Nanoscale clustering of mycobacterial ligands and DC-SIGN host receptors are key determinants for pathogen recognition

The bacterial pathogen Mycobacterium tuberculosis binds to the C-type lectin DC-SIGN (dendritic cell–specific intercellular adhesion molecule 3-grabbing nonintegrin) on dendritic cells to evade the immune system. While DC-SIGN glycoconjugate ligands are ubiquitous among mycobacterial species, the receptor selectively binds pathogenic species from the M. tuberculosis complex (MTBC). Here, we unravel the molecular mechanism behind this intriguing selective recognition by means of a multidisciplinary approach combining single-molecule atomic force microscopy with Förster resonance energy transfer and bioassays. Molecular recognition imaging of mycobacteria demonstrates that the distribution of DC-SIGN ligands markedly differs between Mycobacterium bovis Bacille Calmette-Guérin (BCG) (model MTBC species) and Mycobacterium smegmatis (non-MTBC species), the ligands being concentrated into dense nanodomains on M. bovis BCG. Upon bacteria-host cell adhesion, ligand nanodomains induce the recruitment and clustering of DC-SIGN. Our study highlights the key role of clustering of both ligands on MTBC species and DC-SIGN host receptors in pathogen recognition, a mechanism that might be widespread in host-pathogen interactions.

Consequently, a critical, yet currently unsolved, issue is that while DC-SIGN ligands are redundant on mycobacterial cell surfaces, irrespective of pathogenic and nonpathogenic species, the receptor selectively binds pathogenic MTBC species. As the difference in mycobacterial species recognition by DC-SIGN does not primarily rely on the type of glycoconjugates they produce, other unknown mechanisms must be at play. Here, we address this using singlemolecule atomic force microscopy (AFM) (38), combined with Förster resonance energy transfer (FRET) and bioassays. We find that DC-SIGN ligands are concentrated into densely arranged nanodomains on the surface of Mycobacterium bovis Bacille Calmette-Guérin (BCG) (used as a model MTBC species), while they are essentially randomly distributed on Mycobacterium smegmatis (a non-MTBC species). This is accompanied by the presence of large membrane-expressed DC-SIGN clusters upon adhesion of bacteria to host cells and by adhesion-induced recruitment of DC-SIGN. Our findings demonstrate that the clustering of mycobacterial ligands and the clustering of host DC-SIGN are key determinants for pathogen recognition, therefore rationalizing, at the molecular level, the highly selective recognition of MTBC by DC-SIGN. This mechanism might be widespread among pathogen-immune cell interactions involving DC-SIGN but possibly also other PRRs, with consequences for the modulation of the immune response during infection.

DC-SIGN ligands are surface-localized and available for binding on MTBC and non-MTBC members
We first tested the extent to which the MTBC species model M. bovis BCG and the non-MTBC model nonpathogenic species M. smegmatis are recognized by DC-SIGN expressed at the membrane (mDC-SIGN) of a human embryonic kidney cell line (HEK DC-SIGN ). Using flow cytometry analysis, we found that M. bovis BCG bound to HEK DC-SIGN cells in a multiplicity-of-infection (MOI)dependent fashion, while it did not bind to wild-type HEK cells (HEK WT ) ( Fig. 1A and fig. S1). As expected, M. bovis BCG binding to HEK DC-SIGN cells was inhibited by mannan [a high-affinity DC-SIGN ligand (24,39)] and by the divalent cation chelator EDTA. In contrast, M. smegmatis did not bind the HEK DC-SIGN cells appreciably beyond background control levels (Fig. 1A). These results are in line with previous work in which an mDC-SIGN-expressing HELA cell line bound MTBC species selectively among phylogenetically diverse mycobacteria (21).
Intriguingly, mDC-SIGN binds purified forms of LM, PIM 6 , mannoproteins, and α-glucan (21,23,35,37), all present in M. smegmatis, while it does not recognize M. smegmatis. A possible explanation could be that these potential ligands are located within deeper layers of the envelope and masked from interactions with DC-SIGN. To test this hypothesis, we first produced a recombinant soluble form of the extracellular domain of DC-SIGN [sDC-SIGN; (40)], which presented as a tetramer and was functional in binding known ligands, including ManLAM, LM, PIM 2 , and PIM 6 ( fig. S2). Next, we developed a binding assay using sDC-SIGN and microtitre plates coated with various bacterial species, including non-MTBC species (Mycobacterium chelonae, Mycobacterium kansasii, Mycobacterium avium, and M. smegmatis), the MTBC model M. bovis BCG, and Escherichia coli DH5α as a control bacterium (Fig. 1B) (41,42). Although the highest levels of sDC-SIGN binding occurred for M. bovis BCG, all mycobacterial species substantially bound the extracellular domain of the receptor, strongly supporting the notion that DC-SIGN ligands are surface-exposed on mycobacterial species both within and outside of the MTBC. Therefore, there must be a currently unknown mechanism explaining why mDC-SIGN does not bind non-MTBC species despite the presence of ligands on their surface.

Mechanical strength of single sDC-SIGN-ligand complexes on living mycobacteria
We wondered whether the binding strength between DC-SIGN receptors and their glycoconjugate ligands might differ between MTBC and non-MTBC species. To test this hypothesis, we used AFM, a multifunctional nanotechnique allowing the probing of the surface ultrastructure and molecular interactions of living bacterial cells, in a way that is inaccessible to conventional microscopy Biotinylated mycobacteria were incubated with HEK cells that do not (HEK WT ) or that do express DC-SIGN (HEK DC-SIGN ). Bound bacteria were labeled with allophycocyanin (APC)-conjugated streptavidin and detected by flow cytometry. The data shown are representative of two independent experiments. The bars show the mean fluorescence intensity. Specificity of binding was confirmed through blocking with mannan (3 mg ml −1 ) or EDTA (2 mM). (B) Microtiter plate-bound assay to test the binding of a recombinant soluble form of the extracellular domain of DC-SIGN (sDC-SIGN) to diverse mycobacterial species. Bars indicate means, and error bars show the SE. The data shown are representative of four independent experiments. Specificity of binding was tested by adding 100 mM mannose or 5 mM EDTA. E. coli DH5α produces a lipopolysaccharide unlikely to bind DC-SIGN (41,42). MOI, multiplicity of infection; OD, optical density; HRP, horseradish peroxidase. and biochemical assays (43)(44)(45). Figure 2A shows topographic images of whole M. bovis BCG cells (left) and the high-resolution surface of one such bacterium (right). It evidences a cell surface roughness of 0.6 ± 0.1 nm (means ± SD from seven different cells), indicating that the ultrastructure on the bacterial surface is very flat, as typically observed for mycobacteria (46)(47)(48)(49), a factor that enhances the area available for adhesion (50).
To quantify the strength of single DC-SIGN-ligand complexes, we used single-molecule force spectroscopy (SMFS), in which AFM tips functionalized with sDC-SIGN, carrying an N-terminal immunoglobulin G (IgG)-Fc fusion for optimal orientation, were used to generate multiple force-distance (FD) curves across living mycobacteria. Analyzing FD curves (Fig. 2B, right) yielded binding frequencies (i.e., the percentage of curves with binding events among the total number of curves recorded per cell), rupture forces (bond strengths), and rupture lengths (extension at which the complex ruptures) (Fig. 2, C and D). sDC-SIGN binds ligands on M. bovis BCG with a frequency of 19 ± 5% (means ± SD, from a total of 11,776 curves from n = 13 cells). The rupture forces adopted a multimodal distribution with a first and dominant peak representing the rupture of single bonds, followed by minor peaks centered at multiples of the first peak ( Fig. 2C and fig. S3). From a total of 13 different cells, we obtained an average rupture force for single complexes of 27 ± 3 pN (Fig. 2D). This rupture force is in the range of those reported for other lectin-carbohydrate interactions (51)(52)(53)(54)(55). The force profiles were well fitted by the worm-like chain model of biopolymer extension (Fig. 2B), in agreement with the stretching of receptor-ligand complexes. Our tip functionalization protocol ensures a minimal number of PRRs at the tip apex (56,57), which favors formation of single molecular complexes. Yet, parallel bond formation was observed with larger rupture forces ( fig. S3), most likely arising from the tetrameric state ( fig. S4) of the extracellular domain of DC-SIGN (25,58,59) that allows multivalent interactions with up to four carbohydrate moieties. Single complexes extended over an average length of 26 ± 5 nm (from a total of 11,776 curves from n = 13 cells; Fig. 2, C and D).
To prove the specificity of the interactions, we first used bare silicon nitride tips. These bound with a negligible frequency (~3%) to the mycobacterial surfaces ( fig. S5). In addition, mannose-blocking reduced binding threefold (6 ± 1%, from a total of 3072 curves from three cells) but had virtually no effect on rupture forces and rupture lengths (Fig. 2D). Last, injecting a polyclonal antibody raised against a sequence within the C-terminal of DC-SIGN where its C-type lectin domain is located also caused a substantial decrease in binding frequency (11 ± 3%, from a total of 5120 curves from n = 5 cells).
We then asked whether M. bovis BCG ligands interact in a similar way with soluble forms of the extracellular domains of two related C-type lectin PRRs, Dectin-2 (DC-associated C-type lectin-2; sDectin-2) and Mincle (macrophage inducible Ca 2+ -dependent lectin; sMincle). The structures and mycobacterial-glycoconjugate binding specificities of these PRRs diverge from DC-SIGN; Mincle recognizes all mycobacterial species via a plethora of lipid ligands (60,61), whereas Dectin-2, having ManLAM as the sole ligand, recognizes slow-growing mycobacteria only (62,63). Single molecular complexes for both receptors ruptured under similar forces (31 ± 5 pN, n = 11,264 total curves from 11 cells for sDectin-2 and 29 ± 6 pN, n = 6,144 total curves from 6 cells for sMincle) to those formed by sDC-SIGN ( Fig. 2, C (middle and bottom) and D). Bimodal distributions were observed for these receptors ( Fig. 2C and fig. S6), which may be accounted for by their monomeric and dimeric states (64,65). Contrasting with sDC-SIGN, complexes with these PRRs ruptured at shorter extension lengths (18 ± 3 nm for sDectin-2 and 19 ± 4 nm for sMincle), in agreement with the smaller sizes of their extracellular domains (see Materials and Methods). Notably, binding frequencies for both receptors were significantly lower than for sDC-SIGN at 11 ± 4% for sDectin-2 and only 8 ± 3% for sMincle, implying that DC-SIGN is the C-type lectin PRR receptor that most readily interacts with pathogen-associated molecular patterns on the M. bovis BCG surface.
Notably, sDC-SIGN bindsligands on the M. smegmatis surface with a frequency of 8 ± 3%, which is about twofold lower than in M. bovis BCG (n = 10 cells; Fig. 2D), and there were no differences in rupture forces and lengths (Fig. 2D). In addition, as expected, Dectin-2 appeared to interact the weakest with the molecules exposed on the surface of M. smegmatis.

sDC-SIGN binds M. bovis BCG and M. smegmatis ligands with similar kinetics
Next, we aimed to understand why M. smegmatis ligands are bound with a lower frequency by sDC-SIGN compared to M. bovis BCG. One explanation may be that sDC-SIGN binds ligands on M. smegmatis with a lower overall affinity. DC-SIGN affinity might differ for the different individual ligands, and their relative abundances are not clearly defined between M. bovis BCG and M. smegmatis, which, in addition, does not produce ManLAM (66). Using SMFS, binding kinetics parameters can be estimated (67, 68) for ligands exposed in their native form on living bacteria. A pseudofirst-order kinetics analysis of the relationship between binding frequency and probe-bacterial surface contact time (67) (25). These results show that M. bovis BCG and M. smegmatis exhibit similar DC-SIGN-ligand binding kinetics. This led us to hypothesize that the main factor defining differences in the overall mDC-SIGN binding properties is the spatial distribution of ligands across the bacterial surface, rather than the binding strength or kinetics.

Ligand clusters cover most of the M. bovis BCG surface, while ligands are sparsely distributed on M. smegmatis
To assess ligand surface distribution, we generated molecular recognition maps with sDC-SIGN-modified tips, wherein white and black pixels (16 nm-by-16 nm in size) indicate the presence or absence of a ligand (Fig. 4A). A dense clustered distribution of ligands was observed on M. bovis BCG, while they were essentially scattered on M. smegmatis (Fig. 4A). For a quantitative analysis, we defined a ligand cluster as any contiguous area containing at least two white pixels that are not separated by more than one black pixel, implying a maximal distance between two ligands of~45 nm, which is roughly equivalent to the upper range nearest neighbor spacing of DC-SIGN molecules in lipid rafts on immature DCs (26,27). While M. smegmatis showed ligand clusters always smaller than 0.02 μm 2 (n = 10 cells), M. bovis BCG featured ligand clusters that were much larger, ranging from 0.03 to 0.12 μm 2 (interquartile range, n = 12 cells; Fig. 4B). Because a cluster area of 0.02 μm 2 contains at least 40 ligands (based on a cluster density of~2000 ligands per μm 2 ), our results suggest that there may be an optimal threshold for efficient binding to mycobacterial ligands by cell membrane DC-SIGN.
Knowing that high surface density of DC-SIGN is required for efficient binding of particles such as viruses, whereas it is not for efficient binding of soluble ligands (26,71), we wondered whether a high density of receptors could be important for selective recognition of MTBC species. To test this hypothesis, we developed a binding assay using His-tagged sDC-SIGN coated onto Ni-chelate microplates at saturated levels (fig. S7) and incubated with M. bovis BCG or M. smegmatis. After washing, M. bovis BCG cells remained attached to the sDC-SIGN surface, while M. smegmatis did not (Fig. 4C). M. bovis BCG binding was specific to DC-SIGN as it was inhibited by EDTA and mannose, but not by galactose. These results suggest that the DC-SIGN density plays a role in the selective binding of MTBC species. In addition, it is conceivable that the forces exerted on DC-SIGN-ligand complexes, where whole bacteria are in solution while DC-SIGN is immobilized (Fig. 4C), are much greater than when the bacteria are immobilized, and the receptor is in solution (Fig. 1B). It is possible that the greater shear forces at play in the former configuration (Fig. 4C), which likely approximates the in vivo context, also contribute to differential binding.

M. bovis BCG adhesion to mDC-SIGN-expressing host cells is higher and involves receptor clustering
Single-cell force spectroscopy (SCFS) (38), in which a single bacterial cell is bound to a colloidal AFM probe (Fig. 5A, left), allows the evaluation of force magnitudes and frequencies in bacterial-host cell adhesion. This method allowed us to measure the adhesion forces between single M. bovis BCG or M. smegmatis cells and HEK DC-SIGN or HEK WT (control) cells (Fig. 5A, top right). Like in SMFS, FD curves showed a nonlinear extension profile indicative of biomolecular extension (Fig. 5A, bottom right). Some curves exhibited serial unbinding events resulting from sequential unbinding of multiple bonds (72), with the largest peak being the maximum adhesion force, while the last peak yielded the rupture force of single molecular complexes (Fig. 5B, bottom right).
Average rupture forces of~100 pN were measured for both species (Fig 5B), considerably greater than the SMFS values. This originates from the much higher LR used due to the considerably stiffer (k ≈ 0.1 N m −1 ) colloidal probe cantilevers and to the faster retraction speed (20 μm s −1 ) required to limit HEK cell membrane deformation. M. bovis BCG bound to the HEK DC-SIGN cells with a frequency of 33 ± 18% (means ± SD, n = 9 bacterium-cell pairs), significantly greater than M. smegmatis (15 ± 9%, means ± SD, n = 7 bacterium-cell pairs; Fig. 5, B and C). In addition,~25% of M. bovis BCG cells exhibited maximum adhesion forces of~200 pN and greater, considerably above the population mean (117 ± 11 pN, n = 9 bacterium-cell pairs; Fig. 5B); this was not observed for M. smegmatis. Taking the adhesion force as a rough indicator of valency, on average, only one or two cellular receptors bound the M. bovis BCG probes. We assign this to the rapid probe velocity we used here, resulting in short contact times (~25 ms), which would limit the number of bonds that can form. In control experiments, we found that (i) injection of free mannose strongly reduced binding in HEK DC-SIGN cells and (ii) HEK WT cells poorly bound M. bovis BCG (Fig. 5B), indicating that mDC-SIGN expressed on HEK DC-SIGN cells is surface exposed and fully functional and that it represents the main receptor for mycobacteria. Notably, molecular recognition maps obtained with M. bovis BCG probes revealed stark segregation of DC-SIGN on the host cell surfaces, with clear DC-SIGN clusters surrounded by zones practically devoid of the receptor (Fig. 5C). This phenomenon was much less apparent with M. smegmatis probes. These results show that a clustered distribution of DC-SIGN must play a role in the selective recognition of MTBC species.

Large ligand clusters on M. bovis BCG correlates with mDC-SIGN recruitment
Last, we wondered whether ligand clustering on MTBC mycobacteria might induce local recruitment of DC-SIGN. To test this, we used FRET, which relies on the nonradiative energy transfer between a donor and acceptor fluorophore and presents extreme sensitivity to monitor minute changes in distances between the two molecules when in proximity, <~10 nm (Fig. 6, A and B). We made use of our HEK cell line coexpressing enhanced green fluorescent protein (EGFP)-and DsRed-fused DC-SIGN (HEK EGFP-DC-SIGN/DsRed-DC-SIGN ) and confocal microspectrofluorimetry (73), which relies on FRET measurements taken locally at the plasma membrane (~1-μm 2 zone) of a single cell. Cells were scored as FRET + according to established criteria ( fig. S8) (73). Incubation of the HEK EGFP-DC-SIGN/DsRed-DC-SIGN cells with M. bovis BCG rendered the cells FRET positive in an MOI-dependent fashion, with 54% of the cells scoring as FRET + at MOI 20 (Fig. 6C). This effect was blocked by the addition of mannan or EDTA. In sharp contrast, only 4% of HEK EGFP-DC-SIGN/DsRed-DC-SIGN cells exposed to M. smegmatis at MOI 20 showed a FRET shift. These results lead us to conclude that M. bovis BCG adhesion to HEK DC-SIGN cells induces the local recruitment of mDC-SIGN, most likely via ligand clustering on the bacterial cell surface, therefore explaining selective and efficient attachment of MTBC species to DC-SIGN.

DISCUSSION
Hijacking DC-SIGN expressed at the surface of antigen-presenting DCs represents an escape mechanism for several important pathogens (74). Therefore, a thorough understanding of the molecular bases of efficient ligand binding leading to bacterial recognition is critical. Here, we show that, beyond ligand specificity, selective and efficient adhesion of pathogenic mycobacteria to host cell membrane DC-SIGN relies on the nanoscale clustering of glycoconjugate ligands on the bacterial cell surface and on adhesion-induced recruitment of the receptor (summarized in Fig. 7).
Binding strength and kinetics of single DC-SIGN-ligand complexes is similar for both pathogenic M. bovis BCG (an MTBC species) and nonpathogenic M. smegmatis (a non-MTBC species). However, the distribution of surface-exposed ligands strongly differs between the two species, with dense ligand nanoclusters being observed only on M. bovis BCG (Fig. 7). As Mtb has coevolved with its human host to evade the immune system (75,76), it is tempting to speculate that the organization of DC-SIGN ligands into dense clusters has provided an evolutionary advantage for MTBC pathogens.
Until recently, nanoscale heterogeneities on living microbial cells were not accessible to study. However, the emergence of livecell nanoscopy (45) has revolutionized the way microbiologists explore the constituents and machineries of living bacteria to molecular resolution. Whereas super-resolution fluorescence nanoscopy enables to study the dynamics of biomolecules and particles inside cells, AFM is capable of imaging and force probing single-cell surface components. In the past years, there have been exciting discoveries demonstrating that many specific molecules such as receptors or ligands on pathogen surfaces are not distributed homogenously but form nanodomains [reviewed in (77,78)]. Early single-molecule AFM studies with mycobacteria found that the M. tuberculosis heparin-binding hemagglutinin adhesin is segregated within nanodomains on the bacterial cell surface (79), while its sulphated proteoglycan ligands were distributed homogenously on pneumocytes (80). More recently, nanoclustering of various staphylococcal adhesins has been reported, suggesting that they use this phenomenon to favor enhanced, multivalent interactions with host extracellular matrix proteins such as fibrinogen (81)(82)(83)(84). Using hydrophobic tips, we found that a smooth variant of the non-MTBC pathogen Mycobacterium abscessus exhibits hydrophilic and hydrophobic nanodomains associated with glycopeptidolids (46,85). Surface compartmentalization of rough lipopolysaccharide classes has also been reported for Brucella abortus (86). Fungal pathogens also form cell surface nanodomains, as demonstrated for the cell-wall adhesion protein Als5 from Candida albicans (87). Pulling on single adhesins with AFM tips functionalized with specific antibodies induced the formation of Als5 domains of 100 to 500 nm, resulting from force-induced conformational changes in the protein, and the domains were shown to propagate over the entire cell surface.
An important unsolved question is how clusters are formed on bacterial surfaces. Nano-/micro-domains within bacterial membranes is an emerging, fast-moving field, and their origins, compositions, and functional roles in bacterial physiology and pathology are being unraveled (88)(89)(90). Yet, in many species including mycobacteria, more external envelope layers mask the inner and outer membranes, and, to our knowledge, practically nothing is known about nanodomains within these outermost surface layers where interactions with the host most likely occur. Unraveling how these domains are formed, and how they change over time and in response to chemical or mechanical stresses, constitutes an exciting untapped field of research.
Another key finding is that the clustering of DC-SIGN on the host cell membrane also contributes to efficient and selective binding of M. bovis BCG and that adhesion of the latter stimulates local recruitment of the receptor (Fig. 7). This recruitment may involve passive diffusion of DC-SIGN where receptor molecules binding a bacterium are retained, leading to an increase in their local density. However, a mechanism involving microtubules for the rapid, directed transport of DC-SIGN clusters was recently reported and was proposed to bring bound pathogens on the periphery or projections of DCs toward the perinuclear region for internalization (91). Future work may explore the mechanism underlying adhesion-dependent recruitment of DC-SIGN.
We speculate that clustering of pathogen ligands and of DC-SIGN host receptors might be a general mechanism for activating pathogen recognition and internalization by antigen-presenting DCs. Recruitment at the adhesion site was reported for zymosan particles derived from the fungal pathogen-surrogate, Saccharomyces cerevisiae (92), although it remains to be investigated whether binding of fungal pathogens, such as C. albicans, to DC-SIGN (93) involves ligand clustering or DC-SIGN recruitment. A detailed understanding of this phenomenon could open new avenues in therapeutics, e.g., immunomodulatory or anti-adhesive antimicrobial strategies.
Collectively, our work sheds light on the importance and complexity of surface distribution of both ligands and DC-SIGN in binding of pathogens by this receptor through high-avidity interactions. Recent progress in the development of DC-SIGN antagonists includes multivalent glycomimetic modulators showing great promise (40,(94)(95)(96)(97)(98)(99)(100)(101).  (21)] were cultured as surface pellicles at the appropriate temperature in Middlebrook 7H9 medium (Sigma-Aldrich) supplemented with 0.1% (w/v) glycerol, 0.1% (w/v) D-glucose, sodium chloride (0.425 g liter −1 ), catalase (2 mg liter −1 ), and bovine serum albumin fraction IV (BSA; 2.5 g liter −1 ). Mature surface pellicles were washed in 7H9 medium (not supplemented with BSA or catalase) and gently dispersed using 3-mm-diameter glass beads yielding suspensions containing single bacterial cells (as observed by microscopy and AFM). Freshly preparedsingle-bacterium suspensions werestained (or not) with the green fluorescent dye fluorescein isothiocyanate (FITC) following a described (102) protocol with the exception that buffer solutions contained no detergent. These suspensions were either used immediately for binding experiments or snap frozen in liquid nitrogen and stored at −80°C until their use. For SMFS experiments, an aliquot was thawed, appropriately diluted in nonsupplemented 7H9 medium and seeded in a hydrophobic nontreated polystyrene culture dish (35 mm). After 30-min incubation at 37°C, nonadherent bacteria were washed away, and fresh medium was added before incubation at 37°C for an additional 30 min (this ensured strong immobilization of the bacterial cells). The bacteria were then washed with phosphate-buffered saline (PBS) and the petri dish filled with PBS supplemented with 0.1% BSA, 1 mM CaCl 2 , and 1 mM MgSO 4 . For SCFS experiments, an aliquot was thawed, diluted appropriately in PBS, and seeded on a treated (hydrophilic) polystyrene dish. Bacteria were allowed to settle down and weakly adhere to the surface before they were caught with a hydrophobic beaded AFM cantilever (see SCFS section below).

Production of recombinant C-type lectins
The coding sequence of the extracellular portion of the DC-SIGN protein (amino acids 64 to 404, including both the helical neck domain and the C-terminal C-type lectin domain) was cloned into the vector pET19b (Novagen). This plasmid was then transformed into E. coli BL21 (DE3)plysS, which served as expression strain (40). Recombinant protein expression was induced in exponential cultures (optical density at 600 nm =~0.6) with 1 mM isopropyl-β-D-thiogalactopyranoside (Fluka). After 4 hours of induction at 37°C, the bacteria were collected by centrifugation and the pellets were frozen. They were then thawed, resuspended in lysis buffer [100 mM NaCl, 50 mM tris-HCl, and 0.1% (v/v) Triton X-100 (pH 7.8)], and probe sonicated in a bath of melting ice (three cycles of 30 s). The protein, which was expressed insolubly in inclusion bodies, was collected by centrifugation (45,000g, 30 min, 4°C), and the pellet washed twice with washing buffer [1 M NaCl, 25 mM tris, and 1 M urea (pH 7.8)], with intermittent sonication steps. After the last wash, the pellet was resuspended in solubilization buffer [10 mM tris (pH 7.8), 100 mM NaH 2 PO 4 , 6 M guanidine, and 0.01% (v/v) β-mercaptoethanol], and the solution was sonicated (two cycles of 30 s) and ultracentrifuged (55,000g, 30 min, 4°C). The recovered supernatant was mixed with nickel-nitrilotriacetic acid (Ni-NTA) agarose with stirring overnight at 4°C before loading the suspension onto a column. After a wash step [with 50 ml of 30 mM tris-HCl buffer containing 1 M NaCl, 1 mM CaCl 2 , 6 M urea, and 15 mM imidazole (pH 7.8)], the protein was renatured on the Ni-NTA agarose column by successive passages with 30 ml of 30 mM tris-HCl buffer (pH 7.8) containing 1 M NaCl and decreasing concentrations of urea (from 5 to 1 M) at a flow rate of 15 ml hour −1 . The refolded, pure protein was eluted with 30 mM tris-HCl buffer (pH 7.8) containing 1 M NaCl, 1 mM CaCl 2 , and 1 M imidazole, and 10 to 15 1-ml fractions were recovered. The most concentrated fractions were pooled and dialyzed twice overnight at 4°C against 30 mM tris buffer (pH 7.8) containing 1 M NaCl and 1 mM CaCl 2 . The purity and correct tetrameric quaternary state of the produced protein were verified by denaturing and nondenaturing polyacrylamide gel electrophoresis ( fig. S2, A and B).
Recombinant C-type lectins used in SMFS experiments (DC-SIGN, Dectin-2, and Mincle) consisted of their C-terminal extracellular domains (amino acids 59 to 404 for DC-SIGN, amino acids 42 to 209 for Dectin-2, and amino acids 41 to 219 for Mincle) fused at their N termini to the C terminus of the human IgG1-Fc1 fragment. All these Fc-lectin fusions were produced in Chinese hamster ovary cells and obtained in purified form (in PBS) from Invivogen (available on request). The predicted maximum extended length of the polyethylene glycol (PEG)-linked IgG-Fc-tagged sDC-SIGN is 50 nm [~12-nm linker,~6-nm IgG-Fc (103);~32-nm sDC-SIGN (104)], while for both sMincle and sDectin-2, it is~20 nm [with soluble extracellular domains of~4 nm; (105,106)]. However, the linker is expected to attach to lysine residues randomly. In addition, molecules may be attached off-center at the tip with an apex diameter of~20 nm that may give rise to rupture lengths that underestimate the maximum molecular complex extended length.

Microplate binding assays using wells coated with sDC-SIGN
In binding assays using plate-bound sDC-SIGN, the protein (2 μg ml −1 in TBS-2 mM CaCl 2 -His-1% BSA) was allowed to react for 2 hours at RT with Ni-chelate microplates (Nunc). Wells were then blocked and rinsed as indicated above. Dissociated bacterial preparations (0.5 mg, wet weight) or M. bovis BCG ManLAM (107) (at the indicated amount) in TBS-2 mM CaCl 2 -1% BSA were preincubated or not with 50 or 100 mM mannose or galactose or 1 mM EDTA and allowed the reaction with plate-bound sDC-SIGN for 2 hours at RT. After washing, bound mycobacteria or ManLAM was labeled with HRP-conjugated concanavalin A (Sigma-Aldrich). HRP was detected as above.

Construction of HEK-DC-SIGN cell lines
Briefly, a cDNA (gift from O. Neyrolles, IPBS, Toulouse) encoding the gene of DC-SIGN/CD209 was amplified by the polymerase chain reaction (PCR) using primers 5′-GATATCTACGTAAGT GACTCCAAGGAACCAAGACTGC-3′ and 5′-GATATCTCTA GACTACGCAGGAGGGGGGTTTGGGGTG-3′. PCR product lacking the initiation codon was then double-digested with Sna BI and Xba I (underlined sequences in primers). The generated segment was inserted into a modified SK+ bluescript plasmid, containing Sna BI-Xba I restriction sites, downstream of an epitope tag from the bacteriophage T7 fused with either EGFP cDNA or DsRed monomer cDNA (Clontech). Fragments named T7-EGFP-DC-SIGN and T7-DsRedmono-DC-SIGN were then cloned into Ecor V-Xba I-digested PcDNA3.1/Hygro or PcDNA3.1/neo vectors (Invitrogen), respectively. The constructs were verified by restriction enzyme analysis and Sanger sequencing.
HEK 293 cells were seeded in Dulbecco's modified Eagle's medium (DMEM) plus 10% fetal calf serum in six-well culture plates at a density of 5 × 10 5

Binding assay using HEK-DC-SIGN cell lines
Bacteria were first labeled with biotin hydrazide after periodate oxidation (21). Around 1 g (wet weight) of bacteria was washed twice with PBS and resuspended in 200 μl of 0.1 M ammonium acetate buffer (pH 5.5) containing 15 mM sodium metaperiodate (Merck). After a 20-min incubation at 4°C in the dark with gentle rotation, the oxidation reaction was quenched by adding 200 μl of 0.1 M ammonium acetate buffer (pH 5.5) containing 30 mM sodium bisulphite (Sigma-Aldrich). After centrifugation, bacteria were resuspended in 400 μl of PBS containing 5 mM biotin hydrazide (Sigma-Aldrich). After a 2-hour incubation at RT with gentle rotation, cells were washed three times with PBS. Biotinlabeled bacteria were resuspended in PBS and added to 2 × 10 5 HEK cells at the indicated MOI in a total volume of 1 ml of cold DMEM medium. After a 4-hour incubation at 4°C under gentle rotation, the cell suspension was centrifuged at 300g for 10 min. The pellet was suspended in 50 μl of allophycocyanin-conjugated streptavidin (BD Pharmingen). After 20 min at 4°C in the dark, cells were washed twice with PBS, resuspended in 400 μl of PBS, and analyzed by confocal microscopy ( fig. S9) and flow cytometry using a FACS-Calibur CellQuest Pro (BD Biosciences) software.

AFM tip functionalization for SMFS
For SMFS experiments, we functionalized AFM tips with flexible PEG linkers terminated with a reactive aldehyde (Ald) function allowing the covalent coupling of an IgG-Fc-fused lectin (56,57). The net negatively charged extracellular domains of the lectins (sDC-SIGN: isoelectric point (pI) 5.2, sDectin-2: pI 5.7, and sMincle: pI 5.1) in PBS (pH 7.4) are repelled by the net negatively charged Ald-PEG-coated tip surface, while IgG1-Fc1 (pI 7.6) is not. Therefore, the IgG1-Fc1 fusion configuration favored the covalent coupling of a primary amine group within N-terminal IgG1-Fc1 and, hence, optimal geometry of the lectin extracellular domain for interactions with probed ligands. D-cantilevers (nominal k = 10 pN nm −1 ) of Bruker MSCT silicon nitride AFM probes were used. The AFM probes were functionalized at RT with Fc-lectins using Ald-Ph-PEG 24 -NHS ester (BroadPharm) bifunctional linkers. Briefly, the bare silicon nitride AFM probes were washed with chloroform, dried under nitrogen flow, ultraviolet (UV) ozone-treated, amino-functionalized using the 3-aminopropylyriethoxysilane and triethylamine in the gas-phase method (57), and lastly placed in a chloroformic solution of Ald-Ph-PEG 24 -NHS ester (6.6 mg ml −1 ) and triethylamine [6% (v/v)]. After 2 hours, they were washed thoroughly with chloroform, dried under nitrogen flow, and immersed in 50 μl of an Fc-lectin solution [50 μg ml −1 in PBS (pH 7.4)] to which 1 μl of sodium cyanoborohydride (1 M stock solution) was added immediately. After 1 hour of incubation, unreacted free Ald groups were quenched through the addition of 2.5 μl of ethanolamine hydrochloride (1 M stock solution, pH 8.0), and the fully functionalized AFM probes were thoroughly washed with PBS buffer. Fc-lectin AFM probes were stored at 4°C and used within 48 hours.
AFM imaging and single-molecule force spectroscopy AFM imaging was performed with bare MSCT-D cantilevers in PBS buffer using the Quantitative Imaging mode (approach and retract velocity of 40 μm s −1 , z length of 600 nm for whole bacteria, and 150 nm for high-resolution images). Average roughness values were obtained for second-order polynomial line-leveled, high-resolution images (256 pixels by 256 pixels, 300 nm by 300 nm) recorded on top of bacteria.
All SMFS experiments were carried out in PBS supplemented with 0.1% BSA, 1 mM CaCl 2 , and 1 mM MgSO 4 using a JPK Nano-Wizard 4 NanoScience AFM. Force spectroscopy data were collected in force mapping (force-volume) mode using a constant approach and retraction velocity of 1 μm s −1 , a ramp length of 250 nm, a contact force set point and pause of 250 pN and 250 ms, respectively, a closed z-loop, and fast and slow scan axes of 250 nm (16 pixels) and 1 μm (64 pixels), respectively. For dynamic force spectroscopy (dfs) and contact-time versus binding frequency experiments, respectively, retraction velocity and contact pause were varied as indicated in the relevant figures.

Single-cell force spectroscopy
Bacterial probes were prepared as previously described with modifications (108). Using the AFM, the first~5 μm of a triangular tipless cantilever (NP-O10, Bruker) was brought into contact with a thin layer of UV light-curable glue (NOA 63, Norland Edmund Optics). The glue-covered part of the cantilever was then brought into contact for 3 min with a silica bead of 6.1 μm diameter (Bangs Laboratories). Afterward, the colloidal probe was taken out of the AFM, and the glue was cured under UV light for 30 min. Colloidal probes were rendered hydrophobic by siliconization. They were placed in a glass dish inside a vacuum chamber along with a separate glass dish containing 100 μl of Sigmacote (Sigma-Aldrich) siliconization reagent, and a vacuum (200 mbar) was applied for 1 hour. Then, they were placed in a 45°C oven for 30 min before being stored in ultrapure water until use. Bacterial probes were made by bringing a hydrophobic colloidal probe in contact (applying 5 nN for 60 s) with a single FITC-stained bacterium (observed through a 40× objective of an inverted epifluorescence microscope) on a tissue culture-treated polymer coverslip bottom dish (iBidi) in PBS. The dish was then exchanged with a similar dish containing HEK cells and gently fixed as follows: Cocultured HEK DC-SIGN and HEK WT cells were first washed three times with PBS supplemented with 1 mM CaCl 2 and 1 mM MgSO 4 (PBS Ca/Mg ) to remove BSA. They were immediately fixed with 4% formaldehyde solution in PBS (Invitrogen) for 15 min at RT, washed (4×, 5 min) with PBS Ca/Mg , and stored at 4°to 8°C until use on the same day. SCFS was carried out in PBS Ca/Mg using a constant approach and retraction velocity of 20 μm s −1 , a ramp length of 1 μm, a contact force set point of 500 pN, no additional pause at contact, a closed z-loop, and a scan area of 3.2 μm by 3.2 μm (32 × 32 pixels) located on the apical portion of a cell. Because their soft membranes and dynamic movements complicated AFM SCFS measurements and because cross-linked sDC-SIGN retains its carbohydrate-binding activity (24,104), we gently formaldehyde-fixed the HEK cells (109).

Force spectroscopy data analysis
In FD curves, the last rupture peak that could be fitted with the worm-like chain model (110) of polymer extension was considered as representing the extension and rupture of a single lectin-ligand complex and was used to obtain rupture forces and lengths (see Fig. 2). FD curve analyses were performed using JPK data processing software. DC-SIGN interaction with ligands was approximated with pseudo-first-order kinetics (67), allowing the estimation of k on according to the formula k on ¼ ðτc eff Þ À 1 where τ is the interaction time and c eff is the effective concentration. τ was determined from a fit of binding frequency (B F ) versus contact time (T c ) data with the following function B F ðtÞ ¼ A 1 À e Tc À T 0 τ � � S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E c eff was calculated using the following formula where n b is the number of binding pairs (≈1), N A is the Avogadro constant, and r eff is equal to the radius (in decimeters) of a half sphere whose diameter is equal to the combined approximate equilibrium lengths of the PEG 24 linker (6 × 10 −8 dm), the N-ter IgG1-Fc1 fusion (~6.5 × 10 −8 dm), and the soluble extracellular domain of DC-SIGN (~32 × 10 −8 dm). For dfs data, the most probable rupture forces (F ) representing single molecular bonds for five log-equispaced LR bins were fit with the Bell-Evans model (70) where k B is the Boltzmann constant, T is the temperature (≈293 K), x β is the distance along the reaction coordinate to the transition between bound and unbound states, and k off is the off-rate constant.

FRET experiments
FRET measurements on single cells were performed using a spectrofluorimeter and analysis procedure described elsewhere (73). Briefly, the measuring system consisted of a microscope (Zeiss Axioplan) equipped with a 40× oil immersion objective (numerical aperture of 1.3) and pinholes to improve spatial resolution, an excitation laser line (Coherent Inova 90C), and a spectrograph for fluorescence recording (Horiba Jobin-Yvon Symphony). Fluorescence spectra for GFP/DsRed FRET experiments were recorded from 495 to 700 nm with 0.59-nm spectral resolution upon donor excitation at 476 nm. Recorded fluorescence spectra being different from one cell to the other, depending on levels of dye expression and the autofluorescence contribution, were converted into trichromatic coordinates according to the CIE 1931 international standard (111). These coordinates from recorded spectra allowed their classification for those cells that express only the donor or the acceptor and for those that express variable amounts of FRET.

Statistical analyses
All statistical analyses were conducted using the R programming language, and graphs were drawn in RStudio. Sample sizes and replicates are reported in figure legends. Experiments were repeated a minimum of two times.

Supplementary Materials
This PDF file includes: Figs. S1 to S9 Legend for data S1 Other Supplementary Material for this manuscript includes the following: Data S1 View/request a protocol for this paper from Bio-protocol.