Chemosensitive Relapse in Small Cell Lung Cancer Proceeds through an EZH2-SLFN11 Axis

Authors

Eric E. Gardner

Benjamin H. Lok

Valentina E. Schneeberger

Patrice Desmeules

Linde A. Miles

Paige K. Arnold

Andy Ni

Inna Khodos

Elisa de Stanchina

Thuyen Nguyen

Julien Sage

John E. Campbell

Scott Ribich

Natasha Rekhtman

Afshin Dowlati

Pierre P. Massion

Charles M. Rudin

John T. Poirier

Doi

10.1016/j.ccell.2017.01.006

PMID: 28196596 · DOI: 10.1016/j.ccell.2017.01.006 · Journal: Cancer Cell (2017)

TL;DR

Gardner et al. modeled acquired chemoresistance to first-line cisplatin/etoposide in small cell lung cancer (SCLC) by serially treating 10 patient-derived xenograft (PDX) models in vivo to derive paired chemonaive/chemoresistant tumors. Whole-exome sequencing showed no recurrent acquired mutations, but RNA-seq identified two mutually exclusive expression programs across 7/10 resistant models: induction of TWIST1 (3/10) and suppression of SLFN11 (4/10). ChIP-seq localized EZH2-mediated H3K27me3 spreading across the SLFN11 gene body in resistant tumors, and the EZH2 inhibitor EPZ011989 (a structural homolog of clinical-stage tazemetostat) re-expressed SLFN11, restored chemosensitivity ex vivo, prevented emergence of resistance when combined with cisplatin/etoposide, and re-sensitized chemoresistant tumors to irinotecan in vivo.

Cohort & data

  • 10 paired SCLC PDX models (chemonaive vs. chemoresistant), majority derived from chemonaive patients; two models had matched normal DNA (MSK-LX40 and MSK-LX95). Models include JHU-LX102, SCRX-Lu149, JHU-LX44 (chemorefractory), and others. PDXs were generated under MSKCC and JHMI IRB-approved protocols.
  • Murine SCLC models: Rb1/Trp53-null (DKO), Rb1/Rbl2/Trp53-null (TKO) cell lines, plus a chemonaive allograft (TKO-A) and its chemoresistant derivative (TKO-AR).
  • SCLC cell lines: NCI-H82, NCI-H446, NCI-H526; CCLE bi-modal SLFN11 expression analysis (cellline_ccle_broad).
  • Clinical tissue microarrays: untreated SCLC (Vanderbilt Medical Center, VMC) and previously-treated SCLC (Case Western Reserve University, CWRU); SLFN11 IHC scored by H-score by a blinded pathologist.
  • Assays: whole-exome sequencing, RNA-seq (paired chemonaive/resistant), ChIP-seq (EZH2, H3K27me3, H3K27Ac in SCRX-Lu149), ChIP-qPCR, SLFN11 IHC, shRNA knockdown of EZH2/SLFN11/TWIST1, doxycycline-inducible expression. Reads aligned to a custom human/mouse hybrid reference (GRCh38 + GRCm38.p3).
  • Data deposit: dbGaP accession phs001249.v1.p1. cBioPortal study ID: sclc_cancercell_gardner_2017.

Key findings

  • Acquired resistance is reproducible across PDX models. 10/10 SCLC PDXs were sensitive to 6–8 weekly cycles of cisplatin/etoposide (C/E); 2/10 had complete responses, 8/10 partial responses (65–95% tumor growth inhibition). Repeated in vivo selection across three mouse generations produced significantly reduced median survival in the chemoresistant state (p=0.0008) and required fewer C/E cycles to fail (p=0.026). The shift was not explained by Ki67 changes or randomization timing.
  • No recurrent acquired mutations. WES of 10 paired models showed expected SCLC alterations (frequent TP53 and RB1); resistance models retained parental driver mutations. Tobacco-associated C>A transversion signature predominated; a minor mutational Signature 3 (HR-deficiency-associated) was acquired in MSK-LX40 and MSK-LX95. Most mutations were shared between paired models, with private mutations in the resistant setting deemed passengers. Copy-number profiles were concordant; no significant focal CNAs or genome-doubling events emerged.
  • Two mutually exclusive expression programs of resistance, present in 7/10 models. Meta-p-value analysis across paired RNA-seq identified SLFN11 as among the most significantly downregulated genes (4/10 models) and TWIST1 plus cancer-testis antigens as upregulated (3/10 models). The two programs were mutually exclusive within models.
  • TWIST1 is a biomarker, not a driver. Doxycycline-inducible WT or K145E DNA-binding-mutant Twist1 did not change etoposide IC50 in murine SCLC cells (TKO-A IC50 ~0.15 µM vs. resistant TKO-AR IC50 ~3.0 µM). shRNA suppression of TWIST1 did not rescue chemosensitivity in murine or TWIST1-high human SCLC lines, nor reverse EMT markers (e.g., E-cadherin loss).
  • SLFN11 is bi-modally expressed in SCLC and tracks prior treatment. SLFN11 expression was lower in cell lines from previously-treated patients (p=0.031), in chemoresistant PDXs vs. paired chemonaive (p=0.003), and (by IHC) in tumors from treated patients. SCLC SLFN11 was higher than in lung adenocarcinoma or squamous carcinoma. SLFN11 H-score >75 predicted response in 10/12 untreated and 6/6 previously-treated patients; SLFN11 was higher in responders vs. non-responders (p=0.0192) and in limited- vs. extensive-stage disease (p=0.0397). Dichotomized H-score (Youden 68.8) did not significantly stratify overall survival (log-rank p=0.884).
  • EZH2 (not DNA methylation) silences SLFN11 in SCLC. Ex vivo treatment of paired PDX cells with the EZH2 inhibitor EPZ011989 (EPZ; a tazemetostat structural homolog) but not 5-azacitidine (azacitidine) re-expressed SLFN11 and restored ex vivo etoposide sensitivity. In NCI-H82, EPZ caused dose-dependent SLFN11 protein re-expression with concomitant H3K27me2/3 loss; SLFN11 stayed elevated through a 10-day washout even as global H3K27me3 returned to baseline. SLFN11 re-expression by EPZ correlated with topotecan re-sensitization across cell lines (Pearson r=0.916). Ectopic SLFN11 was sufficient to sensitize NCI-H82 and NCI-H446 to topotecan; shRNA against SLFN11 reversed EPZ-mediated topotecan sensitization in NCI-H82 and reduced cleaved-PARP induction.
  • ChIP-seq mechanism: H3K27me3 spreads across the SLFN11 gene body in resistance. In SCRX-Lu149 chemoresistant tumors, EZH2 and H3K27me3 were focally concentrated near the SLFN11 TSS and H3K27me3 spread across the gene body; H3K27Ac at the TSS was lost (ChIP-qPCR p<0.0001). EPZ erased gene-body H3K27me3 (confirmed exon-by-exon for exons 2–7) and increased H3K27Ac at and around the TSS, while TSS H3K27me3 remained largely unchanged. Globally, chemoresistant tumors showed increased H3K27me3 and decreased H3K27Ac that were both reversed by EPZ; super-enhancer regions showed global signal changes but no selective remodeling of discrete super-enhancers.
  • DNA damage induces EZH2 activity. In NCI-H446, topotecan exposure progressively increased EZH2 protein and H3K27me3 over 48 hr; EPZ blocked the H3K27me3 increase, and shRNA against EZH2 abolished both the H3K27me3 gain and accompanying H3K27me2 loss. The induction was dose-dependent, not cell-line-specific, and stronger for topotecan than equimolar cisplatin or etoposide. The same EZH2/H3K27me3 induction with H3K27me2 loss was observed in vivo after a single irinotecan dose.
  • EPZ + chemotherapy is potently combinatorial in vivo.
    • In SLFN11-high chemonaive JHU-LX102 and SCRX-Lu149, EPZ + 6 cycles C/E enhanced disease control vs. either alone, without added weight loss; randomizing animals on the C/E arm to add EPZ after 3 cycles induced tumor regression in the C/E+EPZ arm.
    • In chemoresistant JHU-LX102 and SCRX-Lu149, EPZ + irinotecan (IRI) over 6 weekly cycles produced potent combinatorial activity vs. either agent alone. Strong cross-resistance to IRI (but not ionizing radiation) was confirmed in chemoresistant models.
    • SLFN11 was quantitatively suppressed after as few as 3 cycles of C/E (chemonaive setting) and further suppressed after 3 cycles of IRI in the chemoresistant setting; concurrent EPZ rescued SLFN11 to baseline in the chemoresistant setting.
    • EPZ + IRI in chemonaive JHU-LX102 produced complete responses in 5/5 animals over 6 cycles; benefit was not additive when EPZ was started as adjuvant after maximal IRI consolidation. SLFN11-low models did not show the strikingly enhanced EPZ+IRI activity that SLFN11-high models did.
  • EPZ as a single agent has modest efficacy in vivo. In 4 PDX models over 21 days at 250 mg/kg PO bid, EPZ slowed growth in 3/4 but did not produce major tumor regression — consistent with the slow pace of epigenetic remodeling and the protocol-limited duration of treatment.

Genes & alterations

  • SLFN11 — epigenetic silencing in chemoresistant SCLC via H3K27me3 spreading across the gene body; loss confers cross-resistance to DNA damaging agents (cisplatin, etoposide, topotecan, irinotecan) but not ionizing radiation. Necessary and sufficient for chemosensitivity in tested SCLC lines (ectopic expression sensitizes; shRNA reverses EPZ-induced sensitization).
  • EZH2 — induced by cytotoxic chemotherapy (especially topotecan) within 48 hr; deposits H3K27me3 across SLFN11 gene body driving resistance. Pharmacologic inhibition (EPZ011989, GSK126) or shRNA suppression blocks H3K27me3 induction and re-expresses SLFN11. EZH2 expression is higher in SCLC than any TCGA tumor type (Poirier 2015).
  • TWIST1 — upregulated in 3/10 chemoresistant PDX models and accompanied by EMT-like changes (E-cadherin loss). Functional gain/loss experiments show TWIST1 is a biomarker rather than direct driver of acquired resistance in these models.
  • TP53 and RB1 — recurrent baseline alterations consistent with SCLC; preserved through acquired resistance. No evidence of acquired driver mutations under chemotherapy selection.
  • RBL2 — co-deleted with Rb1/Trp53 in murine TKO models used to corroborate findings.
  • SOX2 and MYC — referenced as known SCLC oncogenes contributing to transcriptional addictions / super-enhancer biology in cited prior work.

Clinical implications

  • Biomarker: SLFN11 expression (IHC H-score) is a candidate predictive biomarker for chemotherapy response in SCLC. SLFN11 H-score >75 predicted response in 10/12 untreated and 6/6 previously-treated patients in the VMC/CWRU TMA cohorts. SLFN11 dichotomized H-score did not stratify overall survival in this cohort.
  • Therapeutic strategy: Combining a clinical-stage EZH2 inhibitor (e.g., tazemetostat, structurally related to the tool compound EPZ011989 used here) with first-line cisplatin/etoposide (cisplatin + etoposide) or with second-line irinotecan/topotecan is proposed to prevent emergence of acquired chemoresistance and re-sensitize relapsed disease — particularly in SLFN11-expressing tumors.
  • Single-agent EZH2 inhibition is unlikely to be active in SCLC; the paper argues for combination strategies and recommends combinations be evaluated early in clinical development.
  • Tazemetostat tolerability in 82 NHL patients (cited from ASH Lymphoma 2016) was favorable, with grade ≥3 thrombocytopenia in 11% and grade ≥3 neutropenia in 6%, supporting the feasibility of combination with cytotoxic chemotherapy in SCLC.
  • DNA methylation inhibitors such as 5-azacitidine (azacitidine) did NOT re-express SLFN11 in SCLC PDX cells, in contrast to a prior report (Nogales et al. 2016) — implicating histone (rather than DNA) methylation as the dominant SLFN11 silencing mechanism in this disease.

Limitations & open questions

  • Epigenetic mechanism for resistance is not the only one. 3/10 PDX models did not fit the SLFN11-loss or TWIST1-induction programs; the authors do not characterize the mechanism in those tumors.
  • Models that lacked SLFN11 at baseline did not regain it under EPZ in vivo, suggesting promoter context or additional layers of silencing limit responsiveness. The relative contributions of SLFN11 vs. other EZH2 target genes to EPZ-mediated chemosensitization are not quantified.
  • Clinical correlations rest on archival TMAs with relatively small numbers (e.g., 12 untreated patients with H-score data above the 75 threshold) and did not show an overall survival benefit at the dichotomized cutoff (p=0.884).
  • Single-agent EZH2 inhibitor activity in vivo was limited — partly attributed by the authors to treatment duration constraints (weight-loss-limited) versus the timescale of epigenome remodeling (weeks to months).
  • No prospective clinical trial data are presented; clinical translation is proposed but not tested. The role of SLFN11 IHC as a companion biomarker requires prospective evaluation.
  • EZH2 inhibition has pleiotropic effects. The authors acknowledge that factors beyond SLFN11 likely contribute to EPZ-induced re-sensitization, leaving the full set of EZH2-modulated mediators to be defined.
  • Cross-resistance pattern is selective (IRI cross-resistant; ionizing radiation not), suggesting the SLFN11 axis is specific to certain DNA damage modalities.

Citations from this paper used in the wiki

  • “Multiple chemoresistant models demonstrated suppression of SLFN11, a factor implicated in DNA damage repair deficiency. In vivo silencing of SLFN11 was associated with marked deposition of H3K27me3, a histone modification placed by EZH2, within the gene body of SLFN11…” (SUMMARY)
  • “Comparing the chemonaive to the chemoresistant state across all models, we observed a significant (p=0.0008) difference in the median survival time…” (Results, modeling acquired resistance)
  • “We observed the expected pattern of genetic alterations consistent with SCLC, including frequent alterations in TP53 and RB1… in each case the key genetic alterations identified in the chemonaive model were maintained through acquisition of chemoresistance.” (Acquired chemoresistance is not associated with emergence of recurrent mutations)
  • “Cell lines generated from treated patients have lower levels of SLFN11 expression relative to lines generated from untreated patients (p=0.031)… chemonaive to chemoresistant PDX models (p=0.003).” (SLFN11 expression is decreased…)
  • “ChIP-seq data demonstrate focally concentrated EZH2 and H3K27me3 in the immediate vicinity of the transcription start site (TSS) in vehicle-treated tumors, with spreading of H3K27me3 across the gene body in the context of acquired chemoresistance.” (EZH2 silences SLFN11)
  • “Addition of EPZ to 6 cycles of C/E strongly enhanced disease control in the chemonaive setting in both JHU-LX102 and SCRX-Lu149 relative to either EPZ or C/E alone…” (Pharmacologic EZH2 inhibition prevents…)
  • “All raw data resulting from RNA-seq, whole exome sequencing, and targeted sequencing is available from the database of Genotypes and Phenotypes (dbGaP) under accession number phs001249.v1.p1.” (Accession number)

This page was processed by crosslinker on 2026-05-14.