Open probability of homomeric murine 5-HT_3A serotonin receptors depends on subunit occupancy

Abstract

1. The time course of macroscopic current responses of homomeric murine serotonin 5-HT_3A receptors was studied in whole cells and excised membrane patches under voltage clamp in response to rapid application of serotonin.

2. Serotonin activated whole cell currents with an EC₅₀ value for the peak response of 2 μm and a Hill slope of 3.0 (n = 12), suggesting that the binding of at least three agonist molecules is required to open the channel.

3. Homomeric 5-HT_3A receptors in excised membrane patches had a slow activation time course (mean ±s.e.m. 10-90 % rise time 12.5 ± 1.6 ms; n = 9 patches) for 100 μm serotonin. The apparent activation rate was estimated by fitting an exponential function to the rising phase of responses to supramaximal serotonin to be 136 s⁻¹.

4. The 5-HT_3A receptor response to 100 μm serotonin in outside-out patches (n = 19) and whole cells (n = 41) desensitized with a variable rate that accelerated throughout the experiment. The time course for desensitization was described by two exponential components (for patches τ_slow 1006 ± 139 ms, amplitude 31 % τ_fast 176 ± 25 ms, amplitude 69 %).

5. Deactivation of the response following serotonin removal from excised membrane patches (n = 8) and whole cells (n = 29) was described by a dual exponential time course with time constants similar to those for desensitization (for patches τ_slow 838 ± 217 ms, 55 % amplitude; τ_fast 213 ± 44 ms, 45 % amplitude).

6. In most patches (6 of 8), the deactivation time course in response to a brief 1-5 ms pulse of serotonin was similar to or slower than desensitization. This suggests that the continued presence of agonist can induce desensitization with a similar or more rapid time course than agonist unbinding. The difference between the time course for deactivation and desensitization was voltage independent over the range -100 to -40 mV in patches (n = 4) and -100 to +50 mV in whole cells (n = 4), suggesting desensitization of these receptors in the presence of serotonin does not reflect a voltage-dependent block of the channel by agonist.

7. Simultaneously fitting the macroscopic 5-HT_3A receptor responses in patches to submaximal (2 μm) and maximal (100 μm) concentrations of serotonin to a variety of state models suggests that homomeric 5-HT_3A receptors require the binding of three agonists to open and possess a peak open probability greater than 0.8. Our modelling also suggests that channel open probability varies with the number of serotonin molecules bound to the receptor, with a reduced open probability for fully liganded receptors. Increasing the desensitization rate constants in this model can generate desensitization that is more rapid than deactivation, as observed in a subpopulation of our patches.

Serotonin is a biogenic amine that activates a large and heterogeneous family of receptors that are widely distributed throughout the central and peripheral nervous systems (Boess & Martin, 1994; Jackson & Yakel, 1995). The 5-hydroxytryptamine type 3 (5-HT₃) receptor was first defined pharmacologically (Richardson & Engel, 1986), and subsequently found to be the only ligand-gated ion channel in the serotonin receptor family (Derkach et al. 1989). A cDNA encoding the 5-HT_3A receptor subunit that was initially isolated (Maricq et al. 1991) appears structurally related to the neuronal nicotinic α7 and α8 subunits (Eisele et al. 1993; Ortells & Lunt, 1995). The 5-HT_3A subunit assembles as a pentameric (Boess et al. 1995) homo-oligomeric complex in heterologous expression systems. It is therefore likely that the receptor has five equivalent binding sites, although these may not all need to be occupied to efficiently activate the receptor. Sequence analysis (Ortells & Lunt, 1995) and transmembrane topology (Mukerji et al. 1996; Spier et al. 1999) suggest that the 5-HT_3A receptor is a member of a superfamily of pentameric protein complexes that includes the nicotinic acetylcholine, γ-aminobutyric acid type A and glycine receptors. Recombinant homomeric 5-HT_3A receptors reproduce many but not all of the properties of pharmacologically defined 5-HT₃ receptors in peripheral and central neurons, suggesting additional heterogeneity may exist through assembly with other subunits (Hussy et al. 1994; Gill et al. 1995; Van Hooft et al. 1997). Recent studies have identified a cDNA encoding a second 5-HT₃ receptor subunit (5-HT_3B) that does not form functional homomeric receptors, but can co-assemble with 5-HT_3A subunits (Davies et al. 1999; Dubin et al. 1999). In addition, the neuronal nicotinic α4 receptor subunit can functionally coassemble with 5-HT_3A subunits (Van Hooft et al. 1998; Kriegler et al. 1999).

A high density of 5-HT₃ receptor binding sites, 5-HT_3A mRNA, and 5-HT_3A receptor immunoreactivity have been found in hindbrain areas, with moderate densities in cortex, amygdala and hippocampus (Bufton et al. 1993; Tecott et al. 1993, 1995; Johnson & Heinemann, 1995; Spier et al. 1999). In these latter areas 5-HT₃ receptors often appear localized in inhibitory GABAergic interneurons (Kawa, 1994; Bloom & Morales, 1998). Although the extent to which 5-HT₃ receptors participate in fast synaptic transmission remains unclear (Ropert & Guy, 1991; Sugita et al. 1992; Roerig et al. 1997), these receptors have received attention for their role in several clinical settings. 5-HT₃ receptors in area postrema are the likely target of clinically useful antagonists that prevent cytotoxic drug-evoked or irradiation-induced emesis (Blower, 1990; Viner et al. 1990; Aapro, 1991; Hasler, 1999; Lehoczky, 1999; Perez, 1999), as well as emesis in relation to migraine (Dahlof & Hargreaves, 1998). In addition, agents acting at these receptors might also have anxiolytic and antipsychotic properties that potentially could be of therapeutic interest (Costall et al. 1988; Jones et al. 1988; Greenshaw & Silverstone, 1997; Tancer & Uhde, 1997). Furthermore, the role of 5-HT₃ receptors in modulating activity in the myenteric plexus suggests that 5-HT₃ receptor antagonists might be useful for treatment of irritable bowel syndrome (Greenshaw & Silverstone, 1997; Farthing, 1999).

In this study we have explored serotonin 5-HT_3A receptor kinetics using a rapid agonist application system capable of submillisecond solution exchange. We describe here in detail the time course for serotonin activation and deactivation of homomeric 5-HT_3A receptors in excised membrane patches. In addition, we propose that in some circumstances the rate of desensitization can exceed the rate of deactivation. A kinetic model is presented that is consistent with our results, and predicts that open probability of 5-HT_3A receptors varies with subunit occupancy in a biphasic fashion. Some of these results have appeared in preliminary form (Mott & Traynelis, 1996; Traynelis & Mott, 1996, 1997).

METHODS

Tissue culture

Human embryonic kidney cells (HEK 293 cell line) stably transfected with 5-HT_3A cDNA in the expression vector pLNCX and cDNA encoding the murine 5-HT_3A receptors were generously provided by Dr D. Julius (University of California, San Francisco, CA, USA). Cells (passage numbers 5-25) were cultured in standard Dulbecco's modified Eagle's medium (Gibco 11960) containing high glucose and supplemented with 10 % fetal calf serum, glutamine (2 mm), sodium pyruvate (1 mm), penicillin (50 U ml⁻¹) and streptomycin (50 μg ml⁻¹). Stable transformants were selected by supplementing the medium with geneticin (1 mg ml⁻¹) to which the vector confers resistance. Cells were maintained in 60 mm culture plates at 37 °C in a humidified atmosphere containing 5 % CO₂, grown to about 80 % confluency, harvested enzymatically using 0.25 % trypsin, and dissociated further by gentle trituration. For electrophysiological recording, cells were plated at a density of 10⁶ cells ml⁻¹ on 12 mm glass coverslips coated with either poly-d-lysine (180 μg ml⁻¹) or fibronectin (20 μg ml⁻¹). Experiments were performed 2-5 days after plating cells. For some experiments HEK 293 cells that stably expressed 5-HT_3A receptors were transiently transfected as previously described (Traynelis & Wahl, 1997) with 0.1-1.0 mg ml⁻¹ GluR6(621Q) cDNA (provided by Dr Heinemann, Salk Institute, La Jolla, CA, USA) in a CMV-based mammalian expression vector (JG3.6). The reporter gene green fluorescent protein (GFP; 0.2-0.4 mg ml⁻¹; Columbia University) was used to identify individually transfected cells.

Electrophysiological recording

Transfected HEK 293 cells plated on glass coverslips were transferred to a perfusion chamber on the stage of a microscope, and continually perfused at a rate of 0.5 ml min⁻¹ with medium containing (mm): 150 NaCl, 3 KCl, 10 Hepes, 0.25 CaCl₂ and 0.25 MgCl₂ (pH 7.35). All experiments were performed at 23 °C. The concentration of divalent cations in the perfusion medium was lowered because divalent cations inhibit 5-HT₃ receptor responses (Peters et al. 1988). Whole cell and outside-out patch clamp recordings of agonist-evoked membrane currents were performed under voltage clamp conditions with electrodes containing (mm): 110 d-gluconic acid, 110 CsOH, 30 CsCl, 0.5 CaCl₂, 2 MgCl₂, 4 NaCl, 5 BAPTA, 5 Hepes, 2 Na-ATP, 0.3 Tris-GTP, unless otherwise stated. The pH of the electrode filling solution was adjusted to 7.3 using CsOH; osmolality was 290 mosmol kg⁻¹. Recording electrodes were made from borosilicate glass (inner diameter 1.15, outer diameter 1.65) fire-polished to a resistance of 4-6 MΩ. The membrane potential was held between -60 and -75 mV unless otherwise specified. Current recordings were amplified (Axopatch 200 or Warner PC 505A), filtered (1-3 kHz, -3 dB), and digitized at 3-13 kHz (pCLAMP 6.0-8.0). Series resistance errors were corrected either on-line or digitally after the experiment (Traynelis, 1998). Tip potentials were measured after each patch experiment and experiments with slow or multi-phasic exchange time courses were excluded. For whole cell recordings, cells were lifted off of the bottom of the dish to facilitate solution exchange. Cells with current responses larger than 1200 pA or rise times slower than 28 ms were excluded (n = 5) to minimize the chance that uncompensated series resistance errors might influence the time course of the fast component; exclusion of these data did not alter the results.

We used a piezoelectric-driven double-barrelled perfusion system to rapidly apply 5-HT₃ agonists and antagonists to whole cells or non-nucleated excised membrane patches. The application pipette was pulled from theta glass tubing (Hilgenberg, Malsfeld, Germany; 2 mm outer diameter, 0.3 mm wall thickness, 0.22 mm septum) and had a tip diameter of 200-300 μm, with the inner diameter of each barrel being 80-120 μm. Flow in each side of the theta tubing was typically controlled by a solenoid valve. Control solution flowed continuously through one barrel while the agonist solution flowed through the other barrel only during drug application. The agonist application barrel was pre-flushed for 1-2 s before piezo-driven application to clear diluted solution, and moved by means of a piezoelectric device causing the recorded cell or membrane patch to be transiently exposed to the agonist-containing solution. The time course of solution exchange across the laminar flow interface was estimated at the end of each experiment by liquid junction potential measurements, and was found to possess a 20-80 % rise time of 300-500 μs. For whole cell recordings, the solution exchange time was estimated by stepping the cell held at 0 mV in the absence of agonist into a solution in which 150 mm NaCl was replaced by KCl. The time course of the change in the holding current had a 10-90 % rise time of 16 ms and could be fitted with a single exponential component with a time constant of 7.3 ms. The solution flowing through the application pipette could be changed by means of a rotary valve connected to each barrel. During solution changes the agonist application barrel was flushed for 20-60 s to remove the previous solution. Unless otherwise stated, agonist was applied at intervals at which the receptors were shown (Figs 3–4) to completely recover (4 × mean τ 99 %) from desensitization (10 min for 1-(m-chlorophenyl)-biguanide or mCPBG, 1 min for serotonin, and 15 s for glutamate). Since the desensitization kinetics of 5-HT₃ receptors depend upon agonist concentration (Yakel et al. 1991; Fan, 1994; Lobitz et al. 2001), all experiments were performed using a supramaximal concentration (100 μm) of serotonin, unless otherwise indicated. Previous reports have suggested that the desensitization kinetics for 5-HT₃ receptor responses are variable and under the control of second messenger systems and intracellular Ca²⁺ (Yakel et al. 1991; Jones & Yakel, 1998). In our hands the rate of desensitization of whole cell current responses accelerated over the course of an experiment to a value similar to that observed in excised membrane patches. Although we have observed decay time constants of desensitization that varied from cell to cell in whole cell recordings, experiments were completed within 15 min of obtaining the recording to minimize time-dependent changes in response properties. For experiments on patches in which desensitization was more rapid than deactivation, we recorded the more rapid desensitization time course before the slower deactivation time course in order to ensure that rundown-linked acceleration of response decay did not contribute to the acceleration we observed in response to prolonged agonist application.

Figure 3. Time course of desensitization.

A, the onset of serotonin-induced desensitization in a current recording from an excised patch is best described by two exponential components. The thin white curve shows the fitted dual exponential time course. B, the mean ±s.e.m.τ_fast and τ_slow describing serotonin-induced desensitization of patch (□) and whole cell () current responses are shown. C, averaged amplitude ratios (mean ±s.e.m.) obtained using a double pulse protocol (see inset) show the time course of recovery from serotonin-induced desensitization (τ_recovery;n = 12 cells). The continuous line shows the fit of eqn (2) to the data. The time course was best fitted by a single exponential component (N_D = 1.1).

Figure 4. Time course of deactivation.

A, the deactivation time course of a current response evoked by a brief application of serotonin (3 ms) to the same excised patch shown in Fig. 3. The thin white curve shows the fitted dual exponential time course. B, the mean ±s.e.m.τ_fast and τ_slow describing deactivation of serotonin-induced responses in patch (□) and whole cell () current responses are shown. C, averaged amplitude ratios (mean ±s.e.m.) obtained using a double pulse protocol (see inset) show the time course of recovery from desensitization of whole cell current responses induced by a brief application (1-5 ms) of serotonin (n = 9 cells). The continuous line shows the fit of eqn (2) to the data. The time course was best fitted by a single exponential component with N_D= 1.0 (from eqn (2)). For comparison, the recovery time course from desensitization caused by prolonged agonist application in these same cells is superimposed as a dashed line.

Expression of 5-HT_3A receptors in Xenopus laevis oocytes

cRNA was synthesized from linearized cDNA template encoding the murine 5-HT_3A subunit (plasmid pMX201 in pcDNA1/AMP) according to the manufacturer's instructions (Ambion). The quality and amount of cRNA was estimated by gel electrophoresis, and 5-10 ng in 50 nl injected into stage V or VI oocytes isolated from adult female Xenopus laevis as previously described (Traynelis et al. 1998). Oocytes were placed 3-4 days post injection in a hyperosmotic solution of the following composition (mm): 200 potassium aspartate, 20 KCl, 1 MgCl₂, 5 K-EGTA, 10 Hepes (pH adjusted to 7.4 with KOH). Following shrinkage of the oocyte, the vitelline membrane was removed using a fine pair of forceps, and oocytes transferred to our recording chamber. Excised outside-out membrane patches were obtained from oocytes and studied as described above for HEK cells except that the internal electrodes contained (mm): 100 KCl, 10 EGTA, 10 Hepes (pH adjusted to 7.4 with KOH). The external solution comprised (mm): 115 NaCl, 2 KCl, 5 Hepes, 0.25 CaCl₂, 0.25 MgCl₂ (pH adjusted to 7.4 with NaCl). Patch data recordings from oocytes and HEK cells yielded similar results.

Analysis and simulations

Macroscopic current responses were simulated (ChanneLab v1.0, Synaptosoft) from state-dependent models using a 4th order Runga-Kutta algorithm (Press et al. 1992) or a Monte Carlo simulation scheme built from a pseudo-random number generator (Wichmann & Hill, 1985). Microscopic rate constants were determined using a simplex algorithm and least squares criteria to fit different models to the average macroscopic response time course recorded from several patches. During fitting, response simulations from our models were driven by concentration profiles that matched our mean measured solution exchange time course as estimated from tip potentials. For fitting multiple waveforms, the open probability of the maximal recorded response was normalized to that of the largest simulated response; the open probability was scaled appropriately for the submaximal response at each iteration. Simulations were run at the same sampling rate as our recorded data.

The time course of macroscopic 5-HT_3A current deactivation and desensitization were fitted with the equation:

1

where A is amplitude and t is time. The peak/steady state ratio of current responses were the measured peak current divided by the fitted value for the steady state as determined in eqn (1). Recovery from desensitization was determined by the equation:

2

where N_D indicates the number of sequential steps leading to a desensitized state (Van Hooft & Vijverberg, 1996). The rising phase of responses was fitted to the equation:

3

The dose response relationship was determined by recording responses alternately to 100 μm serotonin and 0.1, 1, 2, 3, or 10 μm serotonin. Responses at each concentration were expressed as a fraction of the mean of responses to 100 μm serotonin before and after the serotonin, and the composite curve fitted with the equation:

4

where EC₅₀ is the half-maximally effective concentration and N_H is the Hill slope. The inhibition curve for MDL72222 (3-tropanyl-3,5-dichlorobenzoate) was fitted by the same equation with N_H constrained to be less than 0, IC₅₀ substituted for EC₅₀, and [serotonin] replaced by [MDL7222]. The significance of measurements was assessed using ANOVA, Student's t test, or the variance ratio test where appropriate. Results were considered significant at P < 0.05, and the power of the test was calculated for P > 0.05 (Zar, 1996).

Materials

All tissue culture reagents were obtained from Gibco (Gaithersburg, MD, USA) except human fibronectin (Promega, Madison, WI, USA). MDL72222 and mCPBG were obtained from RBI (Natick, MA, USA). Serotonin, glutamate and other chemicals were obtained from Sigma Chemical Co. (St Louis, MO, USA).

RESULTS

Activation of 5-HT_3A homomeric receptors

We first determined the dose-response relationship for serotonin activation of murine 5-HT_3A homomeric receptors under voltage clamp using a rapid agonist application system capable of submillisecond solution exchange. All experiments were carried out in reduced divalent cations (0.25 mm Ca²⁺, Mg²⁺) and at a holding potential between -60 to -75 mV unless stated otherwise. Although HEK cells showed no serotonin-evoked current responses (data not shown), application of serotonin to HEK cells that stably express 5-HT_3A receptors evoked inward whole cell current responses characteristic of 5-HT₃ receptors (Fig. 1). Responses to low concentrations of serotonin activated slowly over hundreds of milliseconds (Fig. 1A) and endured for seconds, whereas the response to higher concentrations activated and desensitized more rapidly in the continued presence of agonist (Fig. 1). Expressing the HEK whole cell current peak response as a function of serotonin concentration produced a sigmoidal relationship that could be fitted by eqn (4), which gave an EC₅₀ value of 2.0 μm with a Hill slope of 3.0 (n = 12; Fig. 1B). 5-HT_3A receptor responses recorded from excised membrane patches from Xenopus oocytes also had an EC₅₀ of 2.0 μm (n = 3). This EC₅₀ value is similar to that previously determined for murine 5-HT_3A receptors expressed in Xenopus oocytes studied under two-electrode voltage clamp (3.4 μm; Maricq et al. 1991). The Hill slope of 3.0 is slightly steeper than that previously reported for recombinant 5-HT_3A receptors (1.7-2.7; Maricq et al. 1991; Hussy et al. 1994; Van Hooft & Vijverberg, 1996; Barann et al. 1997; Gunthorpe et & Lummis, 1999) and suggests that channel opening requires the binding of multiple agonist molecules and perhaps some cooperativity between agonist binding sites. This implies that several binding steps (and therefore channel closed states) are likely to precede channel opening. Whole cell 5-HT_3A receptor current responses in HEK cells to 10 μm serotonin could be fully blocked by co-application of the 5-HT₃-selective competitive antagonist MDL72222 (IC₅₀ 96 nm; n = 6), confirming that the serotonin-evoked currents that we studied reflect activation of 5-HT₃ receptors.

Figure 1. Activation of 5-HT_3A receptors by serotonin.

A, the upper panel shows whole cell currents recorded from an HEK cell stably expressing 5-HT_3A receptors in response to application of different concentrations of serotonin (-100 mV); bar shows period of agonist application. The lower panel shows normalized current responses to three different concentrations of serotonin. B, dose-response relationship is shown for serotonin activation of 5-HT_3A receptors. ○, mean ±s.e.m. from 12 cells; ▵, data from patches (n = 3). Continuous line is from eqn (4) fitted to the whole cell data.

5-HT_3A receptor whole cell current responses reached peak amplitude within tens of milliseconds (10-90 % rise time 19.6 ± 4.5 ms) in response to rapid application of a maximal concentration of 100 μm serotonin (50 × EC₅₀). Figure 2A shows the rising phase of a current response of an excised membrane patch to rapid application of serotonin for either 5 s or 2 ms. Figure 2C compares on an expanded scale the average current response and open tip potential from three patches. To verify that the 5-HT₃ response time course we observe was unrelated to the perfusion system, HEK cells that stably expressed 5-HT_3A receptors were transiently transfected with cDNA encoding the kainate receptor subunit GluR6 and the response to 10 mm glutamate was measured (20 × EC₅₀; Traynelis & Wahl, 1997). GluR6 receptors that co-existed in the same patches with 5-HT_3A receptors could be activated with a mean 10-90 % rise time of 760 μs (n = 5). Furthermore, analysis of the exchange time at the tip of the pipette after the patch was ruptured also verified that solution exchange was rapid in these experiments (10-90 % rise time for junction potential was 550 μs; n = 5). The average rise time for serotonin-induced responses in excised membrane patches was 12.5 ± 1.6 ms (mean ±s.e.m.; n = 9 patches; Fig. 2B), and was not significantly different for receptors in patches activated by 2 ms or 5 s pulse of serotonin (P < 0.05; power to detect a 4 ms difference was 0.75), suggesting that slow binding of serotonin cannot fully explain the full duration (i.e. 12 ms) of the rising phase. In further support of this idea, we find that rise times for whole cell current responses to 10 and 100 μm serotonin were similar (Fig. 1A). We could detect no voltage dependence to the 5-HT_3A receptor-mediated whole cell current rise time over the voltage range -75 to +50 mV, and only weak voltage dependence at potentials more hyperpolarized than -75 mV (n = 5; data not shown). Fitting the rising phase of the mean 5-HT_3A receptor responses to 100 μm serotonin averaged from nine excised membrane patches to a single exponentially rising waveform (eqn (3)) gave a τ_rise of 7.3 ms, suggesting an apparent activation rate constant of 136 s⁻¹. Activation of 5-HT_3A receptors was also slow for 10 μm of the 5-HT₃-selective agonist mCPBG, which had a 10-90 % rise time that was 2.2 ± 0.2-fold slower than that for serotonin in whole cell recordings (n = 3; P < 0.02; paired t test). Thus, the rise time of 5-HT₃ receptor responses is slower than that reported for glycine receptors (Grewer, 1999), GABA_A receptors (Maconochie et al. 1994), nicotinic acetylcholine muscle receptors (Liu & Dilger, 1991), AMPA receptors (Mosbacher et al. 1994), kainate receptors (Traynelis & Wahl, 1997), and purinergic P_2X receptors (Evans et al. 1992). Among ligand-gated ion channels, only NMDA-selective glutamate receptors activate with a similarly slow time course (10-90 % rise time 10.4 ms; Clements & Westbrook, 1991).

Figure 2. The rise time of the 5-HT_3A current responses in excised patches.

A, current recordings are shown from an outside-out membrane patch from an HEK 293 cell that was stably transfected with 5-HT_3A cDNA and transiently transfected with GluR6(Q) cDNA. In the left panel a 5 s application of serotonin (100 μm or 50 × EC₅₀) produces an inward current at -80 mV with a 10-90 % rise time of 13.7 ms. In this same patch application of 10 mm glutamate (20 × EC₅₀) for an equivalent duration produces an inward current (dashed line) which rises faster than the response to serotonin. In the right panel application of serotonin to this same patch for 1 ms produces a current with similar rising time course to that produced by the 5 s agonist pulse; a 1 ms pulse of glutamate produced a GluR6-mediated current response with faster kinetics. In order to evaluate the agonist application time course, the open pipette junction current (labelled ‘open pipette’) was recorded after the patch was ruptured. B, averaged values for the 10 -90 % rise time of serotonin- and glutamate-induced currents recorded from excised patches (n = 5) exposed to 5 s pulses of 100 μm serotonin. Asterisk (*) indicates that the 5-HT₃ current rose significantly more slowly than the GluR6(Q) current (P < 0.001; paired t test,). C, mean current response (dotted line) and open tip potential (continuous line) from three patches exposed to 5 s application of 100 μm serotonin are superimposed. Traces were aligned on the point of steepest rise of the open tip potential.

Desensitization of 5-HT_3A homomeric receptors

Homomeric 5-HT_3A receptors in excised membrane patches desensitize in the continued presence of 100 μm serotonin or 10 μm mCPBG (τ= 280 ± 40 ms; n = 8) with a ratio of whole cell peak to steady state current that is greater than 20. Serotonin-evoked 5-HT_3A receptor responses in outside-out membrane patches desensitize with a dual exponential time course (Fig. 3A) that can be described by the sum of two exponential components with time constants of τ_slow 1006 ± 139 ms and τ_fast 176 ± 25 ms (A_fast/(A_fast + A_slow) = 0.69 ± 0.08; eqn (1); n = 16; Fig. 3B). Whole cell serotonin-evoked current responses similarly desensitize with a dual exponential time course (τ_slow 2747 ± 236 ms; τ_fast 527 ± 65 ms; n = 41; Fig. 3B), and the time constants describing this desensitization fell within the previously reported range of time constants for 5-HT₃ receptor desensitization (τ_slow 2300-6000 ms; τ_fast 170-570 ms; Yakel et al. 1991; Yang et al. 1992; Werner et al. 1994; Jones & Yakel, 1998). The time constants describing desensitization were only weakly voltage dependent at membrane potentials depolarized to -75 mV (data not shown). The recovery from desensitization was studied for both serotonin- and the mCPBG-evoked whole cell current responses using a double pulse protocol with a variable interval. The time course of desensitization for paired responses recorded in response to 100 μm serotonin at a 10 s interval were identical, suggesting a similar population of receptors underlie each response (data not shown). For both mCPBG and serotonin, recovery from desensitization followed a slow exponential time course and was voltage independent over the range -40 to -120 mV (data not shown; n = 12 serotonin, n = 6 for mCPBG). A plot of the ratio of the peak amplitude of the second response to that of the first response for serotonin vs. the inter-pulse interval (Fig. 3C) could be fitted by the exponential function described by eqn (2) in which N_D corresponds to the number of consecutive desensitization steps (Van Hooft & Vijverberg, 1996). Fits to this equation suggested that maximal concentrations (100 μm) of serotonin (N_D = 1.1) and mCPBG (10 μm; N_D= 1.3) largely promote transition of the receptor into a single desensitized state. The recovery from serotonin-induced desensitization (τ_recovery 17.4 s; n = 12) was faster than from mCPBG-induced desensitization (τ_recovery 96 s; n = 6; data not shown). Responses to mCPBG were not studied further because the slow time course of recovery from desensitization complicated completion of experiments before rundown of the channel response.

Deactivation of 5-HT_3A homomeric receptors

In order to determine how 5-HT_3A receptors might respond to brief synaptic-like stimulation by serotonin, we compared the response to 1-5 ms and 5 s pulses of 100 μm serotonin rapidly applied to excised membrane patches. The mean amplitude of the response to brief agonist application was 83 ± 4 % (n = 8 patches) of the amplitude of the response to prolonged agonist application. The time course of decay of the 5-HT_3A receptor responses in patches following agonist removal (which we assume reflects deactivation) could be best described by a dual exponential time course (Fig. 4A). To our surprise, the time constants describing 5-HT_3A receptor deactivation time course (τ_slow 838 ± 217 ms, τ_fast 213 ± 44 ms; A_fast/(A_fast+A_slow) = 0.45 ± 0.10; eqn (1); n = 8) following brief 1-5 ms serotonin application were slow and not significantly different from those observed for receptor desensitization (P > 0.2; compare Fig. 3B and Fig. 4B). One potential explanation for this could be that brief application of serotonin induces some degree of desensitization, which subsequently slows the time course for deactivation by returning receptor to an activable state long after agonist has been removed. To test this possibility, we used a double pulse protocol in which a variable interval separated two brief pulses of serotonin that were applied to a whole cell. Brief application of agonist induced partial (60 %) desensitization since double pulse protocols showed suppression of the response to a second pulse of serotonin. Desensitization recovered with a time course that was about 2-fold faster (τ_recovery 9.0 s) than that observed for desensitization to prolonged application of serotonin measured in the same cells (τ_recovery 19.4 s; P < 0.05, paired t test; n = 9 cells; Fig. 4C). Both NMDA and GABA_A receptors also enter a desensitized state following brief exposure to agonist (Lester & Jahr, 1992; Jones & Westbrook, 1995). The mean time courses for recovery of 5-HT_3A receptors from desensitization induced by brief and prolonged application of agonist were best fitted by a single exponential component (i.e. N_D= 1.0-1.1; see eqn (2)). The differential rate of recovery from desensitization induced by brief agonist application and desensitization produced by prolonged agonist application suggests that at least two different routes out of desensitized states may exist. The idea that 5-HT₃ receptors possess multiple paths into and out of desensitized states is also consistent with studies suggesting that 5-HT₃ receptors can be predesensitized by preapplication of serotonin with an IC₅₀ value of 0.2 μm and Hill slope of 3.2 (Van Hooft & Vijverberg, 1996).

Variable time course for desensitization of 5-HT_3A homomeric receptors

The time course for desensitization varied among patches and cells for unknown reasons. Upon examination of this variability, we were surprised to find that in half of our patch recordings (4 of 8 patches), deactivation was slower than desensitization. Similarly, 15 of 35 whole cell current responses showed slower deactivation than desensitization. In order to explore the basis of this unexpected result we compared the variability of our measurements of deactivation and desensitization. Figure 5A shows a scatter plot for individual measurements of deactivation and desensitization time constants from whole cell current responses. Figure 5B and C further shows substantial overlap of the time constants for desensitization and deactivation. Moreover, the time constants describing desensitization had a significantly larger variance (3- to 9-fold) than those describing deactivation across our experiments (P < 0.005 for τ_slow and P < 0.001 for τ_fast; D-Agostino-Pearson test for normality and the variance ratio test; Zar, 1996). This suggests that some unknown condition (growth cycle, tonic kinase or phosphatase activity, etc.) at the time of recording controls the time course of desensitization but not deactivation.

Figure 5. Variable time course for 5-HT_3A desensitization.

A, each circle represents the τ_fast (left panel) or τ_slow (right panel) value for either desensitization (▪) or deactivation (□) from serotonin-evoked whole cell currents. The bars show the mean fitted time constants ±s.e.m.B and C, histograms compare for τ_slow (B) and τ_fast (C) the distribution of desensitization time constants (▪) to the distribution of deactivation time constants (□).

Although the underlying basis for why some cells show rapid desensitization is unknown, this behaviour of 5-HT_3A receptors is nevertheless in striking contrast to most other ligand-gated channels. AMPA receptors (Mosbacher et al. 1994), kainate receptors (Heckmann et al. 1996), GABA_A receptors (Jones & Westbrook, 1995), glycine receptors (Legendre, 1998), muscle nicotinic receptors (see Fig. 3 in Liu & Dilger, 1993; Fig. 3 in Maconochie & Steinbach, 1998), P_2X receptors (e.g. Fig. 4A in Taschenberger et al. 1999) and NMDA receptors (Wyllie et al. 1998) invariably desensitize more slowly than they deactivate. Likewise, voltage-gated channels typically inactivate (i.e. desensitize) more slowly than they deactivate (e.g. Meir & Dolphin, 1998). Figure 6A compares fast desensitization to deactivation in one such patch. In other patches in which desensitization was more rapid than deactivation, application of serotonin for a duration that was intermediate between 2 ms and 5 s produced a current response with a distinctly biphasic decay. That is, the decay kinetics of 5-HT_3A receptor currents change rapidly after removal of serotonin (Fig. 6B) to a time course similar to that observed following brief application of agonist. Inspection of the time course of decay of a current response in an excised membrane patch to a 1 s application of serotonin (thick trace) shows an initial accelerated rate of decay that is identical to that observed for longer applications of agonist (lower thin trace). However, at the end of the 1 s pulse there is rapid transformation of the time course to a slower time course observed for deactivation following a 2 ms pulse of agonist (upper thin trace). Independent fits of an exponential function to these two portions of the decay show that they possess the expected time constants for desensitization (when fitting the period during serotonin application) and deactivation (when fitting the time after serotonin removal) as compared to recordings in the same cell using prolonged and brief application of serotonin, respectively. In six experiments in which serotonin was applied for durations sufficient to produce 20-60 % decay in the peak amplitude, the time constants derived from the response during serotonin application were significantly faster than the time constant derived from the portion of the response obtained following removal of serotonin (P < 0.05; ANOVA).

Figure 6. Rapid desensitization of 5-HT_3A receptors.

A, current recordings of 100 μm serotonin-evoked current in an excised outside-out patch from an HEK 293 cell are shown. Responses of similar amplitudes were evoked by brief (1-5 ms, lower panel, thick traces) and long (5 s, lower panel, thin traces) duration applications of agonist. The junction potential changes (labelled ‘open pipette’) in response to brief (upper panel, top trace) and long duration (upper panel, bottom trace) agonist application are shown for this patch. B, application of serotonin for an intermediate duration (1 s in this patch; thick trace) evoked a response that initially declined in amplitude with a time course identical to desensitization induced by 5 s application of serotonin (thin trace). After removal of the agonist (circled area), the decay immediately becomes slower with a time course similar to the decay for a 1-5 ms pulse (thin trace). C, the I-V relationship for whole cell currents (corrected for the junction potential) in response to serotonin application for 1-5 ms and 5 s is shown in the left panel. Each I-V curve represents the average of 5 trials from the same cell. The right panel shows whole cell responses to 1-5 ms and 5 s pulses of serotonin at -80 and +60 mV. At both potentials the onset of desensitization was faster than that of deactivation in this cell. D, the half-width of the response to a 5 s application of serotonin (desensitization) was expressed as a fraction of the half-width of the response to 1-5 ms pulse (deactivation) for cells (○) and patches (•). This ratio changed only minimally over the tested voltage range.

Voltage independence of the difference between deactivation and desensitization

One potential explanation of the finding that deactivation can proceed more slowly than desensitization in some cells and patches is that serotonin, which is a cation at physiological pH, might exert some degree of voltage-dependent channel block within the 5-HT_3A receptor pore since it does not discriminate well among monovalent cations (Yang, 1990). Acetylcholine at concentrations greater than 200 μm has been suggested to cause a voltage-dependent block of the nicotinic receptor at the neuromuscular junction (e.g. Colquhoun & Ogden, 1988). To assess whether such a block occurs at 100 μm serotonin, we evaluated the voltage dependence of the current response to brief and long duration pulses of serotonin applied to excised membrane patches and whole cells. If serotonin acted as a pore-blocking molecule at a site within the transmembrane electric field, its effects on the time course of the response should be voltage dependent. Although the current-voltage relationship for peak serotonin responses (Fig. 6C) showed strong rectification, which may reflect our caesium-based internal solution (Glitsch et al. 1996), we were able to record the 5-HT_3A receptor response time course at a variety of hyperpolarized membrane potentials. In outside out patches, neither the onset of desensitization nor deactivation was affected by voltage over the range -100 to -40 mV. Thus the ratio between deactivation and desensitization half-widths appeared voltage independent (Fig. 6D). Similar results were found in whole cell current recordings (Fig. 6C and D). In addition, we found no difference in the reversal potential for any portion of the response to brief or long duration agonist application (n = 15 cells and 4 patches). We also found no difference in the reversal potential following replacement of the caesium gluconate (4.5 ± 1.5 mV; n = 15) in our internal recording solution with CsCl (4.4 ± 1.0 mV; n = 14; P > 0.5; power to detect a 1 mV difference was 0.75). These results suggest there is no time- or agonist-dependent change in cationic or anionic selectivity of the 5-HT_3A channel during prolonged serotonin applications. These data argue against the possibility that the more rapid desensitization of the current in response to prolonged application of serotonin reflects channel block by serotonin deep within the pore. The experiments shown in Fig. 6C provide strong support for this idea, since they show that prolonged application of serotonin accelerates the decay rate of outward currents recorded at holding potentials positive to the reversal potential. These results, however, do not rule out a block of 5-HT_3A receptors by serotonin acting at a site within the vestibule that is uninfluenced by the direction of net current flow or the transmembrane electric field.

DISCUSSION

In this study we describe the response of 5-HT_3A receptors to brief and prolonged serotonin application with submillisecond exchange times. Using this approach we show that the 5-HT_3A receptor is the slowest member of the nicotinic superfamily of ligand-gated receptors. Activation of the receptor in response to 1-5 ms pulses of serotonin exhibited a ∼12 ms 10-90 % rise time, suggesting that the binding reaction could not explain the full time course for activation since the agonist was removed before the channel response reached 25 % full amplitude. We also provide evidence that brief 1-5 ms pulses of serotonin induce some degree of desensitization. A similar result with GABA_A receptors has been suggested to contribute to the unusually slow deactivation time course described for these receptors (Jones & Westbrook, 1995). We additionally found a large degree of variability in the time constants describing desensitization. One important result that unexpectedly arose from the variable time course of desensitization is that, in half of our patches, 5-HT_3A receptors in 0.25 mm external divalent ion concentrations desensitized more rapidly than they deactivated. The observation that desensitization is often faster than deactivation is unusual and has not been reported for other ligand-gated ion channels (Liu & Dilger, 1993; Mosbacher et al. 1994; Jones & Westbrook, 1995; Heckmann et al. 1996; Legendre, 1998; Wyllie et al. 1998; Taschenberger et al. 1999). However, it is important to note that decay kinetics of other ligand-gated ion channels are typically measured at a higher extracellular calcium concentration than was used in our experiments. Thus, it is possible that 5-HT₃ receptors could behave more like other ligand-gated ion channels if the calcium concentration were raised. Indeed, increasing extracellular Ca²⁺ accelerates both desensitization (Peters et al. 1988; Robertson & Bevan, 1991; Yakel et al. 1993; Lobitz et al. 2001) and deactivation (Traynelis & Mott, 1999) of 5-HT_3A receptors in a voltage dependent manner, suggesting that the 5-HT_3A receptor response time course is highly sensitive to extracellular Ca²⁺. It is possible that 5-HT₃ receptor desensitization and deactivation may depend upon intracellular calcium or calcium influx through the 5-HT₃ receptor channel (Jones & Yakel, 1998) and this may become more important at higher calcium concentrations.

Serotonin activates 5-HT_3A receptors with a lower EC₅₀ (2 μm) than the endogenous neurotransmitters for kainate (500 μm; Heckmann et al. 1996; Traynelis & Wahl, 1997), AMPA (432 μm; Hausser & Roth. 1997), glycine (54 μm; Legendre, 1998), acetylcholine (15-50 μm; Colquhoun & Ogden, 1988), GABA (15 μm; Jones et al. 1998), or ATP receptors (15 μm; Taschenberger et al. 1999). However, slower unbinding due to a higher affinity of serotonin for 5-HT_3A receptors than other neurotransmitters for their receptors alone cannot explain our observation that 5-HT_3A receptor deactivation is often slower than desensitization. For example, slowing the unbinding rate constant for AMPA or kainate receptor models by 1000-fold (Heckmann et al. 1996; Banke et al. 2000) slows the deactivation time course to a rate similar to, but not slower than, desensitization. This is because allowing receptors to retain the agonist for prolonged periods of time simply promotes desensitization in these models. Thus, the formulation of these low affinity receptor models cannot generate a deactivation time course that is slower than desensitization, suggesting there must be features of 5-HT_3A receptors other than agonist affinity that account for the slower deactivation.

Biophysical mechanism of 5-HT_3A receptor activation

In order to explore 5-HT_3A receptor function, we used recordings of responses to both 2 and 100 μm serotonin in the same patch (n = 3) to evaluate various hypothetical reaction schemes. We aligned the recordings to the onset of the agonist application as determined by the tip potential measured after the experiment, and subsequently averaged response waveforms across patches. We then normalized the waveform of the 2 μm response to the peak amplitude of the 100 μm response. Because our dose-response data are consistent with the idea that the pentameric (Boess et al. 1995; Green et al. 1995) 5-HT_3A receptors activate upon binding of three or more agonist molecules (Fig. 1B), we evaluated linear reaction schemes with three to five agonist binding sites and a single open and a single desensitized state. We utilized a simplex algorithm that altered the microscopic rate constants and compared a numerical integration of the reaction scheme to our data (see Methods). Although we could obtain adequate fits for all of these models to either the 2 μm or 100 μm response waveform, we could not obtain satisfactory results when we refitted both response waveforms simultaneously to these models (data not shown). Consistent with this observation, rate constants derived from fits of either waveform in isolation did not produce the correct concentration dependence of 5-HT_3A response amplitude, rise time, or half-width. However, we were able to obtain excellent fits to our two mean response waveforms evaluated simultaneously within the fitting algorithm when we introduced two or more open and desensitized states. We subsequently evaluated the model shown in Fig. 7A, which is based on the assumptions that there are five equivalent agonist binding sites (Boess et al. 1995), that three agonists must bind before channel opening (Fig. 1B), that receptors can be predesensitized with low agonist concentrations (van Hooft & Vijverberg, 1996), that recovery from desensitization does not vary with agonist occupancy (measured recovery rate kr was 0.062 s⁻¹; Fig. 3C), and that mean single channel open times are about 25 ms (α (the closing rate) = 40 s⁻¹; van Hooft & Vijverberg, 1996). Transitions between states in Fig. 7A that we did not evaluate are shown in grey, and transitions for which the rate was fixed are shown as open-headed arrows. Microscopic reversibility for cyclic portions of the reaction scheme was maintained by appropriately scaling the unbinding rates from desensitized states.

Figure 7. A model of murine homomeric 5-HT_3A receptor activation.

A, hypothetical binding scheme for a homo-pentameric receptor with five equivalent agonist binding sites. States and transitions shown with dotted arrows were not included in the fitting procedure, but are presumed to exist on the basis of published or experimental data (see text). Transitions shown as open-headed arrows were not varied as free parameters in the fitting procedure. The unbinding rate constants for the desensitized states were adjusted to maintain microscopic reversibility of cyclic portions of the model. Asterisks indicate open states. B, two different sets of rate constants were identified from fitting this model simultaneously to averaged current waveforms recorded in response to 2 and 100 μm serotonin applied to the same patch. C-E, the concentration dependence of peak current (continuous line; Fig. 1), rise time (□, continuous line), and half-width (□, continuous line) are shown for experimental whole cell currents (n = 12) and for the two sets of rate constants determined from fitting our model to patch data (▵, ▿). Although the fitted rate constants produce response waveforms that described experimental data from patches well (*), there was some divergence between the simulated and measured half-width from whole cell responses (E), consistent with our observation of different kinetics in patch vs. whole cell recordings. Error bars are s.e.m.

We found two distinct sets of microscopic rate constants for this model (referred to as Fit 1 and Fit 2; Fig. 7B) that produced adequate waveforms with reasonable concentration dependence of the rise time, amplitude, and half-width when compared to our patch data (Fig. 7C-E); these models also predicted a concentration dependence for the rise time and amplitude that was similar to measurements from whole cell recordings. Interestingly, although the rate of desensitization in Fit 1 is similar for conformations with four or five serotonin molecules bound, the decrease in the opening rate for fully liganded receptors favours desensitization from this state. Fit 2 similarly favours desensitization of receptors with four compared to three or five molecules bound, but by a different mechanism. Here the opening rate constants are similar for receptors with three to five agonist molecules bound (Fig. 7B). However, the desensitization rate constants vary with agonist occupancy such that there is more desensitization with four rather than three or five agonist molecules bound. More information on single channel open and shut dwell times is needed before we can distinguish between these models.

Each of the fits to our macroscopic responses that we obtained predicted the dose-response data we observed in patch and whole cell responses. For example, the peak open probability was 0.89 and the EC₅₀ was 1.9 μm with a Hill slope of 2.3 for Fit 1 (Fig. 7B). While this slope is more shallow than that calculated from our data, inspection of Fig. 7C illustrates how even slight variation in the response to 1 μm serotonin can have strong effects on the Hill slope, suggesting that a Hill slope between 2 and 3 is sensitive to any experimental error associated with measurements of low concentration responses. The data could not be fitted with the model shown in Fig. 7A when the microscopic rate constants governing channel opening and the onset of desensitization were constrained to be the same for states with different numbers of agonist molecules bound (data not shown). Rather, the two best fits from the models shown in Fig. 7 suggest that an agonist-dependent open probability is a necessary feature of 5-HT₃ receptor function. Figure 8A shows the quality of Fit 1 to the two 5-HT_3A receptor response waveforms we evaluated and illustrates how receptors with three, four, or five agonist molecules contribute to overall open probability (Fig. 8B and C). Whereas the open probability shown in Fig. 8B and C reflects both the overall fraction of receptors with 3, 4 and 5 agonist molecules bound and their open probability, evaluation of the open probability purely as a function of subunit occupancy can be made if the occupancy of each state is calculated. The insets in Fig. 8B and C show maximum open probabilities calculated for the receptor when either three, four, or five agonist molecules are bound (see Fig. 8 legend), and demonstrate the subunit dependence of open probability predicted by this model.

Figure 8. Comparison of predicted and recorded macroscopic homomeric 5-HT_3A response waveforms.

A, thick black traces show the averaged current waveforms recorded from excised membrane patches (n = 3) in response to prolonged (5 s) application of 2 or 100 μm serotonin. These waveforms were used in the fitting of the model in Fig. 7A. The predicted response waveforms from the fitted rate constants for Fit 1 (Fig. 7B) are superimposed as thin white lines. B and C, the time course of the open probabilities predicted from the rate constants for Fit 1 for receptors with 3, 4 or 5 serotonin molecules bound is shown for simulated responses to 100 μm (B) and 2 μm serotonin (C). The inset shows the maximum fraction of open receptors calculated in a manner independent of subunit occupancy from:

where n is the number of agonist molecules bound from the model shown in Fig. 7. Note the dependence of the peak open probability on the number of agonist-bound subunits. Similar dependence of open probability on the number of agonist-bound subunits was observed for the rate constants for Fit 2 (data not shown).

Several groups have previously raised the idea that differing degrees of fractional agonist occupancy of a multimeric receptor can have differential effects on receptor function (Raman & Trussell, 1992; Jones & Westbrook, 1995; Rosenmund et al. 1998; Papke et al. 2000). The kinetic analysis of our data support the idea that 5-HT_3A receptor open probability differs with three, four, or five serotonin molecules bound. This result has important implications for receptor function, since it suggests that 5-HT_3A receptors must bind three or four ligands before opening but that the binding of additional serotonin molecules favours desensitization either by enhancing its rate of onset or by slowing the opening rate. Thus there appears to be an optimal degree of subunit occupancy by serotonin for 5-HT_3A receptors that lies between three and five molecules. One way to relate the hypothetical model presented in Fig. 7 and Fig. 8 to a physical model of receptor function would be to postulate that each agonist binding event to a 5-HT_3A subunit within a pentameric receptor produced an incremental change in the pore properties. In this scenario differential agonist occupancy of the various subunits might produce different effects on ion permeation, gating, or desensitization (Rosenmund et al. 1998). A mutagenesis study of homomeric 5-HT_3A receptors supports a role in desensitization for a presumed pore-lining residue (Yakel et al. 1993), suggesting that the determinants of desensitization and thus channel gating are shared by each subunit, and thus could be controlled in a discrete fashion by agonist occupancy.

Whereas our model predicts the correct waveforms for responses to 2 and 100 μm serotonin, it of course should not be considered a definitive description of 5-HT_3A receptor kinetics but rather a first degree approximation. Indeed, we find multiple solutions for this kinetic model, suggesting more measurements of receptor properties are needed to create additional constraints for model evaluation. We have also made a number of simplifying assumptions, three of which deserve mention. First, our model was evaluated omitting the concentration-dependent predesensitizing steps (states indicated with grey arrows in Fig. 7A). Second, no data exists showing concentration dependence of single channel open time, forcing us to set the closing rate α equal for all state (40 s⁻¹). Third, we fixed the recovery rate from desensitization to match our measured rate following prolonged application of agonist (0.062 ⁻¹), ignoring the 2-fold more rapid recovery from desensitization induced by a brief pulse of agonist (Fig. 4C). Despite these simplifying assumptions, extension of the approach of fitting multiple waveforms simultaneously with a model should allow estimation of additional parameters if waveforms for predesensitization, recovery from desensitization, and more than two agonist concentrations can be recorded from the same patch.

Biophysical mechanism of rapid 5-HT_3A receptor desensitization

One unexpected finding from our study of 5-HT_3A receptor function is that in some cells and patches, 5-HT_3A receptors desensitize more rapidly than they deactivate. How might this occur? If 5-HT_3A receptors have multiple desensitized states with different agonist occupancy (e.g. the result predicted from our modelling), rapid agonist-dependent desensitization might arise if desensitization becomes faster as fractional occupancy of the agonist binding sites increases. Figure 9A illustrates this idea using a hypothetical linear reaction scheme with binding rate constants similar to those we observe. Evaluation of the probability of individual states shows that brief pulses of agonist cause relatively equal occupancy of receptors with three, four, or five agonist molecules bound, whereas prolonged agonist application drives receptors into the fully liganded state. This suggests that if the fully liganded state exhibits faster desensitization than receptors with three or four agonist bound, such receptors could show a rate of desensitization that is faster than the slow agonist unbinding of the fully liganded receptor. Alternatively, serotonin binding at a site distinct from the agonist recognition site might occur during prolonged application of agonist. For example, serotonin might bind within the channel vestibule such that occupancy occludes ion conduction, an idea that has already been well developed in the channel block literature (e.g. Armstrong, 1971; Adams, 1976; Neher & Steinbach, 1978). The acceleration of response decay in the presence of agonist for outward currents (Fig. 6) argues against a voltage-dependent channel blocking mechanism, as has been proposed to account for the biphasic dose-response curve describing activation of muscle nicotinic acetylcholine receptors (Colquhoun & Ogden, 1988). We therefore favour a model in which the desensitization rate varies with agonist occupancy of the pharmacologically defined 5-HT_3A binding site.

Figure 9. A model of rapid desensitization of murine homomeric 5-HT_3A receptors.

A, hypothetical binding scheme for a pentameric receptor with five equivalent agonist binding sites. The left panel shows the summed occupancy of receptors that have bound 3, 4, or 5 agonist molecules in response to a 2 or 50 ms application of 100 μm agonist. Note that the probability of reaching states with 3, 4, or 5 ligands bound is similar for both agonist stimulation protocols. However, brief application of agonist does not result in a large fraction of fully liganded receptors, but rather has similar proportions of receptors with 3, 4 and 5 agonist molecules bound during and following the agonist pulse (centre panel). By contrast, longer duration of agonist drives most receptors into the fully liganded state (right panel). Grey lines show the summed probability of having 3, 4, or 5 agonist molecules bound. B, serotonin-induced current responses from a subset (4 of 8) of excised membrane patches (n = 4) that showed desensitization that was more rapid than deactivation were averaged by aligning the waveforms on the rise of the measured junction potential. The averaged responses to 3 ms and 5 s application of 100 μm serotonin (points) were fitted to the model shown in Fig. 7A. The smooth lines show waveforms predicted by fitting the model from Fig. 7A, varying only the rate of desensitization. C, fitted rate constants for Fit 1 from Fig. 7 are shown in the left column. The right column shows the fitted result to the curves in B when the binding and channel opening rate constants were fixed equal to those determined for Fit 1 (Fig. 7B) and the rate constants describing the onset of desensitization were allowed to vary in the fitting routine. Thus, the model shown in Fig. 7, which predicts similar deactivation and desensitization (not shown), can be converted into a model showing more rapid desensitization simply by increasing the desensitization rate constants (shown in the box) for receptors with agonist occupancy of all subunits. Other satisfactory fits to the data presented in B can be obtained by changing any combination of rate constants so that desensitization of receptors with 4 or 5 agonists bound is increased (not shown).

The model summarized in Fig. 7, derived from fitting of waveforms from prolonged application of serotonin, predicts different desensitization rates as a function of agonist occupancy. We tested whether it was possible to modify the rate constants of this model to cause desensitization that is faster than deactivation, as observed in a subset of our recordings. If desensitization accelerated with agonist occupancy, this would be sufficient to produce more rapid desensitization than deactivation. In principle, acceleration of the closing rate for fully liganded receptors could also contribute to this effect by reducing the open probability. In order to evaluate these possible mechanisms, we fitted the model shown in Fig. 7A simultaneously to the average response waveforms from four patches in which desensitization was more rapid than deactivation; in each patch we recorded the response to brief and prolonged application of 100 μm serotonin. Figure 9B shows that our working model of 5-HT_3A receptor function (Fig. 7A) can exhibit desensitization that is more rapid than deactivation when the desensitization rate constants are increased (Fig. 9C). Similar fits could be obtained by changing rate constants for opening (βn) and desensitization (kdn) in tandem (data not shown). This result suggests that modification of the desensitization rate and open probability for receptors with five ligands bound is the minimum alteration required to shift 5-HT_3A receptors into a mode from which they desensitize more rapidly than the agonist unbinds. This raises the possibility that the desensitization or open probability of fully liganded 5-HT_3A receptors might be under biological control, and underlie the variability we observe for receptor desensitization.