Physiology of renal glucose handling via SGLT1, SGLT2 and GLUT2

Although the kidneys freely filter plasma glucose, none appears in the urine. However, in diabetes, glucose may appear in the urine, specifically when the plasma glucose concentration is so high that the filtered load exceeds the maximum capacity for sugar reabsorption. Homer Smith and his colleagues were the first to quantify glomerular filtration rates in humans. They showed that all the filtered glucose was normally reabsorbed and, importantly, they also demonstrated that reabsorption of total filtered glucose load could be inhibited by a compound called phlorizin (Fig. 1) [1]. This prompted decades of research into the location and mechanism of kidney glucose transport, mostly in animal models.

The current models for glucose reabsorption from the glomerular filtrate by SGLT2 and SGLT1 in the different segments of the proximal tubule, S1/S2 and S3, are shown in Fig. 3. In both S1/S2 and S3 segments, the first stage is glucose transport across the apical membrane by SGLTs. This leads to glucose accumulation within the epithelium, modulated to some extent by intracellular metabolism. The glucose concentration gradient between the cell and plasma in turn drives the second stage: net passive exit of glucose through the basolateral membrane, towards the plasma, via GLUT2. The basolateral Na⁺/K⁺ pump (which extrudes three sodium ions for every two potassium ions entering the cell) maintains the sodium gradient across the apical membrane by pumping sodium out of the cell, towards plasma. Inhibition of the Na⁺/K⁺ pump by cardiac glycosides blocks the pumping of sodium out of the cell, with the concomitant rise in intracellular sodium concentration. The elimination of the sodium gradient across the apical membrane results in the loss of sodium–glucose cotransport across the apical membrane. Thus, the two-stage process, together with the absorption of glomerular fluid, accounts for the complete absorption of glucose by the time the filtrate reaches the end of the proximal tubule (Fig. 2).

In 1987, members of the Wright laboratory began pioneering work that resulted in the identification of SGLTs and their functional properties. The rabbit intestinal transporter was first identified by expression cloning [12], followed by homology cloning of the human intestinal SGLT1 and renal SGLT2 transporters [13, 14]. The SGLTs belong to a human gene family, SLC5, which contains 12 members, including sodium-coupled transporters for myoinositol, fructose, iodide and short-chain fatty acids [6, 15, 16]. SGLT2 (also known as SLC5A2) mRNA is almost exclusively expressed in the kidney, while SGLT1 (SLC5A1) mRNA is found mainly in the small intestine and is only expressed to a small extent in the kidney [6, 17]. A single-cell transcriptomic study of mouse kidney has revealed that Sglt2 is a unique marker gene for cells of the S1 segment of the proximal tubule [18]. Glut2 (also known as Slc2a2) is expressed in both S1 and S3, while Sglt1 is expressed at a low level along the proximal tubule, with a somewhat higher level in S3. The SGLT genes code for membrane proteins with 14 transmembrane helices, as confirmed by the crystal structures of a bacterial homologue, vSGLT [19, 20]. The crystal structures have also provided important clues about the SGLT transport mechanism (detailed below).

External phlorizin acts as a high-affinity, specific, non-transported, competitive, dead-end inhibitor of sodium:glucose transport in and out of a cell. It contains a glucose structure on one end, which binds to the glucose-binding site on SGLTs, and an aglycone tail (phloretin), which binds to the wall of the hydrophilic cavity leading to the glucose-binding site (described further below). Internal phlorizin, even at high cytoplasmic sodium concentrations, is a poor inhibitor of both forward and reverse glucose transport [31].

Several studies have been carried out to confirm and explain the total inhibition of glucose reabsorption by phlorizin that is reported in humans [1]. For example, using microPET to monitor Me-4FDG excretion into the urinary bladder, we have shown that intravenous injection of phlorizin or the specific SGLT2 inhibitor dapagliflozin rapidly increases excretion of Me-4FDG into the urinary bladder (Fig. 5a) [32]. Dapagliflozin acts specifically on the kidneys, as shown by microPET imaging of mice injected with 4-[¹⁸F]fluoro-dapagliflozin (F-Dapa) [32]. F-Dapa binds specifically to the external surface of functional SGLT2 in the plasma membrane (with a binding constant of 4 nmol/l) [33] and is displaced by phlorizin and cold (non-radioactive) dapagliflozin. In rodents, the kidney is the only organ that shows significant specific F-Dapa binding (Fig. 5b). As SGLT2 inhibitors only bind to functional SGLT2 proteins in plasma membranes [31], this extends the immunohistochemical finding of the kidney-specific location of SGLT2 in rodents [23].

So, where in the kidney does F-Dapa bind? This has been determined by ex vivo autoradiography (Fig. 5c,d). The inhibitor binds to the outer cortex of the whole mouse kidney and (as shown at higher magnification) only within tubules surrounding the glomeruli. This pattern of binding is similar to the hybridisation of Sglt2 (also known as Slc5a2) mRNA in rat kidneys and SGLT2 antibody binding in mouse, rat and human kidneys. We conclude that dapagliflozin is filtered from the plasma by the glomerulus and then binds to SGLT2 in the early proximal tubule, where it inhibits glucose absorption. Biochemical studies in dog kidney have shown that [³H]-phlorizin is filtered by the kidney, binds to the proximal tubule brush border membranes, and is displaced and excreted into the urine by excess cold (non-radioactive) phlorizin [34]. Unlike phlorizin, dapagliflozin and other SGLT2 inhibitors are not excreted into the urine, implying that they are reabsorbed further down the nephron and excreted in bile [32].

A mechanical model of sodium–glucose cotransport by SGLTs is shown in Fig. 6. The vast majority of information on SGLT-mediated sodium–glucose cotransport has come from biochemical and biophysical experiments on human SGLT1 in heterologous expression systems, with a more limited set of data from human SGLT2 experiments. In addition, solving the crystal structure of the bacterial homologue vSGLT and molecular dynamic simulations of these structures has provided further insights into the molecular mechanisms of SGLTs. We have also learnt from the fact that SGLTs belong to a structural family of transporters that have a common five-helix inverted repeat motif, the LeuT structural fold (see [6, 35]). Different cotransporters and exchangers have this common structure and so may have similar transport mechanisms. For cotransporters and exchangers in this family, the substrate binding sites are located in the middle of the protein, with external and internal gates isolating the substrate from the extracellular and intracellular solution on each side of the membrane (Fig. 6). Having determined the structure of GLUT transporters in the SLC2 gene family [36], it is suggested that glucose transport through these proteins has a similar gated mechanism to the cotransporters, but with the difference that opening and closing of the external and internal gates is not controlled by sodium.

SGLT1 normally couples the inward transport of two sodium ions and one glucose molecule in each kinetic cycle. The direction of the transporter is completely reversible. The rate and direction of transport is simply a function of the extracellular and intracellular sodium and glucose concentrations and the polarity and magnitude of the membrane potential. Cotransport of sodium and glucose by SGLT1 and SGLT2 generates an electrical current and this provides a biophysical tool to measure the rate of cotransport. The sugar-activated (inward or outward) current is observed in the presence of sodium and is equivalent to the rate of sodium transport. In the case of SGLT1, capacitive currents are generated by voltage jumps in the presence and absence of sodium, in the absence of sugar. These currents are due to the movement of charged or polar amino acid residues of SGLT1 in the membrane electric field. They have provided a powerful biophysical means of measuring SGLT1 presteady state kinetics and the number of SGLT1 proteins in the plasma membrane. As yet, capacitive currents have not been recorded for human SGLT2.

In the normal forward transport cycle (Fig. 6), extracellular sodium first binds to the Na2 and Na1 sites. The Na2 site is conserved in SGLT2 and other members of the LeuT structural family: in SGLT1 the residues involved in coordinating Na⁺ are A76, I79, S289, S392 and S293. We have evidence that the Na1 site in human SGLT1 involves residues N78, H83, E102, Y290 and W291 [37]. Sodium binding increases the probability of opening the external gate so that external glucose (and phlorizin) can bind. The residues responsible for glucose binding are N78, H83, E102, K321, Q457, Y290 and W291, and these are conserved in human SGLT2. After binding glucose (or phlorizin) the outer gate closes to occlude the substrate from the external solution and then the inner gate opens to allow glucose and sodium to escape into the cytosol. Finally, the inner gate closes and the kinetic cycle continues to the initial starting position. We originally proposed an ordered internal dissociation of glucose and sodium, but recent experiments and molecular dynamic studies favour the simultaneous release of glucose and sodium [38]. One complete cycle takes about 20 ms.

Although kinetic studies of SGLT2 are not as advanced as those for SGLT1, the transport model is similar (Fig. 6). A major reason for the lack of data on SGLT2 is the low expression of this transporter in heterologous expression systems, such as X. laevis oocytes and cultured cells at 22°C. In preliminary kinetic studies on oocytes, the expression level of SGLT2 was less than 1% of that for SGLT1 [39]. On learning from Ana Pajor and Chari Smith that raising the temperature to 37°C dramatically increased SGLT2 activity in cultured cells [40], we were able to perform a robust comparison of the kinetics of SGLT1 and SGLT2 in HEK 293 cells [26, 33]. The major difference between SGLT2 and SGLT1 is that the sodium:glucose coupling ratio is 1:1 for SGLT2 vs 2:1 for SGLT1 [26]. Other differences include the higher affinity of phlorizin to SGLT2 vs SGLT1 (K_i 11 vs 140 nmol/l), the higher affinity of gliflozins, such as dapagliflozin to SGLT2 (K_i 4 vs 400 nmol/l), and a narrower sugar selectivity of SGLT2 (galactose is a poor substrate for SGLT2). Michael Coady and his colleagues recently elucidated the reason for low SGLT2 expression in oocytes with the discovery that co-expression of MAP17 was required [41, 42]. Apart from small differences in kinetic variables (likely owing to the difference in temperature of culture conditions), the kinetic properties of SGLT2 expressed in oocytes (cultured at 22°C) were identical to those in HEK 293 cells (cultured at 37°C). The role of MAP17 in SGLT2 expression in oocytes is not yet understood.

In HEK 293T cells, SGLT2 is regulated over the short term by stimulation of protein kinases, Protein kinase A (PKA) and Protein kinase C (PKC) [43]. This reversible stimulation occurs with a half-time of 10 min and is due to a change in the maximum rate of transport. Insulin mimics the effect of PKA and PKC activation, but this does not occur with deletion of the only phosphorylation site on SGLT2, Ser⁶²⁴Ala. PKC and PKA also regulate human SGLT1 in cultured cells, resulting in rapid changes in the trafficking of an intracellular pool of transporters to the cell membrane. In oocytes, biophysical and electron microscopic studies show that PKA- and PKC-induced increases in SGLT1 activity is due to an increase in the number of SGLT1 proteins in the cell membrane, at a rate of 1 × 10⁷ molecules per second [6, 44]. There was no effect of insulin on human SGLT1 trafficking.

Following the work of Eric Turk in purifying the bacterial transporter, vSGLT [45], the x-ray structure was solved in two conformations [19, 20]. The structures have provided unique insights into the mechanism of sodium-coupled sugar transport, as incorporated into the transport model (Fig. 6). As aforementioned, the protein has 14 transmembrane helices with a core-inverted repeat of transmembrane helices 1–5 and 6–10, each containing a discontinuous helix (transmembrane helix [TM]1 and TM6). The substrate binding site is in the middle of the protein, adjacent to the discontinuous helices (N78, H83, E102, K321 Q457, Y290 and W291), and occluded from the external and internal solutions by external and internal gates. The relatively high sequence identities and similarities between vSGLT, SGLT1 and SGLT2 have made it possible to construct homology models of the human proteins (Fig. 7). In the sugar occluded state (Fig. 6, state 3) glucose is coordinated by N78, H83, E102, K321, Q457, Y290 and W291, and is excluded from contact with the external solution by hydrophobic residues (L84, F98 and F453) and the internal solution, in part, by Y290. A common Na2 sodium-binding site in the SGLT1 membrane and other proteins within the LeuT structural family is proposed at S393 (along with A76, I79, S389 and S392; not shown in Fig. 6). Mutations of each of the SGLT1 sugar and sodium coordination residues dramatically alter sugar binding, whereas mutations of the external gate residues do not [46].

Considerable progress has been made by using the structure of vSGLT to examine the molecular dynamics of SGLTs (interested readers are referred to the literature, e.g. [20, 38]). These and other biophysical studies of human SGLT1 provide insights into the conformational changes associated with sodium binding and isomerisation of the transporter between the outward and inward facing structures (e.g. see [37, 47]).

The lead compound for the development of SGLT drugs was phlorizin [48]. As indicated above, this plant glucoside is a non-transported, specific competitive inhibitor of SGLT2 and SGLT1, with a K_i of 11 nmol/l and 140 nmol/l, respectively. It preferentially binds to the external surface of the SGLTs in the presence of external sodium, blocking glucose transport (inward or outward). The aglycones of phlorizin (phloretin) and dapagliflozin are poor, non-competitive inhibitors of SGLTs, indicating that the glycosides bind to both the glucose-binding pocket and an adjacent lipophilic vestibule leading to the glucose-binding site (Fig. 8). There is remarkable selectivity in the binding of gliflozins to SGLTs; for example, dapagliflozin binds to SGLT2 with a much greater affinity than to SGLT1, and galacto-dapagliflozin has orders of magnitude lower affinity for SGLT1 than SGLT2 [33].

A mutational analysis of phlorizin binding to SGLT1 confirms that the sugar-binding site is important for both glucose and phlorizin binding, in that there was a linear relationship between glucose K_m and phlorizin K_i [46]. A major exception is that mutation of an outer gate residue, F101C, increased phlorizin K_i by 200-fold with no change in glucose K_m. This indicates an interaction, possibly π–π bonding, between F101 and the phlorizin aglycone. Additional clues about the location of the phlorizin aglycone binding site comes from voltage clamp fluorometery of SGLT1, which identified that a 600 Å³ vestibule leads to the glucose-binding site in the sodium-bound open conformation (Fig. 6, state 2, and Fig. 8) [47]. This vestibule is lined with hydrophobic resides on TM1, TM2, TM6, TM9 and TM10 (Fig. 7). We anticipate that gliflozins bind to a similar site in both SGLT1 and SGLT2, and this is supported by molecular dynamic studies of inhibitor binding to SGLTs [49]. The success of SGLT2 drugs is, in large part, due to their high affinity for SGLT2 and the limited expression of this transporter in the kidney cortex (see above).

The development of GLUT2 drugs to treat diabetes has not been practical owing to the close structural and functional similarity between GLUT family members and the vital functions of GLUTs throughout the body [36].

As previously mentioned, there are three known rare, autosomal recessive disorders of SGLTs and GLUT2, glucose–galactose malabsorption (OMIM 182380), familial renal glucosuria (OMIM 233100) and Fanconi–Bickel syndrome (OMIM 227810), all of which result in mild to excessive glucosuria (1–150 g [1.73 m]⁻² day⁻¹ [see textbox below]). In glucose–galactose malabsorption, mutations in SGLT1 cause a defect in intestinal glucose (and galactose) absorption, resulting in mild glucosuria (see [6, 50]). These mutations cause malabsorption due to mistrafficking of SGLT1 to the brush border membrane. In contrast, mutations in SGLT2 cause renal glucosuria (from 1 g [1.73 m]⁻² day⁻¹ to 150 g [1.73 m]⁻² day⁻¹) without defects in intestinal absorption. Only in cases of homogeneous truncation mutations in SGLT2 is it reasonable to expect severe glucosuria, but the reserve capacity of SGLT1 should be taken into account. Unlike glucose–galactose malabsorption, there are no comprehensive studies of the transport properties of SGLT2 mutants, largely due to the low expression of SGLT2 in heterologous expression systems (as detailed above). This, and the incomplete clinical study of individuals with familial renal glucosuria, has led to confusion about the inheritance of this disease. However, we expect familial renal glucosuria to be a simple autosomal recessive disease.

Overall, these genetic disorders, caused by mutations in SGLT1, SGLT2 and GLUT2, support animal studies and confirm the importance of the three affected proteins in reabsorbing the filtered glucose load in healthy humans.

Thirty years ago, the cloning of human SGLT1, SGLT2 and GLUT2 was a major breakthrough in the identification of transporters involved in renal glucose absorption. Subsequent advances in molecular biology permitted the development of genetic mouse models lacking SGLT1, SGLT2 and GLUT2, furthering our understanding of the role of these transporters in the proximal tubule. Based on the assumption that SGLT2 was responsible for the bulk of glucose reabsorption, the pharmaceutical industry in Japan, USA and Europe invented SGLT2 inhibitors to treat type 2 diabetes. Although these drugs do lower blood glucose levels in patients with diabetes without causing hypoglycaemia, they are only capable of a 50% inhibition of glucose reabsorption and are, thus, not so effective as a monotherapy. In retrospect, given the huge reserve capacity of SGLT1 in the late proximal tubule, one can argue that dual SGLT2 and SGLT1 drugs may be more effective in controlling renal glucose excretion.