Rhabdoid Tumors Are Sensitive to the Protein-Translation Inhibitor Homoharringtonine

Rhabdoid tumors are devastating pediatric cancers with poor prognoses, despite the standard use of intensive multimodal therapy (1–3). Primary rhabdoid tumors can arise in the brain [referred to as atypical teratoid/rhabdoid tumors (ATRT)], the kidney [referred to as malignant rhabdoid tumors (MRT)], or various soft tissues (4). The incidence and prognoses of rhabdoid tumors vary by tumor location: ATRTs represent about 6% of all pediatric central nervous system tumors, and the probability of 5-year survival of children with ATRT is 30% to 45% (2, 3, 5); that for children with MRT is 20% to 50%, depending on the stage of disease at diagnosis (1).

Although particularly aggressive pediatric cancers, rhabdoid tumors exhibit one of the lowest mutation rates ever characterized (6–9). The sole recurrent mutation detected in nearly all patient samples is biallelic inactivation of the SMARCB1 gene, and rhabdoid tumors are defined, in part, by the loss of expression of SMARCB1 protein (also known as SNF5, INI1, or BAF47; refs. 10–12). SMARCB1 is a subunit of the SWI/SNF family chromatin-remodeling complexes (4). Sequencing studies have shown that genes encoding at least eight other SWI/SNF subunits are found to be recurrently mutated in more than 20% of all cancers, thereby underscoring the importance of this chromatin regulatory complex in cancer (13). Furthermore, inactivation of Smarcb1 in mice resulted in all animals rapidly developing cancer, thus establishing the gene as a potent tumor suppressor (14, 15). Because rhabdoid tumors lack recurrent targetable mutations and their prognoses and survival are poor, improved therapies are greatly needed.

Homoharringtonine (HHT) is a natural plant alkaloid purified from Cephalotaxus species of Chinese plum yew (16, 17). HHT has long been used in Chinese medicine to treat a variety of blood cancers, most notably acute myeloid leukemia (AML), where it yields an approximately 60% complete response (CR) rate (18). Although HHT has been studied clinically in the United States for decades, the discovery of a semisynthetic method to create a highly purified and consistent version of the drug, termed omacetaxine mepesuccinate (OM), led to its approval by the FDA in 2012 for patients with chronic myelogenous leukemia (CML) resistant to multiple tyrosine kinase inhibitors (17, 19). In addition, a recent phase II trial of omacetaxine mepesuccinate in patients with late-stage myelodysplastic syndrome showed a 33% overall response rate and increased survival (20). Finally, HHT is a safe and effective substitute for anthracyclines in young children with AML (21).

HHT inhibits protein translation by interacting with the A-site of the large ribosomal subunit (22–24). Although translation inhibition most likely leads to widespread changes in protein abundance in cancer cells, the decreased stabilities of a number of specific short-lived proteins have been reported to underlie the sensitivity to HHT, including Mcl-1 (25, 26), c-Myc (17, 27), and BCR–ABL in CML (28). Numerous protein changes most likely contribute to cell-cycle arrest and cell death following HHT treatment; however, the precise mechanisms of HHT-mediated loss of cell viability remain unknown.

Interactions among members of the Bcl-2 family of proteins play key roles in regulating cell death by influencing mitochondrial membrane integrity (29). The Bcl-2 family includes both proapoptotic (Bax, Bad, Noxa, etc.) and antiapoptotic (Bcl-2, Bcl-XL, Mcl-1, etc.) proteins, the balance of which determines whether a cell undergoes apoptosis (29). One particular pair of antiapoptotic proteins, the short-lived Mcl-1 and the more stable Bcl-XL, exhibits a consistent synthetic-lethal relationship in multiple systems (30–32). In general, cells tolerate the loss of one antiapoptotic member, but depletion of both proteins induces apoptosis.

Here, we report the results of our high-throughput screening of small molecules in rhabdoid tumor cell lines to identify compounds and drugs to which these tumor cells show high sensitivity. We also investigated the mechanism(s) underlying this sensitivity to find new therapeutic approaches to this devastating pediatric disease.

Cell lines were obtained from the ATCC (A204, G401, G402), C. David James (BT12, BT16), Children's Oncology Group (CHLA266), Franck Bourdeaut (KD, MON), Yasumichi Kuwahara (KPMRTRY), Cancer Cell Line Encyclopedia (CCLE)/Broad Institute Biological Samples Platform (KYM1, CMLT1, K562, KCL22, KPNYN, LAMA84, LUDLU1, NCIH747, NCIH1755, OVKATE, RMG1, SNU601, SW1783), Bernard Weissman (NCIH2004RT, TM87), and Timothy Triche (TTC549, TTC642, TTC709, TTC1240). Growth conditions are presented in Supplementary Table S1. All cell lines were SNP authenticated before screening and tested for Mycoplasma prior to using new frozen stocks, screening, and in vivo experiments. A luciferase reporter gene was inserted into A204, BT16, and TTC642 cells for in vivo studies and cultured as above.

Screening of 481 small molecules across 840 cancer cell lines has been described previously (33–35). The same procedure was used to screen 47 additional cell lines (5 rhabdoid tumors). Each small molecule was tested over a 16-point concentration range (two-fold dilution) in technical duplicates. Cells were plated at 500 cells/well in 1,536-well plates. Small molecules were added 24 hours later, and viability was assessed using CellTiter-Glo (Promega) after 72 hours.

A cohort of rhabdoid tumor, CML, and control cell lines representing multiple lineages were plated and screened with protein-translation inhibitors in a 384-well format as described previously (33, 34). Cells were plated at 1,000 cells/well and treated 24 hours later. Small molecules (Supplementary Table S2) were added (1:300 dilution) using a CyBi-Well Vario (Endress+Hauser) pin-transfer machine or an HP D300e Digital Dispenser (Hewlett-Packard), and viability was measured using CellTiter-Glo 72 hours later.

We analyzed two independent Project Achilles CRISPR–Cas9 genome-scale loss-of-function screens: the GeCKO screen (19Q1) with 43 cancer cell lines (8 rhabdoid tumors; ref. 36) and the Avana screen (19Q1) with 676 cancer cell lines (11 rhabdoid tumor). Preferential genetic dependencies (i.e., genes required for survival) between rhabdoid tumors and other cancer types were calculated as described previously (37). We then performed gene set enrichment analysis (GSEA; ref. 38) on the list of preferential rhabdoid tumor genetic dependencies using annotated pathways from the Kyoto Encyclopedia of Genes and Genomes (39).

Pearson correlation coefficients were calculated in R by comparing all HHT AUC values from Cancer Therapeutics Response Portal version 2 (CTRPv2) to transcripts per million (TPM) expression values (CCLE) or to the Avana dependency scores for all genes. Correlation coefficients were transformed using Fisher z-transformation to approximate a normal distribution by normalization for cell line number (33).

For time course, cells were plated at 2 million cells/plate in 10-cm plates. After 24 hours, cells were treated with HHT (100 nmol/L; Tocris, catalog no. 1416) and harvested at the indicated times. For concentration response, cells were plated at 300,000 cells/well in 6-well plates, allowed to adhere overnight, and treated with the indicated HHT concentrations for 4 hours. Immunoblots were performed as described previously (37) using antibodies indicated in Supplementary Table S3.

Primary rhabdoid tumor samples were previously profiled using RNA-sequencing through the NCI Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative (http://ocg.cancer.gov/programs/target; ref. 40). Additional TARGET TPM expression data for other primary solid tumors were downloaded from UCSC Xena (http://xena.ucsc.edu, TARGET Pan-Cancer (PANCAN) dataset, and TCGA PANCAN dataset). TARGET rhabdoid tumor and paired normal kidney samples (dbGaP phs000218.v19.p7) were aligned or realigned with STAR, and transcript quantification was performed with RSEM to generate TPM expression per gene. Expression values were log₂-transformed and floored at –3. For the PANCAN datasets, samples annotated as normal, cell line, xenograft, metastatic, or recurrent were excluded. BCL2L1 expression was compared between rhabdoid tumor, normal kidney, and all other primary samples in the dataset.

SMARCB1 was expressed in rhabdoid tumor cells from pLX401-SMARCB1 (Addgene, catalog no. 111182) as described previously (37). SMARCB1 knockout (KO) was performed in 293T cells (ATCC CRL-1573) using the Santa Cruz Ini1 CRISPR–Cas9 KO plasmids (sc-401485). Cells (50% confluent in 60-mm dishes) were transfected with 3 μg total plasmid mixture using Lipofectamine 3000 (Thermo Fisher Scientific, catalog no. L3000015), and sorted 48 hours later into 96-well plates. At least 30 individual clones were expanded and analyzed for deletion of SMARCB1 by immunoblot and Sanger sequencing. Two individual clones were further characterized for SWI/SNF complex composition by immunoblot, glycerol-sedimentation analysis, and immunoprecipitation/mass spectroscopy analysis (41).

Cells were plated at 100,000 cells/well of a 6-well plate in 2 mL culture media, allowed to adhere overnight, and treated with either DMSO (0.1% final) or HHT (100 nmol/L final). Cells were harvested and counted as described previously (37).

BCL2L1 and LacZ ORFs were ordered from the Broad Genetic Perturbation Platform (clone IDs: TRCN0000489920 and ccsbBroad304_99994). The ccsbBroad304_99994 was subcloned into the pLX307 vector (i.e., the same backbone as TRCN0000489920) using Gateway cloning (Life Technologies). After sequence verification, ORFs were lentivirally transduced into G402 rhabdoid tumor cells. Cells were selected in puromycin for 4 days and then plated for concentration-response curves, HHT treatments followed by immunoblots, or IncuCyte imaging during HHT treatment.

For live-cell imaging, 1,000 cells/well were plated into black clear-bottom plates (Corning, catalog no. 3764). Small molecules were added 24 hours later using an HP D300e Digital Dispenser and immediately placed on an IncuCyte S3 for 10× phase imaging every 3 hours for 6 days. The phase % confluence metric (analysis metrics: segmentation adjustment 0.4, minimum area 400 μm²) was used to estimate per-well cell doublings relative to the initial scan.

The pilot TTC642 in vivo studies were performed according to the approved protocols of the Dana-Farber Cancer Institute's Institutional Animal Care and Use Committee (IACUC) using 6- to 8-week-old female NCr nude mice (Taconic, #NCRNU-F). For the tolerability study, three mice/arm were injected once daily with HHT dissolved in normal saline at either 2 or 3 mg/kg intraperitoneally or 3 mg/kg subcutaneously. The body weight of each mouse was recorded daily to monitor drug tolerability. For the efficacy study, two million TTC642 cells in 100 μL [1:1 PBS:Matrigel (Corning, catalog no. 354234)] were implanted subcutaneously into right flanks. When tumors reached approximately 150–250 mm³, mice were randomized [stratified randomization; Study Director 3.1 (Studylog)] to vehicle or HHT treatment (n = 8 mice per arm). HHT (2 mg/kg) or saline vehicle was injected intraperitoneally once daily for 28 days. Tumor volume (volume = length*width²/2; measured with calipers) and body weight were measured twice per week. The primary endpoint was total tumor volume larger than 2,000 mm³.

The follow-up studies were performed upon approval from the St. Jude Children's Research Hospital's IACUC using athymic nude immunodeficient mice (Charles River Laboratories, catalog no. 553). Five million luciferase-labeled A204, BT16, and TTC642 cells or unlabeled SJMRT031055_X1 patient-derived xenograft (PDX) cells [obtained from the Childhood Solid Tumor Network (http://www.stjude.org/CSTN/), dissociated, passaged, and cryopreserved as a single-cell suspension as described previously (42)] were implanted subcutaneously in right flanks with 50% Matrigel. Mice injected with luciferase-labeled cells were screened weekly by Xenogen imaging and enrolled after achieving a target bioluminescence signal of 10⁷ photons/second/cm² or a palpable tumor. Mice injected with SJMRT031055_X1 cells were observed weekly until a palpable tumor was detected. Chemotherapy started the following week.

Clinical grade vincristine (VCR; Hospira 1 mg/mL) and doxorubicin (DOXO; Actavis 2 mg/mL) were sourced from the St. Jude pharmacy. Mice were randomized into the following treatment groups: VCR (0.5 mg/kg i.p. once on days 1, 8, 15, and 22), DOXO + VCR (DOXO 6 mg/kg i.p. once on day 1, VCR as above), HHT 0.75 mg/kg (0.75 mg/kg s.c. twice daily on days 1–14), HHT 0.75 mg/kg + VCR (both as above), HHT 2 mg/kg (2 mg/kg i.p. once daily on days 1–28), HHT 2 mg/kg +VCR (both as above), and placebo (no chemotherapy).

Mice received two 28-day courses of chemotherapy. All mice were monitored daily including twice-weekly caliper measurements. For mice injected with luciferase-labeled cell lines, bioluminescence was monitored weekly. Mice with a signal of less than 10⁵ photons/second/cm² (similar to background) were classified as CR and 10⁵–10⁷ photons/second/cm² as partial response. For SJMRT031055_X1, CR was defined as an undetectable tumor by calipers at endpoint. Endpoints included completion of the two courses and tumor burden at any time greater than 20% of body weight (tumor typically 3–5g). Mice that were found dead or moribund for reason other than tumor burden (e.g., unexpected death or drug toxicity) while not at endpoint were censored from the study but are reported in Supplementary Table S4.

Mice bearing luciferase-labeled cells were injected intraperitoneally with Firefly D-Luciferin (Caliper Life Sciences; 3 mg/mouse). Bioluminescent images were taken 5 minutes later using the IVIS 200 imaging system. Anesthesia was administered throughout image acquisition (isoflurane 1.5% in O₂ at 2 L/minute). The Living Image 4.3 software (Caliper Life Sciences) was used to generate a standard region of interest (ROI) encompassing the largest tumor at maximal bioluminescence signal. The identical ROI was used to determine the average radiance (photons/second/cm²/sr) for all xenografts.

We sought to identify small molecules to which rhabdoid tumors are preferentially sensitive by screening eight rhabdoid tumor cell lines as part of the Cancer Therapeutics Response Portal version 2 [CTRPv2 (33–35)]. In this unbiased screening platform, up to 887 cancer cell lines were treated with 16 concentrations (serial two-fold dilutions) of 481 small molecules, and AUC values were calculated for each cell line–molecule pair. When we compared the AUC values of rhabdoid tumor cell lines with those of all other cancer cell lines, rhabdoid tumor cell lines were particularly sensitive to HHT (Fig. 1A). All eight rhabdoid tumor cell lines scored among the top 20% most sensitive cancer cell lines (Fig. 1B and C). Although individual rhabdoid tumor cell lines were highly sensitive to other small molecules (Fig. 1A and B; refs. 35, 37), HHT was the most consistently effective drug, showing high activity against all rhabdoid tumor cell lines tested.

Because HHT/OM has been approved by the FDA for treatment of CML, we compared the sensitivity of rhabdoid tumor cells with that of CML cells. We performed HHT concentration-response curves in an expanded set of 16 rhabdoid tumor cell lines, four CML cell lines, and eight control cancer cell lines that do not contain mutations in any SWI/SNF subunits (Fig. 1D; Supplementary Fig. S1A and S1B). All additional rhabdoid tumor cell lines were either as sensitive or more sensitive to HHT than were the rhabdoid tumor cell lines included in the initial screening dataset (Fig. 1C). Sensitivity did not differ across the rhabdoid tumor cell lines derived from three different anatomic locations (Supplementary Fig. S1C). These observations suggest that ATRTs, MRTs, and soft-tissue rhabdoid tumors are broadly sensitive to HHT.

Rhabdoid tumor (P < 0.0001) and CML (P = 0.0005) cell lines were significantly more sensitive to HHT than were the control cancer cell lines, but we found no difference (P = 0.9551) in sensitivity between rhabdoid tumor and CML cell lines (Fig. 1E). This result was surprising, given that hematopoietic and lymphoid cell lines (such as CML) were significantly (P < 0.0001) more sensitive, on average, to all small molecules in the CTRPv2 than were other histologic types of cells (such as rhabdoid tumors; Supplementary Fig. S1D). Together, these observations demonstrate that rhabdoid tumor cell lines are highly and consistently sensitive to HHT at levels comparable with CML cell lines, which model the FDA-approved indication for the drug.

Next, we investigated the mechanism underlying the sensitivity of rhabdoid tumor cell lines to HHT. Given the established role of HHT in inhibiting ribosome function, we first evaluated whether rhabdoid tumors are also preferentially sensitive to genetic loss of ribosomal subunits. We recently reported the results of genome-scale CRISPR–Cas9 screens in various rhabdoid tumor cell lines (35, 37, 41). Here, we used GSEA to probe these two independent sets of CRISPR–Cas9 dependency screens, one with 43 cancer cell lines (8 rhabdoid tumor lines) (36) and the other with 676 cancer cell lines (11 rhabdoid tumor lines). In both datasets, ribosomal subunits scored among the top two most preferentially represented pathways (out of 186) in rhabdoid tumor cell lines (Fig. 2A–D). These analyses suggest that rhabdoid tumor cell lines are more sensitive to genetic loss of the ribosome than are other cancer cell lines, which is concordant with rhabdoid tumor cell lines exhibiting greater sensitivity to HHT.

To determine whether this dependency on ribosomal subunits is recapitulated by treatment with additional translation inhibitors, we screened the expanded cohort of rhabdoid tumor and control cell lines introduced above with seven other inhibitors that were not part of the full CTRPv2 screen. Although rhabdoid tumor lines were significantly more sensitive than the control lines to four (anisomycin: P = 0.0026; bruceantin: P < 0.0001; emetine: P = 0.0044; verrucarin: P < 0.0001) of the seven small molecules (Supplementary Fig. S2A–S2G), HHT was the protein-translation inhibitor to which the rhabdoid tumor cell lines were most preferentially sensitive compared with control cell lines (Supplementary Fig. S2H). In addition, clustering analysis of the various protein-translation inhibitors revealed that rhabdoid tumor cell line–specific sensitivity to HHT was most similar to sensitivity to bruceantin (Supplementary Fig. S2I), which also binds the A-site of the ribosome (24). Overall, these findings suggest that rhabdoid tumor lines are generally sensitive to the loss of protein translation, but HHT is particularly effective across all rhabdoid tumor cell lines studied.

Although it is well established that HHT inhibits the ribosome, the specific proteins affected by HHT and, in turn, the effects of those altered proteins on cell viability remain undefined. Examining gene expression and genetic dependency data for hundreds of cancer cell lines, we asked whether any predictive features suggested a plausible mechanism of action underlying HHT sensitivity. Across all cancer cell lines screened and of the nearly 17,000 gene-expression features studied, the greatest predictor of HHT sensitivity was low expression of BCL2L1 (Fig. 3A; Supplementary Fig. S3A). With respect to genetic dependencies, of more than 18,000 genes assessed, the best predictor of HHT sensitivity was vulnerability to the loss of MCL1 (Fig. 3B; Supplementary Fig. S3B). Notably, BCL2L1 and MCL1 have a well-established synthetic-lethal relationship in multiple contexts (30–32). In addition, HHT causes rapid depletion of Mcl-1 (25, 26). These observations suggest that low expression of the more stable Bcl-XL protein sensitizes cells to HHT-mediated depletion of the shorter-lived Mcl-1 protein.

Next, we questioned whether these global correlations across all cancer cell lines explained the observed sensitivity of rhabdoid tumor cells to HHT. Rhabdoid tumor cell lines expressed significantly (P < 0.0001) less BCL2L1 RNA than the average expression of BCL2L1 of all cell lines in the CCLE (43, 44; Fig. 3C). Rhabdoid tumor cell lines were significantly (GeCKO: P = 0.0490; Avana: P = 0.0431) more dependent on MCL1 for cell viability than were the other cancer cell lines (Supplementary Fig. S3C and S3D). In contrast, we did not find any difference (P = 0.6817) in BCL2L1 expression among rhabdoid tumor cell lines derived from different anatomic locations (Supplementary Fig. S3E), suggesting that low expression of BCL2L1 is a common feature of all types of rhabdoid tumor. When we assessed the levels of Bcl-XL in the expanded set of rhabdoid tumor cell lines and control cell lines, we found Bcl-XL at low levels in 14 of 16 rhabdoid tumor cell lines and at moderate levels in the remaining two (Fig. 3D). Bcl-XL was expressed at high levels in some CML cell lines, suggesting that additional mechanisms, such as HHT-mediated loss of the short-lived BCR–ABL fusion protein (28), most likely contribute to CML sensitivity to HHT (Fig. 3D). Together, these results show that rhabdoid tumor cell lines express low levels of BCL2L1, which was the best indicator of sensitivity to HHT across the CTRPv2 dataset.

To investigate whether these observations in cell lines extended to primary rhabdoid tumors, we analyzed RNA-sequencing data of primary MRTs (40) paired with normal kidney samples and other primary solid tumors. BCL2L1 expression was significantly (P = 0.0002) lower in MRT samples than in matched normal kidney cells (Fig. 3E), a finding confirmed (P < 0.0001) when all MRT samples were compared with all normal kidney samples in the dataset (Fig. 3F). In addition, when we compared primary MRT samples with more than 7,500 other primary tumor samples from 35 different lineages, we discovered that MRTs expressed substantially less BCL2L1 than all but one other solid tumor type (Fig. 3G). Interestingly, the three lineages that expressed the lowest level of BCL2L1 were all pediatric kidney cancers. These observations emphasize that the expression of BCL2L1 is particularly low in primary MRTs and preclinical rhabdoid tumor models.

The inverse correlations between HHT sensitivity and BCL2L1 expression prompted us to assess the consequence of HHT treatment on the levels of Bcl-2 family proteins. In rhabdoid tumor (G402) and CML (K562) cell lines, we observed substantial depletion of Mcl-1, which directly preceded the cleavage of caspase-3 only in rhabdoid tumor cells that expressed low levels of Bcl-XL (Fig. 4A). In contrast, Bcl-XL expression was not substantially affected by HHT treatment (Fig. 4A). Mcl-1 was also depleted to a lesser extent upon HHT treatment in a control (NCIH1755) cell line with a medium-level sensitivity to HHT, also without accompanying caspase-3 cleavage (Fig. 4A). Depletion of Mcl-1 and cleavage of caspase-3 occurred in a concentration-dependent manner in rhabdoid tumor cells (Fig. 4B). We confirmed this finding of preferential cytotoxicity by conducting cell-count experiments in additional rhabdoid tumor cell lines (Supplementary Fig. S4A–S4F). These observations suggest that HHT depletes Mcl-1 in many cell types, but a greater apoptotic response occurs in cells that express lower levels of Bcl-XL at baseline. This may explain why rhabdoid tumors, which express especially low baseline levels of Bcl-XL, are particularly sensitive to HHT.

To confirm that the level of BCL2L1 expression determines HHT sensitivity, we overexpressed BCL2L1 in a rhabdoid tumor cell line. While increased expression of BCL2L1 only modestly shifted the IC₅₀ value toward resistance, it substantially reduced the maximal cytotoxicity caused by HHT treatment across multiple logs of concentrations (Fig. 5A and B). In addition, IncuCyte tracking of G402 cells showed that although LacZ-expressing G402 cells exhibited cytotoxicity at about 62 nmol/L HHT (Fig. 5C), similar to the reported C_max of 46 nmol/L from a single dose of HHT in humans, those overexpressing BCL2L1 were resistant to HHT-mediated cytotoxicity at concentrations as high as 1 μmol/L (Fig. 5D). Expression of BCL2L1 also substantially inhibited the cleavage of caspase-3 without affecting the loss of Mcl-1 (Fig. 5E), and did so in a highly concentration-dependent manner (Supplementary Fig. S5A). Overall, these results confirm that the expression of BCL2L1 influences the sensitivity of cancer cell lines to HHT.

Low BCL2L1 expression in rhabdoid tumor cells suggested that either the cell(s) of origin of rhabdoid tumors expresses abnormally low levels of BCL2L1 compared with neighboring normal cells, or SMARCB1 mutation leads to transcriptional repression of BCL2L1. To differentiate between these possibilities, we reexpressed SMARCB1 in three rhabdoid tumor cell lines. We found no substantial differences in Bcl-XL protein expression when SMARCB1 was reexpressed (Supplementary Fig. S5B). Furthermore, while reexpression of SMARCB1 causes resistance to multiple other drugs (37, 45), we found no changes in rhabdoid tumor sensitivity to HHT (Supplementary Fig. S5C–S5E). In addition, CRISPR–Cas9-mediated deletion of SMARCB1 in 293T cells did not increase sensitivity to HHT (Supplementary Fig. S5F and S5G). Overall, these results suggest that rather than SMARCB1 mutation directly conferring sensitivity to HHT, rhabdoid tumor progenitor cells transformed by the loss of SMARCB1 express low BCL2L1, which makes them inherently sensitive to HHT.

To confirm these cell culture–based findings in vivo, we performed preclinical testing using rhabdoid tumor xenografts. Given the wide range of HHT doses used in vivo in the literature, we first performed a pilot tolerability study. A dose of 2 mg/kg HHT injected intraperitoneally was well tolerated, but 3 mg/kg HHT caused substantial weight loss (Supplementary Fig. S6A). We began with TTC642 cells because they had a highly penetrant tumor-formation phenotype in mice (37). We then treated mice bearing rhabdoid tumor xenografts with 2 mg/kg HHT injected intraperitoneally once daily for 28 days; this blocked xenograft growth during the treatment period and significantly increased the survival of mice bearing xenograft tumors (Fig. 6A; Supplementary Fig. S6B; P < 0.0001). The tumors resumed growth once HHT was withdrawn.

We next pursued a preclinical study using a previously described design (46) using four independent models: three luciferase-labeled cell-line xenografts (A204, BT16, and TTC642) and one PDX (SJMRT031055_X1). We designed the study with seven arms, each treated with two 4-week cycles for a total of 8 weeks (Fig. 6B). We used three separate control arms: untreated; standard-of-care therapy vincristine (VCR); and standard-of-care combination therapy doxorubicin (DOXO) + VCR. The four experimental arms contained HHT: HHT at 0.75 mg/kg twice daily for 14 days followed by 14 days off (the calculated mouse equivalent of the FDA-approved schedule for CML); HHT at 0.75 mg/kg BID + VCR; HHT at 2 mg/kg daily for 28 days (equivalent to the pilot study); and HHT at 2 mg/kg QD + VCR. Five to seven mice bearing tumors of each xenograft type were randomized to each treatment arm.

Overall, we observed very similar findings between the two different HHT treatment schedules (Supplementary Fig. S6C–S6F), and so focused on the FDA-approved clinical schedule, but the full dataset is included in Supplementary Table S4. We also observed a higher degree of toxicity in the combined DOXO + VCR arm in this study (Supplementary Table S4).

In all four xenograft models, we observed significant increases in mouse survival upon treatment with HHT alone (A204: P = 0.0155; BT16: P = 0.0387; TTC642: P = 0.0009) or HHT in combination with VCR (A204: P = 0.0004; TTC642: P = 0.0020; SJMRT031055_X1: P = 0.0047; Fig. 6C–E; Supplementary Fig. S6G). CR were observed in some mice bearing all four xenograft types treated with HHT-based regimens (Fig. 6F–I; Supplementary Fig. S6H). Across all models, 16 CR (19.75% of 81) were observed in the four HHT-based treatment arms (Supplementary Table S4). In comparison, 3 CR (10.34% of 29) were observed in standard-of-care arms (Supplementary Table S4). Overall, HHT, alone and in combination, demonstrated efficacy in each of the four models.

We tested 481 small molecules across a large number of rhabdoid tumor cell lines and discovered that rhabdoid tumor cell lines were highly and consistently sensitive to HHT. Because HHT/OM is an FDA-approved drug, these observations may have immediate therapeutic relevance. Rhabdoid tumor cell lines were as sensitive to HHT as were CML cell lines. A recent trial showed the safety and efficacy of HHT in children younger than 2 years with AML (21). This finding is particularly relevant, given the similar young age of patients with rhabdoid tumors.

We found that the baseline expression of BCL2L1 predicted the sensitivity of all cancer cell lines to HHT. This finding is consistent with an earlier report that higher expression of Bcl-XL predicts resistance to apoptosis in some leukemia cell lines (47). These findings may help identify additional types or subtypes of cancer beyond rhabdoid tumors that will most likely respond to HHT. In the case of rhabdoid tumors, all cell lines that we tested expressed low BCL2L1, and primary MRTs expressed among the lowest levels of BCL2L1 of all primary solid tumors tested. This low expression of BCL2L1 makes rhabdoid tumor cells more sensitive to HHT-mediated depletion of Mcl-1, though we cannot eliminate the possibility that the depletion of additional proteins contributes to rhabdoid tumor sensitivity to the drug. Consistent with the reduced Mcl-1 expression in rhabdoid tumor cells upon HHT treatment reported here, an earlier report showed that rhabdoid tumor cell lines are sensitive to the loss of Mcl-1 (48). Together, our experiments provide further evidence that HHT sensitivity depends on the expression of Bcl-2 family proteins.

The experiments reported here suggest that low expression of BCL2L1 is a property of rhabdoid tumor cells independent of mutation in SMARCB1, although it remains possible that mutation in SMARCB1 influences BCL2L1 expression on a different timescale or in contexts not investigated here. One possible mechanism is that the undefined cell(s) of origin that are susceptible to transformation by SMARCB1 loss intrinsically express low levels of BCL2L1. We also observed that the three cancer types that expressed the lowest levels of BCL2L1 in primary samples were all pediatric kidney cancers. Together, these findings may point to a unique regulation of BCL2L1 in a subset of cells in the developing kidney that is reflected in tumors that arise from those cells. However, we also found that rhabdoid tumor cell lines and xenografts derived from tissues other than kidney were susceptible to HHT, and similarly express low levels of BCL2L1, suggesting a broad relationship across the varied origins of these SMARCB1-deficient rhabdoid tumors. Testing of other pediatric kidney cancers for sensitivity to HHT could also be considered in future studies.

Our in vivo studies showed that HHT, either alone or in combination with the standard-of-care therapy VCR, led to a significant increase in average mouse survival in four independent models of rhabdoid tumor, including one PDX. In addition, CRs were observed in HHT-treated mice bearing all four xenograft models and, in some models, HHT-based regimens outperformed standard-of-care arms. Importantly, the FDA-approved treatment schedule of HHT/OM was well-tolerated, both alone and in combination with VCR. We did observe a range of responses with some tumors continuing to progress with HHT treatment. Further studies will be needed to assess whether longer treatment courses, as well as additional standard-of-care and novel therapeutic combinations, could provide even greater benefit.

The equal sensitivities to HHT of rhabdoid tumor and CML cell lines combined with the inhibition of rhabdoid tumor growth mediated by HHT in four independent xenograft models is promising. HHT has exhibited both efficacy and a favorable safety profile in young children; thus, it may be an effective treatment for the often-fatal rhabdoid tumor. Clinical studies with this FDA-approved drug are needed to assess whether HHT, alone and in combination with other chemotherapies, holds promise for patients with rhabdoid tumor.

T.P. Howard reports grants from U.S. NIH (T32GM007753, T32GM007226) during the conduct of the study. B.B. Freeman reports personal fees from University of Kentucky College of Pharmacy outside the submitted work. C.M. Johannessen reports full-time employment with the Novartis Institutes of Biomedical Research. S.L. Schreiber serves on the Board of Directors of the Genomics Institute of the Novartis Research Foundation (GNF); is a shareholder and serves on the Board of Directors of Jnana Therapeutics; is a shareholder of Forma Therapeutics; is a shareholder and advises Kojin Therapeutics, Kisbee Therapeutics, Decibel Therapeutics, and Eikonizo Therapeutics; serves on the Scientific Advisory Boards of Eisai Co., Ltd., Ono Pharma Foundation, Exo Therapeutics, and F-Prime Capital Partners; and is a Novartis Faculty Scholar. W.C. Hahn reports personal fees from Thermo Fisher Scientific, KSQ Therapeutics, MPM, Solasta, iTeos, Frontier Biosciences, and Tyra Biosciences outside the submitted work. C.W.M. Roberts reports grants from NCI during the conduct of the study. No potential conflicts of interests were disclosed by the other authors.

T.P. Howard: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. E.M. Oberlick: Conceptualization, formal analysis, validation, investigation, visualization, writing-original draft, writing-review and editing. M.G. Rees: Conceptualization, data curation, formal analysis, validation, investigation, visualization, writing-original draft, writing-review and editing. T.E. Arnoff: Conceptualization, validation, investigation, visualization, methodology, writing-review and editing. M.-T. Pham: Validation, investigation, writing-review and editing. L. Brenan: Validation, investigation, writing-review and editing. M. DoCarmo: Validation, investigation, writing-review and editing. A.L. Hong: Conceptualization, formal analysis, writing-review and editing. G. Kugener: Formal analysis, investigation, writing-review and editing. H.-C. Chou: Formal analysis, investigation, visualization, writing-review and editing. Y. Drosos: Conceptualization, investigation, writing-review and editing. K.M. Mathias: Validation, investigation, writing-review and editing. P. Ramos: Conceptualization, resources, writing-review and editing. B. Seashore-Ludlow: Conceptualization, formal analysis, writing-review and editing. A.O. Giacomelli: conceptualization, writing-review and editing. X. Wang: Conceptualization, writing-review and editing. B.B. Freeman: Conceptualization, methodology, writing-review and editing. K. Blankenship: Investigation, writing-review and editing. L. Hoffmann: Investigation, writing-review and editing. H.L. Tiv: Investigation, writing-review and editing. P.C. Gokhale: Conceptualization, resources, supervision, investigation, methodology, writing-review and editing. C.M. Johannessen: Supervision, writing-review and editing. E.A. Stewart: Conceptualization, resources, formal analysis, supervision, investigation, visualization, methodology, writing-review and editing. S.L. Schreiber: Conceptualization, resources, supervision, writing-review and editing. W.C. Hahn: Conceptualization, supervision, funding acquisition, visualization, writing-original draft, writing-review and editing. C.W.M. Roberts: Conceptualization, supervision, funding acquisition, visualization, writing-original draft, writing-review and editing.

The authors thank the Roberts, Hahn, and Schreiber labs for insightful discussions. They thank Angela McArthur for manuscript comments. They thank Franck Bordeaut, C. David James, Yasumichi Kuwahara, Timothy Triche, and Bernard Weissman for cell lines. They thank Payton Archer, Childhood Solid Tumor Network, Preclinical Pharmacokinetics Shared Resource, and Center for In Vivo Imaging and Therapeutics at St. Jude for resources and assistance with in vivo studies. Support includes U.S. NIH grants T32GM007753 (to T.P. Howard), T32GM007226 (to T.P. Howard), P50CA101942 (to A.L. Hong), R00CA197640 (to X. Wang), U01CA176058 (to W.C. Hahn), R01CA172152 (to C.W.M. Roberts), and R01CA113794 (to C.W.M. Roberts); ACS Mentored Research Scholar 132943-MRSG-18-202-01-TBG (to A.L. Hong), Cure AT/RT Now (to C.W.M. Roberts), Avalanna Fund (to C.W.M. Roberts), Garrett B. Smith Foundation (to C.W.M. Roberts), and ALSAC (to C.W.M. Roberts and E.A. Stewart). This work is partially based on data generated by the Cancer Target Discovery and Development (CTD2) Network (https://ocg.cancer.gov/programs/ctd2/data-portal) and the TARGET initiative.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.