Targeting PAK4 Inhibits Ras-Mediated Signaling and Multiple Oncogenic Pathways in High-Risk Rhabdomyosarcoma

Rhabdomyosarcoma (RMS) is the most common soft-tissue sarcoma in children and adolescents with an incidence rate of 4.5 cases per million children (1). Although multimodal therapy, including surgery, radiation, and cytotoxic chemotherapy, has improved 5-year survival rates to almost 70%, prognosis for patients with relapsed or metastatic disease remains dismal at 20%–30% (1), indicating a dire need for novel therapeutic targets. RMS is classified into embryonal (ERMS) and alveolar (ARMS) based upon molecular and histologic features. About 70% of RMSs are ERMSs, while the rest are comprised of ARMSs or fusion-positive tumors that frequently contain the PAX3/7-FOXO1 chromosomal translocation, which creates a novel fusion protein (2, 3). Recently, whole-exome sequencing of RMS xenografts identified mutations in the RAS/NF1 pathways in approximately 45% of cases, predominantly within the fusion-negative (ERMS) population (4), and 75% of high-risk patients were found to have RAS pathway mutations (5, 6). These studies have also shown that overall 93% of RMS cases carry alteration in the receptor tyrosine kinase (RTK)/RAS/PI3K axis (4). This clearly shows the importance and the need to further study the significance of RAS signaling in RMS. Thus, there have been several attempts to therapeutically target RAS, but to date no direct inhibitor of RAS mutants has been approved by FDA. This is partly due to the difficulty to directly target RAS, for the lack of deep hydrophobic pockets for binding of small-molecule inhibitors (7). Very recently, promising preclinical results with AMG510, a direct inhibitor targeting the mutant KRAS G12C (8), have prompted a phase I and II clinical trial (NCT03600883) with ongoing recruitments and results are awaited with excitement. Moreover, attempts involving clinical trials to target downstream RAS effectors, such as ERK/MAPK and PI3K/AKT (9), as single agents have not been successful. This highlights the necessity to probe into unexplored RAS effectors in RMS.

A group of serine/threonine kinases, namely p21-activated kinases (PAK), are one such class of effectors. PAKs occupy a hub through which many oncogenic signaling pathways intersect. Initially identified as an effector of Rho GTPases, PAKs play a central role in the reorganization of the cytoskeleton and depolymerization (10) and consequently, cellular architecture, migration, and adhesion (11, 12). More recently they have been shown to regulate several cellular pathways, including growth factor- and steroid receptor–mediated signaling, energy homeostasis, transcription, and mitosis (13, 14). This central role of PAKs allows the linking of several key signaling pathways, which include the Ras-ERK, Wnt/β-catenin, estrogen receptor, and androgen receptor pathways (15), making them potential therapeutic targets. In addition, as recently reported, PAK signaling plays a critical role in acquired resistance. In BRAF-mutant melanoma, MEK inhibitor–resistant cell lines and clinical samples show activation of PAK signaling that can be targeted to synergize with MEK inhibition (16).

So far six isoforms of PAKs have been identified in humans, divided into two families, on the basis of domain structure and regulation mechanisms. Group 1 PAKs, comprising of PAK1–3, and group 2 PAKs, comprising of PAK4–6 (17, 18). These latter PAKs are activated independent of GTPases (19, 20). From the first group, PAK1, by virtue of being discovered first, has been most extensively studied among the PAKs and has been found to be overexpressed in several cancers, including breast, ovarian, bladder, colorectal, lungs, and in T-cell lymphoma and glioblastoma (14, 21, 22). Among the group 2, PAK4 overexpression and/or hyperactivation has been shown to be associated with tumorigenesis in several malignancies, such as prostate, breast, gastric, lung, ovarian, melanoma, and pancreatic cancer (15, 23, 24).

Interestingly, PAK4 has been demonstrated to be essential for Ras-mediated transformation. As reported by Callow and colleagues, inactive PAK4 can inhibit, or significantly attenuate Ras transformation of NIH3T3 and rat intestinal epithelial cells, and abolish anchorage-independent growth of Ras-mutant HCT116 colon cancer cells (25). In addition, overexpression of wild-type or activated PAK4 in NIH3T3 when injected in athymic mice was sufficient for tumor generation. Histologic analysis of these tumors showed morphologic features consistent with sarcoma transformation (26). Furthermore, PAK4 was reported to be required for oncogenic transformation of MDA-MB-231 breast cancer cells (23). Besides these oncogenic functions, PAK4 has also been noted to contribute to chemoresistance and has a role in cancer stem cell phenotypes (14, 27).

As alluded to, while the oncogenic role of PAKs has been studied in several malignancies, there is a still a glaring gap in knowledge regarding their role in sarcomas. Our work in Ewing sarcoma was one of the first studies to report that PAKs are critical regulators of aggressive phenotypes in pediatric sarcomas. We noted that activation of PAK signaling promotes metastatic phenotypes in vitro and metastasis in vivo in Ewing sarcoma (28). A very recent study has shown that guanine nucleotide exchange factor T is correlated with metastatic progression in RMS and in general with a worse prognosis, the underlying mechanism of which is through the Rac1/Cdc42-mediated pathway (29).

Therefore, there is a strong scientific rationale that inhibition of PAKs may be an effective therapeutic strategy in high-risk RMS. To date, there have been no studies investigating the prospects of therapeutically targeting PAKs for the treatment of RMS. PAKs have become a significant target for therapeutic development by pharmaceutical companies (30, 31). Recent studies in lung cancer and melanoma have demonstrated the therapeutic importance of inhibiting PAKs (32–35). Of these targeted inhibitors, PF-3758309 was evaluated in a clinical trial (NCT00932126), but the trial was terminated because of poor bioavailability and undesirable pharmacokinetics characteristics. A second compound, KPT-9274, a dual inhibitor of PAK4 and NAMPT, is currently in clinical trials for refractory/relapsed hematologic and solid tumors (NCT02702492). Hence, in this work we attempt to address the validity to test understudied effectors of oncogenic signaling pathways, in this case PAK4. In conjunction with our PAK4 knockdown (KD) and pharmaceutical targeting of PAK4 in in vivo models, our results provide extremely strong preclinical evidence that targeting PAK4 can be a potential therapeutic approach for advanced RMS.

KPT-9274, provided by Karyopharm Therapeutics Inc., was reconstituted in DMSO and further diluted in culture media. PF-3758309 was procured from MedChemExpress (#HY-13007). Propidium Iodide Flow Cytometry Cell Kit (#ab139418) was procured from Abcam. Dead Cell Apoptosis Assay Kit (#V13242) was procured from Thermo Fisher Scientific. BrdU Cell Proliferation Kit (#6813) was obtained from Cell Signaling Technology. Following antibodies were used for Western blotting and/or immunostaining: PAK4 from ProteinTech (#14685-1-AP), HIF1α from Novus Biological, and β-actin (clone C4) from Millipore; and AXL (Tyr702) (#D12B2), β-catenin (#9582), S675-phospho-β-catenin (#4176), S474-phospho-PAK4 (#3241), and vinculin (#13901) from Cell Signaling Technology.

Human RMS cell lines were procured from Children's Oncology Group. Cells were grown in RPMI media supplemented with 20% FBS. Cells were grown in a humidified atmosphere containing 5% CO₂ at 37°C. TCCC-RMS40, a patient-derived xenograft (PDX) cell line was generated from tumor biopsy from an 8-year-old male with left parameningeal ERMS (stage 3 group III) complicated by intracranial extension with local recurrence 6 months after completion of chemotherapy. To generate stably transfected KD cells, RMS cells were seeded in wells of 6-well plate at 60% density. Lipofectamine 3000 was used to transfect DNA and Lipofectamine RNAiMAX (Invitrogen) was used for siRNAs according to the manufacturer's instruction. A final concentration of 10 nmol/L siRNA pool was used for each transfection. Stable clones were generated by selecting under puromycin. All cell lines were tested for Mycoplasma contamination every 6 months by using MycoAlert Mycoplasma Detection Kit from Lonza. No cell line was used for more than 10 passages after thawing from frozen stocks. All cell lines were authenticated at the Cell Line Characterization Core of MD Anderson Cancer Center (Houston, TX) as recently as 22 July 2019.

Patient samples were obtained under the approval of H-32129, which was approved by the Institutional Review Boards (IRB) for Baylor College of Medicine (BCM, Houston, TX) and Affiliated Institutions (IRB00002649). The BCM IRB approved a waiver of consent/HIPAA authorization and determined that all requirements were met by this protocol to grant the waiver. Informed written consent was not needed from patients. Patient clinical annotations are outlined in Supplementary Table S1.

Migration/invasion assays were analyzed in a 24-well Boyden chamber. For invasion, 30 μg/mL collagen-coated wells were used. Cells were suspended in 100 μL serum-free media and bottom chamber was filled with media containing 10% FBS. After incubation at 37°C for the 24 hours, cells were fixed and stained with 0.1% crystal violet. Random fields were quantified using ImageJ.

Briefly, total cell lysates were isolated in 50 mmol/L Tris-HCl (pH 7.4), 150 mmol/L NaCl, 1 mmol/L EDTA, and 1% Triton X-100 lysis buffer supplemented with Protease Inhibitors (Complete mini, Roche) and phosphatase inhibitors. Proteins were separated on 4%–12% Bis-Tris (Invitrogen) or TGX (Bio-Rad) gels, and transferred to polyvinylidene difluoride membranes. Blots were incubated sequentially overnight with primary and secondary antibodies. Immunoreactive signals were developed with ECL Kit (Thermo Fisher Scientific) or by Li-COR Odyssey System.

Briefly, cells were fixed in ice-cold 70% ethanol at 4°C overnight, followed by PBS wash next day. Cells were resuspended in 1 × buffer with propidium iodide along with RNaseA and incubated for 15 minutes at room temperature. Analysis was performed on a BD LSRII analyzer. Statistical modeling was done with FlowJo software to fit the curves (v10). A total of 10,000 cells were analyzed for each sample.

Growth assays were assessed by plating 1,000 cells per well in triplicate per day for 4 days in a 96-well plate, unless otherwise indicated. Cell proliferation was detected in two ways. In the first method, through the addition of the colorimetric reagent Cell Counting Kit-8 (CCK-8, Dojindo Laboratories), according to the manufacturer's instructions (CCK-8 Assay Kit; Dojindo Laboratories). In the second method, CellTiter-Glo Luminescent Cell Viability Assay (Promega) was used to measure cell viability. Assays were done in triplicates.

Soft agar colony formation assay was done in 6-well plates with 0.5% agarose in 1 × media as base layer. RMS cells (or stable clones) were seeded at the density of 2–5 × 10⁴ cells in a 6-well plate dissolved in the top layer of 0.35% agarose in 1 × media over the bottom layer. Media from the top were refreshed every 3 days. After 14–21 days, cells were stained with crystal violet and methanol. Images of colonies were taken with a scanner and quantified using ImageJ. All experiments were done in triplicate.

Total RNA was harvested using TRIzol extraction method. The qScript cDNA SuperMix (QuantaBio) was used to synthesize cDNA from 200 ng of total RNA from each sample. Relative quantifications of mRNA expression of genes were examined by qRT-PCR with iTaq Universal SYBR Green Supermix (Bio-Rad). Reactions were performed on a StepOnePlus system. All reactions were run in triplicate. Melting curve analysis verified that all primers yielded a single PCR product. Gene expressions were normalized to 18S to yield a 2^−ΔΔCt value. All primers were purchased from Sigma-Aldrich. Primer sequences are included in Supplementary Table S2.

All animal experiments were conducted according to Institutional Animal Care and Use Committee protocols after approval was obtained from the BCM IRB (BCM Animal Protocol AN-5225). KPT-9274 was administered through gavage, while PF-3758309 by subcutaneous administration.

Tissue microarray (TMA) slides with 18 human RMS samples, along with one sample each of smooth muscle, skeletal muscle, and cardiac muscle tissue, were obtained from US Biomax Inc (#SO751). Formalin-fixed, paraffin-embedded tissue sections were deparaffinized followed by rehydration by taking the slides through a graded xylene and ethanol series. VECTASTAIN Elite ABC kit, following the manufacturer's instructions, was used for visualization. For control, IgG isotype was used instead of primary antibody, wherever indicated. ImageJ-based analysis of immunostaining is described in Supplementary Materials and Methods.

Reverse phase protein array (RPPA) assays were carried out as described previously (28, 36).

Gene ontology (GO) analysis was done using Thomson Reuters Metacore program for genes that had a differential expression of 1.5-fold and P values, as evaluated by two tailed t test, P < 0.05.

RNA samples underwent quality control assessment using the RNA tape on TapeStation 4200 (Agilent) and were quantified with Qubit Fluorometer (Thermo Fisher Scientific). The RNA libraries were prepared and sequenced at the University of Houston (Houston, TX) Seq-N-Edit Core per standard protocols. RNA libraries were prepared with QIAseq Stranded Total RNA Library Kit (Qiagen) using 500 ng input RNA. mRNA was enriched with oligo-dT probes attached to Pure mRNA Beads (Qiagen). RNA was fragmented, reverse transcribed into cDNA, and ligated with Illumina sequencing adaptors. The size selection for libraries was analyzed using the DNA 1000 tape TapeStation 4200 (Agilent). The prepared libraries were pooled and sequenced using NextSeq 500 (Illumina); generating approximately 17 × 10⁶, 2 × 76 bp paired-end reads per samples.

Paired-end sequencing reads were trimmed using the trimGalore software. Detailed description of the analysis is described in the Supplementary Materials and Methods.

A P < 0.05 was considered significant was used to calculate the level of statistical significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). PAK1 and PAK4 expression data were pulled from the Schafer Welle mixed RMS dataset in R2: genomics analysis and visualization platform (http://r2.amc.nl) with 30 RMS samples and 26 normal muscle samples. Expression levels were graphed. All statistical significance tests were performed using GraphPad prism.

PAKs are potential central signaling nodes integrating pathways critical to growth for both fusion-positive and -negative RMS tumors, including growth factor receptor signaling and Ras activation (Fig. 1A). We initially evaluated the transcript expression of both PAK1 and PAK4, using the Schafer Welle mixed RMS dataset with the bioinformatics program: R2 (http://r2.amc.nl). We noted significantly increased expression of PAK4 (Fig. 1B), but not PAK1, in both ERMS and ARMS compared with normal human muscle tissue. Analysis of other PAK members revealed that PAK3 was the only other PAK significantly upregulated in RMS tumors compared with all the other PAKs (Supplementary Fig. S1A), but with a much wider variation in expression in the RMS samples.

Furthermore, staining for PAK4 by IHC on a TMA of primary patient tumor specimens showed high expression of total PAK4 compared with nonmalignant tissue specimens (Fig. 1C). A comparison of the ARMS and ERMS with the normal muscle samples on this TMA showed PAK4 to be significantly higher in both ARMS and ERMS samples (Fig. 1D). Subsequently, we confirmed higher expression of PAK4 using direct patient biopsy/surgery samples, as well as PDX tumor samples, by Western blotting (Fig. 1E; Supplementary Fig. S1B). Secondary to these results, we compared the protein expression of total and phosphorylated PAK4 across a panel of fusion-positive and -negative RMS cell lines with primary human skeletal muscles cells (Fig. 1F). We observed that both phosphorylated and total PAK4 protein levels were elevated in the tumor cell lines, albeit with some heterogeneity among the cell lines with respect to the level of this elevation. RD, which is a fusion-negative cell line, harboring mutations in NRAS, p53, and c-Myc amplification, showed higher PAK4 phosphorylation compared with other cell lines tested. The convergence of these dysregulated pathways on the regulation of PAK4 can be a potential explanation for these findings.

To further investigate the importance of PAK4-mediated signaling, we first looked at activation of PAK4 downstream of major growth factor receptor–mediated signaling in RMS by Western blotting in two cell lines, RD and RH30 (Supplementary Fig. S1C). Densitometric quantification (Supplementary Fig. S1D) of the Western blot indicated an increase in activation of PAK4 levels for both cell lines with basal FGF 2 stimulation within 15 minutes, while stimulation with insulin-like growth factor 2 resulted in increase in PAK4 activation in RD cell line only. Concomitantly, we saw an increase in phospho-β-catenin over its total and phospho-ERK1/2. This increase was significant for phospho-ERK1/2 over total ERK1/2 (Supplementary Fig. S1D).

To evaluate the functional significance of PAK4, we undertook multiple molecular and pharmacologic approaches to attenuate PAK4 levels and/or activity in RMS tumor models. We first generated PAK4 KD in two RMS cell lines, RH30 and RD, using two independent short hairpin RNAs (shRNA) against PAK4. KD was confirmed by Western blot analysis (Supplementary Fig. S2A and S2B). No discernable off-target effects were noted, as we did not observe any effects on PAK1 (Supplementary Fig. S2C). Fusion-negative and -positive RMS cell lines models, RH30 (Fig. 2A) and RD (Fig. 2B), were used, multiple clones of both of which showed a significant reduction in cell viability when compared with the control scrambled. In addition, we observed significant inhibition in migratory and invasive potential in RH30 (Fig. 2C and D), as represented by microscopic images (Supplementary Fig. S2D), and in RD cell lines (Fig. 2E and F; Supplementary Fig. S2E). To further ascertain the role of PAK4 in tumorigenesis, we assessed anchorage-independent growth via soft agar colony formation assay in RH30 and RD cell lines, and all KD clones demonstrated significantly fewer colony formation under these conditions, compared with the scramble controls (Fig. 2G; Supplementary Fig. S2F). A complementary, reciprocal experiment with stable overexpression of full-length PAK4 in the RH30 cell lines showed a significant increase in the number of soft agar colonies (Fig. 2H; Supplementary Fig. S2G).

After characterizing the phenotypic contributions of PAK4 in vitro, we investigated the role of PAK4 KD on primary and metastatic tumor development in vivo using orthotropic and tail vein murine models. Upon orthotopic injection into the gastrocnemius muscle of NSG mice, RH30 shPAK4 cells resulted in a significant decrease in tumor growth compared with scrambled control cells, as seen by either mean growth rate of each arm (Fig. 3A) or for individual mice in each group (Supplementary Fig. S3A). We observed similar findings with RD PAK4-KD cells (Fig. 3B; Supplementary Fig. S3B). Tumors samples showed a significant increase in cleaved caspase-3 (CC-3) staining in the RH30-PAK4–KD tumors (Fig. 3C; Supplementary Fig. S3C). On termination of experiment, in the residual tumor, we did not find any significant difference in PAK4 staining (Supplementary Fig. S3D).

To investigate the role of PAK4 in contributing toward RMS metastatic potential, we performed tail vein injections to model tumor dissemination and allowed for cells to circulate, colonize, and grow colonies for 3 weeks (Fig. 3D). At the end of 3 weeks, compared with control RH30 cells, RH30 shPAK4 cells resulted in lower metastatic burden, as we observed fewer metastatic foci from hematoxylin and eosin (H&E) staining (Fig. 3E), along with a significant decrease in overall liver weights and surface area covered by metastatic lesions in the liver sections (Fig. 3F). Similar inhibition of metastatic potential was observed in mice injected with RD PAK4-KD cells (Fig. 3G and H).

After demonstrating that KD of PAK4 has potent antitumor and metastatic properties in RMS, we investigated the efficacy of small-molecule inhibitors of PAK4 to provide additional preclinical assessment for the therapeutic targeting of this kinase in RMS. We utilized two independent PAK4 small-molecule inhibitors, KPT-9274, an allosteric inhibitor of PAK4 and NAMPT (Karyopharm Therapeutics), and PF-3758309 (Pfizer), a potent ATP-competitive, pyrrolopyrazole inhibitor of PAK4 (Supplementary Fig. S4A).

We first assessed the effect of KPT-9274 and PF-3758309 on proliferation of several RMS cell lines in vitro. A dose-dependent reduction in cell proliferation was observed with both drugs at low nanomolar concentrations. For KPT-9274, the IC₅₀ values ranged from 40 to 80 nmol/L for the multiple cell lines studied, whereas for PF-3758309 the IC₅₀ values ranged from 8.4 to 33.7 nmol/L (Supplementary Fig. S4B). As seen in Fig. 1C and D, normal tissue and myoblast cells have very low PAK4 expression and several studies have shown very little toxicity of KPT-9274 on normal tissues (37, 38). To further address this, we investigated the effect of KPT-9274 and PF-3758309 on normal skeletal muscle myoblast (SKMM) cells. We observed (Supplementary Fig. S4B) the IC₅₀ values were 8–10 times higher for both PAK4 inhibitors on the nonmalignant cells. Overexpression of PAK4 in RH30 cells increased the IC₅₀ from 80 to 95 nmol/L under KPT-9274 treatment and from 33.7 to 45 nmol/L under PF-3758309 treatment. Overexpression of PAK4 in RD cells (Supplementary Fig. S4C), on the other hand, was significant (P < 0.001) for both small-molecule inhibitors, with the IC₅₀ increasing from 75.5 to 118 nmol/L with KPT-9274 treatment, and the IC₅₀ increasing from 8.5 to 23.5 nmol/L with PF-3758309 treatment (Supplementary Fig. S4B). To evaluate potential off-target effects, we evaluated PAK1 expression for RMS cells treated with KPT-9274 and PF-3758309 under similar conditions as above and observed no effect on PAK1 levels (Fig. 4A).

To evaluate for the mechanism for this reduction in cell proliferation, we first looked at induction of apoptosis by flow cytometry. Treatment for 72 hours with either drugs resulted in significant increase in apoptotic cells in both RD and RH30 cells (Fig. 4B). Percentage of apoptotic cells in the RD cell line increased from 1.4% with vehicle treatment to 4% and 4.7% with 100 and 200 nmol/L KPT-9274, respectively. Treatment of 100 nmol/L of PF-3758309 had the most effect by raising distribution of apoptotic cells to 5.7%. In the RH30 cell line, KPT-9274 increased apoptotic cells significantly from 0.1% to 4.8% at a concentration of 100 nmol/L and to 5.5% at 200 nmol/L. Treatment with PF-3758309 resulted in an increase to 2.2% (Fig. 4B; Supplementary Fig. S4D). Thus, both drugs significantly induced apoptosis at nanomolar concentrations. Accordingly, investigating the effect of the two PAK4 inhibitors on cell proliferation rate via bromodeoxyuridine (BrdU) incorporation, we found KPT-9274 caused a significant reduction at 200 nmol/L for both cell lines, while RH30 had a significant reduction in BrdU incorporation when treated with 200 nmol/L PF-3758309 (Supplementary Fig. S4E).

In addition, to further account for the reduction in the number of tumor cells, we looked at the stages of the cell cycle by flow cytometry (Fig. 4C; Supplementary Fig. S4F). In RH30 cells, treatment with either the KPT-9274 or PF-3758309 resulted in a significant increase of cells in G₁-phase with a concomitant decrease of cells in S-phase, indicating a block in G₁–S-phase progression. In the RD cell line, like RH30, KPT-9274 induced a G₁–S-phase arrest, by significantly increasing cells in G₁-phase, while reducing the percentage in S-phase. However, treatment with PF-3758309 seemed to have no effect on the distribution of cells in the G₁-phase, but reduced cells in S-phase with an increase in G₂-phase, indicating a G₂–M-phase arrest. This difference in stages of arrest could be attributed to the fact that KPT-9274 is a dual inhibitor of NAMPT and PAK4, and some of observations are possibly due to inhibition of NAMPT or a combination of blocking both targets.

Recurrence and metastasis remain a therapeutic challenge in treating patients with RMS. Indeed, PAK4 signaling has been implicated in cytoskeletal alterations and signaling pathways associated with enhanced metastatic phenotypes. Thus, we were interested in assessing the effect of inhibiting PAK4 on phenotypic properties associated with metastasis. Using the Boyden transwell assay, we probed the effect of KPT-9274 on invasive and migratory properties. Our results show that KPT-9274 significantly inhibited cell motility for both cell lines (Fig. 4D and E; Supplementary Fig. S5A), while PF-3758309 had a significant inhibitory effect on the motility of RD cells (Fig. 4D), but not on RH30 cells (Fig. 4E). Treatment with either of the two drugs induced a significant inhibition in invasive properties for both cell lines tested, with greater reduction in invasive properties noted with PF-3758398 treatment.

Besides the phenotypic properties, we investigated the ramifications on downstream signaling secondary to PAK4 targeting. As noted in Fig. 4F, in vitro treatment with KPT-9274 and PF-3758309 confirmed a reduction of PAK4 activity via decreased PAK4 phosphorylation and decreased activation of the downstream target, β-catenin, as shown by representative Western blot (Supplementary Fig. S5B) from two independent studies (Fig. 4F). Besides the perturbations in signaling, we also noted a significant increase in PARP cleavage in RH30 cells thus, confirming induction of apoptosis. While in RD cells, we noticed a cleavage product at molecular weight 55–60 kDa, which is indicative of necrosis (39, 40). Thus, pharmacologic inhibition of PAK4 leads to potent antitumor and metastatic properties in vitro through downregulating critical signaling pathways involved in tumor development and progression.

Results confirming significant antitumor and antimetastatic effects in vitro, led us to evaluate the preclinical testing of KPT-9274 against RMS primary and metastatic growth in vivo, using orthotropic and metastatic tail vein mouse models. For orthotropic models, RMS cells were injected into the gastrocnemius muscle of NSG mice. As shown in the schematic (Fig. 5A), once a tumor was palpable, mice were randomized to receive either KPT-9274 or vehicle by oral gavage, and administration was continued until the endpoint of tumor burden was reached for one of the cohorts. KPT-9274 treatment resulted in a significant regression in tumor growth with RH30 cell lines (Fig. 5B) as compared with the control cohorts. To confirm that this observed antitumorigenic effect of KPT-9274 was indeed through inhibition of PAK4, we performed Western blot analysis and IHC for PAK4 from tumor samples from the RH30 in vivo studies. PAK4 protein level attenuation was observed in the tumors from animals treated with KPT-9274 compared with the control tumors (Fig. 5C and D; Supplementary Fig. S6A). We also looked at inhibition of β-catenin activation (S675), as a direct downstream target of PAK4 (Supplementary Fig. S6B), and inhibition at the level of protein expression for both activated and total β-catenin was observed. To assess for apoptosis in these tissues, we stained for CC-3 in three random tumor samples from each group and there was a significant increase in CC-3–positive cells in the KPT-9274–treated group (Fig. 5E). Similar antitumorigenic growth result was observed with RD cell lines as well (Fig. 5F; Supplementary Fig. S6C) when treated with KPT-9274. This reduction was not associated with any apparent toxicity, and there was no significant weight loss for animals receiving KPT-9274 when compared with animals receiving vehicle control (Supplementary Fig. S6D). The xenograft experiments were repeated in RH30 with similar antitumorigenic results (Supplementary Fig. S6E), again with no significant effect on body weight (Supplementary Fig. S6F).

Even though PF-3758309, another potent PAK4 inhibitor, did not successfully progress through clinical trials, as a proof of principle and to confirm that the antitumor activities seen in xenograft models can be attributed to blocking PAK4, we injected RH30 cells into the gastrocnemius muscle of NSG mice. Once tumors were palpable, mice were randomized to receive either PF-3758309 or vehicle by oral gavage, and administration was continued until the endpoint of tumor burden was reached for one of the cohorts. Again, we observed a significant reduction in tumor growth with PAK4 inhibition (Supplementary Fig. S6G).

The PAK4-KD and PAK4 small-molecule inhibition in vitro studies suggest a significant role for PAK4 in dictating metastatic phenotypes, thus providing the rationale to investigate pharmacologic efficacy of KPT-9274 in targeting metastatic RMS in vivo. Using the RD cell line in a tail vein metastatic model, we evaluated the efficacy of PAK4 inhibition by KPT-9274. Treatment of KPT-9274 was started 2 weeks after injection and continued for 3–4 weeks, as highlighted in the schema (Fig. 5G). At the end of the treatment period, we evaluated for the presence of metastatic disease. We observed the KPT-9274 treatment group had lower metastatic burden in liver compared with the vehicle-treated group (Fig. 5H) by gross macroscopic visualization, as well as by histologic staining (Fig. 5I) with H&E.

After demonstrating significant antitumor activity on established fusion-positive and -negative RMS cell lines, we tested the potential antitumor effects of KPT-9274 on a high-risk fusion-negative PDX model. Cryopreserved tumor pieces from TCCCRMS40, derived from a patient with relapsed RMS that also had high PAK4 protein expression, were implanted subcutaneously into NSG mice (Fig 6A). As evident in Fig. 6B, KPT-9274 treatment dramatically inhibited tumor growth, with significant differences in tumor volumes (Fig. 6C) at the end of the study. The study was performed without any apparent toxicities or changes in body weight (Fig. 6D).

To further elucidate the downstream signaling effects secondary to PAK4 inhibition, we performed transcriptomic and proteomic analysis of RH30 control- and KPT-9274–treated tumors. Analysis of the transcriptomic data revealed several key cancer-associated pathways to be modulated in KPT-9274–treated tumors. As shown through the volcano plot, with KPT-9274 treatment, 628 genes were significantly upregulated with a fold change of 1.25 or higher (red dots), while 65 genes were significantly downregulated showing fold change of 1/1.25 or lower (blue dots; Fig. 7A). Gene set enrichment analysis (GSEA) using the Kyoto Encyclopedia of Genes and Genomes, REACTOME, GO, and HALLMARK pathway collections revealed multiple oncogenic pathway were enriched with a negative normalized enrichment score (NES; Fig. 7B). Consistent with the mechanism of action of the KPT-9274 compound, we noted GTPase and RAS signaling pathways were significantly inhibited (Fig. 7C). Furthermore, two major pathways previously implicated in RMS biology, namely Notch and Hedgehog signaling, were also significantly downregulated (Fig. 7C). String database was used to analyze for protein–protein interaction (PPI) in Notch signaling, and PPI networks were visualized with Cytoscape software and color coded on the basis of upregulation (red) or downregulation (blue) in KPT-9274–treated samples as compared with control (Fig. 7D). Finally, a network of pathways enriched via hypergeometric distribution (P < 0.05) in the significantly differential genes was depicted using the Cytoscape software (Fig. 7E), with edges indicating the overlap between pathways.

We further assessed the effects of the KPT-9274 treatment on RH30 cells using RPPA profiling. Hierarchical clustering of the differentially expressed proteins revealed distinct sets of proteins being modulated between the two treatment groups (Supplementary Fig. S7A). Several of the oncogenic pathways determined using RNA sequencing were confirmed by the targeted proteomic profiling; specifically, we determined enrichment in the proteomic signature for cell cycle and migration, corroborating our functional in vitro studies (Supplementary Fig. S7B). As a surrogate marker of PAK4 inhibition, β-catenin, a known downstream target of PAK4, was found to be inhibited. Among other groups of differentially expressed proteins revealed by hierarchical clustering, we found proteins involved in metastatic progression, such as AXL and HIF1α, being significantly diminished in the KPT-9274–treated tumors. Several of these key modulations were validated by Western blot analysis from the KPT-9274–treated tumor lysates (Supplementary Fig. S7C).

While advancements in the genetic characterization have propelled our molecular insights into RMS, there are still significant deficiencies in the identification and implementation of effective therapeutic regimens for advanced stages of the disease (4).

Through extensive preclinical studies, our report provides evidence for the first time that, PAK4 is a viable and novel therapeutic target for the treatment of advanced RMS. Using multiple in vitro and in vivo cell lines and PDX models, in conjunction with molecular and pharmacologic targeting of PAK4, we have demonstrated that knocking down or inhibition of PAK4 can reduce cell proliferation in vitro and delay or regress tumor growth.

Blocking PAK4 had a profound effect not just on proliferation, but also on metastasis, as evidenced by diminished in vitro transwell assays and in vivo metastatic foci formation in mouse models. Because KPT-9274 is a dual inhibitor of PAK4 and NAMPT, and, therefore, we cannot exclude that some of the biological ramifications can be secondary to NAMPT inhibition, such as the effects seen on AKT inhibition could partially be contributed by NAMPT inhibition, with the latter being involved in energy metabolism. This is also relevant when we observe a G₁-phase arrest in the RMS cells with both drugs, but a G₂–M-phase arrest in RD cells only with PF-3758309 treatment. Although, a G₁-phase arrest induced by PAK4 inhibition has been reported in breast and ovarian cancer via inhibition of cyclin D1 (41, 42), through the PAK4/c-Src/EGFR pathway, or in pancreatic cancer (43) by downregulation cyclin D1 and upregulation of p21, we cannot rule out the possibility that some of these effects are secondary to, or a combination of, PAK4 and NAMPT inhibition.

However, we believe that the use of a complementary PAK4 inhibitor, PF-3758309, provides convincing evidence through both molecular perturbation that targeting PAK4 has significant antitumor and metastatic activity in RMS. Western blotting and immunostaining of our KPT-9274–treated tumors also confirm PAK4 inhibition (Fig. 5C and D). Furthermore, our comprehensive transcriptomic analysis of KPT-9274–treated tumors leads to molecular signatures consistent with targeting Ras/small GTPases, again signifying target engagement. Additional studies into the role of NAMPT in RMS biology, which is unrelated to the focus of this work, are warranted. Therefore, we believe that our comprehensive studies targeting PAK4 in multiple tumor model systems, including a PDX derived from a patient with relapsed fusion-negative RMS, pave the way to a novel therapeutic strategy for patients with high-risk RMS.

Mechanistically, our studies unveil several novel molecular insights into the role for PAK regulation in RMS. As mentioned above, besides modulation of Ras/GTPase downstream signaling, our comprehensive transcriptomic analysis identified pathways that are significantly modulated by PAK4, including Hedgehog, Notch, and PI3K signaling, which corroborate previous studies that characterized critical roles for these pathways in RMS tumorigenesis (44, 45). Studies investigating the mechanism by which PAK4 regulates, or cross-talks with these pathways are in preparation and will provide additional insights into the interplay between these vital signaling cascades.

Interestingly, besides alterations in intrinsic signaling perturbations, our transcriptomic analysis of KPT-9274–treated tumors provided preliminary evidence for extrinsic tumor microenvironmental regulation dictated by PAK4 signaling, as we identified enhanced immune modulation frequently associated with antitumor activity. Several tumor immunity–related pathways were enriched with a positive NES with KPT-9274 treatment (Supplementary Fig. S8A). Specifically, we observed a significant upregulation in adaptive immunity, IFNα and -γ response, and pathways in antigen processing in KPT-9274–treated tumors (Supplementary Fig. S8B and S8C).

Although validating this observation is beyond the scope of this article, these findings are supported by the recent publication identifying PAK4 inhibition can alter the tumor immune microenvironment and further enhance anti-PD-1 efficacy and increase T-cell infiltration in syngeneic melanoma murine models (46). However, the use of immunodeficient murine models limits our interpretation of the significance of these molecular alterations. We plan to utilize genetically engineered mouse models for RMS, and syngeneic murine RMS cell lines, for future studies that can be used to further evaluate the role of PAK4 in the regulation of the tumor immune microenvironment and immune evasion.

Finally, besides observing the inhibitory effects on oncogenic pathways and activation of immune-stimulatory signatures, we were able to evaluate the association between the KPT-9274 signature and survival in The Cancer Genome Atlas-SARC patient cohort. Although this cohort does not include RMS patient information, the association was significant (P < 0.05), and patients with transcriptome profiles closer to those of the KPT-9274–treated tumors had a significantly better outcome than those with profiles closer to the RH30 control–treated cells (Supplementary Fig. S8D).

In addition, GO analysis of the proteogenomic data indicates signatures associated with metastatic progression and myogenic differentiation, through AXL and HIF1α. Aside from its role in microenvironmental modulation in the context of tumor hypoxia, HIF1α has also been reported to be critical in driving therapy resistance and other metastatic properties (47, 48). AXL is a RTK, found to be overexpressed in many cancer types and implicated for its role in proliferation and invasion and in epithelial–mesenchymal transition processes (49). AXL has been identified as a “semidirect” regulator that can negatively regulate muscle genes in the presence of a sensitizer, and, thus, its KD can promote myogenic differentiation (50). The ligand to AXL, growth arrest–specific 6 (GAS6), has been reported to play a role in tumor dormancy in the bone marrow (51) by downregulating AXL during ambient physiologic conditions, but not under hypoxic conditions, thereby promoting metastasis. In another study in clear-cell renal cell carcinoma, AXL was found to be activated under hypoxia by both HIF1 and HIF2 binding to the hypoxia-response element in the AXL proximal promoter (52). Thus, the GAS6/AXL signaling pathway is a critical player in metastasis and our RPPA data show PAK4 is modulating this pathway. Further mechanistic studies are warranted to gain insights into the regulation of hypoxia-mediated functions by PAK4 in RMS.

Taken together, we are the first to demonstrate that PAK4 is a regulator of critical intrinsic, and potentially extrinsic mechanisms of RMS development and progression, and is a viable therapeutic target for the treatment of advanced stages of RMS.

N. Rainusso reports grants from Cancer Prevention and Research Institute of Texas and Snowdrop Foundation during the conduct of the study, and Cancer Prevention and Research Institute of Texas, Snowdrop Foundation, and St. Baldrick's Foundation outside the submitted work. Y. Landesman reports personal fees from Karyopharm Therapeutics during the conduct of the study. T. Unger reports personal fees from Karyopharm Therapeutics during the conduct of the study and outside the submitted work. No disclosures were reported by the other authors.

A. Dasgupta: Conceptualization, formal analysis, writing-original draft, writing-review and editing. L. Sierra: Conceptualization, investigation. S.V. Tsang: Formal analysis, investigation. L. Kurenbekova: Investigation. T. Patel: Formal analysis. K. Rajapakshe: Formal analysis. R.L. Shuck: Resources, investigation. N. Rainusso: Resources. Y. Landesman: Resources. T. Unger: Resources. C. Coarfa: Formal analysis. J.T. Yustein: Conceptualization, funding acquisition, writing-original draft, writing-review and editing.

We would like to acknowledge the Advanced Technologies Core services at the Baylor College of Medicine. This work was funded by NIH R21 CA234665, Alex's Lemonade Stand Foundation Innovation Award, and The Faris D. Virani Ewing Sarcoma Center (to J.T. Yustein). S.V. Tsang was partially supported by Cancer Prevention Institute of Texas (CPRIT) RP160283, K. Rajapakshe and C. Coarfa were partially supported by (CPRIT) RP170005, NIH P30 shared resource grant CA125123, and NIEHS P30 Center grant 1P30ES030285; CPRIT Core Facility Support Award (CPRIT-RP180672), the NIH (CA125123 and RR024574 to the Cytometry and Cell Sorting Core at Baylor College of Medicine); and Core Facility Award (CPRIT-RP170005 to Proteomics and Metabolomics Core at Baylor College of Medicine).

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.