Deciphering the complex three-way interaction between the non-integrin laminin receptor, galectin-3 and Neisseria meningitidis

The non-integrin laminin receptor (LAMR1/RPSA) and galectin-3 (Gal-3) are multi-functional host molecules with roles in diverse pathological processes, particularly of infectious or oncogenic origins. Using bimolecular fluorescence complementation and confocal imaging, we demonstrate that the two proteins homo- and heterodimerize, and that each isotype forms a distinct cell surface population. We present evidence that the 37 kDa form of LAMR1 (37LRP) is the precursor of the previously described 67 kDa laminin receptor (67LR), whereas the heterodimer represents an entity that is distinct from this molecule. Site-directed mutagenesis confirmed that the single cysteine (C173) of Gal-3 or lysine (K166) of LAMR1 are critical for heterodimerization. Recombinant Gal-3, expressed in normally Gal-3-deficient N2a cells, dimerized with endogenous LAMR1 and led to a significantly increased number of internalized bacteria (Neisseria meningitidis), confirming the role of Gal-3 in bacterial invasion. Contact-dependent cross-linking determined that, in common with LAMR1, Gal-3 binds the meningococcal secretin PilQ, in addition to the major pilin PilE. This study adds significant new mechanistic insights into the bacterial–host cell interaction by clarifying the nature, role and bacterial ligands of LAMR1 and Gal-3 isotypes during colonization.


Summary
The non-integrin laminin receptor (LAMR1/RPSA) and galectin-3 (Gal-3) are multi-functional host molecules with roles in diverse pathological processes, particularly of infectious or oncogenic origins. Using bimolecular fluorescence complementation and confocal imaging, we demonstrate that the two proteins homo-and heterodimerize, and that each isotype forms a distinct cell surface population. We present evidence that the 37 kDa form of LAMR1 (37LRP) is the precursor of the previously described 67 kDa laminin receptor (67LR), whereas the heterodimer represents an entity that is distinct from this molecule. Site-directed mutagenesis confirmed that the single cysteine (C 173 ) of Gal-3 or lysine (K 166 ) of LAMR1 are critical for heterodimerization. Recombinant Gal-3, expressed in normally Gal-3-deficient N2a cells, dimerized with endogenous LAMR1 and led to a significantly increased number of internalized bacteria (Neisseria meningitidis), confirming the role of Gal-3 in bacterial invasion. Contact-dependent cross-linking determined that, in common with LAMR1, Gal-3 binds the meningococcal secretin PilQ, in addition to the major pilin PilE. This study adds significant new mechanistic insights into the bacterial-host cell interaction by clarifying the nature, role and bacterial ligands of LAMR1 and Gal-3 isotypes during colonization. LAMR1 is encoded by the RPSA gene, which is present in 64 copies, although these are predominantly pseudo-genes; it is unclear how many functional copies there are [17], but one or more active genes encodes the 37 kDa 'precursor' protein (37LRP), which migrates at 37 -45 kDa on SDS -PAGE gels. A proportion of 37LRP migrates to the cell surface, where it is thought to 'mature', possibly via dimerization and/or posttranslational modification, into the 67 kDa high-affinity receptor (67LR), which migrates at 60-67 kDa on SDS -PAGE gels. The transition from 37LRP to 67LR has never been directly demonstrated, however. It is the 67LR isoform that is thought to act as the receptor for laminin and/or other substrates including elastin or collagen [2,[18][19][20][21], although the 37LRP isoform has also been suggested to have laminin-binding properties [18,22]. A 120 kDa form of the protein has also been described, but its nature, identity and relationship to 37LRP is unknown [23,24].
37LRP has also been suggested to associate with the b-galactoside-binding lectin galectin-3 (Gal-3), previously known as the 31 kDa human laminin-binding protein (HLBP31), to form a composite receptor for molecules including laminin [23]. Indeed, a heterodimer containing 37LRP and a protein carrying one or more galectin-3 epitopes has been proposed as the basis for 67LR [23]; this has given rise to confusion in the literature between the homo-and heterodimers. Evidence for heterodimerization has so far been circumstantial, based primarily on (i) the cross-reactivity of anti-Gal-3 antibodies with a 67 kDa protein [23], (ii) that Gal-3 and a 67 kDa protein can be co-eluted from laminin affinity columns by lactose, galactose and N-acetyl-lactosamine [25], and (iii) pre-treatment of laminin with b-galactosidase abolishes the interaction of the 67 kDa molecule and Gal-3 with laminin [25]. Nevertheless, 67LR is observed in Neuro 2a (N2a) cells, which do not express Gal-3, suggesting that 67LR and 37LRP-Gal-3 heterodimer exist as distinct, but potentially coexisting cell surface populations.
Together with galectin-1, Gal-3 is an important mediator of inflammation, and is involved in recruitment of neutrophils, bacterial recognition and activation of the phagocytic respiratory burst [26]. Gal-3 is the only chimera-type galectin, and occurs mainly as a monomer in solution [27]; it also self-associates on surfaces and, upon binding to divalent ligands, into homodimers or pentamers [28][29][30]. Gal-3 has one C-terminal 135-amino acid carbohydrate recognition domain (CRD) that is responsible for binding the b-galactoside moiety of mono-or oligosaccharides on several host molecules, including the b-galactosides and the poly-N-acetyl-lactosamine residues of host laminin [31], and structurally similar microbial molecules including neisserial lipooligosaccharide (LOS) [32,33]. The N-terminal domain (NTD) of Gal-3 is nonlectin-binding, but can bind proteins as well as recognizing the lipid A/inner core region of bacterial lipopolysaccharides (LPSs) [34]. It is also thought to be responsible for the protein's self-association, although Gal-3 self-association in solution via its CRD has also been described [35].
A fundamental understanding of the nature and functions of Gal-3 and the multiple isotypes of LAMR1 is central to our understanding of the diverse pathological processes in which these proteins play critical roles. LAMR1 and Gal-3 are both differentially expressed in various cancer cells, and between them play regulatory roles in a broad range of processes, including cancer cell growth, transformation, apoptosis, angiogenesis, adhesion, invasion and metastasis [2,36]. The direct relationship between these proteins remains controversial, and is often confusing. As a consequence, our current lack of understanding of the basic biochemistry, cell biology and ligand-specificity of the homo-and heterodimers of 37LRP and Gal-3 remains a hindrance to further progress towards deciphering these important pathological processes.
In this study, we show evidence for the self-and mutual association of 37LRP and Gal-3, show their distinctive surface distribution and demonstrate that they target an overlapping repertoire of surface ligands on N. meningitidis. We provide the first evidence that, together with LAMR1, Gal-3 expression enhances bacterial host -cell invasion. Our data clarify much of the confusion relating to LAMR1/Gal-3 monomeric and dimeric isotypes, and will have significant implications in the fields of infection and cancer biology.

Molecular modelling supports the proposed association of 37LRP and Gal-3
The proposed interaction of 37LRP and Gal-3 was investigated using the ZDOCK server (zdock.umassmed.edu) and their docking compared with the previously proposed interaction of two 37LRP molecules forming a homodimer [18]. ZDOCK has been reported to achieve high predictive accuracy on protein -protein docking benchmarks for rigid-body cases and consistent success in the international proteinprotein docking experiment [37]. The model revealed a large protein -protein interface between the 37LRP and Gal-3 subunits, involving several prominent 37LRP residues, mainly through salt bridges and hydrogen bonds (figure 1).
In the proposed complex between 37LRP and lactose-associated Gal-3 (figure 1a), four salt bridges are apparent. They are K 166 of 37LRP with D 178 and D 154 of Gal-3, R 80 of 37LRP with D 178 of Gal-3, D 151 of 37LRP with R 151 and K 227 of Gal-3, and D 44 of 37LRP with K 233 of Gal-3. Hydrogen bonds between N 81 of 37LRP and N 180 of Gal-3, N 29 of 37LRP and K 233 and E 230 of Gal-3, and Y 28 of 37LPR and E 230 of Gal-3 are also involved in the interaction (figure 1a). A similar interaction mode occurs between 37LRP and non-liganded Gal-3 (figure 1b). Here, K 166 in 37LRP also interacts with D 178 and D 154 of Gal-3, and S 233 of Gal-3 forms a hydrogen bond with D 44 in 37LRP. Importantly, binding of Gal-3 to the proposed heterodimer interface is incompatible with both homodimer formation (figure 1c) and laminin binding to at least one of the four proposed laminin-binding sites of 37LRP, because the relevant residues (A 114 , F 116 and Y 139 for homodimerization and F 32 , E 35 and R 155 for 37LRP-laminin binding) [38] are either sterically hindered or implicated in the heterodimer interaction. In short, our model supports a hypothesis of the mutual exclusivity of 37LRP homodimerization to form 67LR and 37LRP/Gal-3 heterodimerization for a given 37LRP molecule, while allowing for the possibility that both homo-and heterodimers could coexist as separate molecules. It should be noted, however, that the LAMR1 crystal structure is derived from a truncated protein and the proposed structure of the homodimer is highly speculative; it is possible that the proposed dimer interface may be an artefact of crystallization. rsob.royalsocietypublishing.org Open Biol. 4: 140053 3.2. 37LRP and Gal-3 co-localization on the cell surface To demonstrate cell surface co-localization of Gal-3 and 37LRP, two independent approaches were adopted. First, COS7 cells were co-transfected with constructs encoding recombinant 37LRP and Gal-3 proteins fused at their C-terminal ends with vYFP (YFP) or mCherry, respectively. Both proteins were successfully expressed and co-localized (figure 1d).
As the data presented in figure 1d could also be consistent with a cytosolic location of the protein, to demonstrate surface localization human cerebrovascular endothelial cells (hBMECs) were double-labelled for endogenous 37LRP, 67LR and/or Gal-3. Use of a wide range of antibodies with differing specificities for various isoforms of LAMR1 has resulted in considerable confusion in the literature. We selected antibodies with defined specificities as follows: rabbit IgG polyclonal (IHLR) raised in-house against 37LRPderived peptide aa263-282 [7] and the commercially available mouse IgG 2b monoclonal (A7) raised against the 37LRPderived peptide aa253-289 were employed to detect 37LRP. Both recognized an indistinguishable approximately 40 kDa protein in human hBMECs in immunoblotting experiments  and produced interchangeable staining patterns in immunostaining experiments (electronic supplementary material, figure S1; co-localization coefficient: 81.5%). 37LRP was predominantly cytoplasmic but was also present to a much lesser extent in the membrane fraction of hBMECs and in the soluble nuclear fraction (electronic supplementary material, figure S1). IHLR, but not A7, exhibited very low level cross-reactivity with an approximately 60 kDa band ( possibly 67LR) in the cytoskeleton-enriched cell fraction of hBMECs.
The widely used mouse IgM monoclonal MLuC5, which is known to block several LAMR1-mediated functions, exhibited a distinct immunostaining pattern from the two 37LRP-specific antibodies and predominantly recognized an approximately 60 kDa protein ( presumably 67LR) in hBMEC cellular fractions. This protein was predominantly associated with the cytoskeletal cell fraction, but was also present in small amounts in the membrane and nuclear fractions (electronic supplementary material, figure S1). The commercially available Gal-3-specific mouse monoclonal (MAb; Mac-2 clone M3/38; Biolegends) and goat polyclonal (PAb; AF1154; R&D Systems) antibodies were also employed. Staining of non-permeabilized hBMEC cells clearly demonstrated the co-localization of Gal-3 (using MAb) with 37LRP (figure 2a,b), but not with 67LR on the cell surface (figure 2c). Similar results were obtained using a second Gal-3 MAb (clone 9H3.2; Millipore). Cells stained with anti-Gal-3 PAb and 67LR-specific MLuC5 consistently demonstrated a significant co-localization of Gal-3-specific epitope(s) with 67LR (more than 90%), consistent with the existence of a Gal-3-specific epitope associated with the 67LR molecule [23] (figure 2d). The identity of this epitope is unknown, but it is not localized to galectin-3 as two independent galectin-3-specific monoclonal antibodies do not detect a protein in association with 67LR. Additionally, COS7 cells (which exhibit no surface-localized 67LR (electronic supplementary material, figure S2) demonstrated clear surface co-localization of 37LRP and Gal-3 (figure 2e), suggesting that 37LRP/Gal-3 heterodimers are distinct from 67LR, a finding which is consistent with the observation that 67LR can be detected in N2a cells, which do not express Gal-3 [9]. By determining the co-localization coefficient of 37LRP or 67LR with Gal-3 (using MAb M3/38) on hBMECs, we observed that Gal-3 associated with approximately half of the surfacelocalized 37LRP (50 + 5.9%); by contrast, only 9.4 + 2.1% of the surface-localized 67LR co-localized with Gal-3 (figure 2k).

Gal-3 and 37LRP form homo-and heterodimers
To test the hypothesis that 37LRP is capable of forming homodimers and also heterodimers with Gal-3, as suggested by our model, we employed a bimolecular fluorescence complementation (BiFC) technique. A fluorescing union is formed when two separate non-fluorescent YFP subfragments are brought together by the close association of two intimately interacting proteins [39]. In addition to the full-length YFP, its N-or C-terminal domains (Yn and Yc, respectively) were fused to the C-terminal end of Gal-3 and 37LRP (figure 3a), and transfected into N2a, hBMECs and COS7 cells (figure 3b-g shows N2a cells). Immunoblot analysis of host cells (electronic supplementary material, figure S3) confirmed transfection and protein expression. Although 37LRP was readily observed in COS7 cells by fluorescence microscopy (electronic supplementary material, figure S2) and immunoblotting (electronic supplementary material, figure S3), no 67LR was observed in these cells, suggesting that under the experimental conditions tested, 67LR did not form. As the tagged proteins do not form a higher molecular mass protein equivalent to 67LR, it is possible that additional factors may be required subsequent to homodimerization (as detected by our BiFC experiments) for the maturation into the SDS-stable 67LR form of the protein, which are not present in these cells. This would be consistent with the observation that endogenous 37LRP was apparent in these cells (electronic supplementary material, figures S2 and S3), whereas 67LR was not detected. It should also be noted that in cells expressing endogenous 37LRP and 67LR the latter species runs in SDS-PAGE gels with an apparent molecular mass of around 60 kDa, which is lower than might be expected of a homodimer of 37LRP, which has an apparent molecular mass of around 42 kDa (electronic supplementary material, figure S1). The most likely explanation for this is that 67LR is resistant to denaturation in SDS, and thus maintains a more compact structure than during SDS-PAGE and runs with a higher electrophoretic mobility.
Cells were examined 24 h post-transfection by confocal microscopy after fixation without permeabilization. Transfection with full-length YFP-fused proteins (figure 3c,d, respectively; positive control) resulted in the appearance of punctate fluorescence; non-transfected cells (figure 3b) exhibited negligible fluorescence, as did cells transfected with Yn or Yc constructs alone (figure 5d shows FACS analysis of COS7 cells; see below). Co-transfection of cells with 37LRP-Gal-3, 37LRP-37LRP or Gal-3-Gal-3 (each pair carrying complementary Yn þ Yc) yielded fluorescent cells, indicating both homo-and heterodimerization of both molecules (figure 3e-g). Similar results were observed in COS7 cells, whereas in hBMEC cells, punctate fluorescence could be observed in cells co-transfected with 37LRP-Gal-3 or 37LRP-37LRP, while cells co-transfected with Gal-3-Gal-3 did not fluoresce. The reason for this is not known. Quantification and statistical analysis of surface fluorescence of transfected cells, using confocal and FACS analysis, confirmed the statistical significance of homoand heterodimerization of both 37LRP and Gal-3 (figure 5d shows data for COS7 cells).

Surface 67LR is resilient to siRNA knockdown of Gal-3 or 37LRP
Gal-3 or 37LRP was knocked down by targeting LGALS3 or RPSA, respectively, in hBMEC cells over a 48 h period ( figure 4). RT-qPCR analysis confirmed the effective and specific silencing of Gal-3 and 37LRP siRNA treatment at the level of mRNA (figure 4j). Fixed siRNA-treated cells were co-stained for either Gal-3 or 37LRP and compared with untreated or mock siRNA-treated cells. RPSA siRNA-treated cells exhibited significantly reduced levels of both surface 37LRP and Gal-3 (figure 4c,g), whereas Gal-3 siRNA-treated cells exhibited significantly reduced surface levels of Gal-3 (figure 4d), but not 37LRP (figure 4i). Approximately 50% of surface-localized 37LRP was Gal-3-associated, and levels of surface-localized Gal-3 were significantly reduced (by approx. 37%) in the presence of 37LRP siRNA (figure 4i), implying that 37LRP promotes Gal-3 surface localization. It should be noted, however, that 37LRP is a ribosomal protein and it is also possible that depressed levels of this protein could reduce transcription of Gal-3. Surprisingly, in both 37LRP and      rsob.royalsocietypublishing.org Open Biol. 4: 140053 by this time levels had started to recover. By 96 h, 37LRP levels had recovered further and were not significantly lower than in untreated cells (figure 4k). 67LR levels were not significantly different than in untreated or mock-treated cells by 48 h, as had previously been observed. However, by 72 h, there was an approximately 70% reduction in 67LR levels in RPSA siRNA-treated cells compared with untreated or mock-treated cells, which was highly significant. By 96 h, 67LR levels were still significantly depressed in RPSA siRNA-treated cells, although levels had started to recover by this time.
3.5. Gal-3 residue C 173 and 37LRP residue K 166 are implicated in 37LRP -Gal-3 dimerization The sole cysteine residue of murine Gal-3 (C 186 , corresponding to C 173 in its human homologue) has been implicated in Gal-3/Gal-3 homodimerization [40] (figure 5a). We substituted alanine for C 173 (yielding Gal-3 C173A ), and fused the mutant molecule to Yn and Yc. Co-transfection of COS7 cells with complementary Gal-3 C173A Yn and Yc pairs resulted in very little fluorescence, confirming the central role of C 173 in homodimerization and the validity of the BiFC approach (figure 5d). Dimerization of Gal-3 C173A with wild-type Gal-3 or wild-type 37LRP was also significantly reduced (figure 5d); the latter observation suggests a role for C 173 in Gal-3/37LRP heterodimerization. In contrast, alanine substitution of 37LRP's two cysteines (C 148 and C 163 ) had no statistically significant effect on 37LRP cell surface expression cells as measured by FACS analysis of COS7 cells transfected with the relevant BiFC constructs (electronic supplementary material, figure S4). Our molecular model suggested that K 166 of 37LRP may play an important role in the interaction of this molecule with Gal-3 (figure 1). Substitution of alanine for K 166 in 37LRP is predicted to disrupt several interactions between the two molecules (compare figure 1a and figure 5b; for clarity, the position of 37LRP K 166 is highlighted in the proposed interaction of 37LRP with superimposed lactose and nonlactose-liganded Gal-3 shown in figure 5c). To investigate this, we purified recombinant 37LRP and a 37LRP K166A derivative, and determined their binding to Gal-3 in ELISA experiments. The ability of the 37LRP K166A derivative to bind rGal-3 was significantly reduced, confirming the involvement of this residue in heterodimer formation and supporting the validity of our model of heterodimerization (figure 5e). By contrast, substitution with phenylalanine or alanine, respectively, of key residues implicated in homodimerization or laminin binding (Y 139 and R 155 , respectively) [38] had no significant impact on Gal-3 binding (Y139F), or reduced binding significantly but to a much lower degree (R155A; figure 5e). The involvement of K166 of LRP in binding to Gal-3 was confirmed by BiFC complementation experiments; K166A derivatives of 37LRP fused to either the N-or C-terminal portions of YFP interacted to a significantly lower degree with their corresponding C-or N-terminal YFP-labelled Gal-3 partners compared with the wild-type molecules (figure 5f ).

Gal-3 engages meningococci and potentiates cellular invasion
Gal-3 has previously been shown to be expressed in splenic tissue from patients infected with N. meningitidis, but not that of healthy humans, and co-localization of Gal-3 with meningococcal colonies on splenic tissue was demonstrated [33]. Quattroni et al. also showed that Gal-3 binds to the lactosamine moiety of meningococcal LOS. To determine whether interaction of meningococci with Gal-3 could potentiate cellular invasion, N2a cells, which do not express endogenous Gal-3, were transfected with a Gal-3-YFP construct prior to infection with meningococci. Diffuse expression of Gal-3 was confirmed by confocal microscopy (figure 6a). Expression of endogenous 37LRP, which co-localized with Gal-3.vYFP on the cell surface, was also confirmed (figure 6a), confirming that these cells were an appropriate model to study the role of Gal-3 in the interaction of meningococci with host cells.

Neisseria meningitidis binds Gal-3 in a lipooligosaccharide-dependent and -independent manner
Quattroni et al. [33] showed that high concentrations (100 mM) of lactose significantly reduced, but did not abolish, binding of Gal-3 to meningococci. Similarly, meningococcal mutants not elaborating the a-chain of LOS showed significantly reduced, but not completely abolished Gal-3 binding (figure 6c). Collectively, this suggested the presence of additional meningococcal Gal-3 receptors. To explore this, we used lactose-liganded recombinant Gal-3 (Lac -Gal-3) in further ELISA assays. Lac -Gal-3 was biologically functional, as determined by its ability to bind 37LRP, N. meningitidis MC58 and 25 additional isolates, irrespective of their invasiveness, serotype or country of isolation (electronic supplementary material, figure S5 and table S1). Meningococcal binding to Lac-Gal-3 was not inhibited by increasing concentrations of lactose (up to 150 mM) or by adding a range of other Gal-3-binding sugars (Lewis X, Lewis Y, core H type II, galactose and fucose; electronic supplementary material, figure S6), suggesting that lactose-liganding had successfully blocked the carbohydrate-dependent (and thus LOS-dependent) binding capacity of Gal-3.

Gal-3 and 37LRP target common as well as specific meningococcal targets
To identify non-LOS meningococcal Gal-3-binding surface molecules, we employed a re-tagging (a contact-dependent rsob.royalsocietypublishing.org Open Biol. 4: 140053 cross-linking) approach. MC58 cells were incubated with Lac-Gal-3 conjugated to a light-activated cross-linker. Following photoactivation and subsequent exposure to reducing conditions to allow transfer of the reactive biotin moiety to molecules in close proximity to the Lac-Gal-3, bacteria were lysed and the biotin-tagged proteins purified with streptavidin-coated magnetic beads as previously described [41]. Extracted biotin-tagged proteins were separated by SDS -PAGE and identified using MALDI-TOF. This approach identified PilQ (accession: NMB1812; 82.4 kDa; score: as the only Lac-Gal-3 binding meningococcal surface ligands. PilQ is a conserved outer membrane secretin that is absolutely required for type IV pilus extrusion [42]. PilE is the major glycosylated pilin subunit that forms the shaft of the hair-like and retractile type IV pilus fibre [43,44]. A role for each of these proteins in Lac-Gal-3 binding by whole bacteria was confirmed by testing the Lac -Gal-3 binding activity of pilQ and pilE mutants. As expected, MC58 mutants lacking pilE or pilQ showed significantly reduced Lac -Gal-3 binding when compared with wild-type (figure 6d). Mutation of PorA (a 37LRP-binding, surfaceexposed, outer membrane protein employed as a negative control) had no significant effect on Lac-Gal-3 binding (figure 6d ). Because PilQ is required for secretion and surface expression of the PilE-containing fibre, it is not possible from these data to distinguish between a direct role for PilQ binding of Gal-3 from an indirect effect owing to the lack of surface PilE. However, these data show that at least PilE expression is required for maximal Gal-3 binding.

Discussion
It is increasingly well established that 37LRP and Gal-3 have a complex functional relationship and are likely to play collaborative roles during physiological and pathological events, including the interaction of mammalian host cells with pathogens [2,45]. On the cell surface, each of these molecules has been suggested to be both self-associating as well as associating with each other. However, there is extensive confusion in the literature regarding the relationship between the 37LRP species, assumed to represent the monomer, and a 67 kDa 'mature' form of the protein, which has been described as either a homodimer or a heterodimer of 37LRP with Gal-3. Here, we present evidence for both self-and mutual association of both 37LRP and Gal-3. We employed molecular modelling to visualize the 37LRP -Gal-3 interaction supporting previous predictions of mutual association, and further informed the putative critical interfaces and key residues involved. BiFC experiments supported the hypothesis that 37LRP and Gal-3 both homo-and heterodimerize, and that the single cysteine (C 173 ) of Gal-3 (but not the two cysteines of 37LRP) is critical for these associations, as was previously shown for the murine Gal-3 molecule's equivalent residue C 186 [40]. We showed that monomeric 37LRP, its presumed homodimer 67LR and the 37LRP/Gal-3 heterodimer coexist as distinct populations on the host cell surface. It is of note that 67LR has previously been shown to incorporate a Gal-3-specific epitope that is not present on monomeric 37LRP [23,45]. It is likely that this epitope was responsible for the observed co-localization of 67-LR with one or more Gal-3 specific epitopes when a polyclonal antiserum was employed. We show that this is not due to the presence of Gal-3 in this complex, however, as it is not detected by two independent Gal-3-specific monoclonal antibodies. It is possible that homodimerization results in the creation of a new epitope that mimics a Gal-3 epitope. Alternatively, the polyclonal anti-Gal-3 antiserum may contain 67LR cross-reacting antibodies that are not specific for Gal-3. The latter possibility seems unlikely, however, as the polyclonal antiserum employed was raised against recombinant Gal-3 produced in E. coli. Gal-3 preferentially associated with 37LRP and rarely (if at all) with 67LR, and knockdown of Gal-3 had no discernable qualitative or significant quantitative effects on cell surface 67LR, confirming early models which suggested that Gal-3 is not a component of 67LR [9,23]. Interestingly, while 37LRP and Gal-3 are shown here to associate, these proteins dissociate under denaturing SDS-PAGE and thus cannot represent 67LR, which is known to be resistant to dissociation in the presence of SDS. Elevated cell surface 67LR is a prognostic marker for metastatic potential in cancer, yet overexpression of 37LRP by transfection is not coincident with increased 67LR, which is in line with previous observations [46 -48]. Indeed, the transition of 37LRP to 67LR has not been conclusively demonstrated to date, and recombinant 37LRP has never been unambiguously shown to associate with higher molecular mass isoforms corresponding to the native 67LR isoform [47]. Our data support the hypothesis that 67LR is a homodimer of 37LRP, but it remains a possibility that the ability of 37LRP to homodimerize that we have demonstrated does not represent 67LR, and further experiments will be required to conclusively settle this matter. The delayed reduction in 67LR expression after prolonged siRNA knockdown of RPSA described here represents the clearest evidence yet that 67LR is indeed derived from the smaller protein. It should be born in mind, however, that another plausible explanation for the reduction in 67LR after prolonged knockdown of RPSA is due to a general reduction in protein synthesis that knockdown of 37LRP, a known ribosomal protein, results in a more generalized depression of new protein synthesis. While some 37LRP partitions to the membrane fraction, this precursor protein is predominantly isolated in soluble cytoplasmic cell fractions, where it is known to have a clearly defined ribosomal association and function. By contrast, 67LR has been clearly demonstrated previously to predominate in the membrane and, in particular, preferentially localizes to the environment of 'lipid raft' domains and focal plaques, co-localizing with vinculin and b-actinin [3,49]. Our demonstration that 37LRP and 67LR are essentially discrete populations implies that the majority of membrane-associated 37LRP is not lipid-raft-resident or focal-plaque-associated. While we acknowledge that transfection and high-level expression of recombinant proteins can give rise to anomalous protein localization, and that single amino acid changes can have an effect by altering the overall conformation of a protein, the observation that our data are consistent with the in silico modelling data supports our conclusions based on the BiFC experiments.
Therapeutic strategies based on 67LR currently involve selective drug targeting, antibodies against 37LRP, polysulfated glycans and siRNA treatment. Resolving the relationship and mechanism(s) governing 'maturation' of 67LR may provide new approaches in combinatorial cancer therapy. As interest in LAMR1 derives from its potential for therapeutic intervention, and the molecule is clearly a pathologically significant multi-functional protein involved in a wide variety of processes, assigning roles to differing receptor populations and determining any potential pleiotropism or functional redundancy will be critical for effective and selective targeting.
While the precursor -product relationship between 37LRP and 67LR is widely accepted, the mechanism of 67LR maturation remains unresolved. As we show that 37LRP is the predominant Gal-3-associated LAMR1 population, it is tempting to speculate that Gal-3 may play a regulatory or rsob.royalsocietypublishing.org Open Biol. 4: 140053 mechanistic role in this process. Indeed, modelling revealed that the interaction domains for 37LRP homo-and heterodimer formation may overlap, and thus form the basis for two mutually exclusive LAMR1 cell surface populations. In support of our hypothesis that a proportion of cell surface Gal-3 is spatially restricted to 37LRP, previous reports by Buto et al. [23] showed that treatment of A431 cells with cerulenin, to inhibit 37LRP acylation, resulted in a marked decrease in detectable Gal-3 and 67 kDa LAMR1 in the NP-40 soluble membrane fraction [23].
The fact that 37LRP and Gal-3 bind common and separate antigens in the human host further supports the hypothesis that the two proteins engage in a joint and mechanistically synchronized interaction during physiological and pathological events. Here, we show that this is paralleled during bacterial-host cell interactions. Using N. meningitidis as a model invasive organism that is known to interact with both Gal-3 and 37LRP [7,33], we showed that both host molecules are targeted by a common bacterial surface ligand (PilQ). The latter protein forms a large (960 kDa) homo-dodecamer complex with an apparent cavity through which the type IV pilus fibre is exported and retracted [50]. PilE is the major subunit that forms the shaft of that fibre [43]. PilQ and PilE remain intimately positioned and are surrounded by LOS, which is the predominant lipid moiety of the outer leaflet of the outer membrane. PorA, the other specific meningococcal ligand for 37LRP, is also a pore-forming outer membrane protein present in large numbers, and juxtapositioned with both PilQ and LOS [7,51]. Meningococci, and perhaps other bacterial pathogens, are likely to have evolved a closely positioned array of 37LRP-and Gal-3-binding surface ligands, with consequences for the virulence of these pathogens. Ligand-receptor interactions may act in concert to enhance cell-cell intimacy and affinity, and trigger downstream host cell-signalling and cytoskeletal rearrangements as a prelude to cellular invasion and tissue penetration [52].
Taken together, these findings lead us to propose a model for the dynamic and mechanistic interaction of the host and bacterial surface proteins (figure 7). We propose that upon entry into the host (mucosa or blood stream), N. meningitidis would initiate contact with host cell surfaces via the type IV pili, which extend up to several micrometres, and at the tip may contain the adhesive PilC1 [44]. Retracting pili induces twitching motility that allows the organism to move over host cell surfaces against the tide of mucus or blood flow, until it comes into intimate contact with specific receptors [53]. Secreted forms of Gal-3 may bind LOS and/or bacterial protein ligands, a process that is likely to protect the host from bacterial colonization. However, meningococci may use Gal-3 to enhance its adherence to the host cell surface via Gal-3/ 37LRP heterodimerization, or shed the Gal-3-LOS complex as part of the outer membrane vesicles (which are well known to be shed continuously). Interestingly, surface Gal-3 is thought to be a negative regulator of the LOS-induced inflammatory response, and protects the host from endotoxin shock while also promoting survival of invading bacteria [54]. Similarly, when triggered by specific molecules (e.g. the LAMR1-binding green tea-derived polyphenol, EGCG), surface LAMR1 is believed to inhibit TLR4 and TLR2 responses in mouse macrophages [55,56].
Given that both 37LRP and Gal-3 are broadly distributed in different cellular compartments, their collaborative interaction with the meningococcus or its ligands may not be limited to the cell surface. In the cytoplasm, Gal-3 can bind to Bcl-2 and inhibit cell apoptosis, and in the nucleus it can regulate gene transcription [31,57]. Suppression of Gal-3 in highly malignant human breast carcinoma cells resulted in reversion of the transformed phenotype and led to the inhibition of tumour growth in immunologically suppressed mice [58]. Extracellular Gal-3 can bind to cell surface N-glycans and induce monocyte and T-cell apoptosis, which may help bacterial or tumour cell evasion from the immune surveillance. Gal-3 can suppress IL-5 production and inhibit B-lymphocyte differentiation, and increase phagocytosis of neutrophils. In human monocytes, Gal-3 is chemotactic and increases calcium influx at high concentrations, whereas at low concentrations it promotes chemokinesis [59].
In conclusion, we demonstrate the presence of discrete populations of LAMR1 on the epithelial and endothelial cell surface. We show that 37LRP can form homodimers (67LR) as well as heterodimers with Gal-3, which is also capable of forming homodimers. These surface molecules  Figure 7. Illustration of meningococcal binding to host cells using surface ligands. Initial contact mediated by PilC1, which binds its putative receptor, followed by 37LRP/Gal-3-PilE binding, which strengthens the initial contact, and pilus retraction, which brings the two cells closer together. Decametric PilQ on the meningococci surface binds to both 37LRP and Gal-3, establishes intimate contact, strengthens interactions, and triggers host cell signalling events.
rsob.royalsocietypublishing.org Open Biol. 4: 140053 are likely to play distinct roles in a number of oncogenic, neurodegenerative and infectious diseases. 37LRP and Gal-3 target common and specific meningococcal surface ligands, events that are exploited by the organism to enhance its adhesion and cell invasion. Our data enhance our understanding of the roles of Gal-3 and 37LRP in host cell biology, which can be exploited for the design of therapeutic and/or preventive strategies against invasive bacterial and/ or oncogenic diseases. Coverslips were mounted with Prolong Gold anti-fade with/without DAPI (Invitrogen) and images obtained as sections, unless stated otherwise, by confocal microscope. For each experiment, unstained and secondary antibody-alone samples were processed in parallel to control for non-specific staining. Additionally, all possible primary/secondary combinations were checked for cross-reactivity and sequential staining used as required. For all co-localization studies, samples stained for localization of each protein individually were processed in parallel and, in all cases, the fluorescence of adjacent channels was monitored for bleed through.

Confocal microscopy
Images (400-600 nm optical sections) were acquired using a Zeiss LSM 700 AxioObserver confocal microscope using a Plan-Apochromat 63Â/1.40 Oil DIC M27 objective with ZEN 2009 operating software. Images were processed using IMAGEJ and Adobe PHOTOSHOP software. Raw data are available upon request. Co-localization analyses used the ZEN software function with an FI threshold of 50; the data were not intensity-weighted. Mean FI data of individual cells or fields of cells were also obtained using ZEN software.

siRNA of galectin-3 and LAMR1
hBMECs were cultured as described and transfected with predesigned siGENOME SMARTpool (mix of four siRNAs targeting one ORF) targeting LAMR1 (M-013303-01-0005), galectin-3 (M-010606-02-0005) or ON-TARGETplus nontargeting pool control siRNA (Dharmacon, Thermo Scientific). Cells were transfected on day 1 with a final concentration of 37.5 nM siRNA using the DharmaFECT transfection reagents (Dharmacon, USA). The inhibition of LAMR1 and Gal-3 expression was assessed at 48, 72 or 96 h post-transfection 3 by confocal analysis, quantitative polymerase chain reaction (RT-qPCR) and immunoblot analysis. To determine transfection efficiency, non-transfected and siGLO-transfected cells were analysed 1 day after transfection on a FACSCaliber machine (Beckton-Coulter). All galectin-3 or LAMR1 confocal images were acquired at the same resolution and scale, with the same hardware/laser settings used to image untreated control cells. Identical image manipulations were performed on all images stained with the same antibody set.

Purification of recombinant protein
Cells from 50 ml culture were re-suspended in 5

Immunoblotting
Cells were lysed with RIPA buffer supplemented with Phosstop (Merck Millipore) and complete mini EDTA-free protease inhibitor cocktail. Alternatively, cells were fractionated using a cell fractionation kit (Thermo Fisher Scientific) as described in the manufacturer's instructions. Protein samples were separated on 4-20% gradient SDS -PAGE (Thermo Scientific) at 125 V and calibrated with Colourplus broad pre-stained marker. Gels were transferred to nitrocellulose membranes (BioRad) at 10 V for 30 min on a BioRad semi-dry transfer system. Membranes were blocked in Tris-buffered saline with 0.1% Tween20 (TBS-T) containing 5% bovine serum albumin (BSA, w/v). Primary antibodies (in TBS-T/5% BSA) were incubated with the membrane at either 48C overnight or 1 h at room temperature. Membranes were washed (3 Â 15 min) in TBS-T and subsequently probed with conjugated secondary antibody (in TBS-T/5% BSA) for 45 min at room temperature. The membrane was washed (5 Â 10 min) with TBS-T before membranes were exposed to ECL substrate (Luminata Crescendo; Millipore) for visualization of immuno-reactive proteins. Antibodies used included anti-67LR (MLuC5), anti-37-kDa LAMR1 (A7 or IHLR) or anti-Galectin-3 (9H3.2).

Modelling the molecular interaction of galectin-3 with LAMR1
The structures of the Gal-3 CRD and LAMR1 were obtained from the RCSB protein data bank (www.rcsb.org). The molecular docking of Gal-3 (3zsj.pdb: lactose-liganded structure; 3zsm.pdb: non-liganded structure) and LAMR1 (3bch.pdb) was performed employing a server in zdock.umassmed.edu [60,61] based on current knowledge of crystal structures. RASMOL [62] and UCSF CHIMERA [63] were used to generate images of the molecular interactions and docking.

Bacterial strains, growth conditions and invasion assays
Neisseria meningitidis clinical isolates (electronic supplementary material, table S1) were grown on chocolate horse blood (Oxoid) at 378C, in an atmosphere of 5% CO 2 . Mutagenesis of N. meningitidis MC58 pilQ and porA was described previously [7]. To mutate pilE, chromosomal DNA extracted from N. meningitidis C311DpilE (kindly provided by Prof. C. Tang, University of Oxford, UK) was used to mutate MC58 by natural transformation and allelic exchange as described previously [64]. The MC58DlgtF strain used in this study was described previously [65]. For selection of mutants, meningococcal cells were cultured on Mueller-Hinton agar plates supplemented with 1% Vitox (Oxoid) and, where appropriate, with streptomycin and spectinomycin (100 mg ml 21 ) or kanamycin (50 mg ml 21 ). Invasion assays were performed as previously described [66].

Enzyme-linked immunosorbant assays
100 ml lactose-purified recombinant human Gal-3 (Calbiochem) or BSA (5 mg ml 21 ) in PBS was used to coat amino-reactive 96-well microtiter plates (Immobilizer Amino; Nunc) overnight at 48C. Bacterial strains were grown in liquid culture, washed and labelled with digoxigenin (Roche) as described previously [51]. Labelled bacteria were added to ELISA plates for 2-4 h at room temperature. Plates were washed with PBS/T and incubated with 100 ml polyclonal anti-digoxigenin F ab fragment-alkaline phosphatase antibody (1 : 5000; Roche) in PBS/1% BSA for 1 h, and then washed several times as described above. 100 ml of alkaline phosphatase substrate (5 mg ml 21 ; Roche) was added to each well, and the absorbance measured at 405 nm after 15 min using an ELISA reader (Biotek EL800). Inhibition assays were performed as described above, except that bacteria were pre-incubated with lactose or sugars for 2 h at room temperature before being added. For LamR/Gal-3 binding, 100 ml aliquots of 8.7 mg ml 21 37LRP proteins were immobilized as above.
Following washing in PBS/T, wells were blocked with 1% BSA/PBS for 1 h. 100 ml of 5 mg ml 21 Gal-3 was then added and incubated at room temperature for 1 h. After washing, 100 ml of mouse anti-Gal-3 (9H3.2; 1 : 8000 diluted in 1% BSA/PBS) was added and incubated at 48C overnight. After washing, 100 ml anti-mouse IgG-HRP conjugate (1 : 8000 diluted in 1% BSA/PBS) was added and incubated at 48C overnight. Plates were again vigorously washed and colour developed by adding 100 ml ABTS substrate (Roche). Plates were read at an absorbance of 405 nm.

Cross-linking
Cross-linking was performed as described previously [41]. Briefly, N. meningitidis was incubated with Gal-3 conjugated to the light-activated cross-linker Sulfo-SBED. After photoactivation, in which the reactive biotin moiety is transferred to molecules in close proximity to the cross-linking agent, cells were washed, lysed, subjected to SDS -PAGE and immunoblotting, and probed with streptavidin before molecules were identified using MALDI-TOF.