Cryo-electron tomography of Birbeck granules reveals the molecular mechanism of langerin lattice formation

Langerhans cells are specialized antigen-presenting cells localized within the epidermis and mucosal epithelium. Upon contact with Langerhans cells, pathogens are captured by the C-type lectin langerin and internalized into a structurally unique vesicle known as a Birbeck granule. Although the immunological role of Langerhans cells and Birbeck granules have been extensively studied, the mechanism by which the characteristic zippered membrane structure of Birbeck granules is formed remains elusive. In this study, we observed isolated Birbeck granules using cryo-electron tomography and reconstructed the 3D structure of the repeating unit of the honeycomb lattice of langerin at 6.4 Å resolution. We found that the interaction between the two langerin trimers was mediated by docking the flexible loop at residues 258–263 into the secondary carbohydrate-binding cleft. Mutations within the loop inhibited Birbeck granule formation and the internalization of HIV pseudovirus. These findings suggest a molecular mechanism for membrane zippering during Birbeck granule biogenesis and provide insight into the role of langerin in the defense against viral infection.


Introduction
The epidermis and mucosa provide the first line of defense against pathogens. In these tissues, specialized antigen-presenting cells called Langerhans cells capture pathogens by recognizing their surface mannose-containing oligosaccharides via the C-type lectin langerin (Valladeau et al., 2003;Doebel et al., 2017). Pathogens internalized via langerin are enclosed in a specialized organelle called a Birbeck granule, which has a central striated lamella between closely apposed membranes (Birbeck et al., 1961;Valladeau et al., 2000;McDermott et al., 2004;Valladeau et al., 1999). Once pathogens are captured, Langerhans cells undergo maturation and migrate to peripheral lymph nodes, where they present processed antigens to T cells (Merad et al., 2008;Rhodes et al., 2021).
Langerin is a type II transmembrane protein that is composed of a cytoplasmic tail, coiledcoil neck region, and carbohydrate recognition domain (CRD) (Stambach and Taylor, 2003;Valladeau et al., 1999;Valladeau et al., 2000) ( Figure 1A). Through the coiled-coil neck region, langerin forms a trimer, which is functionally important for oligosaccharide binding (Stambach and Taylor, 2003;Weis and Drickamer, 1996;Weis and Drickamer, 1994). The calcium-dependent carbohydrate-binding capacity of langerin CRD is essential for the formation of Birbeck granules because the removal of calcium ions causes them to unzip and disintegrate (Stambach and Taylor, 2003;Bartosik, 1992). It has been proposed that the "faceto-face" interaction between the langerin trimers via CRDs confers the characteristic zippered membranes with a central striation of Birbeck granules (Thepaut et al., 2009) (Figure 1B and 1C). However, the precise molecular mechanism of Birbeck granule formation is unknown because of the lack of a high-resolution structure.

The resistance of Langerhans cells to HIV-1 infection is mediated by the efficient internalization
of HIV-1 into Birbeck granules and subsequent autophagic degradation (de_Witte et al., 2007;Ribeiro et al., 2016). Therefore, elucidating the mechanism by which HIV-1 is internalized into Birbeck granules is essential to understanding the role of Langerhans cells as a barrier to HIV-1 infection.
In the present study, we reconstructed the 3D structure of isolated Birbeck granules using cryoelectron tomography and identified a flexible loop region that is crucial for Birbeck granule formation and virus internalization.

Cryo-electron microscopy of isolated Birbeck granules
Exogenous expression of langerin in mammalian cell lines readily reproduces the formation of Birbeck granules (McDermott et al., 2004). We found that the addition of yeast mannan to langerin-expressing cells enhances the formation of large Birbeck granules ( Figure 1C). To purify these Birbeck granules, we expressed SNAP-tagged human langerin in 293T cells, biotinylated the tag by incubation with benzylguanine-biotin, and immunoprecipitated the Birbeck granules using streptavidin agarose (Figure S1A and S1B). Under cryo-electron microscopy, the isolated Birbeck granules were observed to be thin lamellar vesicles of assorted sizes, similar to those reported previously ( Figure 1D, 1E, and S1C) (Sagebiel and Reed, 1968;Lenormand et al., 2013). Unlike regular 2D crystals of proteins, Birbeck granules appear wavy, suggesting structural heterogeneity. The 2D classification of the segmented images extracted from the micrographs showed nearly three-fold symmetrical arrangements of pores with a diameter of approximately 10 nm; however, no clear crystalline lattice was observed ( Figure 1F). Although most Birbeck granules were oriented in approximately the same direction with respect to the beam axis, some granules were twisted, providing different views that were sufficient for 3D reconstruction ( Figure S1D).

Subtomogram averaging of the repeating unit of Birbeck granules
We conducted cryo-electron tomography of the isolated Birbeck granules and reconstructed the repeating units by averaging the subtomograms (Figure 2 and S2A). The initial density map ( Figure 2A) shows a langerin trimer (blue) with three inverted trimers (red) attached to the CRDs. However, because of the flexibility between each langerin trimer, the resolution was limited to ~24 Å. To improve the resolution, we generated a mask enclosing the central langerin trimer and three attached CRDs and performed focused refinement and 3D classification ( Figure 2B). The resolution was improved to 6.4 Å ( Figure S2B), which clearly visualized the alpha helices and enabled us to generate a model based on the previously reported crystal structure of the langerin trimer (Feinberg et al., 2010). We then performed a two-body refinement to visualize the spatial relationship between the two opposing langerin trimers (Nakane et al., 2018). Based on this refinement ( Figure 2C), we built a honeycomb lattice model ( Figure 2D and Movie S1), which agrees with the striation of the Birbeck granules observed by ultra-thin section electron microscopy and the ~10 nm pores visualized in the 2D classification images ( Figure 1C and 1F) (Birbeck et al., 1961). However, in reality, the honeycomb lattice is non-uniform because of the structural heterogeneity.

Interaction between the two langerin trimers via flexible loops
An in-depth analysis of the interaction region between the two CRDs ( Figure 3) revealed that the interaction is mediated by a loop at residues 258-262, which is highly flexible in the crystal structure (Feinberg et al., 2010). The interaction between the two CRDs appears to be maintained by docking the loops of two langerin molecules into each other's lateral cleft, which constitutes the secondary carbohydrate-binding site ( Figure 3A, pink circles) (Chatwell et al., 2008). The interaction between the two CRDs is flexible because the first eigenvector of the multi-body analysis represents an approximately 30° tilting motion with the loop as the fulcrum ( Figure 3B, and Movie S2). Interestingly, loop 258-262 of the inverted molecule appears to participate in the secondary carbohydrate-binding site of the upright molecule ( Figure 3A). A previous study found that the secondary calcium-independent binding site has a lower affinity for carbohydrate ligands than the primary calcium-dependent binding site (Chatwell et al., 2008). The docked loop 258-262 may provide additional interactions with the hydroxyl group of mannose via Glu261 and Asp263, thus increasing the affinity of the secondary binding site when two langerin trimers form a complex. Furthermore, the Trp264Arg mutation is associated with a lack of Birbeck granules (Verdijk et al., 2005;Figdor et al., 2002). Trp264 may also be involved in docking the loop into the cleft; however, it is more likely that disruption of the hydrophobic core of the CRD is responsible for this defect in Birbeck granule formation (Chatwell et al., 2008).

Loop 258-262 is essential for Birbeck granule formation
To investigate the importance of loop 258-262 in the biogenesis of Birbeck granules, we mutated three residues in the loop: Met260Ala, Glu261Arg, and Asp263Lys, and generated three mutated langerin constructs, each carrying one (MRGD), two (MRGK), or three mutations (ARGK). We expressed these three langerin mutants in 293T cells and induced Birbeck granule formation by adding yeast mannan ( Figure 4A). Under ultra-thin section microscopy, mannan-induced Birbeck granules usually formed a cluster, and wild-type langerin produced Birbeck granules a few micrometers in length ( Figure 4A, MEGD WT).
As the number of mutations increased, the length of Birbeck granules decreased. With two point mutations (MRGK), some cross-sections became spherical rather than rod-shaped, suggesting that the Birbeck granules failed to extend. With three mutations (ARGK), clusters of Birbeck granules were not observed, with isolated individual granules observed instead.
The length of individual Birbeck granules and the total length of granules per area were measured to confirm the above observations ( Figure 4B and 4C). These results demonstrate the importance of loop 258-262 in the biogenesis of Birbeck granules.

Loop 258-262 is crucial for the internalization of HIV pseudovirus
To investigate the impact of the mutations in loop 258-262 on virus internalization by Birbeck granules, we added lentiviruses pseudo-typed with the HIV-1 env protein to langerinexpressing 293T cells. After 30 min of incubation, we retrieved viruses attached to the surface langerin by washing with EDTA buffer, removed non-specifically attached viruses by trypsinization, and isolated the Birbeck granules using streptavidin-agarose ( Figure 5). A langerin mutant that lost its calcium-binding capacity (Stambach and Taylor, 2003) was used as the negative control ( Figure 5, lectin (-)). The HIV pseudoviruses attached to both wildtype and ARGK mutant langerin and the attachment was abolished by the addition of mannan ( Figure 5A). This indicated that the mutation of the loop did not disrupt the binding of the glycosylated envelope proteins to the primary carbohydrate-binding sites. In contrast, the internalization of the pseudovirus was significantly inhibited by the ARGK mutation ( Figure   5B and 5C). These results suggest that disruption of Birbeck granule formation by the mutation of loop 258-262 inhibited the internalization of HIV pseudovirus.

Virus internalization mechanism of Birbeck granules
In the present study, we reconstructed the 3D structure of isolated Birbeck granules and found that docking of the flexible loop 258-262 into the secondary carbohydrate-binding cleft is essential for assembly of the langerin lattice. In agreement with this observation, we showed that mutations introduced into the loop inhibited Birbeck granule formation and internalization of HIV pseudovirus. Based on the structure of the glycosylated HIV env protein (Stewart-Jones et al., 2016;Pan et al., 2020), we propose a model of HIV internalization by langerin ( Figure 6A and 6B). A previous study proposed that N-linked high-mannose oligosaccharides bind to langerin at both primary and secondary carbohydratebinding sites (Chatwell et al., 2008). Our model requires bending of the coiled-coil necks of the two langerin trimers to form a complex with HIV-1 env via the high-mannose oligosaccharide ( Figure 6B). Although the triple coiled-coil structure of the neck domain does not seem to bend strongly, the coiled-coil prediction profile suggests that the N-terminal part of the neck does not necessarily form a rigid triple coiled coil (Lupas et al., 1991) ( Figure   S1E). Thus, we speculate that the neck region of the langerin trimer is flexible and swings on the cell surface, allowing loop 258-262 of one trimer to dock into the cleft of an adjacent trimer. Once the triad structure, consisting of two langerin trimers and one env protein, is formed, the elasticity of the neck region can deform the cell membrane and initiate invagination. The formation of clathrin pits may also be involved in membrane invagination, because the cytoplasmic ends of Birbeck granules are coated with clathrin (Mc et al., 2002).
Additionally, it is possible that other intracellular mechanisms such as the actin cytoskeleton, are involved in this membrane deformation process because annexin A2 is required for the proper formation of Birbeck granules (Thornton et al., 2020).

Significance of langerin-specific secondary carbohydrate-binding cleft
It is noteworthy that the docking of loop 258-262 into the secondary calcium-independent carbohydrate-binding cleft, which is not found in other C-type lectins (Chatwell et al., 2008), mediates the assembly of the langerin lattice. In DC-SIGN, which is another C-type lectin expressed in dendritic cells, and is closely related to langerin, the region corresponding to the secondary carbohydrate-binding site of langerin forms a shallow pocket, that is occupied by two calcium ions ( Figure 7A) (Guo et al., 2004;Geijtenbeek et al., 2000;Geijtenbeek et al., 2000). This structural difference between langerin and DC-SIGN may account for the inability of DC-SIGN to form Birbeck granules despite its high sequence similarity to langerin (Cambi et al., 2009). Another characteristic of the secondary carbohydrate-binding site of langerin is its weak affinity for mannose and maltose (Chatwell et al., 2008). This may reflect that the sugar-binding surface of langerin CRD is less negatively charged than that of DC-SIGN CRD ( Figure 7A) (Chatwell et al., 2008). Moreover, the interaction between the secondary binding site and mannose/maltose is mainly mediated by backbone amide and carbonyl groups (Chatwell et al., 2008). It is possible that the side chains of Glu261 and Asp263 of loop 258-262 docked into the cleft of an adjacent langerin CRD provide the negative charge required for stabilizing oligo-mannose bound to the secondary binding site ( Figure 7A and B). In agreement with this hypothesis, in the present study the Asp263Lys mutation caused a significant defect in Birbeck granule formation. The secondary carbohydrate-binding site may only be fully functional when two langerin trimers are bound to each other (Figure 6 and 7B). The effect of the Met260Ala mutation on Birbeck granule formation was unexpected, because the mutation was supposed to be silent in terms of side chain charge. The observed structural variation of the langerin trimer complex suggests that the side chain of Met260 acts as a hook to stabilize the interface between the two intersecting loops.
Our model of Birbeck granule formation suggests that oligosaccharide ligands are captured by the lateral surface of langerin ( Figure 6). In contrast, DC-SIGN binds to dengue virus on its upper surface (Pokidysheva et al., 2006). Considering that DC-SIGN serves as an attachment receptor for various viruses but not as an entry receptor (Bermejo-Jambrina et al., 2018;Lempp et al., 2021), the possible structural coupling of lateral ligand binding and internalization ( Figure 6B) implies that langerin functions as a more efficient entry receptor for viruses than DC-SIGN or other C-type lectins. Further studies are needed to investigate the physiological significance of the unique carbohydrate-binding mechanism of langerin and the honeycomb lattice structure of Birbeck granules.
The washed cells were homogenized with a Dounce glass homogenizer and were further disrupted by sonication for 8 seconds using a Q125 sonicator (Qsonica, Newtown, CT). The cell homogenate was loaded on top of a discontinuous Opti-prep density gradient (10-20-40% layers, ThermoFisher Scientific) and ultracentrifuged for 1 hour at 100,000×g. The fraction at the interface of 20/40% layers was collected and diluted 10-fold with HMC buffer and was again disrupted by sonication for 2 min (120 cycles of 1-sec pulse and 3-sec wait) at 4℃. Large debris was removed by filtration through a 20 µm-pore polyethylene mesh (GL Science, Tokyo, Japan) and the cell lysate was incubated with a 30 µl slurry of streptavidin agarose (Solulink, San Diego, CA) overnight at 4℃. The streptavidin agarose beads were washed five times with HMC buffer and the bound Birbeck granules were eluted by overnight digestion using home-made E. coli overexpressed His×6-HRV3C protease in HMC buffer at 4℃. After digestion and elution, HRV3C protease was removed by incubating the elute with Ni-NTA resin (Qiagen, Düsseldorf, Germany). About 50 µl of elute containing released Birbeck granules (approximately 0.05 mg/ml) were obtained from twelve 100-mm dishes of cell culture.

Cryo-electron microscopy of isolated Birbeck granules
5 µl of the purified Birbeck granules was applied onto freshly glow-discharged holey carbon grids, Quantifoil R1.2/1.3 Cu/Rh 200 mesh (Quantifoil Micro Tools GmbH, Großlöbichau, Germany), blotted from the backside to concentrate the granules by filtration through the holes, and 4 µl of HMC buffer containing cytochrome c-stabilized 15-nm colloidal gold (BBI Solutions) was applied for attachment of fiducial markers (Oda et al., 2015). Grids were blotted from the backside for 10 seconds at 4℃ under 99% humidity and plunge frozen in liquid ethane using Vitrobot Mark IV (Thermo Fisher Scientific). We removed one of the blotting arms for one-side blotting because Birbeck granules would be absorbed onto the filter paper. Images were recorded using a Titan Krios G3i microscope at the University of Tokyo (Thermo Fisher Scientific) at 300 keV, a Gatan Quantum-LS Energy Filter (Gatan, Pleasanton, CA) with a slit width of 20 eV, and a Gatan K3 Summit direct electron detector in the electron counting mode. The nominal magnification was set to 33,000× with a physical pixel size of 2.67 Å/pixel. Movies were acquired using the SerialEM software (Mastronarde, 2005), and the target defocus was set to 3-6 µm. The angular range of the tilt series was from -60° to 60° with 3.0° increments. Each movie was recorded for 0.74 sec with a total dose of 1.24 electrons/Å 2 and subdivided into 20 frames. The total dose for one tilt series acquisition is thus 50 electrons/Å 2 .

Data processing
Movies were subjected to beam-induced motion correction using MotionCor2 (Zheng et al., 2017), and tilt series images were aligned, CTF corrected, and back-projected to reconstruct 3D tomograms using the IMOD software package (Kremer et al., 1996). For the subsequent analysis using Relion-4.0 (Kimanius et al., 2021), the tomogram positioning step was skipped to keep the angle offset and the Z-shift zero. Tomograms were 4×binned to reduce loads of the calculation. The areas of the Birbeck granules were manually selected using the 3dmod model tool (Kremer et al., 1996), and the center coordinates for the subtomograms were determined by placing points at 8-pixel (~8.5 nm) intervals within the selected area using a custom-made Ruby script (Metlagel et al., 2007). For twisting Birbeck granules ( Figure S1D), center points were picked manually by selecting the mid-portion of the lamella. Volumes with 32×32×32 pixel-dimensions were extracted from 8×binned tomograms and were averaged using the PEET software suite (Nicastro et al., 2006). A randomly selected subtomogram was used for the initial reference. Alignments were repeated five times for 8×binned and twice for 4×binned tomograms with 64×64×64 pixel-dimensions. The coordinates and the rotation angles of the aligned subtomograms were exported to Relion-4.0. Initial pseudosubtomograms were generated with the box size of 128 pixels, the cropped box size of 64 pixels, and the binning factor of 4. The subtomograms were 3D refined using a spherical mask of 320-Å diameter. After 3D classification, 2×binned pseudosubtomograms were generated with the box size of 256 pixels and the cropped box size of 128 pixels based on the refined coordinates and angles and were 3D refined using the same spherical mask. Finally, un-binned pseudosubtomograms were generated with a box size of 256 pixels and a cropped box size of 128 pixels. 3D refinement was done using a tighter mask enclosing the central upright langerin molecule and three bound CRDs, and low-resolution subtomograms were removed by 3D classification. The alignment was further refined using CTF refinement and frame alignment. We initially refined the surrounding inverted langerin molecule by 3D refinement masking out the central upright molecule, but the refinement was biased toward the center. To improve the map quality of the inverted molecule, we applied 3D multi-body refinement (Nakane et al., 2018), which successfully compensated for the flexibility between the upright and the inverted molecules. For model building, the crystal structure of langerin trimer (Feinberg et al., 2010) (PDB ID: 3KQG) was fitted into the map using UCSF Chimera (Pettersen et al., 2004). Then, the fitted model was further refined using PHENIX real-space refinement tool, and the refined model was validated using the comprehensive validation tool (Adams et al., 2010). The Coulombic surfaces rendering was conducted using UCSF Chimera and other model presentations, surface rendering, and morph movie generation were conducted using ChimeraX (Goddard et al., 2018). The data processing and the statistics of the model validation are summarized in Figure S2A and Table S1.

Ultrathin section electron microscopy of Birbeck granules
Pellets of 293T cells with mannan-induced Birbeck granules were fixed with 4% paraformaldehyde for 1 hour at 4℃ and stained with 1% osmium tetroxide and subsequently with 1% uranium acetate. After dehydration in ethanol and acetone, the samples were embedded in Quetol 812 resin (Nissin EM, Tokyo, Japan). 60-nm ultrathin sections were cut using a ULTRACUT microtome (Reichert Leica) and mounted onto Formvar-coated copper grid. Images were recorded using a JEM-2100F microscope (JEOL, Tokyo, Japan) at University of Yamanashi operated at 200 keV equipped with an F216 CMOS camera (TVIPS GmbH, Gauting, Germany). Length of Birbeck granules were measured using Fiji software (Schindelin et al., 2012), and plots were generated using PlotsOfData web tool (Postma and Goedhart, 2019). For total length of Birbeck granules per 100 µm 2 , cross sections of cells with nuclei of > 5 µm in diameter were selected, and the sum of the lengths of Birbeck granules within the cells were measured. The areas of the cross sections were also measured using Fiji software. 100 µm 2 corresponds to the approximate mean of the cross-sectional area of one cell.
The obtained supernatants were immediately used for subsequent assays.

Data availability
The map and the model will be available on the EMDB upon publication under the following accession numbers: EMD-32906, and PDB 7WZ8. Although 2D classification did not converge well due to the continuity of the structure, some classes showed a porous structure with a honeycomb-like lattice.     The inset on the top shows an overview of antigen processing within a Langerhans cell.
Viruses are captured by langerin, internalized into a Birbeck granule, and degraded by the autophagic system. The red square indicates the viral entry process described in the lower panels. The high-mannose oligosaccharide on HIV-1 env first binds to the primary carbohydrate-binding site, and one of the branches of the oligo-mannose weakly interacts with the secondary binding site. The flexible neck region allows docking of the loop 258-262 of an adjacent trimer into the secondary carbohydrate-binding cleft, reinforcing the binding of the oligo-mannose. As the elastic neck regions straighten, the plasma membrane is deformed into a lamellar invagination, which internalizes the virus.