Oligomeric interactions maintain active‐site structure in a noncooperative enzyme family

Abstract

The evolutionary benefit accounting for widespread conservation of oligomeric structures in proteins lacking evidence of intersubunit cooperativity remains unclear. Here, crystal and cryo‐EM structures, and enzymological data, demonstrate that a conserved tetramer interface maintains the active‐site structure in one such class of proteins, the short‐chain dehydrogenase/reductase (SDR) superfamily. Phylogenetic comparisons support a significantly longer polypeptide being required to maintain an equivalent active‐site structure in the context of a single subunit. Oligomerization therefore enhances evolutionary fitness by reducing the metabolic cost of enzyme biosynthesis. The large surface area of the structure‐stabilizing oligomeric interface yields a synergistic gain in fitness by increasing tolerance to activity‐enhancing yet destabilizing mutations. We demonstrate that two paralogous SDR superfamily enzymes with different specificities can form mixed heterotetramers that combine their individual enzymological properties. This suggests that oligomerization can also diversify the functions generated by a given metabolic investment, enhancing the fitness advantage provided by this architectural strategy.

Oligomeric structures are conserved in many proteins that do not show evidence of intersubunit cooperativity and contribute to enhancing evolutionary fitness.

Introduction

Protein oligomerization, or assembly into stable multisubunit complexes, is extremely common and often preserved in protein families and even superfamilies (Goodsell & Olson, 2000; Pereira‐Leal et al, 2007; Zhang et al, 2010; Finnigan et al, 2012; Matthews, 2012; Garcia‐Seisdedos et al, 2017). However, the structural and functional factors driving the evolution and conservation of oligomeric interactions remain unclear for many protein families. One well understood reason for oligomerization is to produce cooperativity in ligand binding or enzyme activity, which changes the shape of the activity profile versus ligand/substrate concentration (Perutz et al, 1998; Sigler et al, 1998; Williams et al, 1998; Hunter & Anderson, 2009; Finnigan et al, 2012). Cooperativity derives from a global conformational change in an oligomeric assembly that simultaneously increases the affinity of the ligand/substrate‐binding sites in all subunits (Monod et al, 1965). A seminal example is provided by mammalian hemoglobin, a heterotetramer in which ~4‐fold cooperativity produces a steeply sigmoidal binding curve that directly contributes to maintaining tight oxygen homeostasis (Perutz et al, 1998). Another well‐studied example is HIV protease, a dimer that exhibits weaker cooperativity (< 2‐fold) (Alvizo et al, 2012). However, the functional significance of the cooperativity remains unclear in this case, suggesting that it could be a secondary property of a conserved dimer structure that evolved for a different reason. In some protein superfamilies, functional binding sites reside directly in conserved intersubunit interfaces (Finnigan et al, 2012; Venkatesan et al, 2014), or the conserved oligomeric assembly makes functional multivalent interactions (Lin & Rye, 2006; Hochberg et al, 2018), providing obvious explanations for conservation of oligomeric structure. Heterooligomeric complexes can also combine activities of the constituent domains and improve spatial coordination of their activities (Goodsell & Olson, 2000; Ramsay et al, 2002; Pereira‐Leal et al, 2007; Venkatesan et al, 2014; Hochberg et al, 2018). However, there are many protein families and superfamilies with conserved homo‐oligomeric structures that do not exhibit any evidence of cooperativity, that have all of their functional interaction sites far from their intersubunit interfaces, and that do not mediate multivalent interactions (Oppermann et al, 2003; Kristan et al, 2005). The reason for the evolution and conservation of these oligomeric structures remains unclear, especially in the case of enzymes acting on small molecule substrates.

It has frequently been suggested that oligomerization helps stabilize the functional protein structure (Hoffmann et al, 2007; Venkatesan et al, 2014; Alvarez‐Cao et al, 2018). However, the experimental support for this hypothesis relies on indirect evidence, e.g., higher order oligomerization of hyperthermophilic proteins compared with mesophilic orthologs (Sterner & Liebl, 2001; Vieille & Zeikus, 2001) or mutations that could potentially destabilize subunit structure in addition to disrupting intersubunit interactions (Schliebs et al, 1997; Sterner & Liebl, 2001; Vieille & Zeikus, 2001; Schwab et al, 2008). The specific physicochemical mechanisms and architectural features responsible for the stabilization of protein structure by oligomeric interactions remain incompletely characterized.

Recent studies have employed elegant biophysical approaches (Garcia‐Seisdedos et al, 2017) and large‐scale computational (Hoffmann et al, 2007; Akiva et al, 2008; Levy et al, 2008; Dayhoff et al, 2010; Hashimoto & Panchenko, 2010; Zhang et al, 2010; Hashimoto et al, 2011; Chen et al, 2019) and experimental (Levy et al, 2008; Aakre et al, 2015; Chen et al, 2019) methods to characterize the physicochemical factors and evolutionary pathways controlling stable oligomer formation. Despite the impressively deep mechanistic understanding provided by these studies, the selective advantage provided by oligomerization has not been established for many protein superfamilies and especially for noncooperative oligomeric enzyme families. Notably, a recent paper concludes that some conserved oligomeric interactions are gratuitous and represent a vestigial architectural feature that persists evolutionarily due to the inability of point mutations to generate a soluble monomer once a large hydrophobic surface area is buried in an intersubunit interface (Hochberg et al, 2020). We herein establish clear counterexamples to both these interrelated structural and physicochemical inferences about the evolution of oligomeric interactions. We experimentally demonstrate that homo‐oligomeric interactions can directly stabilize functional protein structures in a manner that promotes evolutionary fitness by enabling smaller protein chains requiring less metabolic investment to mediate more diverse biochemical functions, and we furthermore demonstrate that mutational disruption of conserved oligomeric interactions can readily generate biochemically well‐behaved sub‐assemblies.

One notable example of the conservation of oligomeric interactions in a noncooperative protein superfamily is provided by the Short‐chain dehydrogenases/reductases (SDRs) (Oppermann et al, 2003; Matthews, 2012), which have 146,793 representatives in the PFAM database (PF13561) (Bateman et al, 2002). Enzymes in this pervasively distributed superfamily reduce prochiral ketones to chiral alcohols, many of which are valuable pharmaceutical intermediates used to synthesize drugs (Matsuda et al, 2009; Chen et al, 2016). The core of SDR enzymes comprises a classical Rossmann fold that binds NAD(P)/H via a GxxxGxG motif and activates hydride transfer using a conserved tyr‐ser‐lys triplet (Oppermann et al, 2003). Almost all enzymes in the SDR superfamily form homo‐oligomers (Matthews, 2012), which we demonstrate below to share conserved intersubunit interactions. To dissect the reason for the conservation of homo‐oligomerization in this superfamily, we employed two previously characterized paralogous enzymes, (S)‐carbonyl reductase (SCR) and (S)‐carbonyl reductase 2 (SCR2) from Candida parapsilosis, which share 85% sequence identity (Nie et al, 2011). These enzymes both have broad substrate specificity, but they have significantly different substrate/activity profiles. One specific example is that they both reduce 2‐hydroxyacetophenone to (S)‐1‐phenyl‐1,2‐ethanediol but do so with substantially different catalytic efficiency. The results presented below prove that the conserved homo‐oligomeric interactions in the SDR enzyme superfamily maintain the functional active‐site structure and furthermore that hetero‐oligomerization of two paralogous SDR superfamily enzymes combines both of their activities in a single protein assembly. These results provide a rigorous, but simple demonstration that conserved oligomeric interactions can enhance evolutionary fitness by stabilizing functional protein structure.

Results

Crystal structures of SDR superfamily enzymes show strongly conserved tertiary and quaternary structures

We cloned (Appendix Table S1), purified (Appendix Fig S1), and determined X‐ray crystal structures (Fig 1 and Appendix Tables S2‐S4) for the paralogous SCR and SCR2 enzymes from C. parapsilosis. These proteins were purified using N‐terminal hexahistidine tags that were proteolytically removed prior to crystallization at 20°C in similar low‐salt PEG solutions near‐neutral pH (Appendix Table S2). Synchrotron diffraction data were collected from frozen crystals at 100 K (Appendix Table S3). The structures were solved using molecular replacement and refined to resolutions of 1.75 Å for apo SCR (R _free 22.9%), 1.59 Å for the SCR/NADPH complex (R _free 19.2%), 1.79 Å for the SCR2/NADPH complex (R _free 17.1%), and 1.89 Å for the SCR2/NADP complex (R_free 23.0%) (Appendix Table S4). The models contain all residues in the crystallized constructs from their N‐ through C‐termini, except for a short segment of an internal loop covering the active site that is disordered in some of the structures (residues 222–224 in apo SCR, 222–226 in SCR2/NADPH, and 222–225 in SCR2/NADP). The models have excellent geometry and 94.8–96.5% of residues in the favored regions of the Ramachandran plot (Appendix Table S4). The asymmetric units in all of the crystal structures contain two subunits from the physiological homotetramer formed by both enzymes (Fig 1C and D, Appendix Fig S1). This homotetramer has D2 symmetry, alternatively called 222 symmetry, meaning three orthogonal 2‐fold axes intersect at the center of the homotetramer (Fig 1C).

Figure 1. Comparison of C. parapsilosis SCR and SCR2 crystal structures.

1. The A subunits from apo SCR (left), its binary complex with NADPH (center left), and binary complexes of SCR2 with NADPH and NADP (center right and right, respectively). The core of SCR is shown in shades of red, while its Tetramer Interface Helices (TIHs—residues 122–144 and 172–197) are shown in shades of magenta. The core of SCR2 is shown in shades of blue, while its TIHs are shown in shades of cyan. The active‐site lids (residues 221–227) are colored orange, and the C‐terminal residue in each subunit is shown in green space‐filling representation in all structures. NADPH and NADP are shown in gray and black space‐filling representations, respectively.
2. Least‐squares superpositions of the A subunits according to the colored arrows.
3. The corresponding physiological tetramers.
4. Three orthogonal views of a least‐squares superposition of those four tetramers.

The four crystal structures show extremely strong conservation in both tertiary structure and quaternary structure, and there are minimal changes in backbone conformation (Fig 1) or B‐factors (Fig EV1) between the different substrate/product complexes of the two paralogs. Superpositions of subunits from the different structures show similar Cα root‐mean‐square deviations (RMSDs) as superpositions of the two subunits in the asymmetric units in the same structures. All‐residue superpositions show 0.14–1.31 Å RMSD for subunits in different structures compared with 0.20–0.91 Å RMSD for subunits in the same structures (Appendix Table S5A), while core superpositions including at least 83% of the residues show 0.10–0.29 Å RMSD for subunits in different structures compared with 0.05–0.19 Å RMSD for subunits in the same structures (Appendix Table S5B). There are no significant differences in the backbone B‐factors in the different structures except for local increases in the apo SCR structure in two very short loops that directly contact NADPH (Fig EV1A). Therefore, the active sites in the SCR paralogs are extremely rigid, and there is no significant induced‐fit conformational change upon NADP/H binding.

Figure EV1. B‐factor analysis of SCR and SCR2 crystal structures.

1. B‐factor putty representations of the A subunits from the apo SCR (left), SCR/NADPH (center left), SCR2/NADPH (center right), and SCR2/NADP (right) crystal structures. The color and thickness of the backtraces encode the backbone B‐factors according to the default scheme in PyMOL (Tao et al, 2020), with thin/blue and thick/red representing the segments with the lowest (least dynamic) and highest (most dynamic) B‐factors in each structure, respectively. NADPH and NADP are shown in gray and black space‐filling representations, respectively. The asterisks in the apo SCR structure indicate two short protein segments (T120‐V121 and T67‐H68‐P690) that directly contact the NADPH co‐substrate and show reduced B‐factors and backbone dynamics upon NADPH binding.
2. Least‐squares superpositions of the same representations of the A subunits but with apo SCR colored red, SCR/NADPH colored maroon, SCR2/NADPH colored blue, and SCR2 colored cyan.
3. B‐factor putty representations of physiological tetramers from the four crystal structures using the same color/thickness coding as the corresponding subunits in panel A.

The SCR and SCR2 subunits adopt a standard Rossmann fold, like many NAD(P)/H‐dependent enzymes, but with five significant additions that all contribute to forming the enzyme active site or the intersubunit interfaces in the physiological homotetramer or both (Figs 1 and EV2). The canonical Rossmann fold comprises 6 consecutive α/β super‐secondary structural elements that form a central parallel β‐sheet with β‐strand order 3‐2‐1‐4‐5‐6 (relative to their position in the primary sequence). One addition to the canonical Rossmann fold is a 34‐residue N‐terminal extension (η0 and α0 in Fig EV2) that adopts an extended conformation and makes extensive intersubunit contacts contributing to all interfaces in the homotetramer. Another addition is an extra C‐terminal β‐strand (β7 in Fig EV2) that extends the β‐sheet by making parallel hydrogen‐bonding (H‐bonding) interactions to strand β6 while simultaneously mediating dimer formation via water‐mediated H‐bonds across a proper two‐fold symmetry axis to the equivalent β‐strand in another subunit (right in Fig 1D and Appendix bottom of Fig S2A). This dimer interface buries 2,606 Ų of solvent‐accessible surface per subunit (Appendix Table S6). The intersubunit interactions mediated by these two additions to the canonical Rossmann fold that are conserved among SDR superfamily enzymes suggest that the physiological tetramer contributes in some important way to enzyme evolution or function.

Figure EV2. SCR topology diagram.

1. Topology diagram colored equivalently to apo SCR in Fig 1A in the main text with arrows representing β‐strands and cylinders representing η‐helices (3₁₀‐helices) and α‐helices. Residues making contacts in the dimer and tetramer interface are indicated by the letters D and T, respectively, with residues making hydrogen bonds shown in black bold text and residues making exclusively van der Waals contacts shown in gray plain text. The lengths of the α‐helices and β‐strands in this schematic are proportional to their lengths in the 3‐dimensional structure. The schematic was modified from the output from PDBSUM (Laskowski et al, 2018), while the contact analysis was performed by PISA (Krissinel & Henrick, 2005).
2. Copy of Fig 1A in the main text to facilitate comparison to the topology diagram.

The remaining three additions to the canonical Rossmann fold form most of the active sites in the SCR/SCR2 subunits while simultaneously mediating additional intersubunit interactions in the physiological tetramer (η4, η5, and α(α/β6)1 in Fig EV2). The fourth and fifth β‐strands in the canonical Rossmann fold are followed by, respectively, ~23 and ~26 residue insertions (magenta and cyan in Figs 1 and EV2) that form long loops followed by ~10 residues N‐terminal extensions of the α‐helices in these α/β elements. These insertions form a substantial portion of the active‐site cavity (Figs 1A and B, and EV2), and their α‐helical segments simultaneously mediate tetramer interactions (Figs 1C and EV2) between the dimers formed via the self‐interaction between the C‐terminal β‐strands. The α‐helices in these α/β elements both pack with the equivalent α‐helices in another subunit in the tetramer (Fig 1C and Appendix top of Fig S2A), forming an interface that buries 2,056 Ų of solvent‐accessible surface area per subunit (Appendix Table S6). The interactions mediating the formation of this tetramer interface occur across a proper two‐fold symmetry axis that is orthogonal to the two‐fold symmetry axis that relates the constituent dimers (Fig 1C and D), and these two orthogonal axes produce the overall D2 symmetry of the physiological tetramer.

The dominant contacts in the interfaces mediating tetramer formation are made by the N‐terminal extensions of α‐helices 4 and 5 that follow β‐strands 4 and 5 in the Rossmann fold, leading us to call these segments the “Tetramer Interface Helices” (TIHs—magenta or cyan in Figs 1, 2A, and EV2). The opposite surfaces of these helices make direct contacts to NADP/H and also contribute to forming the binding cavity for the primary ketone/alcohol substrate/product of the enzymes. The contribution of the TIHs to active‐site formation in both SCR paralogs suggests that tetramer formation contributes directly to maintaining active‐site structure, an inference we confirm below.

Figure 2. Conservation of quaternary structure between the SCR paralogs from C. parapsilosis and remote homologs in the short‐chain dehydrogenase/reductase (SDR) enzyme superfamily.

1. Sequence alignment of SCR and SCR2 with four homologs (labeled by PDB ID) having ~30% sequence identity to SCR that were selected based on their wide phylogenetic separation from one another among the 72 homologs in the PDB with ≥ 30% sequence identity to SCR (Appendix Fig S3).
2. Maximum likelihood cladogram generated by the program MEGA (50).
3. Three orthogonal views of a least‐squares superposition of the A subunits from the crystal structures of SCR (red/pink), SCR2 (dark/light blue), Lactobacillus brevis alcohol dehydrogenase (PDB ID 6HLF, teal/cyan), Alcaligenes faecalis D‐3‐hydroxybutyrate dehydrogenase (PBD ID 5B4V, orange/yellow), Thermotoga maritima Gluconate 5‐dehydrogenase (PBD ID 1VL8, brown/tan), and Bacillus megaterium Glucose dehydrogenase (PDB ID 3AUS, dark/light green). TIHs in each structure are shown in a lighter shade of the color used for its core.
4. Three orthogonal views of a least‐squares superposition of the corresponding physiological tetramers.

The active‐site cavity in the SCR paralogs is completed by a 30‐residue protein segment (residues 217–246) inserted between the β‐strand and α‐helix in the sixth α/β element in the Rossmann fold (Fig EV2). Residues 221–227 in this insertion form the flexible lid over the active‐site cavity (Fig 1A) alluded to above, while its C‐terminal segment includes an α‐helix (residues 229–238) that makes extensive intersubunit packing interactions in the dimer interface, again pointing to the architectural linkage between oligomerization and active‐site structure.

The homotetramer interface in the SCR paralogs is strongly conserved in the SDR superfamily

To gain insight into the significance of the homotetramer structure formed by SCR and SCR2, we evaluated its evolutionary conservation. We searched the Protein Data Bank (PDB) for deposited sequences with ≥ 30% identity to SCR that are annotated as being NAD(P)/H‐dependent oxidoreductases, which yielded 72 unique sequences of proteins of known structures (Appendix Fig S3). All of these homologs, which share 30–50% sequence identity with SCR over the structurally aligned region, preserve the same subunit fold, and 68 of them also preserve the same homotetramer structure. Figure 2A–C analyzes the sequence and evolutionary relationships between four representative remote homologs to SCR: 3‐hydroxybutyrate dehydrogenase from Alcaligenes faecalis (PDB ID 5B4V, 30.0% identity) (Kanazawa et al, 2016), glucose dehydrogenase 4 from Bacillus megaterium (PDB ID 3AUS, 30.2% identity) (Nishioka et al, 2012), gluconate‐5‐dehydrogenase from Thermotoga maritima (PDB ID 1VL8, 30.2% identity), and alcohol dehydrogenase from Lactobacillus brevis (PDB ID 6HLF, 32.2% identity) (Nowotny et al, 2019). While some of these enzymes show local structural differences near the active‐site lid, the subunits otherwise share the same topology and tertiary structure (Fig 2C), and they show 0.69–0.96 Å Cα RMSD for least‐squares alignments with SCR covering > 67% of their structures (Appendix Table S5C). Furthermore, all four of these remote homologs form the same homotetramer structure as SCR, and they share geometrically similar interfaces with equivalent interactions between both the C‐terminal β‐strands mediating dimerization and the TIHs in the fourth and fifth β/α units in the Rossmann fold that mediate tetramer formation (Fig 2D).

Three of the four oxidoreductases in the PDB with ≥ 30% sequence identity to SCR that do not conserve its homotetramer architecture still conserve the oligomeric interface formed by the TIHs (e.g., PDB ID 1GZ6) but not the dimer interface formed by the C‐terminal β‐strand in SCR. The loss of this interface otherwise conserved in the SDR superfamily members is compensated in these homologs by an ~60‐residue extension at the C‐terminus of the Rossmann fold that forms additional intersubunit packing interactions across the conserved interface stabilized by the TIHs. Notably, the TIHs in these remote homologs still contribute to forming the active‐site structure.

Replacement of the conserved tetramer interface by an additional subdomain in an exceptional SDR superfamily member

The only oxidoreductase in the PDB with ≥ 30% sequence identity to SCR that does not conserve the symmetrical intersubunit packing interactions made by the TIHs is 3‐α‐hydroxysteriod dehydrogenase/carbonyl reductase (3α‐HSD/CR) from Comamonas testosterone (PDB ID 1FK8, 30.0% identity to the structurally aligned region in SCR) (Grimm et al, 2000). This enzyme conserves the dimer interface in SCR mediated by the C‐terminal β‐strand, but it replaces the intersubunit packing interactions between the TIHs with packing interactions to a unique 34‐residue subdomain in the same subunit (Fig EV3) that buries 826 Ų of accessible surface area. This subdomain, which blocks the formation of the tetramer, is inserted in the primary sequence immediately preceding the second TIH. The equivalent segment forms part of the active‐site cavity in the homologous enzymes, but there is no active‐site cavity in 3α‐HSD/CR, which instead has the NAD/H co‐substrate directly exposed to bulk solvent. The relaxed structural constraints on the solvent‐exposed active site of 3α‐HSD/CR seem likely to be related to the alternative architectural strategy used to stabilize the equivalent of the TIHs in this enzyme compared with all of the other 71 homologous enzymes in the PDB (Appendix Fig S3).

Figure EV3. Three orthogonal views of the SCR and 3α‐HSD/CR subunits.

The A subunit from the NADPH complex of SCR is shown on the left, while the A subunit from the NAD complex of 3‐α‐hydroxysteriod dehydrogenase/carbonyl reductase (3α‐HSD/CR) from Comamonas testosterone is shown on the right (PDB ID 1KF8). A superposition of the two structures in the same orientation is shown in the center of each row. SCR is colored as in Fig 1. Most of the backbone of 3α‐HSD/CR is colored light gray, except for its TIHs, which are colored light blue, and the inserted 34‐residue subdomain that stabilizes them which is colored dark blue. NADPH and NAD are shown in gray and black space‐filling representations, respectively.

The replacement of the tetramer interface in SCR with an alternative structural interaction of similar spatial extent in the only homologous enzyme with ≥ 30% sequence identity that does conserve this interface suggests that the TIHs require stabilization from some other structural element to maintain their conformation, as previously suggested without experimental support (Hoffmann et al, 2007). The 34‐residue size of the structural element in 3α‐HSD/CR that replaces the intersubunit interactions in the tetramer interface in the other SDR superfamily enzymes provides an estimate of the reduction in domain size and metabolic cost (~12%) provided by their conserved oligomeric interactions. Notably, the C‐terminal extension in SCR and other SDR superfamily enzymes compared to the canonical Rossman fold is ~60 residues longer in the three homologs alluded to above that do not conserve the dimer interface while conserving the tetramer interface. Therefore, the relatively rare (4/72) homologs in the PDB with ≥ 30% identity sequence to SCR that do not conserve an equivalent tetramer structure all contain significantly longer subunits, establishing an empirical evolutionary correlation between loss of oligomeric interactions and an increase in subunit length and therefore the metabolic cost of subunit biosynthesis.

A C‐terminal hexahistidine tag disrupts the phylogenetically conserved tetramer interface and alters enzyme activity

To investigate the role of tetramer formation in maintaining active‐site structure in SCR and other SDR superfamily enzymes, we selectively disrupted its tetramer interface. Given that the C‐terminus of the enzyme is partially buried in this interface, C‐terminal extensions would be expected to disrupt tetramer formation. Therefore, we characterized the oligomerization (Fig EV4A), structural (Fig 3), and enzymatic (Fig 4) properties of an SCR construct with a C‐terminal hexahistidine tag. Analytical gel filtration chromatography with in‐line multiangle light scattering confirms our inference that this extension disrupts the native tetramer structure (Fig EV4A); specifically, 17 µM apo SCR‐his₆ forms a mixture of 24% tetramer and 76% dimer compared with 98% tetramer and 2% dimer for an equivalent concentration of SCR with the native C‐terminus. The clean separation of these species in gel filtration indicates that the equilibration between the biochemically well‐behaved tetramer and dimer occurs relatively slowly on the ~30‐min timescale of this analysis, suggesting that a conformational change in the tetramer interface in the dimer state may create a kinetic barrier to tetramerization.

Figure EV4. Hydrodynamic and mass‐spectrometric analyses of the homo‐oligomers and hetero‐oligomers.

1. Size‐exclusion chromatography—multiangle light scattering (SEC‐MALS) analysis of homo‐oligomers. A 100 µl volume of protein at 0.5 mg/ml (17 µM subunit concentration) was injected on a Superdex 200 Increase 10/300 column equilibrated at 4°C in 150 mM NaCl, 50 mM Tris–Cl, pH 6.0 flowing at 0.5 ml/min. The calculated molecular weight is plotted in green according to the scale on the right vertical axis, while 90° static light scattering (red) and refractive index chromatograms (blue) are plotted on arbitrary scales adjusted so that their maximum and minimum values coincide. The calculated molecular weight is proportional to the ratio of the 90° static light scattering and refractive index signals above their baseline values.
2. SEC‐MALS analysis of hetero‐oligomers. A 100 µl volume of protein at 0.5 mg/ml (17 µM subunit concentration) was injected on a Superdex 200 Increase 10/300 column equilibrated at 25°C in 150 mM NaCl, 50 mM Tris–Cl, pH 8.0 flowing at 0.5 ml/min. The calculated molecular weight is plotted in black according to the scale on the left vertical axis, while 90° static light scattering (red) and refractive index chromatograms (blue) are plotted the same way as in panel A. The inset shows Coomassie Blue‐stained SDS–PAGE analysis of the same protein sample, which retained both affinity tags.
3. Liquid‐chromatography time‐of‐flight mass spectrometry analysis of the SCR‐SCR2 heterotetramer after removal of the affinity tags using TEV protease following purification of the heterotetramer using sequential affinity chromatography.

Figure 3. Appending a C‐terminal hexahistidine tag disrupts the native tetramer interface in SCR, resulting in distortion of active‐site structure in the constituent subunits.

1. Comparison of crystal structures of the tetramers formed by untagged SCR (WT) versus SCR‐his₆ (C‐tag—PDB ID 3CTM) in the NADP/H‐free (apo) state. Untagged SCR is shown in red (core) and magenta (tetramerization helices), while SCR‐his₆ is shown in gray (core), purple (TIHs in subunits A/D), and cyan (TIHs in subunits B/C). The C‐terminal residue in untagged SCR (pro‐279) is shown in light green space‐filling representation, while the corresponding residue plus the crystallographically observed residues from the C‐terminal tags in SCR‐his₆ are shown in black space‐filling representation. The structurally unique dimers in each tetramer are shown above/below the complete tetramers in the center. A superposition of the full tetramers is shown at center left, and a zoomed‐in view of the interface between the B/D subunits in that superposition is shown at center right.
2. Superpositions of the subunits colored in the same way. Residues S172 and Y187 in the conserved catalytic tyr‐ser‐lys triplet are shown in space‐filling representation colored like the TIHs in each subunit.

Figure 4. Kinetic analyses of NADPH‐dependent reduction of prochiral ketones by SCR and SCR‐his₆  .

1. Michaelis‐Menten analysis of the reduction of 2‐hydroxyacetophenone (2‐HAP).
2. Comparison of k _cat values obtained from Michaelis‐Menten analyses of the reduction of four different substrates. Their chemical structures are shown in Appendix Fig S6A. The initial velocities (V₀/[E_tot]) and turnover rates (k _cat) are given normalized to subunit, not oligomer, concentration. Enzyme assays were conducted at 30°C in 100 mM sodium phosphate (pH 6.0) using 500 µM NADPH and 100 nM SCR or 1 µM SCR‐his₆. The specific activity of SCR does not change when enzyme subunit concentration is varied over a 10‐fold range from 10–100 nM (Appendix Fig S11). Enzyme activity at each substrate concentration was assayed at least three and usually four times. The data points show the average values for these technical replicates, while the error bars show the standard deviations.

Source data are available online for this figure.

The crystal structure of apo SCR‐his₆ shows an asymmetrical tetramer related to that formed by native apo SCR but with two of its four subunits pushed apart across the tetramer interface (Fig 3A). That interface is expanded by ~7 Å at its midpoint and ~11 Å at one edge (the lower edge of the tetramer as oriented in the bottom half of Fig 3A), disrupting the packing interaction across the tetramer interface for two of the four subunits and therefore the D2 symmetry of the native tetramer. The tetramer interfaces in this structure bury 342 Ų and 1,048 Ų of solvent‐accessible surface area per subunit compared with an average of 2,052 Ų in the equivalent interfaces in the native tetramer, while the dimer interfaces bury an average of 2,740 Ų compared with 2,606 Ų in the equivalent interfaces in the native tetramer (Appendix Table S6). Therefore, while the dimer interfaces show only minor distortion, both of the tetramer interfaces show dramatic structural changes (Appendix Fig S2B), and the two subunits that are pushed further apart in SCR‐his₆ make minimal contacts across the tetramer interface. Analysis of the sequences of the 72 SDR enzymes described above with structures in the Protein Data Bank demonstrates that lengthening of the C‐terminus in a small number of superfamily members is coupled to a shortening of the protein loop that it contacts across the tetramer interface (e.g., PDB ID 3AUS in Appendix Fig S4), consistent with evolutionary pressure to preserve the intersubunit separation and thereby the complementarity of the packing interactions across this interface.

Examining the conformation of the constituent subunits in the apo SCR‐his₆ structure shows that they all have significant conformational changes compared with native apo SCR (Fig 3B). Least‐squares alignment of all Cα atoms in the SCR‐his₆ subunits with the native subunits shows 1.87–3.77 Å RMSD compared with 0.63 Å when aligning the two subunits in the native structure with each other (Appendix Table S5A). In contrast, the least‐squares alignment of Cα atoms in the core of the Rossmann fold shows 0.31–0.36 Å RMSD after outlier rejection for 71–78% of the residues in the native SCR subunits compared with the SCR‐his₆ subunits (Appendix Table S5B). Therefore, disruption of the tetramer interface by the C‐terminal hexahistidine tag distorts the structure of some segments in all of the constituent subunits, while their core structure is preserved. Examining the superimposed structures shows that the subunits that retain significant contact across the disrupted tetramer interface in SCR‐his₆ show significant distortions in the conformations of the loops preceding the two TIHs but not in the conformation of those α‐helices themselves (purple in Fig 3B). Those loops make significant contributions to the active‐site structure, suggesting that even these subunits are likely to have altered catalytic properties. The subunits in the SCR‐his₆ structure that are pushed further apart across the tetramer interface show even greater structural distortions in the vicinity of the active site (cyan in Fig 3B), including significant perturbations in the NADP/H binding site. The N‐termini of the TIHs, which represent extensions compared with the α‐helices at the equivalent positions in the canonical Rossmann fold, are both unfolded, and the loops preceding them both adopt severely distorted conformations. Therefore, disruption of the tetramer interface in SCR leads to a complete collapse in active‐site structure due to destabilization of the native conformation of the TIHs. These observations confirm our inference that the conserved tetramer interactions in the SDR superfamily stabilize the conformation of the TIHs and the enzyme active site. This inference is additionally supported by the observation that the only other SDR superfamily member for which the structure of a C‐terminally tagged construct is available, enoyl‐ACP reductases III from Bacillus subtilis (PDB ID 3OIC) (Kim et al, 2011), shows a disruption of the tetramer interface coupled to a collapse in active‐site structure analogous to that observed in the SCR‐his₆ structure (Appendix Fig S5).

Michaelis‐Menten kinetic analyses on SCR‐his₆ demonstrate that its k _cat for the reduction of 2‐hydroxyacetophenone (2‐HAP) by NADPH is reduced over 100‐fold (Fig 4A) compared with native SCR. Notably, SCR‐his₆ completely loses activity on some substrates but not others compared with the native enzyme, and the rank‐order of k _cat is reversed for the substrates for which it retains activity, indicating that a clear change in specificity accompanies the large reduction in catalytic efficiency (Fig 4B). Therefore, the distortions and collapse in active‐site structure caused by disruption of the tetramer interface severely impair enzyme activity and also alter enzyme specificity. SCR‐his₆ exhibits a 4‐fold lower K _M for 2‐HAP compared with native SCR (Appendix Fig S6), indicating substrate‐binding affinity may be improved slightly by the distortion of the active‐site conformation, possibly because increased flexibility of the active‐site loops in SCR‐his₆ allows induced‐fit accommodation of the substrate in a catalytically sub‐optimal geometry. The tetramer interface in SCR thus plays a critical role in maintaining high catalytic efficiency by controlling the structure of the active site via stabilization of the native conformation of the TIHs.

Single‐particle cryo‐EM reveals the dynamic opening of the tetramerization interface of SCR

To gain further insight into the oligomerization behavior of native SCR in solution, we collected a cryogenic electron microscopy (cryo‐EM) dataset on a mixed product/substrate complex with the enzyme bound to NADP and ethyl 4‐chloroacetoacetate, an industrially important ketone substrate (Chen et al, 2016). While the 2‐dimensional class averages show predominantly a symmetrical tetramer resembling the tetramer observed in its crystal structures (Fig 1C), they also show that a significant proportion of the particles adopt an alternate conformation in which one half of the tetramer interface is pried open and EM density is lost for the TIHs at this site (top of Fig 5). Therefore, the simplest processing of the cryo‐EM dataset suggests that the tetramer interface in SCR dynamically fluctuates into a partially open conformation coupled to disordering of the TIHs, strongly supporting the conclusion above that the tetramer interface is required to maintain their structure and the structure of the active site in native SCR.

Figure 5. Cryo‐EM analysis of the conformation of the native SCR tetramer in solution.

1. Image‐processing workflow and statistics for 2D classification and 3D reconstruction/refinement in cryoSPARC (Punjani et al, 2017) of SCR deposited on holey gold grids at 3 mg/ml in 5 mM ethyl 4‐chloroacetoacetate, 0.5 mM NADP, 150 mM NaCl, 50 mM Tris–Cl, pH 8.0. Particles assigned to protein volumes during heterorefinement runs were retained, while those assigned to dummy volumes were discarded. The final maps (lower left) were generated using a nonuniform refinement of the particles and volumes after the 6^th iterative heterorefinement cycle (middle left). The final Open Tetramer Interface (OTI) and Asymmetrical Tetramer (AT) maps were generated using C1 symmetry applied to all particles assigned to those classes, while the final ST map was generated using D2 symmetry applied to particles with > 0.99 probability of being in that class (lower left). The OTI, AT, and ST maps are shown in blue, orange, and red, respectively. The gray ribbon diagrams under the transparent renderings of the maps at lower left show models from real‐space refinement in PHENIX (Liebschner et al, 2019). The X‐ray crystal structure of SCR/NADPH (PDB ID 7DLM) was used as the initial model for these refinements, and their final Ramachadran distributions are shown in the last row of the table below the maps. Least‐squares alignment of the refined structures indicates a 5.25° inter‐dimer rotation between the ST and OTI conformations, as schematized in the graphical abstract for this paper.
2. Frames showing the opening of the tetramer interface based on 3D Variability Analysis in cryoSPARC of combined particles from the ST, AT, and OTI classes. Three frames are shown representing the two extremes and the middle of a 10‐step intermediates analysis. The map is colored cyan, and the ribbon diagram shows the refined ST structure with the TIHs colored magenta and the rest of the structure colored red. A movie showing the full analysis is presented as Movie EV1 included within the Expanded View material in the online version of this paper.
3. Local resolution maps for the OTI, AT, and ST classes calculated by cryoSPARC.
4. Superposition of the ST cryo‐EM structure after real‐space refinement in PHENIX with the X‐ray crystal structure of the SCR/NADPH tetramer determined in space group C222₁ with two subunits in the asymmetric unit (PDB ID 7DLM). The backbones are represented by ribbons with their diameter and color representing the B‐factors in the refined structures according to the default encoding in PyMOL.

Complete 3‐dimensional classification of the cryo‐EM dataset was performed as schematized in Fig 5A, and the corresponding Fourier Shell Correlation (FSC) and orientation distribution plots are shown in Appendix Fig S7A–C. This analysis indicates that ~50% of the protein population adopts a Symmetrical Tetramer (ST) conformation closely matching the oligomer with D2 symmetry observed in SCR crystal structures (Fig 5D). The remainder of the population adopts conformations with altered intersubunit packing interactions and substantially weaker density in some or all of the active sites in the tetramer (Fig 5A–C). The failure to observe conformations of this kind in the extensive crystallographic studies of SDR superfamily enzymes suggests that the requirement for stable packing in a regular lattice may have suppressed observation of these conformational states in crystal structures and thereby biased understanding of the functional dynamics in this enzyme superfamily.

Cryo‐EM classification indicates that ~23% of the SCR population in solution adopts a diversity of Asymmetrical Tetramer (AT) conformations in which contacts are maintained in all of the interfaces, but the intersubunit packing geometry deviates from regular D2 symmetry. The final ~27% of the native enzyme population adopts the Open Tetramer Interface (OTI) conformation described above. The similarity in their refined orientational distributions (Appendix Fig S7C) suggests all three conformations have roughly equivalent interactions with the air‐water interfaces, making it unlikely that differential interactions with the interfaces substantially bias the conformational distribution observed in the cryo‐EM sample. The AT and OTI conformations exhibit substantially lower local resolution throughout the maps than the ST conformation (Fig 5C), suggesting that these states both exhibit significant conformational variability. Notably, the systematically weaker density in the active‐site region in the AT and OTI states suggests that the intersubunit fluctuations in these global conformational states are coupled to substantially greater conformational dynamics in the active sites in the tetramer. The most dramatic of these effects is the ostensibly complete disordering of the TIHs in the OTI state, as described above and illustrated in the 3D‐variability analysis movie in Movie EV1; keyframes from this movie are shown in Fig 5B.

Two SCR paralogs expressed by C. parapsilosis form a stable heterotetramer combining their individual enzymological properties

In earlier studies using affinity chromatography to purify epitope‐tagged SCR2 expressed in the native host C. parapsilopsis, we obtained mass‐spectrometric evidence for heterotetramer formation with SCR in vivo (unpublished data). Given the results above establishing the importance of oligomeric interactions in maintaining the catalytic activity of SDR superfamily enzymes, heterotetramer formation could represent a strategy to diversify the catalytic specificity of the minimal functional assembly of the enzyme. To evaluate this possibility, we co‐expressed SCR2 and SCR in E. coli with different short N‐terminal affinity tags (his₆ on SCR2 and a Strep‐tag on SCR), and we used sequential affinity chromatography to isolate his₆‐SCR2/strep‐SCR heterotetramers (Fig EV4B). Mass spectrometry analysis of these heterotetramers, which represented ~1/8^th of the total amount of protein expressed in vivo (~10 mg/l versus ~80 mg/l of all oligomeric species combined), shows equivalent intensities for his₆‐SCR2 and strep‐SCR (Fig EV4C), consistent with an equal content of the two subunits in the heterotetramer population.

To enable the subunit composition in individual heterotetramers to be analyzed using gel filtration chromatography, we co‐expressed in E. coli the same strep‐SCR construct with an alternative SCR2 construct with an N‐terminal his₈‐MBP fusion, which adds ~43 kDa of mass to that subunit. Gel filtration analysis of the resulting heterotetramers shows predominantly one hydrodynamic species migrating as expected for an oligomer containing two his₈‐MBP‐SCR2 subunits and two strep‐SCR subunits (Fig 6A). Random mixing of these differentially tagged paralogous subunits would be expected to yield three distinct hydrodynamic species (i.e., containing 1, 2, or 3 his₈‐MBP‐SCR2 subunits). Therefore, the dominance of a single oligomer likely to contain two SCR2 and two SCR subunits combined with the lability of the tetramer interface observed by cryo‐EM (Fig 5A–C) suggests the dominant heterotetramer species formed in vivo contains one his₈‐MBP‐SCR2 homodimer combined with one strep‐SCR homodimer.

Figure 6. Enzymological and structural characterization of the SCR‐SCR2 heterotetramer.

1. Analytical gel filtration of purified his₈‐MBP‐SCR2 (black), strep‐SCR (yellow), and his₈‐MBP‐SCR2/strep‐SCR on a Superdex 200 Increase column at 20°C in 150 mM NaCl, 50 mM Tris–HCl, pH 8.0. The inset shows Coomassie Blue‐stained SDS–PAGE analysis of the same protein samples.
2. Michaelis‐Menten kinetic analyses of NADPH‐dependent reduction of 2‐hydroxyacetophenone and methyl acetoacetate at 30°C in 100 mM sodium phosphate, pH 6.0. The turnover rate (V₀/[E_tot]) is given for the individual subunits (as in Fig 4). Enzyme activity at each substrate concentration was assayed at least three and usually four times. The data points show the average values for these technical replicates, while the error bars show the standard deviations.
3. Sequence alignment of SCR and SCR2 from residue 81 to residue 279. Residues 1–80 are identical. The cyan stars highlight substitutions of sidechains exposed in the active‐site cavity.
4. 2Fo‐Fc (blue) and Fo‐Fc (positive/negative in green/red, respectively) electron density maps for residues 268–272 in the X‐ray crystal structure of the NADPH complex of the SCR‐SCR2 heterotetramer refined using superimposed SCR and SCR2 homodimers (left) or alternatively single SCR‐SCR2 (center) or SCR2‐SCR (right) heterodimers as the coordinate model for the two subunits in the asymmetric unit of the crystal lattice (Appendix Table S7). The leftmost panel here corresponds to the statically disordered refinement reported in the leftmost column of Appendix Table S7, while the center and right panels here correspond to the refinements without static disorder reported in 4th and 5th columns from the left in that table.
5. Superposition of subunits of SCR (yellow) and SCR2 (green) in the heterotetramer crystal structure. All residues differing between the two paralogs are shown in space‐filling representation, in cyan in the active site and blue elsewhere. The bound NADPH and the C‐terminal residue in each subunit are shown in gray and black space‐filling representations, respectively.
6. The complete heterotetramer model from the SCR‐SCR2 crystal structure colored the same way. This model correspond to the statically disordered refinement reported in the leftmost column of Appendix Table S7. It represents the most likely tetrameric structure in the lattice based on the hydrodynamic data shown in panel A combined with analysis of the alternative refinement strategies reported in Appendix Table S7.
7. Superposition of the SCR‐SCR2 heterotetramer colored the same way with the SCR homotetramer colored as in Fig 3.

Source data are available online for this figure.

The same differentially tagged constructs were also used to conduct in vitro homotetramer mixing experiments (Appendix Fig S8B), which demonstrate that subunit exchange occurs spontaneously but inefficiently during 8 h at 25°C. These conditions were chosen to match the time period and temperature used for in vivo protein expression. Therefore, the 7:1 ratio of homotetramers versus heterotetramers produced in vivo suggests preferential oligomerization of subunits translated from the same mRNA molecule or a local pool of mRNAs transcribed from the same plasmid encoding one of the two paralogs. However, there is still a relatively efficient assembly of heterotetramers from paralogous subunits translated from different mRNAs in vivo compared to the amount formed when mixing homotetramers in vitro (Fig 6A versus Appendix Fig S8B). The slow interconversion of SCR homodimers and homotetramers, as demonstrated by gel filtration chromatography (top of Fig EV4A), suggests that there is a kinetic barrier to tetramer formation that may have evolved to facilitate the formation of heterotetramers in vivo.

Michaelis‐Menten analyses were conducted on the homotetramers and heterotetramers expressed in vivo with short N‐terminal tags after proteolytic removal of the tags (Fig 6B). These assays demonstrate that the SCR homotetramer has high activity reducing 2‐hydroxyacetophenone (2‐HAP) but not methyl acetoacetate, while the SCR2 homotetramer shows the opposite pattern. In contrast, the SCR2‐SCR heterotetramer has high activity on both of these substrates, indicating that it combines the substrate specificities and catalytic properties of the two paralogs.

We determined the X‐ray crystal structure of the tag‐cleaved SCR2‐SCR heterotetramer at 1.74 Å and refined it to an R _free of 20.7% with excellent geometrical characteristics (Fig 6C–G and Appendix Tables S2–S4). Careful analysis of the electron density maps, including F_o‐F_c difference density maps (Fig 6D and Appendix Fig S9), indicates that the dimeric assembly in the asymmetric unit in the crystal structure contains a mixture of SCR and SCR2 subunits at both positions. Conducting refinements using different homodimer and heterodimer models with different occupancy constraints supports an equal content of SCR and SCR2 subunits at both positions in the asymmetric unit (Appendix Table S7 and Appendix Fig S9). Comparisons of the SCR versus SCR2 subunits between their respective homotetramer crystal structures show 0.33–1.31 Å RMSD for least‐squares alignment of all Cα atoms, while equivalent alignments of those same homotetramer subunits to the subunits in the heterotetramer crystal structure show 0.47–1.30 Å RMSD (Appendix Table S5A). This close stereochemical similarity is consistent with the two paralogous subunits packing interchangeably to create a statically disordered crystal lattice with equal occupancy of both paralogs at each of the two positions in the asymmetric unit of the lattice. While our crystallographic analyses do not establish the exact distribution of the paralogous subunits in the heterotetramers in solution, they are consistent with heterotetramers containing a combination of SCR and SCR2 subunits stochastically docking into the crystal lattice in random orientations.

Comparing the structures of SCR and SCR2 shows that 31 of the 32 residues that differ between the two paralogs are clustered around the active site (Fig 6C, E and F, Appendix Fig S10). This strikingly nonrandom spatial pattern supports positive evolutionary selection for these substitutions, and, taken together with the maintenance of both paralogs in the genome of C. parapsilosis, it provides strong evidence that most of these substitutions modulate enzyme specificity. Notably, only four of the substitutions are in residues directly lining the active site, while 27 are in proximal protein segments. This pattern of substitution suggests that residues in structural layers surrounding the active site but not directly lining it significantly modulate enzyme activity and/or specificity (Nie et al, 2011).

Evolutionary modification and diversification of enzyme activity are therefore likely to be facilitated by high tolerance for mutations in this region of the protein, which is proximal to and, based on the data presented above, stabilized by the tetramer interface in SDR superfamily enzymes (Figs 1C, 6F and G, and Appendix Fig S2C). The extensive packing interactions in the tetramer interface can therefore energetically buffer the active‐site structure so that a wider range of destabilizing mutations that enhance activity can be introduced without disrupting the active‐site structure (Brien & Herschlag, 1999). Stabilization of active‐site structure by a large interaction surface should enhance evolutionary plasticity by providing a reservoir of stabilization energy that increases tolerance for diverse activity‐enhancing mutations. Intersubunit interactions can provide a large interaction surface providing high stabilization potential without the metabolic investment required to add a sizable domain specifically to stabilize a functionally important protein segment. Stabilization of active‐site structure by intersubunit interactions in a protein oligomer can thereby offer a significantly larger metabolic and evolutionary advantage than estimated based on the minimum subdomain size capable of stabilizing the structure of one SDR superfamily member that is described above.

Discussion

It has frequently been proposed that intersubunit interactions in oligomeric assemblies stabilize protein structure (Chen & Stites, 2001), but little experimental data have been presented to support this inference or to establish the specific mechanisms involved or their impact on protein function. In this paper, we provide direct experimental evidence that the evolutionarily conserved tetramer interface in the noncooperative short‐chain dehydrogenase/reductase (SDR) enzyme superfamily stabilizes the active‐site structure (Figs 3, 4, 5). One important component of this evidence is the observation in cryo‐EM reconstructions of native SCR, a paradigmatic representative of the SDR superfamily, of the spontaneous opening of its tetramer interface coupled to the disordering of the two Tetramer Interface Helices (TIHs) (Fig 5A–C) that form a substantial portion of the enzyme active site while making the principal packing interactions stabilizing the tetramer interface (Fig 1C and D, and 2D). Differential interactions with the air‐water interfaces could potentially perturb the relative populations of the different conformations of the SCR tetramer observed in the cryo‐EM sample compared with bulk solution, but the similar orientational distributions observed for the conformation with the open tetramer interface and the canonical symmetrical conformation of the tetramer (Appendix Fig S7C) suggest such perturbations are likely to be limited in magnitude. Furthermore, analysis of the spatial pattern of amino acid substitutions in the paralogous SCR and SCR2 enzymes from this superfamily (Fig 6E) suggests that the stability imparted by the tetramer interface provides an energetic buffer that enhances evolutionary plasticity by enabling a wider range of specificity‐modulating mutations to be tolerated without disrupting the active‐site structure. Finally, we demonstrate that extending the C‐terminus of SCR by 8 residues at its tetramer interface results in an equilibrium between biochemically well‐behaved dimer and tetramer species (Fig EV4A) that together have a different activity profile from the native tetramer (Fig 4).

These results provide a counter‐example to a recent study of steroid hormone receptors that concluded that oligomeric interactions can be functionally gratuitous but evolutionarily maintained because of the difficulty of generating biochemically well‐behaved proteins when an oligomeric interface is mutated (Hochberg et al, 2020). Disrupting the native tetramer interface in SCR clearly perturbs both the local structure in the active site (Figs 3 and 5A–C) and enzyme function (Fig 4) but without compromising qualitative biochemical behavior. The formation of a stable subassembly with an altered activity profile upon disruption of an oligomeric assembly provides an additional source of plasticity that can be exploited for the evolution of enzyme function.

The conclusion that the conserved tetramer interface in the SDR superfamily stabilizes the active‐site structure is reinforced by a survey of homologous enzymes of known structure (Fig EV3), which demonstrates that a 34‐residue extension in the primary sequence, representing an ~12% increase in protein length, is needed to stabilize the active‐site structure in the absence of the conserved oligomeric interactions in the tetramer interface. This observation provides an estimate of the reduced metabolic cost of stabilizing the active‐site structure via oligomeric interactions versus an expansion in subunit structure. We demonstrate that this metabolic benefit can be achieved while simultaneously diversifying enzyme function by showing that hetero‐oligomerization enables the different catalytic properties of paralogous enzymes in the superfamily to be combined in a single oligomeric assembly (Fig 6). Another example of this phenomenon is provided by the hetero‐oligomeric retinoid oxidoreductase complex, which contains two remote paralogs in the SDR superfamily with distinct activities (Adams et al, 2021). Therefore, oligomerization can expand chemical capabilities in parallel with enhancing metabolic efficiency during protein evolution.

As alluded to above, our data furthermore suggest that stabilizing active‐site structure via oligomeric interactions expands functional potential in an indirect way by offsetting the destabilizing effect of activity‐enhancing mutations in protein segments surrounding the residues that directly line the active site (Fig 6E). Comparing the structures of the paralogous SCR and SCR2 enzymes from C. parapsilosis, which share 85% sequence identity (Fig 6C) but have significantly different substrate specificity (Fig 6B), shows strong evidence for positive functional selection in “second‐shell” residues that simultaneously make packing interactions across the conserved tetramer interface and with residues that line the active site (Fig 6E). This strikingly localized pattern of evolutionary divergence indicates that relatively small changes in active‐site stereochemistry or regional changes in protein dynamics can significantly modulate enzyme specificity, at least for the NAD(P)/H‐dependent oxidoreductases in the SDR superfamily. In this context, it is notable that, to date, extensive efforts at de novo enzyme design focused on optimizing the static conformation of residues directly lining the active site have not produced an enzyme with catalytic efficiency approaching that of a naturally evolved enzyme (Huang et al, 2016; Chen et al, 2019; Basanta et al, 2020; Yang et al, 2021). The pattern of sequence divergence in the SCR versus SCR2 enzymes suggests that second‐shell structural interactions can play an important role in controlling enzyme activity and specificity (Fig 6E), which echoes conclusions from a recent study showing that mutation of some residues located even further from an active site can modulate activity (Wang et al, 2020).

Using intersubunit packing interactions to stabilize the protein segments making functionally important second‐shell interactions is likely to enable locally destabilizing amino acid substitutions that enhance catalytic activity to be made without destabilizing the tertiary structure forming the active site. The extensive surface area (Appendix Table S6) buried in intersubunit packing interactions proximal to the active site in SDR superfamily enzymes (Figs 1C and D, 2D, and 6F) provides a reservoir of free energy stabilizing the active‐site structure that will buffer it to mutations. The conserved oligomeric interfaces thereby facilitate the introduction of individually destabilizing mutations that increase enzyme activity or alter specificity (Junmei Brien & Herschlag, 1999; Chen & Stites, 2001; Sneddon & Tobias, 2002). Therefore, conservation of oligomeric interactions in some enzyme superfamilies may reflect their ability to enhance the efficiency of the evolution of new and altered enzyme activities (D'Alessio, 1999; Hashimoto & Panchenko, 2010; Hashimoto et al, 2011) by strongly stabilizing active‐site structure at low metabolic cost.

The observation of a small population of dimers in wild‐type SCR preparations (Fig EV4A) suggests oligomeric interactions may enhance the evolutionary plasticity and potential of enzymes in a second essentially independent way. Given that perturbation of the tetramer interface in SCR alters specificity (Fig 4B), the dimer is likely to have altered enzymatic properties that can influence fitness and contribute to enzyme evolution. The observation of spontaneous opening of the tetramer interface in SCR in solution coupled to a substantial change in the structure of two of its four active sites (Fig 5A–C) suggests dynamic fluctuations in oligomer geometry could also contribute to the diversification of enzyme specificity. The lattice‐packing constraints imposed in crystal structures may have suppressed observation and understanding of the mechanistic and evolutionary significance of these dynamic states. Reversible equilibration between different oligomeric states with different specificity properties will enable evolutionary selection to operate on all states in parallel, enabling simultaneous exploration of multiple pathways for the evolution of enzyme activity. The frequent conservation of oligomeric interactions in enzyme superfamilies thus may reflect not only their ability to increase metabolic efficiency in producing active protein structures but also their ability to enhance the efficiency of the evolution of new and altered enzyme activities via multiple distinct physicochemical mechanisms.

Elegant mutational studies have shown it is easy to identify surface mutations that produce some degree of regular oligomeric interaction between protein molecules (Li et al, 2014; Jubb et al, 2017; Empereur‐Mot et al, 2019). The data presented in this paper establish specific biophysical mechanisms that enable the formation of such structures to enhance evolutionary fitness in several distinct and complementary ways. They can reduce the metabolic cost of enzyme synthesis (Fig EV3) while simultaneously diversifying enzyme specificity via hetero‐oligomerization of paralogs (Fig 6B) (Adams et al, 2021). Alterations in specificity (Fig 4B) can also occur upon dynamic fluctuations in intersubunit interactions (Fig 5A–C) or mutational perturbation of those interactions (Fig 3), providing parallel physicochemical mechanisms by which oligomeric interactions can enhance the evolutionary plasticity and potential of enzymes. Promotion of evolutionary adaptability via multiple physicochemical mechanisms is likely to contribute to the frequent conservation of oligomeric interactions in enzyme superfamilies. Therefore, in conjunction with recent studies illuminating the pathways by which oligomeric interactions can evolve (Levy et al, 2008; Faure et al, 2012; Perica et al, 2012, 2014), our work deepens understanding of basic processes involved in the evolution of protein structure and enzyme activity.

Materials and Methods

Purification of heterotetramers and analysis of heterotetramer formation kinetics

The strep‐SCR/his₆‐SCR2 heterotetramers were purified after co‐expression in vivo in E. coli by sequential Ni‐NTA and Strep‐Tactin affinity chromatography (Cai & Inouye, 2002; DellaVecchia et al, 2007) followed by gel filtration on a Superdex 200 Increase 10/300 column (Cytivia). The kinetics of heterotetramer formation were characterized by mixing separately purified his₈‐MBP‐SCR2 and strep‐SCR at a concentration of 2 μM each at 25°C for 0, 2, 4, or 8 h prior to analysis on a Superdex 200 increase 10/300 column.

Phylogenomic analyses

The SCR sequence was used to perform a search in the PDB database for sequence identity above 30%. A maximum‐likelihood phylogenetic tree was generated by MEGA X (Kumar et al, 2018) which is shown as a bootstrap consensus tree. Four of the most remote PDB structures were selected to compare their crystal structures with SCR and SCR2, and their sequence alignments using ESPript 3.0 (Gouet, 2003).

Molecular graphics images

Figures were prepared using PyMOL (Tao et al, 2020).

Preparation of homo‐oligomers of SCR proteins and determination of their molecular weights

Recombinant SCR and SCR2 proteins from C. parapsilosis CCTCC M203011 were overexpressed in Escherichia coli using plasmids ligated with fragments generated by PCR amplifications of the scr gene (DQ675534) with a strep tag at its N‐terminus and the scr2 gene (FJ939563) with a hexahistidine tag at its N‐terminus. A cleavage site for TEV protease was added between the strep tag or his tag and the protein‐coding sequence (Nie et al, 2011). The primers used for the amplification of scr and scr2 genes are given in Appendix Table S1. The PCR fragments were inserted into pET21c and pET28a vectors to generate pET21c‐strep‐TEV‐SCR and pET28a‐his₆‐TEV‐SCR2. Recombinant SCR and SCR2 were heterologously overexpressed in E. coli BL21 (DE3) in LB broth. Following growth at 37˚C, SCR and SCR2 were induced with 0.1 mM IPTG at OD 0.6 − 0.8. The cells were then incubated at 28°C for 10 h prior to harvesting by centrifugation at 8,000 g for 10 min. The cell pellets were suspended in 0.9% NaCl and washed three times. Later, the cell pellets were suspended in buffer A (40 mM Tris, 150 mM NaCl, pH 8.0) for strep‐tagged SCR (strep‐TEV‐SCR) or in buffer B (20 mM Tris, 150 mM NaCl, pH 8.2) for his‐tagged SCR (his₆‐TEV‐SCR2) and then disrupted by sonication. The N‐terminally strep‐tagged SCR (strep‐TEV‐SCR) was purified as follows: (i) The supernatant was applied onto the Strep‐Tactin column equilibrated with buffer A and eluted with buffer A containing 2.5 mM d‐Desthiobiotin. (ii) The eluate from Strep‐Tactin column was desalted on a PD‐10 column with buffer B. (iii) The eluate from the PD‐10 column was then mixed with TEV protease at a molar ratio of 1:10 TEV:SCR, and incubated at 4°C for 16 h. (iv) The mixture was then passed through a Ni‐NTA column. (v) That flow‐through was further purified using gel filtration chromatography on a Superdex increase 200 10/300 GE column preequilibrated with buffer C (50 mM Tris, 150 mM NaCl, pH 8.0). The N‐terminally his‐tagged SCR2 (his₆‐TEV‐SCR2) was purified with Ni‐NTA affinity chromatography using a gradient from buffer A (20 mM Tris, 150 mM NaCl, pH 8.2) to buffer B (20 mM Tris, 150 mM NaCl, 1 M imidazole, pH 8.2). After affinity chromatography, the purification procedure followed the same steps used for strep‐TEV‐SCR. The purified protein was concentrated to 5–8 mg/ml as determined using a Nanodrop spectrophotometer (Thermo Fisher Scientific) to measure the absorbance at 280 nm, and it was stored at −80°C after adding 2 mM DTT.

Purified untagged SCR and SCR2 were analyzed by gel filtration chromatography on a Superdex 200 increase column (Appendix Fig S1). Protein markers used for gel filtration were from Oriental Yeast Co., LTD: (i) glutamate dehydrogenase (yeast) (290 kDa); (ii) lactate dehydrogenase (pig heart) (142 kDa); (iii) enolase (yeast) (67 kDa); (iv) myokinase (yeast) (32 kDa); and (v) cytochrome c (horse heart) (12.4 kDa) (Xia et al, 2016). The theoretical molecular weights were calculated by expasy (http://web.expasy.org/protparam/).

Preparation of heterotetramers of SCR‐SCR2 proteins

The hetero‐oligomers of his₆‐SCR2/strep‐SCR were prepared by co‐expression of plasmids of pET28‐his‐SCR2 and pET21‐strep‐SCR in E. coli BL21 (DE3) induced at 25°C for 10 h. The cellular extracts from the co‐expressed system were purified as described by Levy et al (2016) with some modifications. In brief, the sample was first purified by Ni‐NTA chromatography. The eluate with 200 mM imidazole was collected, desalted on a PD‐10 column in buffer A (40 mM Tris, 150 mM NaCl, pH 8.0), and then loaded on a Strep‐Tactin column equilibrated in Buffer A. The Strep‐Tactin column was washed with buffer A until the UV280 leveled off, and the column was then eluted with buffer A containing 2.5 mM D‐desthiobiotin. The peak was collected and further purified by gel filtration using a Superdex 200 Increase 10/300 column equilibrated in 150 mM NaCl, 50 mM Tris, pH 8.0. The peak from gel filtration chromatography was collected for further analysis and crystallization.

Enzyme activity assays

The kinetics of 2‐hydroxyacetophenone and methyl acetoacetate reduction (k _cat and K _M) were determined at 500 μM NADPH concentration at various concentrations of 2‐hydroxyacetophenone (0.5–30 mM) and methyl acetoacetate (2.0–120 mM). Assays were conducted at 30°C in a 96‐well plate with NADPH consumption monitored for 10 min at 340 nm using a UV absorbance plate reader; the plate was shaken in situ in the plate reader for 6 s prior to the initiation of the absorbance measurements. The buffer used for the enzyme assays was 100 mM sodium phosphate. All measurements were made in triplicate.

Multiangle light scattering methods

The molecular weight of SCR‐SCR2 hetero‐oligomers was analyzed by size‐exclusion chromatography—multiangle light scattering (Wyatt Technology) as described by Kendrick et al (2001). The weight average molar mass (M_w) was determined using a dn/dc value of 0.185 (Roberts et al, 2014) in the Astra software (Trathnigg, 1995).

Mass spectrometric methods

The molecular weights of subunits in hetero‐oligomers of SCR‐SCR2 then were analyzed by TOF‐LC/MS on an Agilent Technologies 6224 mass spectrometer as described by Sala et al (Tzeli et al, 2015) with 10 μl (1 mg/ml) per injection. The mobile phase contained 90% A (0.1% HCOOH + 99.9% H₂O) and 10% B (0.1% HCOOH + 99.9% acetonitrile), and the flow rate was 0.5 ml/min.

Crystallization

Initial crystallization trials were conducted with commercially available kits (Hampton Research and Molecular Dimensions) and using the high‐throughput screening service at the Hauptman‐Woodward Medical Research Institute (hwi.buffalo.edu/crystallization‐cocktails). Microbatch crystallization reactions under oil contained equal volumes (0.8 μl) of protein sample (5–8 mg/ml in 50 mM Tris, 150 mM NaCl, 2 mM DTT, pH 8.0) and reservoir solution (0.8 μl) mixed using a multichannel pipette in a 96 well plate. Initial crystals of apo SCR, binary complexes of SCR/NADPH, and binary complexes of SCR2/NADP and SCR2/NADPH grew after 1–4 weeks in different precipitant conditions at 20°C. Optimization of the crystals yeilded the following final conditions for microbatch crystallization using equal 1 μl volumes of protein sample and reservoir solution mixed manually with a single pipette in a 72 well plate at 20°C:

1. apo SCR: 200 mM Ammonium Chloride, 20% (w/v) PEG3350, pH 6.3;

2. binary complexes of SCR/NADPH: 200 mM Ammonium Chloride, 20% PEG3350, pH 7.0;

3. binary complexes of SCR2/NADPH: 200 mM Ammonium acetate, 20% PEG3350, pH 7.1;

4. binary complexes of SCR2/NADP: 200 mM Ammonium acetate, 20% PEG3350, pH 7.1.

The dimensions of the crystals from the optimized conditions were about 150 × 100 × 20 μm for apo SCR, 50 × 50 × 20 μm for binary complexes of SCR/NADPH, and 100 × 20 × 20 μm for binary complexes of SCR2/NADPH and SCR2/NADP. Single crystals were transferred to cryoprotectant solution (Paratone) and flash‐cooled in liquid nitrogen. Crystallization information is summarized in Appendix Table S2.

Data collection and processing

Diffraction data were recorded on EIGER, ADSC HF, or Q315CCD detector at 100 K (in a nitrogen stream) on beamlines as shown in Table S3. Data were processed with XDS and were merged and scaled with AIMLESS (Evans, 2005; Kabsch, 2010). Data collection and processing statistics are summarized in Appendix Table S3.

Structure solution and refinement

Apo SCR was solved by molecular replacement using PHASER within PHENIX (Kavscek et al, 2015) with a search model comprising one subunit from C‐terminally his‐tagged SCR (PDB ID 3CTM) (Zhang et al, 2008). SCR/NADPH, SCR2/NADPH, and SCR2/NADP were solved by molecular replacement using PHASER within PHENIX with a search model comprising chain A of apo SCR. NADPH and NADP were fit into the corresponding electron densities using the LigandFit module in PHENIX (Roy et al, 2010). The structure of the SCR‐SCR2/NADPH heterotetramer was also solved by molecular replacement using PHASER/PHENIX with a search model comprising a homodimer of SCR/NADPH. To conduct statically disordered refinements, a model with superimposed subunits was created by editing the coordinate file to have two residues at the same position for every amino acid that differs between SCR and SCR2 (Fig 6C), with the residues in each model described in Appendix Table S7 all being assigned the same insertion code to put them in the same occupancy refinement group. The occupancies for these two groups were constrained by PHENIX to sum to 1.0. Alternating rounds of computational refinement and automated water addition in PHENIX and manual building in COOT continued until R _free, R _work, and the number of water molecules converged. Model quality was analyzed by MOLPROBITY (Laskowski & Swindells, 2011). Refinement statistics are summarized in Appendix Table S4.

Cryogenic electron microscopy data collection and reconstruction (cryo‐EM) methods

Before preparing cryo‐EM grids, the purified protein was filtered using a 0.22 um Spin‐X centrifugal filter (Corning), and the concentration of the flow‐through was measured with a NanoDrop spectrophotometer (Thermos Fisher Scientific). The protein sample was diluted to 3 mg/ml in 5 mM ethyl 4‐chloroacetoacetate, 0.5 mM NADP, 150 mM NaCl, 50 mM Tris–Cl, pH 8.0, and 3.5 µl was deposited on an UltraAuFoil 300 mesh R 0.6/1 grid (Quantifoil Micro Tools GmbH, Großlöbichau, DE) using a Vitrobot Mark IV robot (Thermofisher Inc., USA) with its chamber set to 4°C and 100% humidity. Before sample deposition, the grid was treated in a Solarus Plasma Cleaner 950 (Gatan Inc., USA) using for 25 s with O2/H₂ flow rates of 27.5/6.4 sccm and 15 W cleaning power. Following protein deposition on the grid, it was blotted for 7 s with Whatman grade 1 filter paper (Whatman Inc., USA) at force 3 and then plunged into cooled liquid ethane (~90 K). Cryo‐EM micrographs were collected on a 300 kV Titan Krios electron microscope (ThermoFisher Inc., USA) at the Cryo‐Electron Microscopy Center of Columbia University using Leginon (Carragher et al, 2000). The images were recorded using a K3 camera and imaging filter (Gatan Inc., USA) in counting mode at a nominal magnification of 105,000× (nominal pixel size of 0.844 Å/pixel, which was refined to 0.837 Å/pixel based on calibration of the final ST map against the crystal structure of the SCR/NADPH complex) with a defocus range from −1.34 to −2.17 µm. Exposures of 2.5 s were dose‐fractionated into 50 frames (50 ms/frame), with an exposure rate of 16 electrons/pixel/sec, resulting in a total exposure of 56.15 electrons/Ų. Image‐processing and 3‐dimensional reconstruction were performed using cryoSPARC 3.2 (Punjani et al, 2017); the processing flowchart is shown in Fig 5A in the main text. Model refinement was conducted using PHENIX (Liebschner et al, 2019) with initial coordinates from the X‐ray crystal structure of SCR/NADPH(PDB ID 7DLM).

Solvent‐accessible surface area calculations

PISA (Krissinel & Henrick, 2005) was used to calculate buried surface areas.

Author contributions

Yaohui Li: Data curation; Software; Formal analysis; Validation; Investigation; Visualization; Methodology; Writing—original draft; Project administration; Writing—review & editing; cryoEM data processing. Rongzhen Zhang: Funding acquisition; Project administration. Chi Wang: Data curation; Software; cryoEM data processing. Farhad Forouhar: Data curation; Software. Oliver B Clarke: cryo‐EM data processing. Sergey Vorobiev: cryo‐EM data processing. Shikha Singh: Data curation; Software. Gaetano T Montelione: Methodology. Thomas Szyperski: Methodology. Yan Xu: Funding acquisition; Project administration. John F Hunt: Formal analysis; Supervision; Validation; Investigation; Methodology; Writing—original draft; Project administration; Writing—review & editing; cryoEM data processing.

In addition to the CRediT author contributions listed above, the contributions in detail are:

YX, RZ, JFH, and YL conceived the project and designed the structural experiments, which were performed and analyzed by YL, FF, CW, SV, SS, and OBC. The oligomerization and kinetic experiments were designed by JFH, GTM, TS, and YL, who performed these experiments. YL and JFH drafted the manuscript and finalized it with help from all authors.

Disclosure and competing interests statement

GTM is a founder and JFH is a member of the Scientific Advisory Board of Nexomics Biosciences, Inc. JFH consults on cryo‐EM projects for Cyrus Biotechnology. These affiliations do not constitute a conflict of interest with respect to the work described in this study. The other authors declare no competing financial interests.

Supporting information

Appendix

Expanded View Figures PDF

Movie EV1

Source Data for Appendix

Source Data for Figure 4

Source Data for Figure 6

The EMBO Journal (2022) 41: e108368

Data availability

Crystallographic data and Cryo‐EM data have been deposited to the PDB, EMDB, and EMPIAR databases, with the following identifiers:

• apo SCR crystal structure 7DLD (10.2210/pdb7DLD/pdb) https://www.rcsb.org/structure/7DLD;

• SCR/NADPH crystal structure 7DLM (10.2210/pdb7DLM/pdb) https://www.rcsb.org/structure/7DLM;

• SCR2/NADPH crystal structure 7DLL (10.2210/pdb7DLL/pdb) https://www.rcsb.org/structure/7DLL;

• SCR2/NADP crystal structure 7DMG (10.2210/pdb7DMG/pdb) https://www.rcsb.org/structure/7DMG;

• SCR‐SCR2/NADPH crystal structure 7DN1 (10.2210/pdb7DN1/pdb) https://www.rcsb.org/structure/7DN1;

• SCR/NADP/ethyl 4‐chloroacetoacetate ST cryo‐EM coordinate model 7VYQ (10.2210/pdb7VYQ/pdb) https://www.rcsb.org/structure/7VYQ;

• SCR ST cryo‐EM map EMD‐32211 (https://www.ebi.ac.uk/emdb/EMD‐32211);

• SCR AT cryo‐EM map EMD‐32212 (https://www.ebi.ac.uk/emdb/EMD‐32212);

• SCR OTI cryo‐EM map EMD‐32213 (http://www.ebi.ac.uk/pdbe/entry/EMD‐32213);

• SCR cryo‐EM dataset EMPIAR‐10872 (https://www.ebi.ac.uk/empiar/EMPIAR‐10872/).