Structural basis of VHH-mediated neutralization of the food-borne pathogen Listeria monocytogenes

Listeria monocytogenes causes listeriosis, a potentially fatal food-borne disease. The condition is especially harmful to pregnant women. Listeria outbreaks can originate from diverse foods, highlighting the need for novel strategies to improve food safety. The first step in Listeria invasion is internalization of the bacteria, which is mediated by the interaction of the internalin family of virulence factors with host cell receptors. A crucial interaction for Listeria invasion of the placenta, and thus a target for therapeutic intervention, is between internalin B (InlB) and the receptor c-Met. Single-domain antibodies (VHH, also called nanobodies, or sdAbs) from camel heavy-chain antibodies are a novel solution for preventing Listeria infections. The VHH R303, R330, and R326 all bind InlB with high affinity; however, the molecular mechanism behind their mode of action was unknown. We demonstrate that despite a high degree of sequence and structural diversity, the VHH bind a single epitope on InlB. A combination of gentamicin protection assays and florescent microscopy establish that InlB-specific VHH inhibit Listeria invasion of HeLa cells. A high-resolution X-ray structure of VHH R303 in complex with InlB showed that the VHH binds at the c-Met interaction site on InlB, thereby acting as a competitive inhibitor preventing bacterial invasion. These results point to the potential of VHH as a novel class of therapeutics for the prevention of listeriosis.

Listeriosis is a potentially lethal food-borne disease caused by the Gram-positive bacteria Listeria monocytogenes. Although infections have a low rate of incidence in the general population, the disease has an unusually high mortality rate of 20 -30% (1). Listeria is transmitted by consumption of contaminated foods. Soft cheeses, deli meats, and ready-to-eat foods have historically been considered at high risk of Listeria contamination. Clinical presentation of listeriosis includes severe gastroenteritis; however, invasive infections can cross the blood-brain barrier, leading to central nervous system infections and fatal meningitis (2). Pregnant women are especially susceptible to Listeria infection due to T-cell suppression (3). The danger dur-ing pregnancy is further compounded by the capacity of the bacteria to cross the placental barrier, which can result in termination of the developing fetus (4). The pathogenesis of L. monocytogenes infection and invasion is well characterized (5) and points to potential avenues for the generation of novel therapeutic interventions.
The invasion of nonphagocytic cells by L. monocytogenes occurs through the action of a complex set of virulence factors that allow the bacteria to enter host cells, escape the vacuole, and hijack the actin network to spread from cell to cell (5). Listeria host cell entry is the initial step in pathogenesis, and it is mediated by two members of the internalin family of virulence factors (InlA 2 and InlB) (6,7). Binding of InlA and InlB to host cell receptors activates signaling cascades that trigger receptormediated endocytosis and internalization of the bacteria. InlA and InlB have different receptors and are responsible for mediating entry into different cell types and biological barriers.
The interaction of InlA with the host receptor E-cadherin is important for Listeria penetration of the intestinal barrier and invasion of several epithelial cell types (8). On the other hand, InlB binds the receptor tyrosine kinase c-Met (9), which permits Listeria internalization into a variety of cell types, including HeLa, Vero, and hepatocyte cell lines (7, 10 -13). c-Met functions as the receptor for the hepatocyte growth factor and is required for normal embryonic development, pointing to the importance of InlB in pregnancy-related listeriosis. Indeed, synergistic action of InlA and InlB is required for L. monocytogenes to cross and penetrate the placental barrier (14). Given the importance of InlB receptor interaction in fetal listeriosis, disruption of this interaction may represent a target for therapeutic intervention.
Although interruption of the InlB-c-Met interaction is an intriguing approach for preventing Listeria cellular invasion, one potential pitfall is that the protein is buried in the peptidoglycan layer. One innovative solution is to use single-domain antibodies (V H H), derived from the antigen-binding fragment of the heavy-chain antibodies found in camelids (17). V H H are 10 times smaller (12-15 kDa) than conventional IgG antibodies (150 kDa) and may be able to penetrate the Listeria cell wall to bind InlB.
Previously, four V H H (R303, R326, R330, and R419) that bind the LRR domain of InlB with nanomolar affinity were isolated from a nonimmune phage display library (18,19). As the InlB-LRR domain is crucial for interaction with c-Met, we hypothesized that these V H H could inhibit bacterial endocytosis and protect the cells from Listeria invasion. We demonstrate that InlB-specific V H H effectively neutralize Listeria invasion in vitro. Furthermore, high-resolution X-ray structures reveal that the mechanism behind V H H-mediated Listeria inhibition is competitive inhibition of the InlB-c-Met interaction.

V H H bind overlapping epitopes on InlB
Previous work had identified several V H H (R303, R330, R419, and R326) from a preimmune (naive) phage display library that bound the LRR domain of InlB (residues 36 -249; InlB 249 ) (18). Using an indirect ELISA the relative affinity of the V H H for the LRR domain of InlB was compared (Fig. 1A). A variable concentration of immobilized InlB 249 was detected using a fixed concentration of biotinylated V H H. Consistent with previously reported surface plasmon resonance results (18), V H H R303, R326, and R330 bound to immobilized InlB 249 with a similar apparent affinity (Fig. 1A). However, no binding of V H H R419 to InlB 249 was observed.
Because the V H H were originally generated by screening a phage display library against a truncated version of InlB with only the cap and the LRR domain (InlB 249 ) , we next investigated whether the V H H would also bind InlB if the IR domain was also present (InlB 321 ) (Fig. 1B). R303 and R330 both bound to InlB 321 with similar affinity; however, R326 bound with ϳ2-fold lower affinity to InlB 321 when compared with R303 and R330 (Fig. 1B).
The V H H R303, R330, and R326 displayed variability in their CDR sequences and canonical CDR cluster classification (Table  1). Furthermore, based on nucleotide sequence alignments with antibody germ line segments, the V H H may be derived from different species of Camelid. This was expected, as the phage display library used to isolate these V H H is originated from the immune repertoire of three species of Camelid (18,19). R303 is unquestionably from Camelus dromedaries, whereas R330, R326, and R419 are derived from either Llama glama or Vicugna pacos. Given this diversity, we next investigated the epitope specificity of the V H H.
The possibility that the V H H bound distinct InlB epitopes was investigated using a competitive ELISA. A single fixed concentration of InlB 249 or InlB 321 was immobilized, and a mixture of biotinylated R303, R326, or R330 was added along with an 80-fold higher concentration of unlabeled R303 to act as an inhibitor (Fig. 1C).
Assuming the V H H bound to spatially distinct epitopes, the expectation was that R303 would not act as an inhibitor for the other V H H. On the other hand, if the V H H bound to overlapping epitopes, R303 should inhibit binding. In all cases, R303

V H H neutralization of Listeria
acted as an inhibitor to binding, suggesting that all three V H H (R303, R330, and R326) bound overlapping epitopes (Fig. 1C).

InlB-specific V H H inhibit Listeria invasion of HeLa cells in vitro
Interaction of InlB with the host cell c-Met receptor is essential for Listeria invasion of epithelial cells (11,12). Interference of this interaction may provide a site for therapeutic intervention by preventing Listeria colonization and invasion. Gentamicin protection assays were employed to determine whether InlB-specific V H H could inhibit Listeria invasion in vitro.
L. monocytogenes were treated with the four V H H (R303, R330, R326, and R419) and allowed to invade HeLa cells in vitro. InlB 249 was used as a positive control, as it has been previously shown to inhibit Listeria invasion of HeLa cells (10), and an irrelevant anti-GFP V H H (20) was used as a negative control. Following protein treatment, gentamicin was added to eradicate noninternalized Listeria. HeLa cells were lysed, the internalized bacteria were counted, and the efficiency of Listeria invasion was calculated (Fig. 2).
As R303, R326, and R330 bound overlapping epitopes on InlB 249 and InlB 321 (Fig. 1), it was hypothesized that these three V H H would perform similarly in the invasion assay. However, there were some differences in the ability of V H H to inhibit Listeria invasion. R303 and R330 were both highly effective at inhibiting Listeria internalization of HeLa cells (94 Ϯ 1.4 and 75 Ϯ 1.9%, respectively; Fig. 2). However, R326 exhibited a reduced ability to inhibit Listeria invasion of HeLa cells (36 Ϯ 5.5%). Given that R419 did not bind InlB ( Fig. 1), it was not surprising that the V H H resulted in a level of invasion inhibition similar to that of the irrelevant V H H control (Fig. 2).
As a second line of evidence to evaluate V H H neutralization of Listeria, a fluorescence microscopy-based invasion assay was conducted (21). A strain of constitutively expressing GFP Listeria was constructed, followed by biotinylation and subsequent invasion of the strain into HeLa cells. Listeria cells were treated with PBS (negative control), an irrelevant GFP-specific V H H (negative control), InlB 249 (positive control), and each of the InlB-specific V H H. Following invasion, the cells were treated with streptavidin conjugated to DyLight550. If the GFPexpressing Listeria invaded the HeLa cells, they would not be detected by the red-labeled streptavidin; however, if the Listeria were impeded from cell invasion, they would be available for detection by the labeled streptavidin and would thus be stained red.
The negative controls, PBS and irrelevant V H H, resulted in minimal red staining of the GFP-expressing Listeria, indicating that the strain had invaded the HeLa cells (Fig. 3). When treated with the InlB 249 -positive control and with R303, R330, and R326, the majority of the Listeria cells were stained red, indicating that they remained extracellular to the HeLa cells and had been inhibited from invasion (Fig. 3). R419 did not inhibit Listeria invasion, consistent with the results of the ELISA and gentamicin protection assays (Figs. 1-3).

Structures of V H H R303, R326, and R330
X-ray structures of V H H R303, R326, and R330 were determined at resolutions ranging from 1.3 to 1.8 Å (Table 2). Consistent with the divergent amino acid sequences of the V H H CDRs (Table 1), the X-ray structures revealed variability in the antigen-binding sites. The CDRs were assigned using the definitions reported by North et al. (22). The CDR loop conformations were assigned from the X-ray structures using the PyIgClassify CDR loop database (23) ( Table 1).

V H H neutralization of Listeria
R303 was solved to a resolution of 1.3 Å, and the structure contains two molecules in the asymmetric unit arranged in a head-to-tail fashion. R303 had the longest CDR-3 of the three V H H with a length of 16 amino acids (Table 1 and Fig. 4A). A noncanonical disulfide bond was formed between CDR-1 and CDR-3 (residues 33-102) that linked the long 16-residue CDR-3 loop against the framework region of the antibody (Fig.  4A). CDR-3 formed a short helical segment (residues 102-107) in proximity to the noncanonical disulfide bond. The fixing of CDR-3 against the framework region resulted in a large solventaccessible surface area (1970 Å 2 ) available for antigen recognition. The CDR-1 loop bisects the antibody paratope, creating two relatively flat interaction surfaces on either side of the loop. The paratope region between CDR-1 and CDR-3 showed a positively charged electrostatic surface, with a wide pocket-like structure forming (Fig. 4A).
The structure of V H H R326 was solved to a resolution of 1.8 Å and contained a tetramer in the asymmetric unit. Unlike R303, R326 had no disulfide bond connecting CDR-3 to CDR-1. Structurally, R326 was distinct from the other two V H H with the three CDR loops protruding from the framework region, forming a convex paratope structure (Fig. 4B). The paratope was a large solvent-accessible surface area (1650 Å 2 ) with a positively charged electrostatic surface (Fig. 4B).
The structure of V H H R330 was solved to a resolution of 1.6 Å and contained a dimer in the asymmetric unit. Similar to R326, the paratope of R330 was a wide, roughly convex shape with a positively charged solvent-accessible surface area of 2050 Å 2 (Fig. 4C). Interestingly, the structure of CDR-1 of R330 did not fall into one of the previously characterized structural clus-ters identified by North et al. (22) (Table 1). CDR-1 also was disordered at the apex of the loop (residue 28 in chain A; residues 29 and 30 in chain B).

Structure of R303-InlB 249 and R303-InlB 321
To determine the molecular mechanism behind V H H neutralization of Listeria invasion, the structures of V H H R303 in complex with the LRR domain of InlB (InlB 249 ) and the longer InlB fragment of the LRR domain linked to the IR region (InlB 321 ) were both determined to a resolution of ϳ1.5 Å. The two complex structures crystallized in different space groups ( Table 2). R303 in complex with InlB 249 crystallized as a monomer, whereas R303 with InlB 321 was a dimer in the asymmetric unit.
The overall binding interactions between R303 and InlB 249 and InlB 321 were identical (Fig. 5A), indicating that the IR domain of InlB 321 played no role in binding. This finding is consistent with the observation that R303 binds to both proteins (InlB 249 and InlB 321 ) with similar affinity (Fig. 1, A and B).
The entire interaction between R303 and InlB occurs on an electronegative cavity on the concave face of the InlB-LRR domain, resulting an approximate buried surface area of 1400 Å 2 . The bulk of the binding interactions are mediated by CDR-3 and CDR-2, with CDR-1 displaying only limited contact with InlB (Fig. 5B).
Consistent with the picomolar affinity of R303 for the InlB-LRR domain (18), there were extensive polar and nonpolar contacts between the antibody and InlB. Interactions originating from CDR-3 on R303 are of central importance and form the majority of the binding interactions (Fig. 5). There were a series of salt bridges that likely contribute significantly to the highaffinity binding of the V H H. The salt bridges are formed between Arg-100 vhh on CDR-3 of R303 and Glu-194 inl and Glu-236 inl on InlB (where the superscript "vhh" denotes residues on V H H R303 and the superscript "inl" denotes residues on InlB) (Fig. 5B). This central arginine residue on R303 also forms a hydrogen bond to Tyr-214 inl . Additional polar interactions include 12 hydrogen bonds between the antibody and InlB. On CDR-3, Asn-103 vhh hydrogen-bonds to Ser-168 inl , Asp-189 inl , and Thr-190 inl . The adjacent residue on CDR-3, Thr-104 vhh , hydrogen-bonds to the hydroxyl side chain of Tyr-170 inl (Fig.  5B). On CDR-2, Ser-56 vhh and Ser-57 vhh form hydrogen bonds to Asp-233 inl (Fig. 5B). In addition to the polar contacts, there are aromatic stacking interactions, with the side chain of Phe-104 vhh on CDR-3 inserting between Tyr-214 inl and Tyr-170 inl (Fig. 5B).

Specificity of V H H isolated from a nonimmune library
The anti-InlB V H H (R303, R330, and R326) used in this study were isolated from a preimmune phage display library from the naive immune repertoires of camels, alpacas, and llamas (18). Each of the isolated V H H was unique in terms of primary sequence diversity and CDR canonical structure (Table 1). Furthermore, based on alignment with germ line gene segments, the V H H originate from different species of Camelid (R303 (camel), R326 (llama or alpaca), and R330 (llama or alpaca)). However, despite this structural and sequence diversity, the

V H H neutralization of Listeria
specificity of the V H H converged onto a single epitope (Fig. 1C). This epitope was centralized to a negatively charged cavity on the concave face of the LRR domain of InlB (Fig. 5A).
The specific structural features of the InlB antigen and the particular binding properties associated with V H H in general may be responsible for the observed V H H specificity. It has been

V H H neutralization of Listeria
observed previously that V H H often bind concave features on protein antigens due to the convex shape of the paratope formed on the three CDR loops (24,25). Given this preference, the V H H specificity toward the InlB-LRR electronegative cavity may be the result of the protein only having this one concave surface feature.

V H H properties facilitate neutralization of Listeria
The biophysical and binding properties of V H H are distinct compared with traditional monoclonal antibodies. V H H are small and stable, and their convex shape allows V H H to bind protein cavities, which are frequently inaccessible to traditional monoclonal antibodies (17). This combination of properties provides several advantages that may have contributed to the effectiveness of V H H R303, R330, and R326 for the in vitro neutralization of L. monocytogenes (Figs. 2 and 3). In particular, the small size and preferential binding of V H H toward protein cavities may explain the success of V H H at Listeria neutralization compared with traditional antibody formats.
Several mouse anti-InlB antibodies displayed variable effectiveness at inhibiting Listeria invasion of Vera cells, suggesting that specific epitopes must be recognized for neutralization to occur (10). In some cases, InlB epitopes may be inaccessible; an InlB-specific ScFv was only able to bind InlB following enzymatic digestion of the bacterial cell wall, suggesting that the epitopes were buried in the cell wall (26). As V H H R303, R330, and R326 are all able to neutralize Listeria invasion (Figs. 2 and  3), it can be inferred that the immunodominant epitope must be accessible to the V H H. The small size of the V H H may facilitate penetration of the bacterial peptidoglycan layer to access the protein-protein interaction surface on InlB. This further highlights the specific advantages of using V H H in targeting difficult-to-access cell surface epitopes.

InlB-specific V H H inhibit Listeria invasion through competitive inhibition
The neutralization of Listeria invasion by V H H R303, R330, and R326 could potentially be mediated by two different mechanisms. The V H H could bind InlB and inhibit its interaction with c-Met simply through steric effects, or the V H H could competitively inhibit the native interaction of InlB with c-Met. The X-ray structure of R303 in complex with InlB (Fig. 5) permits an analysis of the molecular mechanism behind the antibacterial activity of the V H H. c-Met is a receptor tyrosine kinase whose ectodomain consists of six domains: Sema, Psi, and four Ig-like domains (Ig1-4) (27). The natural ligand for c-Met is the hepatocyte growth factor/scatter factor (HGF/SF). In healthy cells, the c-Met-HGF/SF interaction mediates cell signaling related to embryogenesis and tissue regeneration, and deregulation of c-Met is also important in carcinogenesis (28). Interestingly, whereas L. monocytogenes hijacks c-Met as a vehicle for bacterial entry, the interaction of InlB with c-Met does not mimic the natural HGF/SF ligand, as the two proteins bind c-Met at distinct sites (9,27).
InlB-c-Met receptor binding and subsequent cell signaling events that ultimately result in bacterial internalization are mediated by different domain-domain interactions. A frag-ment comprising the cap region and LRR domain of InlB (InlB 241 ) is the minimum unit for c-Met receptor binding (9). The binding of the InlB 241 fragment to the Ig1 domain of c-Met occurs at the electronegative cavity on the concave face of the InlB-LRR domain (Fig. 6A) (27). However, c-Met receptor activation and cell invasion by L. monocytogenes require a larger fragment of InlB, consisting of the cap region, LRR domain, and interrepeat (InlB 321 ) (16). The secondary, weaker interaction of the InlB-IR domain with the c-Met Sema domain (Fig. 6A) is required for receptor activation and not binding (27).
V H H R303 binds InlB directly at the c-Met receptor-binding site: the electronegative cavity on the concave face of InlB  (Figs. 5A and 6A). Overlap of the structure of R303-InlB 321 with that of the c-Met ectodomain in complex with InlB 321 (27) demonstrated that R303 would directly occupy the same physical space as the c-Met Ig1 domain, mimicking the interaction with the Ig1 domain (Fig. 6B).

V H H neutralization of Listeria
The binding of the c-Met Ig1 domain to InlB is mediated by many of the same residues involved in the R303-InlB interaction. There are five residues on InlB that are important for c-Met receptor binding: Asp-128 inl , Glu-150 inl , Tyr-170 inl , Tyr-214 inl , and Trp-124 inl (27). Of these five InlB residues, four are either interacting with R303 directly through hydrogen bond interactions (Tyr-170 inl and Tyr-214 inl ; Figs. 5B and 6C) or are buried upon complex formation (Glu-150 inl and Trp-124 inl ; Fig. 6C).
The high-affinity binding of R303 to the c-Met receptorbinding site on InlB provides a clear molecular mechanism for the neutralization of L. monocytogenes by the V H H used in this study. By mimicking the interactions of c-Met, the natural ligand of InlB, the V H H are acting as high-affinity competitive inhibitors, neutralizing bacterial invasion.

Therapeutic potential of Listeria-specific V H H
Listeria infections are a particular challenge facing pregnant women. Maternal infection is frequently asymptomatic or displays nonspecific symptoms, making diagnosis a serious challenge during prenatal care (29,30). Even in cases with diagnosis, antibiotic treatment is not always successful, presumably due to the intracellular nature of the pathogen (31).
Prevention of Listeria infection is currently the most effective strategy for safeguarding women from the disease during pregnancy. Typically, pregnant women are advised to avoid consumption of foods at high risk of Listeria contamination. However, a series of deadly Listeria outbreaks in fresh produce, fruit, and other foods traditionally at low risk of Listeria contamination, highlight the need for alternative and novel approaches to safeguarding the food supply (32)(33)(34).
A prophylactic strategy of blocking Listeria entry into nonphagocytic cells by inhibiting the interaction of InlB with the c-Met receptor is a potential venue of Listeria treatment or prophylactic. A recent report using the c-Met inhibitor tanespimycin as a Listeria antibiotic suggests that this approach may represent a viable therapeutic strategy (35).
The ability of InlB-specific V H H to neutralize Listeria invasion in vitro points to a therapeutic potential for the prevention or treatment of listeriosis. There have been several recent reports of using V H H as anti-bacterial agents against a variety of bacterial pathogens, including Clostridium difficile, Bacillus anthracis, Shigella, botulism, and Bordetella pertussis (36 -40). In each of these cases, the anti-bacterial strategy was to employ the high-affinity binding of V H H to neutralize secreted bacterial toxins. The use of InlB-specific V H H represents a novel approach to combating bacterial disease using V H H. The dependence of the internalin-host cell receptor interaction in Listeria pathogenesis provides a novel mechanism of V H H-mediated therapeutic intervention by inhibiting host cell invasion (Figs. 2 and 3). Although further in vivo studies are required to validate the therapeutic potential of V H H for the treatment and prevention of listeriosis, the results presented here highlight the future potential of V H H as anti-bacterial agents.

Expression and purification of V H H
The plasmids (pSJF2H) for V H H R303 and R330 were a generous gift of Dr. Roger MacKenzie (National Research Council, Ottawa, Canada). Genes for V H H R419 and R326 were codonoptimized and synthesized as double-stranded gene blocks (GenScript, Piscataway, NJ). R326 and R419 were cloned into the plasmid pET22b using the restriction enzyme sites NcoI and XhoI.
Plasmids for R303 and R330 were transformed into Escherichia coli TG1, whereas R419 and R326 were transformed into E. coli BL21 (DE3) for protein expression. All of the InlB-specific V H H were extracted from the periplasm using an osmotic shock procedure and purified using Ni-NTA chromatography and size-exclusion chromatography, as described previously (41). The control anti-GFP V H H was expressed and purified as described previously (20).

Indirect and competitive ELISA
For the indirect ELISA, a 96-well plate was coated with serial dilutions of InlB (5-0.02 ng/l) in PBS overnight at 4°C. The wells were blocked for 1 h with BSA (3% in PBS). Biotinylated V H H (R303, R330, R326, and R419) were used as a primary antibody (15 g/ml, 1 h). The plate was washed three times with PBS-Tween (0.05% Tween 20) followed by the addition of streptavidin horseradish peroxidase (Fisher) (1:50,000 dilution in 3% BSA 1 h). Finally, detection was carried out by the addition of 3,3,5,5-tetramethyl benzidine (15 min). The reaction was stopped by the addition of 0.18 M H 2 SO 4 , and the absorbance was measured at 450 nm using a plate reader (BioTek Instruments Inc., Winooski, VT).
A similar procedure was carried out for the competitive ELISA except that InlB was immobilized at a fixed concentra-V H H neutralization of Listeria tion (10 g/ml), and a mixture of biotinylated V H H (15 g/ml) and unlabeled R303 (80 g/ml) was added as a competitor.
Treatment solutions of InlB 249 and V H H diluted to 100 g/ml in unsupplemented RPMI 1640 were added to a 24-well cell plate containing 1 ϫ 10 5 HeLa cells/well and incubated for 30 min at 37°C, 5% CO 2 . Log phase L. monocytogenes (MOI of 50:1) were then added to the wells, and the plate was centrifuged (1000 rpm for 3 min) and incubated at 37°C with 5% CO 2 for 1 h. Infected cells were washed twice with PBS to remove nonadherent bacteria. To kill extracellular bacteria, RPMI 1640 containing 100 g/ml gentamicin was added and incubated for 60 min (37°C, 5% CO 2 ). To enumerate intracellular bacteria, wells were washed once with PBS and then lysed with 1% Triton X (Sigma) in PBS at the appropriate times. Recovered intracellular bacteria were quantified by plating serial dilutions on LB agar plates and enumerating colony counts.
Replicate wells were included in which total and surfaceadherent Listeria were enumerated by harvesting the supernatant immediately after incubation of bacteria with HeLa cells (total) or collecting the Triton X-100 lysate before treatment with gentamicin (adherent). Each experiment was done in duplicates, and duplicates were performed at least three times independently.

Fluorescence microscopy
GFP-expressing L. monocytogenes were created as described previously (42). HeLa cells were cultured in 1ϫ RPMI 1640 culture medium (HyClone) containing 2.05 mM L-glutamine, 10% FBS, and penicillin/streptomycin and incubated at 37°C with 5% CO 2 . HeLa cells were seeded at a density of 4 ϫ 10 5 cells/ml onto a microscope coverglass placed in each well of a 24-well plate. GFP-Listeria was grown overnight in BHI broth containing antibiotics, and the concentration was measured at A 600 . The bacteria were washed three times with sterile 1ϫ PBS (pH 7.4) and labeled with 0.5 mg/ml EZ-Link Sulfo-NHS-LC-Biotin (Thermo Scientific). After quenching excess biotin by washing three times with 1% BSA, the bacteria were incubated with 100 g/ml nanobodies at 37°C for 30 min. HeLa cells were stained with 1 l of 10 g/ml 4Ј,6-diamidino-2-phenylindole and infected with biotinylated GFP-Listeria at an MOI of 50:1. After centrifuging for 15 min at 300 rpm, the plate was incubated for 1 h at 37°C with 5% CO 2 followed by three washes with unsupplemented RPMI 1640. Biotinylated GFP-Listeria were detected by the addition of 2.5 l of 1 mg/ml Streptavidin-Dylight550 (Thermo Scientific) to each well, and the plate was incubated for 30 min at 37°C with 5% CO 2 . The wells were washed with RPMI 1640, and the coverslips were fixed with 4% p-formaldehyde for 30 min at 4°C. After washing the wells three times with 1ϫ PBS (pH 7.4), the coverslips were removed from the plate, and Fluoromount-G (SouthernBiotech) was added to mount them onto slides. The slides were analyzed in a Leica DMI3000 B fluorescence microscope at ϫ63 magnification.
Crystallization trials were carried out in Intelli 96-well sitting-drop plates using a Gryphon crystallization robot (Art Robbins Instruments). Sitting-crystal drops were set up using 1 l of protein and 1 l of reservoir solution. The proteins were screened using the PEGs, PEG II, and PACT crystallization suites (Qiagen Inc.

Data collection and X-ray structure determination
Crystals were dipped in cryoprotectant (mother liquor supplemented with 25% glycerol) and flash-frozen in liquid nitrogen. X-ray data were collected at the Canadian Light Source on beamline 08ID-1 (43). Diffraction data were processed using Xia2 (44). All structures were solved by molecular replacement using Phaser as implemented in Phenix (45). For molecular replacement, the previously solved structures of R303 (41) and InlB 241 (46) and InlB 321 (27) were used as search models. The structure was automatically built and refined using Phenix. Manual fitting of A-weighted F o Ϫ F c electron density maps was carried out using Coot (47). The final model and refinement statistics are given in Table 2.