects, as well as following dose delay and reduction guidelines, was effective in keeping patients in the study. The global health status was analyzed using items 29 and 30 of the QLQ-C30, following the EORTC recommendations. There was a moderate improvement in the perception of global health status (GHS)/ quality of life (QoL), throughout the treatment, especially after 4 cycles of treatment (Supplementary Table S2).

Table 2.

Safety profile

Adverse event term	Any grade	Grade 1-2	Grade 3-4
Haematological
Neutropenia	14 (82.4%)	3 (17.6%)	11 (64.7%)
Lymphocytopenia	13 (76.5%)	7 (41.2%)	6 (35.3%)
Thrombocytopenia	12 (70.6%)	4 (23.5%)	8 (47.1%)
Anemia	12 (70.6%)	10 (58.8%)	2 (11.8%)
Leukopenia	10 (58.8%)	3 (17.6%)	7 (41.2%)
Febrile Neutropenia	2 (11.8%)	0	2 (11.8%)
Lymphocytosis	1 (5.9%)	1 (5.9%)	0
Non-Haematological
Nausea	12 (70.6%)	10 (58.8%)	2 (11.8%)
Vomiting	11 (64.7%)	10 (58.8%)	1 (5.9%)
Asthenia	11 (64.7%)	10 (58.8%)	1 (5.9%)
ALT Increased	8 (47.1%)	7 (41.2%)	1 (5.9%)
Diarrhea	6 (35.3%)	6 (35.3%)	0
Anorexia	6 (35.3%)	6 (35.3%)	0
AST Increased	6 (35.3%)	6 (35.3%)	0
Alopecia	4 (23.5%)	3 (17.6%)	1 (5.9%)
Rash	3 (17.6%)	3 (17.6%)	0
Lethargy	3 (17.6%)	3 (17.6%)	0
Dysgeusia	3 (17.6%)	3 (17.6%)	0
Constipation	3 (17.6%)	3 (17.6%)	0
Paresthesia	2 (11.8%)	2 (11.8%)	0
Myalgia	2 (11.8%)	2 (11.8%)	0
GGT Increased	2 (11.8%)	2 (11.8%)	0
Creatinine Increased	2 (11.8%)	1 (5.9%)	1 (5.9%)
Pain of skin	1 (5.9%)	1 (5.9%)	0
Mucositis Oral	1 (5.9%)	1 (5.9%)	0
Lipase Increased	1 (5.9%)	0	1 (5.9%)
Hyponatremia	1 (5.9%)	1 (5.9%)	0
Hypokalemia	1 (5.9%)	0	1 (5.9%)
Hypoglycemia	1 (5.9%)	1 (5.9%)	0
Hypermagnesemia	1 (5.9%)	0	1 (5.9%)
Hyperhidrosis	1 (5.9%)	1 (5.9%)	0
Fever	1 (5.9%)	1 (5.9%)	0
Epistaxis	1 (5.9%)	1 (5.9%)	0
Abdominal Pain	1 (5.9%)	1 (5.9%)	0
ALP Increased	1 (5.9%)	1 (5.9%)	0

ALT alanine aminotransferase, AST Aspartate aminotransferase, GGT gamma glutamyltransferase, ALP alkaline transferase.

Efficacy results

Based on the central radiological assessment, 5 of 16 evaluable patients obtained a partial response (31.25 %), 5 (31.25 %) had stable disease, and 6 (37.5 %) progressed, for an ORR of 31.25% by RECIST 1.1 criteria (Fig. 3A). There were only 9 out of 17 cases with sufficient quality in the CT scan to be considered evaluable by Choi criteria. Following these latter criteria, the distribution of partial response, stable disease, and progression was 5 (55.6%), 3 (33.3%), and 1 (11.1%), respectively (Supplementary Fig. S2). Focusing on the leiomyosarcoma subset, 4 of 9 (44.4%) patients achieved a PR (Fig. 3A).

Fig. 3. Efficacy of the combination of selinexor and gemcitabine in patients diagnosed with advanced sarcomas.

A Waterfall plot of response to treatment according to RECIST 1.1 (central radiological review). All evaluable patients (n = 16) are shown. Tumor diameter was measured in mm. The dashed lines represent a 20 % diameter increase and a 30% diameter decrease (RECIST progression and response cut-offs, respectively). PD progressive disease, SD stable disease, PR partial response, DL dose level, LMS leiomyosarcoma, ASPS alveolar soft-part sarcoma, OS osteosarcoma, SS synovial sarcoma. B Swimmer plot illustrating progression-free survival (PFS) of all the patients treated in the study. PI Principal investigator.

With a median follow-up of 30 months (95% CI, 18–41), the median OS was 39.5 months (95% CI, 12.4–67), with a 36-month OS rate of 50.2%. The median PFS (mPFS) was 5.6 months (95% CI, 1.6–9.5) for the whole cohort of 17 patients. In the subset of patients diagnosed with leiomyosarcoma, the mPFS was 7.6 months (95% CI 1.9–13.2), while the median OS (mOS) was 39.5 months (95% CI, 8.3–70.7). At the opposite end, the osteosarcoma group exhibited an mPFS of merely 1.2 months (95% CI, 1–1.3)—Fig. 3B and Supplementary Fig. S3.

The achievement of a response by RECIST criteria was the only clinical factor related to a significantly longer PFS in the univariate analysis (Supplementary Table S3).

Exploratory non-prespecified outcomes

Protein expression analyses by immunostaining were fully assessable for all the biomarkers in 15 out of 17 cases for p53, Calbindin 1, and survivin, and 14 out of 17 for IκBα antibodies (Supplementary Table S4). Among the four proteins selected for immunohistochemistry, Calbindin 1 expression was negative in all the cases analyzed, while survivin and p53 did not significantly correlate with clinical outcomes in the complete cohort. The high intensity of IκBα nuclear expression (score 4) was significantly associated with worse PFS [1.4 months (95% CI 0–3.9) vs 9.6 months (95% CI 0–25.2), p = 0.047] in the whole cohort (n = 14). In the leiomyosarcoma subgroup of patients, high expression of survivin (>50%) correlated significantly with worse PFS [(n = 9); 3.4 months (95% CI 0–7.6) vs. 15.4 months (95% CI 3.1–27.7), p = 0.022], whereas the high intensity of IκBα expression (level 4) was significantly associated with worse PFS [(n = 8); 3.4 months (95% CI 0–7.6) vs. 15.4 months (95% CI NR), p = 0.007], (Supplementary Table S5 and Supplementary Fig. S4). No clinical or molecular factors showed a correlation with ORR or clinical benefit (defined as PR or SD), as shown in Supplementary Table S6.

Discussion

In this phase I trial, the combination of gemcitabine plus selinexor was feasible and manageable. Even when just one case of DLT was detected in the highest dose-level with gemcitabine 1200 mg/m² at 10 mg/m²/min followed by selinexor 80 mg once per week in 6 evaluable patients, it seemed more reasonable to recommend the lower dose-level with gemcitabine 1200 mg/m² at 10 mg/m²/min followed by selinexor 60 mg as the RP2D based on the considerable need for delay, omission or reduction related to the highest dose-level. Although our study followed the standard Phase I trial design, where the RP2D is selected based on a 20–33% dose-limiting toxicity (DLT) rate in the first cycle, we recommended dose level +2 (below the suggested level by DLTs) due to observed cumulative toxicity¹⁷. While several patients (65%) required G-CSF after the first cycle, 65% and 47% experienced grade 3 or 4 neutropenia and thrombocytopenia, respectively. We recommend the systematic use of G-CSF for 3 and 5 consecutive days from days 2 and 9 of each cycle. Gastrointestinal side effects were the most frequent non-hematologic toxicities, in particular nausea and vomiting. Prophylactic ondansetron and olanzapine recommended in the protocol, were not systematically prescribed in all patients. However, the personalized application of different antiemetics, mainly ondansetron and olanzapine, along with metoclopramide and dexamethasone, made the difference for achieving adequate gastrointestinal tolerance. Quality of life analyses indicated that patients perceived general health status did not significantly worsen throughout the study.

This study shows notable activity with an ORR of 31% based on RECIST criteria and central review, and a mPFS of 5.7 months. These results warrant further investigation of this treatment regimen as a second-line option for patients with advanced, progressing sarcoma. This activity favorably compares with the outcome of other gemcitabine-based combinations in the same setting. For example, gemcitabine plus docetaxel achieved an ORR and an mPFS of 16% and 6.2 months, respectively, in a randomized phase II trial in a cohort of individuals with STS, with 40% being leiomyosarcomas, showing superiority for the combination over gemcitabine alone¹⁸. Gemcitabine plus dacarbazine, exhibited an ORR and a mPFS of 12% and 4.2 months, respectively, in a positive randomized phase II trial compared with dacarbazine alone, in the context of individuals diagnosed with STS, of which 28% were leiomyosarcoma¹⁹. In osteosarcoma, the combination of gemcitabine plus rapamycin obtained an ORR of 6% and an mPFS of 2.8 months in a single-arm phase II trial⁴.

The activity signals detected in this study were mainly related to the leiomyosarcoma subgroup, showing an ORR of 44.4% and an mPFS of almost 8 months. This data is in line with the observations of our preclinical study, where we detected signals of synergy, in cell viability studies, between both drugs in 3 out of 4 leiomyosarcoma cell lines, and despite the limited number of patients, it seems promising compared with other gemcitabine-based combinations in patients diagnosed with leiomyosarcoma. Gemcitabine and docetaxel showed an ORR of 24% and an mPFS of 4.7 months in the context of uterine leiomyosarcoma, while they showed an ORR of 5% and a mPFS of 3.8 months for non-uterine leiomyosarcoma in a comparative study unable to demonstrate the superiority of the combination over gemcitabine alone²⁰. Likewise, the mOS of 39.5 months in our study was found in the leiomyosarcoma subset. The mOS of gemcitabine-based studies in progressing patients with STS ranged from 16.8 to 17.9 months, while in the specific scenario of leiomyosarcoma, the mOS was 13 and 23 months for non-uterine and uterine leiomyosarcoma populations, respectively. The long-term survival observed in our study may largely be attributed to chance, given the limited number of patients. We can speculate about a potential immunomodulatory effect associated with XPO1 inhibition^21,22. This is an area that requires further investigation. In any case, the results of activity in this study, including survival, should be interpreted with caution due to the small number of patients analyzed.

Our results indicate a significant correlation between increased survivin expression, and reduced efficacy of the combination of gemcitabine and selinexor. This could prove crucial for improved patient selection towards more personalized treatment, pending validation in future studies, at least in patients with a leiomyosarcoma diagnosis. Of note, the dynamic decrease of survivin expression has been previously associated with the induction of apoptosis by selinexor, in dedifferentiated liposarcoma²³. In our preclinical models, the complete depletion of survivin seemed to correlate with the synergy observed between both experimental drugs, whereas in cell lines in which selinexor did not completely abrogate survivin, the combination showed signs of antagonism. Thus, it is possible that in cases where survivin is not fully depleted, it maintains certain levels of surviving signals, which could correlate with drug resistance and worse efficacy of the combination. However, we cannot rule out the possibility that survivin expression is related to selinexor activity rather than to the combination, since selinexor monotherapy abrogated this protein’s expression in several cell lines that showed similar sensitivity to the compound in our preclinical experiments. Nonetheless, survivin is also a key protein involved in cell cycle progression, participating in the formation of the chromosome passenger complex (CPC), a master regulator of cell division, alongside Aurora Kinase B, the scaffolding protein inner centromere protein (INCENP), and borealin^24,25. Thus, it is possible that in cases with survivin overexpression, selinexor treatment can significantly decrease the levels of this protein in the cytoplasm, and reduce the survival signaling mediated by survivin; however, the nuclear overexpression of survivin may promote cell division and tumor proliferation. Moreover, survivin loss can cause merotelic kinetochore attachments and polyploidy, generating double-strand breaks (DSBs) during aberrant mitosis, and triggering γH2A.X foci formation²⁶. These observations are in line with our results, in which we observe that the complete loss of survivin increased the protein levels of γH2A.X induced by the combination of selinexor plus gemcitabine. The complete loss of survivin seemed to be associated with the nuclear accumulation of IκBα, which was only observed in our preclinical experiments in the CP0024 cell line. IκBα had been reported to regulate survivin by sequestering NF-κB in the nucleus, after selinexor treatment, thereby suppressing its transcriptional activation of survivin, an anti-apoptotic protein²⁷. It is worth noting that the low intensity of IκBα expression, before treatment, was associated with better PFS in our cohort. A possible hypothesis to be further tested is that survivin binds the IKKβ promoter, increasing its transcriptional activity. This could elevate IKKβ expression, leading to IκBα phosphorylation/degradation and subsequent NF-κB nuclear translocation. NF-κB then could upregulate survivin, creating a feed-forward loop²⁸. Future preclinical studies should address these hypotheses in order to improve clinical study design. On the other hand, the expression of calbindin 1 (CALB1) has been described to be a predictive biomarker for selinexor in dedifferentiated liposarcoma¹⁶. However, in our cohort, we could not detect calbindin 1 protein expression, even when positive controls showed calbindin 1 expression, which can suggest that this biomarker could be specific for dedifferentiated liposarcoma.

Weaknesses of this study include the limited number of enrolled patients, which could somewhat overestimate the activity in patients diagnosed with leiomyosarcoma, and the lack of pharmacokinetics due to budget limitations. The low number of patients could also justify the lack of association between clinical outcomes and the expression of p53. This protein has been described as predictive of response to selinexor²⁹, but in our patients, it lacked a significant association with PFS or OS. Patients with negative expression of p53, which is typically associated with TP53 deletions³⁰, showed a non-significantly lower PFS compared to patients with positive p53 expression, but a larger cohort of patients would be necessary to increase the statistical power. Another limitation of the study is that, at the April 2024 database cutoff, four patients were lost to follow-up; any death among them before this date could have led to a shorter median OS, so the estimated OS for the trial population should be interpreted with caution. An additional limitation is the absence of paired tumor samples from patients treated in the clinical trial, which would enable validation of the effect of selinexor on nuclear IκBα accumulation and survivin protein downregulation, as well as further exploration of the mechanisms underlying the lack of IκBα nuclear localization in certain tumors. The potential application of selinexor as a maintenance therapy has already shown activity in patients with p53 wild-type endometrial carcinoma²⁹, and represents a promising future strategy also in sarcoma, particularly following its combination with gemcitabine during the induction phase. This approach has proven effective in the first-line treatment of advanced leiomyosarcoma, as demonstrated by the use of doxorubicin combined with trabectedin, followed by trabectedin maintenance³¹.

In summary, the combination of Gemcitabine 1200 mg/m² in a 10 mg/m²/min I.V. infusion on days 1 and 8, followed by selinexor P.O. at a 60 mg flat dose on days 1, 8, and 15 in cycles repeated every 21 days warrants investigation in phase II and III trials for certain sarcomas. We have currently initiated a phase II trial to further evaluate activity in selected sarcoma groups.

Method

Preclinical studies

Cell culture

Leiomyosarcoma cell lines included CP0024, IEC005 (established in the Martin-Broto lab from female patients), AA (provided by Dr. Carnero, from IBiS, Seville, Spain, obtained from a female patient), and SK-UT-1 (ATCC HTB-114). Osteosarcoma cell lines were MG-63 (ATCC CRL-1427), U2OS (ATCC HTB-96), and SAOS-2 (ATCC HTB-85). MPNST cell lines comprised ICP060 (established in the Dr. Martin-Broto lab from a male patient), S462, 88-14 (provided by Dr. Romagosa, from Hospital Vall d’Hebron, Barcelona, Spain), and sNF96.2 (ATCC CRL-2884). All cells were cultured in media supplemented with 10% fetal bovine serum (FBS; Gibco) and antibiotics (100 U/mL penicillin, 100 μg/mL streptomycin; Sigma-Aldrich, San Luis, MO, USA). CP0024, IEC005, and ICP060 were maintained in RPMI-1640, while SK-UT-1, MG-63, U2OS, SAOS-2, S462, sNF96.2, and 88–14 were cultured in DMEM. The AA cell line was maintained in Ham’s F-10. Cells were incubated at 37 °C in a humidified atmosphere with 5% CO₂ and passaged as needed using 0.5% trypsin-EDTA (Gibco). Routine screenings for Mycoplasma and fungal contamination were conducted. Cell lines were discarded after 15 passages and replaced with fresh stocks from frozen vials. All procedures were performed in a Class II biological safety cabinet (Cruma, Barcelona, Spain).

IC₅₀ and synergy determinations

IC₅₀ concentrations, representing the drug concentration required to reduce cell viability by 50%, were determined for selinexor and gemcitabine in leiomyosarcoma, osteosarcoma, and MPNST cell lines. Cells were seeded at a density of 2.5 × 10³ cells per well in 96-well plates and treated with selinexor (10⁻⁹ to 10⁻⁵ M) and gemcitabine (10⁻¹¹ to 10⁻⁷ M) alone or in combination for 72 h. DMSO served as the vehicle control. Combination treatments were prepared by pairing the lowest concentration of selinexor with the lowest concentration of gemcitabine, followed by incremental increases in concentration for both compounds.

Cell viability was assessed using the colorimetric MTS assay (Promega), and absorbance was measured at 492 nm with a Heales MB-580 microplate reader (Shenzhen Huisong Technology Development) until control wells reached absorbance values around 1 (approximately 1 h of incubation). Cell viability data were normalized to the control condition and analyzed using the non-linear model fit “log(inhibitor) vs. response—Variable slope (four parameters)” in GraphPad Prism 8.0 software (RRID: SCR_002798, San Diego, CA, USA).

The combination index (CI) at IC₅₀ was calculated according to the Chou-Talalay method³²: CI = (D)₁/(Dx)₁ + (D)₂/(Dx), where (D)₁ and (D)₂ are the concentrations of selinexor and gemcitabine, respectively, in combination that achieve a specified effect. (Dx)₁ and (Dx)₂ are the concentrations of selinexor and gemcitabine, respectively, as single agents that achieve the same effect. CI values were interpreted as follows³³: strong synergism: CI < 0.3; synergism: CI between 0.3 and 0.7; moderate to slight synergism: CI between 0.7 and 0.9; nearly additive: CI between 0.9 and 1.1; slight to moderate antagonism: CI between 1.1 and 1.45 strong antagonism: CI > 1.45.

Apoptosis assay

Leiomyosarcoma, osteosarcoma, and MPNST cell lines were treated with 1 µM selinexor and 1.5 nM gemcitabine for 72 h. DMSO was used as the vehicle control. Cells were stained using a FITC Annexin V/propidium iodide (PI) Apoptosis Detection Kit (Immunostep, Salamanca, Spain) following the manufacturer’s protocol. Flow cytometry was performed on a BD FACSCanto II or BD Accuri C6 Plus system (BD Biosciences), with FITC fluorescence used for detecting annexin V-positive apoptotic cells and PerCP for PI-stained necrotic cells. Data analysis was conducted using BD FACS Diva (RRID: SCR_001456) and Floreada.io software (RRID: SCR_025286). (Supplementary Fig. S5).

Western blot

For the western blot, cells were treated for 48 h with 1 uM selinexor and 1.5 nM gemcitabine. Then, cells were lysed using 1× RIPA buffer (1 M Tris–HCl pH 8, 0.5 M EDTA, Triton™ X-100, 10% sodium deoxycholate, 10% SDS, and 3 M NaCl), supplemented with protease and phosphatase inhibitors (Sigma-Aldrich). The protein samples (20 μg) were separated by SDS-PAGE using a constant current of 90 V and 120 V for stacking and resolving acrylamide gels, respectively. Proteins were transferred to 0.2 μm pore-size Amersham nitrocellulose membranes (Cytiva) at 4 °C for 150 min at a constant current of 200 mA. Membranes were blocked for 1 h with 5% bovine serum albumin (BSA) or non-fat milk (PanReac AppliChem ITW Reagents) in 1× TBS 0.1% Tween-20 (Bio-Rad), according to the primary antibodies recommendations. Next, primary antibodies were applied overnight at 4 °C in BSA or milk, as recommended by the manufacturer. On the following day, after washing with 1× TBS-T, the membranes were incubated with the secondary antibodies: Rabbit Anti-Mouse IgG–Peroxidase (Sigma-Aldrich) or Goat Anti-Rabbit IgG H&L Peroxidase-conjugated (Abcam). Chemiluminescent detection was performed using ECL Prime (Cytiva), and images were acquired with a ChemiDoc Imaging System (Bio-Rad). Blot protein quantification was performed with ImageJ.

Immunofluorescence

For the immunofluorescence (IF), cells were treated for 48 h with 1 uM selinexor and 1.5 nM gemcitabine. After treatment, coverslips were fixed with 3% paraformaldehyde (PFA) and then permeabilized using a 0.2% Triton X-100 (Sigma-Aldrich) 2× PBS solution. The blocking was performed with 1% BSA, 2× PBS for 30 min at room temperature (RT). The IκBα antibody (1:1000; Proteintech) was incubated at 4 °C overnight. A Goat Anti-Mouse IgG1 Cross-Adsorbed Secondary Antibody, Alexa Fluor 546 (Invitrogen; Waltham, MA, USA) was used as a secondary fluorescent antibody at 1:1000 for 1 h at RT. DAPI (Invitrogen) was subsequently added for 15 min to stain nucleic acid. Then, coverslips were carefully washed and mounted onto microscope slides with ProLong™ Gold mounting media (ThermoFisher; Waltham, MA, USA). An LSM900 fluorescence microscope (Zeiss; Oberkochen, Germany) was used for image acquisition, and the data were analyzed with ImageJ.

Phase I part

Patients and methods

Adult patients diagnosed with sarcoma, preferably leiomyosarcoma or osteosarcoma, previously treated with at least one previous line based on anthracyclines, and progressing in the previous 6 months, were enrolled at five hospitals in Spain with expertise in sarcoma care. Before study initiation, approval from the reference Ethics Committee (Ethics Committee from Hospital Clínico San Carlos, approval code C.I. 20/152- EC_M) was obtained, and every patient signed the informed consent to participate in the study. Patients did not receive any compensation for their participation in the trial. Demographic data (including gender determined based on self-report), diagnostic information, previous treatment, and concomitant medication were collected.

Patients started receiving gemcitabine at 1000 mg/m² intravenous (IV) in a 30-min infusion once on days 1 and 8, followed by selinexor 60 mg (dose-level 0), flat dose, once per week orally on days 1, 8, and 15 in 21-day cycles. Dose-level +1 consisted of gemcitabine 1000 mg/m² at 10 mg/m² per minute and selinexor 60 mg. Dose-level +2 consisted of gemcitabine 1200 mg/m² at 10 mg/m² per minute and selinexor 60 mg. Finally, dose-level +3 consisted of gemcitabine 1200 mg/m² at 10 mg/m² per minute and selinexor 80 mg. A −1 dose-level was designed with gemcitabine 800 mg/m² in a 30-min infusion and selinexor 60 mg. Cycles were repeated every 21 days up to progression or intolerance, whichever came first.

Thorough dose modification and delay rules were defined in the protocol (available at Supplementary Information). For day 1, grade 2 neutrophil and/or platelet count decrease resulted in a delay of the gemcitabine administration, while selinexor administration was at the investigator’s discretion. For day 1, a neutrophil and/or platelet count decrease ≥grade 3 required a delay in both drugs. The rules and recommendations for day 8 are as follows: if the neutrophil count was ≥1000/μL and the platelet count ≥100,000/μL, then both drugs were administered at full dose. If the neutropenia was grade 3 or thrombocytopenia was ≤grade 2, then gemcitabine was administered at 80% of the total dose, and selinexor at the investigator’s discretion, while if thrombocytopenia was ≥grade 3 or neutropenia was grade 4, gemcitabine was omitted or delayed and selinexor delayed until grade 2 or less. The rules for day 15 were the same for selinexor administration as for day 8. The protocol provides detailed recommendations for dose adjustments and delays for both drugs in response to hematological and non-hematological toxicities.

This phase I trial had a 3 + 3 design³⁴. Any observed adverse event during the first 21-day period of observation (from day 1 to 21) was considered dose-limiting toxicity (DLT) if the following conditions were met. For hematologic toxicities, grade 4 neutropenia for ≥7 days, grade 4 thrombocytopenia or grade 3 complicated by hemorrhage higher than grade 1, or febrile neutropenia with documented grade ≥ 3 infection or sepsis. For non-hematological toxicities, this was any grade ≥ 3, excluding nausea/vomiting or diarrhea unless uncontrolled by appropriate antiemetic or antidiarrheal therapy. Additionally, the following were not considered as DLT: grade 3 fatigue persisting less than 14 days, grade 3 increased serum creatinine or electrolyte abnormalities deemed not clinically significant and not requiring treatment, and grade 3 transaminitis unless this would delay the treatment administration. If the DLT rate was greater than 0.33 among the patients with the initial dose level, the treatment dose was lowered to dose -1.

The main endpoint of this academic phase I was to determine the RP2D for the combination of gemcitabine and selinexor. Secondary endpoints included establishing the safety profile through assessment of secondary side effects based on the Common Terminology Criteria for Adverse Events (CTCAE) 5.0 assessment, ORR by RECIST 1.0 and by Choi criteria, OS, and quality of life measured with EORTC QLQ-C30 core, version 3, focusing on items 29 and 30. All the pre-specified secondary endpoints were reported in this manuscript. The analyses of these items [Global Health Status (GHS)/ Quality of Life (QoL)] followed two steps. Firstly, the raw data of the average of the two items was estimated according to the equation: RawScore: RS = (I₁ + I₂ + …I_n)/n. Then, the GHS/QoL was transformed to a standardized score (S; 0–100), according to the equation: S = [(RS−1/range] × 100.

This study is registered with ClinicalTrials.gov, identifier NCT04595994, and with the European Clinical Trials Database, EudraCT number 2019-000652-33.

Non-prespecified exploratory outcomes

Other endpoints included post hoc correlative studies searching for potential predictive biomarkers. For translational purposes, the performance of a tumor biopsy was mandatory at baseline, and blood samples were collected at baseline and coincident with every radiological assessment.

Immunohistochemistry

Four-micrometer-thick sections from paraffin blocks were baked for 30 min at 65 °C. Antigen retrieval was performed with a PT Link instrument (Agilent; Santa Clara, CA, USA), using an EDTA buffer (97 °C, 20 min). Sections were immersed in H₂O₂ aqueous solution (Blocking peroxidase reagent, Leica Biosystems; Nussloch, Germany) for 10 min to exhaust endogenous peroxidase activity and then covered with 1% blocking reagent (Roche, Mannheim, Germany) in PBS to block nonspecific binding sites. Sections were then incubated with Calbedin-1 (1:300; D1I4Q; Cell Signaling; Danvers, MA, USA), IκBα (1:300; 66418-1-Ig; Proteintech; Rosemont, IL, USA) or Survivin (1:300; 10508-1-AP; Proteintech; Rosemont, IL, USA) antibodies, O/N, at 4 °C in a humid chamber. Later, HRP polymer conjugated secondary antibodies (Leica Biosystems) were applied for 1 h, at room temperature in a humid chamber, and 3, 3’-diaminobenzidine was applied for 5 min to develop immunoreactivity. Slides were then counterstained with hematoxylin and mounted in DPX. For the anti-p53 (DO-7) primary antibody, immunohistochemistry was performed automatically using the Benchmark ULTRA platform (Roche), according to the manufacturer’s protocol.

Protein expression was assessed as negative, positive + (extension between 5% and 25%), positive ++ (extension between 26% and 50%), and positive +++ (extension over 51%). The intensity of IκBα protein expression was assessed as negative (score 0), weak (score 1), moderate (score 2), strong (score 3), and very strong (score 4).

Statistics

Variables following binomial distributions (i.e., response rate) and qualitative variables were expressed as frequencies and percentages. Quantitative variables were reported as medians and ranges. Time-to-event variables (OS and PFS) were measured from the date of therapy onset and were estimated according to the Kaplan–Meier method. Comparisons between the variables of interest were performed by the log-rank test. Prognostic factors related to response were assessed using binary logistic regression. All p-values reported were 2-sided, and statistical significance was defined at p < 0.05. Statistical comparisons in correlative and preclinical studies were performed using two-tailed t-tests. The analyses were performed with SPSS Statistics version 29.0.2.0 (IBM; Armonk, NY, USA). The investigators were not blinded to allocation during experiments and outcome assessment due to the nature of the trial.

Ethics approval and consent to participate

Ethics approval was obtained from each participating institution before study initiation. Patients signed the informed consent to participate in the study.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.M.B. and D.S.M. contributed to the conception and design of the study. J.M.B., A.G., J.L.M.H., D.S.M., and N.H. analyzed the data. All the authors participated in data and material acquisition. J.M.B. and D.S.M. wrote the initial draft of the manuscript. All authors reviewed and approved the submitted version.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The clinical dataset, with anonymized participant data regarding demographics, treatment information and outcomes will be available upon request to corresponding author for related research and following the signature of an specific agreement to this end. The study protocol and the statistical analysis plan will be available in the Supplementary data. In addition, data regarding the preclinical experiments generated in this study are provided in the Supplementary Information and Source Data file. Source data are provided with this paper.

Competing interests

Javier Martin-Broto has received honoraria for consulting or advisory board participation and expert testimony from PharmaMar, Bayer, GSK, Deciphera, Boehringer Ingelheim, Cogent Biosciences, Roche, Tecnofarma and Asofarma; and research funding for clinical studies (institutional) from Deciphera, PharmaMar, Eli Lilly and Company, BMS, Pfizer, Boehringer Ingelheim, Synox, ABBISKO, Biosplice, Lixte, Karyopharm, Rain Therapeutics, INHIBRX, Immunome, Philogen, Cebiotex, PTC Therapeutics, Inc. and SpringWorks therapeutics. Gloria Marquina has received financial support for advisory board participation: Pharmamar, Clovis oncology, GSK/Tesaro, Lilly, Pfizer, Boehringer Ingelheim, Deciphera, EISAI; lectures: Roche, AstraZeneca, GSK/Tesaro, Clovis Oncology, Medicamenta, EISAI, Pharmamar, Bayer, Lilly, IPSEN, MSD, Inhibrx biosciences INC; and congress attendance: Roche, Pharmamar, GSK/Tesaro, MSD, Medicamenta, Pfizer, Angelini, Novartis, Merck, Lilly, EISAI. Andres Redondo reports receiving honoraria and providing advisory/consultancy services for MSD, AstraZeneca, GSK, Pharma&, and AbbVie, travel/accommodation/congress registration from AstraZeneca and GSK, and participating in a speaker’s bureau for MSD, AstraZeneca, GSK, Pharma& and AbbVie outside the submitted work. David S. Moura has received institutional research grants from PharmaMar and Synox outside the submitted work; travel support from PharmaMar, and personal fees from Tecnopharma, outside the submitted work. Nadia Hindi has received grants, personal fees, and non-financial support from PharmaMar, personal fees from Deciphera, and research funding for clinical studies (institutional) Deciphera, PharmaMar, Eli Lilly and Company, BMS, Pfizer, Boehringer Ingelheim, Synox, ABBISKO, Biosplice, Lixte, Karyopharm, Rain Therapeutics, INHIBRX, Immunome, Philogen, Cebiotex, PTC Therapeutics, Inc. and SpringWorks Therapeutics. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68729-1.