Opposite physiological and pathological mTORC1-mediated roles of the CB1 receptor in regulating renal tubular function

Activation of the cannabinoid-1 receptor (CB1R) and the mammalian target of rapamycin complex 1 (mTORC1) in the renal proximal tubular cells (RPTCs) contributes to the development of diabetic kidney disease (DKD). However, the CB1R/mTORC1 signaling axis in the kidney has not been described yet. We show here that hyperglycemia-induced endocannabinoid/CB1R stimulation increased mTORC1 activity, enhancing the transcription of the facilitative glucose transporter 2 (GLUT2) and leading to the development of DKD in mice; this effect was ameliorated by specific RPTCs ablation of GLUT2. Conversely, CB1R maintained the normal activity of mTORC1 by preventing the cellular excess of amino acids during normoglycemia. Our findings highlight a novel molecular mechanism by which the activation of mTORC1 in RPTCs is tightly controlled by CB1R, either by enhancing the reabsorption of glucose and inducing kidney dysfunction in diabetes or by preventing amino acid uptake and maintaining normal kidney function in healthy conditions.

D iabetes is a chronic disease that is now reaching epidemic proportions 1 , and has been described as a catalyst for a number of conditions, most notably cardiovascular disease, retinopathy, and diabetic kidney disease (DKD). The latter affects approximately 30% of patients with diabetes, and it is strongly associated with morbidity and mortality 2 . DKD is manifested by glomerular hypertrophy, transient hyperfiltration, albuminuria, kidney fibrosis, and ultimately a progressive decline in the glomerular filtration rate 3 . Recently, a paradigm shift in our understanding of kidney dysfunction in diabetic patients has emerged, indicating that tubulopathy precedes glomerular alterations 4,5 . The renal proximal tubular cells (RPTCs) are uniquely susceptible to a variety of metabolic and hemodynamic factors associated with diabetes, especially to hyperglycemia. In fact, glucose entry into RPTCs is insulin-independent, making these cells particularly sensitive to the deleterious effects of chronic hyperglycemia in diabetic patients, that in turn, may lead to enhanced O 2 consumption and increased hypoxic tubular damage 5 . During normoglycemia, however, the RPTCs account for 25% of whole-body gluconeogenesis, and play a key role in glucose 6 and amino acid 7 reabsorption, thus maintaining adequate blood glucose levels and metabolic homeostasis.
A growing body of evidence indicates that hyperglycemia induces changes in glucose transport via GLUT2 and may negatively affect kidney function and the associated tubulo-interstitial changes seen in DKD 48,49 . In contrast to the reduced expression of GLUT2 in pancreatic β cells in different models of diabetes [50][51][52] , its levels in RPTCs rise in diabetic patients 53 as well as in murine models of diabetes and obesity [54][55][56] . Plasma or luminal glucose concentrations are known to regulate its expression and/or translocation 57 , accounting for the deleterious effects of hyperglycemia on the proximal tubule. Recently, we have demonstrated that diabetes-induced upregulation in kidney GLUT2 expression and dynamic translocation is mitigated by peripheral pharmacological blockade or genetic deletion of CB 1 R in RPTCs, and that it reduces glucose reabsorption and prevents the development of DKD 25 . However, there are several fundamental open questions that should be addressed: What are the cellular mechanisms by which hyperglycemia activates CB 1 R? How do CB 1 Rs specifically regulate the transcriptional levels of GLUT2? And what is the explicit role of GLUT2 in RPTCs in the development of DKD? Addressing these important questions will eventually advance our understanding of the interaction of CB 1 R and GLUT2 in the development of DKD. Based on the key roles of CB 1 R and mTORC1 in whole-body energy utilization, and their similar involvement in the pathogenesis of DKD, we aimed to determine whether these two pathways interact in the proximal tubules and affect kidney homeostasis in health and disease. Here, we show a profound inactivation of mTORC1 signaling and, consequently, downregulation of GLUT2 in diabetic mice lacking CB 1 R specifically in RPTCs; these molecular events were associated with preserved kidney function in the diabetic mice. Furthermore, we show that hyperglycemia specifically enhances the endocannabinoid (eCB)/CB 1 R 'tone' in RPTCs, promoting a cascade of molecular events that activate mTORC1, which in turn, triggers transcriptional changes in the GLUT2 promoter, ultimately enhancing its expression. On the other hand, non-diabetic mice lacking CB 1 R in RPTCs display enhanced mTORC1 activation as a result of elevated amino acid transport, causing morphological abnormalities in the kidney and renal dysfunction. These observations reveal the critical role of CB 1 R in maintaining normal mTORC1 activity in RPTCs both under pathological and physiological conditions. Finally, we demonstrate that a specific reduction in GLUT2 expression in RPTCs is sufficient to protect diabetic mice from developing DKD, emphasizing this transporter's role in regulating renal glucose homeostasis and the pathogenesis of DKD. Overall, our data describe for the first time an opposing CB 1 R/mTORC1 signaling axis in healthy and diseased kidney.
( Fig. 2b, e). Importantly, GLUT2 expression was decreased in the Akita-RPTC-CB 1 R −/− mice (Fig. 2c, f), concomitant with a significant increase in glycosuria (Fig. 2g). In fact, significant positive correlations between pS6 or pAKT and GLUT2 expression were found (Fig. 2h, i). Conversely, inducing acute CB 1 R activation by administering WIN-55,212 to WT mice increased kidney mTORC1 activity and elevated GLUT2 expression levels ( Supplementary  Fig. 2a-c). These results are in agreement with our previous findings, demonstrating that peripheral pharmacological blockade of CB 1 R ameliorates DKD in Akita and in streptozotocin (STZ)induced diabetic mice via downregulating GLUT2 expression and its dynamic translocation 25 .
Next, we aimed to determine whether acute exposure of human primary RPTCs to hyperglycemic conditions induces an elevation in GLUT2 expression in a CB 1 R-dependent manner. Indeed, GLUT2 protein levels were upregulated by the high glucose (HG) levels and normalized with CB 1 R antagonism ( Fig. 2n-q). The HG-induced increase in GLUT2 expression was associated with enhanced 2-arachidonoylglycerol (2-AG) synthesis, manifested by reduced levels of the precursor 1-stearoyl-2-arachidonoyl-snglycerol (SAG) and elevated 2-AG and arachidonic acid (AA) levels ( Fig. 2r-t and Supplementary Fig. 4a-f). However, no changes were found in the expression of the two isoforms of diacylglycerol lipase (DAGL), the rate-limiting enzyme in the biosynthesis of 2-AG ( Supplementary Fig. 5a-e). In fact, inhibiting its activity by DO34 in HG-treated hRPTCs normalized the elevated GLUT2 protein levels ( Fig. 2u, v), suggesting a vicious cycle by which hyperglycemia induces CB 1 R reactivation by 2-AG production and subsequently increases GLUT2 expression through activation of the CB 1 R/mTORC1 signaling pathway.
CB 1 R modulates mTORC1 via a Gi-PI3K-dependent signaling pathway. As a G protein-coupled receptor (GPCR), CB 1 R may affect mTORC1 activity and consequently, GLUT2 via several cellular signaling pathways. In fact, CB 1 R-induced phosphorylation of AKT is known to be regulated upstream via the activation (phosphorylation) of PIP2 by phosphoinositide 3-kinases (PI3K) induced through the βγ subunit of GPCR [59][60][61] . Investigating the effect of CB 1 R activation either by the synthetic cannabinoid arachidonyl-2′-chloroethylamide (ACEA) or by hyperglycemia in hRPTCs revealed that both conditions increased pAKT and pS6 in a PI3K-dependent manner, since their effects were completely prevented by the PI3K inhibitor wortmannin and partially by the CB 1 R inverse agonist JD5037 (Fig. 3a-f). CB 1 R G-protein coupling may be biased according to the means of its activation; it stabilizes its conformation either towards Gi-or Gq-coupled signaling 62,63 . Therefore, we next examined which of the two possible molecular events is triggered by CB 1 R agonism and hyperglycemia. We found that pretreatment of RPTCs with the Gi inhibitor, pertussis toxin (PTX), but not with the Gq inhibitor, YM254890, abolished the effect of CB 1 R activation or HG on mTORC1 activation (Fig. 3g-l). Collectively, these results suggest that CB 1 R activates mTORC1 and increases GLUT2 expression in Gi-and PI3K-dependent manner.
CB 1 R interacts with mTORC1 to maintain cellular homeostasis under normoglycemic conditions. To further assess the link between CB 1 R and mTORC1 in RPTCs, we followed their association under normoglycemic conditions, and surprisingly, we noted the robust activation of mTORC1 in non-diabetic WT-RPTC-CB 1 R −/− mice, as manifested by the elevated pS6 and pAKT protein levels ( Fig. 5a-d). This effect was accompanied by a significant elevation in the SREBP1 and GLUT2 protein expression levels ( Fig. 5e-g), indicating a strong link between CB 1 R and mTORC1 and their impact on GLUT2 transcription also during healthy conditions. Whereas reduced mTORC1 activity in normoglycemic WT-RPTC-RPTOR −/− resulted in reduced SREBP1 expression, it did not yield a significant effect on the GLUT2 levels ( Fig. 5h-k).
These surprising findings of an opposite regulation of mTORC1 by CB 1 R in normal conditions led us to investigate the nutritional mediators (other than glucose) that may explain how CB 1 R affects mTORC1 signaling. As it is already known, mTORC1 can be activated via signals coming from branched-chain amino acids (BCAAs) and protein degradation products 67 . Indeed, CB 1 R deletion in RPTCs resulted in an upregulation in the kidney mRNA and protein levels of megalin (LRP2) (Fig. 5l-o), a cellular membrane transporter responsible for the reuptake of lipoproteins, amino acids, vitamin-binding proteins, and hormones, as well as to a significant increase in the Na + -dependent neutral amino acid transporter B(0)AT1 (SLC6A19) (Fig. 5p, r, s) along with a marked upregulation in the expression of the large neutral amino acid transporter LAT1 (SLC7A5) (Fig. 5p, r, s). No significant changes were detected in the BCAA degrading enzymes (Fig. 5q). Interestingly, these changes were accompanied by elevated levels of kidney amino acids in WT-RPTC-CB 1 R −/− mice ( Fig. 5t and Supplementary Table 7), which also persisted in the diabetic Akita-RPTC-CB 1 R −/− mice (Supplementary Fig. 8 and Supplementary Table 7), emphasizing the novel role of CB 1 R in modulating amino acid uptake. SLC6A19 amino acids transporter is Na + dependent, and indeed we found decreased urinary levels of Na + (Fig. 5u), implicating a higher Na + utilization in the absence of CB 1 R in RPTCs. Similarly, in hRPTCs under normoglycemic conditions, the CB 1 R antagonist JD5037 upregulated the protein levels of SLC6A19 and SLC7A5 ( Supplementary Fig. 9a-c), further enhancing amino acid bioavailability to activate mTORC1 ( Supplementary Fig. 9d, e). Moreover, CB 1 R antagonism enhanced BCAAs uptake in hRPTCs (Supplementary Fig. 9f-i). The overactivation of kidney mTORC1 in WT-RPTC-CB 1 R −/− mice was accompanied by increased albuminuria and kidney injury ( Fig. 6a-e), evident by abnormal morphological alterations, such as enlarged glomerular and Bowman's space cross sectional areas ( Fig. 6f-i). Taken together, these findings indicate that under normal conditions RPTC CB 1 R could restrain mTORC1 activity via regulating nutrient absorption to maintain normal kidney morphology and function.
Genetic reduction of GLUT2 in RPTCs protects mice from developing DKD. Following our finding that the CB 1 R/ mTORC1/SREBP1c signaling pathway regulates GLUT2 transcriptional expression in RPTCs, we next assessed whether a reduction in GLUT2 expression is sufficient to prevent DKD. To that end, we utilized the same breeding paradigm described earlier to generate diabetic mice lacking GLUT2 in RPTCs (Akita-RPTC-GLUT2 −/− ) and their littermate control mice (Akita-RPTC-GLUT2 +/+ ) (Supplementary Fig. 10a-e). Whereas genetic reduction of GLUT2 specifically in RPTCs did not affect the susceptibility of the mice to develop insulin deficient diabetes ( Fig. 7a-e), the null mice exhibited increased glycosuria (Fig. 7f) and preserved kidney function, manifested by reduced kidney-tobody weight ratio, urine excretion-to-water consumption ratio, as well as reduced proteinuria, albuminuria, urine creatinine, ACR, and urinary KIM-1 (Fig. 7g-m), without affecting BUN or CCr (Fig. 7n, o). Moreover, kidney morphological anomalies were prevented, including glomerular and Bowman's space areas hypertrophy and mesangial expansion (Fig. 7p-s). The expression of kidney injury, fibrosis, and inflammation markers was reduced in Akita-RPTC-GLUT2 −/− mice (Fig. 7t-v), suggesting that GLUT2 plays an obligatory role in the development of DKD. To further determine the role of RPTC-GLUT2 in glucose reabsorption by the kidney under diabetic conditions, we utilized an in vivo PET-MRI assay in which the GLUT high-affinity substrate, 2-deoxy-2-[ 18 F]-fluoro-D-glucose (FDG), was injected into the tail vein and its uptake by different organs was analyzed. Kidney [ 18 F]-FDG uptake was significantly increased in the diabetic mice (Fig. 7w-x), consequently increasing its concentration in the circulation and uptake in peripheral organs (e.g., the liver and muscle) while reducing its accumulation in the bladder (Supplementary Fig. 11a-h). Notably, these alterations were completely reversed in Akita-RPTC-GLUT2 −/− mice (Fig. 7w-x and Supplementary Fig. 11a-h), indicating the key role played by RPTC-GLUT2 in kidney glucose reabsorption. Interestingly, we found that deletion of GLUT2 in RPTCs significantly R modulates mTORC1 via a Gi-PI3K-dependent signaling pathway. a-c Representative immunoblots for pAKT and pS6 in primary hRPTCs treated with or without ACEA (10 μM), ACEA + JD5037 (100 nM) or ACEA + Wortmannin (500 nM) for 1 h. For b, n = 3 for Control group, n = 4 for ACEA and ACEA + JD5037 groups, n = 5 for ACEA + Wort group. For c, n = 3 for Control, ACEA and ACEA + JD5037 groups, n = 5 for ACEA + Wort group (*P < 0.0116, # P < 0.0141). d-f Representative immunoblots for pAKT and pS6 in primary hRPTCs treated with or without HG (30 mM), HG + JD5037 (100 nM) or HG + Wortmannin (500 nM) for 1 h. For e, n = 15 for Control group, n = 16 for HG group and n = 12 for HG + JD5037 and HG + Wort groups. For f, n = 15 for Control group, n = 16 for HG group and n = 12 for HG + JD5037 and HG + Wort groups (*P < 0.0307, # P < 0.0003). g-i Representative immunoblots for pAKT and pS6 in primary hRPTCs treated with or without ACEA (10 μM), ACEA + JD5037 (100 nM), ACEA + Pertussis toxin (100 ng/mL) or ACEA + YM254890 (1 μM) for 1 h. For h, n = 4 for Control, ACEA, ACEA + JD and ACEA + YM groups, n = 5 for ACEA + PTX group. For i, n = 8 per group (*P < 0.0009, # P < 0.0025). j-l Representative immunoblots for pAKT and pS6 in primary hRPTCs treated with or without HG (30 mM), HG + JD5037 (100 nM), HG + Pertussis toxin (100 ng/mL) or HG + YM254890 (1 μM) for 1 h. For k, n = 9 for Control, HG, HG + JD and HG + YM groups, n = 8 for ACEA + PTX group. For l, n = 8 per group (*P < 0.0422, # P < 0.0020). Data represent the mean ± SEM and were analyzed by One-way ANOVA followed by Tukey test (one-sided). *P < 0.05 relative to the corresponding control group. # P < 0.05 relative to the ACEA-or HG-treated group. Source data are provided as a Source Data file.  Fig. 12a-d).
These findings may suggest that GLUT2 has a central role in modulating kidney glucose reabsorption and homeostasis by governing glucose transporters in the proximal tubules.

Discussion
The present study reveals, for the first time, the existence of a CB 1 R/mTORC1 signaling axis in RPTCs and its importance in regulating kidney function in health and disease. Whereas CB 1 R may restrain potentially deleterious mTORC1 over-activation during normoglycemia by preventing amino acid flux, in diabetes, enhanced activity of CB 1 R by eCBs, stimulates mTORC1 to further increase glucose uptake via upregulating SREBP1-mediated GLUT2 expression, thus contributing to the development of DKD (Fig. 8). These novel findings have developmental, functional, and translational implications.
Under normoglycemic physiological condition, kidney CB 1 R regulates hemodynamic by inducing afferent arterioles vasodilation and reducing GFR 68,69 . CB 1 R has been suggested to regulate Na + /K + -ATPase activity in RPTCs 63,70 , which is required for maintaining the electrochemical sodium gradient across the brush border membrane of the RPTCs, and it is therefore essential for the reabsorption process that occurs in the kidney. Similar to CB 1 R, mTORC1 also plays an essential physiological role in regulating nutrient transport in RPTCs 71 . Mice lacking mTORC1 in RPTCs 71 or treated with rapamycin develop glycosuria, phosphaturia, aminoaciduria, low-molecular weight proteinuria, and albuminuria 72 , findings observed in individuals suffering from the Fanconi-Bickel syndrome, caused by mutations in the GLUT2 gene, SLC2A2 73,74 . Interestingly, we found that deletion of CB 1 R from RPTCs in nondiabetic animals stimulated mTORC1 signaling resulting in kidney dysfunction. This effect was associated with increased amino acid content and bioavailability in RPTC-CB 1 R −/− mice and hRPTCs treated with a CB 1 R antagonist. Increased BCAA content in the kidney of these animals and cells is most likely mediated via megalin and the neutral amino acid transporter SLC6A19 as well as the large neutral amino acid transporter, SLC7A5. CB 1 R has been previously shown to regulate kidney megalin expression in dietinduced obesity 75,76 . SLC6A19 was also suggested as a potential target for the treatment of metabolic disorders, since its absence in mice was associated with decreased mTORC1 activity 77,78 . In fact, the absence of CB 1 R may alter Na + /K + -ATPase 63,70 , required for SLC6A19 activity. SLC7A5, known to be expressed in proximal tubular cells 79 , was shown to activate mTORC1 via transporting Leucine into the lysosome 80 . We show here that these transporters are regulated by CB 1 R. Although Cystine levels were highly upregulated in mice lacking CB 1 R, no changes were detected in the levels of the Cystine transporter SLC3A1. Grahammer and colleagues showed that genetic depletion of mTORC1 in the proximal tubules decreased the expression and/or phosphorylation of multiple amino acid transporters, leading to aminoaciduria 71 , emphasizing a direct role for mTORC1 in amino acid transport. We have previously reported that genetic manipulations of mTORC1 in RPTCs, both constant activation, and inhibition by TSC1 KO and RPTOR KO respectively, modulate SLC6A19 expression 58 . Therefore, we cannot exclude the possibility that prolonged mTORC1 activation in RPTC-CB 1 R −/− mice per se lead to enhanced amino acid transport. Nevertheless, megalin and cubilin expression was unchanged in mTORC1-depleted tubules 71 ; therefore, the specific role of CB 1 R in modulating mTORC1 activity via megalin/cubilin mediated amino acids transport cannot be negated. Moreover, we previously reported that genetic deletion of CB 1 R in RPTCs increases kidney weight and glomerular area 21 , implying that CB 1 R may play a developmental role in the kidney. The present study offers an explanation for this phenomenon through a hitherto unrecognized CB 1 R-mediated regulation of mTORC1, known to control cell metabolism, growth, and proliferation. Thus, our current findings raise a concern for using CB 1 R antagonists in normoglycemic patients because they may enhance mTORC1 activity and induce deleterious effects on the kidney.
Whereas under normoglycemic conditions CB 1 R controls mTORC1 activation in RPTCs via restraining amino acids flux, our findings demonstrate that glucotoxicity enhances CB 1 R-induced mTORC1 activity in the same type of cells. In accordance with previous findings with CB 1 R blockers 17,19,25 , specific deletion of RPTC-CB 1 R in Akita diabetic mice, although not affecting the diabetic phenotype of the animals, had a profound beneficial effect on the kidney, which can be attributed to the inhibition of mTORC1 signaling. These findings were associated with reduced GLUT2 expression and glucosuria, effects that were previously described by us as being modulated by kidney CB 1 R 25 . In agreement with our findings, it was recently demonstrated that pancreatic GLUT2 levels are reduced in CB 1 R/CB 2 R-deficient mice 81 . Mechanistically, our findings show here that CB 1 R regulates mTORC1 in RPTCs via the PI3K-Akt pathway under diabetic Fig. 4 GLUT2 transcription is directly regulated by CB 1 R and mTORC1. a-e Luminescence relative response ratio (RRR) analyses of HEK293 cells cotransfected transiently with pGL3-GLUT and Renilla luciferase plasmids (T2) or with pGL3-GLUT, Renilla luciferase, and CB 1 R-TK-d64 plasmids (T3), treated with or without ACEA (10 μM; b, c) or with HG (30 mM; a, d, e) for 3 h, in the presence or absence of JD5037 (100 nM) or rapamycin (100 nM). For a, n = 6 for T2, n = 8 for T2+HG (*P = 0.0034). For b, n = 16 per group (*P < 0.0001). For c, n = 19 for T3 group, n = 16 for T3 + ACEA group and n = 20 for T3 + ACEA + JD5037 and T3 + JD5037 groups (*P < 0.0001, # P = 0.0002). For d, n = 23 for T3 and T3 + HG + JD5037 groups, n = 19 for T3 + HG group and n = 14 for T3 + JD5037 group (*P = 0.0003, # P = 0.0002). For e, n = 18 for T3 group, n = 16 for T3 + HG group, n = 8 for T3 + HG + RAP group and n = 14 for T3 + RAP group (*P < 0.0001, # P = 0.0207). f Promoter-binding transcription-factor (TF) profiling assay of the eight most prominent candidate TFs present in nuclear extracts (NEs) collected from primary hRPTCs treated with or without HG (30 mM) for 3 h in the presence or absence of hGLUT2 promoter (P1). g Transcription-factor profiling assay for eight candidate TFs present at the NEs of primary hRPTCs treated with or without HG (30 mM) in the presence or absence of JD5037 (100 nM) or rapamycin (100 nM) for 3 h. n = 4 per group (*P < 0.0019, # P < 0.0001). h qPCR analyses of kidney Glut2 candidate TFs in kidney lysate from Akita-RPTC CB1R+/+ and Akita-RPTC CB1R−/− diabetic mice. n = 13 mice per group (*P < 0.0018). i, j Immunoblotting analysis and quantification of SREBP1 in kidney lysate from Akita-RPTC CB1R+/+ and Akita-RPTC CB1R−/− diabetic mice. n = 4 mice for Akita-RPTC CB1R+/+ , n = 5 mice for Akita-RPTC CB1R−/− (*P = 0.0264). k qPCR analysis and quantification of kidney Glut2 candidate TFs in kidney lysate from Akita-RPTC RPTOR+/+ and Akita-RPTC RPTOR−/− diabetic mice. n = 7 mice per group (*P < 0.0324). l, m Immunoblotting analysis and quantification of SREBP1 in kidney lysate from Akita-RPTC RPTOR+/+ and Akita-RPTC RPTOR−/− diabetic mice. n = 5 mice per group (*P = 0.0191). n-q Immunoblotting analysis of SREBP1c and GLUT2 in primary hRPTCs treated with or without HG (30 mM) in the presence or absence of SREBP1 siRNA for 24 h. n = 3 biological replicates per group (*P < 0.0385, # P < 0.0051). r qPCR analysis of SREBP1c in primary hRPTCs treated with or without HG (30 mM) in the presence or absence of JD5037 (100 nM) for 1 h. n = 8 for Control and HG groups, n = 12 for HG + JD group (*P = 0.0031, # P < 0.0001). s, t Representative immunoblotting analysis and quantification of nuclear SREBP1 in extracts collected from primary hRPTCs treated with or without HG (30 mM) in the presence or absence of JD5037 (100 nM) or rapamycin (100 nM) for 3 h. SREBP1 expression was normalized to the nuclear marker Fibrillarin. n = 4 per group (*P = 0.0196, # P < 0.0497). u, v Representative immunofluorescence and quantification of cytoplasmic SREBP1 expression in primary hRPTCs treated with or without HG (30 mM) in the presence or absence of JD5037 (100 nM) or rapamycin (100 nM) for 3 h. 100× magnification, scale bar: 10 µm. n = 14 for Control group, n = 19 for HG group, n = 15 for HG + JD5037 group and n = 11 for HG + RAP group (*P < 0.0001, # P < 0.0012). Data represent the mean ± SEM and were analyzed by unpaired two-tailed Student's t-test or one-way ANOVA followed by Tukey test (one-sided). *P < 0.05 relative to the corresponding control group. # P < 0.05 relative to the ACEA-or HG-treated group. Source data are provided as a Source Data file.
conditions. Dalton and colleagues suggested that the βγ subunits of the CB 1 R-coupled G i/o protein in neuronal cells stimulate PI3K to induce ERK phosphorylation 59 . On the contrary, we found that stimulation of CB 1 R by glucose did not affect (or even reduced) the phosphorylation of ERK in RPTCs and that mTORC1 activation is mediated via a G i/o protein coupled CB 1 R pathway. We and others 82,83 have previously shown that CB 1 R may also signal via G q signaling, which enhances cellular calcium influx, contributing to increased PKC-β1 expression, resulting in dynamic translocation of GLUT2 into the brush border membrane of the RPTCs 25 . Such a duality in the cellular signaling cascades modulated by CB 1 R can exist in RPTCs 63 , allowing CB 1 R to modulate both the expression and the translocation of GLUT2. Alternatively, CB 1 R activation can stimulate phospholipase C (PLC) via the G i/o βγ subunit, thereby increasing intracellular [Ca 2+ ] influx and PKC activation 61 . Consistently, we found (i) significant positive correlation between pS6 and pAKT and GLUT2, (ii) modulation of GLUT2 expression in mice lacking RPTOR or TSC1 specifically in RPTCs, and (iii) increased transcriptional expression of GLUT2 by HG and/or CB 1 R activation, effects that were completely prevented by CB 1 R or mTORC1 inhibition.
As per the transcriptional regulation of GLUT2 in RPTCs 65,66 , we found 24 TFs that were enriched in the nuclear compartment of hRPTCs exposed to HG and that selectively bind to the GLUT2 promoter. The most prominent one, SREBP1c, has been previously shown to promote kidney lipotoxicity [84][85][86][87] . It has been also reported that in RPTCs, hepatocytes, and Schwann cells, HG stimulates SREBP1c via the PI3K/AKT/mTORC1 pathway 42,87-90 . Moreover, SREBP1c enhances GLUT2 expression in hepatocytes exposed to hyperglycemia 91 . In line with the recent reports, we also found that specific genetic deletion of CB 1 R, RPTOR, or TSC1 in the RPTCs modulated SREBP1c levels in parallel to GLUT2 expression. Furthermore, treatment of hRPTCs with rapamycin stimulated GLUT2 transcriptional activity. Collectively, these findings demonstrate the importance of SREPB1c in modulating the CB 1 R/mTORC1 effect on GLUT2.
Here, we also demonstrate a direct effect of glucose on the production and/or degradation of 2-AG, which in turn, may activate its own receptor to further promote glucose transport resulting in RPTC's dysfunction. These findings are in line with previous reports showing increased eCB/CB 1 R 'tone' in different compartments/cells within the kidney as well as in pancreatic islets under gluco-and/or lipo-toxic conditions 17,20,21,92 . As previously reported by Sampaio and colleagues, the two isoforms of DAGL are differentially expressed in the renal proximal tubule 63 . While we did not find any changes in their expression in response to hyperglycemia, the functional inhibition of DAGL R regulates mTORC1 activation under normoglycemic conditions. a, c Immunoblotting analysis and quantification of pS6 in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 5 mice per group (*P = 0.0166). b, d Immunoblotting analysis and quantification of pAKT in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 4 mice per group (*P = 0.0275). e-g Immunoblotting analysis and quantification of SREBP1c and GLUT2 in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 6 mice per group (*P < 0.0264). h-k Immunoblotting analysis and quantification of pS6, SREBP1c, and GLUT2 in kidney lysates from WT-RPTC RPTOR+/+ and WT-RPTC RPTOR−/− normoglycemic mice. n = 6 mice per group (*P < 0.0136). l qPCR analyses of megalin (Lrp2) in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 9 mice for WT-RPTC CB1R+/+ , n = 8 mice for WT-RPTC CB1R−/− (*P = 0.0395). m Representative LRP2 immunoblotting and immunofluorescence staining of kidney sections from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. 10× magnification, scale bar 100 µm. n = 5 mice per group. n Immunoblotting quantification of LRP2 in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 5 mice per group (*P = 0.0007). o Immunofluorescence quantification for LRP2 in kidney sections from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. 10× magnification, scale bar 100 µm. n = 5 mice per group (*P = 0.0164). p qPCR analysis of amino acid transporters in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 8 mice per group (*P = 0.0007). q qPCR analysis of BCAA degrading enzymes in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 8 mice per group. r, s Immunoblotting analysis and quantification of SLC6A19 and SLC7A5 in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 5 mice per group (*P < 0.0017). t LC-MS/MS quantification of amino acids in kidney lysates from WT-RPTC CB1R+/+ and WT-RPTC CB1R−/− normoglycemic mice. n = 5 mice per group (*P < 0.0354). u Urinary Na + levels normalized to the creatinine concentration. n = 11 mice for WT-RPTC CB1R+/+ , n = 17 mice for WT-RPTC CB1R−/− (*P = 0.0031). Data represent the mean ± SEM and were analyzed by unpaired two-tailed Student's t-test. *P < 0.05 relative to the corresponding control groups. Source data are provided as a Source Data file. Global GLUT2 −/− mice display severe glycosuria 93 , and the complete absence of GLUT2 at the basolateral membrane of the RPTCs supposedly leads to enhanced intracellular glucose levels and proximal tubular damage, even in the absence of diabetes. This effect has been well-documented in Fanconi-Bickel Syndrome patients who exhibit tubular dysfunction 94 . However, our current study revealed additional surprising phenomena. Specific reduction, but not full ablation of GLUT2 in RPTCs ameliorates Proximal tubular GLUT2 is known to be upregulated and translocated to the brush border membrane of the RPTCs where it facilitates glucose reabsorption during diabetes 55,95 , and may increase glucose-induced tubular damage and tubule-interstitial fibrosis. The contribution of GLUT2 recruitment to the brush border membrane of the RPTCs in facilitating increased glucose reabsorption in diabetes is not clear, since kidney glucose reabsorption is also mediated via SGLT2 and SGLT1 96 . Possibly, in diabetes, luminal glucose concentrations exceed a point by which SGLT's glucose uptake is in its full capacity, GLUT2 recruitment to the apical membrane enables glucose "leakiness" back to the lumen, facilitating apical glucose recycling to promote sodium reabsorption through SGLTs 6 . Moreover, rising blood glucose levels hamper the tubule-to-blood glucose gradient, which is essential for glucose flux through basolateral GLUT2 97 .  Fig. 8 A proposed mechanism by which renal proximal tubule CB 1 R regulates mTORC1 activity in health and disease. In normoglycemia CB 1 R is responsible to maintain mTORC1 normal activation by preventing excess of amino acids uptake (left). In hyperglycemia, CB 1 R-mediated mTORC1 overactivation results in upregulating GLUT2 transcription, consequently enhancing glucose absorption and tubular damage (right). Therefore, to maintain 'normal' tubular glucose levels, GLUT2 is recruited to the apical brush border membrane to assist in glucose transfer into the cell. Marks and colleagues found that STZinduced diabetes in rats increases facilitative glucose transport at the brush border membrane by 67.5%, mainly due to GLUT2 translocation 55 . Our novel in vivo PET-MRI data emphasize the key role of GLUT2 in glucose reabsorption in diabetes, since its ablation in RPTCs normalized the diabetes-induced glucose uptake to the kidney. In addition, Umino et al. suggested that activation of the GLUT2/importin-α1/HNF-1α pathway by basolateral HG enhances SGLT2 expression in RPTCs; this was preventable by GLUT2 silencing/inhibition 98 . These findings are also supported by our data, demonstrating that genetic deletion of GLUT2 in RPTCs decreased SGLT2 expression, regardless the status of glycemia, suggesting that tight crosstalk between these two glucose transporters regulates kidney glucose handling and function.
Our results may also have translational aspects when one considers the therapeutic potential of pharmacologically targeting GLUT2 for the treatment of DKD. However, GLUT2 inhibition is unlikely to discriminate between apical and basal GLUT2 and might have major systemic side-effects as GLUT2 plays a role in glucose transport in other tissues, including the gut, islets, the liver, and neurons. Moreover, kidney glucose toxicity may remain an issue if glucose is taken up into RPTCs via SGLT2 but cannot exit to the interstitium and the blood at the basolateral side. Therefore, from this perspective, targeting the CB 1 R and its downstream signaling molecules identified here may be a more fruitful therapeutic strategy. An alternative approach would be to target the recruitment machinery of GLUT2 to the brush border membrane to restrain the development of DKD rather than targeting the transporter itself. Further work is required to test which option might become a vital therapeutic approach for DKD.

Methods
Animals and the experimental protocol. The Institutional Animal Care and Use Committee of the Hebrew University (AAALAC accreditation #1285; Ethic approval number MD-19-15784) approved the experimental protocol used. Animal studies are reported, in compliance with the ARRIVE guidelines 99 . The current experiment is based on the rule of the replacement, refinement, or reduction. All the animals used in this study were males under C57BL/6J background that were housed under specific pathogen-free (SPF) conditions, up to five per cage, in standard plastic cages with natural soft sawdust as bedding. The animals were maintained under controlled temperature of 22-24°C, humidity at 55 ± 5%, and alternating 12 h light/dark cycles (lights were on between 7:00 and 19:00 h), and provided with chow food (Cat#, NIH-31 rodent diet) and water ad libitum. To elucidate the role of the proximal tubular CB 1 R in the pathogenesis of DKD, we generated a novel diabetic mouse strain that lacks CB 1 R specifically in the RPTCs (AK-RPTC-CB 1 R −/− ) by crossing the Akita Ins2+/ C96Y mouse model for type I diabetes with RPTC-CB 1 R fl/fl;sglt2Cre [RPTC-CB 1 R −/− ] characterized previously 21,25 . To generate diabetic mice lacking GLUT2 in RPTCs, we first crossed mice containing two loxP sites flanking the open reading frame of the GLUT2 gene (GLUT2 fl/fl ; described in 100 ) with the iL1-sglt2-Cre line 101  At three weeks of age, littermates were divided into four groups according to their genotypes: WT-Cre − , WT-Cre + , AK-Cre − , and AK-Cre + . Cre − or Cre + refers to the presence or deletion of the gene of interest (CB 1 R, GLUT2, Raptor, and TSC), respectively. All animals were homozygous for flox, except for TSC. The mice were monitored weekly for their blood glucose levels and body weight until they were sixteen weeks of age. Mice were subjected to a complete metabolic and kidney analysis. 24 h urine was collected one week before euthanasia using mouse metabolic cages (CCS2000 Chiller System, Hatteras Instruments, NC, USA). At week 16, the mice were euthanized by a cervical dislocation under anesthesia, the kidneys and pancreases were removed and weighed, and samples were either snapfrozen or fixed in buffered 4% formalin. Trunk blood was collected for determining the biochemical parameters. CB 1 R activation in WT mice was done in 8-week-old, C57Bl/6J male mice. WIN 55,212-2 (3 mg/kg), or a vehicle (Veh; 1% Tween80, 4% DMSO in saline) was injected ip, and after 6 h the mice were euthanized by a cervical dislocation under anesthesia, the kidneys were removed and were either snap-frozen or fixed in buffered 4% formalin.
Blood and urine biochemistry. Serum insulin and urine albumin were measured by ELISAs (Millipore and Bethyl Laboratories, respectively). Serum urea, urine/ serum glucose, and creatinine were determined using a Cobas C-111 chemistry analyzer (Roche, Switzerland). Blood urea nitrogen (BUN) was calculated by the serum urea levels (BUN mg/dL = Urea mM × 2.801). CCr was calculated using urine and serum creatinine levels (CCr mL h −1 = Urine creatinine mg/dL × Urine volume/Serum creatinine mg/dL × 24 h). The urine levels of KIM-1 were measured by ELISA kit (R&D Systems, MN, USA).
Histopathological Analyses. Paraffin-embedded kidney and pancreas sections (5 µm) from each group (5-6 animals per group) were stained with hematoxylin and eosin or Periodic Acid-Schiff (PAS), followed by hematoxylin staining. In addition, kidney sections were stained for collagen types I and III deposition by using Masson's trichrome (Abcam, ab150686) staining according to the manufacturer's procedures. Kidney and pancreas images were captured with a Zeiss AxioCam ICc5 color camera mounted on a Zeiss Axio Scope.A1 light microscope, and taken from 10 random 40× fields of each animal. Mesangial expansion, glomerular, and Bowman's space cross-sectional areas were quantified using ZEN imaging software (Zeiss, Germany). Masson's trichrome positive areas were quantified using Image J software with a minimum of 10 random kidney images per mouse.
Vectors. The hGLUT2 promoter 1043 bp site with the XohI/HindIII restriction sites were amplified via PCR from hRPTC DNA and cloned into the multiple cloning region of the pGL3 luciferase reporter vector (Promega, USA), generating the pGL3-GLUT2 vector. The Renilla luciferase vector (Promega, USA) was used as a co-reporter vector and for expression control. The CB 1 R-TK-d64 plasmid contained a short version of the CB 1 R gene under the thymidine kinase (TK) promoter, enabling CB 1 R to translocate to the plasma membrane more easily than the full-length CB 1 R 64 (Supplementary Fig. 6).
GLUT2 transcription luciferase reporter assay. HEK-293 cells were transiently transfected using the Lipofectamine™ 3000 transfection reagent (Thermo Fisher Scientific, USA) with the pGL3-GLUT2 and Renilla vectors with/without the CB 1 R-TK-d64 plasmid 64 . One day following the transfection, cells were starved overnight in SFM and treated for 3 h with HG (30 mM) or CB 1 R agonist-ACEA (10 μM), with/without 1 h pretreatment with JD5037 (100 nM) or Rapamycin (100 nM). The Dual-Glo Luciferase Assay System (Promega, USA) was used to detect GLUT2 promoter transcriptional activity. The Relative Response Ratio (RRR) was calculated for each treatment and compared to the control group.
Promoter-binding transcription-factor profiling assay. A list of~120 TFs that can bind to the1043 bp hGLUT2 promoter site was prepared using TFBIND software. A modified list of 47 relevant TFs was screened for their ability to bind to the hGLUT2 promoter using a promoter-binding TF profiling array (Signosis, USA). An assay was performed according to the manufacturer's instructions. Briefly, nuclear extract (NE) was isolated from 1 × 10 7 hRPTCs, and treated/ untreated with HG for 3 h, by using a nuclear protein extraction kit (Signosis, USA). The reaction mixture was prepared using 15 μL of the TF binding buffer, 3 μL of the probe, 10 μg of nuclear extract, and 5 μL of the hRPTC GLUT2 promoter fragment (1043 bp), and incubated at room temperature for 30 min to allow for the formation of the TF-DNA complex. Unbound probes were separated from the complex, whereas bound probes were eluted and then hybridized to the plate and incubated overnight at 42°C. Bound probes were detected using an HRPstreptavidin conjugate incubated with the chemiluminescent substrate. Luminescence is reported as relative light units (RLUs) on the Multi-Mode Microplate Reader SpectraMax iD3 (Molecular Devices, USA).
Small interfering RNA treatment. Small interfering RNA (siRNA) transfection against SREBP1 (sc-36557, Santa Cruz) was performed in hRPTCs using the siRNA reagent system (sc-45064, Santa Cruz) according to the manufacturer's instructions.
In vivo micro PET-MRI scanning. Experiments were performed at the Wohl Institute for Translational Medicine at Hadassah Hebrew University Medical Center. PET-MRI images were acquired on a 7T 24 cm bore, a cryogen-free MR scanner based on the proprietary dry magnet technology (MR Solutions, Guildford, UK) with a 3-ring PET insert that uses the latest silicon photomultiplier (SiPM) technology 102 . The PET subsystem contains 24 detector heads arranged in three octagons of 116 mm in diameter. For MRI acquisition, a mouse quadrature RF volume coil was used. Mice were anesthetized with isoflurane vaporized with O 2 . Isoflurane was used at 3.0% for induction and at 1.0-2.0% for maintenance. The mice were positioned on a heated bed, which allowed for continuous anesthesia and breathing rate monitoring. To determine the distribution of [ 18 F]-FDG in mice, the tracer was injected into the tail vein (230 ± 30 mCi in 200 mL). Mice were subjected to 31 min dynamic PET scans; a homemade small catheter was inserted into the proximal tail vein and the tracer was injected after positioning the mouse in the micro-PET/MRI scanner. For dynamic scans, the acquired data were binned into 25 image frames (1 × 60, 6 × 10, 8 × 30, 5 × 60, and 4 × 300 s). During the PET, scans were acquired, T1 & T2; weighted coronal spin echo images were collected for anatomical evaluation. Coronal T1 weighted images were acquired using the following parameters: TR = 1100 ms, TE = 11 ms, echo spacing = 11 ms, FOV = 6 × 3 cm, slice thickness = 1 mm, 4 averages. Coronal T2-weighted images were acquired using the following parameters: TR = 4000 ms, TE = 45 ms, echo spacing = 15 ms, FOV = 6 × 3 cm, slice thickness = 1 mm, and 4 averages. Images were analyzed using VivoQuant pre-clinical image post-processing software (Invicro). PET-MRI raw data were processed using the standard software provided by the manufacturers. PET data were acquired in list-mode, histogrammed by Fourier rebinning, and reconstructed using the 3D-OSEM algorithm, with standard corrections for random coincidences, system response, and physical decay applied. The reconstructed PET images from the PET/MR scanner were quantitated using a measured system-specific 18 F calibration factor to convert reconstructed count rates per voxel to activity concentrations (%ID/g). Manual tissue segmentation of kidneys, liver, muscle, inferior vena cava (IVC), and bladder was carried out on coregistered 3D MR images. The regions of interest were then used to calculate tissue radiotracer uptake from the reconstructed PET images.
Real-time PCR. Total kidney or cell mRNA was extracted using Bio-Tri RNA lysis buffer (Bio-Lab, Israel), followed by DNase I treatment (Thermo Scientific, IL, USA), and reverse transcribed using the Iscript cDNA kit (Bio-Rad, CA). Real-time PCR was performed using iTaq Universal SYBR Green Supermix (Bio-Rad, CA) and the CFX connect ST system (Bio-Rad, CA). The primers used to detect mouse or human genes are listed in Supplementary Tables 1, 2. Mouse and human genes were normalized to Ubc or RPLP, respectively.
Sample preparation and endocannabinoid measurements by LC-MS/MS. The eCBs and related lipid measurements were performed in two independently laboratories. eCBs were extracted, purified, and quantified from RPTC lysates. In brief, A. (Tam Laboratory, Hebrew University) hRPTCs were scraped from the culture plates in ice-cold Tris Buffer, homogenized using sonication, and protein concentration was determined. Samples were then supplemented with an ice-cold Extraction buffer (1:1 Methanol/Tris Buffer+ Internal Standard) and Chloroform/ Methanol (2:1), vortexed and centrifuged. The lower organic phase was transferred into borosilicate tubes; this step was repeated three times by adding ice-cold Chloroform to the samples and transferring the lower organic phase into the same borosilicate tubes. The samples were then dried and kept overnight in −80°C, and then reconstitute with ice-cold Chloroform and Acetone then kept in −20°C for 30 min and centrifuged to precipitate proteins. The supernatant was then dried and reconstituted in ice-cold LC/MS grade Methanol. LC-MS/MS was analyzed on an AB Sciex (Framingham, MA, USA) QTRAP® 6500 + mass spectrometer coupled with a Shimadzu (Kyoto, Japan) UHPLC System. Liquid chromatographic separation was obtained using 5 μL injections of samples onto a Kinetex 2.6 µm C18 (100*2.1 mm) column from Phenomenex (Torrance, CA, USA). The autosampler was set at 4°C and the column was maintained at 40°C during the entire analysis. Gradient elution mobile phases consisted of 0.1% formic acid in water (phase A) and 0.1% formic acid in acetonitrile (phase B). eCBs were detected in a positive ion mode using electron spray ionization (ESI) and the multiple reaction monitoring (MRM) mode of acquisition, using d 4 -AEA as internal standard (IS). The collision energy (CE), declustering potential (DP), and collision cell exit potential (CXP) for the monitored transitions are given in Supplementary Table 5. The levels of AEA, 2-AG, OEA, PEA, and AA in samples were measured against standard curves, and normalized to the RPTC lysate protein concentration.
B. (Gertsch Laboratory, University of Bern) similar to Schuele et al. 103 , the scraped hRPTCs were homogenized in 0.1 M FA by three times freeze-thaw cycles and sonification. The protein concentration was determined with an aliquot of the sample using BCA. The remaining homogenate was extracted in 9:1 ethyl acetate:hexane 0.1% FA solution (3:1 ratio organic to water part). After a drying the organic layer, the sample was reconstituted in 35 µL 80% acetonitrile and 10 µL was injected per measurement into the system. LC-MS/MS was conducted on an AB Sciex (Framingham, MA, USA) Triple Quad™ 5500 mass spectrometer coupled to a Shimadzu (Kyoto, Japan) UHPLC System. For the chromatographic separation, a C18 column (3 μm particle size; 2 × 50 mm, Dr. A. Maisch HPLC GmbH, Ammerbuch-Entringen, Germany) was used. In a positive mode, the mobile phase A was water containing 2 mM NH 4 Ac 0.1% FA, and mobile phase B contained methanol and 2 mM NH 4 Ac. In a negative mode, mobile phase A consisted of water 2 mM NH 4 Ac 0.1% FA, and mobile phase B was acetonitrile 0.1% FA. The flow rate was set to 0.3 mL/min. eCBs were detected in a positive ion mode (for SAG, 2-AG, and ethanolamines) and AA in a negative mode using ESI and the MRM mode. The molecular ions and fragments for each compound are presented in Supplementary Table 5. Data acquisition and analysis were performed using Analyst software 1.6. For all analyses, a linear regression without weighing was applied. The levels of AEA, 2-AG, SAG, OEA, PEA, and AA in samples were