Main

The ocean is our planet’s largest biome, where pelagic microbial Thaumarchaeota (syn. Nitrososphaerota) represent one of the most abundant organisms3. The numerical dominance of marine thaumarchaea suggests that they have a major role in global biogeochemical cycles1,2. N. maritimus, an intensely studied marine thaumarchaeon, grows chemolithoautotrophically by aerobically oxidizing ammonia to nitrite1. This organism has also been shown to regenerate oxygen under anoxic conditions, and to fix CO24,5, placing it in an important position in the global nitrogen and carbon dioxide biogeochemical cycles.

Owing to the low concentration of ammonia (NH3) or ammonium (NH4+) ions in the oceans, reported in the 10−8 to 10−9 M range6, marine archaea such as N. maritimus have evolved specialized molecular machinery to attract ammonium ions on their cell surface to facilitate ammonium oxidation at the cell membrane7. N. maritimus, like most archaeal cells, is encased by a paracrystalline, proteinaceous surface layer or S-layer8,9,10,11. Bulk modelling of the N. maritimus S-layer has predicted that this cellular organelle might have a key role in elevating the ammonium concentrations in the pseudoperiplasmic space between the S-layer and the cell membrane12,13; however, how this occurs on a mechanistic and molecular level is unclear. At the overall morphological scale, the N. maritimus S-layer has been reported to have a hexagonal arrangement9,12 and is postulated to consist of repeating subunits of the proteins Nmar_1547 or Nmar_1201 (two proteins with 91% sequence identity), based on transcriptomic and proteomic data14,15. Here, to understand the rules governing ammonium binding and enrichment by N. maritimus and related marine Thaumarchaeota, we investigated the molecular structure of the N. maritimus S-layer using structural, biochemical, cellular and bioinformatic methods.

Cryo-ET analysis of the N. maritimus S-layer

To gain insights into this problem, we used electron cryotomography (cryo-ET) and subtomogram averaging techniques. We have previously applied these methods to determine in situ structures of prokaryotic S-layers from in vitro purified specimens containing cellular fragments16,17,18. Our goal was to solve the structure of the N. maritimus S-layer directly from whole cells. Cryo-ET analysis of N. maritimus cells revealed an S-layer surrounding the approximately 300-nm-wide elongated cells, which contained a compact nucleoid and several cytosolic ribosomes (Fig. 1a). Using subtomogram averaging, we mapped the locations of the S-layer repeating units on the cell, which were arranged in a hexagonal array (Fig. 1b). We found that the S-layer hexamers coated the cells with near-perfect continuity along their lengths, while local pentameric defects were present on the cell edges, completing the S-layer (Fig. 1b), which was confirmed by quantification of hexamer and pentamer positions relative to the centre of the cell (Fig. 1c).

Fig. 1: The molecular structure and assembly of the N. maritimus S-layer in intact cells.
figure 1

a, A denoised36,37 tomographic slice through a N. maritimus cell shows ultrastructural details of this marine archaeon (annotated). Inset: top and side views of the subtomogram average of the S-layer. Scale bars, 500 Å (main image) and 100 Å (inset). Cellular tomography was performed at least 27 times (Extended Data Table 1). b, Map of the subtomogram positions in the cellular S-layer showing the presence of pentameric defects at the edge of the cell. Scale bar, 500 Å. c, A histogram of subtomogram positions from all tomograms relative to the three-dimensional centre of the cell body (n = 41,303 hexamers (grey) and n = 203 pentamers (pink); one out of two biological replicates shown). d, The subtomogram averaging map enables derivation of a molecular model directly from cellular data. Amino acid residue side chains resolved are marked. See also Extended Data Figs. 1 and 2. e, A ribbon model of the cashew-shaped NmSLP monomer is shown in two orthogonal views. f,g, The structure of the S-layer hexamer displayed in two orthogonal views shows that NmSLP monomers are arranged as an array of Ig-like domains; each domain (D) is coloured differently (a schematic is shown below). The first eight Ig-like domains are resolved in the cryo-ET and subtomogram averaging map.

Source Data

Full size image

Next, we used state-of-the-art cryo-ET imaging and image-processing workflows, which have been shown to support high-resolution in situ structure determination from purified specimens19. As a result, we produced a high-resolution map of the N. maritimus S-layer hexamer from intact cells (Fig. 1d, Extended Data Fig. 1, Extended Data Table 1 and Supplementary Videos 1 and 2). The central region of the S-layer hexamer in the map had a resolution of 3.3 Å, with the resolution decaying to around 4.5 Å towards the periphery (Extended Data Fig. 1). The subtomogram averaging map contained sufficient details to enable us to derive an atomic model of the S-layer (Fig. 1d–g and Extended Data Fig. 2). The structure shows that the S-layer is pseudohexagonal (Fig. 1f,g) and consists of the repeated interactions of the Nmar_1547 (hereafter, NmSLP) S-layer protein. Despite the high sequence similarity of NmSLP to the other previously predicted SLP, Nmar_1201, a unique segment of NmSLP between residues 911 and 977 was clearly resolved in our map. This enabled us to identify NmSLP as the primary N. maritimus SLP on cells through direct structure determination (Extended Data Fig. 2). This observation confirms previous transcriptomic data showing high expression levels of the Nmar_1547 gene compared with Nmar_120114. However, we cannot rule out that Nmar_1201 could be present at lower copy numbers.

At the sequence level, NmSLP is arranged into ten immunoglobulin-like (Ig-like) domains (Fig. 1f,g). The first eight domains were well resolved in our 3.3–4.5-Å-resolution subtomogram averaging map. By contrast, the last two domains appeared less ordered, with the local resolution decaying towards the C terminus of NmSLP, away from the centre of the S-layer hexamer (Extended Data Figs. 1 and 2). At the N terminus, the C6 symmetry of the hexamer is broken, revealing a distinctly two-fold symmetric central pore (Extended Data Fig. 2). Each monomer of NmSLP in the S-layer adopts a rough ‘cashew’ shape, facilitating several interactions around the hexamer of NmSLP (Fig. 1e–g). Domain 4, nestled within domain 3 and linked by short connectors, is slightly raised relative to the base of the cashew-shaped Ig-array of NmSLP, projecting towards the extracellular milieu.

Although the amino acid residues (37–1499) from the first eight Ig-like domains could be unambiguously built into the subtomogram averaging map, several unexplained densities were observed, emanating from surface-exposed asparagine residues (Extended Data Fig. 2). Given that archaeal SLPs are known to be heavily glycosylated20, we hypothesized that these densities might correspond to glycans. Another notable set of unexplained densities was observed near the negatively charged amino acid residue side chains of Asp73 and Glu74 at the central C2 pore, as well as between NmSLP monomers around the hexamer (Extended Data Fig. 2). Considering that positively charged ions have been previously observed bound to S-layers21,22, we hypothesized that these additional densities on the N. maritimus S-layer could potentially correspond to bound cations, although the resolution of the map prevented us from unambiguously assigning their chemical identities.

Cryo-EM shows a porous S-layer

The N. maritimus S-layer has previously been predicted through bulk biophysical modelling to help attract ammonium ions12. Considering the tight sheath formed by the S-layer around cells (Fig. 1b,c), providing an extremely large surface area for interaction with the marine environment, we postulated that the S-layer functions as a negatively charged ammonium trap. This would facilitate the movement of cations such as ammonium towards the cell membrane, specifically to the sites of ammonium oxidation. To test these hypotheses with higher resolution structures, in which ion and other densities would be better resolved, we purified N. maritimus cell envelopes for cryo-electron microscopy (cryo-EM) analysis. We then used single-particle techniques, as applied previously to two-dimensional S-layer sheets for structure determination18,23, to solve a 2.7-Å-resolution structure of the S-layer (Fig. 2a, Extended Data Fig. 3 and Extended Data Table 2).

Fig. 2: Cryo-EM structure of isolated N. maritimus S-layer sheets.
figure 2

a, In vitro cryo-EM structure (global resolution, 2.7 Å) of isolated S-layer sheets from N. maritimus. The colour scheme for ribbon diagrams is the same as in Fig. 1; domains of one NmSLP are marked. b–j, Magnified views of the pores lined with negatively charged residues, which are ubiquitous in the S-layer sheet. The location of the pores is given in the titles of the panels. k, The sharpened38 cryo-EM map shows 17 glycans decorating each NmSLP; the map shown at a lower contour level in two different orientations. A schematic of the glycan locations on the NmSLP sequence is shown below.

Full size image

The single-particle structure was very similar (root mean squared deviation of 2.21 Å for the full composite model, 0.54 Å for residues 35–455 (refined in C2) and 1.39 Å for residues 466–1498 (refined in C6)) to the subtomogram averaging structure (Fig. 2, Extended Data Fig. 4 and Supplementary Video 3), enabling us to extend our structure by modelling the ninth Ig-like domain, reaching up to residue 1616 out of 1734 (Fig. 2a and Extended Data Fig. 3). The last Ig-like domain remains unresolved in our map, with only disordered, diffuse density observed beyond the ninth Ig-like domain of NmSLP in the pseudoperiplasmic space, indicating flexibility relative to the rigid part of the S-layer. The NmSLP hexamer in the single-particle structure appears to be slightly expanded compared with the subtomogram averaging structure (Supplementary Video 3), perhaps due to differences in the S-layer curvature.

The N. maritimus genome also encodes a homologue for a cell-anchoring SlaB protein (Extended Data Fig. 5a,b and Supplementary Table 1) that is known to bind the S-layer of the archaeon Sulfolobus acidocaldarius to the cell membrane24. However, proteomic data on N. maritimus indicate that this protein is considerably less abundant in the cell than NmSLP15. We speculate that the last (tenth), unresolved Ig-like domain of NmSLP may have a role in anchoring the S-layer to molecules present on the cell envelope, therefore partially reducing the need for stoichiometric anchoring by a SlaB or a SlaB-like protein in N. maritimus.

In our single-particle structure, the subunit contacts between NmSLP hexamers (Fig. 2b–j) were better resolved compared with in our subtomogram averaging map (Fig. 1). Notably, several contact sites, both between hexamers and within each hexamer, contain pores lined predominantly by rows of negatively charged amino acid residues (Fig. 2b–j). These pores are relatively small (around 5 Å) but are compatible with the size of small chemical species, fitting with the idea that the pores of the S-layer may function as cation channels. Owing to the repeating pattern of the S-layer, these pores span the entirety of the N. maritimus cell surface. Supporting the anticipated cation-binding properties of the S-layer, a bioinformatics comparison of the amino acid composition of NmSLP with that of the N. maritimus proteome and all archaeal proteins revealed a substantial increase in aspartic acid residues (10.9% in NmSLP versus 6.1% in the N. maritimus proteome versus 7.3% in all archaea). Concurrent with this, the percentages of lysine and arginine residues are reduced (1.7% and 2.2% in NmSLP versus 8.4% and 3.4% in the N. maritimus proteome versus 3.8% and 5.9% in all archaea), resulting in an S-layer that is highly negatively charged.

Overall, our structure shows that the NmSLP monomers densely populate the S-layer sheet. This arrangement is reminiscent of the S-layer in the Dead Sea archaeon Haloferax volcanii, which is composed of an SLP called csg, which also consists of tandemly repeated Ig-like domains18. Such arrays of Ig-like domains have been observed in archaeal18,25, monoderm bacterial22,26 and diderm bacterial S-layers23,27. Although these SLPs share some structural similarities, they diverge notably at the sequence level, as well as at the overall organizational level containing different number of domains (Extended Data Fig. 5c), enabling them to assemble into unique two-dimensional sheets23,24, each with distinctly different glycosylation patterns and cell anchoring mechanisms.

In total, 17 glycan densities per monomer of NmSLP were also resolved in the single-particle map (Fig. 2k and Extended Data Fig. 4). Although these densities do not support direct derivation of the chemical structure of the glycans, they project away from the cell surface at asparagine residues, which are followed by a threonine or serine residue at the +2 position (Fig. 2k). Enshrouding the outer domains of the proteinaceous NmSLP S-layer, these glycans form a thick shell, encasing the cell in a sugar-rich coat (Fig. 2k). The mesh-like arrangement of the glycans probably provides protection, potentially shielding the cell from phages28. It might also enhance the hydrophilicity of the cell surface, making it suitable for marine environments. Most of these glycans are located in the N-terminal segment of NmSLP, primarily in outermost domains 1, 2 and 4, with a single glycan present in domain 6 (Fig. 2k).

Ammonium binding of the S-layer

We confirmed that the concentration of ammonium ions in the medium strongly influences the growth of N. maritimus (Supplementary Fig. 1a), as shown previously4,9. To directly measure ammonium ion binding to the cell surface, we performed isothermal titration calorimetry (ITC), titrating a medium containing NH4Cl against whole N. maritimus cells. As expected, growing N. maritimus cells showed strong and robust ammonium binding (Fig. 3a,b and Supplementary Fig. 1b; n = 3). We subsequently performed cryo-ET analysis of the same sample after ITC measurements, revealing normal cell morphology with an intact S-layer coating the cells (Fig. 3c,d; n = 150). As ammonium is the sole energy source for these growing cells, ammonium ions must somehow reach the cell for oxidation; we therefore inferred that the measured ammonium binding occurs either directly to the S-layer or to the underlying cell after passage through the S-layer (Fig. 3c,d). We then perturbed the S-layer by pretreating the cells with ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), which is known to impair several prokaryotic S-layers16. N. maritimus cells that were pretreated with 2.5 mM EGTA showed altered ammonium binding (Fig. 3a,b and Supplementary Fig. 1b; n = 3), concurrent with observed gaps and partial S-layer disruptions seen in cryo-ET in a heterogeneous population (Fig. 3e; n = 83). Near-complete disruption of the S-layer with 5 mM EGTA entirely abolished ammonium binding (Fig. 3a,b and Supplementary Fig. 1b; n = 3), leading to rounding up of cells with exposed, uncoated membranes (Fig. 3f; n = 152). These experiments indicate that an intact S-layer is critical for ammonium binding and may also be important for cell shape maintenance in N. maritimus in a calcium-dependent manner.

Fig. 3: Ammonium ions either bind directly to the S-layer or pass through it to bind to the underlying cell.
figure 3

a, ITC signal of intact N. maritimus cells titrated with ammonium chloride with different pretreatments with EGTA. b, Quantification of the ITC curves showing the total heat (Q) released in the different experiments. Data are mean ± s.d. n = 3 biologically independent replicates. Statistical analysis was performed using ordinary one-way analysis of variance (ANOVA) with correction for multiple comparisons; ****P < 0.0001, **P = 0.0075. c, Schematic of the ITC and cryo-ET experiment presented. d, Cryo-ET analysis of N. maritimus cells after titration with ammonium chloride shows complete coating with an S-layer. n = 150 from two biological replicates. e, Cryo-ET analysis of N. maritimus cells treated with 2.5 mM EGTA after titration with ammonium chloride shows gaps in the S-layer in a heterogeneous cell population. n = 83 from two biological replicates. f, Cryo-ET analysis of N. maritimus cells treated with 5 mM EGTA after titration with ammonium chloride showing round cells with a naked membrane. n = 152 from three biological replicates. For d–f, scale bars, 500 Å.

Source Data

Full size image

To examine the biochemically observed ammonium-binding properties of the S-layer structurally, we purified S-layer sheets in vitro, at a higher concentration of ammonium (2.5 mM compared to 1 mM NH4Cl) and resolved a 3.1-Å-resolution cryo-EM structure of the S-layer enriched in ammonium ions. Compared with our original cryo-ET structure (Supplementary Fig. 1c–f), this ammonium-enriched S-layer structure showed an increased number of unexplained densities at several S-layer pores, indicative of ammonium ion binding to the negatively charged amino acid residues lining these pores (Supplementary Fig. 1c–f).

To further investigate the ability of the lattice to bind to cations, we performed atomistic molecular dynamics (MD) simulations of the N. maritimus S-layer in the presence of ammonium (Fig. 4a, Methods, Extended Data Fig. 6 and Supplementary Fig. 2), using a framework that has recently been shown to yield strong agreement with X-ray crystallography and cryo-EM studies of cation binding to S-layers21. Analysis of residue-based ammonium occupancies over the course of three 500 ns simulations using PyLipID29 clearly identified the primary acidic residues that mediate ammonium binding (Supplementary Table 2 and Supplementary Video 4). Moreover, mapping the high-ammonium occupancy (>50%) residues onto the hexamer structure revealed multiple acidic residue clusters in excellent agreement with the pores identified in our cryo-EM and cryo-ET maps (Fig. 4b and Extended Data Fig. 7). These observations are compatible with the idea that the multiple pores in the S-layer sheet, lined with negatively charged residues, facilitate the movement of ammonium ions across the S-layer. Furthermore, in our simulations, we also found that ammonium ions could be replaced with sodium ions, suggesting that the S-layer might not discriminate between different small positively charged ions for binding (Supplementary Fig. 2).

Fig. 4: Ammonium (NH4+) binding to the negatively charged S-layer lattice.
figure 4

a, MD simulations support ammonium-ion binding at the S-layer pores. Residue-based ammonium occupancies during the 0.1 M NH4+ MD simulations are plotted onto the S-layer structure on a relative scale from white to purple. b, The distribution of negatively charged residues (shown in red) in the S-layer matches well with the MD simulations showing predicted ammonium-binding residues (Extended Data Fig. 6). c, A histogram of negatively charged residues along the S-layer, overlaid onto the ammonium-ion positions in the 0.1 M NH4+ MD simulations (distance was calculated from the closest, membrane-proximal amino acid residue in the S-layer structure). For the averaged ammonium-ion residence from three independent MD simulations (averaged over the last 400 ns of each simulation), the error bars denote ±1 s.d. (Supplementary Fig. 2). d, The mean local charge of NmSLP plotted along the sequence shows a gradual but continual increase in negative charge. The z position of the NmSLP residues, derived from the S-layer structure, is indicated, with the ninth Ig-like domain forming the base of the S-layer, proximal to the cell membrane.

Source Data

Full size image

Collectively, our structural data reveal that the S-layer of N. maritimus functions as a multichannel exchanger for ammonium ions, featuring negatively charged residues that line several pores in the S-layer (Fig. 2b–j). In this context, it is interesting that the negative charge of the S-layer increases when moving from the extracellular environment toward the cell membrane (Fig. 4c), specifically from the N to the C terminus of each NmSLP. This charge gradient probably facilitates the movement of ammonium ions through the S-layer towards the cell membrane. Our MD simulations support this hypothesis, showing an accumulation of ammonium ions at the cell-facing side of the S-layer (Fig. 4c, Supplementary Fig. 2 and Supplementary Video 4). To ensure that these observations did not depend on the amount of ammonium present, we conducted further MD simulations of the S-layer system in 0.05 M or 0.2 M ammonium, reproducing in both cases strong ammonium binding by the S-layer and relative accumulation of ammonium ions towards the cell-facing side of the S-layer (Supplementary Fig. 2).

The observed increase in negative charge at the structural level is also mirrored at the sequence level. There is a continuous and substantial increase in negative charge from the N to the C terminus of the NmSLP protein sequence (Fig. 4d), with the membrane-proximal C terminus being highly negatively charged. Using sensitive sequence-based homology searches and structure prediction, we found that Ig-domain-containing SLPs with a charge gradient are common across all described groups of ammonia-oxidizing archaea (AOA), suggesting similar S-layer arrangements in these species (Extended Data Fig. 7). Although the overall domain organization of such AOA and other archaeal SLPs could be similar (Extended Data Fig. 5a,b), the number of Ig-like domains and the charge distribution in each SLP vary, possibly reflecting differences in the function of these S-layers in binding positively charged molecules and ions (Extended Data Fig. 7d).

Molecular modelling of the cell surface

Together, our data enabled us to construct a molecular model of the N. maritimus cellular S-layer (Fig. 5a,b). Subtomogram position mapping of the S-layer hexamers and pentamers demonstrates that the S-layer coats the cell surface with near-perfect continuity (Fig. 5a,b), a characteristic that is also observed in other archaeal S-layers18. The continuous S-layer is heavily glycosylated, possibly protecting the cell from phages in the harsh marine environment. The dome-shaped structure of the S-layer hexamer probably supports flexibility, allowing NmSLP to coat different parts of the cell membrane with varied curvature. The S-layer is closed around cells by pentameric positions, which also appear to be composed of NmSLP (Extended Data Fig. 8). At the technical level, this study highlights the power of modern structural biology—molecular structures obtained from whole-cell cryo-ET, in conjunction with in-cell biochemistry and MD simulations, can provide key biochemical and mechanistic insights. With advancements in data collection and image-processing methodologies19,30,31, we anticipate that such in situ structural techniques will significantly enhance our molecular understanding of cells (Supplementary Fig. 3).

Fig. 5: The in situ structure of the N. maritimus S-layer shows a multichannel sieve nearly perfectly coating cells at high-copy numbers.
figure 5

a, The S-layer lattice coats nearly the entire outer surface of N. maritimus cells. A small part of the lattice map (black dashed line) has been cut out from the top of the cell for clarity. The pseudohexagonal lattice is joined together by pentameric defects (Extended Data Fig. 8). Scale bar, 500 Å. Cellular tomography was performed at least 32 times (Extended Data Table 1). b, A model of the S-layer lattice is shown as a space-filling representation with the same colour scheme as in Fig. 1. c, The ion-exchanging model for the N. maritimus S-layer. The highly negatively charged S-layer entraps ammonium ions, and these ions may move along multiple pores in the lattice, aided by increasing negative charge along the lattice. Once beyond the S-layer, the ammonium ions can diffuse to the membrane predominantly composed of crenarchaeol39, where they are converted to nitrite through hydroxylamine (NH2OH) as an intermediate by an unknown protein complex (pink)34, setting up an ion sink and concentration gradient between the S-layer and the membrane.

Source Data

Full size image

Conclusion

Our data are consistent with a scenario in which ammonium and other cations are bound and enriched at the cell surface by the negatively charged S-layer (Fig. 5c). The S-layer provides an extremely large surface area for interaction with the surrounding marine environment, where it acts as a multichannel cation exchanger and, due to its gradually increasing negative charge, leads to the accumulation of ammonium ions on the cell-facing side of the S-layer. An ‘ammonium sink’ exists at the cell membrane, where the integral membrane machinery, ammonia monooxygenase32,33, converts NH4+ to NO2− with hydroxylamine (NH2OH) as an intermediate34. This activity probably establishes an ammonium concentration gradient, extending from the membrane-proximal, ammonium-rich side of the S-layer to areas of lower concentrations near the cell membrane. A previous biophysical model has outlined how such a sink could function12, potentially initiating a chain of ammonium uptake that provides energy to the cell (Fig. 5c). This passive charge gradient of the S-layer could explain why marine AOA exhibit a 200-fold greater affinity for ammonia compared with ammonia-oxidizing bacteria7, which generally lack S-layers. This increased affinity enables AOA to thrive in low-ammonia marine environments. The energy-efficient, passive enrichment mechanism is particularly advantageous in the resource-limited, harsh conditions that are found in the ocean depths. Moreover, as S-layers similar to that of N. maritimus are found in nearly all AOA, including those in soil ecosystems (Extended Data Fig. 5a,b), this ammonium enrichment mechanism is probably conserved among these organisms35, contributing to the biogeochemically important nitrogen cycle (Supplementary Fig. 3). A comprehensive understanding of these biogeochemical processes is critical for preserving the vital ecological functions that sustain life on Earth.

Methods

Growth of N. maritimus cells

A culture of N. maritimus (SCM1) was provided by F. Elling and A. Pearson. Continuous cultures of SCM1 were grown on modified synthetic crenarchaeota medium supplemented with 1 mM NH4Cl, as previously described7,40 at 28 °C in a standing incubator that was covered to prevent excessive exposure to light. The growth of SCM1 N. maritimus cells was monitored using a NO2−-detection assay reported previously1. The same assay was used to follow SCM1 growth with differing ammonium concentrations.

Purification of N. maritimus cell envelopes

Native cell envelopes were purified from N. maritimus by adapting a previously described protocol23. A total of 12 l of N. maritimus cultures was prepared and late-log-phase cells were collected by centrifugation (10,000g, 4 °C, 30 min) and frozen and stored at −80 °C until further experimentation. The cell pellet from a 1 l culture was carefully resuspended in 3 ml lysis buffer (50 mM HEPES/NaOH pH 7.5, 500 mM NaCl, 50 mM MgCl2, 10 mM CaCl2, 1% (w/v) CHAPS, supplemented with 1× cOmplete protease inhibitor cocktail (Roche)). The cell suspension was incubated for 1 h on ice, and then lysed using sonication (10×, 5 s pulse, amplitude strength 10%). The sonicated sample was subsequently centrifuged (80,000g, 4 °C, 1 h), forming a very small white pellet at the bottom of the centrifugation tube. The pellet was resuspended into 40 µl of the same buffer and used for cryo-EM experiments. For the ammonium-enriched sample, the buffer was supplemented with 2.5 mM NH4Cl.

Cryo-EM and cryo-ET sample preparation

For cryo-EM and cryo-ET grid preparation, previously reported protocols were used18,23,41. In brief, 2.5 µl of the specimen was applied to a freshly glow discharged Quantifoil R2/2 Cu/Rh 200 mesh grid, adsorbed for 60 s, blotted for 4–5 s and plunge-frozen into liquid ethane in the Vitrobot Mark IV (Thermo Fisher Scientific), while the blotting chamber was maintained at 100% humidity at 10 °C. For tomography, the specimen was additionally supplemented with 10 nm gold conjugated with protein A. The grids were clipped and stored under liquid nitrogen until cryo-EM data collection was performed.

Cryo-ET and cryo-EM data collection

Cryo-ET data

For high-resolution in situ structure determination of the S-layer, a pipeline for high-throughput data collection was adopted42. In brief, a Titan Krios microscope was used to collect tilt-series data with a dose-symmetric tilting scheme43. Tilt series were collected at a pixel size of 1.327 Å, with a total dose of ~121 e− Å−2 applied over entire series collected between ±60° with 3° tilt increments. A total of 160 tilt series were collected with a defocus range of between −2 and −5 µm target defocus, and the samples were subjected to 0.9 s of exposure per tilt video containing 10 frames each (Extended Data Table 1). For visualization of the cellular ultrastructure, tilt-series images were acquired using the SerialEM software44 at a pixel size 3.468 Å with a defocus range of −3 to −10 μm, ±60° oscillation, 1° increments with a total dose of ~172 e− Å−2 as well as at a pixel size of 1.33 Å with a defocus range of −4 to −10 μm, ±60° oscillation, 2° increments with a total final dose of ~160 e− Å−2.

Cryo-EM single-particle data

Single-particle cryo-EM data were collected as described previously16,18,23 on the Titan Krios G3 microscope (Thermo Fisher Scientific) operating at 300 kV fitted with a Quantum energy filter (slit width 20 eV) and a K3 direct electron detector (Gatan) with a sampling pixel size of 0.546 Å running in counting super-resolution mode. For the N. maritimus purified sheets sample, a total of 12,557 videos over three sessions was collected with a dose rate of around 3.5 e− per super-resolution pixel per s on the camera level. The sample was subjected to 4.2 s of exposure, during which a total dose of around 48–51 e− Å−2 was applied, and 40 frames were recorded per video (Extended Data Table 2).

Subtomogram averaging of whole cells for structure determination

To obtain initial lattice maps, a previously described strategy was used16, in which tilt-series alignment using gold fiducials and tomogram generation was performed using IMOD45 and initial contrast transfer functions (CTFs) were estimated using CTFFIND446. Tomograms for visualization were generated using the simultaneous iterative reconstruction technique (SIRT) implemented in Tomo3D47 and denoised using Cryo-CARE36,37. Subtomogram averaging was performed using custom scripts written in MATLAB, described in detail elsewhere42,48. For initial cryo-ET structure determination, we used previously published methods17, with the major difference being the use of a recently developed 3D-CTF correction method for tomographic data49. The roughly aligned subtomogram coordinates were then imported into RELION-4 for further analysis19. We used the tilt series after video frame alignment from the initial analysis above, without additional preprocessing, along with the tilt-series alignments performed within IMOD, CTF parameters from CTFFIND446 and the Euler angle assignments and subtomogram coordinates from the original analysis to proceed with refinement. The imported parameters into RELION-4 were used for multiple cycles of pseudosubtomogram generation and realignment as described recently19. Accounting for per-particle motions with additional cycles of pseudosubtomogram improvements and realignments increased the resolution of the NmSLP hexamer to 3.4 Å in C6 symmetry. Relaxation of the symmetry50,51 led to an improved (3.3 Å) resolution overall, and 3.2 Å at the pseudohexameric axis, but decreased the resolution (~4.5 Å) at the periphery of the hexamer (Extended Data Table 1 and Extended Data Fig. 1). For spatial analysis of hexameric and pentameric S-layer positions with respect to the cell centre, the distance of each position from the cell centre was normalized by the maximally distanced hexamer/pentamer in every cell in the tomogram.

Cryo-EM single-particle analysis

For S-layer structure from two-dimensional sheets, cryo-EM data processing was performed as described previously for S-layers in our laboratory18,23. Videos collected at the scope were clustered into optics groups based on the XML metadata of the data-collection software EPU (Thermo Fisher Scientific) using a k-means algorithm implemented in EPU_group_AFIS (https://github.com/DustinMorado/EPU_group_AFIS). Imported videos were motion-corrected, dose-weighted and Fourier cropped (2×) with MotionCor252 implemented in RELION-3.153. CTFs of the resulting motion-corrected micrographs were estimated using CTFFIND446. Initially, side views of S-layer sheets were first manually picked along the edge of the lattice using the helical picking tab in RELION while setting the helical rise to 60 Å. Top and tilted views were manually picked at the central hexameric axis. Manually picked particles were extracted in 4× downsampled 128 × 128 px2 boxes and classified using reference-free 2D classification inside RELION-3.1. Class averages centred at a hexameric axis were used to automatically pick particles inside RELION-3.1. Automatically picked particles were extracted in 4× downsampled 128 × 128 px2 boxes and classified using reference-free 2D classification. Particle coordinates belonging to class averages centred at the hexameric axis were used to train TOPAZ54 in 5× downsampled micrographs using the neural network architecture conv127. For the final reconstruction, particles were picked using TOPAZ and the previously trained neural network above. Furthermore, top, bottom and side views were picked using the reference-based autopicker inside RELION-3.1, which TOPAZ did not readily identify. Particles were extracted in 4× downsampled 128 px × 128 px boxes and classified using reference-free 2D classification inside RELION-3.1. Particles belonging to class averages centred at the pseudohexameric axis were combined, and particles within 30 Å were removed to prevent duplication after alignment. All of the resulting particles were then re-extracted in 4× downsampled 128 × 128 px2 boxes. All of the side views and a subset of the top and bottom views were used for initial model generation in RELION-3.1. The scaled and low-pass filtered output was then used as a starting model for 3D auto refinement in a 512 × 512 px2 box. Per-particle defocus, anisotropy magnification and higher-order aberrations55 were refined inside RELION-3.1, followed by three rounds of focused 3D autorefinement. Bayesian particle polishing was performed subsequently in a 640 px × 640 px box55 followed by autorefinement and symmetry relaxation50,51. The final map was obtained from 354,860 particles and post-processed using a soft mask focused on the central hexamer, yielding a global resolution of 2.7 Å according to the Fourier shell correlation criterion between two independently refined half-maps at a threshold value at 0.143 (ref. 56) and a local resolution of up to 2.5 Å (Extended Data Fig. 3 and Extended Data Table 2). The two-dimensional sheet-like arrangement led to anisotropy in resolution, with lower resolution perpendicular to the plane as estimated by directional FSCs57. Further details are provided in Extended Data Table 2 and Extended Data Fig. 3.

Data visualization, analysis and model building

For model building, a previously described strategy was used18,23. For the single-particle cryo-EM map, the original 640 × 640 × 640 voxel box was cropped into a 320 × 320 × 320 voxel box. In both the cryo-ET and cryo-EM maps, and the protein backbone of NmSLP was manually traced as a poly-alanine model through a single NmSLP subunit using Coot58. Side chains were assigned at clearly identifiable positions which allowed deduction of the protein sequence register. The model was then placed into the hexameric map as six copies and subjected to several rounds of refinement using refmac559 inside the CCP-EM software suite60 and PHENIX61, followed by manually rebuilding in Coot58. At the N terminus, the C2 maps were better resolved compared to the C6 maps at the C termini of NmSLPs; therefore, multimap atomic model refinement was performed in servalcat62. Model validation was performed in PHENIX and CCP-EM, and data visualization was performed in UCSF Chimera63, UCSF ChimeraX64 and PyMOL65. To analyse lattice interfaces, multiple copies of the hexameric structure were placed in the cryo-EM map prepared with a larger box size. Figure panels containing cryo-EM or cryo-ET images were prepared using IMOD and Fiji66. Lattice maps of S-layers for visual inspection were plotted inside UCSF Chimera63 with the PlaceObject plugin67 and model coordinates were plotted inside UCSF ChimeraX64 with the sym function and the BIOMATRIX PDB file header or directly using the ArtiaX plugin68. The SPA and STA maps were postprocessed using deepEMhancer38 for visualization of the N-glycan densities (Fig. 2k and Extended Data Fig. 4g–h,k–l). Composite maps from focused refinements of the two-fold (C2) and six-fold (C6) symmetrized maps were generated using refmac559 and PHENIX61 and then converted using mtz2mrc implemented in PHENIX61.

Bioinformatic analysis

A previously described strategy for detection and analysis of SLPs was used18,23. All sequence similarity searches were performed in the MPI Bioinformatics Toolkit69 using BLAST70 and HHpred71. BLAST searches were performed against the nr_arc database, a specialized subset of the NCBI non-redundant protein sequence database filtered specifically for archaeal sequences, using the default settings to identify homologues of NmSLP in archaea. The searches were seeded with the protein sequence of N. maritimus SLP. The domain organization of several obtained matches and many experimentally characterized SLPs (Supplementary Table 1) were analysed using HHpred searches with the default settings over the PDB70 and ECOD70 databases, which are versions of the PDB and ECOD databases filtered for a maximum pairwise identity of 70%, and using structural models built using AlphaFold (v.2.2.0)72. Signal peptides were predicted using SignalP (v.6.0)73. The mean local charge of the protein sequences was calculated using the EMBOSS charge tool74, using a window length of 7.

ITC analysis

ITC measurements were made using Malvern Panalytical ITC200 instruments at 25 °C in SCM buffer without ammonium chloride. Experiments were performed at a reference power of 10 µcal s−1 and with injections at 300 s intervals to capture the large exothermic heats and broad peak profiles. The ITC cell contained N. maritimus at an optical density at 600 nm (OD600) of 1.0 and the syringe contained 10 mM ammonium chloride in the SCM buffer. In total, ten injections, with the first injection corresponding to 0.5 µl, followed by nine injections of 1 µl were performed, resulting in a final ammonium chloride concentration of 0.475 mM in the ITC cell. N. maritimus cells were pretreated with 0, 2.5 or 5 mM EGTA for 30 min, and were then centrifuged at 16,000g for 15 min and resuspended in SCM medium lacking ammonium chloride to recover before adjusting to an OD600 of 1.0. Control measurements of injections of ammonium chloride into buffer were performed and these heats were small and close to the values seen for buffer into buffer control experiments. This control heat was subtracted from the N. maritimus experiments before peak integration using Malvern Panalytical PEAQ software. Experiments were performed three times with different batches of N. maritimus prepared from cells in log-growth phase in SCM medium with 1 mM ammonium chloride as nutrient source. These cultures were centrifuged and resuspended in SCM buffer lacking ammonium chloride before adjusting to an OD600 of 1 and loading into the ITC cell.

MD simulations

The NmSLP hexamer structure was prepared for atomistic MD simulation using VMD (v.1.94)67. The system was first solvated with TIP3P water molecules and 0.5 M NaCl to mimic the salinity of sea water. Next, 312 ammonium ions (0.1 M NH4+) were randomly distributed throughout the solvent, along with an equal number of chloride counter ions to maintain a charge neutral system. Simulation parameters for NH4+ were derived through analogy with existing CHARMM parameters for methylammonium. Note that, to better help identify specific ion binding sites, no structural ions apparent from the NmSLP cryo-EM and cryo-ET structures were included. The resulting system contained 566,371 atoms, including 136,236 protein atoms, 141,657 water molecules, 2,246 sodium ions, 1,358 chloride ions and 312 ammonium ions, within a hexagonal box of dimensions x = y = 217 Å, z = 150 Å and axial angles α = β = 90°, γ = 60°. The geometry of the simulation box was chosen so that the molecular interfaces observed between neighbouring hexamers in our tomography data would be reproduced through the interactions of the NmSLP hexamer with its periodic images in the x–y plane. The system was then subjected to a series of conjugant gradient energy minimizations followed by three 500 ns MD simulations. To prevent potential distortions in the NmSLP hexamer due to the absence of structural ions offsetting its highly negative charge, protein atoms (excluding hydrogens) were harmonically restrained during simulation. Unless otherwise indicated, analyses were performed after disregarding the first 100 ns of each simulation to ensure equilibrium sampling. To assess the robustness of the observed ammonium-binding pattern, we further constructed hexamer systems containing ammonium at concentrations of 0.05 M (156 NH4+ ions) and 0.2 M (624 NH4+ ions) using an identical procedure, and these systems were subjected to a single 500 ns production simulation. Note that lower concentrations of ammonium ions could not be used due to few total ammonium ions in the box in every simulation. All simulations were conducted using NAMD (v.2.14)68 and the CHARMM36 force field38. Production simulations used the NPT ensemble with conditions maintained at 1 atm and 310 K using the Nosé–Hoover Langevin piston and Langevin thermostat, respectively. The r-RESPA integrator scheme was used with an integration time step of 2 fs and SHAKE constraints applied to all hydrogen atoms. Short-range, non-bonded interactions were calculated every 2 fs with a cut-off of 12 Å; long-range electrostatics were evaluated every 6 fs using the particle-mesh-Ewald method. Further details are provided in Supplementary Table 3.

Reporting summary

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