Functional coupling of serotonin and noradrenaline transporters

Authors


Address correspondence and reprint requests to P. Schloss, Biochemical Laboratory, Central Institute of Mental Health, J5, 68159 Mannheim, Germany. E-mail: schloss@zi-mannheim.de

Abstract

Re-uptake of the neurotransmitters serotonin and noradrenaline out of the synaptic cleft is mediated by selective transporter proteins, the serotonin transporter and the noradrenaline transporter respectively. Both are integral membrane proteins that are have a high degree of homology and represent members of a larger neurotransmitter transporter superfamily. Several studies have indicated that the serotonin transporter has an an oligomeric structure. To determine whether monoamine transporters can also function in oligomeric structures in situ, we constructed a concatenate consisting of one molecule of serotonin transporter covalently linked to one molecule of noradrenaline transporter. Heterologous expression of this hybrid construct allowed us to analyse the function, i.e. transport activity, and the structure, i.e. the molecular weight of the total construct and of its single components, at the same time. We showed that serotonin–noradrenaline transporter fusion proteins are fully active and exhibit the pharmacological profile of both their individual components. These findings support the hypothesis that monoamine transporters are expressed and may function as oligomeric proteins composed of non-interacting monomers.

Abbreviations used
DA

dopamine

DAT

dopamine transporter

DTT

dithiothreitol

HEK

human embryonic kidney

5-HT

5-hydroxytryptamine

NE

noradrenaline

NET

noradrenaline transporter

PAGE

polyacrylamide gel electrophoresis

SERT

serotonin transporter

SDS

sodium dodecyl sulfate

SNRI

selective noradrenaline re-uptake inhibitor

SSRI

selective serotonin re-uptake inhibitor

TCA

tricyclic antidepressant

TMD

transmembrane domain

The postsynaptic action of the monoamines serotonin (5-hydroxytryptamine, 5-HT), noradrenaline (NE) and dopamine (DA) is controlled by specific, high-affinity re-uptake of these neurotransmitters out of the synaptic cleft back into the presynaptic terminal. This process is mediated by highly selective neurotransmitter proteins, the serotonin transporter (SERT), NE transporter (NET) and DA transporter (DAT) respectively. These polypeptides are members of a gene family that also includes transporters for other neurotransmitters such as GABA and glycine (Schloss et al. 1994; Worrall and Williams 1994; Iversen 2000; Reith 2002). The members of this neurotransmitter transporter family are integral transmembrane proteins and share a common putative 12-transmembrane domain (TMD) structure. Heterologous expression of the cloned neurotransmitter transporters in Xenopus oocytes and mammalian cells generates specific neurotransmitter uptake which exhibits a pharmacological profile similar to that of the native transport systems. Thus, transfection with a single cDNA confers expression of functional neurotransmitter transport of well defined pharmacology on to non-neuronal cells.

Based on studies of both intact membranes and solubilized proteins, there is growing evidence that transporter proteins exist in oligomeric (dimeric and/or tetrameric) forms (Klingenberg 1981). In this context, it has been postulated that the unrelated hexose transporter GLUT1 exists as a mixture of homodimers and homotetramers, with an intermolecular disulphide bridge promoting tetramerization (Hebert and Carruthers 1991; Zottola et al. 1995). Evidence for oligomerization of neurotransmitter transporters was primarily based on radiation inactivation studies on the DAT (Milner et al. 1994). Investigation of the quaternary structure of the SERT using cross-linking methodology revealed that the SERT can form dimeric and tetrameric adducts, indicating that this transporter is an also oligomeric membrane protein (Jess et al. 1996). This finding was also corroborated by analysis of heterologously expressed linear concatenates of SERT (Chang et al. 1998), co-immunoprecipitation of two forms of SERT tagged with different epitopes (Kilic and Rudnick 2000) and by fluorescence resonance energy transfer analysis (Schmid et al. 2000).

In this study, we designed a heterodimeric transporter by constructing a linear, chimeric ‘head to tail’ concatenate with 24 TMDs, consisting of one molecule of SERT and one of NET. Immunological analysis suggested a disulphide-stabilized dimerization of this construct and heterologous expression in human embryonic kidney (HEK)-293 cells proved its functionality as [3H]5-HT and [3H]NE transport variables were comparable to those of the wild-type SERT and NET proteins. 5-HT transport was sensitive to the selective serotonin re-uptake inhibitor (SSRI) citalopram but insensitive to the selective NE re-uptake inhibitor (SNRI) nisoxetine; NE transport was sensitive to nisoxetine, but not to citalopram; the tricyclic antidepressant (TCA) imipramine inhibited the uptake of both monoamines. These data provide the first demonstration of a dimeric, functional SERT–NET hybrid membrane protein and corroborate the hypothesis that monoamine transporters can function as oligomeric proteins.

Materials and methods

Materials

The plasmid pcDNA3.1(−)/Myc-His was purchased from Invitrogen (Carlsbad, CA, USA). Gel purification and plasmid preparation kits were purchased from Qiagen (Valencia, CA, USA). Taq DNA polymerase was purchased from Promega (Madison, WI, USA). Restriction enzymes and T4 DNA ligase were obtained from MBI Fermentas (St. Leon-Rot, Germany). Fetal bovine serum was purchased from Sigma (Taufkirchen, Germany), Dulbecco's modified Eagle's medium (DMEM) and other cell culture supplements and DH5α bacteria were obtained from GibcoBRL (Rockville, MD, USA). [3H]Citalopram (82 Ci/mmol) was purchased from PerkinElmer (Zaventem, Belgium), and [3H]5-HT (6.1 Ci/mmol), [3H]NE (12.0 Ci/mmol) and [3H]imipramine (21 Ci/mmol) were obtained from Amersham (Amersham, UK). Unlabelled imipramine was purchased from ICN Biochemicals (Asse, Belgium), nisoxetine from Tocris (Köln, Germany) and unlabelled citalopram was a gift from J. Hyttel (H. Lundbeck A/S, Copenhagen-Valby, Denmark). The antibody against NET (AB5066P) was obtained from Chemicon (Temecula, CA, USA) and the antibody against the myc-epitope was from Clontech (Heidelberg, Germany). All other chemicals were of analytical grade.

Construction of chimeric concatenates

The cDNAs of rat SERT and rat NET were excised from pCis-SERT (Schloss and Betz 1995) and from pEUKC-NET (a gift from H. Bönisch) respectively. They were cloned subsequently between an XbaI and a NotI site (SERT) and an EcoRV site (NET) into the polylinker of pcDNA3.1(−)/Myc-His to obtain construct [A]. The NotI site at the C-terminus of SERT was inserted by PCR thereby eliminating the SERT stop codon. For the cloning of construct [B] the linker between the two transporters of construct [A] was replaced by the myc tag. This tag was generated by PCR using Taq DNA polymerase and the oligonucleotide primer pair 1 (sense 5′-TGCTCATCGTGGTCATC-3′; antisense 5′-ATTCAGATCCTCTTCTGAGATGAGTTTTTGTTCCACAGCATTCATGCG-3′) that encode a C-terminal region of SERT and the oligonucleotide primer pair 2 (sense 5′-GAACAAAAACTCATCTCAGAAGAGGATCTGAATATTCTTCTGGCACGA-3′; antisense 5′-CACAGACAGCAGGAAAT-3′) that encode an N-terminal region of NET. The antisense primer of pair 1 and the sense primer of pair 2 both have the oligonucleotides encoding the myc tag added to the coding oligonucleotides for the transporters so that they anneal together in a second PCR step. This step was performed with the DNA products generated with primer pair 1 and with primer pair 2 using the sense primer of pair 1 and the antisense primer of pair 2. Conditions for all reactions were 34 cycles at 94°C for 30 s, 50°C for 30 s and 72°C for 1 min, with a 10-min extension time at cycle 34. Construct [A] and the PCR product of the second step were restricted with Eco72I and Bsu36I and the linker of construct [A] replaced by the PCR product. Products of all cloning steps were verified by sequencing.

Heterologous expression

HEK-293 cells were transfected with the cDNA of rat NET (rNET) and the constructs, by lipofection with the FuGENE 6 reagent according to the supplier's protocol (Roche, Indianapolis, IN, USA). Stably transfected cells were selected with G418 as described (Sur et al. 1998). HEK-SERT, HEK-NET, HEK-S/N (construct [A]) and HEK-SmycN (construct [B]) were maintained in DMEM supplemented with 10% fetal bovine serum, penicillin (100 U/mL), streptomycin (100 µg/mL) and G418 (200 µg/mL) at 37°C in a humidified atmosphere (5% CO2).

Measurement of [3H]5-HT and [3H]NE transport

For uptake measurements, HEK-293 cells expressing either the SERT, NET, both or no transporter were plated into 24-well dishes (2 cm in diameter), and the culture medium was replaced with 0.2 mL TBB (125 mm NaCl, 2 mm KCl, 1 mm CaCl2, 1 mm MgCl2 and 10 mm HEPES, pH 7.5) containing 250 nm[3H]5-HT or [3H]NE (unless otherwise stated). After 6 min at 18°C–24°C the solution was rapidly removed, and the cells were washed twice with cold TBB before extraction with 0.4 mL 10% (w/v) sodium dodecyl sulfate. Radioactivity was determined by scintillation counting using a Packard 1600TR scintillation counter (Dreieich, Germany). Specific [3H]5-HT or [3H]NE uptake was defined as the difference between monoamine transporter-mediated minus control uptake into non-transfected or HEK-293 cells transfected with insertless pRC vector.

Membrane preparation and radioligand binding

Cell membranes were prepared as described elsewhere and stored in aliquots at − 80°C (Schloss and Betz 1995). Protein concentration was determined according to Markwell et al. (1978). Binding of radioligands was performed as described previously for binding of [3H]citalopram to HEK-SERT membranes (Schloss and Betz 1995). Saturation binding was performed at room temperature using 25 µg membrane protein in a total volume of 200 µL assay buffer containing 0.2–24 nm[3H]citalopram or [3H]nisoxetine, or 1.5–96 nm[3H]imipramine. Specific binding was defined as the difference between the binding to membranes from cells transfected with SERT, NET and SmycN constructs and the binding to membranes from HEK-293 cells (there was no difference in the binding to membranes from non-transfected HEK-293 cells and HEK-293 cells transfected with an insertless pcDNA vector). Non-specific binding in the presence of high concentrations (> 50 µm) of unlabelled citalopram and nisoxetine or 1.5 mm unlabelled imipramine did not differ from that obtained in membrane preparations from non-transfected cells.

Electrophoresis and immunoblotting

SDS–polyacrylamide gel electrophoresis (PAGE) in 10% or in 4–12% gradient gels was performed according to the method of Laemmli (1970). In all experiments, the samples were heated to 37°C for 30 min before SDS–PAGE. For immunoblotting, separated proteins were transferred to nitrocellulose and blots were incubated with primary antibody overnight at 4°C. Detection was performed after incubation with a horseradish peroxidase-coupled secondary antibody using the ECL detection system (NEN).

Data analysis

All uptake data represent the mean ± SEM of quadruplicate determinations; each experiment was repeated at least twice. SEMs were routinely < 5% and are indicated where larger than the symbols used. Data were analysed by a non-linear regression analysis program (Sigma plot V, SPSS Inc, Richmond, CA, USA), which fitted sigmoidal uptake curves to the following equations: V = Vmax/(1 + [KM/S]n, and V/Vmax = inline image for competition experiments. V represents the transport rate, Vmax the maximal transport rate, KM the Michaelis Menten constant, S the substrate concentration, I the inhibitor concentration, IC50 the inhibitor concentration for half-maximal transport inhibition, and n the Hill coefficient.

Results

Characterization of SERT–NET fusion proteins by immunoblotting

In a first attempt the SERT and NET were cloned in-frame into the polylinker of pcDNA3.1 vector and the resulting fusion protein was termed HEK-S/N (Fig. 1). Transient expression of this construct in HEK-293 cells generated specific uptake of [3H]5-HT and [3H]NE (data not shown). Before cloning a stable cell line, we checked the integrity of the S/N polypeptide by western blotting. Membrane preparations from cells expressing the S/N fusion protein were subjected to SDS–PAGE under reducing (100 mm dithiothreitol; DTT) and non-reducing conditions, transferred to nitrocellulose and reacted with a SERT-specific antibody. No immunoreactive band was detected at ∼120 kDa, the expected molecular weight of the dimeric transporter. Instead, under reducing and non-reducing conditions, monomeric SERT molecules were detected with an apparent molecular weight of 65 kDa (Fig. 2). In addition, a higher molecular weight form was visualized at > 200 kDa, which was strongly increased under non-reducing conditions; comparable results were obtained with a NET-specific antibody (data not shown).

Figure 1.

Scheme of construction of SERT–NET concatenates. To achieve construct [A] rat SERT was cloned via XbaI and a PCR-generated NotI site at the C-terminus of SERT into pcDNA3.1(−)/Myc-HisA. Rat NET was cloned via an EcoRV site. The last amino acids of SERT and the first amino acids of NET are underlined and italicized. For the generation of construct [B], the linker of construct [A] was excised with Eco72I and Bsu36I, both cutting in the coding region of one of the transporters, and replaced by a PCR-generated linker coding for the myc epitope. For details see Materials and methods.

Figure 2.

Immunoblotting of membrane preparations from HEK-293 cells expressing the fusion proteins S/N and SmycN under reducing (+ DTT) and non-reducing (– DTT) conditions. Membranes (∼ 15 µg/lane) prepared from HEK-293 cells transfected with the S/N cDNA were subjected to SDS–PAGE (10%) and reacted with an anti-SERT antibody. Membranes (∼ 15 µg/lane) prepared from HEK-293 cells transfected with the SmycN cDNA were electrophoresed on a 4–10% polyacrylamide gel before reaction with an anti-myc antibody.

In a second approach we then exchanged the linker between the two transporters with the myc tag; the resulting construct was termed SmycN (Fig. 1). Western blotting of cell membranes from HEK-293 expressing SmycN with an anti-myc antibody under reducing conditions detected specific SmycN immunoreactivity at about 120 kDa and 250 kDa (Fig. 2). Under non-reducing conditions nearly all immunoreactivity was concentrated at 250 kDa. No signal was seen at the molecular weight of monomeric transporters. The 120-kDa band observed in the presence of DTT corresponds to the calculated molecular weight of SmycN and thus confirmed the integrity of the construct.

Characterization of [3H]5HT and [3H]NE uptake into HEK-SmycN cells

In order to characterize the transport parameters of the SmycN fusion protein, we first performed saturation experiments with increasing concentrations of [3H]5-HT and [3H]NE (Fig. 3). SERT-mediated [3H]5-HT uptake was characterized by a KM of 728.4 ± 129.7 nm, a Vmax of 8.2 ± 0.6 pmol per min per well and nH = 1.5 ± 0.3; and NET-mediated [3H]NE uptake by a KM of 1.2 ± 0.2 µm, a Vmax of 15.3 ± 1.1 pmol per min per well and nH = 1.4 ± 0.2. SmycN mediated [3H]5-HT transport with a KM of 1.5 ± 0.3 µm and a Vmax of 14.1 ± 1.0 pmol per min per well, nH = 1.4 ± 0.2; and [3H]NE uptake with a KM of 1.2 ± 0.2 µm and a Vmax of 9.2 ± 0.6 pmol per min per well, nH = 1.3 ± 0.1 (data from three experiments performed in quadruplicate). Uptake studies of [3H]5HT in the presence of NE and vice versa revealed that the two monoamines did not affect each other's transport parameters at concentrations up to 10 µm (data not shown). Interestingly, at higher micromolar concentrations 5-HT competed with [3H]NE for uptake by the NET, but NE did not compete with [3H]5-HT for uptake by the SERT.

Figure 3.

[3H]5-HT uptake into HEK-293 cells expressing SERT (a) or SmycN (c) and [3H]NE uptake into HEK-293 cells expressing NET (b) or SmycN (c). In the experiments shown, saturation analysis revealed mean ± SEM apparent affinities for 5-HT (▪) of 658 ± 63 nm for the SERT and 1.7 ± 0.3 µm for SmycN-expressing cells, and for NE (•) of 1.2 ± 0.17 µm for the NET and 1.1 ± 0.1 µm for SmycN-expressing cells. The mean ± SEM maximal transport rates were 7.8 ± 0.8 and 14.6 ± 1.4 pmol per min per well for 5-HT into SERT and SmycN cells respectively, and 15.3 ± 1.1 and 9.6 ± 0.5 pmol per min per well respectively for NE into NET and SmycN cells.

We then compared the inhibitory potencies of the SSRI citalopram, the SNRI nisoxetine and the TCA imipramine at the SERT, NET and SmycN, by performing competition experiments at various inhibitor concentrations; the resulting IC50 values are given in Table 1. The results can be summarized as follows: (i) citalopram inhibited [3H]5-HT uptake with similar efficiency at both the SERT and SmycN, but did not significantly affect [3H]NE uptake at the NET or SmycN; (ii) nisoxetine inhibited [3H]NE uptake with similar efficiency at both the NET and SmycN, but did not significantly affect [3H]5-HT uptake at the SERT or SmycN; (iii) imipramine inhibited [3H]5-HT uptake with similar potency at both the SERT and SmycN as well as [3H]NE uptake at the NET and SmycN; and (iv) 5-HT and NE did not influence each other's transport mediated by SmycN.

Table 1.  Inhibition of [3H]5-HT and [3H]NE transport by citalopram, imipramine and nisoxetine
SubstrateInhibitorIC50 (nm)
SERTSmycNNET
  1. Data represent the mean ±SEM of at least three independent experiments; IC50 values for [3H]5-HT or [3H]NE transport were determined using 250 nm[3H]5-HT and 250 nm[3H]NE in the uptake assays. n.t., No transport.

[3H]5-HTCitalopram5.0 ± 0.59.4 ± 2.6n.t.
Imipramine254.5 ± 12.0284.3 ± 27.2n.t.
Nisoxetine> 1000> 1000n.t.
[3H]NECitalopramn.t.> 1000> 1000
Imipraminen.t.65.0 ± 24.798.7 ± 36.3
Nisoxetinen.t.5.1 ± 2.06.6 ± 1.5

Characterization of antidepressant binding to HEK-SmycN membranes

To gain further insight into the properties of the SmycN protein, we characterized the binding of [3H]citalopram, [3H]imipramine and [3H]nisoxetine membranes of cells expressing this construct and compared the resulting equilibrium parameters with those obtained with the recombinant SERT and NET proteins (Fig. 4 and Table 2). Citalopram and nisoxetine displayed binding parameters at the SmycN dimer comparable to those obtained at the SERT and NET respectively. Imipramine, which binds to the NET with about fourfold lower affinity than the SERT, bound to SmycN with an intermediate, but not significantly different, affinity (Fig. 4, Table 2).

Figure 4.

Saturation analysis of [3H]citalopram (a), [3H]nisoxetine (b) and [3H]imipramine (c) binding to membrane preparations from stable HEK-SERT (▪), HEK-NET (•) and HEK-SmycN (▴) cells. KD, nH and Bmax values were: [3H]citalopram to SERT: 1.5 nm, 5.0 pmol per mg protein and 1.3 respectively; [3H]citalopram to to SmycN: 2.1 nm, 9.3 pmol per mg protein, 1.4; [3H]imipramine to SERT: 10.7 nm, 5.4 pmol per mg protein, 1.2; [3H]imipramine to to NET: 41.8 nm, 11.0 pmol per mg protein, 1.0; [3H]imipramine to SmycN: 20.2 nm, 16.3 pmol/mg protein, 1.0; [3H]nisoxetine to NET: 4.8 nm, 12.7 pmol per mg protein, 1.1; and [3H]nisoxetine to SmycN: 7.0 nm, 4.4 pmol per mg protein, 1.0.

Table 2.  Comparison of antidepressant-binding parameters of SERT, NET and SmycN
AntidepressantTransporterKD (nm)nHBmax (pmol per mg protein)
  1. Data represent the mean ± SEM of at least three independent experiments.

[3H]citalopramSERT1.1 ± 0.31.4 ± 0.15.2 ± 0.7
SmycN4.1 ± 2.01.3 ± 0.116.1 ± 6.4
[3H]imipramineSERT12.1 ± 0.90.9 ± 0.25.3 ± 0.8
NET39.2 ± 20.20.9 ± 0.111.2 ± 2.1
SmycN19.7 ± 9.30.9 ± 0.214.0 ± 2.8
[3H]nisoxetineNET3.8 ± 0.81.2 ± 0.19.7 ± 2.7
SmycN5.7 ± 2.80.9 ± 0.034.9 ± 1.1

Mutual competition experiments showed that imipramine fully displaced [3H]citalopram binding to SERT (IC50 61.9 ± 11.0 nm) and to SmycN (IC50 71.9 ± 1.3 nm), and also [3H]nisoxetine binding to NET (IC50 65.6 ± 3.3 nm) and to SmycN (IC50 = 68.0 ± 6.4 nm).

[3H]Imipramine binding to the SERT was fully displaceable by citalopram with an IC50 value similar to that of its binding affinity (IC50 1.4 ± 0.4 nm). At SmycN, however, citalopram at a concentration of 96 nm only displaced 73.6 ± 3.8% (n = 3) of 20 nm[3H]imipramine binding (Fig. 5a). Nisoxetine inhibited [3H]imipramine binding to the NET with an IC50 of 8.4 ± 2.6 nm. At SmycN, nisoxetine (96 nm) displaced 29.9 ± 5.9% (n = 3) of 20 nm[3H]imipramine binding (Fig. 5b).

Figure 5.

Displacement of [3H]imipramine binding to SERT (▪) and SmycN (•) by citalopram (a) and to NET (▴) and SmycN (•) by nisoxetine (b). In the experiments shown, [3H]imipramine binding at 10 nm to the SERT was inhibited by citalopram with an IC50 of 3.2 nm; [3H]imipramine binding at 40 nm to NET was inhibited by nisoxetine with an IC50 of 13.8 nm. At SmycN, 73% of 20 nm[3H]imipramine binding was displacable by 96 nm citalopram, and 32% by 96 nm nisoxetine.

Discussion

Biochemical and biophysical approaches to the determination of the quaternary structure of proteins are often difficult because during the experimental procedures the respective proteins may be characterized as functionally active on the cell surface, but at the same time functionally inactive ‘en route’, for example in the endoplasmic reticulum or the trans golgi network. Furthermore, when using immunological means to detect different forms of transporters one might identify fully translated, but functionally inactive proteins. To avoid these obstacles, we used a functional analysis in which we expressed two pharmacologically and immunologically defined transporters as a fusion protein consisting of one molecule of SERT covalently linked to one molecule of NET. This allowed us to compare the activity and the molecular weight of the total construct and of its single components at the same time. Western blot analysis combined with functional assays of this construct provided correlated information on possible oligomeric structure together with the transport characteristics of this fusion protein.

First we cloned SERT and NET in-frame into the polylinker of a pcDNA3.1 vector. Analysis of the integrity of the expressed SERT–NET polypeptide was checked by immunoblotting. Unexpectedly, although computer-assisted analysis did not reveal any proteolytic cleavage sites in our construct, under reducing and non-reducing conditions monomeric SERT was detected with an apparent molecular weight of 65 kDa. In addition, a higher molecular weight form was visualized at > 200 kDa, which was strongly increased under non-reducing conditions; comparable results were obtained with a NET-specific antibody. Because PCR analysis of the construct had proven the integrity of the mRNA, one might conclude that the resulting polypeptide was cleaved into its two components during or after translation.

In a second approach we then exchanged the linker between the two transporter molecules by a myc epitope tag resulting in a SERT-myc-NET concatenate (SmycN). The integrity of the resulting fusion protein was confirmed by SDS–PAGE and immunoblotting with an antibody directed against the myc epitope, which connects the two transporter monomers in our construct. Under reducing conditions this antibody detected SmycN at about 120 kDa and 250 kDa; no bands were seen at molecular weights corresponding to monomeric SERT or NET. Under non-reducing conditions, almost all myc-specific immunoreactivity was seen at 250 kDa. These data suggest that the 120-kDa band represents the intact ‘monomeric’ SERT–NET dimer containing 24 TMDs and the 250-kDa band a disulphide-stabilized dimeric form of this construct, that is a tetrameric complex consisting of two molecules each of SERT and NET. This finding supports earlier studies on plasma membrane-bound monoamine transporters which also suggested dimeric and/or tetrameric structures for these membrane proteins (Milner et al. 1994; Jess et al. 1996; Chang et al. 1998; Kilic and Rudnick 2000; Schmid et al. 2000). Our findings are also consistent with those obtained for the unrelated hexose transporter GLUT1 which has been shown to exist as a mixture of homodimers and homotetramers with an intermolecular disulphide bridge promoting tetramerization (Hebert and Carruthers 1991; Zottola et al. 1995). This polypeptide is also an integral membrane protein with 12 putative TMDs but does not display any sequence homology to the related monoamine transporters DAT, NET and SERT. One might hypothesize therefore that di/tetramerization might reflect a general mode of function of transporter proteins. In addition to ligand-gated ion channels G protein-coupled receptors, which are intrinsic transmembrane proteins with seven TMDs, have also been shown to function as oligomers in vivo (Overton and Blumer 2000). Interestingly, the DA D2 receptor and the somatostatin SSTR5 receptor interact physically by hetero-oligomerization to create a novel hybrid receptor with enhanced functional activity (Rocheville et al. 2000).

Many studies on the quaternary structure of monoamine transporters were based on biophysical and/or biochemical methods and did not include investigation of the functionality of the transporter proteins. Heterologous expression of a fusion protein consisting of one molecule each of SERT and NET allowed analysis of the transport parameters of both monoamines, i.e. [3H]5-HT and [3H]NE, mediated by the fusion protein. In addition, the existence of the highly selective 5-HT and NE transport inhibitors citalopram and nisoxetine as well as of the non-selective antidepressant imipramine made a detailed pharmacological characterization of the fusion protein feasible. Kinetic studies revealed that [3H]NE transport parameters of SmycN were comparable to those of NET and that the affinity of [3H]5HT was twofold lower at SmycN than at the SERT. Performing [3H]5HT uptake studies in the presence of unlabelled NE and vice versa showed that the two monoamines did not influence each other's transport mediated by SmycN up to 10 µm. Interestingly, at higher micromolar concentrations 5-HT competed with [3H]NE for uptake by NET, but NE did not compete with [3H]5-HT for uptake by SERT. This was also observed when analysing the single transporters. These findings are in agreement with the NE uptake by the NET in isolated rat tissues (Burgen and Iversen 1965; Paczkowski et al. 1996). In addition, we found that NE did not affect [3H]5HT uptake in the presence of nisoxetine, i.e. when the NE-binding site was blocked, and that 5-HT did not inhibit [3H]NE uptake in the presence of citalopram, i.e. when the 5-HT-binding site was blocked. These findings confirm the specificity of two distinct substrate-binding sites of the two components of the hybrid construct.

We then investigated the specificity and inhibitory affinity of citalopram, nisoxetine and imipramine. Citalopram inhibited [3H]5-HT uptake with similar efficiency at both the SERT and SmycN, but had no significant effect on [3H]NE uptake at the NET or SmycN. Comparably, nisoxetine inhibited [3H]NE uptake with similar efficiency at both the NET and SmycN, but did not significantly affect [3H]5-HT uptake at the SERT or SmycN. The TCA imipramine blocked [3H]5-HT uptake with similar efficiency at the SERT and SmycN, and [3H]NE uptake at the NET and SmycN. These data further support the existence of distinct, non-interacting substrate-binding sites on the chimeric construct. To analyse the antidepressant-binding sites, we first characterized the binding of labelled antidepressants to membrane preparations of cellsS expressing mycN. Saturation analyses revealed that citalopram and nisoxetine displayed binding parameters at SmycN comparable to those obtained at the SERT and NET, when expressed individually. Imipramine, which bound to the NET with about fourfold lower affinity than the SERT, bound to SmycN with an intermediate, but not significantly different, affinity. This might reflect the sum of [3H]imipramine binding to its distinct binding sites on the SERT and NET components of SmycN. Non-linear regression analysis for two binding sites, however, could not distinguish these sites, because their affinities for imipramine are of the same order of magnitude.

Analysis of the stable lines revealed, as expected, the same number of binding sites for citalopram and imipramine on the SERT as of nisoxetine and imipramine on the NET. Actually, we would have also expected the same number of the three ligands to the SmycN construct, unless imipramine binding to both sites on SmycN is additive; then one might have expected twice the Bmax for imipramine. As shown in Table 2, the Bmax values of imipramine and citalopram for the SmycN construct were the same, whereas that for nisoxetine was significantly lower. This might be explained by the fact that the NET part of our SmycN construct is partly hindered with respect to imipramine and nisoxetine binding as well as in NE uptake (see Fig. 3c). The finding that the KD values for nisoxetine and imipramine and the KM value for NE uptake did not differ significantly from the respective values observed in the wild-type NET suggests that in some portion of the SmycN molecules the NET part is not correctly expressed or properly folded. It would be interesting to see whether comparable results are obtained for citalopram and imipramine binding to NmycS, i.e. a construct in which the SERT part is covalently linked behind the NET molecule.

To analyse the interaction of antidepressant binding at SmycN in more detail, mutual competition experiments were performed. Imipramine inhibited [3H]citalopram binding to membranes expressing SERT and SmycN with the same potency, and also [3H]nisoxetine binding to the NET and SmycN. [3H]Imipramine binding to the SERT and NET was completely displaced by citalopram and nisoxetine respectively (Fig. 5). At SmycN, however, citalopram at a concentration of 96 nm displaced about 70%, and nisoxetin (1 µm) about 30%, of 20 nm[3H]imipramine binding (Fig. 5). Calculating the occupation of imipramine-binding sites as L/(KD + L) × 100, where L represents the ligand concentration used, reveals that 50% of imipramine-binding sites on SmycN are occupied at 20 nm. As the KD value for the imipramine-binding site at the SERT molecule is 10 nm, and that at the NET molecule is 40 nm, at 20 nm 66% of the SERT-related imipramine-binding sites, and 33% of the NET-related imipramine-binding sites are occupied. Given that these sites do not interact with each other, citalopram theoretically displaces about 66%, and nisoxetine about 33%, of bound imipramine at the SmycN construct. These findings are depicted in the two-dimensional drawing in Fig. 6.

Figure 6.

Proposed interactions of antidepressants at SmycN. The fusion protein transports [3H]5-HT and [3H]NE with parameters comparable to those of the wild-type SERT and NET proteins. 5-HT transport is sensitive to citalopram but insensitive to nisoxetine; NE transport is sensitive to nisoxetine, but not to citalopram. Imipramine fully inhibits both, the uptake of the monoamines, as well as the binding of [3H]citalopram and [3H]nisoxetin. [3H]Imipramine, however, is only partly displaced from SmycN by citalopram and nisoxetine, suggesting that the distinct antagonist-binding sites on the single components of the hybrid construct do not interact with each other.

Taken together, our studies clearly show that upon covalent linkage the related transporters for 5-HT and NE can form fully functional fusion proteins which form disulphide-stabilized oligomers under non-reducing conditions. The finding that the distinct substrate- and antagonist-binding sites on the single components of the hybrid construct do not interact with each other implies that these components can also function as monomers.

Acknowledgements

We thank Werner Köhl for expert technical assistence, Drs Heinz Bönisch and Michael Brüss for the generous gift of the NET cDNA, and Drs Peter Gebicke and Rainer Spanagel and Clive Williams for critical reading of the manuscript and helpful discussions. This work was supported by the Deutsche Forschungsgesellschaft (DFG to SCHL 353/4–1).

Ancillary