Anti-tumor Effects of the eIF4A Inhibitor Didesmethylrocaglamide and Its Derivatives in Human and Canine Osteosarcomas

Inhibition of translation initiation using eIF4A inhibitors like (−)-didesmethylrocaglamide [(−)-DDR] and (−)-rocaglamide [(−)-Roc] is a potential cancer treatment strategy as they simultaneously diminish multiple oncogenic drivers. We showed that human and dog osteosarcoma cells expressed high levels of eIF4A1/2, particularly eIF4A2. Genetic depletion of eIF4A1 and/or 2 slowed osteosarcoma cell growth. To advance preclinical development of eIF4A inhibitors, we demonstrated the importance of (−)-chirality in DDR for growth-inhibitory activity. Bromination of DDR at carbon-5 abolished growth-inhibitory activity, while acetylating DDR at carbon-1 was tolerated. Like DDR and Roc, DDR-acetate increased the γH2A.X levels and induced G2/M arrest and apoptosis. Consistent with translation inhibition, these rocaglates decreased the levels of several mitogenic kinases, the STAT3 transcription factor, and the stress-activated protein kinase p38. However, phosphorylated p38 was greatly enhanced in treated cells, suggesting activation of stress response pathways. RNA sequencing identified RHOB as a top upregulated gene in both DDR- and Roc-treated osteosarcoma cells, but the Rho inhibitor Rhosin did not enhance the growth-inhibitory activity of (−)-DDR or (−)-Roc. Nonetheless, these rocaglates potently suppressed tumor growth in a canine osteosarcoma patient-derived xenograft model. These results suggest that these eIF4A inhibitors can be leveraged to treat both human and dog osteosarcomas.


Introduction
Osteosarcomas are highly aggressive bone malignancies typically found in patients younger than 20 years old 1 .These cancers tend to develop in the long bones of the axial skeleton.For decades, the standard-of-care has been extensive surgery, possibly including amputation of the affected limb, along with multi-agent cytotoxic chemotherapy.Although the cure rate for localized disease is ~ 70%, this cancer recurs in about one-third of patients.In addition, metastatic disease frequently occurs and usually targets the lungs.Patients with recurrent or metastatic disease have a dismal prognosis with < 20% survival beyond three years.Identi cation of a more effective medical therapy for osteosarcomas is direly needed.
While osteosarcomas exhibit low mutational burden 2 , they frequently have highly individualized copynumber alterations and structural rearrangements 3 .Due to complex genomic aberrations, osteosarcomas show dysregulation of multiple mitogenic signaling pathways, which are potential therapeutic targets 4 .However, clinical trials of therapies inhibiting individual targets have been disappointing, showing at best partial e cacy.Also, osteosarcomas harbor a dynamic tumor microenvironment (TME), which plays an important role in driving tumor growth and promoting treatment resistance 5 .These ndings highlight the likely need to block simultaneously multiple signaling pathways to effectively treat these aggressive bone cancers.
Although osteosarcomas are the most common primary bone malignancies, they are relatively rare in humans with an incidence of less than 1000 cases per year in the United States.In contrast, osteosarcomas are more frequently found in dogs, particularly in certain large breeds 6 .The incidence of osteosarcomas in dogs is over 10-fold higher than in humans, with > 10,000 new canine patients each year.Remarkably, dog osteosarcomas closely recapitulate the human disease in many key features, including histopathology, clinical course, molecular genetics, and deregulated signaling pathways 7 .These large animals also have intact immune systems, and the cellular composition in their TME resembles that of humans 8 .For these reasons, dogs have been regarded as a valuable large animal model for evaluating drug toxicology and e cacy against osteosarcomas in immune-competent conditions.
Protein translation has long been considered as a prime vulnerability in cancers, as several oncogenic drivers are regulated at the level of translation initiation 9 .In nontumorigenic cells, mRNA translation is stringently controlled by the eukaryotic initiation factor 4F (eIF4F) complex, which consists of eIF4A, eIF4E, and eIF4G.As an RNA helicase, eIF4A is critical in unwinding the complex purine-rich 5′untranslated regions (UTRs) found in the mRNAs that encode growth-promoting kinases, cyclins, transcription factors, and epigenetic regulators 10 .Included among these oncogenic kinases are members of the growth-and survival-promoting PI3K/AKT/mTOR and Ras/RAF/MEK/ERK signaling pathways that are frequently deregulated in cancer, including osteosarcoma 3,4 .These signal transduction cascades converge upon eIF4F and facilitate its assembly and activity, such as promoting the helicase processivity of eIF4A.In cancer cells, this process can establish positive feedback loops whereby oncogenic kinases enhance their own protein synthesis.Thus, blocking eIF4A may simultaneously diminish multiple oncogenic signaling molecules.Indeed, we previously showed that the natural eIF4A inhibitors (-)-rocaglamide [(-)-Roc or RocA] and (-)-didesmethylrocaglamide [(-)-DDR] concurrently reduced the levels of IGF-1R, AKT, ERK1/2, and survivin in osteosarcoma cells and potently suppressed the growth of human osteosarcoma patient-derived xenografts (PDXs) 11 .
Roc and DDR belong to a family of compounds called cyclopenta[b]benzofurans (rocaglates or avaglines).Roc and other bioactive rocaglates selectively bind to eIF4A1/2 12 .Mutational studies in yeast and mouse eIF4A orthologs have identi ed amino acid substitutions in Phe163 and Ile199 that confer resistance to rocaglates 13,14 .The crystal structure of Roc complexed to eIF4A1 and purine-rich RNA con rmed that these residues are crucial for forming a rocaglate-binding pocket on eIF4A 15 .The high-a nity interactions among Roc, eIF4A, and polypurine RNA sequences effectively prevent intact preinitiation complexes from successfully translating selected mRNAs.Importantly, the Phe163Leu and Ile199Met substitutions found in the eIF4A homologs of rocaglate-producing Aglaia plants prevent Roc toxicity by abolishing Roc-induced polypurine RNA clamping and translation repression.An Ophiocordyceps fungus species parasitic on Aglaia plants has a glycine substitution at His153 (equivalent to Phe163 in human eIF4A1) that likewise disrupts Roc binding 16 .Collectively, these studies highlight the exquisite speci city of rocaglates as eIF4A inhibitors when they bind to the highlystructured, purine-rich 5′-UTRs of selective mRNAs.
Structure-activity relationship (SAR) analyses from comparing several natural and synthetic rocaglates have identi ed functional groups important for eIF4A inhibition and positions on the rocaglate scaffold that may be modi ed to improve drug-like properties 11,[17][18][19] .By side-by-side comparison of 10 naturallyoccurring rocaglates, we found (-)-DDR to be the most potent in suppressing the growth of multiple types of sarcoma cells 11 .Here, we showed that human and dog osteosarcoma cells expressed high levels of eIF4A1/2.Genetic depletion of eIF4A1 and/or 2 slowed osteosarcoma cell growth.To advance preclinical development of eIF4A inhibitors, we synthesized racemic (±)-DDR, separated the enantiomers of DDR, and found (-)-DDR to be ≥ 500-fold more active than (+)-DDR in inhibiting the growth of a panel of human and dog osteosarcoma cell lines.We also synthesized two DDR derivatives, (±)-bromo-DDR and (±)-DDR-acetate (Supplementary Figure S1) and found that bromination of DDR at carbon-5 to abolish growth-inhibitory activity, while acetylating DDR at carbon-1 is tolerated.Like DDR and Roc, DDR-acetate promoted G 2 /M arrest and apoptosis in part by decreasing the expression of multiple mitogenic kinases, the STAT3 transcription factor, and the stress-activated protein kinase (SAPK) p38, while enhancing the phosphorylated p38 levels.Since rocaglates can decrease the levels of certain transcription factors 10 , we performed RNA-sequencing (RNA-seq) analysis and identi ed RHOB as a common top upregulated gene, but the Rho inhibitor Rhosin did not enhance the growth-inhibitory activity of (-)-DDR or (-)-Roc.
Finally, we showed that these rocaglates potently suppressed the growth of canine osteosarcoma PDXs.

Results
Human bone sarcoma cells express high levels of eIF4F components, particularly eIF4A2.We previously demonstrated that malignant peripheral nerve sheath tumor (MPNST), another aggressive type of sarcoma, overexpress components of the eIF4F complex 20 .To determine if this feature is characteristic of other types of sarcomas, we assessed the protein levels of both eIF4A isoforms 1 and 2 and eIF4E in four human and four canine osteosarcoma and two human Ewing sarcoma cell lines.Compared to MSCs, which have been postulated to be the cell-of-origin for bone sarcomas 21 , all human osteosarcoma and Ewing sarcoma cell lines examined overexpressed both eIF4A1 and 2 and eIF4E (Fig. 1A).In particular, the amounts of eIF4A2 were substantially higher in these sarcoma cells relative to MSCs.Similarly, all four dog osteosarcoma cell lines studied expressed high levels of eIF4A1/2 and eIF4E.
These results indicate that osteosarcoma and Ewing sarcoma cells frequently overexpress eIF4F components.
Silencing of EIF4A1 or EIF4A2 reduces osteosarcoma cell growth.To examine whether eIF4A1/2 are important for osteosarcoma cell proliferation, we performed shRNA-mediated silencing in three human osteosarcoma cell lines.EIF4A1 knockdown in Saos2, 143B, and OS17 cells reduced the eIF4A1 protein levels by 78, 65, and 47%, respectively (Fig. 1B).As was previously reported by us and others 20,22 , depletion of eIF4A1 in osteosarcoma cells caused a feedback-mediated compensatory increase in eIF4A2 protein levels, particularly in 143B and OS17 cells.EIF4A2 knockdown in the same three osteosarcoma cell lines caused a 94-99% reduction of eIF4A2 protein levels.Also, depletion of eIF4A2 was associated with increased eIF4A1 protein levels in OS17 and 143B cells.However, simultaneous knockdown of both EIF4A1 and EIF4A2 resulted in higher expression levels of eIF4A1/2 proteins in all three osteosarcoma cell lines, compared to EIF4A1 or EIF4A2 knockdown alone.It is possible that this effect may be due to feedback compensation and warrants further investigation.
Importantly, silencing EIF4A1 or EIF4A2 slowed the growth of all three osteosarcoma cell lines, and the amount of growth inhibition correlated with the magnitude in the decrease of target protein levels (Fig. 1B).Co-silencing of both EIF4A1 and EIF4A2 also resulted in reduced cell growth in all three osteosarcoma cells, but the effect was slightly less than that by EIF4A1 silencing alone.Additionally, the effect of growth inhibition by combined EIF4A1 and EIF4A2 knockdown was less than that by EIF4A2 knockdown in 143B cells.These attenuated antiproliferative effects may be due to less reduction in the levels of eIF4A1 and eIF4A2 proteins in double-knockdown cells.Nonetheless, the results indicate that both eIF4A1 and eIF4A2 are important for osteosarcoma cell proliferation.
To examine the effects of modifying the C1 and C5 positions, we synthesized DDR-acetate, which is acetylated at C1, and bromo-DDR, which is brominated at C5 of the A-ring (Supplementary Figure S1).In various human and dog osteosarcoma cell lines tested, (±)-DDR-acetate had an IC 50 value (20 ~ 35 nM) roughly 2-3-fold higher than that of (±)-DDR (10-12 nM) (Figs.2A,C).In contrast, (±)-bromo-DDR exhibited little growth-inhibitory activity, with an IC 50 value ranging from 8.5 µM in the dog PDX-derived K9-OS6 cell line to > 20 µM in human Saos2 cells.These results indicate that acetylating the hydroxyl group at C1 of DDR is well-tolerated, while modifying the C5 position on A-ring aryl group substantially reduces antiproliferative activity.
(±)-DDR-acetate, like (-)-DDR and (-)-Roc, induces G 2 /M arrest and apoptosis in human and canine osteosarcoma cells.To examine the mechanisms of growth inhibition, we rst compared the effects of (±)-DDR-acetate, (±)-DDR, and (-)-DDR on cell cycle distribution of human and dog osteosarcoma cells.
An increase in the G 2 /M population was observed in MG-63 cells treated with 1x and 2x IC 50 concentrations of all three rocaglates for three or ve days (Figs.3A,B).Increased sub-G 1 fractions were also observed in all three rocaglate-treated MG-63 cells, suggesting induction of cell death.Similarly, we detected prominent G 2 /M arrest and an increase in the sub-G 1 fraction in OS17 cells treated with all three rocaglates (Fig. 3C).Like in human osteosarcoma cells, all three rocaglates also increased the G 2 /M and sub-G 1 populations in K9-OS6 cells treated for two days (Fig. 3D).
To determine whether induction of cell death by DDR-acetate, (±)-DDR, and (-)-DDR is due to apoptosis, we performed Incucyte® live cell imaging combining with caspase-3/7 cleavage assay in K9-OS6 cells treated with each rocaglate.Growth inhibition was observed with concomitantly increased caspase-3/7 cleavage in cells treated with all three rocaglates, when compared to DMSO-treated controls (Fig. 4A).Signi cant increases in caspase cleavage were seen, starting from ~ 12 hours of treatment and becoming more pronounced thereafter.Like (-)-DDR, (-)-Roc also inhibited the growth of K9-OS6 cells while simultaneously enhancing caspase-3/7 cleavage (Fig. 4B).These results were corroborated by ow cytometry analysis showing prominent increases in the G 2 /M and sub-G 1 populations in DDR or Roctreated cells (Supplementary Figure S2).
Together, these results indicate that apoptosis induction contributes to the antiproliferative effects of DDR-acetate, (±)-DDR, (-)-DDR, and (-)-Roc in human and dog osteosarcoma cells.
(±)-DDR-acetate also suppress multiple signaling molecules important for osteosarcoma growth while activating cellular stress and DNA damage response pathways.We previously showed that (-)-Roc and (-)-DDR inhibit the expression of several mitogenic kinases in sarcoma cells 11 .To examine whether the growth-inhibitory activity of (±)-DDR-acetate acted through similar molecular mechanisms, we performed Western blot analysis on human osteosarcoma cells treated with (±)-DDR-acetate, (±)-DDR, or (-)-DDR.All three rocaglates markedly diminished the levels of IGF-1R and AKT (total and phospho-AKT [p-AKT]) in both MG-63 and OS17 cells (Fig. 5).Also, reduced expression of PDGFRβ, p38 SAPK, FAK, and the transcription factor STAT3 (total and p-STAT3) was observed in treated MG-63 cells.In addition, these rocaglates decreased the MET levels in OS17 cells.Intriguingly, while p38 expression was diminished, phosphorylated p38 was notably enhanced, and the γH2A.X levels were elevated in treated MG-63 and OS17 cells, indicating activation of genotoxic stress signaling and DNA damage responses.In line with induction of G 2 /M arrest, we detected abundant p-histone H3, a G 2 /M marker, particularly in cells treated with 1x IC 50 .Further, all three rocaglates increased cleavage of caspase-3 and PARP in both osteosarcoma cell lines, signifying enhanced apoptosis.Notably, the protein levels of housekeeping proteins GAPDH and tubulin, as well as eIF4A1, and eIF4A2 were unaffected by these rocaglate treatments.These results reinforce that eIF4A inhibition does not indiscriminately inhibit translation initiation but selectively affects transcripts encoding mitogenic proteins.
Overall, these data indicate that (±)-DDR-acetate, (±)-DDR, (-)-DDR, and (-)-Roc suppress human and dog osteosarcoma cell growth by decreasing the levels of key growth-promoting molecules and inducing cell stress and DNA damage responses, ultimately leading to apoptosis.
(-)-DDR and (-)-Roc treatment upregulate genes and pathways linked to in ammatory responses, cell stress and death, and growth-related signaling.Prior studies demonstrate that the transcripts encoding transcription factors like MYC and JUN have long purine-rich 5'-UTRs and are sensitive to inhibition by Roc and related rocaglates 23,24 .As we found that DDR and Roc also decreased the STAT3 levels in osteosarcoma cells (Fig. 5), we investigated the effects of these rocaglates on the transcriptomes of MG-63 and Saos2 cells treated with (-)-DDR or (-)-Roc for one day by RNA-seq.Comparing the set of protein-coding DEGs in rocaglate vs. DMSO-treated cells, we identi ed 173 differentially-expressed transcripts in both rocaglate-treated MG-63 and Saos2 cells (Fig. 8A and Supplementary Table S1).Of these shared transcripts, 96% were similarly up-or downregulated by (-)-DDR and (-)-Roc treatment of MG-63 and Saos2 cells, including 121 upregulated (~ 70%) and 45 downregulated (~ 26%) DEGs (Supplementary Figure S4).Of these 173 shared DEGs, many were also found among the top-50 proteincoding DEGs in both osteosarcoma cell lines treated with either (-)-DDR or (-)-Roc (compare Fig. 8B with Supplementary Table S1).These include upregulation of transcripts encoding the RhoB small GTPase, the dual-phosphatase DUSP16 which dephosphorylates and inactivates ERK and JNK, and the transcription factors JUN, FOSL2, MAFB, and HEY1; several of these proteins are important for osteosarcoma cell growth and differentiation.
Although (-)-DDR or (-)-Roc potently suppressed osteosarcoma growth, they did not eliminate all tumor cells (Fig. 7).Since RHOB was identi ed among the top ten upregulated DEGs in both MG-63 and Saos2 treated with either rocaglate (Figs.8B,C), we tested whether increased RHOB expression provided any survival bene ts.All four human osteosarcoma cell lines were tested with (-)-DDR and (-)-Roc in a dosematrix combination with the Rho inhibitor Rhosin 25 .However, Rhosin alone had little or no growthinhibitory activity even at 50 µM (Supplementary Figure S7), and combining Rhosin with (-)-DDR or (-)-Roc did not give rise to enhanced growth inhibition according to Bliss synergy scores.These results suggest that RHOB upregulation in (-)-DDR or (-)-Roc treated osteosarcoma cells does not confer a survival bene t.

Discussion
With the standard-of-care that has not changed over the last four decades, patients with osteosarcoma have not yet bene ted from recent advances in genomic pro ling and therapeutic development.Although patients with localized disease have a 70% ve-year survival rate, this comes at the cost of severely decreased quality-of-life.Radical limb amputation is frequently used, and the harsh multi-cytotoxic chemotherapy incurs acute and chronic systemic toxicities 26,27 .Further, patients suffering relapse and those with metastatic disease have an especially bleak prognosis.Efforts to identify effective drugs have been impeded by both intra-and inter-tumor heterogeneity of this aggressive bone cancer 1,3,4 .A therapy that simultaneously inhibits multiple oncogenic drivers is likely to be needed to eradicate osteosarcoma.
To sustain uncontrolled growth, cancer cells often exhibit enhanced protein translation by upregulation of the translation machinery.Previously, we found that MPNST, another type of sarcoma, frequently overexpresses the three eIF4F components 20 .Along with this line, we showed that both human and dog osteosarcoma cells, as well as Ewing sarcoma cells expressed high levels of eIF4A1, eIF4A2, and eIF4E (Fig. 1).Also, silencing EIF4A1 or EIF4A2 impeded osteosarcoma cell proliferation.However, the role of eIF4A2 in cell proliferation may depend on the cell type.A previous study showed that while EIF4A1 knockdown reduces the growth of HeLa cells, depletion of eIF4A2 has no effects 28 .On the contrary, silencing EIF4A2 in embryonic stem cells compromised their proliferation and disrupted the stem cell maintenance program 29 .Also, Eif4a2 knockout is embryonically lethal in mice.Our nding that silencing EIF4A1 and/or EIF4A2 impeded the growth of multiple osteosarcoma cell lines indicate that both eIF4A isoforms are required for osteosarcoma cell proliferation.Curiously, genetic suppression of EIF4A1 resulted in increased eIF4A2 levels in 143B and OS17 cells, likely due to increased transcription of EIF4A2 as previously reported 22 .Likewise, we observed modestly increased eIF4A1 expression in osteosarcoma cells after EIF4A2 silencing, suggesting potential feedback interactions between EIF4A1 and EIF4A2.Together, out results further emphasize the importance of eIF4A in sarcoma growth.
Previous studies have established the speci city of Roc and related rocaglates as eIF4A inhibitors using biological and biochemical assays and X-ray crystallography 12,14,15 .To further understand the SAR of rocaglates, we con rmed the importance of chirality, as the (+)-DDR enantiomer is 500-to-800-fold less bioactive than (-)-DDR in various osteosarcoma cell lines (Fig. 2).Consistently, racemic (±)-DDR was generally half as potent at reducing osteosarcoma cell proliferation as (-)-DDR.We also observed the importance of the C5 position of the A-ring in rocaglate as bromination of this position, as in (±)-bromo-DDR, almost abolished its antiproliferative activity.Based on the crystal structure of (-)-Roc bound to eIF4A1 and polypurine RNA 15 , we posit that the bulky C5-bromine prevents the A-ring from stacking in parallel with the polypurine tract.Interestingly, functionalizing the C1 position with an acetyl group, as in (±)-DDR-acetate, was well-tolerated, with only mildly decreased growth-inhibitory activity compared to (±)-DDR.An identical substitution in a natural rocaglamide analog, (-)-Roc-AB (1-O-acetylrocaglamide), was reported to show a similar IC 50 to (-)-Roc and inhibit leukemia cell growth 30 .However, azarocaglamide with a C1-substituted aminomethyl group oriented in syn to the 8b-hydroxy group exhibited 30-fold greater antiproliferative activity than the stereoisomer with the anti-con guration 31,32 .Curiously, our DDR-acetate has the same 1,8b-anti-con gured acetyl group as that in (-)-Roc-AB.Thus, it will be worthwhile to see if functionalizing the C1 position with an acetyl or aminomethyl group with the 1,8bsyn-con guration improves the antiproliferative effects of DDR.
Mechanistically, (±)-DDR-acetate, like (-)-DDR and (-)-Roc, diminished the expression of several receptor tyrosine kinases (RTKs) important for osteosarcoma growth, including IGF-1R, PDGFRβ, EGFR, and MET, in human and dog osteosarcoma cells (Figs. 5 and 6).Additionally, the levels of other mitogenic kinase, such as AKT and FAK, and the transcription factor STAT3, which is important for osteosarcoma cell growth, metabolism, survival, and metastatic behavior 33 , were greatly reduced.Moreover, these rocaglates induced G 2 /M arrest as evidenced by increased G 2 /M population and phospho-histone H3 expression in treated osteosarcoma cells.Also, they elevated the levels of γH2A.X and cleaved caspase 3 and PARP (Figs. 3-6), suggestive of induction of DNA damage responses and apoptosis.
Interestingly, treatment of human and dog osteosarcoma cells with (-)-Roc-, (-)-DDR-, (±)-DDR, and (±)-DDR-acetate increased phosphorylated and activated p38 SAPK (Figs. 5 and 6), despite marked declines in the total p38 protein level, consistent with translation inhibition.Presently, the mechanism by which rocaglate treatments activate p38 is not understood.As a mediator of intrinsic apoptosis, p38 is activated by a variety of stimuli, including oxidative and genotoxic stressors 34 .Activated p38 can phosphorylate transcription factors, such as p53, and initiate caspase activation to execute apoptosis.Our GSEA of RNA-seq data substantiates these ndings, with several signi cantly upregulated pathways including those associated with UV responses, hypoxia, p53 pathway, mitotic spindle assembly, G 2 /M checkpoint, and apoptosis in both (-)-DDR and (-)-Roc treated MG-63 and Saos2 cells (Supplementary Figure S5 and Supplementary Table S2).Thus, it is possible that rocaglate treatments induce these stressors, leading to p38 activation.Curiously, we also identi ed several pathways that were associated with in ammatory responses and cytokine signaling, especially activation of type I interferon-stimulated genes (ISGs), among the common top upregulated pathways.Induction of ISGs is linked to cell stress responses, but they may also have effects on the tumor immune microenvironment when osteosarcomas are grown in immune-competent conditions 35 .
Importantly, we showed that (-)-DDR and (-)-Roc potently suppress tumor growth in a canine osteosarcoma PDX model (Fig. 7).Treated tumors exhibited greatly reduced cellularity with degenerative changes and, consistent with in vitro ndings, had increased phospho-p38 expression and high apoptotic labelling.Together with the anti-tumor activity of (-)-Roc in human sarcoma PDX models 11 , these results indicate that these rocaglates, as eIF4A inhibitors, can be used to treat both human and dog osteosarcomas.It should be noted that a synthetic derivative of Roc, eFT-226 (Zotati n) 36,37 , has demonstrated safety and tolerability and is now in phase 2 clinical trial to treat patients with K-Ras or RTK-driven advanced breast cancer and non-small cell lung carcinoma (ClinicalTrials.govidenti er NCT04092673).Thus, a clinical trial in canine patients with spontaneous osteosarcomas and other softtissue sarcomas is warranted.
However, (-)-DDR or (-)-Roc did not eliminate all osteosarcoma cells (Fig. 7), suggesting possible emergence of compensatory survival mechanisms that would require a combination therapy.Our DEG analysis identi ed RHOB, which encodes a small GTPase associated with cell motility, membrane tra cking, and cell proliferation 38 , within the top 10 upregulated genes in both (-)-DDR-and (-)-Roctreated MG-63 and Saos2 cells (Fig. 8C).Despite this nding, we did not observe enhanced growth suppression in osteosarcoma cells after combined inhibition of eIF4A and Rho (Supplementary Figure S7).Previously by CRISPR screening, a novel interaction between rocaglates and the KEAP1-CUL3-NRF2 axis was identi ed 39,40 .Under normal conditions, KEAP1, which binds to CUL3 and NRF2, promotes NRF2 ubiquitination and proteasomal degradation 41 .Upon exposure to reactive oxygen species or other stressors, KEAP1 undergoes conformational changes, liberating NRF2 from the complex.Free NRF2 translocates to the nucleus where it heterodimerizes with MAF proteins and binds to antioxidant responsive elements to activate transcription of antioxidant and metabolic genes.Some of these NRF2activated gene products can enhance translation of eIF4A-dependent transcripts.Intriguingly, we detected MAFB as one of the common top upregulated protein-coding genes, while the RNA levels of NFE2L2, which encodes NRF2, were only slightly elevated in (-)-DDR-and (-)-Roc-treated osteosarcoma cells (Fig. 8 and Supplementary Tables S1 and S3).It will be interesting to see if MAFB provides any survival bene ts to osteosarcoma cells.Furthermore, we identi ed the NOTCH, Hedgehog, TGFβ, and β-catenin pathways among the common top upregulated pathways in (-)-DDR-and (-)-Roc-treated MG-63 and Saos2 cells.These signaling pathways have been reported to drive osteosarcoma cell proliferation, and high expression of proteins in these pathways correlates with tumor aggressiveness and patient survival 42 .Also, signaling from TGFβ receptor family members can enhance osteosarcoma malignant behavior, with tumors expressing high TGFB1 levels being more likely to respond poorly to chemotherapy 43 .Among the TGFβ signaling gene set, the signal transducers for TGFβ receptors SMAD1, SMAD3, and SMAD6 and the bone-morphogenetic protein receptor 2 (BMPR2) were highly upregulated in rocaglate-treated osteosarcoma cells (Supplementary Table S2).Studies have shown that the SMAD-dependent TGFβ and BMP signaling is important for bone differentiation and formation.There are cross-talks between TGFβ/BMP and WNT signaling and between WNT and NOTCH signaling 44 .Thus, it is tempting to speculate that upregulation of these pathways may provide compensatory survival bene ts in rocaglate-treated osteosarcoma cells.
We are presently testing whether combining the inhibitors of these pathways synergizes with (-)-DDR and (-)-Roc to kill osteosarcoma cells.
In summary, our data established the importance of the (-) chiral con guration of DDR for growth suppression of human and dog osteosarcoma cells.We also demonstrated that the C1, but not C5, position of DDR may be modi ed without signi cantly affecting its antiproliferative activity.The ability of this class of translation inhibitors, which simultaneously diminishes multiple key mitogenic molecules and induces DNA damage response, G 2 /M arrest, and apoptosis, warrants further consideration as potential therapies for cancers that lack a de ned oncogenic driver, such as osteosarcoma.

Methods
Compounds.Isolation of (-)-DDR, (-)-Roc, and (-)-methyl rocaglate from the tropical Aglaia plants as part of a multi-institutional collaborative project on the discovery of new antineoplastic natural compounds was as previously described 11,45 .(±)-DDR was synthesized from (±)-methyl rocaglate and separated into its (+)-and (-)-enantiomers on a ChiralPak IB3 and Diacel Chiralcel OD-H columns (Supplementary Figure S1A).Bromo-DDR was generated from (±)-methyl rocaglate, and (±)-DDR-acetate was derived by acetylating (±)-DDR (Supplementary Figure S1B).The details on chemical synthesis and puri cation of compounds and the veri cation of their chemical structures and absolute con gurations by nuclear magnetic resonance spectroscopy and mass spectrometry are provided in Supplementary Methods.Rhosin, a small molecule inhibitor targeting the RhoA subfamily of Rho-guanosine triphosphatases (GTPases) 25 , was purchased from MilliporeSigma.For cell culture studies, all compounds were prepared as 10 mM (rocaglates) or 25 mM (Rhosin) stock solutions in DMSO and stored at -20°C.For animal studies and some in vitro work, (-)-Roc (NSC326408) and (-)-DDR (NSC705956) were chemically synthesized and provided by the NCI Experimental Therapeutics (NExT) Program.
Cell lines, cell proliferation assay, and ow cytometry.Human osteosarcoma cells MG-63, OS17, Saos2, and 143B, human Ewing sarcoma cells TC32 and A673, and canine osteosarcoma cells Abrams, OSA8, and OSA16 were previously described 11,46 .Human bone marrow-derived mesenchymal stem cells (MSCs) were kindly provided by Nilay Shah of Nationwide Children's Hospital.Also, we generated the K9-OS6 cell line from an OSU-K9-OS6 dog osteosarcoma PDX tumor by serial passaging.All cells were grown in Dulbecco's Modi ed Eagle's medium (MilliporeSigma) supplemented with 10% fetal bovine serum (R&D Systems).For single-drug dose-response analysis, osteosarcoma cells seeded in 96-well plates (Sarstedt) were treated for 3 days with compounds added as 9-point, 2-fold serial dilutions.Percent cell viability was assessed by adding resazurin and measuring metabolic conversion to uorescent resoru n, then averaging uorescence values of drug-treated wells and normalizing to the DMSO controls set as 100%.Dose-response curves were plotted on GraphPad Prism v9 and the mean absolute 50% growth-inhibitory concentration (IC 50 ) values were estimated.For drug combination testing, osteosarcoma cells seeded in 96-well plates were treated for 3 days with (-)-Roc or (-)-DDR arrayed in combination with Rhosin.Viability was assessed by resazurin assays and Bliss synergy scores for each combination calculated using the SynergyFinder v3.0 web-based application (https://synergy nder.mm./).For cell cycle analysis, propidium iodide-labeled cells were run on a LSR II ow cytometer (BD Biosciences), followed by gating single cells 47 .The cell cycle distributions were calculated using FlowJo v10 (TreeStar).Cell cycle histograms were deconvoluted using the Dean-Jett-Fox algorithm, and the percentages of cell populations in sub-G 1 , G 1 , S, and G 2 /M calculated.
Lentiviral-mediated knockdown by short-hairpin RNA (shRNA).Osteosarcoma cells were seeded at 9,000 cells per well in 6-well plates.The next day, cells in duplicate wells were transduced with 10 multiplicities of infection (MOI) of lentiviruses expressing shRNAs targeting eIF4A1 (TRCN0000288729), eIF4A2 (TRCN0000051869), or both eIF4A1 and eIF4A2 in fresh growth medium containing 8 µg/mL of polybrene (MilliporeSigma).Lentiviruses expressing a nontargeting shRNA sequence (SHC002V) were used as a negative control.Transduced cells were selected with puromycin (2 µg/mL), and when cells expressing the nontargeting shRNA reached con uence, all wells in the experiment were trypsinized and counted using a hemocytometer.Relative cell numbers were calculated as a percentage of cells transduced with nontargeting shRNA normalized to 100%.Then, cell lysates were prepared from transduced wells and analyzed for protein expression of the eIF4F components (see below).

Figure 1
Figures