Expanding the substrate scope of a bacterial nucleotidyltransferase via allosteric mutations

Abstract

Bacterial glycoconjugates, such as cell surface polysaccharides and glycoproteins, play important roles in cellular interactions and survival. Enzymes called nucleotidyltransferases use sugar-1-phosphates and nucleotide triphosphates (NTPs) to produce nucleotide diphosphate sugars (NDP-sugars), which serve as building blocks for most glycoconjugates. Research spanning several decades has shown that some bacterial nucleotidyltransferases have broad substrate tolerance and can be exploited to produce a variety of NDP-sugars in vitro. While these enzymes are known to be allosterically regulated by NDP-sugars and their fragments, much work has focused on the effect of active site mutations alone. Here, we show that rational mutations in the allosteric site of the nucleotidyltransferase RmlA leads to expanded substrate tolerance and improvements in catalytic activity that can be explained by subtle changes in quaternary structure and interactions with ligands. These observations will help inform future studies on the directed biosynthesis of diverse bacterial NDP-sugars and downstream glycoconjugates.

Graphical Abstract:

Authors are required to submit a graphic entry for the Table of Contents (TOC) that, in conjunction with the manuscript title, should give the reader a representative idea of one of the following: A key structure, reaction, equation, concept, or theorem, etc., that is discussed in the manuscript. Consult the journal’s Instructions for Authors for TOC graphic specifications.

Across bacteria, an estimated 700 distinct activated monosaccharides are used to build cell surface glycoconjugates, as well as to decorate intracellular proteins and natural products.^1–8 Bacterial cell envelopes are enriched in glycan motifs that are pivotal to survival and infection.⁹ These motifs include the cell wall layer (peptidoglycan) and surface polysaccharides that make-up the exposed glycocalyx layer, such as lipopolysaccharides with attached O-antigens in the outer membrane of Gram-negative bacteria^10–15 (Figure 1A). Glycoconjugates are typically built by dedicated glycosyltransferases that largely use nucleotide diphosphate sugars (NDP-sugars), and to a lesser extent nucleotide monophosphate sugars, as donor substrates to react with biomolecular acceptors. Enzymes called sugar-1-phosphate nucleotidyltransferases couple sugar-1-phosphates to nucleotide triphosphates (NTPs) to create NDP-sugars. Several laboratories have taken advantage of bacterial nucleotidyltransferase activity to produce libraries of NDP-sugars from sugar-1-phosphates as an alternative to chemical activation strategies that require numerous protecting group manipulations and suffer from low yields.^16,17 Past work has highlighted the application of these NDP-sugar pools to natural product and inhibitor synthesis,^18–21 as well as to the characterization of diverse glycosyltransferases.^14,22–26

Figure 1. Bacterial nucleotidyltransferases synthesize NDP-sugars that serve as building blocks for glycoconjugates.

(a) Schematic of a cross-section of a Gram-negative bacterial cell indicating the use of sugar-1-phosphates to make NDP-sugars, which are incorporated into various cellular compartments, as indicated. (b) Structure of a Gram-negative nucleotidyltransferase, RmlA (top), which is regulated at an allosteric site (enlarged) by the final product of the Rml biosynthetic pathway, dTDP-β-l-Rha (bottom) (PDB ID: 1G3L, residues are indicated for P. aeruginosa RmlA, with homologous SALTY RmlA residues in parentheses).

In particular, the sugar-1-phosphate nucleotidyltransferase RmlA from the Gram-negative bacterium Salmonella enterica LT2 (SALTY) (Figure 1B, top) is known to exhibit broad substrate promiscuity to produce various NDP-sugars.^27,28 In many bacteria, RmlA catalyzes the coupling of glucose-α-1-phosphate (Glc-α-1-P) and deoxythymidine triphosphate (dTTP) to form dTDP-α-glucose (dTDP-α-Glc) and pyrophosphate in the first of four steps to synthesize the activated prokaryote-specific 6-deoxy-L-sugar dTDP-β-l-Rhamnose (dTDP-β-l-Rha) (Figure 1B, bottom).^29–31 dTDP-β-l-Rha then serves as the precursor for l-Rha incorporation into various bacterial glycoconjugates. Wild-type SALTY RmlA accepts diverse α-d-hexose-1-phosphates, including all monodeoxy α-d-Glc-1-phosphates, to produce various dTDP-sugars.²⁸ Elegant work by Thorson and coworkers demonstrated that rational and evolved mutations in SALTY RmlA could provide access to an even broader pool of NDP-sugars, including ADP- and GDP-sugars and dTDP-unnatural sugars.^32–36 Others have expanded this nucleotide sugar pool through the use of RmlA homologs that also demonstrate a lack of substrate specificity.^2,37–39 While diverse nucleotide sugars have been produced through chemoenzymatic routes,⁴⁰ there have been limitations in the efficiency of a single enzyme to produce different NDP-sugars, especially with non-canonical substrates.^25,26

One key limitation to the catalytic efficiency of RmlA is product inhibition by dTDP-α-Glc or negative feedback regulation by dTDP-β-l-Rha via allosteric site binding (Figure 1B).^41–44 Other bacterial nucleotidyltransferases are also regulated via allosteric mechanisms by pathway intermediates.^41,45–48 While residues around the active site have been varied in studies aimed at expanding the substrate promiscuity of RmlA,^32–35,49 mutagenesis analysis of allosteric site residues has yet to be performed. We hypothesized that manipulation of RmlA regulation would improve nucleotidyl transfer activity and provide a general route to the production of a variety of NDP-sugars, consisting of both α-d- and β-l-sugars, since these molecules can act as inhibitors. Here we show that by mitigating allosteric inhibition through rational mutations, SALTY RmlA can perform nucleotidyl transfer with higher yields than the corresponding wild-type enzymes, and offer access to additional NDP-sugars, due to small conformational changes in the overall structure relative to wild-type protein. These studies offer improved tools to produce NDP-sugars in one step for studies on glycosyltransferases, and provide greater insight into engineering allosterically regulated bacterial nucleotidyltransferases.

We sought to evaluate the behavior of RmlA from relevant Gram-negatives (SALTY, Pseudomonas aeruginosa, Escherichia coli), which are of clinical interest.^32,42,44,50 Initial work was performed on SALTY RmlA, since it has been the target of past engineering efforts.^32–36 During the optimization of recombinant RmlA overexpression in E. coli, we found that SALTY RmlA underwent in vivo cleavage events as part of several fusion constructs with either an N-terminal SUMO tag, or an N- or C-terminal His₆ tag (Figure S1A). Edman sequencing of purified truncations revealed that proteolytic cleavage occurred between D104 and C105 (Figure S1B–C). Mutation of D104 to Asn or Ala resulted in greater yields of full-length protein (Figure S1D). Since D104N showed initial reaction rates similar to wild-type (Figure S1E), this mutation was preserved in subsequent experiments.

Solved crystal structures of RmlA proteins have revealed three functional subdomains: a core subdomain, sugar-binding subdomain, and a dimerization subdomain (Figure 1B, top left).^42,44,50 The core subdomain and sugar-binding subdomain shapes the main active site.⁴⁴ The dimerization subdomain of RmlA assists in the assembly of RmlA homotetramers and helps to shape a distinct allosteric site that can bind to dTDP-β-l-Rha, and other thymidine-containing compounds.^43,44,51 We aimed to introduce mutations into the allosteric site of SALTY RmlA to abrogate ligand binding and to analyze the resulting effects on activity. While inhibition of SALTY RmlA by allosteric ligands is not well-characterized, there are several solved structures of P. aeruginosa RmlA with bound allosteric inhibitors.^43,44,51 SALTY RmlA and P. aeruginosa RmlA possess high sequence identities (Figure S2) (~74 %), and structural alignment (Figure S3) of SALTY RmlA and P. aeruginosa RmlA co-complexed to dTDP-β-l-Rha indicated several conserved residues that make contacts with the allosteric ligand (Figure 1B, top right). Based on inspection of the overlaid structures, we made several mutations within the allosteric site of SALTY RmlA. The mutations E256D and H117A were each introduced to examine the effect of weakened H-bonds with the ligand; S252F was installed to prevent ligand binding through the introduction of bulky side chains; finally R220A was added to remove H-bonds between RmlA monomers at the protein-protein interface that occurs at the allosteric site.

To study the effect of these mutations on RmlA function, we reconstituted nucleotidyltransferase activity using the native substrates α-Glc-1-phosphate (α-Glc-1-P) and dTTP, and analyzed the reaction by HPLC and mass spectrometry (Figure 2A). Initial rates were measured using a coupled assay with inorganic pyrophosphatase and malachite green that detects the release of Pi, which is produced from the byproduct PPi.⁵² Initial rate analysis indicated that the RmlA D104N mutation in combination with H117A, Y146F, R220A, or E256D all show similar activity to wild-type (Figure S4). Only S252F showed ~50% decrease in initial rate relative to wild-type, suggesting that the introduction of a bulky aromatic residue at this position led to structural changes detrimental to turnover at the active site.

Figure 2. Mutations in the allosteric site of RmlA prevent modulation by a known inhibitor.

(a) Reaction scheme for native RmlA reaction (top) with HPLC and mass spectrometry (MS) analysis shown (bottom) confirm production of dTDP-α-Glc. ([M-H]⁻_calcd: 563.0685) (b) Structure of P. aeruginosa RmlA bound to the allosteric inhibitor 1 showing all residues within 5 Å (PDB ID: 4ASJ, homologous SALTY residues in parentheses for context). (c) Percent inhibition analysis of wild-type and mutant RmlA activity by 1 (thymine portion in red) shows reduced sensitivity of some allosteric mutants. Error bars represent standard deviation (SD, n=3).

Past work has shown that derivatized thymine-based molecules can inhibit P. aeruginosa RmlA via allosteric site binding.^43,44,51 We synthesized the most potent small molecule inhibitor shown to complex RmlA from an initial report⁴³ (Figure 2B), and tested its activity against wild-type and mutant RmlA constructs (Figure 2C). Surprisingly, D104N showed decreased sensitivity to 1, even though the mutation is distant from the allosteric and active sites (Figure S1C). As predicted by comparison of the structures of P. aeruginosa and SALTY RmlA (Figures 1B, top, 2B), mutation of residues S252 and E256, which contact the thymine ring led to reduced sensitivity to 1. On the other hand, the R220A mutant showed similar sensitivity to 1 as H117A. This observation confirms R220 is not as important for interactions with 1, as indicated by Figure 2B, even though it is predicted to contact the thymine ring of dTDP-β-l-Rha (Figure 1B). For comparison, when the competitive inhibitor dTDP-α-Glc^41,42 was added in excess to RmlA reactions, the inhibitory activity was similar in both wild-type and mutant reactions since binding occurs at the active site (Figure S5). Collectively, these data verified that mutation of RmlA at the allosteric site does not halt catalytic turnover, but can prevent modulation by allosteric ligands.

As RmlA has been shown to demonstrate enzymatic promiscuity,^27,28 we next examined the role of allosteric mutations on the substrate selectivity of RmlA at both the sugar-1-P and nucleotide recognition pockets. We first explored substrate tolerance at the sugar-1-P site, while using the native dNTP, dTTP (Figure 3A, S6). Wild-type and mutant RmlA proteins were able to activate α-Gal-1-P, α-GlcNAc-1-P, and α-Man-1-P to their respective dTDP-sugars. Allosteric site mutants of RmlA, but not wild-type, showed nearly full activation of α-Man-1-P at t = 6 hr, indicating mutants could better tolerate inverted stereochemistry at the C2 position relative to that of α-Glc-1-P (Figure 3B, blue circles). Of the allosteric site mutants, D104N/H117A, D104N/S252F and D104N/R220A showed the greatest conversion of α-Gal-1-P (~ 50%, t = 6 hr), demonstrating that these mutations expanded tolerance for altered stereochemistry at the C4 position compared to wild-type (Figure 3B, red circles). D104N/S252F and D104N/R220A showed full activation of α-GlcNAc-1-P (t = 6 hr), showing enlarged accommodation of a bulky group at C2 (Figure 3B, blue circles). Hence, modification of the allosteric site can alter recognition at the sugar binding domain.

Figure 3. Allosteric site modification of RmlA leads to broadened substrate tolerance.

(a) Analysis of percent dTDP-sugar produced by indicated RmlA constructs with different sugar-1-Ps and dTTP (t = 6 hr). (b) Chemical structures of sugar-1-Ps tested in part (a) with relevant differences highlighted. (c) HPLC and mass spectrometry analysis of RmlA D104N/Y146F/E256D reaction indicates production of dTDP-β-l-Fuc ([M-H]⁻calc_d: 547.0736). (d) Analysis of percent NDP-Glc produced by indicated RmlA constructs with Glc-1-P and different NTPs (t = 6 hr). The triple mutant shows the broadest access to dNDP/NDP-sugars. Inset: chemical structures of C and T bases. Error bars represent SD (n=3).

While activation of α-d-sugar-1-P substrates by nucleotidyltransferases has been thoroughly explored, we were also interested in one-step activation of β-l-sugar-1-Ps, since the biosynthesis of these NDP-sugars typically occurs over several enzymatic steps.^{10–12,14,29,30} Past directed evolution studies have shown that introduction of the mutation(s) Y146F +/− W224F facilitates usage of β-l-Fucose-1-P (β-l-Fuc-1-P) by SALTY RmlA.³⁴ We anticipated that addition of an active site mutation would likely expand the substrate promiscuity exhibited by our allosteric site mutants, which alone showed poor activation of β-l-Fuc-1-P (Figure 3A). We compared the activity of RmlA Y146F versus W224F and found that the former showed greater percent activation of β-l-fuc-1-P by dTTP (Figure S7), as previously reported;³⁴ hence, this mutant was used in follow-up experiments.

Since dTDP-β-l-Fuc resembles the allosteric inhibitor dTDP-β-l-Rha (compare Figures 1B and 3B), we sought to introduce Y146F into an RmlA mutant background that was unaffected by allosteric inhibition. Of the mutants tested, RmlA D104N/E256D showed the lowest sensitivity to 1 (Figure 2C) while maintaining near wild-type initial rates (Figure S4), and so we combined the Y146F mutation with D104N/E256D. While the D104N/Y146F mutant was able to provide ~20% dTDP-β-l-Fuc (t = 6 hr), addition of the E256D mutation showed a 2-fold increase in the activation of β-l-Fuc-1-P relative to the double mutant (Figure 3A, compare purple and green bars; Table S4, rows 3 and 8, respectively), as validated by HPLC/mass spectrometry (Figure 3C). Notably, the triple mutant also showed low levels of inhibition by the allosteric inhibitor 1 (Figure 2C). To directly assess if our allosteric mutants were less vulnerable to inhibition by dTDP-β-l-sugar product, we purified milligram-quantities of dTDP-β-l-Fuc to add to RmlA reactions containing native substrates. We found that excess dTDP-β-l-Fuc (400 μM) leads to about ~40% inhibition of D104N reactions and ~15% inhibition of D104N/Y146F reactions; however, mutants carrying the E256D mutation are insensitive to inhibition by dTDP-β-l-Fuc (Figure S8), verifying that allosteric mutation can mitigate modulation by NDP-β-l-sugar products.

Different dNTPs and NTPs were next tested as substrates with α-Glc-1-P to explore selectivity at the nucleotide binding pocket (Figure 3D, S9, Table S5). As expected based on the structural similarity of UTP and dTTP, all RmlA constructs showed nearly full activation using either substrate, which has been commented on in previous reports on RmlA homologs.^28,41,44 However, wild-type RmlA and several mutants showed decreased activation of α-Glc-1-P with dCTP, indicating that the conversion of a carbonyl to an amine at the C4-position of the pyrimidine likely leads to loss of H-bonding interactions within the active site (Figure 3D, inset). However, some allosteric mutants can overcome this issue to utilize dCTP as a substrate. Mutation of H117 appears to weaken recognition requirements in the nucleotide-binding pocket, leading to production of dCDP-α-Glc, and the purine nucleotide-sugars dADP-α-Glc and dGDP-α-Glc. Similar to the observations for sugar-1-P substrate usage, the broadest tolerance of all dNTP/NTP-α-Glc combinations was offered by the triple mutant (D104N/Y146F/E256D). Access to GDP-α-Glc was only possible with constructs carrying the Y146F mutation, which led to 3- to 4-fold higher yields than that of the H117A mutant.

Since Man is often activated as GDP-α-Man in bacterial pathways that lead to various GDP-sugars,^14,20,53,54 we also examined whether RmlA mutants could activate α-Man-1-P using GTP as a nucleotide substrate; however, none of our tested constructs were able to produce GDP-α-Man, even after extended time periods. Nonetheless, these results suggest that even a single allosteric mutation offers access to NDP products that could only be accessed by active site mutations based on past reports.^32–35,49

To better understand why mutation at the allosteric site led to expanded substrate tolerance in both the sugar-1-P and nucleotide pockets in SALTY RmlA, we performed Michaelis-Menten kinetic analysis on select mutants. Inspection of the promiscuity data above indicated that the D104N/Y146F/E256D triple mutant was able to produce the widest variety of dNTP/NTP-sugars. When compared to the D104N background, triple mutation led to a decrease in the K_M for both α-Glc-1-P and dTTP, while the k_cat was similar to that of D104N for each substrate, leading to enhanced specificity constants (k_cat/K_M) (Table 1, Figure S10). Examination of RmlA D104N with individual mutations in the active (Y146F) or allosteric (E256D) sites shows that each contributes to changes in the kinetic parameters observed in the triple mutant. Active site mutation led to a lowered K_M for Glc-1-P compared to the D104N background (nearly 3-fold), while the allosteric site mutation led to a reduction in the K_M for both Glc-1-P and dTTP relative to D104N (nearly 1.5-fold for each substrate). Hence, mutation of the allosteric site leads to changes in recognition at both substrate pockets, as reflected in the broad substrate tolerance observed for these mutants (Figure 3). In contrast to our observations, others have noted decreased specificity constants for native substrates in RmlA mutants that confer broadened nucleotide recognition;³³ however analysis is limited across published RmlA mutants using α-Glc-1-P and dTTP as substrates.

Table 1.

Kinetic parameters of indicated substrates with select RmlA mutants.^a

Substrates	Vmax	K m	k cat	Kcat/Km
D104N	μM/min	μM	min−1	min−1μM−1
dTTP	5.47 (0.10)	41.9 (3.0)	438(8)	10.4
Glc-1-P	5.89 (0.15)	57.2 (5.1)	471 (12)	8.24
D104N/Y146F
dTTP	4.14 (0.12)	41.4 (4.3)	332 (9)	8.00
Glc-1-P	4.08 (0.15)	20.2 (3.3)	327 (12)	16.2
β-Fuc-1-P	1.57 (0.12)	1445 (371)	0.314 (0.024)	2.17×10−4
D104N/E256D
dTTP	4.17 (0.15)	30.8 (4.5)	333 (12)	10.8
Glc-1-P	4.74 (0.15)	38.4 (4.9)	379 (12)	9.88
D104N/Y146F/E256D
dTTP	6.00 (0.16)	33.1 (3.5)	480 (12)	14.5
Glc-1-P	5.22 (0.13)	29.0 (3)	417 (10)	14.4
β-Fuc-1-P	2.44 (0.11)	1332 (207)	0.488 (0.023)	3.66×10–4

^aFor kinetic analysis of dTTP, RmlA mutants were incubated with 100 μM Glc-1-P and 0–800 μM dTTP in 35 mM Tris-HCl (pH 7.5), 2.2 mM MgCl₂ at 25 °C. For kinetic analysis of Glc-1-P, RmlA mutants (12.5 nM) were incubated with 100 μM dTTP with 0–800 μM of Glc-1-P under similar conditions. For β-L-Fuc-1-P analysis, RmlA mutants (5 μM were incubated with 2 mM dTTP with 0–10 mM of β—L-Fuc1-P under similar conditions Standard errors (SE, n = 8–9 are shown in parentheses.

For the non-canonical sugar-1-P substrate, β-l-Fuc-1-P, past pseudo-first order kinetic analysis of the RmlA mutant Y146F/W224F showed that active site mutations did not alter the K_M, but did lead to a dramatic increase in the k_cat relative to wild-type.³⁴ Changes in k_cat leading to alteration of substrate specificity have also been noted by others.⁵⁵ Similarly, we found that addition of the E256D mutation to the D104N/Y146F background also led to an increase in the k_cat without a significant change in the K_M (Table 1), which has been noted to reflect enhanced flexibility in the sugar-1-P pocket to promote an induced fit mechanism.³⁴

The allosteric pocket of each RmlA monomer borders the dimerization interface (Figure 4A),^32,44 hinting that allosteric site mutations may lead to changes in quaternary structure that affect activity. Since the RmlA triple mutant D104N/Y146F/E256D showed the greatest promiscuity across all substrates, we evaluated its overall structure in comparison to the D104N background. Size-exclusion chromatography with multi-angle light scattering (SEC/MALS) analysis confirmed that both D104N and D104N/Y146F/E256D formed approximate tetramers in solution (Figure S11, observed: ~123 kD and 130 kD, respectively; calculated ~133 kD), similar to previous structural studies on RmlA homologs.^32,42,44,50 To examine the importance of the tetrameric state of RmlA for activity, we removed the C-terminal a-helix, which is predicted to be involved in oligomerization at the allosteric site (Figure 1B, C-terminus highlighted in pink). The resulting truncated RmlA (lacking residues N279-L292) was analyzed by SEC and compared to wild-type. As opposed to the single tetrameric peak observed for wild-type, we identified two peaks for the C-terminal truncation, corresponding to an approximate trimer and dimer (Figure 4B, S12A), which was confirmed by native-PAGE analysis of each peak (Figure S12B). Neither the dimer nor the trimer state of truncated RmlA was catalytically active (Figure 4C). These data indicate that the C-terminal α-helix is vital in mediating formation of a tetrameric state of RmlA, which is essential for enzyme activity.

Figure 4. A combination of active and allosteric site mutations in RmlA causes changes in quaternary structure.

(a) Quaternary structure of SALTY RmlA bound to UDP-Glc with relevant residues highlighted (PDB ID: 1IIN). (b) Size exclusion analysis of full-length SALTY RmlA (WT) and C-terminal truncated RmlA (ΔN279-L292) indicates that tetrameric structure is lost upon truncation. (c) Activity analysis of fractions indicated in part B with native substrates demonstrates that only tetrameric RmlA (WT) is active. (d) SDS-PAGE followed by Coomassie staining of crosslinking reactions of indicated RmlA mutants using increasing concentrations of glutaraldehyde (left) and quantification of resulting bands (right) demonstrates that additional mutations promote oligomer formation (n=3). M = monomer; D = dimer; T= tetramer. (e) SDS-PAGE analysis followed by fluorescence gel imaging (ex: 493 nm; em: 516 nm) after maleimide-Alexa FL 488 labeling of each RmlA construct from part D carrying an additional S85C mutation in the presence of increasing dTTP (left) and quantification of resulting bands (right) suggests that the nucleotide-binding loop is more flexible in the D104N/Y146F/E256D background (n = 4). Note that for parts D and E, replicate gels are shown in the SI.

Protein crosslinking experiments were then performed to evaluate the propensity of different oligomeric states to form before and after mutation of RmlA. RmlA D104N or D104N/Y146F/E256D were each treated with varying concentrations of glutaraldehyde to promote covalent crosslinking over a short time period followed by SDS-PAGE analysis.⁵⁶ We found that dimeric, trimeric and tetrameric states of RmlA D104N/Y146F/E256D were crosslinked at lower concentrations of glutaraldehyde than required for the D104N background (Figure 4D, S13). These data suggest that higher ordered states of RmlA form more readily upon mutation of Y146 and E256.

Crystal structures of SALTY RmlA point mutants that confer expanded tolerance of nucleotides has indicated enhanced flexibility in a single loop (Q83-L89) within the NTP recognition domain compared to that of the wild-type structure (Figure S14A).³⁴ In particular, Ser85 exhibits different rotamer conformations when bound to nucleotides in promiscuous RmlA mutants. We postulated that allosteric site mutation might lead to similar structural flexibility within this site, since many of our mutants showed expanded nucleotide tolerance. Ser85 was mutated to Cys in RmlA D104N and D104N/E256D/Y146F to facilitate comparison of loop flexibility via protein labeling analysis using a fluorescent maleimide probe. While native RmlA has two Cys residues, we established Cys-labeling conditions that could differentiate between wild-type and S85C mutants by SDS-PAGE analysis (Figure S14B–C). Titration of increasing concentrations of dTTP to RmlA D104N/S85C and D104N/E256D/Y146F/S85C followed by labeling shows increased fluorescence in the latter when compared to a BSA standard (Figure 4E, S15). These results indicate there is greater structural mobility at the nucleotide binding site upon mutation of distal residues in the sugar-1-P (Y146) and allosteric (E256) sites (see Figures 1B and 4A).

Glu256 is a conserved allosteric residue across various bacterial RmlA homologs (Figure S2). To evaluate if our observations on the effect of combined active/allosteric site mutations extended to the behavior of other Gram-negative RmlA proteins, we introduced mutations homologous to Y146 and/or E256 into P. aeruginosa and E. coli RmlA and tested their activity with different NTP and sugar-1-P substrates (Figure S16). We found that introduction of the allosteric site mutation E255D in P. aeruginosa RmlA did not expand substrate tolerance beyond that of the Y145F background (FigureS16A–B). E. coli RmlA wild-type showed the broadest substrate usage when compared to all tested mutants (Figure S16C–D). These data suggest that SALTY RmlA shows different degrees of allosteric communication than that of P. aeruginosa and E. coli. Importantly, all tested wild-type proteins exhibit substrate promiscuity, as noted with other homologs.^38,41

In conclusion, we have shown that allosteric mutation of a Gram-negative RmlA can expand substrate promiscuity observed with and without active site mutation. The active site of each RmlA monomer lies at the protein-protein interface between two dimers, such that these sites are believed to participate in cross-talk (Figures 1B, 4A).⁴⁴ The allosteric site lies at the interface between two monomers to form dimers. Accordingly, we found that subtle structural modifications in these sites, namely conversion of Tyr146 to a Phe residue and Glu256 to an Asp residue, leads to changes in the formation of oligomer states, which has not been reported in previous RmlA engineering efforts. These changes are accompanied by improvements in specific activity and alteration of sugar-1-P and nucleotide substrate specificity at the active site, which offers a range of NDP-sugar products from a single enzyme. As we hypothesized, allosteric mutation leads to reduced inhibition by a dTDP-β-l-sugar, which helps explain improvements in β-l-sugar-1-phosphate activation.

While others have reported that mutations that promote desired enzyme function often occur near the active site,^34,57–59 this work provides examples of productive mutations rationally selected from an allosteric ligand binding site. In line with our observations, small molecule metabolite pathways are rich in examples of enzymes that show altered substrate recognition due to allosteric ligand binding.^60–62 Additionally, some kinases and proteases show different substrate preferences after modification of a distal site.^63,64 Since we found that homologous allosteric modification does not produce identical results in all Gram-negative RmlA proteins, it would be interesting to see how similar changes affect other bacterial nucleotidyltransferases regulated through feedback inhibition.^{41,42,45–48} The findings described here will help inform future studies aimed at modulating cellular NDP-sugar pathways, perhaps through the use of nucleotidyltransferase mutant strains for the discovery of other allosteric inhibitors of bacterial glycoconjugate precursors.

METHODS

General Information.

Primers were purchased from Invitrogen and sequencing was done by Genewiz. All chemicals were purchased from Sigma Aldrich, Alfa Aesar, or Fisher Scientific without further purification. FPLC purification and analysis was performed using an AKTA pure 15 L instrument (UNICORN^™ software, GE Healthcare). ¹H NMR spectra was recorded on Burker Avance III 400MHz, calibrated using residual non-deuterated solvent as internal reference, processed by MestReNova (authorized to NYU). HRMS analyses were acquired on an Agilent 6224 Accurate-Mass time-of-flight LC/MS (LC-TOF) spectrometer with an electrospray (ESI) ionization source equipped with an autosampler. Molecular graphics and analyses were performed with PyMOL 2.5. GraphPad Prism software 9.0 was used for data analysis and plotting. pET His6 Sumo TEV LIC cloning vector (2S-T) was a gift from Scott Gradia (Addgene plasmid # 29711; http://n2t.net/addgene:29711; RRID:Addgene_29711).

Analysis of wild-type and mutant RmlA reaction rates +/− inhibitor using a coupled assay.

Rates of nucleotidyltransferase activity were measured using a reported coupled malachite green assay.52 Reactions were conducted with 12.5 nM wild-type or mutant RmlA and 0.008 mg/mL inorganic pyrophosphatase (PPiase) in buffer (35 mM Tris-HCl, pH 7.5, 2.2 mM MgCl₂)³⁸ with 100 μM of dTTP and α-d-glucose-1-phosphate at 25 °C in a final volume of 60 μL. In a typical reaction, wild-type or mutant RmlA and PPiase were preincubated at 25 °C for 5 min before the initiation of the enzymatic reaction with the addition of 100 μM substrates. At t = 6 min post-initiation of each reaction, 60 μL of 0.05% formic acid was added to quench the reaction. In each set of reactions, a blank reaction that only contained 100 μM each of dTTP and α-d-glucose-1-phosphate in reaction buffer was prepared as a “blank” control and was treated the same as the reaction described above. Reactions were analyzed using malachite green as described below. Relative percent activity was calculated by normalizing mutant RmlA activity to wild-type activity, which was set to 100% activity.

For reactions containing allosteric inhibitor, 100 μM of 1 or 400 μM of dTDP-β-L-fucose was added to 12.5 nM wild-type or mutant RmlA and 0.008 mg/mL PPiase in buffer (35 mM Tris-HCl pH 7.5, 2.2 mM MgCl₂), which were pre-incubated at 25 °C (t = 5 min). Reactions were then initiated by addition of 100 μM of substrates, yielding final reaction volume of 30 μL. At t = 6 min post-initiation of reaction, 30 μL of 0.05% formic acid was added to quench the reaction. In each set of reactions, two “blank” reactions were prepared as controls: one reaction contained 100 μM of dTTP and α-d-glucose-1-phosphate in reaction buffer and a second reaction contained 100 μM of dTTP and α-d-glucose-1-phosphate and 100 μM of 1 or 400 μM of dTDP-β-l-fucose. Reactions were analyzed using malachite green as described below. Relative percent inhibition was calculated by normalizing each wild-type and mutant RmlA reaction plus 1 or dTDP-β-l-fucose to the corresponding percent conversion without addition of inhibitors, which was set to 100% activity.

For the coupled malachite green colorimetric assay, a calibration curve was first generated via serial dilution of monophosphate (1 mM of monophosphate (Pi) to 0.976 μM of Pi) using the reaction buffer mentioned above (wells A1-A11); 50 μL of reaction buffer was then added to well A12. Each quenched reaction (50 μL each) was then added to a 96 well clear plate (Greiner) with 50 μL of premixed malachite green reagents. The premixed malachite green reagent contained 0.0812% malachite green, 2.32% polyvinyl alcohol, 5.72% ammonium molybdate in 6N HCl, and water (2:1:1:2, volume ratio respectively). Reagents were preincubated at 25°C for t = 10–15 min until the color appeared gold/yellow prior to addition to reactions. After addition of reagents, the plate was incubated at 25 °C for t = 30 min and subsequently analyzed using a plate reader (SpectraMax iD5) by measuring absorbance at 630 nm. Data was exported from SoftMax Pro 7.1 software.

The concentration of PPi was calculated using a monophosphate standard curve based on measurements made on the same plate for each experiment, and graphed in GraphPad Prism 9 (Figure S10A). To generate the standard curve, absorbance of water was first subtracted from the obtained absorbance of each well (wells A1-A12), followed by normalizing the absorbance value to buffer (well A12). The resulting delta absorbance was plotted against corresponding concentration of Pi to generate calibration curve. A linear fit was often found between 0–62.5 μΜ Pi. To calculate the concentration of PPi produced in each reaction, absorbance of water was first subtracted from each reaction absorbance. The delta absorbance was then used to calculate PPi produced using the standard curve equation. Appropriate blanks were then subtracted from the obtained PPi (μM) to normalize absorbances based on background PPi production by PPiase.

Analysis of RmlA wild-type and mutant nucleotidyltransferase activity by analytical HPLC.

For analysis of different sugar-1-phosphates, reactions were carried out in a final volume of 60 μL containing 5 μM of wild-type or mutant RmlA, 0.008 mg/mL PPiase, 2 mM dTTP, and 2 mM sugar-1-phosphate in buffer (100 mM 3-(N-morpholino)propanesulfonic acid (MOPS) pH 7.5, and 7.5mM MgCl₂) based on conditions from a previous report.³⁴ For analysis of different nucleotide triphosphates (NTPs), reactions were carried out under same reaction conditions mentioned above with 2 mM each α-d-glucose-1-phosphate and NTPs. Reactions were incubated at 37 °C for 6 hr, and quenched by adding 60 μL of 0.05% formic acid. Quenched reactions were centrifuged using a Beckman Coulter Microfuge 20R at 15,493 × g for 2 min and 50 μL of supernatant was loaded to a 96-well non-coated polypropylene microplate (Thermo Scientific). Analytical HPLC analysis was performed on a Thermo Scientific Dionex UltiMate 3000 UHPLC+ system equipped with an auto sampler through a Phenomenex 5 μm, 4.6 × 150 mm, NX-C18, 110 Å Gemini column. Each sample (15 uL) was injected and eluted using 50 mM TEAB buffer (Buffer A) and acetonitrile (Buffer B) using a linear gradient of 0–5% Buffer B or 0–10% Buffer B at a flow rate of 1.0 mL/min for t = 16 min using 254 nm wavelength for detection. To obtain the percent NDP-sugar, the area of peak representing NDP-sugar was normalized to the total area of peaks containing nucleotides (NTP, NDP, NMP, and NDP-sugar) using Chromeleon software (Thermo Scientific). For E.coli and P. aeruginosa RmlA, analysis of reactions with NTPs and sugar-1-phosphates were carried out under similar conditions to those described above.

Kinetic analysis of wild-type and mutant RmlA constructs.

Kinetic analysis of RmlA constructs were performed using the colorimetric assay described above. Steady-state kinetic parameters of natural substrates were obtained by fixing either glucose-1-phosphate or dTTP at a saturating concentration of 100 μΜ and titrating the substrate of interest. For each substrate of interest, eight concentrations (0–800 μΜ) were assayed in three separate experiments, with each containing three replicas. For each reaction, 12.5 nM of RmlA enzyme and 0.008 mg/mL PPiase was pre-incubated at 25 °C (t = 5 min) in buffer (35 mM Tris-HCl pH 7.5, 2.2 mM MgCl₂). Reaction was subsequently initiated by adding corresponding substrates (reaction volume 50 μL), incubated at 25°C (t = 6 min), and quenched using 50 μL of 0.05% formic acid.

Steady-state kinetic parameters of β-l-fucose-1-phosphate were obtained by fixing dTTP at a saturating concentration of 2 mΜ and titrating β-l-fucose-1-phosphate at eight concentrations (0–10 mM) under conditions similar to those described above. For each reaction, 5 μM of enzyme and 0.008 mg/mL PPiase was incubated with corresponding substrates at 37 °C (t = 180 min) in buffer (100 mM MOPS pH 7.5, 7.5 mM MgCl₂). Reaction was quenched using 50 μL of 0.05% formic acid. Reaction was diluted in 20-fold with water prior to analysis using the coupled colorimetric assay.

Kinetic parameters of RmlA constructs with various substrates of interest were obtained using Prism with the following equation (“determine k_cat”): Y = Et * k_cat * X / (K_M + X) (X is the substrate concentration in μM; Y is the enzyme velocity in μM/min; Et is enzyme concentration in μM). A representative Michaelis-Menten kinetic curve is shown (Figure S10B–C).

Glutaraldehyde-mediated crosslinking of RmlA proteins to assess oligomer state.

Crosslinking of RmlA proteins with glutaraldehyde was carried out by following a published protocol with modifications.⁵⁶ RmlA proteins were first diluted to 2 mg/mL in 0.1X PBS (pH 7.4) buffer and pre-incubated (37 °C, t = 10min) with various concentrations of glutaraldehyde (0.0001%, 0.0005% 0.001%, 0.0025%, 0.005%) in final reaction volume of 20 μL. Reaction was quenched by adding 8 μL of 1M Tris-HCl (pH 8.0), followed by addition of 12 μL of 2X Laemmli sample buffer (Cold Spring Harbor protocols). Samples were loaded at a volume equivalent of 10 μg each and resolved by SDS-PAGE (9% acrylamide gel) at 200 V for t = 43 min followed by standard Coomassie staining/de-staining protocol. Image was recorded using BioRad ChemiDoc Imaging System, the band percentage was analyzed using Image Lab (BioRad) by normalizing to total band percentage for each lane.

Labeling and analysis Cys residues on RmlA mutant proteins to assess loop flexibility.

Labeling of Cys residues on RmlA using Alexa Fluor^™ 488 C₅ Maleimide (Thermo Fisher) was carried out by following the manufacturer’s protocol with minor changes. BSA, chosen as loading control since it contains 35 Cys residues, was diluted to 20 μM and incubated with 100 μΜ of Alexa Fluor^™ 488 C₅ Maleimide at 25°C for t = 15 min. RmlA proteins were diluted to 10 μM in 1X PBS buffer and incubated with Alexa Fluor^™ 488 C₅ Maleimide (5μΜ or 50μM) at 25°C for t = 15 min. Reactions were quenched by adding 2X Laemmli sample buffer. 7.5 μL of BSA, used as a loading control, was added to 30 μL of each quenched reaction, and 25 μL of each mixed sample was loaded and resolved on 12% acrylamide SDS-PAGE gel for t = 43 min under constant voltage (200 V). The gel was gently washed with ddH₂O 10 minutes, and then imaged using an Amersham Typhoon scanner (GE healthcare) equipped with a laser line at 488 nm (excitation at 493 nm and emission at 516 nm). Image was exported, the band percentage was obtained for each lane using Image Lab (BioRad) by comparison to the labeled BSA band.