Introduction

Molybdenum cofactor (Moco) biosynthesis is catalyzed by an ancient and highly conserved multi-step biosynthesis pathway [1], with the general steps of Moco biosynthesis being highly similar amongst eukaryotes and prokaryotes. Initially GTP is converted to cyclic pyranopterin monophosphate (cPMP), a reaction that – in all eukaryotes – was suggested to reside in the mitochondrial matrix [2]. Upon formation, cPMP is exported into the cytosol [3] where all subsequent Moco biosynthesis steps take place. In the second step of Moco biosynthesis, cPMP is converted to molybdopterin (MPT), the metal free precursor of Moco. Upon formation, MPT is used as a substrate by the molybdenum insertase (Mo-insertase). Here, notable differences exist regarding the domain organization of eukaryotic and prokaryotic Mo-insertases. Pioneering work identified E. coli molybdate utilization to depend on the two separate enzymes MoeA and MogA [4, 5], while the homologous domains are fused as a single polypeptide in most eukaryotes [6]. For consistency, in eukaryotes the prokaryotic nomenclature [4] has been retained as the MoeA homologous domain of eukaryotic Mo-insertases is referred to as E-domain while the MogA homologous domain is referred to as G-domain. Work with the plant Mo-insertase Cnx1 identified the G-domain to adenylylate MPT, yielding MPT-AMP (adenylated MPT [7,8], Fig 1) which is used as the substrate for the subsequent molybdate insertion reaction catalyzed by the E-domain. Both Cnx1 domains form a complex which was suggested to ensure the directed and protein protected MPT-AMP transfer from G- to E-domain [9]. Upon binding to the E-domain, molybdate is incorporated into the MPT dithiolene moiety [10]. This reaction, precisely the initial molybdate binding and its transfer to the active site bound MPT (dithiolene), requires a defined set of surface exposed residues which were first identified in the eukaryotic model Mo-insertase Cnx1E from the higher plant Arabidopsis thaliana (summarized in [6]). Consistent with the essential function of these residues for Cnx1 catalytic activity, a high degree of conservation of these residues has been reported amongst various eukaryotes [11]. Molybdate insertion into the MPT dithiolene moiety results in the formation of adenylylated Moco (Moco-AMP, [12]). Upon formation, the phosphor-anhydride bond within Moco-AMP is hydrolyzed and physiologically active Moco is released (Fig 1, [10,12]).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Functionalization of molybdenum.

(A). Cnx1E residues involved in molybdate binding to its initial binding site [11] are shown, directed interactions are indicated by dashed lines with distances in Ångström given above; numbering refers to the A. thaliana Mo-insertase Cnx1E (Q39054.2 [13]). (B) The Mo-insertase G-domain catalyzes the adenylylation of molybdopterin (MPT), yielding MPT-AMP. MPT-AMP and molybdate are the E-domain substrates and most likely, a minor backbone flip within the active site results in the movement of active site bound molybdate (A) into the MPT dithiolene moiety [11] yielding Moco-AMP. A minor rearrangement of the phosphor-anhydride bond in the Moco-AMP molecule is believed to initiate its hydrolysis, resulting in the release of physiologically active Moco [10,12].

https://doi.org/10.1371/journal.pone.0350191.g001

Thus formed Moco may be transferred to Moco dependent enzymes or contribute to the cell’s insertion-competent Moco pool (summarized in [14]). Moco is the active site prosthetic group of Moco dependent enzymes (Mo-enzymes) which catalyze a diverse set of vitally important redox reactions (summarized, e.g. in [15]). For plants, loss of nitrate reductase is most critical, as it is essentially required for plant survival [1,16]. From the mammalian Mo-enzymes, sulfite oxidase activity is most crucial, as its depletion results in severe neurological phenotypes and ultimately leads to death of the affected individuum [17,18]. Next to its importance for metabolic processes, Mo-metabolism is otherwise essential for mammals: Here the Mo-insertase gephyrin was first identified as a receptor clustering protein in the post synapse, but not as enzyme essential for the cellular Mo-metabolism [19,20]. Hence explaining the name of the protein (gephyrin, greek for bridge, [21]). Precisely, gephyrin is required for clustering of glycine- and GABA- (γ-aminobutyric acid, type A) receptors in the postsynaptic membrane of inhibitory synapses [19,20]. Other than the plant Mo-insertase Cnx1 which forms a compact, asymmetric hexameric complex [9], gephyrin is suggested to form a lateral network which is essential for receptor clustering [22,23] and references therein.

With very few exceptions known, eukaryotic Mo-insertases are generally assumed to possess E- and G-domains fused together in one protein, while for prokaryotes both domains exist as separate entities. To shed light on the domain organization of eukaryotic Mo-insertases in the present work we carried out an in silico-based approach to identify Mo-insertase sequences both in eukaryotes and prokaryotes which revealed a great number of sequences of hitherto not described putative Mo-insertases. Whilst prokaryotic Mo-insertases were identified to assemble from two separate domains, the vast majority of identified eukaryotic Mo-insertases assembled from two fused domains. Surprisingly exceptions were found in some invertebrate species, algae and protists, where both domains exist as separate entities.

All eukaryotic Mo-insertases identified within this work were found to possess a highly conserved surface patch which forms the active site. Catalytically important residues located here were found to be strictly conserved with few identified exceptions. Assessment of structural models of these exceptions suggests, that in six out of seven cases these will not impair functionality.

As an unexpected peculiarity, vertebrate type Mo-insertases were identified to possess a tremendously high degree of sequence conservation which – referring to the current available knowledge – is not explainable by its functions for Mo-metabolism and receptor clustering.

Materials and methods

Retrieval of Bait and Reference Sequences – In an initial setup, the protein sequence of Escherichia coli MoeA (WP_003903624.1) was used as query in a BLASTp search see Fig 2 for an overview.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Simplified taxon sampling scheme: Essentially, the MoeA protein sequence was used as query in BLASTp searches against the NCBI database (db).

Doing so revealed numerous prokaryotic and eukaryotic phylae to contain MoeA homologs from which the corresponding proteomes were downloaded and included into a new, combined db. *For vertebrate and invertebrates a random sampling approach was chosen which allowed it to include at least one representative proteome per phylum. **The taxonomic groups included into the combined db from ‘other eukaryotes’ are described in detail in the following section. A readily established plant db was likewise included into the combined db. If not stated otherwise, subsequent analysis was carried out exclusively with this established combined db. Species names and sources of the sequence data sets per species used for BLAST-based analyses are deposited in S1 Data File. Please see the following sections for details.

https://doi.org/10.1371/journal.pone.0350191.g002

Initially, BLASTp v2.15 searches were carried out, using the NCBI protein database (non-redundant protein sequences (nr) with default settings between March and June 2024. Initially, the top two hits (referring to the bit score of the NCBI BLASTp search result) for all bacterial phylae were identified. Next the annotated polypeptide sequences of the respective species were downloaded from the NCBI database [24]. In few cases, no annotated polypeptide sequences were available for the species of the top two hits obtained from BLASTp searches. Here, the first two hits for which annotated polypeptide sequences were deposited were considered. Thus obtained sequences were used for downstream analysis. For fungi, MoeA (WP_003903624.1) was used as query in BLASTp searches using the NCBI protein database (non-redundant sequences (nr), default settings) in the following taxonomic groups: Ascomycota, Basidiomycota, Dikarya incertae sedis, Blastocladiomycota, Chytridiomycota, Cryptomycota, Microsporidia, Mucoromycota, Nephridiophagidae, Olpidiomycota, Sanchytriomycota, and Zoopagomycota. Referring to the NCBI taxonomy browser [25] Ascomycota, Basidomycota, Chytridiomycota, Cryptomycota, Microsporidia, Mucoromycota, Olpidiomycota, Sanchytriomycota and Zoopagomycota are fungal phyla, while Nephridiophagidae is a fungal subclass ranked as phylum and Dikarya incerta sedis are unranked species. From each taxonomic group, again the top two hits (referring to the bit score of the BLASTp search result) were identified and the annotated polypeptide sequences were downloaded from the NCBI database [24] and used for downstream analysis. As we sought to use plants as a control group, we decided to use an available dataset comprising the annotated polypeptide sequences of in total 147 plant species. For animals, the annotated protein sequences of well characterized model organisms, i.e. Mus musculus, Gallus, Rattus norvegicus, Danio rerio, Xenopus laevis, Drosophila melanogaster, Caenorhabditis elegans and that of Homo sapiens were downloaded from the NCBI database [24] and used for downstream analysis. Upon identification of hits with highest sequence similarities amongst the analyzed jawed vertebrate (Gnathostomata) species, the number of jawed vertebrate species included into analysis was significantly expanded to cover various species from different clades aiming to obtain a comprehensive dataset. Therefore, an extension of the animal dataset was based on a BLASTp v2.16 analysis against nr with the same settings in January 2025. In total the analyzed vertebrate dataset encompassed 63 sequences from 59 geni and 63 species, respectively. The annotated polypeptide sequences of selected Gnathostomata species were downloaded from the NCBI database [24] and used for downstream analysis. As highest Mo-insertase sequence similarity was only identified amongst (jawed) vertebrate species, we sought to analyze Mo-insertase sequences from invertebrate species for comparison. As compared to initial taxon sampling, the number of invertebrate species included into analysis was significantly expanded to cover various species from the phyla Nematoda and Mollusca and the class Insecta [25]. Annotated protein sequences of selected invertebrate species were downloaded from the NCBI database [24] and used for downstream analysis. Finally, the MoeA sequence (WP_003903624.1) was used as query in BLASTp searches targeting eukaryotic taxons not covered by the hitherto described approaches, thus identifying MoeA related sequences in various protists and algae. From each taxonomic group [25], precisely Amoebozoa (clade), Ancyromonadida (clade), Apusozoa (class), Breviatea (class), Dephylleia (genus), Regifilida (order), Mantamonadidae (genus), Cryptophyceae (class), Discoba (clade), Glaucocystopheae (class), Haptophyta (phylum), Centroplasthelida (class), Malawimonadida (order), Metamonada (clade), Opisthokonta (class), Rhodophyta (phylum), Alveolata (clade), Rhizaria (clade), Stramenopiles (clade), Telonemia (genus), Kathablepharidaceae (order), Palpitomonas (genus), Virdiplantae (kingdom), Ancoracysta (genus), Picozoa (genus) and Hemimastigophora (phylum), the top two hits (referring to the bit score of the NCBI BLASTp search result) were identified and the respective, annotated polypeptide sequences were downloaded from the NCBI database [24] and used for downstream analysis.

BLAST – The Escherichia coli MoeA sequence (WP_003903624.1) was blasted against all downloaded polypeptide sequences using the script collect_best_BLAST_hits.py and default settings [26]. The obtained sequences were subsequently used for phylogenetic analysis.

Phylogenetic analysis – Generation of alignments was carried out using MAFFT v7.526 [27]. Phylogenetic trees were constructed with IQ-TREE2 v2.3.4 [28] using the maximum likelihood estimation and 1000 bootstrap replicates. The model used was LG + R8 (MoeA and ADH-trees). For the MOCS2B tree, the model used was JTTDCMut + R7. ITOL v6 was used to visualize the phylogenetic trees [29]. Patristic distances within clades were calculated using the Python script branch_length_comparison.py (https://github.com/bpucker/molyb) based on the dendropy module [30].

Elimination of contamination and splice variants in the dataset – After initial tree building, protein sequences from the same species, possessing highest sequence similarities, have been manually inspected for obvious annotation errors (eukaryotic and prokaryotic sequences) and splice variants (eukaryotic sequences). Given that protein sequences originating from multiple splice forms were identified, subsequent analysis was carried out with a single sequence per species. Here the top hit obtained from the BLAST using the specified script (see above) was used for further analysis.

Afterwards, the dataset was inspected manually a second time, to identify species harboring more than one protein sequence as identified by BLASTp analysis. Identified sequences were then checked for their origin to rule out contaminations using BLASTp. Thus identified contaminants were tabulated in S2 Table.

When more than one sequence per species remained in the data set upon application of the above-described regime, routinely all sequences from a single species were aligned to the Escherichia coli MoeA protein sequence (WP_003903624.1) and the Homo sapiens gephyrin sequence (NP_001019389.1) using MultAlin and the BLOSUM62-12-2 matrix [31]. Sequences were considered to be bona fide Mo-insertase E-domains if aligning to both, the E. coli and gephyrin E-domains. Application of this final control step revealed few sequences to be Mo-insertase G-domain like proteins (tabulated in S2 Table).

Mo-insertase domain classification – In order to identify E- and G-domain comprising regions in Mo-insertase fusion proteins, the following work flow was carried out: Upon initial identification, obtained sequences were routinely aligned with the G-domain encoding sequence of gephyrin ([22], invertebrate type Mo-insertases), Nit-9 [32], fungal type Mo-insertases) and Cnx1 [33], plant type Mo-insertases). Due to highest sequence conservation, for Gnathostome-type Mo-insertases these have been generally assumed to possess the domain organization identified for mammalian gephyrin [22].

Visualization of conservation grades – To visualize the degree of Mo-insertase surface conservation, the degree of amino acid conservation of vertebrate- / invertebrate- and plant-type Mo-insertases (see the results-section for details) was calculated to percent (https://github.com/bpucker/molyb). The conservation scores were subsequently written to the b-factor column of the A. thaliana Mo-insertase (PDB code: 6Q32) and the R. norvegicus Mo-insertase A-chain (PDB code: 2FU3), respectively. The modified R. norvergicus Mo-insertase A-chain has subsequently been duplicated to replace the B-chain in the figures shown by using PyMol [34]. For A. thaliana the second monomer was built by crystallographic symmetry. All structure visualization was carried out using PyMol [34].

3D structural prediction of proteins – Prediction of monomeric and dimeric Cnx1E models was performed using AlphaFold2 [35] and AlphaFold3 [36]. Modelling with AlphaFold3 was used to generate wt Cnx1E and active site variants S328T, S400T, K297R, K297L, K294R, and R369K. The G296A substitution was predicted to induce backbone rearrangements that could impact functionality (as described in [11]). To reveal any impact, the two deposited conformations of the G296-K297 segment from the Cnx1E crystal structure (PDB ID: 6ETF) were extracted and used as templates for structure prediction of the G296A variant using AlphaFold2 with default parameters [35]. Only the ‘relaxed’ backbone conformation present in 6ETF [11] yielded a viable template for modeling G296A. For structural comparison, Moco-AMP (PDB ID: 6Q32) [12] or molybdate (PDB ID: 6ETF) was superimposed from the solved Cnx1E structures onto the predicted models to visualize substrate/product positioning within the active site.

Analysis of catalytically important residues – Vitally important Mo-insertase active site residues were recently summarized for eukaryotic Mo-insertases [6]. The conservation amongst active site residues of various Mo-insertases reported here were checked using KIPEs v3 [37] (https://github.com/bpucker/KIPEs) and the bait input sequences as well as the analyzed residues are listed in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Reference sequences, residues and their positions for the inspection of active site amino acid residues in Mo-insertase candidates.

https://doi.org/10.1371/journal.pone.0350191.t001

Analysis of other conserved proteins (MOCS2B and ADH) – The Homo sapiens ADH (AAA19002.1) and MOCS2B (NP_004522.1) sequences were searched with BLAST against all downloaded polypeptide sequences using the script collect_best_BLAST_hits.py [26] and default settings. The top hit obtained from the BLAST was used for further analysis. The obtained sequences were subsequently used for the generation of alignments, using MAFFT v7 [27]. Phylogenetic trees were constructed with IQ-TREE2 [28] using the maximum likelihood estimation and 1000 bootstrap replicates. The model finder integrated in IQ-TREE2 identified LG + R8 as the best model. ITOL was used to visualize the phylogenetic trees [29].

Results

We used the E. coli MoeA sequence for BLASTp searches, which allows for the identification of putative Mo-insertases from both, eukaryotes and prokaryotes. The obtained sequences were used to create a phylogenetic tree comprising a total of 327 Mo-insertases from 289 geni (Figs 3, S1–S5, S5 and S6 Data Files). For better readability, we will in the following use the term Mo-insertase instead of putative Mo-insertase throughout this work.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Simplified phylogenetic tree obtained from maximum likelihood analysis.

The Mo-insertase families are color-coded as follows: Prokaryotic-type Mo-insertases (brownish), fungi-type Mo-insertases including protists (bluish), animal-type Mo-insertases (reddish), plant-type Mo-insertases including protists (greenish). For the detailed tree, please see S1–S5 Figs as indicated in the figure. For the sake of clarity, the prokaryotic clade has been collapsed. The dashed black lines are meant to group together Mo-insertases of different species based on their domain organization inferred within this work (see S2–S4 Data Files). As references, the domain organization of the Mo-insertases from Escherichia coli (MoeA: NCBI Reference Sequence NP_415348.1, MogA: NCBI Reference Sequence: NP_414550.1), Drosophila melanogaster (NCBI Reference Sequence: NP_726659.1, Rattus norvegicus (both annotations according to [13]), Caenorhabditis elegans (CAA90069 and CCD74267, annotation according to [38], Aspergillus nidulans (annotation according to [32]), Chlamydomonas reinhardtii (E-domain: GenBank entry DQ311646, G-domain: GenBank entry DQ311645.1, E-domain, [39]) and the model plant Arabidopsis thaliana (annotation according to [33]) is shown schematized. The dual function of the Mo-insertase gephyrin from vertebrates (i.e. clustering of γ-Aminobutyric acid type A and glycine receptors in the post synapse and functionalization of molybdenum [19, 20]) is indicated (see Fig 1 for a detailed representation of molybdate interacting residues). The number of sequences possessing the indicated domain organization is given within the respective colored box. Asterisks indicate species which possess a differing domain organization, summarized in S1 Table and S6 Fig. The tree bootstrap values are shown in S5 Data File, the alignment file is deposited as S6 Data File.

https://doi.org/10.1371/journal.pone.0350191.g003

As expected, Mo-insertases from different taxonomic lineages group into clades, which we annotated as animal-, fungi- and plant-type Mo-insertase family respectively. On the contrary, for prokaryotes, the diversity of the identified sequences did not result in forming family-level trees (S1 Fig).

The animal-type Mo-insertase family comprises 84 members, which in most cases possess an N-terminal G-domain fused to a C-terminal E-domain (Fig 3). Interestingly, exceptions were identified for the Mo-insertases from invertebrates namely Trichuris trichiura (T. trichiura), Caenorhabditis elegans (C. elegans) and Pinctada imbricata (P. imbricata), S1 Table. In these organisms, E- and G-domain were found to occur separately.

The fungi-type Mo-insertase family comprises in total 26 members including unicellular organisms from the SAR clade (i.e. Tribonema minus, Symbiodinium necroappetens and Reticulomyxa filosa) and the zooflagellate Tecamonas trahens. We found the vast majority of fungal Mo-insertases identified in this work to possess an N-terminal G-domain and a C-terminal E-domain (Figs 3 and S2). However, few exceptions were identified, summarized S1 Table and S6 Fig.

The vast majority of plant Mo-insertases is assigned to members of the Streptophyta phylum and possess an N-terminal E-domain fused to a C-terminal G-domain (see Fig 3 for comparison). A few Streptophyta were identified to harbor a separate E-domain (S1 Table) however, here no G-domain was identified which may be best explained by a provisional and/or incomplete annotation in the employed databases. Several unicellular organisms are summarized in S1 Table are part of the plant type Mo-insertase group, including the alga C. reinhardtii for which was reported, that here the E- and G-domains are expressed separately ([39], see Fig 3 for comparison). We identified this to hold true for all unicellular organisms included into the plant-type Mo-insertase group for which a G-domain sequence could be identified (S1 Table).

Having identified previously undescribed putative Mo-insertases in animals, fungi and plants, we next went on to confirm that these are indeed bona fide Mo-insertases. Therefore positional homologs to A. thaliana critical active site residues [6] were identified as described previously (detailed in the material and method section and [37]; Tables 1 and 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Conserved active site residues of the eukaryotic Mo-insertase E-domain. The catalytically important residues of A. thaliana Cnx1E (Cnx1 sequence NP_197599.1) have been tabulated. Positional homologs of invertebrates, vertebrates, fungi and plants have been identified and the number of identical residues as compared to the Cnx1 active site residues has been calculated as rounded percentage. Non-conserved residues are likewise included and the rounded frequency (percentage) with which these occur is given. In total, 21 invertebrate-type, 62 vertebrate-type, 26 fungi-type, and 155 plant-type Mo-insertases were considered. X = alignment gap.

https://doi.org/10.1371/journal.pone.0350191.t002

To elucidate the potential impact of the identified non conserved residues on Mo-insertase functionality we carried out an in silico-based approach in which the structure of respective Cnx1E variants was predicted using AlphaFold [35,36] (Fig 4). Modelling of all variants was possible (see also S9 Fig for comparison), however for variant G296A, only the ‘relaxed’ backbone conformation present in 6ETF [11] yielded a viable template for modeling.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. AlphaFold-Based modelling of Mo-insertase active site variants.

Cnx1 variants possessing the active site variations identified and tabulated in Table 2 were modelled using AlphaFold (as specified in the material and methods section). View of the active site structures of the modelled Cnx1 variants. Cnx1 wildtype (WT) and variants K297L/R, S328T, S400T: Moco-AMP is shown derived from the structural superimposition with protein structure 6Q32 [12]; Cnx1 variants G296A, K294R and R369K: Molybdate is shown derived from the structural superimposition of modelled variants with protein structure 6ETF [11]. The modelled active site structures are shown in ribbon representation; exchanged residues are depicted as sticks and colored blue. The outlines of the respective positional homologous wildtype residues were derived from the structural superimposition of modelled variants with the protein structure 6ETF [11] and are shown superimposed. Moco-AMP and molybdate are shown in ball and stick representation. Dashed lines indicate directed interactions between modelled and non-modelled active site residues with Moco-AMP and/or molybdate respectively, with distances given in Ångström (Å). Brackets indicate distances between modelled residues, while distance measurements involving wildtype residues are given without brackets.

https://doi.org/10.1371/journal.pone.0350191.g004

Within the animal-type Mo-insertase family, a large number of sequences groups together extremely close (detailed in Fig 5). Interestingly, these are all assigned to jawed vertebrates (Gnathostomata) where the Mo-insertase (named gephyrin here, [21]) is known to possess a neuronal function next to its function in Mo-metabolism.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Phylogenetic distance tree of animal-type Mo-insertases.

S3 Fig has been modified and is shown here. Mo-insertases identified in invertebrates possess sequence deviations which result in their arrangement within the phylogenetic tree allowing it to assign a phylogenetic affiliation, i.e. Nematodes, Insecta and Mollusca. Jawed vertebrates Mo-insertases do not arrange group-wise, hence preventing the assignment of any phylogenetic affiliation. Clades harboring plant and fungal orthologs have been omitted in this figure due to space constrains (see supplements for details). Numbers refer to the number of sequences forming the respective groups.

https://doi.org/10.1371/journal.pone.0350191.g005

We are fully aware, that this extremely close grouping – at first sight – may suggest corrupted database entries or subsequent processing errors. While this might appear as an artifact, all our investigations support the validity of these sequence records.

To exclude any influence of the phylogenetic age of the vertebrate trait on the degree of gephyrin’s sequence conservation, we next analyzed the patristic distances between species in the taxon Gnathostomata (jawed vertebrates, Fig 6) and compared these with values observed for taxa of similar age, documenting almost no changes in gephyrin during Gnathostomata evolution. To further substantiate this finding we next went on and constructed two additional protein phylogenies as controls, one for the Moco biosynthesis enzyme MOCS2B [40], (S7 Data File) and one for alcohol dehydrogenase ([41], (S8 Data File) and analyzed the respective patristic distances (S7 Fig). As can be seen from S7 Fig, the patristic distances of ADH sequences in the taxon Gnathostomata (1.67) are higher than those of gephyrin (0.04). The MOCS2B protein phylogeny revealed a recent splitting as two large metazoan protein families, i.e. Amphibia/Dinosauria and Mammalia were identified, while no Gnathostomata trait as observed for gephyrin / ADH were observed. Comparison of the patristic distances of both families again revealed a higher patristic distance, i.e. 0.94 for Amphibia/Dinosauria and 0.45 for Mammalia, respectively. Hence even within this – as compared to Gnathostomata (462 MYA, Fig 6) – younger taxa (i.e. ca. 300 MYA, Amphibia/Dinosauria and ca. 180 MYA, Mammalia) a significantly higher patristic distance was identified for an enzyme from the same biosynthesis pathway.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Average patristic distance between Mo-insertases in different taxa.

(A) Patristic distance of the Mo-insertases within indicated taxa. The estimated age of the compared taxa is indicated (mya = million years ago). Colors correlate with the patristic distance as indicated within the figure. (B) Simplified timetree visualizing the phylogenetic relationship of selected eukaryotic non-plant taxa. The numbers specify the estimated taxon-ages. Asterisks indicate the taxa compared in (A). (A) and (B): Estimates of taxon ages were extracted from the evolutionary time tree of life [42,43].

https://doi.org/10.1371/journal.pone.0350191.g006

In summary, our patristic distance analysis documents that the highest degree of gephyrins sequence conservation is not a result of the phylogenetic age of the vertebrate trait.

In the following known residues that are important for the two functions of gephyrin (i.e. receptor clustering and Moco synthesis) were identified and plotted onto the protein surface. As can be seen from Fig 7, a significant part but not the complete protein surface is associated with these functions while the overall surface conservation was found to be very high (S8 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Functionally relevant residues of the gephyrin E-domain.

Surface representation of the mammalian Mo-insertase E-domain (PDB code: 2FU3). (A) Residues involved in receptor binding (according to [44]) are shown in blue. (B) Interfacing residues of the interaction 2 model of the plant Cnx1 Mo-insertase complex [9] are shown in green. (C) Residues involved in molybdenum functionalization of the plant Mo-insertase Cnx1 are shown in red (recently summarized in [6]).

https://doi.org/10.1371/journal.pone.0350191.g007

Discussion

Within this work we carried out an in silico-based approach to identify Mo-insertases in the species of life. Our work revealed in total 327 Mo-insertases from 289 (in total) eukaryotic and prokaryotic geni. The finding that for prokaryotes the diversity of the identified sequences did not result in forming family-level clades may be best explained by the prevalent horizontal gene transfer here [45] leading to conflicting gene histories [46]. However, as known for the model Mo-insertase MoeA from E. coli, the prokaryotic Mo-insertases identified all assemble from a separate E- and G-domain.

Relating to the domain organization, our work classified all novel eukaryotic Mo-insertases identified either as plant-type (N-terminal E-domain, C-terminal G-domain) or animal/fungi-type Mo-insertases (N-terminal G-domain, C-terminal E-domain), with the notable peculiarity of Mo-insertases identified in some protists, algae and invertebrates which – like prokaryotes – all possess the E- and G-domain as separate entities (as documented by 190 manually carried out and inspected pairwise sequence alignments (see S2–S4 Data Files) and when indicated in silico based complementary structure predictions (see S1 Table).

Very recently, Megrian and colleagues [47] described the fusion of E- and G-domain to be present in all studied animal type Mo-insertases which they suggested to be due to the finding that this is mandatory for gephyrin’s neuronal function (i.e. receptor clustering in the post synapse of inhibitory neurons). Our study revealed that in some animals (i.e. C. elegans (see also [38]), T. trichuria and the mollusk P. imbricata, see S1 Table for comparison) the Mo-insertase E- and G-domains exist as separate entities, hence the tendency of E- and G-domain to occur fused together in animals requires a careful review. As the invertebrate-type Mo-insertases reported within this work essentially lack conserved E-domain residues required for receptor interaction (S10 Fig), we reason, that any vertebrate-like neuronal function of invertebrate-type Mo-insertases is unlikely, which is consistent with the lack of critical residues for receptor clustering in C. elegans and Drosophila Mo-insertases [48].

Why do eukaryotes possess inconsistent E-G domain arrangements? Other than eukaryotes, all prokaryotes studied thus far possess the identical domain arrangement: E- and G-domains are separate entities, documenting that domain fusion is not mandatory for Mo-insertase functionality (here). Interestingly, when recombinant plant Cnx1E- and G-domains are expressed separately, the MPT-AMP transfer from G- to E-domain as well as the subsequent metal insertion reaction occur under fully defined in vitro conditions when both enzymes are co-incubated [10,12,33]. Further, the Chlamydomonas reinhardtii E-domain was found to complement an E. coli moeA mutant strain [49]. These findings point towards, that i) eukaryotic Mo-insertase functionality does not mandatorily depend on domain fusion and ii) metabolite transfer from G- to E-domain likely underlies a common principle in eukaryotes and prokaryotes (see also [49]). The reason for domain fusion in eukaryotic Mo-insertases may be best explained by assuming that this allows for a directed [9] and efficient MPT-AMP transfer from G- to E-domain [50], which may be beneficial for the vast majority of eukaryotic species reported within this work (see also [47] for comparison). It is not clear, when the E- G- domain fusion occurred, however it appears to have occurred independently at least twice during the evolution of eukaryotes, documented by the finding that plants possess an inverted domain orientation as compared to animals and fungi [32,51]. While extensive data sets support findings regarding domain fusion, significantly less eukaryotic sequences were identified that encode for separate E- and G-domains. We suggest that the identified separate E- and G-domains in C. elegans, T. trichuria and the mollusk P. imbricata (see also S1 Table) may result from the division of the fused domains during speciation. The other possible explanation – an independently occurred domain fusion during speciation of the various animal species – appears unlikely. Taken together, the current view on eukaryotic Mo-insertase domain arrangement may be refined as obviously domain fusion is not essential for both, in vivo and in vitro functionality. However, the findings about domain fusion prevalence and evolutionary conservation are based on the sequence data that is currently available and may be improved as more genome sequences and derived polypeptide sequences (particularly from invertebrates) become available.

According to our analysis, a number of catalytically significant residues for Mo-insertases of the plant, fungal, and invertebrate types are conserved across species, which is in line with their suggested functional roles. For the exceptions identified we carried out AlphaFold based modelling and subsequently carried out structural plausibility assessments to reveal any impact of these residues on functionality. Doing so (see Figs 4 and S9) identified alterations S400T and S328T to most likely not impact functionality, as the Ser OH-group is essentially ‘replaced’ by the threonine OH-group within the active site and no obvious impact of threonine’s methyl group on active site chemistry could be deduced. Further, our assessment of the respective modelled structures suggests that the exchange of Lys 297 to Arg or Lys 297 to Leu will not have any impact on the interaction with the Mo-center: It is the amide nitrogen atom which is involved in a directed interaction here [11]. However in Cnx1E variant S269D D274S (co-crystallized with Moco-AMP, [12]), the side chain of K297 is involved in a single, directed interaction with the Moco-AMP pterin part. For variant K297R our modelling revealed that a directed interaction likely will occur, while our modelling data suggests, that this will not be the case for variant K297L. However, obviously this does not impair functionality, as all vertebrate type Mo-insertases and 95% of the invertebrate type Mo-insertases reported here (Table 2) possess the K297L variation. Cnx1E residues K294 and R369 are involved in (initial) molybdate binding [11,12] to the active site. Our assessment of the respective modelled structures suggests, that the exchange of Lys 294 to Arg or Arg 369 to Lys will very likely have no impact on the (positive) charge of the binding site which leads us to the assumption that binding of molybdate is most likely not impaired here. Our assessment of the modelled structure(s) suggests that solely the G296A variant likely will possess an impaired functionality: The Cnx1E G296-K297 segment exists in two conformations [11] from which variant G296A cannot adopt the ‘tensed’ conformation as documented by the finding that modelling with the respective template was not possible. As (initial) molybdate binding is linked to the ‘tensed’ conformation [11], we conclude that G296A is impaired in molybdate binding which most likely will exclude the subsequent molybdate insertion. Taken together, the current view on eukaryotic Mo-insertase functionality may be refined as our study suggests, that six out of seven functionally essential residues could be replaced by (type conserved) residues putatively without impairing functionality.

From all Mo-insertases identified and characterized thus far, vertebrate type Mo-insertases possess an extraordinarily high degree of sequence conservation (see also S8 Fig). As other vertebrate enzymes such as ADH and MOCS2B possess a significantly lower degree of sequence conservation (documented by the respective patristic distances, see Fig 6), we conclude that the evolutionary rate of the bifunctional gephyrin is lower. We note, that gephyrin residues involved in its metabolic (Mo-functionalization and complex formation) and receptor-clustering function locate nearly exclusively to the ‘front-side’ side of the enzyme, while merely no known functional relevant residue(s) locate to the ‘back-side’ (Fig 8).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. Functional relevant residues of the gephyrin E-domain.

All functional relevant residues (clustering, interaction, synthesis) shown individually in Fig 7 are shown in a combined image here. Please see legend of Fig 7 for details. Residues shown in yellow result from an overlap of the functional relevant residues.

https://doi.org/10.1371/journal.pone.0350191.g008

We speculate that putatively the remaining conserved surface area (see also S8 Fig) is crucial for other known evolutionary conserved protein interactions and/or yet unidentified post-translational modifications (summarized, e.g. in [52]). However, as hitherto to the best of our knowledge no functional relevant gephyrin ‘backside’ residue is known, it will require systematic mutagenesis work to attribute function(s) of gephyrin’s ‘backside’ residues to its functionality.

Supporting information

Acknowledgments

This work was supported by the de.NBI Cloud within the German Network for Bioinformatics Infrastructure (de.NBI) and ELIXIR-DE (Forschungszentrum Jülich and W-de.NBI-001, W-de.NBI-004, W-de.NBI-008, W-de.NBI-010, W-de.NBI-013, W-de.NBI-014, W-de.NBI-016, W-de.NBI-022). We acknowledge support by the Open Access Publication Funds of Technische Universität Braunschweig.