RYR1 is an unfavorable prognostic marker for USC
To exploit the promising therapeutic target candidates associated with advanced disease stages and prognosis in USC, we first identified disease-stage related prognostic markers specific for USC using Significance Analysis of Microarray (SAM) on TCGA dataset (Uterine Corpus Endometrial Carcinoma). 2064 and 1014 genes were upregulated in USC vs Normal control and USC vs EEC comparisons, respectively. In addition, 1977 genes were either significantly correlated with advanced disease stages or overall survival in USC patients by Cox regression analysis. Among these genes, 51 genes overlapped with those that showed significantly higher expression in USC than normal control and EEC, suggesting that they represent prognostic markers specific for USC. In order to identify a repositioned drug for the treatment of USC, we further performed a DrugBank search and found 6 out of 51 genes (RYR1, SCN4A, ANKLE1, KCNQ3, PSAT1 and SLC16A5) can be targeted by FDA-approved drugs (Fig. 1A). We selected RYR1 for further study as it showed the highest fold difference (9.42 folds) between USC and EEC (Table S1).
Next, we asked whether USC cells demonstrated higher RYR1 expression levels than normal counterparts in vitro. ddPCR showed that USC cell lines (ARK1, ARK2, and HEC50) all had significantly higher RYR1 mRNA levels than EEC cell lines (HEC-1A, HEC-1B, HEC-59, and ECC) and normal uterine epithelial (UE) cells (Fig. 1B). Western blot also demonstrated significantly higher RYR1 protein in USC cell lines than UE cells (Fig. 1C).
RYR1 promotes tumor progression and inhibits apoptosis in USC
RYR1 encodes the ryanodine receptor type 1 protein predominately found in skeletal muscle cells [21]. It functions as a Ca2+ release channel located in the membrane of sarcoplasmic reticulum (SR, muscle cells) or endoplasmic reticulum (ER, non-muscle cells) to mediate the release of Ca2+ from the SR/ER. Several drugs or compounds are known to modulate RYR1 activity, such as dantrolene [11], ryanodine [22], ruthenium red [23] and others. Of these, dantrolene is the only specific RYR1 antagonist that is approved by FDA for the treatment of malignant hyperthermia [12]. To determine the role of RYR1 in conferring malignant phenotypes in USC, stable knockdown of RYR1 was performed in USC cell lines ARK1, ARK2 and HEC50 using RYR1-specific small hairpin RNA (shRNA), and RYR1 overexpression was generated by transfecting the full-length RYR1 cDNA construct. Knockdown and overexpression of RYR1 were confirmed by ddPCR and Western blot, respectively (Supplementary Fig. 1A and B). MTT assay demonstrated that RYR1 silencing suppressed cell proliferation in the 3 USC cell lines with higher proliferation rates in RYR1-overexpressing cells (Fig. 2A and B). Depletion of RYR1 caused a reduction in USC cell migration as detected by the transwell assay (Fig. 2C). The apoptotic cell ratio was increased in USC cell lines after RYR1 knockdown as assessed by flow cytometry (Fig. 2D). Conversely, cell migration was enhanced, and apoptosis was suppressed in RYR1 overexpressing cells (Supplementary Fig. 1C and D). In contrast, dantrolene treatment did not significantly decrease cell viability in EEC cell lines HEC1A and HEC1B (Supplementary Fig. 1E), which do not express detectable levels of RYR1 mRNA. These findings indicate RYR1 is a potential common pharmaco-therapeutic target for suppressing the malignant phenotypes of USC cells.
We examined whether dantrolene could potentially be repurposed for the suppression of the USC phenotypes, as observed in RYR1-silenced USC cells. Application of dantrolene exhibited a significant inhibitory effect on the proliferation and mobility of USC cells (Fig. 2E and F). Furthermore, flow cytometry demonstrated that dantrolene-treated USC cells had significantly higher numbers of apoptotic cells than controls (Fig. 2G). To confirm the on-target effect of dantrolene in vitro, RYR1-silenced ARK1 cells were treated with dantrolene (50 μM) or vehicle solution, and cell growth was monitored. Significant inhibitory effect of dantrolene was observed in the mock transfectants instead of RYR1 shRNA transfected ARK1 cells (Supplementary Fig. 1F). These findings imply that RYR1 mediates the effect of dantrolene in suppressing the malignant phenotype in USC cells.
To examine the tumor-promoting effect of RYR1 in vivo, luciferase-labeled RYR1 shRNA or mock transfected ARK1 cells were intraperitoneally injected into 6-week-old female nude mice. Tumor progression was examined by measuring bioluminescence every 2 weeks (Fig. 2H). Mice injected with RYR1-silenced ARK1 cells had significantly lower bioluminescence signals detected at week 10, compared to those injected with control cells (Fig. 2I). Survival analysis showed that mice injected with RYR1-silenced ARK1 cells had significantly longer overall survival times than controls (Fig. 2J), supporting the hypothesis that RYR1 plays a crucial role in USC progression. Similarly, mice were injected intraperitoneally with dantrolene (5 mg/kg) or vehicle 3 times per week for 5 weeks, 28 days after luciferase-labeled ARK1 cell inoculation to check dantrolene effect in vivo. Dantrolene-treated group had significantly lower bioluminescence signals than control group (Fig. 2K and L). There was no significant difference in the body weight of dantrolene- and vehicle-treated mice (Supplementary Fig. 1H), suggesting minimal toxicity of dantrolene in the mice.
To further determine the effect of RYR1 silencing and dantrolene treatment on tumor cell proliferation, angiogenesis, and apoptosis, immunolocalization (IHC) of Ki-67, CD31, and cleaved caspase 3 (CCP3) were performed on tumor tissues harvested in the mouse models. USC tumors of the RYR1 knockdown group and the dantrolene-treated group had significantly lower numbers of Ki-67- and CD31-positive cells and higher numbers of CCP3 cells compared to controls (Fig. 2M and N, and Supplementary Fig. 1I). We also confirm the on-target effect in vivo, female athymic nude mice were first injected with luciferase-labeled ARK1 cells stably expressing RYR1 shRNAs or control shRNAs intraperitoneally to establish tumor. They were then treated with dantrolene or vehicle every other day for 6 weeks. There was a significant decrease in bioluminescence signals in the RYR1-silenced group and dantrolene-treatment group compared to the control group. However, there is no significant difference in tumor growth in RYR1-silenced group with or without dantrolene treatment (Supplementary Fig. 1G), which is similar to what we observed in vitro. Taken together, our results strongly support the tumorigenic role of RYR1 in USC and the feasibility of repurposing dantrolene as an effective agent for the treatment of USC.
RYR1 modulates intracellular calcium signaling in USC cells
To investigate the RYR1-dependent changes in Ca2+ homeostasis of USC cells, cytosolic [Ca2+] was first determined by radiometric measurement of Fura-2 fluorescence in ARK1-shCtrl, ARK1-shRYR1, ARK1-Control and ARK1-RYR1 cells. The resting [Ca2+]i was significantly suppressed in the shRYR1 cells (shCtrl: 99.4 ± 2.1 nM, n = 58; shRYR1: 74.4 ± 3 nM, n = 87; p < 0.001) and significantly elevated in RYR1-overexpressing cells (Control: 94 ± 3.2 nM, n = 171; RYR1: 104.5 ± 2.2 nM, n = 217; p < 0.001; Fig. 3A), suggesting the differences in resting [Ca2+]i in the indicated cells were related to Ca2+ signals mediated by RYR1.
The activity of RYRs was further examined by the RYR agonist 4-Chloro-m-cresol (4-CMC). Application of 4-CMC (5 μM) caused a rapid transient increase in cytosolic [Ca2+]. The responses were significantly suppressed in the shRYR1 cells and accentuated in the RYR1-overexpressing cells (Fig. 3B). These results suggested that functional RYR1 is expressed and operated as an important determinant of Ca2+ homeostasis in ARK1 cells. RYR1 may modulate mitochondrial [Ca2+] ([Ca2+]m) through ER-mitochondrial Ca2+ transfer [24]. ER-mitochondrial Ca2+ transfer was monitored in ARK1 cells transfected with the mitochondrial Ca2+ biosensor CEPIA2mt. Application of 4-CMC caused a rapid increase in [Ca2+]m in the control ARK1 cells. The response was enhanced in the RYR1-overexpressing cells and was significantly suppressed in ARK1 cells transfected with shRYR1 (Fig. 3C). In a separate set of experiments, ARK1 cells were transfected with both the ER specific Ca2+ biosensors G-CEPIA1er and CEPIA2mt for simultaneous recording of ER [Ca2+] ([Ca2+]ER) and [Ca2+]m, respectively. Activation of RYRs by photorelease of caged cADP-ribose (cADPR), an endogenous activator of RYRs, at a subcellular region of interest (ROI) caused an immediate transient reduction in [Ca2+]ER, which was associated with a fast, transient increase in [Ca2+]m (5–10 s) followed by a reduction in the mitochondrial Ca2+ signal (Fig. 3D and E). The ER-mitochondrial Ca2+ transfer occurred locally within the ROI undergoing photolysis, without affecting [Ca2+]ER or [Ca2+]m in the other ROIs. These experiments clearly demonstrated that RYR is capable to mediate ER-mitochondrial Ca2+ transfer in ARK1 cells.
RYR1 silencing down-regulates mitochondrial genes in USC
To identify the RYR1-mediated signaling network that modulates the malignant phenotypes of USC, unbiased reverse-phase protein array (RPPA) was performed on over 400 proteins of key signaling networks. Cell lysates from RYR1 shRNA and scramble shRNA-transfected ARK1 cells as well as dantrolene- and vehicle-treated ARK1 cells were used to identify the overlapping differentially expressed proteins that are associated with down-regulation of RYR1 expression and activity. A total of 305 downregulated and 292 upregulated proteins were found. Among them, 64 downregulated and 38 upregulated proteins were common in both experimental treatments (Supplementary Fig. 2A).
Our attention was drawn to proteins in the top rank of the overlapping list (Fig. 4A and B). They included components of mitochondrial electron transport chain (mETC) (e.g., COXIV, NDUFB4, ATP5a, SDHA), suggesting the potential role of RYR1 in modulating mitochondrial functions. The RPPA results were validated by qPCR and Western blot demonstrating that the USC cells stably transfected with RYR1 shRNA had significantly lower mRNA and protein levels of NDUFB4 (Complex I), SDHA (Complex II), COXIV (Complex IV) and ATP5a (Complex V) than the mock transfectants (Fig. 4C and D). Furthermore, dantrolene (50 μM) treatment of USC cells caused significant time-dependent reduction in these mETC gene mRNA levels (Fig. 4E).
To further determine the role of RYR1 in regulating mETC gene expression in USC tissue samples, regression analysis was performed using TCGA datasets to determine the correlation between mRNA expression of RYR1 and the mETC genes. The results showed significantly higher expression of NDUFB4, SDHA and ATP5a in USC than in grade 1 and grade 3 EECs (Fig. 4F). Significant positive correlations were also observed between the expression of RYR1 and the mETC genes in USC, except COXIV (Fig. 4G). In addition to the above 4 mETC genes, we found positive correlations between RYR1 and some other members of NDUF, SDH, COX, and mitochondrial ATPase subfamilies in GEO datasets (GSE24537) (Table S2), suggesting that these mETC genes may play a role in mediating the tumor-promoting effect of RYR1 in USC.
RYR1 modulates mitochondrial bioenergetics properties in USC
Our results demonstrated that RYR1 regulated the expression of key mETC components (Fig. 5A) in USC cells, suggesting that RYR1 may modulate mitochondrial bioenergetics in these cancer cells. To test this hypothesis, the RYR1-dependent regulation of oxidative phosphorylation (OXPHOS) was determined in ARK1 cells. Oxygen consumption rate (OCR) and ATP production rate were first determined by a Seahorse Analyzer. The results showed that there was a significant reduction in OCR (Fig. 5B and C), and basal mitochondrial and whole cell ATP production rates (pmol/min) (Fig. 5D and E) in shRYR1-transfected or 50 μM dantrolene-treated ARK1 cells. In contrast, ATP production rate was significantly higher in RYR1 stably transfected cells than the mock transfectants (Fig. 5F).
Next, we determined total live-cell ATP production and mitochondrial membrane potential (MMP) in USC cells using CellTiter-Glo kit and JC-1 assay. Tetrachloro-teraethyl benzimidazole carbocyanine iodide (JC-1) is a cationic dye that accumulates in mitochondria, which is commonly used for monitoring MMP (Δψm). Usually, under low MMP, JC-1 exists as monomer and yields green fluorescence. The dye will aggregates under MMP leading to red fluorescence [25]. RYR1 depleted ARK1 cells had significantly lower total live-cell ATP and MMP (Fig. 5G and H) with contrast result in RYR1 overexpression cells (Fig. 5I and Supplementary Fig. 3A). Furthermore, USC cell lines treated with dantrolene (50 μM) for different durations showed that dantrolene decreased live-cell ATP levels in a time-dependent manner (Fig. 5J). A significant decrease in MMP was also observed in dantrolene-treated USC cells (Fig. 5K). ARK1-shRYR1 cells induced mouse tumor tissues also demonstrated a significantly reduced ATP production compared to control (Fig. 5L).
To further examine the role of RYR1 on OXPHOS, the effect of RYR1 expression on 2 additional key OXPHOS-related parameters—NAD+/NADH ratio and accumulation of reactive oxygen species (ROS) were determined. The results showed significantly higher NAD+/NADH ratio, mitochondrial and total cellular ROS in RYR1-silenced or dantrolene-treated USC cells than control cells (Fig. 5M-P) with contrasting results in RYR1-overexpressing cells (Supplementary Fig. 3B and C). Taken together, these results demonstrate that RYR1 alters mitochondrial biogenetics properties through modulating OXPHOS process in USC.
RYR1 silencing suppresses glycolysis and TCA cycle in USC
To exploit energy metabolism alterations driven by RYR1, targeted metabolomics analysis was performed using liquid chromatography-tandem mass spectrometry (LC-MS/MS) in RYR1-deficient ARK1 cells and control cells. The results revealed a markedly lower concentration of glycolytic and TCA metabolites, such as G6P/F6P, FBP/GBP, citrate, succinate and malate in both RYR1 knockdown and 50 μM dantrolene-treated cells, suggesting that removal of RYR1 impeded the normal operation of the glycolysis and TCA cycle in ARK1 cells (Fig. 6A-C). Extracellular acidification rate (ECAR), which measures the excretion of lactic acid per unit time in RYR1 knockdown ARK1 cells was determined. ECAR of the shRYR1-transfected ARK1 cells was significantly lower than controls (Fig. 6D).
Interestingly, we found that glutamine levels were significantly increased in RYR1-knockdown and dantrolene-treated ARK1 cells from LC-MS/MS (Fig. 6A and B), suggesting that increased glutamine uptake could be a feedback mechanism by which the cells compensated for the shutting down of both OXPHOS and glycolysis due to RYR1 depletion. In congruent with the LC-MS experiments, glutamine level was significantly elevated after RYR1 silencing or blockage by dantrolene (Fig. 6E and F) and decreased in RYR1-overexpressing cells (Fig. 6G), suggesting that high levels of RYR1 can probably speed up glycolysis and TCA cycle, which does not need glutamine compensation.
Ca2+ dependent AKT/CREB/PGC-1α and AKT/HK1/2 axes is essential for RYR1-mediated USC progression
Peroxisome proliferator-activated receptor coactivator protein-1alpha (PGC-1α) has been shown to bind directly to the promoter regions of mETC genes to upregulate transcription in breast and ovarian cancer cells [26]. We examined whether PGC-1α mediated the effect of RYR1 in regulating mETC expression. Both PGC-1α mRNA and protein levels were significantly lower in ARK1 and ARK2 cells transfected with RYR1 shRNA or treated with dantrolene (50 μM) (Fig. 7A-C). Furthermore, PGC-1α silencing and rescue experiments were performed by overexpressing RYR1 in PGC-1α–silenced ARK1 cells, and mETC genes expression was determined using qPCR. We found that the decrease of mETC genes in PGC-1α-knockdown cells can be compensated by overexpressing RYR1 in those cells (Fig. 7D). These results clearly suggest that RYR1 may indeed regulate PGC-1α expression to modulate the expression of mETC genes in USC cells.
Next, we determined whether hexokinases (HKs), which are key enzymes that catalyze the first step of glycose metabolism by converting glucose to glucose-6-phosphate (G6P), mediate the effect of RYR1 in regulating glycolysis and TCA cycle. Of the HKs, HK1 and HK2 are the 2 predominant isoforms in most human tissues [27]. Western blot showed markedly lower levels of total HK1 and HK2 in RYR1-silenced ARK1 cells than in control cells (Fig. 7E). To further determine the molecular mechanism by which PGC-1α and HK1/2 are regulated by RYR1, we screened all the known Ca2+-dependent signaling molecules and transcription factors from our RPPA and public datasets. We focused on Ca2+/cAMP response element (CRE)-binding protein (CREB), which upon activation can form a functionally active dimer that binds to the cis-acting CRE element within the promoters of target genes. The PPARGC1A gene contains a CREB binding domain within its promoter. CREB phosphorylated by AKT is able to directly interact with this domain and trigger PGC-1α transcription [28]. Our RPPA data demonstrated that RYR1 shRNA-transfected and dantrolene-treated ARK1 cells had significantly lower levels of phosphorylated AKT and CREB than controls (Supplementary Fig. 4A). This finding was validated by Western blot (Fig. 7F).
To further confirm that AKT and CREB play a role in regulating PGC-1α expression, Western blot were performed on ARK1 cells treated with either AKT inhibitor Akti-1/2 or CBP/CREB inhibitor. The results showed markedly lower phosphorylated AKT (pS473) and CREB (pS133) levels as well as PGC-1α levels after treatment of Akti-1/2 or CBP/CREB inhibitor (Supplementary Fig. 4B). Western blot also showed that both HK1 and HK2 expression were reduced in a dose-dependent manner after treatment with Akti-1/2 but not CBP/CREB inhibitor (Supplementary Fig. 4C), suggesting that HK1/2 were specifically regulated by the CREB-independent AKT pathway. Furthermore, RYR1-overexpressing ARK1 cells were treated with Akti1-1/2 or CBP/CREB inhibitor and controls followed by Western blot for PGC-1α. The results showed that the effect of RYR1-enhanced PGC-1α expression was abrogated by Akti1-1/2 and CBP/CREB inhibitor (Fig. 7G), suggesting that RYR1 modulates PGC-1α expression through the AKT/CREB/PGC-1α pathway. To confirm that PGC-1α mediates the promoting effect of RYR1 on ATP production and mETC gene expression, PGC-1α-specific siRNAs or control siRNA was transfected into USC cells with or without stably overexpressing RYR1. The silencing efficiency of PGC-1α siRNAs in USC cells was validated by qPCR (Supplementary Fig. 4D). Cells transfected with full-length RYR1 and control siRNA had significantly higher levels of ATP production, which was reversed by treatment with PGC-1α siRNAs (Fig. 7H). In addition, the levels of mETC genes were also significantly higher in the RYR1-overexpressing cells compared to the mock transfectants. The enhanced expression of mETC genes in RYR1-overexpressing cells was abrogated by Akti1-1/2, CBP/CREB inhibitor, and PGC-1α siRNA in ARK1 cells (Fig. 7I). These results established the causal relationship of the RYR1-dependent AKT/CREB/PGC-1α pathway with elevated ATP level and mETC gene upregulation. In addition, the enhanced cell proliferation rate induced by RYR1 overexpression in ARK1 cells was abolished by inhibitor treatment (Fig. 7J).
To further confirm that RYR1-dependent regulation of the AKT/CREB/PGC-1α pathway is dependent on Ca2+ signals modulated by RYR1, [Ca2+]i was buffered by incubating ARK1 cells with the cell permeant Ca2+ chelator BAPTA-AM (10 μM) for 1, 12, and 24 hours. The levels of p-CREB (S133) and p-AKT (S473) were significantly lowered in cells incubated with BAPTA-AM, accompanied with a significant reduction in the level of PGC-1α. The decrease in p-CREB (S133), p-AKT (S473), and PGC-1α protein levels after BAPTA treatment was attenuated in RYR1 knockdown ARK1 cells (Fig. 7K and Supplementary Fig. 4E). Additionally, the enhanced phosphorylation of AKT and CREB and overexpression of PGC-1α were again effectively obliterated by the Ca2+ chelators (Supplementary Fig. 4F). The mRNA level of PGC-1α induced by RYR1 overexpression was also attenuated by BAPTA treatment (Supplementary Fig. 4G). These results provide further evidence that RYR1-dependent Ca2+ signals mediate the enhanced AKT and CREB activities, leading to the upregulation of PGC-1α (Fig. 7L).