Human monocarboxylate transporters accept and relay protons via the bound substrate for selectivity and activity at physiological pH

Abstract Human monocarboxylate/H+ transporters, MCT, facilitate the transmembrane translocation of vital weak acid metabolites, mainly l-lactate. Tumors exhibiting a Warburg effect rely on MCT activity for l-lactate release. Recently, high-resolution MCT structures revealed binding sites for anticancer drug candidates and the substrate. Three charged residues, Lys 38, Asp 309, and Arg 313 (MCT1 numbering) are essential for substrate binding and initiation of the alternating access conformational change. However, the mechanism by which the proton cosubstrate binds and traverses MCTs remained elusive. Here, we report that substitution of Lys 38 by neutral residues maintained MCT functionality in principle, yet required strongly acidic pH conditions for wildtype-like transport velocity. We determined pH-dependent biophysical transport properties, Michaelis–Menten kinetics, and heavy water effects for MCT1 wildtype and Lys 38 mutants. Our experimental data provide evidence for the bound substrate itself to accept and shuttle a proton from Lys 38 to Asp 309 initiating transport. We have shown before that substrate protonation is a pivotal step in the mechanisms of other MCT-unrelated weak acid translocating proteins. In connection with this study, we conclude that utilization of the proton binding and transfer capabilities of the transporter-bound substrate is probably a universal theme for weak acid anion/H+ cotransport.


Introduction
Metabolic monocarboxylic acids largely deprotonate in the human body due to an acidic pK a around 4 (1). The resulting monocarboxylates represent typically more than 99% of the equilibrium species at physiological pH (2). Transmembrane diffusion of the anions is strongly impeded and dedicated transporters are required for release as metabolic waste or uptake as nutrients. Transport of the monocarboxylate alone, however, would have unfavorable consequences in a physiological setting. First, anion transport is electrogenic. While the negative membrane potential would drive the export of monocarboxylates, import would be strongly hindered (3). This would basically prohibit the use of monocarboxylates as nutrients. Second, monocarboxylates, e.g. pyruvate or L-lactate, that derive from metabolic redox processes such as glycolysis are produced together with one proton each. Export of the monocarboxylate alone would result in detrimental acidification of the cytosol by the remaining protons if not resolved by energy-consuming H + -ATPase activity (4).
Four monocarboxylate transporters, MCT1-4, of the SLC16 family carry the main load of human transmembrane L-lactate transport by a proton cotransport mechanism (3)(4)(5)(6). Their physiological role is to maintain energy metabolism. MCTs are further at the core of the Warburg effect of glycolytic tumors that is characterized by the production of large amounts of L-lactate despite an adequate supply of oxygen (7). Circulating L-lactate can fuel oxidative tumor cells via MCTs (8). Therefore, MCT inhibitors are being developed as anticancer drugs (9).
Recently, high-resolution MCT structures were generated to elucidate the molecular transport mechanism and drug candidate binding sites (10)(11)(12). The overall MCT protein structure comprises 12 transmembrane helices with an internal symmetry of the first and second helix bundle probably originating from gene duplication (13). MCTs undergo a large conformational change during transport by rigid-body rotation supporting an alternating access transport mechanism. Drug-like molecules lock MCT1 in its outward-or inward-open conformation, and a D309N mutation fixed MCT1 in the inward-open state in the absence of an inhibitor (12). In wildtype MCT1, Asp 309 forms a salt bridge with Arg 313 (Fig. 1A) (14). Breakage of the interaction by protonation of Asp 309 has been hypothesized to initiate the alternating access transition (12). A third charged amino acid residue, Lys 38, is present at the MCT1 substrate binding site. Lys 38 with a protonated ɛ-amine was suspected before to represent a major binding site for the monocarboxylate substrate (15). Mutation of either of the three charged residues reportedly rendered MCT1 nonfunctional (10,12,15,16). While the previous functional and structural studies identified essential residues for substrate binding and the transport process, the proton binding site and cotransport mechanism remained elusive or speculative.
Here, we expressed human MCT1 and Lys 38/Asp 309/Arg 313 mutation variants in a Saccharomyces cerevisiae strain lacking endogenous L-lactate transporting proteins (17,18). Contrary to other expression systems, background transport is marginal in this strain, and the absence of L-lactate metabolizing enzymes allows for prolonged assay times to reach the transport equilibrium state (17). Key to this study, yeast tolerates strongly acidic external pH conditions. When we assayed a MCT1 K38M mutant at pH 3.8 and pH 2.8, it unexpectedly exhibited transport velocities above wildtype MCT1 at its pH 4.8 optimum. Starting out from this result, we related pH-dependent biophysical transport properties, Michaelis-Menten kinetics, and heavy water effects of MCT1 wildtype to mutation variants of Lys 38. This revealed the proton cotransport mechanism in which the bound substrate shuttles a proton from Lys 38 to Asp 309 initiating the alternating access conformational change. The required proton transfer across the substrate is a key element in the selectivity mechanism of the MCT family of monocarboxylate transporters.

Results
We started out by generating point mutants of human MCT1 in which the charged residues of the L-lactate binding site were individually replaced by amino acids with neutral sidechains, i.e. MCT1 K38M, MCT1 D309N, and MCT1 R313N. The Arg 313 position was mutated to asparagine because this exchange is naturally present in another human MCT, i.e. MCT12 a transporter for zwitterionic creatine. We expressed the MCT1 mutants in S. cerevisiae W303-1A Δjen1 Δady2 lacking endogenous monocarboxylate transporters (Figs. S1A and S1B).
We generated more MCT1 Lys 38 mutants, namely MCT1 K38A with a shorter neutral sidechain, MCT1 K38E to reverse the sidechain charge to negative, and MCT1 K38R carrying the two orders of magnitude stronger guanidine base to prevent potential proton release (Figs. S1A and S1B). MCT1 K38A behaved as MCT1 K38M (Fig. S1D). MCT1 K38E was inactive at pH 6.8 and 4.8; yet pH 2.8, which is about one log-unit below the pK a of the glutamate γ-carboxyl, initiated transport (Fig. S1E). This indicates that a glutamate at position 38 is solvent accessible, and partial neutralization by protonation allows for transport similar in fashion to K38M and K38A. MCT1 K38R was nonfunctional at all tested pH conditions (Fig. S1F). Possibly, a permanent charge at position 38 prohibits transport, whereas Lys 38 in MCT1 wildtype may not be charged at all times but could be part of a proton transfer process. It cannot be excluded, though, that the larger size of the arginine sidechain may contribute to the loss of function.

Lys 38 enables MCT1 transport at physiological pH
Next, we carried out a fine-grained titration of MCT1 wildtype and K38M transport rates against pH to determine the number of accessible protonation sites relevant to L-lactate/H + transport (Fig. 1C). The titration curves appeared strongly shifted, and differed in complexity. The sigmoidal curve of MCT1 K38M indicated a single transport-relevant protonation step. The inflection point of the sigmoidal function (pH 50 ) was determined to 3.97 (Fig. 1C, blue curve), which corresponded to the intrinsic pK a of L-lactic acid (pK a 3.86; Fig. 1D). The curve shape and pH 50 were highly reminiscent of the L-lactic acid permeability of the solute channel aquaporin-9, AQP9 (19,21) supporting the view that MCT1 K38M equally uses neutral L-lactic acid as a substrate. Accordingly, at sufficiently acidic buffer, the L-lactate anion itself binds and carries the proton cosubstrate for MCT1 K38M.
With a lysine present at position 38 of MCT1 wildtype, L-lactate/ H + transport functionality was shifted to neutral and mildly acidic physiological pH (Fig. 1C, orange curve). Further, the curve can be sectioned: a monotonically increasing part depicting an initial raise in the transport rate from neutral pH towards mildly acidic conditions followed by a prominent boost in transport velocity at more acidic conditions. This part was fitted best with a doublesigmoidal function (r 2 = 0.9978) with inflection points at pH 7.43 and pH 5.61 (Fig. 1C). Transport velocity was maximal at pH 4.8, and exponentially decreased inverse in shape to the pH curve of MCT1 K38M (Fig. 1C). The opposite pH dependence in the most acidic curve section suggests that MCT1 wildtype accepts the L-lactate anion as a substrate rather than neutral L-lactic acid.
A double-sigmoidal curve shape may indicate two bufferaccessible protonation sites that influence the overall transport process. Since this part is absent in the Lys 38-less MCT1 K38M mutant, the protonation sites may be linked to the Lys 38 sidechain ɛ-amine in the substrate-unbound state and to the complex of Lys 38 with L-lactate. While the protonation status of the Lys 38 sidechain appears to be determined by the buffer pH (see also MCT1 K38E mutant; Fig. S1E), we did not notice direct protonation of Asp 309 in our titrations by external protons. This is fostered by solvation simulations based on the protein structure of MCT1 (12) in which bulk-linked water clusters appeared close to Lys 38, whereas water was absent around the Asp 309 site (Fig. S2). Together, the pH data show that MCT1 with replaced Lys 38 is functional in principle; yet, the basicity of the lysine ɛ-amine is required to enable transport at physiological pH.

Direct evidence for binding of L-lactate to the protonated Lys 38 ɛ-amine
We figured that if the observed pH 50 of 7.43 was derived from Lys 38, L-lactate affinity should be strongly decreased at alkaline pH due to deprotonation of the ɛ-amine. Therefore, we determined Michaelis-Menten kinetics (Fig. 2). Indeed, the L-lactate affinity at pH 7.8 appeared one order of magnitude lower (K m 19.2 ± 4.1 mM) than at physiological pH conditions (K m 2.32 ± 0.36 mM) (Figs. 2A-C). The decrease in affinity at pH 3.8 (K m 7.29 ± 0.40 mM) is due to the shift of the L-lactate/lactic acid equilibrium towards the neutral species (Fig. 1D). A pH of 8 was shown to lock MCT1 in the outward-open conformation (12), and a strong decrease in L-lactate affinity at alkaline pH was seen with human erythrocytes in which MCT1 is the prominent L-lactate transporter (22). Together with our data, this provides evidence for a binding mechanism in which a proton binds first to the Lys 38 ɛ-amine and generates a positive charge for the subsequent binding of the L-lactate anion. The observed decrease in affinity at the most acidic buffer pH of 3.8 (Fig. 2C) is due to the shifted L-lactate/lactic acid equilibrium towards the neutral form (Fig. 1D).
The maximum velocity of transport, v max , also correlated with the MCT1 Lys 38 protonation status (Fig. 2D). At pH 7.8, i.e. in the mainly uncharged state of Lys 38, transport was up to one order of magnitude faster than at lower pH conditions. We attribute this to less impedance of the substrate at the binding site due to weaker interaction with the neutral Lys 38 sidechain.
To meet the acidic pH requirements of MCT1 K38M, we measured K m at pH 3.8 (Fig. 2E). The substrate affinity (K m 2.13 ± 0.19 mM; Fig. 2C) was equal to MCT1 wildtype at a physiological pH, and v max was higher than that of MCT1 wildtype at pH 6.8 and 4.8, yet not as high as at pH 7.8 (Fig. 2D). An uncharged residue at position 38 is beneficial for transport velocity, and polar interactions with the amine of Lys 38 further accelerate transport. The generally comparable K m and v max values of MCT1 wildtype and K38M suggest that the proton-driven alternating access transport mechanism remained unaltered in principle despite entry of the substrate in different forms, i.e. L-lactate/H + (MCT1 wildtype) vs.

MCT1 is selective for protonatable weak acid anions
Next, we tested whether MCT1 wildtype or MCT1 K38M would accept L-lactamide, a permanently neutral mimic of L-lactic acid ( Fig. 3A), as a transport substrate. We chose a buffer pH of 3.8 for equal concentrations of L-lactate and L-lactic acid (Fig. 1D). At an equimolar concentration or 10-fold excess of L-lactamide, neither MCT1 wildtype nor MCT1 K38M showed a decrease in the uptake rate by competition (Fig. 3A), indicating that L-lactamide is not an MCT1 or MCT1 K38M substrate. We employed a second method to test more directly for L-lactamide transport by MCT1 using stopped-flow light scattering ( Fig. 3B) (21,23). This assay requires a steep substrate gradient to elicit sufficiently large changes in cell volume for detection. In the initial phase, the hyperosmotic conditions rapidly drive water out of the yeast eliciting an increase in light scattering (Fig. 3B). The slower import of MCT substrates leads to partial re-swelling and a decrease in light scattering. As in the radiolabeled competition assays, L-lactamide was not transported (Fig. 3C). Testing of further similarly sized compounds fostered the mechanistic findings of this study. We noted that the transport rate of L-lactate and chloroacetate (24) decreased with decreasing pK a , i.e. 3.86 and 2.83, respectively. This again indicates that direct substrate protonation is relevant for transport. Neutral glycerol, which is incapable of binding and transferring protons, was excluded from transport by MCT1, as was zwitterionic L-proline (Fig. 3C) (25).

Bound weak acid substrates shuttle protons from Lys 38 to Asp309
The distance between the Lys 38 ɛ-amine and the Asp 309/Arg 313 salt bridge was determined to about 8 Å (Fig. S2) (12). A substrate carboxyl group appears suited to fill the gap and may shuttle a proton from Lys 38 to Asp 309. Protonation of Asp 309 was suggested to break the salt bridge initiating the alternating access mechanism (12). To gain insight into relevant proton transfer events, we determined the effect of heavy water, D 2 O, on MCT1 transport.
Deuterons have twice the mass of protons, which increases the chemical bond strength resulting in a characteristic pK a shift of monocarboxylic acids by approx. 0.5 (26), and a decrease of the deuteron-hopping velocity (27). The degree by which a protondependent chemical process is retarded in heavy water, thus, is indicative of the complexity of the involved transfer chain (28).
We changed the solvent of the yeast suspensions from light water to heavy water at least 30 min prior to the transport assays for adjustment (29,30). Other than that, the buffers were identically composed and contained the same concentrations of protons or deuterons, i.e. pH or pD. The height of the plateaus derived from the obtained or extrapolated uptake curves (Fig. 3D) indicates the total amount of intracellular substrate in the equilibrium states, termed capacity. Throughout, the capacity of L-lactate/D + transport was higher than with L-lactate/H + (Fig. 3D, Fig. S3A). The heavy water effect was most pronounced at pD 6.8, and higher free deuteron concentrations somewhat compensated the effect (Fig. 3E, top). The rate constants indicated slower transport of MCT1 wildtype in heavy water at all pH/pD conditions (Fig. 3E, bottom, Fig. S3B). MCT1 K38M transport, in turn, appeared slightly accelerated in heavy water. Inverse isotope effects are rare and can indicate that substrate protonation takes place outside of the protein, and other rate limiting events determine the velocity of the overall process (31).
The alternating access transition is usually a major ratelimiting process in the transporter class. Making use of the preprotonated pool of L-lactic acid in the external buffer may explain the mild inverse isotope effect observed with MCT1 K38M. The normal isotope effect with MCT1 wildtype, in turn, indicates that proton transfers inside the protein are relevant for the overall transport velocity. With the previous observations of this study, this suggests that L-lactate binds to Lys 38 and accepts a proton from the lysine ɛ-amine. The final and common step for MCT1 wildtype and the K38M mutant would be a proton transfer from the substrate to Asp 309 that breaks the salt bridge and initiates the alternative access conformational change.

Discussion
The MCT family of monocarboxylate/H + transporters evolved in a way that efficiently enables utilization of the prevalent pool of monocarboxylate anions by exploiting the transmembrane pH gradient generated by the accompanying protons. The data of this study suggest that the bound substrate itself contributes to proton transfer. Since the amino acid residues of the binding site are conserved among the mammalian MCT-type monocarboxylate transporters (Fig. S4), we view the general transport mechanism as follows: (I) The substrate anion is electrostatically steered to the MCT binding site by a net charge of plus one originating from the protonated Lys 38 ɛ-amine (Fig. 4, left). The charge compensation results in a neutral protein interior and possibly prevents differently charged organic anions, such as dicarboxylic acids or phosphates, from being accepted as substrates. The principle of transporter charge compensation has been shown before, e.g. for the mitochondrial alternating access ADP/ATP carrier of the SLC25 family (32). ADP has a charge of minus three matching three positively charged residues in the binding pocket for import into the mitochondrial matrix. ATP with a net charge of minus four is transported in the opposite direction driven by the negative membrane potential. (II) In a second step, a proton from MCT1 Lys 38 neutralizes L-lactate to form L-lactic acid (Fig. 4, center). This is facilitated by the more lipophilic environment within the protein that leads to convergence of both pK a . Only substrates that have the capability to accept and transfer a proton under these conditions can initiate the next step in the transport mechanism. This process would further account for the strict selectivity of MCT1-4 for monocarboxylates. (III) Eventually, a proton transfer from L-lactic acid to Asp 309 leads to breakage of the Asp 309/Arg 313 salt-bridge (Fig. 4, right). The MCT protein converts into the inward open state from which L-lactate and the Asp 309-bound proton are released. The transport cycle is complete with the re-protonation of Lys 38. At an effective pK a of 7.43, this should readily occur from the aqueous bulk even at neutral pH conditions. Generally, water molecules can be part of proton transfer chains. The observed mild heavy water effect, however, is in favor of a transfer chain with a small number of hopping steps. This transport model is compatible with the bidirectionality of MCT transport. In the inward open cryo-EM structure of MCT1 D309N, the Lys 38 ɛ-amine seems accessible to substrate molecules ( Fig. S5) (12). Substrate binding from the cytosolic side would reverse the above-described process. Therefore, directionality of transport is determined by the transmembrane substrate and proton gradients via the probability of binding from the extra-and intracellular space. Further, accessory proteins, i.e. the extracellular domain of basigin and isoforms of the carbonic anhydrase family, modulate MCT directionality by exerting antenna functions for substrate anions and/or protons (17,33).
Currently, three types of membrane proteins have been functionally and structurally described that facilitate transmembrane translocation of a weak acid anion plus a proton. These are members of the ubiquitous AQP family (19,21,34,35), microbial formate/nitrate transporters, FNT (2,(36)(37)(38), and MCT-type monocarboxylate transporters. All use electrostatic attraction of the weak acid anion by positively charged amino acid residues (39). The level of intricacy of the translocation mechanism increases from the AQPs via FNTs to the MCTs.
AQPs represent the least efficient weak acid facilitators and require strongly acidic conditions. Clusters of positively charged amino acids at the protein surface of, e.g. Lactobacillus AQPs attract L-lactate anions (34,35) and locally enhance the probability of substrate protonation for passing the lipophilic channel (21). The FNTs are lookalikes of the AQPs (36). Despite unrelated protein sequences, they adopted the same channel-like fold, yet a positively charged lysine is positioned deep inside funnel-like vestibules on either side of the central channel section (2). A weak acid anion is attracted into the increasingly hydrophobic environment within a vestibule by the "dielectric slide" mechanism (37). At a certain point, the shifting substrate pK a allows a proton to bind. The neutral acid passes a lipophilic constriction which additionally selects by size exclusion (38). The FNT layout enables proton-driven weak acid transport at physiological pH of similar efficiency as the MCTs. Yet, the FNTs exhibit electrogenic anion leakage at neutral pH (40). With the advent of the metazoa, the channel-like FNTs (41) were replaced by the alternating access MCTs (13). In the MCTs, two central steps of weak acid transport appear to be maintained, namely attraction of the acid anion and subsequent proton binding by the substrate. The additional employment of a proton transfer via the bound substrate to the switching amino acid Asp 309 is probably responsible for eliminating anion leakage altogether.
In conclusion, the transport mechanism of the MCTs supports the notion, that, throughout species, a hallmark of membrane protein-facilitated weak acid transport is the exploitation of the substrate-inherent physicochemical properties as a proton acceptor and/or transfer site for selectivity and activity in the physiological pH range.

Plasmids, cloning, and mutagenesis
Human MCT1 in pDRT196 with N-terminal hemagglutinin-and C-terminal His 10 -tags was generated previously (17). Primers for site-directed mutagenesis are listed in Table S1. All constructs were sequenced.
Radiolabeled L-lactate uptake assay Uptake of radiolabeled L-lactate was assayed as previously (45). Briefly, yeast was harvested (4,000 g, 5 min, 4°C, OD 600 of 1 ± 0.1), and washed with water. The cells were resuspended in pH/ pD-adjusted buffer to an OD 600 of 50 ± 5, and kept on ice for 30 min. For the assay, the cells were prewarmed to 19°C. Substrate uptake was initiated by adding 14 C-spiked (0.04 µCi Hartmann Analytic, Braunschweig, Germany) L-lactate solution (1 mM final). Transport was stopped with 1 ml ice-cold water, and suction through GF/C glass microfiber filters (GE Healthcare, Solingen, Germany). The washed filters were placed in 3 ml of ROTISZINTeco plus (Carl Roth, Karlsruhe, Germany) overnight, and counted for 2 min (Packard TriCarb 2900 TR, Perkin Elmer). For Michaelis-Menten kinetics, substrate and radiolabel were increased up to 50 mM and 0.14 µCi. L-Lactamide competition was assayed at 4 mM L-lactate plus 4 mM or 40 mM L-lactamide. Batches in light and heavy water were prepared from the same yeast culture. Internalized L-lactate was normalized to 1 mg of yeast, and background radiolabel on non-expressing cells was subtracted. Uptake rates were determined from the initial linear phase of the curves (usually 2 min, 8 min for slow transport). Michaelis-Menten parameters were calculated from exponential curve fittings (K m = ln(0.5)/rate constant; v max = limit of the curve). Measurements were done in duplicates with two to six biological replicates.
Monitoring of changes in yeast protoplast volume by stopped-flow light scattering The assay was adapted from previous work (21,23). Here, 50 ml transformed yeast were harvested (2,000 g, 5 min, 4°C, OD 600 of 1 ± 0.1), washed and resuspended (2 ml) in 50 mM MOPS buffer, pH 7.2, with 0.2% of freshly added 2-mercaptoethanol. For protoplastation, zymolyase 20 T (400 U per gram of yeast cells), 100 mg of bovine serum albumin (fraction V), and 1.2 M sucrose (all from Carl Roth, Karlsruhe, Germany) were added and incubated for 1 h at 30°C, 140 rpm. Protoplasts were collected (2,000 g, 5 min, 4°C), washed once in 5 ml of 20 mM MOPS buffer, pH 7.2, plus 1.2 M sucrose, 50 mM NaCl and 5 mM CaCl 2 , and diluted in this buffer to an OD 600 of 2. For the assay, 75 µl of protoplast suspension were mixed using a stopped-flow device (SFM-200, BioLogic, Seyssinet-Pariset, France) with an equal volume of buffer supplemented with 600 mM L-lactate, chloroacetate, glycerol, L-lactamide, or L-proline (300 mM inward gradient). For an assay pH of 5.1 ± 0.3 the protoplast suspension was mixed with a pH 4.5 ± 0.3 buffer. Further parameters were: 20°C, dead time 2.7 ms, flow rate 8 ml s −1 . The intensity of 90° light scattering (λ = 524 nm) was monitored. Nine traces were averaged, normalized, and done in three biological replicates.
Substrate uptake velocity was derived from by linear fits (20-90 s range) of background-subtracted traces. Significance of was tested using one-tailed one-sample t-tests. Asterisks indicate P-values <0.05.

Protein structure visualization and solvation simulation
MCT1 protein structure displays were done with the Chimera software (46). Solvation of the MCT1 binding pocket in the outward open conformation was simulated by placing the PDB structure file 6LZ0 (12) in an SPC/Fw water shell with an extent of 5 Å using the AMBER tools (47) from within the Chimera software.