Research ArticleBIOCHEMISTRY

Nitrogen isotope signature evidences ammonium deprotonation as a common transport mechanism for the AMT-Mep-Rh protein superfamily

See allHide authors and affiliations

Science Advances  12 Sep 2018:
Vol. 4, no. 9, eaar3599
DOI: 10.1126/sciadv.aar3599


Ammonium is an important nitrogen (N) source for living organisms, a key metabolite for pH control, and a potent cytotoxic compound. Ammonium is transported by the widespread AMT-Mep-Rh membrane proteins, and despite their significance in physiological processes, the nature of substrate translocation (NH3/NH4+) by the distinct members of this family is still a matter of controversy. Using Saccharomyces cerevisiae cells expressing representative AMT-Mep-Rh ammonium carriers and taking advantage of the natural chemical-physical property of the N isotopic signature linked to NH4+/NH3 conversion, this study shows that only cells expressing AMT-Mep-Rh proteins were depleted in 15N relative to 14N when compared to the external ammonium source. We observed 15N depletion over a wide range of external pH, indicating its independence of NH3 formation in solution. On the basis of inhibitor studies, ammonium transport by nonspecific cation channels did not show isotope fractionation but competition with K+. We propose that kinetic N isotope fractionation is a common feature of AMT-Mep-Rh–type proteins, which favor 14N over 15N, owing to the dissociation of NH4+ into NH3 + H+ in the protein, leading to 15N depletion in the cell and allowing NH3 passage or NH3/H+ cotransport. This deprotonation mechanism explains these proteins’ essential functions in environments under a low NH4+/K+ ratio, allowing organisms to specifically scavenge NH4+. We show that 15N isotope fractionation may be used in vivo not only to determine the molecular species being transported by ammonium transport proteins, but also to track ammonium toxicity and associated amino acids excretion.


Ammonium (unless a distinction is made, the term ammonium used in this paper refers to both chemical forms: NH4+ and NH3) is a major nitrogen (N) source for living organisms and supports tumor proliferation (13). It is not only a key metabolite in the control of systemic pH by the kidney but also a potent cytotoxic compound (14). The conversion of the ionic (NH4+) to the neutral (NH3) form depends on pH and temperature, with a pKa (where Ka is the acid dissociation constant) ≈ 9.25 at 25°C (pKa ≈ 9.09 at 30°C); thus, at pH 7 and below, more than 99% of ammonium in solution is present in its ionic form. Ammonium transport across biological membranes was initially thought to occur mainly through NH3 diffusion while it was later reconsidered to also proceed via carrier-mediated NH4+ transport (57). At low external substrate concentrations, ammonium transport across membranes occurs via AMT-Mep-Rh proteins, which are highly selective against K+, despite the latter sharing a similar ionic radius with hydrated NH4+ ions (68). The AMT-Mep-Rh protein family is divided into two main subfamilies: ammonium transporters (AMT)–methylammonium/ammonium permeases (Meps) and rhesus (Rh) proteins. Most AMT-Mep proteins have been characterized as high-affinity ammonium transport systems, with Km values generally in the micromolar range (7, 8). Rh proteins are phylogenetically more distantly related and transport ammonium with lower affinity (Km values in the millimolar range) (1, 9, 10). This highly conserved superfamily of membrane proteins is present in organisms belonging to all main lineages of life, including bacteria, fungi, algae, plants, and animals (9). Following the discovery and functional characterization of this protein family and the solving of the first three-dimensional (3D) structures, the question whether NH3 or NH4+ is the transported substrate arose (68, 11, 12).

Heterologous expression of plant AMT transporters in Xenopus laevis oocytes and electrophysiological characterization using the two-electrode voltage clamp technique suggest an electrogenic transport mechanism for AMT1-type proteins (9, 13). This mechanism may, for instance, correspond to the transport of NH4+ or the cotransport of H+ and NH3. In contrast, an electroneutral transport mechanism is proposed for AMT2-type transporters (14).

On the basis of the structures of AMT-Mep-Rh proteins, NH3 is proposed to be the form transported for several members of the AMT-Mep-Rh protein family, typically characterized by a narrow hydrophobic pore between two hydrophilic regions, which constitute the periplasmic and cytoplasmic vestibules (15). The ionic form NH4+ appears to be the preferential form recruited in the periplasmic vestibule, and the constriction is proposed to allow NH3, but not NH4+, to permeate the transporter (11, 15). However, these static views are not sufficient to accurately determine the transport mechanism. Overall, several possibilities for electrogenic or electroneutral ammonium transport exist. For electrogenic transport, NH4+ may be carried alone (uniport) or together with an H+ (NH4+-H+ symport). Alternatively, NH4+ may be recruited, deprotonated in the periplasmic vestibule, and carried as NH3 together with an H+ (NH3-H+ symport). In case of electroneutral transport, the possibilities include NH3 uniport following direct recruitment of NH3 from the external medium, NH4+ recruitment and deprotonation before NH3 transport through the pore, or NH4+/H+ antiport (fig. S1) (9).

In solution, NH4+ and NH3 present distinct N isotopic compositions owing to differences in the equilibrium constants for the two isotopologues 15N and 14N (16). NH4+ ions containing 14N are more easily deprotonated than those containing 15N, leading to an isotopic fractionation linked to an isotope effect, α, at 30°C, of 1.044 between NH4+ and NH3 (17). Thus, in equilibrium, NH3 is depleted for 15N by 44.3 milliUreys (mUr) relative to NH4+. Isotope fractionation occurs, in general, against the heavier isotope (for example, 15N) and may originate from a kinetic or a thermodynamic isotope effect. The kinetic effect relates to the difference in reaction rate constants, and the thermodynamic one relates to the equilibrium constants (16). Therefore, cells transporting NH3 or NH3 + H+ across their membranes will become 15N-depleted compared to the external ammonium source as long as the ammonium source is not limited. To date, this isotopic fractionation against 15N in the specific conversion of NH4+ into NH3 has not yet been used to address AMT-Mep-Rh–mediated transport processes.

Here, we used 15N isotope fractionation associated with pH-dependent NH4+ deprotonation, that is, equilibrium isotope effect of ammonium, to study the transport mechanisms of different ammonium-transporting proteins in living cells. This approach allowed us to assign NH4+ and NH3 as substrates for different types of ammonium-transporting proteins in a noninvasive way. Furthermore, we studied the presence/absence of a kinetic isotope effect linked to the protein transport as a means to determine the N species transported (NH4+ or NH3 + H+) by ammonium membrane transport proteins. These studies support the acquisition and subsequent dissociation of NH4+ into NH3 and H+ by AMT-Mep-Rh–type proteins by a kinetic N isotope effect as a common transport mechanism of this protein superfamily.


Yeast 15N depletion depends on the presence of ScMep proteins

We first investigated the influence of the three endogenous Mep ammonium transport proteins, ScMep1 to ScMep3, from Saccharomyces cerevisiae using Σ1278b as a wild-type strain grown under two different nutritional conditions. After 48 hours of growth, all tested yeast strains showed a similar optical density at 600 nm (OD600) when grown at high ammonium concentrations (Fig. 1A). However, at low ammonium, triple-mepΔ cells were not able to grow (Fig. 1B). Consistent with previous reports, the presence of at least one of the three ScMep proteins was essential for yeast growth at low ammonium concentrations (6, 18).

Fig. 1 Functional AMT-Mep–type proteins deplete cells in 15N.

Yeast growth (top) and natural N isotopic abundance (bottom) in yeast strains grown for 48 hours (A and B) or 20 hours (C) at pH 4.3 in the presence of 13 mM K+ and 76 mM ammonium (A) or 1 mM ammonium (B and C). Wild type, Σ1278b; wild type (ura3), 23344c; wild type (ura3) + pFL38, wild type (ura3) + empty plasmid; triple mepΔ, 31019b; triple mepΔ + pFL38, triple mepΔ + empty plasmid; ScMep1, triple mepΔ + ScMep1; ScMep2, triple mepΔ + ScMep2; ScMep2H194E, triple mepΔ + ScMep2H194E; ScMep3, triple mepΔ + ScMep3 yeast strains. Data are presented as means ± SE (n ≥ 3). Letters represent significant differences among yeast strains for a given ammonium concentration (P ≤ 0.05). *Significant differences between ammonium concentrations for a given yeast strain. δ15N of (NH4)2SO4, 0.0 to 0.5 mUr. a.u., arbitrary units.

We next used isotope-ratio mass spectrometry (IRMS) to determine 15N versus 14N isotope discrimination during ammonium uptake. When grown at 76 mM ammonium for 48 hours (Fig. 1A), only ScMep-containing yeast cells, either wild type or the triple mepΔ complemented with either ScMep1, ScMep2, or ScMep3, showed 15N depletion relative to the N source; that is, the cellular δ15N was more negative than the value for the N source; δ15N of (NH4)2SO4 ≈ 0 to 0.5 mUr. After growth at 1 mM ammonium for 48 hours, that is, OD600 > 0.3, pH 4.3 yeast cells did not show significant depletion of 15N compared to their N source (Fig. 1B), as they consumed almost all the ammonium available in the culture medium (fig. S2). However, the cell 15N depletion determined, at OD600 < 0.3, when N in the medium was not depleted (fig. S2) in relation to the N source (Fig. 1C), was similar to that observed for cells grown at 76 mM for 48 hours (Fig. 1A).

Cells expressing different types of ammonium-transporting proteins (that is, endogenous ScMep proteins; a tonoplast intrinsic protein, TaTIP2;1, from Triticum aestivum, known to transport NH3; and AtAKT1, a K+ permease from Arabidopsis thaliana, further capable of NH4+ transport) were used to evaluate the relationship between the mechanism of ammonium transport, the associated cell δ15N, and pH dependency across a 2.5 to 6.5 range (Fig. 2) (19, 20). For the particular case of TaTIP2;1, cells were grown with galactose as carbon source instead of glucose, as the corresponding gene is placed under the control of a Gal promoter. In addition, cells were grown at higher pH values (6.5 to 8.0) to achieve a better complementation of the triple-mepΔ cells’ growth defect by TaTIP2;1 (19). At high external ammonium concentrations (76 mM), cell growth was strongly dependent on the presence of ScMep and pH (table S1). ScMep-expressing cells grew similarly in the range of pH from 2.5 to 6.5 (Fig. 2A). In contrast, the growth of triple-mepΔ cells, thus deprived of ScMep proteins, and of triple-mepΔ cells expressing AtAKT1 or TaTIP2;1 was dependent on pH (Fig. 2, A and B). Both ScMep presence and pH also influenced the cell N content (table S1). In general, cells lacking Mep proteins showed lower N content (Fig. 2, C and D), suggesting that at least a part of cell N supply may be provided by Mep proteins under high ammonium conditions. In addition, the cell δ15N response to pH of the growth medium was differentially influenced by the pH range and the type of ammonium transport proteins (table S1 and Fig. 2, E and F). At pH below 6.5 (Fig. 2E), the strong dependence of cellular δ15N on the presence of ScMep and its independence on pH, with no interaction of both factors (table S1), may indicate that the Mep-associated fractionation process is independent of the external pH and, therefore, of the concentration of NH3 in the culture medium. However, triple-mepΔ cells, triple-mepΔ cells expressing AtAKT1 (Fig. 2E), and triple-mepΔ cells expressingTaTIP2;1 grown on galactose at pH from 6.5 to 8 (Fig. 2F) showed different pH responses. At pH below 6.5, in triple-mepΔ and triple-mepΔ AtAKT1-expressing cells, pH had a contrasting effect on growth and on N content but had no effect on N isotope signature (Fig. 2). Neither of these strains, mepΔ and triple-mepΔ AtAKT1, showed 15N depletion compared to the N source after ammonium acquisition, indicating that NH4+ was the main form transported under these conditions. The data further indicate that the increase of external NH3 with increasing pH did not significantly contribute to N acquisition in none of these cell strains. TaTIP2;1-expressing cells were able to grow at pH values of up to 8.0 (Fig. 2B), and their growth and N content increased as the external pH in the culture medium increased (Fig. 2, B and D). The δ15N values decreased as the biomass and N content increased with external pH and therefore with the increasing external NH3 concentration in the culture medium (fig. S3), reaching values of approximately −13 mUr, similar to those shown by ScMep-expressing cells (Fig. 2, E and F). This augmented 15N depletion with increasing pH is likely due to the thermodynamic isotope effect occurring when the chemical equilibrium NH4+/NH3 in solution is displaced toward NH3. This indicates that the increase of external NH3 derived from the increasing pH contributes significantly to N acquisition in TaTIP2;1-expressing cells, indicating that the substrate acquired by TaTIP2;1 is NH3, and not NH4+, in line with a previous report (19).

Fig. 2 Cellular δ15N derived from Mep-, TIP-, and AKT-type transporters show differential pH dependence.

Cell growth (OD600; A and B), N content (%; C and D), and cellular N isotopic composition (δ15N; E and F). in Mep-containing cells and triple-mepΔ and yeast cells expressing other ammonium transporters. Yeast cells were grown at pH 4.3 in the presence of 13 mM K+ and 76 mM ammonium. Yeast strains used in this study include wild-type (Σ1278b), triple mepΔ (31019b), wild-type (ura3) (23344c), triple mepΔ + ScMep2, and triple mepΔ + ScMep2H194E. Other putative carriers of NH4+ (such as AtAKT1 from A. thaliana, triple mepΔ + AtAKT1) or NH3 (such as TaTIP2;1 from T. aestivum, triple mepΔ + TaTIP2;1) that are not members of the AMT-Mep-Rh family were also tested and considered controls for different ammonium transport mechanisms. TaTIP2;1 from T. aestivum–containing cells and triple-mepΔ cells were grown on galactose (3%) at pH above 6.5 (B and D). Equations for regression curves displayed in (A): wild type: y = 0.21x − 14.69, R2 = 0.5755; wild type (ura3): y = 0.43x − 13.70, R2 = 0.8480; triple mepΔ (on glucose; pH 2.5 to 6.5): y = −0.2x − 1.87, R2 = 0.7711; triple mepΔ + ScMep2: y = 0.52x − 9.98, R2 = 0.5694; triple mepΔ + ScMep2H194E: y = 0.79x − 13.38, R2 = 0.6476; triple mepΔ + TaTIP2;1: y = −2.42x + 5.99, R2 = 0.8135; triple mepΔ (on galactose; pH 6.5 to 8): y = −0.04x − 0.70, R2 = 0.0531; triple mepΔ + AtAKT1: y = −0.21x − 0.61, R2 = 0.7516. δ15N of (NH4)2SO4, −1.0 to 0.5 mUr. Data are presented as means ± SE (A to D; n = 3).

We next addressed whether a point mutation in the pore of ScMep2 previously shown to alter the kinetics of substrate translocation could also influence the δ15N response. We used the ScMep2H194E protein, in which a conserved histidine (His) residue (H194E) in the pore region is replaced by glutamate (21). Consistent with previous studies, similar to ScMep2, ScMep2H194E restored the growth defect of triple-mepΔ cells grown on 1 mM ammonium (Figs. 1B and 2) (21). Thus, the ScMep2H194E variant maintained transport activity, confirming that the conserved His residue at position 194 is not indispensable for the transport function of ScMep2 (21). Although cells expressing the ScMep2 and ScMep2H194E showed similar δ15N responses to pH, cells expressing the mutated protein always exhibited lower δ15N values (Fig. 2E). These observations indicate that the protein structure influences the ability of ScMep2 to fractionate between 14N and 15N isotopes during ammonium transport.

Ammonium transport via ScMeps can enable intracellular alkalinization in metabolically starved cells

We used an experimental setup, complementary to the IRMS technique described above, that could enable one to evidence an NH3 translocation via ScMep proteins. Stopped-flow fluorescence spectroscopy allows one to follow changes in intracellular pH in response to NH3 or NH4+ uptake by monitoring fluorescence variations of the pH-sensitive probe 5-carboxyfluorescein diacetate (CFDA). To avoid a contribution of H+ extrusion systems in the interpretations, we used metabolically starved cells, glucose-depleted and adenosine 5′-triphosphate (ATP)–exhausted, formerly grown with 76 mM ammonium and loaded with CFDA. Metabolically starved cells were suddenly exposed to an external concentration of 25 mM NH4Cl (pH 8 and 23°C) before internal pH changes were recorded over time (Fig. 3A). We used these conditions of concentration and pH to apply an NH3 gradient toward cytosol (22). As indicated by an increase in the intensity of carboxyfluorescein (CF) fluorescence, the addition of NH4Cl to the external medium induced intracellular alkalinization, which indicated NH3 transport in all tested strains. All strains containing ScMep proteins showed an instantaneous increase in fluorescence intensity and a plateau (pH alkalinization) after exposure to external ammonium. Only the triple-mepΔ strain displayed a linear time-dependent increase in fluorescence intensity (Fig. 3A). Because the yeast cells were metabolically inactive, these observations indicate that changes in internal pH are caused by the internal consumption of H+ associated with NH3 entrance and were unlikely due to active extrusion of H+ from the cytoplasm. The weak alkalinization observed in the triple-mepΔ cells most likely reflected the diffusion of NH3 through the cell membrane at pH 8 and 23°C. A slow and passive diffusion of NH3 through the lipid bilayer, as observed in the triple-mepΔ strain, and an involvement of ScMep proteins in the transport of NH3 with subsequent cytoplasmic alkalinization were supported by calculating the activation energy (Ea) of NH3 transport. In the Arrhenius plot shown in Fig. 3B, diffusion of NH3 through the lipid bilayer mediated by the triple-mepΔ cells (Ea = 38.0 ± 2.5 kcal mol−1) exhibited a greater temperature dependence than transport occurring in the presence of ScMep proteins (Ea = 23.4 ± 1.3 kcal mol−1 for the wild-type cells and Ea = 8.76 ± 0.75, 12.43 ± 0.64, and 14.68 ± 1.35 kcal mol−1 for the triple-mepΔ cells expressing ScMep1, ScMep2, and ScMep3, respectively). These data indicate that the presence of ScMep proteins favors NH3 translocation in metabolically starved cells.

Fig. 3 Mep-type proteins mediate intracellular alkalinization upon ammonium uptake.

Stopped-flow fluorescence signals obtained for intact triple-mepΔ (31019b) cells transformed with YCpMep1, YCpMep2, or YCpMep3 were compared with the negative control (31019b) and wild-type strain (Σ1278b). (A) Intracellular alkalinization was indicated by an increase in the fluorescence intensity resulting from the addition of 25 mM NH4Cl (pH 8.0) to the external medium at t0 = 23°C. (B) Arrhenius plot displaying the ln(k) of the changes in intracellular pH resulting from the transport of NH3 as a function of the inverse of the temperature. Data are presented as means ± SE (n = 3).

15N-depletion during uptake is a general feature of AMT-Mep-Rh proteins

We used yeast triple-mepΔ cells to heterologously express ammonium permeases belonging to the AMT-Mep-Rh protein family from different organisms (bacteria, fungi, green algae, plants, and animals) to test whether cellular 15N depletion was only associated with yeast ScMep proteins or to a general feature of members of the AMT-Mep-Rh protein family. At high and low external ammonium supplies, growth of all AMT, Mep, or Rh protein–expressing cells was associated with 15N depletion (Fig. 4), indicating that 15N depletion is a general feature of members of AMT-Mep-Rh proteins. Again, expression of the potassium channel AtAKT1 from A. thaliana in the triple-mepΔ strain did not lead to 15N depletion at high ammonium concentrations (Fig. 4A). Our data (not shown) indicate that a concentration of at least 1 mM ammonium is required to observe the function of AtAKT1 in triple-mepΔ cells and that, under these conditions, growth is not associated with 15N depletion, indicating that NH4+ is transported. Expression of the tonoplast intrinsic protein TaTIP2;1 from T. aestivum (19) favored growth at high ammonium and was associated with cellular 15N depletion (Fig. 4A).

Fig. 4 Functional AMT-Rh protein superfamily depletes cells in 15N.

Isotopic N signature of triple-mepΔ (31019b) cells transformed with different AMT-Rh subfamily transporters from protista, eubacteria, plants, and animals. Yeast growth (top) and natural N isotope abundance (bottom) of different yeast strains grown at pH 4.3 in the presence of 13 mM K+ and 76 mM NH4+ (A) or 0.5 mM NH4+ (B). *TaTIP2;1-containing cells were grown at pH 7.0. AMT-Rh carrier-containing strains grown in the presence of 76 mM NH4+ were compared with their respective negative control strains (31019b + pDR196 and 31019b + p426MET25) and other putative carriers of NH4+ (such as AtAKT1 from A. thaliana) or NH3 (such as TaTIP2;1), which are not members of the AMT-Mep-Rh family, as controls. Control strains did not show growth in the presence of 0.5 mM NH4+. EcAmtB (AMT from Escherichia coli), triple mepΔ + EcAmtB; Cr-AMT1;1 (Amt1;1 from Chlamydomonas reinhardtii), triple mepΔ + CrAmt1;1; Cr-AMT1;2 (AMT1;2 from C. reinhardtii), triple mepΔ + CrAMT1;2; At-AMT1;1 (AMT1;1 from A. thaliana), triple mepΔ + AtAMT1;1; At-AMT1;2 (AMT1;2 from A. thaliana), triple mepΔ + AtAMT1;2; At-AMT1;3 (AMT1;3 from A. thaliana), triple mepΔ + AtAMT1;3; At-AMT2;1 (AMT2;1 from A. thaliana), triple mepΔ + AtAMT2;1; Hs-RhCG (RhCG from Homo sapiens), triple mepΔ + HsRhCG. Data are presented as means ± SE (n ≥ 3). Letters represent significant differences (P ≤ 0.05). δ15N of (NH4)2SO4, −0.8 to 0.1 mUr. nd, not determined.

Using natural δ15N to distinguish between different ammonium transport mechanisms in vivo

15N depletion in yeast cells depended on the expression of functional ScMep proteins and was observed at low and high ammonium concentrations (Fig. 1), suggesting that ScMeps were active under both conditions. At high ammonium concentrations, ammonium is also transported in a ScMep-independent manner, which was not associated with cellular 15N depletion (Fig. 5). The latter ammonium transport process could occur via nonspecific cation channels (NSCCs). However, the molecular nature of these NSCC proteins remains unknown (23). When transported through NSCC, ammonium tends to compete with K+ (4). We used a described inhibitor of the function of NSC1-type channels, hygromycin B (23), to distinguish different ammonium transport mechanisms, that is, ScMep-dependent and ScMep-independent mechanisms, and track the function of ScMep proteins at high ammonium concentrations. The no-interaction term ScMep × hygromycin B on cell growth supports the idea that hygromycin B does not interfere with ScMep (table S2). The wild-type and triple-mepΔ cells were grown in the presence of 13 mM K+ and 76 mM ammonium at pH 4.3, conditions that are commonly used in minimal yeast medium (Fig. 5) (4). The triple-mepΔ cells showed weaker growth associated with almost no 15N depletion (Figs. 1 and 5). However, in the presence of 300 μM hygromycin B, there was no growth of the triple-mepΔ cells (Fig. 5A), and cells were not able to accumulate N any longer (Fig. 5C). In contrast, the wild-type cells displayed reduced growth, along with a similar reduction in total N content (Fig. 5, A and C). Hygromycin B application to the wild-type strain preserved cellular 15N depletion (Fig. 5B). This indicates that at high external ammonium concentrations, only ScMep-mediated ammonium transport was associated with significant 15N depletion (Fig. 5B), whereas ammonium transport through NSC1-type channels showed no N isotope discrimination (Fig. 5 and fig. S1C, top).

Fig. 5 Inhibition of NSC1-type channels does not affect cellular 15N depletion.

Inhibition of NSC1-type channels with hygromycin B in wild-type (Σ1278b) and triple-mepΔ (31019b) cells grown in the presence of 13 mM K+ and 76 mM ammonium. (A) Yeast growth (OD600), (B) natural N isotopic abundance (in milliUreys), and (C) total N content (in micromoles) in wild-type and triple-mepΔ yeast grown under control conditions (black bars) or in the presence of 300 μM hygromycin B, the NSCC type 1 inhibitor (white bars), at the end of the growth period. Yeast cells were grown for 48 hours at 185 rpm, 30°C, and pH 4.3. Data are presented as means ± SE (n = 3). δ15N of (NH4)2SO4, 0.0 mUr.

Using natural δ15N to address K+/NH4+ interdependence, ammonium toxicity, and associated amino acid excretion

Following the identification of the two major functional components of ammonium transport at high substrate concentrations under the growth conditions tested, that is, NSC1-type channels and ScMep proteins (Fig. 5), we altered K+ availability to assess the relative contribution of each component to the total amount of N transported (4). Irrespective of the K+ concentration in the growth medium (1.3, 6, or 13 mM), wild-type cells were able to grow in the presence of any ammonium concentration (0.5 to 300 mM), whereas triple-mepΔ cells were able to grow only in the presence of ammonium concentrations greater than 5 mM (Fig. 6). As expected, decreasing the K+ concentration in the growth medium from 13 to 1.3 mM decreased growth. At low ammonium concentrations, the growth of wild-type cells, but not of triple-mepΔ cells, was associated with 15N depletion of the cells. Notably, cells became enriched with 15N when wild-type or triple-mepΔ cells were grown in the presence of high ammonium and low K+ concentrations (Fig. 6, C and G). This may be related to extrusion of N compounds enriched in 14N, such as amino acids (4, 24). We consistently detected higher amino acid concentrations in the external medium along with growth of both wild-type and triple-mepΔ cells at high ammonium and low K+ concentrations (Fig. 6, C and G, and fig. S4).

Fig. 6 External ammonium/potassium ratio affects cellular δ15N.

Tracking δ15N values in the presence of different external ammonium and potassium concentrations in S. cerevisiae (A to D) wild-type (Σ1278b; black) and (E to H) triple-mepΔ (31019b; red) strains. (A and E) Yeast growth (OD600). (B and F) Yeast δ15N (in milliUreys). (C and G) Amino acid concentrations detected in the media after yeast growth (μmol g−1 DW). (D and H) Yeast N content (%). Yeast cells were grown for 48 hours at 185 rpm, 30°C, and pH 4.3. Data are presented as means ± SE (n = 3). δ15N of (NH4)2SO4, 0.0 to 0.044 mUr. DW, dry weight.

At each K+ concentration, cell 15N depletion tended to increase with rising external ammonium until cell growth started to decrease, at which point 15N depletion also started to decrease, reflecting ammonium toxicity (Fig. 6 and table S3). The interaction term between both factors, K+ and NH4+ availabilities, on cell growth, isotopic composition, extracellular excretion of amino acids, and cell N content shows the strong dependence of these parameters on the ratio of external K+/NH4+ concentrations (table S3). Our data illustrate how measurements of δ15N can enable one to address K+/NH4+ interdependence during growth, ammonium toxicity, and its associated amino acid excretion, the latter being inferred by cells’ 15N enrichment at high, toxic, ammonium concentrations.


Here, we show that the use of the natural N isotopic signature allows one to distinguish different functional ammonium transport mechanisms in vivo in yeast cells. At low ammonium concentrations (<5 mM, pH 4.3), the main proteins responsible for ammonium uptake are AMT-Mep-Rh–type proteins, whose N transport mechanism is associated with cellular 15N depletion relative to the N source. However, at higher ammonium concentrations (>5 mM, pH 4.3), further transport systems contribute to and dominate ammonium uptake, which are not related to cellular 15N depletion and likely reflect the activity of NSCCs (23). Because the δ15N value of a total organism, like a yeast cell, is composed of the δ15N values of a wide range of organic and inorganic compounds with quite different isotope characteristics including residual nonreduced N, amino acids, or nucleotides (25), the low level of 15N depletion that is observed in the triple-mepΔ cells exposed to ammonium (Figs. 1 to 6) makes these cells a true negative control for studying ScMep-related ammonium transport mechanisms.

The slightly negative δ15N values displayed by the triple-mepΔ strain when it was exposed to high ammonium (Figs. 1 and 5) may be the combined outcome of NH3 diffusion through the lipid bilayer and a metabolic effect. However, the almost constant and highly negative δ15N values displayed by ScMep-expressing yeast cells grown in the presence of low (1 mM) and high (76 mM) ammonium concentrations in this study may be considered an independent event from the inherent interfering and overlapping fractionation processes linked to N metabolism. When fractionation processes linked to N metabolism interfere with the total cellular biomass δ15N value, the δ15N value changes in a manner dependent on factors like K+ availability and extracellular excretion of 15N-depleted organic compounds, such as amino acids (Fig. 6 and fig. S4; see below).

15N depletion supports deprotonation during transport through AMT-Mep-Rh proteins in yeast cells

In solution, NH4+ and NH3 are in equilibrium. Because of the equilibrium isotope effect on the deprotonation of NH4+ into NH3 + H+, the heavier isotope is discriminated at a rate of ≈−44.3 mUr (17). At high NH4+ concentrations, when at least one of the ScMep proteins was expressed in yeast, cells became 15N-depleted relative to the N source (Fig. 1). Thus, NH3 is involved in the ammonium transport process mediated by ScMep proteins. The depletion of 15N in yeast cells due to ammonium transport through ScMep proteins may be due to the following mechanisms (fig. S1): (i) the recruitment of NH3 in solution, (ii) preferential recruitment of 14NH4+ versus 15NH4+ in solution followed by nondiscriminating NH4+ transport, or (iii) recruitment of NH4+ followed by deprotonation of NH4+ into NH3 and H+ before transport through the pore.

Recruitment of NH3 in solution. If ScMep proteins recruit NH3 in solution, then the cellular δ15N value is expected to decrease with increasing pH because the NH3 concentration increases exponentially with pH (fig. S3). However, ScMep-expressing cells showed an opposite trend, as cellular δ15N values increased with increasing pH (Fig. 2). NH3 is therefore not likely to be the chemical species recruited by ScMeps, which is consistent with the mechanism suggested for EcAmtB and other members of the AMT-Mep-Rh family (11, 12, 15, 26). The thermodynamic isotope effect linked to NH4+/NH3 equilibrium in the external solution appears not to be the origin of the 15N depletion of Mep-expressing yeast cells. In contrast, a slight decrease in δ15N values was observed in the triple-mepΔ cells, which passively permeate NH3 through the lipid bilayer (Figs. 2 and 3), while a larger decrease was observed in TaTIP2;1-expressing yeast cells (Fig. 2). TaTIP2;1, whose ability to transport ammonium is mainly driven by the transmembrane NH3 concentration gradient, directly recruits and transports NH3 from the external solution (Fig. 2) (9, 19).

Preferential recruitment of 14NH4+ versus 15NH4+ in solution. On the basis of the crystal structure of EcAmtB and modeling of substrate–amino acid interactions, it has been proposed that EcAmtB recruits NH4+ rather than NH3 at its outer vestibule (15). Moreover, the predominance of NH4+ over NH3 in solution would render NH4+ recruitment by ScMep proteins more efficient. Because of its lower mass, 14N is more energetic than the 15N isotope and has a higher probability of reacting (27); therefore, 14NH4+ may putatively be recruited faster than 15NH4+ in the external vestibule of ScMep proteins. However, when a kinetic isotope effect occurs, mass transfer or bonding changes must occur during the process or reaction (16). For this reason, the simple preferential recruitment of 14NH4+ versus 15NH4+ in solution by Mep-type proteins is not expected to be the major factor determining the 15N depletion in Mep-expressing cells. The NH4+ recruited must experience changes in its molecule to give an isotopic fractionation linked to a kinetic effect. In this way, this fractionation must be related to the transformation, that is, dissociation of the molecule during the transport process through the protein.

Recruitment of NH4+ followed by deprotonation. When grown on ammonium, 15N depletion may be related to a kinetic isotope effect associated with different dissociation rates of both isotopologues (16). The assays performed with the triple-mepΔ cells expressing ScMep2H194E provided support for this assumption because they showed that (i) the conserved His residue at position H194 was not indispensable for the transport function of ScMep2 (21) or fractionation mechanism (Figs. 1 and 2) and (ii) the structure of the molecule was involved in the isotope fractionation process (Figs. 1 and 2). These results are compatible with the hypothesis that ScMep2 recruits NH4+, which is then deprotonated in the core of the ScMep molecule (21, 26). The H194E mutation has been shown to modify the transport kinetics of ScMep2, leading to a reduced affinity and an increased substrate transport rate (21). The stronger 15N depletion observed in ScMep2H194E-expressing cells compared to cells expressing ScMep2 indicated that the protein structure is involved in the isotopic fractionation of ScMep2, a kinetic isotope effect, and therefore in the NH4+ deprotonation event (Fig. 1).

The stopped-flow experiments we designed with metabolically starved cells suggest that ScMeps are capable of translocating NH3 (Fig. 3), further supporting the NH4+ deprotonation model. However, the observed cytoplasmic alkalization associated with NH4+ deprotonation does not exclude the fact that the transport may be electrogenic (cotransport of NH3 and H+) in metabolically active cells, as has previously been shown for plant AMT1-type proteins (9, 13). The yeast cells were metabolically starved in our experimental setup. Moreover, it has been proposed that the timing of H+ and NH3 transport via AMT-Mep proteins through the membrane may be different (28). Therefore, while the 15N isotope fractionation evidences the ammonium deprotonation process, electrophysiology studies are required to elucidate whether ScMeps perform electrogenic or electroneutral transport.

Several studies have reported the regulated inhibition, at the transcriptional level and/or at the transport activity level, of ammonium transport through AMT-Mep-Rh proteins by elevated ammonium concentrations in yeast (18, 29, 30), bacteria (31), and plants (32). However, our data indicate that some activity of AMT-Mep-Rh proteins, either endogenous or heterologous, is conserved at high ammonium concentrations in yeast cells. Regulatory mechanisms present in the original cells most likely do not function when plant and animal proteins are heterologously expressed in yeast. In the case of the wild-type yeast strain (Fig. 1), ScMeps, or at least one of the three, were not totally inhibited, showing an activity responsible for the discrimination observed in the presence of 76 mM ammonium, and the inhibition of the NSC1-type channels with hygromycin B did not affect δ15N values (Fig. 5). In agreement with these results, Feller et al. (33) showed a reduction in the growth of triple-mepΔ cells at 20 mM and even at 80 mM ammonium in relation to that shown by wild-type cells grown on yeast nitrogen base medium, which is also visible in this study (Figs. 2 and 6). The ScMep2-mediated transport at low pH, below 4, appears sufficient to ensure growth at 76 mM ammonium (Fig. 2). Marini et al. (18), using a specific minimal buffered medium with different N sources, showed that all three MEP genes are submitted to N catabolite repression and thus repressed when cells were provided with a high ammonium concentration compared to cells grown on nonpreferred N sources such as proline. However, MEP expression was not totally shut down on high ammonium. LacZ data showed that all three MEP genes are still expressed, although at reduced levels, on high ammonium concentrations and that the maximal repression was only achieved when glutamine was combined with ammonium (18). Our results support the idea that the reduced level of ScMep proteins produced at high ammonium concentrations contributes to ammonium transport and growth under these conditions.

δ15N allows differentiating between different ammonium transport mechanisms in vivo and highlights excretion of amino acids (15N-depleted compounds) in yeasts under ammonium toxicity

Excess uptake of ammonium causes toxicity in plants, animals, and microorganisms (1, 4, 7). The similarities between the physicochemical properties of K+ and NH4+ in solution in terms of charge, radius, and hydration energies support the hypothesis that large amounts of NH4+ might enter the cells via NSCCs, particularly when external K+ concentrations are low (4, 23). In addition, experiments in yeast with different combinations of external NH4+/K+ ratios showed a nexus between concentrations of NH4+ and K+ in the growth medium (table S3). For instance, for each considered external K+ concentration, cellular δ15N values, growth, and amino acids in the culture media highlighted the fact that yeast cells experiencing ammonium toxicity excreted amino acids (Fig. 6 and fig. S4). Some important metabolic processes and reactions, such as primary N assimilation by glutamine synthetase, amidation, and/or transamination, discriminate against the heavier N isotope, leading to 15N depletion from organic compound products and the subsequent enrichment of 15N in the remaining unassimilated substrates (24). Similar responses to K+/NH4+ imbalances have also been observed in plants (20) and animals (1) and account for cellular 15N enrichment at low K+/NH4+ ratios.


Here, on the basis of N isotope discrimination associated with ammonium transport by ScMeps at distinct pH, we showed that (i) NH4+, and not NH3, is most likely the chemical species recruited by ScMep proteins and (ii) the transport of NH4+ by AMT-Mep-Rh proteins is associated with NH4+ deprotonation, which confers specificity for NH4+ transport. In view of the physical-chemical similarities between NH4+ and K+, the NH4+ deprotonation process may be a mechanistic feature conferring selectivity against K+ and for NH3 transport after NH4+ recruitment. If deprotonation occurring inside the AMT-Mep-Rh protein is the step providing selectivity for NH4+ transport, then mutations in the pore that allow the transport of other ions such as K+ may be identified. A recent study by Hall and Yan (34) reveals that mutations in the twin His residues in the EcAmtB central pore allow the K+ translocation against a concentration gradient. Notably, human Rh-associated glycoprotein variants associated with overhydrated spherostomatocytosis were reported to mediate K+ transport (35). The link between NH4+ deprotonation within AMT-Mep-Rh proteins and their specificity for NH4+ transport may be a central common feature in this large, widespread family of transporters present across all life domains. We propose that this deprotonation step is independent of the net transport of NH3 (electroneutral transport) or NH3 + H+ (electrogenic transport) through the protein (fig. S1B). The natural N isotope fractionation method presented here will become complementary to net charge movement assessment and 3D structure to achieve a precise characterization of the mechanism of ammonium transport for each considered ammonium membrane transport protein. These findings provide new insights into the controversy surrounding ammonium transport systems and mechanisms, leading to a unifying hypothesis for ammonium transport in living organisms.


Strains and growth conditions

The S. cerevisiae strains used in this study were all isogenic with the wild-type Σ1278b, with the exception of the mutations and transformations specified throughout the text (plasmids used in this study are listed in table S4). All cultures were started with an initial inoculum of ≈104 cells ml−1. Cells were grown in minimal medium [CaCl2 2H2O (0.1 g liter−1), NaCl (0.1 g liter−1), MgSO4 7H2O (0.5 g liter−1), boric acid (0.5 mg liter−1), CuSO4 5H2O (0.04 mg liter−1), KI (0.1 mg liter−1), FeCl3 6H2O (0.2 mg liter−1), MnSO4 H2O (0.4 mg liter−1), sodium molybdate 2H2O (0.2 mg liter−1), biotin (10−3 mg liter−1), calcium pantothenate (0.2 mg liter−1), folic acid (10−3 mg liter−1), inositol (1 mg liter−1), nicotinic acid (0.2 mg liter−1), p-aminobenzoic acid (0.1 mg liter−1), pyridoxine HCl (0.2 mg liter−1), riboflavin (0.1 mg liter−1), thiamine HCl (0.2 mg liter−1), and ZnSO4 (0.4 mg liter−1) (pH 4.3)] with 0.5% glucose as the carbon source (4). Adjustments to the pH in some experiments (with NaOH) are indicated in the text and figure captions. Nitrogen and potassium sources were added to this minimal medium as specified in the text and figures. Various concentrations of (NH4)2SO4 were added as a nitrogen source (ranging from 0.5 to 300; concentrations specified for each experiment/figure). The K+ concentrations (1.3, 6, and 13 mM K+) used in the experiments were prepared from a stock solution of 1.3 M K+ [KCl (98 g liter−1) and KH2PO4 (2 g liter−1)] (4). For tests requiring NSC1-type channel blockers, hygromycin B was added to the minimal medium at 300 μM (23). The TaTIP2;1-containing yeast strain with a Gal promoter was grown in the presence of galactose (3%) as the C source instead of glucose (0.5%). In this case, because of its scarce complementation at pH values below 6.5, the culture media for the TaTIP2;1-containing yeast strain with a Gal promoter were adjusted to higher pH values (6.5, 7, and 8) (19). The triple-mepΔ yeast strain (31019b) was also grown on galactose 3% as the C source instead of glucose (0.5%) in this experiment (pH 6.5, 7, and 8). Thus, the final culture media formulations were sterilized by filtration using a sterile Corning 250-ml vacuum filter/storage bottle system with a 0.22-μm pore and polyethersulfone membrane (Corning Incorporated) to avoid unwanted reactions during the thermal sterilization process (autoclave). The triple mutant strain used in this study, 31019b (triple mepΔ), is deficient in ammonium permeases (yeast ScMep proteins) and was therefore expected to be the ideal strain for analyzing the transport of NH3/NH4+ mediated by heterologously expressed ammonium carriers (18).

Isotopic N composition, isotope effect estimation, and N content

Yeast cells were grown under the specified conditions for each experiment (see the text and figure captions), then collected by centrifugation (for 15 min at 5000g and 20°C), and washed at least twice with autoclaved Milli-Q water. After centrifugation (for 15 min at 5000g and 20°C), the pellets were frozen at −80°C and freeze-dried before the isotope determinations. Freeze-dried material was also weighed to record the biomass. The supernatants from yeast cells collected in some experiments were frozen at −80°C until further analysis (see the “Determination of amino acid contents in the external culture medium” section).

Approximately 0.5 to 1 mg of freeze-dried yeast material from each sample was separately packed in tin capsules. The 15N/14N ratios of the samples were determined by continuous-flow IRMS (36) using an Isoprime (GV Instruments) stable IRMS coupled to a EuroEA (EuroVector) elemental analyzer for online sample preparation via Dumas combustion. The standards used in this study were IAEA-N1 and USGS-35, and δ15N results were normalized to air. The precision of the isotope-ratio analysis, which was calculated using values obtained from six to nine replicates from laboratory standard materials interspersed among samples from every batch analysis, was ≤0.2 mUr. The results of the analysis of the N isotope composition are expressed as δ15N (in milliUreys) normalized to atmospheric N2 and were calculated using the following equation: δ15N (in Ureys) = (Rsample/Rstandard) − 1, where Rsample is the 15N/14N ratio of the sample and Rstandard is the 15N/14N ratio of atmospheric N2.

This approach has the following advantages over other previously used methods: (i) It integrates the N taken up over time in vivo, not only over a short timeframe (seconds), which is used by most electrophysiological approaches conducted in vitro (not during the growth period); (ii) it allows the amounts of N taken up as NH3 (or NH3/H+) or NH4+ to be estimated when both forms are taken up simultaneously; (iii) it is a direct determination of the N concentration (not indirectly, as in measures of parallel events such as pH change or uptake of an analog, that is, methylammonium); and (iv) it is not intrusive. However, the two main limitations are as follows: (i) Unlimited and steady-state ammonium and potassium conditions are required to obtain clear evidence of NH4+ deprotonation, and (ii) additional and complementary techniques, such as net charge movement assessment and 3D-structure, are required to discern whether NH3 or NH3 + H+ is being transported.

Stopped-flow fluorescence assays

Stopped-flow assays were based on the methods reported by Bertl and Kaldenhoff (22), with some modifications. Briefly, cells were grown in the minimal medium described above in the presence of 76 mM ammonium and 13 mM K+ on an orbital shaker at 185 rpm and 30°C up to an OD600 < 1. Yeast cells were harvested by centrifugation (for 15 min at 5000g and 20°C), washed (×2), resuspended in loading buffer [50 mM Hepes and NaOH (pH 5.7); wet weight, 3 ml g−1], and incubated for at least 1 hour on ice (to ensure complete glucose depletion and ATP exhaustion, which represents metabolic starvation). Cells were then preloaded (15 min at 30°C) with the membrane-permeable nonfluorescent dye CFDA (1 mM), which was hydrolyzed after entering the cell, releasing the membrane-impermeable fluorescent form CF. Finally, 1 ml of this cell suspension was centrifuged (for 10 min at 5000g and 4°C), resuspended in 10 ml of ice-cold incubation buffer [50 mM NaCl and 10 mM tris-HCl (pH 8)], and maintained on ice until assay.

Transport experiments were performed in a stopped-flow spectrophotometer (Hi-TECH Scientific PQ/SF-53) at a controlled temperature with 2 ms of dead time using a 470-nm excitation filter and a >530-nm cutoff filter. Experiments were performed at various temperatures. For each experimental condition, four to six runs were usually stored and analyzed. Intact yeast cells equilibrated in incubation buffer were mixed with an equal volume of test solution [50 mM NH4Cl and 10 mM tris-HCl (pH 8)] with the same osmolarity (preventing volume changes due to osmotic fluxes that could interfere with the signal) but a different ionic composition (50 mM NH4Cl instead of 50 mM NaCl) to measure NH3/NH4+ transport, resulting in an inwardly directed 25 mM NH3/NH4+ gradient. NH4+ or NH3 uptake induces a change in the intracellular pH that was monitored by measuring fluorescence intensity over time (22). The kinetic rate constant of the process (k) was estimated by fitting the signals to a single exponential function. The Ea of transport was evaluated from the slope of the Arrhenius plot [ln(k) as a function of 1/T, where T is the temperature in kelvin] multiplied by the gas constant R. Three stopped-flow experiments were performed independently, and the results obtained from all of them were coincident with those shown in Fig. 3. For each experiment, yeast cells grown from three independent batches were blended. Each determination was replicated at least three times.

Determination of amino acid contents in the external culture medium

Following yeast growth, the culture medium was centrifuged (for 15 min at 5000g and 20°C) and then filtered (Ø = 20 μm). The soluble amino acid contents were then determined by high-performance capillary electrophoresis using a Beckman Coulter PA-800 apparatus (Beckman Coulter Inc.) with a laser-induced fluorescence detector (488 nm) equipped with a fused silica capillary (diameter, 50 μm; length, 43/53.2 cm) in an electrophoresis buffer containing 50 mM borax and 45 mM α-cyclodextrin at pH 9.2. Analyses were performed at 30 kV and 20°C. Known concentrations of norvaline and homoglutamic acid were added to all samples as internal references. The filtered supernatants were diluted (1:5) with 20 mM borate buffer (pH 10) and derivatized before detection with 1 mM fluorescein isothiocyanate in acetone. Gly + Ser contents were detected as a single peak.

Statistical analyses

All statistical analyses were performed with IBM SPSS for Windows (IBM). According to the number of independent factors considered for each experiment, one-way, two-way, or three-way analysis of variance was performed (see figures and tables for further information). F value in combination with P value was used to determine significant effects of factors (independent variables) on the parameters considered (dependent variables). Whole-model R2 and interaction term effects between independent factors on each parameter have been presented in all tables of analysis of variance. All statistical analyses were conducted at a significance level of 5% (P ≤ 0.05) and under normality and homoscedasticity assumption. In post hoc analysis tests, least significant difference statistics was used.


Supplementary material for this article is available at

Fig. S1. Ammonium transport mechanisms (fluxes) associated with charge movement.

Fig. S2. Growth and N content of the distinct yeast strains.

Fig. S3. Relative concentrations of NH4+ and NH3 in solution expressed as a percentage of the total ammonium concentration (NH4+ + NH3) as a function of solution pH.

Fig. S4. Amino acids excreted into the medium by wild-type (black bars) and triple-mepΔ (red bars) cells expressed as a percentage of the total amino acid pool (shown in Fig. 6).

Table S1. Two-way analysis of variance of ScMep/NH3 transporter presence and pH.

Table S2. Analysis of variance of ScMep presence and hygromycin B application.

Table S3. Three-way analysis of variance of ScMep presence and external potassium and ammonium availabilities.

Table S4. Plasmids used in this study.

References (3747)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank A. Rodríguez-Navarro (Universidad Politécnica de Madrid, Spain) for providing the plasmid pFL61 AtAKT1 and G. Garijo [Universidad Pública de Navarra (UPNa)] and J. Melo [Faculdade de Ciências da Universidade de Lisboa (FCUL)] for technical assistance. Stable isotope ratio analyses were performed at the Stable Isotopes and Instrumental Analysis Facility of the Centre for Ecology, Evolution and Environmental Changes (cE3c), University of Lisbon–Portugal. Funding: I.A. was supported by a postdoctoral fellowship from the Government of Navarra, Spain (Anabasid outgoing Programme, 2011) and by a postdoctoral fellowship from the Portuguese Fundação para a Ciência e a Tecnologia (SFRH/BPD/90436/2012). A.M.M. is a senior research associate of the Belgian Fonds de la Recherche Scientifique Fonds de la Recherche Scientifique–FNRS (grants CDR J017617F, PDR T011515F, and ARC) and a WELBIO investigator, and M.B. is a scientific research worker supported by WELBIO. This work was also developed in the context of the following projects: PTDC/BIA-BEC/099323/2008 and PTDC/AGR-PRO/115888/2009 to cE3c and FCUL, UID/DTP/04138/2013 to iMed.ULisboa, and AGL2015-64582-C3-1-R and AGL2012-37815-C05-05 to UPNa. Author contributions: I.A., C.C., and P.M.A.-T. conceived and designed all the experiments. A.M.M., M.B., and N.v.W. contributed to design growth conditions and pH experiments. I.A. and C.G. performed experiments. M.B. performed yeast transformations (except to CrAMTs) and helped to analyze results of pH experiments. E.S.-L. performed yeast transformations with CrAMTs. A.P.M., G.S., I.A., and P.M.A.-T. performed stopped-flow experiments. A.P.M. and G.S. analyzed data and contributed to writing the part of stopped-flow experiments. I.A. analyzed the data of the rest of the experiments. I.A., C.C., P.M.A.-T., A.M.M., and N.v.W. wrote the paper with helpful contribution from M.B., E.F., and E.S.-L. I.A. performed statistical analysis. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article