INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sepsis is a hyperdynamic syndrome characterized by maldistribution of blood flow (1, 2). Disturbances in microvascular flow appear to play a major role in contributing to tissue hypoperfusion in this syndrome (3). The decrease in effective nutrient blood flow in sepsis may be related to loss of capillary cross-sectional area resulting from the endothelial inflammation and changes in cell rheology that occur during sepsis (6). Rheologic changes in patients with severe sepsis and septic shock include increased erythrocyte (red blood cell; RBC) and neutrophil (polmorphonuclear neutrophil; PMN) aggregation, impaired RBC and PMN deformability, increased PMN- endothelial cell adherence, and increased plasma viscosity (7- 9). In addition, platelet aggregation and the formation of platelet-PMN aggregates may contribute to microvascular obstruction (in sepsis 10, 11).

Cell filtration through 5-µm pore-size polycarbonate filters is an in vitro experimental model for assessing cell movement through capillary-size apertures (12, 13). Recording of the filtration pressure (Pi) with time during constant flow reflects the plugging of filter pores by blood cells. The resistance (pressure:flow ratio) of the suspension relative to that of the suspending medium provides a measure of cell deformability (13).

The purpose of this study was to examine the effects on cell filtration of changes in whole blood and plasma viscosity, RBC aggregation and deformability, PMN deformability, and platelet-PMN interactions during sepsis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

The study was approved by the Institutional Research Board of Saint Vincents Hospital and Medical Center. Informed consent was obtained from either the patient or a surrogate. We studied normal, healthy control subjects (C), critically ill patients requiring mechanical ventilation who were not infected or in shock (CINS), and critically ill septic patients (CIS). CIS met the following criteria: (1) A positive blood culture or an identifiable site of infection. (2) Evidence of a systemic inflammatory response, as manifested by a combination of any three of a temperature greater than 38.3° C or hypothermia less than 35.5° C; a respiratory rate above 20 breaths/min; a heart rate above 90 beats/min; and a leukocyte count > 12,000/mm³. Systemic hypoperfusion, defined by an arterial lactate level > 2 mEq/L, or hypotension, defined by a mean arterial pressure (MAP) < 60 mm Hg despite fluid infusion to a pulmonary artery occlusion pressure (Ppao) > 12 and requiring vasopressors to maintain a systolic arterial pressure above 90 mm Hg (14). All measurements were made within 24 h of a patient's meeting the criteria for entry into the study. Exclusion criteria included blood transfusions in the prior 48 h, presence of human immunodeficiency virus, or corticosteroid administration.

We successively entered six subjects who met the entry criteria as CIS, six CINS, and six C into Group 1. A second group of patients (Group 2) was subsequently added for examination of platelet-PMN and platelet-RBC interactions. Into this group we successively entered 10 CIS, seven CINS, and 7 C.

Experimental Protocol

Using viscometry, we examined whole-blood and plasma viscosity, RBC aggregation, and RBC deformability of all subjects in Group 1. Venous blood samples from all subjects were sent for complete blood counts and mean corpuscular hemoglobin concentration measurements. Fibrinogen measurements were made on blood from CIS. PMN and RBC isolated from whole blood were reconstituted in different combinations, using PMN at concentrations of 5 × 10³/mm³ and 15 × 10³/mm³, and RBC suspended in 5% albumin for filtration experiments, at hematrocrit (Hct) values of 10%, 25%, and 50%. In addition, whole blood samples from all subjects in Group 1 underwent cell filtration. In subjects of Group 2, platelet-PMN interactions were assessed through cell filtration. PMN and RBC were isolated from whole blood and were reconstituted in combinations, using PMN at a concentration of 9 × 10³/mm³ and RBC at an Hct of 25%, with the two types of cells resuspended in platelet-poor plasma (PPP), platelet-rich plasma (PRP), or 5% albumin. Platelet activation and platelet- PMN interactions were examined with flow cytometry.

Isolated PMN and RBC Preparations

Human PMN were collected as described by Haslett and coworkers (15). Percoll (Pharmacia, Piscataway, NJ) was prepared in 0.9% saline as a 9:1 vol/vol Percoll/saline solution. A total of 25 ml of venous blood was collected in sodium citrate-buffered tubes and the blood was then centrifuged to produce PRP in the supernatant and a cell-rich pellet. The PRP was aspirated and centrifuged at 2,500 × g to yield PPP. A volume of 5 ml of 6% dextran (Sigma, St. Louis, MO) and 0.9% saline was added to the cell-rich pellet to produce a total of 30 ml of solution. This collection was mixed and allowed to stand for 30 min, yielding a leukocyte-rich supernatant. The leukocyte-rich plasma was aspirated and centrifuged at 275 × g. The pellet was resuspended in 3 ml of PPP and underlayered successively with 2 ml of freshly prepared 42% Percoll in PPP and 2 ml of 52% Percoll in PPP. The gradients were then centrifuged at 280 × g. The mononuclear cells and platelet-rich layer at the PPP-42% Percoll interface were aspirated. The PMN-rich layer was collected at the 42%-52% Percoll interface and the PMN were washed once in PPP and once in Krebs-Ringer phosphate buffer with 0.2% dextrose (GIBCO, Grand Island, NY). The resulting PMN were found to be pure and viable as determined by Wright's staining and trypan blue exclusion. The cell count of the PMN-rich sample was determined with a Coulter counter (Coulter, Hialeah, FL) and was used to derive the desired PMN concentrations for cell filtration experiments. PMN suspensions were > 95% pure.

RBC were isolated from 20 ml of venous blood obtained from study subjects and anticoagulated with ethylenediamine tetraacetic acid. The venous blood was centrifuged at 3,000 rpm for 10 min at 4° C. The plasma and buffy coat were removed and the RBC were washed twice with phosphate-buffered saline (PBS) at 2,500 rpm for 10 min at 4° C. The buffer was prepared by adding five PBS tablets (Sigma Chemical Co.) to 11 ml of sterile water and 2 ml of D₅₀W, adjusting the pH with a 1 N solution of NaOH. A 1% solution of bovine serum albumin (BSA) was added to complete the buffer. RBC suspensions were prepared at different Hct values for cell filtration and viscometry experiments. RBC suspensions were > 95% pure as determined by Coulter counting and microscopic examination.

Viscometry Measurements

All samples were measured in a Wells-Brookfield Cone/Plate LVTDV-II+ viscometer with a No. CP-42 cone (Brookfield Engineering Laboratories, Inc., Stoughton, MA) at a temperature of 37° C. Viscosity was measured at two different shear rates: 192 s¹ and 1 s¹. Whole-blood viscosity (VWB) was determined at each subject's actual Hct. Plasma viscosity (VPlasma) was measured from plasma isolated from each subject after the subject's blood was centrifuged (Centra MP 4R; International Equipment Co., Needham Heights, MA) for 10 min at 3,000 rpm and the plasma and buffy coat were separated. Measurements were made in triplicate and a simple mean was determined from them. The viscometer was calibrated against a standard fluid prior to measurement for each subject. Viscosity is expressed in centipoises (mPa).

RBC aggregation was determined from RBC suspended in autologous plasma at an Hct of 45%, from the ratio of the viscosity measured at a low shear rate to that measured at a high shear rate. At low shear rates, RBC form aggregates, whereas at shear rates above 100 s¹, blood behaves as a Newtonian fluid. The ratio of low-shear viscosity to high-shear viscosity is used as an index of RBC aggregation (16).

RBC deformability has been measured with cell suspensions at high, low, and normal Hct values. When an a Hct above 60% is used, viscosity measurements at high shear rates are mainly determined by the internal fluid viscosity of the RBC themselves, whereas at low shear rates the viscosity measurements are influenced by membrane properties and cell geometry (16, 17). We measured the deformability of RBC suspended in PBS, at Hct values of 45% and 70%, at shear rates of 192 s ¹ and 1 s¹.

Cell Filtration Measurements

The filtration apparatus used for measuring Pi consisted of polypropylene filter chambers (Millipore, Bedford, MA) and sterile plastic tubing, incubated with 1% organosilane (PCR Inc., Gainesville, FL) for 5 min at 20° C (18). Immediately before use, the chambers and tubing were incubated with 25% heat-inactivated serum for 2 h at 37° C. Polycarbonate filters (Costar, Cambridge, MA) with a pore size of 5 µm were then placed into the filter chambers.

The infusion system consisted of an adjustable-flow infusion pump (KDS Model 210; Stoelting, Wood Dale, IL) attached to siliconized tubing. A pressure transducer and digital monitor coupled to a strip-chart recorder were linked to the system downstream of the pump. The system terminated in the filtration chamber apparatus, to which it was connected via a stopcock that opened into a collection vial. Pi (mm Hg) represents the maximal pressure achieved for each cell suspension at a constant volume and a fixed flow rate of 1 ml/min.

Flow Cytometry

Platelet activation and platelet-PMN interactions were assessed with murine monoclonal antibodies (mAb), all of which were fluorescein conjugated to either phycoerythrin (PE) or fluorescein isothiocyanate (FITC). One such mAb, anti-CD41-PE (Pharmingen, San Diego, CA), is a platelet-specific anti-glycoprotein IIb-IIIa mAb that recognizes this glycoprotein complex on resting and activated platelets. Another mAb, anti-CD63-PE (Pharmingen, San Diego, CA), recognizes a lysosomal glycoprotein expressed on activated platelets and monocytes. The thried mAb, anti-CD66b-FITC (Immunotech, Westbrook, ME), recognizes a glycosyl-phosphatidylinositol-anchored glycoprotein strongly expressed by PMN. A fourth mAb, anti-CD14, is a monocyte-specific mAb. The purities of PMN preparations were monitored with the anti-CD14 antibody to exclude monocytes, and with the anti-CD66b antibody to include PMN.

Platelet activation was assessed as follows: PRP was obtained during the isolation of PMN. Thereafter, PRP aliquots of 5 µl each were added to polypropylene tubes containing 45 µl of conjugated mAb (anti-CD63-PE or anti-CD41-PE) in Tyrode's buffer (0.1% BSA, 0.1% glucose, 2 mmol/L MgCl₂, 137.5 mmol/L NaCO₃, 2.6 mmol/L KCl, pH 7.4). After a 15-min incubation period in the dark at room temperature (23° C), the sample was diluted to 0.5 ml in Tyrodes buffer. Fluorescence-activated cell sorting (FACS) analysis was done within 6 h of sampling with a FACScan flow cytometer (Becton-Dickinson, Mountain View, CA) and Consort 30 (Becton-Dickinson) software. Platelets were identified by forward and side scatter plotting and CD41 positivity. Platelet activation was assessed by CD63 positivity.

Platelet-PMN interactions were assessed as previously described by Gawaz and associates (11). Fifty-microliter aliquots of whole blood were added to tubes containing saturating concentrations of FITC-conjugated mAb (anti-CD63-PE and anti-CD66b-FITC) in 25 µl of Tyrode's buffer, and the resulting mixtures were incubated for 30 min at room temperature (23° C) in the dark. Thereafter, red cells were lysed and fixed with SIGMA lysing reagent (0.2% saponin; Sigma, St. Louis, MO) according to the manufacturer's protocol, and the lysate was kept on ice before analysis. For flow-cytometric analysis, PMN subgroups were identified by size and granularity in the forward-versus side scatter plots and by binding with a subgroup-specific mAb, anti-CD66b. The mean intensity of the anti-CD63-PE immunofluorescence of each PMN subgroup was used as an index of activated platelet adhesion to PMN.

Statistical Analysis

Overall differences between groups were evaluated by analysis of variance, and individual paired comparisons were analyzed post hoc with Fisher's protected least-squares difference test. Results were considered significant at p < 0.05. Data are expressed as mean ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient information and demographics are presented in Table 1. CIS had significantly lower arterial pressures and higher mortality than other subjects in both Group 1 and Group 2 (p < 0.05). All Acute Physiology and Chronic Health Evaluation scores were significantly higher in the CIS than in the CINS group. In subjects in Group 1, Hct values were significantly lower in CIS and CINS than in C (p < 0.05). The PMN count in whole blood was significantly greater in CIS as than in C (p < 0.05). The mean fibrinogen level measured in CIS was 478 ± 77 mg/dl. In Group 1, all CIS required vasopressors, and in Group 2, six of the 10 CIS required vasopressors to maintain MAP > 60 mm Hg after fluid infusion.

Filtration pressures for Group 1 patients are presented in Table 2. The whole-blood peak Pi (Pi_WB) of CIS was significantly greater than that of C or CINS, at 27 ± 6.5 mm Hg versus 7 ± 2.0 mm Hg and 8 ± 0.6 mm Hg, respectively (p < 0.05). The Pi of CIS was also significantly greater than the Pi of C or CINS at all PMN concentrations and at all Hct concentrations examined (p < 0.05). There appeared to be no intragroup differences in Pi at high and low PMN concentrations. However, in C and CINS, the Pi for Hct values of 25% and 50% was significantly greater than the Pi for an Hct of 10% (p < 0.05). At all PMN and RBC concentrations studied in CIS, PMN and RBC rheologic changes appeared to contribute equally to Pi.

Viscometry measurements for Group 1 patients are presented in Table 3. Viscosity of RBC reconstituted to an Hct of 45% was significantly greater for CIS than for C or CINS (p < 0.05), as is consistent with decreased RBC deformabilty. At an Hct of 70% and a high shear rate (192 s¹), viscosity did not differ among the subject groups, whereas at a low shear rate (1 s¹), viscosity was significantly greater in CIS, at 36.5 ± 1.4 mPa, than in C or CINS, at 30.3 ± 2.1 mPa and 27.0 ± 1.4 mPa, respectively (p < 0.05). These data suggest that decreased RBC deformability may be related to changes in RBC membrane properties. RBC aggregation was determined with RBC suspended in autologous plasma at an Hct of 45%, from the ratio of the viscosity measured at a low shear rate to that at a high shear rate. RBC aggregation was significantly increased in CIS, at 5.2 ± 0.2 mPa, as compared with 3.3 ± 0.2 mPa and 3.8 ± 0.2 mPa for CINS and C, respectively (p < 0.05).

A significant difference in the Pi generated from whole blood and that generated from the combination of PMN with RBC (Hct = 25%, PMN = 15 × 10³/mm³) suspended in albumin (27 ± 6.5 mm Hg versus 11 ± 2.3 mm Hg, respectively, p < 0.05) was present in blood from CIS (Figure 1). In contrast, the Pi values generated from whole blood and combined PMN/RBC were not significantly different for either CINS or C. These data suggested that important interactions in septic blood, between platelets and either PMN or RBC, were affecting cell filtration. Studies with Group 2 patients were designed to address these issues.

View larger version (8K): [in this window] [in a new window]   Figure 1.   Filtration pressures (Pi, mm Hg) in six critically ill patients with sepsis (CIS) six critically ill noninfected patients (CINS), and six healthy controls (C). Open bars are whole blood (WB); closed bars are combined neutrophils and erythrocytes (PMN/RBC). Values are given as mean ± SE (*p < 0.05 versus PMN/RBC).

Platelet counts were not significantly different between the study subject groups, at 147 ± 22 × 10³/mm³ in C, 149 ± 30 × 10³/mm³ in CINS, and 92 ± 17 × 10³/mm³ in CIS. Pi values for Group 2 are presented in Table 4. Pi was significantly greater in CIS for all cell types and combinations, whether suspended in PPP or PRP (p < 0.05) (Table 4). There were no intragroup differences in Pi for any group of patients when RBC were added to albumin, PRP, or PPP. The addition of PPP also did not significantly affect PMN filtration in any group of patients. In contrast, when PRP was added to PMN from all patients, Pi increased significantly as compared with that for PMN suspended in albumin (p < 0.05). Significant increases in Pi were observed for all subjects when PRP was added to PMN/RBC combinations, as compared with the Pi values for albumin-suspended cells (p < 0.05). With both isolated PMN and PMN/RBC combinations, the increases in Pi associated with PRP were significantly greater in CIS than in CINS or C (Figure 2, p < 0.05).

View larger version (17K): [in this window] [in a new window]   Figure 2.   Differences in filtration pressures (Pi, mm Hg) between albumin (ALB) and PRP in 10 critically ill patients with sepsis (CIS), seven critically ill noninfected patients (CINS), and seven healthy controls (C). PMN = neutrophils; PMN/RBC = combined erythrocytes and neutrophils. Values are given as mean ± SE (*p < 0.05 versus C; p < 0.05 versus CINS).

Flow-cytometric measurements were made to examine for the presence of activated platelets and platelet-PMN interactions, and for the purity of cell preparations. There was no difference in the percentage or mean fluorescence of platelets isolated from the three subject groups as identified by CD41- PE positivity. However, platelet activation, identified by CD63- PE positivity, was significantly greater in CIS than in CINS or C (mean fluorescence: 39.0 ± 9.0 lfu, versus 18.7 ± 4.0 lfu and 17.1 ± 2.3 lfu, respectively, p < 0.05) (Figure 3). The percentage of PMN with bound, activated platelets as measured by CD63 positivity was 32.9 ± 7.5% for CIS (p < 0.05 versus C), as compared with 28.1 ± 7.3% for CINS and 8.5 ± 1.7% for C. Similarly, the binding of activated platelets to PMN as measured by the CD63-PE mean fluorescence of CD66b-positive PMN was significantly greater for CIS than for C (mean fluorescence: 44.7 ± 3.6 lfu, versus 23.0 ± 4.1 lfu, p < 0.05). CINS demonstrated intermediate values (mean fluorescence: 34.2 ± 5.8 lfu) between those of CIS and C (Figure 4).

View larger version (24K): [in this window] [in a new window]   Figure 3.   Mean anti-CD63 fluorescence (lfu) of activated platelets measured in 10 critically ill patients with sepsis (CIS), seven critically ill noninfected patients (CINS), and seven healthy controls (C). Values are given as mean ± SE (*p < 0.05 versus C; p < 0.05 versus CINS).

View larger version (24K): [in this window] [in a new window]   Figure 4.   Mean anti-CD63 platelet fluorescence (lfu) measured on anti-CD66b-positive PMN from 10 critically ill patients with sepsis (CIS), seven critically ill noninfected patients (CINS), and seven healthy controls (C). Values are given as mean ± SE (*p < 0.05 versus C).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sepsis is a syndrome characterized by maldistribution of blood flow (1, 2). Disturbances in microvascular blood flow appear to play a major role in contributing to tissue hypoperfusion in this setting (3). The impairment of microvascular blood flow and loss of capillary cross-sectional area may be related to impaired vasodilation or to intravascular obstruction resulting from changes in RBC and PMN rheology that occur during sepsis (6, 7).

Our aim was to elucidate the relative contributions of rheologic changes in sepsis to cell filtration. Both RBC and PMN must alter their shape to pass through the smaller systemic capillaries. Cell filtration through 5-µm pore-size polycarbonate filters, which approximate systemic capillary diameter, is a test commonly used to assess red cell and white cell deformability in clinical investigations (11, 12, 18). Less deformable cells increase the overall resistance of the filter by obstructing pores and thereby reducing the number of pores available for flow of the suspension (18, 19). In experimental studies, a correlation has been established between decreased cell filtration and in vivo impairment of microvascular blood flow (20).

The findings in our study are consistent with previous reports of decreased RBC and PMN deformability in severe sepsis and septic shock (6, 21). As in these earlier studies, our data suggest that the decreases in RBC deformability are primarily related to changes in RBC membrane properties, rather than to changes in internal cellular viscosity (21, 22). However, we observed that filtration pressures were not significantly augmented when septic PMN and RBC were combined. Studies with isolated cells suggest that PMN significantly enhance endotoxin mediated decreases in RBC deformability (22, 23).

The formation of cell aggregates may also impair cell filtration and lead to microvascular obstruction (3, 10). An important finding in our study was evidence of increased RBC aggregation in critically ill patients with sepsis. RBC aggregation in the microvasculature has been described in experimental sepsis, but similar observations have not been reported in patients with sepsis (3, 10). Indeed, decreases in RBC aggregation have been reported in sepsis (22). The difference between our results and this latter finding is probably due to the different suspending media that were used. Whole blood or suspension of RBC in plasma is recommended for measurement of RBC aggregation because of the importance of plasma proteins, and particularly fibrinogen, in enhancing RBC aggregation (16, 24). Whereas Voerman and coworkers (21) suspended RBC in a phosphate buffer, we suspended the RBC in autologous plasma and corrected for baseline differences in Hct. In addition, decreases in RBC deformability associated with sepsis may have contributed to increased RBC aggregation (24).

This study confirms previous observations of decreased PMN filtration in sepsis. Both decreases in PMN deformability and increases in leukoaggregation have been reported to contribute to this decrease (7). Our data suggest that changes in PMN and RBC rheology appeared to contribute equally to increases in Pi during sepsis. No significant increments in Pi were associated with increasing the PMN count from 5,000/ mm³ to 15,000/mm³. This observation is similar to that of Tanner and associates (25) that PMN counts above 25,000/mm³ were required to impair PMN filtration. In addition, for septic patients, we did not observe a significant reduction in Pi among Hct values of 50%, 25%, and 10%, suggesting that decreases in RBC deformability and increases in RBC aggregation offset the decreases in Pi expected with hemodilution (26).

An unexpected finding in our first group of septic patients was the marked difference between Pi in whole blood and with the PMN/RBC combinations (Figure 1). This difference was not observed in C or CINS. Moreover, the magnitude of the increment was greater than the combined effects of the rheologic changes produced by RBC and PMN. Since the initial study was done in albumin suspension, we hypothesized that either platelet and/or plasma interactions with PMN and RBC were significantly affecting cell filtration. The data for Group 2 patients show that an important interaction occurs between PMN and platelets in sepsis, and that it affects cell filtration. Adding PRP to PMN and combinations of PMN and RBC significantly increased filtration pressure for all groups of subjects. The increase in pressure that occurred was significantly greater for septic patients than for noninfected patients or control subjects, and may have been somewhat underestimated by virtue of the lower platelet count in the septic patients. Plasma alone did not affect either PMN or RBC filtration. Similarly, there was no significant effect on filtration pressures when PRP was added to isolated RBC, thus arguing against significant platelet-RBC interactions.

Abnormalities in platelet function have been observed in patients with multiple organ system failure and sepsis (11). Platelet activation and mediator release plays a role in PMN activation and aggregation (27). Conversely, activated PMN have been shown to increase the response of platelets to stimulation, and to directly cause platelet activation (28). Of particular relevance to the present study are reports that platelets can adhere to PMN and increase PMN aggregation (29).

The glycoprotein complex GPIIb-IIIa, the inducible fibrinogen receptor, which is expressed on the surface of resting platelets and after cell activation, plays an essential role in platelet aggregation. Upon activation, platelets degranulate and release glycoproteins such as thrombospondin and GMP140. These glycoproteins remain associated with the platelet membrane and stabilize platelet microaggregates (30). Moreover, platelet membrane glycoproteins such as GPIIb-IIIa and GMP140 are involved in platelet-leukocyte interactions (28). In our study, flow-cytometric measurements corroborated the presence of activated platelets and platelet- PMN interactions. Platelets were significantly activated in septic as compared with noninfected patients and control subjects. Similarly, platelet-PMN interactions were significantly greater in patients with sepsis than in control subjects. Of interest is the observation that although the percent of PMN neutrophils with bound activated platelets was similar for septic and noninfected critically ill patients, the mean CD63 fluorescence of the platelet-PMN aggregates was higher for the septic patients. This suggests that either larger numbers of activated platelets and/or platelets with a greater degree of activation are interacting with PMN in this patient group, thereby accounting for the increased filtration pressures generated by the platelet-PMN combinations from the septic patients.

Increases in circulating platelet-PMN aggregates have been previously reported in sepsis (11). Significantly lower concentrations of circulating platelet-PMN aggregates have been observed in patients with multiple organ failure, leading to speculation that sequestration of these aggregates contributed to microvascular obstruction and the development of organ dysfunction in these patients (11). Our observation of increased platelet activation and platelet-PMN adhesion associated with significantly increased resistance to cell filtration in septic patients is consistent with this theory.

Although cross-contamination may have affected our results, this influence should have been minimized by the more than 95% purity of our cell preparations and the lack of differences in cell concentrations among our study groups. Similarly, changes in monocyte and lymphocyte rheology might also contribute to alterations in cell filtration in sepsis.

In conclusion, we found that increases in the Pi of whole blood from critically ill patients with sepsis appeared to be related to decreases in RBC and PMN deformability, increases in RBC aggregation, and plateletPMN interactions. Of these factors, the formation of platelet-PMN aggregates had the greatest effect in impairing cell filtration. The resulting rheologic abnormalities may contribute to impaired microvascular blood flow in critically ill patients with sepsis.