Structure of Geobacter cytochrome OmcZ identifies mechanism of nanowire assembly and conductivity

Abstract

OmcZ nanowires produced by Geobacter species have high electron conductivity (>30 S cm⁻¹). Of 111 cytochromes present in G. sulfurreducens, OmcZ is the only known nanowire-forming cytochrome essential for the formation of high-current-density biofilms that require long-distance (>10 μm) extracellular electron transport. However, the mechanisms underlying OmcZ nanowire assembly and high conductivity are unknown. Here we report a 3.5-Å-resolution cryogenic electron microscopy structure for OmcZ nanowires. Our structure reveals linear and closely stacked haems that may account for conductivity. Surface-exposed haems and charge interactions explain how OmcZ nanowires bind to diverse extracellular electron acceptors and how organization of nanowire network re-arranges in different biochemical environments. In vitro studies explain how G. sulfurreducens employ a serine protease to control the assembly of OmcZ monomers into nanowires. We find that both OmcZ and serine protease are widespread in environmentally important bacteria and archaea, thus establishing a prevalence of nanowire biogenesis across diverse species and environments.

Long-range (>10 μm) extracellular transport of respiratory electrons along networks of protein filaments, called microbial nanowires, has applications for bioremediation, bioenergy and bioelectronics¹ and drives a wide range of globally important environmental phenomena that influence carbon and mineral cycling in soils and sediments, bioremediation, corrosion and anaerobic conversion of organic wastes to methane or electricity^2,3. The common soil bacterium Geobacter has proven to be a useful model for studying long-range microbial electron transport because it directly contacts extracellular electron acceptors⁴ or syntrophic partner species³ using protein nanowires^5-7. Structural, functional and cellular localization studies have shown that nanowires present on the surface of electricity-producing wild-type (WT) Geobacter sulfurreducens comprise cytochromes OmcS^8,9 and OmcZ^10,11, whereas strains with genetically modified OmcS grown under conditions that do not require extracellular electron transport produce filaments of cytochrome OmcE¹². Furthermore, G. sulfurreducens pili are required for the secretion of OmcS and OmcZ nanowires^1,13.

Out of 111 predicted c-type cytochromes present in the 3.8 Mb genome of G. sulfurreducens, OmcZ is the only nanowire-forming outer-surface cytochrome that has been experimentally proven to enable long-range (>10 μm) electron transport¹⁴, highlighting its importance over other OmcS nanowire-like cytochromes that are abundant in soil¹⁵.

Using a combination of immunogold labelling with a suite of biochemical and biophysical experiments, we previously reported that G. sulfurreducens can assemble highly conductive OmcZ nanowires, and that external application of electric fields to biofilms grown on electrodes stimulates bacterial production of OmcZ nanowires¹⁰. The extremely high conductivity (>30 S cm⁻¹) of OmcZ nanowires and their concentration¹⁶ adjacent to the biofilm–electrode interface is correlated with maximum observed metabolic activity of G. sulfurreducens cells¹⁷. Prior genetic, electrochemical and spectroscopic studies have suggested a crucial role for OmcZ in extracellular respiration^11,18,19 by a mechanism that differs from diffusing shuttle molecules that are produced by other bacteria²⁰. Shuttles limit the electrocatalytic performance of other bacteria^20,21. Synthetically embedding silver nanoparticles in the outer cell membrane improves electric power for shuttle-producing species²¹, whereas similar power is generated by WT G. sulfurreducens via nanowires²². Thus, in addition to physiological role in long-range respiration and ecological role in enabling bacteria to form biofilm communities, OmcZ nanowires enable bacteria to produce the highest electric power by pure culture^23,24 and form >100-μm-thick conductive biofilm communities^14,25,26. However, the mechanism underlying the ability of OmcZ nanowire networks in biofilms to transport electrons over 100 times bacterial length is not known. Understanding the molecular mechanisms of OmcZ nanowire’s assembly and high conductivity will inspire novel strategies to improve bacterial extracellular respiration using these living electronic materials.

In this Article, we use a combination of structural and functional studies using a suite of genetic, biochemical, biophysical and computational methods to identify the mechanism of how OmcZ nanowires assemble and conduct electricity.

Results

OmcZ structure reveals linear and closely stacked haems

All prior studies on OmcZ nanowires also contained nanowires of other cytochromes^27,28, making it difficult to achieve high-resolution structure or perform any bulk functional studies. To obtain pure OmcZ nanowires without any contamination from other nanowires, we disrupted omcS in G. sulfurreducens CL-1 that forms highly conductive biofilms²⁹ (Supplementary Table 1, Methods). We sheared nanowires from pelleted cells using blending to purify them away from pili filaments which are present in the supernatant¹³. Next we purified OmcZ n anowires using a sucrose gradient and size exclusion chromatography (Fig. 1a). We used sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) gel with Coomassie staining and negative-stain transmission electron microscopy (TEM) to validate the purity of nanowires (Fig. 1b,c) and proceeded to analyse pure OmcZ nanowires using cryogenic electron microscopy (cryo-EM) (Fig. 1d,e)

Fig. 1 ∣. Cryo-EM structure of purified OmcZ nanowires.

a, Size-exclusion chromatography of OmcZ nanowires from G. sulfurreducens. The chromatogram of purified sample, showing traces for protein at 280 nm (purple) and haem c at 410 nm (red). Inset: sucrose gradient. b, Coomassie-stained SDS–PAGE gel of purified OmcZ nanowires. c, Negative-stain TEM image of purified OmcZ nanowires. d,e, Representative cryo-EM micrograph (d) and 2D average (e) of purified OmcZ nanowires. Scale bars, 50 nm (c), 20 nm (d) and 5 nm (e). f, Highly linear haem chain in OmcZ. g, Structure of OmcZ protomer. h, Illustration of cryo-EM map quality by fitting sidechains into the EM density.

We obtained a 3.5-Å-resolution cryo-EM structure for purified OmcZ nanowires using iterative helical real space reconstruction (Fig. 1f,g, Extended Data Fig. 1 and Extended Data Table 1a). The structure has a left-handed one-start helix, with a helical rise of 57.4 Å and a rotation of −159.1°. At this resolution, the backbone density along with many bulky sidechain densities, such as aromatic residues and all eight haem densities, are visible (Fig. 1h and Extended Data Fig. 1a-j). Cells produce two forms of OmcZ: the 50 kDa precursor (OmcZ₅₀) and nanowire-forming 30 kDa (OmcZ₃₀). The cryo-EM model enabled us to trace the C_α backbone for each protomer subunit and thread OmcZ₃₀ sequence with all 258 amino acids and 8 haems³⁰. OmcZ contains ~12% helices and ~28% β-strands and turns (~21% strands and ~7% turns), which is higher than ~22% β-strands and turns found in OmcS (~5.7% β strands, 16.7% turns), consistent with prior secondary structure studies¹⁰.

OmcZ subunits have a compact and linear structure with a smaller diameter (~23 Å) than OmcS nanowires⁸ (~35 Å) (Fig. 1f). Owing to its compactness, the haems in the OmcZ nanowire are closely stacked with edge-to-edge distances substantially smaller than haems in OmcS nanowires, consistent with data by grazing-incidence X-ray diffraction that detects π-stacking in soft materials¹⁰. Parallel-stacked haems are ~0.3 Å closer and perpendicularly (T)-stacked haems are ~1 Å closer in OmcZ compared with in OmcS (Fig. 2a,b and Extended Data Table 1b).

Fig. 2 ∣. Linear haem arrangement and closer stacking in OmcZ than in OmcS.

a,b, Haem arrangement of OmcZ (a) versus OmcS protomer (b). c,d, Evolutionary conservation map of all OmcZ residues (c) and haem surrounding residues within 4 Å from the haem (d). All three consecutive histidine pairs are evolutionarily conserved (e) and lock haems more tightly (f).

Closer stacking of haems in OmcZ may be due to conserved amino acid residues adjacent to haems (Fig. 2c,d). In particular, three pairs of histidines are evolutionarily conserved across all OmcZ homologues analysed (Fig. 2e,f and Extended Data Fig. 2). Structure revealed that these histidine pairs lock T-stacked haems more tightly together than those found in OmcS (Fig. 2a,b and Extended Data Table 1b). We hypothesize that high electronic conductivity of OmcZ may be partly due to closer distance between T-stacked haems, which typically have ~10 times lower electronic coupling (ceteris paribus, 100 times lower electron transfer rate) than parallel-stacked haems³¹. c3 cytochromes from Desulfovibrio species also have similar haem structures (Protein Data Bank (PDB) IDs: 1gyo, 2bq4 and 2e84) (ref. ¹⁰). The amino acid sequence of c3 contains CXXCHH motif with one proximal and one distal histidine in the haem pairs while OmcZ contains two distal histidines. All four structures have consecutive histidines causing tight haem T-junction with similar haem–haem distances¹⁰. However, none of the c3 cytochromes is known to form nanowires.

To compare the haem–haem interactions in OmcZ and OmcS nanowires, we used circular dichroism (CD) spectroscopy. CD spectra of OmcZ nanowires at pH 7 were similar to pH 10.5 used to solve the atomic structure (Extended Data Fig. 3a), indicating similar OmcZ structures at these pH values. CD measurements of OmcZ nanowires and computational predictions of haem interactions in OmcZ using density functional theory (DFT) identified a peak instead of valley at 406 nm, which indicates that haem–haem interactions in OmcZ are different than in OmcS (Extended Data Fig. 3b-i). These analyses confirm that the haem arrangement in OmcZ enables specific haem–haem interactions.

Haem loop traps diverse extracellular electron acceptors

In OmcZ, haems I–V and VII form a linear, central, electron transport chain whereas haem VI is surface exposed (Figs. 2a and 3a). In addition, haems in OmcZ have more solvent exposure than haems in OmcS (Fig. 3a,b). For example, the structure reveals fully exposed haem VI and partially exposed haems I and II that can make more contact with electron acceptors and electrodes for low contact resistance and higher conductivity of OmcZ nanowires than OmcS nanowires. Therefore, both the closer stacking of haems and their higher surface exposure in OmcZ may account for the higher conductivity measured for individual OmcZ¹⁰ than OmcS³¹ nanowires using conducting-probe atomic force microscopy as well as high conductivity measured for individual nanowires produced by strain W51W57 (ref. ³²) (shown to be OmcZ nanowires¹⁰).

Fig. 3 ∣. OmcZ nanowires show solvent-exposed haems and non-canonical loop that are involved in electron transfer.

a,b, Higher solvent exposure of haems in OmcZ (a) and OmcS (b) nanowires. c,d, Haem II binding motif is non-canonical (CX₁₄CH) in OmcZ (c) versus conventional (CXXCH) in OmcS (d). e, OmcZ nanowires fully reduce soluble extracellular electron acceptor Fe(III) citrate. f, OmcZ without the haem II loop shows a slower reduction rate of Fe(III) citrate than the WT OmcZ nanowires. g, Electrostatic potential at pH 7 for OmcZ and OmcS nanowires. h, OmcZ nanowires reduce insoluble electron acceptor Fe(III) oxide slower than OmcS nanowires. Values show fraction of reduced cytochrome nanowires as measured by area of α-band absorbance (λ = 543–562 nm) with mean ± standard deviation (n = 3 biological replicates).

Furthermore, all haems in OmcZ are anchored to the protein backbone by haem-binding motifs (CXXCH) except for haem II, which is attached by a non-canonical haem-binding motif that has 12 additional residues (CX₁₄CH) between cystines 53 and 68 in OmcZ but not in OmcS (Fig. 3c,d). These 12 residues in OmcZ form a solvent-exposed loop, which might trap small molecules as electron acceptors and transfer electrons from haem II. Using reconstituted OmcZ, we find that deletion of the haem II loop reduces the rate of electron transfer by OmcZ nanowires to extracellular electron acceptors compared with the WT (Fig. 3e,f). Therefore, our structure explains the ability of OmcZ to reduce diverse physiologically important electron acceptors³⁰ with high conductivity. Higher solvent exposure in OmcZ compared with OmcS is also consistent with more negative reduction potential of OmcZ³⁰ compared with OmcS³³.

OmcZ with severely modified sequence³⁴ can reduce riboflavin in vitro. Sequence modification could have also prevented nanowire formation. We found that purified OmcZ nanowires do not reduce riboflavin (Extended Data Fig. 4b). Our finding is consistent with the prior studies that redox cycling of flavins is primarily associated with the cytochromes in the outer membrane rather than nanowires³⁵.

Despite higher conductivity and solvent exposure of haems, OmcZ is not essential for bacterial reduction of Fe(III) oxide, an extracellular electron acceptor abundant in soil³⁶. Furthermore, in prior studies, OmcZ monomers could not reduce Fe(III) oxide³⁰, which could be due to harsh conditions used in purification altering the OmcZ structure. To understand the binding of OmcZ and OmcS nanowires to Fe(III) oxide, we compared the surface charge of OmcZ and OmcS with Fe(III) oxide. At physiological pH, the surface charge of OmcZ nanowires is mostly negative, whereas OmcS nanowires have a positively charged pocket (Fig. 3g), in comparison with negatively charged Fe(III) oxide³⁷. Furthermore, OmcZ was not fully reduced by dithiothreitol (DTT) (Extended Data Fig. 4b), suggesting that OmcZ contains low-reduction-potential haems that could affect the thermodynamics and kinetics of electron transfer. These differences in surface charge and redox potentials could explain slower reduction of Fe(III) oxide by OmcZ nanowires than OmcS nanowires (Fig. 3h).

Environment re-arranges organization of nanowire network

Despite their negative surface charge, cell-attached OmcZ nanowires form a network under physiological conditions (Fig. 4b), such as in electrode-grown biofilms¹⁰, possibly due to the stacking of surface-exposed haems. This ability of OmcZ nanowires to form network has important implications in bacterial physiology. For example, our studies show that such a network can trap and reduce diverse extracellular electron acceptors^30,36 such as Fe(III) citrate and Fe(III) oxide (Fig. 3e,f,h and Extended Data Fig. 4). Furthermore, network of OmcZ nanowires in biofilms could explain their ability to conduct electrons over 100 times bacterial length².

Fig. 4 ∣. Organization of OmcZ nanowire network changes with biochemical environment.

a, Solubility of OmcZ nanowires decreases with increasing ionic strength of salt (NaCl). b, WT cells form a network of OmcZ nanowires under physiological conditions. c, OmcZ nanowires are only soluble in a buffer with low ionic strength d, OmcZ nanowires precipitate with increased ionic strength unless the pH is further increased (all buffers are at 20 mM). e–k, Schematics and corresponding negative-stain TEM images of supramolecular ordering of OmcZ nanowires: OmcZ nanowires remain aggregated in the presence of 6-aminocapric acid (e,f); OmcZ nanowires form dense networks with addition of pimelic acid (g,h); OmcZ nanowires form ordered arrays in the presence of hexamethylenediamine (i–k). Scale bars, 20 nm (k), 100 nm (b,e,f), 200 nm (g,h) and 500 nm (j).

We further found that the organization of OmcZ nanowire network re-arranges in different biochemical environments. The solubility of OmcZ nanowires decreases with increasing salt concentration (Fig. 4a,c,d). Nanowires remain soluble at low (<10 mM) salt concentrations and start aggregating into precipitating networks at high concentrations. Notably, the precipitation is fully reversible, which could help bacteria to reversibly switch the aggregation state of the network in response to the environmental changes. We built ordered arrays using the ability of OmcZ nanowires to aggregate into networks (Fig. 4c-e). For example, we assembled OmcZ nanowires into C-shaped arrays using a positively charged hexamethylenediamine-based self-assembly³⁸ (Fig. 4e).

Furthermore, we found that purified OmcZ nanowires remain stable when treated with various denaturants including urea, sodium dodecyl sulfate (SDS), and guanidine chloride (GuHCl), or incubated at acidic pH 1.6 (Extended Data Fig. 5). The stability of OmcZ nanowires is probably due to its negative surface charge (Fig.3a), which may prevent binding to denaturing agents (such as SDS) that are also negatively charged. The propensity of OmcZ nanowires to form networks (Fig. 4b) might also prevent other denaturing agents (such as GuHCl) from accessing the OmcZ surface. Furthermore, OmcZ nanowires show unique self-assembly mechanism by polymerization primarily using hydrophobic interactions and haem and aromatic stacking at the interface (Fig. 5a-d), whereas most denaturing agents, such as urea, GuHCl and low pH, destabilize protein by targeting hydrogen bonding. Therefore, stability of OmcZ nanowires could be due to negative surface charge, propensity to form networks and polymerization primarily via stacking interactions. We propose that purified OmcZ nanowires could form the basis of electronic biomaterials for use in acidic or protein-denaturing environments.

Fig. 5 ∣. A protease-mediated switch regulates OmcZ nanowire assembly.

a–d, Interactions at the OmcZ protomer interface, driving assembly of OmcZ into nanowires, identified in the cryo-EM structure. a, Haem-haem stacking interaction. b, Electrostatic interaction. c, Hydrophobic interaction. d, Haemaromatic stacking interaction. e, Haem-stained gel showing OmcZ₅₀ cleavage by OzpA into nanowire-forming OmcZ₃₀. Lane 1: marker. Lane 2: OmcZ₅₀ without OzpA. Lane 3: OmcZ₅₀ + OzpA (full gel in Supplementary Fig. 1). f,g, Negative-stain TEM image of OmcZ₅₀ into OmcZ nanowires before (f) and after (g) cleavage by OzpA. h,i, 2D average (h) and averaged power spectrum (i) showing reconstituted OmcZ nanowires having nearly identical structure to native nanowires. Scale bars, 100 nm (f), 50 nm (g) and 2 nm (h). j,k, Side view (j) and top view (k) of AlphaFold-predicted structure of OmcZ₅₀ with confidence (pLDDT) score. l, Superimposition of the cryo-EM structure of OmcZ protomer (cyan) and the predicted model of OmcZ (red) showing similar structure. m, AlphaFold model of OmcZ₅₀ and serine protease OzpA (pink) forming a complex (animations in Supplementary Video 1). OmcZ₅₀ contains a cytochrome OmcZ₃₀ (yellow) and a sugar binding domain (purple). n, Catalytic triad of OzpA is predicted near the cleavage site of sugar-binding domain.

Assembly of OmcZ nanowire is proteolytically controlled

Our structural analyses showed that OmcZ nanowires have a protomer interface stabilized by four different interactions (Fig. 5a-d), but lack the inter-protomer metal coordination present in OmcS nanowires⁸ or OmcE filaments¹². To understand how OmcZ nanowires assemble, we characterized self-assembly of OmcZ nanowires in vitro. Previously, it was reported that G. sulfurreducens uses a serine protease, named OzpA (OmcZ protease), to cleave the precursor OmcZ₅₀ into nanowire-forming OmcZ₃₀ and that this cleavage is essential for extracellular electron transfer³⁹. However, the cleavage process and its role in assembly of OmcZ into nanowires was not fully elucidated.

We found that purified OzpA cleaves both native and recombinant OmcZ₅₀ into OmcZ₃₀ which then self-assembled into nanowires (Fig. 5e-i and Extended Data Figs. 6 and 7). Notably, reconstituted OmcZ nanowires after proteolytic cleavage had structures that were indistinguishable from native nanowires on the basis of a power spectrum analysis (Fourier transform of the sample images) and 2D average (Fig. 5h,i and Extended Data Fig. 6e,f). Our in vitro studies therefore suggest that the protease serves as a molecular switch that regulates the assembly of OmcZ nanowires.

We next built a model of nanowire precursor OmcZ₅₀ using AlphaFold⁴⁰ (Fig. 5j and Supplementary Video 1). The predicted structure of the OmcZ₃₀ domain in OmcZ₅₀ was similar to the cryo-EM structure (Fig. 5l). We identified an extra domain in the predicted structure of OmcZ₅₀ (shaded in purple in Fig. 5j,k) that resembles a sugar-binding domain from other bacterial proteins such as dietary fibre used by Bacteroidetes⁴¹. This domain is expected to prevent polymerization of OmcZ₃₀ into nanowires by blocking a region where two protomers interact to form nanowires (Fig. 5a-d) but to increase the solubility of OmcZ₅₀. The presence of this extra domain might therefore explain the inability of OmcZ₅₀ to polymerize into nanowires. In addition to preventing polymerization, the sugar-binding domain may enable localization and secretion of OmcZ₅₀. This domain might localize OmcZ₅₀ in the periplasmic glycan layer, or in the outer membrane (where it might bind to OzpA to initiate OmcZ polymerization). Interestingly G. sulfurreducens pili have surface sugars and are required for OmcZ secretion¹³. We propose that this sugar-binding domain of OmcZ₅₀ may bind to pili during OmcZ nanowire secretion.

We next built a computational model of the OzpA-OmcZ₅₀ complex using AlphaFold⁴⁰ (Fig. 5m,n and Supplementary Video 1). Our model of OzpA placed the catalytic triad (His, Asp and Ser) of OzpA near Asn 258, in a position suitable for cleaving the extra domain from OmcZ₅₀ to form OmcZ₃₀. OmcZ₅₀ is found in the periplasmic and outer-membrane spaces of G. sulfurreducens, whereas OmcZ₃₀ is mainly found in the outer-membrane space rather than the periplasm³⁰. OzpA is predicted to localize primarily in the outer-membrane and extracellular space³⁹. We propose that OzpA cleaves OmcZ₅₀ in the outer membrane so that OmcZ nanowires can self-assemble on the cell surface (Fig. 4e-i).

omcZ homologues are widespread in bacteria and archaea

Cytochrome nanowires have been reported only in G. sulfurreducens and omcZ homologues were thought to be rare because not all Geobacter species show OmcZ homologues⁴². To evaluate whether there are omcZ-like genes in other microbes, we analysed the entire omcZ gene cluster, and our bioinformatic analysis revealed that diverse environmentally important bacteria and archaea have omcZ-like genes next to ozpA (Fig. 6 and Extended Data Fig. 9).

Fig. 6 ∣. omcZ operon homologues are widespread in environmentally important bacteria and archaea.

a, Phylogenetic tree derived from the alignment of amino acid sequences of OmcZ homologues using cut-off score of 100 bit. b, The omcZ operon homologues are found in diverse bacterial and archaeal species. c, Highly similar haem arrangement revealed through superimposed structures of OmcZ nanowire (solved by cryo-EM, red) and omcZ homologues (modelled with AlphaFold) from methane oxidizing ANME-2 (cyan) and methanogenic archaea (Methanosacinales, yellow) with GenBank IDs MRG77204.1 and CAG1009295.1, respectively.

Using PCR with reverse transcription (RT–PCR) (Extended Data Fig. 8b), we confirmed co-transcription of omcZ and ozpA in G. sulfurreducens and found both genes in an operon of three genes encoding genes for nanowire precursor OmcZ₅₀, OzpA³⁹, which cleaves OmcZ₅₀ into nanowire-forming OmcZ₃₀ (Fig. 5e-i), and a peptidylprolyl cis–trans isomerase (GSU2074) required for the translation and processing of proline-rich proteins⁴³ (Fig. 5c). The amino acid sequence of OmcZ₃₀ contains 22 prolines³⁰. Therefore, the prolyl isomerase is probably required for proper folding of OmcZ₃₀ into the correct conformation pre-polymerization which might explain why isomerase is essential for bacterial growth on electrodes⁴³. All three genes in the omcZ operon are essential for the bacterial growth on electrodes. The 121 bp intergenic region between omcZ and ozpA is predicted to form extensive stem–loops on the messenger RNA and may suppress translation of ozpA by preventing the ribosome binding, unless triggered by an external stimulus¹⁰, thus controlling the assembly of OmcZ nanowires (Extended Data Fig. 8a,c).

Our amino acid sequence alignment of omcZ homologues revealed complete conservation of all eight haem binding motifs (Extended Data Fig. 2)⁴⁴. Notably, substantial sequence identity for several other residues was also observed throughout the sequence. For example, conserved residues include consecutive histidine pairs, which form a tight T-junction (Fig. 2e,f), and several prolines (in agreement with the conserved prolyl isomerase in the omcZ operon) (Fig. 6c).

Remarkably, using a combination of bioinformatics and AlphaFold, we found omcZ operon-like gene clusters, and OmcZ-like nanowire structures predicted in diverse methanogenic archaea that are the primary source of biogenic methane, such as Methanosarcinales (isolate METP1), as well as anaerobic methanotrophic (ANME) archaea, and nitrate-reducing freshwater ANME relatives (Candidatus Methanoperedens nitroreducens, ANME-2d), which consume methane released in marine sediments⁴⁴ that show high conductivity⁴⁵ (Fig. 6a,c and Extended Data Fig. 9). Furthermore, omcZ was one of the most highly transcribed genes in cultures grown under the conditions of anaerobic methane oxidation⁴⁴. AlphaFold predictions of these omcZ-homologue nanowires showed haem arrangement similar to that in cryo-EM structure of OmcZ nanowire (Fig. 6c). As methane is the most abundant hydrocarbon⁴⁴, we propose that diverse microbes might use OmcZ-like nanowires for globally important processes, such as regulating atmospheric methane levels. High nanowire conductivity would explain the ability of these microbes to transfer electrons rapidly at rates too fast for molecular diffusion^46,47.

Our in vitro experiments showed that OzpA cleaves OmcZ₅₀ to release OmcZ₃₀, which self-assembles into nanowires that have similar structures to native OmcZ nanowires (Fig. 4e-i). Furthermore, along with OmcZ, the OzpA serine protease is essential for extracellular electron transfer to electrodes by G. sulfurreducens³⁹. We found that the ozpA gene is conserved across all bacterial species that contain OmcZ homologues (Fig. 6c and Extended Data Fig. 2). Finally, amino acid sequences of predicted archaeal omcZ homologues show remarkably high (up to 68%) similarity to OmcZ, with conservation of all haem-binding motifs and residues involved in high conductivity (Extended Data Fig. 2). Therefore, it is likely that diverse bacteria and archaea use mechanisms similar to G. sulfurreducens to control the assembly of OmcZ nanowires.

Conclusions

Using a combination of structural, bioinformatic, computational and functional studies, we identify the mechanism of high conductivity and self-assembly of OmcZ nanowires to explain their physiological and ecological roles. Our studies also suggest a previously unknown strategy by which bacteria prevent aggregation of OmcZ₅₀ and assemble OmcZ nanowires on-demand. Our in vitro studies showed that the serine-protease OzpA cleaves OmcZ₅₀ to release OmcZ₃₀, which self-assembles into nanowires for long-distance electron transport in biofilms. Protein nanowires can serve as a template for other conductive materials^38,48. As G. sulfurreducens biofilms also function as living transistors², supercapacitors⁴⁹ and photoconductors, due to ultrafast (~200 fs) electron transfer in cytochrome nanowires⁵⁰, we predict that robust performance of OmcZ nanowires may find novel bioelectronic applications.

Methods

Bacterial strain construction for purification of OmcZ nanowires

OmcZ nanowires were purified by constructing a G. sulfurreducens omcS:pk18 strain using CL1 as a background strain²⁹. omcS:pk18 was constructed using a plasmid disruption protocol as previously described⁵¹. Briefly, we introduced an external plasmid pK18mobsacB to disrupt the omcS gene so that the strain will still produce OmcZ nanowires but not express OmcS. By the forward primer (5′-GTGACACCGGTCCTTCCAGCTA-3′) and the reverse primer (5′-CAGTTCACGTTGCTGGTGGC-3′), 820 bp of central region of omcS was cloned to the plasmid. The constructed plasmid was first introduced into Escherichia coli S17 strain, which further conjugated with G. sulfurreducens CL1 strain. All G. sulfurreducens colonies grew anaerobically on NBAF agar plate under 30 °C. A single-step recombination occurred between plasmid gene and genomic omcS with the constant supply of 200 μg ml⁻¹ kanamycin, which could disrupt the genomic omcS. Successful disruption was confirmed by PCR and nanowire preparation.

OmcZ nanowire purification

The CL1 omcS:pk18 strain was grown anaerobically in 10 l NBAF medium. Cells reaching the stationary phase were collected by centrifugation and the cells were homogenized in 150 mM ethanolamine (ETA) buffer. Then nanowires were collected by blending for 2 min at a low speed. The mixture was then subjected to 8,000g centrifugation for 10 min to remove the cell debris. The supernatant containing nanowires was subjected to ultracentrifugation at 100,000g for 1 h to pellet OmcZ nanowires. The pelleted OmcZ nanowires were further resuspended in the appropriate amount of 15 mM pH 10.5 ETA buffer and loaded for sucrose gradient. The sucrose fractions were loaded accordingly with the following formula without the disturbance between each layer, and all the sucrose solutions were made with 15 mM pH 10.5 ETA:

2 ml 60% sucrose, 7 ml 50% sucrose, 12 ml 40% sucrose and 12 ml 30% sucrose.

No more than 600 μl of OmcZ sample was mixed with 20% sucrose to the final volume of 2 ml and layered on top of the gradient. The tube was then loaded to SW 32 Ti swing-bucket rotor (Beckman) and ultra-centrifuged at 100,000g overnight (>16 h). After ultracentrifugation, most OmcZ nanowires would stay at the density of ~30% to form a red ring. The red ring containing OmcZ nanowires was gently separated from the rest of the solution, and the sucrose was removed either through dialysis or buffer exchange against 15 mM ETA buffer. To further remove small protein contaminations, the sample was loaded to a gel filtration column packed with Sephacryl 500-HR resin (Cytiva) and OmcZ nanowires were collected at ~0.4–0.5 fraction of the volume. The final sample was concentrated and stored in 15 mM ETA buffer at pH 10.5. The sample purity was assessed by both Coomassie blue and haem staining.

Cryo-EM sample preparation conditions for OmcZ

A total of 0.125 M urea was added to the OmcZ sample 1 h before the sample preparation to disperse the OmcZ nanowires. Three-hundred mesh C-flat grids were first glow discharged with air inflow for 20 s, and then 4 μl of the sample was cast on the grids. The grids were then transferred to an FEI Vitrobot Mark IV operating at 22 °C with 100% humidity and blotted for 4 s with the force strength of 2 and plunge frozen in liquid ethane.

OmcZ data collection

Frozen grids were imaged on Titan Krios G3i+, operating at 300 kV equipped with a K3 detector (Gatan). Micrographs were recorded with super-resolution mode at 0.828 Å per pixel, with a total dose of ~60 e⁻ Å⁻² fractioned into 50 frames. A defocus range of 1–2.5 μm was used to collect ~14,500 images in total. Quantum energy filter with a slit width at 20 eV was applied to remove inelastically scattered electrons.

OmcZ data processing

All movies were first motion corrected with MotionCorr implemented in RELION 3.0.8 (ref. ⁵²) using all the frames. The contrast transfer function was estimated using GCTF v1.10 (ref. ⁵³). The particles were first manually picked to generate a reliable 2D classification, which was then used as the matching templates for auto picking implemented in RELION. Approximately 900,000 particles were selected with the box size of 384 pixels, and after 2D classification, ~280,000 particles were selected. All ~280,000 particles were first used to generate a 3D map by imposing the helical symmetry (helical rise ~57 Å, helical twist ~160°), and a featureless cylinder was used as a starting reference. The consensus map was then used as the reference to run 3D classification (K = 4) and final ~120,000 particles were combined and used for another round of refinement. Subsequent contrast transfer function refinement and Bayesian polishing improved the resolution to 3.5 Å (gold standard), where protein backbone and haem positions became evident and ready for model building.

OmcZ model building

The density map was first segmented using volume segmentation in Chimera 1.15 (ref. ⁵⁴) and the backbone of the monomer was manually traced against the density using Baton mode implemented in Coot. All eight haem molecules are manually placed into the density and adjusted in Coot with the haem geometry imposed. All the thioether bonds and the histidine coordination were manually created in PDB files and fixed for the downstream refinement. The monomer coordinates were firstly refined in Phenix with defined haem geometry and ideal bond length for the covalent bonds attached to the haems. The refined monomer coordinate was then expanded into trimers and the trimmer model was further refined in Phenix with NCS operator and imposed haem geometry. The refined model was further manually inspected and corrected in Coot. Then the model was subject to another round of real space refinement in Phenix, and the final model was selected based on the lowest MolProbity score.

Conservation map calculation

Conservation of the residues was calculated using the Consurf Server⁵⁵ and the PDB structure of monomeric OmcZ was used as a user-provided input. The homologous proteins of OmcZ were searched using HHMER against the Uniref90 database. The output result was further rendered in Chimera X⁵⁶.

Purification of OmcS nanowires from G. sulfurreducens

OmcS nanowires were obtained from G. sulfurreducens strain²⁹ CL-1 and nanowires were purified as described previously⁵⁰.

CD

Samples containing OmcS or OmcZ was first dialysed into 15 mM ETA buffer. The protein concentration was determined using both bicinchoninic acid assay and ultraviolet–visible (UV–vis) spectroscopy using the previously published extinction coefficient. CD scans of protein and buffer background were collected between 350 and 480 nm with a 1 mm or 2 mm path length quartz cuvette using a spectrometer (Chirascan, Applied Photophysics). Each buffer or protein sample was scanned with three replicates. All replicate spectra were averaged, and the averaged buffer background was subtracted from all averaged nanowire spectra. To convert the raw ellipticity values (millidegrees (mdeg)) to molar ellipticity values ((θ); deg cm² dmol⁻¹), the following formula was used:

$$\theta =\frac{\text{mdeg}\times \text{MW}}{l\times C}$$

where MW is the mean residue weight, l is the path length of the cuvette in centimetres and C is the concentration of the protein in mg ml⁻¹.

Molecular dynamics simulation for relaxing structures used in CD spectral computations

System setup and parameterization: Starting from the cryo-EM structure of an OmcZ nanowire (PDB accession code: 7LQ5), a trimer model was constructed from chains A, B and C. Each of these chains contains eight c-type, bis-histidine-coordinated haem co-factors, which were modelled in the oxidized (formal Fe(III) centre) state.

The trimer model has 45 Asp, 12 Glu, 51 His (only three of which are not coordinated to the iron centre of a haem co-factor), 21 Lys, 12 Arg and 12 Tyr residues that are titratable. These residues were assigned standard protonation states (for example, deprotonated Asp and Glu; singly protonated His; protonated Lys, Arg and Tyr) to simulate circumneutral pH conditions.

Using tLEaP in the AmberTools20 package^57,58, hydrogen atoms were added to the trimeric OmcZ structure assuming standard or full protonation of titratable residues. The structure was placed at the centre of a box of explicit water with at least a 15 Å buffer region to the boundary of the box. A sufficient number of counterions (45 Na⁺ at circumneutral pH) was added to achieve charge neutrality. Standard proteinogenic residues were modelled with the AMBER FF99SB force field⁵⁹. Parameters for the haem co-factor were adopted from Crespo et al.⁶⁰ and Henriques et al.⁶¹, and used as in previous studies for single- and multi-haem systems^62,63. The TIP3P water model⁶⁴ and the monovalent ion parameters of Joung and Cheatham⁶⁵ were used to model the solution state.

Pre-production simulations: The trimeric OmcZ structure was subjected to 20,000 steps of steepest descent, followed by a maximum of 100,000 steps of conjugate gradient minimization. A 10 kcal per mol Ų restraint was applied to the heavy atoms of the protein backbone, as well as selected atoms (PDB names FE, NA, NB, NC, ND, C3D, C2A, C3B, C2C, CA and CB) of each haem group during the minimization. The system was subsequently heated from 0 to 300 K at a rate of 0.3 K ps⁻¹ in the constant amount, volume and temperature (NVT) ensemble and held at the final temperature for 1.0 ns. The restraints on the protein backbone and haem groups were reduced to 1.0 kcal per mol Ų for the heating stage. These restraints were further reduced to 0.1 kcal per mol Ų for a subsequent 4.0 ns simulation at 300 K and 1 bar in the NPT ensemble to equilibrate the density of the system. This protocol was applied to prepare for the production-stage simulation under circumneutral pH conditions.

All NVT and NPT simulations (including the production-stage trajectories described below) employed periodic boundary conditions, the particle Mesh Ewald⁶⁶ treatment of electrostatic interactions with a direct sum cut-off of 10.0 Å, the SHAKE algorithm^67,68 to rigidify bonds to hydrogen atoms, a Langevin thermostat with a collision frequency of 2.0 ps⁻¹, and an integration timestep for the Langevin equation of motion of 2.0 fs. Pressure in NPT simulations was regulated with a Monte Carlo barostat having a relaxation time of 1.0 ps. Particle mesh Ewald molecular dynamics in its central processing unit and graphics processing unit⁶⁹ implementations in the Amber20 package⁵⁸ was used to perform the minimization and dynamical simulations, respectively.

Conventional (unbiased) production-stage simulations: The simulation at circumneutral pH were propagated for 72 ns, respectively, with conventional (unbiased) molecular dynamics in the NVT ensemble at 300 K. The above-mentioned protocol was also used to prepare a structural model of OmcS for comparison in the CD simulations.

CD spectra simulations

CD spectra were computed for the core of haem co-factors in OmcS and OmcZ. It is presently not possible to include more than a fraction of the entire protein at the DFT level of computation, and doing so actually worsens agreement with experiment (preliminary data not shown). The geometry of the haems was taken from the last frame of the production stage molecular dynamics simulation. We use a single frame to approximate the ensemble average that should ideally be used to compute the CD spectra because the geometry of the haem core was largely invariant under thermal fluctuations (that is, no large-scale protein motions occurred to appreciably change the distance and orientation of the haem co-factors).

For each haem in the trimeric models of OmcS and OmcZ, the C_α-C_β bonds of the ligated histidine residues were cleaved, and the dangling bonds were capped with hydrogen atoms to give bis-methylimidazole ligated co-factors. The C_β-S bond of the thioether linkages to the haems were similarly each cleaved and replaced with a capping hydrogen atom.

The CD spectra for these multi-haem systems were modelled at the time-dependent (TD)-DFT level using the B3LYP^70,71 functional and def2-SVP⁷² basis set for haems in the reduced (closed shell singlet) state. The reduced state was chosen because (1) the experimental sign-pattern of the CD spectra for OmcS and OmcZ was very similar in both oxidation states, and (2) the problem of spin contamination for open-shell species with TD-DFT⁷³ was avoided.

The CD spectra were computed using the fragment-based excitonic model^74,75 implemented in Gaussian 16 Rev. A.03 (ref. ⁷⁶) and analysed with the Excitonic Analysis Tool^77-79. The lowest two excited states with an energy no lower than 3.05 eV (406 nm) were computed for each co-factor. literature precedents⁸⁰ with TD-DFT calculations on haem systems revealed that the Soret transitions of interest for the CD analysis were always blue-shifted relative to experiment and reside in the 380–400 nm region. A plethora of low-intensity (probably spurious charge transfer) states were typically computed at lower energies (longer wavelengths). The number of excited states considered for each co-factor needed to be restricted for computational tractability because the excitonic model considers systematically the (Coulombic, overlap and exchange correlation) interactions between the transition densities for each excitation on every fragment. There are 18 or 24 haems in the OmcS and OmcZ models, respectively, giving a total of 612 or 1,104 evaluations of the interactions between transition densities on pairs of haems.

In comparing theory with experiment, it is important to note: (1) Interactions of the haems with the aqueously solvated protein environment have been entirely neglected. Intensity may be redistributed among the excitations in the presence of these interactions. Simulated intensity is also dependent on the chosen approximate DFT functional⁸¹, but we found no difference in the spectra when using B3LYP or CAM-B3LYP. (2) Simulated wavelengths are expected to be blue-shifted relative to experiment because vertical electronic (not 0–0 vibronic) transitions were computed, and inaccuracies exist in available DFT functionals. The blue shift of the theoretical versus experimental spectra for OmcS and OmcZ was an acceptable ~0.06 eV (ref. ⁸²) and consistent with literature on haem systems⁸⁰. Remarkably, the different sign patterns in the CD spectra for OmcS and OmcZ were found to be encoded in the haem arrangement alone (Extended Data Fig. 4d). These sign patterns result from haem-haem interactions, because the rotatory strength of a single haem group due to the relative orientation of the axial histidine ligands is minimal.

For more insight into the nature of haem–haem interactions responsible for the spectral differences, we analysed the evolution of the CD spectrum for OmcS and OmcZ as adjacent haems were sequentially added to the computations (Extended Data Fig. 4e). The optical response in the computed CD spectra only results when multiple haems are present, indicating the effect of haem–haem interactions. Differences between the simulated spectra of OmcS and OmcZ emerge with including as few as three haems in the computations.

Notably, the simulated spectra for OmcS and OmcZ were similar when six haems are present in the computations. This result for OmcZ is independent of which six of the eight haems are selected (Extended Data Fig. 4f). A substantial change in the simulated CD spectrum was found when a seventh haem (either the mainchain haem II or branched haem VI) was added to the computations. The addition of an eighth haem does not have an appreciable effect on the sign pattern. Thus, unique interactions specific to an assembly of seven haems are responsible for the different sign patterns between the simulated spectra for OmcS and OmcZ.

Substrate reduction test

All the substrates were freshly made in 20 mM Bis-Tris buffer at pH 7.2 followed by 1 h of degassing using a pure N₂ gas to scavenge the oxygen. Iron oxide reduction was performed using freshly precipitated β-FeO(OH) (akaganite) (110 mM). The Fe(III) concentration was measured with a ferrozine assay.

Reduction and oxidation of the cytochromes were monitored by UV–vis spectroscopy using a 1 cm path length cuvette. Briefly, 5 μM of OmcZ or OmcS sample was first dialysed into 20 mM Bis-Tris buffer at pH 7.2. Dialysed samples were then reduced with 10 mM of DTT in an anaerobic chamber overnight. Q bands are the UV–vis absorption of porphyrins observed in the range 500–750 nm, and their splitting corresponds to α- and β-bands³⁴. The α-band is commonly used to monitor substrate reduction³⁴. The successful reduction was confirmed by UV–vis spectra showing the splitting of Q band with no further change over time. The excessive amount of the substrate of substrates (final concentration ~1 mM) was directly added to the reduced nanowire samples. Upon addition, the UV–vis spectra were recorded in a timely manner. The recording stopped after a complete diminish of splitting of Q band. Control was made with degassed buffer to ensure no random oxidation happening within the period.

To calculate the area of α-band, the area under the curve between wavelength 543 nm and 562 nm was first calculated by integration. Then the baseline area was calculated by first fitting a straight line between these two wavelengths and the area was integrated under the line³³. Then the area of α-band was finally calculated as the difference between these two areas. The calculation was repeated for each timepoint, and all the calculations were implemented as a single Python script for automation.

Stability experiments

Purified OmcZ nanowires were subjected to adding denaturants to the final concentration of 6 M urea, 5 M GuHCl, 2% SDS and phosphoric acid at low pH (pH 1.6) respectively and incubated at room temperature overnight. Their stability was assessed by negative-stain TEM, CD spectroscopy and UV–vis spectroscopy (Extended Data Fig. 6). In addition, visual inspection confirmed that the red colour due to haems was retained under all denaturants.

Purification of OmcZ₅₀ from G. sulfurreducens

OmcZ₅₀ were obtained from G. sulfurreducens strain CL1-ΔomcS. Twenty litres of culture was grown until OD₆₀₀ ~0.6, and the cells were collected via centrifugation. Then periplasmic fraction was extracted following the previous established protocol³⁰. The periplasmic fraction was then loaded to a gel filtration column embedded with 500 S-HR Sephacryl resin (Cytiva) and running in 40 mM Tris buffer with 250 mM sucrose. The sample with peak corresponds to the non-filamentous form of OmcZ was combined and ran through an anion exchange column. The majority of OmcZ₅₀ did not bind to the column and was collected in the flow-through. Then the sample was further loaded to 200 high-resolution columns for purification. The sample with peak corresponds to a non-filament form of OmcZ was combined, and OmcZ₅₀ was the only OmcZ protein confirmed by western blot.

Expression and purification of OmcZ₅₀ in E. coli

A piece of genomic DNA coding for omcZ (gsu2076), ozpA (gsu2075) and prolyl isomerase (gsu2074) was amplified with Phusion DNA polymerase (Thermo Fisher) and primers OmcZ-F and 2074-R (Supplementary Table 2). Then the backbone of plasmid pCK32 (ref. ⁸³) was amplified with primers Vec-F and Vec-R (the amplified backbone included everything but the ppcA gene). The two PCR fragments were assembled with In-Fusion HD cloning kit (Takara Bio). The resulting plasmid was named pOmcZ-3.

Next, this plasmid was mutated to shorten the second haem-binding motif, which is non-conventional and required modification for the expression in E. coli. It was amplified with primers D12-F and D12-R, and the PCR product (the entire plasmid except for 36 bp in the haem-binding motif) was re-circularized with the same cloning kit. The resulting plasmid was named pOmcZ-4.

Finally, pOmcZ-4 was mutated to insert the Strep-tag II⁸⁴ at the very C-terminus of OmcZ. The plasmid was amplified with primers Str-F and Str-Rev and the PCR product was re-circularized in the same way. The new plasmid was named pOmcZ-9.

All plasmids were verified by sequencing at Keck Sequencing Facility, Yale University.

E. coli strain BL21 was used for all experiments. To express OmcZ₅₀ in E. coli, the frozen stock containing cytochrome maturation machinery⁸⁵ was transformed with pOmcZ-9 and then streaked on the plate supplied with both ampicillin (50 μg ml⁻¹) and chloramphenicol (37 μg ml⁻¹) and grew overnight at 37 °C. A single colony was picked and transferred to a small Terrific Broth culture (50 ml) with both antibiotics and cultured at 30 °C with ~150 r.p.m. shaking overnight (~16 h). The pre-culture was then transferred to a 6 × 2 l large Terrific Broth culture with 1:100 dilution and was first cultured to maximum OD (~5) with normal shaking speed (~110 r.p.m.) (~12 h) and then reduced to a lower shaking speed (~65 r.p.m.) and cultured for another 24 h for induction without any inducer.

To purify the protein, the cells were first pelleted by centrifugation (12,000g, 10 min) and the cells were lysed by sonication with 50 mM Bis-Tris buffer (pH 7.2) and 250 mM sucrose. Then the soluble lysate was separated from the insoluble part by ultracentrifugation (100,000g, 1 h). The soluble lysate was loaded to a Strep Trap XT column (Cytiva) and the bound protein was eluted with the gradient of biotin buffer (up to 50 mM) prepared in 40 mM ETA buffer with the addition of 250 mM sucrose. OmcZ₅₀ was then subject to gel filtration (Superdex 200 Increase HiScale 16/40, Cytiva) to further purify the sample.

OzpA expression and purification

OzpA gene and gsu2074 were cloned as one amplicon from pOmcZ-9 and an N-terminal StrepII tag and flexible linker with TEV cleavage site were added to ozpA while a 6× His tag was added to the C-terminus of gsu2074. Briefly, ozpA and gsu2074 was amplified with primers 20745_his_F and 20745_his_R (Supplementary Table 3). The backbone of pOmcZ-9 plasmid (without OmcZ, OzpA and GSU2074) was amplified with primers P_F and P_R. The signalling peptide of OzpA is fused to a StrepII tag with a flexible linker attached to a TEV cleavage site and it was directly synthesized as a DNA oligo with the sequence (5′-ATGAGGTATCTGCTCGCCGTTACGGCTCTTTTTATCCT GCTGCCTCCTCTGGGTGACGCTTTTGCCTGGAGCCATCCGCAATTT GAGAAGAATTCGAGCTCGAACAACAACAACAATAACAATAACAACAAC GAAAACCTGTACTTCCAGGGT -3′) and the overhang was added by PCR with primers TEV_strep_F and TEV_strep_R. All the fragments with the plasmid backbone were assembled to the expression plasmid by Gibson assembly.

The constructed plasmid was first transformed into E. coli DH5α strain and streaked on an LB plate with ampicillin. The correct construct was verified by sequencing.

E. coli BL21 expression strain was transformed with the plasmid containing the correct insertion of ozpA and gsu2074 with the supply of 200 μg ml⁻¹ of ampicillin. The expression condition was first screened with different concentrations of isopropyl-β-d-thiogalactoside (IPTG) as the inducer and two different growing temperatures (20 °C and 30 °C). On the basis of the screening result, 50 μM of IPTG with the induction temperature at 20 °C was used for downstream expression.

For expression, the frozen stock of BL21 strain with the correct plasmid was first streaked on the LB plate with ampicillin. A single colony was picked and subcultured at 30 °C overnight until confluent. Then the subculture was diluted 1:100 into three bottles of 2 l LB medium containing 200 μg ml⁻¹ ampicillin and grew at 30 °C with shaking at 110 r.p.m. 50 μM of IPTG was added once the OD reached ~1.8, and the temperature was lowered to 20 °C and induced overnight.

For purification of OzpA, the cells were first spun down at 10,000g and lysed in 40 mM Tris buffer at pH 7.5 with 50 mM of sodium chloride via a microbial fluidizer. The soluble part of the lysate was separated from the insoluble by ultracentrifugation at 100,000g for 45 min. The soluble lysate was loaded onto FPLC with a 5 ml StrepTrap XT column (Cytiva). The bound protein was eluted with a gradient of biotin up to 50 mM. The majority of the eluted proteins precipitated overnight, and the soluble part was separated from the insoluble by centrifugation. The soluble protein was then subject to gel filtration column to further purify the premature OzpA and dialysed against 20 mM sodium phosphate buffer (pH 7.2).

Premature OzpA underwent auto-maturation at room temperature overnight as seen in other subtilase⁸⁶. The mature protein was further purified via anion exchange column (HiTrap Q FF, Cytiva) and the purified mature protein was confirmed by SDS–PAGE gel and mass spectrometry.

In vitro digestion of OmcZ₅₀ for self-assembly test

Nine microlitres of purified OmcZ₅₀ (~10 μM) was incubated with 1 μl of purified OzpA enzyme (0.5 mg ml⁻¹) overnight at the room temperature. The digestion of OmcZ₅₀ was confirmed by western blot or haem staining and the self-assembled OmcZ nanowires were confirmed by TEM and immunogold labelling as previously described¹⁰ using two different peptide antibodies targeting OmcZ₃₀ and OmcZ₅₀ specifically. Both antibodies were synthesized by LifeTein by immunizing rabbits with synthetic peptide sequence (OmcZ₃₀: DSPNAANLGTVKPGL; OmcZ₅₀:KTINGPLSGKILAAPKKVKR).

Negative-stain TEM

CF400-CU grids were first plasma cleaned at low power setting for 40 s, and the samples containing nanowires or bacteria were drop cast on the grids and absorbed to the grids for 5 min. Then the excess liquid was removed by filter paper, and then the grids were floated on top of 50 μl droplet of 1% phosphotungstic acid (pH adjusted to 7) with the carbon side facing the liquid and stained for 2 min for protein sample or 30 s for bacterial sample. Then the excessive stain was removed by filter paper and the grids were allowed to be air dried before checking under TEM. TEM images were taken with JEM-1400 series (JEOL) microscope located on Yale West Campus with 80 kV operating voltage.

Power spectrum analysis

TEM images with the same pixel size were used for power spectrum analysis. e2helixboxer.py implemented in EMAN2 (ref. ⁸⁷) was used for picking particles along the nanowires. The boxed nanowires were further segmented into particles with 384 pixel size in SPRING⁸⁸. The averaged power spectrum was generated based on the picked particles by using segmentexam implemented in SPRING.

AlphaFold prediction of OmcZ and OzpA structure

AlphaFold suite^40,89 was downloaded and installed from https://github.com/deepmind/alphafold. For the prediction of OmcZ, the sequences were downloaded from the National Center for Biotechnology Information (GenBank ID: MRG77204.1 and CAG1009295.1). The signalling peptide was predicted by SignalP 6.0 online server and removed from the sequence based on the prediction. Then the sequences were used as input for AlphaFold prediction, and five models were generated for each sequence and the top-ranked models, based on the model confidence generated by the program⁸⁹, were used for all the subsequent analysis. Haems were then manually built inside the model structure based on the predicted haem binding motifs and a trimer model was built for each monomer structure. The haem arrangement in the models was obtained based on the position of cystines and histidines that form haem binding motif. A pymol command ‘pair_fit’ was used to insert haems in the AlphaFold structure that showed a good agreement with the cryo-EM structure. The root-mean-square deviation score was computed using the Local–Global Alignment method⁹⁰ from the online server AS2TS http://linum.proteinmodel.org/.

For the prediction of OzpA and OmcZ multimer, annotated catalytic domain of OzpA from Uniprot was used as input sequence for OzpA, and the full sequence of OmcZ without the signalling peptide was used as the input for OmcZ sequence. Then the multimer structure was predicted using AlphaFold multimer suite⁸⁹. Out of five predicted models, the top-ranked model was used for figure generation and interpretation. RNA structure of the intergenic region of mRNA between omcZ and ozpA was modeled using ViennaRNA suite Package 2.094.

Phylogenetic tree

OmcZ homologues were searched via protein basic local alignment search tool, and the top 100 hits were downloaded from the National Center for Biotechnology Information. The alignment of OmcZ homologues was performed in MEGAX⁹¹ using the MUSCLE algorithm⁹². Then the aligned sequences were further used to generate phylogenetic tree in MEGAX using maximum likelihood approach with bootstrap. The final tree was visualized and rendered in web-based Interactive Tree of Life⁹³.

RT–PCR

Log-phase WT G. sulfurreducens cells (OD 0.5) were used for RNA extraction. The cells were collected by centrifugation at 4,000 r.p.m. for 5 min from 5 ml of cell-containing medium. The pellet was resuspended in 200 μl of lysis buffer (100 mM Tris, 1 mM EDTA, 20 mg ml⁻¹ proteinase K and 1 mg ml⁻¹ lysozyme) in a 1.5 ml Eppendorf tube by pipetting. The tube was vortexed for 10 s and then incubated at room temperature for 10 min with a vortex every 2 min. After lysis, the total RNA was extracted, and clean-up was performed using RNeasy Mini Kit (Qiagen) following the instructions provided by the manufacturer.

Complementary DNA was obtained by reverse transcribing ~200 ng clean RNA using superscript II kit with random primers. The control cDNA was generated using the same protocol without adding the reverse transcriptase. Then the cDNA was used as the template to amplify the intergenic region between omcZ and ozpA using a forward primer (5′-CCAGCCCGGACTTTGCAACCAA-3′) and a reverse primer (5′-CGCCGTGCCGATAGACCTTCTC-3′). The expected size of the PCR product is 373 bp and was visualized on the 1% DNA agarose gel.

Native mass spectrometry for haem quantification

The proteins were buffer exchanged to 200 mM ammonium acetate (MP Biomedicals), 2 mM DTT with Zeba Spin Desalting Columns (Thermo Fisher Scientific). The protein concentration in the analysed sample was ~5 μM. OmcZ nanowires purified from G. sulfurreducens was monomerized by boiling in SDS and used as a positive control. Native mass spectrometry was performed on Q Exactive UHMR (Thermo Fisher Scientific) using in-house nano-emitter capillaries. The tips in the capillaries were formed by pulling borosilicate glass capillaries (outer diameter 1.2 mm, inner diameter 0.69 mm, length 10 cm; Sutter Instruments) using a Flaming/Brown micropipette puller (ModelP-1000, Sutter Instruments). Then the nano-emitters were coated with gold using rotary pumped coater Q150R Plus (Quorum Technologies). To perform the measurement the emitter filled with the sample was installed into Nanospray Flex Ion Source (Thermo Fisher Scientific). Mass spectrometry parameters for the analysis of the proteins include spray voltage 1.5 kV, capillary temperature 275 °C, resolving power 3,125 at m/z of 400, ultrahigh vacuum pressure 4.6 × 10⁻¹⁰ torr to 8.18 × 10⁻¹⁰ torr, in-source trapping between 100 V and 200 V. Data were visualized with the Xcalibur software and assembled into figures using Adobe Illustrator.

Using native mass spectrometry, we confirmed that the observed mass for OmcZ₅₀ is consistent with the successful incorporation of all eight haems and OmcZ₅₀ tends to dimerize. However, haem incorporation varied, probably due to differences in the cultures. Therefore, we validated haem incorporation for each culture. Then we performed the same digestion assay in vitro with OzpA on the OmcZ₅₀ with eight haems as described above. OmcZ purified from G. sulfurreducens was used as a control for the validation of the method.

Immunoblotting and antibody characterization

As described previously³⁰, custom polyclonal anti-OmcZ antibody was synthesized by LifeTein by immunizing two rabbits with synthetic peptide sequence (DSPNAANLGTVKPGL) containing targeted epitope on the native protein, OmcZ, and then affinity purifying the serum against that peptide sequence. The antibody was used at a dilution of 1:5,000 for immunoblotting. Filament preps were normalized to the initial cell mass of the starting material. Antibodies were designed to target the 30 kDa and 50 kDa domains separately, and their specificity was established using either western blot or immunogold labelling.

Statistics and reproducibility

All experiments were performed at least in triplicate and yielded similar results.

Extended Data

Extended Data Fig. 1 ∣. Illustration of cryo-EM map quality.

a–g, Density fitting of individual heme cofactors. Low iron density is due to the non-selection of coordinated histidine when rendering the figures at the given threshold of the density. i, Resolution distribution of the cryo-EM map. j, Representative density of sidechains. k, Map to map FSC. l, Map to model FSC. m, Real space correlation coefficient (RSCC) for each residue. n, Heme RSCC.

Extended Data Fig. 2 ∣. omcZ homologues show high sequence similarity.

omcZ homologues show up to 68% sequence similarity. In addition to all heme binding motifs, key residues are also highly conserved which we found to be critical for high conductivity. For example, a consecutive pair of histidines that brings T-stacked hemes closer is also highly conserved in all omcZ homologs. Thus significant sequence identities are present throughout the protein.

Extended Data Fig. 3 ∣. Circular Dichroism (CD) spectroscopy shows unique heme arrangements for OmcZ nanowires.

a, CD spectra OmcZ nanowire shows similar secondary structure at pH 10.5 and pH 7 buffer. b, CD spectra of OmcS and OmcZ nanowire hemes. Reduced spectrum show narrowing, and a red-shift of the peaks compared to the oxidized spectrum for c, OmcZ. d, OmcS nanowires as expected. e, The double positive peak feature is unique to the OmcZ nanowires, while the non-filamentous OmcZ₃₀ fails to show a such feature. f, Simulated CD spectra for various sextets of hemes from OmcZ showing that the line shape is independent of six hemes included in the computations. Simulated CD spectra of g, OmcS (green) and h, OmcZ (pink), using only hemes and heme binding motifs show the peak features observed in the experimental data (black). The simulated spectra were uniformly shifted to longer wavelength by 0.06 eV (see methods for details) i, Resolving the contribution of individual hemes to the CD spectra for OmcS and OmcZ reveals that the double peak is a specific feature of hemes in OmcZ nanowires. Evolution of the CD spectra as adjacent hemes in OmcS (green) and OmcZ (pink) were sequentially added to the computations. The labels in each panel indicate which hemes were added for the spectra. The sequence of addition is from left-to-right in both rows.

Extended Data Fig. 4 ∣. OmcZ nanowires reduce diverse electron acceptors.

a, Buffer Control: OmcZ nanowire spectra do not change after 1 hour upon addition of degassed deionized bis-tris buffer (20 mM, pH=7.2) used for all substrate reduction experiments. b, OmcZ nanowires could not be fully reduced by DTT but could be reduced by Sodium dithionite. c, OmcZ cannot reduce riboflavin.

Extended Data Fig. 5 ∣. OmcZ nanowires are robust electronic biomaterials that function under extreme environments.

Negative staining TEM images of OmcZ nanowires treated with following protein denaturants. a, No denaturant. b, 6 M urea, c, 5 M GuHCl, d, Phosphoric acid at low pH (pH = 1.6), and e, boiled in 2% SDS. All scale bars: 100 nm. f, OmcZ nanowires are more stable than OmcS at low pH (pH = 1.6) as evident by retention of its red colour due to hemes. OmcZ nanowires also retained their red colour characteristic of haems under all other denaturants. g, Solution CD spectra and h, UV-vis spectra show the stability of protein and electronic structure of OmcZ nanowires in denaturants.

Extended Data Fig. 6 ∣. In vitro Assembly of OmcZ nanowires from native OmcZ₅₀.

a, TEM image of native G. sulfurreducens OmcZ₅₀ showing lack of nanowire assembly. b, Western immunoblot using OmcZ antibody showing OzpA digesting OmcZ₅₀ into OmcZ₃₀. c, Digested OmcZ₃₀. self-assembles into nanowires in vitro. d, SDS-PAGE gel of purified OzpA showing three major bands: premature protein, mature protein, and the cleaved inhibitor domain (I9). e, Power spectrum of the reconstituted OmcZ nanowires shows the same helical parameters as cell-produced OmcZ nanowires. f, 2D average of reconstituted OmcZ nanowires. g, CD spectra of reconstituted and native OmcZ nanowires is similar. Immuno-gold labeling of reconstituted OmcZ nanowires using h, No-primary antibody, i, antibody for OmcZ₃₀ and j, antibody for OmcZ₅₀. Scale bars, a, 100 nm. c, 50 nm. c, 2 nm. h, 100 nm, i, 50 nm, j, 100 nm.

Extended Data Fig. 7 ∣. Native mass spectrometry of OmcZ₅₀ purified from E. coli shows molecular weight and heme groups consistent with native OmcZ₅₀.

a, Coomassie and Heme staining gel of OmcZ₅₀ purified from E. coli. b, Mass spectrum of OmcZ purified from G. sulfurreducens. Detected mass corresponds to the OmcZ monomer covalently bound to eight hemes. c-d, Mass spectrum of OmcZ₅₀ purified from E. coli. Detected proteins exist as monomer and homodimer. Mass analysis shows that OmcZ₅₀ monomer is bound covalently to primarily c, eight haems but sometimes also to d, six hemes.

Extended Data Fig. 8 ∣. OmcZ nanowire assembly could be environmentally controlled.

a, omcZ operon genes with intergenic region in G. sulfurreducens. b, RT-PCR shows that omcZ and ozpA are co-transcribed. Lanes 1: genomic DNA (gDNA), 2: cDNA, 3: cDNA control. c, Predicted RNA structure of the intergenic region of mRNA using ViennaRNA suite. Color code shows the probability of each base or base pair’s placement in the secondary structure.

Extended Data Fig. 9 ∣. OmcZ operon is found in diverse microbes.

Phylogenetic tree derived from OmcZ amino acid sequence alignments using homology cut-off score of 100 bit. Light blue circles represent bootstrap values.

Extended Data Table 1 ∣.

Statistics of OmcZ refinement and heme distances

a	OmcZ nanowire
	(EMD-23481)
	(PDB 7LQ5)
Data collection and processing	81,000
Magnification (kV)	300
Electron exposure (e-/Å)	50
Defocus range (μm)	1.0-2.5
Pixel Size (Å)	0.828
Symmetry imposed	−159.1 /57.4 Å
Initial particle images (No.)	900,000
Final particle images (No.)	128,000
Map resolution (Å)	3.4
FSC threshold	0.143
Map resolution range (Å)	3.3-5.4
Refinement
PDB code	7LQ5
Model resolution	3.6
FSC threshold	0.5
Map sharpening B factor	98
Model composition
Non-hydrogen atoms	6843
Protein residues	774
Ligands	24
B factors	NA
Protein	NA
Ligand	NA
R.m.s deviation
Bond length (Å)	0.56 (0)
Bond angles (°)	0.64 (0)
Validation
MolProbity score	2.5
Clash score	26
Poor rotamers (%)	2.6
Ramachandran plot
Favoured (%)	87.9
Allowed (%)	12.1
Disallowed (%)	0

b	stack	OmcS (Å)	OmcZ (Å)
	T-stack	6.1 ± 0.1	5.1 ± 0.3
	Parallel	4.3 ± 0.3	4.0 ± 0.3

a, Statistics of OmcZ refinement b, Hemes in OmcZ are closer than in OmcS as revealed by edge-to-edge distance. Values represent mean ± standard deviation (n = 4 T-stack heme pairs in OmcZ and n = 3 T-stack heme pairs in OmcS, n = 3 parallel-stacked hemes for both OmcS and OmcZ)

Data availability

The key relevant datasets generated during and/or analysed during the current study are publicly available. Cryo-EM data were deposited with the Electron Microscopy Data Bank (ID EMD-23481) and with the Protein Data Bank (ID 7LQ5). Source data are provided with this paper.