Structural insights into VLR fine specificity for blood group carbohydrates

SUMMARY

High-quality reagents to study and detect glycans with high specificity for research and clinic applications are severely lacking. Here, we structurally and functionally characterize several variable lymphocyte receptor (VLR) based antibodies from lampreys immunized with O erythrocytes that specifically recognize the blood group H-trisaccharide type II antigen. Glycan microarray analysis and biophysical data reveal that these VLRs exhibit greater specificity for H-trisaccharide compared to the plant lectin UEA-1, which is widely used in blood typing. Amongst these antibodies, O13 exhibits superior specificity for H-trisaccharide, the basis for which is revealed by comparative analysis of high-resolution VLR:glycan crystal structures. Using a structure-guided approach, we designed an O13 mutant with further enhanced specificity for H-trisaccharide. These insights into glycan recognition by VLRs suggest that lampreys can produce highly specific glycan antibodies and are a valuable resource for the production of next-generation glycan reagents for biological and biomedical research and as diagnostics and therapeutics.

eTOC

Collins et al. identify and characterize glycan-specific VLR antibodies from immunized lamprey. Glycan array, biophysical, and structural analyses show their highly specific glycan-binding properties. A VLR with superior specificity for the major O blood group antigen, H-trisaccharide, is identified by coupling the unconventional adaptive immune system of lamprey with structure-guided design.

INTRODUCTION

Glycans are an important class of macromolecules that are found in every organism, from viruses to bacteria to humans. They are critical for many important biological functions, including cell-to-cell adhesion, cell signaling, as well as protein folding, function and trafficking (Varki et al., 2009). Glycans are also involved in many pathogenic processes, such as viral immune evasion, adhesion of bacteria to host cells, and tumor metastasis. Changes in glycan expression are also currently used as a readout to detect and monitor several cancers (Stowell et al., 2015) and as therapeutic targets as in the case of neuroblastoma treatment (Ahmed and Cheung, 2014).

The study of glycans is inherently challenging. Glycans are often complex in composition and structure and vary widely amongst organisms, tissues, and cell types (Varki et al., 2009). The complexity and diversity of glycans arise from several phenomena: the variety of sugars in different tissues and organisms, the variability of the glycosidic linkages that connect individual monosaccharides, branching of polysaccharide chains, postglycosylational modification of individual sugar moieties, and attachment of glycans to proteins and lipids. Additionally, glycans are not directly genetically encoded, making them difficult to study by genetic approaches.

Due in part to the inherent complexity of glycans, the biological tools to detect, purify, and analyze glycans are scarce and lacking in specificity (Sterner et al., 2016). Lectins and glycan-specific antibodies are the most commonly used tools. However, useful lectins are difficult to identify and have sub-optimal specificities. For example, a plant lectin used to identify the human O blood type, Ulex europaeus agglutinin-1 (UEA-1), binds to the major O antigen H-trisaccharide (Fucα1-2Galβ1-4GlcNAc), but also to unrelated glycans such as 2′-fucosyllactose (Fucα1-2Galβ1-4GlcNAc), GalNAcα1-4(Fucα1-2)Galβ1-4GlcNAc, Galα1-4(Fucα1-2)Galβ1-4GlcNAc, and the Lewis-Y antigen (Fucα1-2Galβ1-4(Fucα1-3)GlcNac (Figure S1). Most anti-glycan antibodies are low affinity IgM antibodies that may also be cross-reactive. Currently, the available antibodies recognize only <4% of the estimated 7000 mammalian glycan determinants, and many of the epitopes recognized by these antibodies are ill-defined (Sterner et al., 2016). Thus, there is a great need for novel reagents to analyze and characterize glycans. The lack of such available reagents and methodologies has created a lag in the fields of glycobiology and glycomics, compared to other major fields of research such as genomics and proteomics. Additionally, there is an enormous untapped potential for the use of glycan-specific reagents in diagnostic and therapeutic applications.

The sea lamprey (Petromyzon marinus) represents a promising source of glycan reagents. The adaptive immune system of sea lamprey is functionally similar to that of other vertebrates (Boehm et al., 2012). However, sea lamprey antigen receptors, known as variable lymphocyte receptors (VLRs), are assembled using gene segments that encode leucine-rich repeats (LRRs), not immunoglobulin domains (Alder et al., 2005; Pancer et al., 2004). Mammalian carbohydrates are highly immunogenic in lamprey (Boffa et al., 1967; Litman et al., 1970; Pollara et al., 1970) and VLR-based antibodies that exhibit high affinity and selectivity for glycans have been derived from immunized and even non-immunized lamprey (Alder et al., 2008; Han et al., 2008; Hong et al., 2013; Tasumi et al., 2009). We demonstrated previously that H-trisaccharide-specific VLR antibodies were produced in lampreys immunized with type O erythrocytes (Alder et al., 2008). Our previous structural studies revealed how one monoclonal VLR, RBC36, recognizes the H-trisaccharide, the major blood group antigen expressed on O erythrocytes (Han et al., 2008). RBC36 bound the glycan antigen on the hypervariable concave face using additional contacts mediated by the LRR C-terminal (LRRCT) loop. Contacts between the concave face and H-trisaccharide were mostly mediated by hydrogen bonding to glycan hydroxyl groups, whereas in the LRRCT interface, a tryptophan residue stacked on top of the glycan. A similar binding mode was observed with another glycan-specific VLR (Luo et al., 2013).

To further characterize the recognition of glycans by VLRs and to examine the utility of VLRs as glycome analysis reagents, we performed a detailed analysis of the glycan binding properties of RBC36 and two additional monoclonal VLRs (O13 and Tn4-22) derived from lampreys immunized with human O erythrocytes or the human B-cell line Tn4 (Mi et al., 2012). Tn4 cells exhibit abnormal cell surface glycosylation profiles, including aberrant expression of the Tn antigen. Thus, to expand the repertoire of glycan-binding VLRs, lamprey were immunized with Tn4 cells (Mi et al., 2012). We used purified recombinant VLRs to quantitatively assess the binding affinities and specificities of these antibodies for glycans. We also used high-resolution structural analyses to determine the basis for their glycan specificities. Binding studies confirmed that O13 exhibits enhanced selectivity for the type II H-trisaccharide compared to RBC36 and Tn4-22 or the plant lectin UEA-1. Structural studies demonstrated that the three VLRs have a similar LRR domain fold and that all three bind glycan antigens using their concave surface and LRRCT loop. However, O13 selectively binds to H-trisaccharide by recognizing an acetamide moiety unique to the H-trisaccharide. Lastly, we demonstrated that the selectivity of O13 could be further enhanced by mutating only two residues. These data provide additional insights into glycan recognition by VLRs and indicate that VLR specificities can be finely tuned by the introduction of rationally designed mutations. Altogether, our data suggest that O13 and similar monoclonal VLRs are valuable novel reagents for detection of glycans.

RESULTS

Glycan array profiling of monoclonal VLRs

Analysis of the glycan binding profiles of the three clones (RBC36, O13, and Tn4-22) by glycan array demonstrated that these VLRs specifically recognize the type II H-trisaccharide (Fucα1–2Galβ1–4GlcNac) referred to as H-trisaccharide hereon unless otherwise indicated (Figure S2). In particular, O13 showed enhanced specificity for H-trisaccharide (Figure S2A) when compared to Tn4-22 (Figure S2B) or to the H-trisaccharide-specific plant lectin UEA-1 (Figure S1), which is currently used to detect the O antigen in blood type screening.

Comparison of basic structural features

Alignment of the amino-acid sequences of RBC36, O13, and Tn4-22 ectodomains (Figure 1A,B) illustrates that these VLRs are closely related, with amino-acid identities of 80–90% (Table S1). The C-terminal halves are essentially identical (98.3–100%), suggesting that recognition of H-trisaccharide by O13 and Tn4-22 is similar to that of RBC36. However, several differences in amino-acid sequence can be observed in the N-terminal halves of these VLRs. Additionally, O13 and Tn4-22 have one more LRRV segment compared to RBC36, which likely corresponds to LRRV1.

Figure 1. Crystal structures of monoclonal VLR antibodies.

(A) Schematic showing domain organization of the VLRs in this study. O13 and Tn4-22 VLRs here have an extra LRR compared to RBC36. SP, signal peptide; CP, connecting peptide. (B) Alignment of VLR ectodomain amino-acid sequences by LRR. Conserved residues are highlighted in yellow. Residues involved in direct interactions with ligands are underlined and marked by triangles (red triangles, if the interaction is common to all VLRs, or black triangles if the interaction is unique to a particular VLR). Based on alignment generated with ClustalO. (C) Alignment of apo structures of RBC36 (cyan), O13 (green), and Tn4-22 (magenta).

To perform binding and structural studies, the ectodomains of RBC36, O13, and Tn4-22 (Figure 1A) were expressed as soluble monomers in the baculovirus expression system or in 293S cells. We determined crystal structures of O13, RBC36, and Tn4-22 in their apo form and complexes of O13:H-trisaccharide, O13:Galβ1-4GlcNacβ1-3Galβ1-4Glc (lacto-N-neotetraose, LNnT), RBC36:Fucα1-2Galb1-4Glc (2′-fucosyllactose), and Tn4-22:H-trisaccharide (Table 1). Well-defined electron density was visible for all ligands (Figure S3). All three VLRs, whether in apo (Figure 1C) or ligand-bound form, show a similar solenoid structure, with nearly identical LRRCT loop positions and conformations. RMSDs on all atoms for all of the alignments were below 1Å (Table S1). These results demonstrate that this group of glycan-specific VLRs exhibit the typical solenoid structure observed in other VLRs and that any differences in glycan binding specificities likely arise from specific amino-acid differences.

Table 1.

X-ray data collection and refinement statistics

	RBC36 Apo	RBC36 2′fucosyllactose	O13 Apo	O13 H-trisaccharide	O13 LNnT	Tn4-22 Apo	Tn4-22 H-trisaccharide
Data Collection
Beamline	APS 23ID-D	SSRL 12-2	SSRL 12-2	APS 23-ID-B	APS 23-ID-B	SSRL 12-2	APS 23-ID-D
Wavelength (Å)	1.03316	0.97840	0.97946	1.03320	1.03320	1.03320	1.03320
Space Group	P212121	P21	P41	P21212	P4322	I41	P3121
Unit Cell (Å;°)	a=43.3, b=78.9, c=158.8	a=40.0, b=47.5, c=57.3; β=85.7	a=b=86.7, c=38.1	a=87.1, b=79.0, c=83.4	a=b=82.4, c=227.5	a=b=83.7, c=115.1	a=b=98.9, c=162.9
Resolution (Å)a	41.8-1.70 (1.76-1.70)	36.5-2.14 (2.21-2.14)	43.3-1.70 (1.73-1.70)	47.90-2.03 (2.10-2.03)	46.8-2.04 (2.08-2.04)	41.3-1.85 (1.91-1.85)	47.3-1.89 (1.96-1.89)
Observations	686,465 (29,524)	76,282 (7003)	356,846 (17,989)	191,001 (3518)	332,438 (2234)	225,038 (20,103)	71,5406 (57174)
Unique	60,711	11,758	31,376	32,373	46,172	33,959	73,483
Reflections	(5444)	(1106)	(1573)	(1466)	(1117)	(3343)	(6177)
Redundancy	11.3 (5.4)	6.5 (6.2)	11.4 (11.4)	5.9 (2.4)	7.2 (2.0)	6.6 (6.0)	9.7 (9.2)
Completeness(%)	99.0 (90.0)	97.0 (94.0)	100.0 (99.9)	85.5 (39.8)	90.4 (44.4)	100.0 (99.0)	98.0 (85.0)
< I/σi>	23.6 (3.7)	8.0 (1.0)	37.4 (2.0)	13.7 (1.6)	10.0 (1.5)	15.1 (2.0)	14.8 (1.7)
Rsymb	0.07 (0.36)	0.16 (1.73)	0.09 (0.97)	0.13 (0.79)	0.23 (0.79)	0.07 (0.93)	0.10 (1.01)
Rpimc	0.02 (0.16)	0.07 (0.70)	0.03 (0.29)	0.05 (0.48)	0.08 (0.49)	0.03 (0.41)	0.03 (0.35)
CC1/2d	1.00 (0.95)	1.00 (0.91)	0.97 (0.82)	0.97 (0.62)	0.99 (0.49)	1.00 (0.77)	1.00 (0.76)
Refinement
Resolution (Å)	41.8-1.70 (1.76-1.70)	36.5-2.14 (2.21-2.14)	43.33-1.70 (1.75-1.70)	47.90-2.03 (2.08-2.03)	46.8-2.04 (2.08-2.04)	41.3-1.85 (1.91-1.85)	47.3-1.89 (1.96-1.89)
Reflections-work	60,686 (5439)	11,654 (1098)	29,837 (2647)	32,335 (1019)	43,726 (2284)	33,930 (3340)	73,412 (6177)
Reflections-test	3002 (246)	560 (56)	1524 (157)	1813 (65)	2281 (122)	1762 (145)	3698 (298)
Rcryst(%)e	18.6 (23.8)	22.6 (38.3)	16.1 (22.5)	18.7 (26.7)	19.4 (27.1)	20.9 (37.1)	20.2 (27.9)
Rfree (%)f	21.8 (28.2)	28.3 (44.1)	18.9 (26.6)	21.1 (29.9)	22.4 (27.9)	21.4 (40.8)	23.2 (29.5)
Av. B-value (Å2)	22.5	37.6	33.6	25.4	17.4	40.6	32.7
Protein	22.0	37.6	32.0	25.0	16.3	40.5	32.7
Ligand	-	39.0	-	21.9	18.5	-	29.1
Solvent	27.5	39.1	43.7	30.0	25.0	42.5	33.1
Wilson B (Å2)	18.0	30.7	26.9	21.2	15.5	35.3	30.4
Residues	457	216	247	482	724	241	720
Ligands	none	FUC-GAL-GLC	none	FUC-GAL-NAG	GAL-NAG-GAL-GLC	none	FUC-GAL-NAG
Ions	Mg+2	-	Cl−	Cl−	-	-	-
Waters	373	61	278	371	799	92	267
R.m.s.d. from ideal geometry
Bond Length (Å)	0.006	0.007	0.012	0.002	0.004	0.007	0.008
Bond Angles (°)	1.09	1.34	1.31	0.62	0.76	1.22	1.19
Ramachandran Statistics(%)g
Favored	97.0	92.0	94.7	96.7	96.4	96.0	95
Allowed	3.0	7.5	4.9	3.3	3.6	4.0	5.0
Outliers	0	0.5	0.4	0	0	0	0

^aNumbers in parentheses refer to the highest resolution shell.

^bR_sym = Σ_hklΣ_i | I_hkl,i − <I_hkl> |/Σ_hklΣ_i I_hkl,I, where I_hkl,i is the scaled intensity of the i^th measurement of reflection h, k, l, < I_hkl> is the average intensity for that reflection, and n is the redundancy.

^cR_pim is a redundancy-independent measure of the quality of intensity measurements. R_pim = Σ_hkl (1/(n−1))^1/2Σ_i | I_hkl,i − <I_hkl> |/Σ_hklΣ_i I_hkl,I, where I_hkl,i is the scaled intensity of the i^th measurement of reflection h, k, l, < I_hkl> is the average intensity for that reflection, and n is the redundancy.

^dCC_1/2 = Pearson Correlation Coefficient between two random half datasets

^eR_cryst = Σ_hkl | F_o − F_c |/Σ_hkl | F_o | x 100

^fR_free was calculated as for R_cryst, but on a test set comprising 5% of the data excluded from refinement.

^gThese values were calculated using Phenix.

Quantitative binding studies

To quantitatively assess the specificities of RBC36, O13, and Tn4-22, we performed Isothermal Titration Calorimetry (ITC) studies with the recombinant VLR ectodomains and H-trisaccharide, 2′-fucosyllactose, Lewis-Y, and LNnT. O13 bound to H-trisaccharide with a slightly higher affinity (2.6 μM) compared to RBC36 (8.3 μM) and Tn4-22 (6.9 μM) (Table 2, Figure S4). The specificity of O13 was also greater as O13 did not bind to 2′-fucosyllactose or Lewis-Y and bound to LNnT with very low affinity (160 μM). In contrast, RBC36 and Tn4-22 bound to 2′-fucosyllactose and LNnT, respectively, with much higher affinities (57.1 μM and 16.2 μM, respectively) than O13. The glycans interacted with the VLRs in an enthalpically driven, entropically disfavored process, as revealed by thermodynamic ITC measurements (Table S2). These results are in agreement with our glycan array analyses demonstrating highly restricted glycan binding specificity of O13 among these VLRs (Figure S2). UEA-1 bound the H-trisaccharide with similar affinity as the VLRs (8.1 μM), but UEA-1 also bound 2′-fucosyllactose and Lewis-Y with nearly the same affinity (12.5 μM and 19.8 μM, respectively). Thus, the specificity of O13 for H-trisaccharide is far greater than that of UEA-1.

Table 2.

Binding of VLRs and lectin UEA-1 to glycans

	H-trisaccharide FUC-GAL-NAG	2′fucosyllactose FUC-GAL-GLC	Lewis-Y FUC-NAG-GAL-FUC	LNnT GAL-NAG-GAL-GLC
O13	2.6	N.B.	N.B.	160
RBC36	8.3	57.1	N.B.	N.B.
Tn4-22	6.9	N.B.	N.B.	16.2
UEA-1	8.1	12.5	19.8	N.B.

Values shown are Kd (μM), N.B. = No binding

Common features of glycan binding modes

Because of their high amino-acid sequence identity (82.6–88.5%) (Figure 1B and Table S1), the VLRs in this study were expected to share many common features of ligand binding. The identity is especially high in the C-terminal halves of each protein (98.3–100%). Accordingly, the conformations of the variable loops of all LRRCTs are nearly identical (Figure 1C), as is the tryptophan side chain at the apex of each LRRCT loop (Figure 2). For RBC36, this tryptophan (W204) was previously shown to make contacts with H-trisaccharide (Han et al., 2008). Structures of these VLRs bound to glycan ligands (Figure 3A) show that all three VLRs bind to glycans via their concave face (Figure S5). The crystal structures also illustrated that, similar to the RBC36:H-trisaccharide interaction (Figure 3B), these VLRs bind glycans using a complex network of interactions (Figure 3C–F), including direct and water-mediated hydrogen bonds and hydrophobic interactions. A comparison of the glycan-bound structures reveals a common binding mode mediated by the C-terminal portions of each VLR and therefore close overlap of glycan positions (Figure 4A, monosaccharides occupying identical positions are labeled in red or green). In each of the five glycan-bound VLR structures, a GAL ring stacks under the conserved tryptophan side chain in the LRRCT loop (Figure 4B–F, residues labeled in red). In the trisaccharide-bound structures (H-trisaccharide and 2′-fucosyllactose), the FUC ring is also positioned below the LRRCT loop. The third sugar ring, NAG (H-trisaccharide) or GLC (2′-fucosyllactose), extends toward the LRRV2/3 loops. In all trisaccharide-bound VLR structures, two adjacent residues in LRRVe, an aspartic acid and a glutamine, hydrogen bond with hydroxyl groups on the FUC and GAL (Figure 4B,C,D,F; residues labeled in red). In the O13:LNnT structure (Figure 4E), the terminal GAL and adjacent NAG occupy positions identical to those of GAL and NAG in the H-trisaccharide-bound complexes. The other two sugar rings, GAL and GLC, extend out along the elongated N-terminal part of the binding pocket of O13. These results demonstrate that O13, RBC36, and Tn4-22 interact with glycan ligands via common binding modes, particularly in the VLR C-terminal half for binding of the first two sugars.

Figure 2. Conserved Trp residue in glycan binding.

Location of the conserved Trp found in the extended loops of the LRRCTs. H-trisaccharide-bound structures are shown and colored as in Figure 1C. For clarity, glycan ligands are not shown.

Figure 3. VLR-glycan interactions.

(A) Glycans used in structural studies. Monosaccharides that occupy the same positions in each structure are labeled with the same color. The red box highlights the acetamide group of H-trisaccharide that is not present in 2′-fucosyllactose (equivalent position of 2′-fucosyllactose is marked by a dashed box). FUC, fucose; GLC, glucose; GAL, galactose; NAG, N-acetyl-glucosamine. (B–E) Schematic showing the interactions of VLR residues with glycan ligands. Modified from diagrams generated with MOE (Chemical Computing Group, 2016). Green dashed lines and arrows indicate hydrogen bonds; thick green dashed arrows indicate H-Pi interactions. For clarity, residues and waters that do not interact directly with glycans were omitted.

Figure 4. Detailed view of the VLR-glycan interactions.

(A) Positions of overlapped glycan ligands in aligned structures. Colored as in Figure 1 and 2. (B–E) Structures of VLR:glycan complexes shown as cartoon representations. Side chains involved in direct interactions are shown as sticks. Residue numbers shown in red refer to residues involved in VLR-glycan interactions that are common to all three VLRs. Green dashed lines indicate hydrogen bonds. Black dashed lines indicate H-pi interactions.

Structural basis for the selectivity of O13

Although common features of ligand binding were observed in these crystal structures, our binding studies and glycan array data demonstrate that O13 exhibits greater specificity for H-trisaccharide versus other similar glycans when compared to RBC36 and Tn4-22. The high level of sequence identity in the C-terminal halves of these VLRs (98.3–100%) and the similar network of contacts made by residues of the LRRCT and LRRVe suggest that differences in glycan-binding specificity are mediated by the central and N-terminal regions of the VLRs. Indeed, unique contacts with ligands are made by each VLR that involve residues of the central LRRV segments and N-terminal LRRNT and LRR1 segments.

These differences explain why O13 binds only H-trisaccharide while RBC36 binds both H-trisaccharide and 2′-fucosyllactose. H-trisaccharide and 2′-fucosyllactose are closely related trisaccharides, differing by just one substituent on the GLC ring. In H-trisaccharide, an acetamide group is present at the C2 position of the GLC ring (Figure 3A, red box), whereas, in 2′-fucosyllactose, a hydroxyl group occupies this position (Figure 3A, red dashed box).

The increased selectivity of O13 for H-trisaccharide compared to RBC36 can be explained by differences in the hydrophobic pockets that accommodate the methyl group of the acetamide moiety. In O13, the methyl group fits snugly into a deep pocket lined by the hydrophobic side chains of W79, I101, and Y127, with A103 forming the base of the pocket (Figure 5A,B). An H-pi interaction between the methyl group and Y127 further stabilizes H-trisaccharide binding to O13 (Figures 4D and 5A). Additionally, O13 D105 on the neighboring LRR segment hydrogen bonds with the nitrogen of the H-trisaccharide acetamide moiety (Figures 4D and 5A).

Figure 5. Recognition of the acetamide group of H-trisaccharide by O13.

Close-up view of the acetamide binding pocket of the O13:H-trisaccharide complex represented as (A) sticks and (B) a surface representation. (C,D) The same views of the acetamide binding pocket of the RBC36:H-trisaccharide complex.

Relative to O13, RBC36 has a shallower, less hydrophobic acetamide binding pocket. The base of the pocket in RBC36 is lined with the relatively bulky Y79 side chain, in place of O13 A103 (Figure 5C,D). Unlike O13 Y127, RBC36 Y79 is positioned further from the acetamide methyl group and its side-chain orientation is not suitable for an analogous H-pi interaction. In RBC36, D103 replaces O13 Y127 and hydrogen bonds with the C3-hydroxyl of the GLC ring in both the H-trisaccharide and 2′-fucosyllactose structures (Figure 4B,C). RBC36 D103 also hydrogen bonds with the C2-hydroxyl of GLC, which is specific to 2′-fucosyllactose, further contributing to cross-reactivity of RBC36 with this glycan. Thus, RBC36 binds to both H-trisaccharide and 2′-fucosyllactose because almost all contacts by RBC36 are made to the glycan functional groups that are common to both glycans.

Amino-acid sequence differences in the N-terminal and central LRRs also explain how O13 binds to LNnT, albeit with low affinity, whereas RBC36 does not. LNnT is an extended tetrasaccharide that has two adjacent sugar rings (NAG-GAL) in common with H-trisaccharide, but that lacks the FUC sugar attached to the GAL (Figure 3A). In the aligned structures, the NAG-GAL of LNnT in the O13:LNnT structure overlaps with the NAG-GAL of the other glycans in the glycan-bound VLR structures (Figure 4A). The O13 and RBC36 glycan binding interfaces are similar in size suggesting that RBC36 could also accommodate the extended LNnT tetrasaccharide. Modeling of LNnT in the RBC36 glycan binding pocket also suggests that RBC36 could accommodate LNnT (Figure 6). However, our binding studies show that RBC36 does not bind LNnT (Table 2). A detailed structural comparison shows that sequence differences in the N-terminal LRRs of O13 and RBC36 explain why O13 binds to LNnT, albeit weakly, while RBC36 does not. To bind LNnT, which lacks the FUC sugar in H-trisaccharide or 2′-fucosyllactose, each VLR would have to make up for a loss of interactions with the FUC sugar (which are mediated by the C-terminal portion of the VLR). In the case of O13, but not RBC36, residues of the N-terminal portion compensate for the loss of interactions with FUC. The N-terminal portion of O13 interacts with the additional GLC and GAL of LNnT (Figure 7A). O13 N81 and N82 make water-mediated hydrogen bonds to the internal GAL of LNnT and K36 and D58 hydrogen bond with the terminal GLC of LNnT (Figure 7A). The amino acids of RBC36 differ at all four positions. RBC36 H57 and D58 replace O13 N81 and N82. Whether RBC36 H57 and D58 would also form hydrogen bonds to the GAL is unclear. The shorter RBC36 S37 side chain (compared to O13 D58 in the same position) is too far from the terminal GLC hydroxyl groups to form a hydrogen bond. RBC36 side chains equivalent to O13 K36 are also too far from the terminal GLC to hydrogen bond. RBC36 is smaller than O13, because O13 contains one additional LRRV segment (LRRV1) that shifts the LRR register, such that O13 LRR1 now superimposes with RBC36-LRRNT and O13-LRRNT extends beyond the N-terminus of RBC36 (Figures 1A, C, and 7B). The position of RBC36 that is equivalent to K36 of O13 is part of the RBC36 LRRNT that forms a much smaller turn and caps the solenoid. This portion of the LRRNT is much smaller than other LRRs so that the backbone is positioned 7.5Å below the adjacent LRRs that comprise the binding surface (Figure 7A,B). Lastly, it is also possible that the bulky R36 side chain of RBC36 (Figure 7A) interferes with LNnT binding. A smaller residue, N57, occupies a similar position in O13. Together, these results demonstrate that the observed differences in specificities between O13 and RBC36 are due to specific interactions that are unique to each VLR, many of which are mediated by residues from the central and N-terminal LRR segments. Similar differences in the N-terminal portion of Tn4-22 likely also explain its ability to interact with LNnT; however, additional structural studies will be required to address that question.

Figure 6. Model of LNnT bound to RBC36.

RBC36 (ribbon) with LNnT (yellow sticks) modeled in the glycan binding pocket (top). Close-up view of the modeled fit shows predicted hydrogen bonds (green dashed lines) (bottom).

Figure 7. Improved specificity of an engineered O13 VLR with two mutations.

(A) Detailed view of the structural alignment of O13:LNnT complex (green) with apo RBC36 (cyan). (B) Zoomed-out view of the O13 and RBC36 structural alignment highlighting the N-terminal regions of each VLR. Arrows indicate the distance between the backbone at O13 K36 and the backbone of the equivalent position in RBC36. Hydrogen bonds are colored magenta. (C) Glycan binding affinity of RBC36, O13 wild type and O13 mutant, as determined by ITC (N.B., no binding). (D) Raw ITC data (top of each panel) and binding isotherms (bottom of each panel) from ITC experiments with the O13 mutant and H-trisaccharide (left) or LNnT (right).

Structure-guided engineering of O13

To further enhance the specificity of O13 for H-trisaccharide, we introduced mutations in O13 designed to eliminate cross-reactivity with LNnT, while retaining high affinity interactions with H-trisaccharide. Several mutants were screened for expression in 293S cells. One double mutant, N81H-N82Q, expressed at a level high enough to enable binding studies. The double mutant was designed to decrease the size of the O13 glycan binding pocket and eliminate a water-mediated hydrogen bond between N81, N82, and the C2 hydroxyl group of the internal GAL residue of LNnT (Figure 7A). Each Asn was mutated to a residue that was similar or identical to the equivalent residue in RBC36 (Figure 7C), which does not bind to LNnT. Indeed, ITC binding studies demonstrated that the N81H-N82Q mutant bound to H-trisaccharide with the same affinity as wild-type O13 and did not bind to LNnT (Figure 7C, D).

DISCUSSION

Although glycans play important roles in many biological processes for all extant organisms, tools to support glycobiology and glycomics research are sparse and could be greatly enhanced in both quantity and quality. Lampreys represent a valuable source for such new and improved glycan reagents. Here, we carried out glycan array analyses, binding studies, and structural analyses using recombinant VLRs and glycan epitopes. Our glycan array data illustrate that this group of H-trisaccharide-specific monoclonal lamprey VLRs exhibit superior specificity over current reagents. Our binding studies confirm that VLRs bind to glycan antigens with exquisite specificity and low micromolar affinity. Our structural characterization of VLRs bound to related tri- and tetrasaccharides reveal that this specificity is a result of the extensive network of interactions between the VLR and functional groups of each monosaccharide.

RBC36, O13, and Tn4-22, which are 82 to 89% identical in amino-acid sequence (Table S1), share several structural and biophysical characteristics. The three VLRs have similar affinities for H-trisaccharide in the single digit μM range, recapitulating the affinities of other glycan-specific VLRs (Luo et al., 2013) and the affinity of VLR4 for the Bacillus anthracis spore protein BclA (Kirchdoerfer et al., 2012). The crystal structures of RBC36, O13, and Tn4-22 are nearly identical to each other, with RMSDs for all atoms below 0.5 Å (Table S1). The glycan binding modes of RBC36, O13, and Tn4-22 are also similar to each other, with each VLR using an extensive network of interactions on the concave face to engage glycans. A comparison of RBC36 (Han et al., 2008) and the disaccharide-specific (TFα; Galβ1-3GalNAcα)VLRBaGPA23 structures (Luo et al., 2013) shows that the ligand binding modes of RBC36 (and by extension O13 and Tn4-22) are, as expected, similar to that of VLRBaGPA23 in general location and types of binding interactions, although the conserved Trp in the LRRCT loop is slightly displaced (Figure S6).

Importantly, the VLRs in this study can distinguish between closely related glycans. In particular, the enhanced specificity of O13 for H-trisaccharide, compared to that of RBC36 and Tn4-22, demonstrates that some VLRs can distinguish between glycans that differ by a single functional group on one monosaccharide subunit. O13 exhibits greater specificity for H-trisaccharide by making contacts with the acetamide group of the NAG moiety. We generated structural models to analyze the binding of other closely related blood group antigens to these VLRs. Using the O13:H-trisaccharide structure as a template, several alternative blood group antigens were superimposed onto H-trisaccharide in the O13 glycan binding pocket (Figure S7). In all cases, the alternative ligands had at least two sugar rings in common with Type II H-trisaccharide, often a Galβ1-4GlcNac, that superimposes well with the Galβ1-4GlcNac of H-trisaccharide (n.b. this approach gave the most realistic superposition of ligands, although other orientations of the ligands are possible). Our glycan array data (Figure S2) suggest that all three VLRs in this study bind to Type II H-trisaccharide (Fucα1-2Galβ1-4GlcNac), but not to Type I H-trisaccharide (Fucα1-2Galβ1-3GlcNac), which differs only in the Gal-to-GlcNac glycosidic linkage. Modeling of the Type I H-trisaccharide in the O13 glycan binding pocket suggests that the NAG ring would be inverted 180° relative to its position in the O13:H-trisaccharide (Type II) structure (Figure S7, Panel A). This inversion would prevent O13 from binding to Type I H-trisaccharide by disrupting the interactions that would otherwise focus on the acetamide group and by placing a polar hydroxyl group into the hydrophobic pocket. Our binding experiments show that none of the VLRs in this study bind to the Lewis-Y antigen. Modeling of Lewis-Y and Lewis-X in the glycan binding site of O13 shows that, in both cases, the FUC ring attached to the NAG points downward and clashes with residues at the bottom of the binding pocket (Figure S7, Panel B,C). Because the O13 clone recognizes a single small functional group in H-trisaccharide (the acetamide group), we postulate that O13 could also distinguish between glycans bearing/lacking other modifications such as sulfation. Sulfate groups can be found on blood group glycans and are known to play important roles in lymphocyte trafficking and tumor cell adhesion (Varki et al., 2009). 3S-lactose is a disaccharide that is similar to H-trisaccharide but contains a sulfate group on the GAL ring. Modeling of 3S-lactose in the glycan binding pocket demonstrates that the 3′ sulfate group of the GAL ring would clash with O13, preventing binding to 3S-lactose (Figure S7, Panel D). By extension, this suggests that O13 is specific for unmodified H-trisaccharide. It also suggests that VLRs may be able to distinguish between glycans bearing a sulfate (or other) modification at distinct sites.

When compared to the specificity of current glycan detection reagents, glycan-specific VLRs appear to be superior. A comparison of the glycan binding profiles of RBC36, O13, and Tn4-22 with the lectin UEA-1, which is the reagent currently used to identify the O blood group antigen, supports this notion. Our glycan array data suggest that UEA-1 binds far more promiscuously to glycans than these VLRs (Figures S1 and S2). The binding studies here show that O13 binds H-trisaccharide, but not 2′-fucosyllactose or Lewis-Y (Table 2). UEA-1, however, binds all three glycans with similar affinity. Binding of both O13 and UEA-1 to glycans involves direct and water-mediated hydrogen bonding to monosaccharide hydroxyl groups (Figures 3D,E and S8A); however, O13 makes hydrogen bonds with each sugar moiety in H-trisaccharide (Figure 4D), while UEA-1 may focus primarily on the FUC moiety of its glycan ligands (Figure S8B) (Audette et al., 2002; Audette et al., 2000) (PDB 1JXN). H-trisaccharide, 2′-fucosyllactose, and Lewis-Y all contain a FUC moiety, whereas LNnT does not. Accordingly, UEA-1 binds to H-trisaccharide, 2′-fucosyllactose, and Lewis-Y, but not to LNnT (Table 2). Notably, UEA-1 is characterized as a FUC-specific lectin because of its higher affinity for the FUC monosaccharide compared to NAG or GAL (Loris et al., 1998). Like lectins, many antibodies to blood group antigens are cross-reactive. At least 10% of anti-Lewis antigen antibodies in the DAGR database bind to more than one type of Lewis antigen (Sterner et al., 2016). The structural basis for antibody cross-reactivity is unknown, as few structures of antibodies in complex with blood group antigens are available.

This collection of VLR:glycan structures also allows us to determine whether glycan binding by VLRs occurs via an induced fit or rather by binding of glycans to a rigid preformed binding site. Structures of unliganded RBC36, O13, and Tn4-22 closely align with their glycan-bound structures, with RMSDs for all atoms below 0.5 Å (Figure S9A,B). Additionally, the B-values are relatively low for apo VLR structures (Figure S9C). These data suggest that glycan binding by VLRs does not occur via an induced fit but rather by binding of glycans to a rigid, preformed, binding site, as previously suggested for VLR-antigen interactions (Velikovsky et al., 2009). These results are also in agreement with our ITC studies showing that glycan binding by VLRs is primarily an enthalpy-driven process (Table S2). The thermodynamic profile of our ITC data, namely favorable enthalpy and unfavorable entropy, is typical for other glycan binding proteins, including the UEA-1 lectin (Dam and Brewer, 2002; Hindsgaul et al., 1985). That glycans bind to a preformed site in these VLRs is consistent with β-solenoid proteins having densely packed hydrophobic cores that could be expected to be relatively rigid (Bella et al., 2008). It should be noted, however, that the structures presented here are crystal structures, and the dynamics of VLRs, and LRRCT loops in particular, should be examined further by solution studies. Binding of glycans to lectins may occur in a similar manner, as the structure of the unliganded lectin UEA-1 closely aligns with that of the glycan-bound structure (Figure S9B, middle) and the B-values are relatively low for the apo lectin structure (Figure S9C, middle). However, binding of glycans to antibodies may occur in a manner more akin to induced fit. RMSDs between apo and glycan-bound structures are greater than those of VLRs and the lectin UEA-1 (Figure S9B, right). B-values for glycan-binding pocket residues are also higher in apo structures of the antibodies (Figure S9C, right). The greater apparent conformational flexibility in the glycan binding sites of antibodies may explain the observed cross-reactivity of antibodies as compared to that of VLRs.

VLRs are very versatile glycan reagents because their specificity can be further modified or enhanced in several ways. First, the specificity of the lamprey response and, therefore, the specificities of the VLRs can be directed by the choice of immunogen. Notably, glycan-reactive VLRs are present in the lamprey even before immunization (Hong et al., 2013). Also, when using a yeast surface display library approach to select glycan-reactive VLRs, enrichment for VLRs with a given specificity can be accomplished by sorting with the appropriate glycan antigen(s). It is also possible to use structure-guided mutagenesis to design VLRs with finely tuned specificities. Our data demonstrate that substitution of just two residues in O13 is sufficient to eliminate cross-reactivity to LNnT, effectively increasing the specificity of O13 for H-trisaccharide, while maintaining the same affinity for H-trisaccharide (Figure 7). Furthermore, the modular structure of the LRR domain makes it possible to generate designer proteins. In support of this notion of LRR protein versatility is the observation that all organisms use the LRR fold to accomplish a variety of functions, all of which involve direct binding to every general class of macromolecule and many types of small molecules. The vertebrate Toll-like receptors (TLRs), for example, detect the presence of pathogens by binding directly to numerous pathogen-associated molecular signatures: flagellin protein, lipopeptides, glycolipids, double-stranded RNA, single-stranded DNA, and single nucleosides (Kawai and Akira, 2011). There are several examples of engineered LRR proteins with modified specificities, including those that use a VLR-based scaffold (Lee et al., 2012; Park et al., 2015; Parker et al., 2014; Parmeggiani et al., 2015; Stumpp et al., 2003). The modification of VLR specificity can be accomplished in several ways. For example, LRRs could be inserted or removed to make the glycan-binding site larger or smaller. The former approach could allow VLRs to accommodate larger, branched glycans. Although unusual, VLRBs containing as many as five or even eight LRRV segments have been documented (Boehm et al., 2012; Holland et al., 2014). Additionally, LRRNT loops, which have been observed in VLRC-type receptors (Kanda et al., 2014), could be inserted to modify the glycan binding site, and thus the specificity. Likewise, it should be possible to generate collections of VLRs with a range of glycan specificities by generating libraries with varied LRRNT/LRRCT loop sequences. In addition to modifying VLR specificities, the sensitivity of VLR-based glycan reagents could be improved by mutation (Herrin et al., 2008; Tasumi et al., 2009; Velikovsky et al., 2009) or by multimerizing the VLRs, which is observed in vivo (Boehm et al., 2012).

In summary, we present an extensive structural characterization of glycan recognition by VLRs and reveal the structural basis for the fine specificity of VLRs for glycan ligands. We also present evidence to show that the specificity of VLRs for glycans surpasses that of lectin reagents. Lastly, we demonstrate that VLR specificity can be further enhanced by engineering of the glycan binding site. Together, our results suggest that VLRs can serve as a major source of glycan detection reagents for both research and clinical applications.

STAR METHODS

Detailed methods are provided in the online version of this paper and include the following:

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Myc-tag-488 monoclonal antibody, Clone 4A6	EMD-Millipore	Cat# 05-724
Alexa-Fluor 488 labeled goat anti-mouse IgG mAb	Molecular Probes (ThermoFisher)	Cat# R37120
Bacterial and Virus Strains
DH5α	ThermoFisher	Cat# 18258012
DH10Bac	ThermoFisher	Cat# 10361012
Biological Samples
Human O erythrocytes	Vaccine Center at Emory University	N/A
Human B-cell line Tn4	(Mi et al., 2012)	N/A
Chemicals, Peptides, and Recombinant Proteins
RNeasy mini kit	Qiagen	Cat No./ID: 74104
SuperScript III Reverse Transcriptase	ThermoFisher
EZ-link Sulfo-NHS-LC-Biotin	ThermoFisher	Cat# 21335
QuickChange Lightning Multi Mutagenesis Kit	Agilent Technologies	Cat# 210515
UEA-1	Vector Laboratories	Cat# L-1060
H-trisaccharide Type II	Sigma	Cat# F7297
LNnT	Consortium for Functional Glycomics	N/A
2′-fucosyllactose	Consortium for Functional Glycomics	N/A
Deposited Data
Crystal Structure, RBC36 (APO)	This paper	PDB 5UFD
Crystal Structure, RBC36 + 2′-fucosyllactose	This paper	PDB 5UFF
Crystal Structure, O13 (APO)	This paper	PDB 5UEI
Crystal Structure, O13 + H-trisaccharide	This paper	PDB 5UF1
Crystal Structure, O13 + LNnT	This paper	PDB 5UF4
Crystal Structure, Tn4-22 (APO)	This paper	PDB 5UFB
Crystal Structure, Tn4-22 + H-trisaccharide	This paper	PDB 5UFC
Experimental Models: Cell Lines
Human cells: HEK-293F (Freestyle 293F)	ThemoFisher	Cat# R79007
Human cells: HEK-293S (GnTI−)	ATCC	RRID:CVCL_A785
Insect cells: Sf9 (Spodoptera frugiperda)	ThermoFisher	Cat# 11496015
Insect cells: Hi5 cells (Trichoplusia ni)	ThermoFisher	Cat# B85502
Experimental Models: Organisms/Strains
Petromyzon marinus	Lamprey Services	Larval lampreys
Oligonucleotides
VLRb LRRNT: GCATGTCCCTCGCAGTG	IDT	www.idtdna.com
VLRb LRRCT: CGTGGTCGTAGCAACGTAG	IDT	www.idtdna.com
O13-N81H, N82Q mutagenesis primer: CAGACCCTGTGGCTGCACCAGAACCAGCTGACCTCC	IDT	www.idtdna.com
Recombinant DNA
Plasmid: pCT-ESO-BDNF YSD vector	(Burns et al., 2014)	Gift from Eric Shusta
Plasmid: pCDH	Systems Biosciences	Cat# CD500A-1
Plasmid: pFastBac1	ThermoFisher	Cat# 10360014
Plasmid: phCMV3	Amsbio	P003300
Software and Algorithms
HKL2000	(Otwinowski and Minor, 1997)	hkl-xray.com/hkl-2000
XDS	(Kabsch, 2010)	xds.mpimf-heidelberg.mpg.de/
PHENIX	(Adams et al., 2010)	hwww.phenix-online.org/
Coot	(Emsley et al., 2010)	www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PDB validation server	PDB	validate.wwpdb.org
Other
CFG Glycan array	Consortium for Function Glycomics	Versions 5.0, 5.1
Stanford Synchrotron Radiation Lightsource (SSRL)	Stanford	Beamline 12-2
Advanced Photon Source (APS)	Argonne National Lab	23ID-B, 23ID-D

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ian A. Wilson (wilson@scripps.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Petromyzon marinus

Petromyzon marinus larvae (8–15 cm in length, approximately 2–4 years in age) were collected from tributaries to Lake Michigan (Lamprey Services, Ludington, MI) and housed in sand-lined aquarium tanks at 20°C in the animal facility at Emory University. All experiments were approved by the Institutional Animal Care and Use Committee at Emory University.

HEK293 Cell Culture

HEK293F and HEK293S cells were grown in DMEM medium at 37 C, 8% CO₂.

Insect cell culture

Hi5 and Sf9 cells were grown in HyClone SFX-Insect media at 28 C.

METHOD DETAILS

Lamprey Immunization and Leukocyte Isolation

Lampreys (n=3) were anesthetized with MS-222 (0.1g/L) and given intracoelomic injections (n=3) at two-week intervals with blood type O human erythrocytes isolated from healthy volunteers of the Vaccine Center at Emory University, Atlanta GA or the human B-cell line Tn4. Two weeks after the final injection, the animals were sacrificed with MS-222 (1g/L) and exsanguinated. Total lamprey blood was collected into 0.67xPBS with 30 mM EDTA and layered onto 55% Percoll and centrifuged for 20 minutes at 400xg. Total leukocytes were collected, washed three times in 0.67xPBS, total RNA was extracted from the leukocytes using the RNeasy mini kit (Qiagen) and a leukocyte cDNA library was generated according to standard methods using SuperScript III Reverse Transcriptase (ThermoFisher).

For O13 and Tn4-22, VLRB cDNA libraries were PCR amplified from total lymphocyte cDNA of immunized lampreys using primers specific for the VLRB antigen-binding domain from the LRRNT (GCATGTCCCTCGCAGTG) to the LRRCT (CGTGGTCGTAGCAACGTAG) that each contained 50 bp of sequence homology to the pCT-ESO-BDNF yeast surface display (YSD) vector to facilitate cloning by in vivo homologous recombination in transfected yeast cells. The pCT-ESO-BDNF YSD vector (Burns et al., 2014), a generous gift from Eric Shusta, was digested with NheI, BamHI and NcoI to linearize the vector and remove the BDNF insert, and co-electroporated with the gel-purified VLRB PCR products into the EBY100 strain of Saccharomyces cerevisiae. YSD libraries were grown in tryptophan-deficient SD-CAA selection media, and the size of the YSD library was determined by plating serial dilutions of the YSD libraries onto solid SD-CAA agar plates (~5×10⁶ transformants). To induce surface expression of the VLRB clones, the YSD libraries were grown to log-phase in SD-CAA at 30°C, shaking at 260 RPM and then transferred into SG-CAA induction media containing galactose for 18–24 hours at 20°C, shaking at 260RPM. To detect VLRB surface expression, the YSD library was labeled with an anti-Myc-tag-488 monoclonal antibody (clone 4A6, EMD-Millipore) that recognizes a C-terminal Myc epitope tag in the YSD vector and analyzed by flow cytometry. The isolation of the RBC36 clone was performed by previously described methods (Herrin et al., 2008).

The YSD library was then enriched for human type O erythrocyte VLRB clones by labeling cell surface proteins on red blood cells with biotin (EZ-link Sulfo-NHS-LC-Biotin, Thermo-Fisher Pierce) and incubating with the induced library for one hour at room temperature on rotation. The yeast bound to the labeled erythrocytes were then captured by magnetic-activated cell sorting (MACS) using streptavidin (SA)-conjugated magnetic microbeads (Miltenyi Biotec) and eluted directly into SD-CAA media and grown overnight. This library was further enriched by FACS using the same approach. The success of the MACS and FACS enrichment was monitored by flow cytometry using anti-Myc-tag 488 and SA-PE secondary reagents. After one round of MACS and FACS enrichment, the type O erythrocyte YSD library was plated onto SD-CAA agar plates and individual clones were sequenced. Yeast clones that were bound to the array were transferred to solid media and sequenced.

Glycan Array Profiling

Lamprey VLR sequences for RBC36, O13, and Tn4-22 were cloned into a modified pCDH mammalian expression vector (Systems Biosciences) containing the IgG2a mouse Fc constant region and a 6xHis tag. The chimeric proteins were produced by transiently transfecting the expression vector into 293F cells using 25 kDa, linear polyethylenimine (PEI), 1:3 DNA to PEI) and 2.2 mM valproic acid. The proteins were then purified from the media according to the standard methods recommended from the manufacturer of HisPur Colbalt Resin column (ThermoFisher Scientific). The chimeric VLRB-mFc proteins were screened on the Consortium for Function Glycomics (CFG) array (version 5.0 and 5.1) at 10-fold concentration intervals ranging from 0.02 μg/ml to 20 μg/ml to confirm dose-dependence. Alexa-Fluor 488 labeled goat anti-mouse IgG mAb (Thermo Fisher- Molecular Probes) was used for detection of the VLR-Fc recombinant fusion proteins. The binding conditions, analysis and screening of CFG arrays were done according to previously described methodology (Heimburg-Molinaro et al., 2011).

Protein Expression and Purification

N-terminally His₆-tagged ectodomains of RBC36, O13, and Tn4-22 were cloned into modified pFastBac1 (Thermo Fisher Scientific, Inc.) or phCMV3 vectors. The O13 (N81H, N82Q) double mutant was generated with a QuickChange kit (Agilent Technologies) using phCMV3-O13 as a template. Proteins were expressed in Hi5 cells using standard methods or in 293S cells by transient transfection. Secreted proteins were harvested from the media and purified with a Ni-NTA column (Qiagen) followed by size-exclusion chromatography (Superdex75, GE Healthcare) in 25mM HEPES pH7.5 and 200mM NaCl. Proteins were concentrated and used immediately or stored at −80°C before use.

Crystallization

RBC36 (apo) was concentrated to 17.3 mg/mL for crystallization. Crystals were grown at 4°C in 100mM Tris pH 8.5, 200mM MgCl₂, 20% PEG-8000. To obtain crystals of RBC36 in complex with 2′-fucosyllactose, RBC36 (11.4 mg/mL) was mixed with 2′-fucosyllactose at a final concentration of 6.9 mM prior to crystallization. Crystals were grown at 4°C in 140 mM ammonium chloride and 22% PEG-3350. Tn4-22 (apo) was concentrated to 21.6 mg/mL for crystallization. Crystals were grown at 20°C in 100 mM sodium acetate pH 4.6, 200 mM Li₂SO₄, and 2.7 M NaCl. For Tn4-22 in complex with H-trisaccharide, Tn4-22 (11 mg/mL) was mixed with H-trisaccharide at a final concentration of 1 mM prior to crystallization. Crystals were grown at 4°C in 85 mM sodium acetate pH 4.6, 170 mM ammonium acetate, 15% glycerol, and 25.5% PEG-4000. O13 (apo) was concentrated to 3.7 mg/mL for crystallization. Crystals were grown at 4°C in 140mM ammonium sulfate and 22% PEG-4000. For O13 in complex with H-trisaccharide, O13 (10 mg/mL) was mixed with H-trisaccharide at a final concentration of 1.9 mM prior to crystallization. Crystals were grown at 4°C in 100 mM citrate pH 5.2, 1 M LiCl₂, and 14% PEG-6000. For O13 in complex with LNnT, O13 (7.5 mg/mL) was mixed with LNnT at a final concentration of 1.25 mM prior to crystallization. Crystals were grown at 4°C in 100 mM sodium acetate pH 5.0, 200mM ammonium sulfate, and 20% PEG-4000. All crystals were grown in sitting drops, setup up using our robotic Rigaku CrystalMation^™ system (100nL + 100nL drops) or by hand (1μL + 1μL drops), using the vapor diffusion method.

Data Collection and Structure Determination

All crystals were flash-cooled in liquid nitrogen. Crystals of RBC36 (apo), RBC36: 2′-fucosyllactose, and Tn4-22:H-trisaccharide did not require additional cryoprotectant(s). Tn4-22 (apo) crystals were cryoprotected with addition of 25% ethylene glycol and all O13 crystals cryoprotected with 15% ethylene glycol. Diffraction data from single crystals were collected on beamlines BL12-2 at the Stanford Synchrotron Radiation Lightsource (SSRL) and 23ID-D or 23ID-B at the Advanced Photon Source (APS). Data were processed using XDS (Kabsch, 2010) or HKL2000 (Otwinowski and Minor, 1997). Data collection and processing statistics are reported in Table 1. Molecular replacement (MR) solutions for the initial O13 and Tn4-22 structures were found with Phaser in the Phenix suite (Adams et al., 2010) using RBC36 (with H-trisaccharide removed) (PDB ID 3E6J) as the search model. Refinement was carried out using Phenix and iterative re-building of the models into 2Fo-Fc and Fo-Fc difference electron density maps was carried out with Coot (Emsley et al., 2010). Minor improvements to the H-trisaccharide ligand conformation in the RBC36:H-trisaccharide complex (PDB 3E6J) were accomplished with a single round of refinement performed with Phenix using the deposited coordinates and structure factors. Progression of the refinement process was judged by monitoring Rcryst/Rfree. Structure validation was performed using the PDB Validation Server (validate.wwpdb.org). Refinement statistics are reported in Table 1.

Isothermal Titration Calorimetry

A Microcal Auto-iTC200 instrument (GE Healthcare) was used to perform isothermal titration calorimetry (ITC). Proteins were purified and dialyzed overnight in a buffer containing 25 mM HEPES pH7.5 and 250 mM sodium chloride. Lyophilized glycans (obtained from the Consortium for Functional Glycomics or purchased from Sigma) were resuspended in dialysis buffer. Glycans were placed in the syringe at a concentration of 1 to 2 mM. VLR or UEA-1 (purchased from Vector Laboratories) proteins were placed in the cell at 50–60 μM or120 μM, respectively. Protein concentrations were determined by UV absorbance at 280 nm using calculated extinction coefficients. Experiments were carried out at 10°C and consisted of 16 injections of 2.45 μL, with injection duration of 1 s, injection interval of 180 s, and reference power of 5 μCal. Fitting of integrated titration peaks was performed with Origin 7.0 software using a single-site binding model, which allowed for direct determination of the interaction affinity constant (Kd) and molar reaction enthalpy (ΔH).

DATA AND SOFTWARE AVAILABILITY

Structure factors and atomic coordinates of the VLR crystal structures have been deposited in the RCSB Protein Data Bank (http://www.rcsb.org/) under PDB ID codes 5UFD (RBC36 apo), 5UFF (RBC36: 2′-fucosyllactose), 5UEI (O13 apo), 5UF1 (O13:Htri), 5UF4 (O13:LNnT), 5UFB (Tn4-22 apo), 5UFC (Tn4-22:Htri).

Highlights.

1. Highly-specific anti-H-trisaccharide (O antigen) VLR antibodies can be identified

2. VLR specificity is superior to the widely used O blood typing reagent lectin UEA-1

3. Structural studies reveal how VLRs distinguish between closely related glycans

4. Structure-guided mutation can enhance VLR specificity