acetylated low-density lipoprotein
bio-gel-elicited peritoneal macrophage
bone marrow culture-derived macrophage
macrophage receptor with collagenous domain
- Poly C:
polycytidilic acid , Poly I: polyinosinic acid
resident peritoneal macrophage
rhodamine green X
The scavenger receptor-A I/II (SR-A) and macrophage receptor with collagenous domain (MARCO) share a common domain organisation and ligand repertoire, including selected polyanions and gram-positive and -negative organisms, but differ in fine specificity of ligand binding, tissue distribution and regulation. Neisseria meningitidis (NM) is a selective ligand for SR-A, but there is evidence for an additional SR-A-independent, polyanion-sensitive component for NM recognition. We therefore studied the relative contribution of MARCO and SR-A to binding of NM by resident and elicited peritoneal macrophages obtained from MARCO–/–, SR-A–/– and SR-A-MARCO–/– mice. Results confirmed that both mouse and human MARCO are able to bind NM independently of NM LPS. MARCO and SR-A contributed independently to NM binding, correlating with their expression levels in different cell populations, but neither of these two molecules was required for release of TNF-α and nitric oxide. We propose that the TLR-dependent induction of MARCO by innate immune stimulation enhances recognition and uptake of pathogenic organisms such as NM, thus contributing to host defence against infection.
Scavenger receptors (SR) are a family of structurally unrelated receptor proteins functionally defined by their ability to recognise modified low-density lipoprotein 1, 2. Although a large body of research has been directed to identify the role of SR in atherosclerosis, the description of SR as microbial pattern recognition receptors has only recently been appreciated. SR-A(I/II) 2, the first member of the class-A SR family to be cloned, is a trimeric myeloid-restricted cell surface glycoprotein, able to recognise multiple ligands and perform varied cellular functions. Macrophage receptor with collagenous domain (MARCO), another class-A SR, but a distinct gene product, shares partially common domain structures, but contains a longer collagenous domain and lacks the coiled coil domain of the classical SR-A molecule 3. Similarly, SR-A and MARCO share a range of overlapping, but distinct polyanionic ligands, including microbial products 4.
Hampton and colleagues 5 first reported that SR-A could bind the lipid A moiety of lipopolysaccharide (LPS), an integral component of gram-negative cell walls. Subsequently, a large number of investigators confirmed that SR-A could bind intact gram-positive and gram-negative organisms or their isolated components 6–8.
In contrast to SR-A, which is expressed by almost all macrophage (MΦ) populations, MARCO expression is restricted to subpopulations of MΦ, for example spleen marginal zone and resident peritoneal MΦ (RPM) 3. MARCO can be readily induced by a variety of infectious (BCG, bacterial infection, sepsis, LPS) 9, 10 and non-infectious inflammatory stimuli (joint inflammation, murine atherosclerotic plaques) 11, 12 in most tissue MΦ, both in vivo and in vitro. However, a variety of pro- and anti-inflammatory cytokines showed no effect on MARCO induction 10. By contrast, SR-A, although induced by selected cytokines, such as MΦ colony-stimulating factor, remains mostly unaltered by bacterial stimuli 13–15. Comparative phenotypic analysis of MΦ activation by the microbial or cytokine stimuli Neisseria meningitidis (NM) and interferon gamma (IFN)-γ confirmed that MARCO is a selective marker for innately activated MΦ, whose induction depends on Toll-like receptor (TLR)-4, but not on SR-A 15.
NM, a gram-negative diplococcus, is an important cause of bacterial meningitis and septic shock in humans. The organism and its cell wall derivatives are potent immune adjuvants 16, able to stimulate pro-inflammatory cytokine secretion and Ab production, and to activate complement 17. Generally, the humoral system plays a major role in anti-neisserial immunity, but non-opsonic phagocytosis in the opsonin-poor nasopharynx and meninges, two entry points for NM, may play an important role in local resistance to infection 18, 19.
We have previously shown that non-opsonic phagocytosis of NM was mediated almost exclusively via SR-A in bone marrow culture-derived MΦ (BMDM) 20, which lack MARCO and express high levels of SR-A. The contribution of SR-A to NM binding is much less in bio-gel-elicited peritoneal MΦ (Bg-PM); however, SR-A-independent residual binding is still sensitive to general polyanionic SR inhibitors and can be up-regulated by LPS pre-treatment of the cells. Furthermore, NM readily induce MARCO in MΦ during innate activation 15.
To test the hypothesis that MARCO contributes to SR function of peritoneal MΦ, we set out to investigate (1) the ability of MARCO to recognise NM, and (2) the relative contribution of MARCO and SR-A in NM recognition. We utilised SR-A–/–, MARCO–/– or doubly deficient SR-A-MARCO–/– mice, and a range of specific reagents to distinguish between SR-A and MARCO. Selection of model organisms and microbial mutants allowed us to compare ligand recognition by these two SR molecules. We show that MARCO and SR-A are both able to contribute to binding of NM, correlating with their level of expression by different MΦ populations. However, their role in innate microbial resistance can be distinguished by differential dependence on prior engagement of TLR.
MARCO binds NM in a cell-free system
It has been shown before that recombinant soluble mouse MARCO (sMARCO) can bind selected bacteria in a cell-free system 21. Therefore, to investigate whether sMARCO could bind NM in a cell-free system, glass coverslips were coated with sMARCO or nephrin, an unrelated recombinant control protein, and incubated at 37°C with rhodamine green X (RdGnX)-labelled NM (RdGnX-NM) or fluorescently labelled Escherichia coli bioparticles, in the absence of serum. Some coverslips were pre-incubated with the general SR inhibitor polyinosinic acid (Poly I) or a cognate non-ligand, polycytidylic acid (Poly C), which was retained throughout the experiment. Our results confirmed that sMARCO bound both organisms (Fig. 1), whereas the control protein showed no binding of either bacterium. Poly I, but not Poly C, blocked the binding of both organisms, indicating that such binding was polyanion-dependent.
Cells expressing mouse/human MARCO bind NM
To see whether MARCO expressed on the cell surface could recognise NM, CHO cells were transiently transfected with either full-length mouse or human MARCO cDNA or with an empty vector on 10-cm plate. After 24 h, cells were detached and approximately 105 cells were seeded on glass coverslips and cultured overnight. Fluorescently labelled RdGnX-NM were added in the absence of serum and incubated at 37°C for 90 min. In some cases, cells were pre-incubated with Poly I and Poly C. Another set of coverslips was incubated with fluorescent E. coli bioparticles in identical conditions, as positive control (data not shown). Finally, we stained the cells with anti-mouse MARCO specific mAb ED31 and anti-human polyclonal Ab 2082, respectively, to distinguish mouse or human MARCO-expressing cells from untransfected cells.
Our results confirmed that both mouse and human MARCO-expressing CHO cells, but not the untransfected cells, bound both NM and E. coli (Fig. 2). While Poly I almost completely inhibited both RdGnX-NM and E. coli binding by both species of MARCO, Poly C had no effect on binding of either bacterium (data not shown). Using a FACS-based quantitative assay, we confirmed that cell lines stably transfected with full-length mouse MARCO also show significantly increased binding of RdGNX-NM and E. coli bioparticles compared to control cells (data not shown). Confocal microscopic analysis of the above experiments revealed that a subset of the MARCO-expressing CHO cells do not only bind RdGnX-NM, but also ingest the bacteria, confirming the possible role of MARCO in phagocytosis (Fig. 2, top insert).
Mouse, human and soluble MARCO recognise NM independent of LPS
Lipid A is thought to be a ligand for SR-A 5 and recently LPS has also been reported to be a ligand for MARCO 21. We have previously shown that although lipid A is a ligand for SR-A, it is not required for SR-A-mediated binding of NM, indicating the presence of ligand(s) other than lipid A 20. In the present study, we wanted to test the role of lipid A in MARCO-mediated binding of NM. Therefore, we studied the binding of WT or an lpxA mutant NM strain (which lacks lipid A) by mouse or human MARCO-transfected CHO cells, or by sMARCO in the cell-free system described above. Our results revealed that surface-expressed mouse and human MARCO, as well as sMARCO, all bound both strains of NM, indicating that, similarly to SR-A, MARCO does not require lipid A to bind NM (Fig. 3). As before, Poly I, but not Poly C, blocked the binding of both strains in all three systems studied (not shown), confirming that binding of non-lipid A ligand is also sensitive to known SR polyanion inhibitors.
Resident and Bg-PM of SR-A–/–, MARCO–/– and SR-A-MARCO–/– animals show reduced binding of NM, E. coli and acetylated low-density lipoprotein
We have previously reported that Bg-PM from SR-A–/– mice show a decreased ability to take up NM, but not as marked a difference as in BMDM. Residual uptake by SR-A–/– MΦ was selectively sensitive to polyanion inhibition, suggesting that (an) other SR such as MARCO contribute to NM uptake. MARCO–/– and doubly deficient SR-A-MARCO–/– mice have recently been generated, making it possible to test this hypothesis. To establish that MARCO–/– MΦ indeed do not express MARCO, we stained resident peritoneal cells to be used for further experiments, and splenic tissue sections of WT and MARCO–/– animals with anti mouse mAb ED31, since RPM and spleen marginal zone MΦ express MARCO constitutively 3. As expected, unlike the WT cells, MARCO–/– cells did not label with MARCO mAb (not shown). Further control experiments with mAb ED31 staining established that Bg-PM from WT animals express very low levels of MARCO, but RPM expressed significantly higher levels of MARCO antigen than Bg-PM (not shown).
To study whether the differential expression of MARCO in Bg-PM and RPM has functional consequences and determine the relative contribution of SR-A and MARCO in ligand recognition by elicited and resident cells, we harvested either Bg-PM (Fig. 4A) or RPM (Fig. 4B) from WT, SR-A–/–, MARCO–/– or SR-A-MARCO–/– animals. Adherent MΦ were incubated with RdGnX-NM, E. coli bioparticles or acetylated low-density lipoprotein (Ac-LDL), respectively, in the presence and absence of Poly I or Poly C. Ac-LDL is a prototypic SR ligand, and previously MARCO has been shown to bind E. coli. Therefore in our study, Ac-LDL and E. coli have served as a measure of general SR function and control for MARCO-mediated bacterial binding, respectively.
Our results confirmed that both Bg-PM and RPM from SR-A–/–, but not MARCO–/– animals, showed significantly reduced binding of E. coli, Ac-LDL and NM compared to WT cells. However, SR-A-MARCO–/– MΦ from either source showed a relatively greater reduction in binding of E. coli, NM or Ac-LDL compared to SR-A–/– MΦ. Furthermore, we found that the contribution of MARCO in Ac-LDL and bacteria binding in RPM is higher than Bg-PM, which correlated with the differential expression of MARCO in these two MΦ populations (not shown).
SR-A and MARCO do not contribute to the secretion of TNF-α or nitric oxide
Previously, using BMDM and Bg-PM, we reported that phagocytosis of NM was dissociated from NM-induced release of mediators, where phagocytosis depended on SR-A, but TNF-α and NO secretion depended on TLR-4 and not SR-A 20. To extend this observation to MARCO, we incubated Bg-PM or RPM from SR-A–/–, MARCO–/–, SR-A-MARCO–/– and the corresponding WT strain with different concentrations of NM in the presence or absence of IFN-γ (Fig. 5). After 24 h of incubation, the supernatant was harvested and TNF-α and NO release measured by cytokine ELISA or the Greiss reaction.
Our results confirmed that IFN-γ significantly (p⩽0.05) augmented the release of both mediators in WT and all KO strains in every NM concentration studied, irrespective of the presence or absence of SR-A or MARCO. However, in individual experiments we observed some variability in TNF-α/NO release among the strains but these are not consistent over different concentrations and no statistically significant difference was observed between the strains in different experiments.
In order to compare the roles of MARCO and SR-A in recognition of NM, we utilised in vitro models and primary MΦ populations from WT and KO strains, with receptor-specific reagents. Our main results showed that both mouse and human MARCO avidly bound NM in cell-free conditions and in transfected cells, and that lipid A was not required for binding. In primary MΦ, both MARCO and SR-A contributed to NM and E. coli binding, but their relative contribution in different peritoneal MΦ populations varied with the level of MARCO expression.
Apart from particulate bacteria, we utilised soluble Ac-LDL to compare the polyanionic ligand specificity and endocytic capacity of these two receptors. The role of SR-A in modified LDL uptake and atherosclerosis is well established. The contribution of MARCO is less well understood, although MARCO is expressed in mouse atherosclerotic plaques and has been reported to bind Ac-LDL 10, 12. Our results shed further light on possible involvement of MARCO in the uptake of polyanionic ligands. The relatively substantial reduction in uptake of NM and E. coli and also of Ac-LDL by SR-A-MARCO–/– MΦ compared with SR-A–/– and MARCO–/– MΦ alone, indicate shared or overlapping ligand binding as well as non-redundancy between these two receptors.
The model we have characterized is efficient and reproducible. Use of Bg-PM enables access to defined murine strains, with high yields of elicited MΦ, which are readily purified by adhesion. Bg-PM are more representative of MΦ recruited by inflammatory stimuli in vivo than BMDM, which express high levels of SR-A as a result of CSF-1 present in the L cell-conditioned growth medium 13. Comparison of NM, E. coli and Ac-LDL uptake between Bg-PM and RPM confirmed that higher expression of MARCO correlated with enhanced binding. The use of bacteriologic plastic (BP) vessels and serum-free media avoided ligation of SR-A, which promotes divalent cation-independent adhesion to tissue-culture plastic 22 and facilitated detachment for further analysis by flow cytometry.
In our earlier studies we showed that SR-A plays a comparable role in uptake of live and fixed NM. Ethanol fixation preserves key ligand functions for SR-A and TLR and limits the surface variation in the organism 20. Lipid A was previously thought to be a ligand for both SR-A and MARCO 5, 21, but our present and previous studies have established that LPS is not required for binding of NM by either receptor. It would be interesting to determine whether SR-A and MARCO recognise the same or different ligand(s) on the NM surface and to establish their conservation among a broader range of microorganisms.
The results reported here did not distinguish between binding of NM and their uptake. We have not extended the comparison of SR-A and MARCO expression to other functions such as adhesion or antigen presentation. However, experiments with confocal microscopy of transfected CHO cells revealed that MARCO does not only bind NM, but also internalises the organism. CHO cells are not professional phagocytes and lack the intracellular signalling molecules necessary for efficient phagocytosis. Further studies in professional phagocytes will elucidate the true phagocytic potential of MARCO and its signalling requirement in this process.
SR-A I/II plays a crucial role in host defense against listerial and staphylococcal infections in vivo23, 24. Using an in vivo model, Haworth et al.25 showed that BCG-primed SR-A–/– animals developed normal granulomata, but were highly susceptible to subsequent LPS challenge and septic shock compared with WT animals, associated with enhanced levels of circulating TNF-α, confirming another host-protective role for SR-A in endotoxic shock25. Recently, a role has also been proposed for MARCO in host defence, as MARCO–/– animals displayed an impaired ability to clear bacteria from the lungs, resulting in increased inflammation and diminished survival 26. Peiser et al.20 showed that SR-A-mediated uptake of NM in vitro can be dissociated from subsequent TNF-α and NO secretion. The present study extended this observation to MARCO, confirming that neither SR-A nor MARCO is required for NM-mediated TNF-α or NO induction in vitro, either alone or in combination.
We have previously reported that NM induced MARCO via an SR-A-independent, but TLR-4-mediated pathway 15. Doyle et al.27 showed that agonists for other TLR family members also induced MARCO. They showed further that TLR-mediated MARCO induction depends on the intracellular adapter protein myeloid differentiation factor 88 and signalling molecule P38 MAPK.
In the present study we analysed the binding and uptake of ligands by constitutively or over-expressed MARCO. Surprisingly, we were unable to induce MARCO by NM or LPS treatment in Bg-PM or RPM of C57BL/6 strain (not shown), although other TLR-4-mediated responses (TNF-α and NO release) remained intact (not shown). However, we could induce MARCO in the same cell populations of other strains (ICR/129, BALB/c) 15. Unlike the inability of the mAb 2F8 to detect the C57BL/6 mutant SR-A molecule efficiently 28, ED31 recognised the constitutively expressed MARCO in this strain normally, indicating that there is no obvious defect in MARCO protein in the C57BL/6 strain.
Consistent with possible strain differences in MARCO induction by NM, Su et al. 29 showed that severe exercise can induce MARCO expression and enhance phagocytosis in broncho-alveolar MΦ of BALB/c, but not C57BL/6 mice. Inducibility of MARCO is well documented in other strains of mice or many other species. The use of C57BL/6 as background strain for SR-A and MARCO phenotypic studies may introduce difficulties in analysis of functions in WT animals. The mouse does not provide a natural model for NM infection in vivo unlike humans. Further studies are needed with a range of different microorganisms to establish whether the in vitro data reported here can be extended to the living host.
Materials and methods
1,1′-dioctadecyl-1-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI)-labelled Ac-LDL (DiI-Ac-LDL) was obtained from Intracell (Rockville, MD), RdGnX-NM and FITC-labelled E. coli bioparticles from Molecular Probes (Eugene, OR). ELISA kits for murine TNF-α were obtained from Pharmingen (San Diego, CA). OPTIMEM-1 culture media were obtained from Gibco (Paisley, UK). Unless stated otherwise, all reagents were from Sigma (Poole, UK). BP products were obtained from Greiner (Gloucester, UK) and other plastic products from Becton Dickinson Labware (Oxford, UK).
We used the following KO mouse strains; SR-A (SR-A–/–), MARCO (MARCO–/–), SR-A-MARCO double-KO (SR-A-MARCO–/–) and their WT control C57BL/6 strain. All animals were bred and housed under specific pathogen-free conditions, according to local ethical guidelines. MARCO–/– animals were developed and bred onto C57BL/6 background using standard molecular biology techniques and will be reported in detail elsewhere. SR-A-MARCO–/– animals were generated by mating SR-A–/– and MARCO–/– animals.
MΦ isolation and culture
Bg-PM were prepared by i.p. injection of 1 mL polyacrylamide gel P-100 (Bio-Rad) beads (2% w/v in endotoxin-free water). After 4 days, peritoneal cells were harvested by lavage with phosphate-buffered saline (PBS). RPM were isolated from uninjected animals by lavage with PBS. Both Bg-PM and RPM were plated on BP dishes in a defined serum-free medium, OPTIMEM, supplemented with 50 IU/mL penicillin-streptomycin and 2 mM L-glutamine. After 3–4 h, adherent cells were washed three times to remove non-adherent cells and bio-gel beads. After washing, purity of MΦ in adherent monolayer was >95%, as characterized previously 30, 31.
Bacterial culture and fluorescent labelling
The capsulated B serogroup strain of NM, MC58, 44/76 32 and lipid A-mutant strain 44/76lpxA33 were cultured as described before. For fluorescent labelling, NM were harvested and resuspended in 70% ethanol overnight at 4°C and labelled with RdGnX (RdGnX-NM), according to the manufacturer's protocol. Briefly, ethanol was removed by washing with PBS. Bacteria were resuspended in 0.2 M sodium bicarbonate; 100 μL of RdGnX (dissolved in DMSO) were added per 1010 bacteria and the mixture incubated at room temperature for 1–1.5 h; finally, 400 μL of hydroxylamine hydrochloride were added per 1010 bacteria and the organisms incubated for a further 1 h to stop the reaction. Bacteria were washed with ice-cold saline, resuspended in PBS and stored at –80°C.
Cell-free bacteria-binding assay
Thirteen-mm glass coverslips were coated with 5 µg of sMARCO 21 or a control protein, recombinant nephrin, prepared as described before 34, for 1 h at room temperature, or overnight at 4°C. An additional coverslip was incubated with PBS only. Coverslips were then incubated with 1 mg/mL bovine serum albumin in PBS for 30 min at room temperature and washed three times with 20 mM Tris-HCl, pH 7.5, 0.1 M NaCl containing buffer. FITC-labelled E. coli bioparticles or RdGnX-NM were added in the absence of serum and incubated for 1 h at 37°C. Some coverslips were pre-incubated with 50 μg/mL Poly I or Poly C for 30 min and retained throughout the assay. After incubation, coverslips were washed three times with the same buffer, and bacterial binding analysed by fluorescent microscopy.
Quantitative analysis of bacterial association by flow cytometry
Bg-PM or RPM were obtained from different KO mice; cells (106) were plated in six-well BP dishes 24 h before assay, along with respective WT control cells. To examine the uptake of bacteria, cells were washed twice with PBS and incubated with OPTIMEM containing ethanol-fixed RdGnX-NM or FITC-labelled E. coli bioparticles (20 bacteria/cell), for 90 min at 37°C. Some wells were pre-incubated for 30 min with 50 μg/mL of Poly I (a general SR inhibitor) or a cognate non-ligand, Poly C, and retained throughout the assay. Endocytosis of DiI-Ac-LDL was routinely examined in all assays as a positive control to measure SR function, by incubating cells with 5 μg/mL DiI-Ac-LDL in the presence and absence of Poly I and Poly C, as described above. After incubation, all the cells were washed three times with PBS and detached from BP surfaces by means of PBS containing 5 mM EDTA and 4 mg/mL lidocaine before fixation with 4% w/v formaldehyde. Mean fluorescence was analysed by flow cytometry using Cell Quest software.
Analysis of bacterial binding by microscopy
Cells were transfected with either a full-length mouse or human MARCO cDNA or an empty vector by means of calcium phosphate or lipofectamine, using 20 µg of DNA per 100-mm dish. Precipitates were incubated with cells overnight, the cells seeded on glass coverslips 24 h after transfection and incubated with RdGnX-NM or E. coli bioparticles for 90 min at 37°C in the presence or absence of Poly I and Poly C. After incubation with bacteria, cells were washed three times with PBS and fixed in 4% paraformaldehyde. Cells were blocked with PBS containing 5% v/v goat serum, 5% v/v rabbit serum and 1% w/v bovine serum albumin, and stained with anti-mouse MARCO mAb ED31 or anti-human MARCO polyclonal Ab 2082, respectively, followed by the appropriate rhodamine-conjugated secondary Ab. Fluorescence microscopy was performed using a Zeiss Axiovert 25 CTL inverted microscope equipped with a 50-W mercury vapor lamp fitted with standard filter sets for viewing FITC and rhodamine fluorescence.
Measurement of TNF-α and NO release
MΦ were plated in 96-well BP culture dishes. The monolayers were washed three times with PBS to remove non-adherent cells and bio-gel beads, then incubated with different doses of NM in the presence or absence of 20 ng/mL IFN-γ for 24 h. E. coli LPS (1 μg/mL) was used as a positive control stimulus for both TNF-α and NO. N-monomethyl arginine was included as a control to block NO release. The tissue culture supernatants were harvested and centrifuged at 1300×g to remove particulate matter. TNF-α was measured by ELISA, according to the manufacturer's protocol, with recombinant murine TNF-α as standard. NO release was detected using the colorimetric Greiss reaction and absorbance measured at 550 nm, with sodium nitrite as standard 35.
All experiments were done in triplicate and results are representative of three independent experiments. The statistical significance of results was determined by Student's paired t-test and significance tested at the 95% confidence level (p⩽0.05).
We thank Prof. Richard Moxon and Dr. Katherine Makepeace, Department of Pediatrics, John Radcliffe Hospital, Oxford, for the kind gift of the MC58+ and lpxA mutant strains of NM. S.M. was supported by an Usher Cunnigham Postgraduate Studentship, Exeter College, University of Oxford. S.G.’s laboratory is supported by a grant from the Medical Research Council, UK.