Introduction

Cancer cells are known to adopt metabolic switch for energy supply and building blocks necessary for rapid proliferation under different stressed conditions. As two vital energy sources in favor of Warburg effect, glucose and glutamine are avidly used by proliferating cancer cells to provide not only ATP but also carbon and nitrogen sources for the biosynthesis of proteins, nucleotides and lipids, and reducing powers, etc.1,2,3,4. Glucose can sustain rapid cancer cell proliferation through high levels of glycolysis or providing metabolic intermediates for PPP pathway to synthesize biomacromolecules5, whereas glutamine contributes to multiple biosynthetic pathways and governs redox homeostasis in addition to supporting the TCA cycle and lipogenesis by decomposing into α-ketoglutarate (α-KG), especially under hypoxic conditions2,6,7,8. As a result, cancer cells are highly dependent on glucose and glutamine for survival and proliferation, and their deprivation leads to severe cell death. However, the metabolism of glucose and glutamine is complicated and their regulation mechanisms are not fully understood. Specifically, we are still unaware of what specific and detailed metabolic responses that cancer cells would undertake when deprived of glucose or glutamine.

Also critical for cell proliferation are serine and glycine, two non-essential amino acids in mammals9,10, which are known for their functions in phospholipid synthesis and cystathionine formation in central nervous system11,12. Interestingly, it has been reported recently that these two amino acids play important roles in cancer progression13,14,15,16. Glycolytic metabolism of glucose provides metabolite of 3-PG which can be catalyzed by 3-phosphoglycerate dehydrogenase (PHGDH), phosphoserine aminotransferase (PSAT1) and phosphoserine phosphatase (PSPH) into serine, during which glycine is produced through the inter-conversion between serine and glycine by serine hydroxymethyltransferase (SHMT1/2)17,18,19. Serine and glycine synthesis is accompanied by the production of GSH, a tripeptide that is comprised of glycine, glutamate and cysteine, contributing to the redox balance in cells as reactive oxygen species (ROS) scavenger20,21,22,23. Serine and glycine are also involved in nucleotide synthesis, thus entering into one-carbon metabolism and promoting cancer cell proliferation14,18,24. It was reported recently that, upon serine starvation, p53 activates p21 to promote GSH production to combat ROS25,26. Additionally, p73, ATF4, G9A, HIF1 and PKCζ were also reported to regulate serine biosynthesis and metabolism27,28,29,30,31. However, compared to the profound understanding of glycolysis and glutaminolysis in cancer cells, we are only beginning to appreciate the critical impact of serine synthesis pathway (SSP) on cancer progression. It remains largely unexplored as to how this pathway is regulated under normal or stressed conditions.

Among numerous notable transcriptional factors, cMyc is well known to regulate up to 10%-15% human genes involved in cell cycle, development, apoptosis and metabolism32,33,34. Interestingly, it has been suggested recently that cMyc is posed to amplify all the extant expressed genes35,36. More recently, new evidence emerges to reveal that cMyc selectively fine-tunes expression of many important genes essential for cell growth and cancer progression37,38,39. All these results point to the enigmatic roles and underlying mechanisms for cMyc yet to be further elucidated under developmental or oncogenic circumstances. cMyc has been extensively documented to regulate glucose, glutamine and nucleotides metabolism7,40,41,42,43, nevertheless, it is unclear whether cMyc also controls the serine/glycine metabolism, especially under nutrient-deprived conditions.

As cancer cells normally undergo a comprehensive metabolic reprogramming that involves multiple major metabolic pathways, we surmised that the metabolic networks of glucose, glutamine and serine/glycine are coordinately regulated and intertwined during cancer progression under stressed conditions. Here, we report that the expressions of all the major enzymes involved in SSP are comprehensively activated when cancer cells are deprived of glucose or glutamine. We further demonstrate that oncogenic cMyc plays a vital role in this process by coordination of glycolysis, glutaminolysis, and importantly, SSP to promote cancer cell survival and progression under the nutrient-deprived conditions. Moreover, we uncover that aberrant expression of SSP enzyme PSPH is critical for cMyc-mediated cancer progression. Taken together, our findings illustrate a previously unappreciated role for cMyc to orchestrate a comprehensive metabolic changes leading to SSP activation and cancer proliferation.

Results

The pathways leading to serine biosynthesis are activated under nutrient deprivation conditions

Previous studies have shown that serine is important for cancer proliferation and that serine/glycine deprivation triggers activation of SSP25. However, while glucose and glutamine are the major sources for energy supply and anabolic metabolisms for cancer cell proliferation, the effects of glucose/glutamine deprivation on activation of SSP and the underlying mechanisms are still largely unknown. Hence we first examined SSP enzyme expressions under glucose/glutamine deprivation conditions compared with serine/glycine deprivation conditions. Western blot assay revealed that the protein levels of 5 major SSP enzymes, PHGDH, PSAT1, PSPH, SHMT1 and SHMT2, were all elevated under glucose- or glutamine- or serine/glycine-deprived conditions in Hep3B human hepatoma cells. Similar results were observed in SK-hep-1 and Hela cells (Figure 1A). Additionally, the mRNA levels of these enzymes were also increased in Hep3B cells (Figure 1B), suggesting transcriptional activation of SSP genes under glucose- or glutamine- or serine/glycine-deprived conditions.

Figure 1
figure 1

The pathways leading to serine biosynthesis are activated under nutrient deprivation conditions. (A, B) Western blot (A) and qRT-PCR (B) analyzed the expression of SSP enzymes and cMyc in Hep3B, SK-hep-1 and Hela cells cultured with or without glucose, glutamine or serine/glycine for 48 h, respectively. β-actin serves as loading control. Data were presented as mean ± SD of three independent experiments. Schematic drawing on the right of A indicates the SSP pathway. *P < 0.05 as compared to control groups. (C, D) Western blot analyzed glutaminolysis enzymes (C) and glycolysis enzymes (D) in Hep3B cells cultured with or without glucose, glutamine or serine/glycine for 48 h, respectively. β-actin serves as loading control. Schematic drawing indicates glutaminolysis pathway (C) and glycolysis pathway (D).

Full size image

The pathways leading to SSP are complicated. In brief, glucose and glutamine can furnish the precursors 3-phosphoglycerate (3-PG) and glutamate, respectively, to fuel serine synthesis (Figure 1A, right). To determine which nutrient(s) or metabolic pathway(s) are upstream and essential for SSP activation under nutrient deprivation conditions, we studied more global changes in protein levels of enzymes involved in glycolysis and glutaminolysis along with that of SSP. As expected, we detected higher expression of GLS1 under glucose-free conditions and elevated HK2 level under glutamine-deprived conditions (Figure 1C). And the elevated GLS1 and HK2 levels were also found under serine/glycine starvation conditions (Figure 1C). It is intriguing that, while Glutamic-oxaloacetic transaminase-Malate dehydrogenase-NADP-dependent malic enzyme (GOT1-MDH1-ME1) pathway was activated, there is no major difference of GLUD1 and GOT2 under 3 nutrient-deprived conditions, suggesting that the flux shifted to gluconeogenesis, ultimately producing 3-PG to fuel SSP under glucose- or glutamine-deprived conditions (Figure 1C). We also examined glycolytic enzymes and found that protein levels of the major glycolytic enzymes increased under glucose- or glutamine-free conditions (Figure 1D). It should be noted that PGK1 and PGAM1 also increased under serine/glycine starvation conditions (Figure 1C). Taken together, the activation of enzymes involved in glucose/glutamine metabolism suggests potential different routes by which glucose or glutamine sustains SSP: (1) glucose → 3-PG → SSP (via PGK1 and PGAM1); (2) glutamine → glutamate → SSP (via GLS1); (3) glucose/glutamine → Aspartate (Asp) → Oxaloacetate (OAA) → malate → pyruvate → 3-PG → SSP (via GOT1-MDH1-ME1; Figure 1A, 1C and 1D). Collectively, these results suggest that SSP is activated under nutrient stressed conditions by integration of glycolysis or glutaminolysis pathways.

cMyc transactivates the expression of enzymes involved in serine biosynthesis

Next, we were prompted to further study what oncogenic lesions might coordinate the changes in glycolysis, glutaminolysis and SSP activation. Owing to the well-established fact that cMyc deregulation positively correlates with elevated rates of glycolysis and glutaminolysis in various human cancers41,44, we performed microarray profiling to determine the global roles that cMyc might have to coordinate the SSP activation. Our array data revealed that cMyc had profound effects on expression of SSP genes as well as genes involved in glycolysis and glutaminolysis (Supplementary information, Figure S1A). More intriguingly, both cMyc mRNA and protein expression were significantly induced by glucose or glutamine starvation, whereas serine/glycine starvation elevated cMyc protein levels without significant effect on its mRNA level (Figure 1A and 1B), further indicating that cMyc is involved in cellular responses to nutrient starvation stress.

Quantitative real-time PCR (qRT-PCR) and western blot with cMyc-overexpressing or -knockdown Hep3B cells further demonstrated that PHGDH, PSAT1, PSPH, SHMT1 and SHMT2 are positively regulated by cMyc (Figure 2A and 2B). In human P493 B lymphoma cells that bear a tet-repressible cMyc construct, we observed that mRNA levels of PHGDH, PSAT1, PSPH, SHMT1 and SHMT2 were decreased in the presence of tet in a time-dependent manner (Supplementary information, Figure S1B). Protein expressions of PHGDH, PSAT1 and PSPH were markedly reduced when tet was added in culture medium and recovered after washing off tet to allow cMyc expression in P493 cells (Figure 2C). ChIP-qPCR analysis with Hep3B cells demonstrated direct binding of cMyc to the E-box sequences close to the transcription start sites of PHGDH, PSAT1 and PSPH genes (Figure 2D). Similar results were observed in P493 cells (Supplementary information, Figure S1C). Furthermore, downregulation of cMyc by shRNAs attenuated nutrient deprivation-induced high expression of SSP enzymes in Hep3B cells (Figure 2E), suggesting that cMyc is critical for nutrient starvation-induced SSP activation. Finally, NMR analysis using 13C-labeled glucose demonstrated that 13C incorporation into both glycine and GSH was significantly increased in cMyc-overexpressing Hep3B cells (Figure 2F), further demonstrating that cMyc promotes SSP activation, driving metabolic flux to glycine and GSH. Our data also showed that cMyc regulates glutaminolysis and glycolysis enzymes both at mRNA and protein levels in Hep3B cells (Supplementary information, Figure S1D-S1F). Moreover, cMyc deficiency abolished their induction under glucose or glutamine starvation conditions (Figure 2G). Taken together, our data indicate that cMyc plays important roles in cellular response to glucose or glutamine starvation conditions.

Figure 2
figure 2

cMyc transactivates the expression of enzymes involved in serine biosynthesis. (A, B) qRT-PCR (A) and western blot (B) analyzed the expression of SSP enzymes in Hep3B cells expressing cMyc or shRNAs against cMyc. β-actin serves as loading control. Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to empty vector (EV) or non-target control (NTC) groups. (C) Western blot analyzed the expression of SSP enzymes in P493 cells treated with tetracycline (Tet) for the indicated hours with or without washing off Tet followed by continued culturing for the indicated hours. β-actin serves as loading control. (D) ChIP-qPCR analyzed the occupancy of potential E-boxes by cMyc in the genes of PHGDH, PSAT1 and PSPH in Flag-cMyc-overexpressing Hep3B cells using IgG or anti-cMyc antibody. Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to IgG group. (E) Western blot analyzed the expression of cMyc and SSP enzymes in Hep3B cells expressing cMyc shRNAs cultured with or without glucose, glutamine or serine/glycine for 48 h, respectively. β-actin serves as loading control. (F) NMR data for [U-13C] glucose incorporated into GSH and glycine in Hep3B cells overexpressing cMyc. Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to the corresponding EV group. (G) Western blot analyzed the expression of glycolysis and glutaminolysis enzymes critical for SSP in Hep3B cells expressing cMyc shRNAs cultured with or without glucose, glutamine or serine/glycine for 48 h, respectively. β-actin serves as loading control.

Full size image

cMyc-mediated PSPH expression and SSP activation are critical for cancer cell proliferation by regulating GSH, ROS, apoptosis and nucleotide synthesis

SSP supports anabolism by providing precursors for biosynthesis of proteins, nucleotides, phospholipids and GSH15,45. Our NMR data from 13C-labeled glucose showed that cMyc facilitates cellular GSH production (Figure 2F). This result is consistent with a previous study by Valencia et al.46 demonstrating that downregulation of cMyc results in reduced glucose flux into serine in stromal cells. To further confirm the NMR results, we measured GSH level and GSH/GSSG ratio in cultured cells in vitro and the results demonstrated that both GSH level and GSH/GSSG ratio were significantly decreased in Hep3B cells with cMyc shRNA expression (Figure 3A). As expected, knockdown of PSPH, the final rate-limiting enzyme of SSP pathway, reduced both GSH level and GSH/GSSG ratio in Hep3B cells (Figure 3B). Similar results were obtained in Hep3B cells with PHGDH or PSAT1 knockdown (Figure 3C). Further, suppression of PSPH abolished cMyc overexpression-induced increase in GSH level and GSH/GSSG ratio (Figure 3D). On the contrary, overexpression of PSPH can only partially rescue the decreased GSH level, GSH/GSSG ratio and cell growth rate induced by cMyc knockdown, accompanied by alleviated apoptosis (Figure 3E-3G). This is not surprising, considering the enormous influence of cMyc on the proliferation and survival of cancer cells. These results indicate that SSP activation is, at least partially, involved in cMyc-mediated regulation of GSH production. Consistent with the GSH results, Hep3B-shPSPH cells exhibited increased cellular ROS levels and cell apoptosis, accompanied by decreased population of S phase that is responsible for DNA biosynthesis (Figure 3H-3J). Similar results were observed in SK-hep-1 cells with PSPH knockdown (Supplementary information, Figure S2A-S2D). Moreover, in Hep3B cells with PHGDH or PSAT1 knockdown, we obtained essentially the same results of the variation of cellular ROS levels, apoptosis and S phase of cell cycle (Supplementary information, Figure S2F-S2H). Consistently, our NMR data using 13C-labeled glucose showed that PSPH overexpression enhanced the 13C incorporation into GSH as well as glycine (Figure 3K), supporting that PSPH indeed promotes SSP and GSH production.

Figure 3
figure 3

cMyc-mediated PSPH expression and SSP activation are critical for cancer cell proliferation by regulating GSH, ROS, apoptosis and nucleotide synthesis. (A-C) GSH level and the GSH/GSSG ratio were determined in Hep3B cells expressing shRNAs targeting cMyc (A), PSPH (B), PHGDH or PSAT1 (C). Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to NTC group. (D) GSH level and GSH/GSSG ratio were determined in cMyc-overexpressing Hep3B cells with PSPH knockdown by shRNAs. Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups. (E-G) GSH level and GSH/GSSG ratio (E), cell growth and western blot (F), apoptosis (G) were determined in cMyc-knockdown Hep3B cells with Flag-PSPH overexpression. Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups. (H-J) ROS (H), apoptosis (I) and cell cycle (J) were analyzed by flow cytometry in Hep3B cells expressing PSPH shRNAs. Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to NTC group. (K) NMR data for [U-13C] glucose incorporated into GSH and glycine in Hep3B cells overexpressing PSPH. Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to the corresponding EV group. (L) NMR data for [U-13C] glucose incorporated into AMP and UMP in Hep3B cells overexpressing cMyc or PSPH (left), or in Hep3B cells expressing shRNAs against cMyc or PSPH (right). Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared to the corresponding EV or NTC group. (M, N) Growth curves were determined by trypan blue counting in Hep3B cells expressing NTC or PSPH shRNAs (M), or in those cells further treated with or without 5 mM NAC, 5 mM GSH or 2 mM serine (N). Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups.

Full size image

Since serine and glycine are major mediators for nucleotide synthesis18,45, we examined the effects of cMyc and PSPH on the synthesis of AMP and UMP, two typical nucleotides, by NMR analysis using 13C-labeled glucose. As shown in Figure 3L, 13C incorporation into both AMP and UMP was significantly increased in cMyc- or PSPH-overexpressing cells, and knocking down cMyc or PSPH markedly reduced 13C incorporation into both AMP and UMP, suggesting that both cMyc and PSPH facilitate nucleotide synthesis, which could possibly account for the decreased S phase populations in PSPH-deficient cells (Figure 3J). Importantly, knockdown of PSPH markedly inhibited cell proliferation in Hep3B and SK-hep-1 cells (Figure 3M and Supplementary information, Figure S2E). Similar results were observed when PHGDH or PSAT1 were knocked down (Supplementary information, Figure S2I). Moreover, addition of ROS scavengers, NAC (5 mM) or GSH (5 mM), to the culture medium partially rescued cell proliferation inhibited by knocking down PSPH in Hep3B cells (Figure 3N), suggesting that PSPH affects cell proliferation in part via ROS. Collectively, our results demonstrate that cMyc-mediated PSPH expression and SSP activation are critical for cancer cell proliferation by regulating redox homeostasis between GSH and ROS levels and by regulating nucleotide synthesis as well as cell cycle progression.

cMyc and PSPH are critical for cancer cell growth under nutrient deprivation conditions

Next, we investigated further the effects of cMyc and PSPH on cell survival and proliferation under nutrient deprivation conditions. As we have observed that PSPH expression and SSP activation were critical for cancer cell proliferation by regulating redox homeostasis and that cMyc stimulated PSPH expression as well as SSP activation under nutrient starvation conditions, we surmise that redox imbalance may be a cause for cancer cell death under starvation conditions. Indeed, our results demonstrated that knockdown of cMyc or PSPH accelerated cell death of Hep3B cells under glucose- or glutamine-free conditions (Figure 4A). However, addition of GSH (5 mM) or Nuc (4 ribonucleosides plus 4 deoxynucleosides)47 to the glucose- or glutamine-free culture medium partially recovered the cell survival under nutrient starvation conditions (Figure 4B and 4C). Importantly, the rescue effect is more obvious in the presence of both GSH and Nuc (Figure 4B and 4C), suggesting that cMyc and SSP mediated redox balance and nucleotides synthesis play important roles in cell survival under starvation conditions. Consistently, we observed that glucose or glutamine starvation resulted in reduced cellular GSH level, decreased GSH/GSSG ratio, increased cell apoptosis and increased cellular ROS level (Figure 4D and 4E), which were further augmented by suppression of PSPH or other SSP genes under glucose- or glutamine-deprived conditions (Figure 4D, 4E and Supplementary information, Figure S3A-S3C). Similarly, knockdown of cMyc also augmented the effect of glucose or glutamine deprivation on cellular GSH level, GSH/GSSG ratio and apoptosis (Figure 4D and 4E). Thus, these results suggested that redox imbalance resulting from SSP deregulation accounts for, at least partially, the accelerated cell death phenotype induced by glucose or glutamine starvation, underlining the significance of SSP for cancer cell survival under nutrient-deprived conditions.

Figure 4
figure 4

cMyc and PSPH are critical for cancer cell growth under nutrient deprivation conditions. (A) Growth curves were determined by trypan blue counting in Hep3B cells expressing shRNAs against cMyc or PSPH starved of glucose (top) or glutamine (bottom). Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups. (B, C) Growth analysis (B) and crystal violet assay (C) were performed in Hep3B cells starved of glucose or glutamine without or with supplement of GSH, Nuc (4 ribonucleosides and 4 deoxyribonucleosides) or both. Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups. (D, E) GSH level, GSH/GSSG ratio, apoptosis and cellular ROS level were determined in Hep3B cells expressing shRNAs targeting PSPH or cMyc starved of glucose (D) or glutamine (E). Data were presented as mean ± SD of three independent experiments. *P < 0.05 as compared with NTC+Glc or NTC+Gln group; #P < 0.05 as compared with NTC–Glc or NTC–Gln group. (F, G) Hep3B cells expressing shRNAs targeting PSPH (F) or cMyc (G) were cultured in medium without glucose (left) or glutamine (right) for 4 days, followed by culturing in complete medium for the indicated days. Cell growth was determined by trypan blue counting. Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups.

Full size image

To further address the effects of the activated cMyc and PSPH on cancer cell survival under nutrient deprivation stress, we starved cancer cells of glucose or glutamine for 4 days followed by reconstitution of glucose or glutamine in the culture medium. We observed that reconstitution of glucose or glutamine in the culture medium can rescue the starved control cells but not the cells expressing shPSPH or shcMyc (Figure 4F and 4G), demonstrating that loss of cMyc or PSPH caused permanent damage to the cells that could not recover even when the nutrient conditions were improved. Taken together, our results demonstrate that cMyc-mediated SSP activation and PSPH expression contribute significantly to the survival of cancer cells deprived of glucose or glutamine.

PSPH is vital for the tumorigenesis capacity of cMyc in vitro and in vivo

Next we asked whether PSPH is vital for cancer progression in vitro and in vivo. Forced expression of PSPH in Hep3B cells increased cell proliferation in vitro (Figure 3F), indicating that high PSPH expression alone is sufficient to confer a growth advantage to cancer cells. Xenograft experiment in nude mice showed significantly accelerated tumor growth in Hep3B cells stably overexpressing PSPH compared to EV control group (Figure 5A and Supplementary information, Figure S4A and S4B). Western blot using lysates from xenograft tumor tissues confirmed forced expression of PSPH in tumors (Figure 5B). Thus, in vivo animal study provided further evidence for the importance of PSPH in cancer progression.

Figure 5
figure 5

PSPH is vital for the tumorigenesis capacity of cMyc in vitro and in vivo. (A, B) Hep3B cells stably expressing EV or PSPH were injected subcutaneously into nude mice (n = 6 for each group). Tumor sizes were measured starting from 8 days after inoculation (A). Tumors were extracted at the end of experiment, and protein expression of PSPH in tumors was analyzed by western blot using anti-PSPH antibody (B). β-actin serves as loading control. Data were presented as mean ± SEM. *P < 0.05 compared between the indicated groups. (C, D) cMyc-overexpressing Hep3B cells were infected with viruses expressing NTC or PSPH shRNAs. Western blot and trypan blue counting were performed to analyze protein expression (C) and cell growth (D), respectively. Data were presented as mean ± SD of three independent experiments. *P < 0.05 compared between the indicated groups. (E-G) cMyc-overexpressing Hep3B cells with or without PSPH knockdown were injected subcutaneously into nude mice (n = 6 for each group). Tumor growth curves were determined (E). At the end of experiment, incidence of tumor formation was calculated (F) and tumors were extracted followed by western blot analyzing cMyc and PSPH protein expressions in tumors (G). β-actin serves as loading control. Data were presented as mean ± SEM. *P < 0.05 compared between the indicated groups.

Full size image

To further determine whether activated SSP is critical for the oncogenic role of cMyc, PSPH was stably knocked down by shRNAs in cMyc-overexpressing Hep3B cells (Figure 5C). Cell growth analysis revealed that PSPH knockdown significantly abolished the enhanced cell proliferation by cMyc overexpression in vitro (Figure 5D). It is interesting to note that, consistent with early observations in PHGDH-knockdown cells48, addition of exogenous serine into culture medium could not rescue the growth deficiency of PSPH-knockdown cells (Figure 3N), suggesting that activation of SSP rather than addition of serine itself is critical for cancer cell proliferation. Furthermore, xenograft study showed delayed tumor initiation, decreased tumor size and tumor mass when PSPH was knocked down in cMyc-overexpressing cells (Figure 5E, 5F and Supplementary information, Figure S4C). Intriguingly, there was no tumor formation in all 6 mice injected with cells stably expressing shPSPH and the tumorigenesis capacity of cMyc-overexpressing cells was dramatically damaged when PSPH was knocked down, with tumor formation incidence declining to 50% (Figure 5F). Western blot confirmed cMyc overexpression and PSPH knockdown in xenograft tumor lysates (Figure 5G). Taken together, these in vivo data indicate that activated SSP is critical for cMyc-mediated cell proliferation and cancer progression both in vitro and in vivo.

Aberrant PSPH expression in human hepatocellular cancer predicts patient mortality

Analysis of cell proliferation in vitro and tumor growth in vivo indicated that PSPH is critical for the oncogenic roles of cMyc. To investigate the physiological significance of aberrant PSPH expression or SSP activation in human malignancies, PSPH expression was further studied in 20 paired hepatocellular carcinoma (HCC) lesions and adjacent non-cancerous tissue samples and the results showed that both PSPH mRNA and protein levels were significantly increased in HCC lesions as compared to adjacent normal tissue (Figure 6A and 6B), whereas the expression of PHGDH and PSAT1 was decreased in HCC lesions as compared to adjacent normal tissue (Supplementary information, Figure S4D), suggesting that it is important to combine clinical data and experimental evidence systematically to decipher mechanisms underlying SSP activation during cancer progression.

Figure 6
figure 6

Prognostic significance of aberrant PSPH expression in human hepatocellular cancer. (A) qRT-PCR analyzed PSPH mRNA expression in 20 pairs of clinically matched tumor adjacent non-cancerous liver tissues (normal) and human HCC tissues (tumor). mRNA levels were normalized to 18S rRNA. *P < 0.05 as compared to normal tissue group. (B) Western blot analyzed PSPH protein expression in the paired tumor adjacent non-cancerous liver tissues (N) and human HCC tissues (T). GAPDH serves as loading control. (C) Representative IHC analysis of PSPH expression in normal liver tissues (normal) and HCC specimens of different clinical stages (I-IV) was shown. (D) Statistical quantification of the mean optical density (MOD) values of PSPH staining in IHC assay between normal liver tissues and HCC specimens of different clinical stages (I-IV). The MOD of PSPH staining increases as HCC progresses to a higher clinical stage. Data were presented as mean ± SD. *P < 0.05 vs normal control group. (E) Kaplan-Meier curves with univariate analyses for patients with low vs high PSPH expression. (F) Summary: cMyc activates serine biosynthesis pathway directly by regulating SSP enzymes at transcriptional level or indirectly by coordinating glycolysis (PGK1, PGAM1) and glutaminolysis (GOT1, MDH1, ME1) under nutrient stressed conditions, leading to cancer cell survival and proliferation via regulation of ROS, GSH, apoptosis and cell cycle.

Full size image

Next, immunohistochemistry (IHC) was employed to analyze PSPH expression in a retrospective cohort of 158 clinicopathologically characterized HCC cases, including 10 cases of clinical stage I (6.3%), 106 cases of clinical stage II (67.1%), 33 cases of clinical stage III (20.9%) and 9 cases of clinical stage IV (5.7%) liver cancer samples, based on the TNM staging (Supplementary information, Table S1). The IHC results revealed that PSPH protein expression was generally negative in normal human liver tissues, weak in early stage HCC (TNM stages I and II) and abundant in late stage HCC (TNM stages III and IV) tissues (Figure 6C). Quantitative analysis of the IHC results revealed that PSPH expression in clinical stage I-IV primary liver tumors was significantly increased compared to normal liver (Figure 6D and Supplementary information, Table S2). Furthermore, PSPH was markedly upregulated in late stage HCC (stages III and IV) as compared to early stage HCC (stages I and II; Figure 6D and Supplementary information, Table S2), suggesting that PSPH expression was gradually increased as HCC progressed to a higher clinical stage. Moreover, Spearman analysis revealed correlations between PSPH expression and patient clinicopathological characteristics, including survival time (P < 0.001), vital status (P < 0.001) and clinical stage (P < 0.001; Supplementary information, Table S3), further suggesting a strong association of PSPH expression with HCC clinical staging and patient survival. Univariate and multivariate analysis revealed that, together with TNM stage and tumor size, PSPH level is a significant prognostic factor and is predictive of the overall survival of HCC patients (Supplementary information, Table S4). Finally, Kaplan-Meier test indicated that PSPH expression in HCC patients was significantly associated with survival time, with patients expressing low PSPH in their HCC lesions surviving much longer than those with high PSPH expression (Figure 6E), suggesting that PSPH protein may represent a promising prognostic biomarker for HCC.

Discussion

An important progress recently in the field of cancer metabolism is the appreciation of serine/glycine metabolism in cancer cells. Accumulating studies have established serine/glycine metabolism as one of the central hubs for cancer metabolism, however, it is far from clear how serine/glycine metabolism is regulated under normal or stressed conditions. Hence, in this study, we explored how activation of SSP is regulated under nutrient stressed conditions, especially when glucose or glutamine is deprived. We demonstrated that, under glucose/glutamine-deprived conditions, cMyc is activated to enhance the expression of enzymes involved in SSP activation to promote cellular survival and proliferation.

The current study focused on the effects of two major nutrient substrates, glucose and glutamine, on serine biosynthesis because cancer cells are highly dependent on these two substrates for survival and proliferation1,13,14,15. It was described almost 90 years ago by Otto Warburg that cancer cells avidly use the less efficient glycolysis pathway for energy sources even with ambient oxygen supply4. More recently, cancer cells have been reported to be highly addicted to glutamine metabolism2,41. Obviously, glycolysis and glutaminolysis in cancer cells are being extensively sought largely because of the consensus that elucidation of cancer-specific metabolisms may hold the key to understand and control cancer progression. Glucose and glutamine are also the major sources for serine/glycine biosynthesis. Nevertheless, it is very little appreciated to date how glycolysis or glutaminolysis pathways intertwine with SSP activation to coordinate cellular survival and proliferation under stressed conditions. Possemato et al.48 and Locasale et al.49 reported that elevated SSP enzyme expression was constantly detected in human breast cancer and that SSP activation was observed to sustain cellular survival during serine starvation. Interestingly, PHGDH suppression inhibits cell proliferation even in cells cultured in media containing exogenous serine, suggesting that de novo synthesis rather than exogenous serine is important for cell proliferation48. We demonstrated here that SSP was activated under glucose/glutamine-deprived conditions (Figure 1A and 1B). We further observed that several pathways connecting glycolysis or glutaminolysis to SSP were enhanced under such conditions (Figures 1C, 1D and 6F), underlining the central importance of this pathway for cellular survival under stressed conditions. Moreover, we also provide evidence which is consistent with the previous study that GSH, a metabolite of SSP, is critical for cell survival, proliferation and apoptosis control through redox balance (Figures 3, 4B and 4C).

cMyc is known to be extensively involved in the regulation of glycolysis and glutaminolysis in cancer cells40. Previous study by Ma et al.30 demonstrated that glucose deprivation allows not only the nitrogen, but also the carbon of glutamine to be incorporated into serine. Here our results also indicated that cMyc is critically involved in the regulation of SSP (Figure 2F and Supplementary information, Figure S1), hence coordinating metabolism of glucose, glutamine and serine/glycine for cell survival and proliferation. It is striking to note that cMyc comprehensively regulates all the major pathways leading to serine biosynthesis (Supplementary information, Figure S5A). Intriguingly, we observed enhanced expression of multiple enzymes in cancer cells where cMyc expression is deregulated under stressed conditions (Figure 1C and 1D). More significantly, we provided evidence that expression of PSPH, the final rate-limiting enzyme for serine biosynthesis, is critical and sufficient for cMyc-mediated cancer progression (Figure 5E-5G), highlighting again that the activation of SSP is critical for cancer progression in general.

As most of cancer cells are sensitive to glucose/glutamine deprivation and die eventually under conditions without these substrates, one would wonder what the significance of the proposed adaptation in metabolism is for cancer cell survival and progression. In fact, as a cancer develops, its microenvironment might vary dynamically over time. As a consequence, cancer's comprehensive metabolic adaptation might represent a strategy to allow cancer cells to survive various stresses before conditions improve. Indeed, we have observed that reconstitution of glucose and glutamine in cell culture would revive the growth of cancer cells expressing cMyc and PSPH, however, depletion of cMyc or PSPH caused permanent damage to the cells and they would not revive even when the conditions were improved (Figure 4F and 4G). Consistently, we observed clinical correlation of PSPH aberrant expression with HCC development and patient survival, suggesting that aberrant expression of PSPH predicts patient mortality (Figure 6C-6E). This correlation is highly relevant for understanding and treatment of human cancers, pointing to a potential rationale to target PSPH or SSP pathway in general for cancer therapy in human malignancies with aberrant activation of SSP and cMyc deregulation.

Materials and Methods

Cell culture and reagents

Unless otherwise stated, chemicals are obtained from Sigma-Aldrich, and cell culture reagents are from Gibco (Invitrogen). Cells were cultured in DMEM (Hep3B, SK-hep-1, Hela, HT1080) or RPMI-1640 (P493) supplemented with 10% FBS and 1% Pen/Strep Cells were kept at 37 °C in humidified 5% CO2 in air. For glucose, glutamine or serine/glycine starvation experiments, cells were fed with DMEM deprived of corresponding nutrients. Trypan blue exclusion assay and colony formation assay were used to assess cell growth and survival.

Western blot

Cell lysates were prepared in RIPA buffer (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.1% SDS, 1% NP-40) supplemented with protease cocktails. Equal amounts of proteins were fractionated by SDS-PAGE. Primary antibodies against the following proteins were used: PHGDH, PSAT1, PSPH, SHMT2, GLS1, GLUD1, GOT1, GOT2, MDH1, ME1, ME2, HK2, GPI, PFKL, ALDB, TPI1, GAPDH, PGK1, PGAM1, ENO1 and LDHA (Proteintech); cMyc (Zymed); PKM2 (Cell Signaling Technology); SHMT1 (Abgent), Flag (Sigma-Aldrich). β-actin (Abmart) served as loading control. HRP-conjugated anti-rabbit and anti-mouse (Bio-Rad) secondary antibodies were used. Signal was detected using Western ECL Substrate (Bio-Rad).

qRT-PCR

Total RNA was isolated using Trizol followed by DNase (Ambion) treatment and reverse transcription with iScript cDNA Synthesis Kit (Bio-rad). qRT-PCR was performed using SYBR Green master mix (Vazyme) on a Bio-Rad iCycler. Primer sequences used were shown in Supplementary information, Table S5. All samples were normalized to 18S rRNA. The fold change of target mRNA expression was calculated based on threshold cycle (Ct), where ΔCt = Cttarget − Ct18S and Δ(ΔCt) = ΔCtControl − ΔCtIndicated condition.

Plasmids and stable cells

All shRNAs in PLKO vector against cMyc, PHGDH, PSAT1, PSPH were commercially purchased (Sigma-Aldrich). cMyc was subcloned into pBabeMNires GFP empty vector50 or PMX-3×Flag empty vector. PSPH was subcloned into pSin-3×Flag vector. Hep3B cells were infected with retrovirus (cMyc) or lentivirus (PSPH and all shRNAs) followed by antibiotic selection to establish stable cells. Detailed shRNA information is shown in Supplementary information, Table S6.

Chromatin immunoprecipitation (ChIP) assay

The ChIP assay was performed with an EZ-ChIP kit (Millipore) following the manufacturer's instruction. Briefly, cells were fixed with 1% formaldehyde and quenched in 0.125 M glycine. Cells were sonicated by Bioruptor Sonication System UCD-300. DNA was immunoprecipitated by either control IgG or cMyc primary antibody (Cell Signaling Technology). RNA and protein were digested using RNase A and Proteinase K, respectively, followed by qRT-PCR analysis using SYBR Green master mix (Vazyme). Oligos used for this analysis are tabulated in Supplementary information, Table S7.

ROS detection

Cells were stained with 5-(and -6)-carboxy-2′, 7′-dichlorodihydrofluorescein diacetate (carboxy-H2DCFDA; Molecular Probes) and analyzed by flow cytometers (BD Bioscience) following the manufacturer's instruction. Gain and amplifier settings were held constant during the experiment50.

GSH and GSH/GSSG ratio determination

The intracellular GSH levels, and reduced and oxidized glutathione (GSH/GSSG) ratio were measured using assay kit (V6611 and V6911) purchased from Promega following the manufacturer's instruction.

Cell cycle analysis

Cells were harvested by trypsinization, washed twice with PBS containing 5% FBS, fixed in 70% ethanol followed by staining with 20 μg/ml propidium iodide containing 20 μg/ml RNase (DNase free). Stained cells were analyzed by flow cytometry.

Apoptosis assays

Apoptosis was measured by flow cytometry using the Annexin V-PI Apoptosis Detection Kit (Bestbio) according to the manufacturer's instructions.

NMR analysis of 13C-labeled metabolites

13C-labeled metabolites were measured quantitatively with one-dimensional heteronuclear multiquantum coherence NMR method and proton detection using a gradient selected pulse sequence. All these 1H NMR spectra were acquired at 298 K on a Bruker Avance III 600 MHz NMR spectrometer (operating at 600.13 MHz for 1H and 150.90 MHz for 13C) equipped with an inverse cryogenic probe. The 90° pulse length was adjusted to about 12 and 13.5 μs for 1H and 13C, respectively, and 13C decoupling was achieved with an adiabatic scheme (p5m4sp180.2) during acquisition period. Proton spectral width was 8 400 Hz (14 ppm) whilst 13C decoupling carrier frequency was set on about 70 ppm and JCH set to 140 Hz. 2 048 data points were collected to ensure the 13C decoupling period short enough (i.e., acquisition time of 0.122 s) to avoid stressing probe. Sixty four scans were collected with the recycle delay of 20 s (to ensure complete proton relaxation) with the water signal pre-saturated 2 s prior to the first 1H 90° pulse. FIDs were zero-filled to 16 k data points and applied with an exponential window function, line-broadening factor of 5 Hz and forward linear prediction prior to Fourier transformation. So-obtained spectra only contained signals of protons directly attached to 13C nuclei.

Concentration of 13C-containing metabolites were quantified by tabulating the proton signal integrals against that of sodium 3-trimethylsilyl [2,2,3,3-d4] propionate (TSP) with a known concentration, which had natural 13C abundance and was also employed as proton chemical shift reference (δ0.00). During such calculation, polarization transfer efficiency for each concerned moieties was taken into account with a factor of sin2 (πJCHÄ), where Ä was the polarization transfer delay and JCH was one-band coupling constant. Such strategy was validated experimentally with a model mixture of 13C-labeled metabolites with known concentration.

Animal studies

All animal studies were conducted with approval from the Animal Research Ethics Committee of the University of Science and Technology of China. For xenograft experiments, 5 × 106 Hep3B cells stably expressing EV, PSPH, or cMyc-overexpressing Hep3B cells with or without PSPH knockdown were injected subcutaneously into 5-week-old male nude mice (SJA Laboratory Animal Company, China). The tumor volumes were measured using digital calipers every 3 days and calculated using the equation: length (mm) × width (mm) × depth (mm) × 0.52.

Clinical human HCC specimens

Human tissue specimens, including 20 pairs of freshly snap-frozen HCC tissues and paired normal adjacent tissues, were histopathologically diagnosed at Sun Yat-sen University Cancer Center (Guangzhou, China). Formalin-fixed, paraffin-embedded primary HCC specimens obtained from 158 patients were randomly selected from the archives of the Sun Yat-sen University Cancer Center. For using these clinical materials for research purposes, prior patients' written informed consents and approval from the Institutional Research Ethics Committee of Sun Yat-sen University Cancer Center were obtained. Tumor clinical stages were defined according to the 2002 American Joint Committee on Cancer/International Union against Cancer tumor/lymph node metastasis/distal metastasis (TNM) classification system51. Total RNA and protein were extracted from paired tissues and detected by qRT-PCR and western blot, respectively.

Immunohistochemistry

The IHC procedure for PSPH and scoring of PSPH expression were performed as previously reported52. IHC staining was quantitatively analyzed with the AxioVision Rel.4.6 computerized image analysis system assisted with the automatic measurement program (Carl Zeiss, Oberkochen, Germany). Ten representative staining fields of each section were analyzed to verify the Mean Optical Density (MOD). The MOD data were statistically analyzed by t-test to compare the average MOD difference between different groups of tissues.

Statistical analysis

The data were presented as mean ± SD or mean ± SEM as stated. Student's t-test was used to calculate P values. Statistical significance is displayed as P < 0.05.