Inhibiting Translation Elongation with SVC112 Suppresses Cancer Stem Cells and Inhibits Growth in Head and Neck Squamous Carcinoma

Head and neck squamous cell carcinoma (HNSCC) incidence has increased due to the rise in HPV-related tumors (1, 2), and radiotherapy remains a mainstay of both first-line and postsurgical management. Given current treatment failures and toxicity, more effective therapies are needed. Cancer stem cells (CSC) sit atop the hierarchical tumor model and are the candidate population that sustains tumor growth, are resistant to therapy, and drive repopulation after treatment (3). We recently defined CSCs across HNSCC subtypes (4), and a concerted effort is underway to identify and target conserved cellular processes and signaling pathways that maintain CSCs (5, 6). Tissue SCs are maintained through tight regulation of protein synthesis (7), and there is growing evidence that translation regulates the CSC phenotype (8). Like normal SCs, the CSC phenotype require a discrete set of factors (e.g., Myc, Sox2; ref. 9) that are challenging to target (10).

The Myc protein has a short half-life (20–30 minutes; ref. 11), suggesting Myc can be targeted, and depleted quickly, through the inhibition of de novo protein synthesis (12). We recently showed that the Sox2 pluripotency factor is regulated at the point of translation (4), and also has a relatively short half-life (approximately 5 hours; refs. 13, 14), which supports that blocking translation may target CSCs (15). The dual reliance on protein synthesis to maintain core CSC factors and specific effectors of CSC signaling led us to exploit this vulnerability, or protein addiction. The translational elongation inhibitor bouvardin, identified in a Drosophila regeneration screen (16), impacted primordial cell repopulation following radiation (17) by locking eukaryotic elongation factor 2 (eEF2) to ribosomes (18). Improvement of bouvardin yielded the fully synthetic derivative SVC112 (SuviCa Inc.). Unlike translation initiation inhibitors (12, 15), elongation inhibitors such as SVC112 block both cap-dependent and cap-independent internal ribosome entry site (IRES)-dependent translation. Cap-independent translation is preferentially activated during stress and hypoxia for mRNAs containing an IRES, which include those encoding Myc and cyclin D1 (19–21). Translation elongation inhibition was validated by omacetaxine mepesuccinate [semisynthetic homoharringtonine (HHT)], which is FDA-approved in chronic myeloid leukemia where it suppresses the BCR–ABL fusion protein, however its use is limited by toxicity (22).

Here we explored how protein elongation inhibition with SVC112 impacts basic cell features such as viability and proliferation and also more complex properties such as sphere formation, response to radiation damage, and in vivo tumor growth, using established HNSCC cell lines, patient-matched HNSCC cell lines and cancer-associated fibroblasts (CAF), and patient-derived xenografts (PDX). SVC112 had an improved therapeutic index and preferentially targeted cancer cells compared with patient-matched CAFs, as opposed to HHT that inhibited growth of cancer and noncancer cells alike. SVC112 enhanced radiation effects in HNSCC cells but not in nontransformed cells, suppressed Myc, cyclin D1, and Sox2 proteins, and inhibited sphere formation. SVC112 alone and combined with radiation inhibited HPV-negative and HPV-positive PDX tumors. This is the first report of a translation elongation inhibitor that reduces HNSCC PDX growth by reducing the in vivo CSC fraction beyond a critical threshold.

Studies involving human subjects were approved by the Colorado Multiple Institutional Review Board (COMIRB-08-0552). Informed written consent was obtained from all patients whose tissues were used for this study. The University of Colorado Institutional Animal Care and Use Committee approved all mouse experiments.

013C, 036C, 049C, and 067C cells were derived and maintained as described previously (4). 013CAF, 036CAF, and 067CAF cells were derived from tumor tissue using DMEM with 10% FBS, penicillin (200 units/mL), and streptomycin (200 μg/mL) and immortalized using SV40 LgT and hTERT expression (23). To generate resistant cell lines, 013C and 036C were cultured in media containing increasing concentrations of drug until they grew normally at 1,000 and 100 nmol/L, respectively. Established HNSCC cell lines (e.g., Fadu, Detroit562) were obtained by SuviCa from Drs. David Raben and Barbara Frederick, University of Colorado. Cell lines were authenticated by DNA fingerprinting (STR analysis) before and during use.

HHT was acquired commercially (Sigma). Cells were irradiated in a Fa xitron Cabinet X-ray System Model RX-650 at 115 kV and 319 cGy/minute.

Meta Br-N-29-H derivative of RA-VII (PubChem CID: 3034401) was synthesized, characterized by 1D and 2D NMR and by LC/MS, and used at 98% or greater purity (by high-performance liquid chromatography).

These were performed by WuXi Apptec under contract to SuviCa, Inc. Female CD-1 mice were dosed with 40 mg/kg SVC112. Vein blood was drawn and plasma levels quantified by LC/MS-MS.

For gene overexpression experiments HEK293T cells were transfected with an empty (control) pMICH-mCherry retroviral vector, or vector containing cDNA for SOX2, and the pCL-Ampho packaging plasmid. 013C, 036C, 049C, and 067C cells were transduced with the resulting viral media and cells were selected by mCherry expression (FACS).

The toxic effect of SVC112 was assessed in nontumor bearing female Balb/C mice. The studies were performed under a contract at the University of Colorado Pharmacology Shared Resource, University of Colorado Cancer Center. Mice were randomly assigned to treatment groups and treated with vehicle or SVC112 (intraperitoneally at 60 mg/kg, every six hours ×2, once per week). Vehicle solution [1.03% D-α-tocopherol polyethylene glycol 1000 succinate (TPGS) and 1% poloxamer in water] was dosed at 100 μL per 25 g body weight at the time of SVC112 dosing. SVC112 nanosuspension dosing solution was freshly prepared and dosed intraperitoneally at 60 mg/kg (every six hours ×2, once per week) at 100 μL per 25 g body weight. Body weight was measured thrice weekly to grossly assess toxicity. Organs were collected 24 hours from the last SVC112 dose.

For cell injection and PDX tumor implantation animals were anesthetized with Isoflurane (induction at 5%, maintained at 1%–2%). In preparation for surgical implantation of PDX tumor tissue, and for 48 hours following the procedure, animals received buprenorphine injections (1 mg/kg) every 12 hours.

PDX generation and characterization was reported previously (24). Tumor pieces were implanted on both flanks of 6- to 10-week-old female Athymic Nude-Foxn1^nu mice (Envigo). Cases (4th–8th generation) were expanded into larger cohorts. For efficacy studies, anesthetized mice were shielded leaving only the flank tumors exposed, and irradiated (RS-2000 irradiator; Rad Source Technologies) at 115 cGy/min, twice a week (3 Gy) for 4 weeks. SVC112 was delivered intraperitoneally (60 mg/kg) immediately following radiation treatment and again 6 hours later. Tumors were measured twice a week.

Cell lines or PDX-derived CSCs were plated in ultra-low attachment 96-well plates at a concentration of 3 × 10³ (cell lines) or 1 × 10⁴ (PDX CSCs) per well in serum-free DMEM/F12 media supplemented with 10 ng/mL EGF and 10 ng/mL bFGF. Media was supplemented 4, 7, and 10 days following cell seeding. Cells were allowed to form spheres for 10 or 14 days for cell lines or CSCs, respectively. All spheres (>50 μm) in a well were manually imaged, counted, and measured (diameter) using a Zeiss Axio Observer Z1 inverted microscope (Zeiss software Rel. 4.8).

For bromodeoxyuridine (BrdUrd) labeling, cells were incubated for 1 hour in media containing 10 μmol/L BrdUrd, then washed twice with warm PBS, and fresh media was added. Cells were immediately irradiated followed by the addition of SVC112. Cells were collected at multiple time points following irradiation, suspended in 2 mL of cold PBS, and fixed by the dropwise addition of cold 100% ethanol to 7 mL total volume. BrdUrd incorporation was labeled with anti-BrdUrd-Alexa Fluor 488 antibodies (#B35189; Invitrogen) for 25 minutes. For analysis, cells were suspended in 10 μg/mL propidium iodide (PI) in PBS plus 5 μL Rnase A per mL. For PI only cell-cycle analysis, fixed cells were suspended in 50 μg/mL PI in PBS plus 5 μL Rnase A per mL. Resulting histograms were analyzed by Modfit LT software, v3.3 (BD Biosciences).

To quantify de novo protein synthesis in HNSCC cell lines, cells were incubated in methionine-free media (Invitrogen) for 1 hour, followed by addition of L-azidohomoalanine (AHA; 50 μmol/L final; Invitrogen) and the indicated SVC112 drug concentrations for 2 hours, prior to harvest. Labeling of de novo proteins containing AHA with tetramethylrhodamine was performed using Click-iT Protein Analysis Detection Kit (Invitrogen) following manufacture's protocols and as described previously (18).

For early-passage HNSCC cell lines, the methionine analog incorporation assay was completed following the manufacturer's instructions (Invitrogen) unless noted below. Cells were seeded in 96-well plates in normal growth media and incubated for 48 hours. Cells were then washed with PBS before adding L-methionine free DMEM (#21013024; Thermo Fisher Scientific, supplemented with 53 ng/L L-cystine and 584 ng/L L-glutamine) containing methionine analog [L-homoproparglyglycine (HPG)] and either control vehicle, SVC112, or HHT. Cells were then incubated for 1.5 hours before fixation and labeling for HPG incorporation. HPG incorporation was assessed by measuring fluorescence intensity (Alexa Fluor-488) for 30 cells (10 cells for three replicate regions within each well) per treatment condition, normalized to background using a Zeiss Axio Observer Z1 inverted microscope (Zeiss software Rel. 4.8).

Cell pellets were lysed in 30 to 100 μL RIPA lysis buffer containing 5 μL/mL PMSF. Protein concentration was measured using Bradford Assay and the ELx800 absorbance microplate reader (BioTek) according to the manufacturer's instructions. Thirty nanogram of protein was loaded per well into NuPage Novex 4% to 12% Bis-Tris Midi Gel (Life Technologies), transferred using the iBlot Gel Transfer Stack System (Life Technologies). Primary antibodies: Myc (#5605S; Cell Signaling Technology), cyclin D1 (#2978S; Cell Signaling Technology), Sox2 (#3579S; Cell Signaling Technology), eEF2 (#2332S; Cell Signaling Technology), Actin (4968S; Cell Signaling Technology), and vinculin (#ab129002; Abcam). Secondary anti-rabbit IgG was purchased from Jackson ImmunoResearch, and used at a 1:50,000 dilution. Signal was visualized using Immobilon Western chemiluminescent HRP substrate (EMD Millipore) on X-ray film. Quantification of relative protein levels was completed using ImageJ software version 1.5 (NIH, imagej.nih.gov).

Cells were seeded in chamber slides and incubated for 2 days before irradiation. Slides were fixed in 4% paraformaldehyde at 0, 0.5, 2, and 6 hours following 4Gy X-ray treatment. Slides were washed with PBS+0.5% Tween20, permeabilized in 0.2% Triton X-100 for 10 minutes at room temperature, and blocked in PBS+5% milk for 1 hour. Samples were labeled in PBS+1%BSA with the primary antibody mouse anti-γH2AX (#05–636; Millipore) at 1:500 for 1.5 hours, and secondary goat anti-mouse-Alexa Fluor 488 (#A11029; Thermo Fisher Scientific) for 1.5 hours. ProLong Gold antifade with DAPI (#P36931; Invitrogen) was added to slides before covering. Nuclei were randomly imaged by DAPI staining using a Zeiss Axio Observer Z1 inverted microscope (Zeiss software Rel. 4.8) at 1,008× magnification. Foci (punctate fluorescent nuclear staining) were scored as foci/nuclei and 50 nuclei were assessed per treatment. Results presented are from three independent experiments.

Primary antibodies and dilutions; 1:200 cleaved caspase-3 (#9664; Cell Signaling Technology), 1:100 Ki67 (#RM-9106-S1; Thermo Fisher Scientific), 1:200 Myc (#AB32072; Abcam), and 1:100 cyclin D1 (#2978; Cell Signaling Technology). Slides were deparaffinized and rehydrated in graded concentrations of alcohol by standard techniques before antigen retrieval in citrate buffer pH 6.0 (#S1699; Dako/Agilent) for Myc, cyclin D1, and cleaved caspase-3 at 121°F for 10 minutes. All staining was done in a Dako Autostainer, and slides were incubated in Dual Endogenous Enzyme Block (#S2003; Dako/Agilent) for 10 minutes, and in protein free blocking solution (#X0909; Dako/Agilent) for 20 minutes, and followed by the appropriate dilution of primary antibody (60 minutes at room temperature). Staining was developed using the following conditions: EnVision + Dual Link System HRP (#K4061; Dako/Agilent) for 30 minutes and substrate-chromogen (DAB+) Solution (#K3468; Dako/Agilent) for 5 minutes. Slides were then counterstained with Automated Hematoxylin (#S3301; Dako/Agilent) for 10 minutes.

To assess changes in tumor tissue following treatment, four independent tumors per treatment arm were hematoxylin and eosin stained and then 10 random regions of the proliferating tumor edge were imaged (5×) using a Zeiss Axio Observer Z1 inverted microscope. Proliferating cell depth for each image was measured using Zeiss software Rel. 4.8.

To quantify changes Myc following treatment, four tumors per treatment arm were analyzed by Myc IHC staining. Three random images of viable tumor were scored manually (number of Myc positively stained nuclei per 100 nuclei).

For spheres treated with SVC112, ImageJ Fiji software was used to measure changes in Myc and cyclin D1 IHC staining (DAB intensity). Individual spheres were gated as regions of Interest. Image color was deconvoluted and the color 2 (DAB spectrum) channel was analyzed for mean intensity of the image (0 = full brown, 255 = white). DAB optical density (OD) was calculated by OD = log(max intensity/mean intensity). For each treatment, DAB intensity for ten spheres, five from two independent replicates, was quantified.

Cells (013C, 049C, 067C) were seeded in 10 cm cell culture dishes and incubated for 48 hours to reach nearly 75% confluency. Cells were then treated with media containing DMSO (control) or SVC112 (1,000 nmol/L) for 6 hours before harvesting. Cells were collected and libraries were generated following the manufacturer's (Illumina) instructions in the Mammalian TruSeq Ribo Profile Reference Guide. Isolated total mRNA fragments and ribosomal protected mRNA fragments (RPF) were sequenced on an Illumina HiSEQ instrument at the University of Colorado Cancer Center's Genomic Core. Data processing and analysis were completed as reported previously (25).

Experiments were compared by two-tailed Student t test. Kaplan–Meier survival curves were analyzed by log-rank Mantel–Cox test. Correlation between CSC changes and tumor response was analyzed by linear regression. Calculations were done using GraphPad Prism version 8.0. Data are represented graphically as mean ±SD or ±SEM. GSEA estimates statistical significance by a modified Kolmogorov–Smirnov permutation test. P of less than 0.05 were statistically significant. All statistical tests were two-sided.

Materials will be shared per the University of Colorado's Office for Technology Transfer policies and Institutional Review Board.

SVC112, a fully synthetic derivative of the cyclic hexapeptide bouvardin (Fig. 1A). SVC112 inhibits in vitro cap-independent translation of capless luciferase mRNA using rabbit reticulocyte lysates (IC₅₀ = 81 ± 16 nmol/L; Fig. 1B). SVC112 blocked the release of eEF2 from the ribosome (Fig. 1C) like bouvardin (18), as measured by immunoprecipitation of eEF2 with ribosomal protein l13a (Rpl13a). SVC112 inhibited de novo protein synthesis in HNSCC cells (Fig. 1D), which led to the rapid depletion of cyclin D1 (26) and Myc proteins (Fig. 1E; Supplementary Fig. S1A). SVC112 inhibited growth in 19 HNSCC lines with an average IC₅₀ of 155 nmol/L (Fig. 1F) and SVC112 enhanced the anticlonogenic effect of radiation (Fig. 1G and H; Supplementary Fig. S1B; Supplementary Table S1). The dose-modifying factor (DMF) is a measure of how the antibiological effects (e.g., clonogenicity) of a radiation dose are enhanced by drug (e.g., SVC112) treatment (27). For multiple SVC112 doses, the DMF was >1 in cancer cells, but near or below 1 in four nontransformed normal lines (Fig. 1I), suggesting that SVC112 enhances radiation only in cancer cells.

We assessed SVC112 antiproliferative effects in four low passage cell lines, three of which have patient-matched CAFs (013CAF, 036CAF, 067CAF) as a nontransformed comparison. 036C, 067C, 049C, and 013C had IC₅₀s of 3.8, 9.3, 24.1, and 50.5 nmol/L, respectively (Fig. 1J and K; Supplementary Fig. S1C), and 036C, 067C, and 049C were 13.9-, 2.1-, and 1.7-fold more sensitive to SVC112 than HHT. Patient-matched 036CAFs and 067CAFs were 3.8- and 5.6-fold less sensitive to SVC112 than cancer cells, whereas there was a lesser or opposite effect for HHT (Fig. 1J and K). Sensitivity did not correlate with cell proliferation (Supplementary Fig. S1D), and SVC112 induced a dose-dependent reduction in new protein synthesis in 036C, 067C, and 013C (Fig. 1L). Although SVC112 and HHT both have substantial antiproliferative effects at nanomolar concentrations, SVC112 has a superior therapeutic index (relatively sparing normal cells) at doses achievable in vivo (Supplementary Fig. S1E). This has significant implications from both a selectivity of action and a translational perspective.

We next assessed SVC112 effect on tumor spheres, which is a standard in vitro method to assess CSC properties. SVC112 suppressed Myc and cyclin D1 in preformed spheres (Fig. 2A). Seeding cells in media containing SVC112 (1,000 nmol/L) blocked sphere formation (P < 0.001) by 100%, 79%, 74%, and 48% in 036C, 067C, 013C, and 049C, respectively. Although 100 and 1,000 nmol/L SVC112 also decreased the size of resulting 067C spheres, 013C sphere size increased at 100 nm and was unaffected at 1,000 nmol/L (P = 0.320; Fig. 2B). SVC112 and HHT had a similar impact at 100 nmol/L, whereas 1,000 nmol/L HHT blocked nearly all sphere formation. SVC112 (100 nmol/L) increased sphere inhibition by radiation in 036C and 067C cells, with a lower effect in 049C, and no impact in 013C (Supplementary Figs. S1F and S1G).

We sorted stringently defined CSCs (Aldefluor⁺CD44^high) from PDX tumors for ex vivo nonadherent cultures. CSCs had higher Myc and Sox2 levels compared with non-CSC tumor cells, and addition of SVC112 to nonadherent CSC cultures suppressed Myc and Sox2 proteins (Supplementary Fig. S1H and S1I). SVC112 (1,000 nmol/L) blocked sphere formation by CUHN036 (P = 0.001) and CUHN004 (P = 0.001) CSCs, whereas CUHN013 (P = 0.001) CSCs formed fewer, albeit normal-sized spheres (Fig. 2C).

We conducted PDX efficacy studies to explore the SVC112 inhibitory potential on CSCs, both alone and with radiation, in a preclinical setting that is stringent and as close as possible to a patient. CUHN013 and CUHN036 PDXs were chosen based on the divergent effects of SVC112 on matched (013C, 036C) cells in vitro. CUHN004 and CUHN047 were derived from a relapsed HPV-negative subject and a HPV-positive patient, respectively (Supplementary Table S2). Treatment groups were control, twice weekly radiation (3Gy), twice weekly SVC112, or the combination of radiation followed by SVC112. Radiation alone suppressed growth in CUHN036 (T/C = 0.31 ± 0.10, P < 0.001) and CUHN047 (T/C = 0.27 ± 0.09, P < 0.001). SVC112 alone suppressed growth in CUHN036 (T/C = 0.45 ± 0.12, P = 0.005) and CUHN047 (T/C = 0.37 ± 0.07, P < 0.001), and radiation plus SVC112 inhibited growth in CUHN047 (T/C = −0.02 ± 0.03, P < 0.001), CUHN036 (T/C = 0.08 ± 0.08, P < 0.001), and CUHN004 (T/C = 0.19 ± 0.05, P < 0.001; Fig. 3A). Combination treatment resulted in complete regression in 6 of 10 CUHN047 tumors and 4 of 12 CUHN036 tumors by day 28. After ending treatment, 6 of 8 CUHN036 tumors continued to regress and did not regrow (Fig. 3B and C; Supplementary Fig. S2A). CUHN013 did not respond to any treatment (Fig. 3A). Initial studies treating CUHN013 and CUHN036 once weekly generated less dramatic inhibition in CUHN036 (P = 0.003), which suggests dose dependency (Supplementary Fig. S2B). Weekly SVC112 increased inhibition of FaDu xenografts when combined with radiation (2Gy twice weekly) and cisplatin (1 mg/kg weekly; Supplementary Fig. S2C). Significant weight loss compared with control animals was only observed in the combination arm of animals bearing CUHN013 tumors treated twice weekly (P = 0.003; Supplementary Fig. S2D), and no abnormal histopathology was observed in mice treated with SVC112 for 4 weeks (Supplementary Table S3).

Histologic analysis of tumors from end of treatment (day 28) noted increased keratinization and necrosis as well as fewer viable tumor cells in all treatment arms. Notably, combination treated CUHN036 (P < 0.001) and CUHN047 (P < 0.001) tumors had only thin rims of viable tumor remaining (Fig. 3D–F). SVC112 suppressed Myc (IHC) in CUHN036 (P < 0.001), and combination treatment suppressed Myc in CUHN036 (P < 0.001) and CUHN047 (P < 0.001) tumors (Fig. 3G and H). Cleaved caspase-3 increased in irradiated or combination treated tumors (CUHN036, CUHN047, CUHN004, but not SVC112 treated tumors, suggesting the SVC112 mechanism of action is not apoptosis (Supplementary Fig. S2E). These results showed in vitro and in vivo consistency for patient-matched CUHN013 and CUHN036 PDX tumors compared with 013C and 036C cells, respectively.

To assess a potentially anti-CSC effect in vivo, we analyzed the CSC subpopulation (ALDH⁺CD44^hi) within tumors following treatment, noting that CUHN013 had the highest baseline CSC population. Radiation decreased the CSC fraction only within CUHN036 tumors, whereas SVC112 decreased CSCs in all four cases. Combination treatment decreased CSCs by 10.4-, 7.6-, and 7.9-fold in CUHN036 (P < 0.001), CUHN047 (P = 0.012), and CUHN004 (P = 0.018) respectively, but only 2.4-fold in CUHN013 (P = 0.005). Tumor inhibition was associated with lowering the CSC population below 1% (Fig. 4A and B). There was a strong relationship (measured by linear regression) between the fold change reduction of CSCs in vivo (CSC-FC) and tumor response following SVC112 treatment (alone and with radiation) both by pooled treatment arms (R² = 0.543, P = 0.037; Fig. 4C) or by individual tumors (R² = 0.224, P = 0.002; Fig. 4D), which suggests that surpassing a threshold in reducing CSCs is required to achieve tumor arrest.

To understand the biological effects of SVC112 on tumors, and specifically how it impedes normal ribosomal processing of transcripts, we sequenced RPF in control and SVC112-treated cells (013C, 049C, 067C; 6-hour exposure) using Ribo-seq. Seventy RPFs were enriched >2-fold after normalization to RNA-seq across SVC112-treated cells, indicating they are captive in the ribosome; ten were related to cellular stress response (e.g., DUSP1, MAFF, DUSP5, ATF3), four with TNFα signaling (e.g., TNFAIP2, TRAF1), four with NF-κB signaling (e.g., NFKBIA, RELB, NFKB2), and four were ribosomal/translational proteins (RPL18, RPS9, RPS16, EEF1D; Supplementary Fig. S3A). Ribo-seq identified 1699-, 666-, 346-, and 238-fold increases following SVC112 in captured RFPs of SERTAD1, NUAK2, MAFF, and DUSP1 respectively, three of which (SERTAD1, NUAK2, DUSP1) have putative IRESs within their transcribed mRNA sequence (Supplementary Fig. S3A). GSEA analysis of RPFs (captive mRNAs) identified hallmark pathways related to stress response, including TNFα Signaling via NF-κB and p53 pathway (Supplementary Fig.S3B). Overall, this indicates that SVC112 impedes ribosomal processing of mRNAs that are particularly critical for cancer cells, including IRES-containing response to cell stress and inflammation genes.

Acute SVC112 (24 hours) did not decrease the mRNA levels of Myc, cyclin D1, or Sox2 (Supplementary Fig. S3C), supporting that SVC112 effects are posttranscriptional. SVC112 (10 nmol/L) significantly suppressed ALDH1A1 mRNA (036C P = 0.008, 067C P < 0.001, 049C P < 0.001, 013C P = 0.001), which is transcriptionally regulated by the Sox2 protein (4). Global RNA-seq identified decreased transcription of GSEA hallmark pathways (Myc Targets, Wnt β-catenin Signaling, Angiogenesis), likely due to depletion of proteins like Myc by SVC112. Conversely, increased expression of apoptosis, p53 pathway, and epithelial to mesenchymal transition signaling (Supplementary Table S4) was also noted.

To further assess the cancer selectivity of SVC112, we used three patient-matched pairs of HNSCC and CAF cell lines, and compared SVC112 with HHT. CAFs had similar (013CAF) or higher (036CAF, 067CAF) protein synthesis than their patient-matched cancer lines (Supplementary Fig. S4A). Both SVC112 and HHT suppressed translation in all cells by >85% at 100 nmol/L, although only SVC112 decreased protein synthesis by >50% in 036C (P < 0.001) and 067C (P < 0.001) cells at 1 nmol/L. 013C was the least sensitive requiring a 10-fold higher dose (10 nmol/L) to achieve nearly 50% inhibition. 036C and 067C were more susceptible (lower IC₅₀s) to SVC112 than their patient-matched CAFs (Fig. 2C; Supplementary Fig. S4B).

Myc and cyclin D1 have short half-lives (<1 hour) and are expressed at different levels in HNSCC cells, whereas eEF2 is higher in cancer cells compared with CAFs (Supplementary Fig. S4C and S4D). SVC112 decreased cyclin D1 protein within 1 hour, which remained suppressed at 3 and 6 hours for all cell lines. Cyclin D1 levels started to recover by 24 hours in 013C, 049C, and 067C cells, but remained suppressed in 036C. Myc decreased in all lines within 1 hour, although complete suppression of Myc was slower (∼6 hours) in 013C and 067C cells. Myc remained suppressed (>80% reduction) at 24 hours for 036C, 049C, 067C, but returned to >60% of baseline in 013C (Fig. 5A). SVC112 had no effect on eEF2 levels in cancer cell lines (Supplementary Fig. S4E). Following incubation (24 hours) with cells, transferred media containing SVC112 inhibited protein synthesis, proliferation, and suppressed protein levels like fresh drug, indicating SVC112 stability (Supplementary Figs. S4F–S4H). However, these suggest that cells can recover protein synthesis even in the continued presence of inhibitor.

To test treatment reversibility cells were either maintained in SVC112-containing media (100 nmol/L) for 6 and 24 hours, or treated for 6 hours followed by an 18 hours washout period (6 hours: 24 hours). Myc and cyclin D1 decreased with continuous exposure (6 or 24 hours) but returned to baseline within 18 hours following washout (6 hours: 24 hours; Supplementary Fig. S5A). To assess the duration of SVC112 inhibitory effects, cells were treated and then seeded (clonogenic, spheres) following washout periods of 6, 24, and 48 hours. Sustained reduction of clonogenicity and sphere formation was observed for 24 to 48 hours after SVC112 removal (Supplementary Figs. S5B–S5D).

When comparing the effect of SVC112 and HHT on Myc and cyclin D1, sensitive strains had an idiosyncratic response (Supplementary Fig. S5E and S5F); in the less sensitive 013C (highest baseline levels of Sox2 and Myc; Supplementary Fig. S4C) HHT consistently achieved higher protein inhibition.

Resistant cell lines were generated by continuous culture in increasing concentrations of SVC112 until 036C and 013C cells proliferated normally (compared with DMSO-treated controls) in 100 and 1,000 nmol/L SVC112, respectively. Originating 013C cells were intrinsically resistant to SVC112 compared 036C, and continuous treatment generated a highly SVC112-resistant 013C strain. Both generated resistant cell lines were dramatically less sensitive to both growth and protein synthesis inhibition by SVC112, with no difference between chronically treated or cells washed out of drug 7 days before assessment (Fig. 5B and C). Sphere forming potential significantly increased in resistant 036C (P < 0.001) and 013C (P<0.001) resistant strains, and 036C-resistant spheres were larger (Fig. 5D and E). There was no change in to the ability of HHT or radiation to inhibit growth or protein synthesis in SVC112-resistant cell lines (Fig. 5F and G; Supplementary Fig. S6A). Both resistant lines had increased cyclin D1 levels, whereas resistant 013C had higher Sox2 and resistant 036C had lower eEF2 compared with control (Supplementary Figs. S6B and S6C).

Because Sox2 has been previously shown to enable and enhance CSC properties in HNSCC, we assessed its ability to counter SVC112 effects. Forced Sox2 expression blunted the anti-sphering effects of 100 nmol/L SVC112 in three of four cell lines (036C P = 0.004, 049C P < 0.001, 013C P < 0.001). Sox2 expression did not rescue SVC112 sphering inhibition in 067C cells, but decreased the impact on sphere size (P < 0.001; Fig. 6A; Supplementary Fig. S7A). Sox2 expression did not consistently alter total protein synthesis or the ability of SVC112 to inhibit protein synthesis (Fig. 6B; Supplementary Fig. S7B). Myc and cyclin D1 baseline levels did not change with expression, and SVC112 continued to suppress their levels (Fig. 6C). Sox2 expression did not affect cell-cycle kinetics following radiation and SVC112 (Supplementary Fig. S7C). Together, these suggest that the effects of Sox2 on SVC112 treatment is limited to its role as a stemness factor.

To characterize radiation enhancement, we analyzed the cell cycle for 013C, 036C, 049C, and 067C up to 48 hours after 4Gy ± 100 nmol/L SVC112. SVC112 alone increased the fraction of cells in G₁–G₀ phase by 12 hours in 049C (P < 0.001) and 24 hours in 013C (P = 0.022), 036C (P = 0.011), and 067C (P = 0.003) cells. Radiation induced G₂–M arrest by 12 hours in all cell lines, but cells began cycling within 24 hours. SVC112 after radiation led to a persistent G₂–M block at 36 hours and >48 hours in 013C and 036C cells, respectively (Fig. 7A and B; Supplementary Figs. S7D and S7E).

Cells were pulse-labeled with BrdUrd prior to treatment (4Gy ± 100 nmol/L SVC112). SVC112 decreased dividing (G₁ phase) BrdUrd-positive 036C and 013C cells by 15.6- and 2.1-fold, respectively (Fig. 7C, black arrows). Twelve-hour following radiation, nearly all BrdUrd labeled cells were arrested in G₂–M, which was delayed by SVC112 treatment; note BrdUrd-positive cells remaining in S phase (Fig. 7D, blue arrows). SVC112 decreased the number of BrdUrd-positive cells that had cleared the G₂–M arrest at 24 hours after radiation by 7.8-fold in 036C and 2.8-fold in 013C (Fig. 7D, red arrows). These data indicate that part of the SVC112 effect is by interference with checkpoint activation and recovery, and that 036C is more sensitive to these effects than 013C.

To assess SVC112 impact on DNA double-strand breaks following radiation, we next measured γH2AX foci formation finding that SVC112 alone or with radiation did not increase DNA double-strand breaks (γH2AX foci formation). However, SVC112 significantly delayed the loss of γH2AX foci (DNA repair) at 2 hours (036C and 013C, P < 0.001) and 6 hours (036C and 013C, P < 0.001) postradiation in both 013C and 036C (Fig. 7E and F). These data indicate SVC112 interferes with checkpoint activation and recovery, does not increase DNA damage, but delays DNA repair, which delays cell-cycle recovery.

Aberrant translation can increase CSC-related proteins and properties (8, 28), in contrast, tightly regulated translation maintains the balance between normal tissue SCs and proliferative progenitors (7). There have been significant efforts to target CSCs (5, 6) and to leverage translation inhibition as therapy (12, 15). HHT suppresses the BCR–ABL fusion protein (29) and depletes leukemia SCs more effectively than fusion-targeted tyrosine kinase inhibitors (30). Therefore, we tested the ability of SVC112 to suppress HNSCC CSCs and thus HNSCC growth.

SVC112 reduced translation more potently than HHT, which is likely due to differences in how each compound interacts with the ribosome; HTT binds to the ribosome A-site (31, 32) whereas SVC112 locks eEF2 on the ribosome (18). HHT had a similar impact on patient-matched cancer cells and nontumorigenic CAFs, whereas SVC112 had a superior therapeutic index, particularly at clinically achievable concentrations (10–100 nmol/L). Sparing normal tissues while blocking tumor growth is an ultimate therapeutic goal (33, 34).

A key question is if SVC112 induces selective inhibition of specific transcripts, or if it impacts transcripts more critical to HNSCC. By sequencing total mRNAs and RPFs from SVC112-treated cells, we identified changes in gene expression and actively translated ORFs, respectively. RNA-seq analysis identified decreased expression of Myc targets in treated cells, which follows the suppression of Myc by SVC112. Ribo-seq analysis of RPFs after treatment evidenced that transcripts of proteins related to inflammatory response pathways like TNFα/NF-κB signaling, as well as apoptosis and p53 pathways were captive in the ribosome at a dramatically higher proportion compared with RNA-seq also after SVC112. Of the top four genes with this translation/transcription discordance, three (SERTAD1, DUSP1, NUAK2) contained a putative IRES, suggesting they are regulated at the point of translation more so than transcription. DUSP phosphatases are of particular interest because they negatively regulate MAPKs (35), which are known to decrease rapamycin-induced autophagy (36) and gemcitabine effect (37). Other putative targets of SVC112 preferentially captured in ribosomes included MAFF and ATF3 that are known stress response transcription factors and promote radio-resistance (38, 39), and TNFα/NF-κB signaling components, activated following cellular stress (40, 41). Finally, we observed increased ribosome-association of ribosomal proteins, which activate the p53 pathway in response to ribosomal stress (42). Overall, these findings suggest that underlying the effect of SVC112 is the suppression of proteins that are critical for cancer cells, that is, SVC112 takes advantage of the protein addiction of HNSCC. Similar to how HHT suppression of BCR–ABL (29) depletes leukemic SCs (30), targeting CSC-related proteins in solid tumors using translation initiation inhibitors has been proposed (12, 15). However, stress and hypoxia promote cap-independent translation of IRES containing mRNAs (e.g., Myc, cyclin D1), which is not blocked by cap-dependent initiation inhibitors like HHT. Elongation inhibition by SVC112 significantly suppressed Myc, cyclin D1, and Sox2 in HNSCC cells and CSCs in vitro, ex vivo, and in vivo. Sox2 maintains CSCs (4) and rescued cells from SVC112 anti-sphering effect, although it did not alter Myc/cyclin D1 levels.

Tissue SC depletion leads to organ failure (e.g., cirrhosis, aplastic disease; refs. 43, 44) and we hypothesized that sufficient CSC depletion would inhibit tumors. Combination treatment significantly decreased CSCs in all four cases (>7-fold in responding cases, nearly 2-fold in resistant tumors) and CSC suppression below 1% was associated with tumor response. This may explain how, in addition to enhancing radiation efficacy in two radiation-sensitive cases, SVC112 increased radiation effectiveness in the resistant CUHN004. Despite a modest decrease in 013C sphering, SVC112 had no impact in vivo, likely because the CSC threshold was not achieved in CUHN013. These findings reinforce the importance of deploying adequately complex preclinical systems that avoid the overestimation of efficacy seen with cell lines (45).

The tested PDXs covered the HNSCC spectrum by etiology (HPV-positive/HPV-negative), response to therapy (radiation sensitive/resistant), and clinical stage (primary/relapse). We have reported that PDX susceptibility to approved agents such as cetuximab suggests they are representative and possibly predictive of patient outcomes (24). Overall, in vivo testing mirrored the in vitro results, and the models reproduced the radiation therapy susceptibility documented in the patients.

SVC112 inhibited translation elongation and preferentially blocked proliferation of HNSCC compared with noncancer cells. SVC112 suppressed CSC-related proteins, resulting in decreased proliferation and sphere formation. In PDXs representative of the clinical spectrum of HNSCC, SVC112 inhibited growth and led to tumor regression when combined with radiation. The degree of CSC depletion in tumors was associated with response to SVC112 across PDXs tested. Exogenous Sox2 partially rescued the anti-sphering effects of SVC112, confirming that suppression of high turnover “stemness” factors contributes to its anticancer effects. SVC112 is a novel agent with promising efficacy in HNSCC.

N. Gomes is a Senior Scientist at SuviCa Inc. D. Raben is a franchise lead for LASR in Genentech and has received speakers bureau honoraria from AstraZeneca, Nanobiotix, Merck, EMD Serono, Suvica, and Regeneron. T.T. Su is a consultant at SuviCa, Inc., has ownership interest (including patents) in SuviCa, Inc., and has a unpaid consultant/advisory board relationship with SuviCa, Inc. A. Jimeno has ownership interest (including patents) in SuviCa, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.B. Keysar, J.J. Morton, G.J. Pronk, D. Raben, T.T. Su, A. Jimeno

Development of methodology: S.B. Keysar, J.J. Morton, G.J. Pronk, T.T. Su, A. Jimeno

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.B. Keysar, N. Gomes, B. Miller, B.C. Jackson, P.N. Le, J. Reisinger, B. Frederick, A.-C. Tan, A. Jimeno

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.B. Keysar, N. Gomes, B. Miller, A.-C. Tan, T.T. Su, A. Jimeno

Writing, review, and/or revision of the manuscript: S.B. Keysar, N. Gomes, B.C. Jackson, P.N. Le, J.J. Morton, T.-S. Chimed, K.E. Gomez, C. Nieto, G.J. Pronk, A.-C. Tan, X.-J. Wang, D. Raben, T.T. Su, A. Jimeno

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.B. Keysar, B. Miller, J. Reisinger, H.L. Somerset, T.T. Su, A. Jimeno

Study supervision: G.J. Pronk, T.T. Su, A. Jimeno

The authors wish to thank the patients who donated their tissue, blood, and time, and to the clinical teams who facilitated patient informed consent, as well as sample and data acquisition. This work was supported by NCI SBIR Contract HHSN261201500010C (to SuviCa Inc.; subcontract to S.B. Keysar), P30-CA046934 (to A. Jimeno; University of Colorado Cancer Center Support Grant), NIH R01 GM106317 (to T.T. Su), R35 GM130374 (to T.T. Su), Ruth L. Kirschstein National Research Service Award T32CA17468 (to P.N. Le, trainee), the Daniel and Janet Mordecai Foundation (to A. Jimeno), the Charles C. Gates Center for Stem Cell Biology (to A. Jimeno), and the Peter and Rhonda Grant Foundation (to A. Jimeno).

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.