N-linked glycosylation sites of the motor protein prestin: effects on membrane targeting and electrophysiological function

Authors


Address correspondence and reprint requests to Dr Peter Dallos, Northwestern University, Frances Searle Building, 2240 Campus Drive, Evanston, IL 60208 USA. E-mail: p-dallos@northwestern.edu

Abstract

Prestin is a motor protein of outer hair cells (OHC) that plays a crucial role in mammalian hearing. Prestin is a putative N-glycoprotein with three potential N-linked glycosylation sites. It is not known whether glycosylation affects the function and activity of prestin. Therefore, the effects of N-glycosylation were investigated by producing single-point (N163Q and N166Q) or double-point mutations (NN163/166QQ and NN163/166AA) at putative N-glycosylation sites. Further, treatment with tunicamycin or glycopeptidase-F was used to determine the consequences of removing N-linked glycosylation in wild-type prestin. We determined the effects of these manipulations on prestin's cell surface expression, molecular mass, glycosylation pattern, and electrophysiological properties in different cell-types. Data indicate that prestin is a glycoprotein with N-linked glycosylation sites at N163 and N166. N163 and N166 may have differential programs for synthesis and trimming of the glycans. The N166 site appears to have greater extent of glycosylation than its companion. N-linked glycosylation is not required for plasma membrane targeting of prestin. Both glycosylated and deglycosylated prestin demonstrate non-linear capacitance, a signature of prestin's motor function. Compared to glycosylated prestin, the fully de-glycosylated protein has altered electrophysiological function, with a change in membrane potential at most effective charge transfer to more depolarized values. These data suggest that glycosylation of prestin may quantitatively affect OHC electromotility.

Abbreviations used
CHO

Chinese hamster ovary cells

GFP

Green fluorescent protein

GHRHR

growth hormone releasing hormone receptor

gPrestin

gerbil prestin

IHC

inner hair cell

mPrestin

mouse prestin

NLC

non-linear capacitance

OHC

outer hair cells

OK cells

opossum kidney cells

SLC26

Solute Carrier Family 26

Outer hair cells (OHC) in the mammalian organ of Corti possess a unique feature called ‘electromotility’. The OHC can rapidly alter its length (Brownell et al. 1985) and stiffness (He and Dallos 1999) when its transmembrane voltage is changed. This electromechanical process is assumed to play a critical role in frequency resolution and signal amplification in the mammalian cochlea (Dallos 1992). A widely accepted explanation for the cellular mechanism of electromotility is that an unusual ‘motor protein’, located in the cylindrical OHC's lateral membrane, changes conformation when the cell's membrane potential is modulated, leading to a change in OHC length. We recently identified the gene Prestin from an inner hair cell (IHC) subtracted OHC cDNA library that codes for the motor protein prestin (Zheng et al. 2000). Prestin is comprised of 744 amino acids and expressed in the lateral plasma membrane of OHCs. Prestin-expressing mammalian cells change their shape upon alteration of their membrane potential and show non-linear capacitance and electro-mechanical reciprocity, not unlike OHCs (Zheng et al. 2000; Ludwig et al. 2001; Santos-Sacchi et al. 2001). The expression of prestin in OHCs progressively increases from birth and reaches mature levels at 12 days (rat and mouse) to 20 days (gerbil) after birth (Zheng et al. 2000; Belyantseva et al. 2000), similar to the development of electromotility (He et al. 1994). Examination of a recently available prestin knockout mouse demonstrated that OHCs from prestin-null mice lack electromotility and the animals show 40–60 dB loss of hearing sensitivity (Liberman et al. 2002). Collectively, these results demonstrate that prestin is the motor protein of cochlear OHCs.

Prestin is a member of an anion transporter protein family termed Solute Carrier Family 26 (SLC26), which has 11 identified members, SLC26A1 through SLC26A11 (Lohi et al. 2002; Vincourt et al. 2003). Prestin (SLC26A5) uniquely demonstrates non-linear charge transfer under voltage stimulation, manifested in non-linear capacitance (NLC), an accepted signature of OHC electromotility (Ashmore 1990). Other members of the SLC26 family, such as pendrin (SLC26A4) and PAT1 (SLC26A6), do not show NLC under similar conditions (Zheng et al. 2000; Oliver et al. 2001). The unique features of prestin make understanding the mechanism of its function compellingly interesting as well as potentially useful in nanotechnology applications. There are at least two putative functional domains in prestin: the ‘voltage sensor’ that detects changes in the transmembrane potential of the cell, and the ‘actuator’ that undergoes a conformational change and thereby facilitates cell contraction or elongation in response to depolarization and hyperpolarization, respectively (Dallos and Fakler 2002). Prestin probably utilizes intracellular Cl ions as extrinsic voltage sensors for promoting charge translocation (Oliver et al. 2001). After binding to prestin, Cl anions are partially translocated across the membrane in response to changes in the transmembrane voltage. This translocation is assumed to trigger conformational changes in the molecule, resulting in an increased (decreased) membrane area, and elongation (contraction) of cells. Prestin therefore appears to behave as an incomplete anion transporter (Dallos and Fakler 2002).

According to computer model predictions, the amino acid sequence and conservation of relative locations in SLC26 proteins suggests several putative N-linked glycosylation sites. Table 1 lists the glycosylation sites (Asn-X-Ser/Thr) predicted by the PROSITE program (Bairoch and Apweiler 1997). X denotes any amino acid except proline. One to four potential N-linked glycosylation sites are consistently predicted in the second extracellular loop except in the newly discovered SLC26A11. The relative locations of the putative N-linked glycosylation sites in the second extracellular loop of the SLC26 family are highly conserved, which intimates a potentially significant role for glycosylation of these transporters.

Table 1.  Putative N-glycosylation sites in SLC26 family predicted by the PROSITE program
Gene nameAccession no.ReferencePotential N-glycosylation sites (Aa position)
  1. *The conserved N-glycosylation sites predicted in the second extracellular loop.

SLC26A1 (Sat-1)AF297659Lohi et al. 2000162*, 167*
SLC26A2 (DTD)U14528Hastbacka et al. 199420, 199*, 205*, 357
SLC26A3 (DRA, CLD)NM_00011Schweinfest et al. 1993153*, 161*, 164*, 165*, 621
SLC26A4 (PDS)AF030880Everett et al. 1997167*, 172*, 241
SLC26A5 (Prestin)AF230376Zheng et al. 2000163*, 166*, 603
SLC26A6 (PAT-1)NM_022911Lohi et al. 2000146*, 151*, 593
SLC26A7aAF331522Vincourt et al. 2002131*, 137*
SLC26A8 (Tat-1)AF331522Toure et al. 200157, 197*, 282, 389, 600, 656, 692, 693
SLC26A9AF331525Lohi et al. 2002130*, 136*
SLC26A11AJ544073Vincourt et al. 2003298

In general, N-glycans are important structural components that affect plasma membrane localization, protein folding, and the conformation and stability of proteins. Many biological activities, such as cell adhesion, migration, and ligand-specificity in receptors, are also affected by N-glycans. There is evidence suggesting that the down-regulated in adenoma (DRA, SLC26A3) and Tat1 (SLC26A8) are glycoproteins (Byeon et al. 1998; Toure et al. 2001). It is not clear which of the Asn residues actually constitute an N-linked glycosylation site, or what their role might be in relation to the anion transporter function of DRA and Tat1. There is also no information about post-transcriptional modification of prestin or the possible role of N-linked glycosylation in the function and activity of prestin. In this work, we have determined the number and location of N-linked glycosylation sites in prestin. Possible effects of N-linked glycosylation on prestin's function were studied with the aid of four N-linked glycosylation mutants, N163Q, N166Q, NN163/166QQ, and NN163/166AA, as well as with tunicamycin or glycopeptidase-F, two agents used to prevent or remove glycosylation, respectively. Physiological and structural features of these mutants were investigated, including the extent of glycosylation (changes in apparent mass) in different cell lines, prestin protein distribution pattern at the cellular level, and electrophysiological properties.

Experimental procedures

Antibodies

Anti-N-gPres polyclonal antibody was raised in two rabbits against a synthetic oligopeptide derived from the amino terminus of gerbil prestin (gPrestin: MDHAEENEIPVATQKYHVER). Anti-C-mPres (anti-COOH-mPres) polyclonal antibody was raised against a synthetic oligopeptide derived from the carboxy terminus of mouse prestin (mPrestin: TASLPQEDMEPNATPTTPEA). Anti-C-mPres serum was further affinity-purified using the same peptide. Anti-N-gPres and anti-C-mPres antibodies were used in a 1 : 2000 dilution, anti-V5 (Invitrogen, Carlsbad, CA, USA) and anti-Xpress (Invitrogen) antibodies were used in a 1 : 200 dilution in immunofluorescence experiments, and 1 : 5000 in western blot. Anti-Flag antibody (Sigma, St Louis, MO, USA) was used in 1 : 1000 dilution in the immunofluorescence experiments. FITC-conjugated anti-rabbit IgG, FITC and rhodamine conjugated anti-mouse IgG, donkey anti-rabbit IgG-HRP (horseradish peroxidase) and donkey anti-mouse IgG-HRP were purchased from Pierce (Rockford, IL, USA) or Jackson ImmunoResearch (West Grove, PA, USA).

DNA constructs and generation of sugar mutations of prestin

Gerbil prestin (gPrestin) was cloned into the vector pcDNA3.1 (Zheng et al. 2000). The glycosylation mutants were generated by replacing the Asn residue at positions 163 and/or 166 with Gln or Ala using a QuickChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). The nucleotide exchanges were confirmed by DNA sequencing. gPrestin was also cloned into pcDNA 6/V5 HisA (Invitrogen), which adds a V5 epitope at the COOH terminus and pcDNA3.1/HisB (Invitrogen), which places the Xpress epitope at the NH2 terminus of prestin (Zheng et al. 2001). The V5 and Xpress tags were detected by the anti-V5 and anti-Xpress antibodies, respectively. In addition, GFP (Green fluorescent protein) was fused to the rat (rPrestin-GFP) and gerbil (gPrestin-GFP) prestin C-termini. Another N-linked glycosylation mutant: NN163/166/AA was generated from rPrestin-GFP. An irrelevant expressed plasmid (CAT in pcDNA3.1) was used as a negative control. A Flag-growth hormone releasing hormone (GHRH) receptor construct was used as a positive membrane insertion control (DeAlmeida and Mayo 1998) (kindly provided by Dr K. Mayo).

Cell culture and transient transfection

Prestin and four N-linked glycosylation mutants were transiently expressed in TSA-201 cells, a subclone of human embryonic kidney 293 cells (Margolskee et al. 1993), OK (opossum kidney) cells, or CHO (Chinese hamster ovary) cells. For transient transfection, cells were cultured in 35-mm culture dishes with or without cover glasses. After 24 h incubation, the cells were transfected with the different prestin constructs using transfection reagents either Effectene (Qiagen, Valencia, CA, USA) or ExGen 500 (Fermentas, Hanover, MD, USA). For transfection with Effectene, we used 0.4 µg/35 mm dish of prestin cDNA, for using ExGen, we used 1 µg/35 mm dish of cDNA. In some experiments, the prestin construct was co-transfected with 1/10 the amount of a green fluorescence protein cDNA expressing plasmid pEGFP-N2 (Clontech), which was used as an independent marker for successful transfection of individual cells. After 24–48 h incubation, the transfected cells were used for electrophysiological measurements or immunofluorescence experiments.

Immunofluorescence experiments

The transiently transfected cells were fixed with 1% formaldehyde in phosphate-buffered saline for 10 min at room temperature. The cells were incubated with phosphate-buffered saline containing anti-N-prestin or anti-Flag antibody, 0.1% saponin and 5% bovine serum albumin. After washing with phosphate-buffered saline, the samples were then incubated with the corresponding 2nd antibody, anti-mouse IgG or anti-rabbit IgG, conjugated with different fluorescent labels in phosphate-buffered saline containing 5% bovine serum albumin, 0.1% saponin and 10% normal donkey serum or goat serum. The samples were mounted on glass slides with mounting solution (Fluoromount-G) and observed using a Leica confocal system with a standard configuration DMRXE7 microscope.

Tissue preparation and western blot

All surgical and experimental procedures were conducted in accordance with the policies of Northwestern University's Animal Care and Use Committee and the NIH. Prior to a lethal injection of pentobarbital (∼200 mg/kg), followed by decapitation, adult mice were sedated with chloroform for approximately 30 s. Cochleae were collected in 10 mm Tris lysis buffer (50 mm sucrose, 1 mm EDTA, protease inhibitor cocktail (Sigma, P8340) and phenylmethylsulfonyl fluoride, pH 7.5). Equal amounts of transiently transfected CHO cells or TSA cells were treated in the same manner after physical removal from tissue culture dishes with a rubber policeman. Cells in lysis buffer were frozen, thawed, and then subjected to Dounce homogenization. Lysates were centrifuged at 2900 g, 4°C for 10 min, to remove nuclei and un-lysed cells. The supernatant was then centrifuged at 4°C, 16 000 g for 1.5 h. These membrane-enriched pellets were suspended in glycopeptidase-F reaction buffer (50 mm Tris-HCl, pH 7.5, 1% Triton X100, 0.1% sodium dodecyl sulfate, 1.6% 2-mercaptoethanol, protease inhibitor cocktail and phenylmethylsulfonyl fluoride) with or without glycopeptidase-F. The samples were incubated overnight at 37°C with gentle agitation. Protein concentration was determined by using Bio-Rad protein assay reagent (Bio-Rad, Hercules, CA, USA) based on Coomassie Brilliant Blue G250 dye-binding assay. Equal volume of 2 × Laemmli buffer with 10 mm dithiothreitol was added to the samples before loading onto a 6% or 7.5% sodium dodecyl sulfate–acrylamide gel. After sodium dodecyl sulfate–polyacrylamide gel electrophoresis, the gel proteins were electrotransferred onto a nitrocellulose membrane. The membrane was blocked with 2% non-fat dry milk/2% bovine serum albumin/phosphate-buffered saline, then incubated with primary antibodies: anti-Xpress, or anti-C-prestin, or anti-V5 containing 5% bovine serum albumin. After washing with Tris-buffered saline with 0.2% Tween 20, the membrane was incubated with the secondary antibody (donkey anti-rabbit IgG-HRP or anti-mouse IgG-HRP) in phosphate-buffered saline containing 4% normal donkey serum, 2% non-fat milk and 2% bovine serum albumin. Finally, after thoroughly washing with Tris-buffered saline with 0.2% Tween 20, the membranes were then used to detect prestin expression using an enhanced chemiluminescence western blotting detection system (Pharmacia, Uppsala, Sweden). The apparent molecular masses were calculated by non-linear curve-fitting of the molecular mass standards indicated at the left of each gel.

De-glycosylation treatment

De-glycosylation of prestin-expressing TSA cells during whole cell recordings was accomplished by treatment with glycopeptidase-F from Chryseobacterium meningosepticum (Sigma). Prestin/GFP-transfected TSA cells were incubated for 1.5 h at 37°C with glycopeptidase-F at a concentration of 0.06 U/µL in the culture medium. The non-linear capacitance of prestin-transfected TSA cells with or without glycopeptidase-F digestion was determined. The apparent molecular masses of partial or completely de-glycosylated prestin were analyzed through western blotting. Prestin-containing membrane-enriched pellets were treated with glycopeptidase-F (0.04 U/µL) before solublilization in Laemmli buffer on sodium dodecyl sulfate–polyacrylamide gel for western blot analysis. Tunicamycin was also used to inhibit the N-glycosylation of prestin-expressing TSA cells. Tunicamycin (0.25–2.0 µg/mL) was added to the culture medium immediately after transfection and subsequent incubation continued for 48 h. The prestin expression pattern in TSA cells was observed using a Leica confocal system with a standard configuration DMRXE7 microscope.

In vitro transcription and translation

In vitro transcription/translation was performed according to the manufacturer's instructions using gPrestin and Xpress-tagged gPrestin (X-gP) cDNA template, the TnT System (Promega, Madison, WI, USA) and 35S Met (NEN, Boston, MA, USA), followed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis on 10% gel and autoradiography.

Whole cell recordings in TSA cells

At 48–72 h after transfection, a cluster of TSA cells in the bottom of a culture dish were separated into single cells by trituration with a non-enzymatic Cell Dissociative Agent (Sigma) and transferred to another dish for measurement. Whole cell voltage clamp recordings were performed on the GFP-positive cells with an Axopatch 200B amplifier (Axon Instruments) at room temperature. Recording pipettes had initial bath resistances of 2–3 MΩ and were filled with internal solution containing (in mm): 140 CsCl, 2 MgCl2, 10 EGTA, 10 HEPES; at pH 7.2. TSA cells were bathed in external solution containing (in mm): 120 NaCl, 20 TEA-Cl, 2 CoCl2, 2 MgCl2, 10 HEPES, 5 glucose; at pH 7.2. Osmolarity was adjusted to 300 mosml−1 with glucose. After gigaseal formation whole-cell recording configuration was established. Whole-cell series resistances ranged from 3 to 10 MΩ and membrane resistance exceeded 100 MΩ. Cells were voltage clamped at 0 mV. Voltage-dependent capacitance was measured using a Windows-based whole-cell voltage-clamp program, jClamp (SciSoft, New Haven, CT, USA). A continuous high-resolution two-sine voltage stimulus protocol (20 mV peak at both 390.6 and 781.2 Hz) was used to detect the presence of non-linear charge movement (Santos-Sacchi and Navarrete 2002). TSA cells expressing gPres, or the mutants N163Q, N166Q, NN163/166QQ, and NN163/166AA were studied.

Data evaluation

Capacitance data were fitted with the first derivative of a two-state Boltzmann function (Santos-Sacchi 1991; Oliver et al. 2001),

image(1)

where Clin is the linear capacitance, obtained at the extremes of the membrane potential range. Qmax is the maximum charge transferred, V1/2 is voltage at half-maximal non-linear charge transfer, V is membrane potential, and α is the slope factor of the voltage dependence. Because the size of cells varies, we normalized Qmax by dividing with Clin, which is proportional to the membrane area. Qmax/Clin will be termed ‘charge density’; it is used as a normalized metric of charge transfer (Oliver and Fakler 1999). The unit of charge density is fC/pF. The magnitudes of voltage-dependent capacitance curves were normalized as 100(Cmax − Clin)/Clin (%). Cmax was the non-linear peak capacitance. The normalized data in each group was shown as relative Cm (membrane capacitance) in comparison with the control set as 1. We used IgorPro (WaveMetrics, Lake Oswego, OR, USA) for processing and fitting eqn 1 to the data.

All data are given as median (+ upper interquartile range, – lower interquartile range). The upper quartile range is defined as Q3 (the 75th percentile) minus median. The lower quartile range is defined as median minus Q1 (the 25th percentile). Statistical significance between groups was evaluated by the non-parametric Wilcoxon–Mann–Whitney test. Levels of statistical significance are indicated by asterisks (*p < 0.05; **p < 0.01).

Results

Prestin is a glycoprotein

Prestin is a highly conserved protein found only in mammals. Among mouse, rat, gerbil and human, 689 (∼93%) of the 744 residues of the prestin molecule are identical. In contrast, the most closely related proteins in the SLC26A family are ∼40% conserved. According to computer model predictions, prestin has potential N-linked glycosylation sites in the second extracellular loop, which are conserved among different species, including mouse, rat, gerbil, and human. It has been demonstrated previously that the fusion of the Xpress, V5-tag, or GFP at either the N-terminus or C-terminus of prestin does not affect its function (Ludwig et al. 2001; Zheng et al. 2001). Therefore, gPrestin constructs with or without the V5-tag or Xpress were transiently transfected into CHO or TSA-201 cells in order to study the glycosylation of prestin. The CHO and TSA cell lines were chosen because they have been used previously to study prestin's function (Zheng et al. 2000; Oliver et al. 2001). OK cells were also used because of their large cytosol-to-nucleus ratio, which makes membrane targeting of prestin relatively easy to detect. Prestin has a predicted molecular weight of 81.4 kDa. V5-tagged Prestin (V5-gP) and Xpress-tagged Prestin (X-gP) consist of 785 and 783 amino acids, respectively, with predicted molecular weight around 86 kDa. Western blot analysis showed a dominant ∼106 kDa band in prestin-expressing TSA cells (Fig. 1a), two dominant bands of ∼158 and ∼86 kDa in prestin-expressing CHO cells (Fig. 1b), and a dominant band of ∼86 kDa in native OHCs (Fig. 1c). It seems that the apparent molecular mass, as determined by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, of prestin can differ when expressed in different cell types. When prestin-containing membrane-enriched pellets were treated with glycopeptidase-F, an enzyme that digests all N-linked carbohydrate moieties, all of the high molecular mass (> 86 kDa) bands disappeared. A new immunoband representing the fully deglycosylated prestin appeared in proteins from OHCs (∼72 kDa band, Fig. 1c), V5-tagged prestin (V5-gP) transfected CHO cells (∼72 kDa in Fig. 2a, lane 1) and Xpress-tagged prestin (X-gP) transfected TSA cells (Fig. 2b, lane 3). These data suggest the presence of different lengths of N-linked carbohydrate moieties associated with prestin's post-translational modification, and also intimates that its glycosylation is variable between particular cell lines. The complex high molecular mass material that appeared in different cell types before treatment with glycopeptidase-F, may correspond to varying degrees of glycosylation or aggregation occurring during biosynthesis or sample preparation in different species of cells and tissues. Prestin appeared to possess more carbohydrate mass when it was expressed in CHO cells than it did in TSA cells.

Figure 1.

Western blot of prestin expressed in (a) TSA cells, (b) CHO cells, and (c) OHC from cochleae. Cells were lysed and the plasma membrane proteins from the three different cell types were resolved on polyacrylamide gels, transferred to nitrocellulose membranes and immunoblotted with either anti-V5 (a and b) or anti-C-Pres (c). (d)  In vitro transcription and translation. Protein markers are indicated by arrows located on the left of each image. OHCs from cochleae were also treated with (+ E) glycopeptidase-F (c). *Indicates the prestin band in OHC and in vitro transcription and translation experiments. V5-gP: V5-tagged gerbil prestin; X-gP: Xpress-tagged gerbil prestin; Vector: a vector alone.

Figure 2.

Western blot of wild-type gPrestin and its associated N-linked glycosylation mutants: N163Q, N166Q and NN163/166QQ. Plasma membrane proteins from various prestin-expressing CHO cells (a) or TSA cell (b) were treated with glycopeptidase-F (+ E) or untreated (– E) before loading onto polyacrylamide gels. The gels were transferred to nitrocellulose membranes and immunoblotted with either anti-V5 (a) or anti-Xpress (b). Protein markers are indicated by arrows located on the left of each image. *Indicates the deglycosylated prestin. V5-gP: V5-tagged gPrestin; X-gP: Xpress-tagged gPrestin; NN/QQ: NN163/166QQ.

The observed ∼72 kDa deglycosylated protein band is smaller in mass than the ∼86 kDa band predicted from the wild-type gPrestin, Xpress-tagged and V5-tagged prestin open reading frame translation. Indeed, when prestin cDNA is transcribed and translated in an in vitro system, where no glycosylation occurs, it produces a protein of around ∼72 kDa (Fig. 1d), similar to the deglycosylated prestin band found in OHC (Fig. 1c). Similar mass discrepancies have been seen for Tat1 and DRA proteins, two other members of the SLC26 family (Byeon et al. 1998; Toure et al. 2001). DRA consists of 764 amino acids with an expected theoretical molecular weight of around 84.5 kDa. After treatment with glycopeptidase-F, a band around ∼75 kDa is consistently observed in the DRA-expressing mammalian cell lines COS (Byeon et al. 1996) and SF9 cells (Byeon et al. 1998). It was suggested that DRA protein might undergo post-translational cleavage of approximately 100 N-terminal amino acids (Byeon et al. 1998). In order to investigate the possibility of cleavage of prestin, the glycopeptidase-F treated proteins, extracted from OHC, prestin-transfected CHO or TSA cells, were tested with anti-Xpress and anti-C-Pres antibodies that bind to epitopes at the ends of the N-terminus and C-terminus of prestin, respectively. A single ∼72 kDa immunoband was consistently observed using either anti-Xpress (Fig. 2b), anti-C-Pres (Fig. 1c), or anti-V5 antibody (Fig. 1a). These data suggest that there is no cleavage at either the N- or C-terminus of the prestin protein. A possible explanation for the discrepancy in mass is related to the extremely hydrophobic nature of the SLC26A proteins. The high degree of hydrophobicity in these proteins could result in more sodium dodecyl sulfate binding to the protein compared to an average soluble protein, resulting in a faster-moving negatively charged protein–sodium dodecyl sulfate complex. Alternatively, the hydrophobic portions of prestin may be tightly bound upon themselves, making the prestin–sodium dodecyl sulfate conformation smaller than an average protein, resulting in faster gel permeation.

Identification of N-linked glycosylation sites in prestin protein

Prestin has three potential N-linked glycosylation sites as listed in Table 1: N163, N166 and N603. The two N-linked glycosylation sites in the second extracellular loop (N163, N166) are conserved among members of the SLC26 family. A study of one SLC26 member (DRA) suggests that N-linked glycosylation occurred at the Asn located in the second extracellular loop, not at the N621 site, though it was not clear which Asn(s) in the second loop was involved in N-linked glycosylation (Byeon et al. 1996). The C-terminus of prestin, where N603 is located, is on the cytoplasmic side of plasma membrane (Zheng et al. 2001) and is an unlikely N-glycosylation site for prestin. Therefore, we mutated both potential N-linked glycosylation sites located in the second extracellular loop of prestin. Wild-type gPrestin and mutant constructs N163Q, N166Q and NN163/166QQ were transiently transfected into either TSA cells or CHO cells. The calculated molecular mass of wild-type prestin protein and its associated mutants were compared by western blot analysis. As shown in Fig. 2(a), both N163Q (Fig. 2a, lane 6) and N166Q (Fig. 2a, lane 4) have diffuse immnobands with larger apparent masses than the deglycosylated prestin band (∼72 kDa). Importantly, a ∼72 kDa prestin band, similar to deglycosylated wild-type gPrestin (Fig. 2a, lane 1), appeared after deglycosylation by glycopeptidase-F digestion (Fig. 2a, lanes 3 and 5). These data suggest that both N163 and N166 sites are involved in N-linked glycosylation. This conclusion is further confirmed by the result from experiments with the double-mutant NN163/166QQ. X-gPrestin and Xpress-tagged NN163/166QQ were transiently transfected in TSA cells. As seen in Fig. 2(b), an ∼86 kDa band, equivalent to glycosylated prestin (Fig. 2b, lane 4), appeared only in TSA cells transfected with wild-type gPrestin (X-gP). The ∼86 kDa band was replaced with the immunoband, around ∼72 kDa (deglycosylated form of prestin), after treatment with glycopeptidase-F (Fig. 2b, lane 3). In the western blot corresponding to NN163/166QQ (NN/QQ), no glycosylated prestin band (∼86 kDa) was seen. Instead, a band similar in size to the deglycosylated form of prestin (∼72 kDa) appeared, and did not change in mass after glycopeptidase-F digestion. Although higher molecular mass bands were observed in NN163/166QQ (Fig. 2b, lanes 1 and 2), similar size protein bands (less intense) were also observed in deglycosylated form of wild-type prestin (Fig. 2b, lane 3). Unlike the wild-type prestin (X-gP), the pattern of higher molecular mass bands observed in NN163/166QQ did not change upon glycopeptidase-F digestion. This result indicates that the mutant NN163/166QQ expresses a deglycosylated form of prestin. Since there are no other consensus glycosylations sites in the extracellular loop of prestin, and glycopeptidase-F treatment did not change the prestin band pattern of NN163/166QQ, it is likely that there are no other N-linked glycans on prestin.

Although both N163Q and N166Q mutants of prestin expressed multiple higher molecular mass bands, N163Q produced protein had a much larger apparent mass. N163Q had several diffuse bands around ∼102 kDa, whereas N166Q had a single dominant band around ∼82 kDa. With the N166Q mutant, asparagine 166 was replaced with glutamine so that N-glycosylation at that site is no longer possible. Consequently, the multiple higher molecular mass bands in lane 4 of Fig. 2(a) for mutant N166Q must represent glycosylation of N163. A similar line of reasoning suggests the multiple higher molecular bands in lane 6 of Fig. 2(a) for the N163Q mutant were due to glycosylation of N166. The molecular masses of the bands for lane 6 were greater than those bands of lane 4. Furthermore, all of the bands in lanes 4 and 6 are larger than the deglycosylated prestin protein in lanes 1, 3 and 5. These data suggested that both N163Q and N166Q constructs produced N-glycosylated prestin proteins and that N163 and N166 may have differential programs for synthesis and trimming of the glycans. The N166 site appears to have greater extent of glycosylation than its companion.

Role of N-linked glycosylation in cell surface expression

N-linked glycosylation is an enzyme-catalyzed, co-translational protein modification reaction. In many cases, N-linked glycosylation is important for the protein's membrane localization (Nagayama et al. 1998). However, there are also data suggesting the opposite. For example, N-linked glycosylation is required for plasma membrane localization of D5, but not D1, dopamine receptor in transfected mammalian cells (Karpa et al. 1999). Tunicamycin inhibits the first step of N-linked glycosylation, the acquisition of N-linked oligosaccharides (Elbein 1984). Therefore, we compared prestin's cellular expression pattern with or without tunicamycin treatment using confocal microscopy. GFP-fused gPrestin was transiently transfected into TSA or CHO cells. The green fluorescence of GFP indicated the location of prestin. A dose–response study of tunicamycin concentrations determined the non-toxic dose in TSA cells (250 ng/mL). In both tunicamycin treated and untreated cells, we consistently found some cells with strong green fluorescence at cell margins as marked by the arrows in Fig. 3, indicating the presence of prestin at the plasma membrane. In the same dishes, we also found some cells with green fluorescent diffuse staining throughout the cell and along the borders of the cells, and at projecting membrane segments of the cells. Figures 3(a) and (b) shows prestin-transfected TSA cells without tunicamycin treatment, and in Figs 3(c) and (d), prestin-transfected cells after 250 ng/mL tunicamycin treatment for 48 h. No difference is seen in the green fluorescence distribution patterns between these two groups regarding prestin membrane targeting. These data suggest that glycosylation of prestin is not a requirement for membrane targeting of this protein.

Figure 3.

Immunofluourescence images of TSA cells transiently transfected with GFP fused with wild-type gPrestin (wt) without (a and b) or with (c and d) tunicamycin treatment. The green color of GFP indicates the location of prestin proteins. (b) and (d) are DIC images superimposed on the corresponding images (a) and (c). Arrows indicate the cells with prestin located at the plasma membrane.

N-linked glycosylation is required for many cellular biosynthesis processes and a deficiency in this process may play a role in apoptosis (Yoshimi et al. 2000; Lukowitz et al. 2001). In our case, at tunicamycin concentrations > 1 µg/mL, most TSA cells detached from the culture dish and appeared to be dead. In order to avoid the complexity possibly induced by the cellular toxicity of tunicamycin, the role of N-linked glycosylation was examined using N-linked glycosylation site-directed mutants: N163Q, N166Q, NN163/166QQ andNN163/166AA. These mutants produced a partially or completely deglycosylated prestin as demonstrated in Fig. 2 and discussed above. In order to check the proper expression of wild-type prestin or N-linked glycosylation mutants in the plasma membrane, Flag-tagged growth hormone releasing hormone receptor (GHRHR) was co-transfected with prestin in a 1 : 1 ratio into TSA or CHO cells. GHRHR is known as a G-protein coupled receptor located in the plasma membrane, featuring seven transmembrane domains with an extracellular N-terminus and a cytoplasmic C-terminus (DeAlmeida and Mayo 1998). Prestin/GHRHR transfected cells were labeled with both anti-N-prestin and anti-Flag antibodies, with green color indicating prestin expression and red color showing GHRHR expression. Combined fluorescence of red and green is seen as a yellow/orange color. As seen in Fig. 4(a), red and green fluorescence was nearly identical, indicating that all of the N-linked glycosylation mutant molecules, including the fully deglycosylated prestin (NN163/166QQ), were able to target to the plasma membrane, just as the GHRH receptor. To further confirm these results, GFP tagged-prestin double mutants NN163/166AA were expressed in an opossum kidney cell (OK) line, in which membrane-localization is easier to see than in TSA cells. Both glycosylated prestin (wt) and deglycosylated prestin (N/A) were located in the plasma membrane, as shown in Fig. 4(b). OK cells, transfected with the GFP construct alone, were used as negative control. They showed light green fluorescent staining in the entire cytoplasm area. These data further confirm that N-linked glycosylation is not required for membrane targeting of prestin. Mutating the glycosylation sites (N) to either polar amino acid (Q) or hydrophobic amino acid (A) did not change the prestin membrane-targeting process.

Figure 4.

(a) Immunofluorescence images of TSA cells transiently transfected with wild-type gPrestin (wt) and three N-linked glycosylation mutant constructs: N163Q, N166Q and NN163/166QQ (N/Q). Red color, anti-GNRH receptor staining, indicates the existence of growth hormone releasing hormone receptor (GHRHR) protein, known to be localized in the cell membrane. Green color, FITC-labeled anti-N-Pres, indicates the location of prestin and N-linked glycosylation mutant proteins. Yellow images are superimposed from green and red images, indicating the co-localization of prestin and GHRH-receptor in the cell membrane. (b) OK cells transiently transfected with GFP-tagged rPrestin and mutant NN163/166AA(N/A), as well as GFP alone.

Role of N-linked glycosylation in the function of prestin

Aside from cellular membrane targeting, N-linked glycans in general can affect protein structure by facilitating the protein folding process and stabilizing the mature protein structure. Therefore, we next examined whether N-linked glycosylation affects prestin's function, as assayed by NLC. NLC is an accurate reflection of motility function, and we used it to measure the electromechanical effects produced by prestin and its associated N-linked glycosylation mutants. We measured the voltage-dependent capacitance (C; see eqn 1) in transfected cells with wild-type gPrestin or N-linked glycosylation mutant constructs. Capacitance data points were fitted with eqn 1 to derive the values of V1/2, α (slope), Clin, Qmax, and charge density (Qmax/Clin). Fit parameters for wild type (n = 12) were: V1/2 = −77.4 (+ 5.1, − 4.8) mV, α = 31.4 (+ 1.2, −0.5) mV, Clin = 20.4 (+ 6.6, −1.6) pF, Qmax = 535 (+ 422, −272) fC, and charge density (Qmax/Clin) = 26.3 (+ 5.5, − 11.2) fC/pF. Data points for each parameter represent median (+ upper interquartile range, – lower interquartile range). The data from TSA cells expressing wild-type gPrestin are consistent with those previously reported (Zheng et al. 2000; Santos-Sacchi et al. 2001). Figure 5(a) shows the voltage-dependent NLC curves observed for the single N-linked glycosylation-site mutants. The NLC functions recorded from N163Q and N166Q are basically the same as for wild-type gPrestin. Figure 5(b) shows the voltage-dependent NLC curves observed for the double mutants. There were statistically significant differences in V1/2 between wild-type gPrestin and the double-point mutations, NN163/166QQ and NN163/166AA (Fig. 5b). The V1/2 values of double-points mutants were shifted to the depolarizing side at −62.2 (+3.8, −6.0) mV for NN163/166QQ and −68.5 (+1.6, −4.0) mV for NN163/166AA (Figs 5b and c). Figure 5(c) shows the V1/2 of the NLC of the double mutants, NN163/166QQ and NN163/166AA. No statistical differences of charge densities were found among the five groups (Fig. 5d).

Figure 5.

Non-linear capacitance of TSA cells transiently transfected with wild-type gPrestin (WT: control) and four N-linked glycosylation mutant constructs. (a) Capacitance from single point mutations (N163Q, N166Q) normalized to the magnitude at Cmax of control. The line is the fit of eqn 1 to the data. Fit parameters were: for wild-type (n = 12): V1/2 = −77.4 (+5.1, −4.8) mV, α = 31.4 (+1.2, −0.5) mV, Qmax = 535 (+422, −272) fC, and charge density = 26.3 (+5.5, −11.2) fC/pF; for the N163Q mutant (n = 11). Also, V1/2 = −74.2 (+4.8, −6.4) mV, α = 39.5 (+2.1, −0.9) mV, Qmax = 398 (+347, −78) fC, and charge density = 23.4 (+6.2, −6.4) fC/pF; for the N166Q mutant (n = 11). Further, V1/2 = −84.1 (+5.9, −8.9) mV, α = 36.3 (+1.5, −4.3) mV, Qmax = 588 (+211, −163) fC, and charge density = 27.8 (+9.1, −8.5) fC/pF. (b) Capacitance measures from double mutants (NN163/166AA and NN163/166AA) normalized to the control magnitude at Cmax. The line is the fit of eqn 1 to the data. Fit parameters were: for wild type (n = 12) is used the same data shown above (a); for NN163/166AA (n = 11): V1/2 = −68.5** (+1.6, −4.0) mV, α = 38.0 (+1.4, −1.7) mV, Qmax = 431 (+129, −189) fC, and charge density = 16.6 (+4.6, −4.2) fC/pF; for NN163/166QQ (n = 10): V1/2 = −62.2** (+3.8, −6.0) mV, α = 38.6 (+3.2, −2.8) mV, Qmax = 309 (+89, −98) fC, Clin = 17.6 (+0.8, −4.3) pF and charge density = 15.1 (+14.5, −3.2) fC/pF. (c) Voltage at peak capacitance (V1/2) of each mutant. Both single point mutations showed voltage at peak capacitance almost identical to wild type. However, the double mutations displayed statistically significant shift of V1/2 to the depolarizing side. (d) Charge density [Qmax/Clin (fC/pF)]. Although the charge densities of double mutants are lower, there is no statistical difference between them and the wild-type control. Data points represent median (+ upper interquartile range, – lower interquartile range) for 10–12 TSA cells per experiment. Statistical significance was tested using Wilcoxon–Mann–Whitney test, and significance levels are indicated by asterisks (*p < 0.05; **p < 0.01). Asterisks represent significance vs. control (wild-type prestin).

It is generally believed that glycan structure plays an important role in correct protein folding, but it is not essential for maintaining the overall structure after the protein has been made (Helenius and Aebi 2001). Because the deglycosylated prestin (the double mutants) shifted their V1/2 in the depolarizing direction compared to wild-type gPrestin, the role of N-linked glycans in prestin's function was further investigated by removing N-linked glycans from the glycosylated form of prestin that was already inserted in the plasma membrane. Thus, we treated wild-type prestin expressing TSA cells with glycopeptidase-F. After the removal of N-linked glycans by glycopeptidase-F, the charge density of gPrestin-transfected cells remains the same (Fig. 6c), but these glycopeptidase-F treated prestin-transfected cells showed a small but statistically significant shift of V1/2 toward the depolarized side in comparison with non-treated gPrestin-expressing cells (Figs 6a and b). This change was similar to the V1/2 shift observed in TSA cells transfected with the double mutants (Fig. 5b). We also noticed that the degree of V1/2 shift caused by glycopeptidase-F treatment showed a time-dependence, with greater V1/2 shift toward the depolarized direction as the glycopeptidase-F digestion time increased (data not shown). Of course, enzymatic digestion might affect not only prestin but also other membrane proteins that may have N-glycans. It should be noted, however, that prestin has a high level of expression in transfected TSA cells due to the powerful viral promoter driving the expression. It is likely that prestin is the major plasma membrane protein present in expressing cells. Although there are limitations in comparing function between enzyme treatment and N-glycosylation residue mutations, it is observed that the V1/2 of enzyme-treated cells was shifted in the same direction as in the double mutants (NN163/166QQ and NN163/166AA). While this result may be coincidental, it is possible that N-glycans in prestin might influence the establishment of V1/2. It is emphasized that functional comparisons between wild-type prestin and mutant forms of prestin were always done in parallel, i.e. from the same batch of cells each time. This enabled us to effectively compare differences in glycosylation efficiencies among different cell groups

Figure 6.

Effect of glycopeptidase-F on NLC of prestin-transfected TSA cells. (a) Capacitance from prestin-transfected TSA cells before and 1.5 h after enzymatic digestion by glycopeptidase-F. The line is the fit of eqn 1 to the data. Fit parameters were for control cells (n = 20): V1/2 = −66.2 (+3.9, −6.2) mV, α = 34.5 (+2.4, −2.2) mV, Qmax = 217 (+104, −89) fC, and charge density = 17.7 (+3.5, −5.7) fC/pF; for glycopeptidase-treated cells (n = 14): V1/2 = − 56.0 (+5.4, −4.0) mV, α = 35.6 (+4.1, −4.2) mV, Qmax = 256 (+80, −73) fC, and charge density = 17.3 (+4.3, −8.2) fC/pF. (b) Effect of glycopeptidase-F on V1/2. The digested cells displayed statistically significant shift of V1/2 toward the depolarizing side. (c) Charge density [Qmax/Clin (fC/pF)] was unchanged after glycopeptidase-F treatment.

Discussion

N-linked glycans can play an important role in protein structure by facilitating the protein-folding process and in the degradation of misfolded or mutant proteins (Imperiali and O'Connor 1999; Helenius and Aebi 2001). The values of the measured charge density (Qmax/Clin), as shown in Fig. 5(d), reflect the number of functional prestin molecules located at a unit cell surface. No statistical differences of charge densities were found among the glycosylated wild-type prestin (WT), and partially or fully deglycosylated prestin. These data indicate that the glycans do not change the expression efficiency of prestin and its transport to the plasma membrane.

The partially glycosylated prestin mutants (the single N-linked mutants) had similar voltage-dependent capacitance (NLC) as that of the wild-type prestin when they were transiently transfected into TSA cells. The N166 site appears to have a greater degree of glycosylation than N163 based on the mass loss, but we did not find significant alterations in prestin's functional assay comparing N163Q, N166Q and wild-type prestin. Furthermore the deglycosylated prestin (double-mutant) maintained its NLC, a signature of prestin's electrophysiological function. However, the deglycosylated prestin did show altered functional properties, specifically a small positive shift of the voltage at peak capacitance without diminishing maximal charge transfer (Qmax). It is possible that prestin has the ability to compensate for a single N-glycosylation site mutation. However, a total lack of both N-glycosylation sites might result in a molecule with somewhat altered function (Helenius and Aebi 2001).

For in vitro analysis of the glycans, deglycosylation of prestin was achieved either from the absence of N-linked sites (double mutants) or through the glycopeptidase-F treatment that removes all N-linked carbohydrate moieties from the cell surface. Although the glycan is generally not essential for maintaining a protein's overall structure, it was suggested that glycans do affect protein stability (Imperiali and O'Connor 1999). It is not clear whether the small V1/2 shift observed in the deglycosylated prestin expressing cells is due to an unstable prestin conformation caused by lacking of N-glycans, or changed membrane properties caused by N-glycan structure, or both. The net change in charge of the extracellular portion of prestin molecule probably decreases as a result of deglycosylation, which may vary depending primarily on the number of sialic acid residues (Hermentin et al. 1996). Consequently, it is possible that the charges of N-glycans on the prestin molecule contribute to the determination of V1/2. Whatever the reason may be, it seems that N-glycans play some role in maintaining the proper electrophysiological features of prestin.

Unlike other motor proteins such as myosin, motile function of prestin is driven directly by voltage change rather than enzymatic activity. Non-linear capacitance observed in OHCs and prestin-transfected cells reflects a voltage-dependent gating current caused by charge displacement in the prestin molecule, which in turn results in conformational change of the molecule and, ultimately, in cell contraction and elongation. The integral of the NLC function represents the probability of mobile charges being translocated across the plasma membrane as a consequence of the binding of a monovalent anion (Cl). Therefore, V1/2 is the voltage at which charges are moved with the smallest voltage-increment. There are multiple variables identified that influence V1/2. Among these are tension of cell surface and change in internal pressure of prestin-transfected cells or OHCs (Gale and Ashmore 1994; Kakehata and Santos-Sacchi 1995). Similar to deglycosylated prestin, more than half of the mutated prestin proteins to date also have altered V1/2 and the shift can be either in depolarizing or hyperpolarizing directions (Oliver et al. 2001). Even for wild-type prestin, we often observed a V1/2 shift when different batches of TSA cells were used for transfection. For example, the V1/2 for wild-type prestin-transfected TSA cells shown in Fig. 5(c) is approximately −77 mV, whereas the same prestin construct measured −67 mV (Fig. 6b) when TSA cells from low-passage frozen stock were used for transfection. It is possible that the metabolic condition of the cell line affects prestin's glycosylation, similar to other proteins reported previously (Gawlitzek et al. 1995; Nyberg et al. 1999). Similar to prestin-transfected TSA or CHO cells, OHCs at P2 and P12 ages demonstrated different V1/2 (Oliver and Fakler 1999; Ludwig et al. 2001). These data indicate that the cell's metabolic or physiological condition can influence prestin's V1/2 as well. It was demonstrated that prestin mRNA expression, along with other genes expressed in OHCs, are developmentally regulated along both radial and basal-to-apical longitudinal gradients (Judice et al. 2002). It is possible that the glycosylation of prestin varies within the organ of Corti or during developmental stages. These differing glycosylations of prestin, resulting in various V1/2 along the organ of Corti, may create a spatial and temporal spectrum of functional molecules (cells) with different responsiveness. Such variation may contribute to the cochlea's exquisite tuning in a subtle but important way. In conclusion, prestin's native protein structure and its surrounding environment appear to play a role in determining the sensitivity of its functional activity.

Acknowledgments

Mr Alex Orem and Mr Kevin Long prepared some of the molecular biology experiments, Dr W. Russin at the Biological Imaging Facility of Northwestern University is thanked for his help in image processing and Dr Kelly Mayo for providing the GHRH receptor construct. This work is supported by the National Institutes of Health, NIDCD, DC00089.

Ancillary