TAp73 Modifies Metabolism and Positively Regulates Growth of Cancer Stem–Like Cells in a Redox-Sensitive Manner Free

Cancer stem–like cells (CSLC) constitute a distinct population of cells within a heterogeneous cancer mass. CSLCs possess the ability to self-renew and differentiate into various cell lineages (1), which is mainly regulated by aberrant expression of stem cell pluripotency transcription factors such as Oct4, Sox-2, and Nanog (2). As such, CSLCs are extremely tumorigenic and remain more resistant to various therapeutic insults than differentiated cancer cells, and are believed to be the main culprits behind cancer persistence and relapse (3, 4). Thus, owing to their stem cell features, CSLCs pose a major obstacle in achieving effective therapeutic outcomes from cancer treatments (5).

Tumor suppressor TAp73/TP73, the first identified p53/TP53 homolog, has been extensively studied in the context of cancer and is known for its growth-suppressing activities (6, 7). Unlike p53, which is lost or mutated in over 70% of cancers, TAp73 mutations in cancer are rare, making it readily available for possible therapeutic modulations (8). Therefore, activation of TAp73 is deemed a valid strategy to treat cancers by promoting its tumor suppression function (6, 7). In addition to its tumor-suppressing role, TAp73 has been identified as a major regulator of development (9). Moreover, TAp73 has been shown to play a crucial role in neuronal stem cell maintenance and differentiation, yet its involvement in the stemness of cancer remains unexplored (10).

Here, we report that tumor suppressor TAp73 is required for the maintenance of stemness in cancer cells. This unexpected role for TAp73 is in total contrast to the recently reported anticipated role for ΔNp73, a dominant-negative variant of the tumor-suppressor TAp73 (11). Using both human (NT2/D1) and murine (P19) models of embryonal carcinoma stem-like cells (ECSLC; refs. 12, 13), we found that following the knockdown (KD) of TAp73, ECSLCs lose the expression of pluripotency factors as well as tumorsphere-forming capacity, become differentiated, and demonstrate severe impairment in their potential to form tumors in vivo. Interestingly, this role of TAp73 in ECSLCs is opposite to that reported in differentiated cancer cells and is generalizable in other non–Oct4-driven, human mammary epithelial (HMLE)–based CSLC models (14). We also provide direct clinical evidence to support our findings, where we observed similar effects of TAp73 KD on growth and stem cell markers in patient-derived brain tumor–initiating cells (BTIC) with CSLC characteristics. In differentiated cancer cells, TAp73 overexpression (OE) has been shown to inhibit p-Akt and promote autophagy, senescence, and p21 expression (6). In contrast, we observed these effects following the depletion of TAp73 in ECSLCs. Mechanistically, we reveal that TAp73 regulates proline/glutamine metabolism and antioxidant defense system. KD of TAp73 causes a shift in the proline regulatory axis (PRA) that inhibits proline and glutamine synthesis, and decreases the levels of the antioxidants glutathione and superoxide dismutase 1 (SOD1) that results in elevated levels of reactive oxygen species (ROS). Taken together, our findings define a novel role in which TAp73 regulates stemness through redox-sensitive metabolic reprogramming. Most importantly, these findings identify a need for differential therapeutic targeting of TAp73 in stem-like versus differentiated cancer cells in clinical settings.

This study was conducted in accordance with the Declaration of Helsinki. Animal work was approved by the Ethics Committee on Laboratory Animals (UCLA) at Dalhousie University. Human glioblastoma (GBM) samples were obtained from consenting patients, as approved by the Hamilton Health Sciences/McMaster Health Sciences Research Ethics Board.

For all cell lines used in this study, an in-house Mycoplasma testing regimen was performed monthly (MycoAlert, Lonza; LT07-703). Prior to in vivo experiments, NT2/D1 and HMLER^shECad cells were sent for external pathogen testing to Charles Rivers Laboratories and verified to be pathogen-free before introduction in animals. NT2/D1 and P19 were maintained in DMEM or DMEM without glutamine, HMLE and HMLER were maintained in DMEM F12, while HMLER^shECad were maintained in serum-free HUMEC Ready Media (Gibco). DMEM and McCoy's 5A media were supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin, and DMEM F12 was supplemented with 5% heat-inactivated FBS, 20 ng/mL EGF, 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, and 1% penicillin/streptomycin. Media for NT2/D1 and P19 were additionally supplemented with 1% nonessential amino acids (NEAA; ThermoFisher). Actively growing cells were treated with 12 μmol/L of chloroquine (Sigma), Proline (Invitrogen), 2 mmol/L Glutamax (Invitrogen), 2 μmol/L of N-acetyl cysteine (NAC; Sigma), 10 μmol/L of Mitotempo (Sigma), or 50 μmol/L of H₂O₂ (Sigma) for 24 hours.

In order to comprehensively study the role of TAp73 in cancer stem cells, we wanted to use a comprehensive approach to determine whether our findings were applicable in different cancer stem cell models. All of the models employed in this study have also been extensively used and validated by other groups and offer a wide range of very different biological models to study the role of TAp73 cancer stem cell biology (12, 13, 15–20). The NT2/D1 and P19 cell lines used throughout most of this study are embryonal carcinoma cell lines that possess all of the traditional stem cell properties such as the ability to self-renew and maintain an undifferentiated state (12, 13, 15, 16). These cells express all of the major Yamanaka pluripotency factors (Oct4, Nanog, and Sox-2; ref. 21), and have potential to differentiate into many different downstream derivatives from ectoderm, mesoderm, and endodermal lineages, and are commonly used to study neuroendothelial differentiation (22, 23). Therefore, these cells are a useful model to study the role of p73 in regulating stemness and differentiation. To compare the role of p73 in more traditional cancer stem–like cell models, which are commonly characterized based on expression of certain cell surface markers, the ability to form self-renewing tumorspheres, and seed tumor formation at low cell density, we obtained the HMLER^shECad cell line (CD44⁺/CD24⁻) as a generous gift from Dr. Robert Weinberg (17, 18). Finally, we validate our findings in the clinically relevant CD133^High patient-derived brain-tumor initiating cell lines, where we also confirmed the effect of TAp73 on stemness features and several mechanistic targets (19, 20, 24).

Human GBM samples were obtained from consenting patients, as approved by the Hamilton Health Sciences/McMaster Health Sciences Research Ethics Board. Tumor tissues were dissociated and cultured in NeuroCult NS-A Proliferation Medium (STEMcell Technologies) supplemented with epidermal growth factor (20 ng/mL), basic fibroblast growth factor (10 ng/mL), 2 μg/mL of Heparin, and antibiotic–antimycotic (1X Wisent) on ultralow attachment plates (Corning) and propagated as tumorspheres as previously described (25).

Flow cytometric analysis of BTICs was performed as previously described (25) using APC-conjugated anti-CD133 or a matched isotype control (Miltenyi) and run on a MoFlo XDP Cell Sorter (Beckman Coulter). Dead cells were excluded using the viability dye 7AAD (1%; Beckman Coulter). Compensation was performed using mouse IgG CompBeads (BD Biosciences). Expression of CD133 was defined as positive or negative based on the analysis regions set on the isotype control.

Tumorsphere formation assays were performed as previously described (25). Cells were seeded at a density of 60,000 cells per well for NT2/D1 and 40,000 cells per well for HMLER^shECad in a serum-free HUMEC Ready Media for stem cells (Gibco). Spheres with a diameter equal or higher than 50 μm were deemed tumorspheres, and average number of tumorspheres per 10⁵ μm of plate surface area was determined and quantified using ImageJ. Data are representative of at least three independent experiments and quantified from more than three microscopic fields of view per experiment.

Animal work was approved by the Ethics Committee on Laboratory Animals (UCLA) at Dalhousie University, protocol approval number: 16–137. shNS nonsilencing control or TAp73 KD NT2/D1 cells (2 × 10⁶ cells) were injected subcutaneously with Matrigel (ThermoFisher) into the left hind flank of 6- to 8-week-old female NOD/SCID mice weighing 20 to 25 g (purchased from Charles River Laboratory). Ten mice were allocated to each experimental group (shNS control and shTAp73). Mice were monitored for 42 days to observe animal health and tumor growth. Weight of mice was monitored every week, mice were sacrificed if less than 10% of body mass was lost, and no mice had to be sacrificed due to weight loss in this study. Tumors were then excised, weighed, and dissociated for protein extraction and subjected to Western blot analysis as described. Three tumors were excluded from Western blot and qRT-PCR analysis from TAp73 group due to lack of sufficient tumor tissue for protein and RNA extraction. Note that 1 × 10⁶ HMLER^shECad cells with either nonsilencing shNS control or shTAp73 KD from two different clones (shp73 #1 and #2) were injected subcutaneously with Matrigel (ThermoFisher) into the mammary fat pad of 6- to 8-week-old female NOD/SCID mice weighing 20 to 25 g (Charles River Laboratory). Six mice were allocated to each experimental group (shNS, shTAp73 #1, and shTAp73 #2). Mice were monitored for 12 days to observe animal health and tumor growth. Weight of mice was monitored every week, and mice were sacrificed if less than 10% of body mass was lost, and no mice had to be sacrificed due to weight loss in this study. Tumors were then excised and weighed.

Lentiviral vectors with shRNA sequence targeting p73 or Oct4 were purchased from GE Dharmacon. The lentiviral vector shSOD1 was a gift from Patrick Aebischer (Addgene, 10880). A nonsilencing shRNA vector was also used as a control, denoted as scrambled. The envelope plasmid pMD2G and the packaging plasmid psPAX2 were obtained from Addgene. For lentiviral, production was performed as previously described (25).

Primary patient-derived brain tumor–initiating cells were transduced and maintained using a modified protocol from Venugopal and colleagues (24). Briefly, at the time of transduction, tumorspheres were broken down to single-cell suspensions by manual pipetting, sequabrene was added at a final concentration of 1 μg/mL, and lentivirus carrying the specified shRNA vectors was added to the cells. After transduction, cells were cultured in a final concentration of 0.5 μg/mL of puromycin and maintained for one passage before collection for Western blot assays (2–3 days after transduction).

For transient OE, specified plasmids were used and NT2/D1 cells were transfected by polyethylenimine. Cells were harvested 24 hours after transfection for further analysis as described. Empty vector (EV) pCDNA3 plasmid was used as a control. TAp73α-overexpressing plasmid was a gift from William Kaelin (Addgene plasmid #22102). β-Actin full-length overexpressing plasmid was a gift from Robert Singer (Addgene plasmid #27123). SOD1 WT overexpressing plasmid was a gift from Elizabeth Fisher (Addgene plasmid #26397). Oct4-overexpressing plasmid was a gift from Ruldolf Jaenisch (Addgene plasmid #20728).

Equal numbers of cells from each sample were seeded in 6-well plates containing 2 mL of culture medium. After 24-hour incubation, cells were treated with chemicals at the indicated concentrations. Adherent cells were dissociated at the indicated times with 0.05% trypsin-EDTA and then counted by trypan blue dye exclusion. The numbers of viable cells are presented as mean ± SD of three replicates for each sample. 5–(and–6) carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled cells were cultured for 4 days and then monitored in flow cytometry (FACSCalibur, BD Biosciences) for cell division through halving of CFSE fluorescence. The CSFE fluorescence halving was deconvoluted using Flowing Software (Turku Bioimaging).

Cells were lysed and proteins were extracted and resolved by SDS-PAGE as previously described (25). Protein was transferred onto nitrocellulose membranes (BioRad). Specific primary antibodies against the following proteins were used for immunoblotting: p73 (human: BD Biosciences, 558787; murine: Abcam, ab14430), Oct4 (Santa Cruz Biotechnology, sc-5279), Nanog (Santa Cruz Biotechnology, sc-134218), Sox-2 (Cell Signaling Technology, 2748), β3-tubulin (Santa Cruz Biotechnology, sc-80005), β-actin (Santa Cruz Biotechnology, sc-47778), caspase-3 (Cell Signaling Technology, 8G10), SQSTM1 (Cell Signaling Technology, 5114), LC3A (Cell Signaling Technology, 4599), LC3B (Cell Signaling Technology, 3868), p16 (Santa Cruz Biotechnology, sc-390485), p21 (Santa Cruz Biotechnology, sc-56335), SOD1 (Cell Signaling Technology, 4266), SOD2 (Cell Signaling Technology, 13141), ATG5 (Cell Signaling Technology, 8540), ATG7 (Cell Signaling Technology, 8558), GAPDH (Santa Cruz Biotechnology, sc-365062), and β-tubulin (Cell Signaling Technology, 2146). Secondary antibodies were as follows: horseradish peroxidase (HRP)–conjugated anti-mouse IgG (Jackson ImmunoResearch) and HRP-conjugated anti-rabbit IgG (Jackson ImmunoResearch). Detection was by chemiluminescence (ECL, ThermoFisher) using ChemiDoc Touch Imaging System (BioRad). Quantification was by densitometry using ImageJ software (NIH).

RNA was extracted from cultured cells using Trizol, and cDNA was synthesized using enzyme Superscript II (ThermoFisher) as previously described (25). The BioRad CFX96 PCR machine was used for the qRT-PCR, using SYBR Green Supermix (BioRad). All primers were purchased from Invitrogen. GAPDH was used for normalization of the genes of interest. The results were analyzed using 2^−ΔΔCT method and expressed as fold change to respective nontreated or scrambled controls.

Protein extraction and peptide digestion were performed as previously described (26). Dried peptides were labeled using 10-plex Tandem Mass Tag (TMT) reagents (Thermo Fisher) as previously described (27). Samples were mixed equally, desalted using solid-phase C18 extraction cartridges (Waters), and lyophilized.

TMT10-labeled samples were fractionated using high-pH reversed phase chromatography performed with an Onyx monolithic 100 × 4.6 mm C18 column (Phenomenex) previously described (26). Fractions were desalted using homemade stage-tips as previously described (28) and lyophilized. Each fraction was analyzed using an Orbitrap Velos Pro mass spectrometer (Thermo Fisher) using an MS3 method as previously described (29). Protein identification was performed using a database search against a human proteome database (downloaded from UniprotKB September, 2014) concatenated to a database of common proteomic contaminants. All FDR filtering and protein quantitation were performed as previously described (27). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers PXD005407 and 10.6019/PXD005407 (Reviewer login details: Username: [email protected]; Password: e9gSfbwl).

Senescence was detected by using the senescence β-galactosidase staining kit (Cell Signaling Technology) following the manufacturer's protocol as previously described (25). After 24 hours, the images were captured using a light microscope.

NT2/D1 cells were serum starved for 12 hours to synchronize their cell cycle to G0, and then 5 × 10⁵ cells were seeded in 6-well plates. After 36 hours, cells were harvested, washed, and resuspended in 0.5 mL of cold PBS. Note that 4.5 mL of 70% ethanol was added drop-wise while the samples were gently vortexed. Following 24-hour storage in −20°C, cells were washed and stained with a propidium iodide (PI) solution containing 0.1% v/v Triton X-100 in PBS, 0.02 mg/mL PI (ThermoFisher), and 0.2 mg/mL DNase-free RNase A (ThermoFisher). Samples were incubated at room temperature for 30 minutes, and PI expression was detected by flow cytometry (BD FACSCalibur). ModFit LT 4.1.7 (Verity Software House) software was used to evaluate the DNA content of single, live cells.

Cell death of scrambled and shp73 NT2/D1 cells was measured using AlexaFluor 488-anti-Annexin-V and peridinine chlorophyll protein (PerCP)-anti-7AAD (both from Biolegend) as per the manufacturer's instructions.

Adherent cells were dissociated using trypsin and incubated with 5 μmol/L of DCF prepared in FACS buffer (1% FBS and 5 mmol/L in PBS) for 30 minutes at 37°C (30). Cells were then analyzed using FL1 channel of FACSCalibur. Treating cells with 1 mmol/L of H₂O₂ for 30 minutes was used as positive control.

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured from cells in XF assay media (unbuffered DMEM containing 10 mmol/L glucose, 2 mmol/L glutamine, and 1 mmol/L pyruvate) under basal conditions and in response to 1 μmol/L oligomycin (O4876; Sigma), 1.5 μmol/L carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP; C2920; Sigma), 1 μmol/L rotenone (R8875; Sigma), and 1 μmol/L antimycin A (A8674; Sigma) on the XF24 extracellular flux analyzer (Seahorse Bioscience). Cells (4 × 10⁴) were grown in XF24 plates for 24 hours prior to analysis; OCR and ECAR were normalized to final cell number calculated after the completion of the assay. Basal OCR was calculated by subtraction of the residual rate after antimycin A treatment, maximal rate was calculated by subtraction of the residual rate after antimycin A treatment from FCCP-induced OCR, proton leak was calculated as the difference between OCR after oligomycin treatment and OCR after antimycin A treatment, ATP production was calculated by subtraction of OCR after oligomycin treatment from basal OCR, spare respiration capacity was calculated by the difference between maximal OCR and basal OCR, spare respiration capacity coupling efficiency was calculated by the dividend of basal OCR and ATP production, glycolytic capacity was calculated as the ECAR after oligomycin treatment, glycolytic reserve was calculated by the difference between glycolytic capacity and ECAR, and glycolytic reserve percentage was calculated by the dividend of glycolytic capacity and ECAR.

All values are expressed as mean ± SEM of three independent experiments. Statistical evaluation was performed with using two-tailed, Student t test with 95% confidence interval. P < 0.05 was considered as significant.

Considering the “tumor suppressing” activities of TAp73, we began by transiently overexpressing TAp73 in the well-characterized human ECSLCs, NT2/D1 (12, 31), and then measured their viability and proliferation. Unexpectedly, TAp73 OE in NT2/D1 ECSLCs only slightly decreased cell viability and proliferation as compared with EV control (Fig. 1A and B). Furthermore, TAp73 OE had upregulated the protein levels of stemness factors Nanog and Sox-2 (Fig. 1C) without affecting the mRNA (Fig. 1D) levels. Similarly, TAp73 OE had no effect on the expression of differentiation markers (Fig. 1E and F) or morphologic characteristics in ECSLCs (Fig. 1G). These findings raised the possibility that TAp73 may have additional functions other than its tumor-suppressing role in the context of ECSLCs; therefore, we next investigated the effect of TAp73 KD in ECSLCs. We screened a panel of human and murine shRNA clones obtained from Dharmacon (Supplementary Tables S1 and S2), and preliminary phenotypes were confirmed using at least two distinct shRNA clones. Surprisingly, TAp73 KD using two distinct shRNA clones drastically decreased the viability (Fig. 1H) and proliferation (Supplementary Fig. S1A) of NT2/D1 ECSLCs, and strongly downregulated the expression of pluripotency factors Oct4/POU5F1, Nanog/NANOG, and Sox-2/SOX2 at both protein (Fig. 1I) and mRNA (Fig. 1J) levels. To assess the self-renewal capacity of NT2/D1 cells following TAp73 depletion, the cells were transduced with control, nontargeting (shNS) or TAp73 (shp73) shRNAs, and then cultured in parallel using serum-free, low attachment tumorsphere culture conditions. We found that control cells formed significantly more tumorspheres as compared with TAp73 KD from two distinct shRNA clones. (Fig. 1K). We also found that control cells form tumorspheres that can be enriched by multiple passages; however, TAp73 KD cells were unable to form tumorspheres efficiently, and there was no increase in the number of tumorspheres formed over several passages (refs. 17, 32; Supplementary Fig. S1B). Together, these results indicate that TAp73 is required to maintain the stemness and self-renewal capacity of NT2/D1 ECSLCs.

When examined under microscope, TAp73 KD cells from two distinct clones showed extensive morphologic changes and displayed long dendritic outgrowths, a characteristic of a more differentiated phenotype, compared with the control NT2/D1 cells (Fig. 2A). This change in morphology and induction of differentiation was confirmed using immunofluorescent analysis of the distribution of the differentiation marker β3-tubulin (Fig. 2B). We found that the loss of TAp73 using two distinct shRNA clones increased the expression of the differentiation marker β3-tubulin (Fig. 2B and C), and we observed that β3-tubulin was localized in the periphery of the cells in the dendritic-like structures as compared with being more concentrated in the nucleus in control cells (Fig 2B). This change in morphology following TAp73 KD was complemented by substantial upregulation of differentiation markers from various cell lineages, including β3-tubulin/TUBB3 (neuronal progenitor lineage; both at protein and mRNA levels) and SPP1 and GATA6 (endodermal lineage), T (mesodermal lineage), and CDX2 (ectodermal lineage; at mRNA level; Fig. 2C and D). In addition to upregulating many standard differentiation markers, we noticed that both clones of TAp73 KD also increased the protein and mRNA levels of β-Actin/ACTB (Fig. 2D; Supplementary Fig. S1C). Although β-Actin is commonly used as a housekeeping gene, there are reports that β-Actin levels increase when cells undergo differentiation into muscle cells or neuronal lineage (33). Concurring with these observations, β-Actin OE alone significantly decreased the viability of ECSLCs (Supplementary Fig. S2A) and downregulated pluripotency factors while promoting differentiation markers (Supplementary Fig. S2B and S2C). In order to conclusively link the induction of differentiation following TAp73 KD in NT2/D1 ECSLCs with the downregulation of pluripotency genes, we replenished the major pluripotency factor Oct4 in TAp73 deficient cells, from two different shRNA clones, and analyzed the effect on growth and differentiation. We found that restoration of Oct4 expression restored the growth of TAp73 KD cells and suppressed the upregulated expression of differentiation marker β3-tubulin [Fig. 2E(i) and (ii); Fig. 2F]. These findings demonstrate that upregulation of differentiation is linked with a of loss of stemness genes following TAp73 KD in NT2/D1 ECSLCs. Taken together, these data reveal a new role for TAp73, usually known for its tumor-suppressing activities, in the regulation of pluripotency and differentiation of ECSLCs.

To further confirm and generalize the role of TAp73 in the regulation of stemness and differentiation, we extended our investigations in the mouse ECSLC line, P19 (12, 13), and the well-characterized human breast CSLC system (14). In line with our data from human NT2/D1 ECSLCs, TAp73 KD in P19 cells significantly inhibited cell viability (Supplementary Fig. S3A and S3B), decreased the expression of pluripotency factors, and upregulated the levels of differentiation markers (Supplementary Fig. S3C and S3D).

We also found that HMLER^shECad CSLCs has higher levels of TAp73 when compared with HMLE and HMLER (Fig. 2G). TAp73 KD using two distinct shRNA clones in HMLER^shECad CSLCs also reduced their growth (Fig. 2H), expression of the epithelial-to-mesenchymal marker N-Cadherin (Supplementary Fig. S3E), and capacity to form tumorspheres (Fig. 2I). These results demonstrate that tumor suppressor TAp73 is required for maintaining CSLC growth and stem-like properties.

To confirm whether our in vitro findings were applicable in an in vivo model, we xenografted NOD/SCID mice with NT2/D1 ECSLCs that were stably transduced with lentiviruses expressing either nontargeting shRNA (shNS) control or TAp73 shRNA, and monitored tumor growth. As shown in Fig. 3A and B, TAp73 KD NT2/D1 cells were significantly less tumorigenic, grew drastically slower, and had significantly lower tumor weights as compared with that of control NT2/D1 cells. Figure 3C shows the remarkable size difference between the tumors generated with NT2/D1 ECSLCs expressing either nontargeting control or TAp73 shRNAs. When analyzed directly in situ, tumors with TAp73 KD harbored significantly lower protein and mRNA levels of pluripotency factor Oct4/POU5F1 than those found in tumors with control shRNA (Fig. 3D–F). Furthermore, we also found that tumors with TAp73 KD harbored significantly lower protein and mRNA levels of and oncogene Survivin/BIRC5 (Fig. 3D–F). These results conclusively recapitulated the importance of TAp73 in the maintenance of ECSLCs growth and the expression of pluripotency factors, and demonstrated that our in vitro findings are translatable to an in vivo setting.

To confirm these findings in an additional in vivo model, we generated TAp73 KD using two distinct shRNA clones in HMLER^shECad BCSLCs and xenografted NOD/SCID mice in the mammary fat pad with either of these KDs or shNS control, and monitored tumor growth. Tumors developed from both clones of TAp73 KD HMLER^shECad cells grew significantly slower and weighed significantly less as compared with the controls (Fig. 3G and H). Photographs demonstrating the size difference between tumors generated from shNS control or TAp73 KD clones are shown in Fig. 3I.

To comprehensively understand all the possible major pathways affected by TAp73 KD in ECSLCs, we quantitatively compared the global proteomes of NT2/D1 cells expressing nontargeting control and TAp73 shRNAs, using TMT-based multiplexed mass spectrometry (ref. 29; Fig. 3J). In line with our observations thus far, we found that TAp73 KD greatly decreased Oct4/POU5F1 levels and simultaneously upregulated many differentiation markers including AHNAK, known to inhibit stemness by regulating the expression of c-myc (34), and EFHD1, involved in neuronal differentiation (35). In addition, TAp73 KD affected many cellular processes involved in tumorigenesis and development including, metabolism, mitochondrial maintenance (TIMM9 and MRPS34), signal transduction (VSNL1 and GNAI3), positive regulation of immune response (LTF and C4BPA), cell cycle (TRIM36 and UBEC2C), and metastasis (Fig. 3J; Supplementary Fig. S4A). Overall, our proteomics data identify TAp73 as a regulator of various essential cellular processes involved in stemness regulation.

To address the possibility that the observed effects on pluripotency may be due to an increase in programmed cell death in response to TAp73 deprivation, we measured the levels of apoptosis in ECSLCs following TAp73 KD. We were unable to detect any discernible changes in the expression of cleaved caspase-3 by Western blot analysis in two distinct clones of TAp73 KD cells (Fig. 3K), or the expression of extracellularly exposed phosphatidyl serine (PS) by flow cytometry using Annexin-V/7-AAD stain (Supplementary Fig. S4B) as compared with nontargeting shRNA control, indicating no effect on apoptosis following TAp73 KD in ECSLCs.

Although TAp73 KD did not promote apoptosis in ECSLCs, we found that TAp73 KD drastically upregulated both protein and mRNA levels of the cell cycle and senescence regulators p21/CDKN1A and p16/CDKN2A (cyclin dependent kinase inhibitor 2A; Fig. 3L; Supplementary Fig. S4C and S4D). Using a widely performed β-galactosidase staining assay to detect the presence of senescent cells, we observed that ECSLCs with TAp73 KD from two distinct clones contained greater β-galactosidase activity as compared with the control (Fig. 3M; Supplementary Fig. S4E). We also discovered that TAp73 KD cells had a higher percentage of cells in G₁–G₀ phase and a lower percentage of cells in G₂–M phase (Supplementary Fig. S4F and S4G) compared with control cells, indicating increased cell-cycle arrest at the G₁–S checkpoint.

On the same note, TAp73 KD from two distinct clones in ECSLCs reduced phosphorylation of Akt at Ser473 (Fig. 4A), a major regulator of cell growth as well as Oct4-dependent stemness of ECSLCs (12). Similarly, TAp73 also promoted the phosphorylation of PTEN, a major antagonist of p-Akt (Fig. 4A). These results together suggest that one mechanism by which TAp73 decreases ECSLC growth is via inhibition of AKT phosphorylation by possibly promoting the activation of PTEN.

Recent reports have shown that senescence can be induced in stem cells through fluctuations in autophagy, an important homeostatic process required for maintaining stemness (31, 36). In line with these reports, our data show that TAp73 KD from two different clones increased the mRNA and/or protein expression of autophagy markers ATG5, ATG7, BECN1, LC3A-II, and LC3B-II and decreased levels of the autophagy target protein SQSTM1/p62 (Supplementary Fig. S5A and S5B). In autophagy flux analysis assay [Fig. 4B(i) and (ii)], treatment with chloroquine simultaneously enhanced the levels of LC3A-II and LC3B-II as well as SQSTM1 in ECSLCs with TAp73 KD from two different clones, as compared with the control, and further confirmed that TAp73 KD promotes autophagy in ECSLCs. (37). This increase in autophagy following TAp73 KD can be attributed to inhibition of a major negative upstream regulator of autophagy, phosphorylated mTOR, and its downstream targets p-P70S6 kinase/RBPS6KB and p-4EBP1/EIF4EBP1 (Fig. 4C; Supplementary Fig. S5C; refs. 38, 39). Furthermore, TAp73 KD in NT2/D1 cells increased the levels of negative upstream regulators of p-mTOR, p-AMPK, p-TSC2, and TSC1, indicating that the increase in autophagy following TAp73 KD could stem from energy depletion (Fig. 4C; ref. 40). To confirm this, we measured ATP levels in cells cultured in normal media containing glucose as well as galactose-supplemented media to promote oxidative phosphorylation, and we found that TAp73 KD decreased cellular ATP levels in both conditions (Fig. 4D). As summarized in Supplementary Fig. 5SD, our data reveal that TAp73 KD–related effects on stemness are linked with the dephosphorylation of Akt, senescence, cell-cycle arrest, and energy deprivation–induced autophagy. Taken together, these findings also suggest that TAp73-mediated effects on stemness in ECSLCs are not linked with apoptosis.

TAp73 KD–mediated ATP decrease and subsequent autophagy induction suggested that TAp73 deficiency–related effects on stemness may have a link with energy metabolism in ECSLCs (41). Hence, using a Seahorse extracellular flux analyzer, we performed real-time cellular bioenergetics analysis. For this, a Mito Stress test was performed, and the OCR and the ECAR were assessed as a measure of mitochondrial respiration and lactate secretion, respectively. As shown in Fig. 4E (Supplementary Fig. S6A–S6D), KD of TAp73 decreased lactate secretion, as evidenced through significantly decreased ECAR, and a trend of decreased OCR in ECSLCs. This decrease in lactate secretion is likely due to a decrease in glycolysis as opposed to redirection of pyruvate into the mitochondria and oxidative phosphorylation because we did not observe an increase in mitochondrial respiration. The data suggest that the TAp73 KD–induced drop in cellular and mitochondrial ATP levels could arise from lowered glycolysis and oxidative phosphorylation.

To further elucidate all the metabolites affected by KD of TAp73 in ECSLCs, we performed comprehensive mass spectrometry–based metabolomics analysis (Fig. 4F–K; Supplementary Fig. S6E). We found that TAp73 KD decreased the levels of intracellular NAD⁺, an important electron carrier required to maintain glycolysis during cell proliferation (ref. 42; Fig. 4G). This decrease in NAD⁺ may be due to decreased NAD⁺ recycling from pyruvate to lactate conversion (Supplementary Fig. S6F), as we observed that TAp73 KD decreases lactate secretion (ECAR; Fig. 4E). Our metabolomics data also revealed that TAp73 KD specifically downregulated the levels of the NEAAs proline and glutamate without any discernible effect on the other nine NEAAs (Fig. 4H and I; Supplementary Fig. S6E). In line with this, we also found that glutamate-derived metabolites such as glutamine and the antioxidant glutathione disulfide were decreased in TAp73 KD ECSLCs (Fig. 4H and J). Proline and glutamine metabolism is linked through the PRA, where glutamine can be converted to glutamate, proline, or urea (43). We found that the levels of urea cycle metabolites, arginine, aspartate, citrulline, ornithine, and urea were increased following TAp73 KD (Fig. 4K), indicating a shift toward increased urea cycle metabolism. In line with our metabolomics data, we found that TAp73 KD from two distinct clones decreased the mRNA levels of pyrroline-5-carboxylate reductase 1 (PYCR1; catalyzes proline synthesis from P5C) and glutaminase (GLS; converts glutamine to glutamate) and increased the expression of ornithine aminotransferase (OAT; reversible conversion of P5C into ornithine) with no effect on the other enzymes in the interconversion between proline, glutamine, and urea (Fig. 4L; Supplementary Fig. S6G). We further confirmed these findings in our in vivo tumor samples, where TAp73-deficient tumors had significantly less PYCR1 and GLS mRNA and higher levels of OAT mRNA (Fig. 4M). These data suggest that TAp73 KD leads to a shift from proline and glutamate synthesis to increased shuttling of metabolites into the urea cycle.

Considering the fact that TAp73 modified enzymes involved in proline and glutamine synthesis, we attempted to rescue TAp73 KD–mediated effects on ECSLCs by replenishing these depleted metabolites. We measured the growth of TAp73 KD ECSLCs in the presence or absence of additional glutamine or proline, and found that with the replenishment of glutamine (not proline) by addition of Glutamax to p73 KD cells from two different shRNA clones, we were able to rescue the growth of TAp73 KD ECSLCs to levels comparable with control cells (Fig. 5A; Supplementary Fig. S7A). Furthermore, replenishment of glutamine in TAp73 KD ECSLCs from two distinct shRNA clones restored the expression of pluripotency factor Oct4 and reversed the upregulation of differentiation marker β3-tubulin and autophagy protein LC3B-II (Fig. 5B and C; Supplementary Fig. S7B and S7C). Together, these results highlight that glutamine deprivation, as opposed to proline, is more critical for TAp73 KD–mediated effects and indicate that TAp73 KD cells may require supply of external glutamine to maintain their growth rate, pluripotency, and undifferentiated state. Supporting this hypothesis, we also found that culturing in glutamine-free media showed exacerbated inhibition of pluripotency (as evident by further decrease in Oct4) in TAp73 KD ECSLs (Supplementary Fig. S7D). We attributed this decrease in pluripotency to an increase in apoptosis as opposed to an increase in differentiation, as we found that the combination of TAp73 KD and glutamine starvation decreased β3-tubulin expression and promoted the activation of cleaved caspase-3 (Supplementary Fig. S7D and S7E). Of note, this increase in apoptosis was accompanied by further increase in autophagy, as the combination of TAp73 KD and glutamine starvation further enhanced the degradation of SQSTM1 and LC3B-II proteins (Supplementary Fig. S7F) as compared with TAp73 KD alone. Together, these data suggest that the combination of TAp73 KD and glutamine starvation promotes both apoptosis and autophagy and thus simultaneously reduces both pluripotency and differentiation in ECSLCs.

In light of our metabolomics and qRT-PCR data showing the decreased level of glutamine in TAp73 KD cells, we asked if TAp73 ECSLCs have impaired glutamine uptake. The oncogenic transcription factor Myc is a major regulator of proline and glutamine (44) and is known to promote glutamine uptake by increasing the expression of the glutamine receptor SLC1A5 (45). We found that TAp73 KD decreased the protein expression of Myc, which was accompanied by a decrease in the mRNA expression of SLC1A5 (Supplementary Fig. S7G and S7H). These findings suggest that the decreased glutamine levels in TAp73 KD ECSLCs may arise from the impaired glutamine uptake resulting from the decreased levels of Myc-dependent SLC1A5 expression.

A major characteristic of cells undergoing metabolic stress and deficiency of glutamate is the decrease of glutathiones and accumulation of ROS. We found that ECSLCs with TAp73 KD contained significantly higher levels of ROS, as measured using a DCF-based ROS assay, as compared with the control (Supplementary Fig. S8A and S8B). Further mechanistic analysis revealed that TAp73 KD decreased the protein and mRNA levels of the antioxidant enzyme SOD1 but not of SOD2 (Fig. 5D; Supplementary Fig. S8C). Of note, the finding that TAp73 KD decreases SOD1 levels was also confirmed in our quantitative proteomics database (Supplementary Fig. S8D) as well as in HMLER^shECad CSLCs (Supplementary Fig. S8E). In contrast, TAp73 KD had no effect on SOD1 expression in breast cancer cells, HMLER, indicating that TAp73 differentially regulates SOD1 in stem-like cancer cells as compared with their differentiated, non–stem-like counterparts (Supplementary Fig. S8F). Most importantly, Western blot analysis of tumor lysates collected from our in vivo experiment described in Fig. 3A–D conclusively showed that tumors with TAp73 KD had significantly lower SOD1 levels as compared with the control (Fig. 5E). Altogether, these data show that TAp73 regulates the expression of SOD1 and plays an important role in the regulation of the antioxidant defense system in ECSLCs. In support of this notion, we found that SOD1 KD alone was able to reproduce the phenotype of TAp73 KD on ECSLC growth, pluripotency, and differentiation (Supplementary Fig. S8G–S8L).

To further probe the interplay between ROS, TAp73, and the pluripotency of ECSLCs, we analyzed TAp73 levels following SOD1 KD, H₂O₂ treatment, and Oct4 KD. We found that SOD1 KD as well as H₂O₂ treatment increased the levels of TAp73 (Supplementary Fig. S8M and S8N). Furthermore, we found that Oct4 KD decreased the levels of SOD1 and promoted ROS and increased TAp73 expression in ECSLCs (Supplementary Fig. S8O and S8P). These observations further support the notion that TAp73 acts as a redox-sensitive protein (46, 47) in response to increased accumulation of ROS and is required for maintaining the antioxidant defense system in ECSLCs.

To understand the importance of SOD1 downregulation in TAp73 deficiency–related effects on growth and pluripotency in ECSLCs, we replenished the expression of SOD1 in TAp73 KD cells from two different clones and measured the effect on growth and the expression of pluripotency factor Oct4. We found that restoration of SOD1 expression effectively rescued the growth of TAp73-deficient ECSLCs (Fig. 5F) and restored the expression of Oct4 (Fig. 5G and H). Finally, to conclusively support the role of ROS in TAp73 KD–mediated decrease in growth and stemness, we found that the treatment of two different clones of TAp73 KD ECSLCs with the pharmacologic antioxidant compounds NAC or Mitotempo (48, 49) partially rescued the growth and expression of Oct4 (Fig. 5I and J; Supplementary Fig. S9A–S9C). Altogether, these findings further highlight the significance of TAp73 in antioxidant defense in the context of pluripotency of ECSLCs.

To confirm that the observed effects of TAp73 KD on ECSLCs are due to TAp73 deficiency, we restored TAp73 levels using transient TAp73 OE in NT2/D1 cells with TAp73 KD from two different clones. We found that OE of TAp73 in TAp73 KD NT2/D1 cells rescued the growth and the downregulation of pluripotency factor Oct4, reversed the upregulation differentiation marker β3-tubulin and autophagy protein LC3B-II, and importantly restored the expression of antioxidant enzyme SOD1 (Fig. 5K and L; Supplementary Fig. S9D and S9E).

To understand the clinical implications of our in vitro and in vivo findings, we extended our investigations in primary patient-derived BTICs. These BTICs are derived from surgically resected GBMs, sorted based on the expression of the surface marker CD133 using flow cytometric analysis (Supplementary Fig. S10) and display characteristics of stem-like features such as self-renewal and tumor initiation capacity (19, 24). We found that TAp73 KD in three distinct CD133^HI patient-derived BTICs (BT698, BT935, and BT954) severely inhibited their tumorsphere formation capacity [Fig. 6A(i) and (ii)]. Moreover, TAp73 KD in CD133^HI patient-derived BTICs strongly inhibited the expression of the established BTIC-associated stem cell markers Sox-2 and Bmi-1 (50) as well as p-Akt [Fig. 6B(i)]. In line with our data from ECSLCs, TAp73 KD in CD133^HI patient-derived BTICs leads to energy deprivation and subsequent autophagy activation as evidenced by decreased levels of p-mTOR and increased expression of p-AMPK, LC3A-II, and LC3B-II [Fig. 6B(ii)]. Furthermore, TAp73 KD in CD133^HI patient-derived BTICs strongly decreased the expression of SOD1 [Fig. 6B(iii)].

In contrast to CD133^HI GBM cells, TAp73 KD in CD133^LO patient-derived GBM cells BT602 regulated the expression of metabolic, antioxidant, autophagy, and growth-related proteins in a opposite manner. We found that depletion of TAp73 in CD133^LO BT602 GBM cells promoted the expression of growth-related proteins Bmi-1 and p-Akt, in-line with the well-known role of TAp73 as a tumor suppressor [Fig. 6C(i)]. However, TAp73 KD in CD133^LO BT602 GBM cells slightly decreased the expression of the stem cell marker Sox-2 [Fig. 6C(i)]. Furthermore, TAp73 KD inhibited autophagy in CD133^LO BT602 GBM cells as evidenced by decreased expression of p-AMPK, LC3A-II, and LC3B-II with no discernable effect on p-mTOR levels [Fig. 6C(ii)]. Lastly, we found that TAp73 depletion in CD133^LO BT602 GBM cells led to a substantial upregulation of SOD1 expression [Fig. 6C(iii)]. Taken together, this comparative analysis of patient-derived CD133^HI and CD133^LO GBM cells directly demonstrates the differential role of TAp73 in stem-like cancer cells versus non–stem-like differentiated cancer cells, and confirmed the clinical relevance of TAp73 inhibition-mediated effects on the growth and stemness of stem-like cancer cells.

Contrasting traditionally acknowledged antitumor activities of tumor suppressors, we describe that TAp73 is required to maintain the stemness of ECSLCs and CSLCs. Compared with differentiated cancer cells where TAp73 OE previously decreased cell viability (37, 51), we demonstrate a novel role for TAp73 KD in inhibiting ECSLCs' viability and tumorigenesis. Considering that therapeutic strategies aim to promote tumor suppressors while targeting heterogeneous cancers, our observation that TAp73 KD, and not OE, inhibits stem-like cancer cells is clinically relevant.

Regarding the role of TAp73 in maintaining cancer cell stemness, we observed that TAp73 KD inhibits the expression of master regulators of pluripotency and promotes multilineage differentiation of human and murine ECSLCs. Furthermore, xenografts of TAp73-deficient ECSLCs demonstrated drastically reduced tumorigenicity, with tumors harboring significantly lower levels of pluripotency factor Oct4. It is possible that the drastic inhibition of tumor growth observed in TAp73 KD cell xenografts could be due to either poor survival of TAp73 KD NT2/D1 cells following transplantation or low proliferative rate of TAp73 cells. However, we found no upregulation of activated cleaved caspase-3 or RIP3K in shp73 tumors compared with control tumors, indicating the absence of apoptosis or necroptosis following TAp73 KD (Supplementary Fig. S11). These results indicate that the decreased tumor size in shp73 NT2/D1 cells most likely stems from their lower proliferative capacity.

Numerous mechanisms of TAp73 tumor-suppressing functions have been identified (6, 52); however, our study demonstrates that the mechanisms by which TAp73 regulates ECSLCs' development are opposite to those observed in cancer cells. Although TAp73 has been shown to negatively regulate oncogene Akt in cancer cells (51), we report that inhibition of TAp73 blocks Akt phosphorylation by promoting the tumor suppressor p-PTEN in ECSLCs. Of note, p-Akt has been shown to be an important regulator of stem cell growth and to reciprocally regulate master pluripotency factor Oct4 (12) and autophagy (53). In concurrence, our TAp73 KD in NT2/D1 cells inhibited p-Akt phosphorylation while also lowered Oct4 levels and enhanced autophagy. Further, in line with a recent report (36), upregulation of autophagy in TAp73 KD cells was complemented by increased senescence and cell-cycle arrest via p21/CDKN1A and p16/CDKN2A. These findings in ECSLCs stand in contrast to the observations made in differentiated cancer cells, wherein TAp73 OE positively regulated autophagic cell death (51, 54) and p21/CDNK1A expression to promote senescence and cell-cycle arrest (55).

Unlike differentiated cancer cells, ECSLCs maintain low ROS levels and high expression of antioxidant proteins (56). ROS accumulation is known to promote stem cell differentiation (57). TAp73 is a redox-sensitive protein that has been shown to respond to oxidative stress, although the mechanism by which TAp73 regulated redox homeostasis was unclear (46). Our findings demonstrate that TAp73 maintains the antioxidant defense system in ECSLCs by regulating the expression of antioxidant enzyme SOD1 (but not SOD2). TAp73 KD decreased SOD1 levels in ECSLCs, and KD of SOD1 alone was sufficient to inhibit the expression of pluripotency factors and promote differentiation, highlighting its importance in TAp73 KD–related effects. Addition of antioxidants Mitotempo or NAC partly rescued levels of Oct4 in TAp73 KD cells, indicating that upregulation of ROS is important for TAp73 KD–mediated effects on ECSLC pluripotency. Moreover, TAp73 is strongly upregulated during oxidative stress induced by SOD1 KD or H₂O₂ treatment, and in response to pluripotency inhibition via Oct4 KD. These results further highlight the importance of TAp73 in maintaining redox homeostasis in ECSLCs.

Our findings also complement a recent study which showed TAp73 positively regulates metabolic pathways during oxidative stress in mouse embryonic fibroblasts (47). We also found that TAp73 KD–mediated oxidative stress was complemented by changes in the metabolic profile of ECSLCs, where TAp73 KD cells contained decreased ATP levels, glycolysis, and NAD⁺ levels. Furthermore, TAp73 KD downregulated the levels of proline, glutamine, and glutamate, without any significant effect on the other nine NEAAs. Interestingly, reduced proline levels were also accompanied by reduced mRNA levels of proline-synthesizing enzyme PYCR1 in ECSLCs with TAp73 KD. Recent reports have highlighted the importance of NEAAs in tumorigenesis (43, 58), and our findings are in line with a study by Sahu and colleagues (59), which showed that proline deprivation inhibits cancer cell growth by inactivating mTOR and promoting autophagy. Our findings also complement a recent study which showed that TAp73 positively regulates the growth of aggressive medulloblastoma (MB) cells by promoting glutamine addiction (60). This study demonstrated that specifically TAp73α (not ΔNp73) is overexpressed in aggressive MB subgroups and that the depletion of TAp73 in primary human MB cell lines deregulates glutamine metabolism by inhibiting the expression of GLS-2 and impairs mitochondrial bioenergetics, leading to activation of p-AMPK and inhibition of p-mTOR (60). Although this study did not directly examine the effect of TAp73 on tumor-initiating cell populations in MB, aggressive MB cell lines have been shown to contain cancer stem–like cell populations (61, 62). The findings from this report lend further credence to our hypothesis that TAp73 plays a unique role in stem-like cancer cells. Our results support previous findings that TAp73 regulates the enzyme GLS (60, 63); however, we also reveal a novel mechanism by which TAp73 KD decreases levels of glutamine and glutamate by impairing glutamine uptake via SLC1A5. It has been demonstrated that Myc is a major regulator of the glutamine to proline biosynthesis pathway (43). By providing a link between the TAp73 and Myc, our findings implicate a signaling network between TAp73, Myc, and glutamine uptake in the tumorigenic potential of ECSLCs. Moreover, our findings are in line with previous published results indicating that glutamine metabolism regulates the pluripotency transcription factor Oct4 (64); however, our results have linked TAp73-related glutamine deficiency with the pluripotency factor Oct4, differentiation marker β3-tubulin, and autophagy regulation, which to the best of our knowledge has not been reported previously in cancer stem–like cells. Our data support a model whereby TAp73 KD leads to decreased glucose and amino acid metabolism, which is accompanied by altered antioxidant defense system and autophagy, resulting in compromised cell viability and stemness within ECSLCs.

Altogether, the findings in this report bear immediate clinical implications aimed at cancer eradication. In recent years, the concept of eliminating CSLCs' populations within tumors has emerged as a valid strategy to prevent cancer recurrence and relapse (5). It should be noted that many currently practiced therapeutic strategies usually eliminate differentiated cancer cells while leaving highly tumorigenic stem-like cancer cells, suggesting that the complete eradication of cancers in clinics would probably require differential, and yet complimentary, approaches specifically designed against these two distinct populations. In line with these clinical observations, our study highlights the key differences between stem-like and differentiated cancer cell biology and advocates for multimodal therapeutic modulation of tumor suppressor TAp73. Although promoting TAp73's tumor-suppressing function is an attractive strategy for targeting differentiated cancer cells, the exact opposite approach, i.e., inhibition of TAp73, would be required to target constituting ECSLC and CSLC populations.

S.K. Singh is an employee of and has ownership interests (including patents) at Empirica Therapeutics, reports receiving speakers bureau honoraria from Amgen, is a consultant/advisory board member for Longbow Therapeutics, reports receiving commercial research grants from PTC Therapeutics, and reports receiving commercial research support from JanPix. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Sharif, S. Gujar

Development of methodology: T. Sharif, P.J. Murphy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Sharif, E. Martell, M.S. Ghassemi-Rad, M.R. Hanes, P.J. Murphy, B.E. Kennedy, C. Venugopal, S.K. Singh, S. Gujar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Sharif, C. Dai, E. Martell, M.R. Hanes, P.J. Murphy, M. Subapanditha, C.A. Giacomantonio, P. Marcato, S.K. Singh, S. Gujar

Writing, review, and/or revision of the manuscript: T. Sharif, C. Dai, E. Martell, M.S. Ghassemi-Rad, M.R. Hanes, C.A. Giacomantonio, P. Marcato, S.K. Singh, S. Gujar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Sharif, M.S. Ghassemi-Rad, M.R. Hanes, P.J. Murphy, S. Gujar

Study supervision: T. Sharif, S. Gujar

Others (performed experiments): C. Dai

This work was supported by grants from the Canadian Institute of Health Research (CIHR) and from the Canadian Breast Cancer Foundation—Atlantic. T. Sharif was previously supported by a CIHR Postdoctoral Fellowship and is currently funded by the Cancer Research Training Program (CRTP) of the Beatrice Hunter Cancer Research Institute (BHCRI). C. Dai was supported by the Masters Fellowship from CIHR, whereas E. Martell is supported through Nova Scotia Graduate Scholarship and Nova Scotia Health Research Foundation Masters award. CRTP awards from the BHCRI have funded B.E. Kennedy, P.J. Murphy, and M.R. Hanes in the past. S. Gujar is supported by the Dalhousie Medical Research Foundation. The authors are grateful to Dr. Patrick Lee for his support of this study. The authors thank Drs. Devanand Pinto (National Research Council) and Alehandro Cohen (Dalhousie Proteomics Core) for their assistance with proteomics and metabolomics analysis.

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.