Resilience of biochemical activity in protein domains in the face of structural divergence

Abstract

Recent studies point to the prevalence of the evolutionary phenomenon of drastic structural transformation of protein domains while continuing to preserve their basic biochemical function. These transformations span a wide spectrum, including simple domains incorporated into larger structural scaffolds, changes in the structural core, major active site shifts, topological rewiring and extensive structural transmogrifications. Proteins from biological conflict systems, such as toxinantitoxin, restriction-modification, CRISPR/Cas, polymorphic toxin and secondary metabolism systems commonly display such transformations. These include endoDNases, metal-independent RNases, deaminases, ADP ribosyltransferases, immunity proteins, kinases and E1-like enzymes. In eukaryotes such transformations are seen in domains involved in chromatin-related peptide recognition and protein/DNA-modification. Intense selective pressures from “arm-race”-like situations in conflict and macromolecular modification systems could favor drastic structural divergence while preserving function.

Introduction

Since the formulation of the evolutionary theory the question of convergent versus divergent evolution has been widely studied and debated [1]. Analysis of organismal structure revealed that certain structurally similar, archetypal forms have repeatedly evolved from distinct ancestral forms (convergence): for instance, the same “fish-like” body shape has independently evolved among vertebrates on multiple occasions in fishes, ichthyosaurs, and whales [1,2]. In addition to such global convergence of form, organisms also display more limited forms of convergent evolution, such as functionally similar organs constituted from structurally different precursors (e.g. wings of insects, pterosaurs, birds and bats)[1]. With the advent of modern structural studies on proteins the question of convergent versus divergent evolution entered the molecular realm [3].

Protein structure and function: Convergence, divergence, and more complex relationships

From their inception, structural studies furnished clear cases of convergence of biochemical function [3,4], which might be defined as the primary biochemical activity of a given protein, i.e., catalysis or binding of another biomolecule or solute. A classic example are the tRNA synthetases, all of which catalyze the same basic reaction (ATP-dependent ligation of the acyl group of an amino acid of the 2’ or 3’ hydroxyl of the cognate tRNA), but belong to two distinct protein superfamilies with structurally unrelated folds [5]. On the other hand, structural studies also revealed that divergent evolution is rampant among protein domains – enzymatic domains, which were originally considered to be unrelated, such as actin, hexokinase, RNase H, PIWI, and diverse integrases of transposons and retroviruses were shown to contain a common fold (the RNase H fold) with comparable active site residues [6]. Discovery of divergent relationships via structural comparisons has provided a robust framework for understanding the evolutionary “exploration” of substrate space, wherein certain folds have been utilized as platforms for extensive biochemical diversification [7-9]. Development of increasingly sensitive sequence-profile and profile-profile search methods has greatly enabled the detection of such divergent relationships between protein domains [10-12].

The explosion in the number of structures in the past decade is now revealing the prevalence of more complex scenarios beyond straightforward functional convergence and divergent evolution. One of these is local structural convergence in active sites of protein domains with similar functions [13,14], such as glycine rich loops in the nucleotide-binding loops of structurally distinct folds [15,16]. This can be seen as a molecular analog of the localized convergence in functionally comparable body parts of organisms. With the protein structure universe being increasingly populated, more evolutionary intermediates are emerging that help understand better the major structural transmogrifications among domains [7,8]. As a result, an underappreciated phenomenon in the structure-function relationships of proteins has come to light: domains sharing a common ancestry undergoing increasingly drastic divergence of structure while at the same time preserving their ancestral catalytic/binding activity and general substrate specificity. This tendency is distinguishable from conventional divergent evolution, wherein divergence occurs primarily at the level of sequence similarity while generally preserving the core fold [17-19]. Importantly, this situation runs contrary to conventional divergent evolution, wherein sequence divergence goes hand-in-hand with diversification of biochemical function, both in terms of broad substrate specificity and types of reactions catalyzed [13,20-22].

As the above-defined evolutionary phenomenon apparently violates the basic predictive logic used in extrapolating function based on protein sequence/structure divergence it is worth a more detailed documentation. Accordingly, we review the recent literature to illustrate and synthesize information regarding this phenomenon.

The spectrum of structural divergence in the face of biochemical function preservation

The structural divergence of domains sharing a common ancestry but preserving biochemical function is best approximated by a spectrum (Figure 1). This premise stems from the observation that structural change, while objectively measurable in different ways, becomes subjective in terms of the threshold of what might be considered a “drastic” modification. Thus, on one end of this apparent spectrum we have changes that fall with the realm of conventional divergent evolution i.e., sequence divergence and simple elaborations of the core fold, typically in the loops connecting core secondary structure elements. Beyond this “baseline”, for descriptive convenience the spectrum can be broken into the following classes of modifications defining the more drastic changes (Figure 1): 1) A simple, ancestral fold being combined with new secondary structure elements to form structurally distinct scaffolds; 2) Modification of the ancestral fold via deletion/degeneration or insertion or transformation of secondary structure elements in the core; 3) Reorganization of active site residues without change in ancestral catalytic/binding activity; 4) Rewiring of secondary structure elements in the topology while preserving their overall spatial positions; 5) Transmogrification of the ancestral fold with major spatial and/or topological reorganization.

Figure 1. Spectrum of structural divergence which preserves biochemical function.

The spectrum is broken up into distinct classes of structural divergence separated by dotted lines. Example structures are depicted as topology diagrams with arrows representing β-strands and coils representing α-helices. ‘B’ represents the ‘baseline’ class of structural divergence. Class 1: representatives of papain-like peptidases; Class 2: restriction endonuclease fold with modified part of the secondary structural elements colored in yellow; Class 3: BECR fold members with divergent active site residues highlighted in green; Class 4: transition between NFACT and Fpg-MutM-EndoVIII DNA glycosylase proteins are accompanied by linear arrays of secondary structural elements to show rewiring, each duplicated basic 4-stranded element is given a distinct color; Class 5: topological transmogrification observed in the obligate dimer-forming BcbF family of HAD domains, strands are labeled, and monomers are given distinct colors.

It must be emphasized that, as these changes span a spectrum of variation, some domains might show more than one of the above classes of modifications. Below we illustrate these classes of modifications using the wealth of examples uncovered by recent structural studies.

Incorporation of simple conserved cores into complex scaffolds

A striking case of this type of structural modification is furnished by the HNH endonucleases that are found in wide variety of DNA-cleaving and, on some occasions, RNA-cleaving enzymes (Figure 2) [23,24]. Simplest versions of this domain display a standalone treble clef fold (Figure 2A), which is stabilized by 4 conserved Zinc-chelating residues and contains an endonuclease active site between the β-hairpin and the C-terminal helix characteristic of this fold (e.g. the restriction endonuclease Hpy99I; PDB 3GOX) [25]. This core is incorporated into increasingly complex structures in various nucleases. In the Cas9 endoDNases of the type-II CRISPR systems it is incorporated into a larger structure formed by α-helices packing around from both the N- and C-termini [26](Figure 2B). In homing endonucleases (e.g. I-PpoI), the original stabilizing Zn²⁺ of the treble-clef core is lost, but it is incorporated into a larger structure that is stabilized by two newly acquired Zn-chelating sites [27] (Figure 2C). A further variation is seen in the colicin E9 DNase domain, where again the core treble clef has lost its metal and is incorporated into a larger globular domain stabilized by augmentation of the β-hairpin into a sheet by stacking of additional strands (Figure 2D)[28]. Finally, an even more dramatic change is seen in the NucA/non-specific endonuclease family, wherein the treble clef core, which has lost its stabilizing metal, is embedded into an extended sheet formed by two copies of a three-stranded domain with characteristic loop-like C-terminal extensions [24] (Figure 2E). This profusion of structural modifications has often resulted in the relationship between these domains being identified only upon careful sequence structure comparisons [7]. Yet, strikingly, all these domains are rather conservative in terms of their endonuclease activity.

Figure 2. Structural diversity of HNH endonucleases.

The α-helix and β-sheet of the HNH structural core are shown in red and aquamarine respectively. Metals, active site (blue) and zinc chelating (in green) residues are shown in the ball and stick mode. Other incorporated structural elements are in light blue. The duplicated three-stranded units of NucA are shown in light blue and light brown respectively.

The wider prevalence of this type of modification is indicated by another homologous group of domains, the papain-like peptidases, which catalyze hydrolysis of peptide and related amide linkages (e.g. the isopeptide bond formed by conjugated ubiquitin) (Figure 1) [29]. Here the core fold is formed by a helix bearing the catalytic cysteine, and three strands bearing a histidine and a polar residue, which together constitute the catalytic triad of these enzymes [30]. This core might occur more or less in a simple standalone form (e.g. the Pasteurella Dermonecrotic toxin peptidase [31]) or is incorporated into a wide range of more elaborate scaffolds where the above core is part of an augmented β-sheet (e.g. the ATG8-deconjugating enzyme ATG4B [32]) or β-barrels with varying number of strands (e.g. NlpC/P60 peptidase domains)[30,33]. Indeed, a great diversity of such structural elaborations is observed among papain-like peptidase families, which deconjugate ubiquitin and ubiquitin-like proteins from targets despite the fact they catalyze an essentially equivalent peptidase/isopeptidase reaction [29,30].

Modification of the ancestral fold by degeneration or insertion or transformation of secondary structure elements

Hints for this form of structural transformation were first obtained in the serine/threonine/tyrosine (STY) kinase domains, which are found in enzymes catalyzing a stereotypic reaction of ATP-dependent phosphorylation of hydroxyl or carboxyl groups in peptide side chains, lipid head groups, and various small molecules [34]. This domain displays a complex fold with a 5-stranded unit at the “top”, a smaller central sheet comprised of 3 strands, and a “bottom” formed by a mixed α+β structure related to the equivalent unit of the ATP-grasp fold [7,34]; all three units contribute residues for nucleotide binding or catalysis (Figure 3). Phyletic pattern and structural analysis [35] suggests that the most ancient version of this fold is likely to be the SAICAR synthetase (PurC), which is a kinase catalyzing the formation of a peptide-like bond in purine biosynthesis. Here, the “bottom” element has an extended sheet of 5 strands [36] (Figure 3A). In kinases which phosphorylate aminoglycoside antibiotics, as part of resistance strategies of bacteria, this sheet in the “bottom” element is reduced to 4 strands (Figure 3B)[37,38]. In the Rio1/2 kinases, which are ancient protein kinases shared by archaea and eukaryotes as part of their ribosome maturation apparatus [39], the said sheet is reduced to a mere dyad of strands (Figure 3C) [40]. Finally, in STY kinases, which are the mainstay of signal transduction in eukaryotes and certain bacteria (e.g. myxobacteria, actinobacteria and cyanobacteria), and lipid kinases (e.g. PI3 kinase) the strands are entirely lost in the bottom unit of the domain [34,41,42]. Thus, structures solved to date recapitulate the entire trajectory of structural transformation from the pronounced sheet of the ATP-grasp like precursor to the total loss of the strands (Figure 3D). However, through this major transformation the basic phosphorylation reaction catalyzed by these domains has remained the same in large part because the active site residues lying at the inter-β strand connectors have remained intact, even as the secondary structure of the proximal elements have changed [42]. This change has probably allowed these enzymes to explore a wider substrate space while keeping the basic reaction constant [21].

Figure 3. Cartoon representations of various STY kinase domains illustrating the structural transformations in the superfamily.

Helices are colored red, whereas strands are colored based on their structural unit. Strands of the top unit are colored yellow and those in the central sheet, magenta. The cyan and pink strands of the bottom unit show the equivalence of these strands between the structures.

New structures and results from sequence analysis suggest that such transformations might be more prevalent, as illustrated by the deaminase-like fold [43]. This fold includes catalytic domains with biochemically distinct activities such as, base deaminase (e.g. the mutagenic and RNA-editing AID/APOBEC deaminases), peptidase, possible nuclease, AICAR transformylase and ADP-ribosyl transferase [43,44]. Despite this diversification the spatial location of the active site is retained throughout the fold along with a core β-sheet with 5 β-strands [43]. However, among these enzymes, recent studies show that deaminases are prone to rampant structural plasticity of the core sheet. First, deaminases might come in two types wherein the 5^th strand of the sheet might exist either in parallel or antiparallel configuration [45]. Moreover, the α-helices following this strand might often be lost in several representatives. In other cases not just helices but the entire 5^th strand might be lost. Given that other enzymes of this fold closely conform to the ancestral type, these drastic changes appear to be a more recent feature occurring within the deaminases alone [43]. Interestingly, closer examination of their evolution has shown that this rampant structural modification is primarily associated with their diversification as toxins delivered into target cells by diverse bacteria [43]. Comparable structural changes involving strand insertion or deletion are also observed in nucleases of the restriction endonuclease fold that target specific DNA sequences or structures (Figure 1)[46-48]. Other rich sources of such structural modifications are the chromo-like fold and PHD-finger-like domains, which despite their evolutionary plasticity retain a comparable function of binding methylated peptides in chromatin proteins, such as histones [49-52].

Reorganization of active site residues while retaining ancestral catalytic/binding activity

Usually the constellation of residues constituting the active site is the most conserved sequence feature of a superfamily of enzymes catalyzing the same reaction type [17,18,21,22]. Thus, such active sites residues have great predictive value [22] and are typically resilient to a wide range of structural changes in the core fold, as noted in the above-discussed categories. Studies on P-loop NTPases showed that one key active site residue, namely the arginine finger, is not conserved and has independently evolved on multiple occasions, either within the core fold, or in linked domains, or in separate proteins [53]. More drastic alterations of active sites are seen in DNases of the restriction endonuclease fold—the Vsr (very short patch DNA repair)-like endonucleases have undergone a reconfiguration of the ancestral active site via loss of two key residues and acquisition of two new ones from entirely different locations in their structure [47,54-56]. Despite this the spatial position of the active site pocket and their catalytic activity remains comparable to the ancestral versions. Multiple examples of comparable modifications have recently become apparent among RNases of the BECR (Barnase-EndoU-Colicin E5/D-RelE) fold [57]. These RNases include various bacterial and fungal toxins which are deployed against target cells [57,58], toxin components of toxin-antitoxin systems [59,60], and certain RNases involved in splicing such as Endonuclease U [61]. Several families of these RNases possess a histidine that is critical for catalysis [57,62]. While this residue is conserved within a given family (e.g. Ntox21 and colicin E3 families), it is typically not conserved across all members of the BECR fold (Figure 1). In some, such as the RelE family, even the histidine might not be conserved. Yet all families of the BECR fold catalyze a similar metal-independent endoRNase reaction, generating a RNA product with a cyclic 2’-3’ end [62].

Interestingly, at least two other unrelated folds of RNases show a comparable phenomenon. One of these is the recently identified all α-helical HEPN RNase, which is found in type-I and III CRISPR systems and, like the BECR RNases, also in toxin-antitoxin systems [63]. While several of these RNases have a characteristic histidine-containing motif associated with their active site, some families have lost this histidine and acquired distinct catalytic histidines from elsewhere in the sequence [63-65]. Similarly, the RAMP superfamily from the type-I and III CRISPR systems are RNases containing the RRM-like fold and are critical for processing the CRISPR RNAs from their precursor transcript [66,67]. Although several of the individual RAMP families contain conserved histidines, none of these are conserved throughout the superfamily [68]. Structural studies on RAMPs have demonstrated that these independently acquired histidines and other polar residues might constitute distinct but catalytically equivalent active sites for this RNase superfamily [66,68]. Like the BECR RNases, the HEPN and RAMPs are also metal-independent endoRNases that generate products with cyclic 2’-3’ ends [64-66]. Metal-dependent active sites require a precise 3D configuration of metal-chelating residues coming from different parts of the fold probably making them harder to reconfigure. However, the metal-independent RNases are primarily dependent on histidine or other polar residues to facilitate an internal attack on the phosphodiester bond by the 2’OH group in RNA [62]. Hence, it is likely that such residues could independently emerge in a “pre-adapted” RNA-binding scaffold such as those observed in the above-discussed domains.

Rewiring of secondary structure elements while preserving their general spatial positions

Straightforward versions of this type of structural modification are the widely known circular permutations, which are common in domains wherein the N- and C-terminal elements tend to be spatially proximal [13,69]. Circular permutations have occurred on multiple occasions without any change in catalyzed reactions in amide bond-forming ligases of the classical ATP-grasp fold (e.g. glutathione synthetases which ligate glycine to gamma-glutamylcysteine to form glutathione)[70,71]. More dramatic modifications of this type have been reported in the RNA-binding KH domain superfamily [72]. In the ERA GTPases, involved in assembly of the 30S ribosomal subunit, the C-terminal KH domain has undergone a major rewiring of its secondary structure units [73]. However, it still preserves the characteristic sequence motif and overall spatial organization of classic KH domains [72], and binds RNA in a comparable manner [73]. Similar rewiring is also observed in another nucleic acid/nucleotide binding domain, the RAGNYA domain, wherein multiple shifts in inter-secondary structure connectivity are observed [74]. A particularly dramatic case is seen in the nucleic acid ligases [7,75]: whereas in classical amide bond ligases of the ATP-grasp fold a kinase-C terminus-like domain is fused to the C-terminus of a RAGNYA domain, in nucleic acid ligases this domain is inserted into the RAGNYA domain [74]. Despite the consequent rewiring of secondary structure the RAGNYA domain continues to bind ATP just as in the case of unmodified version of the fold.

A recent example of such a modification has come to light with the discovery of the relationship between the Fpg-MutM-EndoVIII family of DNA glycosylases and the NFACT (NEMF-FpbA-caliban- Tae2) proteins that are predicted to be RNA-modifying enzymes [76]. Their shared catalytic domain is a composite fold comprised of an N-terminal 8-stranded β-sandwich with 2 flanking helices and a C-terminal unit formed by two helix-hairpin-helix (HhH) motifs (Figure 1). Here, the respective phyletic patterns of NFACT and DNA glycosylases help in establishing the direction of structural modification – NFACT is found across all three superkingdoms of life, whereas these DNA glycosylases emerged in bacteria and were transferred to eukaryotes but are absent in archaea [76]. This suggests that the latter is likely to have been derived from a precursor like the former. Consistent with this, the β-sandwich in the NFACT domains shows a more ancestral condition of being comprised of a duplication of two 4-stranded elements. Reconstruction of the structural reorganization in the DNA glycosylases indicates that it proceeded via an initial triplication of the basic 4-stranded element, followed by reconstitution of the β-sandwich from secondary structure elements from each of the three copies (Figure 1). The resulting β-sheets are topologically rewired, but the active site is notably conserved across the two versions of the fold, suggesting that RNA modification by NFACT domains is likely to be mechanistically similar to the DNA glycosylases [76]. More generally, while simple circular permutation proceeds via duplicated intermediates, more complex reorganizations might arise from higher order n-plications or domain insertions [69].

Major topological and spatial transmogrifications

The prevalence of these extreme structural modifications is difficult to objectively estimate because evolutionary information necessary for such assessments is typically irretrievably lost [77]. However, in certain cases structural data can be combined with evidence from sequence analysis and contextual information from domain architectures and/or genomic context to make a confident case for such transformations [15,34,72]. Consequently, evidence for several such transformations is emerging from careful case-by-case analysis of structural data. Recently such a transformation was noted in the most common version of the α-helical HEPN RNase domains found in type-IIIA CRISPR systems [78]. Here the HEPN domain is combined with a winged helix-turn-helix domain and has undergone a complex structural reorganization, while retaining the catalytic residues intact [79,80].

Striking examples have emerged from the study of immunity proteins, which neutralize the toxin domains of the recently described prokaryotic polymorphic toxin systems that are deployed by cells to kill intraspecific competitor cells [57]. In these systems there is strong genomic coupling of genes coding for immunity proteins and gene cassettes coding for polymorphic toxin domains. Further, there is evidence for rapid divergence of immunity proteins in response to the polymorphism in the toxin domains that they neutralize [24,57]. Hence, these are fertile grounds for rapid evolutionary changes as immunity proteins diversify to adapt to new toxins. In these systems several distinct types of immunity proteins have been described, of which the SUKH and SuFu domains are the first and second most common immunity proteins [24]. While SUKH and SuFu display superficially different folds [24,81], a closer examination suggests that their core β-sheets are similarly organized (Figure 4): in both cases it is a “split sheet”, i.e. one with a gap between distinct four-stranded and three-stranded elements. This, together with their functional equivalence and cognate genomic organization in the polymorphic toxin loci suggests that they have indeed diverged via drastic structural modification. In the case of the SUKH and SuFu domains this can be reconstructed as involving circular permutation followed by a rotation of the three-stranded element resulting in it being flipped in orientation with respect to the four-stranded sheet (Figure 4). Recently, the structure of a novel immunity protein, CdiI/Imm75, from the Enterobacter cloacae polymorphic toxin system was published [82]. While this was reported as a novel fold, it shows the same structural organization as the SUKH and SuFu domains in being comprised of equivalent four-stranded and three-stranded elements (Figure 4). Thus, CdiI/Imm75 is yet another immunity protein sharing a common ancestry with the SUKH and SuFu domains via major structural modification. In this case it appears to have involved rotation of the three-stranded unit with it now partly stacking with the 4-stranded element to form an open sandwich-like fold (Figure 4).

Figure 4. Topological and spatial transmogrifications in SUKH, SuFu and Imm75 families.

The shared structural elements of these families are equivalently colored in the topological and cartoon representations. The type of structural transition with respect to the SUKH domain is shown above the topology.

Prevalence of function-preserving structural modifications in biological conflict systems and eukaryotic macromolecule modification systems

In the early days of structural studies such drastic modifications were initially regarded as curiosities or rare quirks of evolution [34,70]. However, the sheer wealth of currently-available structure and sequence data are beginning to reveal certain patterns in terms of biological systems where such modifications tend to be overrepresented. While some of the above-mentioned structural modifications, such as that in the nucleic acid ligases and DNA glycosylases, are relatively ancient events in core cellular systems [74,76], the majority of these can be traced to biological conflict systems (Figure 5). These systems are deployed in: 1) intra-genomic conflicts, e.g. toxin-antitoxin and restriction-modification systems [83,84]; 2) inter-genomic conflicts, e.g. the CRISPR/Cas system involved in restricting invasive genomes like bacteriophages [68,78]; 3) intra-specific conflicts, e.g. polymorphic toxin systems [24,57]; 4) inter-specific conflicts, e.g. antibiotics and toxins deployed by bacteria against competitors or hosts, and mechanisms of immunity against them [57,85,86]. Several enzymatic and non-enzymatic domains from these systems, including aminoglycoside kinases, E1-like adenylating enzymes, deaminases, ADP-ribosyltransferases, restriction endonucleases, HNH endonucleases, diverse RNases and immunity proteins from such systems show drastic structural modifications while preserving their biochemical function (Figure 5). Although the majority of instances of such diversification are currently known from prokaryotic conflict systems, traditional immunity systems of animals such as the antibodies and variable lymphocyte receptors also show evidence for such modifications [85,86]. Intense selective pressures found in these conflict systems, which directly impinge on the survival of the organism, have resulted in arms races thereby favoring rapid emergence of innovations [87]. In these situations divergence is often critical for evading counter-adaptations of the rival systems locked in the arms race. Thus, in these systems selection might balance structural divergence (necessary for evasion) with preservation of function (cannot be compromised for organismal survival).

Figure 5. Biological context of proteins showing major structural variations.

Key components of each biological system are illustrated with the proteins and domains showing major structural variations highlighted in light blue.

In eukaryotes such structural diversification is additionally encountered in systems primarily involved in introduction and “reading” of modifications in macromolecules such as DNA, proteins (histones and tubulins) or lipids [49,50,88,89] (Figure 5). In many cases these modifications serve as epigenetic marks that encode information over and beyond the genetic material. Thus, examples of these structural modifications are found in DNA-modifying enzymes, peptide recognition domains and enzymes of the ubiquitin system (Figure 5). The diversification in these systems might be seen as a parallel to the arms race scenario in conflict systems. Here, proliferation of signals based on modified peptides and their utilization as new epigenetic marks probably act as drivers of innovation similar to counter-adaptations of rival systems in biological conflicts.

Concluding remarks

Detection of major structural modifications led to the proposal that the classification of proteins on the basis of folds is not accurate [77,90]. However, as examples of such modifications accumulate it is becoming clear that such an extreme view is unwarranted. Rather, it merely suggests that protein structures can be either plastic while preserving function (Figure 1) or relatively refractory to structural change while diverging in function [17,22]. A closer examination reveals that the former is a strategy that is exploited in specific circumstances, particularly in systems pertaining to biological conflicts or modifications of macromolecules (Figure 5). Thus, plasticity or conservation of a fold is primarily a reflection of the type of selective forces operating on the protein in the context of the biological system in which it functions. In terms of further studies, a better understanding of the interplay between selective forces and such structural transformations would be of particular interest. These transformations could also provide useful information for guiding future protein-engineering efforts. Hence, we hope that this survey of a dramatic but under-appreciated tendency in protein evolution inspires further studies.

Highlights.

Drastic structural transformation of protein domains while retaining their basic biochemical function is a notable phenomenon in protein evolution.

These transformations span a wide spectrum of structural changes including complex modifications of domain structure.

Domains displaying these transformations are most commonly observed in systems involved in biological conflicts, eukaryotic chromatin-related protein/DNA modification and peptide recognition.

These systems show “arm-race”-like situations that could favor drastic structural divergence while preserving function