The sequences coding for the four putative P5C reductase proteins from Bacillus subtilis strain 168 were amplified by PCR using the primers reported in Table 1, and cloned into vector pMCSG68 according to the standard protocol described previously (Eschenfeldt et al., 2013; Forlani et al., 2015b). The pMCSG68 vector introduces a His₆-tag followed by the Tobacco Etch Virus (TEV) protease cleavage site at the N-terminus of the expressed protein. The correctness of the insert was confirmed by DNA sequencing. Overexpression was carried out in BL21 Gold E. coli cells (Agilent Technologies). The bacteria were cultured with shaking at 210 rpm in LB medium supplemented with 150 μg mL⁻¹ ampicillin at 37°C until the OD₆₀₀ reached 1.0. The temperature was lowered to 18°C, and isopropyl-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM. The culture was grown for 18 h and then centrifuged at 4,000 g for 10 min at 4°C. The cell pellet from 1 L culture was resuspended in 35 mL of lysis buffer (50 mM HEPES sodium salt pH 8.0, 500 mM NaCl, 5% glycerol, 20 mM imidazole, 10 mM β-mercaptoethanol) and stored at −80°C.

All proteins were purified according to the standard protocol for Ni-NTA affinity chromatography, as described previously (Forlani et al., 2015c). The His₆-tag was removed by treating each enzyme with His₆-tagged TEV protease for 16 h at 4°C in 50 mM HEPES buffer, pH 8.0. Cleaved protein was separated from TEV protease using Ni-NTA affinity chromatography, further purified by size-exclusion chromatography on a HiLoad 16/600 Superdex 200 Prep Grade (GE Healthcare) in standard lysis buffer, and concentrated to ~10 mg mL⁻¹.

Small-angle X-ray Scattering Studies (SAXS) were performed at the BioCAT/18ID beamline at the Advanced Photon Source, Argonne National Laboratory. Each native protein at a concentration of 5 mg mL⁻¹ was loaded onto a Superdex 200 column, pre-equilibrated with 20 mM HEPES buffer, pH 8.0, containing 150 mM NaCl and 2 mM TCEP, using a size exclusion chromatographic system (AKTA pure, GE Healthcare) and monodisperse samples were routed directly and continuously into a flow cell for subsequent X-ray exposure. A photon-counting PILATUS 3 1M detector was used to record the scattered X-rays at a wavelength of 1.03 Å. The sample-to-detector distance was 3.5 m and yielded a range of 0.005–0.33 Å⁻¹ for the momentum transfer (q = 4π sinΘ/λ, where 2Θ is the angle between the incident and scattered beam and λ is the X-ray wavelength). The standard data reduction procedure for biological SAXS was performed with the programs in the ATSAS package (Petoukhov et al., 2012). The radius of gyration Rg and the maximum dimension of the particle D_max were obtained using Guinier analysis and the calculated pair distance distribution function p(r). Ten bead models were reconstructed ab initio using DAMMIF (Franke and Svergun, 2009) and averaged using DAMAVER (Volkov and Svergun, 2003). Atomic coordinates of P5C reductase from Streptococcus pyogenes (PDB 2AMF; Nocek et al., 2005) were superimposed using the program SUBCOMB of ATSAS package (Petoukhov et al., 2012).

The physiological, forward reaction of P5C reductase was measured at 30°C following the oxidation of NAD(P)H at 340 nm. Under standard conditions, the assay mixture contained 20 mM Tris-HCl buffer, pH 7.0, 0.5 mM NADH or NADPH, and 1 mM l-P5C in a final volume of 0.2 mL. Proteins were column buffer-exchanged just before the analysis to remove Hepes buffer, which was replaced with 10 mM Tris-HCl buffer, pH 7.0. dl-P5C was synthesized by the periodate oxidation of δ-allo-hydroxylysine (Sigma H0377) and purified by cation-exchange chromatography according to Williams and Frank (1975). A limiting amount of enzyme (about 0.1, 2, 0.4, and 10 μg of the purified ProH, ProI, ProG, and ComER, respectively) was added to the pre-warmed mixture, and the decrease in absorbance at 340 nm was recorded at 30-s intervals and up to 25 min through an optical path of 0.5 cm. Unspecific oxidation of pyridine dinucleotides, measured in parallel blanks in which P5C had been omitted, was subtracted, and activity was calculated from the initial linear rate on the assumption of a molar extinction coefficient for NAD(P)H of 6,220 M⁻¹ cm⁻¹. Linear regression analysis was computed with Prism 6 (version 6.03, GraphPad Software, Inc., USA). Protein content was determined by the Coomassie Blue method (Bradford, 1976), using bovine serum albumin as the standard. For the purified proteins, direct absorbance at 280 nm was used instead, and the concentration was calculated by a deduced (http://web.expasy.org/protparam/) molar extinction coefficient of 9,065, 10,555, 14,565, and 10,805 M⁻¹ cm⁻¹ for ProH, ProI, ProG, and ComER, respectively.

To calculate substrate affinity constants and enzyme maximum rates, non-variable substrates were fixed at the same levels as in standard assay. Depending on the protein, the concentration of l-P5C ranged from 40 to 2,000 μM, the concentration of NADH was varied from 50 to 800 μM and that of NADPH ranged from 12 to 400 μM. To evaluate the occurrence of product inhibition, proline and NAD(P)⁺ were added to the standard assay mixture at increasing levels, ranging from 5 to 500 mM (proline), from 0.02 to 12.5 mM (NADP⁺), or from 0.8 to 25 mM (NAD⁺). To assess the effect of ions on either the NADH- or the NADPH-dependent activity, increasing concentrations of NaCl from 10 to 1,000 mM were added to the reaction mixture. In all cases, the resulting activity was expressed as percent of that in untreated controls. All assays were performed in triplicate. K_M and V_max values and the concentrations causing 50% inhibition (IC₅₀) of P5C reductase activity, as well as their confidence limits, were estimated by non-linear regression analysis using Prism 6. Catalytic constants were calculated from V_max values for a single monomer, not taking into account the oligomeric composition of the native holoenzyme.

Bacillus subtilis genome contains four genes that can be predicted by homology as possible P5C reductases. Based on the sequence alignment blueprint (Figure 2) outlined in a previous study (Forlani et al., 2015c), the presence of the two conserved cofactor-binding fingerprint motifs B and C was clearly detectable in all four deduced proteins. The motif B, the so-called Rossman fold hallmark motif with the G-x-x-G-x-G-x-M consensus sequence, is conserved in all representatives except ProH, in which the first glycine is replaced by an alanine residue. This motif identifies the presence of the glycine-rich loop, which interacts with the cofactor phosphate moiety. There are other three residues (V–K–P) that are a part of the motif B, and impact the interaction with the pyrophosphate moiety of the cofactor, even though their position is ~60 residues from the hallmark motif (Forlani et al., 2015c). The middle residue (a lysine) is conserved in all P5C reductases except ProG. The motif C forms a conserved loop that contributes to the formation of the hydrophobic region around the cofactor nicotinamide ring. The sequence motif R-x-M-x-N shows a presence of methionine residue, which in concert with the methionine of motif B, surrounds the cofactor's nicotinamide ring. The exact match is found in ProI and ProH, while ProG and ComER appear to have a variant of that motif with sequence R-x-I-x-A/S, the methionine being replaced by another hydrophobic residue, isoleucine. In contrast to the cofactor binding which is positioned at the N-terminal end, the binding loop for P5C (motif E) with the consensus sequence S-P-G-G-T-T (Nocek et al., 2005; Forlani et al., 2015c; Ruszkowski et al., 2015), is positioned at the C-terminal end. The cavity formed by the residues forming this motif has been predicted to play a critical role in binding of the pyrroline ring. Based on the structural data it was deducted that the presence of the proline and glycine residues in the center allows positioning of the flanking serine and threonine residues such that they could interact with the carboxylate moiety of the substrate. Interestingly, there are irregularities in the sequences of the motif in all four putative P5C reductases. While ProI has an exact matching sequence in the motif region, ComER is on the opposite side of the spectrum with only one threonine present. In ComER, valines replaced two other conserved residues. It may indicate that ComER lost elements that allow substrate binding, or could use an alternative mechanism. This is even clearer while comparing motif D sequences. Motifs D were shown to form hinge loops with the sequence of G-S-x-P-A (Forlani et al., 2015c) and contribute bonding to stabilized motif E. The sequence of this motif in ProG and ComER appears to be modified; especially the second residue [a serine that has been shown to form the critical interactions with the active site (motif E threonine)] is missing in ComER.

The four B. subtilis genes coding putative P5C reductases were cloned into the expression vector pMCSG68 and heterologously expressed in E. coli. The presence of N-terminal His₆-tag and the adoption of a stepwise elution protocol allowed the attainment of substantially homogeneous preparations in a single step, although the heterologous proteins eluted at different imidazole concentrations (data not shown). The presence of the His₆-tag did not have any apparent effect on the enzymatic activity, as similar specific activity values were obtained before and after the cleavage of the purified proteins with TEV protease. Enzyme preparations were moderately stable; if sterilized by filtration (0.22 μm pore size), more than 80% activity was retained after 2 week-storage at 4°C, but < 20% of the initial activity was evident after 1 month.

In all cases the recombinant enzymes showed the capability of catalyzing the P5C-dependent oxidation of NAD(P)H at neutral pH. However, activity levels strongly differed within the group (Table 2), with ProG and ProH showing the typical values observed for P5C reductases, while ProI and ComER produced detectable levels of P5C reduction only when added to the reaction mixture in microgram amounts. Under standard assay conditions and with NADPH as the electron donor, the resulting specific activity values were about 1,500 and 400 nkat mg⁻¹ protein for ProH and ProG, respectively, but only 5 nkat mg⁻¹ protein in the case of ProI, and <1 nkat mg⁻¹ protein for ComER. NADH was found to serve as a substrate as well, but this caused a generalized 50–75% drop of the catalytic rate.

The kinetic characterization of ComER revealed some unusual features for a P5C reductase. The enzyme showed a highest affinity for NADPH, with an apparent K_M value of about 1 μM. In contrast, a lowest affinity was observed for l-P5C, with an estimated half-maximal rate at concentrations as high as 3 mM (Table 2). This corresponds to an average value of ~10⁵ M⁻¹ s⁻¹ of K_cat/K_M with NADPH, but to a strikingly low value with P5C (K_cat/K_M = 29 M⁻¹ s⁻¹). Taking into account that high levels of P5C are cytotoxic (Deuschle et al., 2004) and that under physiological conditions its intracellular concentration ranges from 10 to 20 μM (Giberti et al., 2017), the resulting activity in vivo of ComER would be negligible (~0.01 nkat mg⁻¹ protein). Moreover, with NADPH as the electron donor the activity was unaffected by either high proline concentrations exceeding 500 mM, or NaCl levels as high as 1 M (Table 3). When the pH-activity relationship was tested, ComER showed an uncommon pattern in which the catalytic rate was as higher as lower the pH over the entire range tested (5.8–8.0; Figure 3). Additionally, the SAXS results (Figure 4) were consistent with a molecular mass of 33 kDa, clearly pointing to ComER as a monomer. The large-scale analysis of all P5C reductases characterized to date showed that at least a dimeric organization is required to be fully functional (Forlani et al., 2015c). Therefore, these results strongly suggest that ComER may be instead an NADPH-dependent reductase using in vivo other aldehyde(s) as the substrate.

Concerning ProI, more canonical properties were found. A higher V_max value with NADH than NADPH as the co-factor was paralleled by a 35-fold higher affinity for the latter, resulting under standard assay conditions in a higher specific activity level with NADPH (Table 2). The measurement of the activity as a function of pH showed two optimal values at pH 7.1 and 6.1 (Figure 3), possibly reflecting the intrinsic enzyme property and the equilibrium in solution between P5C and glutamate semialdehyde (which seems to be the actual substrate of P5C-metabolizing enzymes; Arentson et al., 2012; Forlani et al., 2015a), respectively. P5C reduction was inhibited in a concentration-dependent manner by both proline or NaCl (Table 3). However, the activity rate was extremely low, with only 1 catalytic event every 4–7 s (Table 2), and even with the preferred electron donor the corresponding specificity constant was < 3 × 10³ M⁻¹ s⁻¹. If the relatively low affinity constant for P5C (K_M values in the 10⁻⁴ to 10⁻³ M range) is also considered, the contribution of ProI to the overall P5C reductase activity inside the cell may be questionable.

The specific activity values found for the other two heterologously-expressed proteins were 200–2,000-fold higher (Table 2), thereby similar to those previously described for single P5C reductases in other eubacteria (e.g., Petrollino and Forlani, 2012). In both cases a higher catalytic rate under standard assay conditions was evident with NADPH as the electron donor, and the Michaelis constant for NADH was 15–60-fold higher than for NADPH (Table 2). Moreover, the presence of NADP⁺ in the 10⁻⁴–10⁻³ M range progressively inhibited the NADH-dependent activity without substantially affecting that driven by NADPH (Table 3), a result that also in this species strengthens the possibility that NADPH may be the only co-factor in vivo (Ruszkowski et al., 2015).

Interestingly, some relevant differences were found between the two enzymes. Concerning the pH-activity relationship, ProG showed a similar catalytic rate over the whole range tested, whereas for ProH a prominent maximum was evident around pH 6.5 (Figure 3). More remarkably, completely different patterns were found at increasing substrate concentrations. In the case of ProG, saturating conditions were obtained in the 50–500 μM range for both substrates (Figure 5B), with half maximal activity at 70 μM P5C and 240 μM NADPH (Table 2). On the contrary, ProH showed a higher affinity for the pyridine dinucleotide, with more than 80% V_max achieved at concentrations as low as 50 μM, but a much lower affinity for the specific substrate, so as the enzyme did not reach saturation even at millimolar P5C concentrations (Figure 5A). As a consequence, ProH showed a highest theorical specificity constant of 6 × 10⁷ M⁻¹ s⁻¹ vs. the average constant of 8 × 10⁴ M⁻¹ s⁻¹ for ProG. However, at the physiological P5C concentration of 20 μM the activity rate of the two enzymes would be higher for ProG (about 34 and 148 nkat mg⁻¹ protein for ProH and ProG, respectively). Yet, following the increase of the intracellular level of P5C, the activity of ProH would increase much more than that of ProG, with similar values (about 550 nkat mg⁻¹ protein) at 330 μM. With a further increase of P5C level to 500 μM, the activity of ProG would be only slightly enhanced (about 580 nkat mg⁻¹ protein), whereas that of ProH would be further stimulated by 50% (835 nkat mg⁻¹ protein).

Concerning mechanisms for post-translational regulation, both enzymes were subjected to product inhibition, being proline concentrations in the range 10–100 mM able to inhibit their activity (Table 3), with IC₅₀ values slightly but significantly different. A more remarkable difference was found with respect to salt. Not taking into account the NADH-dependent activity, which seems of marginal importance in vivo, a noteworthy dissimilar pattern was found by adding to the assay mixture increasing concentrations of NaCl in a physiological range (10–100 mM). In the case of ProH the addition was substantially ineffective, and at doses exceeding 100 mM enzyme inhibition took place (Figure 6A). On the contrary the activity of ProG was progressively enhanced by up to 35%, and only over 200 mM the presence of salt became inhibitory (Figure 6B). On the whole, all their features would make ProG more suitable for non-stressful conditions, as its activity is not influenced by pH fluctuations, reaches about 25% of maximal level at the physiologically lowest concentrations of P5C, and may be immediately enhanced by an increase of salt level in the cell. On the contrary, ProH would meet cell requirement under stress, since its activity increases almost linearly as a function of P5C concentration allowing the attainment of a higher rate of proline synthesis, is not significantly influenced by salt concentration, and is further stimulated by cytoplasmic acidification.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.