Extracellular Matrix Lumican Promotes Bacterial Phagocytosis, and Lum−/− Mice Show Increased Pseudomonas aeruginosa Lung Infection Severity*

Background: Lumican, an extracellular matrix protein, promotes host response to LPS endotoxins. Results: Lum−/− mice show increased lung infection. Phagocytosis of nonopsonized bacteria is reduced in Lum−/− macrophages and restored by recombinant wild type lumican but not mutated rLumY20A. Conclusion: Lumican interaction with CD14 at an N-terminal tyrosine is important for bacterial phagocytosis. Significance: A broader function is implicated for lumican and other LRR ECM proteins in host pathogen interactions. Phagocytosis is central to bacterial clearance, but the exact mechanism is incompletely understood. Here, we show a novel and critical role for lumican, the connective tissue extracellular matrix small leucine-rich repeat proteoglycan, in CD14-mediated bacterial phagocytosis. In Psuedomonas aeruginosa lung infections, lumican-deficient (Lum−/−) mice failed to clear the bacterium from lungs, tissues, and showed a dramatic increase in mortality. In vitro, phagocytosis of nonopsonized Gram-negative Escherichia coli and P. aeruginosa was inhibited in Lum−/− peritoneal macrophages (MΦs). Lumican co-localized with CD14, CD18, and bacteria on Lum+/+ MΦ surfaces. Using two different P. aeruginosa strains that require host CD14 (808) or CD18/CR3 (P1) for phagocytosis, we showed that lumican has a larger role in CD14-mediated phagocytosis. Recombinant lumican (rLum) restored phagocytosis in Lum−/− MΦs. Surface plasmon resonance showed specific binding of rLum to CD14 (KA = 2.15 × 106 m−1), whereas rLumY20A, and not rLumY21A, where a tyrosine in each was replaced with an alanine, showed 60-fold decreased binding. The rLumY20A variant also failed to restore phagocytosis in Lum−/− MΦs, indicating Tyr-20 to be functionally important. Thus, in addition to a structural role in connective tissues, lumican has a major protective role in Gram-negative bacterial infections, a novel function for small leucine-rich repeat proteoglycans.

other pathogen-associated molecular patterns were not inhibited (15). Specific binding between lumican and CD14 was indicated by co-immunoprecipitation of lumican and CD14 from elicited peritoneal M⌽s and in pulldown assays using a mammalian expression system. In a solid phase binding assay, lumican bound LPS that could be competed out with soluble recombinant CD14. These in vitro findings, taken together with the hypo-responseness of Lum Ϫ/Ϫ mice to LPS septic shock, prompted us to hypothesize that bacterial infections may have a contrary negative effect in the absence of lumican.
Here, we show that Lum Ϫ/Ϫ mice are unable to clear Pseudomonas aeruginosa in a lung infection model with a marked increase in mortality and morbidity. Lumican associates with bacteria on the surface of primary peritoneal M⌽s and regulates phagocytosis mediated by CD14. We further characterized the lumican-CD14 binding by surface plasmon resonance; substitution of an N-terminal tyrosine (rLumY20A) in lumican led to its rapid dissociation from CD14. Using lumican and lumican-CD14 double-null mouse M⌽s, we determined that lumican regulates bacterial ingestion, whereas CD14 is needed for phagosome progression to late endosomes. This study provides mechanistic and molecular insights into how an ECM protein regulates CD14-dependent phagocytosis and clearance of Gram-negative bacteria in vivo.
Mice and Bacterial Strains-The Lum Ϫ/Ϫ mice generated earlier (11) were backcrossed to the C57BL/6J background for more than 12 generations. Lum Ϫ/Ϫ CD14 Ϫ/Ϫ double null mice were generated by breeding the Lum Ϫ/Ϫ with CD14 Ϫ/Ϫ / C57BL/6J mice (Jackson Laboratory). C57BL/6J WT and the knock-out mice were maintained in a specific pathogen-free mouse facility at the Johns Hopkins University and used according to the institutional animal care and use committee. Mice between the ages of 8 -9 weeks were used in all experiments, and the animals were gender-balanced within an experiment. P. aeruginosa (ATCC, 19660)  Peritoneal M⌽ Isolation and Culture-Peritoneal M⌽s were harvested from peritoneal exudates 4 days after an intraperitoneal injection of 1 ml of 3% thioglycolate. The cells were plated in RPMI 1640 (Invitrogen) containing 1% FBS. After overnight incubation at 37°C and 5% CO 2 , the attached M⌽s were washed three times with PBS, switched to serum-free RPMI, and used for all assays. By flow cytometry, more than 98% of the cells were F4/80 positive (data not shown).
Phagocytosis-M⌽s (2.0 ϫ 10 5 cells/well) were incubated with 6 ϫ 10 6 of pHrodo TM E. coli BioParticles for 1 h at 37°C, 5% CO 2 , and time lapse live cell images were acquired every 2 min, five frames of z-stacking with 2-m optic slice by Zeiss LSM 410 confocal microscopy fitted with 63ϫ/1.2 water objective at 37°C, 5% CO 2 . On average 150 -300 M⌽s were analyzed to determine mean fluorescence intensity (MFI) from pHrodo TM E. coli BioParticles using the Image J software, and the data were expressed as MFI/M⌽.
Bacterial association with the cell surface was determined using Alexa Fluor 488 E. coli BioParticles conjugates or GFP-P. aeruginosa PAO1, where M⌽s were first incubated at 37°C for 20 min with bacteria and then fixed in 4% paraformaldehyde. The plasma membrane was immunostained with an antibody against Na/K-ATPase, nuclei with DAPI, while lumican, and CD14 and CD18 localized by immunostaining as previously described (17), and images were captured with an oil objective of 40ϫ/1.3 or 100ϫ/1.4 in a Zeiss LSM 410 microscope.
To determine bound and ingested P. aeruginosa ATCC19660 strain, M⌽s from WT and Lum Ϫ/Ϫ mice were incubated with bacteria (13/M⌽) at 4°C and 37°C for 1 and 2 h. The M⌽s were centrifuged, washed in PBS (3ϫ), and lysed in water, and serial dilutions were plated on Pseudomonas selective Cetrimide Agar (Sigma-Aldrich) to obtain viable cfu.
Localization of Bacteria in Late Endosome/Lysosome-We incubated M⌽s with Alexa 488 E. coli for 30 min, and the lysosomal compartment was stained with 100 nM LysoTracker Red DND-99 for another 30 min at 37°C with 5% CO 2 in serum-free RPMI. Time lapse images were acquired every 3 min in the 30 -60-min period. Areas of overlaps (yellow) between bacteria (green) and lysosomes (red) were calculated using Image J for three Alexa-E. coli particles per experiment. In a second approach, M⌽s were incubated with Alexa 488 E. coli for 40, 60, and 90 min; fixed; and immunostained with anti-mouse CD107 (LAMP-1) antibody.
Bacterial Killing-M⌽s were incubated with P. aeruginosa ATCC19660 (13 bacteria/M⌽) in 1 ml of serum-free RPMI with gentle shaking for 1 and 2 h. The macrophages were harvested by centrifugation, washed three times in PBS, and resuspended in water to lyse the cells. Serial dilutions of the supernatant and lysed cells were plated for cell-associated viable cfu.
Surface Plasmon Resonance-Human recombinant CD14 (Abcam) was immobilized on a CM5 (carboxymethylated) sensorchip (GE Healthcare) by injecting 50 l of CD14 (25 g/ml in 10 mM sodium acetate, pH 4.5) at the rate of 10 l/min. Blank channels without CD14 served as controls. Between injections, the CD14 immobilized sensorchip was washed with 0.01% SDS and degassed phosphate-buffered saline. The rate constants of association and dissociation were derived at eight different injection concentrations of positive, negative control, and test analytes. The analytes were injected for 1.5 min at 25 l/min with a dissociation interval of 5 min. The differences in the reference and blank cell readings were used to determine specific binding. When anti-CD14 IgG was used as the analyte, the cells were regenerated by washing with 10 mM glycine, pH 2.0, for 30 s. The dissociation constant (K D ) was calculated from the sensorgram data by fitting a 1:1 Lang-muir model using the BIA evaluation 4.1 software (BIAcore; GE Healthcare).
Mouse Model of Lung Infection-Anesthetized mice (Avertin, 350 mg/kg of body weight) were infected by intranasal or intratracheal delivery of 10 8 cfu of P. aeruginosa (ATCC 19660). The mice were harvested 12 and 20 h after infection, and the bronchoalveolar lavage fluid was collected by tracheal intubation. Lung tissues for histology were fixed in 10% buffered formalin and paraffin-embedded for hematoxylin and Periodic-acid Schiff staining. Frozen lung tissue sections (5 m) were incubated overnight with Gr1-FITC diluted in 1% FBS/ phosphate-buffered saline with Tween 20, and nuclei counterstained with DAPI (eBioscience) and washed in phosphate-buffered saline with Tween 20 and mounted in fluorescent mounting media (Dako). To assess viable bacteria, lung tissues were weighed and homogenized with 5 l of PBS/g of tissue, and appropriate dilutions were plated on agar and incubated overnight at 37°C.
Statistical Analysis-The experiments were repeated two or three times as indicated in the figure legend. Statistical significance was calculated using unpaired Student's t test (Graphpad Prism), and p Յ 0.05 was considered to be significant.

RESULTS
Lumican Regulates Phagocytosis by M⌽s-Our previous studies indicated a role for lumican in CD14-TLR4 mediated innate immune response to LPS (15,17). To further investigate its implications in host response to bacteria, we determined the uptake of pHrodo TM E. coli BioParticles by elicited peritoneal M⌽s. The pHrodo TM dye increases in fluorescence in a pH-dependent manner as the BioParticles reach the acidic lysosomal compartments, and acidification of phagosomes can be measured as MFI per M⌽. We first compared phagocytosis of opsonized and nonopsonized E. coli in WT and Lum Ϫ/Ϫ M⌽s to distinguish between phagocytosis mediated by host receptors such as Fc␥ and complement receptors versus those that do not require opsonization of the pathogen. After 1 h, phagocytosis of serum-opsonized E. coli occurred unhampered, whereas that of nonopsonized E. coli was reduced by 52% in the Lum Ϫ/Ϫ M⌽s (Fig. 1A). Thus, lumican does not seem to modulate phagocytosis of serum-opsonized bacteria, and subsequent assays were carried out in serum-free medium. We examined phagosome acidification by time lapse imaging over 1 h, and bacterial uptake was evident in the WT but impaired in the Lum Ϫ/Ϫ M⌽s (Fig. 1B). On average, phagosome acidification after 60 min (Fig. 1C) was 43% lower in the Lum Ϫ/Ϫ compared with wild type M⌽s.
Although lumican contributed to M⌽ functions, we did not detect lumican transcript in M⌽s by quantitative RT-PCR (Table 2), confirming our previous observation (15). Therefore, we hypothesized that exogenous lumican is acquired by M⌽s for optimal phagocytosis. We also found other SLRP members, decorin and biglycan, and not fibromodulin, to be expressed by M⌽s (Table 2). We excluded intrinsic developmental differences as factors influencing phagocytosis, by flow cytometric analyses of cell surface markers in peritoneal exudate F4/80ϩ WT and Lum Ϫ/Ϫ M⌽s that showed similar differentiation markers and surface TLR2 (supplemental Fig. S1). However, there was a 2-fold decrease in TLR4 in the Lum Ϫ/Ϫ M⌽s, consistent with our previously identified link between lumican and TLR4 signaling (15,17). We next developed bone marrow cell chimeras (BMC), which by FACS typing of CD45 had greater than 85% donor contribution (data not shown). M⌽s from Lum Ϫ/Ϫ donor in WT recipient (Lum Ϫ/Ϫ 3 WT) BMC showed wild type-like phagocytic capabilities, whereas M⌽s from WT 3 Lum Ϫ/Ϫ BMC were phagocytosis deficient, like the parental Lum Ϫ/Ϫ mice (Fig. 1D). We also tested induction of TNF-␣ by LPS as a measure of innate immune response in the BMC M⌽s; Lum Ϫ/Ϫ 3 WT showed WT like response, whereas WT 3 Lum Ϫ/Ϫ BMC showed poor induction of TNF-␣ (Fig. 1E). Furthermore, in in vitro assays, exogenous   1F), whereas antibody-mediated blocking of lumican in WT M⌽s inhibited phagocytosis (Fig. 1G), supporting a role for lumican in phagocytosis. Immunofluorescent staining for lumican in WT M⌽s, without permeabilizing the cells, showed its presence on the surface (supplemental Fig. S2), corroborating our study and earlier studies of others showing lumican binding to macrophages (15,18). We also found that lumican is expressed in mouse lung tissues and in culture by HUVECs and fibroblasts (16,17,19), and others have shown its presence in the vascular endothelial ECM (20,21). Our earlier study showed the presence of lumican on mouse peritoneal neutrophils. Thus, the cell surface lumican on M⌽s is likely to be acquired from the ECM during their transport. Lumican Facilitates CD14-dependent Phagocytosis-Because our previous studies showed binding of lumican to CD14 and CD18 by immunoprecipitations (15,17), here we sought to determine whether lumican co-localizes with CD14, CD18, and bacteria on M⌽ surfaces during phagocytosis. WT M⌽s were incubated with Alexa Fluor 488 E. coli BioParticles conjugates (Alexa 488 E. coli) and immunostained for lumican, CD14, and CD18; overlapping staining patterns for lumican, bacteria, CD14, and CD18 were detected (Fig. 2, A-C).
We next tested the role of lumican in CD14-and CD18mediated phagocytosis using two different P. aeruginosa strains with differential host receptor bias. The P1 strain has a CR3 (␤2 or CD18) receptor bias, whereas the 808 strain is taken up by CD14-mediated phagocytosis (22). Because Lum Ϫ/Ϫ M⌽s were fully capable of engulfing serum-opsonized bacteria (Fig. 1A), we envisioned CD18, a complement component 3 receptor subunit, to have lumican-independent phagocytic functions. Consistent with this idea, phagocytosis of the P1 strain (CD18dependent) was reduced slightly, whereas that of the 808 strain (CD14-dependent) was reduced significantly in Lum Ϫ/Ϫ M⌽s after 1 and 2 h of incubation at 37°C (p ϭ 0.0006 and p ϭ 0.0005) (Fig. 2D), indicating that lumican functions in CD14regulated phagocytosis primarily. We further tested this using pHrodo TM E. coli BioParticles and WT M⌽s after antibodymediated blocking of either CD14 or CD18 (supplemental Fig.  S3). Lum Ϫ/Ϫ M⌽s were susceptible to anti-CD18-mediated inhibition of phagocytosis (p Ͻ 0.01), indicating CD18 phagocytosis to be operative, even in the absence of lumican. By contrast, anti-CD14 caused very little additional blockade of phagocytosis in the Lum Ϫ/Ϫ M⌽s (p ϭ 0.07), indicating low CD14 phagocytic activities in the absence of lumican.
Bacterial Uptake Reduced in Lum Ϫ/Ϫ M⌽s-To investigate how lumican and CD14 cooperate in phagocytosis, we broke down overall phagocytosis into two stages: initial bacterial uptake and formation of phagosomes and a later stage where bacteria appear in late endosomes and lysosomes. In actuality, this is a multistage process. To test whether initial bacterial uptake is aided by lumican, WT and Lum Ϫ/Ϫ M⌽s were incubated for 20 min with Alexa 488 E. coli, fixed without permeabilization, and immunostained for Na/K-ATPase as a membrane marker. The Lum Ϫ/Ϫ M⌽s had fewer plasma membrane-associated bacteria compared with WT. However, bacterial association with CD14 Ϫ/Ϫ M⌽s, which are wild type for lumican, was normal and may be even higher (Fig. 3, A and B). This may be because progression of the early phagosomes to late endosomes is delayed in CD14 deficiency; more bacteria can be seen at the cell surface at any one time. The Lum Ϫ/Ϫ CD14 Ϫ/Ϫ showed fewer bacterial phagosomes. We interpreted this as lumican deficiency interfering with the early stages of bacterial association to disrupt phagocytosis altogether (Fig. 3A). Overall, Alexa 488 E. coli association was reduced by 37% in lumican-deficient M⌽s (Lum Ϫ/Ϫ , Lum Ϫ/Ϫ CD14 Ϫ/Ϫ ) but unaffected in the CD14 Ϫ/Ϫ M⌽s (Fig. 3B). We also tested phagocytosis of P. aeruginosa, using a GFP-expressing strain (PAO1) in in vitro assays and found its uptake to be reduced in the Lum Ϫ/Ϫ and Lum Ϫ/Ϫ CD14 Ϫ/Ϫ M⌽s as well (Fig. 3, C and D). On the other hand, uptake of Gram-positive S. aureus was not significantly impaired in the Lum Ϫ/Ϫ , CD14 Ϫ/Ϫ , or Lum Ϫ/Ϫ CD14 Ϫ/Ϫ compared with WT M⌽s (data not shown). Future investigations will use macrophage-like cell lines, manipulated to express lumican, to further investigate regulation of intermediate steps in phagocytosis by lumican.
Appearance of Bacteria in Late Endosome/Lysosome Delayed by CD14 Deficiency-In the second stage, we asked whether the transfer of bacterial phagosomes to late endosome/lysosomes was affected by lumican deficiency. Peritoneal M⌽s from WT, Lum Ϫ/Ϫ , CD14 Ϫ/Ϫ , and Lum Ϫ/Ϫ CD14 Ϫ/Ϫ doublenull mice were incubated with bacteria for 40, 60, and 90 min, fixed, and then immunostained with a late endosome/lysosome marker LAMP-1 (Fig. 4). The WT M⌽s showed multiple Alexa 488 E. coli bioparticles at the cell surface by 40 min (thin arrow), and by 90 min there were several co-localizations (yellow) with LAMP-1 (thick arrow). The number of early phagosomes appearing green, and those co-localizing with LAMP-1 and appearing yellow were counted for each microscopic field ( Table 3). The Lum Ϫ/Ϫ M⌽s showed few early bacterial phagosomes, confirming poor initial uptake in the absence of lumican; but of the few early phagosomes that were formed, by 90 min two of them co-localized with LAMP-1. In contrast, the Lum ϩ/ϩ CD14 Ϫ/Ϫ M⌽s showed as many early phagosomes (thin arrow), as the WT, but no co-localization with LAMP-1, and the double-null M⌽s showed a lack of bacterial phagosomes and their co-localization with LAMP-1 (Table 3). We further confirmed poor fusion of bacterial phagosomes with late endosome/lysosome in CD14 deficiency by prestaining all four genotype of M⌽s with LysoTracker Red DND-99 and testing for uptake of Alexa 488 E. coli and passage to the lysosomal compartment by time lapse live cell confocal microscopy between 30 and 60 min of incubation (supplemental Fig. S4). For WT and Lum Ϫ/Ϫ M⌽s, bacterial passage from the surface to acidic compartments occurred within 30 -40 min of incubation, whereas the CD14-deficient M⌽s (Lum ϩ/ϩ CD14 Ϫ/Ϫ and Lum Ϫ/Ϫ CD14 Ϫ/Ϫ ) did not show phagosome acidification within these time frames. Thus, although bacterial uptake was significantly impeded in Lum Ϫ/Ϫ M⌽s, phagosome acidification and fusion with late endosome/lysosome appeared to be unaffected.

N-terminal Tyrosine of Lumican Involved in CD14
Interactions-We hypothesized that lumican acts as an intermediary between bacteria and CD14 to drive induction of cytokines and phagocytosis by CD14 (and TLR4). We considered four N-terminal tyrosine residues in lumican as functionally important for binding CD14 for the following reasons. These tyrosine residues of lumican have been shown by others to undergo posttranslational sulfation (23). Tyrosine sulfation is a feature of other SLRPs, such as fibromodulin and osteoadherin, and suspected to support protein-protein interactions (23, 24). We developed two variants of rLum, rLumY20A and rLumY21A, where a tyrosine was replaced with an alanine in each (Fig. 5A), and by surface plasmon resonance determined their binding kinetics with CD14. The recombinant wild type protein rLum showed a dose-and time-dependent binding to CD14 immobilized on a CM5 sensorchip (Fig. 5B). As expected, anti-CD14 and nonspecific IgG, used as positive and negative control ana- lytes, showed strong binding (K A ϭ 8.26 ϫ 10 9 M Ϫ1 ) and no measurable binding with CD14, respectively (Fig. 5, C and D). The CD14-rLumY20A interaction showed association kinetics that was comparable with rLum. However, dissociation of rLumY20A was rapid, such that the dissociation constant (K D ) for CD14 and rLumY20A was 60-fold higher than that of CD14-rLum or CD14-rLumY21A ( Fig. 5E and Table 4). Association and dissociation constants with CD14 were not adversely affected by the rLumY21A mutation ( Fig. 5F and Table 4). To discount the loss of CD14 binding in rLumY20A as a purification-related artifact, we tested rLum and rLumY20A in collagen fibrillogenesis assays, often used as a test for biological activity for these ECM proteins (25). Collagen solutions form fibrils spontaneously in vitro, which can be tracked spectrophotometrically. The SLRP core proteins have been shown to bind collagen and inhibit fibrillogenesis. We found comparable inhibition of collagen fibrillogenesis by rLum and rLumY20A, suggesting retention of biological activity in rLumY20A, and the selective loss of its CD14 binding activity is most likely due to the tyrosine substitution (Fig. 5G).
Tyrosine 20 in Lumican Is Biologically Important for CD14 Functions-Because CD14 binding to rLumY20A was markedly reduced, we postulated that the tyrosine at position 20 is important for stable binding to CD14 and its functions in phagocytosis. Wild type rLum and rLumY21A could restore phagocytosis in Lum Ϫ/Ϫ M⌽s, whereas the rLumY20A mutant improved phagocytosis by only 17.5% (Fig. 6A). To determine whether the CD14-binding region on lumican drives CD14 functions further in TLR4-mediated induction of proinflammatory cytokines, we tested rLum, rLumY20A, and rLumY21A for their abilities to restore LPS sensing and induction of TNF-␣ in Lum Ϫ/Ϫ M⌽s. In the presence of rLum or mutated rLumY21A, the Lum Ϫ/Ϫ M⌽s were able to respond to LPS and produce wild type levels of TNF-␣, whereas the rLumY20A variant could not restore LPS response (Fig. 6B). Interestingly, rLumY20A interfered with normal LPS response in WT M⌽s as well.
Pseudomonas Lung Infection Morbidity Increased in the Lum Ϫ/Ϫ Mice-To assess the relevance of lumican-driven phagocytosis in a disease setting, we tested the Lum Ϫ/Ϫ mice in a lung infection model using a virulent, cytotoxic strain of P. aeruginosa (ATCC 19660). By 12 h of infection, 42% of the Lum Ϫ/Ϫ mice died, as opposed to only 14% of the Lum ϩ/ϩ mice. By 20 h, 64% of the Lum Ϫ/Ϫ infected mice had died (Fig.  7A). After 12 or 20 h of infection, bacterial clearance from the lungs was significantly impaired in the Lum Ϫ/Ϫ mice; the infected WT yielded few to no viable bacteria, whereas the Lum Ϫ/Ϫ lung homogenates yielded 10 6 -10 9 cfu (Fig. 7B). In an in vitro bacterial killing assay, whereas both M⌽ types reduced the total bacterial load markedly, the WT were two to three times more effective than Lum Ϫ/Ϫ M⌽s (Fig. 7C). Histology of the 12-h infected Lum Ϫ/Ϫ lung tissues showed degeneration and necrosis/apoptosis of the bronchiolar epithelium, some atelectasis of alveoli, and mild neutrophil infiltrates (supplemental Fig. S5A). By 20 h of infection alveolar tissue damage in the Lum Ϫ/Ϫ mice was extensive (supplemental Fig. S5B). By flow cytometry, we detected increased F4/80 ϩ cells in the bronchoalveolar lavage and increased Gr-1 ϩ neutrophils by histology of lung tissues of the 20-h infected Lum Ϫ/Ϫ mice (supplemental Fig. S5, C and D). Our earlier study has shown that lumican promotes inflammatory cell migration, and in an LPScorneal injury model, the Lum Ϫ/Ϫ corneas had shown delayed influx of neutrophils (26). However, in the lung infection  Table 4 is a quantitative representation of this figure.  OCTOBER 19, 2012 • VOLUME 287 • NUMBER 43  (Table 4). G, collagen fibrillogenesis was analyzed in the presence of wild type rLum and mutated rLumY20A. Both lumican forms demonstrated an inhibition of collagen fibrillogenesis, seen as an increase in the lag times required of initiation of fibrillogenesis as well as a decrease in the final absorbance. Both forms demonstrated comparable effects, indicating no difference in the biological activity. In this representative assay, the collagen I concentration was 100 g/ml PBS, and rLum and rLumAY20A were present at a molar ratio of 1:6 (lum:collagen) at 34°C. The assays were also done using collagen I concentrations of 50 and 100 g/ml PBS at molar ratios of 1:6 and 1:1 (collagen:lum) and 34 -35°C. In all cases, the results were comparable with those presented.

model, by 20 h any delay in leukocyte migration in the infected
Lum Ϫ/Ϫ lungs was clearly overcome, and levels of IFN-␥ and TNF-␣ in the Lum Ϫ/Ϫ mouse bronchoalveolar lavage was WTlike (data not shown). However, increased mortality in the Lum Ϫ/Ϫ mice may be the result of a number of issues, reduced bacterial killing, poor epithelial healing, and reduced apoptotic clearing of inflammatory cells.

DISCUSSION
Our study shows that association between bacteria and lumican and binding between lumican and CD14 promote phagocytosis by CD14. Lum Ϫ/Ϫ M⌽s showed poor uptake of bacteria. Using three different confocal microscopy approaches, we showed that bacterial phagosome progression to lysosomes was delayed in Lum Ϫ/Ϫ M⌽s. In P. aeruginosa-mediated lung infections, the Lum Ϫ/Ϫ mice showed poor bacterial clearance and a dramatic increase in mortality. P. aeruginosa is a common Gram-negative bacterium in cystic fibrosis patients, and a frequent nosocomial pathogen in ventilator-associated lung infections (27)(28)(29)(30)(31). Our study suggests a significant role for lumican in phagocytosis and resolution of bacterial lung infections. Although in vitro bacterial uptake and killing was reduced in Lum Ϫ/Ϫ M⌽s, increased mortality of the infected Lum Ϫ/Ϫ mice may be due to a combination of events. Previous studies show that lumican is transiently expressed by injured epithelial cells, and epithelial wound healing in the cornea is delayed by lumican deficiency (26,32). We have also shown that lumican promotes Fas-FasL-mediated apoptosis, with reduced apoptosis in injured Lum Ϫ/Ϫ mouse corneas (16). Thus, delayed epithelial healing and inefficient apoptotic removal of damaged cells and inflammatory infiltrates via Fas-FasL may also con-tribute to increased mortality and morbidity of Lum Ϫ/Ϫ mice in P. aeruginosa lung infections. Interestingly, an epithelial cell surface heparan sulfate proteoglycan, syndecan-1, regulates P. aeruginosa lung infections by facilitating resolution of neutrophilic inflammation (33). It remains to be seen whether transient expression of lumican by epithelial cells modulates syndecan-1 functions in lung infections.
Lumican is an ECM protein; to interact with the cell surface it requires specific cell surface proteins. Lumican was reported to bind M⌽s in a divalent cation-dependent way that implicated integrins rather than scavenger receptors or CD14 (18). Our earlier studies demonstrated binding of lumican with CD18 and CD14 in immunoprecipitation and pulldown assays (15,17). CD18 is the ␤2 integrin subunit of the complement component 3 receptor, known to regulate phagocytosis of complement-opsonized bacterial particles (34). However, we show that lumican-aided phagocytosis occurs primarily through CD14, and not CD18, as Lum Ϫ/Ϫ M⌽s 1) phagocytose serumopsonized Gram-negative bacteria normally and 2) are unable to phagocytose P. aeruginosa strain 808 (CD14 bias) efficiently. By surface plasmon resonance of lumican and its mutated variants, we further show that tyrosine 20 of the core protein is important in binding CD14. Other SLRPs, fibromodulin, and osteoadherin also have a tyrosine-rich N-terminal region that may be sulfated and involved in protein-protein interactions in the ECM, although no biological functions have been attributed to these sites yet (24,35). CD14 has been shown to promote LPS-binding protein (LBP)-dependent phagocytosis that is also complement-and divalent cation-independent (36). CD14 is one of several proteins at the cell surface that regulate ligand recognition by TLR4 (7). Thus, in addition to interactions with CD14 and CD18, lumican may have other binding partners in this complex that impact its role in phagocytosis and innate immune response. No other SLRP members have been shown to regulate phagocytosis; however, we detected decorin on the surface of peritoneal macrophages also co-localizing with Alexa 488 E. coli, CD14, and CD18 (data not shown).
Functions of CD14 and TLR4 in phagocytosis are somewhat conflicted, and re-evaluation of the reported findings, after taking into consideration the opsonization status of the pathogens, may clarify some of this data. According to one study, CD14 Ϫ/Ϫ M⌽s showed no difference in the uptake of E. coli particles (37). Yet another reported CD14 Ϫ/Ϫ mice to be hypo-responsive to E. coli with poor induction of proinflammatory cytokines and poor dissemination of bacteria introduced intraperitoneally (38). Using nonopsonized bacteria when we dissected the overall phagocytic process into bacterial uptake and phagosome acidification, we found clear distinctions in lumican and CD14 functions. The Lum Ϫ/Ϫ M⌽s clearly had difficulties in bacterial uptake, whereas CD14 Ϫ/Ϫ and the Lum Ϫ/Ϫ CD14 Ϫ/Ϫ double null M⌽s were dramatically delayed in forming late endosomes. The earlier contrasting report of no deficiency in phagocytosis by CD14 Ϫ/Ϫ M⌽s did not exclude serum, thus not accounting for CD14-independent phagocytosis of opsonized bacteria (37). Our current findings on CD14 in phagocytosis may provide new insights into the role of TLR4 in this process.
TLR4 was generally thought to interact with target ligands in phagosomes and activate cytokine and chemokine induction, thus allowing efficient linking of inflammatory responses to microbial internalization (4). A later study reported that TLRs do more than jump start inflammatory cytokine production; TLR4 induced maturation of phagosomes containing bacteria in a MyD88-dependent manner (39). They further concluded that the inducible nature of phagosome maturation was autonomous, because LPS stimulation did not hasten phagocytosis of apoptotic cells, and only bacteria-containing phagosomes had the inducible quality. Taking our observation that phagosome maturation requires CD14, the direct contribution of TLR4 to phagocytosis may actually be due to membrane-anchored CD14 in the same phagosome. In further support of this concept, using an in vitro FRET-based technique, Yates and Russell (40) showed that stimulation of TLR4 or TLR2 had no effect on phagosome lysosome fusion.
An interesting point to consider is the source and form of lumican that regulates CD14 functions in macrophages. Lumican is present widely in many connective tissues (11,19,41,42). We and others detected lumican by Northern blotting and in situ hybridization in mouse lungs (19,21), and in vascular endothelial ECM (20) and HUVEC cultures (17) by immunoblotting. On the other hand, increased deposits of lumican and biglycan were reported in the lungs of asthmatic patients (43). In in vitro assays, we found M⌽s from Lum Ϫ/Ϫ 3 WT BMC to show WT-like phagocytosis. Thus, as our earlier study suggested for neutrophils (17), M⌽s acquire exogenous lumican from one or more of these ECM to enhance its phagocytic capabilities promote resolution of lung infection. Lumican exists in multiple post-translationally modified forms that may have distinctive functions in health and disease. In the collagen-rich stroma of the cornea, lumican exists in its keratan sulfate proteoglycan form, whereas in vascular ECM (20) and as we show in leukocyte exudates, it is a glycoprotein (17). The proteoglycan form of lumican may be less reactive with the cell surface; interactions of collagen fibrils with lumican in connective tissues may also protect the CD14-reactive N-terminal end of lumican from inappropriate amplification of innate immune responses. To this end, lumican and other SLRP fragments have been detected in degenerative joint diseases (44), whereas lumican levels have been variably associated with cancer and metastasis (45)(46)(47)(48)(49). There are only a few reports of other SLRP members displaying immune modulatory influences. Biglycan was reported to mimic danger signals in its soluble form and stimulate proinflammatory cytokines (50). Decorin and fibromodulin core proteins interact with the complement component C1q, and fibromodulin may promote the classical complement pathway (51). No other members of the 18 SLRP family constituents have been shown to have a definitive role in host defense against microbial infections.
In summary, we have elucidated a role for lumican in binding Gram-negative bacteria and driving CD14-mediated bacterial phagocytosis. An N-terminal tyrosine in lumican is required for binding CD14 and subsequent uptake of bacteria for ultimate pathogen clearance, such that lumican-deficient mice are grossly inadequate in fighting P. aeruginosa lung infection. Our study raises the possibility that tight regulations of cell-reactive lumican forms may be important to infections, repair, regeneration, and degenerative diseases.