Remodeling Translation Primes CD8⁺ T-cell Antitumor Immunity

The efficacy of T cells to combat tumor growth is intimately tied to bioenergetics (1–3). Glycolytic T effectors are rapidly disarmed in the tumor microenvironment by nutrient deprivation (4) and inhibition of mitochondrial biogenesis (5). In contrast, T cells with memory properties induce tumor control highlighted by an exquisite ability to persist long-term in vivo (6) and the antitumor immunity conferred by memory T cells is attributed in part to their intrinsic metabolic path (7). mRNA translation is an essential feature of cell bioenergetics (8), but has not been applied to T-cell biology for applications in tumor immunotherapy.

In response to nutrient and amino acid availability, mTOR catalyzes mRNA translation through the activation of ribosomal protein S6 kinase beta-1 (p70S6K), stimulating a component of the 40S ribosomal subunit, ribosomal protein S6 (S6; ref. 9). In response to metabolic stress, AMP-activated protein kinase (AMPK) limits protein synthesis to conserve metabolic energy through two mechanisms. AMPK inhibits mTOR-directed p70S6K protein translation (10), limiting activation of S6. AMPK sequesters the elongation step of protein translation through phosphorylation of the kinase eukaryotic elongation factor 2 (eEF2K), which in turn phosphorylates eEF2 (p-eEF2; refs. 11, 12). p-eEF2 restricts protein translation through the inhibition of GTP-dependent translocation of the A-site–bound peptidyl-tRNA to the P-site during the elongation phase (13).

Here, we demonstrated that modulation of translation was a requisite for T-cell tumor control. Using IL2 and IL15 cytokine priming to study translation in effector and memory-like T cells, respectively, we demonstrated that AMPK restricted protein translation in IL15-primed T cells through canonical mechanisms. Studying the paths of protein translation in T cells revealed that modulation of protein translation elongation improved T-cell antitumor immunity, providing a unique treatment strategy.

Pmel (B6.CgThy1^a/CyTg(TcraTcrb)8Rest/J) and C57BL/6J mice were obtained from the Jackson Laboratory. Six- to 8-week-old mice were used in the study. All animal experiments were approved by the Medical University of South Carolina (MUSC) Institutional Animal Care and Use Committee and the Division of Laboratory Animal Resources at MUSC maintained all mice.

For T-cell cultures, splenocytes were manually dissociated to single-cell suspension, red blood cells were lysed, and whole splenocytes from Pmel mice were activated with hgp100 peptide (1 μg/mL, GenScript) and expanded with 200 U rhIL2 (NCI) in complete T-cell media (RPMI1640, 10% FBS, 300 mg/L l-glutamine, 100 U/mL Penicillin, 100μg/mL Streptomycin, 1 mmol/L Sodium Pyruvate, 100 μmol/L NEAA, 1 mmol/L HEPES, 55 μmol/L β-mercaptoethanol, and 0.2% Plasmocin Mycoplasma Prophylactic). For cytokine differentiation, after 3 days of expansion, T cells were split into complete T-cell media containing 200 U rhIL2 or 50 ng/mL rhIL15 (Shenandoah) and harvested for analysis at time points indicated throughout the text and figure legends. Vehicle (DMSO), Compound C (Sigma), or Rapamycin (Sigma) were added at 1 μmol/L and 10 nmol/L, respectively. For drug treatments, whole splenocytes from Pmel mice were activated (as described above) in the presence of vehicle, 10 μmol/L AICAR (dissolved in H₂O, Sigma), or 20 nmol/L Homoharringtonine (dissolved in DMSO, Sigma).

B16F1 tumor cells obtained from ATCC were maintained in RMPI complete T-cell media and passaged three times prior to in vivo inoculations. Tumor cells were determined to be Mycoplasma free in November 2019. The tumor cell line has not been authenticated in the past year.

Fluorochrome-conjugated mAbs CD8-APC (53-6.7), CD62L-FITC (MEL-14), CD44-Pe-Cy7 (IM7), and Thy1.1-Percp-Cy5.5 (HIS51) and respective isotype controls were purchased from Thermo Fisher Scientific. Live-or-Dye-PE Fixable Viability Stain was purchased from Biotium. Extracellular stains were performed in FACS buffer (PBS, 10% FBS). p-S6 staining was performed using Intracellular Fixation and Permeabilization Buffer Set (eBioscience, 88-8824-00). Cells were fixed and then stained with p-S6-APC (cupk43k, Thermo Fisher Scientific) or isotype control in permeabilization buffer. Protein synthesis was measured using the Click-iT HPG Alexa Fluor 488 Protein Synthesis Assay Kit according to the manufacturer's protocol (Thermo Fisher Scientific, C10428). Briefly, cells were incubated in methionine-free media and control cells were treated with 100 μg/5 × 10⁵ cells cycloheximide (Cayman Chemical, 601105). Cells were washed with PBS and stained extracellularly for 15 minutes and subsequently incubated in 50 μmol/L l-homopropargylglycine for 1 hour. Live-or-Dye Fixable Viability Stain (Biotium, 32005-T) was added for 10 minutes at the end of this incubation. Cells were fixed using 4% paraformaldehyde, washed with 3% BSA, and permeabilized using 0.5% Triton X-100 solution. Cells were washed in 3% BSA, followed by staining with Alexa Fluor 488 azide to label active protein synthesis. Isotype or cycloheximide controls were used in all experiments. Samples were run directly on a BD Accuri C6 flow cytometer and analysis was performed with FlowJo Software (TreeStar).

T cells were lysed in RIPA Buffer (Sigma) supplemented with Protease Inhibitor Cocktail (Cell Signaling Technology) and Phosphatase Inhibitors I and II (Sigma). Protein concentrations were normalized using Pierce BCA Kit (Thermo Fisher Scientific, 23227) and loaded to 4%–10% agarose gels (Bio-Rad, 4568084). p-eEF2 (Thr56), eEF2, p-p70S6K (Thr389), p70S6K, p-AMPK (Thr172), AMPK, p-mTOR (S2448), p-p44/42 MAPK (Erk1/2) (Thr202/Tyr204), Ubiquitin (P4D1), β-actin, and HRP-linked anti-rabbit and mouse secondaries were obtained from Cell Signaling Technology. p-p90RSK was obtained from BioLegend. Phospho proteins were developed with Pierce ECL Plus Western Blotting Substrate (Thermo Fisher Scientific, 32134). Total proteins and p-p44/42 MAPK (Erk1/2) were probed with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific, 32109).

RNA was isolated with RNeasy Mini Kit (Qiagen, 74104) and concentration was measured using the SpectraDrop Micro-Volume Microplate (Molecular Devices). Single-strand cDNA was made with 500 ng RNA using the High Capacity RNA-to-cDNA Kit (Applied Biosystems, 4387406, Thermo Fisher Scientific).Mouse TaqMan Gene Probes (Applied Biosystems, Thermo Fisher Scientific) were used to perform real-time PCR in triplicate using the StepOnePlus Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). Gene expression for Tcf7 was normalized to Gapdh and calculated using ΔΔC_t method.

For transwell coculture assays (Corning), B16F1 tumor cells were seeded for 24 hours, then T-cells cultured in IL2 or IL15 were added to transwells in complete T-cell media supplemented with respective cytokine, and harvested 36 hours later. Vehicle or p-p44/42 MAPK (Erk1/2) inhibitor U0126 (10 μmol/L, Cell Signaling Technology) was added to transwell cocultures 4 hours prior to T-cell harvest.

A total of 2.5 × 10⁵ B16F1 melanomas-injected 200 μL sterile PBS were established subcutaneously on the right flank of female C57BL/6 mice. Tumor-bearing mice were 5 Gy irradiated 24 hours prior to T-cell transfer. After 7 days of tumor growth, 2 × 10⁶ Pmel T cells were infused via tail vein in 100μL of sterile PBS to melanoma-bearing mice. Tumor growth was measured every other day with calipers, area was calculated by tumor length x tumor width, and survival was monitored with an experimental endpoint of tumor size ≥ 400 mm². Vehicle, AICAR, Lys6K2 S6 Kinase Inhibitor (DMSO, 10 μmol/L; Focus Biomolecules), or homoharringtonine-treated T cells were transferred after 4 days of ex vivo expansion following the aforementioned protocol. For ex vivo analysis of transferred T cells, mice were scarified 5 days after transfer and tumors and spleens were processed to single-cell suspensions. Tumors were digested using the Mouse Tumor Dissociation Kit (Miltenyi Biotec, 130-096-730) according to the manufacturer's protocol.

Oxygen consumption rate was measured in Seahorse XF Medium (Agilent) supplemented with 100 nmol/L insulin, 1 mmol/L sodium pyruvate, 5.6 mmol/L glucose, 4 mmol/L glutamine, and 1% FCS under basal conditions and in response to 1 μmol/L oligomycin, 1.5 μmol/L FCCP, and 2 μmol/L rotenone + 1 μmol/L Antimycin A using the XFe96 Extracellular Flux Analyzer (Seahorse Bioscience). Cell-Tak (Corning) was used for adherence of 4 × 10⁵ T cells per well.

T cells were lysed in 9 mol/L urea, 50 mmol/L Tris pH 8, and 100 units/mL Pierce Universal Nuclease (Thermo Fisher Scientific) and the concentration of protein was measured using a BCA Assay (Thermo Fisher Scientific). Protein was trypsin (Sigma) digested at 37°C for 18 hours, and the resulting peptides were desalted using C18 ZipTips (Millipore).

Peptides were separated and analyzed on an EASY nLC 1200 System (Thermo Fisher Scientific) in-line with the Orbitrap Fusion Lumos Tribrid Mass Spectrometer (Thermo Fisher Scientific) with instrument control software v. 4.2.28.14. Peptides were pressure loaded at 1,180 bar, and separated on a C18 Reversed Phase Column [Acclaim PepMap RSLC, 75 μm x 50 cm (C18, 2 μm, 100 Å); Thermo Fisher Scientific] using a gradient of 2%–35% B in 120 minutes (Solvent A: 0.1% FA; Solvent B: 80% ACN/0.1% FA) at a flow rate of 300 nL/minute at 45°C.

Mass spectra were acquired in data-dependent mode with a high resolution (60,000) FTMS survey scan, mass range of m/z 375–1,575, followed by tandem mass spectra of the most intense precursors with a cycle time of 3 seconds. The automatic gain control target value was 4.0e5 for the survey MS scan. Fragmentation was performed with a precursor isolation window of 1.6 m/z, a maximum injection time of 50 ms, and HCD collision energy of 35%; the fragments were detected in the Orbitrap at a 15,000 resolution. Monoisotopic precursor selection was set to “peptide.” Apex detection was not enabled. Precursors were dynamically excluded from resequencing for 20 seconds and a mass tolerance of 10 ppm. Advanced peak determination was not enabled. Precursor ions with charge states that were undetermined, 1, or >7 were excluded.

Protein identification and quantification were extracted from raw LC/MS-MS data using the MaxQuant platform v.1.6.3.3 with the Andromeda database searching algorithm and label free quantification (LFQ) algorithm (14–16). Data were searched against a mouse Uniprot reference database UP000000589 with 54,425 proteins (March, 2019) and a database of common contaminants. The FDR, determined using a reversed database strategy, was set at <1% at the protein and peptide level. Fully tryptic peptides with a minimum of seven residues were required including cleavage between lysine and proline. Two missed cleavages were permitted. LC/MS-MS analyses were performed in triplicate (IL2 and IL15) or duplicate (vehicle and homoharringtonine-primed T cells) for each biological replicate with match between runs enabled. The “fast LFQ” was disabled and “stabilize large ratios” features were enabled. The first search was performed with a 20 ppm mass tolerance, after recalibration a 4.5 ppm tolerance was used for the main search. A minimum ratio count of 2 was required for protein quantification with at least one unique peptide. Parameters included static modification of cysteine with carbamidomethyl and variable N-terminal acetylation.

The protein groups' text file from MaxQuant was processed in Perseus v. 1.6.5.0 (16). Identified proteins were filtered to remove proteins only identified by a modified peptide, matches to the reversed database, and potential contaminants.

Normalized LFQ intensities were log₂ transformed. The LFQ intensities of technical replicates were averaged for each biological replicate. Quantitative measurements were required in at least three of five biological replicates in each treatment group.

For all experiments in which protein synthesis was measured, mean fluorescent intensity (MFI) of L-homopropargylglycine (HPG) incorporation was divided by MFI of HPG incorporation in a condition-matched cycloheximide-treated well. Tumor growth was analyzed by linear regression of growth curves of vehicle versus drug-treated T cells. Survival to 40 days or tumor size of 400 mm² was the experimental endpoint with log-rank test for survival proportions of mice treated with vehicle versus drug-treated T cells used for analysis. Statistical analyses were performed with GraphPad Prism v8.3.0. Statistical parameters can be found in the figure legends.

For proteomics, log₂ transformed, protein LFQ intensities, normalized in MaxQuant, exhibited normal distributions. Binary comparisons of each LC/MS-MS analysis yielded Pearson correlation coefficients >0.90 between all technical and biological replicates within each experiment. For comparison of IL2- and IL15-conditioned T cells, quantitative measurements were required in at least three of five biological replicates in each group yielding 1,521 quantified protein groups (Supplementary Table S1). Using a Student t test, 634 protein groups had a P < 0.05, with more stringent criteria of a permutation-based FDR of 5% and S0 of 0.1, 546 proteins were considered significantly different. For comparison of vehicle and homoharringtonine-primed T cells, quantitative measurements were required in all four biological replicates in each group yielding 2,217 protein groups (Supplementary Table S2). Using a Student t test, 426 proteins had a P < 0.05. Volcano and enrichment plots were generated using R software. The −log₁₀ P value versus the difference in log₂ protein intensities (IL15 − IL12) or (homoharringtonine – Vehicle) are shown. Unless otherwise noted, significance was assessed by two-tailed Student t test.

IL2 and IL15 cytokine–mediated T-cell differentiation is a strategy to study the molecular mechanisms that affect effector and memory-associated T-cell lineages, respectively (7, 17, 18). Here, we used IL2 and IL15 conditioning to assess the canonical mechanisms that regulate protein translation in effector and memory-like T cells as depicted in Fig. 1A. Culture of T cells with IL15 resulted in a pool of T cells with memory traits, specifically CD62L/CD44 coexpression and transcription factor 7 (Tcf7) gene expression (ref. 19; Supplementary Fig. S1A and S1B). Consistent with reports that the energy sensor AMPK is integral for memory T-cell formation (20), AMPK activation was observed within 4 hours of IL15 addition to T cells (Fig. 1B). Concordant with activation of p-AMPK, canonical mechanisms of AMPK that modulated protein translation (Fig. 1A) were evident. Specifically, reduced p-p70S6K and enhanced p-eEF2 were observed in IL15-treated T cells relative to IL2 effector cells (Fig. 1B). Since p-p70S6K generates activation of S6 ribosomal subunit (pS6), we measured pS6 using flow cytometry. Within 6 hours of cytokine addition, pS6 was diminished in IL15-treated T cells relative to IL2 effectors (Fig. 1C).

To measure how protein translation was altered in T-cell subsets, we employed a rigorously validated FACS-based protein synthesis assay, in which a fluorescent amino acid analogue of methionine (HPG) is incorporated into actively translating cells (21–23). IL15 T cells exhibited reduced protein synthesis relative to IL2-treated effectors within 6 hours of cytokine addition (Fig. 1D). To test whether AMPK directly inhibits translation in IL15-primed T cells, we added the AMPK inhibitor Compound C to T-cell cultures at the time of cytokine addition. Translation was restored in IL15-primed T cells treated with Compound C (Fig. 1E). In a similar manner, treatment with mTOR inhibitor, rapamycin, illustrated that translation in T cells at the time of cytokine addition was dependent on mTOR (Fig. 1F).

We confirmed that IL2 and IL15 T cells sustained increased and reduced translation, respectively, after 3 days of cytokine differentiation (Fig. 1G). Protein degradation assessed by ubiquitination was modestly reduced in IL15-treated T cells relative to IL2 (Supplementary Fig. S1C). These data prompted us to use LC/MS-MS analysis of IL2- and IL15-differentiated T cells to assess global protein differences among T-cell groups (Supplementary Table S1). Proteins necessary for translation initiation (eukaryotic initiation factor 4a, eIF4a) and elongation (eEF2) were increased in IL2-derived T cells relative to IL15 as were proteins associated with an increased unfolded/misfolded protein burden (endoplasmic reticulum oxidoreductase, ERO1α and Bax; refs. 24, 25; Fig. 1H). Gene ontology analysis of biological functions demonstrated that IL2 effector T cells harbored a proteome committed to sustaining translation (Fig. 1I). Consistent with memory T-cell metabolism, IL15-treated T cells were enriched for β-oxidation of fatty acids (20). Metabolic processes of generation of precursor metabolites, cellular respiration, and ATP synthesis–coupled proton transport were biological processes enriched in IL15-treated T cells relative to IL2 (Fig. 1I).

To test whether stimulation of AMPK in T cells modulated protein translation, we treated T cells with AMPK stimulant 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR). This treatment increased p-AMPK and enhanced and reduced the AMPK targets p-eEF2 and p-p70S6K, respectively (Fig. 2A). Protein synthesis and pS6 were diminished in AICAR-treated T cells (Fig. 2B and C). To test the effect of AMPK stimulation on T-cell–mediated tumor control, we activated and expanded Pmel T cells in the presence of AICAR. Protein synthesis and pS6 were diminished relative to vehicle controls (Fig. 2D and E). We infused B16 melanoma–bearing mice with vehicle or AICAR-treated T cells. Infusion of AMPK-primed T cells improved tumor control and extended animal survival relative to vehicle controls (Fig. 2F and G).

In tumors, memory T cells can acquire effector cell functions (26), but how translation is regulated in this process is unknown. We transferred IL2- or IL15-differentiated T cells into tumor-bearing mice and measured translation in transferred T cells. In tumors, IL15-primed T cells exhibited enhanced translation relative to IL2-derived tumor-infiltrating lymphocytes (TIL; Fig. 3A).

To assess molecular changes associated with translation in IL2- and IL15-derived T cells in tumors, we created a transwell coculture tumor T-cell assay in which T cells and tumors cells shared media without contact. This assay assessed the effect of nutrient competition or secreted factors on T-cell translation, but does not measure the effect of antigen encounter or contact-dependent tumor inhibition on T-cell translation. Within 36 hours of coculture we found that translation in IL2 T cells was impaired. Similar to IL15-primed T-cell patterns in vivo (Fig. 3A), IL15-treated T cells exhibited a remarkable capacity to increase translation in coculture with tumor cells (Fig. 3B). In accordance with enhanced translation, IL2 T cells cocultured with tumor cells diminished pS6; whereas, IL15-conditioned T cells from tumor coculture increased pS6 relative to tumor-free controls (Fig. 3C).

These data prompted us to assess molecular changes that affect translation in tumor cocultures within T-cell groups. In IL15-derived T cells, p-AMPK was modestly reduced in tumor coculture and p-eEF2 was released (Fig. 3D). P-p70S6K was further reduced in IL15-primed T cells upon coculture with tumor cells (Fig. 3D). Given that pS6 was increased in IL15 T cells cocultured with tumor cells, we assessed activation of other molecular factors known to stimulate pS6 at S235/S236. A noncanonical signaling pathway that activates pS6 occurs through p90-ribosomal S6 kinase (p-p90RSK) that lies downstream of ERK (27). Memory T cells are poised to signal Erk1/2 and IL15 conditioning enhances Erk1/2 activation (28, 29). We observed that IL15 T cells express increased p-p90RSK relative to IL2 effectors. We were unable to detect differences in p-p90RSK in IL15 T cells responding to tumor cell coculture (Supplementary Fig. S2). To establish that reinvigoration of translation of IL15-primed T cells in the tumor microenvironment was dependent on ERK1/2, we treated T cells in transwell cocultures with the p-p44/42 MAPK (Erk1/2) inhibitor U0126. IL15-derived T cells expressed more p-p44/42 MAPK (Erk1/2) relative to IL2-treated cells (Fig. 3E) and 10 μmol/L U0126 abrogated p-p44/42 MAPK (Erk1/2) completely in IL2- and IL15-primed T cells. Inhibition of p-p44/42 MAPK (Erk1/2) reduced translation and p-S6 in IL15-derived T cells cocultured with tumor cells relative to controls (Fig. 3F and G). Our data indicate that the tumor microenvironment dysregulates protein synthesis in effector T cells, suggesting that IL15-primed T cells harbor alternate mechanisms to support translation in tumors.

p-p70S6K appeared dispensable for enhanced translation in IL15-primed T cells in coculture (Fig. 3). These data propelled us to measure how p-p70S6K–mediated activation of S6 affects T-cell tumor control. We activated T cells in the presence of p70S6K inhibitor Lys6K2 (30) and confirmed that pS6 was extinguished (Supplementary Fig. S3A). Unexpectedly, protein synthesis was not impaired in p70S6K-treated T cells relative to vehicle (Supplementary Fig. S3B). Transfer of vehicle or Lys6K2-treated T cells to melanoma-bearing mice did not improve tumor control (Supplementary Fig. S3C).

In IL15- and AICAR-treated T cells, we observed a robust increase in p-eEF2 (Figs. 1 and 2). p-eEF2 inhibits the elongation step of protein translation through blockade of GTP-dependent translocation of the A-site–bound peptidyl-tRNA to the P-site (13). We reasoned that protein translation elongation may be a critical metabolic process that inhibits T-cell function in tumors. We targeted protein translation elongation using the protein synthesis inhibitor, homoharringtonine, which interacts with the A-site of the ribosome to prohibit A-site–binding of peptidyl-tRNA (31). T cells treated with homoharringtonine underwent reduced translation relative to controls (Fig. 4A). Live cell numbers of T cells were reduced in the homoharringtonine-treated condition 24 and 48 hours after activation (Fig. 4B). We found that T cells treated with homoharringtonine at the time of activation continued to undergo limited translation relative to vehicle controls after homoharringtonine was removed from cultures and T cells were reexpanded in vitro (Fig. 4C).

We confirmed by immunoblot that homoharringtonine-mediated translation inhibition was not due to stimulation of AMPK, as homoharringtonine-primed T cells expressed reduced p-AMPK, increased p-p70S6K, and reduced p-eEF2. In-line with these data, the metabolic energy sensor p-mTOR and pS6 were enhanced in homoharringtonine-treated T cells (Supplementary Fig. S4A and S4B). These data suggested that inhibition of the elongation step of translation overrode activation of S6 to control polypeptide chain formation in T cells. To measure the capacity to control tumors, we transferred vehicle or homoharringtonine-treated Pmel T cells to mice bearing 7-day established B16F1 melanomas. We found that homoharringtonine T cells exhibited profound tumor control in contrast to vehicle-matched controls (Fig. 4D) and extended animal survival (Fig. 4E). In vivo homoharringtonine-primed TILs exhibited more translation than vehicle-matched TILs (Fig. 4F).

We next measured the bioenergetic traits that may account for the ability of homoharringtonine T cells to produce more protein in response to tumor antigen. Homoharringtonine T cells did not harbor a greater spare respiratory capacity. However, compared with vehicle controls, homoharringtonine-treated T cells exhibited increased mitochondrial coupling efficiency and reduced proton leak synonymous with a heighted efficiency to generate ATP from the available metabolic pool (Fig. 4G and H). We used LC/MS-MS proteomics to profile homoharringtonine-primed T cells (Supplementary Table S2). In-line with the bioenergetics data, generation of precursor metabolites, cellular respiration, and ATP synthesis–coupled proton transport were enriched biological processes in homoharringtonine-primed T cells relative to vehicle controls (Fig. 4I). These data bear a striking resemblance to the enriched processes found in IL15-primed T cells (Fig. 1); however, a memory phenotype was not evident after expansion of homoharringtonine-treated T cells relative to vehicle (Supplementary Fig. S4C). The proteome showed an abundance of deubiquitylation enzymes increased in homoharringtonine-treated T cells relative to vehicle cells and increased antiapoptotic mitochondrial hallmark Bcl-2 (Fig. 4J; ref. 7). Synonymous with reduced protein degradation, Western blotting confirmed that homoharringtonine-primed T cells harbored fewer ubiquitinated proteins (Fig. 4K). The data show that translation in T cells is tied to bioenergetics that shape tumor control. Targeting translation is thus a potential strategy to augment antitumor immunity.

J.E. Thaxton reports receiving a commercial research grant from TEVA Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.E. Hurst, J.E. Thaxton

Development of methodology: K.E. Hurst, L.E. Ball, J.E. Thaxton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.E. Hurst, K.A. Lawrence, L.E. Ball

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.A. Lawrence, L.E. Ball, D. Chung, J.E. Thaxton

Writing, review, and/or revision of the manuscript: K.E. Hurst, L.E. Ball, J.E. Thaxton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.E. Hurst, R.A. Robino

Study supervision: J.E. Thaxton

The authors thank Drs. Michael Lilly and Zihai Li for mentorship within the Hollings Cancer Center. We acknowledge technical assistance from the Mass Spectrometry Facility at MUSC. This study was supported by NIH grant K12 CA157688 and American Cancer Society (ACS) grants IRG-97-219-14 and IRG-16-185-17 and Pilot award from 5P20GM103542-08 to J.E. Thaxton. Proteomics work supported by NIH grants S10 OD010731 and GM103542 to L.E. Ball.