Here, we demonstrate the utility of the signal strapping method by identifying four metalloprotein families, each of which displays diverse 3-dimensional folds, distinct from that the classical LPMOs. For clarity, we refer to this method as "signal strapping" -- a term derived from "bootstrapping," reflecting the concept of initiating discovery using a minimal, adapted sequence.

First, we assessed whether signal-peptide-containing (SP) proteins were more likely to expose a metal-coordinating amino acid at the N-terminus after SP cleavage. In doing so, we hypothesised that cytosolic proteins with a free N-terminal histidine would chelate redox-active metals (e.g., Cu), potentially generating harmful reactive oxygen species via redox-cycling. As such, the expectation is that such proteins would be less prevalent than those that are secreted. To test this hypothesis, we analysed the SignalP 6.0 training set comparing amino acid frequencies at position two (post-methionine cleavage) versus position one of the mature protein (post-SP cleavage; Fig. 1d). This analysis shows that alanine is the most common residue adjacent to the N-terminus in both cytosolic (20%) and signal peptide-cleaved proteins (12%), followed by serine (18% and 9%). In contrast, cysteine (0.5%, 2.0%), histidine (0.7%, 2.6%), and tryptophan (0.6%, 0.6%) are the least common (cysteine and histidine are metal-chelating residues). Extending this analysis, Fig. 1e shows that cysteine, histidine, glutamine, proline, and tyrosine occur 3-4 times more frequently at the N-terminus of secreted proteins compared to cytosolic ones. Notably, the first three of these residues are known metal chelators, supporting our premise that signal peptide cleavage can reveal residues suitable for metal coordination.

Accordingly, we define "signal-strapping" (Fig. 1f) as follows: I - Select the sequence of a known secreted protein from an organism (e.g., cellulase) and identify the SP using SignalP 6.0 or use a consensus sequence likewise. II - Append a histidine (H) or dipeptide (HX, where X is any residue) to the C-terminus of the SP. III - Use the modified sequence in a blastp search of the NCBI or Uniprot databases, excluding the original organism. (Blastp at NCBI or Uniprot automatically adjusts search parameters for a 'short sequence'. No other greater adjustments were made to these parameters. Please, see the material and methods section for more details). IV - Review hits for conservation of N-terminal histidines and for the presence, although not necessarily conservation, of other potential metal-coordinating residues, e.g., histidine, cysteine, aspartate (an N-terminal histidine is capable of stably chelating a transition metal ion alone without further coordination from other amino acid residues). V - Confirm conservation of the histidine following predicted SP cleavage and classify the resulting sequence using InterPro and Pfam.

Using the signal strapping approach, we identified several candidate metalloproteins. Here, we focus on four representative examples (Table 1). Each protein was heterologously expressed, purified, and characterised for metal-binding properties using thermal shift assays (TSA) and spectroscopy. Crystal structures were also determined for selected representatives of the DUF4198 and Anglerase-1 families.

DUF4198 -- a secreted nickel-binding protein

We initially selected a signal peptide from a GH5 family protein (GenBank: OAS21937.1) from Pseudomonas putida and manually appended "HG" at the C-terminus (Table 1). GH5 proteins are widely secreted across both eukaryotes and prokaryotes, offering broad taxonomic coverage in downstream searches. This sequence was used in a standard blastp search against all organisms excluding P. putida. The search recovered mostly glycoside hydrolases from several bacteria, but also sequences of uncharacterised proteins. Amongst these was a DUF4198 domain-containing protein from Hydrogenophaga sp. IBVHS1 (WP_086125165.1) with an N-terminal histidine. This protein sequence was then used in a search which established that DUF4198 is found broadly distributed in bacteria (mostly Pseudomonadota) with some archaeal sequences (Supplementary Fig. 1a) and that these contain, following SP cleavage, a mostly conserved (79%) N-terminal histidine along with other potential metal-coordinating residues further in the sequences, exhibiting good levels of conservation (Supplementary Fig. 1b and Supplementary Data 1).

To infer the potential function, we examined the genomic neighbourhoods (GN) of DUF4198 using the EFI webserver. Adjacent genes included TonB receptors and NikR_C family transcriptional regulators, both linked to nickel metabolism (Supplementary Fig. 2a). This association is further supported by transcriptomic data showing DUF4198 expression within nickel and cobalt ATP-binding cassette transporter operons. Based on this evidence, we selected a DUF4198 protein from Ideonella sakaiensis (WP_082368692.1, "IsDUF4198") for recombinant expression. Though I. sakaiensis is known for plastic degradation, genes involved in that process (e.g., PETases, MHETases) are not linked to DUF4198 (Supplementary Fig. 2b).

Purified IsDUF4198 (Supplementary Fig. 3) was assessed for its metal-binding capacity using TSA. At a 1:1 protein:metal molar ratio, the greatest thermal stabilisation was observed with Ni(ΔT = + 7.0 °C, Fig. 2a), followed by Zn (+ 4.6 °C), Co (+ 4.0 °C) and Cu (- 0.2 °C). These relative shifts, save for Cu, are in accord with those expected from the Irving-Williams series. Further TSA analysis at multiple stoichiometries (200:1 and 10:1, Supplementary Fig. 4) and a measured dissociation constant for Ni²⁺ (K= 18 ± 3 nM; Supplementary Fig. 5) confirm nickel binding, consistent with the GN analysis described above. The potential role for DUF4198 in metal transport rather than as an active metalloenzyme finds further support from assays described herein (Methods), which showed only weak peroxidase activity of isolated Ni-IsDUF4198 (Supplementary Discussion and Supplementary Fig. 6).

The IsDUF4198 protein was co-crystallised with Ni²⁺ and its structure resolved at 1.4 Å resolution using molecular replacement (Supplementary Table 1). The structure reveals a β-sandwich fold composed of thirteen β-strands connected by flexible loops (Fig. 2b) (PDB: 9GCB). Structural comparison using the DALI server identified a cytoplasmic sulphur-carrier protein from Chlorobium limicola (PDB ID: 2NNC) as the closest match, though it shares only 15% sequence identity and lacks an N-terminal histidine (Z-score: 5.1).

The IsDUF4198 structure features a cleft measuring ~ 12 × 20 Å, which houses the metal-binding site (Fig. 2c, d). As anticipated, a Ni²⁺ ion is located within this positively charged cleft (Fig. 2e), coordinated by the N-terminal histidine and the τN-atom of a conserved His18 residue -- both of which are conserved across homologues (Fig. 2f). Nickel coordination is completed by a water molecule and a chelating malonate (from crystallisation buffer), resulting in near-octahedral coordination geometry (Supplementary Table 2). Unlike the canonical histidine brace found in LPMOs, the three nitrogen atoms of the coordinating histidine groups are arranged such that they facially cap one half of the metal ion coordination sphere (Fig. 2c, d). Unlike the canonical T-shaped histidine brace seen in LPMOs, the three nitrogen donors in IsDUF4198 form a facially capping arrangement around the Ni²⁺ ion (Fig. 2c, d). This coordination geometry likely underlies the protein's preference for Ni²⁺ over Cu²⁺ -- a notable inversion of the Irving-Williams series.

DUF6702 -- a metal sequesterase

Building on the discovery of Ni-containing DUF4198 proteins and their association with metal-dependent biochemistry, we selected the SP of the DUF4198-domain protein from Hydrogenophaga sp. IBVHS1 for signal strapping (Table 1). A blastp search (excluding Hydrogenophaga sp.) retrieved DUF4198 homologues and several uncharacterised proteins. From the latter, a DUF6702 family protein from Pseudoalteromonas sp. A757 (WP_128727020) was identified, containing an N-terminal histidine residue. Analysis of DUF6702 family members in the InterPro/Pfam databases revealed that 65% of sequences retain an N-terminal histidine after predicted SP cleavage, and potential metal-coordinating amino acids are present in the rest of the sequences, exhibiting variable levels of conservation (Supplementary Fig. 7a and Supplementary Data 2). This domain appears to be exclusive to bacteria, particularly within the Bacteroidota phylum (Supplementary Fig. 7b).

To investigate the potential biological function of DUF6702, we analysed its GN. While many neighbouring genes encode proteins of unknown function (Supplementary Fig. 8a), several encode M1 peptidases and HAD_2 domain-containing proteins, the latter is predicted to be haloacid dehalogenase-like hydrolases. M1 peptidases utilise different metal ions -- zinc, cobalt, manganese or copper -- within their active sites. Similarly, HAD superfamily enzymes (for example, phosphatases, phosphonatases, P-type ATPases, and beta phosphoglucomutases) also rely on metal ions for their function. This context supports the hypothesis that DUF6702 is metal-associated.

Given its likely metalloprotein nature, we selected a DUF6702 protein from pathogenic bacterium Pseudomonas aeruginosa (VZT40374; hereafter PaDUF6702) for recombinant expression in E. coli (Supplementary Fig. 9). The neighbouring genes are depicted in Supplementary Fig. 8b. TSA analysis at a 1:1 molar ratio revealed a large stabilising shift upon Ni²⁺ binding (ΔT = + 17.2 °C), with notable shifts also observed for Co²⁺ (+ 10.7 °C) and Cu²⁺ (+ 7.0 °C) (Fig. 3a and Supplementary Fig. 10). Dissociation constants (Supplementary Fig. 11) confirmed high affinities: Ni²⁺ (5.60 ± 0.1 pM), Co²⁺ (3.6 ± 0.2 nM), and Cu²⁺ (3.26 ± 0.11 nM). These findings support the designation of PaDUF6702 as a metal sequesterase with a particular affinity for Ni²⁺. Importantly, oxidase and peroxidase assays showed no significant activity for Ni-, Co-, or Cu-bound forms (Supplementary Fig. 6 and Supplementary Discussion), suggesting a non-enzymatic metal-binding or transport function.

Crystallisation of metal-bound PaDUF6702 was unsuccessful. Therefore, structure prediction was performed using AlphaFold3 (AF3). Unfortunately, AF3 is not parameterised for Ni-containing proteins, so as alternatives, five models (0-4) were generated for the apo, Cu-loaded, and Co-loaded states. All models exhibited high global pLDDT confidence scores (Supplementary Table 3). One of which, model 3, was selected based on metal ion structure parameters described in the supplementary results (Supplementary Fig. 12 and Supplementary Tables 4 and 5). The overall structure of the PaDUF6702 model displayed a beta-sandwich architecture with an immunoglobulin-like fold, composed of nine beta-strands connected by several loops (Fig. 3b). In the apo form, the second beta-strand connects with the third through two alpha-helices, whereas the Cu-bound form exhibits three alpha-helices, suggesting a structural change upon metal binding (Supplementary Fig. 12a-c).

The PaDUF6702 structure features a partially flat surface at the predicted metal-binding site, resembling that of some LPMOs and metal-binding carrier proteins (Fig. 3b). In both Cu- and Co-bound models, the predicted site comprises three histidine residues: His1, His28, and His32 (Supplementary Fig. 13a, b and Supplementary Discussion). The coordination involves the N-terminal His1, consistent with the expectations of the signal strapping approach. However, in the apo form, the side chain of His1 is modelled with low confidence (Supplementary Fig. 13c and Supplementary Discussion).

As no crystal structure was available, EPR spectroscopy was used to probe the metal coordination geometry of Cu-PaDUF6702 at pH 7.0. X-band CW-EPR of Cu-PaDUF6702 (Fig. 3d) gave spin-Hamiltonian parameters with g > g ~ g (g 2.050, g, 2.065, g 2.256 and A 550 MHz), consistent with a near square planar coordination geometry, supporting the selected structural model (Fig. 3c, d). Electrostatic surface potential (ESP) calculations of PaDUF6702 at pH 7.0 revealed a positively charged metal-binding site, indicating a potential docking site for a protein partner or substrate (Fig. 3e). ConSurf analysis further confirmed the conservation of His1 and adjacent residues across homologous sequences (Fig. 3f). These structural features are consistent with the DUF6702 family acting as a metal transport protein, insofar as the metal binding site can potentially directly interact with other proteins.

Ang-1 and Ang-2 -- metal "anglerases" appended to membrane-bound permeases

We applied the signal strapping method using the SP from a nickel superoxide dismutase (Pfam 09055, WP_230780104.1) of Roseiconus lacunae (Table 1). Nickel superoxide dismutases are known to coordinate Ni²⁺ via an N-terminal histidine. This search retrieved many Ni-superoxide dismutases alongside proteins of unknown function and cytochrome c and copper chaperones (CopC) proteins; the last of which is known for an N-terminal histidine that binds copper. In addition, further sequences annotated as putative hydrogenase/urease accessory proteins (HupE/UreJ-2) were found with an N-terminal histidine after the SP. These last domains are known nickel permeases.

We selected a representative HupE/UreJ-2 protein (Pfam 13795, InterPro IPR032809) from a Rhodobacteraceae bacterium (ETA49561.1) and analysed it using InterPro to evaluate the protein domain conservation. The search revealed that the query protein contains two separate domains: a non-cytoplasmic domain containing the N-terminal histidine (hereafter termed Ang-1), and a C-terminal transmembrane (TM) HupE/UreJ-2 domain. A BLAST search using this full sequence confirmed the two-domain architecture (Supplementary Fig. 14a). Further analysis with ConservFold showed 100% conservation of the N-terminal histidine following SP cleavage, along with highly conserved potential metal-coordinating amino acids in other parts of the sequences (Supplementary Fig. 14b and Supplementary Data 3).

In a separate search, we used the SP of a GH16 family hemicellulase (IPR000757; SUPfam 49899; WP_230780104.1) from Streptomyces sp. YIM 130001 (Table 1). In contrast to the known metalloprotein SP used for Ang-1, GH16s are broadly secreted across bacteria and eukaryotes. A BLASTp search (excluding Streptomyces sp.) returned GH16 proteins, metallopeptidase inhibitors (I36), cellulases, and again, proteins with a C-terminal HupE/UreJ-2 domain, with N-terminal domains distinct from Ang-1 (hereafter Ang-2). Like Ang-1, Ang-2 displays a non-cytoplasmic domain after the SP, followed by the TM HupE/UreJ-2 at the C-terminal (Supplementary Fig. 15a), and ConservFold results showed 100% conservation of His1 immediately following SP cleavage, along with highly conserved potential metal-coordinating amino acids in other parts of the sequences (Supplementary Fig. 15b and Supplementary Data 4).

As both Ang-1 and Ang-2 possess similar C-terminal HupE/UreJ-2 domains (42% of identity), we also performed full-length and only Ang domain sequence alignments (in Uniprot-CLUSTAL) to assess their similarity. The two full-length proteins share 34% of sequence identity (Supplementary Fig. 15c), while only the Ang domains share 24% identity, suggesting distantly related for the different Ang domains. Ang-1 proteins are found exclusively in bacteria, particularly within the Pseudomonadota phylum, while Ang-2 sequences are restricted to Actinomycetota (Supplementary Fig. 15d).

To gain insight into the biological functions of Ang-1 and Ang-2, we analysed their genomic neighbourhoods. For Ang-1, the most frequently co-occurring genes encode proteins of unknown function (Fig. 4a). DUF4198 domains (see above), Indigoidine A-like proteins (InterPro IPR022830), and PfkB kinases (Pfam 00294) were also prevalent. Indigoidine synthase proteins are known to exhibit glycosidase activity, requiring manganese in the active site. For Ang-2, proteins with unknown functions were again the most frequently encountered (Fig. 4b), followed by metallophosphatases and DUF916 domains. Metallophosphatases possess a conserved bimetallic active site (typically Mn, Fe, or Zn). DUF916 proteins (Pfam 06030) have been recently implicated as WxL-interacting proteins (WxLIPs), serving as scaffolds for anchoring WxL domains to bacterial cell wall peptidoglycan. For both Ang-1 and Ang-2, therefore, the gene neighbourhoods encode some metal-dependent proteins.

The Ang-1 domain of the HupE/UreJ-2 protein from R. bacterium (hereafter RbAng-1a), excluding its transmembrane (TM) region, was cloned for recombinant expression in E. coli (Supplementary Fig. 16). The corresponding gene locus and neighbouring genes are shown in Fig. 4c. Similarly, the Ang-2 domain from Microbacterium maritypicum MF109 (MmAng-2a), also lacking the TM region, was expressed in E. coli (Supplementary Fig. 17), with its genomic context shown in Fig. 4d.

TSAs were used to assess metal binding by Ang-1 and Ang-2. At a 1:1 molar ratio, the largest increase in melting temperature (ΔTm = + 2.1 °C) was observed with Cu²⁺, followed by Ni²⁺ (+ 1.2 °C). Other metals caused modest destabilisation: Fe²⁺ (- 0.9 °C), Zn²⁺ (- 0.5 °C), Mn²⁺ (- 0.4 °C), and Co²⁺ (- 0.4 °C) (Fig. 4e and Supplementary Fig. 18). Based on these results, RbAng-1a is designated a Cu-metalloprotein, though Ni²⁺ binding is also significant. Dissociation constants (Kd) determined by TSA were 73.6 ± 2.3 nM for Cu²⁺ and 220 ± 10 nM for Ni²⁺ (Supplementary Fig. 19), suggesting physiological relevance for both ions.

MmAng-2a was also assessed for metal-binding capacity (Fig. 2f). The largest thermal shift at 1:1 metal:protein ratio occurred with Cu²⁺ (+ 1.9 °C), followed by Fe²⁺ (+ 0.5 °C) (Supplementary Fig. 20). The dissociation constant for Cu²⁺ was 62.3 ± 2.8 nM (Supplementary Fig. 21), supporting its classification as a Cu-binding metalloprotein. Fe²⁺ binding may also be relevant, consistent with the metal dependencies of nearby genes in its operon.

Oxidase and peroxidase assays of metal-loaded RbAng-1a and MmAng-2a (M = Cu, Co, or Ni) revealed no substantial activity relative to LPMOs, though the Cu-loaded forms showed weak oxidase activity (Supplementary Fig. 6 and Supplementary Discussion). These levels are comparable to those of free Cu²⁺ and known weak LPMO oxidases, suggesting that the primary function of Ang-1 and Ang-2 is metal binding or transport, though a modest enzymatic role cannot be fully excluded.

RbAng-1a was crystallised in the presence of Cu and its structure resolved to 1.9 Å using molecular replacement with an AlphaFold2 (AF2) model (Supplementary Table 6). The protein adopts a beta-sandwich architecture with an immunoglobulin-like fold, composed of eight ß-strands connected by several loops (Fig. 5a) (PDB id:9GCE), somewhat resembling PaDUF6702 (RMSD of ~ 6.7 Å, Supplementary Fig. 22). The second beta-strand connects to the third via three alpha-helices. A DALI search identified the closest structural match as a zinc-binding protein from Vibrio cholerae (PDB ID: 8F1B).

The copper-binding site of RbAng-1a lies on a flat surface (Fig. 5a, b), where Cu⁺ is coordinated by the N-terminal His1 and the O atom of Asn41 in a T-shaped geometry (Fig. 5c). This coordination motif has been reported as a 'mono-histidine brace'. The RbAng-1a Cu-ligand bond lengths range from 2.1 to 2.7 Å (Supplementary Table 7), and the copper ion is three-coordinate. The geometry suggests a Cu oxidation state, following photoreduction of Cu in the X-ray beam. To verify that the site can also support Cu²⁺, we performed X-band CW-EPR spectroscopy at pH 5. The resulting spin-Hamiltonian parameters of g = 2.28 and A = 525 MHz and superhyperfine couplings from two Cu-N interactions A = 35, 42 MHz (Fig. 3d) are consistent with Cu-RbAng-1a binding Cu in a T-shaped coordination from the protein-based ligands (NO coordinating atoms), augmented by a coordinating ligand, likely an exogenous water molecule. This would yield an overall four-coordinate (NO) near-square planar geometry at the Cu (Fig. 5c).

We also modelled the full-length Cu-RbAng-1a-HupE/UreJ-2 complex using AF3 which achieved a high confidence prediction pLDDT score (> 90.0) (Supplementary Fig. 23a). The model closely aligns with the crystal structure (RMSD ~ 0.3 Å; Supplementary Fig. 23b), though discrepancies were noted in the coordination geometry at the metal ion. Whereas the AF3 model shows bond lengths of 1.9 to 3.4 Å with a pyramidal geometry at the Cu, while the crystal structure has 2.0 to 2.7 Å and a T-shaped geometry (RMS difference 0.14 Å, Supplementary Fig. 24). Furthermore, the AF3 prediction did not match the observed EPR-derived parameters. These findings underscore the value of AF3 for overall fold prediction while highlighting its limitations in accurately modelling metal coordination.

Crystallisation of Cu-loaded MmAng-2a was unsuccessful despite extensive screening. Therefore, structural modelling was performed using AF3 for both apo and Cu-bound states. Five models were generated for each condition, with high pLDDT scores (≤ 80). Model 4 (apo) and Model 1 (Cu-bound) were selected as the most reliable (Supplementary Table 8). The predicted MmAng-2a-HupE/UreJ-2 structure comprises a β-sandwich MmAng-2 domain and a HupE/UreJ-2 domain containing eight α-helices (Supplementary Fig. 25a).

Two candidate metal binding sites were identified, one in the MmAng-2 domain (His1/Asp33) (Supplementary Fig. 25b) and another in the HupE/UreJ-2 domain (His194, His201, His229) (Supplementary Fig. 25c). Notwithstanding the overall high confidence in the structure, however, some uncertainty in the His1 site of the MmAng-2 domain was evident with a low pLDDT score and lower inter-chain prediction TM-score (Supplementary Table 9). In addition, the MmAng-2a predicted metal binding exhibited bond lengths outside of normal ranges, suggesting again that AF3 metal coordination prediction must be interpreted cautiously.

Given this uncertainty, to validate Cu binding experimentally, EPR spectroscopy was conducted at pH 5.0. The resulting spin-Hamiltonian parameters for the Cu of g = 2.27 and A = 530 MHz, along with superhyperfine couplings from two Cu-N interactions A = 45, 35 MHz (Supplementary Fig. 25d), are consistent with the coordination geometry seen in RbAng-1a, showing that MmAng-2a is a bona fide metalloprotein, and also that the detailed modelling of the metal binding site by AF3 is likely inaccurate. ESP analysis of MmAng-2a at pH 7.0 revealed several charged side chains around the mono-histidine brace (Supplementary Fig. 25e), with negative charges dominating the metal-binding site. Consurf analysis showed that His1 and Asp33 are highly conserved among Ang-2 sequences (Supplementary Fig. 25f).

Based on structural, biochemical, and genomic data presented here, we propose that both Ang-1 and Ang-2 function as metal-binding domains directly linked to membrane-anchored HupE/UreJ-2 permeases. This architecture suggests their involvement in metal capture and assimilation in bacteria.

Using the validated AF3 model of Ang-1 (which aligns closely with its crystal structure), we modelled the full-length Ang-1- and Ang-2-HupE/UreJ-2 apparatus (Fig. 6). In both cases, the metal-binding domain faces the HupE/UreJ-2 domain in a compact "closed" conformation relative to the periplasm (Fig. 6a, b). This is consistent with the modelling context, which lacks solvent and membrane constraints, and thus steers the model to maximise protein-protein interactions, favouring close domain interactions.

A flexible linker region (residues 176-184 in RbAng-1a, 175-180 in MmAng-2a) connects the anglerase and HupE/UreJ-2 domains. Flexibility analysis (Supplementary Fig. 26) and normal mode simulations (Supplementary Fig. 27) revealed significant conformational dynamics. These models suggest that the anglerase domain can rotate away from the channel "mouth" of the permease (Supplementary Movie 1), exposing its N-terminal His residue to the surrounding milieu, from which it can sequester any adventitious transition metal ions. Indeed, both conformations are modelled by AF3, which further predicts low likelihoods of homo-multimeric complexes for both Ang-1- and Ang-2-HupE/UreJ-2 (Supplementary Fig. 28).

This open-closed transition resembles systems used for glycan acquisition in Bacteroidetes, where a periplasmic receptor feeds substrates into membrane channels, but until now, not proposed for metal ion uptake systems. Put more colloquially, the anglerase domains 'fish' for metal ions that are fed into the mouth of the permease, thus forming part of a mechanism for transition-metal ion capture and uptake by both in Gram-positive and Gram-negative bacteria. This model is consistent with previous knockout studies showing the essentiality of HupE/UreJ-family transporters in metal homoeostasis in nitrogen-fixing bacteria. In this context, we thus suggest that Ang-1 and Ang-2 are named anglerases.

Phylogenetic relationships among the herein identified metalloproteins

To explore the evolutionary context of the herein identified metalloproteins, we performed a phylogenetic analysis (Fig. 6c). Sequences containing DUF4198 form a distinct clade with a high bootstrap value (100), consistent with their divergent fold compared to the other protein families described herein. DUF6702 sequences also form a strong clade (bootstrap 92), but with a shorter phylogenetic distance to the anglerases, reflecting their shared β-sandwich architecture. In addition, a separate DUF6702 subgroup (bootstrap 91) likely represents a subfamily with a divergent fold. The anglerases cluster within four different clades, all with high bootstrap support (75, 99, 97, 92). This result indicates that anglerases, exemplified here by RbAng-1a and MmAng-2a, may constitute a broad and structurally diverse family of metalloproteins with variable metal binding preferences and selectivity.

Signal strapping is a protein-sequence search technique that identifies mature, secreted proteins that have amino acids at the N-terminus capable of chelating metal ions. The approach begins with a known signal peptide (consensus sequences could be used), which is manually appended with a metal-coordinating residue (e.g., histidine), and used to bootstrap a proteomic search. Following removal of the signal peptide, the appended residue becomes the first residue of the mature protein and can coordinate transition metals via its amino group and side chain (e.g., imidazole in histidine). As such, amino acid chains with an N-terminal histidine are likely metalloproteins.

This strategy is generalisable to other chelating N-terminal residues, such as cysteine and aspartate. Herein we exemplified this approach for the discovery and characterisation of four unknow metalloproteins. Structural and spectroscopic analyses confirmed that these proteins bind transition metals (Cu²⁺ or Ni²⁺) via N-terminal histidines. Genomic context further revealed that two of these families -- designated Ang-1 and Ang-2 -- are likely components of bacterial metal acquisition systems. These proteins, which we term anglerases, appear to "fish" for metal ions in the extracellular space and deliver them to membrane-bound permeases, highlighting a potentially widespread mechanism for metal ion uptake in bacteria.