Vertical Sleeve Gastrectomy Lowers SGLT2/Slc5a2 Expression in the Mouse Kidney

Type 2 diabetes is characterized by decreased β-cell mass and function, leading to defective insulin secretion during the progression of insulin resistance (1). Despite advances in recent years (2,3), there is still no permanent cure for type 2 diabetes based purely on pharmacology. Originally conceived as an intervention for weight loss (4,5), bariatric surgery has been shown to be remarkably effective in reversing type 2 diabetes (6). Several randomized clinical trials (7,8) have reported that surgery results in health benefits beyond weight loss, including improved outcomes in cardiovascular disease, steatohepatitis, and cancer. Such findings point toward the existence of multiple interorgan signaling mechanisms emanating from the gastrointestinal tract. Numerous studies (9) have attempted to shed light on these lines of communication and consequently on the mechanisms of disease remission, offering the hope of replacing bariatric surgery with less invasive pharmacological treatments. Possible pathways include a postoperative increase of the incretin glucagon-like peptide 1 (GLP-1), increased bile acid secretion (10), or reduction of glucocorticoid activity (11).

A change in intestinal glucose absorption is one of the mechanisms hypothesized to contribute to type 2 diabetes remission following surgery (12,13). Sodium–glucose transport protein 1 (SGLT1/Slc5a1) provides the principal means of glucose resorption in the intestine (14), with SGLT2/Slc5a2 playing a more minor role in this tissue (15). On the other hand, SGLT2 plays the predominant role in the resorption of the sugar in the kidney (15). However, studies of the effects of bariatric surgery on SGLT1 expression levels have provided contradictory results (14,16).

The gliflozins, a recently released class of antidiabetes drugs, target renal glucose reabsorption by inhibiting SGLT2 (17). These agents have proved to be highly efficacious in patients with type 2 diabetes (18,19), and bariatric surgery (20,21) and SGLT2 inhibitors (22,23) both lead to marked improvements in cardiovascular health.

Bariatric surgery is associated with a significant increase in the risk for urinary tract infections (24). Although these are often associated with operative time, antibiotics, age, urolithiasis, dehydration, and use of a urinary catheter, the high occurrence could also point to glycosuria, conceivably the result of SGLT2 inhibition. Correspondingly, in preliminary findings (25), E.A. reported that SGLT2 expression is downregulated in the kidney cortex of lean rats following duodenal jejunal bypass. Using a combination of surgical, genetic, and pharmacological approaches, the current study tested the hypotheses that 1) SGLT2 expression in the kidney may be altered by a high-fat diet (HFD) or by vertical sleeve gastrectomy (VSG), and 2) that altered SGLT2 activity may contribute to the euglycemic effects of VSG.

All animal procedures undertaken were approved by the British Home Office under the UK Animal (Scientific Procedures) Act 1986 (Project License PPL PA03F7F07 to I.L.), with approval from the Animal Welfare and Ethics Review Board at the Central Biological Services unit at the Hammersmith Campus of Imperial College London. Adult male C57BL/6J mice (Envigo, Huntingdon, U.K.) were maintained under controlled temperature (21–23°C) and light (12:12-h light-dark schedule, lights on at 0700 h). From the age of 8 weeks, C57BL/6J and SGLT2^−/− mice were fed a 58 kcal% fat and sucrose diet (D12331; Research Diets, New Brunswick, NJ) to induce obesity and diabetes. The lean animals were fed PMI Nutrition International Certified Rodent Chow No. 5CR4 (Research Diets) ad libitum. All mice were divided in two groups, VSG and sham. Lean animals were returned on rodent chow following surgery, and obese animals were returned on the D12331 diet ad libitum. Lean animals were euthanized and harvested 4 weeks after surgery, while obese animals were euthanized and harvested at 10 weeks after surgery. Kidneys and gut were harvested from all mice following sham or VSG surgery in the fed state and were snap frozen in −80°C or fixed in formalin, or both.

During the semaglutide study, lean adult male C57BL/6 mice were treated with either a single subcutaneous (SC) injection of semaglutide (Novo Nordisk UK) at 5 nmol/kg (n = 6) or saline (n = 6), for 7 and 30 days. Body weight was measured daily, and all animals were euthanized and harvested on day 7 or day 30.

Two guide (g)RNAs flanking the exon 1 region were designed using the software CRISPOR (http://crispor.tefor.net). gRNAs were synthesized, mixed with clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 protein, and microinjected into 30–50 mouse zygotes at the Medical Research Council transgenic facility in Imperial College London. F0 compound homozygous mice were crossed with wild-type (WT). F1 heterozygous mice were sequenced to determine whether exon 1 is deleted. Heterozygous mice positive for exon 1 deletion were then crossed to generate WT, heterozygous, and homozygous littermates. Genotyping PCR reaction was performed using the primer set listed in Table 1. PCR products were then purified (PCR purification, Qiagen) and sent for Sanger sequencing (Genewiz).

VSG and sham surgery was performed as previously described (26).

Glucose tolerance tests were performed as previously described (26).

Insulin tolerance tests were performed as previously described (26).

Mice were fasted overnight (total 13 h) and given free access to water. At 0800 h, mice received an oral gavage of dapagliflozin (10 mg/kg) dissolved in methyl cellulose and were returned to their cages. At 1100 h, glucose (3 g/kg body wt) was administered via intraperitoneal injection. Blood was sampled from the tail vein at 0, 5, 15, 30, 60, and 90 min after glucose administration. Blood glucose was measured with an automatic glucometer (Accuchek; Roche, Burgess Hill, U.K.).

Plasma was collected as previously described (26). Plasma insulin and GLP-1(1–37) levels were measured by ELISA kits from Crystal Chem (Zaandam, the Netherlands), plasma glucagon was measured by ELISA kit from Mercodia (Uppsala, Sweden), and GLP-1(1–37) and adiponectin levels were measured using an ELISA kit from Abcam (Cambridge, MA).

Urine was collected from mice during glucose tolerance tests by placing them on a cage without bedding. Glucose was measured by glucose assay (Abcam), and creatinine was measured by creatinine assay (Crystal Chem). Urinary glucose-to-creatinine ratio (UGCR) was calculated as (mg/mg) = [urine glucose (mg/dL)/urine creatinine (mg/dL)].

Immunohistochemistry was performed as previously described (26). Permeabilized kidney slices were blotted with anti-rabbit SGLT2 antibody (1:1,000; Cell Signaling Technology, Danvers, MA).

Quantification of the protein levels of SGLT2 in kidney cortex was performed as previously described (27). The following antibodies were used: anti-rabbit SGLT2 (1:200; (Abcam) and anti-mouse GAPDH (1:1,000; Sigma-Aldrich). Intensities were quantified using ImageJ.

Quantification of the expression level of SGLT2 in kidney cortex, adiponectin, and peroxisome proliferator–activated receptor-δ (PPAR-δ) in SC adipose tissue and SGLT1 in the duodenum was performed as previously described (11). Primers, which crossed a splice junction, were designed using Primer Express (Invitrogen, Waltham, MA) (Table 1).

Excised intestinal tissues were taken from age-matched, lean WT, HFD WT sham, HFD WT VSG, lean SGLT2^−/−, and HFD SGLT2^−/− mice. Resultant changes in ion transport short-circuit current (Isc) were measured as described previously (27). SGLT1 and SGLT2 were investigated in mucosae from WT lean mice and compared with SGLT2^−/− lean and SGLT2^−/− HFD mucosae by apical additions of the SGLT inhibitors, phloridzin, which blocks SGLT1 and SGLT 2 (50 µmol/L; Sigma-Aldrich, St. Louis, MO), or canagliflozin, which inhibits SGLT2 preferentially (0 µmol/L; Cayman Chemicals, Ann Arbor, MI). Maximal changes in Isc were measured within 15 min of drug addition. All Isc responses were transformed to µA/cm², pooled (with each observation representing data from a single animal), and expressed as mean ± SD.

HK-2 cells were obtained from ATCC (Manassas, VA) and were cultured with keratinocyte serum-free medium supplemented with 0.05 mg/mL bovine pituitary extract and 5 ng/mL human recombinant epidermal growth factor (Invitrogen). Cells were seeded at a density of ×10⁵ per well and treated for 48 h with glucose, insulin, exendin-4, glucagon, or a combination. Medium was changed every 12 h. A total of 3–10 different replicates were used for each condition.

Data were analyzed using GraphPad Prism 7.0 software. Significance was tested using unpaired Student two-tailed t tests with Bonferroni or Dunnett posttests, where applicable for multiple comparisons, or two-way ANOVA as indicated. P < 0.05 was considered significant, and errors signify ± SD.

All data generated during this study are included in the published article (and its online supplementary files). The SGLT2^−/− mouse line generated during the current study is available from the corresponding authors upon reasonable request.

We first performed VSG in lean mice to confirm that the improvement in glucose tolerance was independent of weight loss. Indeed, despite there being no significant difference in body weight between VSG and sham-operated-on animals 4 weeks postsurgery (Fig. 1A), intraperitoneal glucose tolerance test (IPGTT) (Fig. 1B), glucose-induced changes in circulating insulin (Fig. 1C), and C-peptide (Supplementary Fig. 1) levels were significantly improved by VSG (ratio area under the curve [AUC] insulin-to-glucagon: sham = 0.027, VSG = 0.068).

Surgery was associated with an increase in glycosuria (Fig. 1D). VSG was associated with striking decreases in both SGLT2/Slc5a2 mRNA levels (Fig. 1E) and immunoreactivity (Fig. 1F and G) in the kidney cortex. Consistent with these findings, treatment with the SGLT2 inhibitor dapagliflozin exerted no additive effect on glucose tolerance in VSG-treated mice compared with sham-operated-on mice, with only the latter responding significantly to the drug (Fig. 1H and I). No difference was observed in SGLT1/Slc5a1 expression (Fig. 1J), and expression of the renal gluconeogenesis markers Pck1 and G6pc1 was unchanged (Fig. 1K and L). Expression of a specific marker of proximal tubular injury (Kim-1) was measured to ensure no renal tubular injury was caused as a result of the surgery (Fig. 1M). Similar findings were observed in mice maintained on the HFD for 12 weeks (Fig. 2A–M), although in this case, surgery also affected body weight (Fig. 2A). Moreover, glycosuria was more pronounced in HFD animals (Fig. 2D), and a correlation was observed between urinary glucose and SGLT2 protein expression. The expression of Pck1 and G6pc1 was also lowered in HFD animals following VSG (Fig. 2K and L).

To determine whether the observed inhibited SGLT2 expression after bariatric surgery might be driven by lower glycemia or higher GLP-1 plasma concentration, lean nonoperated-on mice were acutely treated daily with a SC dose (5 nmol/kg) of the GLP-1 receptor agonist semaglutide for 7 days and chronically for 30 days. Although body weight remained stable in both studies (Fig. 3A and D), glycemia was significantly lower in semaglutide-treated animals compared with the saline-treated control group, as expected (Fig. 3B and E). Moreover, insulin secretion was significantly increased in chronically treated animals at 15 min postglucose injection (Fig. 3F). Although acutely treated animals did not display significant differences in kidney cortex SGLT2 gene expression between the groups, chronic administration of semaglutide did lower SGLT2 expression, consistent with potential roles for lowered glucose, elevated GLP-1, or insulin in the effects of surgery on SGLT2 expression. However, urinary glucose did not increase as a result, indicating that this lowering of SGLT2 expression was not sufficient to trigger glycosuria (Fig. 3H). Pck1 and G6pc1 were unchanged by the treatment (Fig. 3I and J).

In order to assess the role played by the lowered SGLT2 expression in the effects of VSG, we generated a whole-body SGLT2/Slc5a2-knockout mouse model using CRISPR/Cas9 (Research Design and Methods) (Fig. 4A and B). The strategy was designed to remove the whole of exon 1 and thus the ability of the transporter to bind Na⁺ ions (Fig. 4B) and required for glucose transport (15). Deletion of exon 1 was confirmed by genomic PCR and subsequent Sanger sequencing (Fig. 4C and Supplementary Fig. 2A). The loss of mRNA encoded by exon 1 (Supplementary Fig. 2B), but retained expression of mRNA encoded by exons 2–3 (Supplementary Fig. 2C) and exons 4–5 (Supplementary Fig. 2D), was confirmed via real-time quantitative PCR. Loss of SGLT2 protein was confirmed by Western immunoblotting (Fig. 4D and E). Profound glycosuria was observed in SGLT2-deficient mice (Fig. 4F) (28). Finally, consistent with the elimination of functional SGLT2 transporters, electrophysiological recordings in intestinal mucosae from SGLT2^−/− mice were insensitive to the inhibitor canagliflozin. The inhibitor blocked Na⁺ absorption (alongside glucose absorption) and thus reduced Isc levels in mucosae from the small and large intestine of lean WT mice only (Fig. 4G), revealing a small but significant SGLT2 signal (∼10–20%) in WT colon compared with SGLT1 activity (Fig. 5H), as observed previously in WT colon (see Fig. 4 in Cox et al. [29]).

SGLT2^−/− mice were fed the HFD for 12 weeks and then separated into sham (n = 7) and VSG (n = 8) groups. Both groups displayed very similar initial weight loss and weight gain (Fig. 5A). Four weeks after surgery, VSG and sham SGLT2^−/− mice displayed no significant differences in the time course of blood glucose following an IPGTT (8-h fasting, 1 g/kg). Importantly, the two SGLT2^−/− groups did not differ compared with obese WT VSG-treated mice, indicating that after the low glucose challenge (1 g/kg), the absence of SGLT2 may be behind the improved glucose tolerance. However, all groups (WT VSG, SGLT2^−/− VSG, and sham) displayed significantly enhanced glucose tolerance compared with WT, HFD sham surgery mice (Fig. 5B). Insulin secretion was also similar between VSG and sham SGLT2^−/− mice (Fig. 5C). During oral glucose tolerance tests (OGTTs), which compared SGLT2^−/− and WT mice, there was no significant difference between SGLT2^−/− mice that had undergone VSG and SGLT2^−/− sham mice, even though VSG led to significantly improved glucose tolerance in WT mice (Fig. 5D). However, during an IPGTT (16-h fasting, 3 g/kg), there was a clear difference between VSG-treated SGLT2^−/− and sham-operated-on SGLT2^−/− mice (Fig. 5E), revealing that VSG exerts a further, SGLT2-independent impact on glucose homeostasis under these conditions where the high glucose load presumably saturates the capacity for SGLT2-dependent glomerular filtration in WT sham mice.

Interestingly, in SGLT2^−/− mice that had undergone VSG, a further improvement in glucose tolerance was still observed following dapagliflozin gavage (10 mg/kg), indicating an apparent off-target (SGLT2-independent) effect of the drug (Fig. 5F).

We also measured glucose tolerance in HFD WT (Fig. 2H) and SGLT2^−/− (Fig. 5F) mice following surgery. Explored in the absence of dapagliflozin, surgery reduced the AUC during IPGTT by 29.4% in WT mice (Fig. 2H and I), whereas after SGLT2 deletion, this reduction was lowered to 23.3% (Fig. 5F).

There was no compensatory increase in SGLT1/Slc5a1 gene expression in SGLT2^−/− mice (Fig. 5G). As expected, the SGLT1 inhibitor phloridzin reduced Isc levels, particularly in the jejunum of SGLT2^−/− lean mice (Fig. 5H). Why this area-specific SGLT1 signal was enhanced significantly is not clear, but it may be due to the higher basal levels of Isc recorded in the jejunum mucosa of lean SGLT2^−/− mice (Fig. 5I). HFD feeding blunted this jejunum-specific phloridzin effect in SGLT2^−/− mice (Fig. 5H and I).

We sought next to explore whether surgery may reflect an altered number or activity of L cells in the intestine, which may in turn enhance the production of incretins. L-cell–derived endogenous peptide YY (PYY)-Y1 receptor tone was measured in mucosae from WT and SGLT2^−/− mice using a competitive Y1 receptor antagonist (29). Both WT and SGLT2^−/− colonic mucosa exhibited significantly elevated levels of PYY-Y1 tone compared with the rest of the gut, following the established pattern of L-cell distribution (30) (Fig. 6A and B). Moreover, Y1 tone was significantly higher in the colon of obese WT mice following VSG, indicating increased L-cell activity postsurgery (Fig. 6A) and in line with previous reports of increased GLP-1 secretion after VSG (31). Lean and obese SGLT2^−/− mice both displayed higher Y1 tone in the colon compared with lean WT mice (Fig. 6B). GLP-1 levels were measured after OGTT in the SGLT2^−/− mice. Indeed, SGLT2^−/− VSG and sham mice both showed a significant increase of GLP-1 15 min following a glucose oral gavage (Fig. 6C). These observations are in contrast to earlier findings in the same surgical model of HFD obesity (26) showing that WT obese sham-operated-on mice do not respond to oral gavage with a significant increase of GLP-1.

We explored the possibility that the effects of dapagliflozin may, in part, be due to an action on the pancreatic α-cell, as suggested by Bonner et al. (32,33) but contradicted by findings from Kuhre et al. (34) and Chae et al. (35). Following an overnight fast, plasma glucagon was increased by dapagliflozin treatment in both sham WT mice and in both groups of SGLT2^−/− mice (sham and VSG) (Supplementary Fig. 3), arguing against an action of the inhibitor via SGLT2. Glucagon levels were also raised by VSG in both WT and SGLT2^−/− mice even in the absence of dapagliflozin.

In order to explore potential mechanisms through which SGLT2 expression is lowered following VSG, we used the human kidney cell line HK-2 (36). Cells were treated for 24 h with a range of molecules whose levels have previously been observed to change following bariatric surgery, including glucose, glucagon, GLP-1, and insulin (9). None of those tested affected SGLT2 expression (Supplementary Fig. 4).

It has been reported previously that increased adiponectin reduces renal SGLT2 through activation of PPAR-δ (37). Although PPAR-δ gene expression in subcutaneous adipose tissue was significantly increased in WT lean mice (Supplementary Fig. 5B), adiponectin levels in the plasma of lean mice (Supplementary Fig. 5C) were not changed by VSG. Adiponectin gene expression in lean and obese WT mice (Supplementary Fig. 5A and D) in SC adipose tissue were also not significantly increased post-VSG, pointing toward alternative mechanisms of SGLT2 regulation. Similar to HFD WT mice, levels of PPAR-δ and adiponectin mRNA were not altered in adipose tissue of SGLT2^−/− sham versus VSG mice (Supplementary Fig. 5F and G).

We show here that the expression of SGLT2/Slc5a2 downregulated in the kidney cortex in mice after bariatric surgery. Interestingly, our findings also suggest that an HFD may increase SGLT2/Slc5a2 expression in the kidney (Fig. 2E vs. Fig. 1E), possibly contributing to hyperglycemia in these conditions.

Animals that underwent VSG displayed a lowered excursion in glucose and augmented insulin secretion in response to intraperitoneal delivery of glucose compared with sham controls. Obese animals additionally displayed high glucose content in their urine following oral delivery of glucose. These findings provide important insights into the mechanisms underlying type 2 diabetes remission following bariatric surgery and point to a potential gut-kidney axis whereby (presently undefined) factors, likely to be released from the intestine, influence gene expression in the kidney (Supplementary Fig. 6).

We previously reported in a meetings proceedings (25) that stomach-sparing duodenal-jejunal bypass in lean rats caused no significant weight loss compared with sham animals, while SGLT2 expression was lowered in the kidney cortex. In the current study, we used lean and obese mice and performed a different type of bariatric surgery (VSG) that reduces the size of the stomach rather than rerouting the intestinal tract. This approach also lowered SGLT2 expression and in each case points toward a mechanism that involves communication—either direct or indirect—between the gastrointestinal tract and the kidney.

SGLT2 deletion in the mouse was previously shown to reduce obesity-associated hyperglycemia and increase glucose-stimulated insulin secretion in vivo (38). These changes are similar to those after treatment with pharmacological SGLT2 inhibitors (39,40). In order to assess the extent to which lowering of SGLT2 levels post-VSG resulted in the same phenotype as SGLT2 inhibitor treatment, we administered dapagliflozin to VSG and sham-operated-on mice. The time course of blood glucose following an intraperitoneal glucose load in the VSG-treated mice remained significantly greater than in dapagliflozin-treated sham-operated-on mice. This observation suggested that SGLT2 inhibition may contribute to bariatric-associated euglycemia, but it is not the sole mechanism involved.

How might VSG lower SGLT2 expression in the kidney? There is currently conflicting evidence regarding the role of glucose levels in the glomerular filtrate of the kidney (41). A recent study reported (37) that hyperglycemia reduces urinary sodium excretion by enhancing SGLT2 activity. Adipokines were found to upregulate SGLT2 in vitro (42), and increased adiponectin was shown to cause activation of PPAR-δ in the adipose tissue and reduce renal SGLT2 (37). Recent evidence (43) showed potential cross talk between the sympathetic nervous system and SGLT2 (44). At the transcriptional level, hepatocyte nuclear factor 1-α (HNF1-α) is known to bind to the SGLT2 promoter in the kidney and is considered a vital transcription factor for SGLT2 expression (45). Moreover, both cAMP and protein kinase A were shown to upregulate SGLT2 at the posttranslational level (46).

We explored a number of different mechanisms that may be involved in the downregulation of SGLT2 after VSG. We first explored a role for postoperative lowered glucose levels or increased GLP-1 by administering semaglutide treatment to lean mice. Although acute administration exerted no significant effect on SGLT2 mRNA, chronic treatment reduced SGLT2 levels significantly. However, the effects of semaglutide on SGLT2 expression (<40%) were less pronounced than those of VSG (>75%) and may explain why semaglutide-treated mice did not demonstrate glycosuria. Nevertheless, this finding suggests that metabolic improvements, as also seen after incretin treatment (lowered glycemia and increased circulating insulin. etc.), are likely to contribute to the effects of surgery on renal SGLT2 expression. This view is further supported by the observation that SGLT2/Slc5a2 expression is twofold higher in HFD animals compared with lean mice. Further evidence, through in vitro studies or the generation of suitable knockout models in the relevant proximal tubular cell types in the mouse (e.g., for insulin or GLP1 receptors), will be required to reach firm conclusions on the exact mechanisms involved. Changes in circulating adiponectin, which were not observed after surgery in the current study, would seem to be unlikely to be involved.

Since VSG lowers renal SGLT2 gene expression and dapagliflozin antagonizes the transporter, the action of the two interventions may not be the same. In order to evaluate this, we developed an SGLT2-deficient mouse line (SGLT2^−/−) by deleting exon 1 of the Slc5a2 gene. VSG or sham-treated SGLT2^−/− mice did not display any difference in body weight. This may initially suggest that VSG is unable to cause weight loss in SGLT2^−/− mice. However, when compared with our previous study that tracked body weight weekly in obese WT mice treated with VSG or sham surgery (26), we note that SGLT2^−/− sham mice gained less weight than WT. This implies that SGLT2^−/− mice are partially protected from significant weight gain after HFD, confirming previous findings (38).

Importantly, there was no significant difference in insulin secretion between SGLT2^−/− mice treated with VSG and sham operations up to 30 min following a glucose load (insulin AUC 0–30 min VSG: 73.8 ± 16.9 vs. sham: 50.5 ± 13.4). This is in contrast to VSG-treated WT mice that demonstrated a spike in insulin secretion increase postglucose load compared with sham-treated WT mice (insulin AUC 0–30 min VSG: 56.3 ± 17.27 vs. sham: 25.5 ± 6.9). Our analyses indicate that the contribution of the observed reduction in SGLT2 in bariatric surgery is dependent upon the glucose load. After exposure to a high (3 g/kg) glucose load, we observed an ∼23% reduction in the effects of surgery (SGLT2^−/− vs. WT), while the analogous value was ∼14.4% when mice were treated with a lower glucose load (1 g/kg).

An important observation made during the current study was that of an apparent off-target effect of dapaglifozin, whose actions were still apparent in SGLT2-null mice. We note that these effects of the inhibitor were less apparent in WT mice. One potential reason for this is that SGLT2-null mice are more metabolically healthy that WT mice, in line with their lower body weight, thus allowing off-target effects to be more readily detected. Further and more detailed studies will be needed in the future to explore the basis of these effects of the drug.

A further observation made in this study relates to results of previous work, which have suggested that inhibition of SGLT2 in the α-cell may contribute to increase glucagon secretion (32,33). However, this proposal has been contested based on the absence of effects of SGLT2 inhibitors on isolated islets (34). Our findings suggest that observations of Bonner et al. (32) might reflect off-target effects of dapaglifozin. Thus, both sham- and VSG-treated SGLT2^−/− mice displayed increased glucagon levels following dapagliflozin treatment.

One likely contributor to the effects of VSG is altered GLP-1 secretion. We previously reported that WT obese mice have a blunted GLP-1 response to oral glucose versus lean animals and that this deficiency is rescued following VSG (26). However, both SGLT2^−/− obese sham and VSG obese mice displayed significantly increased GLP-1 secretion. This confirms previous studies showing higher GLP-1 plasma levels as a response to SGLT2 inhibition treatment (47) in addition to the importance of SGLT1/SGLT2 inhibition for enteropancreatic hormone concentrations and glucose metabolism after bariatric surgery (48). Combined with our findings of lowered SGLT2 expression after chronic semaglutide treatment, these results may indicate a feedback mechanism between GLP-1 and SGLT2 levels.

The findings described here provide important new insights into the mechanisms of action of bariatric surgery. We demonstrate that bariatric surgery in rodents downregulates SGLT2/Slc5a2 expression in a weight-independent manner. We also report that exogenous inhibition of SGLT2 action does not fully replicate the positive effects of surgery but that genetic elimination of SGLT2 reduces the effect of surgery on glucose clearance. Finally, we present evidence for potentially beneficial off-target effects of dapagliflozin in glucose clearance.

To date, we are not aware of any clinical trials that have reported on urinary glucose levels following bariatric surgery in humans. Future studies will therefore now be important to determine 1) whether these mechanisms are translated in humans by exploring the effect of bariatric surgery on glycosuria following a glucose challenge; 2) the molecular mechanisms that drive lowered SGLT2 expression following bariatric surgery, and 3) the possible off-target effects of SGLT2 inhibitors. Understanding the relationship between the gastrointestinal tract and the kidney may also shed light on a possible feedback loop between SGLT2-driven glucose absorption and GLP-1 secretion.

Funding. E.A. was supported by a grant from the Rosetrees Trust (M825) and from the British Society for Neuroendocrinology. G.A.R. was supported by a Wellcome Trust Investigator Award (212625/Z/18/Z), Medical Research Council Programme grants (MR/R022259/1, MR/J0003042/1, MR/L020149/1), and Experimental Challenge Grant (DIVA, MR/L02036X/1), MRC (MR/N00275X/1), and Diabetes UK (BDA/11/0004210, BDA/15/0005275, BDA 16/0005485) grants. This project has received funding from the European Union’s Horizon 2020 Research and Innovation Programme via the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No. 115881 (RHAPSODY) to G.R. I.L. was supported by a project grant from Diabetes UK (16/0005485). I.R.T was funded by a BBSRC project grant (BB/N006763/1).

Duality of Interest. G.A.R. has received grant funding from Sun Pharmaceuticals Inc. and Laboratoires Servier and is a consultant for Sun Pharmaceuticals Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. E.A. undertook all mouse studies. E.A. and I.R.T. undertook all data analyses. E.A. and G.A.R. designed and supervised the study. E.A. and G.A.R. wrote the manuscript. L.L.-N., M.H., and I.L. assisted with mouse studies. Additional experiments were performed by I.R.T. and supervised by H.M.C. All authors contributed to writing the manuscript. G.A.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the analysis.