|Home | About | Journals | Submit | Contact Us | Français|
Surfactant protein D (SP-D), a lectin that recognizes carbohydrates via its C-type carbohydrate recognition domains (CRDs), regulates Mycobacterium tuberculosis (M.tb)–macrophage interactions via recognition of M.tb mannosylated cell wall components. SP-D binds to, agglutinates, and reduces phagocytosis and intracellular growth of M.tb. Species-specific variations in the CRD amino acid sequence contribute to carbohydrate recognition preferences and have been exploited to enhance the antimicrobial properties of SP-D in vitro. Here, we characterized the binding interaction between several wild-type and mutant SP-D neck + CRD trimeric subunits (NCRDs) and pathogenic and nonpathogenic mycobacterial species. Specific amino acid substitutions (i.e., the 343-amino-acid position) that flank the carbohydrate binding groove led to significant increases in binding of only virulent and attenuated M.tb strains and to a lesser extent M. marinum, whereas there was negligible binding to M. avium complex and M. smegmatis. Moreover, a nonconserved mutation at the critical 321-amino-acid position (involved in Ca2+ coordination) abrogated binding to M.tb and M. marinum. We further characterized the binding of NCRDs to the predominant surface-exposed mannosylated lipoglycans of the M.tb cell envelope. Results showed a binding pattern that is dependent on the nature of the side chain of the 343-amino-acid position flanking the SP-D CRD binding groove and the nature of the terminal mannosyl sugar linkages of the mycobacterial lipoglycans. We conclude that the 343 position is critical in defining the binding pattern of SP-D proteins to M.tb and its mannosylated cell envelope components.
Mycobacterium tuberculosis (M.tb) infects one third of the world's population and is one of the leading causes of death due to a single infectious disease agent (WHO Tuberculosis fact sheet). It is a pulmonary pathogen that travels on a 1–5 μm respiratory droplet into the distal airways and alveoli where it encounters its natural niche, the alveolar macrophage (AM). AMs are professional phagocytic cells which engulf particulate matter and microbes into phagosomes that later fuse with lysosomes containing lytic compounds leading to destruction of the phagosomal contents. M.tb has developed mechanisms to avoid this intracellular destruction, in part by its interaction with soluble and cellular receptors leading to subsequent phagosome maturation arrest (Chua et al. 2004; Kang et al. 2005).
The alveolar environment of the lung, which is rich in surfactant, plays a major role in regulating the interaction between M.tb and the macrophage at the time of infection. The multimolecular surfactant complex contains several innate immune defense components including surfactant protein D (SP-D). SP-D is a calcium-dependent, collagenous lectin (collectin) that consists of varying multimers of a basic trimeric subunit (Figure (Figure1A)1A) (Crouch et al. 1994). Each SP-D trimer consists of three disulfide-linked monomers (Figure (Figure1B),1B), each of which is composed of a carbohydrate recognition domain (CRD), an α-helical coiled-coil neck domain, a collagen-like domain, and an N-terminal domain (reviewed in Crouch (1998)). Optimal SP-D antimicrobial activity requires binding to microbial polysaccharides through the three closely associated CRDs on a trimer (Ogasawara and Voelker 1995; Hakansson et al. 1999; Shrive et al. 2003). This binding is influenced by the amino acid sequences flanking the carbohydrate binding groove of SP-D, and amino acid positions 325–328 and 343–349 (Figure (Figure1C).1C). Incompletely conserved amino acid residues along the binding groove, such as the unique arginine at position 343 in the native human protein, contribute to species-dependent variation in binding preference to carbohydrates and pathogens (Crouch et al. 2006). Mutations to position 343 (Figure (Figure1C,1C, supplementary Table I) lead to altered SP-D carbohydrate binding affinity. For example, substitution of lysine 343 (murine SP-D) for arginine 343 on a human SP-D background changes human SP-D to resemble the rodent protein (Crouch et al. 2006) by enhancing its binding to phosphatidyl-myo-inositol (Crouch et al. 2007) and certain forms of LPS (Wang et al. 2008). Similarly, substituting valine 343, which is found in a similar C-type lectin, mannose binding protein (MBP), alters the monosaccharide binding affinity of SP-D to mimic that of MBP (Allen et al. 2004). Mutations to the CRD have also been associated with greater interaction and antimicrobial activity of the protein against intracellular pathogens such as influenza virus (Crouch et al. 2005, 2009) indicating that specific mutations to the CRD can be used to enhance antimicrobial properties of the protein.
Work from our laboratory has shown that native SP-D and an SP-D collagen-deletion mutant bind to and agglutinate M.tb. The resultant SP-D-coated bacteria are reduced in phagocytosis by human macrophages and those bacteria that do enter demonstrate reduced survival and increased phagosome–lysosome (P-L) fusion (Ferguson et al. 1999, 2002, 2006). Thus, these studies support the idea that SP-D acts as a host defense molecule in the local environment of the lung alveolus. The SP-D-M.tb interaction is defined by SP-D recognition of the surface-exposed mannosyl units of mannose-capped lipoarabinomannan (ManLAM) (Ferguson et al. 1999) and is Ca2+-dependent and carbohydrate inhibitable, providing evidence that SP-D interactions with M.tb occur through its CRD.
The cell envelope of M.tb plays a critical role in the survival of the bacteria within macrophages (Briken et al. 2004; Fenton et al. 2005; Torrelles et al. 2009). It is composed of a multilayered, complex structure consisting of peptidoglycan, arabinogalactan, mycolic acids, peripheral lipids, and a surface exposed outer layer (Crick et al. 2003). The outer layer consists mainly of polysaccharides and proteins and is thought to contain exposed mannosylated moieties from lipoglycoconjugates such as the biosynthetically related ManLAM, lipomannan (LM), and phosphatidyl-myo-inositol mannosides (PIMs) (Crick et al. 2003; Pitarque et al. 2008) (Figure (Figure2).2). ManLAM has only been described in pathogenic Mycobacterium spp. such as M.tb, M. leprae, M. avium, and M. marinum among others (Vercellone et al. 1998; Khoo et al. 2001; Pitarque et al. 2005). In contrast, nonpathogenic Mycobacterium spp., such as M. smegmatis, contain phosphoinositol-capped LAM (PILAM) (Khoo et al. 1995a) (Figure (Figure2).2). PIMs are a heterogeneous mixture of families, which differ by the number of mannosyl residues in their structure (PIMxf, x = number of mannoses from 1 to 6). Each family is composed of several species that differ in their fatty acid content (AcyPIMxf, y = number of fatty acids where 0, 1, or 2 denotes di-, tri-, or tetra-acylated species, respectively) (Khoo et al. 1995b). We recently showed that mannose-containing components on the surface of M.tb (i.e., ManLAM and higher-order PIMs (i.e., PIM5f and PIM6f)) play an important role in dictating the early intracellular fate of the M.tb bacillus by interacting with the human macrophage mannose receptor (MR) (Schlesinger et al. 1994; Torrelles et al. 2006). This interaction leads to limited P-L fusion following phagocytosis (Kang et al. 2005).
Recombinant fragments of SP-D trimers consisting of the neck + CRD (NCRD) with an S-protein binding tag have recently been constructed (Figure (Figure1A,1A, C and supplement Table 1) (Crouch et al. 2005, 2006, 2007) allowing for refined analysis of SP-D CRD–carbohydrate interactions (Hakansson et al. 1999; Shrive et al. 2003). These NCRD proteins have been shown to retain mycobacterial binding activity but lack the agglutination function attributed to the native molecule (Ferguson et al. 2002). Here, using NCRD wild-type and mutant proteins with alterations near the CRD binding groove, we defined specific CRD mutations that enhance binding to pathogenic and attenuated M.tb strains and M. marinum but not to M. avium complex and the nonpathogenic M. smegmatis. We determined the M.tb mannosylated cell envelope components that are directly recognized by these SP-D mutants and further we defined the motifs of these cell wall components involved in NCRD binding. Our results demonstrate that the side chain of amino acid 343 in the SP-D CRD binding groove is essential in regulating the binding of this protein to mycobacterial species and their mannosylated cell envelope components.
To test the effect of site-directed mutations to the CRD on the relative binding of SP-D to mycobacteria, we used a series of NCRD mutants (supplementary Table I) derived from the human wild-type amino acid sequence and evaluated their binding profiles to different mycobacterial species using an ELISA. We focused on mutants in the 343 position which flanks the binding groove of the CRD. We also compared human (hNCRD) and rat (rNCRD) wild-type NCRD proteins. The human mutants included some that mimic the amino acid sequence of the SP-D CRD from other species, such as the R343K mutant, which mimics the rat and mouse NCRD. Other mutants resemble related serum collectins including human MBP and bovine serum conglutinin (R343V) and rat MBP (R343I). Single mycobacteria cell suspensions were plated onto wells of 96-well microtiter plates, and NCRD mutant binding was assessed by ELISA using maltose as a specific competitor for the CRD binding site.
Results in Figure Figure33 (A, representative figure, B cumulative data) show the greatest binding for the R343V mutant to virulent and attenuated M.tb strains when compared to hNCRD. The low relative binding of hNCRD is consistent with published results reporting lower affinity binding of the collagen deletion mutant (Ferguson et al. 2002). The R343I and R343A mutants also demonstrated a high level of binding (supplementary Figure 1A). The E321K mutant failed to bind to the mycobacterial species. This mutation corresponds to a site that contributes to Ca2+ coordination. All NCRD mutants tested showed a similar binding profile to virulent (Erdman and H37Rv) and attenuated (H37Ra) strains of M.tb. This binding profile was also observed for M. marinum, although at a lower level. In contrast, NCRD mutant binding to M. smegmatis and M. avium 104 was negligible (Figure (Figure33 and supplementary Figure 1A).
SP-D has been shown to bind to the mannosyl units of M.tb ManLAM but not the phosphoinositol caps of PILAM from M. smegmatis (Ferguson et al. 1999). Since ManLAM is a major mannosylated lipoglycan exposed on the surface of M.tb (Pitarque et al. 2008), we next tested whether NCRD mutants that demonstrate high binding to M.tb strains also demonstrate enhanced binding to ManLAM, as compared to PILAM. LAMs from different mycobacterial species were purified, and NCRD protein binding to LAMs was assessed by ELISA. Results in Figure Figure4A4A show that the NCRD proteins bind to ManLAMs from the M.tb strains tested. Similar to the results with intact bacteria, the R343V mutant showed the best binding to all of the ManLAMs. The R343I mutant also showed a high level of binding along with the R343A mutant, although the latter mutant showed differences in binding among the LAM types (supplementary Figure 1B). rNCRD bound better to the ManLAMs than did the hNCRD, but interestingly no binding was appreciated for the R343K mutant which mimics rNCRD at this amino acid position. Finally, the E321K mutant lacked binding to the LAMs consistent with the results using bacteria. Generally, SP-D NCRD mutants bound similarly to ManLAMs from the different M.tb strains. All M.tb ManLAMs are reported to have primarily di- and tri-mannose-capped LAM (i.e., t-Manp-α-(1→2)-Manp-β-(1→5)-Araf and t-Manp-α-(1→2)-Manp-α-(1→2)-Manp-β-(1→5)-Araf, respectively, see Figure Figure2)2) (Nigou et al. 2003). None of the SP-D mutants bound to PILAM in agreement with the binding data using M. smegmatis whole bacteria.
M.tb LM also contains surface exposed mannosyl units (Pitarque et al. 2008) and is an important immunomodulatory molecule (Vignal et al. 2003). Structurally, M.tb LM consists of a mannosyl phosphatidyl-myo-inositol (MPI) anchor with a long α(1→6) mannan chain with α(1→2) mannosyl branches (see Figure Figure2).2). Due to its abundance in the mycobacterial cell wall and its immunomodulatory function, we studied the interaction of LM with SP-D NCRD mutants by ELISA. Results in Figure Figure4B4B show that the binding to LM was highest with R343V. In addition, binding of rNCRD (Figure (Figure4B)4B) and that of the R343I mutant (supplementary Figure 2A) were significantly increased compared to hNCRD.
Like LM, PIMs are biosynthetically related to ManLAM and are thought to expose their terminal mannosyl moieties on the M.tb cell envelope surface (Villeneuve et al. 2005). Higher-order PIMs (i.e., PIM5f and PIM6f) are expressed in abundance only in pathogenic M.tb strains (Torrelles et al. 2006) and contain a similar nonreducing mannosyl terminus as found in ManLAM (Torrelles et al. 2006) (see Figure Figure2).2). Thus, we examined the binding of NCRD proteins to PIM families by ELISA.
Binding of NCRDs to the PIMs was highly dependent on the specific mutation and the number of mannosyl units present in the M.tb PIM (Figure (Figure5).5). For example, the R343I mutant showed the highest binding to the PIM2f. rNCRD and the R343K and R343V mutants showed detectable binding to both the PIM2f and PIM5f with rNCRD preferentially binding to the PIM2f and the R343V mutant preferentially binding to the PIM5f. Unexpectedly, no NCRD protein showed significant binding to the PIM6f.
PIMs are significantly smaller than LAMs (Figure (Figure2),2), and their binding is potentially influenced by their phosphatidyl-myo-inositol (PI) anchor to a greater extent. We previously determined that the presence of the fourth fatty acid in the PI anchor of PIMs influenced the ability of the mannosyl termini to bind to the MR (Torrelles et al. 2006). Thus, we separated PIMs by their polarity into species using 2D TLC (Figure (Figure6A)6A) and further evaluated their recognition by the NCRD proteins by immuno-lectin blotting (Figure (Figure66B).
Consistent with our ELISA results, R343V bound preferentially to the PIM5f (Figure (Figure6B)6B) and the immuno-lectin blot showed that the binding was independent of the degree of acylation of the PIMs (i.e., Ac1PIM5 and Ac2PIM5). Also consistent with the PIM ELISA, rNCRD bound to both the PIM2f and PIM5f and once again the binding was independent of the degree of acylation (species) of these PIM families.
Contrary to the ELISA results, R343K did not appreciably bind to the PIM2f. However, an additional spot was detected for rNCRD and the R343K mutant (Figure (Figure6B)6B) and, to a certain extent, R343I (supplementary Figure 2B). Co-migration experiments by 2D TLC using commercial PI identified this spot as PI. This is consistent with published work showing that the basic side chain for the 343 position seen in lysine is involved in increasing binding to PI (Crouch et al. 2007).
The binding properties of the SP-D CRD to carbohydrates are determined, in part, by the orientation of the presenting residues, which relates to their linkage pattern (Allen et al. 2001). In order to evaluate the linkage preferences of select NCRDs and compare them to linkage patterns found on mycobacterial lipoglycans, a high-mannose carbohydrate microarray was employed which models a microbial surface, with the reducing end attached to the slide and the terminal carbohydrate residues freely available in the aqueous phase. Residues of varying carbohydrate moieties and linkages (Figure (Figure7A)7A) were exposed to soluble human and rat wild-type NCRDs, and the R343V and E321K mutants either with or without a competitive inhibitor.
The R343V mutant bound to α(1→6)-Man compounds (Figure (Figure7B,7B, compounds 6, 10, and 11); nomenclature used as previously published (Crouch et al. 2009) with increased binding that correlated with the number of mannosyl units. Compounds 6, 10, and 11 mimic motifs present in the LM structure, i.e., a linear α(1→6) mannan chain with several single α[1→2] mannosyl branches (Figure (Figure7A).7A). These results are consistent with the LM ELISA which showed an increase in binding of the R343V mutant to LM (Figure (Figure44B).
The nonreducing terminal sugars of the t-Manp-α-(1→2)-Manp-α-(1→2)-Manp-R (Figure (Figure7A,7A, compound 4) approximate the appearance of the tri-mannoside caps of ManLAM. This linkage promoted strong binding to R343V and hNCRD (Figure (Figure7B).7B). Significantly less binding was noted for the single bound mannose, t-Manp-R (Figure (Figure7A,7A, compound 7) which approximates a mono-capped ManLAM. This is consistent with the binding results of the whole bacteria and LAM ELISAs (Figures (Figures33 and and4A)4A) where the NCRDs show greater binding to the predominant di- and tri-mannoside-capped ManLAM present in M.tb compared to the predominant mono-mannoside-capped ManLAM defined in M. marinum (Pitarque et al. 2005). Interestingly, R343V binding to compound 5 (Figure (Figure7)7) was increased compared to hNCRD. This compound, t-Manp-α-(1→2)-Manp-α-(1→6)-Manp-R, mimics the nonreducing terminal sugars of PIM5f. Therefore, these results are consistent with the PIM ELISA results (Figure (Figure5)5) showing increased R343V binding for PIM5f compared to hNCRD. Binding to the arabinose chains (Figure (Figure7A,7A, compounds 8 and 9) was minimal to absent for all tested NCRDs (data not shown). This confirms our previous studies showing that enzymatic removal of the mannosyl caps of LAM abolished the binding of SP-D to the underlying arabinose units of ManLAM (Ferguson et al. 1999). The E321K mutant showed no binding to any of the carbohydrates (Figure (Figure77B).
The serum bovine collectin, CL-43 has greater affinity for mannose compared to other carbohydrates (Hartshorn et al. 2002). Modifications to the CRD of SP-D to mimic CL-43 successfully enhance binding of the protein to mannose thereby increasing the antimicrobial effects of SP-D against high-mannose containing pathogens (Crouch et al. 2005). Thus, we hypothesized that modifying the amino acid sequence to mimic bovine CL-43 would enhance the binding of SP-D to M.tb. To test this hypothesis, two additional mutants of the SP-D NCRD were created to mimic CL-43. The first mutant had incorporated a three-amino-acid insertion N-terminal to the binding groove (RAK). The second mutant contained the RAK insertion with the addition of a single substitution at 343 (RAK+R343I) (Figure (Figure11C).
Results in Figure Figure88 show the addition of the RAK insertion-enhanced binding for the RAK+R343I mutant (compared to the R343I mutant) to M.tb H37Rv, as well as to M.tb Erdman and H37Ra (data not shown). Results also demonstrate increased binding of this mutant to M.tb ManLAM and LM, respectively, in a pattern that is very similar to whole bacteria. However, the RAK insertion did not enhance the binding of RAK+R343V to mycobacteria and mycobacterial cell wall components compared to R343V (data not shown). Together these results suggest that the RAK insertion permits enhanced interactions of the 343 side chain of isoleucine with mycobacteria and their mannosylated surface structures, but does not enhance the binding of SP-D with valine at this position.
SP-D is an immunomodulatory protein that has antimicrobial effects against many pulmonary pathogens including M.tb (Ferguson et al. 1999, 2002; Kishore et al. 2006). Interactions of SP-D with microorganisms are often mediated through attachment of its CRD to microbial polysaccharides such as the mannosyl caps of ManLAM. Recently, mutations to the SP-D CRD have been demonstrated to increase its antimicrobial properties likely by increasing association with high-mannose surface moieties (Crouch et al. 2005, 2009). Our results show that specific point or insertion mutations flanking the primary carbohydrate binding site alter and, in some cases, greatly enhance recognition of M.tb and their surface exposed mannosylated glycoconjugates. To our knowledge, this is the first direct evidence that interactions of SP-D with M.tb occur specifically via amino acids associated with Ca2+ in the CRD and that these interactions can be enhanced by genetic manipulation of the CRD.
Binding of SP-D to carbohydrates is Ca2+-dependent (Persson et al. 1990). Thus, substitutions that interfere with Ca2+ coordination at the carbohydrate binding site are predicted to interfere with carbohydrate binding and microbial recognition. Our results show that an amino acid substitution replacing the wild-type glutamic acid (negatively charged, moderately sized) for lysine (positively charged, bulky) at position 321, completely abrogated binding of the NCRD to mycobacteria which we hypothesize is likely as a result of electrostatic repulsion of the Ca2+. The complete lack of binding of the E321K mutant provides further evidence that binding to mycobacterial ligands involves the CRD and requires the calcium-dependent lectin activity of SP-D (Persson et al. 1990).
Within our panel of mutants of the 343 position (Figure (Figure11 and see supplementary Table I), the R343V mutant demonstrates marked binding to virulent (i.e., H37Rv and Erdman) and an attenuated (i.e., H37Ra) M.tb strain. In general, our results show that smaller, nonpolar, aliphatic amino acids such as valine (V), isoleucine (I), and alanine (A) (R343V, R343I, and R343A; Figure Figure1,1, supplementary Figures 1 and 2) are more permissive to binding M.tb and/or its cell wall components compared to the larger, more polar amino acids such as lysine (K) and wild-type arginine (R). This is consistent with computational studies that suggest that R343V decreases steric hindrance and may allow for greater accessibility of the carbohydrate residues to the binding groove and/or increase the number of binding orientations possible at this site (Allen et al. 2004). Our results with these mutants illustrate the importance of the amino acid characteristics at position 343 of the NCRD of SP-D in the binding of M.tb. This is supported by several studies showing the influence of position 343 on CRD binding to other ligands (Allen et al. 2001; Crouch et al. 2006, 2007, 2009).
The major surface exposed lipoglycan of M.tb, ManLAM, is of particular interest among the exposed mannosylated molecules due to its abundance, immunomodulatory properties, and roles in the pathogenesis of tuberculosis infection (Briken et al. 2004). Significant species and strain variations exist in the number of mannosyl units per cap on the ManLAMs in pathogenic mycobacteria (Khoo et al. 2001; Pitarque et al. 2005), where the number of mannosyl units per cap directly correlates with ManLAM association with specific C-type lectins (Schlesinger et al. 1994, 1996; Ferguson et al. 1999; Maeda et al. 2003; Koppel et al. 2004). M.tb ManLAMs contain more tri-mannoside (i.e., t-Manp-α-(1→2)-Manp-α-(1→2)-Manp-β-(1→5)-Araf-) and di-mannoside (i.e., t-Manp-α-(1→2)-Manp-β-(1→5)-Araf-) caps than mono-mannoside (i.e., t-Manp-β-(1→5)-Araf) caps (Nigou et al. 1997, 2003) while M. marinum and M. avium ManLAMs are characterized by mono-mannoside caps (Khoo et al. 2001; Pitarque et al. 2005). There were no differences in the pattern observed of the NCRDs binding to the different M.tb strains; however, there was less binding to M. marinum and no appreciable binding to M. avium (Figure (Figure4).4). High-mannose carbohydrate array results (Figure (Figure7B)7B) for rNCRD, R343V, and the hNCRD show greater binding of the R343V mutant to the (t-Manp-α(1→2)-Manp-α(1→2)-Manp-) linear tri-mannoside (approximates the terminal mannosyl residues of the tri-mannoside ManLAM cap) than hNCRD. Taken together, these results allow us to speculate that the enhanced binding of the R343V NCRD to M.tb results in part from a binding preference for the structure and presentation of the tri-mannoside cap of ManLAM.
Binding of the NCRD mutants to ManLAM by ELISA showed a relative decrease in R343V binding compared to whole bacteria, and R343V binding to LM was enhanced in a similar type of assay. This allows us to conjecture that LM exposed on the surface of the bacteria is involved in M.tb recognition by R343V. LM binding was also corroborated by results from the microarray showing enhanced binding of R343V to the LM-like branched hexa-mannoside and linear tri- and tetra-mannoside (Figure (Figure7A7A and B, compounds 6, 10 and 11). In the case of R343I, the binding to ManLAM was increased compared to hNCRD although the comparative increase in LM binding was much smaller (supplementary Figures 1 and 2). This suggests that, in contrast to R343V, R343I preferentially recognizes ManLAM over LM and emphasizes the importance of the characteristics of the 343 position in carbohydrate binding selectivity.
Evaluation of NCRD mutant binding to PIMs (Figures (Figures55 and and6)6) showed different profiles of binding, with rNCRD, R343I, and R343K binding to both the PIM2f (t-Manp-α-(1→6)-myo-inositol or t-Manp-α-(1→2)-myo-inositol) and the PIM5f (i.e. t-Manp-α(1→2)-[Manp-α-(1→6)]2-Manp- or t-Manp-α-(1→2)-myo-inositol) and R343V binding preferentially to the PIM5f (Figures (Figures55 and and6).6). Interestingly, despite the structural similarity of the terminal α-(1→2)-linked mannosyl units between PIM6f and the di-mannoside-capped ManLAM (Figure (Figure2)2) (Torrelles et al. 2006), none of the mutant NCRDs tested bound to the PIM6f, i.e. (t-Manp-α(1→2)-Manp-α(1→2)-[Manp-α-(1→6)]2-Manp or t-Manp-α-(1→2)-myo-inositol). This may be explained by differences in the structure, spatial conformation and size between the two molecules or to variable presentation of the molecules in the context of the ELISA or TLC plates. However, it does not depend upon the acylation state of the PIM which is in contrast to the binding observed for the MR, where the degree of PIM acylation dictates the ideal carbohydrate spatial conformation for binding to this receptor (i.e., Ac1PIM6 versus Ac2PIM6, where only the former binds to the MR) (Torrelles et al. 2006). In this regard, another C-type lectin, DC-SIGN, binds to all PIM species irrespective of their acylation state (Torrelles et al. 2006). This may be a reflection of similar clustering of the CRDs of DC-SIGN and SP-D, both of which multimerize in the fully functional form (Mitchell et al. 2001; McGreal et al. 2005) compared to the CRDs of the MR which are arranged in sequence on a single polypeptide, and high binding is dependent upon the engagement of several CRDs (East and Isacke 2002; McGreal et al. 2005).
We ascertained that the strong binding of R343K to the PIM2f was partially due to traces of PI that were undetectable on a control 2D TLC but seen by mass spectrometry in this sample. PI was shown to react with R343K, R343I (see supplementary Figure 2B), and rNCRD in our immuno-lectin blots (Figure (Figure6),6), in agreement with previous studies where R343K and rNCRD have been shown to bind strongly to PI (Crouch et al. 2007). These results suggest that the binding of NCRDs to mycobacteria could potentially be mediated by PI. However, there was minimal binding of R343K to whole mycobacteria (supplementary Figure 1) or to the PI-caps of PILAM from M. smegmatis suggesting that PI is either not exposed on the mycobacterial surface (obstructed by other cell envelope surface components) and/or is in an unfavorable conformation for binding to PILAM. In this context, it has been shown that the addition of a phosphate on C1 of myo-inositol to create phospho-myo-inositol leads to an interaction between the C4 and C5 of the inositol ring and the CRD (Crouch et al. 2007). This becomes particularly important when evaluating the binding of CRDs to PIMs, which have α(1→2) and α(1→6)-linked mannosyl units at C3 and C5 of PI, respectively (potential CRD binding sites) and are likely to obstruct binding interactions (Figure (Figure2).2). For PILAM, the lack of binding may also be explained by the spatial conformation adopted by its PI caps (Figure (Figure2).2). Specifically, the proximity of long arabinan chains may sterically interfere with SP-D binding or lead to an unfavorable binding conformation. Also, the low level of PI capping in M. smegmatis PILAM (Khoo et al. 1995a) may allow for only a few SP-D molecules to bind which are potentially below the sensitivity of the assay.
In addition to the 343 binding site, studies have shown that a three-amino-acid (RAK) insertion can enhance binding of SP-D to high-mannose containing pathogens (Crouch et al. 2005). This mutation creates an amino acid sequence that mimics the bovine serum protein CL-43, a protein with particularly high affinity for mannan (Rothmann et al. 1997; Hartshorn et al. 2002). Our results with mutants containing the RAK insertion showed that RAK+R343I (both mutations mimic CL-43) NCRD increases binding to virulent M.tb Erdman and H37Rv strains and the attenuated M.tb H37Ra strain as well as to each strain's ManLAM. In addition, RAK+R343I showed increased binding to LM. These results suggest that binding of RAK+R343I to virulent M.tb strains may be due to enhanced affinity for mannosylated ligands which are highly prevalent on the surface of M.tb compared to other mycobacterial species (Vercellone et al. 1998). Interestingly, the increased binding seen with RAK+R343I is not seen with the RAK+R343V mutant. This confirms earlier studies illustrating the importance of the 343 position for SP-D binding.
Together, our data demonstrate that binding of SP-D to mannosylated molecules on the M.tb surface is determined by the specific linkage pattern of terminal mannosyl units of the mycobacterial lipoglycans with a preference for trimannoside caps of ManLAM and the α(1→2) branches of LM. Additionally, this binding to mycobacteria can be manipulated through specific site-directed mutations in the CRD of SP-D that modify the characteristics of the side chain in position 343 (summarized in supplementary Table II). As SP-D has been shown to be a potentially important host defense molecule against M.tb infection, the enhanced interaction between SP-D mutants with substitutions in key amino acids in position 343 of the CRD and M.tb may be useful in further increasing the anti-mycobacterial effects seen with this collectin. Such SP-D mutants may prove useful as a targeted immunotherapy for pulmonary tuberculosis.
All chemical reagents were the highest grade from Sigma (St. Louis, MO) unless otherwise specified. Endotoxin-free water was used for all chemical reactions (Hospira, Lake Forest, IL). M.tb Erdman #35801, M.tb H37Rv #25618, M.tb H37Ra #25177, M. smegmatis #700084, and M. marinum #TMC 1218 were obtained from American Type Culture Collection. M. avium 104 was kindly provided by Dr. Andrea Cooper, Trudeau Institute, Saranac Lake, NY. Mycobacterial lipoglycoconjugates (i.e., ManLAM, PILAM, M.tb H37Rv LM, and M.tb H37Rv PIM families) were purified and characterized from M.tb H37Rv, M.tb H37Ra, M.tb Erdman, and M. smegmatis as previously described (Torrelles et al. 2006). Endotoxin levels were <18 pg/mg/sample.
Mycobacterial strains were plated on 7H11 agar (Difco, Detroit, MI) enriched with 5% oleic acid-albumin-dextrose-catalase (OADC) at 37°C/5%CO2 and grown for 9–14 days, except for M. smegmatis, which was grown for 2 days under the same conditions. Bacterial single cell suspensions were prepared as previously described (Ferguson et al. 1999). Mycobacterial lipoglycans and glycolipids were extracted, purified, and identified as previously described (Torrelles et al. 2004, 2006). In brief, purified ManLAM and LM were analyzed by SDS–PAGE, sugar analysis, fatty acid content, and 1H-nuclear magnetic resonance prior to detailed analyses. PIMs were purified from total lipid extracts as previously described (Torrelles et al. 2006). Briefly, total lipid extracts were precipitated in cold acetone for 24 h at −20°C, resulting in a pellet containing a mixture of phospholipids. PIMs were separated by silica column chromatography using a gradient from 100% chloroform to 100% methanol and eluted in the 60% methanol in chloroform fraction. PIM families and species were further purified by preparative TLC and purified components were confirmed by 2D TLC and mass spectrometry as we have previously described (McCarthy et al. 2005; Torrelles et al. 2006).
NCRD mutants (R343K, R343V, R343A, R343I, hNCRD, and rNCRD) have been described (Crouch et al. 2006, 2009). Unpublished and known mutants were generated (Crouch et al. 2005) using a bacterial expression system to produce N-terminally tagged, mutant trimeric human NCRD fusion proteins as previously described (Crouch et al. 2005, 2006). Trimeric fusion proteins were isolated by sequential chelation affinity and gel filtration chromatography. In all cases, site-directed mutants were sequenced for verification and the purified proteins were examined by SDS–PAGE and protein staining. The N-terminus was tagged with an S-protein binding sequence and His-tag, which were used for detection in binding assays. These tags are remote from the binding sites on the CRD and previous studies have shown that they allow for equivalent detection of various wild-type and mutant trimeric NCRDs (Crouch et al. 2005) Lectin activity was verified in mannan binding assays, which employ S-protein HRP conjugates for detection (Crouch et al. 2005). For microarray studies, integrity of the N-terminal His-tags was confirmed by immunoblotting using equivalent amounts of the fusion protein and the anti-His Ab.
Binding of SP-D NCRD mutants to mycobacteria and/or their cell envelope components was performed by ELISA (Schlesinger et al. 1990). Bacterial single cell suspensions (5 × 105) in 50 μL TBS (50 mM Tris-hydrochloride + 150 mM sodium chloride, pH 7.5) were added and dried onto triplicate wells of a 96-well microtiter plate (Immulon 1; Thermo Electron Corporation, Milford, MA) followed by overnight UV irradiation treatment to kill the bacteria. To confirm that NCRD binding was independent of the UV exposure, ELISA experiments were also performed without UV treatment with no significant differences seen in the results (data not shown). Alternatively, 50 μL containing 0.1 μg/μL of purified mycobacterial cell wall components were dried in ethanol onto duplicate or triplicate wells of a 96-well microtiter plate (EIA/RIA, CoStar, Corning, NY). All plates were blocked with the blocking buffer (3% fatty acid free BSA, 20 mM Tris, 140 mM sodium chloride, 5 mM calcium chloride, pH 7.4) for 1 h at room temperature (RT) under humid conditions. Wells were then incubated with SP-D NCRD mutants at 5 μg/mL in the blocking buffer for 1 h at RT either in the presence or absence of maltose (100 mM) as a binding competitor for the CRD of the NCRD mutants (Ferguson et al. 2002). Plates were washed three times with the washing buffer (20 mM Tris-HCl, 140 mM sodium chloride, 5 mM calcium chloride, 0.05% Tween 20, pH 7.4) prior to incubation with an S-protein HRP conjugate in the blocking buffer (1:5000 dilution) for 1 h at RT. ABTS/HRP (Bio-Rad, Hercules, CA) was used as a substrate following the manufacturer's instructions. Reactions were stopped with 2% oxalic acid and read on an ELISA reader (SPECTRAmax M5; Molecular Devices, Sunnydale, CA) at a wavelength of 405 nm.
Total lipid from M.tb H37Rv was extracted by using mixtures of chloroform:methanol (2:1 and 1:2, v/v) and chloroform:methanol:water (10:10:3, v/v/v) at 37°C for 24 h. Phospholipids were purified by cold-acetone precipitation as previously described (Besra 1998). Up to 300 μg of acetone insoluble lipids containing mainly PIM mixtures were spotted on the origin of 10 cm × 10 cm silica gel 60 aluminum-based TLC plates (Merck, Dardstadt, Germany) and resolved using a solvent system of chloroform:methanol:water (60:30:6, v/v/v) for the first dimension and chloroform:aceticacid:methanol:water (40:25:3:6, v/v/v/v) for the second dimension. Plates were then dried at RT prior to treatment with 0.1% polyisobutylmethacrylate in acetone for 90 s. TLCs were then blocked using 5% BSA blocking buffer for 3–5 h at RT and washed gently with PBS. Plates were incubated overnight at 4°C with 5 μg/mL NCRD mutants in the 1% BSA blocking buffer using 65 μL/cm2 of TLC plate surface. Plates were washed with PBS and incubated with 6.5 mL of a 1:5000 solution of S-protein HRP conjugate in the 1% BSA blocking buffer for 1 h at RT. TLC plates were thoroughly washed with PBS and incubated with 6.5 mL of ECL-HRP detection kit (Amersham, Buckinghamshire, UK) for 1 min at RT. Amersham hyperfilm ECL high-performance chemiluminescence film (GE Healthcare Systems, Piscataway, NJ) was then exposed to the immunostaining TLCs and developed. For analysis of the binding of NCRDs to phosphatidyl-myo-inositol (PI), some blots were performed with the addition of 50 μg of commercial PI (Sigma, St. Louis, MO) added to the origin of the TLC prior to resolving the plates.
GAPS II slides (Corning, Lowell, MA) were submerged in dimethylformamide containing 2% diisopropylethylamine and 0.5 mg/mL maleimido-N-hydroxysuccinimide hexanoic acid for 14 h at RT. Functionalized slides were washed three times with methanol, dried, and stored under an argon atmosphere. The high-mannose containing oligosaccharides were synthesized as previously described (Ratner et al. 2002). Compounds were dissolved in PBS with one molar equivalent TCEP (tris(2-carboxyethyl)phosphine) to reduce disulfides. One nanoliter of the solution was printed onto each spot of the maleimide-functionalized slides using an automated printing robot. In this context, the ligands are reproducibly oriented in relation to the solution phase, with the reducing end attached to the slide. Slides were incubated in a humid chamber to complete reaction for 24 h and stored in a dessicator until usage.
Microarray slides were processed as previously described (Ratner et al. 2004). Slides were washed with water and methanol, and quenched in 0.1% (v/v) β-mercaptoethanol in PBS for 1 h at RT. Slides were blocked with 2.5% (w/v) BSA in 20 mM HEPES, pH 7.5, with 100 mM sodium chloride and 20 mM calcium chloride for 1 h at RT, washed twice with the buffer, and centrifuged for 5 min at 50 × g at RT. Blocked slides were incubated with NCRD solutions (20 μg/mL of NCRD in binding buffer: 20 mM HEPES, pH 7.5, with 100 mM sodium chloride, 20 mM calcium chloride, 1% (w/v) BSA and 0.5% (v/v) Tween-20) for 1 h at RT. Quality of the microarray was confirmed in parallel assays using Concanavalin A (10 μg/mL in binding buffer). For inhibition, 50 μg/mL yeast mannan (Sigma), 125 mM maltose, and 20 mM EDTA were added to the binding buffer lacking calcium chloride. Incubated slides were washed twice with the buffer, centrifuged, and overlaid with 2 μg/mL PENTA-HIS Alexa Fluor 555 antibody conjugate (Qiagen, Valencia, CA) in the binding buffer in a similar fashion. After two washes and centrifugation, slides were scanned with a fluorescent microarray scanner (Tecan, Mannedorf, Switzerland). Binding intensities were analyzed using microarray evaluation software (Genespotter, Microdiscovery San Diego, CA).
Statistical analyses were performed with GraphPad Prism 4.01 (www.graphpad.com) using an unpaired, one-tailed Student's t-test.
National Institutes of Health (grant number AI059639 to L.S.S., RR007073-05 to T.K.C., HL-44015, HL-29594 to E.C.C.) and the ETH Zurich and Swiss National Science Foundation (T.H., P.H.S.).
Supplementary data for this article is available online at http://glycob.oxfordjournals.org/.