Digestion of Streptococcus pneumoniae Cell Walls with Its Major Peptidoglycan Hydrolase Releases Branched Stem Peptides Carrying Proinflammatory Activity*

The peptidoglycan of Gram-positive bacteria is known to trigger cytokine release from peripheral blood mononuclear cells (PBMCs). However, it requires 100–1000 times more Gram-positive peptidoglycan than Gram-negative lipopolysaccharide to release the same amounts of cytokines from target cells. Thus, either peptidoglycan is poorly active or only part of it is required for PBMC activation. To test this hypothesis, purified Streptococcus pneumoniaewalls were digested with their major autolysinN-acetylmuramoyl-l-alanine amidase, and/or muramidase. Solubilized walls were separated by reverse phase high pressure chromatography. Individual fractions were tested for their PBMC-stimulating activity, and their composition was determined. Soluble components had a M r between 600 and 1500. These primarily comprised stem peptides cross-linked to various extents. Simple stem peptides (M r <750) were 10-fold less active than undigested peptidoglycan. In contrast, tripeptides (M r >1000) were ≥100-fold more potent than the native material. One dipeptide (inactive) and two tripeptides (active) were confirmed by post-source decay analysis. Complex branched peptides represented ≤2% of the total material, but their activity (w/w) was almost equal to that of LPS. This is the first observation suggesting that peptidoglycan stem peptides carry high tumor necrosis factor-stimulating activity. These types of structures are conserved among Gram-positive bacteria and will provide new material to help elucidate the mechanism of peptidoglycan-induced inflammation.

Inflammation is a complex, nonspecific host defense response to pathological situations such as infection. When localized, inflammation is beneficial because it helps eradicate the invading organisms at the infected site (1). In severe situations, however, the deregulation of numerous host defense factors and soluble cellular mediators, such as cytokines, may result in a dramatic cardiovascular collapse called septic shock (2,3). The cytokine tumor necrosis factor (TNF) 1 was shown to play a critical role in triggering such an event (4,5). Septic shock is a classical response to severe infections due to Gram-negative bacteria, in which the outer membrane lipopolysaccharide (LPS) is the principal bacterial component triggering cytokine release from peripheral blood mononuclear cells (PBMCs) and other responsive cells (6 -9). To stimulate cytokine release, LPS binds to the cell surface protein CD14 (10), a glycosylphosphatidylinositol-linked determinant that presumably requires additional transmembrane elements such as the recently described Tol receptor (11) to trigger cell response. In vivo LPS also binds to lipopolysaccharide-binding protein, an acute phase protein that catalyzes the binding of LPS to CD14 (12).
Septic shock is also a feature of severe Gram-positive infections (13)(14)(15)(16). In recent years, Gram-positive pathogens have become more frequent than Gram-negative bacteria as a cause of nosocomial infections and shock (17,18). This is particularly true in intensive care unit patients and in neutropenic patients, two subgroups with impaired host defenses that are especially prone to develop severe sepsis. However, in contrast to their Gram-negative counterparts, Gram-positive bacteria do not contain LPS. In certain cases, highly pathogenic agents such as Staphylococcus aureus and Streptococcus pyogenes produce toxins that may act as superantigens and trigger an overwhelming nonspecific T-cell-mediated inflammatory response (7,13). With less pathogenic agents, such as viridans group streptococci, triggering of inflammation and shock might be due to other cellular components, including the thick Grampositive peptidoglycan. Gram-positive cell walls were shown to stimulate TNF release and inflammation in several in vitro and in vivo systems, including experimental meningitis, experimental arthritis, and experimental septic shock (19 -21).
To study the mechanism of peptidoglycan-induced inflammation, different types of Gram-positive wall materials were utilized. On the one hand, large molecules such as insoluble cell walls and soluble peptidoglycan released from penicillintreated staphylococci allowed the demonstration that CD14 was also a receptor for Gram-positive wall products (22,23). On the other hand, the small molecule muramoyl dipeptide was shown to be the minimal common peptidoglycan structure carrying immunomodulatory activity (24). However, while very useful, these compounds might not be ideal for solving the structure-activity relationship between the wall degradation products occurring in the nature and their cytokine-releasing capacity. First, molecules such as soluble peptidoglycan are very large (M r Ն125,000) (25,26) and might be too complicated for refined analysis. Second, muramoyl dipeptide is not a nat-ural product of wall degradation, and it has only limited cytokine-stimulatory power in vitro (27).
To further explore this question, we attempted to isolate and characterize small proinflammatory peptidoglycan fragments resulting from the digestion of pneumococcal walls with their native autolysins, thereby liberating products of digestion that are likely to be found in vivo. Specifically, Streptococcus pneumoniae contains a major autolytic enzyme (an N-acetylmuramoyl-L-alanine amidase, which hydrolyzes the bonds between the glycan chain and the stem peptides) that is responsible both for wall solubilization and cell lysis during the stationary growth phase or during ␤-lactam treatment (28 -30). The role of this autolytic enzyme in pneumococcal pathogenesis has been demonstrated (31). In certain experiments, a second enzyme (M1-muramidase) that hydrolyzes the glycan chain was used in parallel. M1-muramidase is not formally a pneumococcal enzyme, but it has a functional equivalent in these organisms (32). The resulting soluble fragments were separated by reverse phase high pressure chromatography (HPLC). Individual fractions were tested for their TNF-releasing activity, and the molecular structure of single components carrying inflammatory power was analyzed by mass spectrometry (MS), amino acid and amino sugar analysis, and postsource decay analysis.

EXPERIMENTAL PROCEDURES
Bacterial Strain and Growth Conditions-The noncapsulated lysisprone positive S. pneumoniae strain R6 was used (33). Bacteria were grown in the chemically defined liquid medium Cden (44). Growth of the cultures was followed by their optical density at 620 nm (A 620 ) with a spectrophotometer (Sequoia-Turner, Moutainville, CA). The cells were stored at Ϫ70°C in Cden containing 20% glycerol.
In certain experiments a label was introduced into the cell walls by either of the following methods. (i) To label the glycan chains, N-acetyl- H]glucosamine (Amersham Pharmacia Biotech) was added to the culture to a final concentration of 250 Ci/liter and 1.23 g/liter. (ii) To label the stem peptides, L-[4,5-3 H]lysine (Amersham Pharmacia Biotech) was added to a final concentration of 250 Ci/liter and 0.45 g/liter. The cultures were allowed to grow for 4 -5 generation times in the presence of the label before being harvested as described below.
Preparation and Purification of Cell Walls-Cell walls were prepared as described previously (34). Briefly, one liter cultures grown to the late logarithmic phase (A 620 of 0.5) were rapidly cooled in an ice/ethanol bath and harvested by centrifugation for 10 min at 16,000 ϫ g and 4°C. The cells were resuspended in endotoxin-free water, and intrinsic autolytic enzymes were inactivated by pouring the suspension dropwise into boiling SDS (4%, final concentration). After 15 min of boiling, the SDS was removed by extensive washing with water, and the cells were broken open by vortexing with an equal volume of acid-washed glass beads with a 106-m diameter (Sigma). The cells walls were pelleted by centrifugation at 25,000 ϫ g for 30 min at 4°C; resuspended in 0.1 M Tris-HCl buffer (pH 7.5) containing 0.05% sodium azide; and treated with DNase, RNase, and trypsin as described (34). Peptides from protease digestion were extracted by boiling for another 10 min with SDS (1% final concentration), and the walls were incubated with 8 M lithium chloride and then 100 mM EDTA to remove material bound by ionic interactions. To remove possible contaminating endotoxin, the insoluble preparation was washed extensively with acetone followed by resuspension in endotoxin-free water prior to drying by rotary evaporation. The absence of contaminating endotoxin was confirmed by a negative result of the Limulus amebocyte test (Chromogenix, Molndal, Sweden) with a sensitivity of 4 pg/ml. Digestion of Cell Walls by Autolytic Enzymes-To hydrolyze the bonds between the glycan chains and the stem peptides, the natural pneumococcal autolysin N-acetylmuramoyl-L-alanine amidase (referred to as amidase in the following) was used. This enzyme was purified by affinity chromatography (35) from Escherichia coli CM21 lysates, into which the pneumococcal amidase gene (lytA) had been cloned (36). Purified pneumococcal walls were suspended in 50 mM MOPS buffer (pH 6.9) at a concentration of 1 mg/ml. The mixture was incubated with 100 units/ml (final concentration) of amidase for 18 h with shaking at 37°C. The digest was spun for 10 min at 15,000 ϫ g, and the supernatant, containing the solubilized walls, was collected. In certain experiments, the amount of radioactivity was counted to determine the extent of solubilization. The supernatant was then boiled for 10 min and centrifuged at 15,000 ϫ g and 4°C to denature and remove the amidase. Eventually, the soluble walls were washed with acetone and water, as described, to remove any contaminating endotoxin and subsequently dried by rotary evaporation and stored at Ϫ20°C. This enzymatic treatment resulted in Ͼ90% solubilization of the purified material.
To hydrolyze the bonds between the N-acetylmuramoyl and N-acetylglucosamine residues, mutanolysin from Streptomyces globiosporus (Sigma; EC 3.2.1.17) was used. The enzyme was added to a final concentration of 500 units/ml and 1 mg/ml substrate in 0.25 mM potassium phosphate buffer (pH 6.5) containing 10 mM magnesium chloride. The mixture was incubated for 18 h with gentle shaking at 37°C before boiling and centrifugation to remove the enzyme as above. The supernatant was subsequently washed with acetone, dried by rotary evaporation, and stored at Ϫ20°C as above.
In certain experiments, the glycan chains and stem peptides were separated by differential precipitation in a mixture of water/acetonitrile/propan-2-ol (in the ratio 50:25:25) containing 0.1% trifluoroacetic acid. Typically, 600 g of amidase-solubilized cell walls were suspended in this solvent and kept on ice for 30 min. Insoluble material containing the glycan-rich fraction was then collected by centrifugation at 15,000 ϫ g for 15 min (4°C). The supernatant contained the stem peptides. Solvents were removed by rotary evaporation, and the peptides and glycan were dissolved in water. The amount of material present in each fraction was quantified by liquid scintillation counting.
Separation of Digested Peptidoglycan by HPLC and Identification of Specific Fractions-The HPLC system (Hitachi Instruments, Ichige, Hitachinaka, Japan) consisted of the L-7200 autosampler, the L-7100 gradient pump, with low pressure mixing, and the L-7400 UV detector. Column temperature was maintained at 25°C using a pelcooler (Lab-Source, Reinach, Switzerland). The results were analyzed using the D-7000 HPLC System Manager program (Hitachi).
Separation was performed by injection of a 100-l sample, containing 100 g of wall digest, into a C18 reverse phase column (SuperPac Sephasil C18, 5 m, 4 ϫ 250-mm column, Amersham Pharmacia Biotech) protected with a guard cartridge (C18, 5 m, 4 ϫ 10 mm). The mixture was separated using a linear gradient of 0 -15% acetonitrile in 0.1% trifluoroacetic acid over 100 min with a flow rate of 0.5 ml/min. Detection was at 210 nm. 1-min fractions were collected. The solvent was removed by rotary evaporation. The amount of material present in each fraction was too small to be measured either by weight or by radioactive counts and was therefore calculated by the percentage area method. Knowing the weight of material loaded, the amount of material present in each fraction was deduced. The samples were dissolved in 100 l of water and stored at Ϫ20°C.
The content of specific fractions was analyzed by the following means. First their content in amino acids was determined using the OPAC postcolumn derivatization method following hydrolysis of the samples for 24 h in 6 M HCl at 110°C (37). Second, their molecular mass was measured by matrix-assisted laser desorption ionization time of flight MS, using a Voyager-Elite TOF mass spectrometer (PerSpective Biosystems Inc., Farmingham, MA) operating in the positive ion mode and using an ␣-cyano-4-hydroxycinnamic acid matrix. The laser power was set to 1,600 units, and the accelerating voltage was set to 20,000 V. Postsource decay analysis was performed on certain fractions.
Preparation, Stimulation of Human PBMCs or Whole Blood from Rabbits, and Measurement of TNF-␣ Release-Human PBMCs were extracted from heparinized blood of healthy volunteers by Ficoll-Hypaque (Seromed, Munich, Germany) density gradient centrifugation as described by Heumann et al. (22). The cells were suspended in RPMI 1640 medium (Life Technologies, Inc.) and distributed into the wells of a flat bottomed 96-well tissue culture plate at a concentration of 0.5 ϫ 10 6 cells/well. Each well contained a final volume of 200 l, which comprised the RPMI medium (140 l), plasma from the donor (20 l), and sample (20 l). LPS from E. coli O111 (Sigma) in the concentration range 0.01-100 ng/ml was used as a positive control. PBMCs incubated with plasma or with medium alone were used as negative controls. The plates were incubated at 37°C in an atmosphere of 5% CO 2 . After 8 h of incubation, samples (20 l) of the supernatants were taken, diluted 20-fold in RPMI medium, and stored at Ϫ80°C for measurement of TNF concentrations.
For the whole blood assay, heparinized rabbit blood was diluted 4-fold in RPMI 1640 medium supplemented with 10% fetal calf serum as described (38). The diluted blood was distributed into tissue culture plates (200 l/well). Samples (20 l) were added, and the plates were incubated as above for 6 h . The plates were centrifuged for 5 min, and samples of the supernatant were diluted and stored as described above for subsequent measurement of TNF concentrations.
WEHI clone 13 murine fibroblast cells (1 ϫ 10 4 /well) were used for quantitation of TNF-␣ as described (39). Recombinant mouse TNF-␣ were used as a standard. The sensitivity of the assay was 25 pg/ml.

TNF-stimulating Activity of LPS and Purified Pneumococcal
Walls- Fig. 1 depicts a typical titration experiment of the TNFreleasing activity of LPS and insoluble pneumococcal walls exposed to human PBMCs in the presence of 10% plasma. Both LPS and purified cell walls triggered TNF release from the monocytes in a concentration-dependent fashion. However, it required ϳ1000 times more pneumococcal walls than LPS to release the same amounts of TNF from the target cells. This was in agreement with previous observations reported by Timmerman et al. (40) with insoluble staphylococcal peptidoglycan and Heumann et al. (22) with pneumococcal peptidoglycan.
To attempt to expose more active pneumococcal wall subcomponents that might be embedded in the insoluble peptidoglycan, we first solubilized the purified walls using the bacteria's own major autolysin amidase, purified as described (35), and then sought inflammatory subcomponents in the soluble material. Amidase solubilized Ͼ90% of the purified walls. Nevertheless, the inflammatory power of this soluble mixture was not greater than that of insoluble material. Similar observations were made by Riesenfeld-Orn et al. (14). Indeed, when soluble and insoluble walls were titrated for their TNF-releasing activity, as in Fig. 1, they produced very similar stimulation profiles and had the same stimulatory potency (see upper part of Table I). This observation came in support of the hypothesis that only part of the peptidoglycan might be inflammatory and suggested that such specific inflammatory constituents might now be present in the soluble mixture. Therefore, the wall constituents were further separated by reverse phase HPLC, and individual fractions were tested for their PBMC and/or whole blood-stimulating activity.
Solubilization and HPLC Separation of Cell Walls- Fig. 2 depicts the solubilization steps used in these experiments. Purified insoluble pneumococcal walls (step 1 in Fig. 2) digested with amidase (step 2 in Fig. 2) produced a wealth of soluble fragments comprising free stem peptides, separated from the glycan backbone at the N-acetylmuramoyl-L-alanine bonds, and glycan chains decorated or not with teichoic acids and/or residual peptides (34). However, although amidase solubilized Ͼ90% of the walls (step 2 in Fig. 2), the soluble fragments generated by this treatment could not be properly resolved by the chromatography conditions used in these experiments. The profile of this separation is shown in Fig. 3a.
We reasoned that this might be due to some interference of the undigested glycan backbone (step 2 in Fig. 2) with the C18 column. To circumvent this problem, two strategies were adopted. In the first, the amidase-solubilized walls were submitted to a second digestion with muramidase (step 3A in Fig.  2), an enzyme that hydrolyzes the 1-4 ␤-bonds between the N-acetylmuramoyl and N-acetylglucosamine residues of the glycan chain. The profile of this second separation is depicted in Fig. 3b. In contrast to Fig. 3a, Fig. 3b indicates that doubledigested wall material could now be resolved by HPLC. The chromatogram was complex, indicating that the peptidoglycan contained a variety of subcomponents. Nevertheless, this profile was perfectly reproducible in terms of both interrun and intercell wall batch resolution and over at least four independent experiments.
A second strategy (step 3B in Fig. 2) to encompass the interference of the undigested glycan chains consisted of removing the glycans from the amidase digest by serial precipitation in water/acetonitrile/propan-2-ol (as described under "Experimental Procedures") and then analyzing only the peptide-enriched fraction by HPLC. Fig. 3c depicts the result of this separation. It can be seen that this chromatogram resembled very much that of Fig. 3b, indicating that the stem peptides of the cell walls were the prevalent species resolved with both of these methods.
TNF-releasing Activity of Individual HPLC Fractions-In a first series of experiments, we determined the TNF-releasing activity of 1-min fractions collected from the chromatogram presented in Fig. 3b. Individual fractions were exposed to human PBMCs as described. Fig. 4 indicates that the TNF-releasing activity was very low for material eluting before 30 min, whereas it rose considerably later on, i.e. up to 100-fold above background levels. Since the chromatograms resulting from the separation of either double-digested or single amidase-digested peptidoglycan were almost superimposable, we repeated the stimulation experiment with 1-min fractions of both of these enzymatic digests. In this second series of experiments, a new column was used, which resulted in a slight increase in retention times. Fig. 5 presents these results. In these experiments, the TNF assay was performed with whole blood from rabbits, a system that gave similar results as human PBMCs but was much more convenient for screening large numbers of fractions. It can be seen that not only the chromatograms of doubledigested or single amidase-digested peptidoglycan were very FIG. 1. TNF-releasing activity of LPS and pneumococcal walls exposed to human PBMCs. Human PBMCs were isolated as described and exposed for 8 h to LPS or purified pneumoccal walls in the presence of 10% plasma. TNF production was measured by its cytotoxic activity on murine fibroblasts. Recombinant human TNF was used as a standard. The black bars indicate the levels of TNF produced by stimulation with LPS from E. coli O111, and the open bars indicate the amounts of TNF produced by stimulation with purified pneumoccal walls. Concentrations of TNF and LPS or cell wall stimulants are indicated.  Table  II). similar, but their profiles of TNF-release were comparable as well. This was consistent with the fact that the peptide-enriched material contained active subcomponents similar to those observed in the double-digested material.
Quantitation of TNF-releasing Activity in Separate HPLC Fractions-To more precisely evaluate the specific TNF-releasing activity of the material present in individual HPLC fractions (Fig. 4), the content of these fractions was quantified as follows. First, the relative contribution of each peak to the whole chromatogram was determined by comparing their areas under the curve to that of the whole HPLC profile using the D-7000 HPLC System Manager software (Hitachi). This permitted the calculation of the percentage of material present in each fraction as compared with the total material loaded on the column. Using this information, the concentration of material present in each fraction was then calculated, and the specific TNF-releasing activity of the fractions was determined. Table I presents these results. It indicates an appraisal of the relative stimulatory potency of the various materials tested in the present experiments by denoting the minimum concentra-  Fig. 3b. Individual fractions were tested both for stimulation of PBMCs and structural analysis based on MS and amino acid and amino sugar analysis. Similar profiles were obtained irrespective of whether the double digest or peptide-enriched fractions were tested (see Fig. 5).

FIG. 2. Steps used in the solubilization of the cell walls and the separation of its components.
The diagram represents the principal wall components. The horizontal lines indicate the glycan chains. The vertical lines indicate the stem peptides and their interconnecting bonds. The black circles represent the location of glycan-attached teichoic acids. Insoluble cell walls were purified from log phase cultures of S. pneumoniae (step 1). Purified walls were digested with amidase (step 2), which hydrolyzed the bonds between the glycan chain and the stem peptides. Note that amidase hydrolyzes these bonds only in the presence of an adjacent teichoic acid (black circle) (34). This digestion resulted in soluble stem peptides with various degrees of cross-linkage and glycan chains that still contained a few peptides. Amidase-digested walls were then treated with muramidase (step 3A), which resulted in the breakdown of the glycan chains. This preparation was submitted to reverse phase HPLC separation (step 4A) and TNF stimulation assay (step 5A). In separate experiments, amidase-digested walls were also separated into their peptide and glycan constituents by serial precipitation (step 3B). Soluble peptides were then separated by HPLC and analyzed for TNF-stimulating activity as described (steps 4A and 5A). On the other hand, the glycan-enriched fraction was submitted to muramidase digestion (step 4B) and tested in the TNF-stimulation assay (step 5B).

FIG. 3. Separation of soluble cell wall components by HPLC.
Equal amounts (100 g) of the following pneumococcal wall preparations were loaded onto a C18 Sephasil column: purified walls digested with amidase alone (a), purified walls digested with both amidase and muramidase (b), and stem peptides prepared by sequential precipitation of amidase-digested walls (c) (see Fig. 2). HPLC failed to resolve cell walls solubilized by amidase alone (a). In contrast, the amidase/ muramidase, double-digested wall (b) and the peptide-enriched fractions (c) could be resolved and gave reproducible patterns that were very similar in both cases. The insets show the putative composition of the unseparated mixture. tion of stimulants necessary to release at least a 10-fold increase in the TNF levels above background (defined as the minimal stimulatory concentrations 10ϫ, or MSC 10ϫ ). It can be seen that the most active peaks eluted between 38 and 60 min. Additional material eluting until 105 min was also active but less well resolved. Together, these active fractions of the chromatogram represented only 2% of the total material. However, their specific (w/w) activity approached that of LPS.
For instance, the fraction containing the inactive peak 5 (Fig.  4) represented ϳ150 ng/ml wall material, but this concentration released only 500 pg/ml TNF in the cytokine stimulation assay. In contrast, the active fractions containing peaks 13 and 20 (Fig. 4) represented only ϳ27 and 23 ng/ml wall material, respectively, but these concentrations triggered the release of Ͼ3000 and 1800 pg/ml TNF from PBMCs. This indicated a difference of at least 20 -30-fold in the specific activity of these two types of material.
On the other hand, material eluting before 37 min was poorly active (peaks 1-8 in Table I), and represented another 43% of the loaded material. The rest (45%) of the loaded material eluted after 105 min and was essentially inactive. Taken together, these results indicated that the most active fractions in the solubilized pneumococcal were found in only 2% of the material.
Molecular Characterization of Specific HPLC Fractions-The HPLC fractions corresponding to peaks 1-21 were further analyzed for their molecular content by amino acid and amino sugar analysis and for the molecular weights of their components by MS. All of the HPLC fractions contained the basic pneumococcal wall amino acids alanine, glutamine, and lysine. Some peaks also contained lower amounts of serine, glycine, and aspartic acid, which are amino acids previously reported to be present in the pneumococcal wall (41). No aminosugars were found, confirming that the present experiments preferentially analyzed cell wall stem peptides.
The MS analysis is presented in Fig. 6. It indicates that most fractions contained more than one component. Moreover, the M r of these components increased with the retention time, from 600 to 950 in peaks 1-8 (retention time 15-37 min), to 950 -1500 in peaks 9 -21 (retention time 38 -55 min). This MS spectrum was similar when run with five consecutive HPLC separations, three analyzing double-digested walls (step 3A in Fig. 2) and two analyzing peptide-enriched materials (step 3B in Fig. 2).
On one occasion, three fractions (peaks 5, 13, and 20 in Fig.  4) gave a single major ion by MS (Fig. 7), thus allowing further characterization by postsource decay analysis. Table II presents the molecular structures of these components as tentatively deduced from (i) their content in amino acids, (ii) their measured and calculated molecular weights, and (iii) their postsource decay spectrum. Peak 5, which carried a very low inflammatory activity (Table I and Fig. 4), consisted of a dipeptide previously described in the pneumococcal wall (42). Peaks 13 and 20, which carried TNF-releasing activities almost equal to that of LPS (Table I and Fig. 4), consisted of two tripeptides. One of them (peak 20) resembled a previously described structure (43), whereas the other (peak 13) was new.
TNF-releasing Activity of Glycan-enriched Fractions-In the present experiments, the double-digested walls (step 3A in Fig.  2) and the peptide-enriched materials (step 3B in Fig. 2) gave very similar HPLC and MS.
This indicated that the present chromatography conditions preferentially resolved the soluble wall peptides and did not provide much information on the activity of the glycan moiety contained in the original material. To address this question more specifically, the glycan-enriched fraction released during amidase digestion was precipitated as described (step 3B in Fig. 2) and further tested for its TNF-releasing activity before and after digestion with muramidase.
It is noteworthy, however, that this fraction did not only contain the glycan backbone but also some residual peptides that invariably escape detachment by amidase (41). Table III presents the type of material tested and the results of this experiment. First, insoluble and amidase-digested walls had comparable TNF-releasing activities as already described. Sec-  (Table II). The dashed line shows the retention time cut-off above which high TNFstimulating activity was observed, as depicted in Fig. 4. The many components found at the beginning of the chromatogram indicate that this fraction contained the flow-through.
ond, the glycan-enriched fraction had an unaltered TNF-releasing activity as compared with insoluble and amidase-digested walls. Third, muramidase-treatment of this glycan-enriched material resulted in a 100-fold decrease in its specific TNFreleasing power. These results indicated a role for the integrity of the glycan backbone to ensure TNF-releasing activity. DISCUSSION The present results unraveled the existence of very active proinflammatory stem peptides in the pneumococcal peptidoglycan by digesting it with its natural autolytic enzyme amidase. Moreover, quantitative and structural analysis of these stem peptides indicated that a minimal degree of crosslinking between them was necessary to confer TNF-releasing activity. Indeed, one dimeric form of branched peptides (peak 5 with a M r of 744 in Fig. 3, b and c), which represented a large fraction on the chromatogram, carried no activity. Moreover, material with a lower molecular weight did not carry inflammatory power either. This suggested that simpler structures such as monomeric stem peptides, which also exist in the pneumococcal wall (44), were not inflammatory. This was in agreement with Timmerman et al. (40), who did not detect TNF-stimulating activity with the synthetic monomeric peptide L-alanyl-D-isoglutaminyl-L-lysyl-D-alanyl-D-alanine acetate. On the other hand, wall components with a molecular weight greater than 1000 were highly proinflammatory. Several of them had specific (w/w) activities Ն100-fold greater than that of the total wall and almost equal to that of LPS. Two of them were structurally characterized and identified as branched tripeptides.
In the present experiments, the very active wall fragments released by digestion with amidase or digestion with amidase plus muramidase represented only ϳ2% of the total soluble material. However, this 2% of material was 10 -100-fold more active than the 98% of the remaining wall. Therefore, it was tempting to speculate that this 2% fraction comprised the totality of the active wall subcomponents. Although plausible, this hypothesis might greatly oversimplify the reality. Indeed, the wall of Gram-negative bacteria is complex and may contain different types of proinflammatory subcomponents. During enzymatic digestion, it is possible that certain of these inflammatory constituents were destroyed, while others (such as branched stem peptides) were revealed by making them accessible to the target cells.
For instance, Dokter et al. (27) reported that the disaccharide tetrapeptide from E. coli peptidoglycan was almost as powerful as LPS as an inducer of interleukin-6 and -1␤ secretion by PBMCs. In the present experiments, we did not find such  Table  II.  structures in the pneumococcal wall. However, it is noteworthy that putative pneumococcal disaccharide tetrapeptides could have been destroyed during amidase digestion, because this enzyme hydrolyzes the bonds between N-acetylmuramic acid and the stem peptide. In this case, they would not have been revealed in the present experiments. Alternatively, it is also possible that disaccharide tetrapeptides of pneumococci were less active than those of E. coli. Indeed, differences exist between these two constituents of the wall. In many Gram-positive bacteria, such as pneumococci and staphylococci, the third amino acid in the stem peptide is an L-lysine, whereas this residue is a diaminopimelic acid in E. coli. Whether this subtle difference might affect the proinflammatory activity of these molecules is as yet undetermined. However, Kengatharan et al. (45) recently reported that the closely related disaccharide pentapeptide of S. aureus was poorly proinflammatory on its own. A synergism between lipoteichoic acids and disaccharide pentapeptide was required for this staphylococcal component to be active.
The proinflammatory potential of cross-linked peptidoglycan stem peptides has never been tested before. Nonetheless, an indirect clue for the importance of stem peptide cross-linkage may be drawn from previous experiments, utilizing another soluble wall material that held stem peptides close together. This material, so-called soluble peptidoglycan, was released during penicillin treatment of staphylococci (23) and thus consisted of disaccharide-pentapeptide polymers linked together via 1-4 ␤-glycosidic bonds, but not interpeptide bridges (due to the action of penicillin). Staphylococcal soluble peptidoglycan approached/ resembled somewhat the pneumococcal glycan-enriched fraction schematized in Fig. 2 (step 3B). Both of them carried intrinsic inflammatory activity. However, both of them lost their inflammatory power after destruction of the glycan chain with muramidase, a process that also disconnected the stem peptides.
While these experiments indicated the importance of glycan chain integrity to ensure a proinflammatory activity, they also might help reconcile previous results with soluble peptidoglycan and the present observations with cross-linked peptides under a common concept. It is possible that in order to stimulate target cells, both soluble peptidoglycan and cross-linked peptides must present at least three stem peptides in close proximity. Below this level of polymerization, peptidoglycan fragments are inactive. Above this level of polymerization, they are active. This hypothesis does not exclude the existence of other inflammatory components in the wall, and/or the existence of components that might interact in synergism with lesser active fragments, as described by Kengatharan et al. (45). However, if true, this hypothesis would provide an important clue as to the structure-activity relationship of some very active constituents of the Gram-positive peptidoglycan. Further studies with synthetic peptides or disaccharide peptide oligomers will be necessary to confirm or disprove this model.
In conclusion, the observation that complex peptidoglycan branched peptides carried high TNF-stimulating activity was novel. The extent of the existence of these peptides occurring freely during infection remains to be seen. Nonetheless, amidase-induced autolysis has been implicated in the pathogenesis of pneumococci (31). The fact that trimeric forms of these peptides were a minimal requirement for activity might provide a clue as to the structural constraint necessary for target cell stimulation. These types of cross-linked peptides are likely to be conserved in the walls of many different Gram-positive pathogens. Moreover, although some intrinsic variation may exist between different bacteria, these cross-linked peptides might be structurally close enough to obey a common stimulatory pathway of target cells. Hence, they might represent a common target for potential anti-inflammatory agents. The existence of such a common stimulatory pathway and of freely circulating cross-linked peptides in infections is currently under investigation.