Biophysical investigation of type A PutAs reveals a conserved core oligomeric structure

Abstract

Many enzymes form homooligomers, yet the functional significance of self-association is seldom obvious. Herein, we examine the connection between oligomerization and catalytic function for proline utilization A (PutA) enzymes. PutAs are bifunctional enzymes that catalyze both reactions of proline catabolism. Type A PutAs are the smallest members of the family, possessing a minimal domain architecture consisting of N-terminal proline dehydrogenase and C-terminal L-glutamate-γ-semialdehyde dehydrogenase modules. Type A PutAs form domain-swapped dimers, and in one case (Bradyrhizobium japonicum PutA), two of the dimers assemble into a ring-shaped tetramer. Whereas the dimer has a clear role in substrate channeling, the functional significance of the tetramer is unknown. To address this question, we performed structural studies of four type A PutAs from two clades of the PutA tree. The crystal structure of Bdellovibrio bacteriovorus PutA covalently inactivated by N-propargylglycine revealed a fold and substrate-channeling tunnel similar to other PutAs. Small-angle X-ray scattering (SAXS) and analytical ultracentrifugation indicated Bdellovibrio PutA is dimeric in solution, in contrast to the prediction from crystal packing of a stable tetrameric assembly. SAXS studies of two other type A PutAs from separate clades also suggested the dimer predominates in solution. To assess whether the tetramer of B. japonicum PutA is necessary for catalytic function, a hot spot disruption mutant that cleanly produces dimeric protein was generated. The dimeric variant exhibited kinetic parameters similar to the wild-type enzyme. These results implicate the domain-swapped dimer as the core structural and functional unit of type A PutAs.

Graphical abstract

Many enzymes form homooligomers, yet the functional significance of self-association is not always obvious. We examine the connection between oligomerization and catalytic function for type A PutA bifunctional enzymes using an integrative approach involving crystallography, small-angle X-ray scattering, analytical ultracentrifugation, kinetics, and hot spot disruption mutagenesis. Our results implicate a domain-swapped dimer as the core functional unit of these enzymes.

Introduction

The proline catabolic pathway in bacteria and eukaryotes consists of the enzymes proline dehydrogenase (PRODH) and L-glutamate-γ-semialdehyde aldehyde dehydrogenase (GSALDH) (Fig. 1) [1]. PRODH catalyzes the FAD-dependent oxidation of L-proline to Δ¹-pyrroline-5-carboxylate (P5C). Hydrolysis of P5C generates the substrate for GSALDH, L-glutamate-γ-semialdehyde. GSALDH is an aldehyde dehydrogenase (ALDH) superfamily enzyme that catalyzes the NAD⁺-dependent oxidation of L-glutamate-γ-semialdehyde to L-glutamate. In total, the pathway produces a 4-electron oxidation. The electrons abstracted from proline flow into the electron transport chain, while the carbon skeleton of L-proline ultimately enters the citric acid cycle as α-ketoglutarate. Proline catabolic enzymes have been implicated in many aspects of health and disease, including tumor suppression [2], hyperprolinemia metabolic disorders [3], schizophrenia susceptibility [4–8], life-span extension [9], production of fungal virulence factors [10], and virulence and survival of pathogenic bacteria [11–14].

Fig. 1.

The reactions catalyzed by PutA. PutAs consist of two modules that catalyze the oxidation of L-proline to L-glutamate. The first module, proline dehydrogenase (PRODH), utilizes FAD to catalyze the conversion of L-proline to Δ¹-pyrroline-5-carboxylate, which is subject to non-enzymatic hydrolysis to L-glutamate-γ-semialdehyde. The second module, L-glutamate-γ-semialdehyde dehydrogenase (GSALDH), performs the NAD⁺-dependent oxidation of L-glutamate-γ-semialdehyde to the final product, L-glutamate.

PRODH and GSALSH (aka P5CDH and ALDH4A1) are combined in some bacteria into a single protein known as proline utilization A (PutA) [1, 15]. The name PutA refers to early studies of proline utilization in bacteria, which led to the discovery that a single gene encodes both PRODH and GSALDH [16, 17]. The covalent linking of two enzymes into one polypeptide chain suggests the possibility of substrate channeling, which can improve kinetic efficiency, protect reactive intermediates, and prevent crosstalk between competing pathways, such as proline catabolism and biosynthesis [18]. Indeed, kinetic evidence in support of substrate channeling has been found for several PutAs [19–23].

PutAs can be classified according to global sequence similarity and domain architectures (Fig. 2). The PutA phylogenetic tree has three prominent branches (1, 2, and 3), and there are three types of PutA domain architectures (A, B, and C). Type A PutAs have N-terminal PRODH and C-terminal GSALDH modules (Fig. 2A). Type B PutAs have an additional C-terminal domain, which was recently shown to have the ALDH superfamily fold [24, 25]. Type C PutAs have an N-terminal ribbon-helix-helix DNA-binding domain in addition to the C-terminal domain. These PutAs function as transcriptional repressors of the genes encoding PutA and the proline transporter PutP. Types A, B, and C PutAs are found in branch 1 (Fig. 2B). Only type A PutAs are found in branch 2, and only type B PutAs are observed in branch 3. Thus, combining the phylogenetic tree and domain analysis yields five classes of PutA: 1A, 1B, 1C, 2A, and 3B.

Fig. 2.

Classification of PutAs according to domain architecture and global sequence identity. (A) The three domain architectures of PutAs. Abbreviations: RHH, ribbon-helix-helix; ALDHSF, aldehyde dehydrogenase superfamily. The small N-terminal domain of type C PutAs is a ribbon-helix-helix DNA-binding domain. (B) Phylogenetic tree based on a global sequence alignment of PutAs. Architecture types A, B, and C are indicated by black, blue, and red font, respectively. The PutAs mentioned in the text are noted in large font. The alignment was calculated with Clustal Omega [74] and visualized with DrawTree [75]. Abbreviation not listed in the text: EcPutA, Escherichia coli proline utilization A.

The PutAs studied to date show surprising diversity in oligomeric state and quaternary structure. Crystal structures and oligomeric states in solution have been determined for the class 1A PutA from Bradyrhizobium japonicum (BjPutA) [20], the class 2A PutA from Geobacter sulfurreducens (GsPutA) [23], the class 1B PutA from Sinorhizobium meliloti (SmPutA) [24], and the class 3B PutA from Corynebacterium freiburgense (CfPutA) [25]. These structures showed that the PRODH and GSALDH active sites are spatially separated and connected by a tunnel. Mutagenesis studies showed the tunnel is used for substrate channeling [26]. Although the arrangement of domains within the protomer is similar in all three structures, the oligomeric states and quaternary structures differ. GsPutA forms a domain-swapped dimer in solution. In BjPutA - another type A PutA - two of the dimers assemble into a ring-shaped tetramer. The type B PutAs SmPutA and CfPutA form monomer-dimer equilibria in solution; however, the quaternary structure of the dimer is completely different from the type A domain-swapped dimer [24, 25]. Small-angle X-ray scattering (SAXS) showed that the class 1C PutA from Escherichia coli forms yet a third type of dimer, which is mediated by the DNA-binding domain [27]. The functional relevance of these various oligomeric states and quaternary structures has not been systematically studied in detail.

The observation of different oligomeric states in two type A PutAs from different branches of the phylogenetic tree motivated us to study the self-association of other type A PutAs to understand the relationship between oligomeric state and catalytic function. In particular, we sought to determine whether oligomeric state distinguishes class 1A from class 2A and whether tetramerization of BjPutA is required for catalytic activity.

Herein, we report a crystal structure of the class 2A PutA from Bdellovibrio bacteriovorus (BbPutA), along with SAXS data for BbPutA and three other type A PutAs (from classes 1A and 2A). We also employed analytical ultracentrifugation, size-exclusion chromatography, and enzyme kinetics for an extensive biophysical characterization of these enzymes. The data suggest type A PutAs are predominantly dimeric in solution, and global sequence identity is not a reliable indicator of the oligomeric state. Further, we investigate the necessity of higher order oligomeric states by generating a dimeric hot spot mutant of the tetrameric BjPutA. Overall, these results reveal a conserved dimer as the essential oligomer required for catalysis and channeling in type A PutAs. Our results also demonstrate the challenges of predicting the solution oligomeric state from crystal packing.

Results

Crystal structure of covalently inactivated BbPutA

The crystal structure of BbPutA inactivated by the mechanism-based inactivator N-propargylglycine (NPPG) was determined at 2.2 Å resolution (Table 1). Similar to other PutAs, BbPutA has spatially-separated PRODH and GSALDH active sites (Fig. 3A). The PRODH active site resides at the C-termini of the strands of a (βα)₈ barrel. The GSALDH active site is located in the crevice between the Rossmann dinucleotide binding domain and the GSALDH catalytic domain. In addition to the catalytic domains, the fold includes an N-terminal arm domain and an α-domain in the PRODH half of the chain, as well as an oligomerization flap at the C-terminus. These ancillary domains are also observed in other type A PutA structures. The root mean square deviation between BbPutA and GsPutA is 1.1 Å, as expected for two proteins with high sequence identity (47%, Table 2). The deviations of BbPutA from BjPutA (class 1A) and type B PutAs are also low (1.8–2.0 Å), despite sharing only ~30% sequence identity with these proteins. This result attests to the high structural conservation of the PutA fold.

Table 1.

Data Collection and Refinement Statistics^a

Space group	P21221
Beamline	APS 24-ID-E
Unit cell parameters (Å)	a = 144.3, b = 158.6, c = 221.1
Wavelength (Å)	0.97920
Resolution (Å)	128.9– 2.23 (2.35 – 2.23)
Observations	930561
Unique reflections	241211
Rmerge(I)	0.095 (0.493)
Rmeas(I)	0.124 (0.644)
Rpim(I)	0.061 (0.320)
Mean I/σ	10.0 (2.4)
Completeness (%)	97.9 (98.3)
Multiplicity	3.9 (3.9)
No. of protein residues	3829
No. of atoms
Protein	29525
FAD	212
Modified Lys	48
Water	1048
Rcryst	0.183 (0.230)
Rfreeb	0.228 (0.291)
rmsd bond lengths (Å)	0.008
rmsd bond angles (°)	0.872
Ramachandran plotc
Favored (%)	97.95
Outliers (%)	0.00
Clashscore (PR)c	1.73 (100)
MolProbity score (PR)c	0.94 (100)
Average B (Å2)
Protein	31.2
FAD	31.4
Modified Lys	43.0
Water	30.2
Coordinate error (Å)d	0.27
PDB code	5UR2

^aValues for the outer resolution shell of data are given in parenthesis.

^b5% test set.

^cFrom MolProbity. The percentile ranks (PR) for Clashscore and MolProbity score are given in parentheses.

^dMaximum likelihood-based coordinate error estimate from PHENIX.

Fig. 3.

Structure of BbPutA. (A) A protomer of BbPutA with the domains colored according to the domain diagram. The pink surface represents the substrate-channeling tunnel, which connects the two active sites. The FAD is show in yellow sticks. Catalytic Cys778 is drawn in spheres. (B) The domain-swapped dimer of BbPutA. The two protomers have different colors. The molecular 2-fold axis is vertical. (C) Close-up view of a portion of the dimer interface where the oligomerization flap of one protomer covers the substrate-channeling tunnel of the opposite protomer. (D) Surface representation of the dimer interface shown in panel C, highlighting how dimerization seals the substrate-channeling tunnel from the bulk medium. (E) Results of a PRODH-GSALDH coupled assay for BbPutA. The reaction mixture contained BbPutA (0.2 μM), proline (40 mM), menadione bisulfite (0.1 mM), and NAD⁺ (0.2 mM) in a buffer containing 50 mM potassium phosphate, 25 mM NaCl, and 10 mM MgCl₂ at pH 7.5. Production of NADH was monitored at 340 nm. The data points represent the average of assays performed in triplicate.

Table 2.

Pairwise amino acid sequence identities of type A PutAs

	LpPutA	BjPutA	BbPutA	GsPutA	DvPutA
LpPutA	100	50	31	31	30
BjPutA		100	32	31	29
BbPutA			100	47	48
GsPutA				100	67

BbPutA has a substrate-channeling tunnel. The FAD in the PRODH active site and the catalytic Cys (Cys778) of the GSALDH site are separated by a linear distance of 45 Å and connected by a 65-Å curved tunnel that varies in diameter from 1.5 Å near the active sites to 4.5 Å in the middle section (Fig. 3A). The dimensions of the tunnel are similar to those of other PutAs. A coupled PRODH-GSALDH enzyme activity assay, which tests both active sites and provides preliminary evidence of substrate channeling [26], confirmed that both actives sites are functional (Fig. 3E). Additionally, no lag phase is apparent in the time-dependence of NADH production, consistent with the presence of a functional substrate-channeling tunnel.

The structure of BbPutA was determined from a crystal of the enzyme that had been inactivated with NPPG (Fig. 4A, 1). Previous studies of NPPG-inactivated PRODHs and PutAs show that inactivation is due to a covalent link between the FAD N5 atom and a conserved active site Lys (Fig. 4A, 2). Electron density maps for BbPutA show evidence of this inactivation mechanism. An electron density feature connects the ε-amino group of Lys196 with the N5 atom of the FAD (Fig. 4B). The strength of this feature varied among the four chains in the asymmetric unit, with the strongest density observed in chain D (Fig. 4). The electron density feature could be modeled satisfactorily as a 3-carbon link between Lys196 and the N5 of the FAD consistent with the presence of a covalent NPPG modification in the active site.

Fig. 4.

Covalent modification of the FAD. (A) Structures of (1) NPPG and (2) the covalently-modified FAD resulting from inactivation by NPPG. (B) Electron density evidence of inactivation of BbPutA by NPPG. This view of the N5 edge of the isoalloxazine shows the 25° butterfly bend induced by inactivation. The mesh represents a simulated annealing F_o-F_c omit map contoured at 2.5 σ. (C) Electron density for the modified FAD of BbPutA. The mesh represents a simulated annealing F_o-F_c omit map contoured at 2.5 σ.

The FAD exhibits the structural hallmarks of reduction. Previous studies revealed that NPPG locks the flavin into a conformation that resembles the 2-electron reduced state [28]. These features are seen in the BbPutA structure. In particular, the isoalloxazine moiety exhibits a butterfly bend of 25° (si face convex) (Fig. 4B). For reference, the bend angles of other NPPG-inactivated PutAs and PRODHs are 25° – 35° [23, 28, 29]. Furthermore, the ribityl chain conformation of NPPG-inactivated BbPutA is indicative of reduced PutA/PRODH. The 2´-OH and 3´-OH groups of the inactivated FAD are rotated toward the pyrimidine side of the isoalloxazine, while the 4´-OH is beneath the dimethylbenzene ring (Fig. 4C). This conformation is also observed in other NPPG-inactivated or dithionite-reduced PutAs [23, 28] and PRODHs [29, 30].

Solution oligomeric state analysis of BbPutA by SAXS and analytical ultracentrifugation

The oligomeric state and quaternary structure of BbPutA in solution were determined using SAXS and analytical ultracentrifugation (Fig. 5, Table 3). Guinier analysis of experimental SAXS data yields a radius of gyration (R_g) of 45 Å (Fig. 5A). For reference, the 2-body assembly in the crystallographic asymmetric unit also has R_g of 45 Å (Fig. 3B). The real space R_g from calculations of the distance distribution function is 45.9 – 46.7 Å for assumed maximum particle dimension (D_max) of 140 – 150 Å (Fig. 5B, Table 3). Thus, the reciprocal space and real space radii of gyration are in good agreement. The oligomeric state was estimated using the volume of correlation method [31]. This analysis yields molecular mass (M_r) of 190 kDa, which is within 13% of the theoretical M_r of a dimer (218 kDa, Table 3). Similarly, M_r estimated from the SAXS MoW2 server [32] is 226 kDa, which is within 4% of the dimer (Table 3). Altogether, the SAXS data are consistent with BbPutA forming a dimer under the solution conditions and protein concentration (1.8 mg mL⁻¹) used for SAXS.

Fig. 5.

SAXS and analytical ultracentrifugation of BbPutA. (A) Experimental SAXS curve measured at 1.8 mg mL⁻¹ (open circles). The inset shows a Guinier plot. Theoretical curves calculated from the BbPutA dimer (Fig. 5C) and tetramer (Fig. 5D) are shown in solid red (FoXS χ = 1.3) and red dashes (FoXS χ = 18.1), respectively. (B) Experimental SAXS distance distribution function. (C) The crystallographic dimer of BbPutA. The surface represents the shape reconstruction from DAMMIF. The correlation coefficient between the crystal structure and the shape reconstruction volumetric map is 0.85. (D) The BbPutA tetramer predicted by PDBePISA from analysis of crystal packing. (E) Sedimentation velocity analysis for BbPutA at 4 mg mL⁻¹ (~37 μM). The distribution of apparent sedimentation coefficients (red) and the corresponding molecular mass distribution (black) are shown.

Table 3.

Structural and molecular mass (M_r) parameters from SAXS for BbPutA, DvPutA and LpPutA

BbPutA	1.8 mg mL−1
Rg from Guinier (Å)	44.9 ± 0.4
I(0) from Guinier (Å−1)	1130 ± 7
Dmax (Å)	140 – 150
Rg from P(r) (Å)a	45.9 – 46.7
I(0) from P(r) (Å−1)a	1117 – 1132
Porod volume (Å3)a	288,000 – 289,000
Volume of correlation, Vc (Å2)b	1034
Mr from Vc (kDa)b	190
Mr from MoW2 (kDa)c	226
Monomeric Mr from sequence	109
DvPutA	1.5 mg mL−1	3.0 mg mL−1	4.5 mg mL−1
Rg from Guinier (Å)	43.6 ± 0.4	43.6 ± 0.3	43.9 ± 0.3
I(0) from Guinier (Å−1)	681 ± 4	1600 ± 8	2319 ± 11
Dmax (Å)	150 – 160	150 – 160	150 – 160
Rg from P(r) (Å)a	46.1 – 47.0	45.3 – 45.7	45.7 – 46.5
I(0) from P(r) (Å−1)a	699 – 703	1623 – 1627	2350 – 2364
Porod volume (Å3)a	292,000 – 294,000	294,000 – 295,000	294,000 – 295,000
Volume of correlation, Vc (Å2)b	994	1020	1010
Mr from Vc (kDa)b	180	190	180
Mr from MoW2 (kDa)c	213	224	222
Monomeric Mr from sequence	112	112	112
LpPutA	3 mg mL−1	5 mg mL−1	8 mg mL−1
Rg from Guinier (Å)	45.5 ± 0.2	45.9 ± 0.1	46.3 ± 0.1
I(0) from Guinier (Å−1)	471 ± 1	1010 ± 1	1475 ± 0.2
Dmax (Å)	150 – 160	150 – 166	150 – 170
Rg from P(r) (Å)a	46.5 – 46.9	46.6 – 46.9	46.7 – 47.2
I(0) from P(r) (Å−1)a	473 – 475	1014 – 1018	1470 – 1481
Porod volume (Å3)a	290,000 – 291,000	297,000 – 298,000	294,000 – 296,000
Volume of correlation, Vc (Å2)b	1013	1035	1030
Mr from Vc (kDa)b	180	190	190
Mr from MoW2 (kDa)c	230	241	231
Monomeric Mr from sequence	116	116	116

^aFrom calculations of P(r) using PRIMUS [63] with the indicated range of D_max.

^bCalculated with Scatter 3.0 [69].

^cCalculated from the method of Fischer, et al. [32].

The BbPutA crystal lattice was inspected using PDBePISA [33] to identify plausible oligomers. This analysis revealed two stable assemblies, including the classic type A PutA domain-swapped dimer (R_g = 45 Å, Figs. 3B and 5C) and a ring-shaped tetramer (R_g = 54 Å, Fig. 5D). The predicted tetramer is generated by a crystallographic 2-fold rotation applied to the domain-swapped dimer. Thus, the tetramer is a dimer-of-dimers and has 222 point group symmetry. We note this tetramer resembles the one formed by BjPutA in solution; however, the BbPutA tetramer is slightly larger (R_g of 54 Å versus 51 Å for BjPutA). In addition, PDBePISA analysis returned an octamer consisting of two of the 54 Å tetramers, which was classified in the uncertain region of complex formation criteria (R_g = 68 Å). The experimental R_g of 45 – 47 Å suggests the dimer is the predominant species in solution. FoXS [34, 35] was used to assess the agreements of the dimer and tetramer models to the experimental curve. The scattering curve calculated from the dimer has a goodness-of-fit parameter (χ) of 1.3, whereas the curve calculated from the crystallographic tetramer has a much larger χ of 18 (Fig. 5A). Monomer-dimer and dimer-tetramer equilibria were explored using MultiFoXS [35]. MultiFoXS returned neither monomer-dimer nor dimer-tetramer 2-body fits, indicating that 100% dimer provides the best interpretation of the SAXS data. Finally, shape reconstruction with DAMMIF [36] produced a shape consistent with the crystallographic dimer (Fig. 5C). In summary, SAXS suggests that BbPutA is primarily dimeric in solution, and furthermore the 2-body assembly in the asymmetric unit is the predominant species formed in solution.

Analytical ultracentrifugation was employed to determine whether BbPutA self-associates into higher oligomeric states at a higher concentration than was used in SAXS. A sedimentation velocity experiment performed at 4 mg mL⁻¹ (~37 μM) revealed a distribution of sedimentation coefficients with a major peak at approximately 6.4 S. This apparent sedimentation coefficient corresponds to M_r of approximately 225 kDa (Fig. 5E). The expected M_r of a BbPutA dimer is 218.6 kDa.

Analytical ultracentrifugation was also used to test whether inactivation with NPPG induces tetramerization. This experiment was motivated by the observation of an apparent tetramer in the crystal structure of NPPG-inactivated BbPutA (Fig. 5D). Sedimentation velocity analysis of NPPG-inactivated BbPutA revealed one major peak corresponding to M_r of 225 kDa (Fig. 5E), showing that inactivation does not promote tetramerization in solution.

Finally, sedimentation velocity was performed on BbPutA in the presence of the active site ligands L-tetrahydro-2-furoic acid (THFA, 10 mM) and NAD⁺ (1 mM). THFA is a competitive inhibitor (competitive with proline) of PutAs and is known to occupy the proline-binding site of PutAs and monofunctional PRODHs [23, 24, 30, 37, 38]. NAD⁺ is the cofactor of the GSALDH reaction. This experiment was motivated by our recent observation that THFA and NAD⁺ induce monomers of the class 3B CfPutA to form dimers [25]. The c(s) distribution obtained for BbPutA under this condition exhibited just a single c(s) peak at 6.4 S, corresponding to M_r of 225 kDa (Fig. 5E), showing that the binding of THFA and NAD⁺ do not induce higher order oligomerization. Taken together, the SAXS and analytical ultracentrifugation results are consistent with BbPutA being primarily dimeric under solution conditions that are relevant to catalytic function.

SAXS analysis of two other type A PutAs

To further explore the oligomerization of type A PutAs, SAXS was performed on the class 2A PutA from Desulfovibrio vulgaris (DvPutA) and the class 1A PutA from Legionella pneumophila (LpPutA). DvPutA is 48% identical in amino acid sequence to BbPutA and 67% identical to GsPutA (Table 2). Such high sequence homology is expected from PutAs of the same clade (branch 2, Fig. 2B). In contrast, LpPutA, a branch 1 PutA, is only 30% identical to the class 2A PutAs, but 50% identical to the class 1A BjPutA (Table 2). Analysis of DvPutA and LpPutA provides two more data points on the relationship between sequence and oligomeric state for type A PutAs.

SAXS curves measured at three different concentrations of DvPutA and LpPutA are shown in Fig. 6. SAXS-derived parameters are listed in Table 3. The average R_g from Guinier analysis is 44 Å for DvPutA and 46 Å for LpPutA. These values agree more closely with the R_g of 45 Å of dimeric BbPutA than with the R_g expected for a tetrameric type A PutA (51 – 54 Å). The distance distribution functions of DvPutA and LpPutA resemble that of BbPutA. For all three proteins, the distribution has a major peak near the real-space vector length (r) of 40 Å followed by a shoulder peak at r = 80 – 100 Å (Figs. 5B, 6B, and 6D), suggesting the three proteins share a common quaternary structure. Further, the SAXS-derived molecular masses of DvPutA and LpPutA based on the volume of correlation are within 20% of the theoretical dimer masses for each protein (Table 3). Moreover, the masses from the SAXS MoW2 server are within 4% of the theoretical dimer mass (Table 3). Note also the Porod volumes for DvPutA and LpPutA (290,000 – 298,000 ų) are similar to that of BbPutA (289,000 ų) and almost 2 times smaller than expected for tetrameric PutA (~560,000 ų). The theoretical SAXS curves calculated from domain-swapped dimeric homology models show good agreement with the experimental profiles in the region of q = 0 – 0.1 Å⁻¹ (Fig. 6). We note this region is critical for distinguishing between dimeric and tetrameric PutAs (see Fig. 5A and references [20, 23]). Monomer-dimer and dimer-tetramer equilibria were explored using MultiFoXS [35]. As with BbPutA, MultiFoXS returned neither monomer-dimer nor dimer-tetramer 2-body fits, indicating that 100% dimer provides the best interpretation of the SAXS data for DvPutA and LpPutA. Finally, shape reconstructions generated shapes that are consistent with the dimeric models and do not resemble a ring-shaped particle (Figs. 6A and 6C). Altogether, the SAXS results suggest DvPutA and LpPutA are dimeric in solution and form the classic type A PutA domain-swapped dimer.

Fig. 6.

SAXS analysis of DvPutA and LpPutA. (A) Experimental SAXS curves for DvPutA (open circles) at three concentrations: 1.5, 3.0, and 4.5 mg mL⁻¹. The inset shows Guinier plots. The red curve was calculated from a homology model of the DvPutA domain-swapped dimer (shown in the inset). The χ values obtained from FoXS fits of the theoretical scattering curve of the homology model to the experimental scattering data are as follows: χ = 1.3 (1.5 mg mL⁻¹), χ = 1.7 (3.0 mg mL⁻¹), and χ = 2.3 (4.5 mg mL⁻¹). A homology model of the BbPutA dimer is shown inside the DAMMIF shape reconstruction. The correlation coefficient between the homology model and the shape reconstruction volumetric map is 0.79. (B) Experimental SAXS distance distribution functions for DvPutA. (C) SAXS curves for LpPutA (3, 5, 8 mg mL⁻¹). The inset shows Guinier plots. The red curve was calculated from a homology model of the LpPutA domain-swapped dimer (shown in the inset). The χ values obtained from FoXS fits of the theoretical scattering curve of the homology model to the experimental scattering data are as follows: χ = 3.3 (3.0 mg mL⁻¹), χ = 4.9 (5.0 mg mL⁻¹), and χ = 8.7, (8.0 mg mL⁻¹). A homology model of the LpPutA dimer is shown inside the DAMMIF shape reconstruction. The correlation coefficient between the homology model and the shape reconstruction volumetric map is 0.78. (D) Experimental SAXS distance distribution functions for LpPutA.

Identification of the core functional oligomer of BjPutA

All type A PutAs studied thus far appear to be dimeric in solution except for BjPutA, which is tetrameric (Fig. 7A) [20]. To better understand whether the tetrameric assembly is necessary for BjPutA function, we attempted to generate a hot spot mutant to disrupt tetramerization. Structural analysis of BjPutA revealed Arg51 as potentially important for stabilizing the tetramer. Arg51 is located in the dimer-dimer interface on a flexible loop that connects the N-terminal arm to the α-domain (Fig. 7B). Although Arg51 has weak electron density and appears to make no strong interdomain hydrogen bonds or ion pairs, its location in the center of the dimer-dimer interface nevertheless suggested it could be important for tetramer formation. Analysis with PDBePISA indicates Arg51 contributes 76 Ų of surface area to the tetramer interface, which makes it the second-largest contributor, behind only Tyr474 (99 Ų). Interestingly, Arg does not appear at this position in the sequences of the dimeric type A PutAs studied so far; rather, it is replaced with Glu in LpPutA, Gln in BbPutA, and Gly in GsPutA and DvPutA. Therefore we generated a mutant, R51E, to reflect the charge reversal seen in LpPutA. We note the introduction of Glu51 in BjPutA results in three consecutive acidic residues: Glu51-Asp52-Asp53.

Fig. 7.

Structural context of Arg51 of BjPutA. (A) The BjPutA tetramer with the four protomers in different colors. The domain-swapped dimers are colored red-cyan and gold-blue. Arg51 is shown in spheres in the right hand image. The boxed region is expanded in panel B. (B) Close-up view of the dimer-dimer interface. The dashed curves represent disordered residues 52–53.

To understand any effects of this mutation on the quaternary structure of BjPutA, we first purified wild-type BjPutA and subjected it to analysis by sedimentation velocity. Initial sedimentation velocity studies of wild-type BjPutA revealed a major peak near apparent sedimentation coefficient of 10.8 S (Fig. 8A), which corresponds to M_r of 429 kDa (Fig. 8B). The predicted M_r of the BjPutA tetramer is also 429 kDa.

Fig. 8.

Solution oligomeric state analysis of BjPutA and BjPutA R51E. (A) The distribution of apparent sedimentation coefficients from sedimentation velocity observed for BjPutA (28 μM, black) or BjPutA R51E (28 μM, red). (B) The distribution of molecular masses from sedimentation velocity observed for BjPutA (28 μM, black) or BjPutA R51E (28 μM, red). (C) SAXS curves measured at three wild-type BjPutA protein concentrations (2.3, 4.7, 7.0 mg mL⁻¹). The inset shows Guinier plots. The theoretical SAXS curve calculated from the crystallographic tetramer (inset) is shown in cyan (χ values of 4.3 (2.3 mg mL⁻¹), 7.3 (4.7 mg mL⁻¹), and 9.1 (7.0 mg mL⁻¹)). The theoretical curves obtained from MultiFoXS assuming a mixture of the crystallographic tetramer and dimer are shown in red (χ values of 1.6 (2.3 mg mL⁻¹), 3.3 (4.7 mg mL⁻¹), and 4.4 (7.0 mg mL⁻¹)). The optimal tetramer:dimer compositions from MultiFoXS are 74%:26% (2.3 mg mL⁻¹), 77%:23% (4.7 mg mL⁻¹), and 81%:19% (7.0 mg mL⁻¹). The crystallographic tetramer of BjPutA is shown inside the DAMMIF shape reconstruction. The correlation coefficient between the tetramer and the shape reconstruction volumetric map is 0.78. (D) Experimental SAXS distance distribution functions for wild-type BjPutA. (E) SAXS curves measured at three BjPutA R51E protein concentrations. The inset shows Guinier plots. The theoretical SAXS curve calculated from the BjPutA domain-swapped dimer (inset) is shown in red. The χ values obtained from FoXS fits of the theoretical scattering curve of the crystallographic dimer to the experimental scattering data are as follows: χ = 0.85 (2.3 mg mL⁻¹), χ = 1.3 (4.7 mg mL⁻¹), and χ = 2.5 (7.0 mg mL⁻¹). The crystallographic dimer of BjPutA is shown inside the DAMMIF shape reconstruction. The correlation coefficient between the tetramer and the shape reconstruction volumetric map is 0.83. (F) Experimental SAXS distance distribution functions for BjPutA R51E.

Wild-type BjPutA was also studied with SAXS. The SAXS curves for wild-type BjPutA show pronounced trough and peak features in the region of q = 0.05 – 0.1 Å⁻¹ (Fig. 8C). We have shown previously that these features are diagnostic of the ring-shaped tetrameric form of type A PutA [20, 23]. Note these diagnostic features are absent in the SAXS curves for all the other PutAs studied here. Also, the distance distribution function of wild-type BjPutA is very different from those of the other proteins in this study (Fig. 8D). The distribution function of wild-type BjPutA indicates a most-probable real-space vector length of r ~ 80 Å (Fig. 8D), compared to approximately 40 Å for the other PutAs (Figs. 5B, 6B, and 6D). The SAXS R_g of 51 – 53 Å for wild-type BjPutA (Table 4) is in good agreement with the R_g of 51 Å calculated from the BjPutA tetramer. The molecular masses derived from SAXS are within 0.5 – 11 % of the expected mass of a tetramer (Table 4). The SAXS curve calculated from the BjPutA tetramer shows good agreement with the experimental curves (χ = 4.3 – 9.1, Fig. 8C). The fits could be improved somewhat (to χ = 1.6 – 4.4) by using a tetramer:dimer ensemble in MultiFoXS (Fig. 8C). The optimal ratio of tetramer:dimer ranged from 74%:26% for the lowest concentration sample to 81%:19% for the highest concentration sample. Thus, the fitting calculations are consistent with BjPutA being predominantly tetrameric in solution. Finally, shape reconstruction calculations using the SAXS data for the highest concentration sample returned a ring-shaped particle consistent with the crystallographic tetramer of BjPutA (Fig. 8C). In summary, the SAXS data show that wild-type BjPutA exists in solution primarily as a ring-shaped tetramer, in agreement with our previously crystallographic, SAXS, and centrifugation studies [20, 23].

Table 4.

Structural and molecular mass (M_r) parameters from SAXS for BjPutA and BjPutA R51E

BjPutA	2.3 mg mL−1	4.7 mg mL−1	7.0 mg mL−1
Rg from Guinier (Å)	53 ± 2	52 ± 1	51.9 ± 0.7
I(0) from Guinier (Å−1)	183 ± 8	409 ± 7	607 ± 7
Dmax (Å)	140 – 146	140 – 146	136 – 140
Rg from P(r) (Å)a	51.4 – 51.6	51.6 – 51.7	51.4 – 51.5
I(0) from P(r) (Å−1)a	177 – 180	406 – 412	612 – 613
Porod volume (Å3)a	541,000 – 549,000	553,000 – 561,000	560,000
Volume of correlation, Vc (Å2)b	1577	1577	1566
Mr from Vc (kDa)b	380	390	380
Mr from MoW2 (kDa)c	426	424	426
Monomeric Mr from sequence	107	107	107
BjPutA R51E	2.3 mg mL−1	4.7 mg mL−1	7.0 mg mL−1
Rg from Guinier (Å)	44.9 ± 0.7	45.0 ± 0.6	44.9 ± 0.4
I(0) from Guinier (Å−1)	118.9 ± 0.4	232 ± 3	356 ± 3
Dmax (Å)	140 – 150	136 – 145	140 – 150
Rg from P(r) (Å)a	45.3 – 45.7	44.9 – 45.0	45.8 – 46.0
I(0) from P(r) (Å−1)a	116 – 117	224 – 227	356 – 358
Porod volume (Å3)a	280,000 – 281,000	278,000 – 283,000	288,000 – 289,000
Volume of correlation, Vc (Å2)b	1005	1011	1010
Mr from Vc (kDa)b	180	180	180
Mr from MoW2 (kDa)c	231	234	219
Monomeric Mr from sequence	107	107	107

^aFrom calculations of P(r) using PRIMUS [63] with the indicated range of D_max.

^bCalculated with Scatter 3.0 [69].

^cCalculated from the method of Fischer, et al. [32]

The mutation of Arg51 to Glu profoundly changes the oligomeric structure of BjPutA. To determine the effects of the R51E mutation on the oligomeric state of BjPutA, we subjected the BjPutA R51E mutant variant to the same analysis as wild-type BjPutA. Sedimentation velocity experiments on BjPutA R51E revealed a major peak at apparent sedimentation coefficient of 6.5 S, which corresponds to M_r of 216 kDa (Figs. 8A and 8B). The theoretical M_r for the BjPutA dimer is 214.7 kDa, which is within 0.6% of the result from sedimentation velocity, suggesting BjPutA R51E is dimeric in solution. SAXS analysis of BjPutA R51E at three protein concentrations revealed experimental curves statistically very similar (χ = 0.9 – 2.1) to the theoretical curves generated by FoXS using the crystallographic domain-swapped dimer (Fig. 8E). Monomer-dimer and dimer-tetramer equilibria were explored using MultiFoXS [35]. MultiFoXS returned neither monomer-dimer nor dimer-tetramer 2-body fits, indicating that 100% dimer provides the best interpretation of the SAXS data for R51E. Guinier analysis returned R_g of 45 Å, consistent with the R_g of dimeric type A PutAs (Table 4). The real-space R_g of 45 – 46 Å agrees well with the Guinier R_g. The P(r) distribution for R51E (Fig. 8F) is distinctly different from that of wild-type BjPutA (Fig. 8D). Further, fitting a representative experimental R51E SAXS curve to the BjPutA crystallographic dimer yields a greatly improved statistical fit (χ = 1.3) compared to fitting the experimental data to the BjPutA tetramer (χ = 51; Fig. 9A). Moreover, the experimental P(r) distributions of wild-type BjPutA and R51E are qualitatively very different (Fig. 9B). In fact, the area under the distance distribution function of R51E is approximately 54% of the area under the distance distribution function of wild-type BjPutA at the same protein concentration, consistent with R51E being half the size of wild-type BjPutA (Fig. 9B). Overall, these results suggest that the R51E mutation cleanly disrupts tetramerization, resulting in consistently dimeric BjPutA.

Fig. 9.

Comparison of the in-solution properties of BjPutA and BjPutA R51E. (A) A representative experimental SAXS curve of BjPutA R51E (4.7 mg mL⁻¹) is shown in open circles. Overlaid are the theoretical SAXS curves calculated from the BjPutA domain-swapped dimer (solid red) and the BjPutA tetramer (dashed red). The χ values obtained from FoXS fits of the theoretical scattering curve of the crystallographic dimer and crystallographic tetramer to the experimental scattering data are 1.3 and 51.9, respectively. (B) Experimental distance distributions for wild-type BjPutA (black) and R51E (red).

To determine whether the R51E mutation has an effect on the catalytic activity, we performed the coupled PRODH-GSALDH activity assay, which reports on both catalytic activities and provides information on substrate channeling. Previous results have shown that at high concentrations of proline, kinetic data for the coupled PRODH-GSALDH assay fit to a substrate inhibition model [26]. Substrate inhibition was also observed here for both wild-type BjPutA and R51E (Figs. 10A and 10B). Using proline as the variable substrate, wild-type BjPutA displayed a K_m of 4 mM, k_cat of 0.31 s⁻¹, and inhibition constant for proline (K_i) of 400 mM (Fig. 10A, Table 5). Similarly, the kinetic parameters for R51E are K_m = 10 mM, k_cat = 0.4 s⁻¹, and K_i of 220 mM (Fig. 10B, Table 5). The catalytic efficiencies (k_cat/K_m) of wild-type BjPutA and R51E are within a factor of 2 (Table 5). These results suggest that tetramerization is not essential for the in vitro catalytic activity of BjPutA.

Fig. 10.

Kinetic data for wild-type BjPutA and R51E, and the effects of active site ligands on the oligomeric state of R51E. (A) Dependence of the coupled PRODH-GSALDH reaction rate on proline concentration for wild-type BjPutA (black points). Kinetic data were fit to a substrate inhibition model (red curve). (B) Dependence of the coupled PRODH-GSALDH reaction rate on proline concentration for R51E (black points). Kinetic data were fit to a substrate inhibition model (red curve). The assays used fixed concentrations of proline (40 mM), CoQ₁ (0.1 mM), and NAD⁺ (0.2 mM). (C) The distribution of sedimentation coefficients observed from sedimentation velocity experiments performed in the absence of ligands (solid black), after inactivation with NPPG (red), or in the presence of the active site ligands THFA (10 mM) and NAD⁺ (1 mM) (blue). The distribution of sedimentation coefficients observed for wild-type BjPutA (dashed black) in the absence of ligands is provided for reference. The enzyme concentration was 28 μM in all experiments. (D) Dependence of the coupled PRODH-GSALDH reaction rate on protein concentration for R51E (black points). Data were fit to a linear regression model (red line).

Table 5.

Kinetic Parameters for BjPutA and BjPutA R51E

	Km (mM)	kcat (s−1)	kcat/Km (M−1s−1)	Ki (mM)
BjPutA	4 ± 1	0.31 ± 0.03	70 ± 20	400 ± 100
BjPutA R51E	10 ± 2	0.40 ± 0.04	40 ± 10	220 ± 50

We considered the hypothesis that dimeric R51E assembles into the tetramer under the conditions of the activity assay, i.e., that reduction of the FAD or substrate binding enhances the association of two dimers into the tetramer. To test this idea, we first performed sedimentation velocity experiments after inactivation by NPPG. NPPG locks the FAD into the reduced state, thus mimicking a major effect of proline oxidation on the enzyme. NPPG-inactivated R51E displayed an apparent sedimentation coefficient of 5.9 S (Fig. 10C), which is close to that of untreated R51E (6.5 S) and far from that of wild-type BjPutA (10.8 S). We next performed sedimentation velocity of R51E in the presence of the active site ligands THFA and NAD⁺, which bind in the PRODH and GSALDH active sites, respectively. We note the binding of these ligands to a monomeric type B PutA (CfPutA) induces dimerization [25]. R51E in the presence of THFA and NAD⁺ exhibits an apparent sedimentation coefficient (6.2 S) consistent with a dimer (Fig. 10C). Thus, in contrast to monomeric type B PutA, the binding of active site ligands does not induce higher order assembly of R51E. Finally, we conducted the coupled activity assay for R51E as a function the enzyme concentration. The observed rate is linearly proportional to R51E concentration, as expected for an enzyme that does not require assembly into a higher order oligomer for activity (Fig. 10D). Taken together, our results suggest that R51E forms a catalytically competent dimer.

Discussion

Almost 50% of proteins are homooligomers, implying that self-association underlies function [39, 40]. There are many possible reasons for oligomerization. For example, substrate and cofactor binding sites of enzymes can occur in oligomer interfaces [41, 42]. Even when the active site is not in an interface, oligomerization may still be essential for catalytic activity, as in some ALDHs [43, 44]. Many integral membrane transporters function as oligomers [45]. Quaternary structure is also important in allosteric proteins [46], cooperative enzymes [47], and morpheein enzymes [48]. Oligomerization can also contribute to protein stability [41].

Here we investigated the functional significance of oligomerization in type A PutAs. The first type A PutA studied, BjPutA, was found to be a tetramer consisting of a pair of domain-swapped dimers that form a ring [20]. Herein, we showed that several other type A PutAs are dimeric, implying the BjPutA tetramer may be an exception.

The domain-swapped dimer observed in all type A PutA crystal structures, whether as a stand-alone dimer or half of a tetramer, has obvious functional relevance. The oligomerization flap of one protomer seals the substrate-channeling tunnel of the other protomer, as shown for BbPutA (Figs. 3C, 3D). Without this intermolecular lid, the intermediate would have a higher probability of diffusing into the bulk medium. Thus, domain-swapped dimerization in type A PutAs enables the substrate-channeling step of the catalytic mechanism. We note that in type B PutAs, the intermolecular lid is replaced by an intramolecular (tertiary structural) interaction involving the C-terminal ALDH superfamily domain [21, 24, 25].

The functional significance of the BjPutA tetramer is not obvious, prompting the work reported here. The dimer-dimer interface is formed by the N-terminal arm and α-domain, two regions of the protein that lack catalytic residues and ligand binding sites. The interface is far from the active sites and substrate-channeling tunnel. For example, the hot spot residue Arg51 is 30 Å from the nearest flavin N5 and 50 Å from the catalytic Cys of the GSALDH site. Consistent with these observations, the dimeric hot spot variant R51E exhibits wild-type catalytic behavior and displays an apparent sedimentation coefficient consistent with the formation of a dimer in solution. The sedimentation data for R51E contain no evidence for tetramer formation. Further, it was observed that neither flavin reduction, nor incubation with active site ligands results induces tetramer formation by R51E or BbPutA (Figs. 5E and 10C). Therefore, it is concluded that tetramerization is not essential for catalytic function, suggesting that the domain-swapped dimer is the core oligomeric structure of type A PutAs. We note that this conclusion is rendered from in vitro biochemical and biophysical analysis of PutAs, and that, under cellular conditions, molecular crowding or other favorable conditions may promote higher-order oligomerization.

Our results for BjPutA are consistent with the hot spot theory of protein-protein interaction [49, 50]. Hot spots refer to the region of a protein-protein interface that contains a few critical residues that account for most of the association energy. Hot spots have a distinctive amino acid composition – often Trp, Arg, or Tyr [50]. Consistent with the hot spot theory, we could abrogate tetramerization of BjPutA with the single mutation of Arg51 to Glu. It is interesting that this particular Arg residue makes no specific electrostatic interactions in the native tetramer (Fig. 7). Furthermore, Arg51 is located on a disordered loop. Thus, visual inspection of the crystal structure might not have revealed Arg51 as a hot spot residue. However, analysis of the interface with PISA indicated that Arg51 contributes significantly to the dimer-dimer interface surface area, despite the lack of the traditional electrostatic interactions one typically associates with Arg. This result points to buried surface area as a key metric for identifying hot spot residues.

Additionally, our results indicate prediction of quaternary structure and oligomeric state from crystal packing remains challenging in some cases. The PISA algorithm, which is based on physical-chemical models of protein interactions and chemical thermodynamics, is perhaps the gold standard for predicting oligomeric state from crystal packing and is used by the PDB to identify the most likely biological assembly [33]. PISA analysis suggested BbPutA forms a stable tetramer in solution (Fig. 5D). However, our SAXS and sedimentation data showed no evidence of a tetramer at concentrations in the range of 1.8 – 4 mg mL⁻¹ (16 – 37 μM). Although it is possible that the tetramer could form at even higher concentrations, we have observed solubility problems with BbPutA at concentrations above 10 mg mL⁻¹. It is possible that the high protein concentration achieved as the crystallization drop equilibrates promotes formation of the higher order assembly seen in the crystal. Such high concentrations are not possible in solution for BbPutA, so the tetramer may not be observable with standard solution biophysical techniques. Finally, it remains possible that the tetramer could form in the cell, where molecular crowding may enhance association of dimers. We have also encountered this phenomenon - where the crystal and solution data conflict - with monofunctional GSALDHs [51]. In that case, the wild-type enzyme is hexameric both in solution and in the crystal. However, mutation of a hexamerization hot spot residue produced dimeric protein in solution. Curiously, the dimeric protein was observed to form the hexamer in the crystal. These results demonstrate that it remains challenging to predict oligomeric state in solution from crystal structures for some systems.

Materials and methods

Production and crystallization of BbPutA

A synthetic gene encoding PutA from Bdellovibrio bacteriovorus HD100 (BbPutA, 982 residues, NCBI RefSeq number NP_968157.1) with codons optimized for expression in E. coli was purchased from BIO BASIC Inc. (Markham, Ontario CA). The gene was subcloned from pUC57 into pKA8H between NdeI and BamHI sites. The encoded protein has an N-terminal 8xHis tag that is cleavable by tobacco etch virus protease (TEVP).

BbPutA was expressed in BL21-AI cells and purified with affinity chromatography (Ni²⁺-charged HisTRAP; GE Healthcare) using protocols described previously for GsPutA [23]. The His tag was cleaved as described previously for GsPutA [23]. Purified BbPutA was dialyzed into a storage buffer consisting of 50 mM Tris, 125 mM NaCl, 1 mM EDTA, and 1 mM tris(3-hydroxypropyl)phosphine (THP) at pH 7.5, and then concentrated to 4 mg mL⁻¹ using stirred ultrafiltration. The protein concentration was estimated using the bicinchoninic acid assay (BCA).

Crystal screening and optimization trials of active BbPutA produced weakly diffracting crystals, which were unsuitable for structure determination. Therefore, we pursued crystallization of BbPutA inactivated by the mechanism-based inactivator NPPG (Fig. 4A). Previous studies have shown that NPPG covalently modifies the FAD and induces a conformation that resembles the 2-electron reduced enzyme [23, 28, 29]. We hypothesized that the change of conformation might enhance the crystallization properties of BbPutA. This strategy produced crystals that diffracted to 2.2 Å resolution.

BbPutA was incubated with NPPG (gift from Dr. Chris Whitman) at a ratio of 1 mg of enzyme per 1 mg of inactivator for 30 min on ice. The inactivated BbPutA was passed through a 0.22 μM centrifugal filter device at 4°C to remove precipitate. Crystal screening trials were performed at 20°C with commercial kits using the microbatch method with drops formed by mixing 1.5 μL of inactivated BbPutA (3 mg mL⁻¹) and 1.5 μL of crystallization reagent. The drops were covered with Al’s oil (Hampton Research). Several conditions yielded crystals overnight. Based upon the size and ease of reproducibility, the hit from reagent 10 of Wizard III (Emerald Biosystems) was selected for additional optimization. These efforts produced crystals grown in the presence of 19% (w/v) polyethylene glycol (PEG) 3350 and 0.25 M KSCN. This crystal form was improved by using Hampton Index reagents as additives. The base condition (19% (w/v) PEG 3350, 0.25M KSCN) was mixed with each Hampton Index reagent at a ratio of 80:20 (base:additive) and used in microbatch trials. Crystals of equivalent quality were obtained using Index reagents 45, 72, 73, 76, 79, or 83 as additives. In preparation for cryogenic data collection, the crystals were cryoprotected using 24% (w/v) PEG 3350, 0.25 M KSCN, and 25% (v/v) ethylene glycol and plunged into liquid nitrogen. The space group is P2₁22₁ with the unit cell parameters listed in Table 1. The asymmetric unit contains four protein molecules arranged as two dimers. The method of Matthews predicts 57% solvent (V_M = 2.9 ų/Da) [52].

X-ray diffraction data collection, phasing, and refinement

Crystals of inactivated BbPutA were analyzed at Advanced Photon Source beamline 24-ID-E. The data set used for refinement consisted of 400 frames of data collected with an oscillation width of 0.25° and detector distance of 300 mm. The data were processed with XDS [53] and SCALA [54]. Initial phases were determined with molecular replacement as implemented in MOLREP using a search model derived from the coordinates of GsPutA (47% sequence identity, Table 2). The structure was refined in PHENIX [55] and adjusted manually with COOT [56]. The refinement calculations used non-crystallographic symmetry restraints. The B-factor model consisted of one TLS group per protein chain and isotropic B-factors for all non-hydrogen atoms. The model was validated with MolProbity [57, 58].

Production of LpPutA and DvPutA

The gene encoding PutA from Legionella pneumophila subsp. pneumophila (LpPutA, 1054 residues, NCBI RefSeq WP_010947423.1) was cloned from genomic DNA purchased from ATCC (catalogue number 33152D) and inserted into pET151. The expressed protein contains an N-terminal 6xHis tag followed the V5 epitope and TEVP cleavage site. LpPutA was expressed in BL21 DE3 Star cells and purified with affinity chromatography (Ni²⁺-charged HisTRAP) and anion exchange chromatography (HiTrap Q; GE Healthcare) using protocols similar to those described for BjPutA [59]. The His tag was cleaved as described for BjPutA [59]. The purified protein was dialyzed overnight at 4°C into a buffer containing 50 mM Tris-HCl, 50 mM NaCl, 0.5 mM THP, 5% (v/v) glycerol, and 0.5 mM EDTA at pH 7.5.

The gene encoding PutA from Desulfovibrio vulgaris str. Hildenborough (DvPutA, 1006 residues, NCBI RefSeq WP_010940575.1) in the expression plasmid pNIC28-Bsa4 was obtained from the New York Structural Genomics Research Consortium. The expressed protein has an N-terminal 6xHis tag and TEVP cleavage site. DvPutA was expressed in BL21(DE3)pLysS and purified as described above for LpPutA. The His tag was not cleaved. The purified protein was dialyzed overnight at 4°C into a storage buffer containing 50 mM Tris-HCl, 50 mM NaCl, 0.5 mM EDTA, 5% (v/v) glycerol, and 0.5 mM THP at pH 7.5.

Mutagenesis, production, and activity assays of BjPutA and BjPutA R51E

Wild-type and mutant PutA from Bradyrhizobium japonicum (BjPutA) in the pKA8H vector were expressed and purified as previously described [59]. The His tag was removed from both enzymes as previously described [59]. The R51E mutant variant of BjPutA was generated from the pKA8H-BjPutA construct using the QuikChange II XL kit (Agilent). The purified proteins were dialyzed overnight against a storage buffer containing 50 mM Tris (pH 7.8), 50 mM NaCl, 0.5 mM Tris(2-carboxyethyl)phosphine, and 5% (v/v) glycerol.

Assessment of the coupled PRODH-GSALDH activity of both wild-type and R51E BjPutA was carried out as previously described [26]. Briefly, NADH formation was monitored at 340 nm in assays performed at room temperature in an Epoch 2 plate reader (BioTek). The reaction mixture contained BjPutA or BjPutA R51E (0.5 μM), proline (2.5 – 400 mM), coenzyme Q₁ (0.1 mM), and NAD⁺ (0.2 mM) in a buffer containing 50 mM potassium phosphate, 25 mM NaCl, and 10 mM MgCl₂ at pH 7.5 with a final reaction volume of 200 μL. Reactions were performed in the presence and absence of NAD⁺ to correct for CoQ₁ reduction. Linear regression in Origin 2017 was used to determine rates for final analysis. The rate data were fit to a substrate inhibition model in Origin 2017.

Analytical ultracentrifugation

Sedimentation velocity experiments were performed in a Beckman XL-I analytical ultracentrifuge using an An50Ti rotor at 20°C. Aliquots of the protein solution and dialysis buffer (reference buffer) were loaded into a sedimentation velocity cell, bearing a two-sector charcoal-Epon centerpiece. Prior to centrifugation, the sample was allowed to equilibrate for 2-hours. The sample was then centrifuged at 35,000 rpm for 300 radial scans at two-minute intervals acquired using Rayleigh interference optics. Scans 10–300 were used in the analysis. Sedimentation coefficient, c(s), and molecular mass, c(M), distributions were generated using Sedfit [60].

Prior to sedimentation analysis of NPPG-inactivated BbPutA and BjPutA R51E, each protein was treated with 1 mg NPPG (BOC Sciences, Shirley, NY) per 1 mg of protein, incubated on ice for 30 minutes, and then loaded into the sedimentation cell. The reference for sedimentation was the respective storage buffer (listed in the purification protocols above) supplemented with an equal amount of NPPG used in the inactivation step.

Prior to sedimentation analysis of BbPutA and BjPutA R51E in the presence of active site ligands, protein samples were supplemented with 10 mM THFA and 1 mM NAD⁺ and dialyzed for four hours with two buffer exchanges in a Slide-a-lyzer mini-dialysis device (ThermoFisher) against the storage buffer (listed in the purification protocol above) supplemented with 10 mM THFA and 1 mM NAD⁺. The dialysate served as the reference for sedimentation.

For all BjPutA wild-type and mutant samples, the frictional ratio was allowed to vary during global fitting. In the analysis of BbPutA, the frictional ratio was set at 1.75 to account for sample aggregation and precipitation. We note that the frictional ratio of BjPutA R51E was approximately 1.75, which is why this frictional ratio was applied to the BbPutA data.

SAXS

Purified BbPutA was prepared for SAXS by passing it through a pre-packed Superdex-200 10 × 300 mm size exclusion chromatography (SEC) column (GE Life Sciences) equilibrated with 50 mM Tris, 125 mM NaCl, 1 mM EDTA, and 1 mM THP at pH 7.5 at a flow rate of 0.5 mL min⁻¹. The protein eluted as a single peak that was baseline-separated from the void volume of the column (Fig. 11). Fractions under the peak were pooled for SAXS analysis. A sample of the SEC flow-through was reserved for use in the SAXS background measurement.

Fig. 11.

SEC chromatograms for BbPutA (blue), wild-type BjPutA (black), and BjPutA R51E (red) obtained with a pre-packed Superdex-200 10 × 300 mm SEC column (GE Life Sciences) connected to an AKTA pure chromatography instrument (GE Life Sciences).

Purified LpPutA and DvPutA were prepared for SAXS by passing them through the aforementioned SEC column equilibrated with 50 mM Tris, 500 mM NaCl, and 1 mM THP at pH 8.0 (flow rate of 0.5 mL min⁻¹). Each protein eluted as a single peak that was baseline separated from the void volume. The protein in the single peak was pooled and concentrated using a centrifugal concentration device (30-kDa cutoff) to 8 mg mL⁻¹ as monitored by the BCA assay. The SEC equilibration buffer was used for the SAXS background measurement.

Purified wild-type BjPutA and R51E were prepared for SAXS by passing them through the aforementioned SEC column equilibrated with 50 mM Tris (pH 7.8), 50 mM NaCl, 0.5 mM Tris(2-carboxyethyl)phosphine, and 5% (v/v) glycerol (flow rate of 0.5 mL min⁻¹). Each protein eluted as a single peak that was baseline separated from the void volume (Fig. 11). The protein in the single peak was pooled and concentrated using a centrifugal concentration device (50-kDa cutoff). The concentrated protein was dialyzed overnight using the SEC buffer. The final protein concentration after dialysis was 7 mg mL⁻¹ (BCA method). A sample of the dialysate was reserved for the SAXS background measurement.

The protein samples were transferred to 96-well plates and shipped at 4°C to the SIBYLS beamline 12.3.1 of the Advanced Light Source through the mail-in program [61, 62]. Additional SEC steps were not performed at the beamline. The SAXS intensity data – I(q) vs. q, q = 4πsinθ/λ, where 2θ is the scattering angle and λ is the X-ray wavelength in angstroms – were measured at three nominal protein concentrations in the range of 1 – 8 mg mL⁻¹. Note that scattering intensities for BbPutA were calculated only at a single protein concentration, 1.8 mg mL⁻¹. Because of the low concentration of the BbPutA SEC fractions, data were collected at a single concentration of 1.8 mg mL⁻¹. For BbPutA, DvPutA, and LpPutA, data were collected for each protein concentration at exposure times of 0.5 s, 1.0 s, 3.0 s, and 6.0 s. For both wild-type BjPutA and the BjPutA R51E mutant variant, data were collected in shutterless mode using a Pilatus detector with a total of 33 evenly spaced images acquired over 10.2 s (0.3 s/frame). The scattering curves collected from the protein samples were corrected for background scattering using intensity data collected from the aforementioned reference buffers.

For all except the BjPutA and BjPutA R51E samples, composite scattering curves were generated with PRIMUS [63] by scaling and merging the background-corrected low q region data from the 0.5 s exposure with the high q region data from the 3.0 s exposure. For both wild-type BjPutA and BjPutA R51E, composite scattering curves were generated with PRIMUS by averaging and merging the background-corrected low q region data from the first three (0.9 s) exposures with the high q region data from the first 12 (3.6 s) exposures.

The composite SAXS curves were analyzed as follows. PRIMUS was used to perform Guinier analysis. GNOM was used to calculate distance distribution functions [64]. FoXS and MultiFoXS [35] were used to calculate theoretical SAXS curves from atomic models. For BjPutA and BbPutA, crystal structures were input to FoXS. For DvPutA and LpPutA, homology models were generated using SWISS-MODEL [65], RaptorX [66], and Phyre2 [67]. The LpPutA model was improved with AllosMod-FoXS [34, 68], which added missing residues and allowed for inter-domain movements. The molecular mass in solution was determined from SAXS data using the volume of correlation invariant [31] as implemented in Scatter 3.0 [69] and with the SAXS MoW2 server [32].

DAMMIF [36] was used for shape reconstructions. For each reconstruction, 50 independent calculations were performed. Two-fold symmetry was enforced during the reconstructions of the dimeric PutAs (BbPutA, DvPutA, LpPutA, and BjPutA R51E). Point group 222 symmetry was enforced during the shape reconstruction of wild-type BjPutA. The models from DAMMIF were averaged and filtered with DAMAVER [70]. The averaged and filtered dummy atom models (dammif.pdb) were superimposed onto crystal structures or homology models with supcomb [71]. The pdb2vol utility of situs [72] was used to convert dummy atom models (dammif.pdb) into volumetric maps. The colores utility of situs was used to calculate the correlation coefficient between atomic models and volumetric maps. The SAXS data have been deposited in the SASBDB [73] under the following accession codes: SASDCP3 (BbPutA), SASDCQ3 (DvPutA 1.5 mg mL⁻¹), SASDCX3 (DvPutA 3.0 mg mL⁻¹), SASDCY3 (DvPutA 4.5 mg mL⁻¹), SASDCR3 (LpPutA 3.0 mg mL⁻¹), SASDCV3 (LpPutA 5.0 mg mL⁻¹), SASDCW3 (LpPutA 8.0 mg mL⁻¹), SASDCS3 (BjPutA 2.3 mg mL⁻¹), SASDCT3 (BjPutA 4.7 mg mL⁻¹), SASDCU3 (BjPutA 7.0 mg mL⁻¹), SASDCZ3 (R51E 2.3 mg mL⁻¹), SASDC24 (R51E 4.7 mg mL⁻¹), SASDC34 (R51E 7.0 mg mL⁻¹).

Abbreviations

ALDH

aldehyde dehydrogenase

BbPutA

Bdellovibrio bacteriovorus proline utilization A

BCA

bicinchoninic acid assay

BjPutA

Bradyrhizobium japonicum proline utilization A

CfPutA

Corynebacterium freiburgense proline utilization A

DvPutA

Desulfovibrio vulgaris proline utilization A

GSALDH

L-glutamate-γ-semialdehyde dehydrogenase

GsPutA

Geobacter sulfurreducens proline utilization A

LpPutA

Legionella pneumophila proline utilization A

P5C

Δ¹-pyrroline-5-carboxylate

PRODH

proline dehydrogenase

PDB

Protein Data Bank

PutA

proline utilization A

SAXS

small-angle X-ray scattering

SEC

size exclusion chromatography

SmPutA

Sinorhizobium meliloti proline utilization A

THP

Tris(3-hydroxypropyl)phosphine

TEVP

tobacco etch virus protease