Tangled Up in Fibers: How a Multidomain Lytic Polysaccharide Monooxygenase Binds Its Chitin Substrate

Abstract

Lytic polysaccharide monooxygenases (LPMOs) are redox enzymes that bind to and oxidize insoluble carbohydrate substrates such as chitin or cellulose. This class of enzymes has attracted considerable attention due to their ability to convert biomaterials of high abundance into oligosaccharides that can be useful for producing biofuels and bioplastics. However, processes at the interface between solution and insoluble substrates represent a major challenge to biochemical and structural characterization. Here, we investigated the four-domain LPMO from Vibrio cholerae, N-acetyl glucosamine binding protein A (GbpA), to elucidate how it docks onto its insoluble substrate with its two terminal domains. First, we developed a protocol that allowed GbpA and chitin to form a stable complex in suspension, overcoming incompatibilities of the two binding partners with respect to pH. After determining the small-angle neutron scattering (SANS) contrast match point for chitin (47% D₂O), we characterized the mesoscale structure of GbpA in complex with chitin by SANS and complemented the results with negative-stain electron microscopy. We found that GbpA binds rapidly to chitin, where it coats the chitin fibers and smooths their surface. In some locations, GbpA binding induces the formation of protein–chitin clumps containing a large number of GbpA molecules. Together, this suggests how the secretion of GbpA efficiently prepares the ground for microcolony formation by the bacteria.

Introduction

Since the discovery of lytic polysaccharide monooxygenases (LPMOs) in 2010, their ability to degrade crystalline cellulose, xylan, or chitin has drawn massive interest in converting renewable biomaterials to biofuels. LPMOs are found in a wide range of organisms, including bacteria, fungi, algae, and insects as well as viruses. , Some human pathogenic bacteria secrete LPMOs to facilitate their survival on carbohydrate surfaces, both within and outside the host. − These enzymes can be employed for various strategies, including direct interference with the immune system. Given the abundance of these enzymes and their sequence variation, it is likely that other important functions of these proteins are yet to be revealed. The discovery of LPMOs necessitated a reclassification of enzyme families in the Carbohydrate-Active enZYmes database (CAZY; http://www.cazy.org), where they are listed under Auxiliary Activities (AA; families AA9–11 and AA13–17). The first crystal structures of LPMO complexes with carbohydrate ligands (cellotriaose Glc₃ and cellohexaose Glc₆) were determined in 2016 by Frandsen et al. for an AA9 LPMO from Lentinus similis. At approximately the same time, Courtade et al. probed ligand binding to another AA9 LPMO, from Neurospora crassa, by nuclear magnetic resonance (NMR) spectroscopy. These and subsequent studies showed how the ligand stretches across the flat surface of the pyramidal LPMO structure, with the +1 sugar unit binding to the N-terminal histidine of the characteristic copper-coordinating histidine-brace motif present in the active site of LPMOs (Figure A). The catalytic mechanism of LPMOs and current controversies are summarized in the excellent reviews by Forsberg et al. and Munzone et al. However, not all LPMOs can bind to oligosaccharides, and despite accumulating structural and functional data, there is scarce structural information about the molecular interaction of LPMOs with their complex, recalcitrant carbohydrate substrates. In particular, the complexes between the proteins and larger carbohydrate substrates such as chitin and cellulose have not yet been structurally characterized with experimental methods, not even at low resolution.

1.

GbpA and schematic representation of SANS experiment. (A) The LPMO domain (D1; depicted as triangle) of GbpA binds to chitin and produces oxidative breaks in the fiber using its copper-containing histidine brace in the active site. Chitin has been manually modeled in its expected position based on the LPMO structure determined by Tandrup et al., taking into account information from Bissaro et al. and using the atomic coordinates for GbpA from PDB ID 2XWX. (B) Representation of SANS experiment with GbpA and chitin using contrast matching: (1) GbpA binds chitin with its first and fourth domains, undergoing a conformational change; (2) by applying mixtures of D₂O and H₂O, chitin can be matched out in the SANS experiment, leaving GbpA as the only visible scatterer. The two middle domains (D2 and D3) are known to bind to the bacteria.

We are interested in the virulence factors of Vibrio cholerae, the causative agent of cholera. Therefore, we selected N-acetylglucosamine binding protein A (GbpA) from V. cholerae as a model system to study its interaction with chitin, an insoluble homopolymer of N-acetylglucosamine (GlcNAc; for a review of chitin-active LPMOs, see Courtade and Aachmann). GbpA was discovered in 2005  and found to be important for colonization in the aquatic environment ,, as well as in the human host, binding to (and up-regulating) human mucins. , Recent studies further suggest that GbpA may interact with toll-like receptors on host cells and induce an immune response (through its fourth domain), , mediated by IL-8 secretion. GbpA consists of four domains, of which the first domain, which structurally resembles carbohydrate-binding module CBM33, is a chitin-degrading LPMO , with calcium-binding properties and additionally binds to mucins. The fourth domain (classified as a CBM73 in CAZy) also binds chitin, whereas domains 2 and 3, which structurally resemble flagellin and pilin-binding proteins, respectively, mediate attachment to the cell surface of the bacteria. Very recent data suggest that domain 3 can also interact with the reduced catalytic copper in the LPMO domain and that the presence of chitin and a reductant is required to fully activate the LPMO domain. The structure of the first three domains of GbpA has been solved by X-ray crystallography, recently complemented with the full-length crystal structure of a GbpA homologue from Vibrio campbellii. In addition, the solution-structure of GbpA has been characterized with small-angle X-ray scattering (SAXS) and small-angle neutron scattering (SANS), , revealing a monomeric elongated structure. None of the crystal structures contain carbohydrate ligands, and no structural data have been reported for the interaction between GbpA and chitin.

Here, we set out to reveal how GbpA interacts with chitin fibers using SANS. We characterized the bound and unbound GbpA structures by applying SANS to deuterated and nondeuterated GbpA (alone or in complex with chitin) and by varying the contrast using different D₂O/H₂O mixtures, as schematically shown in Figure B. Additional, complementary insights were obtained from negative-stain electron microscopy (EM).

Results

Development of a Protocol for Solution Studies of the GbpA–Chitin Complex

A suitable method for the structural characterization of individual partners in a protein complex is SANS. However, the insoluble nature of chitin poses a challenge for SANS. This is probably why chitin has rarely been used for solution scattering studies. It has, however, been shown that fine β-chitin nanofibers can be suspended in slightly acidic buffers (pH ∼ 3). For shorter periods of time, chitin can be kept at pH ∼ 5; however, this is still low for GbpA, which shows some precipitation at pH ≤ 5. Fortunately, we could overcome these challenges by adding small volumes of 10–12 mg/mL GbpA (at pH 8.0) to chitin nanofibers (at pH 5.0), followed by immediate thorough pipetting and (for SANS experiments) sonication for 15–30 s with a tip sonicator, being careful to avoid overheating, as described in the Experimental Section. We observed that the mixture immediately became much less viscous during this process. The samples treated in this manner were subsequently dialyzed into 20 mM sodium acetate–HAc pH 5.0 during a 24 h period. After this procedure, the chitin–GbpA complex could be kept in suspension for 12–24 h, sufficient for several SANS experiments. For the lengthier USANS measurements, rotating tumbler cells were used to keep the samples in suspension. In the case of delays, the complex was shortly resonicated directly before the experiments. We observed that the solubility of the chitin fibers increased after the addition of GbpA, indicating that a complex had indeed formed. It was even possible to adjust the pH of the complex mixture to higher pH values (∼7) without immediate precipitation.

Small-Angle Scattering of GbpA and Chitin

Before exploring the structure of the chitin–GbpA complex, we investigated the structures of the two individual components. The solution structure of GbpA had already previously been characterized by SAXS and SANS, , showing an elongated, flexible shape with a radius of gyration (R _g ) of 35–40 Å. Indeed, the scattering of GbpA cannot be described by a simple ellipsoid but requires at least five connected ellipsoids to replicate the scattering data well (Figure A). Even though GbpA is a four-domain protein, this is not entirely surprising since the first domain is approximately twice the size of the other domains and pyramidal in shape (not ellipsoidal).

2.

Independent scattering of GbpA and chitin. (A) SANS of D-GbpA in 47% D₂O (the chitin match point). The protein is fairly extended in solution; hence, describing the scattering as an ellipsoid (blue curve) gives a poor fit; a better fit can be achieved with 5 random-walk ellipsoids (red curve). (B) 3 mg/mL chitin was measured with SANS in D₂O. The data can be fitted with a flexible cylinder model including a power-law for the steep upturn at low-q. (C) Scattering of chitin in 47% D₂O is significantly weaker than in 100% D₂O, and is matched out in the mid to high q-range. However, at low-q, chitin could not be matched out perfectly. (D) Comparing the scattering of D-GbpA in 47% D₂O to chitin shows that the protein scatters very well in the mid to high q-range, where both chitin and H-GbpA are well matched out. Deuteration of the protein is thus essential to obtain sufficient contrast between the two components in an interaction study. Error bars represent one standard deviation.

Chitin fibers were modeled with a flexible cylinder model , based on previous SAXS and SANS images of chitin fibers showing cylindrical shapes with some bends. A similar approach has previously been used to model other nanofibers, such as proteins or cellulose; in the latter case, a slightly more complex variation was used, with an elliptical cross-section. , We found a circular cross-section to be adequate for chitin. To take into account chitin’s crystallinity and its sharp interfaces with the solvent, which give rise to a steep slope at low-q in SANS, we also included a power law in the model. The model fits the chitin data well (Figure B; χ² = 2.78). The decay exponent in the power law approximates 4, which is consistent with sharp interfaces that are impenetrable for the solvent. Furthermore, the chitin fibers show a radius of 16.0 ± 0.0 Å (Table S1), consistent with previous studies. , The model gives the Kuhn length (318 ± 12 Å; Table S1), which is twice the persistence length of the cylinder and thus relates to flexibility. A Kuhn length of 318 Å shows that chitin is very rigid. Both radius and Kuhn length are close to what has previously been reported for nanofibers of cellulose. The model can also yield the contour length (overall length of a fiber), but this parameter is too large to be meaningfully approximated from the q-range in SANS.

Chitin Can Be Matched Out for SANS Studies of GbpA–Chitin Complex

To obtain structural information on GbpA in its chitin-bound state with SANS, it is necessary to measure the sample at the chitin D₂O/H₂O match point. Online tools , suggested a match point around 44% D₂O for a chitin density of 1.5 g/mL, which is close to the theoretical match points of most globular proteins (∼40–45% D₂O) and may thus potentially present a challenge for distinguishing chitin from interacting proteins. However, online tools only give estimates, and experimental determination is required. Moreover, chitin has a complex structure, where crystalline domains might have different densities than less crystalline regions, and a small degree of deacetylation is expected. Additionally, there may be residual components from the natural sources (divalent ions, fatty acids, etc.); thus it was not even clear from the outset, if a defined match point could be obtained. We determined the chitin match point with SANS to be at 47% D₂O (Figure S1). This necessitated the use of deuterated GbpA (D-GbpA) in order to distinguish the protein from chitin in the SANS interaction studies. The production of properly folded, chitin-active D-GbpA is described in a previous paper. To confirm that chitin is well matched out in the q-range where D-GbpA scatters, we measured chitin, D-GbpA, and nondeuterated GbpA (H-GbpA) individually at 47% D₂O (Figure C,D). The scattering of 2.5 mg/mL D-GbpA at 47% was much stronger than that for chitin, which was well matched out except at very low q-values.

GbpA Coats Chitin, Thickening and Smoothing the Fibers

Following the establishment of the contrast match point, we progressed to study the complex of chitin and GbpA. An indication of complex formation was already obtained upon mixing the two components, as described above. The chitin suspension was highly viscous, but with the addition of D-GbpA and subsequent sonication, the suspension became much easier to pipet and stuck less to surfaces (images of the chitin suspension and the mixture are shown in Figure A,B). Nevertheless, some sedimentation occurred, which affected the reproducibility and prevented a more detailed analysis of the scattering data. The SANS curve of the complex at the chitin match point (Figure C,D) does not feature the characteristic GbpA form factor bump around q = 0.05 Å^–1, confirming GpbA to be in the chitin-bound state, where an R _g of 35–40 Å does not persist. Interestingly, the SANS curve has features similar to those of the SANS curve for chitin itself, with a good contrast (at 100% D₂O). In particular, a steep slope at low-q values is present for both chitin and the complex mixture (Figure E), indicating that GbpA coats the chitin fibers. The steep slope also indicates that the samples have a protein–protein structural correlation dictated by the chitin structure. These results were confirmed with negative-stain EM by analyzing GbpA and chitin mixed together (without sonication). Here, the addition of full-length GbpA (GbpA_fl) to chitin produces a change in the ultrastructure of the chitin fibers (Figure ). The fibers appear less stiff and have less defined edges, indicating that GbpA coats the fibers.

3.

Sample preparation and SANS experiments. (A, B) Images of 3 mg/mL chitin samples, either without GbpA (A) or with 2.5 mg/mL GbpA (B). (C–G) SANS data of GbpA, chitin, or GbpA–chitin mixtures, plotted as intensities (I(q)) vs the scattering vector (q) on double-logarithmic scale. Unless stated otherwise, we used 3 mg/mL chitin (except samples of GbpA alone), 2.5 mg/mL GbpA (combined concentration of H-GbpA and D-GbpA, except for samples of chitin alone), and 47% D₂O. (C, D) Chitin and D-GbpA clearly form a complex, as the SANS data of the mixture strongly differ from the sum of the scattering of the individual biomolecules. The disappearance of the GbpA feature (large bump) at around 0.05 Å^–1 reveals a change in the GbpA structure upon chitin binding. The 20- to 30-fold increase in intensity at low q compared to the sum of GbpA and chitin scattering (arrow in part D) suggests scattering by very large clusters of GbpA molecules, possibly GbpA bridging multiple chitin fibers together creating thicker layers. (E) Increasing the amount of chitin relative to GbpA changes the overall structure of the GbpA aggregate, but the steep slope at low-q persists, suggesting a structural correlation between the proteins. The bump at mid-q (indicated by red arrow) is related to changes in fiber thickness upon GbpA binding. (F) The SANS data of the GbpA–chitin complex can be fitted with a unified scattering model (see fitting parameters in Table and discussion of the parameters in the main text). (G) Varying the H-GbpA to D-GbpA ratios affected the structure factor (intensities were lower for higher fractions of H-GbpA, since H-GbpA is almost matched at 47% D₂O), but the steep slope at low-q persisted, showing that even at very low D-GbpA to H-GbpA ratios (i.e., over long distances), the structural correlation between molecules does not disappear.

4.

Negative-stain EM of chitin and GbpA_fl. (A) Cluster of chitin fibers. (B) Overview image (zoomed out) of chitin fibers (uncoated). (C) Individual chitin fiber coated with GbpA_fl. The structure is longer and smoother–like a string. (D) Representative dense aggregate of chitin fibers caused by GbpA_fl. Individual fibers spread out from the cluster. The concentration of chitin under all conditions was 1.5 mg/mL, and the concentration of GbpA_fl (when present) was 400 μM. Extended TEM micrographs are shown in Figure S3.

Given the absence of clear features in the scattering that could be related to the protein form factor in the chitin-bound state and the presence of a strong protein–protein structural correlation dictated by the complex chitin network, it was challenging to obtain a reliable structural model of the complex (a discrete structural model for similar systems does not exist to our knowledge). Initial attempts to model the data using the flexible cylinder model (as for chitin itself) did not yield satisfying fits (χ² > 16). We further attempted to fit the data with a random-walk ellipsoidal model to mimic the alignment of protein domains on the chitin fibers. This also did not fit the data well (χ² > 20). However, we could fit the data using the unified scattering model by Beaucage (χ² between 0.43 and 3.88 for different experimental conditions; see Table ), which does not yield a geometric model (cylinders, ellipsoids) but enables extraction of structural parameters such as R _g values even for overlapping structural levels in hierarchical systems. The Beaucage model applies alternating power-law and Guinier regimes, yielding multiple radii of gyration for the different layers in the structure, as well as insight into the structural nature of interfaces. This model has previously been used for several polymer systems. In some cases, this model can be used to obtain aggregation numbers; however, this would require that one of the structural levels corresponds to a monomer and that the low-q Guinier region is covered in the measured q-range, which is not the case for our data. By analyzing the scattering data for the GbpA–chitin complex using the unified scattering model by Beaucage (Figure F, Table ), we obtained R _g values for the two structural levels within the samples. The larger R _g -value (>1000) was generally unreliable, as the Guinier range was outside the limits (and the standard deviations are high), but the large size does confirm scattering from larger clusters. For most of the samples, an R _g in the range 125–220 Å was obtained (compared to an R _g ∼ 16 Å for uncoated chitin fibers; see section on scattering of GbpA and chitin, and Table S1). This could be related to the cross-section of the protein coating the fibers, especially since it involves a multimeric protein, or alternatively related to the distance between GbpA molecules linking the fibers. Thickening of the fibers was also observed in transmission electron microscopy (TEM) images, obtained under similar conditions (Figure ), although measurements of the fiber thickness in negative-stain EM are not very reliable due to variations in staining intensity between samples (even when the samples are stained for similar times).

1. Fitting Parameters, Unified Scattering Model for Two Structural Levels .

sample	decay exponent 1	R g  1 (Å)	decay exponent 2	R g  2 (Å)	χ2	SASBDB ID
2.5 mg/mL D-GbpA, 10 mg/mL chitin	2.69 ± 0.01	1278 ± 11	3.88 ± 1.1	16.2 ± 0.3	2.66	SASDW92
2.5 mg/mL D-GbpA, 3.0 mg/mL chitin	2.64 ± 0.01	1859 ± 71	2.63 ± 0.02	125.2 ± 0.2	4.18	SASDWA2
2 mg/mL D-GbpA, 2.4 mg/mL chitin	3.64 ± 0.05	1600 ± 169	2.98 ± 0.01	202.5 ± 0.3	2.86	SASDWC2
0.5 mg/mL D-GbpA, 2.4 mg/mL chitin	3.40 ± 0.17	1633 ± 387	2.97 ± 0.07	183.1 ± 12.0	0.98	SASDWD2
1.2 mg/mL D-GbpA, 0.48 mg/mL chitin	3.16 ± 0.14	1484 ± 210	3.40 ± 0.06	221.8 ± 6.9	1.60	SASDWF2
1.0 mg/mL D-GbpA, 1.2 mg/mL chitin, 72% D2O	3.03 ± 0.06	1797 ± 223	2.36 ± 0.60	220.3 ± 5.3	1.03	SASDWE2
1.25 mg/mL D-GbpA, 1.25 mg/mL H-GbpA, 3.0 mg/mL chitin	2.81 ± 0.01	1845 ± 131	0.94 ± 0.11	43.4 ± 1.8	1.99	SASDWH2
0.625 mg/mL D-GbpA, 1.875 mg/mL H-GbpA, 3.0 mg/mL chitin	2.78 ± 0.03	2140 ± 165	0.43 ± 0.11	47.2 ± 1.7	1.95	SASDWJ2

^aThe data were fitted with Beaucage unified scattering model with two levels, yielding two R _g values and two decay exponents. The first R _g is not reliable, as the Guinier region is outside the q-range, hence the large standard deviation. The first decay exponent relates to the fractal dimensionality of the aggregate. The second R _g and decay exponent relate to structural features of the samples. An R _g of 125–220 suggests that several protein molecules contribute to the scattering. Except for the first R _g , the standard deviation of most parameters is small, the exception being the second R _g at low protein concentration (0.5–1.2 mg/mL D-GbpA) and the second decay exponent for the sample with high chitin concentration (10 mg/mL). All data sets are available in the SASBDB, with the accession codes in the last column.

^bSamples are in 47% D₂O unless stated otherwise.

GbpA Alters the Chitin Network Structure and Forms Clusters

We observed that the overall scattered intensity (Figure D) increased about a 100-fold at low-q compared to the sum of chitin and GbpA scattering, indicating aggregation of GbpA on chitin at large length-scales or formation of percolating networks (large networks of randomly connected elements). While the data do not contain a well-defined Guinier-region at low-q, a rough estimate can be made of the features around 0.01 Å^–1. The forward scattering in this region (through extrapolation) is of the order of 200–2000 cm^–1, which corresponds to a 100- to 1000-fold increase compared to the monomeric protein. This suggests that the scattering in the complex results from larger aggregates/networks (comprising hundreds or thousands of protein molecules on the chitin network) rather than from GbpA monomers. Some aggregation can be induced by sonication, even using our careful approach (see Experimental Section), however, the clusters were also prominent in the EM micrographs (visible around bundles of GbpA-covered chitin fibers; see Figure D), where the protein was never subject to sonication, suggesting chitin-induced GbpA clustering.

Interestingly, the R _g is not completely reproducible, and a sample with lower protein:chitin ratio instead exhibits an R _g of 16 Å (Table ), which is close to the radius of the chitin nanofiber chains alone (Table S1, Fan et al. and Ma et al.). , This may be because GbpA is largely dispersed across the chitin fibers, unless GbpA is present in excess. However, similar results were not obtained at lower concentrations (while retaining the GbpA:chitin ratio; Table ), possibly due to a low signal at higher q. The overall discrepancy in R _g between samples can also be explained by different parts of the chitin network being in the neutron beam at the time of SANS data acquisition. SANS measurements of the D-GbpA–chitin complex at D₂O levels between the match points of the two components (Figure S2) also yielded an R _g of 220 Å for the protein aggregate (Table ); nevertheless, some of the characteristic features of chitin fibers are still present in the SANS data at high-q.

In addition to the R _g values, the unified scattering model gives decay exponents for the different levels of structures (summarized in Table ). The decay exponent describes the surface scattering or the fractal dimensionality of the aggregate (Porod-like) if D ∼ 4. Chitin itself has a decay exponent of 4.0 (Table S1), indicating very large aggregates with a sharply defined surface , (as seen in Figure A). For the GbpA–chitin complex at the chitin match point, the low-q decay exponent is 2.5–3.6 (Table ), consistent with aggregates in a state between random and globular. The smaller decay exponent compared to that of chitin suggests a partial loss of the sharp interfaces of the chitin fibers when coated with protein, as seen by EM (Figure C). The decay exponent corresponding to the structural levels with R _g values of 125–220 Å is approximately 2.6–3.2 (Table ) and thus considerably higher than 2.0 (which would be expected for an ideal random-walk chain). This indicates that the protein coating is rigid and compact. For the sample yielding a lower R _g of 16.2 ± 0.3 (Table ), the corresponding decay exponent is somewhat higher (3.9 ± 1.1) but also has higher statistically uncertainty as it is derived from the noisier high-q regime.

Individual GbpA Molecules Bound to Chitin Could Not Be Resolved

By decreasing the ratio of GbpA to chitin, we attempted to disperse the GbpA molecules on the chitin fibers in order to analyze the conformations of individual GbpA molecules bound to chitin. This did not work. To address this problem in a different way, we used mixtures of H-GbpA and D-GbpA bound to chitin at the chitin match-point of 47% D₂O (Figure G). Since H-GbpA is well matched-out at the chitin match-point, we expected to still mostly see D-GbpA on the chitin fibers. However, since D-GbpA and H-GbpA compete for the binding sites, the distance between D-GbpA on the fibers should increase, perhaps to the point of scattering independently. As expected, the SANS intensities for the mixtures of D-GbpA and H-GbpA on chitin decreased with a higher proportion of H-GbpA (Figure G). Using the unified scattering model, an R _g of 43–47 Å was obtained (Table ), which is close to that of monomeric GbpA; however, no clear features could be observed in the scattering curve in the corresponding q-range (∼0.05 Å^–1) and overall, the data quality at high-q was significantly weaker for these samples due to the lower signal-to-noise ratio. While the low R _g likely originates from the GbpA-layer on chitin, the actual value may not be indicative of the thickness, as some of it corresponds to matched-out H-GbpA. The decay exponent at low-q was 2.8, similar to the other GbpA–chitin samples; thus, this approach did not remove the structure factor effect. This finding showed that even when the distance between D-GbpA particles on chitin is increased, D-GbpA molecules do not scatter independently (further increasing the distance between D-GbpA molecules by lowering the D-GbpA to H-GbpA ratio was not possible, since the signal became too low). Due to the rigid structure of the chitin network, it is likely that protein–protein structure factor effects will exist even at extremely low protein concentrations, making it technically impossible to determine the conformation of GbpA on chitin with SANS.

GbpA Binds to Large Areas of Chitin Rapidly and Remains Firmly Attached

To further characterize the GbpA aggregates on chitin, we performed ultra-small-angle neutron scattering (USANS) experiments, which give structural information in the hundreds of nanometers range, thus providing an estimate of the size of aggregates. However, we observed that the slope for the chitin-bound GbpA extends to very low angles without flattening out, indicating that GbpA coats very large areas (or domains) on the chitin fibers (beyond the resolution of several micrometers; Figure ) or forms large percolating networks. EM analysis provided a good explanation for this, with GbpA completely coating chitin fibers and inducing aggregation of the fibers in clumps, as observed in Figure D. A more complete set of EM micrographs is presented in Figures S3 and S4. Interestingly, when experimenting with different GbpA:chitin ratios, we found that it made a difference if we prepared samples with individual ratios independently or if we started out with a 1:1 mixture, to which we subsequently added either chitin or GbpA (Figure S5). In the latter case, we saw uncomplexed, rigid chitin fibers next to smooth GbpA-coated chitin fibers with dense protein aggregates when adding chitin to the 1:1 mixture, suggesting that GbpA binds to chitin rapidly, covering its surface until all binding sites are occupied, and thereafter remains firmly attached to chitin.

5.

SANS and USANS data. (A) GbpA bound to chitin and (B) chitin alone. Data were collected at 47% D₂O. USANS data confirmed a structural order in the micrometer regime. Error bars represent one standard deviation.

EM Analysis of Truncated GbpA and V. cholerae Microcolonies

We were curious about the contributions of the different GbpA domains to chitin binding and complex formation and therefore extended the EM analysis to truncated constructs of GbpA. Since both terminal domains (D1 and D4) are known to bind chitin, we selected constructs consisting of domains 1 to 3 (GbpA_D1–3) and domain 1 alone (GbpA_D1) for comparison with full-length GbpA (GbpA_fl) (Figure ). GbpA_D1–3 induced chitin aggregates similar to those of GbpA_fl (Figure B compared to Figure D). However, GbpA_D1 appears to have a weaker effect on the fibers, as seen in Figure C,D. While the fibers are still coated with the protein, they have much sharper edges compared to GbpA_D1–3 and GbpA_fl, and are visually more similar to the free chitin ultrastructure (Figure A and Figure S3), suggesting that domains 2–3 contribute importantly to GbpA network formation on chitin.

6.

Extended negative-stain EM experiments for GbpA truncation variants and V. cholerae cells. (A) Individual chitin fibers coated with GbpA_D1–3 truncation constructs. As for full-length GbpA, the fibers had a smooth appearance. (B) Cluster of aggregated GbpA_D1–3-coated chitin fibers. (C) Individual GbpA_D1-coated chitin fibers extend from an aggregation of fibers and protein. (D) Cluster of chitin bound to GbpA_D1. The concentration of chitin in the experiments shown in panels A–D was 1.5 mg/mL, while the concentration of GbpA_D1–3 and GbpA_D1 in the experiments shown in panels A and B and panels C and D, respectively, were 400 μM (both cases). (E) Microcolony of V. cholerae (observed only for bacteria incubated with chitin). A matrix of extracellular polysaccharides is indicated with black arrowheads; the hazy, fibrillar mass in the center, likely formed by chitin fibers, is indicated with a white arrowhead. Flagella are labeled with smaller black arrows. The concentration of chitin in this experiment was 1.5 mg/mL, while the bacterial culture had an OD₆₀₀ of 2.0.

Additional EM experiments were performed by adding bacteria to chitin fibers, using an inactivated strain ofV. cholerae, in which the toxin coregulated pilus (TCP) was deleted. Interestingly, some of the bacteria were found to create microcolonies (Figure E) that were not observed for the bacteria imaged alone (in the absence of chitin). A more complete set of EM micrographs can be found in Figures S6 and S7.

Discussion

We developed a suspension protocol for the GbpA–chitin complex and carried out SANS, USANS, and negative-stain EM experiments to characterize the structure of GbpA upon complex formation, allowing the first structural investigation of GbpA in the chitin-bound state. We observed that upon GbpA binding to chitin, the suspension of chitin fibers immediately becomes less viscous, suggesting that the properties of the GbpA–chitin complex differ significantly from those of both partners alone. This was confirmed by SANS (Figure ). Using negative-stain EM, we observed that GbpA binding to chitin smoothed the otherwise rigid chitin fibers (Figure ). After binding, GbpA remained firmly attached, and subsequent serial dilution by either chitin or GbpA had little effect on the formed complex (Figure S5).

In the SANS and USANS experiments, we experienced limitations regarding structural modeling. We suspect that it may not be possible to obtain a detailed conformational model of independently scattering GbpA molecules on β-chitin fibers using SANS, due to the persisting protein–protein structural correlation along the chitin fibers. This is not surprising since the fibers are long and likely rigid, inducing correlation even over long distances. Much shorter chitin fibers would be required in order to overcome these technical impediments, if possible at all.

Both SANS and EM revealed that GbpA does not bind evenly to all parts of the chitin network (Figures and ), possibly because some parts of the fibers are less accessible than others. The regions of chitin with high accessibility for GbpA binding likely exhibit protein–protein interactions between GbpA molecules yet to be elucidated. This hypothesis is supported by the weakened effect that shortened forms of GbpA (i.e., GbpA_D1) have on the chitin ultrastructure (as shown in Figure A–D). For example, domain 1 alone does not appear to be able to induce the formation of larger chitin aggregates (Figure C,D). This lessened interaction cannot be fully explained by the loss of the chitin-binding fourth domain, however, as the qualitative, visual effect of GbpA_D1–3 on chitin was similar to GbpA_fl (Figure B vs Figure D). Domains 2 and 3 therefore appear to be highly important for the observed clustering effect. These domains have previously been reported to be important for binding to the bacteria; thus they may have a dual role. In this context, it is interesting that two very recent studies , predict that the LPMO domain (D1) can fold back onto domain 3 through interactions of the LPMO histidine-brace motif with two conserved histidine residues on D3, preventing off-pathway reactions in the absence of substrate. If this model is correct, perhaps different GbpA molecules can also form intermolecular interactions between domains 1 and 3, explaining the network formation that we observe. Domain 4 would (according to this model) serve as primary anchor on chitin. This would not only prevent oxidative damage but additionally increase the activity radius of GbpA. Another factor that has recently been discovered is a calcium binding site close to the LPMO active site which enhances GbpA stability. Calcium has further been shown to affect V. cholerae biofilm formation and may thus modulate environmental survival and infection in more than one way.

Possible Implications for GbpA’s Role as Colonization Factor

Chitin is the most abundant biopolymer in marine environments and is found, for example, on zooplankton, mussels, and crustaceans. V. cholerae would therefore be well served by using a chitin-binding colonization factor, i.e., GbpA, as one of its first anchors when forming microcolonies (as suggested by Almagro-Moreno et al.). Other important factors for microcolony formation are the toxin-co-regulated pilus (TCP), , the chitin-regulated pilus (ChiRP), and outer membrane adhesion factor multivalent adhesion molecule 7 (Mam7). An extracellular matrix consisting of polysaccharides and proteins then encases the microcolonies, completing the protected environment by forming a biofilm, − which can adapt to changing environmental conditions , and raises infectivity. Such a complex may be influenced by calcium, , often found in mineralized chitin. A negative-stain TEM image of a microcolony from an inactivated TCP-deleted strain of V. cholerae is presented in Figure E (revealing that TCP is not essential for microcolony formation). Microcolonies were not observed in the absence of chitin. Chitin binding may thus initiate a cascade of events that culminate in biofilm formation and hyper-infectivity. Moreover, chitin binding has been shown to induce natural competence in V. cholerae, increasing bacterial fitness by exchange of genes. GbpA secretion appears to play an important role in the initial steps of these events. But why would it make sense for the bacteria to throw out anchors without a connecting line and attach to them in a second step? The picture emerging from the present study is that GbpA efficiently prepares the ground for microcolony formation by forming aggregates dispersed throughout the fibers (Figure ) and with this reserves space for V. cholerae over competitors, as previously suggested by Kirn et al. Secreting anchors first and then quickly attaching to them before reproducing may allow the bacteria to very effectively colonize biotic surfaces in the aquatic environment. Additionally, and at least as important, GbpA ensures access to food supplies through its lytic polysaccharide mono- (or per-) oxidase activity, giving the bacteria a solid return on their initial energy investment when producing and secreting GbpA.

7.

Water-color paintings of V. cholerae sketching microcolony formation on chitin aided by GbpA. From left to right: (1) V. cholerae bacterium secretes the four-domain protein GbpA (domains shown in different colors) close to a cluster of chitin fibers (pink, jagged appearance). (2) GbpA coats chitin by attaching with its two terminal domains (LPMO-domain 1 (red triangle) and domain 4 (green)), smoothing the fibers’ appearance. (3) Bacteria attach via GbpA domains 2 and 3 (blue, pink) to GbpA-coated chitin fibers and produce a hazy, fibrillar mass (green halo around the bacteria), initiating biofilm formation. (4) Formation of V. cholerae microcolony attached to chitin via GbpA. Credit: Ayla Coder.

Conclusions

We established a suspension protocol of chitin and the V. cholerae colonization factor GbpA that allowed us to structurally characterize the GbpA–chitin complex using SANS and negative-stain EM. GbpA changes the ultrastructure of chitin upon binding, smoothing its otherwise jagged appearance. Moreover, GbpA forms dense and stable clusters on chitin. Essential for this transformation are the first three domains of the four-domain protein, which points to the importance of interdomain interactions for complex formation. We also observed a V. cholerae microcolony on our EM grids in the presence of chitin, completing the gallery from GbpA secretion to bacterial colonization (Figure ). Together, our data suggest how GbpA paves the way for biofilm formation.

Experimental Section

Protein Production and Deuteration

Perdeuteration of GbpA involved expression in deuterated media, according to the protocol described by Sørensen et al., which was inspired by Cai et al. This protocol relies on BL21 star (DE3) cells containing the GbpA-encoding gene in the pET22b vector. Growth and expression media contained D₂O and d ₈-glycerol (99% D) purchased from ChemSupport AS (Hommelvik, Norway). The production of H-GbpA and the purification of D-GbpA and H-GbpA is also described in Sørensen et al. The constructs of the truncated variants of GbpA (GbpA_D1 and GbpA_D1–3) for EM analysis are described in Wong et al. and Sørensen et al., , and the protocol for protein production and purification is almost the same as that used for the full-length construct, described by Sørensen et al. The only difference concerns the purification of GbpA_D1–3, where a step gradient was introduced during anion-exchange chromatography to improve the yields of the pure protein. Specifically, a step corresponding to 180 mM NaCl (8 column volumes) was added to the isocratic gradient elution.

GbpA was saturated with copper to activate the enzyme prior to the SANS experiments by incubation with CuCl₂ in 5-fold molar excess for at least 30 min. Free copper was then removed by passing the protein through a HiTrap desalting column (GE Healthcare). This procedure was previously used to produce chitin-active D-GbpA and its LPMO domain, which have the same structures as their nondeuterated counterparts. Prior to the SANS experiments, the protein was dialyzed against 100 mM NaCl and 20 mM Tris-HCl pH 8.0 with 47% of the water being D₂O. For dialysis, we used a Pur-A-Lyzer Maxi Dialysis Kit (Merck), with a cut-off of 3.5 kDa or 12 kDa.

Preparation of Chitin, GbpA, and Chitin–GbpA Samples for SANS

β-Chitin nanofibers (∼180 μm in length) from squid pens (France Chitine, Orange, France) were kindly provided by Jennifer Loose and Gustav Vaaje-Kolstad and prepared according to the protocol described by Loose et al., which was based on the original study by Fan et al. The fibers were dissolved in 20 mM acetic acid (pH 3.2) to a concentration of 10 mg/mL and subsequently sonicated at 50–70% power for 10–15 min (3 s on, 3 s off) with a Vibra Cell ultrasonic processor until they were well dispersed for several days. Finally, the fiber suspension was dialyzed against 20 mM sodium acetate-HAc pH 5.0 with the desired amount of D₂O for at least 24 h. At higher pH values or in the presence of salts, it was observed that the fibers precipitated. However, at pH 5, GbpA is not very stable. Nevertheless, it was possible to add GbpA to the fibers from a highly concentrated stock solution of the protein (prepared by dialysis against 10 mM NaCl, 20 mM Tris-HCl pH 8.0), if the mixture was immediately treated by thorough pipetting (i.e., pipetting the suspension up and down 7–10 times with a micropipette, using a cutoff 1 mL pipette tip). To further stabilize the suspension, we sonicated the mixture for a total of 15–30 s, 1 s on, 1 s off with a tip sonicator at 30–50 W, avoiding overheating (for most samples, 15 s total was sufficient), and redialyzed the sample for 24 h against 20 mM sodium acetate-HAc pH 5.0 with relevant amounts of D₂O. After any dilution of the chitin fibers or the mixture, the solutions were resonicated for 15–30 s to ensure proper suspension but avoiding a raise in temperature of the sample. For GbpA–chitin mixtures, this step was only relevant for the measurement at 72% D₂O. In case of delays before measurements, samples were resonicated for 2 s.

GbpA samples without chitin were dialyzed into 100 mM NaCl and 20 mM Tris-HCl pH 8.0, with the desired level of D₂O prior to SANS analysis.

SANS Measurements

Samples were measured with small-angle neutron scattering (SANS) at beamline D11 at Institut Laue-Langevin (ILL), with wavelength λ = 5.6 Å and a wavelength spread of Δλ/λ = 9%. Data were acquired for the q-range of 0.0013–0.4102 Å^–1. Ultra-small-angle neutron scattering (USANS) experiments were carried out at beamline BT5 at the National Institute of Standards and Technology (NIST) Center for Neutron Research, using λ = 2.4 Å and a wavelength spread of 6% Δλ/λ, for the q-range 0.00005 Å^–1 to 0.001 Å^–1. Tumbling cells were used for the USANS measurements. All samples were measured in the oxidized inactive copper state (in the absence of reducing agent) as sustaining a stable reduced state throughout the long measurements may be unfeasible and changes in the sample throughout the measurements would be undesirable for the data analysis. To find the chitin match point, 10 mg/mL chitin was measured in 0%, 20%, 42%, 66%, 80%, and 100% D₂O for 15 min per sample at ILL beamline D11. The data were integrated from 0.03 to 0.4 Å^–1, and the integrated data were fitted with a second-degree polynomial function. The chitin match point was estimated from the minimum of this function (Figure S1B).

While SANS samples were measured for 2–3 h, USANS measurements lasted 11 h per sample. Buffer, empty cell, and H₂O measurements were used for background subtraction and calibration to absolute scale. Data were processed with software at the respective beamlines. , The scattering data of GbpA were fitted with a random-walk ellipsoidal model using five ellipsoids (see Pedersen et al. and equations within). For chitin the scattering was fitted with a flexible-cylinder model as implemented in SASVIEW, including a power-law for fitting the contribution from smooth interfaces between chitin clusters and solvent,

$$I(q)={\alpha }_1{P}_{\mathrm{Flex}}(q,R,L_{\mathrm{C}},b_{\mathrm{Kuhn}})+{\alpha }_2{q}^{-D}+c$$ 1

where α₁ and α₂ are scaling factors, c is a constant background, and D is the Porod decay exponent. P _Flex is the scattering from flexible polymers, depending on radius (R), contour length (L _C), and Kuhn length (b _Kuhn). P _Flex is modeled using the equations described by Pedersen and Schurtenberger, using their method 3 with excluded volume effects. We fixed the contour length to 5000, which is ∼2π/q _min, to not model sizes outside the q-range. The data were uploaded to the Small Angle Scattering Biological Data Bank (SASBDB), where they are accessible with accession code SASDW42.

For the complex of GbpA and chitin, the data were fitted with the Beaucage unified scattering model ,

$$I(q)=c+\sum _{i=1}^N[G_i\exp (-\frac{q^2{R}_{gi}^2}{3})+B_i\exp (-\frac{q^2{R}_{g(i+1)}^2}{3}){\left(\frac{1}{q_i^{*}}\right)}^{D_i}]$$ 2

where G is the Guinier scaling factor, R_g the radius of gyration, B is the Porod constant, and D is the decay exponent. N is the number of levels used and is set to two for all fits. q _i is given by

$$q_i^{*}=q{\left[\mathrm{erf}\left(\frac{q{R}_{gi}}{\sqrt{6}}\right)\right]}^{-3}$$ 3

An upper limit of π/q _min was put on R _g to keep the values within the q-range. Two levels were first fitted individually to get starting parameters for the fitting of the levels combined. G and B are fitting parameters.

Data were plotted with Origin (version 2017, 64 bit, OriginLab Corporation), and the modeling was carried out with SASVIEW 5.0.5 and QtiSAS. The SANS data have been submitted to SASBDB, and the IDs are included in Table and Table S1.

Negative-Stain EM

GbpA and chitin samples were prepared essentially as described above. 10 mL of 3 mg/mL chitin was subjected to sonication in 20 mM acetic acid at pH 3.2. Sonication was performed on ice with a Q500 Sonicator (Qsonica), using the thinnest tip at 30% intensity in 3/3 s on/off pulses until the sample was completely suspended. The solution was visibly inspected between rounds of sonication to ensure that the solution did not overheat by only sonicating for 5–10 min at a time and waiting for 10 min between each sonication. Chitin was dialyzed overnight against 20 mM sodium acetate–HAc pH 5.0 and sonicated on ice again to ensure that it remained suspended. Chitin aliquots were then mixed with a few microliters of GbpA (or its truncated variants) at a final concentration of 400 μM and incubated with agitation for 1 h at 21 °C. Uncharged, carbon-coated 300 hex-mesh Cu TEM grids were exposed to 8 μL of sample, washed in deionized water 10×, and stained with 1% uranyl acetate (UAc) for 30–60 s.

Additionally, two TEM experiments comparing different GbpA:chitin concentration ratios were performed. The chitin solution was resonicated to ensure full suspension for 10–15 min prior to use, using the same sonication method described above. These samples were mounted to 300 hex-mesh carbon-coated Cu TEM grids as described. In the first experiment (Figure S5A–C), two serial dilutions were carried out, starting from a 300 μL 1:1 solution of 1.5 mg/mL GbpA:1.5 mg/mL chitin. In the first dilution series, a solution of 1.5 mg/mL GbpA_fl was used as the diluent to keep the concentration of GbpA_fl at 1.5 mg/mL while decreasing the chitin concentration. In the second dilution series, the diluent used was a 1.5 mg/mL chitin solution (to keep the chitin concentration at 1.5 mg/mL) while decreasing the GbpA_fl concentration. For both series, 30 μL of the 1:1 solution was first mixed by pipetting with 270 μL of diluent, before 30 μL from the resulting 1:10 (or 10:1) GbpA:chitin solution was immediately transferred to a new Eppendorf tube and mixed with 270 μL of diluent. This was repeated two more times in each direction, creating the GbpA:chitin samples shown in Table S2. In the second experiment (Figure S5D–F), instead of performing a dilution series, the samples were prepared individually. Only three samples were prepared like this, as reported in Table S3.

Finally, TEM experiments were performed with an inactivated TCP-deleted V. cholerae strain (V. cholerae _ΔTCP2) and β-chitin. The bacterial samples were prepared by growing 100 mL of liquid colonies in LB media overnight in a Multitron Standard shaker at 37 °C and 120 rpm. These bacterial samples (OD₆₀₀ = 4) were mounted to 300 hex-mesh Cu TEM grids that were negatively charged in a glow-discharger. Each grid was applied to their respective sample drop (8 μL) for 30 min, then washed 5 times in distilled water and stained with 1.1% UAc for 30–60 s.

For all samples, 40–60 micrographs were acquired to ensure the robustness of our conclusions among varying preparations and fields of view. Extended EM micrographs are shown in the Supporting Information, including artifacts found on the grids. All images were taken with a JEOL 1400Plus transmission electron microscope operating at 120 kV. Micrographs were analyzed with ImageJ.

Glossary

Abbreviations

CBM

carbohydrate-binding module

GbpA

N-acetylglucosamine binding protein A

D-GbpA

deuterated GbpA

EM

electron microscopy

H-GbpA

hydrogenated (=nondeuterated) GbpA

HAc

acetic acid

ILL

Institut Laue Langevin

LPMO

lytic polysaccharide monooxygenase

NIST

National Institute of Standards and Technology

PDB

Protein Data Bank

SANS

small-angle neutron scattering

SASBDB

Small Angle Scattering Biological Data Bank

SAXS

small-angle X-ray scattering

TEM

transmission electron microscopy

SANS data are deposited in the Small Angle Scattering Biological Data Bank (SASBDB), with IDs given in Table and Table S1. All other data are published in this manuscript.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.5c24418.

• Figure S1 illustrates SANS experiments carried out to obtain the chitin match point; Figure S2 shows SANS data of the chitin–D-GbpA complex at 73% D₂O; Figure S3 shows extended TEM micrographs of chitin fibers; Figure S4 shows extended TEM micrographs of chitin–GbpA complexes; Figure S5 shows TEM micrographs from supplementary experiments with differing concentration ratios of chitin and GbpA_fl; Figure S6 shows extended TEM micrographs of chitin fibers incubated with GbpA_D1–3; Figure S7 shows extended TEM micrographs of chitin fibers incubated with GbpA_D1; Table S1 lists SANS fitting parameters for chitin cylinder model with power law; Table S2 lists GbpA_fl:chitin ratios in the dilution-series TEM experiment; Table S3 lists GbpA_fl:chitin ratios in the direct-mixing TEM experiment (PDF)

§.

Ottilia vej 9, H. Lundbeck A/S, DK-2500 Valby, Denmark

K.B.-A. and U.K. conceived the study. H.V.S. and M.M.-C. produced the protein, and H.V.S. established a protocol for sample preparation and performed SANS experiments, supervised by K.B.-A., R.L., and U.K. and with assistance of S.P. at ILL and S.K. at NIST. M.M.-C. and A.C. prepared the samples for EM, acquired the data at the EM core facilities at UiO and discussed them with U.K. G.V.-K. supported the work with samples and know-how. The manuscript was written by H.V.S., with contributions from M.M.-C., A.C. and U.K. It was revised with input and approval from all authors. Author Contributions by Credit Taxonomy: Conceptualization: K. Bjerregaard-Andersen, U. Krengel. Data curation: H. V. Sørensen, M. Montserrat-Canals, A. Coder, S. Prévost, S. Krueger, R. Lund. Formal analysis: H. V. Sørensen, M. Montserrat-Canals, A. Coder, S. Prévost, S. Krueger, R. Lund. Funding acquisition: K. Bjerregaard-Andersen, U. Krengel. Investigation: H. V. Sørensen, M. Montserrat-Canals, A. Coder, S. Prévost, S. Krueger. Methodology: H. V. Sørensen, S. Prévost, R. Lund. Project Administration: U. Krengel. Resources: G. Vaaje-Kolstad. Software: H. V. Sørensen, R. Lund. Supervision: K. Bjerregaard-Andersen, R. Lund, U. Krengel. Validation: H. V. Sørensen, S. Prévost, R. Lund. Visualization: H. V. Sørensen, M. Montserrat-Canals, A. Coder. Writingoriginal draft: H. V. Sørensen, U. Krengel. Writingreview and editing: H. V. Sørensen, M. Montserrat-Canals, A. Coder, S. Prévost, S. Krueger, G. Vaaje-Kolstad, R. Lund, K. Bjerregaard-Andersen, U. Krengel.

The project was funded by the Norwegian Research Council (Grant 272201) and by the University of Oslo (postdoc position of K.B.-A. and Ph.D. position of M.M.-C.). Most of the work was carried out at the UiO Structural Biology core facilities, which are part of the Norwegian Macromolecular Crystallography Consortium (NORCRYST) and which received funding from the Norwegian INFRASTRUKTUR-program (Project 245828) as well as from UiO (core facility funds). Work at international facilities was supported through Granting Research Proposals 8-03-988 (ILL), 1920565 (ISIS), and 27117 (NIST). Access to the USANS instrument was provided by the Center for High Resolution Neutron Scattering, a partnership between the National Institute of Standards and Technology and the National Science Foundation under Agreement DMR-2010792. This work benefited from the use of the SasView application, originally developed under NSF Award DMR-0520547. SasView contains code developed with funding from the European Union’s Horizon 2020 research and innovation program under the SINE2020 project, Grant Agreement 654000.

National Institute of Standards and Technology Disclaimer: Certain commercial equipment, instruments, materials, suppliers, or software are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology (NIST) nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.

The authors declare no competing financial interest.