The crystal structure of CbpD clarifies substrate-specificity motifs in chitin-active lytic polysaccharide monooxygenases

Pseudomonas aeruginosa secretes diverse proteins via its type 2 secretion system, including a 39 kDa chitin-binding protein, CbpD. CbpD has recently been shown to be a lytic polysaccharide monooxygenase active on chitin and to contribute substantially to virulence. To date, no structure of this virulence factor has been reported. Its first two domains are homologous to those found in the crystal structure of Vibrio cholerae GbpA, while the third domain is homologous to the NMR structure of the CBM73 domain of Cellvibrio japonicus CjLPMO10A. Here, the 3.0 Å resolution crystal structure of CbpD solved by molecular replacement is reported, which required ab initio models of each CbpD domain generated by the artificial intelligence deep-learning structure-prediction algorithm RoseTTAFold. The structure of CbpD confirms some previously reported substrate-specificity motifs among LPMOAA10s, while challenging the predictive power of others. Additionally, the structure of CbpD shows that post-translational modifications occur on the chitin-binding surface. Moreover, the structure raises interesting possibilities about how type 2 secretion-system substrates may interact with the secretion machinery and demonstrates the utility of new artificial intelligence protein structure-prediction algorithms in making challenging structural targets tractable.

1. Introduction

Pseudomonas aeruginosa is a Gram-negative bacterium that is responsible for many nosocomial infections in the blood (McCarthy & Paterson, 2017), urinary tract (Vincent, 2003) and lungs. It remains one of the primary causes of death in cystic fibrosis patients (Lyczak et al., 2000). A critical virulence factor in P. aeruginosa is its type 2 secretion system (T2SS; Jyot et al., 2011), through which it exports diverse proteins. One T2SS substrate is the 39 kDa chitin-binding protein CbpD (Bleves et al., 2010). CbpD was originally identified by its ability to bind only crystalline chitin (Folders et al., 2000), and was thus classified into carbohydrate-binding module family 33 (CBM33; Cantarel et al., 2009). CbpD is a virulence factor (Askarian et al., 2021) that carries several different post-translational modifications (Ouidir et al., 2014; Gaviard et al., 2018, 2019) and whose function may be modulated by proteolytic cleavage after secretion through the T2SS (Park & Galloway, 1995; Braun et al., 1998; Folders et al., 2000).

A decade after the discovery of CbpD, it was shown that the members of the CBM33 family are mono-copper binding enzymes with oxidative activity on recalcitrant polysaccharide substrates (Vaaje-Kolstad et al., 2010). These enzymes were reclassified as lytic polysaccharide monooxygenases (LPMOs; Horn et al., 2012). Specifically, LPMOs catalyze the oxidation of glycosidic bonds in abundant natural polymers, including chitin (Hamre et al., 2015), xylan (Corrêa et al., 2019), cellulose (Forsberg et al., 2011), starch (Vu, Beeson, Span et al., 2014) and pectin (Sabbadin, Urresti et al., 2021), with a range of substrate selectivity (Isaksen et al., 2014; Agger et al., 2014; Kojima et al., 2016; Zhou et al., 2019) and regioselectivity (Li et al., 2012; Isaksen et al., 2014; Vu, Beeson, Phillips et al., 2014; Forsberg, Røhr et al., 2014; Vaaje-Kolstad et al., 2017). LPMOs utilize a reactive oxygen species to perform H-atom abstraction followed by hydroxylation of a carbon adjacent to the glycosidic bond (C1 in chitin and either C1 or C4 in cellulose) in the polysaccharide substrate (Bissaro et al., 2020). Spontaneous elimination results in glycosidic bond cleavage (Beeson et al., 2012), nicking the polysaccharide and creating new ends for processive degradation of the biopolymer (Agostoni et al., 2017). While there has been some disagreement about the precise mechanism of LPMO oxidative cleavage (Bissaro et al., 2017; Hangasky et al., 2018), a thorough model has recently been proposed for the utilization of H₂O₂ as the reactive co-substrate (Bissaro et al., 2020). LPMOs are further characterized by a solvent-exposed Peisach–Blumberg type 2 copper active site in which a mononuclear copper ion is coordinated in a T-shape by a histidine brace comprising the N-terminal amine and two histidines (Peisach & Blumberg, 1974; Hemsworth, Taylor et al., 2013; Forsberg, Røhr et al., 2014).

The discovery of catalytic activity in CBM33 precipitated a reorganization of the Carbohydrate Active enZYme (CAZy) database and the creation of a new enzyme class, the auxiliary activity (AA) enzymes, into which CBM33 members and other enzymes were sorted (Levasseur et al., 2013). Based upon their sequence similarity, all CBM33 enzymes, including CbpD, were reclassified into the auxiliary activity 10 (AA10) family (Levasseur et al., 2013). LPMOs have now been found in bacterial (Vaaje-Kolstad et al., 2010), plant (Shukla et al., 2016; Yadav et al., 2019), fungal (Quinlan et al., 2011; Beeson et al., 2012), archaeal (Li et al., 2022), insect (Sabbadin et al., 2018) and viral (Chiu et al., 2015) genomes and are classified into families AA9–AA11 and AA13–AA17 (Vandhana et al., 2022).

The majority of LPMOs comprise a single copper-binding catalytic AA domain, while others have additional domains, themselves classified into families based on sequence identity (Horn et al., 2012; Book et al., 2014; Agostoni et al., 2017; Vaaje-Kolstad et al., 2017). These additional domains often contain aromatic residues that contribute to substrate binding and specificity (Cuskin et al., 2012; Gilbert et al., 2013; Crouch et al., 2016) and are also likely to protect LPMOs from autocatalytic inactivation (Courtade et al., 2018; Forsberg et al., 2018). In addition to its AA10 domain, CbpD has a GbpA2 domain and a carbohydrate-binding module 73 (CBM73), as annotated in the PFAM (Mistry et al., 2021) and CAZy (Drula et al., 2022) databases, respectively (Fig. 1).

Figure 1 The domain architecture of CbpD and representative domain structures. CbpD is composed of three domains: an AA10 domain (green), a GbpA2 domain (yellow) and a CBM73 domain (red). The domain structures of the AA10 and GbpA2 families are represented in the crystal structure of V. cholerae GbpA (PDB entry 2xwx), while the domain structure of the CBM73 family is represented in the NMR solution structure of CjLPMO10A (PDB entry 6z40). Hashes represent regions that are not included in the constructs used in the cited structural studies. Dashed orange lines encapsulate the domain(s) resolved in the respective reported structures. Residue boundaries for each domain are labeled based on previous annotations, adjusted so that the N-terminal His in each mature protein is residue 1.

While the catalytic activity of LPMOs is driving techno­logical advances in biomass degradation (Harris et al., 2010; Arfi et al., 2014; Horn et al., 2012; Rani Singhania et al., 2021), their role in virulence has been less thoroughly investigated. LPMOs play a role in virulence (Wong et al., 2012; Loose et al., 2014; Chiu et al., 2015; Agostoni et al., 2017; Sabbadin, Henrissat et al., 2021; Sabbadin, Urresti et al., 2021) and are upregulated in pathogenic bacteria when exposed to human tissues (Vebø et al., 2009, 2010). Similarly, evidence has existed for two decades that CbpD might play a role in P. aeruginosa virulence. The CbpD gene has extensively been found in pathogenic clinical isolates (Folders et al., 2000) and is up­regulated in the lungs of cystic fibrosis patients (Sriramulu et al., 2005). Recently, Askarian and coworkers demonstrated that catalytically active CbpD promotes the survival of P. aeruginosa in human whole blood, extends the survival of some other bacteria that do not endogenously express an LPMO, prevents cell lysis by the terminal complement pathway and is likely to aid the resistance of P. aeruginosa to immune clearing in mouse infection models (Askarian et al., 2021).

These insights into the roles that LPMOs play in virulence have prompted studies of the structure and function of chitin-binding and virulence-associated LPMOs in pathogenic bacteria. The AA10 and GbpA2 domains of CbpD are represented structurally in the Vibrio cholerae colon­ization factor GbpA (PDB entry 2xwx; Wong et al., 2012), while the CBM73 domain is found in a recently reported standalone NMR solution structure of the CBM73 domain of Cellvibrio japonicus CjLPMO10A (PDB entry 6z40; Madland et al., 2021). Despite the availability of these structures of representatives of the three domains of CbpD, no structure of this virulence factor has been reported.

Concurrently with progress in understanding the role of LPMOs in virulence, recent advances in artificial intelligence (AI)-powered ab initio protein structure prediction (Baek et al., 2021) allowed us to solve the partial structure of the virulence factor CbpD. While structures of other LPMOs have been reported, many are of single-domain LPMOs or are of the catalytic domain of multidomain LPMOs (Vaaje-Kolstad, Houston et al., 2005; Vaaje-Kolstad et al., 2012; Aachmann et al., 2012; Hemsworth, Taylor et al., 2013; Gudmundsson et al., 2014; Forsberg, Røhr et al., 2014; Mekasha et al., 2016). While the structure of V. cholerae GbpA is multimodular, it was generated from a truncated construct lacking the C-terminal domain (a CBM73; Wong et al., 2012; Fig. 1). We report the AI-enabled crystal structure determination of a multidomain LPMO. This structure of CbpD provides insights into domain architecture, substrate-binding motifs, crystal packing, and apparent order and disorder propensities in multidomain LPMOs.

2. Methods and materials

2.1. Protein production and crystallization

CbpD from P. aeruginosa was previously cloned with its native signal peptide into the EcoRI site of pT7.5 with an encoded C-terminal 10×His tag to yield pCbpD (Cadoret et al., 2014). Our purification procedure was modified only slightly from this publication. Overnight cultures of Escherichia coli BL21 (DE3) cells carrying pCbpD were used to inoculate 2 × 2 l cultures of Terrific Broth (Sigma–Aldrich), which were incubated at 37°C at 220 rev min⁻¹ until the OD₆₀₀ reached 0.5–0.6. Protein expression was induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and proceeded overnight at 30°C. Cells were collected using low-speed centrifugation (6000g) and the pellet was resuspended in buffer A (50 mM Tris, 300 mM NaCl, 10 mM imidazole pH 8.2) and frozen at −80°C. After 2 h, the cells were thawed in a 37°C water bath for 20–30 min to induce lysis. DNAse, MgSO₄ and lysozyme were added to final concentrations of 20 µg ml⁻¹, 20 mM and 0.5 µg ml⁻¹, respectively. The mixture was stirred at 4°C for 1 h to complete lysis. The clarified lysate obtained by centrifugation at 20 000g was loaded onto a 5 ml FFcrude prepacked nickel column (GE Healthcare) pre-equilibrated with buffer A. The column was washed with 10% buffer B (50 mM Tris, 300 mM NaCl, 250 mM imidazole pH 8.2), and CbpD was eluted using a gradient from 10 to 50% buffer B over 20 min. The elution fraction containing CbpD was concentrated to 5 ml using a 10K Amicon concentrator and loaded onto a Superdex S200 26/60 column equilibrated with 20 mM Tris, 100 mM NaCl pH 8. Fractions were analyzed by SDS–PAGE, which showed high purity of the purified protein (>95%), implying that no residual CbpD uncleaved by signal peptidase remained (Supplementary Fig. S1). Fractions containing CbpD were pooled, resulting in a yield of 2 mg total purified protein per litre of culture. For crystallization trials, the protein was concentrated to 8 mg ml⁻¹.

Crystals were grown by the hanging-drop method against a reservoir solution consisting of 0.2 mM diammonium tartrate in 20%(w/v) PEG 3350 (condition H2 of the Qiagen PEGs Suite sparse-matrix screen). A crystal was harvested, mounted and flash-cooled in liquid nitrogen without additional cryoprotection.

2.2. Data collection and structure determination

An X-ray diffraction data set was collected to 2.98 Å resolution from a single crystal on the PROXIMA-1 beamline at the SOLEIL synchrotron. The data set was processed in XSCALE (Kabsch, 2010). Initial attempts to solve the structure by molecular replacement (MR) with Phaser (McCoy et al., 2007) in Phenix (Liebschner et al., 2019) using the first two domains of the V. cholerae multidomain LPMO GbpA individually (PDB entry 2xwx; Wong et al., 2012) provided strong rotation-function solutions but no significant translation-function outputs. The amino-acid sequence of mature CbpD was then submitted to the Robetta, Phyre² (Kelley et al., 2015) and PSIPRED (Buchan & Jones, 2019) servers and eight models were generated: five ab initio models from RoseTTA­Fold (Baek et al., 2021), one homology model from Phyre² and two ab initio models from DMPFold (Greener et al., 2019). MR was attempted using each CbpD model as a template. While no full-length model was able to generate a high-confidence solution, we chose the model with the highest LLG and TFZ scores (47 and 9, respectively) to break into domains. The AA10 domain of this RoseTTAFold model provided a partial solution (LLG, 181; TFZ, 18), which was held fixed in a subsequent round of MR using the RoseTTA­Fold GbpA2 output as the search model. This generated a successful MR solution (LLG, 432; TFZ, 22) containing the AA10 and GbpA2 domains of CbpD. An ensemble of CBM73 domains, including models generated by RoseTTAFold, Raptor-X (Askarian et al., 2021) and the NMR solution structure of the Cellvibrio japonicus CBM73 domain (Madland et al., 2021), was then used in subsequent rounds of MR, holding the AA10/GbpA2 partial solution fixed. Despite numerous attempts using varied input parameters, only low-likelihood solutions were found for the CBM73 domain and upon visual inspection were unlikely due to clashes with either the partial model of CbpD or symmetry mates.

The two-domain model of CbpD was rebuilt using phenix.autobuild (Liebschner et al., 2019) to create the linker region between the AA10 and GbpA2 domains. The electron density was good enough to manually extend the model past the GbpA2 domain to residue 296 in Coot (Emsley et al., 2010), which was used with phenix.refine (Liebschner et al., 2019) for iterative rounds of real-space and reciprocal-space refinement. The refinements were performed with a high-resolution cutoff of 3 Å. Atomic displacement parameters were refined as group factors until the final round of refinement, when they were refined using one translation–libration–screw (TLS) group (Urzhumtsev et al., 2013). The final model refined to R_cryst and R_free factors of 0.21 and 0.25, respectively (Table 1). Structure determination was performed using software supported by SBGrid (Morin et al., 2013).

Table 1
Data-collection, processing and refinement statistics

Values in parentheses are for the highest resolution bin.

Data-collection statistics
Diffraction source	PROXIMA-1, SOLEIL
Wavelength (Å)	1.0087
Temperature (K)	100
Detector	Dectris PILATUS 6M
Space group	P3112
a, b, c (Å)	85.77, 85.77, 90.86
α, β, γ (°)	90, 90, 120
Resolution range (Å)	42.89–2.90 (3.00–2.90)
Total No. of reflections	97198
No. of unique reflections	8654 (855)
Completeness (%)	99.9 (99.9)
Multiplicity	11.2
〈I/σ(I)〉	11.9
Wilson B factor (Å^2)	78.9
Matthews coefficient (Å^3 Da^−1)	0.47
Refinement statistics
Resolution range (Å)	38.78–3.00 (3.19–3.00)
Completeness (%)	99.9
No. of reflections, working set	7041 (1157)
No. of reflections, test set	781 (130)
Final Rwork/Rfree	0.21/0.25 (0.35/0.40)
No. of atoms
Protein	2215
Ligand	15
Waters	5
R.m.s.d., bond lengths (Å)	0.003
R.m.s.d., angles (°)	0.57
Average B factor (Å^2)	66.0
Ramachandran plot (%)
Most favored	93.5
Allowed	6.5
Outliers	0

2.3. Modeling and docking

The orientation of CbpD on chitin could confidently be predicted because all chitin-active LPMOAA10s cleave at C1; thus, the reducing end of chitin must align with Glu171 and the nonreducing end must align with Tyr40/Glu20 (for an in-depth review of LPMO oxidation products, regioselectivity and substrate orientation, see Bissaro et al., 2018).

Due to the lack of traceable electron density for the CBM73 domain, we evaluated the possible packing of this domain in the CbpD crystal lattice using the RoseTTAFold model of the CbpD CBM73 domain (Baek et al., 2021). This model was first manually docked into the crystal voids using Coot (Emsley et al., 2010) to minimize clashes with symmetry mates. A surface-model blob of each CBM73 domain was then generated in PyMOL (version 2.5; Schrödinger) by setting the B factor for all atoms to 100 Ų, setting the Gaussian resolution to 8.0, generating a Gaussian map for each CBM73 domain (grid = 1, buffer = 6) and finally generating an isosurface for each CBM73 domain (surface quality = 1).

To assess the fit of structural models to experimental small-angle scattering envelopes, a 15 Å volume map was generated in ChimeraX (Pettersen et al., 2021) by converting the envelope bead model of CbpD (Askarian et al., 2021; SASBDB entry SASDK42) into a density map using the molmap command. CbpD models were then manually placed into the density map and their positions were refined in the density map using the `Fit in Map' tool in ChimeraX.

3D structure figures were generated in PyMOL. Structure-based topology was depicted through the use of Pro-origami (Stivala et al., 2011).

3. Results

Models generated from newly introduced protein structure-prediction AI algorithms enabled us to solve the structure of CbpD using molecular replacement, which had not been successful using the representative domain structures shown in Fig. 1. The structure of mature CbpD (residues 1–296, where 1 is the N-terminal His after signal peptide cleavage) was determined to 3.0 Å resolution with a single molecule in the asymmetric unit (Table 1). The overall structure contains the AA10 and GbpA2 domains, a well ordered random-coil linker between them, and similarly the portion of the linker between the GbpA2 and CBM73 domains to residue 296 (Figs. 2a and 2b). Residues 297–374 (the remainder of the linker and the CBM73 domain) were not visible in the electron-density map. We compared the AI-predicted structure with the final CbpD structure and found an all-atom r.m.s.d. of 11.3 Å (2215 atoms). We then compared each domain and found all-atom r.m.s.d.s of 1.5 Å (1370 atoms) and 1.6 Å (733 atoms) for the AA10 and GbpA2 domains, respectively (Fig. 2c), compared with 8.9 Å (874 atoms) and 4.6 Å (569 atoms) for the refined CbpD domains and the representative domains in GbpA. In hindsight, it is noteworthy that a fern insecticidal LPMOAA10 (Yadav et al., 2019; PDB entry 6if7) has a very low r.m.s.d. of 1.5 Å over 1252 atoms when compared with the CbpD AA10 domain, and is the most structurally similar to CbpD of those LPMOAA10s for which structures are now available (Supplementary Fig. S2). This structure was not yet available when our X-ray data were collected, but would have proved useful in phase determination by molecular replacement. MR with PDB entry 6if7 and the RoseTTAFold GbpA2 domain generated a successful MR solution containing the AA10 and GbpA2 domains of CbpD that had better statistics (LLG, 489; TFZ, 23) than MR with the RoseTTAFold domain predictions alone.

Figure 2 The overall structure of CbpD. (a) The structure of CbpD contains an AA10 domain (green) with an L2 region (slate) between β1 and β2 connected to a GbpA2 domain (yellow) by flexible linker regions (gray). Disulfide bridges and the active-site His residues are shown as sticks. (b) The structure-based topology diagram for CbpD was initially generated with Pro-origami (Stivala et al., 2011), and the long intervening stretches were shortened manually and are not to scale. (c) A comparison of the AA10 and GbpA domains from the CbpD structure and predicted structural domain models from RoseTTAFold (Baek et al., 2021; pink) reveals a high degree of predictive accuracy. The all-atom r.m.s.d.s between the predicted and crystallographic AA10 and GbpA2 domains are 1.5 and 1.6 Å, respectively.

3.1. CbpD AA10 domain

The AA10 domain contains a typical LPMOAA10 fold, with a central seven-stranded β-sandwich comprising two antiparallel β-sheets connected by loop regions and an L2 subdomain between β₁ and β₂ (Figs. 2a and 2b). This L2 region forms the majority of the substrate-binding surface and displays a majority of the sequence variation in LPMOAA10s (Wu et al., 2013; Forsberg et al., 2016; Bissaro et al., 2017; Vaaje-Kolstad et al., 2017). The AA10 domain also contains three disulfide bonds: one linking β₄ to β₅ (Cys135 and Cys151), one linking two helices in the L2 domain (Cys14 and Cys27) and one linking the L2 domain to the β-sandwich through β₇ (Cys64 and Cys179).

3.1.1. Primary coordination sphere of the copper-binding active site

The active site comprises the canonical histidine brace found in LPMOs and is formed by His1 and His104 (Fig. 3a). Although the active site did not appear to contain a coordinated metal, a copper could be modeled into the active site with the electron density contoured to 0.7σ (Fig. 3a). Missing or partially occupied active-site copper is common in LPMOAA10 structures (Frandsen & Lo Leggio, 2016). In this position, the copper would only be coordinated axially in a T-shape by the histidine brace, as expected in an LPMOAA10 (Forsberg, Røhr et al., 2014).

Figure 3 The structure of CbpD reveals that some chitin-specific motifs are preserved while others are more similar to cellulose-specific motifs. (a) The active site of CbpD contains a canonical His-brace motif. 2Fo − Fc electron density contoured at 0.7σ allows copper to be modeled into the active site (sphere_scale 0.6), approximately 2 Å away from the three coordinating N atoms, as expected. (b) The CbpD active site also contains a number of highly conserved residues. (c) Motifs previously identified as conserved among and/or proposed to confer substrate specificity in chitin- and cellulose-active LPMOAA10s are mapped to sequence logos generated from multiple sequence alignments of 13 chitin-active and seven cellulose-active LPMOAA10s listed as experimentally validated in the CAZy database (Drula et al., 2022) and in Zhou & Zhu (2020), with the addition of CbpD to the chitin-active sequences. Motifs were concatenated for display and residues are numbered as in CbpD for the chitin-active sequence logo. In the cellulose-active sequence logo, residues are numbered by the structurally equivalent position in CbpD. The CbpD amino-acid identities are below the chitin logo. MSAs were generated with the T-Coffee web server (Notredame et al., 2000; Di Tommaso et al., 2011). Sequence logos were generated with WebLogo (Crooks et al., 2004).

As in all LPMOAA10s, a secondary coordination sphere includes highly conserved features that help to structure the active site of CbpD (Fig. 3b). These include a phenylalanine (Phe176; Hemsworth, Davies et al., 2013) and an alanine (Ala102; Forsberg, Røhr et al., 2014) in the two axial positions relative to the copper. These are believed to restrict the coordination of cosubstrates, with the alanine in particular blocking the solvent-exposed axial position, and thereby increase the likelihood that an oxygen cosubstrate coordinated at the equatorial position will be activated for catalysis, thus conferring C1/C4 oxidizing specificity (Forsberg, Mac­kenzie et al., 2014; Borisova et al., 2015). A well conserved tryptophan (Trp167) among LPMOAA10s (Forsberg, Røhr et al., 2014; Meier et al., 2018) is thought to protect the active site against oxidative inactivation during uncoupled turnovers (Loose et al., 2018; Paradisi et al., 2019; Gray & Winkler, 2021).

3.1.2. Second shell of the copper-binding active site

In addition to the histidine brace and the characteristic LPMOAA10 residues described above, the CbpD active site contains three highly conserved features of chitin-specific LPMOAA10s. Firstly, in CbpD the reportedly invariant TAXH chitin-recognition motif (Bissaro et al., 2018) is SAPH, with the conservative change of serine in place of threonine (A and H are the active-site Ala102 and histidine-brace His104, respectively; Figs. 3b and 3c). By interacting with Asn43 in the L2 loop, this serine secondly serves as one of the two amino acids in the `Gln–Thr pair' previously identified as conserved among chitin-binding LPMOAA10s (Zhou & Zhu, 2020; Figs. 3b and 3c). In other words, the `Gln–Thr pair' in CbpD becomes an Asn–Ser pair. It provides a polar staple from the L2 subdomain to the β-sandwich domain. In cellulose-active AA10s this pair is replaced by two large, hydrophobic residues, generally Phe–Trp (Forsberg et al., 2016). The important role that the residues in the Gln–Thr pair and TAXH motifs play in substrate specificity have further been demonstrated by directed mutation experiments conferring chitinolytic activity to a cellulose-active LPMOAA10 (ScLPMO10C; Jensen et al., 2019). Thirdly, previous molecular modeling of a chitin-active LPMOAA10 bound to crystalline chitin identified a Glu–Asn pair (Glu60–Asn185 in SmAA10A) that appears to gate a cosubstrate tunnel from the bulk solvent to the copper active site when an LPMOAA10 is bound to crystalline chitin (Fig. 3c; Bissaro et al., 2018). This gated tunnel is proposed to be the route through which only small-molecule cosubstrates (for example O₂, , H₂O₂ or H₂O) could access the active site. All LPMOs have the Glu (sometimes Gln) which is required for enzymatic activity (Vaaje-Kolstad, Horn et al., 2005; Harris et al., 2010), likely by coordinating H₂O₂ and reactive oxygen species intermediates (Span & Marletta, 2015; Bacik et al., 2017; Hedegård & Ryde, 2017; O'Dell et al., 2017; Bissaro et al., 2020). CbpD contains an analogous amino-acid pair Asn46–Glu174, but notably the two amino acids are swapped (Figs. 3b and 3c), a feature that is also present in CjLPMO10A (Forsberg et al., 2016). The swap retains the properties and position of the gating pair in the AA10 domain. Glu174 still points towards the active site of CbpD, approximately 5 Å away from the copper (Figs. 3 and 4a). The maintenance of the polar interaction between the side chains and the proximity of Glu174 to the active site imply that Glu174/Asn46 may also gate the access of cosubstrates or electron donors to the active site when CbpD is bound to crystalline chitin. We also note that this amino-acid pair also serves to stabilize the L2/ β-sandwich interface. In these three chitin-recognition motifs, side-chain properties are conserved.

Figure 4 The chitin-binding surface of CbpD. (a) Looking at the chitin-binding surface from the perspective of the crystalline substrate, equivalent residues to those identified in SmAA10 that interact with chitin (Vaaje-Kolstad, Houston et al., 2005; Aachmann et al., 2012; Bissaro et al., 2018) are labeled. For reference and orientation, copper is modeled into the active-site His brace as a sphere (sphere_scale 0.6) and Phe176 is labeled and shown as sticks. Residues in the putative chitin-specific L2 motif are underlined. Ser172 and Tyr40, which are phosphorylated when CbpD is expressed and secreted by P. aeruginosa (Ouidir et al., 2014), are marked with an asterisk. (b) Side view of the chitin-binding surface showing the relatively flat substrate-binding surface of the CbpD AA10 domain and the large role that the L2 region (slate) plays in substrate binding. The substrate tunnel gating pair Asn46/Glu174 are indicated with a dashed oval. (c) Based on the modeling of SmAA10 interacting with β-chitin (Bissaro et al., 2018), CbpD was manually docked onto crystalline anhydrous β-chitin (Nishiyama et al., 2011; Crystallography Open Database entry 1501776) so that the substrate-binding surface of CbpD (O and N atoms of side chains in red and blue, respectively) was parallel to the chitin surface and aligned such that the chitin runs from one end of the surface (Glu171) across the active site to the opposite end (Tyr40/Glu20). A chito-octaose (NAG8) fragment interacting with subsites −6 to +2 is shown as sticks. Subsites are numbered following the standard practice for glycoside hydrolases (Sunna et al., 1997).

CbpD, however, deviates from or lacks two features previously postulated to be conserved among chitin-active LPMOAA10s. CbpD lacks the cluster of three tryptophans previously identified as being conserved among bacterial chitin-active LPMOAA10s (Beeson et al., 2015). Instead, CbpD has a trio of different aromatic amino acids: Tyr98, Phe109 and Trp167 (Fig. 3c). This pattern of three aromatic residues has been identified in the strictly cellulose-active ScLPMO10C (Forsberg, Røhr et al., 2014). CbpD also contains a reversed putative discriminating feature between chitin and cellulose-active LPMOAA10s. Chitin-active LPMOAA10s were identified to contain a short, aliphatic residue (generally an isoleucine or valine) just below the histidine brace, while this position was an Arg in the cellulose-active LPMOAA10s TfLPMO10B and ScLPMO10C (Forsberg, Mackenzie et al., 2014; Forsberg, Røhr et al., 2014) and is preserved among the strict cellulose-active LPMOAA10s in the CAZy database (Drula et al., 2022; Fig. 3c). In chitin-active CbpD, however, this position is Arg169 (Fig. 3b). It has been postulated that the short aliphatic residue in this position creates a cavity on the substrate-binding surface of chitin-active LPMOAA10s that could accommodate a dioxygen cosubstrate for catalysis (Hemsworth, Taylor et al., 2013) or the N-acetyl group found in chitin (Forsberg, Mackenzie et al., 2014). Arg169, Asp171 and Glu174 eliminate this cavity on the surface of CbpD (Fig. 4a). This is also the case for chitin-active CjLPMO10A (Forsberg et al., 2016). In fact, in 16.7% of reviewed chitin-active LPMOAA10s in the CAZy database (Drula et al., 2022) this position is an Arg (Supplementary Fig. S3). The Arg here as opposed to Ala or another short aliphatic residue implies that a surface cavity may not be required for catalytic activity or substrate discrimination.

3.1.3. Chitin-binding surface

The structure of CbpD reveals a canonical relatively flat surface of ∼1000 Ų encompassing the copper active site that is the presumed chitin-binding surface of CbpD (Figs. 4a and 4b). While no studies have directly examined the role that residues on this surface play in mediating chitin binding by CbpD, the structure of CbpD enables the mapping of homologous chitin-interacting residues from other characterized LPMOAA10s to CbpD. Residues that interact with chitin in chitin-specific LPMOAA10s have been identified through mutagenesis (Vaaje-Kolstad, Houston et al., 2005), NMR studies (Aachmann et al., 2012) and molecular-dynamics (MD) simulations (Bissaro et al., 2018). The analogous putative chitin-binding residues of CbpD could be identified by mapping these sets of residues onto CbpD (Tyr40, Asp41 and His104 from all three methods, Asn46, Ser172 and Glu174 from mutagenesis and MD, Asn43, Gly44, Ser101 and Ala102 from NMR and MD, Leu39, Ala100 and Thr106 from NMR, and Pro173 from MD; Figs. 4a and 4b). These residues form a surface that extends from the conserved single aromatic residue (Tyr40) found on the substrate-binding surface of chitin-active LPMOAA10s (Zhou & Zhu, 2020) to Asp171/Ser172 at the edge of the substrate-binding surface. The structure suggests that Glu20 may further extend the substrate surface beyond Tyr40 in CbpD.

Interestingly, as for the sequence motifs examined above, mapping the features of chitin-specific LPMOAA10 substrate surfaces to CbpD revealed deviations in CbpD from previously identified trends. A motif on the L2 domain, beginning with the conserved substrate-binding Tyr and running down the center of the substrate-binding surface, purportedly exhibits differing polar characteristics in chitin- and cellulose-specific LPMOAA10s (Zhou et al., 2019). In the chitin-specific LPMOAA10s examined this motif contained at least 70% polar residues, with the consensus sequence Y(W)EPQSVE (Fig. 3c). The cellulose-specific LPMOAA10s examined had a motif containing more than 70% hydrophobic residues. In CbpD, this L2 motif is YDWNGVN, beginning with the conserved Tyr40. This sequence differs from the consensus chitin-specific sequence, and the side-chain properties fall between the properties of the chitin- and cellulose-specific motifs (Fig. 3c). This motif in CbpD still forms a flat surface that is an appropriate area for substrate binding (Fig. 4b), and the side-chain characteristics of position 2 in the consensus sequence are conserved. This residue may be particularly important for chitin recognition and enzymatic activity because mutating this residue significantly reduces chitin binding (Vaaje-Kolstad, Horn et al., 2005) and eliminates enzymatic activity (Loose et al., 2018) in another chitin-active LPMOAA10 (SmLPMO10A). The divergence of most L2 motif residues in CbpD from previous trends raises questions about the extent to which the entire motif in LPMOAA10s determines substrate specificity. Interestingly, the chitin-active insecticidal putative LPMOAA10 produced by the fern Tectaria macrodonta (Tma12) has the sequence YEWNEVN that closely matches the outlier CbpD sequence here, providing further evidence that the Y(W)EPQSVE motif is not the determinant of chitin specificity.

3.2. CbpD GbpA2 domain

The second domain of CbpD is, as expected, similar to the GbpA2 domain structure in GbpA (Wong et al., 2012), with an all-atom r.m.s.d. of 4.6 Å for 569 atoms (residues 193–292 in CbpD and 211–314 in GbpA). The GbpA2 domain comprises an α-helix partially sandwiched between two antiparallel β-sheets (β-strands β₉, β₁₀ and β₁₅ and β-strands β₈, β₁₁ and β₁₄), with an unusual bend that allows β₁₁ to pair with β₁₄ while its immediate extension β₁₂ interacts with the short β₁₃ to form a hairpin above the helix (Figs. 2a and 2b). The orientation of the GbpA2 domain relative to the AA10 domain is shifted in the structure of CbpD by 1 Å and 135.6° compared with the structure of GbpA (Fig. 5a). This shift means the CbpD crystal structure is closer to the elongated conformation observed for both GbpA and CbpD by solution scattering (Wong et al., 2012; Askarian et al., 2021; Supplementary Fig. S4). In contrast, GbpA adopts a U-shaped conformation in its crystal, forming an extensive interface in which each subunit contributes 3120 Ų to the dimerization interface, as calculated with PISA (Krissinel & Henrick, 2007). Notably, while CbpD does not form a similar dimer, it does have a crystal contact formed by continuation of the β-sheet from β₁₅ of one CbpD GbpA2 domain into its symmetry mate, burying 990 Ų per monomer (Fig. 5b).

Figure 5 The role of the GbpA2 domain in crystal packing. (a) The GbpA2 domain in the CbpD structure (yellow cylinder) is displaced by 1 Å and rotated by 135.6° relative to the GbpA2 domain in the GbpA structure (light blue cylinder) when their AA10 domains are aligned. The αG of each GbpA2 domain (smaller cylinders), the β12–β13 hairpins (stylized cartoons) and the C-termini are indicated for orientation. The AA10 domain of CbpD is represented as in Fig. 2(a), with copper modeled into the active site as a sphere (sphere_scale 0.6). The displacement of GbpA2 domains is also noted by the ND2 label at the shifted N-terminus of this domain in GbpA. (b) β15 and part of the linker between the GbpA2 and CBM73 domains of one CbpD (yellow and dark gray) engage in extensive β-strand interactions with a symmetry mate (orange and light gray) to form an extended β-sheet, burying 990 Å^2 of surface area per monomer.

3.3. CbpD CBM73 domain

While full-length CbpD-His₁₀ was crystallized, good electron density was only observed to residue 296 (Fig. 6a), leaving nine residues of the linker between the GbpA2 and CBM73 domains, as well as the CBM73 domain itself, un­modeled. Despite the absence of strong density to place the CBM73 domain, it is unlikely that the residues were missing from the protein crystal. Given that the His₁₀ affinity tag was attached to the C-terminus of CbpD, it is clear that full-length CbpD was purified via metal-affinity chromatography (see Section 2 and Supplementary Fig. S1). Crystal solvent-content analysis with phenix.xtriage (Zwart et al., 2005) calculated 46% solvent content for a unit cell containing 374 residues, consistent with the crystallization of full-length CbpD-His₁₀. While there is some evidence that CbpD is processed from its C-terminus by LasB into the 23 kDa LasD when secreted by P. aeruginosa (Braun et al., 1998), CbpD-His₁₀ was expressed in E. coli, which lacks LasB. The processing of CbpD into LasD would, moreover, result in the loss of all but the first 24 residues of the GbpA2 domain, which is fully present in the crystal structure. Additionally, there is a sufficiently large volume remaining in the crystal packing of CbpD to arrange the CBM73 domain (Fig. 6b), although in a more compact configuration than the fully elongated configuration deduced from SAXS data (Askarian et al., 2021).

Figure 6 The CBM73 domain is present but unresolved in the CbpD crystal structure. (a) The 2Fo − Fc electron-density map is contoured at 1.0σ. The Cα trace of the CbpD GbpA2 domain is shown (yellow and gray with residues 287–296 as sticks) with a symmetry mate (orange). The electron density extends to residue 296. (b) Despite the absence of strong density to place the CBM73 domain, there is space in the crystal packing where the domain fits. Looking roughly down the c axis of the unit cell, the left panel depicts one unit cell of the modeled residues of CbpD (ribbons), with each symmetry mate outlined for clarity and reference in the right panel (arbitrary colors). The right panel depicts how the CBM73 domain (red blobs) could pack into the remaining crystal volume.

4. Discussion

The structure of CbpD reported here is the first structurally determined LPMO from P. aeruginosa and highlights the utility that new, AI-powered, ab initio protein structure-prediction algorithms have in the structural biology toolkit. As demonstrated with CbpD, these ab initio models may help to solve previously recalcitrant X-ray crystallographic structures through MR by providing starting models that are substantially closer to the actual protein structure than homologous structures. Because these AI-generated models may provide better MR models, they may also enable protein structures to be solved with lower resolution or lower quality data than was previously routine. Additionally, these models will provide another path towards phasing protein structure targets without the need for additional experimental steps such as selenomethionine labeling, the collection of multiple data sets or heavy-metal soaking (which is also a danger to human and environmental health). These AI-generated models could be particularly beneficial for crystal targets that are difficult to express and purify in large quantities, produce few high-quality diffracting crystals or lack closely homologous structures that otherwise could serve as the basis for MR. The close agreement between AI-generated ab initio models and the crystal structures of each domain of CbpD, specifically the AA10 catalytic domain, also indicates that these AI results may provide high-quality structural models for the thousands of LPMOs that have been identified but lack experimental structures to develop more robust hypotheses about the characteristics of LPMOs that affect catalysis and substrate selectivity.

The structure of CbpD provides greater insight into the conserved features of chitin-active LPMOAA10s. CbpD displays a number of motifs (Glu–Thr, TAXH and the substrate tunnel gating pair) that demonstrate conservation of the structural properties of chitin-active LPMOAA10s despite deviations in motif sequences (Fig. 3c). The similarity of Tyr40, Asp41, Asn43 and Ser101 in the Gln–Thr pair and TAXH motifs to the chitinolytic sequence identified by Jensen et al. (2019) indicates that these motifs near the active site of CbpD are likely to contribute to its chitin specificity. The structure of CbpD reinforces the spatially conserved nature of the Glu residue in the active site that is oriented towards the His brace irrespective of its occurrence in the amino-acid sequence, in line with observations for other LPMOAA10s (O'Dell et al., 2017; Forsberg et al., 2019). While Glu174 and Asn46 are flipped compared with the Glu60/Asn185 pair in SmAA10A, the preservation of the structural position of the pair in the AA10 domain indicates that Glu174/Asn46 may play a similar role in gating the access of small-molecule cosubstrates such as O₂ or H₂O₂ to the active site when CbpD is bound to crystalline chitin as Glu60/Asn185 in SmAA10A (Bissaro et al., 2018). Notably, a previous study found that mutating Glu60 to Asn was deleterious to catalytic activity (Bissaro et al., 2020). This study, however, did not examine the effect of a concomitant mutation to Glu of the other residue in the Glu/Asn pair, which likely explains why CbpD retains chitinolytic activity despite having an apparently deleterious Asn residue at position 46 (equivalent to Glu60) and highlights the need for careful spatial as well as sequence considerations when conducting enzymatic mutation studies.

Other features, however, deviate substantially from previously characterized LPMOAA10s (Fig. 3c). In the proximity of the active site, the presence of Arg169 and the absence of a pocket on the substrate surface is noteworthy because previously an Arg in this position was associated with only cellulose activity in LPMOAA10s (Forsberg, Røhr et al., 2014). Interestingly, an Arg is present in this position in two other chitin-active LPMOAA10s (Tma12 and CjLPMO10A; Supplementary Fig. S3), and all three are structurally similar to the cellulose-active LPMOAA10 ScLPMO10C based on a DALI server `all against all' comparison (Holm, 2020) and are the only members of their structural class to lack cellulolytic activity (Supplementary Fig. S2). This raises the possibility that there are distinct yet convergent features of LPMOAA10s that have evolved to confer chitin-specific activity and indicates that these features may not all have been identified. Interestingly, CbpD may have a shallow pocket just to the side of the active site that may be able to accommodate an N-acetyl group (Fig. 4c), although it is between the putative channel gating resides (Glu174/Asn46) and the copper active site and may be part of the substrate tunnel when CbpD is bound to chitin. The structures of CbpD, Tma12 and CjLPMP10A now provide three examples of LPMOAA10s that are strictly active on chitin, yet lack the pocket previously thought to facilitate either substrate hydrolysis by accommodating the oxygen cosubstrate or substrate binding by accommodating the N-acetyl group present in chitin, and should inform future investigations into the deterministic features that confer substrate specificity in LPMOAA10s.

The cluster of aromatic residues Tyr98, Phe109 and Trp167 also deviate from the previously identified three conserved Trp residues in chitin-active LPMOAA10s (Beeson et al., 2015) and instead match the pattern found in cellulose-active SmLPMO10C (Forsberg, Røhr et al., 2014). This, combined with their location away from the putative substrate-binding surface, indicates that the Trp trio is likely not to be a determinant in substrate specificity and may rather play a role in stabilizing the hydrophobic core of the domain. These aromatic residues near the active site may also be important in protecting CbpD from oxidative damage (Paradisi et al., 2019). It remains to be seen whether this protective mechanism is the reason that aromatic residues appear to be conserved in general, if not with strict identity, near the active site of LPMOs.

A Trp trio in LPMOs was previously suggested to be important for a hypothesized electron-transfer (ET) catalytic mechanism based on O₂ as the oxygen cosubstrate (Vaaje-Kolstad et al., 2012; Forsberg, Røhr et al., 2014). This ET path was putatively identified in fungal LPMOs (Li et al., 2012; Beeson et al., 2012; Book et al., 2014; Lo Leggio et al., 2015; Frandsen et al., 2016), but evidence has since been presented that cellobiose dehydrogenase (CDH) actually binds fungal LPMOAA9 active sites directly as an external electron donor, abrogating the need for an ET path in LPMOAA9s (Courtade et al., 2016; Kracher et al., 2016). An ET mechanism is also unlikely in bacterial LPMOAA10s, including CbpD. An equivalent ET pathway was not identified in bacterial chitin-active LPMOs (Vaaje-Kolstad et al., 2012), although the concept has been revisited in recent literature based on an LMPOAA10 with two copper sites (Fowler et al., 2019). The requirement that electrons would traverse a conserved Phe (Phe176), which does not facilitate ET (Beeson et al., 2015), and the presence of gating residues similar to SmAA10A (Bissaro et al., 2018) indicates that there is a path into the active site of CbpD that is accessible to small-molecule reductants, eliminating the questions about how electrons could reach the copper active site for O₂-based catalysis when bound to a crystalline substrate that originally led to the hypothesis of the ET mechanism (Li et al., 2012; Walton & Davies, 2016). The lack of an identified bacterial CDH (Beeson et al., 2015) and the identification of a number of alternative electron-donor sources, including small molecules (Westereng et al., 2015, 2016; Askarian et al., 2021), photo-active pigments (Cannella et al., 2016) and even substrates themselves (Westereng et al., 2015), rule out that bacterial LPMOAA10s rely on an internal ET path if they utilize O₂ as a cosubstrate.

What remains unclear, however, is whether LPMOs may have reducing-agent specificity in addition to substrate specificity (Frommhagen et al., 2016; Meier et al., 2018). Askarian and coworkers demonstrated that CbpD chitinolytic activity was enhanced by ascorbate and the redox-active compound pyocyanin, which is secreted by P. aeruginosa (Askarian et al., 2021), and Branch and coworkers demonstrated that a c-type cytochrome can activate the cellulose-active CjAA10B in Cellvibrio japonicus (Branch et al., 2021). CbpD and other LPMOs may favor reductants that are secreted by their hosts.

Future mechanistic studies can use the structure of CbpD to investigate whether the proposed mechanisms are universal and could occur in structurally diverse LPMOs with the same and divergent substrates, or whether the mechanisms may be family dependent, and whether LPMOs utilize general or context-specific reductants if an H₂O₂-based mechanism is employed.

The variation of the chitin-specific L2 motif Y(W)EPQSVE on the substrate-binding surface of CbpD further challenges previously strong conclusions about the deterministic features of LPMOAA10s that confer substrate selectivity. It is unclear whether the L2 motif is more tolerant of sequence variation than first supposed (Zhou et al., 2019) or whether the L2 motif plays a critical role in substrate specificity among LPMOAA10s. These discrepancies in the active-site secondary coordination sphere and substrate-binding surface also raise an additional question about whether CbpD may be able to bind and hydrolyze substrates beyond chitin.

Many previous structural studies (Forsberg, Mackenzie et al., 2014; Forsberg, Røhr et al., 2014; Zhou et al., 2019; Zhou & Zhu, 2020) drew their conclusions about discriminatory features from a relatively small number (tens) of evolutionarily closely related LPMO sequences. The structure of CbpD highlights the importance of including an evolutionarily diverse array of LPMOs in the identification and analyses of substrate-specificity motifs. This is particularly true given that the bacterial LPMOAA10 CbpD is structurally most similar to the fern LPMOAA10 Tma12 (Supplementary Fig. S2). The presence of the same L2 motif sequence from CbpD in the structurally similar yet evolutionarily distant fern LPMOAA10 Tma12 demonstrates that the structural features of LPMOAA10s that enable chitin binding and degradation have yet to be fully understood and require further investigation among structurally diverse members of the AA10 family.

Even with the structures of CbpD, Tma12 and CjLPMO10A challenging previously identified substrate-specificity determinants, there are still relatively few LPMOAA10s that have experimentally verified substrate specificity. To date, the CAZy database contains over 8000 LPMOAA10 proteins, yet only 32 are listed as `characterized' and this list does not include CbpD, despite its substrate binding being characterized in 2000 (Folders et al., 2000) and its catalytic activity profile being confirmed in 2021 (Askarian et al., 2021). One characterized LPMOAA10, KpLPMO10A, appears to be able to cleave chitin, cellulose and xylan (Corrêa et al., 2019), expanding the substrate scope of LPMOAA10s beyond just chitin and cellulose, yet no studies have investigated the structural features that may enable this broader substrate scope in an LPMOAA10. Until a larger portion of the LPMOAA10 family has characterized substrate profiles, sequence or structural motifs and features identified to confer absolute substrate specificity must be considered provisional.

Because the catalytic activity of CbpD is necessary for P. aeruginosa virulence (Askarian et al., 2021), discrepancies in putative substrate-specificity features highlight the importance of considering the role of LPMOs in both carbohydrate polymer degradation and virulence when identifying the residues and motifs that confer substrate specificity. Humans do not make the chitin polymer, but do produce highly glycosylated mucin proteins (Corfield, 2015). Mucins play a critical role in cystic fibrosis (Morrison et al., 2019), and CbpD may facilitate P. aeruginosa infection in these patients. Moreover, V. cholerae GbpA, which shares an LPMOAA10 domain and the GbpA2 domain with CbpD, is a demonstrated mucin-binding protein (Bhowmick et al., 2008) which relies on its LPMOAA10 domain for mucin binding (Wong et al., 2012). While mucin glycans are comprised of diverse saccharides and glycosidic bonds, they do contain β1–4 linkages involving N-acetyllactosamine and may imperfectly mimic chitin. Askarian and coworkers found that CbpD does not bind individual glycan structures within mucin glycans, and CbpD may only be able to imperfectly bind full mucin glycans. An alternative biological role for CbpD might be to enable bacteria–fungi interactions in the polymicrobial biofilms and co-infections that P. aeruginosa forms with pathogenic fungi, including Aspergillus fumigatus (Zhao & Yu, 2018) and Candida albicans (Harriott & Noverr, 2011), as fungal cell walls are chitinous.

The presence of a few discrepancies in previously identified substrate-specificity motifs coupled with its role in virulence mean that CbpD should serve as a basis for more extensive modeling of the interaction of LPMOAA10s with chitin. LPMOAA10–chitin modeling to date has only been performed with single-domain LPMOAA10s (Bissaro et al., 2018). Additionally, this modeling was performed without the consideration of post-translational modifications (PTMs). The structure of CbpD, however, reveals that two residues that are phosphorylated when CbpD is secreted from P. aeruginosa (Ouidir et al., 2014), Ser172 and Tyr40, form part of the chitin-binding surface of CbpD (Fig. 4), and in the case of Tyr40 are critical for substrate binding (Beeson et al., 2015; Bissaro et al., 2018). It has been speculated that phosphorylation at Ser172 proximal to Glu174 in the secondary coordination sphere of the copper active site is likely to affect cosubstrate (O₂ or H₂O₂) activation and may alter the copper redox potential (Askarian et al., 2021). A number of these post-transitionally modified residues are also found in the AA10–GbpA2 domain interface of CbpD, likely encouraging a more elongated overall conformation, similar to the observed solution scattering conformation. Additionally, a recent bioinformatics analysis of 27 000 LPMOs found that most of the eight LPMO AA families are enriched in Ser and Thr residues in C-terminal regions that are prime candidates for PTM, specifically O-glycosylation (Tamburrini et al., 2021), which may affect secretion (Vorkapic et al., 2019). Of note, the Tma12 structure has an N-glycosylation at Asn158, structurally equivalent to Ser136 in CbpD, which has been identified as phosphorylated when CbpD is expressed and secreted by P. aeruginosa (Ouidir et al., 2014). A complete understanding of how LMPOAA10s are secreted by their native organisms and bind to their substrates cannot be obtained without taking these PTMs into consideration.

The β-strand interaction observed between the GbpA2 domains of crystallographic symmetry mates raises another question about the possible biological function of structural features of CbpD. Edge β-strands are well known to mediate protein–protein interactions, which in some cases are dynamic and interchangeable (Richardson & Richardson, 2002; Monteiro et al., 2013; Yu et al., 2014; Miyazaki et al., 2021). Given the broad substrate repertoire of the P. aeruginosa T2SS (Cianciotto & White, 2017) and the expectation that the recognition of secreted proteins by multiple proteins along the secretion-machinery pathway relies on dynamic, short-lived interactions (Douzi et al., 2011; Michel-Souzy et al., 2018) as well as inherent disorder (Gu et al., 2017; Pineau et al., 2021), it is interesting to imagine that the β₁₅ edge strand of CbpD may serve as a handle for interaction with secretion-machinery proteins along the pathway through the cell envelope. The CbpD structure may thus provide an experimental entrée into a better understanding of how T2SS substrates are recognized.

5. Conclusions

In addition to demonstrating the utility of new structural biology tools, the structure of CbpD provides evidence to both support and challenge aspects of the current model of how LPMOAA10s bind their substrates, specifically those that are preferential for chitin. The absence of the CBM73 domain in the structure also raises questions about the inherent stability of smaller carbohydrate-binding domains, and a crystal structure of a CBM73 domain remains to be solved. One possibility is to perform high-resolution, single-particle electron microscopy of CbpD bound to chitin fibrils. This could provide the structure of full-length CbpD and shed light on the conformation of CbpD when bound to crystalline chitin. A better understanding of the structure, conformations and dynamics of multimodular LPMOs could lead to advances in understanding the role of these enzymes in biomass degradation and virulence.

Footnotes

‡Current address: Université de Lorraine, INRAE, DynAMic, F-54000 Nancy, France.