Targeting tumor-initiating cells: eliminating anabolic cancer stem cells with inhibitors of protein synthesis or by mimicking caloric restriction.

We have used an unbiased proteomic profiling strategy to identify new potential therapeutic targets in tumor-initiating cells (TICs), a.k.a., cancer stem cells (CSCs). Towards this end, the proteomes of mammospheres from two breast cancer cell lines were directly compared to attached monolayer cells. This allowed us to identify proteins that were highly over-expressed in CSCs and/or progenitor cells. We focused on ribosomal proteins and protein folding chaperones, since they were markedly over-expressed in mammospheres. Overall, we identified >80 molecules specifically associated with protein synthesis that were commonly upregulated in mammospheres. Most of these proteins were also transcriptionally upregulated in human breast cancer cells in vivo, providing evidence for their potential clinical relevance. As such, increased mRNA translation could provide a novel mechanism for enhancing the proliferative clonal expansion of TICs. The proteomic findings were functionally validated using known inhibitors of protein synthesis, via three independent approaches. For example, puromycin (which mimics the structure of tRNAs and competitively inhibits protein synthesis) preferentially targeted CSCs in both mammospheres and monolayer cultures, and was ~10-fold more potent for eradicating TICs, than "bulk" cancer cells. In addition, rapamycin, which inhibits mTOR and hence protein synthesis, was very effective at reducing mammosphere formation, at nanomolar concentrations. Finally, mammosphere formation was also markedly inhibited by methionine restriction, which mimics the positive effects of caloric restriction in cultured cells. Remarkably, mammosphere formation was >18-fold more sensitive to methionine restriction and replacement, as directly compared to monolayer cell proliferation. Methionine is absolutely required for protein synthesis, since every protein sequence starts with a methionine residue. Thus, the proliferation and survival of CSCs is very sensitive to the inhibition of protein synthesis, using multiple independent approaches. Our findings have important clinical implications, since they may also explain the positive therapeutic effects of PI3-kinase inhibitors and AKT inhibitors, as they ultimately converge on mTOR signaling and would block protein synthesis. We conclude that inhibition of mRNA translation by pharmacological or protein/methionine restriction may be effective strategies for eliminating TICs. Our data also indicate a novel mechanism by which caloric/protein restriction may reduce tumor growth, by targeting protein synthesis in anabolic tumor-initiating cancer cells.


INTRODUCTION
Tumor-initiating cells (TICs) are known to be resistant to many conventional therapies, and have been implicated in disease recurrence and metastatic spread [1][2][3]. Residual TICs are linked to poor patient survival in multiple tumor types. As TICs are extremely rare and represent only a small fraction of the total cancer cell population, we still know very little about what allows them to survive and propagate, especially under the harsh conditions associated with chemo-and radio-therapy [1][2][3][4][5][6][7].
Remarkably, TICs are thought to pheno-copy many of the characteristics of normal epithelial stem cells, such as immortalization, asymmetric cell division and resistance to stressors, such as DNA damage [2,3,7]. Hence, the term cancer stem cells (CSCs) is now used virtually interchangeably with TICs [1][2][3][4][5][6][7]. Another hallmark of epithelial TICs is their ability to grow under anchorage-independent conditions, when cultured using low-attachment plates [8]. Under these anchorageindependent growth conditions, CSCs/TICs spontaneously form 3D spheroid structures or "tumor-spheres", that retain stem-like or progenitor cell properties. Conversely, under cell suspension conditions, most non-TICs undergo a form of apoptotic cell death, known as "anoikis". It has been shown that each tumor-sphere is derived directly from the proliferative clonal expansion of a single TIC, and not from the aggregation of bulk cancer cells [8]. Furthermore, the enriched TIC population is more resistant to radiotherapy, showing enhanced DNA damage repair and lower levels of reactive oxygen species (ROS) [9]. As such, the preparation of tumor-spheres represents an efficient and convenient method to selectively enrich for TICs. When these tumor-spheres are specifically generated from primary breast cancer cells or cell lines, they are known as "mammospheres".
Several mechanisms have been proposed to explain the increased resistance of TICs to clinical treatments. Firstly, both radiotherapy and the majority of chemotherapeutic treatments target rapidly dividing cells and it has been proposed that the resistance of TICs was due to them having a quiescent slow-cycling phenotype [10]. However, it has since been demonstrated that CSCs (at least in breast cancer) do cycle and their resistance to stress is not simply a function of quiescence [11,12]. Furthermore, CSCs have a greater capacity to efflux chemotoxins due to an increased expression of ABC transporters possibly explaining their additional resistance to chemotoxins but obviously these cannot account radio-resistance [13]. CSCs have been shown to be resistance to apoptotic stimuli compared to their nonstem cell counter parts and to have an increased capacity for DNA damage repair and this now seems to be the most likely mechanism of radio-and chemo-resistance [14,15]. Several studies have now shown that typical DNA damaging chemotherapeutic agents can even cause an upregulation of stem cell transcription factors and a direct conversion of cancer non-stem cells (bulk cells) into TICs, potentially increasing the TIC burden in patients rather than reducing it [16][17][18][19]. Thus, specifically targeting TIC populations in approaches that circumvent their resistance to DNA damaging therapy is a promising strategy for future cancer treatment.
To begin to understand the phenotypic behavior of TICs at a molecular level, we prepared large numbers of mammospheres from two different ER(+) breast cancer lines (MCF7 & T47D). The mammospheres were subjected to unbiased proteomic profiling to decipher their molecular composition and metabolic characteristics compared to the cells grown in monolayer. Based on proteomics analysis, we observed that mammospheres significantly upregulate molecules associated with protein synthesis, including ribosome-related proteins and proteinfolding chaperones, as well as specific molecules involved in mRNA translation initiation, polypeptide elongation, tRNA synthesis and amino acid uptake. We speculate that TICs are highly anabolic and increase their capacity for protein synthesis, to drive their clonal expansion via cell proliferation. Treatment with well-established inhibitors of protein synthesis (puromycin, rapamycin or methionine restriction) directly validated that mammosphere formation is strictly dependent on nacent protein synthesis. Thus, our results may also explain the anti-cancer health benefits of caloric restriction, intermittent fasting and the vegetarian diet, by systemically reducing protein synthesis in TICs.
Currently, there is a tremendous need to identify a selective "Achilles' Heel" to eliminate TICs. Our new results highlight that TICs are especially functionally dependent on augmented protein synthesis, for their successful survival and continued propagation. This study provides a strong rationale for therapeutically targeting protein synthesis in the CSC population.

RESULTS
Greater than 70 ribosomal protein components, as well as an isoform of S6 Kinase (RPS6KB1), are upregulated in mammospheres First, we performed unbiased label-free proteomic analysis on MCF7 cells, a commonly used ER-positive breast cancer cell line. Table 1 shows a non-redundant list data also indicate a novel mechanism by which caloric/protein restriction may reduce tumor growth, by targeting protein synthesis in anabolic tumor-initiating cancer cells.
Oncotarget 4587 www.impactjournals.com/oncotarget Oncotarget 4588 www.impactjournals.com/oncotarget of the 72 ribosome-related proteins that were selectively upregulated in MCF7 mammospheres, as directly compared with MCF7 cells derived from monolayers. Only proteins with a fold increase of ~1.5 or greater were selected for this analysis. Note that the expression levels of 21 large ribosomal proteins and 15 small ribosomal proteins were increased, as compared with monolayer cultures. Similarly, two large mitochondrial-specific ribosomal proteins were increased. A specific-isoform of ribosomal S6 kinase was also elevated (RPS6KB1) nearly 15-fold. Finally, 34 proteins involved in mRNA translation initiation, polypeptide elongation, tRNA synthesis and amino acid uptake, were all selectively upregulated in MCF7 mammospheres.
For comparison purposes, we also performed unbiased label-free proteomic analysis on a second independent ER-positive breast cancer cell line, namely T47D cells. Our results are summarized in Table 2. Note that 64 ribosome-related proteins were specifically over-expressed in T47D mammospheres, as compared with T47D monolayer cultures processed in parallel. Remarkably, 57 of these proteins overlapped with the proteins that were upregulated in MCF7 mammospheres (57/64 = 89% overlap). See the Venn diagram presented in Figure 1.  Note that 11 heat shock proteins, involved in protein folding, were commonly upregulated in both data sets.  the A-site of the ribosome. During protein synthesis, puromycin transfers to the growing polypeptide chain, leading to the generation of a puromycylated-peptide, which is prematurely released. As such, puromycin inhibits protein synthesis via a premature chain termination mechanism.

Heat shock proteins (HSPs)/protein-folding chaperones are upregulated in MCF7 and T47D mammospheres
Heat shock proteins are important for proper protein folding during protein synthesis. Table 3 shows a list of 13 heat shock proteins that were selectively upregulated in MCF7 mammospheres, relative to MCF7 monolayers. Note that 2 of these heat shock proteins are mitochondrialspecific chaperones (HSPA9 and HSPD1). Similarly, 11 heat shock proteins were specifically over-expressed in T47D mammospheres, as compared with T47D monolayer cultures processed in parallel (Table 4). Importantly, all eleven proteins overlapped with the proteins that were upregulated in MCF7 mammospheres (11/11 = 100% overlap). See the Venn diagram presented in Figure 2.
Taken together, our results predict that protein synthesis may be critical for the survival and propagation of cancer stem cells and/or progenitor cells.

Functional effects of puromycin, a known inhibitor of protein synthesis, on mammosphere formation.
Next, to functionally validate the hypothesis that mammosphere formation strictly requires protein synthesis, we used a highly-specific inhibitor that mimics the nucleotide-polypeptide linkage that occurs in tRNAs, namely puromycin [20]. By mimicking the structure of tRNAs, puromycin competitively inhibits protein synthesis. Also, puromycin is physically transferred to the growing polypeptide chain, leading to the generation of puromycylated-peptides, which are prematurely released ( Figure 3). Thus, puromycin inhibits protein synthesis via a premature chain termination mechanism [20]. Figure 4 shows the effects of increasing concentrations of puromycin on mammosphere formation, using an ER-positive breast cancer cell line (MCF7). Importantly, puromycin significantly reduced mammosphere formation, with an IC-50 of ~ 0.05 µg/ml and mammosphere formation was completed abolished at 0.5 µg/ml.
In striking contrast, monolayer MCF7 cells were ~10 times less sensitive to the effects of puromycin. Puromycin reduced both the i) proliferation and ii) viability of monolayer MCF7 cells, with an IC-50 of ~ 0.5 µg/ml, a concentration 10-times higher than the IC-50 for mammosphere formation ( Figure 4). In addition, mammosphere formation was inhibited by >70% at a concentration of 0.1 µg/ml; this same concentration had no effect on the viability of monolayer MCF7 cells.
In addition, when monolayer MCF7 cells were pretreated with puromycin for 4 days and then trypsinized and plated for mammosphere assays (in the absence of puromycin), mammosphere forming activity was completely eliminated by puromycin pre-treatment at 0.5 µg/ml ( Figure 5). At this same concentration, ~50% of monolayer MCF7 cells still remain viable and proliferate ( Figure 4). This indicates that protein synthesis inhibitors can preferentially target TICs, even in the presence of adjacent non-TIC cells, in the setting of an attached monolayer.
Thus, we conclude that TICs/CSCs are clearly more sensitive to the functional effects of inhibiting protein synthesis, directly supporting the results of our unbiased proteomics analysis.

Functional effects of rapamycin and methionine restriction on mammosphere formation
To further validate that mammosphere formation is functionally dependent on protein synthesis, we next used a well-established FDA-approved drug that potently inhibits protein synthesis, namely rapamycin [21][22][23]. Figure 6 shows the effects of increasing concentrations of rapamycin on mammosphere formation. Note that rapamycin significantly reduces mammosphere formation in MCF7 cells, with an IC-50 < 100 nM.
Complementary results were obtained with methionine restriction. Methionine is absolutely required for protein synthesis, since every new protein sequence    Oncotarget 4594 www.impactjournals.com/oncotarget starts with a methionine residue [24][25][26]. Note that mammosphere formation was dramatically inhibited by methionine deprivation (Figure 7), which mimics the positive effects of caloric restriction in cultured cells. Importantly, mammosphere formation was functionally restored to normal levels by the dose-dependent readdition of methionine to the methionine-free culture media, with maximal effects occurring at 0.01 mM (10 µM). It should be noted that mammosphere media normally contains 0.1 mM methionine (100 µM).
Remarkably, mammosphere formation appeared to be >18-times more sensitive to the effects of methionine deprivation and replacement, when compared directly with MCF7 cell monolayers (Figure 7). More specifically, the re-addition of 0.001 mM methionine stimulated MCF7 mammosphere formation by 22-fold; in contrast, the same concentration of methionine only stimulated MCF7 monolayer proliferation by 1.2-fold (Figure 7). Similar results were also obtained at higher concentrations of methionine (0.01 and 0.1 mM), with mammospheres again showing a >10-fold increase relative to monolayers.
In addition, MCF7 cell monolayers were pretreated with methionine at different concentrations (0, 0.001, 0.01 and 0.1 mM) for 4 days and then trypsinized and re-plated for mammosphere assays (in the presence of normal levels of methionine (0.1 mM)). Under these conditions, mammosphere-forming activity was significantly reduced by up to 3-fold (Figure 8; compare 0 vs. 0.1 mM). Importantly, methionine restriction did not affect MCF7 monolayer viability (as seen in Figure 7). Thus, methionine restriction is also effective at reducing "stemness" in the context of MCF7 cell monolayers.

Relevance of protein synthesis related targets in human breast cancers
To assess the clinical relevance of our results, we also determined whether the proteomic targets that we identified in mammospheres were transcriptionally overexpressed in human breast cancer cells in vivo. Towards this end, we exploited a clinical data set of tumor samples from 28 breast cancer patients [29,30]. These tumor samples were subjected to laser-capture micro-dissection, to separate epithelial cancer cells from adjacent stroma. Tables 5 and 6 present a summary of these findings. Overall, 60 proteomic targets that we identified in mammospheres were also transcriptionally elevated in human breast cancer cells in vivo and the majority of these targets were also upregulated either in MCF7 and/or T47D mammospheres. As such, the new protein targets that we identified in mammospheres may be especially relevant for improving human breast cancer diagnosis and therapy.

DISCUSSION
Here, using unbiased label-free proteomics analysis, we show that the cells of mammospheres (a population which is enriched for TICs and other progenitor cells) functionally overexpress numerous proteins, related to protein synthesis, including ribosomal biogenesis, mRNA translation initiation, polypeptide elongation, tRNA synthesis, amino acid uptake and protein folding. The potential clinical relevance of these targets was further validated using a previously published data set of human  1 mM)). Note that under these conditions, mammosphere-forming activity was significantly reduced by up to 3-fold (compare 0 vs. 0.1 mM). As such, methionine restriction is also effective at reducing "stemness" in the context of MCF7 cell monolayers.
Oncotarget 4596 www.impactjournals.com/oncotarget breast cancer samples (N=28 patients), that were subjected to laser-capture microdissection, to separate the epithelial tumor cells from the adjacent tumor stroma. Thus, these novel anabolic targets reveal a metabolic "Achilles' Heel" to allow the elimination of CSCs. In accordance with this idea, we demonstrate that the therapeutic targeting of protein synthesis in mammospheres (via puromycin, rapamycin or methionine-restriction) is indeed sufficient to prevent their proliferative expansion, as assessed using mammosphere formation as a functional assay (summarized schematically in Figure 9). In accordance with our results, a recent paper has shown that the mTOR inhibitor Torin-1 selectively targets human colon CSCs [31].
Recently, we also reported that mitochondrial oxidative metabolism is also markedly amplified in mammospheres, as evidenced by i) unbiased proteomics analysis and ii) functional validation with inhibitors of mitochondrial OXPHOS [32]. Thus, enhanced mitochondrial energy production could help directly "fuel" increased protein synthesis in CSCs, thereby driving and maintaining the anabolic phenotype of TICs ( Figure 10).

Figure 10: Augmented mitochondrial OXPHOS may help fuel increased protein synthesis.
Recently, we showed that mitochondrial oxidative metabolism is significantly amplified in mammospheres, as evidenced by i) unbiased proteomics analysis and ii) functional validation with inhibitors of mitochondrial OXPHOS. As such, enhanced mitochondrial energy production could help directly "fuel" increased protein synthesis in CSCs, thereby driving and maintaining the anabolic phenotype of TICs. Thus, inhibition of mitochondrial function and protein synthesis may both be beneficial. Figure 9: Anabolic tumor-initiating cells require protein synthesis for clonal expansion. Activation of the PI3-kinase/AKT/ mTOR signaling pathway is known to converge on ribosomal biogenesis and protein synthesis. Thus, the therapeutic effects of PI3-kinase inhibitors and AKT inhibitors, may be explained by their ability to inhibit mTOR signaling and block protein synthesis in TICs. Similarly, direct pharmacological inhibition of protein synthesis and/or caloric restriction or protein restriction may have similar beneficial therapeutic effects. Thus, our current data provide a novel convergent mechanism by which inhibitors of PI3-kinase, AKT, mTOR, as well as caloric restriction, may all directly target tumor-initiating cells, by inhibiting protein synthesis.

Deregulation of protein synthesis
Deregulation of protein synthesis is a relatively unexplored but emerging mechanism of cancer progression. Two of the better documented examples of this phenomenon are increased protein synthesis as a result of c-MYC and mTOR oncogenic signaling. Although better known for targeting genes involved in cell cycle regulation, it has been known for some time that c-MYC directly targets multiple components of the translational machinery including: RNA polymerases I, II and III; ribosomal proteins; translation initiation factors; elongation factors; and rRNA [33][34][35][36][37][38]. Until more recently, the consequence and importance of this protein synthesis up regulation has remained unknown and the complexity and breadth of c-MYC targets has made it a difficult question to address [39]. Specific ribosomal protein haploinsufficiency (L24 +/− and L38 +/− ) have recently helped to address the role of c-MYC induced protein synthesis [40]. It was thus demonstrated ribosomal protein haploinsufficiency is able to rescue mice from increased rates of protein synthesis downstream of oncogenic Eμ-Myc signaling. In these mice, the growth of Myc-overexpressing B cells was returned to normal and remarkably this was coupled with the restoration of cell division rates to near wild-type levels [40]. These results imply that c-MYC directly couples cell growth and cell division, at least in part, by a deregulation of protein synthesis. Furthermore, the oncogenic potential of c-MYC was strongly impaired by ribosomal protein haploinsufficiency genetic backgrounds, with the onset of lymphomas being dramatically delayed Eμ-Myc;L24 +/− and Eμ-Myc;L38 +/− mice compared to Eμ-Myc [40]. This may be due to an increase in the apoptotic response to Myc oncogenic activity observed with ribosomal protein haplo-insufficiency genetic backgrounds.

Methionine and calorie restriction
Methionine-free medium reduced the proliferation and viability of the cells in mammospheres. Reduction of methionine also reduces proliferation and increases apoptosis of embryonal stem cells (ESC) whilst apparently not affecting more differentiated cells [53]. ESC can be induced to form differentiated cardiac cells and a methionine free medium can be used to remove remaining ESCs which, if transplanted with cardiac cells form teratomas [54]. Our data is the first suggestion that epithelial TICs may be similarly sensitive to methionine depletion. It is well known that tumor cell lines and primary tumors in-vitro and when grown in-vivo in rodents are growth inhibited by methionine depletion whereas normal cells are unresponsive. Methionine restriction (MR) was shown to reduce the growth of sarcomas [55,56] [57], adenocarcinomas [25,58] and mammary tumors in rodents [59], as well as of human tumor cell lines [60], human primary tumors in-vitro [61] and tumor growth in nude mice [62,63].
Low methionine diets in rodents produce similar effects as calorie restriction [64]. Tumor formation and tumor growth is reduced and lifespan is increased although MR does not appear as potent as calorie restriction or general protein restriction [26]. MR may produce longevity effects, in part, acting through reduction of the formation of free radicals in mitochondria [64,65] and, in part, by inhibition of protein synthesis in precursor cells as demonstrated in this report. Also, methionine restriction increases the stress tolerance of human fibroblasts, reduces senescence and increases their doubling time [66,67]. In mammals, the effect of MR appears to be growth hormone dependent [68].
Vegans and vegetarians have relatively low intakes of methionine compared with meat eaters and some but not all recent studies suggest that vegans have lower cancer and cardiovascular disease risk [69][70][71]. It has been suggested that low-methionine vegan diets may be used as a feasible approach for life extension [72]. Ornish and his colleagues [73] were able to show that the concentration of prostate specific antigen (PSA) was reduced in men with prostate cancer treated only with lifestyle changes, which included a vegan diet and exercise. A recent phase I study demonstrated that methionine restriction was well tolerated for up to 17 weeks in patients with various solid tumors [74].

CONCLUSIONS
In conclusion, based on our current analysis using mammosphere cultures, we propose that inhibition of protein synthesis is a new therapeutic strategy for eradicating TICs, to potentially prevent tumor recurrence, metastasis and poor clinical outcome in breast cancer patients. This strategy might also be extended to other tumor types, as many of the phenotypic features of TICs are highly conserved between different epithelial cancer types.

Materials
Breast cancer cell lines (MCF7 and T47D) were purchased from the ATCC. Puromycin, rapamycin, methionine and methionine-free media were obtained commercially from Sigma-Aldrich. Gibco-brand cell culture media (DMEM/F12) was purchased from Life Technologies. 50,000 cells were plated in normal medium (DMEM, 10% FCS, L-glutamine, supplemented with Pen-Strep) for 24hr, followed by treatment with increasing concentrations of a specific inhibitor (puromycin) or nutrient (methionine) for a further 4 days. Cells were then collected by trypsinization and centrifugation. To quantitatively determine cell growth, the number of cells after dtreatment was counted using an automatic cell counter (Biorad) and differences compared to untreated cells was calculated and expressed as fold-change. To assess cell viability, cells were incubated for 1 minute with Trypan Blue (Sigma, #T8145) using a 1:1 ratio. The number of Trypan Blue positive cells (non-viable) was measured using an automatic cell counter (Biorad) and compared to untreated controls. For puromycin treatments, cells were also plated into mammosphere culture to assess stem cell-like activity with no further drug treatment. All monolayer experiments were performed in triplicate, three times independently, such that each data point represents the average of 9 replicates.

Mammosphere culture
To directly assess the effects of specific inhibitors (puromycin/rapamycin) or nutrients (methionine) on mammosphere formation, cultures were supplemented with increasing concentrations of puromycin, rapamycin or methionine, as indicated in a given experiment. A single cell suspension was prepared using enzymatic (1x Trypsin-EDTA, Sigma Aldrich, #T3924), and manual disaggregation (25 gauge needle) to create a single cell suspension [8,32]. Cells were plated at a density of 500 cells/cm 2 in mammosphere medium (DMEM-F12/ B27/20ng/ml EGF/PenStrep) in non-adherent conditions, in culture dishes coated with (2-hydroxyethylmethacrylate) (poly-HEMA, Sigma, #P3932). Cells were grown for 4-to-5 days and maintained in a humidified incubator at 37°C at an atmospheric pressure in 5% (v/v) carbon dioxide/air. After 5 days for culture, spheres >50 µm were counted using an eye piece graticule, and the percentage of cells plated which formed spheres was calculated and is referred to as percentage mammosphere formation, and was normalized to one (1 = 100 %MSF). All mammosphere experiments were performed in triplicate, three times independently, such that each data point represents the average of 9 replicates.

Label-free quantitative proteomics analysis
For proteomic analysis, mammospheres were collected by centrifugation at 800 rpm for 10 minutes. Cell lysates were prepared for trypsin digestion by sequential reduction of disulphide bonds with TCEP and alkylation with MMTS [32]. Then, the peptides were extracted and prepared for LC-MS/MS. All LC-MS/MS analyses were performed on an LTQ Orbitrap XL mass spectrometer (Thermo Scientific, San Jose, CA) coupled to an Ultimate 3000 RSLCnano system (Thermo Scientific, formerly Dionex, The Netherlands). Xcalibur raw data files acquired on the LTQ-Orbitrap XL were directly imported into Progenesis LCMS software (Waters Corp., Milford, MA, formerly Non-linear dynamics, Newcastle upon Tyne, UK) for peak detection and alignment. Data were analyzed using the Mascot search engine. Five replicates were analyzed for each sample type (N = 5). Statistical analyses were performed using ANOVA and only foldchanges in proteins with a p-value less than 0.05 were considered significant.

Data mining
To firmly establish the clinical relevance of our results from the quantitative proteomics analysis of mammospheres, we re-analyzed the transcriptional profiles of epithelial breast cancer cells and adjacent tumor stromal cells that were physically separated by lasercapture microdissection (from N=28 human breast cancer patients) [29,30].