Ligand coupling mechanism of the human serotonin transporter differentiates substrates from inhibitors

The presynaptic serotonin transporter (SERT) clears extracellular serotonin following vesicular release to ensure temporal and spatial regulation of serotonergic signalling and neurotransmitter homeostasis. Prescription drugs used to treat neurobehavioral disorders, including depression, anxiety, and obsessive-compulsive disorder, trap SERT by blocking the transport cycle. In contrast, illicit drugs of abuse like amphetamines reverse SERT directionality, causing serotonin efflux. Both processes result in increased extracellular serotonin levels. By combining molecular dynamics simulations with biochemical experiments and using a homologous series of serotonin analogues, we uncovered the coupling mechanism between the substrate and the transporter, which triggers the uptake of serotonin. Free energy analysis showed that only scaffold-bound substrates could initiate SERT occlusion through attractive long-range electrostatic interactions acting on the bundle domain. The associated spatial requirements define substrate and inhibitor properties, enabling additional possibilities for rational drug design approaches.

Supplementary Figure 4: Validation of the free energy analysis.Box plots quantifying the mean value of the average free energy surface of the 2D free energy landscapes are computed by applying the bootstrap methods with replacement.The original population dataset comprising all SERT WT trajectories in complex with compounds was used as input for bootstrapping.The sub-samples originating from this procedure were used first to estimate the MSMs and then the mean value of the free energy surface values of the different sub-populations.A total of 100 sub-populations were generated during bootstrapping to gain 100 mean free energy surface values and estimate the variance.The boxes are defined by the 1st and 3rd quantile, a line within the box representing the median, and two whiskers extending up to 1.5 inter-quartile ranges (IQR) of the lower and upper quartile.Data points plotted as grey squares are supposed to represent outliers as they fall outside the 1.5 IQR range.Conditions are coloured according to the default colour scheme.Methyl (M)serotonin (5HT), 5HT, propyl (P)-5HT, butyl (B)-5HT, and cocaine (COC) are colour-coded in peach, mauve, cyber grape, petrol, and light blue, respectively.The dose dependence of the maximal rate of inhibition (K app ) allows for detecting if a compound stabilises SERT in only one conformation of the transport cycle, such as cocaine or in multiple states, like ibogaine.The K app reaches a stable plateau if only the one state is stabilised, while it does not saturate if it binds multiple states of the transport cycle, e.g.ibogaine.5HT and its derivatives, as well as cocaine, saturate at a K app of approximately 2 s -1 , consistent with a binding mode requiring only one state.) of cocaine (COC), ibogaine (IBO) and the 5HTderivatives (methyl (M)-5HT, propyl (P)-5HT, and butyl (B)-5HT) as a function of the compound concentration.COC and all 5HT-derivatives saturate around the rate-limiting step at 2 s -1 .In contrast, IBO shows a linear dose-dependency, which implies that its binding is not dependent on binding the outward-open conformation 3 .Data are shown as the mean of n=5 (M-5HT), n=8 (P-5HT), n=5 (B-5HT), n=9 (COC), and n=7 (IBO) recordings of independent cells.M-5HT, P-5HT, B-5HT, COC, and IBO are colour-coded in peach, cyber grape, petrol, light blue, and dark blue, respectively.Confidence intervals (95 %) of the curve fittings are plotted within ± SD and delimited by dashed lines.

Supplementary Figure 6: Establishing the double poly-sorted (DPS) YhSERT expressing HEK cell line. Expression level of N-terminally fused eYFP-hSERT (YhSERT) in HEK cells a) 24 hours after transfection, b) after maintaining stable expression, c) after the first round of fluorescence-activated cell sorting (FACS) and d) after the second round of FACS. e) Increase of total fluorescence associated with an increase of expression of
Supplementary Figure 8: Ex-vivo release properties of P-5HT.a) top) Cartoon of a sagittal section of a mouse brain.bottom) Cartoon of a coronal section highlighting in light-petrol the hippocampal tissue used for experiments (Anterior/Posterior: -3.00 mm relative to bregma).As for the in-vitro release assay, the tissue punches were loaded with 0.4 µM [3H]5HT.b) Averaged traces of n=5 (propyl (P)-5HT) and n=4 (escitalopram (S-Cit) + P-5HT and buffer) independent experiment.Data are given as the mean + SD.After collecting three basal efflux fractions, four fractions a' 2 min were collected where either S-Cit or vehicle was added to the superfusion system.At t=14 min, the brain punches were challenged with P-5HT (10 µM) in the presence or absence of S-Cit (10 µM) or vehicle for a total of five 2-minute fractions.c) Percentage of hSERT-mediated total efflux.Plotted is the average total efflux of the last four fractions shown in b) of each independent experiment.Data is shown as the mean ± SD.P-5HT, S-Cit + P-5HT, and Buffer are coloured cyber grape, light cyber grape and white, respectively.

Simulations of hSERT D98E MD simulations show that the D98E mutant alters the balance of key interactions needed for substrate transport
The single point mutation (D98E) was introduced in each of the 50 starting structures of SERT WT apo and in complex with M-5HT, 5HT, P-5HT, and B-5HT by using the Wizard Mutagenesis tool provided by PyMOL.All atom simulations and analyses were performed in accordance with the procedure described in the Material and Methods section.Each system reached a simulation length of 1 µs, hence a total simulation length of 50 µs was computed for the hSERT D98E mutant.

The D98E mutant alters the conformational equilibrium of apo SERT
A comparison between SERT wildtype (WT) and the D98E mutant in the apo state shows that the interactions of residue 98 with residues and ions in close proximity differ between aspartate (D98) and glutamate (E98).The key interacting residues are shown in SI Fig. 9f,g.Binding and stabilisation of Na1 differ as D98 shows one specific interaction with Na1 that completes the interaction shell of Na1 (SI Fig. 9a), while the D98E mutant shows multiple conformations, suggesting that the D98E SERT variant has difficulties to maintain the same interaction with Na1.Consistently, Na1 is not as stably bound (SI Fig. 9b) and can reposition to the region, which overlaps with the region in the S1 that the amino group of the substrate occupies in SERT WT.The carboxylate group of residue 98 (bundle domain) shows one consistent distance to the hydroxyl group of Y176 in SERT WT (SI Fig. 9c), but two distances for the D98E mutant.The sidechain of D98E is sufficiently long to directly interact with R104, which is part of the outer gate bridge (SI Fig. 9d).These local changes alter the global conformation of SERT.We find one broad distribution of distances between TM1b (bundle domain) and TM9up (scaffold domain), while in the D98E mutant, the distribution is changed to a bimodal distribution with the outer vestibule further open or more closed.These data show that the local change of the interaction pattern of residue 98 in the S1 leads to global conformational changes that alter the accessibility to the S1 through the outer vestibule.

The D98E mutant hampers the substrate-triggered occlusion
A comparison of simulations of WT SERT with simulations of the D98E mutant showed that SERT occlusion is altered by the D98E mutation.For M-5HT, the increased interaction between the glutamate in position 98 with the positively charged nitrogen of the compound (SI Fig. 10a), indicated by a smaller distance, leads only to a small repositioning of TM1b (SI Fig. 10c).At the same time the longer glutamate hinders the formation of a continuous interaction between the positively charged nitrogen of M-5HT and the helical dipole of TM6a, specifically with the backbone carbonyl oxygen of F335 (SI Fig. 10b).This lack of interaction prevents occlusion of SERT (SI Fig. 10d).For 5HT, the longer sidechain of the glutamate in position 98 (SI Fig. 10a) can reach to the substrate without the need of a partial closing motions of TM1b (SI Fig. 10c).In parallel, the interaction with the carbonyl oxygen of F335 is hampered (SI Fig. 10b), hence the occlusion motion of TM6a was prevented (SI Fig. 10d), resulting in an inhibition of the occlusion of the D98E variant of SERT in presence of 5HT.For P-5HT and B-5HT, the longer sidechain of the glutamate in position 98 opposes occlusion in addition to the already too long aliphatic chain of the two substrates.

The direction of motions of the bundle domain is unchanged, but the amplitude is reduced
Projection of the trajectories of the D98E mutant of apo SERT onto the two main principal components (PC) as observed for SERT WT (Fig. 2 and SI Fig. 2) revealed the same type of motions (SI Fig. 11a-f).Motions of PC1 describe the main motion of occlusion (SI Fig. 11a,b), while PC2 represents a rotation of the bundle domain perpendicular to the motion of occlusion (SI Fig. 11d,e).The per residue pattern of motion amplitudes of SERT WT (SI Fig. 2) and the D98E mutant (SI Fig. 11c,f) along PC1 and PC2 are the same.The amplitudes of motions (SI Fig. 11c,f) of individual residues of the bundle domain were smaller for PC1, while for PC2 also the amplitudes of motions are consistent between WT SERT and the D98E mutant.The smaller amplitudes of the bundle domain motions are consistent with the measures of distances across the outer vestibule between TM1b-TM9up and TM6a-TM9up and shows that the D98E mutation did not change the mechanics of the closure motions but altered its amplitude.The 2D projections (SI Fig. 11g) of all trajectories of the D98E mutant (SERT apo and in complex with the M-5HT, 5HT, P-5HT and B-5HT) show that the mutation inhibits occlusion as compared to WT SERT (Fig. 2).While M-5HT and B-5HT do not induce occlusion, 5HT and P-5HT have a smaller propensity to induce SERT occlusion.

Supplementary Figure 11: Principal component analysis of hSERT D98E apo and in complex with
5HT and its derivatives.The principal component analysis (PCA) was carried out using all 50 trajectories of the D98E mutant of SERT on the Cα-atoms of all transmembrane helices (TMs) and projected onto the eigenvectors derived for the WT simulations as shown in Fig. 2 and SI Fig. 2

to allow for direct comparison. Front view of the two extreme structures of a) Principal component 1 (PC1) and d) PC2. Top view of the extreme structures of b) PC1 and e) PC2. Structures in white represent the starting structures, whereas the coloured structures (red and blue) depict the conformations with the largest changes. The associated per residue root mean square fluctuation (RMSF) of motions along c) PC1 and f) PC2, coloured in red and blue, respectively. g) 2D projections of all 10 (n=10) simulations of the SERT D98E mutant (apo and in complex with methyl (M)-5HT, 5HT, propyl (P)-5HT, and butyl (B)-5HT) along PC1 and PC2 that describe the largest motions of SERT WT.
Apo, M-5HT, 5HT, P-5HT, and B-5HT are colour-coded in red, peach, mauve, cyber grape, and petrol, respectively.

The D98E mutant alters the conformations of TM1b and TM6a and opening of the outer vestibule
The distances between TM1b (residue 99-111) and TM9up (residue 475-477) as well as between TM6a (residue 328-338) and TM9up are a surrogate measure of the closure of the outer vestibule and of bundle domain motions (SI Fig. 12).TM9 is part of the scaffold domain, while TM1 and TM6 are part of the bundle domain.These helices reside on opposite sides of the vestibule.Consistent with the above measures, the time traces show a changed open geometry of the apo D98E mutant and altered ligand-induced closing dynamics.

Synthesis of 5HT analogues with modified alkyl chain length
Analysis of the simulation data generated the testable hypothesis that a modification of the length of the alkyl chain of 5HT should change transport properties of SERT.To experimentally verify this prediction, we developed a synthetic strategy for the compounds M-5HT, P-5HT and B-5HT, as these compounds were rarely investigated 4,5 and no feasible syntheses were described.The steps of the chemical syntheses are outlined in SI Fig 13 .As starting material, 5-(benzyloxy)-1H-indole (1) was selected, as protecting the 5-hydroxy group was necessary for easier synthetic manipulation and the benzyl group additionally allowed for reductive cleavage, conveniently fitting later synthetic steps.In the first step, an aldehyde moiety was introduced by a regioselective Vilsmeyer-Haack reaction 6 yielding compound 2. Noteworthy, adding catalytic amounts of acetic acid accelerated the reaction considerably by facilitating acid-catalysed hydrolysis of the imine-intermediate.The obtained aldehyde 2 was interconverted to oxime 3 using hydroxylamine hydrochloride 6 .Deprotection of the 5hydroxy group and the reduction of the oxime to the amine could be performed simultaneously to reach the hydrochloride salt of M-5HT (4) in 87% yield over 3 steps 7 .
For the synthesis of P-5HT (7), first, a Wittig reaction 6 was performed on compound 2, simultaneously elongating the carbon chain and introducing a nitrile group yielding compound 5. Next, we had planned to reduce the alkene and the nitrile group while also cleaving the benzyl group by choosing the correct reducing conditions.However, this was not possible as even under harsh conditions the nitrile group would not be reduced.Hence, a sequential approach was employed.First, compound 6 was obtained via reductive hydrogenation that led to deprotection of the 5-hydroxy group and reduction of the double bond 8 .P-5HT (7) was reached through reduction of the nitrile of compound 6 using LiAlH 4 9,10 .The yield of the synthesis of P-5HT (7) was 71% over three steps.The hydrochloride salt of P-5HT (7) could not be obtained as exposure to conc.HCl in dioxane led to decomposition of the material.However, NMR-based stability studies (SI Fig. 18, 18a-e) confirmed high chemical stability of the pure P-5HT as free amine base, as no degradation in solution (DMSO) could be observed even after storage up to five weeks at room temperature.
The synthesis of B-5HT (10) required a different synthetic route as analogous Wittig reactions with compound 2 failed, most likely due to lacking conjugation of the yield to the electron withdrawing nitrile group.Starting from compound 1, compound 8 could be synthesised in 28% yield using Grignard reagent MeMgI as base and a suitable CN-bearing alkyl bromide as reagent 11 .Next, a similar synthetic cascade as performed for compound P-5HT (7) led to the isolation of B-5HT (10) with an overall yield of 15% over three steps [8][9][10] .The stability of B-5HT was comparable to P-5HT as confirmed by NMR stability studies (SI Fig. 19, 19a,b).

3-(4-Aminobutyl)-1H-indol-5-ol (10)
Compound 10 was prepared according to modified literature procedures 9,10 .Starting material 9 (58.8 mg, 0.29 mmol, 1.0 equiv.was dissolved in dry THF (2.5 mL) under argon.The solution was cooled to 0°C and LiAlH 4 (56.0 mg, 1.48 mmol, 5.0 equiv.) was slowly added to it.After complete addition and finished H 2 formation, the reaction mixture was heated to 55°C under argon.After 3 hours TLC monitoring (1:1 LP:EtOAc and MeOH + 1% NH 3 ) showed full conversion.The reaction was cooled with a water bath and quenched by slowly adding MeOH.The heterogeneous grey solution was directly applied on a silica column with about 10 g silica and a diameter of 3 cm for purification with MeOH + 1% NH 3 as solvent.After evaporation of the solvent in vacuo desired compound 10 was obtained as yellow to orange oil (33.2 mg, 55%, 4w% methanol, 90% purity). 1

STABILITY STUDIES OF 5HT-DERIVATIVES
As serotonin, as many substrates with free amine and free hydroxyl groups, is known to be a rather unstable compound, we set out to investigate the stability of our synthesised compounds M-5HT (4), P-5HT (7) and B-5HT (10).It was important to us to establish their stabilities to be more confident not only in storage of the pure compounds but also storage and reuse of sample stock solutions for in vitro testing.Therefore, we conducted stability studies by NMR in several solvents like D 2 O and DMSO-d 6 .
Most importantly, we wanted to check if the samples would show any signs of degradation like non-compound related peaks appearing or decreasing peak integrals in comparison to the solvent peak.NMR samples of the three compounds in the desired solvents were prepared and the same samples in the same tubes remeasured over a timeframe of up to four months.For all the following NMR spectra the integral of the proton H2 peak is set to one.

Stability study in D 2 O
SI Fig. 16 shows the obtained stacked spectra for measurement in D 2 O.As it can be seen, no additional peaks occur during this time period.However, especially after 4 months, we can observe a decrease in the peak integrals compared to the solvent peak.Moreover, even though the integral for the singlet at 7.44 ppm (H2) still shows an integral of one, the peaks at 7.11 ppm (H4) and 6.84 ppm (H6) show diminished integrals, accounting for deuterium exchange with the deuterated NMR solvent (SI Fig. 16a-e).
Nevertheless, M-5HT (4) shows good stability in D 2 O for at least two weeks.Looking at the full 1 H spectra, we see a slight increase in the integral of the solvent peak after five weeks.This points towards either minor degradation to material not visible in the NMR or precipitation of the compound.

3-(3-Aminopropyl)-1H-indol-5-ol (7)
A sample of compound P-5HT (4) in DMSO-d 6 was prepared and 1 H-NMRs measured immediately, after 2 weeks, 3 weeks, and 4 months.SI Fig. 18 shows the obtained stacked spectra for measurement in DMSO-d 6 .As the stability studies were performed on preliminary material, the proton NMR shows some solvent and minor impurity peaks next to the compound peaks.Nevertheless, similar to the behaviour of compound M-5HT (4), no additional peaks occur during the observed time frame of 4 months.Deuterium exchange seems to be happening here too, as many peaks except the 7.09 ppm (H2) peak show slightly decreasing integrals and the 1 H-NMR after 4 months does suggest some precipitation.However, good stability of P-5HT (7) in DMSO-d 6 for at least three weeks was found (SI Fig. 18a-d).

Supplementary Figure 3 :
Markovian behaviour validation.a) Time traces are shown for the relation between the lag time and the implied timescales associated with the ten slowest processes, with the blue trace indicating the slowest process.The solid lines refer to the maximum likelihood result, the dashed lines show the ensemble mean computed with a Bayesian sampling procedure.The black line with the grey area indicates the timescale threshold where the MSM cannot resolve processes.b) displays the Chapman-Kolmogorov (CK) test calculated using a lag time of 25 ns and assuming five macrostates.The solid traces show the predicted values obtained from the MSMs, while the dashed traces show the estimated values for a longer lag time.The superposition of predicted and estimated values indicates that the MSM assures Markovian behaviour.The non-grey shaded areas in both panels indicate 95% confidence intervals computed using the Bayesian sampling procedure mentioned above.
YhSERT.Data are represented as mean ± SD from n=5 independent dishes used to conduct the experiments shown in f. f) Dose-dependent SERT-mediated [3H]-serotonin (5HT) uptake within 1 minute of the 4 different cell lines.Data are shown as mean ± SD from n=5 independent experiments measured in triplicates.g) Plot f normalised to the total fluorescence shown in e. Confidence intervals (95 %) of the curve fittings are plotted within ± SD and delimited by dashed lines.

Supplementary Figure 7 :
Conformational state dependence and its impact on drug binding kinetics.a) Simplified kinetic model of SERT-mediated 5HT reuptake and inhibitor binding adapted from Schicker et al. 1 and Niello et al. 2 .The kinetic model displays main and intermediate states of SERT within the transport cycle as described by the alternating access model.Arrows and numbers display the kinetics between the states.The scaffold domain is kept in grey.Transmembrane helix 1 (TM1), TM2 are coloured in red and TM6, TM7 are highlighted in yellow.The blurred and light blue ellipse highlights the slowest (time-limiting) step of the transport cycle.Serotonin (5HT), Na + , K + , and Cl -are coloured in mauve, blue, purple, and cyan, respectively.Abbreviations: T o : transporter outward-open, T i : transporter inward-open.b) Representative trace of the binding kinetics of 5HT-mediated current inhibition (5HT concertation: 10 µM) by saturating concentrations of M-5HT (200 µM), P-5HT (200 µM), B-5HT (200 µM), cocaine (200 µM) and ibogaine (200 µM).c) Close-up of the kinetics of inhibitor binding shown in b.Mono-exponential curves fitted into the current inhibition by COC and IBO are shown in red.d) Current inhibition rates (s -1