The anti‐inflammatory effect of combined complement and CD14 inhibition is preserved during escalating bacterial load

Combined inhibition of complement and CD14 is known to attenuate bacterial‐induced inflammation, but the dependency of the bacterial load on this effect is unknown. Thus, we investigated whether the effect of such combined inhibition on Escherichia coli‐ and Staphylococcus aureus‐induced inflammation was preserved during increasing bacterial concentrations. Human whole blood was preincubated with anti‐CD14, eculizumab (C5‐inhibitor) or compstatin (C3‐inhibitor), or combinations thereof. Then heat‐inactivated bacteria were added at final concentrations of 5 × 104−1 × 108/ml (E. coli) or 5 × 107−4 × 108/ml (S. aureus). Inflammatory markers were measured using enzyme‐linked immunosorbent assay (ELISA), multiplex technology and flow cytometry. Combined inhibition of complement and CD14 significantly (P < 0.05) reduced E. coli‐induced interleukin (IL)‐6 by 40–92% at all bacterial concentrations. IL‐1β, IL‐8 and macrophage inflammatory protein (MIP)‐1α were significantly (P < 0.05) inhibited by 53–100%, and the effect was lost only at the highest bacterial concentration. Tumour necrosis factor (TNF) and MIP‐1β were significantly (P < 0.05) reduced by 80–97% at the lowest bacterial concentration. Monocyte and granulocyte CD11b were significantly (P < 0.05) reduced by 63–91% at all bacterial doses. Lactoferrin was significantly (P < 0.05) attenuated to the level of background activity at the lowest bacterial concentration. Similar effects were observed for S. aureus, but the attenuation was, in general, less pronounced. Compared to E. coli, much higher concentrations of S. aureus were required to induce the same cytokine responses. This study demonstrates generally preserved effects of combined complement and CD14 inhibition on Gram‐negative and Gram‐positive bacterial‐induced inflammation during escalating bacterial load. The implications of these findings for future therapy of sepsis are discussed.


Introduction
Despite modern treatment including up-to-date intensive care, sepsis still has high morbidity and mortality [1,2]. Massive research and several apparently promising clinical trials have not, so far, led to effective mediator-directed therapy for the systemic inflammatory response syndrome (SIRS), potentially leading to severe sepsis and septic shock [3].
The cardinal trigger of the immune response which leads to SIRS is the recognition of conserved patterns of exogenous as well as endogenous 'danger' molecules by pattern recognition receptors (PRRs) of innate immunity [4,5]. Complement and Toll-like receptors (TLRs) are central upstream sensor-and eventually effector-systems, thus triggering and maintaining the complex and widespread inflammatory response [6,7].
The complement system is a protein cascade with the main goal of defending the host against pathogens and maintaining homeostasis [8]. Complement is activated via the classical, lectin or alternative pathways. The pathways converge on the central complement factor C3, which is cleaved to C3a and C3b. Activation of the terminal pathway leading to the formation of the terminal C5b-9 complement complex (TCC) is triggered as C5 is cleaved to C5a and C5b [9]. C5a is a biologically highly potent anaphylatoxin associated with morbidity and mortality in sepsis [10].
The Toll-like receptor (TLR) family is a group of PRRs which recognize a wide variety of evolutionary conserved molecular structures [11]. Lipopolysaccharides (LPS) of Gram-negative bacteria are recognized by TLR-4, which is essential in inducing the Gram-negative inflammatory response [12]. Several danger molecules have been identified in Gram-positive bacteria, including various lipoproteins and lipoteichoic acid (LTA), and some of these danger molecules are recognized by TLR-2 [13][14][15]. CD14 is a promiscuous co-receptor for a number of TLRs and facilitates the signalling of both TLR-4 and TLR-2 [11,15]. Therefore, CD14 is an attractive target for upstream inhibition of TLR responses.
There is increasing evidence for close cross-talk between complement and TLRs, and recent research suggests possible synergistic anti-inflammatory effects when inhibiting both systems simultaneously [16]. Combined inhibition of C3 and CD14 has proved to be significantly more effective than single inhibition with either of these [17].
In clinical sepsis the bacterial load may increase rapidly, leading to an overwhelming inflammatory response.
Although not equivalent to clinical sepsis, we have used an ex-vivo human whole blood model to simulate bacterial-induced inflammation. In previous studies, using one bacterial load, the combined inhibition of complement and CD14 has proved effective. In the present study we aimed to test whether the effect of combined inhibition would depend upon bacterial load. By incubating escalating loads of Escherichia coli and Staphylococcus aureus in the human whole blood model, we examined to what extent the anti-inflammatory effect of the combined inhibition was preserved.

Ex-vivo whole blood model
The whole blood model has been described in detail previously [20]. In short, blood was drawn into 4-5 ml NUNC tubes containing the anti-coagulant lepirudin (50 mg/ml), which only blocks thrombin and does not interfere with the remaining inflammatory network. All the following experiments were performed with blood from six healthy donors. The different conditions described below were defined after careful pilot titration experiments to obtain optimal concentration and time intervals.
The E. coli experiments

Incubation of whole blood for final plasma analyses
The baseline sample (T0) was processed immediately after the blood was drawn and EDTA added to the whole blood. One tube was preincubated with PBS and served as the negative control. Four tubes were preincubated with PBS for 5 min at 378C before adding E. coli to final concentrations of 5 3 10 4 , 5 3 10 5 , 5 3 10 6 and 5 3 10 7 bacteria/ml and served as positive controls. In the same manner, four tubes were preincubated with eculizumab only, four tubes with anti-CD14 only and four tubes with the combination of eculizumab and anti-CD14 before adding E. coli. All samples were incubated for a total of 4 h before adding EDTA to a final concentration of 20 mM. The tubes receiving 5 3 10 5 , 5 3 10 6 and 5 3 10 7 bacteria/ml were used for measuring proinflammatory cytokines, whereas the tubes receiving 5 3 10 4 , 5 3 10 5 and 5 3 10 6 bacteria/ml were used for measuring chemokines. The tubes receiving 5 3 10 6 and 5 3 10 7 bacteria/ ml were used for measuring TCC. Experiments for analysis of granulocyte activation markers were conducted in the same manner as described above, with three modifications: instead of eculizumab, compstatin Cp40 was used for inhibiting complement, E. coli was added to final concentrations of 1 3 10 6 , 3 3 10 6 and 9 3 10 6 bacteria/ml and all samples were incubated for a total of 1 h. We consistently used Cp40 (a C3-inhibitor) to study the release of granulocyte enzyme release instead of eculizumab, as this effect has been shown to be C3-dependent, in contrast to the other inflammatory readouts studied [21].

Incubation of whole blood for CD11b analysis
Immediately after drawing blood from the donor, the cells from a sample of the whole blood were fixed with 0.5% (v/v) paraformaldehyde in an equal volume for 4 min at 378C to serve as a baseline (T0) sample. The subsequent bacterial activation of the whole blood was performed as described in the experiments for cytokine readout, with two modifications: E. coli was added to a final concentration of 4 3 10 6 , 2 3 10 7 and 1 3 10 8 bacteria/ml and incubated for 15 min. Following 15 min incubation, the cells were fixed with 0.5% (v/v) paraformaldehyde in an equal volume for 4 min at 378C, and then stained with anti-CD11b phycoerythrin (PE) and anti-CD14 fluorescein isothiocyanate (FITC) (Becton Dickinson, San Jose, CA, USA). The red cells were lysed and the samples washed twice using PBS with 0.1% Rinder albumin (300 g for 5 min at 48C) before they were run on a fluorescence activated cell sorter (FACS)Calibur flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA), with threshold on forward-scatter (FSC) to exclude debris. Monocytes were gated in a side-scatter/CD14 dot-plot, whereas granulocytes were gated in a forward-/side-scatter dot-plot. CD11b expression was measured as median fluorescence intensity (MFI).

Incubation of whole blood for final plasma analyses
The experiments were conducted in the same manner as described for E. coli, with two modifications: instead of E. coli, S. aureus was added to final concentrations of 5 3 10 7 , 1 3 10 8 and 2 3 10 8 bacteria/ml, and all samples were incubated for a total of 2 h. In addition, samples containing 1 3 10 8 and 2 3 10 8 bacteria/ml were incubated for a total of 4 h for measuring TCC. Experiments for analysis of granulocyte activation markers were conducted in the same manner as described for E. coli, with one modification: instead of E. coli, S. aureus was added to final concentrations of 5 3 10 7 , 1 3 10 8 and 2 3 10 8 bacteria/ml.

Incubation of whole blood for CD11b analysis
The experiments were conducted in the same manner as described for E. coli, with one modification: Instead of E. coli, S. aureus was added to final concentrations of 1 3 10 8 , 2 3 10 8 and 4 3 10 8 bacteria/ml.

TCC
The soluble terminal C5b-9 complement complex (TCC) was measured in an enzyme immunoassay (EIA), as described previously, and later modified [22,23]. Briefly, the monoclonal antibody (mAb) aE11 reacting with a neoepitope exposed in C9 after incorporation in the C5b-9 complex was used as capture antibody. A biotinylated anti-C6 mAb (clone 9C4) was used as detection antibody. The standard was Stock Complement Standard no. 2 [24], which is normal human serum activated with zymosan and defined to contain 1000 complement arbitrary units (CAU)/ml. Zymosan-activated human serum was used as a positive control.

Granulocyte activation markers
The granulocyte activation markers myeloperoxidase and lactoferrin were quantified using commercial EIA kits from Hycult Biotech (Uden, the Netherlands). The analyses were performed in accordance with the instructions from the manufacturer.

Data presentation and statistics
All results are given as mean and standard error of the mean (s.e.m.). GraphPad Prism version 5.04 (San Diego, CA, USA) was used for statistical analysis. One-way analysis of variance (ANOVA) with Dunnett's multiple comparison test was used to compare single and combined inhibition with positive controls. The background activation (negative control) was subtracted from the positive control and the intervention groups before calculating the percentage inhibition by the different inhibitors. A P-value < 0.05 was considered statistically significant for all analyses.

Ethics
The study was approved by the Regional Ethical Committee and all blood donors gave written informed consent to participate in the study.

Results
The E. coli experiments Effect of bacterial load on single and combined inhibition of C5 and CD14 on proinflammatory cytokines IL-1b. Single inhibition of both C5 and CD14 reduced IL-1b by 93% (P < 0.01) and 66% (P < 0.001), respectively, when whole blood was incubated with 5 3 10 5 Inhibition of escalating bacterial-induced inflammation V C 2015 British Society for Immunology, Clinical and Experimental Immunology, 181: 457-467 E. coli/ml (Fig. 1, upper left panel). No significant reduction of IL-1b was seen by single inhibition when using 5 3 10 6 and 5 3 10 7 E. coli/ml.
TNF. Single inhibition of C5 had no significant effect on TNF at any of the bacterial concentrations, whereas single inhibition of CD14 reduced TNF by 41% (P < 0.01) when whole blood was incubated with 5 3 10 5 E. coli/ml (Fig. 1, upper right panel). No significant reduction of TNF was seen by single inhibition of CD14 when using 5 3 10 6 and 5 3 10 7 E. coli/ml.
Combined inhibition reduced TNF by 80% (P < 0.001) when whole blood was incubated with 5 3 10 5 E. coli/m. (Fig. 1, upper right panel). No significant reduction of TNF was seen when using 5 3 10 6 and 5 3 10 7 E. coli/ml. Effect of bacterial load on single and combined inhibition of C5 and CD14 on chemokines IL-8. Single inhibition of C5 and CD14 had no significant effect on IL-8 at any of the bacterial concentrations. Combined inhibition of C5 and CD14 reduced IL-8 by 100% (P < 0.05) when whole blood was incubated with 5 3 10 5 E. coli/ml (Fig. 1, lower left panel). No significant reduction of IL-8 was seen when using 5 3 10 4 and 5 3 10 6 E. coli/ml. MIP-1a. Single inhibition of C5 reduced MIP-1a by 60% (P < 0.05) when whole blood was incubated with 5 3 10 4 E. coli/ml, whereas no significant reduction was seen when using higher concentrations of E. coli (Fig. 1, lower  middle panel).

Effect of bacterial load on single and combined inhibition of C5 and CD14 on the expression of CD11b
Monocytes. Single inhibition of C5 had no significant effect on CD11b expression at any of the bacterial concentrations, whereas single inhibition of CD14 reduced CD11b expression by 49% (P < 0.05), 55% (P < 0.01) and 29% (P < 0.01) when whole blood was incubated with 4 3 10 6 , 2 3 10 7 and 1 3 10 8 E. coli/ml, respectively (Fig. 2, upper left panel).

Effect of bacterial load on single and combined inhibition of C3 and CD14 on granulocyte enzyme release
We used a C3 inhibitor to study the release of granulocyte enzyme release, as this effect was shown to be C3dependent, in contrast to the other inflammatory readouts studied, which were mainly C5-dependent [21].
Myeloperoxidase. Single and combined inhibition of C3 and CD14 had no significant effect on myeloperoxidase at any of the bacterial concentrations (Fig. 2, lower left panel).
Lactoferrin. Single inhibition of C3 and CD14 had no significant effect on lactoferrin at any of the bacterial concentrations.
Combined inhibition of C3 and CD14 reduced the release of lactoferrin to the level of background activity when whole blood was incubated with 1 3 10 6 E. coli/ml (P < 0.05) (Fig. 2, lower right panel).

Effect of bacterial load on single and combined inhibition of C5 and CD14 on TCC concentration
Combined inhibition of C5 and CD14 and single inhibition of C5, reduced the formation of TCC by 98-99% (P < 0.001) and 99% (P < 0.001) when whole blood was incubated with 5 3 10 6 and 5 3 10 7 E. coli/ml, respectively (Table 1), confirming that C5 inhibition was efficient over the whole range of bacterial concentrations. As expected, no reduction of TCC was seen with selective inhibition of CD14 (Table 1).
Taken together, these data indicate that the effect of combined inhibition of complement and CD14 in general is well preserved by increasing doses of E. coli, although with some reduced effect on cytokines and chemokines in the highest bacterial concentration.

The S. aureus experiments
Effect of bacterial load on single and combined inhibition of C5 and CD14 on proinflammatory cytokines IL-1b. Single and combined inhibition of C5 and CD14 had no significant effect on IL-1b at any of the bacterial concentrations (Fig. 3, upper left panel). IL-6. Single inhibition of C5 and CD14 had no significant effect on IL-1b at any of the bacterial concentrations.
Combined inhibition reduced TNF by 86% (P < 0.05) when whole blood was incubated with 5 3 10 7 S. aureus/ml (Fig. 3, upper right panel). No significant reduction of TNF was seen when using 1 3 10 8 and 2 3 10 8 S. aureus/ml. Fig. 2. Effect of single and combined complement and CD14 inhibition on CD11b expression on monocytes and granulocytes in whole blood (upper panels) and plasma concentrations of granulocyte enzyme release (lower panels) with increasing bacterial load in the Escherichia coli experiments. Data are presented as mean and standard error of the mean. The background activation (negative control) was subtracted from the positive control and the intervention groups before calculating the percentage inhibition by the different inhibitors. Significances are between the positive control group (E. coli) and single and combined inhibition of either C5 (eculizumab in the CD11b experiments) and CD14 or C3 (compstatin in the granulocyte enzyme release experiments) and CD14. *P < 0Á05; **P < 0Á01; ***P < 0Á001. Combined inhibition of C5 and CD14 reduced IL-8 by 80% (P < 0.05) when whole blood was incubated with 1 3 10 8 S. aureus/ml (Fig. 3, lower left panel). No significant reduction of IL-8 was seen when using 5 3 10 7 and 2 3 10 8 S. aureus/ml. MIP-1a. Single and combined inhibition of C5 and CD14 had no significant effect on MIP-1a at any of the bacterial concentrations (Fig. 3, lower middle panel).  Single and combined inhibition of C5 and CD14 had no significant effect on MIP-1b at any of the bacterial concentrations (Fig. 3, lower right panel).

Effect of bacterial load on single and combined inhibition of C5 and CD14 on CD11b expression
Monocytes. Single inhibition of C5 reduced CD11b expression by 79% (P < 0.05) and 74% (P < 0.01) when whole blood was incubated with 1 3 10 8 and 2 3 10 8 S. aureus/ml, respectively (Fig. 4, upper left panel). No significant reduction of CD11b expression was seen when using 4 3 10 8 S. aureus/ml. Single inhibition of CD14 had no significant effect on CD11b expression at any of the bacterial concentrations.
Granulocytes. Single inhibition of C5 reduced CD11b expression by 73% (P < 0.05), 83% (P < 0.001) and Fig. 3. Effect of single and combined complement and CD14 inhibition on plasma concentrations of proinflammatory cytokines (upper panels) and chemokines (lower panels) with increasing bacterial load in the Staphylococcus aureus experiments. Data are presented as mean and standard error of the mean. The background activation (negative control) was subtracted from the positive control and the intervention groups before calculating the percentage inhibition by the different inhibitors. Significances are between the positive control group (S. aureus) and single and combined inhibition of C5 (eculizumab) and CD14. *P < 0Á05; **P < 0Á01; ***P < 0Á001.

Effect of bacterial load on single and combined inhibition of C3 and CD14 on granulocyte enzyme release
We used a C3 inhibitor to study the release of granulocyte enzyme release, as this effect was shown to be C3-dependent, in contrast to the other inflammatory readouts studied, which were mainly C5-dependent [21].
Myeloperoxidase. Single and combined inhibition of C3 and CD14 had no significant effect on myeloperoxidase at any of the bacterial concentrations (Fig. 4, lower left panel).
Lactoferrin. Single and combined inhibition of C3 and CD14 had no significant effect on lactoferrin release at any of the bacterial concentrations (Fig. 4, lower right panel).

Effect of bacterial load on single and combined inhibition of C5 and CD14 on TCC concentration
Combined inhibition of C5 and CD14 and single inhibition of C5 reduced the formation of TCC by 98-99% (P < 0.05) and 99% (P < 0.001) when whole blood was incubated with 1 3 10 8 and 2 3 10 8 S. aureus/ml, Fig. 4. Effect of single and combined complement and CD14 inhibition on CD11b expression on monocytes and granulocytes in whole blood (upper panels) and plasma concentrations of granulocyte enzyme release (lower panels) with increasing bacterial load in the Staphylococcus aureus experiments. Data are presented as mean and standard error of the mean. The background activation (negative control) was subtracted from the positive control and the intervention groups before calculating the percentage inhibition by the different inhibitors. Significances are between the positive control group (S. aureus) and single and combined inhibition of either C5 (eculizumab in the CD11b experiments) and CD14 or C3 (compstatin in the granulocyte enzyme release experiments) and CD14. *P < 0Á05; **P < 0Á01; ***P < 0Á001. respectively (Table 1), confirming that C5 inhibition was efficient over the whole range of bacterial concentrations. As expected, no reduction in TCC formation was seen with single inhibition of CD14.
Taken together, these data indicate that the preserved inhibitory effect of combined inhibition of complement and CD14 seen by increasing doses of the Gram-positive bacterium S. aureus was less pronounced compared to what was found with the Gram-negative bacterium E. coli.

Discussion
The results of this study consistently show a relation between the increased bacterial load and the inflammatory response induced by E. coli and S. aureus. The findings correspond well with the clinical setting in which an increased level of cytokines is correlated with the severity of sepsis [25]. Notably, combined inhibition of complement and CD14 efficiently bacterial-induced inflammation attenuated over a broad range of bacterial concentrations. This effect was preserved quantitatively for most of the inflammatory mediators, although fewer significant results were observed for S. aureus compared with E. coli, and slightly lesser effects were seen for some of the proinflammatory cytokines at the highest bacterial concentration. The latter result is consistent with previous data on Neisseria meningitides [26]. Evidence suggests that high bacterial doses lead to a 'point of no-return' after which the inflammatory cascade cannot be affected by treatment [27].
Taking a 'point of no return' into consideration, the administration of anti-inflammatory drugs must be performed at an early stage, or perhaps even before the onset of sepsis in patients at high risk. As the microorganism is seldom known in early stages of clinical sepsis, we showed equally importantly that the combined inhibition was effective in reducing Gram-negative and -positive induced inflammation. However, compared to E. coli, much higher concentrations of S. aureus were required to induce the same inflammatory responses; the E. coli-induced inflammatory responses, especially chemokine responses and granulocyte enzyme release, were also attenuated more efficiently by the combined inhibition of complement and CD14.
The inflammatory response in Gram-positive sepsis is more complement-driven compared to the Gram-negative response, which is largely LPS-dependent and therefore more TLR-4-driven [28]. Although both complement and CD14 play vital roles mediating E. coli-as well as S. aureus-induced inflammation, it is uncertain whether the anti-CD14 effects observed on the S. aureus-induced responses reflect the involvement of TLR-2, TLR-9 or other TLRs [21,29]. The bacteria causing the infection are most often unknown at the time sepsis develops [30], and quite often the infecting microbe remains unknown. A treatment with combined inhibition of complement and CD14 is meant for early attenuation of the inflammatory response, before the cytokine storm emerges. Thus, from a clinical perspective, the combined approach therefore seems sensible, provided that proper antibiotic therapy is given. Furthermore, no single downstream anti-inflammatory treatment has been able to achieve increased survival in sepsis [31], whereas a caecal ligation and puncture-induced polymicrobial sepsis study in mice clearly showed increased survival in the group receiving combined inhibition [32]. The anti-inflammatory effects observed in the current study are thus linked to in-vivo effects, strengthening possible clinical implications by the combined inhibition.
The proinflammatory cytokines and chemokines included in the present study were selected based on their implication in the pathogenesis of sepsis. Plasma TNF, IL-1b, IL-6 and IL-8 are important mediators of the inflammatory response in SIRS and sepsis and correlate partially with mortality and severity in septic patients [33][34][35]. However, several large clinical trials show that increased survival has not been achieved by targeting these cytokines selectively [36][37][38]. By targeting upstream recognition molecules of innate immunity, i.e. complement and CD14, we observed a significant concurrent reduction of all four cytokines in escalating bacterialinduced inflammation, implicating the benefits of upstream as opposed to downstream single agent antiinflammatory treatments, even in relatively high bacterial concentrations. Interestingly, the release of IL-6 was inhibited significantly even at the highest concentration of E. coli. The inhibitory effect was still significant for IL-1b, IL-8 and MIP-1a at bacterial concentrations close to the highest dose. For S. aureus, less statistically significant effects were observed compared with E. coli, corresponding with a greater variation in the inflammatory response to the Gram-positive bacterium.
For a long time, C5a has been reported to induce IL-8 [39], which may explain the relative complementdependent IL-8 release observed in the current study. IL-8 is one of the major mediators of the inflammatory response in sepsis with the main function to induce chemotaxis of neutrophils to the site of inflammation [25]. As the levels of IL-8 correlate with mortality in septic patients [33], the use of complement inhibition in sepsis is thus logical.
E. coli-induced surface expression of CD11b on granulocytes was reduced significantly to background levels by the combined inhibition even at the highest bacterial concentration. The release of myeloperoxidase was not changed significantly, whereas combined inhibition reduced lactoferrin significantly at the lower dose. In the presence of Gram-positive bacteria, however, this leucocyte inflammatory response was gradually less affected by combined inhibition as the bacterial load increased. Interestingly, E. coli-induced granulocyte CD11b expression Inhibition of escalating bacterial-induced inflammation V C 2015 British Society for Immunology, Clinical and Experimental Immunology, 181: 457-467 was essentially complement-dependent, whereas for monocytes it was mainly CD14-dependent. In contrast, when activating whole blood with S. aureus, CD11b expression seemed to be complement-dependent for both subsets of leucocytes. These results are in line with previous studies and the concept of cross-talk between complement and TLRs [21,40].
In conclusion, this study showed significant broad anti-inflammatory effects by the combined inhibition of complement and CD14 by increasing Gram-negative as well as Gram-positive bacteria to high concentrations. A 'point of no return' in the inflammatory response emerges at last, underscoring the relevance of applying such treatment at an early stage of the syndrome. The results of the present study strengthen the hypothesis that the combined inhibition of complement and CD14 may become a future mediator-directed therapy in sepsis.