A genome-wide CRISPR-Cas9 screen identifies requirement of mitochondrial translation-related genes for liver cancer cells
To identify new vulnerabilities of liver cancer cells, we performed a genome-wide CRISPR-Cas9 genetic screen in SNU398 and HepG2 cells (Fig. 1a-c, Supplementary Fig. 1) and identified 455 genes commonly required in the two cells (Fig. 1d). According to the Gene Ontology (GO) gene set enrichment analyses of cellular component (CC) and biological process (BP), mitochondrial ribosome cellular component and mitochondrial translation biological process were highly enriched (Supplementary Fig. 2a, b). MitoCarta3.0 is an updated inventory of mammalian mitochondrial proteins using multiple experimental and computational approaches15. By comparing the common hits with the MitoCarta3.0 inventory, we found that 26 of these genes were implicated in mitochondrial translation (Fig. 1d, Supplementary Table 1). In view of this, we intend to study the role of mitochondrial translation in liver cancer cells. The differential expression of mitochondrial translation-related genes in The Cancer Genome Atlas (TCGA) liver hepatocellular carcinoma (LIHC) cohort was investigated. 24 mitochondrial translation-related genes were significantly upregulated in 50 tumour (T) tissues compared with corresponding non-tumour (N) tissues (Fig. 1e). Using GSVA to analyse the enrichment scores of the 26 mitochondrial translation-related genes between tumour and non-tumour tissues of HCC in TCGA, we found that mitochondrial translation-related genes were highly enriched in tumour tissues, indicating that mitochondrial translation is an important pathway for liver cancer cells (Supplementary Fig. 2c).
To validate the results of the CRISPR screen, we selected three representative mitochondrial translation-related genes, GFM1 (a mitochondrial translation factor), MRPL4 (a protein of large mitochondrial ribosomal subunit), and MRPS23 (a protein of small mitochondrial ribosomal subunit), by interfering with their expression through shRNA-mediated knockdown. Knockdown of GFM1, MRPL4, or MRPS23 in SNU398 and HepG2 cells resulted in not only short-term and long-term cell proliferation inhibition (Fig. 1f-h, Supplementary Fig. 3a, b), but also the downregulation of translation products, such as CYTB (complex III), MTCO2 (complex IV), and ATP6 (complex V), which are components of the respiratory chain complex (Supplementary Fig. 3c, d), indicating that knockdown of mitochondrial translation-related genes leads to impaired mitochondrial translation.
Prognostic significance of mitochondrial translation-related genes
To investigate the prognostic value of mitochondrial translation-related genes, we collected tumour material and clinical data from 243 HCC patients from the Naval Military Medical University Affiliated Eastern Hepatobiliary Hospital. Univariate and multivariate analyses for protein levels (as determined by immunohistochemistry on tumour material) and clinicopathological features showed that GFM1, MRPL4, and MRPS23 were independent prognostic markers of overall survival (OS) and time to relapse (TTR) (Supplementary Table 2–4). Kaplan-Meier survival analyses showed that HCC patients with high expression of GFM1, MRPL4, or MRPS23, had worse OS and shorter TTR (Fig. 2a-f). Using consensus clustering analysis for the 26 genes, the HCC cohort from TCGA could be divided into three subclusters. Patients in the cluster 2 represented generally low levels of 26 mitochondrial translation-related genes and low GSVA score of mitochondrial translation, but high survival rate (P = 0.0060) (Fig. 2g-i). These data support the notion that high expression of mitochondrial translation-related genes is associated with poor outcome in HCC.
Tigecycline inhibits mitochondrial translation in liver cancer cells
Tigecycline, the only FDA-approved antibiotic of the glycylcyclines, has been shown to specifically inhibit mitochondrial translation11,16. According to cell viability assays, we found tigecycline had significant activity against the majority of liver cancer cells tested (Fig. 3a). Most liver cancer cells were sensitive to tigecycline, especially SNU398, Huh6, and HepG2 cells. In contrast, PLC/PRF/5, Li7, and MHCC97H cells were considered to be relatively insensitive because their half maximal inhibitory concentration (IC50) exceeded 30 µM (Fig. 3a). Moreover, long-term cell proliferation assays using tigecycline also validated that SNU398, Huh6, and HepG2 cells were sensitive to tigecycline, while PLC/PRF/5, Li7, and MHCC97H cells were not (Supplementary Fig. 4).
Tigecycline could inhibit the process of mitochondrial translation of both sensitive and insensitive cells, as evidenced by the notion that this drug can downregulate protein levels of CYTB, MTCO2, and ATP6 in both sensitive and insensitive cells (Fig. 3b). In vivo experiments showed that tigecycline could inhibit the growth of liver cancer cells without affecting the body weight of mice (Fig. 3c-e). IHC staining analyses confirmed that protein levels of mitochondrially-translated MTCO2, CYTB, and ATP6 were decreased in tumour tissues removed from mice treated with tigecycline (Fig. 3f). Therefore, we conclude that tigecycline inhibits the mitochondrial translation of all liver cancer cells but has major anti-proliferative effects in a subset of liver cancer cells.
Compound screen identifies MEK inhibitors as synergistic with tigecycline
To find drugs that have synergistic effects of tigecycline in liver cancer cells, we performed a compound screen in tigecycline-insensitive cells (Fig. 4a). The top 20 compounds with synergistic effects in MHCC97H cells were then selected for the second-round screen in three insensitive cell lines: MHCC97H, PLC/PRF/5, and Li7 (Fig. 4b). The second-round screen showed mitogen-activated protein kinase kinase (MEK) inhibitors and receptor tyrosine kinase (RTK) inhibitors such as trametinib, cobimetinib, selumetinib, ceritinib, crizotinib, and cabozantinib had favorable synergies of tigecycline (Fig. 4c). Incucyte cell proliferation, short-term and long-term cell proliferation assays showed that the MEK inhibitors trametinib and cobimetinib increased the sensitivity of MHCC97H, PLC/PRF/5, and Li7 cells to tigecycline (Fig. 4d, e, Supplementary Fig. 5a-d). However, ceritinib, crizotinib, and cabozantinib had no synergistic effects in long-term cell growth assays (data not shown). Thus, we identify MEK inhibitors as drugs that display synergy with tigecycline.
Combined regimen of tigecycline and trametinib inhibits both oxidative phosphorylation and glycolysis
Given that tigecycline inhibited mitochondrial translation, thereby potentially affecting oxidative phosphorylation (OXPHOS), we measured the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of sensitive and insensitive cells treated with tigecycline. These results showed that tigecycline could reduce the basal OCR, maximal respiration and ATP production of OXPHOS of both tigecycline-sensitive and -insensitive cells (Supplementary Fig. 6a, b). However, the effects on glycolysis were quite different. Compensatory enhancement of glycolysis and glycolytic capacity was only observed in tigecycline-insensitive cells, which might explain the differences in sensitivity of liver cancer cells to tigecycline (Supplementary Fig. 6c, d). Compared to tigecycline treatment alone, the combination of trametinib and tigecycline could further reduce the basal OCR, maximal respiration, and ATP production of OXPHOS in tigecycline-insensitive cells. Additionally, compensatory enhancement of glycolysis and glycolytic capacity was inhibited (Fig. 5a, b, Supplementary Fig. 7a-e). Besides, tigecycline caused changes in the activities of key enzymes in glycolysis pathway. After treatment with tigecycline, the activities of hexokinase, pyruvate kinase, and the content of extracellular lactate increased in MHCC97H and PLC/PRF/5 cells, but decreased after addition of trametinib (Fig. 5c-e).
Tigecycline activates the EGFR-ERK1/2-MYC cascade
To explore the molecular basis of liver cancer cells insensitivity to tigecycline and how trametinib might enhance this sensitivity, we conducted RNA sequencing of MHCC97H and PLC/PRF/5 cells following treatment with tigecycline. Compared tigecycline to control groups, we found 357 upregulated and 261 downregulated genes in MHCC97H cells, and 603 upregulated and 574 downregulated genes in PLC/PRF/5 cells (Supplementary Fig. 8a, Supplementary Table 5–6). KEGG pathways analyses showed shared upregulated pathways and no shared downregulated pathway of MHCC97H and PLC/PRF/5 cells (Fig. 6a, Supplementary Fig. 8b). Sankey diagrams of upregulated genes showed that tigecycline treatment significantly enriched the MAPK signaling pathway and one carbon pool by folate pathway compared to control treated cells (Fig. 6a). qRT-PCR was then performed to investigate the expression of genes involved in one carbon pool by folate pathway, including SHMT2, ALDH1L2, MTHFD1L and MTHFD2. Results showed that mRNA level of SHMT2 was markedly elevated with treatment of tigecycline in MHCC97H and PLC/PRF/5 cells (Fig. 6b, Supplementary Fig. 8c). These results suggest that the MAPK cascade is enhanced in tigecycline-insensitive cells with treatment of tigecycline and SHMT2 may be involved in the low response of liver cancer cells to tigecycline.
Next, we performed an assay to measure activation (tyrosine phosphorylation) of human receptor tyrosine kinases. Results showed that phosphorylation of epidermal growth factor receptor (EGFR) and hepatocyte growth factor receptor (HGFR) are further activated following treatment with tigecycline (Fig. 6c). Combined data analysis for RNA sequencing, mRNA levels of amphiregulin (AREG) and epiregulin (EREG), ligands of EGFR in ERBB signaling pathway, were upregulated in tigecycline-treated cells. Therefore, we focused on EGFR signaling in subsequent studies (Fig. 6a). qRT-PCR experiments confirmed that mRNA levels of AREG and EREG were increased following treatment with tigecycline (Supplementary Fig. 9a). Since AREG and EREG are growth factors that can interact with EGFR in an autocrine fashion, ELISA was then performed to measure the levels of AREG and EREG in the supernatants. Results showed that AREG and EREG were significantly increased in the cell culture supernatants with treatment of tigecyline (Fig. 6d, e, Supplementary Fig. 9b, c). Moreover, Western blot experiments indicated that EGFR pathway was activated, which subsequently the levels of p-MEK1/2 and p-ERK1/2 were increased in tigecycline-treated cells. Inhibition of EGFR by gefitinib could also reduce the levels of p-MEK1/2 and p-ERK1/2 and had a synergistic effect of inhibition on liver cancer cells with tigecycline (Fig. 6f, g, Supplementary Fig. 9d, e). It has been described that MAPK cascade can transcriptionally activate MYC expression, thereby increasing the transcription of MYC downstream genes17,18. Considering the increase of glycolysis after treatment with tigecycline, protein levels of hexokinase 2 (HK2), pyruvate kinase M1/2 (PKM2), and lactate dehydrogenase A (LDHA) were detected by Western blot. Results showed the increase in their protein levels after treatment with tigecycline and a decrease after addition of trametinib (Fig. 6h, Supplementary Fig. 9f). Our data suggest that a feedback activation loop involving in the EGFR-ERK1/2-MYC cascade may mediate insensitivity of liver cancer cells to tigecycline.
To explore whether responses of liver cancer cells to tigecycline could be identified by the level of activation of the MAPK cascade, we performed Western blot analyses and found sustained activation of ERK1/2 signaling by tigecycline in tigecycline-insensitive cells, while this signaling was only transiently activated in tigecycline-sensitive cells, indicating that sustained feedback activation of ERK1/2 is crucial for liver cancer cells insensitivity to tigecycline (Supplementary Fig. 10a, b).
To understand how MYC, an important downstream target of the MAPK cascade, influenced OXPHOS and glycolysis after treatment with tigecycline, we conducted shRNA-mediated knockdown of MYC in MHCC97H and PLC/PRF/5 cells. After addition of tigecycline, knockdown of MYC not only reduced the expression of SHMT2 and glycolytic enzymes-encoding genes such as HK2, PKM2, and LDHA (Fig. 6i, Supplementary Fig. 11a), but also decreased the levels of OCR and ECAR in MHCC97H and PLC/PRF/5 cells (Fig. 6j, k, Supplementary Fig. 11b-d).
To confirm feedback activation of glycolysis when treated with tigecycline, we knocked down HK2 in MHCC97H and PLC/PRF/5 cells with shRNA (Supplementary Fig. 12a). Results exhibited that the levels of OCR were enhanced and ECAR were declined after knockdown of HK2, while the levels of OCR and ECAR were decreased in the presence of tigecycline (Supplementary Fig. 12b, c). Our findings indicate that inhibition of feedback activation of glycolysis might result in energy depletion when treated with tigecycline, which was further supported by the synergistic effect of glycolysis inhibitor 2-Deoxy-D-glucose (2-DG) and tigecycline in liver cancer cells (Supplementary Fig. 12d).
To explore the role of SHMT2 in the treatment of tigecycline, we stably knocked down SHMT2 in liver cancer cells. In the presence of tigecycline, the expression of SHMT2 increased to partially maintain the suppressed mitochondrial function, while knockdown of SHMT2 in MHCC97H and PLC/PRF/5 cells led to downregulation of mitochondrial translation products, including CYTB, MTCO2, and ATP6 (Supplementary Fig. 13a). The usage of tigecycline in SHMT2-knockdown cells further reduced the OCR levels compared with only treated with tigecycline (Supplementary Fig. 13b). Thus, knockdown of mitochondrial folate enzyme SHMT2 in treatment of tigecycline resulted in further impaired mitochondrial translation and defective OXPHOS.
Blocking EGFR-ERK1/2-MYC cascade sensitizes to tigecycline in vivo
To validate the tigecycline drug combination in vivo, we used the tigecycline-insensitive MHCC97H subcutaneous tumour model. Compared to the control group, treatment of tigecycline or trametinib alone represented no obvious tumour suppressive effect, while the combination of tigecycline and trametinib showed a more effective inhibition of tumour growth without affecting the weight of mice (Fig. 7a, b, Supplementary Fig. 14a, b). Protein levels of MYC, glycolytic enzymes such as HK2 and LDHA, SHMT2, and p-ERK1/2 in tumour tissues from tigecycline group were increased compared to control. The combination of tigecycline and trametinib reduced the protein levels of these genes induced by feedback activation (Fig. 7c, Supplementary Fig. 14c, d). On the other hand, the combination treatment further inhibited the expression of mitochondrial translation products such as MTCO2, ATP6, and CYTB compared to tigecycline monotherapy, indicating the involvement of impaired mitochondrial translation and defective OXPHOS in the xenografts treated with the drug combination (Supplementary Fig. 14c, d). Our data suggest that combined regimen of tigecycline and trametinib can significantly suppress tumour growth in vivo by reducing glycolysis and OXPHOS.
It has been described that tigecycline had an immunomodulatory activity under the stimulus of staphylococcal superantigen19. We investigated the combination therapies in immune-competent mice with mouse liver cancer cells (Hepa1-6) transplanted subcutaneously. Trametinib or gefitinib had markedly synergistic effects with tigecycline in immune-competent mice, without affecting the weight of mice (Fig. 7d, e, Supplementary Fig. 15a). IHC analyses of markers of mouse immune cells showed increased CD4+ T cells, CD8+ T cells, and NKG2D+ cells (a marker of NK cells) in the combination groups (tigecycline combined with trametinib or tigecycline combined with gefitinib) compared to control or tigecycline group. Whereas, the number of F4/80+ cells (a marker of macrophages) was decreased in the combination group compared to the tigecycline group (Fig. 7f, Supplementary Fig. 15b). Therefore, this remodeling of immune pattern may further improve the efficacy of the combination therapies.