Ionic Currents in the Human Serotonin Transporter Reveal Inconsistencies in the Alternating Access Hypothesis

We have investigated the conduction states of human serotonin transporter (hSERT) using the voltage clamp, cut-open frog oocyte method under different internal and external ionic conditions. Our data indicate discrepancies in the alternating access model of cotransport, which cannot consistently explain substrate transport and electrophysiological data. We are able simultaneously to isolate distinct external and internal binding sites for substrate, which exert different effects upon currents conducted by hSERT, in contradiction to the alternating access model. External binding sites of coupled Na ions are likewise simultaneously accessible from the internal and external face. Although Na and Cl are putatively cotransported, they have opposite effects on the internal face of the transporter. Finally, the internal K ion does not compete with internal 5-hydroxytryptamine for empty transporters. These data can be explained more readily in the language of ion channels, rather than carrier models distinguished by alternating access mechanisms: in a channel model of coupled transport, the currents represent different states of the same permeation path through hSERT and coupling occurs in a common pore.

In this article, we employ the cut-open Xenopus oocyte voltage clamp (COVC) technique to measure ionic currents conducted by hSERT. COVC ( Stefani and Bezanilla, 1998 ; Kaneko et al., 1998 ; Costa et al., 1994 ) allows simultaneous, dynamic access to the internal and external face of the transporter, so that interactions between substrate, ions, and hSERT may be probed in detail. Ionic current through transporters is an assay for transporter conformation, whether or not the current itself represents 5HT transport. Thus we can use ionic currents to test functional models, such as the alternating access model, which postulate specific relationships between SERT conformational states. This method has revealed inconsistencies in the alternating access model of transporter function.

Radiolabeled tracer flux studies have long been the standard technique employed to study transport and coupling in SERT ( Rudnick, 1998 ). Often, to explain transport data these studies focus on the “alternating access” hypothesis, in which binding sites on hSERT for 5HT and ions alternately face extracellular or cytoplasmic compartments. Within this framework, coupling results from conformational changes in the transporter induced by substrate and ion binding ( Jardetsky, 1966 ). Flux experiments have resulted in a widely accepted electroneutral scheme for SERT in which Na, Cl, and 5HT are cotransported, and K or H is countertransported ( Nelson and Rudnick, 1979 ; Rudnick and Clark, 1993 ; Gu et al., 1994 , 1996 ).

Subtracted electrophysiological currents are shown, except in the raw traces in . Subtraction was performed after averaging each trace over the last 25 ms of each potential step, and was defined as I (presence of activator) − I (R96ext), for 5HT-induced current (I 5HT ), Li-induced current (I Li ), and desipramine (DS)-revealed current (I Leak ). For I Leak , the subtraction procedure results in a signature negative slope conductance ( Mager et al., 1994 ). Subtracted currents are normalized to the most hyperpolarized potential value (−80 or −100 mV) for each oocyte. This allows for comparison and pooling among many oocytes with differing expression levels. Error bars on data plots represent the standard error of the normalized means of 3–6 oocytes.

We employed the cut-open oocyte voltage clamp recording technique ( Stefani and Bezanilla, 1998 ; Kaneko et al., 1998 ) as modified by Costa et al. (1994) . Originally designed for application to voltage-activated ion channels, COVC has been applied previously to the glucose cotransporter SGLT1 ( Chen et al., 1995 , 1996 ) and the amino-acid exchanger rBAT ( Coady et al., 1994 , 1996 ). This technique allows reproducible manipulation of the internal and external ionic environment of the oocyte (see ). N-methyl-D-glucamine (NMDG) replaced cations and methanesulfonate (MES) replaced Cl as described for individual experiments (see Results below). Internal Oocyte Buffer (IOB (mM): 50 KCl, 70 K-MES, 5 Na-MES, 10 HEPES, 2 MgCl 2 , 0.5 EGTA, 0.1 ascorbic acid, 0.1 pargyline; pH 7.4 with NMDG-OH) was perfused at 10–30 μL/min. Total osmolarity, and final pH of all solutions, was always verified to ensure consistency. We found that completely changing the internal ionic concentration took 5–30 min, depending upon pipette diameter and its distance from the internal face of the membrane. Therefore, we performed only one such solution change per oocyte, pairing the experimental condition to a standard baseline. For at least one oocyte in each experimental internal solution, we were able to return the internal solution to the baseline solution, demonstrating that the effects of internally applied ions and substrate are reversible. External solution (R96ext) consisting of R96+Ca, with additional 0.1 mM ascorbic acid and 0.1 mM pargyline, was applied to the oocyte at ∼1 mL/min through a movable, sawed-off 18-gauge syringe placed near the oocyte dome. External solution change was therefore rapid. Reported data report only currents that are activated by external application of 5HT, DS, or Li that washed out (to 10%) upon return to R96ext solution. Typically, each experiment (a single oocyte) lasted 60–90 min. In our apparatus, the agar bridges (1–2% agarose, 120 mM Na-MES, 10 mM HEPES, pH 7.4 with NaOH) connected the solution baths to wells filled with 1 M NaCl and Ag/AgCl pellets, relaying electrophysiological currents to the amplifier (Dagan, Minneapolis, MN). Voltage protocols were applied using AxoClamp 7.0 software (Axon Instruments, Union City, CA) on a desktop PC. Data were filtered at 1 kHz by the amplifier, and digitized at 2 kHz by the acquisition software. In most experiments, we partially compensated the membrane transients to speed the membrane voltage clamp. External solutions were controlled with electronic values and pressurized reservoirs set to 1 PSI (Automate Scientific, San Francisco, CA).

Oocytes were surgically removed from mature Xenopus laevis anaesthetized with 0.2% tricaine. Surgery was performed aseptically and in compliance with Vanderbilt University regulations. After removal from the frog, the follicular tissue containing the oocytes was placed in Ringer’s buffer (R96 (mM): 96 NaCl, 2 KCl, 5 MgCl 2 , 5 HEPES) with 10 μg/mL gentamicin, and gently cut into small pieces. Aliquots of this tissue (3–4 mL each) were added to 10 mL R96 with 0.2% collagenase, and nutated for 90–110 min. Collagenase treatment was terminated by repeated washing in R96 + 0.6 mM CaCl 2 (R96 + Ca) when visual inspection of the oocytes revealed that the majority of the oocytes were not surrounded by follicular tissue. We selected stage V-VI oocytes for cRNA injection within 24 h of oocyte isolation. cRNA was transcribed from cDNA in the pOTV vector (gift of Dr. M. Sonders, Vollum Institute), using Ambion mMessage Machine T7 kit (Ambion, Austin, TX). Each oocyte was injected with 23 ng cRNA and incubated at room temperature (23 ± 3°C) for 4–8 days, in R96+Ca supplemented with 550 μg/mL NaPyruvate, 100 μg/mL streptomycin, 50 μg/mL tetracycline, and 5% dialyzed horse serum ( Elsner et al., 2000 ; Quick et al., 1992 ; Goldin, 1992 ; Stuhmer, 1998 ).

We observed changes in background currents as [Cl] in was increased or decreased from its normal value (near 50 mM) in both uninjected (data not shown) and injected oocytes, presumably due to endogenous Ca-activated Cl channels ( Dascal, 1987 ; Kuruma and Hartzell, 1999 ). IOB contains 0.5 mM EGTA, therefore endogenous Cl channel activity and background currents are moderate. We varied [Cl] in from 0 mM to 120 mM by substitution with MES ions. With [Cl] in = 0 mM, 5HT out activates less I 5HT compared to [Cl] in = 50 mM ( ). In sharp contrast to Na in ( ), increasing [Cl] in from 50 mM to 120 mM caused almost no change in I 5HT ( ). I Li ( ) and I Leak ( ) were little influenced by [Cl] in . In addition, [Cl] in had only a small effect on the potency of 5HT out on hSERT ( ).

To assess the possible interdependence of internal 5HT and internal K we perfuse the cytosol with K-free IOB until obtaining a stable response to external 5HT and external Li. During this stabilization period we observe a steady decrease in evoked currents, consistent with , since the oocyte begins with [K] in ∼120 mM. After reaching [K] in = 0, we record I 5HT , I Li , and I Leak , then exchange the internal solution so that it contains [5HT] in as indicated in . shows that even with [K] in = 0 mM, [5HT] in completely inhibits I 5HT , with IC 50 = 91 ± 21 μM. This value is indistinguishable from the IC 50 obtained with [K] in = 120 mM. Similarly, internal 5HT inhibited I Li at [K] in = 0, with maximal efficacy ∼50% and IC 50 ∼150 μM (Figs. 6 B and 10). Finally, in the absence of internal K, I Leak remains insensitive to internal 5HT ( ). Repetition of this experiment with [K] in = 60 mM (data not shown) showed that [5HT] in inhibits I 5HT at −80 mV, with IC 50 = 119 ± 19 μM; [5HT] in also inhibits I Li , it but has no effect on I Leak .

We next explore the effect of putatively countertransported K, by replacing internal K with NMDG. The effect of reducing [K] in from 120 mM to 0 mM is to inhibit I 5HT by 60% at −80 mV ( ). The EC 50 for internal K is ∼40 mM ( ). Similar to I 5HT , I Li ( ) is decreased by reducing [K] in , with EC 50 ∼40 mM, and maximal efficacy ∼60% at −80 mV. The leak current revealed by DS application appears insensitive to [K] in ( ). Finally, reducing [K] in also increased the potency of external 5HT, as demonstrated by a decreased EC 50 for external 5HT to elicit I 5HT ( ).

We next checked whether the effects of [Na] in and [5HT] in are independent. For this experiment ( ), [Na] in was increased to 25 mM, and hSERT currents measured. We then introduced 5HT to the internal solution. Comparison of with reveals that increasing [Na] in changes the effect of [5HT] in on I 5HT , compared to its effect in low (5 mM) [Na] in . With [Na] in = 25 mM and [5HT] in = 1 mM, I 5HT is outward at hyperpolarized potentials and reverses near V = 0. This current has the negative-slope conductance characteristic of I Leak under the same experimental conditions ( ). shows that elevated [5HT] in inhibits I Li in the presence of elevated [Na] in , consistent with and . Finally, although I Leak is altered by [Na] in , shows that the altered I Leak remains insensitive to the presence of internal 5HT.

Next we investigated the interaction of Na with the cytosolic face of the hSERT ( ). Standard IOB has 5 mM Na and 120 mM K, similar to the native Na content of the oocyte ( Dascal, 1987 ). To reduce complications due to simultaneously lowering [K] in and elevating [Na] in , we set [K] in = 50 mM as the baseline for experiments with elevated [Na] in . As [Na] in increased, we noted a sharp increase in conductance in the oocyte at depolarized potentials. We observed identical behavior in uninjected oocytes (data not shown), but did not attempt to discover the mechanism responsible. Instead, we simply waited for background currents, I 5HT , and I Li to stabilize as [Na] in was altered. The results ( ) show that [Na] in potently inhibits I 5HT , with IC 50 ≪25 mM and with maximal efficacy approaching 100%. Surprisingly, [Na] in has only a weak effect on I Li ( ). In sharp contrast, [Na] in alters I Leak ( ), shifting the reversal potential from +20 mV to −30 mV, consistent with Na permeation contributing to I Leak . However, the apparent outward current at −100 mV is unchanged.

Although many previous studies have examined the properties of ionic currents mediated by SERT when external 5HT is applied (I 5HT ) ( Mager et al., 1994 ; Cao et al., 1997 , 1998 ; Petersen and DeFelice, 1999 ; Ramsey and DeFelice, 2002 ), little is known about the effect of internal 5HT. shows a diagram of the cut-open voltage clamp setup. Because exchange of the internal milieu is relatively slow in COVC, 5HT could not be acutely applied to the internal face of hSERT. Rather, the current evoked by 5HT out was measured in different levels of 5HT in . The voltage protocol was either single pulses to the test potential (separated by 1 s) or the staircase protocol displayed in . The two protocols gave the same results. As shown in , application of [5HT] out = 15 μM induced an inward current at test potentials between −100 mV and +20 mV, compared to the background current measured in the absence of 5HT out . A plots I 5HT as a function of the membrane potential for each [5HT] in tested. Every test [5HT] in was paired to a control at [5HT] in = 0 in the same oocyte and normalized at −80 mV. At hyperpolarized potentials, internal 5HT inhibits I 5HT . At depolarized potentials, however, 5HT in does not significantly inhibit I 5HT . We also determined that elevating [5HT] in increases the potency of [5HT] out to elicit I 5HT , as shown in . In each case, the I 5HT elicited by varying [5HT] out was fit to I 5HT = I max [5HT] out /(EC 50 + [5HT] out ). We additionally measured hSERT-mediated currents evoked by the substitution of Li ions for Na ions in the external buffer (I Li ) ( ). Li-induced current is also inhibited by internal 5HT, with approximately the same potency as I 5HT . However, the efficacy of [5HT] in was lower, as maximal [5HT] in = 1 mM inhibited the Li current by >50% ( ). Previous studies of SERT ( Petersen and DeFelice, 1999 ; Mager et al., 1994 ) show that 5HT out blocks I Li . Finally, we measured the effect of [5HT] in on the constitutive leak current mediated by hSERT, I Leak , by applying 2 μM DS, an hSERT antagonist. Previous experiments showed that 2 μM DS represents a saturating dose of DS (IC 50 = 0.49 ± 0.08 μM; n = 1.5 ± 0.25; data not shown). shows that the constitutive leak current is insensitive to [5HT] in . Uninjected oocytes showed no response to external 5HT or internal 5HT, a small outward (at −40 mV) response to external Li, and no response to external DS (data not shown). shows 5HT in inhibition of I 5HT and I Li at −80 mV. For I 5HT the solid line is fit to the equation: I = 1/(1 + [5HT] in /IC 50 ), with IC 50 = 99 ± 26 μM. For I Li , the solid line is drawn by eye, because this equation produced unsatisfactory fit parameters. Both curves are normalized to 1.0 at 0 5HT in .

DISCUSSION

The COVC technique provides simultaneous access to the internal and external milieu of the oocyte as well as rapid voltage control over a sizable portion (∼10%) of the oocyte membrane. As the ionic and voltage conditions facing the hSERT protein are systematically changed, the ionic currents conducted by hSERT are monitored. These ionic currents are ascribed to hSERT because they resemble currents through SERTs (Mager et al., 1994; Galli et al., 1997; Petersen and DeFelice, 1999) and other GAT/NET family members (Mager et al., 1993; Sonders et al., 1997), they are blocked by desipramine, and they present only in oocytes injected with cRNA encoding hSERT.

hSERT conducts several distinct steady-state currents, including substrate-induced current (I5HT), substrate-independent leak current (ILeak), and Li-induced current (ILi) activated by replacement of Na with Li on the external face (in the absence of 5HT). Other currents have also been noted, for example, in acidic (external) pH (Cao et al., 1997). Furthermore, SERTs exhibit transient currents under appropriate conditions (Mager et al., 1994; Li et al., 2002). We have limited our measurements to I5HT, ILeak, and ILi steady-state currents to test prevalent hypotheses about hSERT function.

Because changes in ionic currents result from changes in the fractional occupancy of the conducting conformational states available to hSERT, the sensitivity of I5HT, ILeak, and ILi to various ions will inform our discussion of how the transport of substrate couples to ionic gradients. Thus, a specific model, such as the alternating access model, which postulates distinct relationships between hSERT conformational states, may be rigorously tested.

ILeak insensitivity to internal substrate

hSERT conducts current in the absence of external substrate ( ; Mager et al., 1994). This leak current requires external Na, but must permeate a state of hSERT without bound 5HT. Therefore, ILeak may be reasonably assigned to the state TN0 of . Measurements of ILeak thus reflect the fractional population of TN0.

Open in a separate window

demonstrates the intriguing result that ILeak is insensitive to internal 5HT. On the basis of a carrier model of SERT function (distinguished by alternating access, such as ), it might be expected that elevation of [5HT]in would draw transporters out of TN0 and reduce the measured leak. On the other hand, if the dissociation constant for external Na is much smaller than the applied [Na]out, the transition TN0→T0 would be forbidden. In the absence of external 5HT, transporters could be expected to accumulate in TN0, and thus be insensitive to the internal milieu. The carrier model can explain the insensitivity of ILeak to [5HT]in by supposing that in the absence of substrate on both sides of the membrane, the molecular rate constants of dictate that all 5HT binding sites are oriented externally, and hence unavailable to 5HTin.

Suppression of I5HT by internal 5HT and internal Na

Unlike ILeak, I5HT is suppressed by elevated [5HT]in ( ) or [Na]in ( ). The alternating access model ( ) explains this suppression by inferring that increased [5HT]in or [Na]in leads to sequestration of substrate binding sites on the internal side of the membrane. Therefore elevated [5HT]in or [Na]in results in an reduction in the average number of external binding sites for 5HT, and a concomitant reduction in the current evoked by 5HTout.

Assuming that SERT conducts I5HT in the state TNS0, this result implies that the transporters, once in the state TNS1, release 5HT and Na to the internal milieu more quickly than they reorient to the outward facing conformation via TNS1→TNS0. If SERT reoriented more quickly, [5HT]in would potentiate I5HT since, on average, SERT would spend more time in the state TNS0. Because elevated [5HT]in or [Na]in decreases the current evoked by external 5HT, a general conclusion is that the rate at which transporters cycle clockwise around exceeds the rate at which loaded transporters exchange substrate across the membrane.

These data are qualitatively in conflict with a standard result obtained with labeled substrate flux experiments, namely, the phenomenon of transacceleration (Stein, 1986). Previous reports (Sitte et al., 2000, 2001; Adams and DeFelice, 2002) demonstrate that efflux of 5HT through SERT is increased as [5HT]out is increased. This effect is referred to as transacceleration, and within the framework of the carrier model of transport it implies that the rate of equilibrium exchange of substrate is greater than the forward cycling of the transporter (Stein, 1986). That is, for cells loaded with high [5HT]in, transacceleration occurs because as [5HT]out increases, the quicker exchange transition TNS1→TNS0 dominates over the rate at which transporters proceed from TNS1 clockwise around the transport cycle ( ). Thus, if transacceleration measurements (Sitte et al., 2000, 2001; Adams and DeFelice, 2002) are accurate and correctly interpreted within the prevailing carrier model ( ), elevated [5HT]in would increase I5HT, because inward-facing transporters (TNS1) would quickly reorient to the conducting state (TNS0). This expectation is, however, opposite to the electrophysiological measurements ( ).

In addition, if the carrier model is evoked to explain the suppression of I5HT by assuming that transporters reorient toward the internal face when [5HT]in or [Na]in is elevated, then the result displayed in is inexplicable. In this experiment, both [Na]in and [5HT]in are elevated. According to the carrier model, suppression of I5HT under these conditions indicates that the transporters are facing inward. However, shows that application of external 5HT still interacts with SERT to effectively block ILeak, implying that many transporters still have external binding sites. Thus a carrier model in which the suppression of I5HT by internal 5HT ( ) or Na ( ) is predicted cannot also predict the result shown in .

The carrier model ( ) is unable to consistently explain the electrophysiological data ( , , and ) and the body of transacceleration data (Sitte et al., 2000, 2001; Adams and DeFelice, 2002). Instead, we speculate that there may be distinct internal and external modes of interaction between hSERT and 5HT. Transacceleration can be explained by a functional model in which external 5HT binds to an external site and opens a pore through hSERT, increasing the rate of efflux through the pore in apparent contradiction to the driving force. The large suppression of I5HT by internal 5HT might not be entirely explained by a straightforward reduction in the driving force for inward 5HT permeation through the pore, because it has been demonstrated that permeating 5HT itself carries, at most, 15% of the measured I5HT (Galli et al., 1997).

However, in a complex, single-file pore, the suppression of current by internal 5HT can be nonlinearly related to the reduction in the driving gradient. Thus, 5HT may in effect occlude the pore for I5HT ( ). Likewise, the suppression of I5HT by [Na]in elevation ( ) may result from decreased driving force for net inward Na permeation through a complex pore. External Li activates a pore, which does not support transport of 5HT, but internal 5HT can enter and block the pore ( ). Additional conformational changes in hSERT may also be necessary to explain the increase in external 5HT potency in the presence of internal 5HT ( ).

Internal K does not compete with internal 5HT for internal hSERT binding sites

The idea that hSERT countertransports K, with respect to 5HT, rests on the observation that elevated [K]in increases the rate of 5HT transport (Rudnick and Nelson, 1978; Nelson and Rudnick, 1979). This has been incorporated into the alternating access model by supposing that internal K accelerates the rate at which internally facing 5HT-binding sites reface to the external orientation (Rudnick and Nelson, 1978; ). We therefore sought to test how changing [K]in affects hSERT currents ( ). Consistent with the transport data, these experiments reveal that increasing [K]in increases I5HT, suggesting a larger average number of available external 5HT binding sites when [K]in is elevated. Furthermore, we also observe a decrease in external 5HT potency with increased [K]in ( ), as predicted by the countertransport hypothesis ( ; Nelson and Rudnick, 1979).

Therefore we combined the experiments of and , to study directly the interaction between internal K and internal 5HT at hSERT. shows that Kin is not required for 5HTin to suppress I5HT, and we also found that internal K does not change the potency with which internal 5HT suppresses I5HT. In the specific K countertransport model proposed previously ( ; Nelson and Rudnick, 1979), internal 5HT and K compete for internally facing, empty transporters. If this were the case, we expect that the potency of 5HT would decrease with increasing [K]in, counter to our observations. Proton (H) substitution for K on the internal face of hSERT does not effect our results since at our experimental pHin = pHout = 7.4, proton stimulation of 5HT transport through SERT is negligible even in the absence of K (Keyes and Rudnick, 1982).

Furthermore, demonstrates that internal K acts to potentiate ILi as well as I5HT. Since ILi cannot be associated with a transport-competent mode of hSERT (Ni et al., 2001; Petersen and DeFelice, 1999), the hypothesis that Kin acts to accelerate refacing of the transporter does not apply.

To explain the acceleration of transport by K (Nelson and Rudnick, 1979) and the reduction of I5HT and ILi by removal of Kin ( ), we postulate that internal K interacts with hSERT at a modulatory site, which is available only when 5HT or Li is bound to their external sites. The mechanism of K modulation could be that K acts to stabilize the open states of the 5HT-activated and Li-activated hSERT pore. In contrast, the internal K site is unavailable when hSERT is in the leakage state.

Na and Cl affect hSERT differentially

Although the putative transport cycle of hSERT cotransports one Na and one Cl (Rudnick and Clark, 1993) we observe that these ions have opposite effects on I5HT from the internal face of the transporter ( and ). If the reduction in I5HT by internal Na were due to sequestration of the binding site to the internal face of the membrane, it is difficult to see how the other co-ion, Cl, could have an opposite effect. GAT1 current is reported to display the opposite behavior with respect to internal Cl (Lu and Hilgemann, 1999). In that case, the measured suppression of GABA-induced current through GAT1 by internal Cl was interpreted as evidence that Cl facilitates the release of GABA from the internally facing transporter substrate site, after cotransport of GABA and Cl (Hilgemann and Lu, 1999). Clearly, such a role for Cl in hSERT is not supported by our data. Rather, Cl− likely plays a regulatory role in hSERT function. In support of our interpretation, Loo et al. (2000) report that Cl is exchanged in GAT1 and uncoupled to substrate transport, and Cl does not contribute to net transported charge.

As we have argued for K, it is possible that Cl may bind hSERT and influence the relative permeability of 5HT and Na. This would alter the magnitude and pharmacology of the observed I5HT, and could make it appear that Cl ions are coupled to 5HT transport. Our data ( ) are consistent with Cl permeation contributing to I5HT, but a previous report strongly argues against this possibility (Lin et al., 1996). In any case, we cannot conceive a way in which both Cl and Na are cotransported with 5HT, yet have opposite effects on I5HT, within the prevailing alternating access models.