Identification of Therapeutic Targets in Rhabdomyosarcoma through Integrated Genomic, Epigenomic, and Proteomic Analyses

Abstract

Personalized cancer therapy targeting somatic mutations in patient tumors is increasingly being incorporated into practice. Other therapeutic vulnerabilities resulting from changes in gene expression due to tumor specific epigenetic perturbations are progressively being recognized. These genomic and epigenomic changes are ultimately manifest in the tumor proteome and phosphoproteome. We integrated transcriptomic, epigenomic, and proteomic/phosphoproteomic data to elucidate the cellular origins and therapeutic vulnerabilities of rhabdomyosarcoma (RMS). We discovered that alveolar RMS occurs further along the developmental program than embryonal RMS. We also identified deregulation of the RAS/MEK/ERK/CDK4/6, G₂/M, and unfolded protein response pathways through our integrated analysis. Comprehensive preclinical testing revealed that targeting the WEE1 kinase in the G₂/M pathway is the most effective approach in vivo for high-risk RMS.

Keywords: epigenetics; molecular targeted therapy; pediatric cancer; preclinical testing; proteomics; rhabdomyosarcoma.

Figure 1.. Differences in DNA Methylation Relative to Gene Expression

(A and B) Representative scatterplot of RNA-seq data (log2 FPKM, A) and the β value (B) from WGBS analysis for an O-PDX sample and corresponding RMS tumor. (C) Venn diagrams depicting overlap of upregulated and downregulated genes (green) in ERMS relative to ARMS tumors that are hypomethylated (blue) or hypermethylated (red). Gene number in each group is noted. (D-G) Examples of a gene that is hypermethylated downstream of the promoter in ARMS relative to ERMS (boxed region) and has reduced gene expression (MFAP2, D), that is hypomethylated in the region flanking the transcriptional start site in ARMS relative to ERMS (boxed region) with increased gene expression (NHLH1, E), that is hypermethylated in ARMS relative to ERMS with increased gene expression in ERMS (MMP16, F), and that is hypermethylated in ARMS relative to ERMS and has increased gene expression in ARMS (NOS1, G). Transcript boundaries indicated by dashed lines, and β values for each CpG are plotted for ERMS (blue) and ARMS (red). Expression in each sample from RNA-seq is shown in the histogram on the right (FPKM). FPKM, fragments per kilobase million. See also Tables S1 and S2.

Figure 2.. Differences in Epigenetic Profiles Relative to Promoter/Enhancer Activity

(A) Heat map of the 18 chromHMM states used in this analysis. (B) Heat map depicting proportion of the 18 chromHMM states in ERMS and ARMS for annotated regions of the genome. Proportion of individual marks or regions is directly proportional to the intensity of the blue color. (C) ChIP-seq peaks and chromHMM for MYOG. Scales are indicated on right, and gene boundaries marked by dashed lines. (D-F) Representative chromHMM for GAS2 that is selectively expressed in ERMS (D), NOS1 that is selectively expressed in ARMS (E), and for AXL that is selectively expressed in myoblasts (F). NOS1 contains a tumor type–specific super-enhancer upstream of the gene (boxed H3K27Ac region). Gene boundaries are marked by dashed lines, and colors for each state are indicated as in (A). The two highest proportion states are full-height bars, and remaining states are half height. Intensity of each bar is proportional to the percentage of each state across all stages for that gene. For half-height bars, intensity is scaled starting at 50% of maximum intensity. (G) Venn diagram for genes upregulated or downregulated (green) in ERMS relative to ARMS. Overlap with genes that have differences in chromHMM state (blue) or tumor type–specific super-enhancers (red) are shown. See also Table S3.

Figure 3.. Profiling the Rhabdomyosarcoma Proteome and Phosphoproteome

(A) Workflow of proteomic and phosphoproteomic profiling of RMS tumors, myotubes, and myoblasts. LC, liquid chromatography; MS, mass spectrometry. (B) Representative immunoblots validating protein and phosphoprotein expression of NOS1, phospho-Ser235/236 of RPS6, and β-actin (loading control). (C) Bar plot of normalized relative fold expression of the proteins in (B) and HMGA2. mb, myoblasts; mt, myotubes. (D and E) Scatterplot of quantitation of proteins (D) or phosphoproteins (E) across replicate samples of SJRHB000026_X1 in the same and in different batches. (F and G) Principle component analyses of myoblasts (light green), myotubes (dark green), ERMS (red), and ARMS (purple) for the whole proteome (F) and phosphoproteome (G). (H and I) Bar plot of RNA expression (FPKM) for MYOG (H) and MYF5 (I) in developing muscle and ARMS and ERMS tumors. Red line indicates expression level in human myoblasts. The ERMS-high group are the samples with highest levels of MYOG and the ERMS-low group are the remaining samples. FQ21, fetal quadriceps at 21 weeks of gestation. (J)Scatterplot of the Log₁₀ protein expression intensity from proteomic analyses of MYOG and MYF5. The two outlier samples (black boxes) were excluded from linear regression (dashed line). (K) Representative immunostaining of samples for MYOG and MYF5 and scatterplot of the of percent MYOG and MYF5 immunopositive cells for each O-PDX tumor with a best fit linear regression (dashed line). Scale bar = 20 μm. See also Figures S1-S5 and Tables S4 and S5.

Figure 4.. Integration of Transcriptomic, Epigenetic, and Proteomic/Phosphoproteomic Data

(A) Scheme for integration of proteomic and transcriptomic data based on differential expression clustering and PPI network modules. (B) Box plot of normalized relative abundance of individual proteins in each IPC from the integrated analyses. The rectangle represents the interquartile range (IQR) for first to third quartile and the horizontal line within the box is the median. The vertical lines that extend above and below the rectangle represent the maximum and minimum values excluding the outliers that are more than 3 times the IQR. Data were normalized to myoblasts, and number of genes/proteins in each IPC are shown. Representative gene/proteins in each IPC are shown in gray. (C) Heat map of the –Log₁₀ enrichment FDR for pathway analysis of the genes/proteins in each IPC. For IPC3, no pathway reached statistical significance (p value < 0.05) in that cluster. (D) Representative PPI modules for IPC1 showing E2F modules (blue circle) and G₂/M checkpoint modules (dashed circle). (E) Expanded view of the G₂/M checkpoint module showing representative genes/proteins. See also Table S6.

Figure 5.. Targeting RAS and CDK4/6 in Rhabdomyosarcoma

(A) Simplified pathway map for the RAS and CDK4/6 pathways. Recurrent mutations in genes encoding proteins highlighted in purple and those in bold were targeted with small-molecule inhibitors. (B) Representative ChIP-seq tracks and chromHMM for CDK1. Scales are indicated on right. Only one splice variant is shown for simplicity. (C) Bar plot of normalized relative CDK1 protein expression for myoblasts, myotubes, and RMS tumors. (D) Representative micrographs of ERMS and ARMS tumor sections with hematoxylin and eosin (H&E) staining or immunohistochemical staining of cyclin D2 (ERMS), cyclin D3 (ARMS), and phosphoRB1 (pRB1). Scale bar = 50 μm in the full image and 10 μm in the magnified inset. (E) Representative combination drug sensitivity study of 10 concentrations of palbociclib and four concentrations of trametinib in RD cells. Survival is plotted relative to untreated cells (100% survival) and complete ablation (0% survival) with a positive-control cytotoxic compound. (F) PK plot of palbociclib in O-PDX tumors (blue) and plasma (red), n = 3 mice per treatment group. Mean and standard deviations are plotted for each time point. (G and H) Representative plot of SJRHB013759_X1 O-PDX tumor burden, as measured by bioluminescence over time (G) and images of mice with PD in the placebo control group and the palbociclib+trametinib treatment group (H). See also Figure S6 and Table S7.

Figure 6.. Targeting HSP90 in Rhabdomyosarcoma

(A) Illustration of the six myogenic pathways deregulated in RMS and modulated by HSP90. (B) Bar plots of HSF1 and S368 phosphorylated HSF1 in myoblasts, myotubes, and RMS tumors. Data normalized to myoblasts (red line). (C and D) Bar plots of quantitative expression of 396 HSF1 target genes (C) and representative examples of HSF1 target gene expression (D) in RMS tumors normalized relative to that of myoblasts, as measured by RNA-seq. Bars are mean and standard deviations across multiple tumors. (E) ChromHMM of an HSF1 target gene (HSPA1A) and corresponding gene expression by RNA-seq (FPKM). (F) Heat map of the log₁₀ 50% effective concentrations of a subset of drugs tested. Bolded terms denote primary cultures of O-PDXs, whereas non-bolded terms denote established cell lines. (G) Immunoblot of HSP70 expression in RD cells after treatment with GSP alone or IRN+VCR. Control indicates untreated cells. HSP70 intensities were quantified and normalized to β-actin before calculating fold differences between treated and untreated samples. See also Table S8.

Figure 7.. Targeting WEE1 in Rhabdomyosarcoma

(A) Box plot of WEE1 gene expression in pediatric cancers from https://pecan.stjude.org/home. The rectangle represents the interquartile range (IQR) for first to third quartile and the vertical line within the box is the median. The horizontal lines that extend from the rectangle represent the maximum and minimum values excluding the outliers (circles) that are more than 3 times the IQR. CPC, choroid plexus carcinoma; EPD, ependymoma; HGG, high-grade glioma; LGG, lowgrade glioma; MB, medulloblastoma; AML, acute myeloid leukemia; B-ALL, B cell acute lymphoblastic leukemia; MLL, mixed lineage leukemia; T-ALL, T cell acute lymphoblastic leukemia; ACT, adrenocortical tumors; MEL, melanoma; NBL, neuroblastoma; OS, osteosarcoma; RB, retinoblastoma; WT, Wilms tumor. (B) Simplified drawing of the pathway regulated by WEE1 at the G₂/M checkpoint. Arrows indicate activation and bars indicate repression. (C) ChromHMM for WEE1 with corresponding RNA-seq expression across samples. (D) Bar plot of WEE1 protein levels in myoblasts, myotubes, and RMS tumors normalized to myoblasts (red line). (E-G) DNA content and cell cycle distribution for RD cells before (black) and after treatment (red) with AZD1775 (E), representative micrographs of DAPI-stained nuclei of these cells highlighting fragmented nuclei/mitotic catastrophe (F) and proportion of cells with nuclear features consistent with mitotic catastrophe after the treatment shown (G). Scale bar in F = 2 μm. Bars in G are mean and standard deviations of duplicate scoring.

Figure 8.. Preclinical Testing of AZD1775 and Ganetespib

(A) Dose response curve for GSP and AZD1775 using primary cultures of SJRHB000026_X1 O-PDX. (B) PK plot of GSP in O-PDX tumors (blue) and plasma (red), n = 3 mice per treatment group. Mean and standard deviations are plotted for each time point. (C) Drug schedule selected for each combination treatment regimen. The daily × 5 × 2 low-dose protracted schedule for IRN is shown but we also tested the alternative daily × 5 schedule. (D) Xenogen images of mice with an O-PDX tumor (SJRHB012_Y) treated with IRN+VCR or GSP+IRN+VCR, which had PD or a CR. (E) Representative plot of tumor burden, as measured by bioluminescence over time, in a preclinical phase 2 study with each of the three O-PDX tumors indicated. Each line is a different mouse. (F) Representative images of mice and their corresponding tumors with PD in the placebo control group, AZD1775 group, and GSP+IRN+VCR group. Mice with SD and a CR are shown for the IRN+VCR and AZD1775+IRN+VCR treatment groups, respectively. (G) Line plot of tumor burden, as measured by bioluminescence over time, in a preclinical phase 3 study of the SJRHB012_Y O-PDX model. (H) Survival curve for the SJRHB012_Y O-PDX for the four indicated treatment groups. (I) Histogram of the percentage of CR and PR for each of the four O-PDX models for all 14 treatment groups. Two clinically relevant doses, high (h) and low (l), of AZD1775 and GSP were tested along with several different schedules and dosing regimens of IRN used in pediatric cancer patients. * indicates p value < 0.05, compared with IRN+VCR. See also Table S9.