Introduction

Abstract Introduction References

Human endotoxemia and sepsis are characterized by the presence in the circulation of various proinflammatory mediators. Blood leukocytes respond to bacteria or bacterial products by secreting various substances, such as cytokines, chemokines, enzymes, oxygen, and nitrogen radicals, which are major mediators responsible for the hypotension and organ failure that characterize sepsis (1, 2).

Among the enzymes secreted by leukocytes, matrix metalloproteinases (MMPs) play an important role in the pathogenesis of several inflammatory diseases (3, 4). MMPs are a family of zinc-containing proteinases with different substrate specificities, cellular sources, and inducibility (4, 5). These enzymes are secreted as inactive proenzymes (or zymogens) and are autoactivated or activated by other proteolytic enzymes on site, resulting in the digestion of collagen. Similar to other enzymatic systems, potent natural enzyme inhibitors (tissue inhibitors of matrix metalloproteinases [TIMPs]) block activated MMPs (6). Thus, collagenase activity depends on the balance between the amount of activated MMPs and the amount of TIMPs.

MMPs play an important role in the migration of tumor and inflammatory cells across the extracellular matrix, as well as in the process of remodelling of tissues (4, 7, 8). Class IV collagenases (MMP-2 and MMP-9, gelatinases A and B, respectively) degrade denatured, type IV collagen. MMP-9 appears to be important for the migration of polymorphonuclear neutrophils (PMNs) across basement membranes in response to the chemoattractant formylmethionylleucylphenylalanine (FMLP) (9).

In this study, we show that various proinflammatory mediators, including endotoxin (lipopolysaccharide [LPS]), tumor necrosis factor- (TNF-), interleukin (IL)-8, and granulocyte colony-stimulating factor (G-CSF) induce a rapid degranulation of PMNs in whole blood with secretion of proMMP-9. Similar results were observed in human subjects studied during experimental endotoxemia. Increased gelatinase activity was found in plasma from patients with sepsis.

	Materials and Methods

Whole Blood Assay

Whole blood from healthy human volunteers was drawn on the day of the assay and heparinized (heparin sulfate 10 U/ml; Liquemine; Roche Pharma, Reinach, Switzerland). For each experiment, blood from two to three different volunteers was assayed in parallel. One milliliter of whole blood diluted 1:2 with RPMI 1640 (Gibco, Basel, Switzerland) was incubated at 37°C for 1 h before stimulation. Escherichia coli O111:B4 LPS (List Biological Laboratories, Campbell, CA) was diluted in RPMI 1640 containing 0.5 mg/ml low-endotoxin human serum albumin (Miles Laboratories, Kankakee, IL), added to the blood (less than 1/10 of the final volume), and incubated for various durations. Final LPS concentration was 50 ng/ml. Stimulated blood was then put on ice, and plasma was collected by centrifugation at 4°C and stored at 70°C. In one experiment, the following inhibitors were added 90 min before LPS stimulation: cycloheximide (Sigma, Basel, Switzerland), herbimycin A (LC Laboratories, Luzerne, Switzerland), dexamethasone (Organon, OSS, The Netherlands), and anti-CD14 monoclonal antibody (mAb) 28C5 (a gift from D. Leturcq, R. W. Johnson Pharmaceutical Research Institute, San Diego, CA). A total of 50 ng/ml LPS was then added for 2 h. In other experiments, the following agonists were tested in both time- and dose-dependent manner; recombinant human TNF- (a gift from G. E. Grau, University Hospital, Geneva, Switzerland), recombinant human IL-8 (kindly provided by T. N. Wells, Glaxo Institute for Molecular Biology, Geneva, Switzerland), phytohemagglutinin (PHA; Sigma), and recombinant G-CSF (Roche Pharma). The TNF-, IL-8, PHA, and G-CSF were found free of endotoxin by the Limulus lysate assay (KabiVitrum, limit of detection 10 pg/ml of equivalent of E. coli endotoxin).

Injection of Endotoxin to Human Volunteers

Plasma samples from 13 normal human young subjects who had participated in experimental endotoxemia studies were evaluated. All subjects were healthy and had normal electrocardiograms, chest radiographs, and blood and urine analyses. These studies were approved by the National Institutes of Health institutional review board on human experimentation, and written consent was obtained from each subject. Study participants were evaluated in the medical intensive care unit after fasting overnight and were given maintenance fluids intravenously.

Six subjects were randomized to receive ibuprofen (800 mg orally at 1.5, 0, and 1.5 h, n = 3) or placebo tablets (n = 4). The same subjects served as their own controls and were given endotoxin (E. coli O113, 4 ng/kg intravenously; Food and Drug Administration, Bethesda, MD) or saline in single-blinded random order separated by 1 wk. Six additional subjects have been studied and described in part in a recent study of the effects of recombinant dimeric TNF (p75) receptor (TNFR:Fc; Immunex, Seattle, WA) during experimental endotoxemia (10). These subjects were given 10 mg/m² of TNFR:Fc, and in contrast to control subjects given endotoxin had no detectable TNF bioactivity after endotoxin administration.

Isolation of Blood Leukocytes

Eight milliliters of 0.12 M sodium citrate and 20 ml of 4% dextran made in 9 g/liter NaCl were sequentially added to 32 ml of fresh whole human blood from healthy volunteers in a 60-ml syringe. Red cells were allowed to decant for 45 min with the syringe in the upright position. The upper leukocyte/plasma mixture fraction was then separated and collected by centrifugation. The leukocyte pellet was resuspended in 5 ml plasma and overlaid onto 5 ml Ficoll (Pharmacia, Alameda, CA). After centrifugation at 400 × g for 30 min at 21°C, the peripheral blood mononuclear cell (PBMC) band and the PMN cell pellet were separated. Each fraction was resuspended in 20 ml RPMI 1640 containing 10% fetal bovine serum (RPMI/FBS; Gibco), and centrifuged at 350 × g for 7 min at room temperature. Finally, PMNs and PBMCs were resuspended in 12 ml of RPMI/FBS; distributed in 24-well plates (Costar, Cambridge, MA), 1 ml per well; and incubated 1 h at 37°C. Typically, PMNs were > 98% pure and PBMC contained 25 to 30% monocytes and 70 to 75% lymphocytes. LPS was then added to the cells at various concentrations and for various times. Conditioned supernatants were collected by centrifugation and stored frozen at 70°C.

Gelatin Zymography

Gelatin zymography was performed according to published methods (9, 11) with minor modifications. Briefly, plasma samples or conditioned supernatants were resuspended in nonreducing 5× sodium dodecyl sulfate (SDS) sample buffer (0.4 M Tris, pH 6.8; 5% SDS; 20% glycerol; bromophenol blue 0.05%) and loaded onto a 15-cm long 8% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) gel in which 0.1% pork skin gelatin (Sigma) was copolymerized. Typically, the equivalent of 3.5 µl of human plasma or 1 µl of conditioned supernatants were loaded per lane. The gel was run at a constant 40 mA until the blue dye had reached the bottom of the gel. The gel was then removed and incubated at room temperature with 100 ml 2.5% Triton-X 100 (Sigma) for 1 h, with one change at 30 min. The gel was washed once with the digestion buffer (50 mM Tris, pH 7.5; 200 mM NaCl; 5 mM CaCl₂) and then incubated in 100 ml of this buffer at 37°C for 18 h in a shaking water bath. Gel staining and destaining were done at room temperature, using a 0.5% coomassie brilliant blue R-250/ 30% methanol/10% acetic acid solution for staining and a 30% methanol/10% acetic acid solution for destaining, with frequent changes of the destaining solution. A digested zone at 92 kD molecular weight (white bands on a blue background) appeared when progelatinase B (proMMP-9) was present in the sample, and the digested 72-kD-band corresponded to progelatinase A (proMMP-2). ProMMPs are not enzymatically active in the sample, but these zymogens can be detected by gel zymography because the procedure activates them, allowing in-gel gelatin lysis. An additional 130-kD band was present when 92 kD proMMP-9 was present in large quantity. This 130-kD band corresponded to proMMP-9 bound to a neutrophil gelatinase-associated lipocalin (12). The gels were then scanned and imaged in black and white, and gelatinolytic bands were quantified using the softwares ImageQuant (Molecular Dynamics, Basel, Switzerland) and Adobe Photoshop (Adobe Software, Seattle, WA). Densities were expressed as arbitrary density units. MMP-9 activity in conditioned plasma from whole blood treated ex vivo with 50 ng/ml LPS for 2 h served as a positive control (= 100% gelatinolytic activity) for measurements of MMP-9 activity in LPS-treated human volunteers' plasma.

Western Blotting

For the experiment with PMNs stimulated with LPS, proteins from conditioned supernates (25 µl) were separated using an 8% SDS-polyacrylamide gel, in reducing (5% 2-mercaptoethanol) and nonreducing conditions. Proteins were then electrotransferred onto nitrocellulose paper. Nitrocellulose membrane was blocked overnight in 5% nonfat dry milk/pH 7.4 Tris-buffered saline (NFDM/TBS) at 4°C, and then incubated with a 1/500 dilution of rabbit antihuman MMP-9 (Anawa, Zurich, Switzerland) in 0.5% NFDM/TBS containing 0.1% Tween-20 for 90 min at 37°C, followed by a 1/10,000 dilution of an affinity-purified peroxidase-conjugated donkey antirabbit immunoglobulin G (Jackson ImmunoResearch Laboratories, West Grove, PA) in the same buffer for 45 min. The membrane was then extensively washed with TBS/0.1% Tween-20, incubated 3 min with enhanced chemiluminescence reagents (Amersham, Zurich, Switzerland), and autoradiographed. Proteins were eletrophoresed in reducing and in nonreducing conditions to test the specificity of the anti-MMP-9 antibody in this Western blot technique. In reducing conditions, the lipocalin-proMMP-9 130-kD complex originating from neutrophil granules should be dissociated, and only one 92-kD band should be detected, whereas in the absence of -mercaptoethanol both the 130- and 92-kD bands should be detected. Densities of the bands were measured, quantified, and expressed as arbitrary density units using the same methods as for gelatin zymography.

Reverse Transcription-Polymerase Chain Reaction

PMNs (~ 10⁶ cells) isolated as described above were stimulated with LPS for different times, and collected by centrifugation at 4°C. Cell pellets were washed once in pH 7.4 phosphate-buffered saline and lysed in 1 ml Trizol (Gibco) containing guanidium chloride and phenol. After the addition of chloroform, total RNA was extracted and treated according to classical methods, with RQ1 DNase (Promega, Madison, WI) to remove any contaminating genomic DNA. cDNAs were obtained with avian Moloney virus reverse transcriptase using the method described by the manufacturer (cDNA Synthesis Kit; Boehringer Mannheim, Indianapolis, IN) and subjected to polymerase chain reaction (PCR) using the following primers: Human MMP-9 3' primer: ACC GCT ATG GTT ACA CTC GG, human MMP-9 5' primer: GCA GGC AGA GTA GGA GCG; human TNF- 3' primer: CCT TGG TCT GGT AGG AGA CG, human TNF- alpha 5' primer: CAG AGG GAA GAG TTC CCC AG; human actin 3' primer: CTC CTT AAT GTC ACG CAC GAT TTC, and human actin 5' primer: GTG GGG CGC CCC AGG CAC CA. PCR consisted of 1 cycle at 94°C for 3 min, 58°C for 30 s and 72°C for 1 min, 30 s, followed by 35 cycles at 94°C for 30 s, 58°C (62°C for MMP-9) for 30 s, and 72°C for 1.5 min. Amplified DNA bands were analyzed using 1% agarose gel containing ethidium bromide under ultraviolet light.

TNF Enzyme-Linked Immunosorbent Assay

In the experiment with inhibitors, TNF- antigen was quantified in conditioned plasma using a "sandwich" enzyme-linked immunosorbent assay with a capture monoclonal antihuman TNF- mAb and a detection system using a biotinylated antihuman TNF- mAb (Pharmingen, San Diego, CA) and horseradish peroxidase-conjugated streptavidin (Zymed, South San Francisco, CA), according to the manufacturer's instructions. A standard curve was done using recombinant human TNF- (a kind gift of G. E. Grau, University of Geneva, Switzerland) diluted in plasma.

Procalcitonin

Procalcitonin levels were measured in patients' plasma according to published methods (13) by C. Bohuon, Villejuif, Paris. Previous work has shown that plasma procalcitonin levels correlate with the severity of human sepsis (14).

Case Reports

A 65-year-old woman with rheumatoid arthritis was admitted with hyperosmolar diabetic coma and E. coli pneumonia. She required antibiotics, mechanical ventilation, and vasopressors. Nine days after admission, she developed hypotension with polymicrobial gram-negative sepsis and underwent a partial hemicolectomy for a perforation of the sigmoid colon. On Day 17 she developed shock, diffuse pulmonary infiltrates, and pancreatitis due to leakage from the colonic anastomosis, and died the following day.

A 67-year-old man with chronic obstructive pulmonary disease and bronchiectasis was admitted with Pseudomonas aeruginosa pneumonia complicated by respiratory failure, shock, and multiple organ failure. Despite antibiotics, vasopressors, and mechanical ventilation, he died 8 d after admission because of refractory shock.

Statistical Analyses

To analyze the repeated measures study in subjects given LPS, a summary measure (area under the curve) was computed for each person, and then analyzed by a nonparametric Wilcoxon's test (15). Data for a single time point were also analyzed using a Wilcoxon's test; however, the P value was not adjusted for multiple time points that may have been selected.

	Results

Whole human blood incubated with LPS resulted in sustained secretion of progelatinase B (proMMP-9, 92 kD). The secretion occurred within 20 min after LPS treatment (Figure 1). The proMMP-9 secretion was blocked by an anti-CD14 mAb (28C5), indicating that the LPS effect was mediated by the myeloid receptor CD14 (Figure 1) (16). A constitutive gelatinolytic band at 72 kD was observed in normal and in conditioned plasma, and correponded to progelatinase A (proMMP-2), which is typically produced by cells other than leukocytes. The intensity of this band was not changed by ex vivo LPS treatment of whole blood. When LPS concentrations were reduced from 50 ng/ml to 1 or 0.1 ng/ml, we observed a delayed secretion of MMP-9 (Figure 2).

View larger version (78K): [in this window] [in a new window]   Figure 1.   Gelatin zymography of conditioned plasma. Upper two gels: whole blood from two normal volunteers without LPS (LPS) or stimulated ex vivo with 50 ng/ml LPS (+LPS) for various times as indicated. The rapid increase in gelatinolytic bands of 92 and 130 kD corresponds to secreted proMMP-9 enzyme (gelatinase B; see text). The constitutive lower 72-kD band corresponds to proMMP-2 (gelatinase A). Lower gel: proMMP-9 secretion in whole blood from normal volunteer stimulated ex vivo with LPS in the presence or in the absence of anti-CD14 monoclonal antibody. A 97-kD molecular weight marker is to the right of the gels.

View larger version (56K): [in this window] [in a new window]   Figure 2.   Gelatin zymography of conditioned plasma. Low doses of LPS (0.1 and 1 ng/ml) induce secretion of proMMP-9 in whole blood from normal volunteers with a delay as compared with the dose of 50 ng/ml (see Figure 1).

Because both neutrophils and monocytes can secrete MMP-9, we fractionated blood leukocytes and stimulated neutrophils separately from PBMC. As shown in Figure 3, the neutrophil was identified as the blood cell responsible for the LPS-induced MMP-9 secretion. No gelatinase activity could be found in PBMC treated with LPS during that time period (0 to 4 h). We confirmed the nature of the gelatinase identified by zymography as being MMP-9, using a Western blot technique on stimulated neutrophil supernatants (Figure 4). We found an upregulation of a 92- and a 130-kD band reacting with the anti-MMP-9 antibody (unreduced condition), which corresponds to the same gelatinolytic bands as those observed in zymograms. After reduction of the samples, a unique 92-kD band was detected. The intensities of the bands were quantified, and we confirmed that proMMP-9 secretion started after 20 min and was maximal at 30 to 60 min. Monocytes required  7 h of LPS treatment to start to secrete proMMP-9 (data not shown). The monocytic proMMP-9 secretion was maximal after 24 to 48 h, with gelatinolytic band intensities always less pronounced than those of neutrophils.

View larger version (45K): [in this window] [in a new window]   Figure 3.   Gelatin zymography of conditioned supernatants from PMN and PBMC purified from whole blood of normal volunteers after stimulation with 50 ng/ml LPS for various times.

View larger version (52K): [in this window] [in a new window]   Figure 4.   Western blot of conditioned supernatants from purified PMN from normal volunteers stimulated ex vivo with 50 ng/ml LPS for various times. (Left panel) Samples in unreduced condition. (Right panel) Samples in reduced conditions, revealing a unique 92-kD band (see text for details). Band densities were measured with a densitometer and expressed as arbitrary density units.

The rapid secretion of proMMP-9 by neutrophils suggested that the enzyme was stored in neutrophil granules and was not the result of de novo protein synthesis. We addressed this point by analyzing the MMP-9 mRNA production in purified neutrophils stimulated with LPS (50 ng/ml), using a reverse transcription-PCR technique. No MMP-9 mRNA was detected in stimulated neutrophils from 0 to 90 min, whereas TNF- mRNA was upregulated at 30, 60, and 90 min, and actin transcript was present at equal quantities over the entire time period tested (data not shown). Figure 5 shows that the addition of cycloheximide did not influence proMMP-9 enzyme secretion, whereas it blocked TNF- antigen production in whole blood. In the latter experiment, we tested three additional inhibitors of molecules from the "classic" LPS signal transduction pathway (17). Only anti-CD14 blocked LPS-induced proMMP-9 production in whole blood. Neither the phosphotyrosine inhibitor herbimycin A nor dexamethasone (an inhibitor of nuclear factor (NF)-B nuclear translocation) was able to influence the LPS-induced proMMP-9 secretion. In contrast, these inhibitors at least partially blocked TNF- production in whole blood (Figure 5).

View larger version (73K): [in this window] [in a new window]   Figure 5.   Gelatin zymography of conditioned plasma from whole blood from two different volunteers stimulated ex vivo with LPS (50 ng/ml) or PHA (10 µg/ml). Effect of various inhibitors on LPS-induced proMMP-9 secretion and TNF production measured in the same plasma (TNF level indicated = mean of the two measurements for each condition).

We then addressed whether other neutrophil stimulants such as TNF-, IL-8, PHA, or G-CSF had the same effect as LPS on neutrophil secretion of MMP-9. In Figure 6 (PHA in Figure 5), it is shown that all these agonists were able to induce a rapid (within 10 to 20 min) proMMP-9 secretion when incubated in whole blood, probably because of neutrophil degranulation. Dose-response studies demonstrated that concentrations as low as 100 pg/ml of TNF- or IL-8, or 100 U/ml of G-CSF were able to induce this effect (not shown).

View larger version (61K): [in this window] [in a new window]   Figure 6.   Gelatin zymography of conditioned plasma from whole blood stimulated ex vivo with IL-8 (100 ng/ml), G-CSF (30,000 U/ml), and TNF (100 ng/ml) for various times.

We next analyzed plasma samples from the subjects given intravenous endotoxin and found that increased levels of proMMP-9 occurred as early as 60 min after LPS administration, with peak levels occurring at 90 min to 3 h (P = 0.0009) (Figure 7). This pattern of initial release of proMMP-9 is similar to that observed in whole blood treated ex vivo (Figure 2). The administration of ibuprofen, a cyclooxygenase inhibitor, prior to endotoxin did not alter this pattern of secretion. In contrast, TNFR:Fc given prior to endotoxin administration delayed the initial rise of proMMP-9 at 60 min (P = 0.053) compared with endotoxin- and endotoxin/ibuprofen-treated subjects, and blunted the overall response to endotoxin by approximately 40% (P = 0.033) (Figure 7).

View larger version (24K): [in this window] [in a new window]   Figure 7.   Time course of proMMP-9 secretion in human volunteers injected with saline and then LPS (n = 4) or saline (n = 3). Effect of pretreatment with soluble TNFR:Fc (n = 6) and ibuprofen (IBU, n = 3). Gelatin zymograms were done with plasma samples from these volunteers, corresponding to different times before and after injections. Gelatinolytic bands of 92 kD were quantified and expressed as percentage of a control (conditioned plasma from a normal volunteer's whole blood treated ex vivo with 50 ng/ml LPS = 100% density). LPS injection significantly increased proMMP-9 secretion (P = 0.0009). TNFR:Fc given prior to LPS administration significantly decreased LPS-induced proMMP-9 secretion (P = 0.033).

Plasma from the two patients with gram-negative sepsis showed increased gelatinase activity as determined by gelatin zymography (Figure 8). In addition to increased proMMP-9 band intensities, we found that the 82- and 62-kD activated forms of MMP-9 were also present in plasma from these patients. Western blot analysis confirmed the MMP-9 origin of the bands (not shown). Increased proMMP-2 activity was also detected in both patients. The increased gelatinase activity paralleled the levels of procalcitonin measured in those samples, as well as the degree of severity of the septic state.

View larger version (62K): [in this window] [in a new window]   Figure 8.   Gelatin zymography of plasma and procalcitonin levels in plasma from two patients with gram-negative sepsis during their clinical course. Clinical events are indicated below. C+ = positive control corresponding to conditioned plasma from a normal volunteer's whole blood stimulated ex vivo with 50 ng/ml LPS; C = negative control corresponding to unstimulated plasma from the same subject.

	Discussion

Here, we show that gelatinase B zymogen (proMMP-9) is rapidly released into the blood after contact with various proinflammatory mediators. MMP-9 can be produced by PMN and monocytes/macrophages (5, 18). Whole-blood fractionation experiments suggest that the major cellular source for rapid proMMP-9 secretion in whole blood is the PMN. More than 7 h was required to detect low levels of proMMP-9 secretion in supernates from stimulated monocytes. In monocytic cells, it was shown by others that there was no stock of preformed enzyme, and the regulation of MMP-9 secretion depended on de novo protein synthesis (19). In contrast, the PMN response to agonists was very rapid (within 10 to 20 min), and inhibition of protein translation by cycloheximide had no effect on proMMP-9 secretion. In addition, the presence of the 130 kD lipocalin-proMMP9 complex in nonreduced SDS-PAGE zymography indicates a PMN origin for this gelatinase, as lipocalin is not present in monocytes (12, 20). These data indicate that proMMP-9 originated from PMN degranulation of preformed enzymes and was not the result of de novo protein synthesis. This was confirmed by demonstrating the absence of an upregulation of MMP-9 mRNA up to 90 min after LPS stimulation. In addition, inhibitors such as herbimycin A (a phosphotyrosine kinase inhibitor) and dexamethasone (an NF-B inhibitor), which have an inhibitory effect of gene activation in response to endotoxin, did not prevent pro-MMP-9 secretion. Only antibody blockade of the agonist-receptor interaction (LPS-CD14) prevented the secretion of the enzyme. This suggests that early signal transduction events take place after ligand binding, with intracellular mediators different from those of the usual pathway leading to cytokine production after LPS ligation to its receptor (17). This is in accordance with studies by different groups who showed that gelatinase B was present in two different types of PMN granulesthe specific and the gelatinase granules (20, 21)and that it could be secreted at the site of inflammation (22). Such a rapid PMN response to LPS was previously demonstrated by Wang and coworkers (23). These authors showed that lactoferrin, another protein present in specific PMN granules, was rapidly released after incubation with endotoxin.

We found that proMMP-9 secretion was not restricted to PMN stimulation by LPS or by the chemoattractant FMLP (9). Other important inflammatory mediators, such as IL-8, TNF-, and G-CSF, were also very potent in producing this effect in PMN assayed in whole blood. This corroborates earlier reports by Masure and colleagues (24) and Kaufman and coworkers (25), who found that IL-8 and GM-CSF induced neutrophil degranulation and gelatinase secretion. The common cellular response of proMMP-9 secretion occurred with agonists that act through very different intracellular signalling pathways. It is not known whether this is the result of converging pathways (a unique mediator for the different receptors) or parallel pathways (different mediators with similar effects on the granules) (26).

Human volunteers injected with LPS showed similar intravascular MMP-9 secretion to that seen with experiments performed with whole blood ex vivo. Soon after LPS injection (60 to 90 min), proMMP-9 could be detected on zymograms of plasma samples. This may have resulted from a direct effect of low-picogram LPS concentrations inducing intravascular PMN degranulation (27) or from early cytokine release at the tissue level. In previous studies of human endotoxemia, TNFR:Fc inhibition of circulating TNF bioactivity blocked neutrophil margination at 1 h, diminished lactoferrin levels, and suppressed the release of IL-8 and G-CSF (10). In the current study, TNFR:Fc decreased the initial rise (60 min, P = 0.053) and significantly blunted the peak response of MMP-9 release by approximately 40% (P = 0.033). These data demonstrate that there are multiple overlapping pathways in vivo that mediate the early release of MMP-9, including neutrophil margination and proinflammatory cytokines such as TNF, IL-8, and G-CSF.

In whole-blood experiments and in normal subjects given endotoxin, only the inactive proenzyme is secreted. Collagenases or serine proteases are likely to be rapidly inactivated by plasma protease inhibitors. In contrast, the adherent PMN prior to transmigration creates a plasma-free microenvironment, and the simultaneous release of elastase and proMMP-9 can result in activation of MMP-9, which is required for basement membrane digestion during transcytosis (9).

In contrast to our observations in experimental endotoxemia, the patients with sepsis had increased plasma activity of both proMMP-9 and proMMP-2, which suggests that cells other than leukocytes (e.g., endothelial cells, a putative source for proMMP-2) were activated during sepsis. Furthermore, activated (proteolytically cleaved) forms of MMP-9 were detected in the plasma from the septic patients. It is unknown whether MMP-9 activation occurred in the blood or extravascular compartments or whether the circulating MMPs are biologically active or complexed with TIMPs. Activated MMP-9 forms could also originate from enzymes spilled over by neutrophils migrating through capillaries. Nonetheless, MMP-2 and -9 were clearly augmented in plasma from septic patients, and their levels seemed to parallel closely other recognized markers of sepsis and the severity of sepsis, such as plasma procalcitonin levels (14). However, the exact function of proMMP-9 secretion by PMN in response to proinflammatory mediators remains to be determined. Additional studies are also needed to determine the cellular origin and the relevance of these enzymes in sepsis.