The osmolyte and folding chaperone betaine is transported by the renal Na+-coupled GABA (γ-aminobutyric acid) symporter BGT-1 (betaine/GABA transporter 1), a member of the SLC6 (solute carrier 6) family. Under hypertonic conditions, the transcription, translation and plasma membrane (PM) insertion of BGT-1 in kidney cells are significantly increased, resulting in elevated betaine and GABA transport. Re-establishing isotonicity involves PM depletion of BGT-1. The molecular mechanism of the regulated PM insertion of BGT-1 during changes in osmotic stress is unknown. In the present study, we reveal a link between regulated PM insertion and N-glycosylation. Based on homology modelling, we identified two sites (Asn171 and Asn183) in the extracellular loop 2 (EL2) of BGT-1, which were investigated with respect to trafficking, insertion and transport by immunogold-labelling, electron microscopy (EM), mutagenesis and two-electrode voltage clamp measurements in Xenopus laevis oocytes and uptake of radiolabelled substrate into MDCK (Madin–Darby canine kidney) and HEK293 (human embryonic kidney) cells. Trafficking and PM insertion of BGT-1 was clearly promoted by N-glycosylation in both oocytes and MDCK cells. Moreover, association with N-glycans at Asn171 and Asn183 contributed equally to protein activity and substrate affinity. Substitution of Asn171 and Asn183 by aspartate individually caused no loss of BGT-1 activity, whereas the double mutant was inactive, suggesting that N-glycosylation of at least one of the sites is required for function. Substitution by alanine or valine at either site caused a dramatic loss in transport activity. Furthermore, in MDCK cells PM insertion of N183D was no longer regulated by osmotic stress, highlighting the impact of N-glycosylation in regulation of this SLC6 transporter.
Secondary active transporters play important roles in maintaining vital functions in mammalian kidneys [1,2], which constantly encounter major changes in osmolality during urinary concentration . The renal hydration state is adjusted by accumulation of several osmolytes, e.g. betaine, taurine, sorbitol, glycerophosphorylcholine (GPC) and myo-inositol . For example, the sodium-myo-inositol transporter (SMIT), SLC5A3 and the betaine/γ-aminobutyric acid (GABA) transporter (BGT-1; alias GAT2, SLC6A12) accomplish cellular accumulation of myo-inostol and betaine respectively and thus contribute to cell volume regulation in the kidneys. Moreover, betaine-mediated volume regulation prevents severe dysfunction of the kidneys and the central nervous system [5–7] and counteracts the destabilizing action of urea on protein structures in renal cells [8,9].
BGT-1, which belongs to the osmolyte branch of the human neurotransmitter sodium symporter (NSS) or SLC6 family  is also found in the brain and in the liver [11,12]. BGT-1 mediates the transport of the neurotransmitter GABA; however, with a lower affinity compared with other GABA transporters . Both betaine and GABA transport are chloride-dependent and coupled to the transport of three sodium ions [13,14], which might be required to provide in the kidneys a sufficient driving force in order to achieve an intracellular renal medullary cell betaine concentration of ∼100 mM [15,16]. During hypertonicity, both bgt-1 transcription and BGT-1 insertion into the basolateral plasma membranes (PMs) are increased so that transport of betaine is up-regulated in Madin–Darby canine kidney cells (MDCK) . Like BGT-1, SMIT1 expression is rapidly stimulated during hypertonicity, leading to increased myo-inositol uptake from the extracellular space .
To date, the genomic regulation of BGT-1 and the other organic osmolyte transporters SMIT1 and of the taurine transporter (TauT) involve the transcription factor tonicity enhancer-binding protein (TonEBP) [19–21]. During hypertonicity, TonEBP is activated and binds to tonicity responsive enhancers (TonE) on the BGT-1 promoter region thereby inducing its transcription . Upon hypertonic conditions, binding of TonEBP to the promotor region of osmo-sensitive genes was therefore considered to be a general mechanism to protect cells from shrinkage. Furthermore, Klaus et al. showed that the cell volume-sensitive protein kinase (serum- and glucocorticoid-inducible kinase, SGK1) modulates SMIT1 function by increasing its PM abundance without affecting transport kinetics .
Beyond that, the molecular mechanisms of stress sensing and regulation of PM insertion of BGT-1 are unknown. Controlled folding by N-glycosylation might be one regulatory mechanism, as N-glycans are reported to affect targeting and functional insertion to the PM in several proteins [23,24]. The role of N-glycosylation sites in the second extracellular loop (EL2; Figure 1A) of SLC6 transporters has been widely discussed [25–30]. EL2 connects transmembrane domains (TMs) 3 and 4 (Figure 1B). In NSS/SLC6 transporters, TM3 is not directly involved in substrate or sodium ion binding. However, structural data on bacterial homologues have revealed that together with TM8 that harbours residues contributing to the Na2 sodium ion-binding site, TM3 undergoes conformational changes during the transport cycle . N-glycosylation sites in EL2 are not conserved across all members of the SLC6 family (Figure 1A) consistent with the fact that individual SLC6 transporters seem to be differently affected by N-glycosylation [25–28,30]. For instance, mutations of two of the three N-glycosylation sites in the GABA transporter GAT1 expressed in Xenopus oocytes yielded a reduction of turnover rates and significant changes in affinity to sodium , whereas the human dopamine transporter (DAT) showed reduced inhibitor affinity when N-glycosylation was altered .
Sequence alignment and homology model of the NSS family transporter BGT-1
In the present study, we present biochemical and functional data describing the role of the two predicted N-glycosylation sites in canine BGT-1. We observe that functional BGT-1 requires at least one of these sites (either Asn171 or Asn183) to be associated with N-glycans, whereas in the other site (either Asn171 or Asn183) the effect of N-glycosylation has to be mimicked by an aspartate, at a minimum. Our data further suggest that the increased PM insertion of BGT-1 upon a hyperosmotic shock requires the association of Asn183 with N-glycans.
MATERIALS AND METHODS
DNA constructs and cRNA synthesis for Xenopus oocytes
Wild-type (WT) cDNA of canine BGT-1 was cloned into pTLN Vector (gift from Professor Bamberg, Department of Biophysical Chemistry, Max Planck Institute of Biophysics, Frankfurt) with XbaI and XhoI restriction sites.
The QuikChangeTM kit (Stratagene), in combination with Pfu Turbo DNA polymerase, was used for the insertion of the desired mutations, N171D, N183D and NN171/183DD in the pTLN-BGT-1 plasmid. All the plasmids were fully sequenced and the specific mutations were confirmed. WT and mutants of pTLN-BGT-1 were linearized using MluI. Linearized DNA was further purified with High Pure PCR Product Purification Kit (Roche) and concentrated up to 0.17 μg/μl for in vitro RNA-synthesis using the mMESSAGE mMACHINE SP6 Kit (Life Technologies).
Expression and two-electrode voltage clamp analysis of WT and mutants in Xenopus laevis oocytes
A standard oocyte Ringer solution (ORi) was used for oocyte preparation, storage and for the electrophysiology measurements. ORi contained (in mM): 110 NaCl, 3 KCl, 2 CaCl2, 5 HEPES/Tris, adjusted to pH 7.5. GABA was added to ORi in the following concentrations: 0.01, 0.025, 0.05, 0.1, 0.25, 0.5 and 1 mM, adjusted to pH 7.5. All chemicals were purchased from Sigma–Aldrich (Taufkirchen).
Oocyte preparation and storage
Stage V and VI oocytes from Xenopus laevis (Nasco) were separated by an overnight treatment with collagenase (Typ CLS II, Biochrom), with subsequent washings in calcium-free ORi and maintained at 16°C–18°C in ORi containing again a calcium concentration of 2 mM. One day after removal from the frog, oocytes were injected either with 23 nl of 1 mg/2 ml of tunicamycin (AppliChem) dissolved in ORi or 23 nl of ORi alone approximately 60 min prior to injection of cRNA coding either for the WT BGT-1 or the mutants. An equivalent amount of ORi was injected as a control (mocks). Tunicamycin inhibits enzymes involved in the first steps of N-linked glycoprotein synthesis in the endoplasmic reticulum (ER). Oocytes were maintained at 16°C–18°C in ORi supplemented with 50 μM gentamycin and 2.5 mM sodium pyruvate, with daily washings and discarding of damaged oocytes.
These studies were carried out 3–4 days after cRNA and tunicamycin injection at room temperature. Oocytes were placed into a 0.5 ml chamber on the stage of a microscope and impaled under direct view with borosilicate glass microelectrodes filled with 3 M KCl (BioMedical Instruments). Current recordings at −60 mV were performed using a two-electrode voltage clamp device (OC725A, Warner) in the voltage clamp mode.
Statistics and calculations
Data are provided as means ± S.E.M. Paired Student's t test was used to show statistically significant difference of the GABA-associated currents in the absence and presence of PNGaseF. Statistical significance was set at P<0.05 (*). Michaelis–Menten constants (KM) for GABA in the absence and presence of tunicamycin were determined by SigmaPlot software (Systat Software) using the Michaelis–Menten equation I=Imax·[S]/(KM+[S]), where I is the current, Imax is the maximum current observed at saturating substrate concentrations, KM is the substrate concentration at half-maximal current and S is the substrate concentration.
Preparation of membranes was carried out as described previously . The lysates were separated on SDS/PAGE (12.5% gel) and then electro-transferred on to PVDF-membranes which were previously activated by methanol. The membrane was blocked with 5% milk powder for 1 h at room temperature and then incubated over night at 4°C with affinity-purified rabbit polyclonal antibody to dog BGT-1 (Proteintech Group), diluted 1:1000 in 0.5% milk powder, followed by a 2 h incubation with affinity-purified polyclonal antibody to rabbit coupled to alkaline phosphatase, diluted 1:1000 in 0.5% milk powder.
Fractionation of oocytes membranes
Fractionation was carried out according to the protocol from Broer  with minor changes. Briefly, 80 oocytes of WT, 100 oocytes of WT+Tun and WT+P and 150 oocytes of NN171/183DD were homogenized in 1, 1.5 and 2 ml ‘homogenization buffer 2’ (in mM): 320 sucrose, 50 Tris, 1 EDTA, 1 Pefabloc, adjusted to pH 7.5 respectively by pipetting up and down. The suspension was centrifuged twice at 1000 g for 10 min at 4°C. The supernatants were transferred on a sucrose gradient: 2 ml [2 M], 3.2 ml [1.3 M], 3.2 ml [1 M], 2 ml [0.6 M] and centrifuged in a SW40 rotor (Beckman Coulter) at 125000 g for 4 h at 4°C. Sucrose solutions were prepared in ‘TE-buffer’ (in mM): 50 Tris, 1 EDTA, 5 MgCl2, adjusted pH 7.5). Fractions (1 ml) were collected from the bottom, diluted 4-fold with 150 mM sucrose in ‘TE-buffer’ and centrifuged at 125000 g in a Ti70 (Beckman) rotor for 2 h at 4°C. Pellets were resuspended in 15 μl of SDS/PAGE sample buffer for electrophoresis and Western blotting as described above. The rough ER (rER) is detected in fractions 2 and 3, the PM in fraction 5 and the trans-Golgi network (TGN) in fractions 9 and 10 .
Fixation and determination of cell surface expression by immunogold-labelling
Post-embedding immunogold labelling and EM followed the method of Haase  and Lörinczi et al. , except that X. laevis oocytes were treated with 4% paraformaldehyde (PFA) in ORi solution, pH 7.5, for 3 h. The antibodies used for immunogold labelling of thin sections were the primary anti-BGT-1 (dog; Proteintech Group) and the second against rabbit coupled to gold particles (diameters 10–12 nm, both diluted 1:100 in PBS supplemented with 0.1% BSA) for visualization.
Cell culture of MDCK and HEK cells and [3H]GABA transport assays
MDCK cells (CCL-34, from A.T.C.C.) and HEK (human embryonic kidney) cells were used as described previously [36,37]. Na+-dependent [3H]GABA (Moravek Biochemicals and Radiochemicals) uptake was determined in six-well plates according to Forrest and Rudnick . Enzymatic de-glycosylation with PNGase F [peptide-N4-(N-acetyl-/3-glucosaminyl)asparagine amidase F] was carried out by adding 10 units/mL PNGase F (New England BioLabs), an enzyme catalysing the complete removal of N-glycan chains from glycoproteins to the isotonic and hypertonic sodium medium respectively, and incubating for 6 h at 37°C prior to measuring. MDCK cells were transiently transfected using GeneJammer (Stratagene) according to the manufacturer's instructions. GraphPad Prism version 5.0c for Mac OS X, GraphPad software  was used for the kinetic constants which were derived by Michaelis–Menten curve fitting of the uptakes rates compared with the substrate concentration. Data are means ± S.D. of at least three separate experiments. In each transport experiment, the mean value was derived from triplicate determinations. Where appropriate, different groups were compared by ANOVA and Tukey's test for multiple comparisons, using GraphPad Prism version 5.0c for Mac OS X, GraphPad software . A probability of P<0.05 was considered statistically significant.
Fluorescence microscopy and Western blotting
Fluorescence microscopy and Western blotting of cell lysates was carried out according to Kempson and Koepsell .
Sequence alignment of EL2 and structural modelling of human BGT-1
The TM of BGT-1 in an outward-occluded state was modelled using the X-ray structure of Drosophila melanogaster dopamine transporter (dmDAT) as template (PDB entry 4M48). The sequence alignment of human and dog BGT-1, human GAT1, GAT2, GAT3, human TauT, human DAT and the dmDAT construct used for X-ray crystallography, was obtained using T-Coffee v10 . The alignment was manually refined in EL2, residues (160–178), to match the predicted N-glycosylation sites (Figure 1A). The sequence identity between dmDAT and human BGT-1 in this alignment is 47.2%. Based on this alignment, a set of 2000 structural models of human BGT-1 was generated using Modeller 9v2  and the selected model was that with the best agreement with the Ramachandran plot according to Procheck  taken from the 10 models with the lowest Modeller molpdf scores. The final BGT-1 model has zero residues in disallowed regions of the Ramachandran plot and just one (Arg413) in a generously allowed region; Arg413 is located in a loop segment.
Implications of N-glycosylation in BGT-1 from homology modelling
A multiple-sequence alignment of representative mammalian NSS transporters (SLC6 family) reveals at least six distinct N-glycosylation motifs in EL2, which we label according to their positions from 1 (N-terminal) to 6 (C-terminal) (Figure 1A). These sites are neither simultaneously conserved in all members of this family nor is there any obvious pattern in distribution or number of sites for different branches of the NSS family (Figure 1A). Focussing on the osmolyte branch, BGT-1 is predicted to have two N-glycosylation sites , 3 and 6, whereas the EL2 of other GABA transporters (GAT1, GAT2 and GAT3) includes the same sites 3 and 6, in addition to either site 4 or site 5. The taurine transporter TauT, which contains sites 3 and 6, also shares site 1 with the human dopamine transporter (hDAT).
In order to identify the positions of the N-glycosylation sites in EL2 in three dimensions, we constructed a homology model of human BGT-1 (Figures 1B and 1C), based on the recently published structure of the dmDAT . This structure revealed important information on the mechanism of anti-depressant binding and substrate inhibition. Although EL2 is truncated by 42 residues (Δ164–206), in dmDAT the structure shows important parts of the architecture of EL2 including the second N-glycosylation site at Asn183. Therefore, EL2 in BGT-1 could be modelled using the dmDAT structure as template. In fact, only nine residues (F169–V178) in BGT-1 have no template in dmDAT (Figure 1). The remaining nine residues of BGT-1 were modelled using Modeller v9.2 . In the resulting model, EL2 covers a large area of the extracellular surface of BGT-1, suggesting an interaction with EL4. Whereas Asn183 is in close proximity to a well-ordered helical segment, Asn171 is located in the long segment of EL2 that lacks notable secondary structure (Figure 1C), suggesting that N-glycan association may have different effects at each of these locations. To investigate whether these sites are indeed functionally distinguishable, we carried out a systematic study of the roles of both predicted N-glycosylation sites in transport and regulation of BGT-1.
Regulation of BGT-1 mediated GABA transport in Xenopus laevis oocytes
We performed localization and transport measurements of chemically, enzymatically and/or by mutagenesis de-glycosylated canine BGT-1 in X. laevis oocytes. This in vivo system was chosen to allow our results to be compared directly to those of other NSS transporters [25–30] as it is known that the transcription and translation machineries affect transporter properties significantly. However, Xenopus laevis oocytes did not survive hypertonic conditions compared with those leading to regulated PM insertion observed in MDCK cells.
This became obvious when comparing the GABA-induced currents measured under hypertonic conditions relative to those under isotonic conditions. Oocytes were superfused with 10 mM GABA (Figure 2, black bar, 10 mM GABA), which led to inward currents in oocytes expressing BGT-1 (Figure 2B), but not in water-injected control oocytes (Figure 2A). After recovery from the effects of GABA, already hypertonic conditions (455 mOsM; Figure 2, black bar; +220 mM sucrose) led to inward currents not only in BGT-1-expressing oocytes (Figure 2B) but also in water-injected control oocytes (Figure 2A) and subsequently no GABA-mediated currents in BGT-1-expressing oocytes were observed (Figure 2B, black bars; +220 mM sucrose and 10 mM GABA). Most of the oocytes did not survive long-term exposure (24 h) to hypertonic conditions and therefore no osmotic-stress regulated insertion could be detected. In the following, therefore, the role of N-glycosylation of BGT-1 expressed in oocytes was investigated exclusively under isotonic (ORi, 235 mOsM) conditions and the data are discussed only in the context of trafficking and transport, not in the context of regulation. The regulatory role of N-glycosylation was considered using measurements performed in MDCK cells (see below).
Response of BGT-1 to GABA in the presence of hypertonic conditions and membrane distribution of N-glycosylated and de-glycosylated BGT-1 in X. oocytes
Trafficking and localization of glycosylated and de-glycosylated BGT-1 in Xenopus laevis oocytes
We investigated the impact of N-glycosylation on the subcellular distribution of BGT-1 in oocytes under isotonic conditions (235 mOsM), focusing on the rER, the TGN and the PM by Western blot. Although the accuracy of a Western blot does not allow for a quantitative analysis, the changes in intensity for similar oocyte numbers (∼150 oocytes) are strong enough for a semi-quantitative statement. After 3 days expression, N-glycosylated BGT-1 is detected at ∼70 kDa on the Western blot (Figure 2C, WT) and appears to be similarly distributed in all three compartments. Nevertheless, the major fraction is found in the PM (Figure 2C, WT, PM). The stability and abundance of PM-inserted BGT-1 was determined after enzymatically removing any surface-exposed N-glycans by applying PNGase F to the extracellular solution. The loss of N-glycans after PM insertion did not affect the amount of BGT-1 found in the rER (Figure 2D, WT+P, rER), but led to a slight increase in the TGN fraction (Figure 2D, WT+P, TGN). The de-glycosylated BGT-1 isoform (∼60 kDa) is still observed in the PM without any apparent degradation (Figure 2D, WT+P, PM), implying that the BGT-1 fraction in the PM is stable without N-glycans, although a partial depletion and internalization is caused by their removal.
Mutagenesis of N-glycosylation sites in BGT-1
Treatment of oocytes with PNGase F reveals the de-glycosylated BGT-1-WT form at ∼60 kDa (Figure 3A, WT, +) and a fully glycosylated form at ∼70 kDa on the Western blot in the absence of PNGase F (Figure 3A, WT). It seems that N-glycosylation is less efficient in oocytes accounting for the remaining fraction of de-glycosylated WT protein.
Expression and distribution of BGT-1 and N-glycosylation site mutants in oocytes
Putative N-glycosylation sites were modified by replacing asparagine residues with aspartate, alanine or valine, either individually (N171D, N183D, N171A, N171V, N183V) or in combination (NN171/183DD). When substituted by aspartate individually (N171D, N183D), a form of BGT-1 is still observed at the same molecular mass as the WT, whereas the alanine mutant (N171A) and two valine mutants (N171V, N183V) all run at lower molecular mass (Figure 3A). N171D and N183D show a similar band shift to the WT after PNGase F treatment (Figure 3A; N171D, N183D, +). Within the accuracy limit of the method it appears that a similar amount of N-glycans is attached to the remaining site as in the WT. For N183D, the glycosylated form is expressed to a lesser extent than that of WT and N171D. The double mutant, NN171/183DD, does not change electrophoretic mobility upon PNGase F treatment (Figure 3A, NNDD, +), representing a fully un-glycosylated form of BGT-1. Both N171A and N171V show band shifts upon PNGase F treatment (Figure 3A; N171A, +; N171V, +) confirming partial N-glycosylation. These two mutations (N171A, N171V) seem to alter the architecture of EL2 in a way that N-glycosylation of the remaining Asn183 is affected as shown by the lower molecular mass prior to PNGase F treatment. N183V behaves similarly, to the double mutant (NN171/183DD) and does not show a band shift upon PNGase F treatment (Figure 3A, N183V, +) suggesting that Asn171 is completely inaccessible without N-glycosylation of Asn183 or the mimicking effect of N-glycosylation (via aspartate) at this position.
NN171/183DD is enriched in the rER and undergoes significant degradation in the TGN (Figure 3B, arrow), perhaps due to a slower targeting to the PM and mis-folding. Cell surface expression was also assayed by immunogold-labelling of thin-sectioned Xenopus oocytes using a BGT-1-specific antibody (Figure 3C, WT, 30±5 gold-labelled BGT-1 molecules). Judged from the electron micrographs of thin sections the amount of N171D (21±3), N183D (15±2) and NN171/183DD (9±1) inserted into the PM is slightly reduced, whereas it is significantly reduced for the double mutant (Figure 3C, NN171/183DD). The remaining fraction of NN171/183DD in the PM does not show degradation (Figure 3B, PM). The reduced PM localization might reflect a disabling effect on PM targeting due to the NN171/183DD double mutation.
We conclude that for proper folding to occur, only one of the sites has to be glycosylated, as long as the remaining site is replaced by aspartate. The amount of BGT-1 trafficked to and inserted into the PM however depends on the presence of N-glycans at Asn183 indicated by the difference in expression levels for N171D and N183D.
GABA transport by BGT-1 and mutants in Xenopus laevis oocytes
Functional analysis of N-glycosylated and de-glycosylated forms of BGT-1 was carried out. GABA transport by BGT-1-WT and mutants was investigated by two-electrode voltage clamp. Apparent KM-values were determined at a holding potential of −60 mV (Table 1). The KM of BGT-1-WT for its substrate GABA was 11.7±0.4 μM, which is in good agreement with a previous report . N183D exhibited an apparent KM value of 9.5±1.2 μM for GABA, which is very close to that of BGT-1-WT. The maximal inducible current of N183D is reduced by a factor of 2 relative to WT, reflecting the lower protein concentration in the PM. NN171/183DD substrate-associated currents were below detection limits (Table 1). Interestingly, the N-glycosylated N171D mutant has a significantly (5-fold) higher GABA affinity even than BGT-1-WT (Table 1, N171D).
|< tex − math/> KM, −60 mV [μM]||ΔImax, −60 mV [nA]|
|< tex − math/> KM, −60 mV [μM]||ΔImax, −60 mV [nA]|
Below detection (BD) limit
The GABA-induced currents of BGT-1-WT were reduced by 80% at −60 mV when N-glycosylation was suppressed by tunicamycin (Table 1, WT+Tun). The apparent KM-value also increased nearly 20-fold (Table 1, WT+Tun). Due to the apoptotic side effects of tunicamycin on all cellular components, these measurements have to be considered with caution. Indeed, compared with N-glycosylated BGT-1-WT only a 3-fold reduction in the apparent KM for GABA was observed when N-glycans were removed by PNGase F after the protein was inserted into the PM (Table 1, WT+P). Therefore, either chemical or enzymatic removal of N-glycans of BGT-1-WT decreases the affinity of BGT-1 for GABA and decreases GABA transport rate, whereas substitution of the N-glycosylation sites by negatively charged residues does not alter affinity and transport kinetics significantly. However, there might also be the possibility that N-glycosylation is indirectly affecting BGT-1 via the action of glycosylated ancillary proteins, although to date there is no indication of any interaction of BGT-1 once inserted in the PM. In addition, our in vitro studies in membrane vesicles (result not shown) have not indicated the necessity of additional interaction partners to facilitate transport. Substitution by either alanine or valine dramatically decreases substrate affinity (Table 1; N171A, N171V, N183V). Association with N-glycans at Asn183 appears not to be crucial as long as the site is mimicked by aspartate, considering the nearly identical affinities of WT and N183D mutant (Table 1). Interestingly, the substitution of Asn171 by aspartate results in a significant increase in substrate affinity (Table 1). In summary, the effect on transport of N-glycan association at the two putative N-glycosylation sites is the opposite of that observed for PM insertion, i.e. N-glycan association with Asn183 is important for insertion whereas association with Asn171 determines transport properties.
Role of N-glycosylation in BGT-1 expressed in MDCK cells
Given that N-glycosylation of BGT-1 at Asn183 is important for PM insertion, we asked whether osmotic-stress-regulated insertion is also affected. Osmotic stress regulation of plasma membrane insertion can only be observed in MDCK cells and therefore, we repeated the key experiments performed in oocytes in MDCK cells. MDCK cells contain endogenous BGT-1 . Total N-glycosylation of Asn171 (site 3, Figure 1) and Asn183 (site 6, Figure 1) were investigated both for the endogenous form (Figure 4A) as well as for BGT-1 fused with a N-terminal 27 kDa EGFP-tag and expressed in addition to the endogenous form (Figure 4B). Unlike oocytes, MDCK cells were able to survive hypertonic conditions and therefore BGT-1 PM insertion was induced in hypertonic medium (500 mOsM; Figures 4C and 4D, Hyp) to up-regulate substrate transport . Endogenous BGT-1 shows an electrophoretic mobility of ∼90 kDa on the Western blot (Figure 4A), whereas the PNGase F-treated cells reveal de-glycosylated endogenous protein running at ∼55 kDa (Figure 4A, WTend+P). The N-glycans in MDCK cells seem to be more complex (larger) compared with the N-glycans in oocytes. The de-glycosylated EGFP-tagged BGT-1 was detected at a molecular mass of ∼90 kDa accounting for the molecular mass of the 27 kDa EGFP tag, whereas the glycosylated form runs at ∼120 kDa (Figure 4B). This corresponds to a comparable shift of ∼40±5 kDa from glycosylated to de-glycosylated both for endogenous and EGFP-tagged BGT-1 in MDCK cells (Figure 4A).
Membrane distribution of de-glycosylated BGT-1 in MDCK cells
In contrast with oocytes, which seem to withstand the apoptotic action of tunicamycin to some extent, MDCK cells were detrimentally affected by tunicamycin, especially under hyperosmotic conditions. Consequently, to assess the effect of N-glycan association on osmo-regulated insertion and withdrawing of N-glycan-depleted BGT-1 from the PM of MDCK cells, we treated them with PNGase F (Figures 4A and 4B,+P and 4D, Hyp+P), but not with tunicamycin. After enzymatic removal of the N-glycans, de-glycosylated BGT-1 remains mainly in the PM (Figure 4D, Hyp+P), suggesting that removal of N-glycans does not trigger depletion of PM-inserted BGT-1. That is, similar to the observation in oocytes, the removal of N-glycans from BGT-1 after insertion into the PM of MDCK cells does not seem to affect the amount and stability of the protein in the PM. However, when PNGase F-treated MDCK cells were exposed to isotonic medium after a hyperosmotic shock, de-glycosylated BGT-1 remained mainly in the PM (Figure 4E, Hyp+Precovery), whereas glycosylated BGT-1-WT is directly depleted from the PM after switching from hyperosmotic conditions to isotonic conditions (Figure 4E, Hyprecovery). This result is the first indication that N-glycosylation of BGT-1 in MDCK cells is involved in regulated PM depletion.
N183D prevents regulation in MDCK cells
Firstly, the effect of N-glycosylation on transport was investigated in MDCK cells by measuring [3H]GABA uptake by endogenous BGT-1 and KM-values were determined (Figure 5). The affinity for GABA was 41.6±23.7 (Table 2). After PNGase F treatment (Figure 5, open symbols), the apparent KM-values increased by a factor of 1.2 (Table 2, BGT-1-WT+P) indicating a smaller effect of N-glycosylation on affinity than observed in oocytes (Table 1, WT, WT+P; KM values increased by a factor of 2.5). The transport rates were reduced by a factor of 2 by PNGase F treatment (Table 2, BGT-1-WT+P), showing a similar trend to the reduction in maximal inducible currents measured in oocytes (Table 1). As MDCK cells express BGT-1-WT endogenously, mutants were measured in HEK cells. Transport kinetics of both aspartic acid mutants (N171D, N183D; Table 2) further supports the localization studies observed for these two mutants in MDCK cells under iso- and hyper-tonic conditions (Figure 6). The apparent KM-value of N171D slightly decreased (Table 2) and a KM value for Asn183 was only measurable under isotonic conditions (Table 2) with a similar KM value as observed for the BGT-1-WT under hypertonic conditions in MDCK cells. Transport rates were reduced by a factor of 2 for N171D but were only slightly affected for N183D. The quantitative differences between oocytes and MDCK measurements can be attributed to the different techniques (inward current compared with radiotracer uptake).
Activity of glycosylated BGT-1 and de-glycosylated BGT-1 in MDCK cells
Expression and distribution of N171D and N183D under iso- and hypertonic conditions in MDCK cells
|KM (μM)||Vmax (pmol/mg/min)|
|KM (μM)||Vmax (pmol/mg/min)|
It can be concluded that GABA transport is only slightly affected by N-glycosylation in both oocytes and MDCK cells, when at least one of the two sites (Asn171, Asn183) can be glycosylated or to some extent mimicked by aspartate. In oocytes, the data suggest that the mutation of Asn183 does affect trafficking and thereby PM insertion (Figure 3C), but not critically transport, when substituted by aspartate. Therefore, we investigated N183D with respect to regulatory expression and PM insertion under both isotonic and hypertonic conditions in MDCK cells (Figures 6A and 6B). The amount of expressed and hence PM inserted N183D is not up-regulated, but instead is strongly down-regulated under hypertonic conditions (Figures 6A and 6B, Hyp). In contrast, both N-glycosylated BGT-1-WT and N171D show strong up-regulation of PM insertion (Figures 4C and 4D, Hyp; and 6C and 6D). The amount of PM inserted N183D is significantly reduced under hypertonic conditions compared with isotonic conditions (Figures 6A and 6B) indicating that this site has an important role in the regulation mechanism in MDCK cells. The pronounced change in PM abundance of N183D indicates furthermore that the association of N-glycans with Asn171 in EL2 is a regulatory parameter during both trafficking and PM insertion respectively.
N-glycosylation is a sophisticated modification of protein structures by which cells control dynamics in conformational space. Domains can be stabilized after translation and during trafficking to fulfil specific requirements of their target locations.
The role of N-glycosylation in trafficking and transport was already described for other SLC6 transporters, too. However, the emerging picture is not entirely conclusive. Loss of N-glycans affected trafficking and PM insertion differently and in some cases decreases the substrate affinity [25–28,30]. A deficiency in N-glycan association with the conserved site 3 was reported to reduce uptake rates in the NSS transporters GAT1 , the creatine transporter (CRT) , the norepinephrine transporter (NET)  and the serotonine transporter (SERT) , which was mainly attributed to a reduction in inserted protein into the PM. On the other hand, the lack of N-glycans at site 5 has been shown to affect protein stability in DAT and GAT1 [26,27]. In a structural context, these data confirm that the architecture of EL2 in SLC6 transporters is crucial for proper targeting and substrate binding.
The situation in BGT-1 is more complex as we now draw a link between N-glycosylation and regulated PM insertion upon osmotic shock. The mutagenesis data point to a regulatory role of the external loop EL2 involving changes in its conformation and thereby affecting insertion and transport of BGT-1. The most surprising result of our study was the exclusive role of Asn183 in EL2 for regulated insertion and depletion (Figures 6A and 6B, Hyp; and 4E). We propose three steps in BGT-1 regulation via N-glycosylation: first, PM insertion and depletion as a function of the amount of added N-glycans; second, increased substrate affinity; and third regulated post-translational modification upon osmotic upshift. In the following text, we will discuss each step in the light of the obtained mutagenesis data and conformational changes in EL2 in the presence and absence of associated N-glycans.
Substitution of asparagine by aspartate for both N-glycosylation sites individually affected only trafficking (Figure 3C), but not transport properties of BGT-1. As a matter of fact, the double mutant did not show any measurable currents. However, this does not automatically mean that the double mutant is inactive, in fact when NN171/183DD was expressed in Pichia pastoris cells, it facilitated significant uptake of radioactive-labelled GABA (result not shown). It might very well be that the stoichiometry of transport was altered in a way that no currents could be detected by two-electrode voltage clamp. Anyway, we conclude that both PM insertion and transport properties are affected when both sites are replaced by aspartate. However, introduced charges at these sites mimic the effect of N-glycosylation to some extent. There are several charged residues in EL2, which could provide the possibility for salt bridges and ionic interactions. The individually-introduced aspartates result in mutants that show comparable substrate transport and affinity to BGT-1-WT and the amounts of N-glycans associated is not strongly affected, when one site is missing (Table 1). The presence of aspartate at Asn183 might alter the conformation of EL2 in a way that N-glycans can still attach to Asn171, whereas this site is not fully accessible when Asn183 is substituted to valine (Figure 3A). Substitution by neutral amino acids had dramatic effects on transport and affinity (Table 1). One could assume that the association of N-glycans, bulky by nature, with EL2, influences the conformation and the flexibility of the loop. In return it can be assumed that the flexibility of the loop itself is a parameter in conformational cycling during transport. In fact, EL2 is located close to the scaffold-bundle interface and also to the vestibule leading to the substrate-binding site (Figure 1B). The presence of bulky N-glycans could therefore block the passage of the substrate or shift the alternating-access equilibrium by populating one conformation more than the other or even by controlling the rate of the conformational change. The decrease in transport rate observed in the present study, when one of the N-glycosylation sites was substituted by alanine or valine, could be explained by a change in the conformation of EL2 upon mutation, to one that is not suitable for transport. Similarly, introducing an aspartate at the conserved position 171 could lead to a conformation of EL2 that is more favourable for transport, yielding a mutant with five times higher apparent substrate affinity (Table 1, N171D).
That the architecture of EL2 could be modified by post-translational modifications such as N-glycosylation or mutations is consistent with earlier observations. For example, the structure of dmDAT reveals a disulfide bond in EL2, formed by cysteine residues that are conserved in most of the transporters in this subfamily (hBGT1, GAT1-3, TauT, DAT and even SERT). In the case of SERT, the disulfide bond formation is crucial for obtaining a functional conformation of EL2 . In the case of human DAT, the conformational state of the entire transporter can be governed by the co-ordination of a zinc ion to EL2 . Moreover, the model of BGT-1 shows a putative interaction of EL2 with other ELs, e.g. with EL4 (Figure 1C), which is known to play a key role in conformational changes in the SLC6 family . Specifically, EL4 was shown to be part of an extracellular gate involved in the conformational change from outside to inside open conformation, as well as being part of the extracellular vestibule that accommodates the inhibitors [46–50]. All together we conclude that for substrate transport mediated by BGT-1 the presence of N-glycans at either site is not essential as long as EL2 can adopt a certain conformation.
The situation is changed when it comes to the regulatory PM insertion during hypertonic stress. Here, the regulatory properties strongly depend on mature N-glycosylation at Asn183 and cannot be compensated by an introduced charged residue at Asn183 or even by N-glycans at the remaining Asn171 (Figure 6).
Homology modelling of BGT-1 suggests that Asn183 is located in the middle of a random-coil segment (most probably a relatively flexible region) and is clearly involved in regulation (Figures 1B and 1C), whereas the more conserved site 3, Asn171, located close to a helical segment and presumed to be a less flexible region, appears primarily to be important for substrate affinity in BGT-1. The number of N-glycans associated at this site might even trigger regulated insertion. We base this assumption on the fact that in oocytes, BGT-1 insertion is not regulated under hyperosmotic conditions (Figure 2B) and we detect only a small amount of N-glycans linked to BGT-1 (Figures 2D and 3A), whereas in MDCK cells, in which PM insertion is regulated, the association with N-glycans appears to be more complex, resulting in an additional mass of ∼35 kDa (Figure 4A).
It is interesting to note that within the SLC6 family, BGT-1 has evolved an EL2 sequence quite distinct from other family members (Figure 1A). However, the mechanism of osmotic stress-dependent PM insertion is not only observed in BGT-1. Kempson  identified the amino acid transport system A and Yorek et al.  the myo-inositol transporter SMIT as being regulated under hypertonic conditions in MDCK and thick ascending limb of Henle's loop (TALH) cells respectively. System A transport activity is increased immediately after switching MDCK cells to hypertonic medium whereas the activation response of BGT-1 occurs after 24 h and coincides with down-regulation of system A . SMIT shows a comparable regulatory time pattern to BGT-1. The amount of both transporters on the PM is increased after exposing the cells to hypertonic medium and they are both depleted from the PM under isotonic conditions.
Interestingly, aspartic acid-linked N-glycosylation sites are located in the EL3 of SMIT, which due to the topology shift of one TM helix occurring between the SLC5 and SLC6 family, would correspond exactly to EL2 in BGT-1. However, to date no data are available concerning N-glycosylation of SMIT1, except PNGase F assays for SMIT2, a sodium-coupled myo-inositol transporter mainly found in the cortex . The latter showed no reduction in molecular mass upon PNGase F treatment suggesting that SMIT2 does not bind N-glycans . We suggest that N-glycosylation might be a regulatory parameter during trafficking under hypertonic conditions for both transporters.
It is now an intriguing question if other SLC6/5 transporters also exploit N-glycosylation in EL2 for some sort of regulation. The very distinctive role of N-glycosylation in BGT-1 might draw attention to the non-conserved regions in EL2, initiating more detailed investigations into this direction of these medically important transporters in future.
Eva Schweikhard performed all experiments with the exception of the thin sectioning, which was carried out by Friederike Joos and the two electrode voltage clamp performed by Birgitta Burckhardt. Cristina Fenollar-Ferrer carried out computational work. Lucy Forrest supervised computational work. Stephen Kemson supervised MDCK measurements. Christine Ziegler designed the research. Christine Ziegler and Eva Schweikhard analysed data and wrote the manuscript. All authors commented on the manuscript.
We thank Manuel Palacín and Baruch Kanner for suggestions and experimental advice, Caroline Koshy, Stefan Köster, Sabrina Schulze, Javier Carrera-Casanova and Ahmadreza Mehdipour for helpful discussions.
This work was supported by the German Research Foundation [grant number DFG ZI-5-2]; the Collaborative Research Center 807 “Transport and Communication across Biological Membranes” (to L.R.F. and C.Z.); the Collaborative Research Center 699 “Strukturelle, physiologische und molekulare Grundlagen der Nierenfunktion” (to E.S. and C.Z.); and by the Division of Intramural Research of the NIH, National Institute of Neurological Disorders and Stroke (L.R.F.).
betaine/GABA transporter 1
Drosophila melanogaster dopamine transporter
human embryonic kidney293
Madin–Darby canine kidney
oocyte Ringer solution
- PNGase F
peptide-N4-(N-acetyl-/3-glucosaminyl)asparagine amidase F
serum- and glucocorticoid-inducible kinase
solute carrier 6
thick ascending limb of Henle's loop
tonicity enhancer-binding protein
Current address: Computational Structural Biology Section, Porter Neuroscience Research Center, National Institutes of Neurological Disorders and Stroke, National Institutes of Health, Bethesda MD 20892, U.S.A.