FAD/NADH Dependent Oxidoreductases: From Different Amino Acid Sequences to Similar Protein Shapes for Playing an Ancient Function

Flavoprotein oxidoreductases are members of a large protein family of specialized dehydrogenases, which include type II NADH dehydrogenase, pyridine nucleotide-disulphide oxidoreductases, ferredoxin-NAD+ reductases, NADH oxidases, and NADH peroxidases, playing a crucial role in the metabolism of several prokaryotes and eukaryotes. Although several studies have been performed on single members or protein subgroups of flavoprotein oxidoreductases, a comprehensive analysis on structure–function relationships among the different members and subgroups of this great dehydrogenase family is still missing. Here, we present a structural comparative analysis showing that the investigated flavoprotein oxidoreductases have a highly similar overall structure, although the investigated dehydrogenases are quite different in functional annotations and global amino acid composition. The different functional annotation is ascribed to their participation in species-specific metabolic pathways based on the same biochemical reaction, i.e., the oxidation of specific cofactors, like NADH and FADH2. Notably, the performed comparative analysis sheds light on conserved sequence features that reflect very similar oxidation mechanisms, conserved among flavoprotein oxidoreductases belonging to phylogenetically distant species, as the bacterial type II NADH dehydrogenases and the mammalian apoptosis-inducing factor protein, until now retained as unique protein entities in Bacteria/Fungi or Animals, respectively. Furthermore, the presented computational analyses will allow consideration of FAD/NADH oxidoreductases as a possible target of new small molecules to be used as modulators of mitochondrial respiration for patients affected by rare diseases or cancer showing mitochondrial dysfunction, or antibiotics for treating bacterial/fungal/protista infections.


Flavoprotein Dehydrogenases
Several oxidative pathways depend on the ability of cells to oxidize NADH (reduced form of nicotinamide adenine dinucleotide cofactor), FADH 2 (reduced form of flavin adenine Cofactor-binding regions were highlighted for comparative purposes by selecting residues within 4 Å from the cofactor crystallized in the sampled structures. When necessary, cofactors were also inserted in some crystal structures by comparative analyses and superimposition by using PyMOL according to what previously reported [27,30,[34][35][36][37][38][39].

Phylogenetic Analysis
The analysis of the evolutionary relationships among the homologous sequences sampled by Blastp or among the sequences of the available homologous crystallized structures sampled by Folding Recognition was conducted using MEGA5 [40] starting from the MSA of the above cited 120 homologous protein sequences and from the MSA of the above cited 49 crystallized structure sequences.
In detail, the two trees were built from the ungapped MSA applying the maximum likelihood method with the JTT model for the amino acid substitutions and a gamma distribution (five discrete gamma categories) for the rates among sites. A total of 100 bootstrap samplings were applied to test the robustness of the tree (similar to what was described in [30]).

Evolutionary Relationships among the Sampled AIF/NDH-2/NDI Homologous Sequences
Although it is commonly accepted that all NDH-2/NDI dehydrogenase sequences might share a common ancestral sequence, based on their conserved similar function and overall structure [13], it has recently been proposed that AIF has a function similar to NDH-2/NDI, i.e., that it is a rotenone-sensitive NADH:UQ oxidoreductase [11,12].
However, the spread of these oxidoreductases in distant taxonomic groups of species and their evolutionary relationships are currently unknown. Thus, we searched for proteins sharing sequence similarity with AIF or NDH-2 or NDI in representative species selected from bacteria, Plants, Fungi, and animals, and we found 120 putative homologous proteins. The evolutionary relationships among all these proteins showed two main clusters referring to AIF-homologous proteins and NDH-2/NDI-homologous proteins, whereas the latter were further divided into two main subclusters ( Figure 1). AIF-homologous proteins were detected in members from all four taxonomic groups while NDH-2-homologous proteins were detected in bacterial and fungal species, and NDI-homologous proteins were detected in fungal, plant, and animal species (Figure 1).
In detail, close to the Mammalia AIF sequences, it is possible to observe a group of Nitrosomonadales and Thermales sequences that group into a subcluster adjacent to the AIF subcluster Metazoan sequences. Then, in the same tree-region, a bit more distant but related to the AIF-like sequence cluster, it is possible to observe a sequence cluster hosting Bacillales, Sulfolobales, and Plant sequences (i.e., XP_022741237.1_Durio_zibethinus, Figure 1).
On the other side of the tree, it is possible to observe a cluster of Bacillales NDH-2-like sequences clustering together with NDH-2-like sequences from Fungi together with two subgroups of Bacteria sequences (the first consisting of Enterobacteriales, Rhodospirillales, Rhodobacterales, and Rhizobiales, and the second consisting of Thermales, Flavobacteriales, and Cytophagales sequences) related to NDH-2-like Bacillales sequences. Finally, a last subgroup containing Fungi, Plant, and Anthozoa NDI-like sequences is detectable, in the same tree region but separated from the NDH-2-like sequence cluster ( Figure 1). NDI-homologous proteins were detected in fungal, plant, and animal species (Figure 1).
In detail, close to the Mammalia AIF sequences, it is possible to observe a group of Nitrosomonadales and Thermales sequences that group into a subcluster adjacent to the AIF subcluster Metazoan sequences. Then, in the same tree-region, a bit more distant but related to the AIF-like sequence cluster, it is possible to observe a sequence cluster hosting Bacillales, Sulfolobales, and Plant sequences (i.e., XP_022741237.1_Durio_zibethinus, Figure  1).

Sequence Features of the Sampled AIF/NDH-2/NDI Protein Sequences
It was expected that the proposed specialization in species-specific biochemical pathways should reflect the existence of further well-conserved sequence features observable by building a MSA, despite the global highly variable percentage of identical residues (in the 15% to 99% range) shared by the sampled full-length sequences (Supplementary Table S1).
In fact, by inspecting the MSA of AIF/NDI/NDH-2-sampled homologs, it is possible to recognize 10 amino acid conserved motifs. Some of those motifs, i.e., the two glycine-rich motifs (respectively the first and fifth sequence motif grouped in Figure 2) and the eighth motif, are conserved in all the investigated sequences, underlining a putative common involvement in the protein function/mechanism. The other sequence motifs show a high degree of amino acid conservation within specific taxonomic groups, reflecting a species-specific acquired function or substrate/cofactor specificity/affinity ( Figure 2).

Figure 2.
AIF/NDI/NDH-2-conserved sequence motifs. Sequence motifs detected by comparing the sampled AIF/NDI/NDH-2 homologous sequences sampled by blastp. The highlighted motifs reveal crucial protein regions involved in cofactor binding and/or the protein function mechanism. The jalviewzappo color style was used for coloring amino acids. Logo representation is also reported to highlight the most conserved residues.
In detail, the most conserved protein regions, at the first and the fifth motifs, consist of two glycine-rich regions strongly conserved in all the sampled sequences. As an example, in metazoa, the two glycine-rich motifs are at the level of residues 137-IGGGTAAFA-146 and 306-IGGGFLGSE-314 (H. sapiens numbering) and align with residues 9-LGAGYGGIV-17 and 161-IGGAGFTGE-169 in Bacteria (C. thermarum numbering) and 59-LGSGWGAIS-67 and 234-VGGGPYGVE-242 in Fungi (S. cerevisiae numbering) ( Figure 2). It is interesting to note that the first highlighted motif hosts an aromatic residue in half of the sampled sequences, whereas the sixth motif always hosts an acidic residue and a further aromatic residue in most of the sampled sequences.
The second highlighted sequence motif, located at residues 162-VSEDPELPYMRPPLSKE-178 (H. sapiens numbering), close to the first glycine-rich motif, appears to be conserved in specific taxonomic groups within the sampled Mammalia AIF sequences. This motif appears to be conserved, with few differences in the AIF-like Figure 2. AIF/NDI/NDH-2-conserved sequence motifs. Sequence motifs detected by comparing the sampled AIF/NDI/NDH-2 homologous sequences sampled by blastp. The highlighted motifs reveal crucial protein regions involved in cofactor binding and/or the protein function mechanism. The jalviewzappo color style was used for coloring amino acids. Logo representation is also reported to highlight the most conserved residues.
In detail, the most conserved protein regions, at the first and the fifth motifs, consist of two glycine-rich regions strongly conserved in all the sampled sequences. As an example, in metazoa, the two glycine-rich motifs are at the level of residues 137-IGGGTAAFA-146 and 306-IGGGFLGSE-314 (H. sapiens numbering) and align with residues 9-LGAGYGGIV-17 and 161-IGGAGFTGE-169 in Bacteria (C. thermarum numbering) and 59-LGSGWGAIS-67 and 234-VGGGPYGVE-242 in Fungi (S. cerevisiae numbering) ( Figure 2). It is interesting to note that the first highlighted motif hosts an aromatic residue in half of the sampled sequences, whereas the sixth motif always hosts an acidic residue and a further aromatic residue in most of the sampled sequences.
The tenth motif consists of residues 482-FWSDLGPDVGYEA-494 (H. sapiens numbering) that aligns with a similar motif conserved with few variations in NDI-sequences from Fungi 436-FKPFKYNDLGALA-448 (S. cerevisiae numbering) and NDH-2 sequences from Bacteria 339-MTPFKPHIRGTVA-351 (C. thermarum numbering). Notably the last alanine residue is conserved in all the sampled sequences. Given the existence of the crystallized structures of the human AIF (4bur.pdb), S. cerevisiae NDI (4g73.pdb), and the C. thermarum NDH-2 (5kmr.pdb), we compared their structures by superimposition and we observed that beyond the shared motifs the investigated proteins show a highly similar overall protein structure and shape ( Figure 3).  sapiens AIF (4bur.pdb) and C. thermarum NDH2 (in magenta cartoon, 5kmr.pdb) that gives an RMSD of 1.891 Å; S. cerevisae NDI (4g73.pdb) and C. thermarum NDH2 (5kmr.pdb) that gives an RMSD of 1.384 Å. Panel (d,e,f) Three 90-degree rotation of the alignment view of 4bur, 4g74, and 5kmr. In the grey cartoon representation, the common features highlighted by the superposition of the three structures. Specific AIF/NDI/NDH-2 structural features highlighted by the proposed superposition are reported in green, blue, and magenta cartoon representations. All the three aligned proteins are represented in complex with the FAD (orange sticks, panels a-c), NAD + (yellow sticks, panels a-c), and UQ (white sticks, panels a and c) cofactors, when the cited cofactors were present in the reported crystallized structures.
The main differences between the three crystallized structures are in the C-terminal domain and in some loops. Indeed, the AIF structure has a longer C-terminal domain containing an extra helix domain (residues 497-612, H. sapiens 4bur.pdb numbering) at variance with the shorter C-terminal domains of NDI (residues 458-513, S. cerevisiae . Comparative analysis of AIF/NDI/NDH-2 3D structure. Panels (a,b,c) Lateral view of superposition of H. sapiens AIF (in green cartoon, 4bur.pdb) and S. cerevisae NDI (in blue cartoon, 4g73.pdb) that gives a root mean square deviation(RMSD) of 2.891 Å; H. sapiens AIF (4bur.pdb) and C. thermarum NDH2 (in magenta cartoon, 5kmr.pdb) that gives an RMSD of 1.891 Å; S. cerevisae NDI (4g73.pdb) and C. thermarum NDH2 (5kmr.pdb) that gives an RMSD of 1.384 Å. Panel (d,e,f) Three 90-degree rotation of the alignment view of 4bur, 4g74, and 5kmr. In the grey cartoon representation, the common features highlighted by the superposition of the three structures. Specific AIF/NDI/NDH-2 structural features highlighted by the proposed superposition are reported in green, blue, and magenta cartoon representations. All the three aligned proteins are represented in complex with the FAD (orange sticks, panels a-c), NAD + (yellow sticks, panels a-c), and UQ (white sticks, panels a and c) cofactors, when the cited cofactors were present in the reported crystallized structures.
The main differences between the three crystallized structures are in the C-terminal domain and in some loops. Indeed, the AIF structure has a longer C-terminal domain containing an extra helix domain (residues 497-612, H. sapiens 4bur.pdb numbering) at variance with the shorter C-terminal domains of NDI (residues 458-513, S. cerevisiae 4g73.pdb numbering) and NDH-2 (residues 363-396, C. thermarum 5kmr.pdb numbering).
Furthermore, AIF, NDI, and NDH-2 show one or two specific loops. AIF-specific loops are observed at the level of residues 210-222 and 440-451 (4bur.pdb numbering, Figure 3), whereas two specific loops are observed in the NDI structure at the level of residues 138-163 and 413-435 (4g73.pdb numbering, Figure 3). A specific loop is observed at the level of residues 302-312 of NDH-2 (5kmr.pdb numbering, Figure 3).

Sampling of Homologous-Crystallized Structures by Folding Recognition Methods
Along the Blastp sequence sampling, it was also noticed that several dehydrogenases out of the sampled homologous protein sequences were annotated as dehydrogenases with different specific functions. i.e., among the sampled homologous AIF/NDH-2/NDI proteins, it was possible to detect a FAD-dependent pyridine-nucleotide disulfide oxidoreductase [41] from N. nitrosa (WP_107789266.1) and a ferredoxine reductase [42] from Nitroso spiralacus (WP_004178167.1), grouping in a subcluster adjacent to AIF sequences sampled from Mammalia. Another example is observed in a cluster adjacent to AIF-like sequences, where it is possible to observe a monodehydroascorbate reductase [43] from D. zibethinus (XP_022741237.1) proposed to be located within peroxisomes.
To verify if some structures of the above cited proteins were available in the protein data bank and showed an overall structure similar to AIF/NDH-2/NDI protein structures, we searched for AIF/NDH-2/NDI 3D-homologous structures by using the fold recognition tools implemented in pGenTHREADER.
Cartoon representations of the superimposed AIF, NDI, and NDH-2, and all the pGenTHREADER-sampled structures are reported in Figure 4.   Table 1). Second to fifth row: superimposition of all the AIF-like proteins, DLD-like proteins, NDI-like proteins, and NDH-2-like proteins, respectively. All the sampled proteins are reported as grey cartoon representations. Specific protein features are colored according to what is reported in the main text at each row. FAD, NADH, UQ, and CoA are reported as orange, yellow, white, and cyan sticks, respectively, where available at the reported crystallized structures, according to Table 1.
Although a common structural trend among all the sampled structures is observed, it is possible to recognize at least three main groups with specific structural features consisting of AIF-like proteins, NDH-2/NDI-like proteins (each with specific features), and lipoamide dehydrogenase-like proteins. The features of each cited group reflect their distribution in the phylogenetic tree (Supplementary Figure S1) built starting from the MSA of the sequences sampled by pGenTHREADER, where it is possible to observe three main groups consisting of AIF-like  Table 1). Second to fifth row: superimposition of all the AIF-like proteins, DLD-like proteins, NDI-like proteins, and NDH-2-like proteins, respectively. All the sampled proteins are reported as grey cartoon representations. Specific protein features are colored according to what is reported in the main text at each row. FAD, NADH, UQ, and CoA are reported as orange, yellow, white, and cyan sticks, respectively, where available at the reported crystallized structures, according to Table 1.
Although a common structural trend among all the sampled structures is observed, it is possible to recognize at least three main groups with specific structural features consisting of AIF-like proteins, NDH-2/NDI-like proteins (each with specific features), and lipoamide dehydrogenase-like proteins. The features of each cited group reflect their distribution in the phylogenetic tree (Supplementary Figure S1) built starting from the MSA of the sequences sampled by pGenTHREADER, where it is possible to observe three main groups consisting of AIF-like proteins (including ferredoxin reductases and pyridine nucleotide disulphide CoA NADPH dehydogeanses), NDI-like/NDH-2-like proteins; and other FAD-dependent dehydrogenases (including lipoamide dehydrogenases, thioredoxin reductases, and glutathione reductases).
In detail, the AIF-like protein group contains 20 of the 49 sampled structures. The functional annotation of these 20 sampled structures includes AIF, NADH-oxidase, ferredoxin reductase, rubredoxin reductase, putidaredoxin reductase, and pyridine nucleotide CoA disulfide reductase. The overall structure of those 18 structures overlaps with the human AIF, with an RMSD ranging between 1.06 and 3.1 Å ( Table 1). The good superimposition allows to highlight a common cavity hosting a FAD molecule in 14 crystallized structures, a NADH molecule in 2 crystallized structures, and a CoA molecule in 2 crystallized structures, according to Table 1.
A peculiar β-sheet motif located at residues 190-200 of the human AIF ( Figure 4) is observed as a specific AIF feature.
Two further structures, namely 3ics.pdb and 3ntd.pdb, annotated as CoA-disulfide reductases, may be associated to the AIF-like protein group. Both the structures host a FAD molecule and a CoA molecule similar to what was observed for 4fx9.pdb and 3cgb.pdb. Nevertheless, both 3ics.pdb and 3ntd.pdb show a 100-aa longer C-terminal domain (residues 461-554 for 3ics.pdb; 471-565 for 3ntd.pdb). Notably, 5vn0.pdb also hosts an oxygen molecule within 3.5 Å from FAD.
The NDH-2/NDI-like protein group contains 7 of the 49 sampled protein structures. Five structures are from bacterial proteins and the other two are from fungal proteins.
The functional annotation of the five sampled bacterial protein structures includes typeII NADH dehydrogenase and sulphide: UQ oxido reductases ( Table 1). The related overall structures overlap with the C. thermarum NDH-2, with an RMSD ranging between 0.82 and 3.33 Å. The good superimposition allows to highlight a common cavity hosting a FAD molecule in all the five sampled crystallized structures and a NADH molecule in one crystallized structure, according to Table 1.
The two NDI-like proteins are both NADH dehydrogenases crystallized from S. cerevisiae. The overall structure of the two proteins (as expected from two crystals of the same protein, crystallized in the presence of similar ligands, i.e., UQ-like molecule and stigmatellin) is very similar, with an estimated RMSD of 0.49 Å. From their superimposition, a common cavity hosting an FAD molecule and a NADH molecule is observed. 4g73.pdb hosts a quinone derivative, whereas 5yjw.pdb [46] hosts a stigmatellin bound at the same level of the quinone-binding region. Notably, stigmatellin is known as a competitive NADH-dehydrogenase inhibitor [47] (Figure 4).
The third group of FAD-dependent dehydrogenases contains 19 dehydrogenase structures, with an overall structure more similar to the ones shown by AIF-like proteins. The functional annotation of these 19 sampled structures includes lipoamide dehydrogenases, glutathione amide reductases, thioredoxin reductases, flavoprotein disulfide reductases, mercuric reductases, and NADPH:2-ketopropyl-coenzyme M oxidoreductase/carboxylases (2-KPCC).
The overall structure of those 19 proteins overlap with the human AIF, with an RMSD ranging between 3.22 and 4.84 Å. The further good superimposition allows to highlight a similarly located common cavity hosting a FAD molecule in 16 crystallized structures and a NADH molecule in 2 crystallized structures, according to Table 1.
Notably, 4m52.pdb hosts a sulfonamide derivative [48] located 9 Å far from the FAD molecule (at a region 3 Å far from the UQ-binding region observed in proteins of the previous groups) whereas 1mo9.pdb hosts a ketopropylthioethanesulphonate [49] bound at the same level of the CoA molecule-binding region observed in the first group of protein structures (Figure 4).
Indeed, all the structures host a FAD molecule and a NAD(P)+ molecule. Nevertheless, 4up3 and 5u63 completely lack a region corresponding to the AIF C-terminal domain, shown by all the AIF-like proteins (residues 491-C-ter, 4bur residues numbering, Supplementary Figure S3), whereas 1ps9 contains a different greater domain in correspondence of the N-terminal domain (residues 1-331, 1ps9 residues numbering). 4up3, 5u63, and 1ps9 were maintained in our comparative analyses as outliers for comparative purposes.

FAD and NADH Binding Regions
Beyond the highly similar overall structures shared between AIF (4bur.pdb), NDH-2 (5kmr.pdb), and NDI (4g73.pdb), highlighting a similarly located binding region for FAD and NADH, we also observed that the three superimposed crystal structures also bind those cofactors with glycine residues of the fifth sequence motif and a set of conserved aromatic/basic/acidic residues (Figures 4 and 5) located among the fifth and ninth sequence motif.
The variability observed in the composition of FAD-and NADH-binding regions does not alter the ability of those pockets in efficiently binding the two cofactors, as much as most of the sampled FAD/NADH dependent dehydrogenase structures show a FAD (32 out of the 49 compared crystal structures, see Table 1) and a NADH (8 out of the 49 compared crystal structures, see Table 1, in correspondence of NADH b , according to NADH cofactor nomenclature reported in [26]) molecule at a very similar position ( Figure 5D; Supplementary Table S3).
In general, all the investigated dehydrogenase 3D structures show several conserved residues, aligning with residues of the proposed 10 sequence motifs (Figure 2 and Supplementary Figure S2), interacting with FAD and NADH cofactors (Supplementary Table S3), despite the relatively low overall percentage of identical residues (15%-30% range, Supplementary Table S2, on the full length sequences) shared with AIF/NDI/NDH-2 sequences.

UQ Binding Site Comparative Analyses between AIF and NDH-2
Given the crystallization of an UQ-like molecule both in S. cerevisiae NDI (4g73.pdb) and in the sulphide: quinone oxidoreductase from Aquifex aeolicus (3hyw.pdb), it is also possible to propose residues putatively involved in the UQ binding in orthologous NDH-2/NDI-like proteins sampled from Bacteria/Fungi (Table 1).
Notably, for an investigation of the UQ-binding region from Fungi NDI-like proteins, it will be sufficient to superpose the crystallized dehydrogenase from S. cerevisiae to a comparative model of the fungal protein under investigation and highlight residues within 4 Å from the ubiquinone obtained from the S. cerevisiae NDI-crystallized structure (the ones in correspondence of UQ I , according to UQ cofactor nomenclature reported in [25]).
Beyond the highly similar overall structures shared between AIF (4bur.pdb), NDH-2 (5kmr.pdb), and NDI (4g73.pdb), highlighting a similarly located binding region for FAD and NADH, we also observed that the three superimposed crystal structures also bind those cofactors with glycine residues of the fifth sequence motif and a set of conserved aromatic/basic/acidic residues (Figure 4 and Figure 5) located among the fifth and ninth sequence motif. Figure 5. AIF/NDI/NDH-2 cofactor-binding regions. Panels (a-c) Residues within 4 Å from NADH (yellow sticks, corresponding to NADHa according to NADH cofactor nomenclature reported in [26]) or FAD (orange sticks) for AIF (green sticks from 4bur.pdb), NDI (blue sticks from 4g73.pdb), and NDH-2 (magenta sticks from 5kmr.pdb) are reported and labeled. Panel (d) Zoomed-in view of cofactor coordinates from all the investigated proteins obtained by superimposing the sampled crystallized structures (see Table 1). UQ (from 4g73.pdb, corresponding to UQI, according to UQ cofactor nomenclature reported in [25]) and DCQ (from 3hyw.pdb) are reported as white sticks. See Supplementary Figure S7 for visualizing the other cofactors observed along the superimposition of the investigated crystallized FAD/NADH oxidoreductases (i.e., heme C, CoA, O2, and H2S).
The variability observed in the composition of FAD-and NADH-binding regions does not alter the ability of those pockets in efficiently binding the two cofactors, as much as most of the sampled FAD/NADH dependent dehydrogenase structures show a FAD (32 out of the 49 compared crystal structures, see Table 1) and a NADH (8 out of the 49 compared crystal structures, see Table 1, in correspondence of NADHb, according to NADH cofactor nomenclature reported in [26]) molecule at a very similar position ( Figure 5D; Supplementary Table S3).
In general, all the investigated dehydrogenase 3D structures show several conserved residues, aligning with residues of the proposed 10 sequence motifs (Figure 2 and Supplementary Figure S2), interacting with FAD and NADH cofactors (Supplementary Table S3), despite the relatively low Figure 5. AIF/NDI/NDH-2 cofactor-binding regions. Panels (a-c) Residues within 4 Å from NADH (yellow sticks, corresponding to NADH a according to NADH cofactor nomenclature reported in [26]) or FAD (orange sticks) for AIF (green sticks from 4bur.pdb), NDI (blue sticks from 4g73.pdb), and NDH-2 (magenta sticks from 5kmr.pdb) are reported and labeled. Panel (d) Zoomed-in view of cofactor coordinates from all the investigated proteins obtained by superimposing the sampled crystallized structures (see Table 1). UQ (from 4g73.pdb, corresponding to UQ I , according to UQ cofactor nomenclature reported in [25]) and DCQ (from 3hyw.pdb) are reported as white sticks. See Supplementary Figure S7 for visualizing the other cofactors observed along the superimposition of the investigated crystallized FAD/NADH oxidoreductases (i.e., heme C, CoA, O 2 , and H 2 S).
Residues forming the UQ-binding region in NDH-2-like crystallized structures were proposed by superimposing the sampled NDH-2-like structures (Table 1) on A. aeolicus sulphide: quinone oxidoreductase and by highlighting residues within 4 Å from DCQ crystallized in complex with the structure of sulphide: quinone oxidoreductase from A. aeolicus (Supplementary Table S4).
Notably, residues involved in the UQ-binding region in NDI from Fungi and NDH-2-like proteins from Bacteria are located between the 9th and 10th sequence motifs and the C-terminal portion.
Given the highly similar overall structure observed between AIF, DLD, and NDH-2/NDI-like proteins and based on recent findings about AIF proteins [11,12], it is proposed that UQ-like molecules may participate in the function of AIF-like proteins and DLD-like proteins. Residues proposed to participate in UQ-like molecule-binding regions in AIF-like proteins and DLD-like proteins were highlighted by selecting residues within 4 Å from the two UQ-like molecules entrapped within AIF-like proteins and DLD-like proteins, after superimposition of the two protein sets with 4g73 and 3hyw (Supplementary Table S4).
Notably, residues involved in the potential UQ-binding region in AIF-like proteins and DLD-like proteins from Bacteria, Metazoa, and Plants are located between 8th, 9th, and 10th sequence motifs and the protein C-terminal portions.

Small Molecules and Other Cofactor-Binding Regions
It should be noticed that most of the crystallized structures among the AIF-like proteins, NDI-like proteins, and NDH-2-like proteins host other small molecules and cofactors in dedicated binding regions or are crystallized in complex with other small proteins (Table 1).
While AMP, ADP, riboflavin, and stigmatellin were crystallized in binding regions generally involved in cofactor binding, or overlapping regions, it was observed that disulfide reductases, among AIF-like proteins, host a CoA molecule in a specific binding region consisting of the residues shown in Supplementary Table S5.
Notably, RYL-552 binds three different binding regions within 5jwc far from the NADH/FAD/UQ cofactor-binding area. While two of those regions appear to be poorly resolved, the third ones appears to be a specific binding cavity [45]. By superimposing 5jwc with the analyzed AIF, NDI, and NDH-2, it appears that NDH-2 and NDI might form several interactions with RYL-552 involving 8 and 11 residues, respectively, that appear to be further well conserved in the sampled NDH-2 and NDI-like proteins (although in the latter, the overlapping binding region appears more buried from the local secondary structure), at variance with the AIF showing a lower number of interacting residues (just four, see Supplementary Table S6 and Supplementary Figure S4). Although AIF-like proteins do not show a well-defined binding pocket in correspondence of RYL-552, at variance with DLD-like proteins showing a sterically hindered/buried region in correspondence withRYL-552, it cannot be excluded that RYL-552 analogs may also bind AIF similarly located accessible binding cavities (Supplementary Figure S4 and Supplementary Figure S5).
The SL827 (a sulphonamide derivative) binding region observed in the lipoamide DH 4m52 consists of few amino acids that appear to be conserved in DLD-like proteins and AIF-like proteins. Notably, by superimposing 4m52 with AIF, NDI, and NDH-2, it appears that NDH-2 and NDI, in correspondence with the sulphonamide derivative binding region, shows a buried region occupied by an a-helical region that most likely will not allow the sulphonamide derivative to penetrate NDI and NDH-2 cofactor-binding regions or affect NDI/NDH-2 activity (Supplementary Figure S4).
Conversely, AIF-like proteins (i.e., 4bur) and DLD-like proteins show an accessible cavity in correspondence with the sulphonamide derivative binding region observed in 4m52 (Supplementary  Table S7), making us to purpose that sulphonamide derivatives may target both DLD-like and AIF-like proteins.
The ketopropylthioethanesulphonate (KPC)-binding region observed in 1mo9 (the 2KPCC) consists of few amino acids that appear to be conserved in DLD-like proteins and AIF-like proteins, similar to what observed for sulphonamide. Thus, also, in this case, NDI and NDH-2 show a buried region not accessible to KPC, whereas AIF-like proteins (i.e., 4bur) and DLD-like proteins show an accessible cavity in correspondence with the KPC-binding region observed in 1mo9 (Supplementary Table S7).
Notably, by superimposing DLD-like proteins with disulfide reductases it is possible to observe that the sulphonamide derivative and the KPC bind DLD-like proteins in correspondence of theCoA-binding region detected in disulfide reductases (Supplementary Figure S4).

Small Protein Subunit-Binding Regions
Some of the sampled dehydrogenases were also crystallized in complex with other related protein subunits. i.e., it was observed that flacocytochromecsulphide dehydrogenase from T. paradoxus (5n1t.pdb) was crystallized in complex with CytC and a dimeric copper-binding protein (COPC), whereas thioredoxin reductase from P. falciparum (4j56.pdb) and DLD from G.stearothermophilus (1ebd.pdb), among DLD-like structures, were crystallized in complex with thioredoxin and dihydrolipoamide acetyltransferase, respectively.
By superimposing flavoCytC: sulphide dehydrogenase from T. paradoxus with AIF and NDI, it is possible to highlight a binding region of AIF and NDI that is aligned with the flavoCytC: sulphide dehydrogenase protein region involved in interactions with CytC. We observed that the two regions consist of several amino acids located at (or very close to) the 1st, 3rd, 9th, and 10th sequence motifs and at the C-terminal region (Supplementary Table S8 and Figure 6).
From the DLD from G.stearothermophilus (1ebd.pdb) crystallized in complex with dihydrolipoamide acetyltransferase, it appears that the 41-aa-long dihydrolipoamide acetyl transferase binding domain may bind at the DLD monomer-monomer interface at the 1ebd C-terminal domain. Some residues close to the 9th and 10th sequence motifs are involved in acetylase binding (Supplementary Table S9) Similarly, by superimposing one of the sampled thioredoxin reductases to the thioredoxin reductase from P. falciparum(4j56.pdb) crystallized in complex with thioredoxin, it is possible to predict a putative binding region for thioredoxin in all the sampled thioredoxin reductases. Notably, by superimposing 4j56 to 4bur, 4g73, and 5kmr, it appears that several residues of 4bur are in the interaction range (below By superimposing flavoCytC: sulphide dehydrogenase from T. paradoxus with AIF and NDI, it is possible to highlight a binding region of AIF and NDI that is aligned with the flavoCytC: sulphide dehydrogenase protein region involved in interactions with CytC. We observed that the two regions consist of several amino acids located at (or very close to) the 1st, 3rd, 9th, and 10th sequence motifs and at the C-terminal region (Supplementary Table S8 and Figure 6). Figure 6. AIF/NDI/NDH-2 possible protein-protein interaction surfaces. Panel (a) Zoomed-in views of the crystallized heterodimeric structure of flavoprotein dehydrogenase (yellow surf representation) and CytC (pink surf representation) from T. paradoxus (5n1t.pdb) is reported in complex with FAD, (orange sticks), the heme C center in magenta sticks and NADH and UQ (yellow and white sticks, respectively, obtained by superimposition with 4g73 (see methods)). Panel (b) FAD, NADH, heme C center (from 5n1t.pdb), and UQ (superimposed from 4g73.pdb) are reported in stick representations. Intermolecular distances are reported by dashed lines and are labeled. Panel (c,d) 3D models of a putative heterodimeric structure of AIF (4bur.pdb) or NDI (4g73) proteins (cyan or white cartoon, respectively) in complex with CytC from T. paradoxus (5n1t.pdb). FAD, NADH, and UQ are reported in stick representations (see previous panels for colors).
From the DLD from G.stearothermophilus (1ebd.pdb) crystallized in complex with dihydrolipoamide acetyltransferase, it appears that the 41-aa-long dihydrolipoamide acetyl transferase binding domain may bind at the DLD monomer-monomer interface at the 1ebd C-terminal domain. Some residues close to the 9th and 10th sequence motifs are involved in acetylase binding (Supplementary Table S9) Similarly, by superimposing one of the sampled thioredoxin reductases to the thioredoxin reductase from P. falciparum(4j56.pdb) crystallized in complex with thioredoxin, it is possible to predict a putative binding region for thioredoxin in all the sampled thioredoxin reductases. Notably, by superimposing 4j56 to 4bur, 4g73, and 5kmr, it appears that several residues of 4bur are in the and CytC (pink surf representation) from T. paradoxus (5n1t.pdb) is reported in complex with FAD, (orange sticks), the heme C center in magenta sticks and NADH and UQ (yellow and white sticks, respectively, obtained by superimposition with 4g73 (see methods)). Panel (b) FAD, NADH, heme C center (from 5n1t.pdb), and UQ (superimposed from 4g73.pdb) are reported in stick representations. Intermolecular distances are reported by dashed lines and are labeled. Panel (c,d) 3D models of a putative heterodimeric structure of AIF (4bur.pdb) or NDI (4g73) proteins (cyan or white cartoon, respectively) in complex with CytC from T. paradoxus (5n1t.pdb). FAD, NADH, and UQ are reported in stick representations (see previous panels for colors).
By superimposing all the investigated dehydrogenase structures, it is possible to observe that all of them have the same overall structure. Moreover, 46 out of the 49 investigated dehydrogenases host a FAD molecule in the same position, whereas 9 of the 49 host an NADH molecule in the same position.
Notably, 3 out of the 49 dehydrogenase structures host a UQ-like molecule, whereas 4 out of the 49 host a CoA molecule at two dedicated similarly located binding regions.

A Similarly Located FAD/NADH-Binding Region for All the Investigated Flavoprotein Oxidoreductases and New Clues about a Putative UQ-Binding Region
Our analysis has allowed to predict the exact localization of NADH and FAD-binding regions among orthologous sequences of the investigated AIF-like, NDI-like, and NDH-2-like proteins. NADH and FAD-binding regions are also similarly located within DLD-like proteins. Furthermore, based on NDI and NDH-2 available structures, MSA, and recently published functional studies about AIF dehydrogenase activity [11], it may be speculated that AIF proteins may also bind a UQ molecule to participate to oxidative pathways crucial for mitochondrial respiration. Notably, our analysis has allowed to propose the existence of a putative binding region for UQ within the different investigated FAD/NADH-dependent oxidoreductases, including AIF-like proteins, that do not show a solved UQ molecule in the available crystallized structures. Considering DLD, it was already proposed that UQ might be reduced by lipoamide DH [50], although there is no evidence of a putative DLD UQ-binding region. Based on our comparative analyses, it may be speculated that the UQ-binding region within DLD is at the same level of the UQ-binding region highlighted within AIF after superimposition with NDI/NDH-2.

Concerns about the Opportunity to Draw New Inhibitors to be Used as Antibiotic/Antiparasitic Drugs, Directed Against the Investigated FAD/NADH Dehydrogenases
New insights have been acquired about the investigated species-specific FAD/NADH dependent oxidoreductases that raise the question about the opportunity to draw new antibiotic/antiparasitic drugs directed against the cited FAD/NADH-dependent oxidoreductases.
Currently, some of the described FAD/NADH-dependent oxidoreductases are considered in microbiology a crucial target of antibiotics or chemotherapeutics (i.e., those structurally related to RYL-552, a quinolinic derivative or to SL827, a sulphonamide derivative [10,48,[51][52][53]) because those enzymes play a fundamental role for the ATP synthesis and oxidative pathways in most microorganisms. The lack of NDH-2-orthologous enzymes in Mammalia was retained an important proof of the selective action of potential drug development [10,[54][55][56][57].
Although RYL-552 and SL827 appear to target specific binding regions within NDH-2/NDI and DLD bacterial proteins, respectively, given the high structural similarity and the similarly located cofactor-binding regions highlighted in the investigated FAD/NADH-dependent oxidoreductases, it will be necessary to ascertain the absence of interactions with the human AIF/DLD-accessible cavities, before using RYL-552/SL827 structurally related ligands in new preclinical/clinical trials.
Similarly, SL827 and KPC analogs were proposed as chemicals for selective targeting of M. tuberculosis DLD and X. autotrophicus NADPH-CoM oxidoreductase. However, human DLD/AIF proteins show accessible cavities in correspondence of theSL827/KPC-binding regions observed within M. tuberculosis DLD and X. autotrophicus NADPH-CoM oxidoreductase. Thus, for SL827/KPC structurally related ligands, it will also be necessary to exclude putative interactions with human AIF/DLD proteins to guarantee the microorganism selectivity of both molecules and to avoid undesirable side effects due to interactions with the human mitochondrial FAD/NADH-dependent oxidoreductases. Notably, antibiotics that might be considered structurally related to SL827, such as sulfamethoxazole/trimethoprim and sulfisoxazole, although commercialized for several years, are known for their serious side effects [58]. It cannot be excluded that those adverse reactions may be ascribed to interactions between the cited antibiotics and AIF/DLD human proteins.
Finally, it should also be noticed that CoA disulfide reductases, structurally related to AIF/DLD-like proteins, host a CoA molecule in correspondence of the SL827/KPC-binding region. The presence of a CoA molecule in CoA disulfide redutases, at the level of the SL827/KPC-binding region observed in the homologous AIF/DLD-like proteins, raises the question about possible competitive effects between SL827/KPC structurally related ligands and disulfide reductases CoA-binding regions that may decrease the efficiency and selectivity of SL827/KPC structurally related ligands.

New Clues in Support of AIF Participation in Mitochondrial Respiration
Based on the performed comparative analyses, it was possible to propose that flavoprotein oxidoreductase interaction surfaces are involved in the binding of other protein subunits involved in respiratory mechanisms (i.e., a putative AIF-binding region for CytC, after comparison with flavocytochrome c sulfide DH from T. paradoxus).
Considering (a) the role played by NDH-2/NDI-like proteins in bacterial/fungal respiration [7,25], (b) the direct interaction between a T. paradoxus flavoprotein oxidoreductase and a CytC-like protein [44], and (c) the recently proposed rotenone sensitive NADH:UQ oxidoreductase activity of AIF [11,12], it could be speculated that a direct interaction between AIF and CytC and other mitochondrial copper-binding proteins of the inner membrane [59][60][61] might also exist in Metazoan mitochondria, with several implications regarding AIF participation in mitochondrial respiration. These data are coherent with previous observations about alternative respiratory protein complexes [62][63][64][65].
Concerning the above reported hypothesis, we should recall that the incomplete NADH consumption along mitochondrial respiration assays performed in the presence of high concentrations of rotenone and other respiratory protein-complex inhibitors, as well as the protective role played by AIF in the maintenance of respiratory chain complexes, are well documented [22,[66][67][68].
Thus, it appears that other mitochondrial enzymes may contribute to the buffering of altered NAD + /NADH ratios by oxidizing alternatively cytosolic NADH or mitochondrial matrix NADH, and here, we might speculate that AIF, being anchored to the mitochondrial inner membrane, protruding into the intermembrane space, may participate in the regulation of the NAD + /NADH or FAD/FADH2 (maybe UQ/UQH2) ratio. AIF participation in the regulation of redox signaling between mitochondria and other cell compartments may be crucial above all in those tissues in which the G3P shuttle [69], nicotinamide nucleotide transhydrogenase (NNT) [70], NADH-b5 oxidoreductase [71,72], or malate/aspartate shuttle [15,73,74] do not work properly or when the cited proteins, mitochondrial oxidative phosphorylation complex subunits, or matrix proteins are mutated or downregulated [75,76]. If AIF behaves as a NADH dehydrogenase, able to oxidize cytosolic NADH, it may also participate in reprogramming metabolic pathways, providing new clues towards the full comprehension of the puzzling Warburg effect condition [77].
Notably, quinoline 3-sulfonamides were recently proposed to be a class of high-affinity (bioactive in the nM range) lactic dehydrogenase inhibitors [78] that may be employed in cancer therapies. However, quinoline 3-sulfonamides are ligands structurally related both to RYL-552 and, although at a lower extent, to quinone-related structures (i.e., stigmatellin or menaquinone) known to be able to inhibit or bind NDI/NDH-2-like proteins. Thus, it appears that it will be necessary in the near future to ascertain putative activity/affinity of quinoline 3-sulfonamides for NDI/NDH-2 and also for AIF/DLD-like proteins.
Based on the above reported observations, the redox path proposed in Figure 7, and the available crystallized structures, we propose that NDH-2, NDI, and AIF proteins may have two binding regions for NADH and two further binding regions for UQ-analogous molecules to work correctly, according to what proposed by [25,26]. The need for having two molecules for each cofactor may reflect the necessity to bind to specific binding pockets with different affinities for the entry/exit of oxidized/reduced cofactors, similar to what proposed for other quinol-dependent oxidases [30,79].
presence of a metal ion working as a further cofactor might reveal crucial for the correct protein function and/or redox mechanism.
More in general solving the structure of the missing fragments at the N-/C-terminal regions and at the level of 509-560 AIF protein region will help in elucidating conformational changes responsible both for cofactor entry in the AIF catalytic site and dimerization/multimerization processes. On this concern it should finally be noticed that the proposed distance between FAD and UQ (obtained by superimposition as previously described) as well as the distance between UQ and CytC (obtained by superimposition as previously described) in AIF would be below 13 Å, which is in the useful distance-range for allowing electron tunneling [30], although the proposed AIF-UQ-binding region appears to be less accessible than its counterpart in NDH-2 and NDI. Also, the presence of an oxygen molecule (obtained by superimposition with 5vn0.pdb, as previously described) as well as an H2S molecule (3hyw.pdb) in the investigated crystallized structures (see Supplementary Figure  S7) makes us think that several investigated flavoprotein dehydrogenases may show in the future other unsuspected abilities. Notably, Ferreira et al. [26] proposed that a high NADH/NAD + ratio may facilitate the formation of the AIF: NADH complex, which may trigger AIF dimerization and cell death induction, through the cleavage of the first 102 residues of AIF [22,26], the following release of AIF ∆1-102 from mitochondria, and its translocation to the nucleus, where AIF (complexed with MIF [24]) mediates chromatinolysis. This hypothesis was based on AIF-crystallized structures solved by Ferreira et al. [26]. In detail, Ferreira et al. have provided a crystal structure of AIF in the presence of one FAD molecule (that they called the oxidized hAIF ∆1-102ox , available as a single monomer under the PDB code 4bv6.pdb, [26]), and one further AIF structure in the presence of both FAD and NADH (that they called the reduced hAIF ∆1-102rd : NAD(H), available as a dimer of homodimers under the PDB code 4bur.pdb, [26]).
Both the obtained crystallized structures 4bur and 4bv6 lack a protein region at the level of the protein segment including residues 509-560, defined by the authors as the apoptogenic segment [26]. In detail, 4bur lacks the 517-552 region (see 4bur, chain B), whereas 4bv6 lacks atomic coordinates for the 546-558 protein region (see the single chain of 4bv6). It should be noticed that the missing regions are close to the NADH b -binding region (according to NADH cofactor nomenclature reported in [26])and in correspondence with theUQ II NDH-2-binding region (according to UQ cofactor nomenclature reported in [25]).
In light of these observations and according to [26], it might also be speculated that NADH presence (i.e., a high NADH/NAD + ratio) can make more exposed/accessible AIF redox catalytic sites for UQ-like ligands that may participate in the re-oxidation of NADH mediated by FAD. Conversely, a high NAD + /NADH ratio may induce conformational changes that lower the affinity of NADH and UQ-like molecules for AIF catalytic pockets.
Nevertheless, Miseviciene et al. [12] showed that AIF has quinone reductase activity (expressed as the k cat /K m ratio) 10 2 -to 10 4 -fold lower than that of cytochrome P450 reductase and NAD(P)H: quinone oxidoreductase (NQO1) and 10 1 -to 10 2 -fold lower than that of Mammalian thioredoxin reductase but is comparable with the activity of glutathione reductase from various sources (see [12] and references therein).
However, the low quinone reductase activity exerted by the recombinant AIF investigated by Miseviciene et al. may also in part be ascribed to the fact that Miseviciene et al. used for their analyses the AIF ∆1-77 protein domain [12] instead of the AIF ∆1-52 protein domain, be the latter the complete protein domain responsible for redox activity [22,80]. It will be necessary to use AIF ∆1-52 recombinant protein for checking/re-estimating AIF quinone reductase activity and compare it with the recently quinone reductase activity mediated by the Mammalian complex I [81].
Concerning the dimerization process, it should also be noticed that, beyond missing residues at the 509-560 AIF region, both 4bur and 4bv6 [12] crystallized structures lack the first 124 and thus it cannot be excluded that the missing fragments at the C-terminal domain (at the level of residues 509-560) and at the N-terminal domain may facilitate a non-physiological multimerization process.
In support of the lack of clarity about dimerization processes and cofactor access, it is observed that in correspondence of the protein region containing the missing portions in 4bur/4bv6 (residues 509-560), 4bv6 also shows the crystallized C-terminal region (Ile610, 4bv6 residues numbering) located between the two termini (Pro545 and Asp559, 4bv6 residues numbering) of the non-solved portion. Furthermore, the mercuric reductase (4k7z.pdb) from P. aeruginosa hosts a mercuric ion in a region partially overlapping with the AIF 546-558 (4bv6 residues numbering) protein region.
All these observations make possible to hypothesize the putative involvement of a metal ion that might locate at the level of the AIF 546-558 (4bv6 residues numbering) missing residues. The presence of a metal ion working as a further cofactor might reveal crucial for the correct protein function and/or redox mechanism.
More in general solving the structure of the missing fragments at the N-/C-terminal regions and at the level of 509-560 AIF protein region will help in elucidating conformational changes responsible both for cofactor entry in the AIF catalytic site and dimerization/multimerization processes.
On this concern it should finally be noticed that the proposed distance between FAD and UQ (obtained by superimposition as previously described) as well as the distance between UQ and CytC (obtained by superimposition as previously described) in AIF would be below 13 Å, which is in the useful distance-range for allowing electron tunneling [30], although the proposed AIF-UQ-binding region appears to be less accessible than its counterpart in NDH-2 and NDI. Also, the presence of an oxygen molecule (obtained by superimposition with 5vn0.pdb, as previously described) as well as an H 2 S molecule (3hyw.pdb) in the investigated crystallized structures (see Supplementary Figure S7) makes us think that several investigated flavoprotein dehydrogenases may show in the future other unsuspected abilities.

Pieces of Evidence about the Possible Targeting of AIF for the Development of New Treatments for Mitochondrial Dysfunction in Rare Diseases
If AIF participation to mitochondrial respiration would be ascertained, AIF protein may be considered as a target for the stimulation of mitochondrial function for the development of new treatments for diseases characterized by mitochondrial dysfunction.
If an UQ structurally related ligand-binding region would be highlighted experimentally on the recombinant protein, according to what reported above, we would expect that several UQ structurally related drugs used for the treatment of mitochondrial diseases may improve patient conditions by stimulating mitochondrial function through AIF. On this concern, it could be speculated that UQ structurally related approved drugs, like EPI-743, Idebenone [82], and KH176 [83], may exert their effect also by targeting the AIF UQ-derivative-binding region.
At the moment, it is proposed that EPI-743 and Idebenone may act on complex I [47,84] whereas KH-176 would interact with the thioredoxin system [83].
Nevertheless, if complex I subunits are mutated, above all, if mutations are located close to the proposed CoQ-binding region, it would be difficult that a greater availability of the cited cofactors (or structurally related ligands) might rescue complex I defect, acting directly on the same complex I activity.
Conversely, because of the similarity between KH-176 (and its precursors, i.e., Trolox), EPI-743, Idebenone, and UQ, it may be speculated that KH-176, EPI-743, and Idebenone may target an UQ analog-binding region in a functional protein participating to mitochondrial respiration.
According to [83], the most probable binding region of KH-176 could be located within the thioredoxin reductase (TrxR1). Nevertheless, TrxR1 shows a highly similar overall structure with AIF-like proteins and appears to be even more similar to DLD-like proteins (Supplementary Figure S6). Thus, it might be speculated that KH-176 may also target AIF and DLD, beyond TrxR1, at the proposed similarly located UQ-binding region (Supplementary Table S4).

Conclusions
The presented computational strategy, based on the combination of sequence database screening and fold recognition methods, will allow clinicians and biomedical researchers to identify/highlight "structurally related proteins", shared from Bacteria, Protista, Fungi, Metazoan, and Plants, not detectable using only sequence alignment-based search tools, to choose the target of new drugs under investigation with greater confidence aiming to reduce off-target effects.
The possibility to target a species/specific enzyme becomes crucial in the battle against antibiotic resistance. Indeed, it is known that several Bacteria have become resistant to antibiotics directed against enzymes involved in cell wall synthesis, cell membrane function, protein and nucleic acid biosynthesis, and antimetabolites.
To overcome the problems related to antimicrobial resistance, medical scientists are starting to target the cellular energy-generating machinery [78] and, in general, enzymes regulating crucial respiration/oxidative pathways in pathogenic microorganisms [54,[85][86][87].
Notably, new chemicals directed against FAD/NADH oxidoreductases are already approved drugs and/or are being tested in preclinical/clinical trials [53,82,83]. Given our findings, clinicians and biomedical researchers could now test drugs designed for the targeting of FAD/NADH oxidoreductases (NDH-2/TrxR1 like, from Bacteria, Fungi, or Protista) on human mitochondria to exclude deleterious interactions with human mitochondrial proteins (i.e., AIF, DLD, and TrxR1), ascribable to the common origin shared by mitochondria from Metazoans and Protista [88] and, in general by mitochondria and Bacteria [89,90].
At the same time, the proposed strategy will allow clinicians and biomedical researchers to link putative adverse effects, following the administration of new drugs directed against bacterial/fungal/protista FAD/NADH oxidoreductase (NDH-2/TrxR1-like) proteins to unpredicted interactions with human structurally related AIF/DLD/TrxR1. Furthermore, the proposed comparative analyses might help clinicians and medical researchers to evaluate the possibility of considering FAD/NADH oxidoreductases as a possible target of new small molecules that are able to modulate mitochondrial respiration in patients affected by rare diseases or cancers characterized by a severe mitochondrial dysfunction.

Patents
The presented data are part of the Italian patent: Pierri et al., 2019; Computational methods for the identification of FAD/NADH dehydrogenases binding regions for drug design and discovery; IT Patent 102019000022545.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.