The landscape of drug sensitivity and resistance in sarcoma

Authors

Ahmad Al Shihabi

Peyton J Tebon

Huyen Thi Lam Nguyen

Jomjit Chantharasamee

Sara Sartini

Ardalan Davarifar

Alexandra Y Jensen

Miranda Diaz-Infante

Hannah Cox

Alfredo Enrique Gonzalez

Summer Norris

Jantzen Sperry

Jonathan Nakashima

Nasrin Tavanaie

Helena Winata

Sorel Fitz-Gibbon

Takafumi Yamaguchi

Jae H. Jeong

Sarah Dry

Arun Singh

Bartosz Chmielowski

Joseph G. Crompton

Anusha Kalbasi

Fritz C Eilber

Francis Hornicek

Nicholas Bernthal

Scott D Nelson

Paul C Boutros

Noah Federman

Jane Yanagawa

Alice Soragni

Doi

10.1016/j.stem.2024.08.010

PMID: 39305899 · DOI: 10.1016/j.stem.2024.08.010 · Journal: Cell Stem Cell (2024)

TL;DR

Al Shihabi et al. assembled a UCLA biobank of n=194 sarcoma specimens from 126 patients spanning 24 distinct bone and soft-tissue subtypes (Feb 2018–May 2022) and built a high-throughput patient-derived tumor organoid (PDTO) “mini-ring” platform that returns drug-sensitivity profiles within one week of surgery. Of the 124 samples seeded, ~93% generated organoids; 92 specimens were screened against ≥10 of >400 single agents and combinations. Drug responses tracked sarcoma subtype, treatment history, age, lesion type, and disease trajectory; intra-patient heterogeneity was substantial. The authors identified at least one FDA-approved or NCCN-recommended effective regimen for 59% of specimens (57/97) and 58% of patients (43/74), and showed organoid responses to neoadjuvant MAP correlated with post-resection necrosis and progression-free survival in three osteosarcoma cases, motivating the prospective PREMOST trial (NCT06064682) PMID:39305899.

Cohort & data

  • 194 sarcoma specimens from 126 patients treated at UCLA between Feb 2018 and May 2022; 11 biopsies and 183 surgical resections; 46% metastatic, 41% primary, 13% recurrent. 27 patients (21%) contributed multiple samples.
  • Patient demographics: 50% adult (63/126), 35% AYA (45/126), 13% pediatric (18/126); 62% male, 38% female; 71% had prior systemic therapy and 30% prior radiation.
  • 24 distinct bone and soft-tissue sarcoma subtypes, dominated by osteosarcoma (73 samples, 28 patients), chordoma (14 samples, 10 patients), chondrosarcoma (13 samples, 12 patients), leiomyosarcoma (12 samples, 10 patients), and undifferentiated pleomorphic sarcoma (MFH, 11 samples, 10 patients). Other histologies present include Ewing sarcoma, synovial sarcoma, MPNST, rhabdomyosarcoma including ARMS / ERMS / SCSRMS, epithelioid sarcoma, DSRCT, PEComa, myxofibrosarcoma, WDLS, DDCHS, and CIC-rearranged sarcoma (no canonical OncoTree code).
  • Dataset: sarcoma_ucla_2024. RNAseq, DNA panel sequencing, and single-agent screening data deposited at Synapse (PDTOSarcoma); panel-sequencing data also slated for cBioPortal sharing.
  • Assays/methods: a custom 96-well PDTO mini-ring drug screen (>400 compounds; viability via CellTiter-Glo at day 5; Z’ / robust-Z’ ≥0.2 inclusion); DFCI-ONCOPANEL-3 targeted DNA panel at Brigham CAMD; whole-genome sequencing (BWA-MEM2, MuSE, Mutect2, SomaticSniper, Strelka2, Battenberg/ASCAT); bulk RNA-seq (KAPA Hyper Prep, NovaSeq6000, STAR, GRCh38.108); FISH for MDM2 and NTRK1/2/3 (NeoGenomics) PMID:39305899.
  • 124 samples seeded; organoid take rate ~93%; mean 117 drugs tested per sample (range 6–423).

Key findings

  • Among 92 specimens screened against ≥10 compounds, 80.4% (74/92) had at least one significant response (top 5% viability score for ≥1 regimen); 19.6% (18/92) had none, and these were typically tested against smaller drug libraries (median 25 vs 58 drugs/combinations) PMID:39305899.
  • 3.3% (3/92) showed broad/non-specific chemosensitivity (top 5% on ≥25% of compounds tested).
  • Osteosarcoma PDTOs were significantly more sensitive than the pan-sarcoma average to ceralasertib (ATR inhibitor, p=0.00026), topotecan (p=0.028), cabozantinib (p=0.024), and everolimus (p=0.012).
  • Chordoma PDTOs were significantly less sensitive than other sarcomas to everolimus (p=0.022), alvocidib (p=0.00046), apitolisib (p=0.0071), and bortezomib (p=1.8×10⁻⁵). Chordomas were preferentially sensitive to TAK-285 (EGFR/ERBB2 kinase inhibitor, p=0.034), BI-D1870 (RSK inhibitor, p=0.047), and A-769662 (AMPK activator, p=0.0015).
  • Chondrosarcoma PDTOs were sensitive to TAE226 (FAK/Pyk2, p=0.04) and BX-912 (PDK1, p=0.038); RMS PDTOs to pazopanib (p=0.049) and trametinib (p=0.013); UPS (MFH) PDTOs to BMS-754807 (IGF-1R, p=0.039); LMS PDTOs were less sensitive to topotecan (p=0.031, consistent with prior negative trials) but more sensitive to dovitinib (p=0.0068).
  • Tumors with prior systemic therapy were more sensitive to everolimus, cabozantinib, lenvatinib, and cediranib than treatment-naïve specimens (p=0.018, 0.036, 0.011, 0.017); ≥3 prior therapies trended with sensitivity to pazopanib (p=0.0062), BI2536 (p=0.046), and danusertib (p=0.011).
  • Pediatric and AYA sarcoma PDTOs were less resistant than adult PDTOs to cediranib (p=0.032 vs pediatric, p=0.013 vs AYA); adult PDTOs were more sensitive to the JAK inhibitor WHI-P154 vs AYA (p=0.0037). Metastatic PDTOs trended toward enhanced response to degrasyn (p=0.028 vs primary, p=0.048 vs recurrent) and OSI-930 (p=0.024).
  • PDTOs from patients with progressive disease at follow-up were more resistant to dasatinib (p=0.039), sirolimus/rapamycin (p=0.014), and gefitinib (p=0.048). Rapidly progressing tumors had increased sensitivity to thiazovivin (ROCK inhibitor), sorafenib, and ruxolitinib (p=0.041, 0.042, 0.043).
  • Pathway analysis (PubChem + WikiPathways and KEGG) showed PDTOs from the same patient consistently clustered together regardless of histology (e.g., SARC0103, SARC0078, SARC0075, SARC0139, SARC0053). EGFR/ErbB signaling was implicated in chordomas (SARC0046_2/3, SARC0053_a, SARC0049), an osteosarcoma subgroup, and individual LMS and RMS cases.
  • Functional vs genomic concordance: SARC0117 (metastatic undifferentiated spindle cell sarcoma, PIK3CA hotspot mutation reported as “H1074L” in extracted text — corresponds to the canonical H1047L hotspot) was a top responder to alpelisib and broadly sensitive across the PI3K/mTOR class (apitolisib, alpelisib, copanlisib, BGT226, vistusertib). SARC0134 (MPNST) carried the same PIK3CA mutation in the primary but did not respond to alpelisib; follow-up sequencing confirmed loss of PIK3CA mutation in the metastatic lesion. SARC0069_2 (primary osteosarcoma, no PIK3CA mutation) was the top alpelisib responder, illustrating biomarker-negative responders.
  • SARC0133 (RMS) was the top responder to the FGFR inhibitor infigratinib; WGS revealed an FGFR1 gain on chromosome 8, but PDTOs were not sensitive to dovitinib, demonstrating intra-class drug differentiation.
  • SARC0127 (suspected infantile fibrosarcoma) PDTOs were resistant to larotrectinib; given that IFS carries ETV6–NTRK3 fusion in 90% of cases, the negative response correctly predicted the eventual diagnosis of high-grade spindle cell/sclerosing rhabdomyosarcoma (SCSRMS) — delivered within a week vs 18 days for pathology.
  • Intra-patient heterogeneity was prominent: SARC0075 (epithelioid sarcoma with undifferentiated pleomorphic features) had three same-surgery metastases with Pearson correlations 0.64–0.84; SARC0075_2 responded to only 7.7% of regimens vs 23% for SARC0075_1/4. WGS showed shared SNVs in ARID2, CRTC1, and MAP3K1 but extensive divergent CNVs including whole-genome duplications.
  • Actionability: of 203 drug-diagnosis pairs ranked among top-five responders (top-10% rank), 4% were FDA-approved for the same sarcoma type, 31% FDA-approved for other cancers, 1% in trial for the matched indication, 21% in trial for other cancers, 43% pre-clinical or terminated. Only 5% were NCCN-preferred and 3% NCCN-recommended for the matched diagnosis.
  • Aggregate actionable hits: 59% of samples (57/97) and 58% of patients (43/74) had ≥1 FDA-approved or NCCN-recommended top-five regimen; 7 patients had four such options.

Genes & alterations

  • PIK3CA: hotspot mutation (reported as “H1074L”; canonical hotspot H1047L) in SARC0117 (USS) drove broad PI3K/mTOR pathway sensitivity (alpelisib, apitolisib, copanlisib, BGT226, vistusertib). SARC0134 (MPNST) had the mutation in primary but lost it in metastasis, predicting non-response. SARC0069_2 (osteosarcoma) was a biomarker-negative top alpelisib responder.
  • MDM2: amplification confirmed by FISH in WDLS SARC0120 in both parent tumor and PDTO, illustrating that PDTOs preserve key sarcoma drivers.
  • FGFR1: gain on chromosome 8 in SARC0133 (RMS) correlated with selective sensitivity to infigratinib but not dovitinib.
  • NTRK1 / NTRK2 / NTRK3, ETV6: canonical ETV6–NTRK3 fusion expected in IFS; SARC0127’s larotrectinib resistance and FISH-negative ETV6 helped reclassify the tumor as SCSRMS.
  • ARID2, CRTC1, MAP3K1: rare shared SNVs across SARC0075’s four metastatic lesions; CNV landscape (whole-genome duplications, arm-level events) was substantially discordant.
  • EGFR / ERBB2: pathway implicated by KEGG analysis in chordomas (SARC0046_2/3, SARC0053_a, SARC0049) and an osteosarcoma subgroup; chordomas preferentially sensitive to TAK-285 (EGFR/ERBB2 kinase inhibitor).
  • CDK4: cited in the introduction as a prevalent CDK aberration in liposarcoma (LIPO) literature; not directly genotyped in this study but motivates palbociclib testing observed across PDTOs (e.g., SARC0086_3 PEComa).

Clinical implications

Limitations & open questions

  • Authors note PDTO viability data were not normalized for proliferation rate; rapidly proliferating advanced/metastatic tumors may bias toward sensitivity to certain agents.
  • Patient outcome data were collected for research without RECIST imaging; sarcoma response is poorly captured by RECIST due to non-spherical growth, edema, fibrosis, necrosis and intra-tumoral hemorrhage. Time-to-next-treatment (TTNT) was used as a surrogate but n=5.
  • Predictive validation cohort for MAP response was n=3 and the matched-treatment cohort n=5; the prospective PREMOST trial (NCT06064682) is required to confirm utility.
  • Intra-patient and intra-tumoral heterogeneity (e.g., SARC0075 four-lesion CNV divergence; SARC0127 18-day biopsy-vs-resection drift) means single-sample PDTO recommendations may misrepresent disseminated disease.
  • 21 of 194 specimens failed cell-yield thresholds and 10 of 124 seeded samples failed to grow; certain indolent subtypes (UPS, LMS, chordoma) showed minimal ex vivo proliferation, limiting screening throughput.
  • 19.6% (18/92) of screened specimens had no top-5% response, frequently because the panel was small (median 25 drugs) — questioning whether functional precision medicine requires a minimum library size for utility.
  • 43% of identified active compounds are pre-clinical or terminated; only 5% are NCCN-preferred for the matched indication, raising drug-access and approval-pathway concerns.
  • The PIK3CA hotspot is rendered as “H1074L” in the manuscript text; this is almost certainly the canonical H1047L hotspot but the discrepancy is recorded as observed.
  • CIC-rearranged sarcoma lacks an OncoTree code in the canonical ontology and is therefore not represented in the cancer-type frontmatter.

Citations from this paper used in the wiki

  • “We collected a total of n=194 sarcoma specimens from 126 patients treated at UCLA for a sarcoma diagnosis between February 2018 and May 2022.”
  • “The remaining 124 samples (80% of all processed) encompassing 21 different sarcoma diagnoses, were used to generate mini-rings… resulting in an organoid take rate of ~93%.”
  • “Among the 92 specimens screened against 10 or more compounds included in our analysis, the majority (74/92, 80.4%) had at least one significant response, defined as ranking in the top 5% of viability scores for one or more tested regimens.”
  • “We found osteosarcoma organoids to be significantly more sensitive than the pan-sarcoma average for the drugs ceralasertib, topotecan, cabozantanib, and everolimus, with p values of 0.00026, 0.028, 0.024, and 0.012 respectively.”
  • “Chordomas were significantly less sensitive than the pan-sarcoma average to everolimus (p = 0.022), alvocidib (p = 0.00046), apitolisib (p = 0.0071), and bortezomib (p = 0.000018).”
  • “A metastatic undifferentiated spindle cell sarcoma (USS, SARC0117) carried the hotspot H1074L mutation in PIK3CA and its PDTOs were among the strongest responders to alpelisib.”
  • “SARC0133 harbors a FGFR1 gain on chromosome 8 contributing to its response. Yet, there was minimal response to other FGFR-targeting molecules such as dovitinib.”
  • “We found that the most effective drugs included at least one FDA approved or NCCN recommended regimen for 59% (57/97) of sarcoma samples we screened.”
  • “Despite the heterogeneities, we could detect a trend where patients whose organoids responded ex vivo had longer TTNT, and those with resistant organoids had shorter TTNT… we initiated a clinical trial (NCT06064682, PREMOST).”

This page was processed by crosslinker on 2026-05-04.