RIP140 inhibits glycolysis-dependent proliferation of breast cancer cells by regulating GLUT3 expression through transcriptional crosstalk between hypoxia induced factor and p53

Glycolysis is essential to support cancer cell proliferation, even in the presence of oxygen. The transcriptional co-regulator RIP140 represses the activity of transcription factors that drive cell proliferation and metabolism and plays a role in mammary tumorigenesis. Here we use cell proliferation and metabolic assays to demonstrate that RIP140-deficiency causes a glycolysis-dependent increase in breast tumor growth. We further demonstrate that RIP140 reduces the transcription of the glucose transporter GLUT3 gene, by inhibiting the transcriptional activity of hypoxia inducible factor HIF-2α in cooperation with p53. Interestingly, RIP140 expression was significantly associated with good prognosis only for breast cancer patients with tumors expressing low GLUT3, low HIF-2α and high p53, thus confirming the mechanism of RIP140 anti-tumor activity provided by our experimental data. Overall, our work establishes RIP140 as a critical modulator of the p53/HIF cross-talk to inhibit breast cancer cell glycolysis and proliferation. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-022-04277-3


Introduction
In a normal resting cell, glycolysis converts glucose to pyruvate, which enters the tricarboxylic acid cycle where it becomes oxidized to generate ATP into mitochondria. In the absence of oxygen, glucose is still degraded into pyruvate, which is now converted into lactate. On switching to proliferative mode, cells increase glycolysis and reduce oxidative phosphorylation, which results in a high rate of glycolysis leading to lactate production. This metabolic switch was first described by Otto Warburg in the 20's, who observed that cancer cells prefer glycolysis to mitochondrial respiration to produce energy, even in the presence of oxygen [1]. This conversion of glucose to lactate is now established as the Warburg effect, or aerobic glycolysis. At first, Warburg hypothesized that cancer arises from impaired mitochondria. However, experimental observations of functional mitochondria in cancer cells have since refuted this hypothesis [2]. In fact, increased aerobic glycolysis has been frequently verified in tumors.
It is now established that cancer cells switch to glycolysis to supply their increased need for biosynthetic precursors and organic resources to synthetize cell components [3]. As glycolysis produces less ATP than oxidative phosphorylation, cancer cells compensate for this energy gap by up-regulating glucose transporters to import more glucose into the cell. Enhanced glycolysis also produces reducing equivalents and many glycolytic intermediates are diverted into the pentose phosphate pathway to produce NADPH [4], allowing cancer cells to fight against reactive oxygen species and oxidative stress [5].
The switch to glycolysis in cancer cells is orchestrated by oncogenes and tumor suppressors [6]. Specifically, the tumor suppressor gene p53 favors oxidative phosphorylation over glycolysis, enhancing the production of reactive oxygen species, to imbalance redox status and promote cell death [7,8]. p53 directly inhibits glycolysis by blocking the expression of glucose transporters [9, 10]. Conversely, inactivating p53 reduces oxygen dependence and allows cancer cells to grow in oxygen-limited conditions, such as hypoxia. Hypoxia induces the stability of hypoxia inducible factors, or HIFs, which have an oxygen sensitive α sub-unit (HIF-1α, HIF-2α or HIF-3α) and one stable β sub-unit (HIF-1β; also known as ARNT) [11]. In cancer cells, HIFs antagonize p53 and stimulate glycolysis. However, the mechanisms of their direct influence on glycolysis needs further clarification [12,13].

RIP140-deficiency promotes cell proliferation and tumor growth.
As mentioned above, opposite effects of RIP140 on human breast cancer cell proliferation have been described. Aiming to clarify this situation, we evaluated the impact of RIP140 silencing after small interfering RNA (siRNA) knockdown with two separate siRNAs on the proliferation of MCF7 and MDA-MB-436 breast cancer cell lines.
As shown by xCELLigence real-time cell analysis and MTT assays (Fig. 1a, b and Supplementary Fig. 1a-e), RIP140 silencing consistently increased cell proliferation confirming our previous results [27,28]. To further validate the antiproliferative activity of RIP140 on human cancer cells, we performed the same type of experiments in prostate (DU145) and colon (RKO) human cancer cell lines. As shown in Supplementary Fig. 1e, knockingdown RIP140 expression in these cancer cells robustly increased proliferation in our experimental conditions. We also studied the proliferation of mouse embryonic fibroblasts (MEFs) knocked-out for the Rip140 gene (RIPKO, Supplementary Fig. 1f). 3T3-immortalized RIPKO MEFs proliferated more than 3T3-immortalized MEFs expressing Rip140 (WT), as shown by xCELLigence and MTT assays ( Fig. 1c and Supplementary Fig. 1g).
Immortalized RIPKO cells expressed more the proliferation markers, Cyclin A and Phosphorylated Serine10-Histone 3, and were richer in ATP ( Fig. 1d and Supplementary Fig. 1h). In SV40-HRas/V12-transformed MEFs, RIP140-deficiency increased the cell proliferation and the number of colonies as shown, respectively, by MTT and soft agar assays ( Supplementary Fig. 1i, j). To confirm the role of RIP140 in the observed effects, we rescued RIP140 expression by generating stable MEFs expressing either GFP or human RIP140 ( Supplementary Fig. 1k).
As the above data built a strong case for a growth inhibitory role of RIP140, we wondered whether RIP140 could also affect tumor growth. We therefore xenografted transformed RIPKO MEFs into nude mice and found that, indeed, these tumors had enhanced growth when compared to transformed MEFs expressing Rip140 (WT) ( Fig. 1g and Supplementary Fig. 1n, o). Altogether, our findings strongly push toward an anti-proliferative effect of RIP140 on cancer cell proliferation.

Inhibiting glycolysis reduces the growth advantage of RIP140-deficient cells.
As glucose homeostasis is a key parameter in the control of tumor cell proliferation, we sought to determine the importance of glucose for RIP140-deficient cell growth. We first performed glucose starvation experiments and monitored cell proliferation. Lowering glucose concentrations affected RIPKO cell proliferation more than WT cells ( Fig. 2a and Supplementary Fig. 2a). As shown in Supplementary Fig. 2b, the growth advantage of RIPKO cells observed in glucose-rich medium was abolished in the absence of glucose.
Upon glucose starvation, previous work reported that transformed MEFs displayed morphological features of cell death, such as loss of plasma membrane integrity and cell fragmentation (Chiaradonna et al., 2006).
Transformed RIPKO MEFs displayed more features of dead cells than transformed WT cells (Fig. 2b).
Furthermore, we found that transformed RIPKO MEFs were more sensitive to glucose limitation than transformed WT MEFs ( Supplementary Fig. 2c). These data suggested that RIP140-deficiency triggered glucose starvationinduced cell death.
Next, we analyzed cell viability after glycolysis blockade with the hexokinase inhibitor 2-deoxyglucose (2DG). As for glucose starvation, 2DG treatment abolished the RIPKO cell growth advantage ( Fig. 2c and Supplementary 2d). Moreover, treatment with the GAPDH inhibitor 3-Bromopyruvate (BrP) induced the same response ( Supplementary Fig. 2e). 2DG treatment reduced cell proliferation and ATP level more in RIPKO MEFs (Fig 2d, e and Supplementary Fig. 2f). Moreover, RIPKO cells formed no colonies in the presence of the drug ( Fig. 2f and Supplementary Fig. 2g). In preclinical models, to treat engrafted nude mice with 2DG led to a greater reduction in volume of RIPKO than WT tumors (Fig 2g). Finally, both glycolysis inhibitors reduced the viability of MCF7 and MDA-MB-436 cells more efficiently when RIP140 was silenced by siRNA ( Fig. 2h and Supplementary Fig. 2h).
Altogether, these results demonstrate that the growth advantage of RIP140-deficient cells is abolished when glycolysis is impaired. Therefore, RIP140-deficient cells rely on glycolysis to grow, suggesting that RIP140 inhibits cell proliferation by blocking glycolysis.

RIP140-deficiency enhances aerobic glycolysis.
Because these results suggested that RIP140 inhibits glycolysis in tumor cells, we characterized the glycolytic properties of RIP140-deficient cells. We first demonstrated that glucose uptake was higher in immortalized and transformed RIPKO MEFs (Fig. 3a). We then performed Seahorse flux analysis that allow the measurement of the glucose-induced extracellular acidification rate, reflecting glycolysis. The glycolytic parameters, such as glycolysis, glycolytic capacity and glycolysis reserve, were higher in RIPKO MEFs ( Fig. 3b and Supplementary   Fig. 3a). The high glycolysis content of RIPKO cells was confirmed in transformed MEFs ( Supplementary Fig.   3b). Furthermore, glucose consumption and lactate production were higher in RIPKO cells, confirming the Seahorse analysis (Fig. 3c, d). Then, rescuing RIP140 expression reduced extracellular acidification, confirming the inhibitory effect of RIP140 on glycolysis (Fig. 3e). Finally, silencing RIP140 by siRNA in MCF7 and MDA-MB-436 human breast cancer cells increased glycolysis ( Fig. 3f and Supplementary 3c). Similar results were obtained in prostate and colon cancer cell lines ( Supplementary Fig. 3d).
Thus, taken together, these results demonstrate that the loss of RIP140 increased glycolysis in cancer cells.
GLUT3 is essential for the growth advantage of RIP140-deficient cells.
To investigate how RIP140 regulates glycolysis in cancer cells, we monitored by RT-qPCR the expression of 23 different metabolic genes involved in glycolysis, in WT and RIPKO immortalized MEFs. The mRNA levels of Glut3, Glut4 and Ldhb (Lactate Dehydrogenase B) were increased more than 2-fold in RIPKO MEFs ( Supplementary Fig. 4a). The increase of Glut3 mRNA expression level was the most important and was recapitulated in transformed MEFs ( Supplementary Fig. 4b). Furthermore, this gene was also significantly increased by RIP140 silencing in the breast cancer cell lines MCF7 and MDA-MB-436 ( Fig. 4a and Supplementary   Fig. 4c), whereas the LDHB and GLUT4 levels did not change significantly (data not shown). Furthermore, rescuing RIP140 expression in RIPKO MEFs down-regulated Glut3 expression (Fig. 4b).
To investigate the mechanism by which RIP140 inhibits GLUT3 expression, we first showed by luciferase reporter assays using the GLUT3 promoter that RIP140 knock-down increased the luciferase activity of this reporter in MDA-MB-436 and MEFs cells (Fig. 4c, d) and that RIP140 overexpression decreased it in MEFs cells (Fig. 4d).
By chromatin immunoprecipitation (ChIP) assay in MDA-MB-436 cells, we observed the recruitment of RIP140 to the GLUT3 promoter, indicating that the regulation by RIP140 occurred directly at the transcriptional level (Fig. 4e). Moreover, by analyzing ChIP-seq data in MCF7 cells [24], we found RIP140 bound in the vicinity of the GLUT3 (SLC2A3) gene. Altogether, these data identify the GLUT3 gene as a transcriptional target of the trans-repressive activity exerted by RIP140.
To evaluate to which extent RIP140-deficient cells needed GLUT3 to grow, we first validated the induction of GLUT3 at the protein level in MEFs ( Supplementary Fig. 4d) and then knocked it down with small hairpin RNA (shRNA) in MEFs, thus confirming the specifity of the GLUT3 antibody ( Supplementary Fig. 4e). Knocking down GLUT3 abrogated the growth advantage induced by RIP140-deficiency as shown by either colony formation Altogether, these data show that the increase of cell proliferation upon RIP140-deficiency requires GLUT3 induction.

RIP140 and p53 inhibit the expression of GLUT3 induced by HIF-2.
One direct DNA binding factor that could likely mediate RIP140 activity is p53, as it is a well-known inhibitor of the Warburg effect (Gomes et al., 2018). Thus, we investigated whether p53 mediated any of the transcriptional activities of RIP140 on GLUT3 expression. Using a luciferase reporter assay, we showed that increasing doses of p53 inhibited the luciferase activity driven by the GLUT3 promoter (Fig. 5a), and that RIP140 reinforced this inhibition (Fig 5b). Of interest, RIP140 deficiency resulted in decreased p53 expression at the protein and mRNA levels ( Fig. 5c and Supplementary Fig. 5a). Therefore, RIP140-deficient cells exhibit many of the characteristics of cells under hypoxia such as cell proliferation, increased glycolysis and inactivation of p53 [12].
To investigate the role of hypoxia in Glut3 overexpression upon RIP140-deficiency, we first up-regulated the HIFs with a luciferase gene reporter driven by the GLUT3 promoter. We found that the overexpression of HIF-2α, but not HIF-1α, increased the activity of the GLUT3 promoter (Fig. 5d). Indeed, HIF-1α regulatory regions on GLUT3 gene are not located within the promoter region [30]. We also found that HIF-2α, and not HIF-1α, was overexpressed in RIP140-deficient cells ( Fig. 5e and Supplementary Fig. 5b). Interestingly, HIF-2α silencing reduced the expression of Glut3, only in RIP140 KO MEFs, suggesting that HIF-2α is responsible, at least in part, for Glut3 overexpression upon RIP140 deficiency ( Fig. 5f and Supplementary Fig. 5b).
To evaluate whether RIP140 could target HIFs activity, we first measured the effect of RIP140 on the transcriptional activity of HIFs using a luciferase reporter assay. We found that the transcriptional activity of a reporter gene containing hypoxia-response elements was higher in RIPKO than in WT MEFs and that increasing doses of RIP140 repressed the basic activity of this reporter gene (Fig. 5g). p53 is known to inhibit the transcriptional activity of HIFs (Blagosklonny et al., 1998). We then wondered whether RIP140 and p53 could cooperate to inhibit HIF transcriptional activity. We first looked at the effect of RIP140 or p53 individually on the transcriptional activity of the reporter gene containing hypoxia-response elements when transactivated by HIF-1α or HIF-2α. RIP140 or p53 were both capable of inhibiting HIF-α transcriptional activity separately ( Fig. 5h and Supplementary Fig. 5c). Of interest, the repressive activity was stronger when RIP140 and p53 were expressed together, suggesting that RIP140 and p53 cooperate to inhibit HIF-α transcriptional activity ( Fig. 5i and Supplementary Fig. 5d). The same synergistic repressive effect was observed on the GLUT3 promoter when transactivated by HIF-2α (Fig. 5j).
Performing a proximity ligation assay demonstrated that HIF-2α, p53 and RIP140 interacted with each other (Fig.   5k). Altogether, these data suggested that the molecular mechanisms by which RIP140 inhibits GLUT3 transcription relies on the repression of HIF transactivation through the formation of a ternary complex with p53.

The prognostic value of RIP140 is correlated with the levels of GLUT3 expression
Our experimental data demonstrated that RIP140 inhibits efficiently GLUT3 expression in synergy with p53, by antagonizing HIF-2 function, leading to a decrease in cancer cell proliferation. We therefore hypothesized that, in breast cancers, low GLUT3 levels could be a marker of such anti-tumor activity of RIP140 and therefore, that RIP140 might be associated with an increased overall survival for patients with tumor expressing a reduced level of GLUT3, as a surrogate of RIP140 anti-tumor activity.
Using Cox proportional hazard regression [32], we analyzed RNAseq data obtained from 1068 breast tumor samples from the TCGA dataset as described previously [33] (Fig. 6). We first checked the prognostic values of the expression of each gene by analyzing patient overall survival at 60 months. High RIP140 (NRIP1) expression was correlated with good prognosis whereas high GLUT3 (SLC2A3) was associated with bad prognosis (Fig. 6a).
Then we used the median as a cutoff value for classification of patients into two groups of tumors with low and high GLUT3 expression, respectively. Using Kaplan-Meier plots, we investigated whether RIP140 expression correlated with OS at 60 months within both groups using the same cutoff values. Interestingly, and as expected, RIP140 expression was significantly associated with increased overall survival in low GLUT3 expression group but not in high GLUT3 expression group (Fig. 6b).
We pursued our hypothesis by analyzing the prognostic value of RIP140 based on p53 (TP53), HIF-2α (EPAS1) and HIF-1α (HIF1A) expression levels. Patients were stratified again into low and high expression groups by using, as cutoff values, the median of either TP53, HIF-2α, or HIF-1α expression ( Fig. 6c and Supplementary Fig.   6a). Strikingly, RIP140 expression was significantly associated with good prognosis in high p53 and low HIF-2 or HIF-1α expression groups, confirming the mechanism of RIP140 anti-tumor activity provided by our experimental data (Fig. 6c). Moreover, the prognostic value of RIP140 was significantly associated with good prognosis in low GLUT3 expression group but not in high GLUT3 expression group in colon and stomach cancers, suggesting that a reduced level of GLUT3 could reflect RIP140 anti-tumor activity in other types of cancer ( Supplementary Fig. 6b).

Discussion
Glycolysis is essential for supporting the rapid proliferation of tumor cells. Our data reveal that the transcription coregulator RIP140 inhibits the glycolysis-dependent proliferation of breast cancer cells by impeding glycolysis through the blockade of GLUT3 expression via a mechanism involving a p53-mediated inhibition of

HIF activation
Our results first demonstrate that the glucose deprivation or glycolysis blockade abolished the gain in cell proliferation caused by the decrease or loss of RIP140 expression, showing the glucose dependency of RIP140deficient cells. Furthermore, our data demonstrate that down-regulating Glut3 expression provoked the same inhibition of RIP140-deficient cell proliferation. H-Ras transformation is well recognised to drive cancer cells towards glycolysis and glycolysis is essential for H-Ras transformation [6]. Of note, transformed RIPKO cells were more sensitive to glucose starvation than immortalized RIPKOMEFs ( Fig. 2b and Supplementary Fig. 2a), suggesting that RIP140-deficiency might influence the transformation process by regulating glycolysis or that Ras transformation could enforce the potential of RIP140. However, RIP140 potential was also observed in Rasindependent cells, such as immortalized MEFs or the human cancer line MDA-MB-436.
RIP140 potential was also independent of PTEN which is another driver of cancer glycolysis. Indeed, we Our tailored experiments describe the mechanism by which RIP140 inhibits GLUT3 expression, which relies on the cooperation of RIP140 and p53 to inhibit the expression of GLUT3 induced by HIF-2α (Fig. 5g). p53 wildtype seems to be dispensable for the cooperation since RIP140 silencing induced GLUT3 expression in p53 mutated cells such as the MDA-MB-436 breast cancer cell line. p53 could act as a facilitator of the RIP140 repressive effect. Proximity ligation assays allowed us to visualize the three partners in close proximity two-bytwo, suggesting that they are involved in a ternary complex (Fig. 5h). Our data reveal for the first time that RIP140 interacts with p53 and HIF. Whether there is a ternary complex or if the interactions follow a temporal order will need to be defined in future studies. It is tempting to speculate that RIP140 could act as an integrator protein such as CBP/p300 as p53 and HIF antagonism relies on competition for p300 (Schmid et al. HIFs remains a complex question and it is still debated whether their reciprocal influence has any direct consequences for metabolism in cancer. Our results add RIP140 as a new major player in this interplay, and provide an additional bond between cell metabolism and cancer progression. The overexpression of GLUT3 has been described in many types of cancer with poor outcomes (Benito et al., 2017). In breast cancer patients, we found that RIP140 is associated with an increased overall survival of patients with tumor expressing a reduced level of GLUT3, as a surrogate of RIP140 anti-tumor activity (Fig. 6). Cellular metabolism in cancer is currently being targeted in clinical trials with some success (Sborov et al., 2015). Our results suggest that these genes could be used as a signature to identify patients that could benefit of therapies targeting glycolysis and/or based on HIF inhibitors.
Altogether, our results enable us to propose a new model ( Figure 7) explaining the transcriptional control of glycolysis-dependent cancer cell proliferation by a nuclear interplay between three actors that might be clinically relevant for breast cancer patients.

Materials and Methods
Plasmids and reagents.  Table 1 for the list of primer sequences. Immunofluorescence and proximity ligation assays (PLA) were performed as described [28].

Protein detection.
Luciferase assay. For gene reporter assays, cells were plated in 96-well plates and transfected with JetPEI (Polyplus, Illkirch-Graffenstaden, France) according to the manufacturer's protocol. Data were expressed as relative firefly luciferase activity normalized to renilla activity. For all experiments, data were collected from at least three biological replicates.
Chromatin immunoprecipitation. Approximately one hundred and twenty million cells were harvested per experiment. Briefly, protein-DNA complexes were first cross-linked, using 2mM Di (N-succinimidyl) Glutarate (DSG), for 50 min on a rotor, then the procedure was performed using High-Sentivity kit (Active Motif, Shanghai) and according to manufacturer's protocol. Chromatin was sonicated for thirty seconds on ice followed by thirty seconds off, for a total of thirty minutes. Immunoprecipitations were performed with 30µg of protein-DNA complexes and 4µg of RIP140 antibody (Ab42126; Abcam). The purified DNAs were amplified with SYBR Green SensiFAST™ SYBR® No-ROX Kit (Bioline, London, UK) by real-time qPCR.

Cell growth analysis
MTT assay. Cell proliferation assays were performed using MTT as previously described [28]. Data were normalized to the absorbance value at day 1. Experiments were performed three times with at least sextuplets. For glucose starvation experiments, cells grew for 24 hours in complete media. Then cells were washed twice with PBS and starvation medium (DMEM5030 supplemented with 10% dialyzed FBS, 2mM Glutamine, 1% Penicillin/Streptomycin, HEPES 10mM, 1mM Sodium Pyruvate, 3.7g/l Sodium Bicarbonate) was added to the cells supplemented, or not, with indicated glucose concentrations. xCELLigence analysis. The Real-time Cell Analyzer DP instrument (Agilent, Santa Clara, USA) was placed in a humidified incubator maintained at 37ºC with 5% CO2. Cells were plated at 1500 cells/well into 16-well E-Plates. The impedance value of each well was automatically monitored every hour for up to the indicated times by the xCELLigence system. Cell proliferation is represented by an index that reflects changes in electrical impedance matching to cellular coverage of the electrode sensors. Seahorse experiments. Extracellular acidification rate (measured in mpH/min) was monitored using an XFe24 extracellular flux analyzer from Seahorse Bioscience (Agilent, Santa Clara, CA) following the manufacturer's protocol. Experiments were carried out on confluent monolayers. Briefly, cells were seeded 24 hours before experiments at a density of 35,000 cells/well (24-well). Before starting measurements, cells were washed once with PBS and medium was replaced with Seahorse XF Base Medium supplemented with 2mM glutamine without glucose at pH 7.4) and were placed into a 37 °C non-CO2 incubator for 1 hour prior to the assay. Glucose, oligomycin, and 2-DG were diluted into XF24 media and loaded into the accompanying cartridge to achieve final concentrations of 10 mM, 1 μm, and 100 mM, respectively. Injections of the drugs into the medium occurred at the time points specified. Each cycle consisted of 3 min mixing, 3 min waiting and 3 min measuring. Data were transformed with Agilent Seahorse Wave software to export glycolysis parameters. Values were expressed after normalization to the protein content of each well and to the values just before glucose addition.
Glucose and Lactate assays. Cells were seeded in culture dishes and cultured for 8 h. The culture medium was then changed and cells were incubated for an additional 16 h. Subsequently, the culture medium was collected for determination of glucose concentration and lactate levels using a Glucose assay kit (Amplite TM Glucose Quantitation Assay kit, AAT Bioquest, Sunnyvale, CA) and a Lactate assay kit (MAK064, Merck, Darmstadt, Germany) according to the manufacturer's instructions. Glucose consumption was calculated as the difference in glucose concentration between fresh medium and cell supernatant. Lactate production was determined as the difference in lactate concentration between cell supernatant and fresh medium. Data were normalized to final cell counts and incubation time. Statistical analysis. Data are expressed as mean ± SD. Statistical analysis was conducted via StatEL (www.adscience.fr). The Mann-Whitney U-test was used to compare two independent groups. The p values less than 0.05 were considered to be statistically significant.

Table1. Murine primer sequences
Name Sequence

Competing Interests
The authors declare no competing financial interests.

Data Availability
The links to publicly archived datasets analyzed in this study and all materials are available on request from the corresponding author.

Ethical approval and consent to participate
All experiments on mice were performed in accordance with the French guidelines for experimental animal studies (agreement n° 201603101538202            RIP140 and p53 cooperate to inhibit the expression of GLUT3 induced by HIF-α. Glycolysis-dependent proliferation of breast cancer cells is reduced, due to a decrease in glycolysis. The prognostic value of RIP140 is associated with good survival in patients with low GLUT3, high p53 and low HIF-α (right panel).
In patients with high HIF-α, low p53 and high GLUT3, RIP140 and p53 do not inhibit the transcriptional activity of HIF-α; GLUT3 is highly expressed and glycolysis is enhanced. In this sub-group, RIP140 expression level is not correlated with good survival.
Double blue lines represent cell membranes. The grey square represents GLUT3 gene, the grey ovoid forms represent GLUT3 protein. The orange circle represents glucose.
b, The slope of the curves was extracted using the RTCA Software from the curves in Fig. 1a and 1b (mean ± SD, n=3, ***p <0.001).