Introduction

Urea serves as a vital nitrogen source for plant growth and can be utilized by plants through multiple pathways1. In soils, urea is maintained at a relatively low level under natural conditions because of the presence of urease, which hydrolyzes urea into ammonia and carbonic acid2. Nevertheless, plants can uptake urea from soil through high-affinity transport3,4. During leaf senescence, urea is transported back to the phloem and reused as a nitrogen source for further growth3. Moreover, urea is extensively used as a fertilizer to boost crop growth in the modern agricultural industry and is considered a rapidly available nitrogen source for plant growth5.

The transport of urea is the central process of urea metabolism in plants, where DUR3, a high-affinity urea transporter, plays a dominant role. Phylogenetic analysis demonstrated that DUR3 is widely distributed in monocots and eudicots, such as Arabidopsis, and in the most common crops, including maize, rice, wheat, and soybean6. DUR3 was originally identified to mediate high-affinity urea transport in yeast7, and later this phenomenon was also confirmed in Arabidopsis (KM ~ 4 μM)4, rice8, and maize9,10. A previous study on rice DUR3 highlighted its contribution to rice production, particularly under nitrogen-deficient and field conditions, underscoring the importance of unraveling its molecular mechanism8.

Investigations of DUR3 from Arabidopsis thaliana (AtDUR3) revealed that this transporter is expressed mainly on the plasma membrane, in agreement with its physiological function in directly mediating urea uptake from the environment2. Secondary structure prediction and sequence alignment analysis further suggested that DUR3 contains 15 transmembrane helices and resembles a LeuT-like transporter11. This structure is completely different from the structural characterization of human urea transporters (UT-A, UT-B)12. A urea uptake assay in oocytes revealed that high-affinity urea transport by DUR3 is energized by a proton gradient and operates independently of sodium4. This is distinct from many LeuT-like transporters, which utilize the electrochemical gradient stored in sodium ions13,14,15. Understanding the molecular mechanism of DUR3, including the urea recognition mechanism and the proton-driven transport cycle, is highly warranted.

In this study, we express and purify recombinant Arabidopsis DUR3 and conduct cryo-EM analysis. The structures of Arabidopsis DUR3 in the inward-facing apo state (DUR3apo) and occluded urea-bound state (DUR3urea) are resolved. Our results elucidate the molecular basis of high-affinity urea transport by DUR3, providing a deeper understanding of proton-driven urea transport and offering insights into the efficient utilization of urea by plants.

Results

Structural determination and architecture of Arabidopsis DUR3

To verify the urea transport capacity of AtDUR3, we cloned the full-length AtDUR3 gene from an Arabidopsis cDNA library and inserted it into a pHXT426 vector, as described by Liu et al.4. The vector was expressed in the Saccharomyces cerevisiae dur3 mutant YNVW1 (Δura3, Δdur3), which has defects in urea absorption and cannot grow on a medium with less than 5 mM urea as the only nitrogen source4. The heterologous expression of AtDUR3 in YNVW1 enhanced its growth in 2 mM urea medium compared to the negative control strain, although it remained notably weaker than the wild-type strain 23346c (Fig. 1a, b, Supplementary Fig. 1a). This indicates that AtDUR3 facilitates urea transport across the plasma membrane. In comparison, expression of the Saccharomyces cerevisiae DUR3 gene (ScDUR3) in the YNVW1 strain resulted in growth comparable to that of the wild-type positive control strain 23346c.

Fig. 1: Structure and topology of DUR3apo.
Fig. 1: Structure and topology of DUR3apo.The alternative text for this image may have been generated using AI.
Full size image

a, b Growth complementation of urea uptake-defective yeast strain (YNVW1) by expressing AtDUR3 and ScDUR3. YNVW1 strain was assessed by transforming the strain with p426 alone or p426 harboring AtDUR3 and ScDUR3. Yeast growth was evaluated on medium containing 2 mM urea as the sole nitrogen source at pH 5. The bars represent the mean ± s.e.m (n = 3), biological independent samples. Two-tailed unpaired t-test, ****P < 0.0001. c Cryo-EM density map of the DUR3apo. Two protomers are colored grey and blue, respectively. Lipid molecules are shown in yellow. TM helices and the horizontal dimension of DUR3apo dimer are labeled. d The atomic model of a DUR3 protomer in the ligand-free state (DUR3apo), viewed in the membrane plane. TM helices, the N-terminus, and the C-terminus of DUR3apo are labeled. The dimensions of the DUR3 protomer are labeled.

To elucidate the molecular mechanisms of DUR3, we cloned the full-length AtDUR3 gene and conducted protein expression using HEK 293-F cells. The recombinant DUR3 was purified using n-dodecyl-β-D-maltoside (DDM). ThermoFluor analysis16 indicated that replacing DDM with lauryl maltose neopentyl glycol (LMNG) increased thermal stability (Supplementary Fig. 1b). This substitution resulted in a purified protein sample that exhibited a monodispersed peak upon size-exclusion chromatography (Supplementary Fig. 1d–g). This suggests that the purified DUR3 may possess good biochemical characteristics for structural determination. We then prepared cryo-EM samples in the absence of urea and performed cryo-EM analysis.

Cryo-EM data were collected on a 300-kV Titan Krios G4. Two-dimensional classification demonstrated that DUR3 particles are arranged within rod-shaped detergent micelles, and two DUR3 protomers can be visualized. More detailed features of the structural elements, such as the continuous density of transmembrane (TM) helices, are also discernible. Iterative 3D classifications, including heterogeneous refinements and 3D classifications without particle alignment, were subsequently performed to improve the map quality (Supplementary Fig. 2). Final non-uniform (NU) refinement gave rise to a cryo-EM map with resolution at 2.8 Å according to the golden standard Fourier shell correlation (GSFSC) at the threshold of 0.143 (Supplementary Fig. 2d). This cryo-EM map displayed clear structural features, such as densities representing backbone atoms, side chains, and lipid molecules, enabling us to build an atomic model. Urea, the substrate of DUR3, was not supplemented during sample preparation. Hereafter, we termed this complex DUR3apo.

The DUR3apo structure is formed with two DUR3 protomers arranged in parallel with a two-fold symmetry, with dimensions of 100 Å × 50 Å × 75 Å (Fig. 1c). Each DUR3 protomer consists of 15 TM helices (TM0–TM14), with the N-terminus starting from the extracellular side and the C-terminus located in the cytosol (Fig. 1d). In agreement with the findings of previous phylogenetic analysis, DUR3 adopts a LeuT-like fold with canonical structural components4,6. This includes discontinuous TM1 and TM6 helices, along with their flanking partners, TM2 and TM7 helices, which may play a dynamic role during conformational transitions. The scaffold helices, namely, the long and tilted helices TM3 and TM8 and the short vertical helices TM4 and TM9, are tightly packed and presumably remain highly stable during conformational transitions. Another TM helix, TM10, is also unwounded in the middle, leading to an approximately 60° counterclockwise rotation in its intracellular half (Fig. 1d). This deformation indicates that TM10 might also undergo conformational changes during the transport cycle. Other helices, such as TM0 and TM11–TM14, are positioned at the outer surface of the transporter. The structure of DUR3apo also revealed the structure of the extracellular linkers EL1–EL3 (N–TM0, TM5–TM6, and TM7–TM8) and the intracellular linkers IL1 (TM2–TM3) and IL2 (TM12–TM13). In particular, two short helices, EL3Ha and EL3Hb, were identified within EL3; a helix, EL2H, was identified within EL2; and a short helix, IL1H, was identified within IL1. A disulfide bond (C264–C275) was identified in the extracellular structure of DUR3. The cysteine residues involved in forming this disulfide bond are conserved across different species of DUR3, potentially contributing to structural integrity and stability (Supplementary Fig. 4)17.

The structure we analyzed demonstrated a homodimer interface mediated by cholesteryl hemisuccinate (CHS) and lipids, with CHS added throughout the protein purification process (Fig. 1c). The dimer formation is primarily achieved through extensive hydrophobic interactions. At the extracellular interface, we identified five additional densities around the dimer interface, attributed to CHS and other hydrophobic lipid molecules (Supplementary Fig. 6a). These CHS and lipid molecules bind tightly to TM6a and TM13, forming the core of the near extracellular dimer interface. Specifically, residues R282 from TM6 (R282TM6), R283TM6, L290TM6 and L612TM13 interact directly with these lipids, establishing lipid-mediated interfaces (Supplementary Fig. 6b, c). No additional hydrophobic interactions were observed at the intracellular interface. Similar dimerization events have been reported in other structural studies of LeuT-like transporters18,19,20,21, but these transporters exhibit distinct homodimer interfaces, implying that dimerization may not be functionally necessary (Supplementary Fig. 7). Consequently, further structural analyses were conducted on a single DUR3 protomer in this study.

Urea binding site

To understand the binding mechanism of urea, we were attempting to add urea to the protein purification process. However, high concentrations of urea can denature proteins by disrupting the internal non-covalent bonds that maintain their tertiary structure22,23. To determine the appropriate urea concentration, we used high-performance liquid chromatography (HPLC) to analyze the effect of various urea concentrations (0-40 mM) on the DUR3 protein (Supplementary Fig. 1c). Based on the final yield of DUR3, we found that over 60% of the protein remained stable at 400 µM urea. Given that 400 µM is 100 times the KM value, we selected this concentration for preparing the cryo-EM samples. As a result, we obtained a cryo-EM map at 3.0 Å according to the GSFSC criterion. (Supplementary Fig. 3). Compared to the density map of DUR3apo, a globular density was identified exclusively in the central pocket of the DUR3 structure in the presence of 400 μM urea (Fig. 2b–d). This density is located in the central cavity of the DUR3, close to TM1, TM2, TM6, TM7 and TM10, and matches the size of urea molecule (Fig. 2a, Supplementary Fig. 8a, b). Therefore, this structure was termed DUR3urea for further analysis.

Fig. 2: Structure and the urea binding site of DUR3urea.
Fig. 2: Structure and the urea binding site of DUR3urea.The alternative text for this image may have been generated using AI.
Full size image

a The atomic model of a DUR3 protomer in the urea-bound state (DUR3urea). b, c The urea molecule and its adjacent residues. DUR3apo (b) and DUR3urea (c) are shown as blue and red cartoons, overlaid with their corresponding cryo-EM density shown in blue mesh. The urea molecule is shown as yellow sticks. A water molecule is shown as red sphere. Residues interacting with the urea molecule are shown as sticks. The urea molecule, water molecule, and adjacent residues are labeled. Two alternative conformations are modeled for Y315 in the structure of DUR3urea and are labeled as Y315A and Y315B. d, e Binding site of the urea molecule in the structure of DUR3urea. The urea molecule and critical residues within the binding pocket or stabilizing the urea molecule are shown as sticks and labeled. Putative hydrogen bonds are shown as black dotted lines. f, g Yeast cell-based analysis of urea transport in the YNVW1 strain, investigates the effects of mutations in the urea binding site on the efficiency of urea transport through mutational analysis of yeast growth. The bars represent the mean ± s.e.m (n = 3), biological independent samples. Two-tailed unpaired t-test, ****P < 0.0001.

The urea molecule is stabilized by multiple side-chain interactions within the central pocket of DUR3urea. Critical residues include W95TM1 and W338TM7, secure the urea molecule in the center between them (Fig. 2e). Additionally, in the cryo-EM map of DUR3urea, Y315 is located at the intracellular entrance of the central pocket (Fig. 2b). This residue exhibited a branched side-chain density, likely due to rotamer variations of its bulky side chain. Therefore, we modeled two alternative rotamers (Y315A and Y315B) into the DUR3urea model at this site (Fig. 2c). Another hydrophilic residue, Q124, is positioned beneath the urea molecule (Fig. 2c, e). Notable differences were observed between the cryo-EM maps of DUR3urea and DUR3apo at this binding site: Q124 showed a regular density typical of glutamine in DUR3apo, whereas, in DUR3urea, the side-chain density of Q124 was elongated, extending significantly beyond the most distal atoms, Nε and Oε (Fig. 2b, c). We speculate that the excess density represents a water molecule forming hydrogen bonds with the urea molecule and bridging it with the surrounding hydrophilic residues, Y315B and Q124.

We subsequently conducted yeast dilution assays to validate the function of critical residues within the urea-binding pocket. Given the low expression and weaker growth complementation function of AtDUR3, we selected the homologous ScDUR3 to facilitate functional verification of mutants. Specifically, we substituted key tryptophan dyad residues (W95, W338) along with other conserved interacting amino acids, with alanine, resulting in the following mutations: W93A, W95A, T98A, Y117A, Q124A, N305A, Y315A, W338A and Y461A. For clarity, residue numbering is based on AtDUR3 (Supplementary Fig. 5). We found that most mutants, including W95A, Y117A, Y315A, W338A and Y461A almost completely lost their ability to transport urea, resulting in significantly impaired growth of the yeast strains (Fig. 2f, g). The remaining amino acid residue mutants (W93A, T98A, Q124A, and N305A) had less impact on transport function (Fig. 2, g, Supplementary Fig. 8c). Urea contains three polar atoms (two nitrogen atoms and one oxygen atom) capable of forming extensive hydrogen bonds with surrounding residues. To elucidate the specific interactions between urea and DUR3, we conducted molecular dynamics (MD) simulations of the DUR3urea. We performed hydrogen bond analysis on the simulated trajectories to investigate these interactions further, calculating the probabilities of different hydrogen bond formations. These results further support hydrogen bond interactions with urea in DUR3urea, involving the main chain oxygen atoms of W93 and the side chain oxygen atoms of T98, Y117, Q124, N305 and Y461 (Supplementary Fig. 9a, b). Moreover, binding free energy predictions suggest that residues Y117, W93, and T98, which have the highest probabilities of forming hydrogen bonds with urea, contribute most significantly to the binding free energy (Supplementary Fig. 9c).

Transition between the apo and urea-bound states

The presence of urea molecules slightly changes the conformational state of DUR3. Structural analysis revealed that the central pocket of DUR3urea is occluded from both sides of the transporter. Several hydrophobic residues, including A97, Y117, P342 and Y461 at the extracellular site, and W93, W95 and Y315 at the intracellular side, stabilize the urea binding site within the central pocket and prevent access from both sides. (Fig. 3a–d). These observations suggest that the transporter is captured in an occluded conformation in the DUR3urea structure. In contrast, the central pocket of DUR3apo is connected to the cytosol via a short tunnel, allowing cytosolic solvent to access the urea binding site. This accessibility indicates that DUR3apo is determined in an inward-facing conformation.

Fig. 3: Structural comparison and conformational transition.
Fig. 3: Structural comparison and conformational transition.The alternative text for this image may have been generated using AI.
Full size image

a, b The solvent accessibility of the urea binding site in the structure of DUR3urea (a) and DUR3apo (b). The urea molecule is shown as spheres, and its adjacent residues are shown as sticks. c, d Structural comparison of the urea binding site between DUR3apo (blue) and DUR3urea (red), shown in two perpendicular views from the membrane plane. The urea molecule and residues interacting with urea are shown as sticks and labeled. Y315 exhibited notable conformational shifts between the two states and are indicated using black dots and arrows. The two alternative conformations of Y315, Y315A ([A]) and Y315B ([B]), are labeled. Putative hydrogen bonds are shown as black dotted lines.

To gain insight of the conformational differences between the two DUR3 structures, we conducted further structural comparison. Most residues at the urea binding site, including T98, W338, and Y461, exhibited a stable conformation between the two states (Fig. 3c). Nevertheless, structural discrepancy was observed on two critical residues, W95 and Y315. Y315 is located near the short tunnel that connects the urea binding site and cytosol. In the structure of DUR3urea, the side chains of Y315 rotate towards the pocket center by 12° and 40° relative to the DUR3apo structure, exhibiting alternative rotamers Y315A and Y315B. The side chain of Y315B, specifically, is oriented to seal the central pocket in the structure and coordinated by a water molecule (Fig. 3d). Meanwhile, the side chain of Y315A showed relatively minor displacement and may not occlude the central pocket by itself. Additionally, in the DUR3urea structure, the side chain of W95TM1 adopts a slightly altered conformation, which interacts with and stabilizes the urea molecule. We also observed that TM10 bends due to the presence of P474, causing the middle of TM10 to unwind near the intracellular region and contributing to the closing of the extracellular cavity (Fig. 4c). Yeast dilution growth assays with the P473A mutation revealed a significant reduction in urea uptake (Fig. 4d, e).

Fig. 4: The potential proton-binding sites and proposed transport mechanism of DUR3.
Fig. 4: The potential proton-binding sites and proposed transport mechanism of DUR3.The alternative text for this image may have been generated using AI.
Full size image

a Titratable residues (aspartates, glutamates and histidines) throughout a DUR3 protomer. Aspartates and glutamates are shown as red spheres, and histidines are shown as pink spheres. D312 and E402 are positioned around the central pocket and labeled. b Structural comparison between DUR3 (red) and vSGLT (green). Na2 sites of vSGLT are shown as spheres. The residue H222 of DUR3 is depicted as sticks and labeled. His222TM5 is located between TM1 and TM8. c Key sites for conformational transformation, residue P474TM4 are shown as sticks. d, e Mutational Analysis of Critical Residues involved in proton-binding and conformational transitions. The bars represent the mean ± s.e.m (n = 3), biological independent samples. Two-tailed unpaired t-test, ****P < 0.0001. f Proposed Proton-Driven Transport Mechanism of DUR3. In the outward-facing conformation, urea enters the binding site with Asp312 and Glu402 protonated. The protonation of Asp312 and Glu402 then promotes isomerization to an inward-facing conformation, where the urea is released. The schematic illustrates three distinct states of DUR3: outward-open, occluded, and inward-open. Transmembrane helices TM1, TM6, TM8 and TM10 are depicted as cylinders, highlighted in red, blue, gray, and brown, respectively. The residues D312TM6, Y315TM6, E402TM8 and P474TM10 are shown as sticks. Urea molecules are represented by blue triangles, and hydrogen ions are shown as purple circles.

Key titratable residues in the urea transport cycle

Furthermore, to unravel the molecular basis of the proton-driven transport cycle of DUR3, we conducted further investigation to search for putative protonation sites that energized the transporter during state transition. In general, the binding sites for protons in transporters are often titratable residues24. Therefore, we screened all titratable residues, namely, aspartates, glutamates, and histidines, through the entire structure to identify possible candidates. It turns out that most titratable residues are located in the extracellular or intracellular loops, as well as in the outer helices. These residues seemed unlikely to be the protonation site responsible for proton transport because they cannot be alternatively exposed to both sides of the membrane. However, near the geometric center of the transporter, just diagonally below the urea binding site and close to the cytoplasmic side, we noticed two titratable residues: D312TM6 and E402TM8 (Fig. 4a). In the structures of other proton-driven APC transporters, such as Nramp (H232TM6), GkApcT (D237TM6), and KimA (E233TM6), a conserved acidic residue is commonly found on TM6 and is considered a key site for protonation. Additionally, a second protonatable residue is typically located near this acidic residue, as observed in Nramp (D131TM3), GkApcT (E115TM3), and KimA (D117TM3). In DUR3, the corresponding secondary protonatable residue is E402 on TM8 (Supplementary Fig. 10)25,26,27. D312TM6 and E402TM8 are highly conserved in DUR3 homologous transporters, and the D312N mutant results in a near-complete loss of protein function (Fig. 4d, e). The TM6 helix, where D312 resides, consists of two half-helices that are crucial for substrate transport in LeuT-fold transporters, underscoring an important role of D312 in proton coupling. Mutation of E402 leads to a partial loss of function, indicating that this residue is also likely to be involved in protonation. This hypothesis is further supported by in silico estimations of residue-wise pKa values, which highlighted the noteworthy high pKa values for D312 (5.5) and E402 (9.4).

A previous study on the mechanism of the sodium-independent amino acid transporter ApcT suggested that a lysine residue on TM5 serves as the proton binding site28. Similarly, the DUR3 protein has an equivalent histidine (H222) on TM5, with its imidazole group occupying the same position as the Na2 site in LeuT29 (Fig. 4b, Supplementary Fig. 11). The predicted pKa of H222TM5 in DUR3 is estimated to be 4.0, indicating that the side chain’s imidazole group readily undergoes deprotonation. We introduced two point mutations, H222L and H222N. The H222L mutant resulted in a complete loss of transport activity and minimal growth in yeast strains, while the H222N mutation exhibited reduced transport activity, though it still lower than that of the wild type (Fig. 4d, Supplementary Fig. 8c). These findings suggest that while H222TM5 is not essential for proton-coupled transport in DUR3, it is likely to play a crucial role in facilitating conformational changes during the transport cycle.

Discussion

Despite the ubiquitous presence of urease in soil matrices, AtDUR3 can transport urea into root cells even at low external concentrations (KM ~ 4 μM). In the resolved DUR3urea structure, we found that several residues, particularly the side chains of aromatic residues such as the tryptophan dyad (W95TM1 and W338TM7), play a crucial role in recognizing the substrate urea. To date, the structures and transport mechanisms of APC superfamily transporters from bacteria, plants, and animals have been extensively studied. These transporters share similar transmembrane domains and conserved folding topologies, known as the LeuT fold, and utilize the common rocking-bundle model30. In nearly all structures of proton-driven APC transporters, proton-titratable amino acid residues can be identified in the central cavity (Supplementary Fig. 10b–d). In DUR3, such residues are likely to be the D312-E402 pair, with the two residues in close proximity and a distance of 3.8 Å between the nearest atoms of their side chains (Supplementary Fig. 10a). This pair is the only titratable acidic residue found within the cavity of the structure.

Based on the provided data and knowledge from other APC transporters, we propose a model for proton-coupled urea transport through DUR3: When the protein is in the outward-facing conformation, extracellular urea binds with high affinity at low concentrations. Asp312TM6 (pKa 5.5) and Glu402 TM8 (pKa 8.0) likely participate in proton binding and undergo protonation (Fig. 4f). Similar to proton-driven MFS proteins31, proton binding generates a structural cue at which the physiological negative-inside membrane potential of the cytoplasmic membrane exerts an electrostatic force on the protonated DUR3, promoting a conformational transition to the inward-facing state. The bending of TM10 aids in closing the extracellular pocket. As seen in the DUR3urea structure, the urea molecule remains enclosed towards the intracellular side by a thin intracellular gate formed by the side chain of Y315. Upon full opening of the inward-facing conformation, the hydrogen bond interactions between Y315 and the urea, mediated by water molecules, are disrupted, exposing the binding site to the cytoplasm and allowing the release of the urea molecule.

However, we have not yet resolved the detailed structure of DUR3 in the outward-facing conformation, making it difficult to determine the helical arrangement and conformational changes of TM6 and TM8, and the relevant amino acid residues in their protonated and deprotonated states through structural comparison. In the growth complementation, substituting AtDUR3 with ScDUR3 for point mutation complementation may be influenced by regulatory genes specific to different expression systems, potentially influencing the yeast growth and should be interpreted with this in mind. Furthermore, the functional complementation in yeast cannot accurately reflect the effects of point mutations on transport reaction kinetics, such as KM and kcat. Additionally, we have not directly addressed the coupling mechanism between DUR3 and protons during the urea binding process. It remains unclear whether protons are transferred between protonatable amino acid residues during protein conformational changes. These questions warrant further structural and mechanistic studies.

Methods

Expression and purification of the Arabidopsis DUR3

The gene of full-length Arabidopsis DUR3 (UniProt ID: F4KD71) was amplified from Arabidopsis complementary DNA (cDNA). The DUR3 gene was subcloned into a modified pEG-BacMam vector with a twin-strep tag at the C-terminus, with a PreScission Protease (PPase) cleavage site and mCherry between them. The DUR3 protein was expressed in HEK293F cells using the Bac-to-Bac system (Invitrogen, USA) at 37 °C with 5% CO2. The recombinant baculoviruses of P1 and P2 were produced in sf9 cells (Gibco, USA). The HEK293F cells were infected with 1% (v/v) P2 viruses and 1% (v/v) fetal bovine serum (FBS) when cultured at a density of 2.8 × 106 cells/mL32. After 12 h, sodium butyrate was added into the medium, 48 h was needed before harvesting. Cells were collected by centrifugation at 1640 × g for 3 min at 4 °C, then frozen in liquid nitrogen and stored at −80 °C.

The cell pellets were resuspended and lysed using a Dounce homogenizer in buffer A [20 mM HEPES (pH 7.4), 150 mM NaCl, 5 mM β-mercaptoethanol (β-ME)] containing protease inhibitors (0.5 mg/mL pepstatin A, 1.4 mg/mL leupeptin and 2 mg/mL aprotinin). The cell membrane was centrifuged at 100,000 × g for 1 h and resuspended again using buffer A. Pellets were solubilized in buffer B (buffer A supplemented with 1% (w/v) n-dodecyl-β-D-maltoside (DDM, Anatrace), 0.15% (w/v) cholesteryl hemisuccinate (CHS, Anatrace) at 4 °C for 2 h with rotation. 1 mM ATP and 5 mM MgCl2 were needed to remove associated heat shock proteins. Pellets were solubilized at 4 °C for 2 h with rotation in buffer B [buffer A supplemented with 1% (w/v) n-dodecyl-β-D-maltoside (DDM, Anatrace), 0.15% (w/v) CHS (Anatrace)]. Associated heat shock proteins were removed with 1 mM ATP and 5 mM MgCl2. The insoluble cell debris was removed by centrifugation at 100,000 × g for 45 min, followed by filtration through a 0.45 μm filter (Millipore, USA). The supernatant was loaded to the Streptactin Beads (Smart-Lifesciences) via gravity flow at 4 °C. The beads were washed with 10-column volumes of buffer C [buffer A supplemented with 0.025% (w/v) DDM]. Subsequently, the DUR3 protein was eluted using 10 mL elution buffer (buffer C additionally containing 5 mM desthiobiotin). The eluate was incubated with PPase for 3 h and then rerun on a Ni-NTA column to adsorb the His-tagged PPase. The flowthrough sample was concentrated to 1 mL using a 100-kDa cut-off Centricon (Millipore, USA) and further purified via size-exclusion chromatography using Superose-6 Increase 10/300GL column (GE Healthcare, USA) with a running buffer containing 20 mM HEPES (pH 7.4), 150 mM NaCl, 5 mM β-ME, and 0.002% LMNG (Anatrace, USA). To determine the urea-bound structure, 400 μM urea (Sigma) was added throughout the entire protein purification process. The elution fractions were analyzed by high performance liquid chromatography (HPLC). The peak fractions between 13.5 and 14.5 mL were pooled, concentrated to 15 mg/mL, and prepared for cryo-EM grids.

Thermal stability assay

Peak fractions from the size-exclusion chromatography were collected, concentrated, and diluted to a final concentration of 0.2 mg/mL in a buffer containing 20 mM HEPES, 150 mM NaCl, and one of three detergents: 0.05% DDM, 0.05% LMNG or 0.05% GDN. A stock solution of N-[4-7-(dimethylamino-4-methyl-3-coumarinyl) phenyl] maleimide (CPM), a thiol-reactive fluorochrome, at 4 mg/mL was diluted to 10 µg/mL33. Samples were incubated at 4 °C for 10 min. Thermal stability data was then collected using a quantitative PCR instrument (Rotor-Gene 6600, Qiagen), programmed to measure dissolution curves from 25 °C to 95 °C, with a temperature increase of 1 °C per minute. Data analysis was performed using GraphPad Prism software.

Cryo-EM sample preparation and data acquisition

For the preparation of apo or urea-bound DUR3 cryo-EM samples, a 2.7 μL droplet of purified protein was applied to glow-discharged holey carbon grids (Cu R1.2/1.3 300 mesh, Quantifoil), which were discharged in H2-O2 condition for 60 s using the Solarus plasma cleaner (Gatan, USA). The grids were then automatically blotted for 5 s at 4 °C under 100% humidity with a Vitrobot Mark IV (Thermo Fisher Scientific, USA) before being plunged into liquid ethane. The grids were loaded into a 300 kV Titan Krios G4 (Thermo Fisher Scientific, UAS) equipped with a K3 Summit direct electron detector (Gatan, USA) and a GIF Quantum LS energy filter (Gatan, USA) with a 20 eV slit width. Images were recorded using EPU software at a calibrated magnification of ×105,000 in the super-resolution mode (0.85 Å pixel size). The nominal defocus range was set from –1.0 to –2.0 μm. Each movie stack was acquired with an exposure time of 4.16 s and dose-fractioned into 32 frames, with a total dose of 60 e−/Å2.

Data processing and model building

The flowcharts for processing apo and urea-bound data are shown in Supplementary Figs. 2 and 3. For DUR3apo and DUR3urea, 718 and 2,705 movie stacks were collected, respectively. Patch motion correction was performed using MotionCor2 and patch contrast transfer function (CTF) estimation was carried out with Gctf 234. All image processing steps were conducted using cryoSPARC35. From the apo-DUR3 dataset, 482,365 particles were picked using the Blob Picker and 2D Class. These particles were extracted and classified through ab initio 3D reconstruction into 4 classes. One class, displaying continuous transmembrane helices and comprising 40.4% of the particles, was re-extracted and subjected to another round of 3D classification without alignment into three classes. Finally, Finally, a map with a resolution of 2.8 Å, determined by the GSFSC criterion, was generated using Non-uniform (NU) refinement (C2) with 103,979 particles. For the DUR3urea dataset, more micrographs were collected due to the influence of urea, but the data processing was similar. A total of 229,911 particles were picked using both blob and template pickers, followed by 2D classifications and ab initio 3D reconstruction. One of the initial models served as a good reference map for heterogenous refinements. The final map, with an overall resolution of 3.0 Å, was generated using Non-uniform (NU) refinement (C2) with 83,751 particles.

The cryo-EM map of DUR3, with a resolution of 2.8 Å, allowed for clear visualization of the side chains of the transmembrane helix. The initial model predicted using AlphaFold236, was docked into the map as a rigid body using UCSF Chimera37. The model was then manually adjusted and rebuilt using Coot438. For DUR3urea, urea was manually added in Coot, and the hydrocarbon chain of the lipid was also generated in Coot. To generate a dimeric DUR3 model, another DUR3 molecule was placed based on double symmetry. The DUR3 model was refined against the cryo-EM map in real space using PHENIX539. Statistics of cryo-EM data collection, refinement, and model validation are shown in Supplementary Table 1. All figures were prepared using PyMOL (Schrödinger) or UCSF ChimeraX.

Expression of DUR3 in S. cerevisiae and functional complementation in yeast

The DUR3 coding sequence with a C-terminal FLAG tag was cloned into the pHXT426 vector, which carries a hexose transporter promoter. Site mutants of the DUR3 construct were induced by overlap extension PCR (Supplementary Table 2). The wild type and mutant DUR3 contracts were transformed into the Saccharomyces cerevisiae strains YNVW1 (Δdur3, Δura3), which has the DUR3 gene deleted and is defective in urea uptake4. The plasmids were introduced using the PEG/LiAc method. Transformants were initially selected on SD agar (Oxid) medium (uracil-deficient yeast nitrogen-base, Lablead) containing 20 mM NH4+ as the sole N source. Single colonies from each transformation were selected for a growth complementation assay using 2 mM urea as the sole nitrogen source. The medium was adjusted to pH 5.6 with 1 M KOH. Cells were resuspended and diluted fivefold in distilled water. Aliquots of 7 μL were spotted onto AP plates and incubated at 30 °C for 3–4 days. For liquid culture, YNVW1 yeast strains carrying the p426-ScDUR3 vector with point mutations were inoculated into YNB medium supplemented with 2 mM urea at an initial optical density (OD600) of 0.1. The cultures were incubated with shaking, and OD600 values were measured after 24 hours.

Western blot analysis

Wild-type AtDUR3 and ScDUR3 were subcloned into the pHXT426 vector with a Flag tag fused to the C-terminus of DUR3. Yeast cultures (2 mL) were grown using the same cultivation method described earlier. After 24 h, cells were harvested by centrifugation at 1500 × g. The cell pellet was resuspended in water, centrifuged, and the supernatant was discarded. To prepare the samples, a buffer containing 0.04 M EDTA and 0.14 M β-mercaptoethanol was added to the pellet, followed by incubation at 30 °C for 30 min. After centrifugation, the pellet was dissolved in a solution of 0.2 g/mL snail enzyme (prepared in 16 mM sodium citrate, 1 M sorbitol and pH 5.8 potassium phosphate buffer) and incubated at 37 °C for 1 h. Yeast cells with removed cell walls were collected by centrifugation. Membrane proteins were solubilized by incubating the cells in 1% DDM buffer (20 mM Tris, 150 mM NaCl, pH 7.5) for 2 h, followed by the addition of 20% sample loading buffer. The extracted proteins were resolved on a 10% SDS-PAGE gel and transferred onto a PVDF membrane (GE Healthcare, USA). The membrane was blocked in skim milk for 1 hour, then incubated for 5 h with primary antibodies in TBST buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl and 0.1% Tween 20). The primary antibodies used were anti-Flag (Bioworld, AP0007M, 1:10,000 dilution) and anti-β-tubulin (Abcam, ab184970, 1:10,000 dilution). After primary antibody incubation, the membrane was washed five times with TBST and incubated for 1 h with horseradish peroxidase (HRP)-conjugated secondary antibodies: goat anti-mouse (CW0102S, CoWin Biotech, 1:10,000 dilution) and goat anti-rabbit (CW0103S, CoWin Biotech, 1:10,000 dilution). Detection was performed using Western Lightning Plus-ECL blotting reagents (Tanon, China), and images were captured accordingly.

Molecular dynamics simulations

We analyzed the DUR3urea structure and observed the density of urea molecules, using this structure as the initial configuration for MD simulations. The all-atom simulated system was prepared using the CHARMM-GUI interface, consisting of a phospholipid bilayer membrane with proteins, POPCs, and ligand urea molecules (Supplementary Table 3). The Amber ff14SB force field was applied to proteins and lipids, the TIP3P model to water molecules, and GAFF2 parameters to ligands40,41. The system included 150 mM NaCl. Energy minimization was performed using the steepest descent method. Pre-equilibration steps followed the default CHARMM-GUI protocols. Three independent 100 ns simulations were conducted following equilibration. The Verlet leapfrog algorithm with a 2 fs time step was used for motion equation integration. Simulations were run under an NPT ensemble at a constant temperature of 303.15 K and a pressure of 1 bar42,43. Temperature control was managed by the Nose-Hoover thermostat, and pressure control by the Parrinello-Rahman barostat. Periodic boundary conditions were applied. Coulomb interactions were calculated with a 0.9 nm truncation length for the Lennard–Jones potential and the real-space part of the Ewald summation. The particle mesh Ewald (PME) method was used for the Fourier space part of the Ewald summation. All simulations were performed using Gromacs 2021.6. MM/GBSA analysis was conducted with the gmx_MMPBSA tool developed by Valdés-Tresanco et al.44,45, MM/GBSA analysis was performed on the final 75 ns of each of the three trajectories.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.