Active site flexibility revealed in crystal structures of Parabacteroides merdae β‐glucuronidase from the human gut microbiome

Abstract

β‐Glucuronidase (GUS) enzymes in the gastrointestinal tract are involved in maintaining mammalian‐microbial symbiosis and can play key roles in drug efficacy and toxicity. Parabacteroides merdae GUS was identified as an abundant mini‐Loop 2 (mL2) type GUS enzyme in the Human Microbiome Project gut metagenomic database. Here, we report the crystal structure of P. merdae GUS and highlight the differences between this enzyme and extant structures of gut microbial GUS proteins. We find that P. merdae GUS exhibits a distinct tetrameric quaternary structure and that the mL2 motif traces a unique path within the active site, which also includes two arginines distinctive to this GUS. We observe two states of the P. merdae GUS active site; a loop repositions itself by more than 50 Å to place a functionally‐relevant residue into the enzyme's catalytic site. Finally, we find that P. merdae GUS is able to bind to homo and heteropolymers of the polysaccharide alginic acid. Together, these data broaden our understanding of the structural and functional diversity in the GUS family of enzymes present in the human gut microbiome and point to specialization as an important feature of microbial GUS orthologs.

Short abstract

PDB Code(s): http://firstglance.jmol.org/fg.htm?mol=6D7J, http://firstglance.jmol.org/fg.htm?mol=6DXU

Introduction

Advancing our molecular understanding of how the trillions of microorganisms that inhabit the human gastrointestinal tract (the GI microbiota) are able to metabolize ingested chemicals (xenobiotics) and endogenous compounds (endobiotics) will have important implications for human physiology and disease. The gut microbiome encodes several types of enzymes that metabolize xenobiotics and have been demonstrated to impact health outcomes.1, 2, 3, 4 GI microbial β‐glucuronidase (GUS) enzymes can reactivate endo and xenobiotic compounds previously inactivated by glucuronidation via phase II drug metabolism.5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 Glucuronidation involves the conjugation of glucuronic acid (GlcA) to xenobiotic and endobiotic compounds via the actions of uridine 5′diphospho‐glucuronosyltransferase enzymes, thus inactivating molecules and marking them for excretion into the GI tract.16, 17, 18, 19, 20, 21, 22 In the gut, microbial GUS enzymes hydrolyze the glucuronide conjugate ether bond and utilize the six‐carbon GlcA sugar to generate energy.6, 7, 8, 10, 11, 13, 14, 15 The previously glucuronidated aglycone parent compound is then either further metabolized or excreted back into the gut lumen, which can result in enterohepatic recirculation or in GI tract toxicity.8, 10, 11, 22, 23, 24 Potent, nonlethal, and microbial GUS‐targeted inhibitors have been used to alleviate drug‐induced gut toxicity and represent a potential new therapeutic paradigm through the selective modulation of bacterial enzymes to impact human health.6, 10, 11, 15, 24, 25

Guided by a conserved active‐site architecture of known GUS crystal structures, we recently used a method to identify, catalogue, and categorize microbial GUS genes from metagenomic data assembled by the Human Microbiome Project.13 A total of 279 unique microbiome‐encoded GUS proteins were discovered using this process, which we have termed Metagenome Analysis by Protein Structure, and sorted into six distinct categories based on structural active‐site loop classifications: Loop 1, Loop 2, Mini‐Loop 1, Mini‐Loop 2, Mini‐Loop 1,2, and No Loop (NL).13 After purifying a representative set of GUS proteins from each loop classification, we found that these enzymes displayed variability in function with respect to their ability to process a small‐molecule glucuronide versus a larger glucuronide polysaccharide substrate. A GUS from Parabacteroides merdae was examined as a part of this in vitro analysis because the gene for this protein was found to have the highest level of reads among this class of GUS enzymes in the HMP stool sample catalog.13 mL2 GUS proteins also appear to be prevalent, as genes for mL2 GUS proteins were identified in 81% of the 139 individuals present in the HMP stool sample dataset employed in this study.13 In vitro substrate analysis of the P. merdae mL2 GUS (PmGUS) revealed that it was the least efficient at processing the small molecule substrate p‐nitrophenyl‐β‐d‐glucuronide (pNPG) and the most efficient at processing the larger substrate employed, a heparan nonasaccharide substrate GlcA‐(GlcNAc‐GlcA)₄‐pNP, where GlcNAc is N‐acetylglucosamine and pNP is p‐nitrophenol.13

Heparan is a member of the glycosaminoglycan family of extended linear polysaccharides consisting of repeating disaccharide units of an amino sugar, often N‐acetylglucosamine (GlcNAc) or N‐acetylgalactosamine (GalNAc), and a uronic sugar acid, such as GlcA or iduronic acid. Heparin, chondroitin and hyaluronic acid contain a GlcA in their repeating disaccharide units, and the carboxyl acid groups at the six positions of the uronic acid sugars results in such glycosaminoglycan chains possessing a negative charge. Additionally, heparan/heparin and chondroitin can be sulfated at multiple positions, further increasing their negative charge density. Glycosaminoglycans are the most abundant heteropolysaccharide found in the human body and are primarily located in the extracellular matrix and on cell surfaces.26, 27 Glycosaminoglycans are both biological and physiological important molecules to human health.28, 29, 30, 31, 32, 33, 34 In addition to such endogenous polymers, exogenous GlcA‐containing polysaccharides, including alginic acid which is either ingested or created by the intestinal microbiota, also have implications for the gut microbiome. Alginic acid can be heteropolymer or homopolymer chain composed of guluronic acid and/or mannuronic acid. Alginic acid polymers are generated by bacteria as part of biofilms, present in dietary plant sources, and used as food additives.35 Thus, the gut microbiota has access to a wide array of endogenous and exogenous polysaccharides.

To gain molecular insights into PmGUS's unique ability to process polysaccharides over a small molecule glucuronide, we present the 1.9 and 2.2 Å resolution crystal structures of PmGUS in two distinct states. We compare PmGUS to other microbial GUS proteins of known structure, from Escherichia coli, Bacteroides fragilis, and Bacteroides uniformis. Additionally, we demonstrate the ability of PmGUS to bind and process polysaccharides, and we describe a large conformational shift of a protein loop region that affects the enzyme's activity. Together, these results advance our understanding of gut microbial GUS enzyme structure and function.

Results

Structure of PmGUS tetramer

Crystals of P. merdae GUS that diffracted X‐rays to 2.2 Å resolution belonged to space group P2₁ with a solvent content of 52% and an asymmetric unit containing of four monomers. The structure was determined with molecular replacement and refined to a final model with a value of R _work of 0.166 and an R _free of 0.200 (Table 1; PDB ID: 6D7J). The four monomers in the asymmetric unit formed a “square” tetramer composed of residues 20–532 and 542–825. Residues 20–532 form the standard glycoside hydrolase 2 (GH2) fold and contain the GUS active site residues E430, E514, N610, and K612, while residues 546–829 follow in the primary sequence and form two C‐terminal domains of unknown function [DUFs; Fig. 1(a)]. The crystallographically observed tetramer was confirmed for PmGUS in solution by size‐exclusion chromatography with multiangle light scattering (SEC‐MALS) analysis (Supporting Information Fig. S2). The four PmGUS monomers share a root‐mean‐square deviation (rmsd) of 0.27 Å between equivalent Cα positions. A Dali‐server search36, 41 performed to identify structural homologs of PmGUS with less than 5.0 Å rmsd values yielded only known GUS protein structures. When compared to PmGUS structure, GUS enzymes from E. coli (a Loop 1 GUS; 3LPF), B. uniformis (Loop 2 GUS; 5UJ6), B. fragilis (Mini‐Loop 1 GUS; 3CMG), and human GUS (3HN3) exhibit 2.5 Å rmsd (over 530 Cα positions), 2.4 Å rmsd (656 Cα positions), 2.2 Å rmsd (623 Cα positions), and 2.3 Å rmsd (526 Cα positions), respectively, but exhibit relatively low sequence identities of 17, 22, 15, and 22%, respectively.

Table 1.

Crystallographic Statistics

Structure	PmGUS insert	PmGUS insert
	Open	Closed
Space group	P21	P21
Unit cell: a, b, c	92.9, 172.0, 125.3,	91.4, 163.4, 120.6,
(Å); α, β, γ (°)	90, 107.8, 90	90, 103.4, 90
Resolution range,	44–2.2	29–1.9
Å (highest shell)	(2.3–2.2)	(1.96–1.90)
Wavelength, Å	0.97943	0.97943
Unique reflections	171,534 (15,583)	269,582 (26,847)
Multiplicity	6.6 (4.3)	6.7 (6.6)
Completeness, %	96.2 (87.4)	99.9 (100)
l/σ	12 (2.2)	15 (3.1)
Wilson B‐factor	25.7	24.3
R merge	0.111 (0.470)	0.08684 (0.6123)
R work	0.166 (0.222)	0.157 (0.202)
R free	0.200 (0.252)	0.184 (0.250)
Molecules in AU	4	4
Waters in AU	1864	2717
Residues in AU	3165	3183
Average B‐factor	30.4	28.7
RMS bonds	0.006	0.008
RMS angles	0.81	1.2
Ramachandran favored, %	96	97
Ramachandran outliers, %	0.6	0.3
Rotomer outliers, %	1.1	1.8
PDB ID	6D7J	6DXU

Figure 1.

Quaternary structure of P. merdae GUS tetramer. (A) P. merdae GUS tetramer observed in the asymmetric unit, with initial (20) and terminal (825) residues in the orange monomer, as well as the position of a missing loop region (532–542). Representative active site residues are rendered in sticks in each monomer, and the locations of the two C‐terminal domains of unknown function (DUF1 and DUF2) are indicated. A twofold axis of symmetry within the plane of the page is indicated. (B) Quaternary structure of the B. fragilis GUS tetramer, with their single DUFs indicated as well as active site residues rendered as spheres. (C) Quaternary structure of the E. coli GUS tetramer with active site residues rendered as spheres. This GUS lacks a C‐terminal domain.

We next compared the PmGUS tetramer with those of tetrameric microbial GUS enzymes reported to date, which are B. fragilis GUS (BfGUS) and E. coli GUS (EcGUS).13 We find that each enzyme forms a unique tetrameric quaternary structural arrangement (Fig. 1). While PmGUS forms quaternary contacts with its DUF2 domains, BfGUS does not involve its single DUF in its quaternary structural contacts, instead exposing them to solvent [Fig. 1(b)]. EcGUS, which lacks a C‐terminal domain, creates its tetramer as a dimer of dimers involving only contacts with its GH2 fold [Fig. 1(c)]. Thus, gut microbial GUS enzymes of the same oligomeric state utilize a variety of distinct monomeric contacts to achieve their final quaternary structures, as observed here for the tetramers of PmGUS, BfGUS, and EcGUS.

Unique structural features of PmGUS monomer

PmGUS maintains the conserved GH2 fold for the first three‐quarters of its primary sequence, but places a 38‐residue insert at residues 521–558 that is unique to this protein among GUS enzymes of known structure [Fig. 2(a)]. As such, the PmGUS monomer's central GH2 region extends to residue 634, longer than the ~600‐residue core GH2 fold observed in EcGUS, the first microbial GUS protein resolved structurally and a representative of the Loop 1‐type enzymes [Fig. 2(b)]. The GH2 fold region of PmGUS is followed by two C‐terminal DUFs, DUF1 (residues 631–724) and DUF2 (residues 725–825) [Fig. 2(a)]. The structures of the two DUF domains from PmGUS were examined using PDBe for fold similarity, and the top hit sidentified were monobody immunoglobulin proteins. Thus, by structure similarity searching, the potential functions of these domains in the microbial PmGUS enzyme cannot be firmly ascribed. As noted above, the core GH2 motif is structurally similar to that seen in EcGUS, with the exception that PmGUS's 521–558 insert replaces a short stretch of seven residues (514–520) in EcGUS [Fig. 2(b)]. While EcGUS lacks a C‐terminal domain, BfGUS, a representative of the Mini‐Loop 1 microbial GUS enzyme, contains one DUF, formed by residues 606–693 that follows its core GH2 region. This single BfGUS DUF aligns well in terms of both secondary structure and tertiary position with DUF1 of PmGUS [Fig. 3(a)]. Still, the 521–558 insert is unique to PmGUS GUS, and replaces 15 residues (520–534) in BfGUS [Fig. 3(a)].

Figure 2.

Tertiary structure of P. merdae GUS monomer. (A) Structure of the P. merdae GUS (PmGUS) monomer with key structural motifs indicated, representative active site residues rendered as sticks, and glucuronic acid (GlcA; green) docked in the active site. (B) Superposition of the PmGUS monomer on the E. coli GUS (EcGUS; blue) monomer, with key differences indicated.

Figure 3.

P. merdae GUS versus B. fragilis and B. uniformis GUS Enzymes. (A) Superposition of the PmGUS monomer on the B. fragilis GUS (BfGUS; green) monomer, with key differences indicated. Note that PmGUS DUF1 aligns well with the single DUF of BfGUS. (B) Superposition of the PmGUS monomer on the B. uniformis GUS (BuGUS; teal) monomer, with key differences indicated. While the DUF1s of both monomers align well, the DUF2s deviate in position relative to one another.

Finally, the core GH2 regions of PmGUS and BuGUS, a representative Loop 2 microbial GUS enzyme,13 align well, although the 521–558 PmGUS insert is distinct when compared to BuGUS, as well [Fig. 3(b)]. However, while both PmGUS and BuGUS contain two C‐terminal DUFs, only the DUF1 regions align well in tertiary structural space, with DUF2 of BuGUS being positioned closer to the active site region of this enzyme. In contrast, DUF2 of PmGUS is located along the side of this protein's monomer fold, well away from the catalytic residues of this enzyme [Fig. 3(b)]. Thus, while the N‐terminal GH2 folds of PmGUS, a Mini‐Loop 2 enzyme and representatives of the Loop 1, Mini‐Loop 1, and Loop 2 microbial GUS proteins are similar, PmGUS displays a unique 38 residue insert within this region, and two C‐terminal DUFs. The first DUF aligns well with other GUS DUFs in this location in primary sequence, and the second deviates in position compared to the two DUF‐encoding BuGUS protein. Taken together, the tertiary and quaternary structural observations related to PmGUS and other microbial GUS orthologs presented here support the conclusion that these enzymes sample unique structural features to decorate their core GH2 folds, and that experimental structural studies are required to resolve these structural details. Indeed, the 521–558 insert in PmGUS is found experimentally to be capable of a significant shift in position, as outlined below.

Mini‐loop 2 cationic PmGUS active site

The active site of PmGUS is framed by a Mini‐Loop 2 region composed of residues 429–439 that extend as a loop adjacent to the catalytic E426 and E510 residues [Fig. 4(a)]. PmGUS contains NL in the Loop 1 region, which can extend to nearly 30 residues in length and was first identified in the Loop 1 GUS enzymes from E. coli and other Firmicutes GUS proteins (e.g., from S. agalactiae, C. perfringens)23 instead, PmGUS exhibits a short turn region composed of residues 389–393. Indeed, when comparing the active site‐adjacent loop regions of PmGUS and that of the Loop 1 enzyme EcGUS, which lacks a Loop 2 region, the distinct loop architectures of the two proteins are evident, and it is noted that the Mini‐Loop 2 of PmGUS partially occupies the position of Loop 1 in EcGUS [Fig. 4(b)]. A similar observation is evident when comparing the active site loop regions of PmGUS and the Mini‐Loop 1 enzyme BfGUS [Fig. 4(c)], which also lacks a Loop 2 insert. While the 14‐residue BfGUS Mini‐Loop 1 region extends further into the active site than the PmGUS Mini‐Loop 2, portions of these two active site‐adjacent loops occupy similar spaces in both structures (e.g., 430–432 in PmGUS, 393–395 in BfGUS) [Fig. 4(c)]. Finally, when comparing the PmGUS and BuGUS, neither of which contain a Loop 1, we find that the 19‐residue Loop 2 region of BuGUS folds into a helix‐turn‐helix motif that is structurally distinct from the loop formed by the 11‐residue Mini‐Loop 2 of PmGUS [Fig. 4(d)]. In spite of this detailed difference in loop 2‐region secondary structure, both PmGUS and BuGUS create less occluded active sites than EcGUS or BfGUS (Figs. 2, 3, 4), potentially for the purpose of accommodating larger substrates, as shown previously for BuGUS and as outlined below for PmGUS.

Figure 4.

Mini‐Loop 2 Region of P. merdae GUS. (A) Active site and Mini‐Loop 2 region (429–439, orange) of P. merdae GUS with the key active site residues rendered (yellow), and the location of the Loop 1 region, missing in this GUS, indicated (329–393, black). (B) Superposition of P. merdae and E. coli GUS active site regions with the intact Loop 1 (356–380, blue) and lack of Loop 2 (416–419, black) in the E. coli enzyme indicated. (C) Superposition of P. merdae and B. fragilis GUS active site regions with the intact Mini‐Loop 1 (385–396, green) and lack of Loop 2 (432–435, black) in the B. fragilis enzyme indicated. (D) Superposition of P. merdae and B. uniformis GUS active site regions with the intact Loop 2 (428–446, teal) and lack of Loop 1 (389–393, black) in the B. uniformis enzyme indicated.

Examination of the active site residues of PmGUS reveals that this enzyme maintains several of the conserved microbial GUS enzyme amino acids, but also exhibits unique features. The catalytic dyad residues of E426 and E510 are present in the PmGUS active site, as are N606 and K608 of the N‐K motif that recognizes the carboxylate unique to the 6‐position of GlcA‐containing substrates.13, 23 W587, which supports the positioning of active site residues, is also present [Fig. 5(a)]. However, the PmGUS active site contains three arginines, R464, R465, and R389 and two tryptophans, W481 and W429, that have not been observed previously in gut microbial GUS enzymes [Fig. 5(a)]. W481 replaces a tyrosine side chain that is almost completely conserved in gut microbial GUS enzymes and in all other GUS protein structures determined to date [Fig. 5(b–d)].13 The W and Y side chains occupy highly similar positions in the active sites of the enzymes compared [Fig. 5(b–d)]. By contrast, W429 in PmGUS is found to replace leucines or a glutamic acid when compared with the active sites of EcGUS [Fig. 4(b)], BfGUS [Fig. 5(c)], and BuGUS [Fig. 5(d)], and these non‐W residues do not occupy the same position as W429 in PmGUS. R464 and R465 in PmGUS largely replace hydrophobic or polar residues when compared to the active sites of EcGUS BfGUS, and BuGUS, with the exception of H473 in BuGUS; thus, these cationic side chains are unique to PmGUS [Fig. 5(b–d)]. Similarly, the cationic nature of the third PmGUS active site arginine, R389, is also unique when compared to related gut microbial GUS enzymes [Fig. 5(b–d)]. Thus, PmGUS creates a distinctly charged active site with two unique tryptophan side chains (Fig. 5). The functional relevance of these unique PmGUS active site residues are outlined below.

Figure 5.

Unique P. merdae GUS active site. (A) Key active site residues of P. merdae GUS rendered, with side chains unique to this GUS labeled in bold. (B) Superposition of P. merdae (orange) and E. coli (blue) GUS active site residues, with non‐identical side chains highlighted; note that F357 in E. coli GUS is rotated out of view by the distance indicated. (C) Superposition of P. merdae (orange) and B. fragilis (green) GUS active site residues, with non‐identical side chains highlighted; note that P386 and E432 in B. fragilis GUS are rotated out of view by the distances indicated. (D) Superposition of P. merdae (orange) and B. uniformis (teal) GUS active site residues, with nonidentical side chains highlighted.

PmGUS conformational change

Given PmGUS's ability to bind to alginate compounds, as outlined below, we co‐crystallized PmGUS in the presence of a 20‐fold molar excess of a homopentamer of mannuronic acid. Crystals were of space group P2₁ and diffracted X‐rays to 1.9 Å resolution (Table 1). The structure was determined by molecular replacement using the PmGUS structure present above as a search model, and refined to R _work and R _free values of 0.157 and 0.184, respectively (Table 1; PDB ID: 6DXU). The mannuronate compound was not observed bound to the enzyme, either at the active site or elsewhere. However, in three of the four monomers in the tetrameric asymmetric unit, the 521–558 insert shifted in position from an open conformation observed in the structure outlined above to a closed position where it enters the PmGUS active site [Fig. 6(a)]. Indeed, the shift is large with respect to the change in position of residues at the tip of the insert: in the open conformation, G542 is well away from the active site, but in the closed conformation this residue's Cα position shifts by 53 Å, placing the 538–545 region of the insert within the active site of PmGUS [Fig. 6(b)]. In spite of these changes, the remainder of the active site residues remain in the same positions in the open and closed PmGUS structures determined [Fig. 6(b)]. E539 on the insert, when in the closed position, is within 3 Å of the hydroxyl oxygen in the 1 position of a modeled GlcA sugar in the PmGUS active site. Thus, the PmGUS 521–558 insert is conformationally flexible and, when in the closed position, places residues within the active site of the enzyme, potentially playing a role in substrate binding or catalysis. This degree of repositioning of secondary structural elements has not been observed in gut microbial GUS enzyme structures resolved to date.

Figure 6.

Structural flexibility in P. merdae GUS. (A) Superposition of the monomers of P. merdae GUS in the “Insert Open” conformation (orange), and the “Insert Closed” conformation (yellow), with the relevant inserts rendered in red and magenta, respectively. G542 shifts in position by 53 Å between the open and closed conformations. (B) In the Closed conformation, residues 538–545 are positioned close to the enzyme's active site such that the side chain of E539 could directly contact a docked glucuronic acid sugar (green).

PmGUS processing of p‐nitrophenyl‐glucuronide

We next examined the ability of PmGUS to process the standard small molecule β‐glucuronidase substrate p‐nitrophenyl‐glucuronide (pNPG). Wild‐type PmGUS exhibited an apparent k _app of 1.5 ± 0.05 min⁻¹ for pNPG (Table 2), which equates to 0.025 s⁻¹. This is the lowest activity observed with gut microbial GUS enzymes when compared to values recently obtained for pNPG processing by Loop 1 (k _app 2.8–122 s⁻¹), Mini‐Loop 1 (23 s⁻¹), Loop 2 (24 s⁻¹), Mini‐Loop1, 2 (0.48 s⁻¹), and NL (7.5 s⁻¹) enzymes.13 These low activities are likely due to the unique residues in the active site of PmGUS relative to the other gut microbial GUS proteins examined. To test this hypothesis, we examined active site mutants of PmGUS with this standard small molecule GUS substrate, focusing here on the residues outlined above that are unique to the PmGUS active site (Table 2). We found that replacing the distinct tryptophan at position 481 with the more typical tyrosine (W481Y) increased PmGUS activity with pNPG by 10‐fold, to 12 min⁻¹, perhaps due to the tyrosine's ability to contact the carboxylate at the 6‐position of the small pNPG substrate. While an individual R464A mutation eliminated activity, the individual R465A and dual mutations of the arginines at 464 and 465 exhibited only slight reductions in activity (Table 2). In contrast, a dual mutation of both arginines to glutamines (R464Q + R465Q) increased PmGUS activity with pNPG by 16‐fold, to 25 min⁻¹, suggesting that active‐site polarity but not cationic charge enhances pNPG processing. Taken together, these data focused on the unique active site residues in PmGUS indicate that the enzyme is not natively tuned for this type of small aromatic substrate, and suggest that more efficient substrates might be larger compounds that can take advantage of the open space and charged character of this gut microbial GUS enzyme.

Table 2.

PNPG Processing Activity of PmGUS Wild‐Type and Designed Protein Variants

PmGUS	k app (min−1)
WT	1.5 ± 0.07
W481Y	12.7 + 0.3
R464A	No activity
R465A	1.1 ± 0.05
R464A + R465A	0.66 ± 0.03
R464Q + R465Q	25.4 + 0.2
R389A	4.1 ± 0.1
E539A	No activity
Delta‐insert	2.0 ± 0.07

We found that deleting the insert observed to undergo a conformational change (Delta‐Insert) did not affect the processing of the small molecule glucuronide pNPG (Table 2). We confirmed that this Delta‐Insert protein remained its tetrameric quaternary state by SEC‐MALS, akin to the wild‐type protein (Supporting Information Fig. S2). To our surprise, however, mutating the glutamate side chain that was observed placed into the active site of PmGUS to alanine (E539A) eliminated activity with this substrate. Thus, we conclude that E539 plays a role in substrate processing by PmGUS. Such a “distant active site” residue has not been observed to date with gut microbial GUS proteins.

PmGUS processing of polysaccharide substrates

We next examined the ability of PmGUS to remove the terminal glucuronic acid moieties from neutral and anionic polysaccharides. We tested carbohydrates molecules five sugars in length composed of alternating glucuronic acid and either N‐acetylglucosamine or a 4‐sulfated glucose moiety, and utilized wild‐type or a select set of PmGUS mutants. We found that wild‐type PmGUS processed both substrates efficiently, as did the W481Y and Delta‐Insert forms of PmGUS (Table 3). In contrast, forms of PmGUS in which the two active site arginines were both replaced with glutamine (R464Q + R465Q), as well as the form of the enzyme in which the glutamate that can be placed into the enzyme's active site is replaced by alanine (E539A), both showed dramatically reduced or eliminated activity with these polysaccharide substrates (Table 3). These data indicate that the active site arginines are essential to polysaccharide cleavage. Taken together, these results suggest that are polysaccharides are PmGUS's native substrates. Furthermore, the fact that E539A dramatically impacted processing with these substrates, as well as with the small molecule substrate pNPG, indicates this side chain plays a role in substrate processing by this gut microbial GUS enzyme. Taken together, these data advance our understanding of substrate preferences and active site motions for the GUS from the human gut commensal species P. merdae.

Table 3.

Polysaccharide Processing by PmGUS Wild‐Type and Designed Protein Variants

PmGUS	NAc 5mer	NS 5mer
	% Cleavage	% Cleavage
WT	95 ± 1	96 ± 0.3
W481Y	99 ± 0.1	97 ± 0.4
R464Q + R465Q	4.3 ± 0.1	0
E539A	4.4 ± 0.1	0
Delta‐insert	98 ± 0.5	97 ± 0

PmGUS binding to alginate

Given the distinct active site architecture of PmGUS, we sought to better understand other types of polysaccharide substrates this enzyme might process. Examination of the genes adjacent to the gus gene in P. merdae revealed the presence of a putative alginate lyase [Supporting Information Fig. S3(a)]. Thus, we hypothesized that PmGUS may bind to or catalytically process alginate compounds, which are composed of repeating or alternating guluronic acid and mannuronic acid sugars, both of which, like GlcA, maintain a carboxylate at their six positions. While guluronate and mannuronate have axial 2‐hydroxyls while GlcA has an equatorial 2‐hydroxyl, there is space in the GUS active site to accommodate this alternative positioning at the 2‐position of a bound substrate. However, importantly, the guluronic acid carboxylate at the 6‐position of the sugar is equatorial, while the carboxylates in mannuronic acid and GlcA are in the axial position in these sugars. We first tested if PmGUS could catalytically process three alginate substrates: penta‐guluronic acid, penta‐mannuronic acid, and a guluronate‐mannuronate disaccharide. We found using isothermal titration calorimetry that PmGUS was not able to utilize these compounds as substrates, as the change in heat was indicative only of binding, as outlined below, and not catalysis.

Second, we examined the ability of PmGUS to bind to the same three alginate‐related compounds using isothermal titration calorimetry, and found that the enzyme exhibited 88–152 μM affinity for the alginate sugars examined [Supporting Information Fig. S3(b)]. Thus, while PmGUS cannot catalytically process these substrates, it exhibits the ability to associate them. Furthermore, examination of representative human gut microbial GUS enzymes from the other loop classifications revealed that no other protein tested bound to any of the three alginate sugar compounds studied [Supporting Information Fig. S3(b)]. Thus, PmGUS is unique in its ability to associate with either a small mannuronate‐guluronate‐mannuronate disaccharide, or relatively short homopolymers of mannuronic acid or guluronic acid.

Discussion

The first structure of a microbial GUS enzyme was reported in 2010, that of the protein from E. coli and one in which a “bacterial loop” was identified that explained GUS‐directed inhibitor potency and selectivity.10 The bacterial loop was not present in the human GUS enzyme ortholog and was hypothesized at the time to be unique to microbial GUS enzymes.10, 37 We now have determined several more structures of gut microbial GUS enzymes,6, 13 and the “bacterial loop” has been renamed “Loop 1” because the 279 microbial GUS enzymes present in the HMP stool sample catalog offer loops in at least two places in the GUS active site.13 As such, our knowledge of GUS enzymes, by both structure and sequence,23 continues to grow and add to our understanding of the diversity of functions and active site conformations that these enzymes sample. A key feature of this advancement of knowledge has been the role experimental structure determination plays in pinpointing new features of protein tertiary and quaternary structure. It is expected that crystal structures and those determined by cryo‐electron microscopy38, 43 will be essential in continuing to resolve the variability present in gut microbial enzymes.

To date, all GUS enzyme structures reported have existed in oligomerization states, either as dimers or tetramers.10, 13, 37, 42 At the quaternary structural level, the PmGUS structure, a tetramer, reveals that gut microbial GUS enzymes sample a variety of packing orientations to create the same oligomerization state (e.g., see Fig. 1). The C‐terminal domains, located in some GUS enzymes following the GH2 fold, appear to influence the contacts individual protein monomers make in assembling into a protein oligomeric state. For example, the second C‐terminal domain of PmGUS contact one another in this protein's tetramer, while the single DUF of the Mini‐Loop 1 GUS from B. fragilis is not involved in quaternary structural contacts. Indeed, these domains may influence function at the active site of some GUS enzymes, as they have been observed in close proximity to the catalytic gorge of BuGUS, for example.13 Here, however, the C‐terminal domains of PmGUS are located well away from the enzyme's catalytic site (e.g., see Figs. 2 and 3), perhaps to allow space for the mobile insert, as discussed below. Thus, GUS C‐terminal domains play roles in the oligomerization state of gut microbial enzymes, but they are not always in intimate contact with GUS active site regions.

PmGUS uniquely places arginines and tryptophans at its active site that have not been observed in other GUS enzymes to date. We find that mutation of these residues affects the catalytic actions of the protein, increasing its activity with the small pNPG glucuronide substrate, for example, with RR‐QQ or W481A mutations. These active site distinctions, along with PmGUS's ability to bind to anionic compounds like the alginic acid carbohydrates made up of mannuronic and guluronic acid, suggest that substrates for PmGUS are sugars with a terminal GlcA moieties and that the enzyme performs exolytic activity. Given the structure of the enzyme, it is not expected that PmGUS could act on a GlcA located within a polysaccharide chain (endolytic activity). Candidate substrates include a vast array of dietary compounds, as well as endogenous polysaccharides like those present on MUC1, for example. Future studies will be required to identify specific substrates for this unique GUS enzyme. The role of the DUFs in PmGUS also remain unclear. However, surface electrostatics reveals that PmGUS maintains a relatively positive active site, as well as a positive patch on its DUF2 domain [Supporting Information Fig. S4(a)]. The residues that create the DUF2 positive region are conserved in some PmGUS orthologs [Supporting Information Fig. S4(a)]. In contrast, other GUS enzymes of known structure do not exhibit positive active sites or DUF‐associated surface positive patches [Supporting Information Figs. S4(b) and S4(c)]. Thus, a role for DUF2 in PmGUS may be to contact anionic polysaccharides and align them for processing.

What is clear, and intriguing, from this structure is its ability to reposition an insert region (521–558) not previously observed in a GUS from an “out” location to an “in” location such that it can place an amino acid side chain, E539, at the enzyme's active site. While PmGUS is a poor enzyme with the small substrate pNPG, elimination of the 521–558 loop region does not affect the protein's processing of this compound. However, mutation of the glutamate on the insert—the E539A mutation—creates a variant that is inactive with pNPG, as well as polysaccharide compounds, which appear to be more native substrates for this enzyme. One wonders if this “triggered” glutamate is capable of being swapped into the active site to perform a catalytic role, as the standard catalytic E side chains are located proximally to neutralizing arginine side chains (e.g., see Fig. 5). Indeed, we found that a rotamer conformation of E539 places it within 2.6 Å of R465 unique to PmGUS [Fig. 7(a)]. We further found that, of the six Mini‐Loop 2 enzymes with the 521–558 insert in the 279 GUS proteins we identified, all have a conserved E within the insert and maintain conserved RR motifs [Fig. 7(b)]. Thus, residues equivalent to E539, R464, and R465 occur together in Mini‐Loop 2 GUS enzymes. It is possible that E539 neutralizes the charge on R465 to facilitate substrate processing.

Figure 7.

mL2 insert glutamate and active site arginines. (A) In the Closed conformation, an E539 side chain rotamer positions it within 2.6 Å of R465. (B) Six mL2 GUS enzymes from the 279 GUS proteins in the human gut microbiome that contain inserts with glutamates also maintain two sequential active site arginines.

Taken together, the data presented provide the first molecular framework in which to understand the potential function(s) of PmGUS. This enzyme most likely works in concert with the other genes proximally located within this organism's chromosome [e.g., see Fig. 6(a)]. It is possible that PmGUS plays dual roles. One role could be in binding to polysaccharide substrates that it does not catalytically process, compounds like alginate that when immobilized by PmGUS are then catabolized by alginate lyase and/or the GH88, with products imported by the SusCD machinery. The second role could be in directly degrading other anionic substrates present in the periplasmic space of this bacterium in the human intestinal tract, compounds such as mucins like MUC1.39, 44 These results, and the hypotheses they generate, advance our understanding of the structural and functional variability present in the gut microbial GUSome.

Materials and Methods

Cloning and expression

A full‐length construct PmGUS was synthesized and codon optimized for E. coli expression by BioBasic and provided in a pUC57 vector. A construct of PmGUS containing only amino acids 20–825, which removes the signal sequence, was cloned into a pLIC expression plasmid vector with an N‐terminal hexa‐histidine tag with a Tobacco etch virus protease cleavage site. The pLIC expression plasmid was transformed into BL21‐Gold competent cells (Agilent Technologies) to create expression cells that were then cultured in Auto‐induction (ZYP‐5052) media (PASM‐5052 media was used for selenomethionine labeled protein) in the presence of antifoam (50 μL) and ampicillin (100 μg/mL), shaking 350 RPM at 37°C.36, 40 When an OD600nm of ~0.8 was reached, the temperature was reduced to 18°C. Induced cells were allowed to grow overnight at 18°C, and the culture pellets were collected the following morning by centrifugation at 4500g for 25 min at 4°C in a Sorvall model RC‐3B swinging bucket centrifuge. Cell pellets were stored at −80°C.

Protein purification

Cell pellets were resuspended in buffer A [20 mM potassium phosphate (pH 7.4), 50 mM imidazole, 500 mM NaCl, 0.5 mM tris(2‐carboxyethyl)phosphine [TCEP]), along with lysozyme, DNase1, and protease inhibitor tablets. Cells were sonicated, and cell lysate was separated into insoluble and soluble fractions by centrifugation at 17,000g for 60 min in a Sorvall model RC‐5B centrifuge. Soluble fractions were syringe‐filtered through a sterile 22 μM polyethersulfone membrane, applied to a Ni‐NTA His‐Trap gravity column (GE Healthcare Life Sciences), and washed with buffer A. Bound protein was eluted with buffer B (20 mM potassium phosphate [pH 7.4], 500 mM imidazole, 500 mM NaCl, 0.5 mM TCEP) and applied to an S200 gel filtration column in buffer C (20 mM 4‐[2‐hydroxyethyl]‐1‐piperazineethanesulfonic acid [HEPES], 300 mM NaCl, 0.5 mM TCEP) on an ÄKTAxpress (GE Healthcare Life Sciences). PmGUS eluted from the S200 column as a single peak in buffer C. Resultant fractions were analyzed by SDS‐PAGE; those with >95% purity were combined and concentrated to 15 mg/mL with 50 kDa molecular weight cutoff centrifugal concentrators (EMD Millipore), and snap‐frozen using liquid nitrogen and stored at −80°C in buffer C.

Size exclusion chromatography with multiangle light scattering

PmGUS was analyzed on a Superdex 200 size exclusion column connected to an Agilent FPLC system, Wyatt DAWN HELEOS II multiangle light scattering instrument and a Trex refractometer. The injection volume was 50 μL, and PmGUS protein was assessed at 24 mg/mL [20 mM HEPES, 10% Glycerol, and 300 mM NaCl at pH 7.4] and 12 mg/mL [20 mM HEPES and 300 mM NaCl at pH 7.4]. A flow rate of 0.5 mL/min was employed. Light scattering and refractive index data were collected and analyzed using Wyatt ASTRA (Ver. 6.1) software. A dn/dc value of 0.185 was used for calculations. Approximately 99% of PmGUS at 24 mg/mL eluted in a single peak with a molar mass of 364 kDa, indicating a tetramer in solution. Approximately 98% of PmGUS at 12 mg/mL eluted in a single peak with a molar mass of 363 kDa, also indicating a tetramer.

Crystallization

Initial crystallization conditions were identified in 96‐well sitting drop trays using a Rigaku Phoenix Liquid Handler with 15 mg/mL of PmGUS in buffer C. Crystallization trays were monitored and tracked by Rigaku Gallery 700 Plate Hotels held at 20°C. Initial crystal hits for PmGUS were pursued for hit conditions: condition 1 (0.15 M Cesium Chloride, 15% PEG 3350) and condition 2 (10% PEG 3350, 0.2 M Proline, 0.1 M HEPES pH 7.4). Crystals of hit condition 1 were reproduced with selenomethionine‐labeled PmGUS in the laboratory by hanging drop vapor diffusion crystallization method at 20°C. Crystals of hit condition 2 were reproduced with PmGUS in the laboratory by hanging drop vapor diffusion at 20°C. Both crystal conditions were looped and soaked in their respective crystallization conditions plus 25% glycerol for cryoprotection. X‐ray data for crystals from condition 1 were collected at 100 K to 2.4 Å resolution at the Advanced Photon Source using the General Medical Sciences and the National Cancer Institute Collaborative Access Team (GM/CA‐CAT) beamline 23ID‐D. Data for crystals from condition 2 were collected at 100 K to 2.24 Å resolution at the APS and using the GM/CA‐CAT beamline 23ID‐D (Table 1). Crystals were also grown in condition 2 in the presence of a hexamer of mannuronic acid, and data from those crystals were collected to 1.9 Å resolution at GM/CA‐CAT beamline 23ID‐D (Table 1).

Structure determination

The 2.2 Å resolution data collected were manually processed by HKL 2000 (Table 1). The structure of PmGUS was determined by the Phaser molecular replacement tool in PHENIX, using a PmGUS monomer partial structure revealing residues (31–526 and 588–630) from a previous 2.4 Å resolution selenomethionine labeled data set. After phasing, the initial model building was performed in PHENIX via the AutoBuild function.37, 41 Subsequent refinements and manual building were completed in PHENIX.REFINE and COOT, respectively.38, 41, 42, 43, 44 The final model consisting of 3165 residues with four molecules in the asymmetric unit, and refined to R _work and R _free values of 0.166 and 0.200, respectively (PmGUS Insert Open; Table 1). Atomic coordinate and structure factors have been deposited in the Protein Data Bank with accession code 6D7J. The data from the crystals grown in the presence of the mannuronic acid pentamer were processed using HKL 2000, and the structure determined by molecular replacement and refined to 1.9 Å resolution using PHENIX.REFINE. The structure contains four molecules in the asymmetric unit, and refined to R _work and R _free values of 0.157 and 0.184, respectively (PmGUS Insert Closed; Table 1). Atomic coordinate and structure factors have been deposited in the Protein Data Bank with accession code 6DXU. The structures are of high quality, as evidenced by the composite annealed omit maps for the active site regions (Supporting Information Fig. S1).

Isothermal titration calorimetry binding studies

All ITC measurements were performed at 37°C using an AutoITC200 microcalorimeter (MicroCal/GE Healthcare). The calorimetry cell was loaded with PmGUS protein at a concentration of 50 μM. The syringe was loaded with ligand (alginic acid disaccharide [S357340, Sigma‐Aldrich], penta‐guluronic acid [OP36504, Carbosynth] or penta‐mannuronic acid [OP61659, Carbosynth]) at concentrations of 3–5 mM in buffer C. A typical injection protocol included a single 0.2 μL first injection followed by 26 1.5 μL injections of the substrate into the calorimeter cell. The spacing between injections was kept at 180 s and the reference power at 8 μcal/s. A control experiment was performed by titrating ligand (alginate substrate) into buffer under identical settings to determine the heat signals that arose from compound dilution; these were subtracted from the heat signals of protein−compound interaction. The data were analyzed using Origin for ITC, version 7.0, software supplied by the manufacturer, and fit well to a one‐site binding model.

Enzyme activity assays

Mutants for PmGUS were created by PCR mutagenesis, confirmed by DNA sequencing, and variant proteins were expressed and purified as described above. pNPG assays (Table 2) and polysaccharide cleavage assays (Table 3) were performed as described previously.13

Conflict of Interest

M.R.R. is the scientific founder of Symberix, Inc., which is developing microbiome‐targeted therapeutics.

Supporting information

Figure S1. Composite Annealed Omit Maps. A. 2.2 Å resolution composite annealed omit electron density map of the active site side chains of the PmGUS Insert Open structure contoured at 1.5 sigma. B. 1.9 Å resolution composite annealed omit map of the active site side chains of the PmGUS Insert Closed structure contoured at 1.5 sigma.

Figure S2. P. merdae GUS SEC‐MALS. A. Size‐exclusion chromatography with multi‐angle light scattering (SEC‐MALS) on wild‐type PmGUS showing that it is a tetramer. B. SEC‐MALS analysis of the Delta‐Insert PmGUS variant showing that its oligomerization state does not change upon this mutation.

Figure S3. P. merdae GUS and Aliginate. A. Gene neighborhood of P. merdae GUS (orange) showing the proximity of a putative alginate lyase gene, AlgA (black). B. ITC studies of the binding of different forms of alginate to P. merdae GUS, as well as to representative GUS enzymes from the six distinct gut microbial GUS structural categories.

Figure S4. Electrostatic Surfaces of P. merdae GUS and Other GUS Enzymes. A. Zoom‐in on top dimer of PmGUS tetramer (same orientation as Figure 1A) showing electrostatic surface (top) and secondary structure (middle), with active site and DUF2 regions highlighted along with positively charged residues. At bottom, positive charge is conserved in some PmGUS orthologs from the 279 human gut microbial GUS enzymes identified to date. B. Electrostatic surface of the EcGUS tetramer with the active site shown. This GUS has no C‐terminal DUFs. C. Electrostatic surface of BfGUS tetramer showing that this enzyme has a relatively neutral active site and lack active site‐proximal positive surfaces arising from its single DUF.