DO NATURE OF BACTEREMIA AND ORIGIN OF SECONDARY SEPSIS IN CRITICALLY ILL PATIENTS DETERMINE SUBSET OF MYELOID- DERIVED SUPPRESSOR CELLS EXPANSION? DA LI VRSTA BAKTERIJA I POREKLO SEKUNDARNE SEPSE KOD KRITIČNO OBOLELIH ODREĐUJU TIP SUPRESORSKIH ĆELIJA MIJELOIDNOG POREKLA?

Background/Aim. Gram-positive and Gram-negative bacteria may induce different inflammatory patterns. The aim of this study was to examine the association of the myeloid-derived suppressor cells (MDSCs) with the type of infecting microorganisms (Gram positive, Gram negative, polymicrobial) and underlying cause of secondary sepsis (peritonitis, pancreatitis, trauma). Methods. Totally, 40 critically ill patients with secondary sepsis were enrolled in the prospective study. Two patients without documented positive blood culture were excluded. We detected and enumerated both main subsets of MDSCs: granulocytic (G)-MDSCs and monocytic (M)-MDSCs on the Days 1 and 5. Blood was simultaneously drawn for a blood culture. The patients with different underlying causes of sepsis (peritonitis, pancreatitis, trauma) were perceived as separated groups and the frequencies and absolute numbers of their G-MDSCs and M-MDSCs were compared. Results. Both main MDSC subpopulations were accumulated significantly in Grampositive sepsis. Univariate logistic regression analyses of investigated variables regarding Gram-positive sepsis on the Day 5 revealed that G-MDSCs absolute number along with both MMDSCs frequency and absolute number had statistically significant power for predicting Gram-positive sepsis. Stepwise multivariate logistic regression analyses of the variables on the Day 5 determined that M-MDSCs absolute number was independent predictor of Gram-positive sepsis [odds ratio (OR) 1.012; p < 0.05]. Clinical accuracy of neutrophil (Ne)/GMDSCs (Ne/G-MDSCs) and monocyte (Mo)/M-MDSCs (Mo/M-MDSCs) ratios in predicting nature of bacteremia and outcome were investigated. Discriminative power of both Ne/G-MDSCs and Mo/M-MDSCs ratios in predicting Grampositive blood culture was statistically significant both on the Day 1 and Day 5 [areas under curve (AUCs): 0.684 and 0.692, and 0.707 and 0.793, respectively). Ne/G-MDSCs both on the Day 1 and Day 5 were statistically significant predictors of lethal outcome (AUCs: 0.694 and 0.678, respectively). There were no statistically significant differences in G-MDSCs and M-MDSCs among different three groups of patients regarding peritonitis, pancreatitis and trauma as causes of sepsis neither on the Day 1 nor on the Day 5. Conclusion. Gram-positive infectious agents were powerful inducers of MDSCs generation in sepsis. Also, underlying causes of secondary sepsis might not seem to influence the MDSCs accumulation.


Introduction
Sepsis is a principal cause of death in critical care units worldwide and consumes considerable healthcare resources.There is evidence to suggest that there are different mechanisms of the clinical manifestations of Gram-positive and Gram-negative sepsis to the extent that they may represent different disease entities (1).Some microbial challenges may elicit levels of mediators that damage both the infecting microorganism and the host.Lipoteichoic acid (LTA) of Gram-positive bacteria as well as lipopolysaccharide (LPS) of Gram-negative bacteria elicits different response from the host.Furthermore, Grampositive and Gram-negative bacteria may induce different inflammatory patterns.But, it is not physiologically or clinically apparent because of the fact that signs of systemic inflammatory response syndrome and routine laboratory markers of infection are nonspecific (1)(2)(3).Myeloid-derived suppressor cells (MDSCs), with its two main subsets being monocytic (M)-MDSCs and granulocytic (G)-MDSCs are important regulators of intricate and complex immuno-inflammatory response to various insults such as bacteria (4).Objectives of our study were to examine the association of the MDSCs with the type of infecting microorganism (Gram positive, Gram negative, polymicrobial) and underlying cause of secondary sepsis (peritonitis, pancreatitis, trauma).

Patients
Total of 40 critically ill patients with secondary sepsis due to peritonitis, pancreatitis and severe trauma, admitted to surgical intensive care unit (SICU), were enrolled in prospective study conducted in a tertiary university hospital (Military Medical Academy, Belgrade, Serbia).Approval in concordance with Declaration of Helsinki was obtained from local ethics committee and informed consent from a patient or first-degree relative.Detailed description of study population is reported elsewhere (5).Blood samples for MDSCs analysis were collected on admission -Day 1 and on Day 5.These two specific time points were chosen because dynamic change in MDSCs function during sepsis were expected.Blood was simultaneously drawn for a blood culture.The Sequential Organ Failure Assessment (SOFA) score, the Simplified Acute Physiology Score (SAPS) II and the Acute Physiology and Chronic Health Evaluation (APACHE) II score were calculated and recorded within the first 24 hours after admission to the SICU (Day 1).SOFA score was recorded daily during SICU stay to assess severity of organ dysfunction in secondary sepsis (6 -8).The use of antibiotics, circulatory volume replacement and vasoactive support were performed according to guidelines (9).Various modes of mechanical ventilation and surgical procedures were performed if and when necessary in all patients.

Sampling and analysis
Fresh peripheral blood samples were analyzed, frequency and absolute number of MDSCs were determined.Both main subsets of MDSCs were detected, the granulocytic (G-MDSCs) and the monocytic subset (M-MDSCs).MDSCs analysis is described elsewhere (5).

Statistical analysis
Complete statistical analysis of data was done with the statistical software package, SPSS Statistics 18.Most of the variables were presented as frequency of certain categories, while statistical significance of differences was tested with the Chi square test.In case of continuous data, variables were presented as mean value ± standard deviation (SD), median, minimal and maximal values.Kolmogorov-Smirnov test was used for evaluation of distribution of continual data.Statistical significance between groups was tested by Wilcoxon or Mann-Whitney test.Spearman's Rank Correlation analyses used to establish the relation between parameters.ROC curves were constructed and analyzed to determine the sensitivity and specificity of variables for prediction of nature of bacteremia and outcome.Calculations of odds ratios and their 95% confidence intervals were done to determine the strength of the association between variables and nature of bacteremia.For that purpose, the most promising independent variables as single or combined variables were incorporated into binary logistic regression analyses.All the analyses were estimated at P<0.05 level of statistical significance.

RESULTS
Demografic and clinical data of forty patients are shown on Table 1.Detailed description is presented in our previous study (5).Two patients with sterile blood cultures were excluded from further analysis.

Both main MDSC subpopulations accumulate significantly in Gram-positive sepsis
We have compared the frequencies and the absolute numbers of G-MDSCs and M-MDSCs in sepsis patients according to the nature of bacteraemia (Gram-positive, Gramnegative and Polymicrobial group) (Figure 1A,B,C,D).Initially, on Day 1, patients with Gram-positive sepsis had more G-MDSCs and M-MDSCs (both relative and absolute number) comparing to other two groups, but without significant difference.Accumulation of G-MDSCs and M-MDSCs in patients with Gram-positive sepsis has become more intensive at 5 th day.This group had significantly more both G-MDSCs and M-MDSCs comparing to Polymicrobial sepsis group (P<0.05)(Figure 1  A,B,C,D).Also, patients with Gram-positive sepsis had significantly more G-MDSCs and M-MDSCs (absolute number) than patients with Gram-negative sepsis (P<0.05)(Figure 1B  and 1D).Univariate logistic regression analyses were performed in order to determine whether associations of each individual variable with Gram-positive sepsis exist.Standardized regression coefficient () and the odds ratio (OR) with 95%CI were calculated for each variable.Forward stepwise multivariate logistic regression model was performed in order to determine the independent predictors of Gram-positive sepsis, without the effect of possible confounders.On Table 3. univariate odds ratios of variables for predicting Grampositive sepsis in patient population, on Day 1 and Day 5 are shown.Univariate logistic regression analyses of investigated variables regarding Gram-positive sepsis on Day 1 revealed that both G-MDSCs frequencies and absolute number had statistically significant power for predicting Gram-positive sepsis.Univariate logistic regression analyses of investigated variables regarding Gram-positive sepsis on Day 5 revealed that G-MDSCs absolute number along with both M-MDSCs frequencies and absolute number had statistically significant power for predicting Gram-positive sepsis.

Fig. 1 -Comparison of MDSCs frequencies between groups of patients with different nature of bacteraemia. A. Relative number of G-MDSC (%). B. Absolute number of G-MDSC (N/µL). C. Relative number of M-MDSC (%). D. Absolute number of M-MDSC (N/µL). (relative and absolute numbers given as
Stepwise multivariate logistic regression analyses of the variables on Day 5 determined that M-MDSCs absolute number was independent predictor of Gram-positive sepsis which is shown on Table 4.A Spearman's rho test of correlation between frequencies and absolute numbers of G-MDSCs and M-MDSCs on one hand and Gram-positive sepsis on the other hand was performed to assess strength of association.On Day 1 absolute numbers of G-MDSCs and M-MDSCs correlated significantly with Gram-positive sepsis.That positive correlation is shown on Table 5 and Figure 2.  On Day 5, there were significantly positive correlations between all investigated variables and Gram-positive sepsis shown on Table 6.Positive correlations between G-MDSCs and M-MDSCs frequencies and Gram-positive sepsis are shown on Figure 3.  7.

Clinical accuracy of Ne/G-MDSCs and Mo/M-MDSCs ratios in predicting nature of bacteremia and outcome
Clinical accuracy of Ne/G-MDSCs and Mo/M-MDSCs ratios in predicting nature of bacteremia and outcome was investigated.Discriminative power of both Ne/G-MDSCs and Mo/M-MDSCs ratios in predicting Gram-positive blood culture was statistically significant both on Day 1 and Day 5. Results are shown on Table 8 and Figures 4 and 5.
Table 8.   9 and Figure 6.Ne/G-MDSCs and Mo/M-MDSCs ratios higher than respective cut-off values are predictors of polymicrobial blood culture on Day 5 in critically ill patients with secondary sepsis.Ne/G-MDSCs ratio has very good discriminative power while Mo/M-MDSCs ratio has moderate one.

Underlying causes of secondary sepsis might not seem to influence the MDSCs accumulation
The underlying causes of secondary sepsis in our patients were pancreatitis, peritonitis and trauma.The patients with different underlying causes of sepsis were perceived as separated groups and the frequencies and absolute numbers of their G-MDSCs and M-MDSCs were compared.There were no statistically significant diffrences between these three groups neither on Day 1 nor on Day 5. So, MDSC expansion was related to secondary infection regardless of nature of primary insult (pancreatitis, peritonitis, trauma).

Discussion
Immune dysfunction is common in critically ill patients and it may modulate immune response and affect patient morbidity and mortality, particularly in severe trauma and/or sepsis.Immune cells' and mediators' role in immune response in critical illness is not yet fully elucidated (10,11).
Expansion and activation of MDSCs, as part of immune response, are under the influence of several different factors, including infectious agents (12)(13)(14).It seems that there is a difference between Gram-positive and Gram-negative sepsis regarding cytokine profile, for instance (1,15).It has also been shown that different types of microbes can induce specific subsets of MDSCs, with different impact on disease outcome 14).A very interesting study of Janols et al. showed predominant accumulation of CD14 low polymorphonuclear MDSCs (PMN-MDSCs) in patients with Gram-positive sepsis and septic shock.They also showed that the CD14 low PMN-MDSCs accumulate in both, Gramnegative and Gram-positive sepsis, but are significantly more potent suppressors of T-cell proliferation when isolated from the Gram-positive sepsis patients (16).The findings of Janols et al. suggest that different types of bacteria can influence myeloid response of the septic host, and accordinglygeneration of specific MDSCs subset with possible distinct functions.In our study, we found significantly higher frequencies of both detected MDSCs subpopulations, the G-MDSCs and the M-MDSCs, in the patients with Gram-positive sepsis when compared with Polymicrobial sepsis patients on the Day 5. Also, patients with Gram-positive sepsis had significantly more both G-MDSCs and M-MDSCs (absolute number) than patients with Gram-negative sepsis (P<0.05).Stepwise multivariate logistic regression analyses of the variables on Day 5 determined that M-MDSCs absolute number was independent predictor of Gram-positive sepsis.Positive correlations between G-MDSCs and M-MDSCs frequencies and Gram-positive sepsis are confirmed by Spearman's rho test.Possible explanation of these differences may lie in the basic understanding of MDSCs expansion seen in malignant diseases and protracted infections (12,17).Prompt reaction of the bone marrow in response to Gram-negative, and possible to Polymicrobial causative infectious agents, may leave no time for the different proinflammatory factors to act on myeloid precursors in different stages of maturation and to activate/convert them into immunosuppressive cells.On the contrary, more indolent, in terms of acute inflammatory response, Gram-positive infectious agents could lead to prolonged bone marrow exposure, creating the environment conducive for MDSCs accumulation (18).In addition, our finding that there were no significant differences in MDSCs accumulation between patients with different underlying causes of secondary sepsis (pancreatitis, peritonitis or trauma injury as primary insults) also speaks in favor of the causative infectious agent being more important for MDSCs generation than the type of primary insult leading to secondary sepsis.Uhel and colleagues performed peripheral blood transcriptomic analysis on 29 patients with sepsis and 15 healthy donors and in a second cohort of 94 patients with sepsis, 11 severitymatched ICU patients and 67 healthy donors they performed functional analysis in order to clarify phenotype, suppressive activity, origin and clinical impact of MDSCs in patients with sepsis (19).Their results showed that MDSCs were major players in sepsis-induced immunosuppression.In sepsis patients they demonstrated up-regulation of gene profile associated with MDSCs regrutation and immunosuppression (MMP8, MMP9, ARG1, S100A8, S100A9, S100A12, PD-L1, IL-4R, and IL-10), but down-regulation of gene profile associated with inflammatory response (CD4, CD20, CD8, CD3, IL-8 and IL-6).They concluded that CD14 + HLA-DR low/-M-MDSCs and CD15 + G-MDSCs strongly contributed to T-cell dysfunction in patients with sepsis.Contrary to our results, they found no association with Gram-staining of the causative organism.Interestingly, they also demonstrated that expression of two among key MDSCs parameters ARG1 and S100A9, significantly directly correlated to granulocyte count and inversely correlated to number of lymphocytes.Furthermore, Uhel et al. showed that beside MDSCs, CD14 + monocytes and CD15 + low density granulocytes from sepsis patients were suppressive in vitro, similar to MDSCs.They also showed that population of low density granulocytes is very heterogenous, being composed of immature and mature granulocytes both expressing degranulation markers.
In other words, beside MDSCs, mature monocytes and granulocytes of investigated patients demonstrated function and phenotype alterations.These findings are hard to explain from the aspect where MDSCs increment is a consequence of emergency myelopoesis followed by export of immature myeloid cells from bone marrow into blood stream.But, several recent articles pointed out that MDSCs increase could be achieved by reprogramming of existing monocytes, arguing that monocyte to M-MDSCs relation is very dynamic and plastic (4,20).Of course, both mechanisms could be operative at the same time, they are not mutualy exclusive.According to this, we have analyzed ratio of monocytes to M-MDSCs and granulocytes to G-MDSCs, in every individual patient and in both time points.All sepsis patients from our study demonstrated decrement of Ne/G-MDSCs ratio and increment of Mo/M-MDSCs ratio from 1 th to 5 th day.But, stratification of patients according to the type of microbial culture demonstrated significant differences.Patients with Gram-positive sepsis demonstrated significant decrement of Ne/G-MDSCs ratio and less prominent decrement of Mo/M-MDSCs ratio from 1 th to 5 th day.Although Gram-negative sepsis patients also demonstrated significant Ne/G-MDSCs ratio decrement, the number of their monocytes increased comparing to detected number of M-MDSCs.Contrary to both previous groups, sepsis patients with polymicrobial cultures on 5 th day demonstrated increase of both Ne/G-MDSCs and Mo/M-MDSCs ratios.All these indicate that type of microbial infection in sepsis is significantly associated with particular profile of MDSCs, dynamic of their change and their relation to mature -like counterpart cells.Finally, in our investigation, decrease of Ne/G-MDSCs ratio was associated with worse outcome, being significantly lower in nonsurvivors comparing to survivors.
Bergenfelz et al. demonstrated that systemic M-MDSCs are generated from monocytes and that their number correlates with disease progression in breast cancer patients (21).Additionaly, they observed significant increase of monocytes with altered phenotype both in breast cancer group as well as in control sepsis group.These monocytes exhibited CD14 + HLA -DR low/-phenotype, which is specific for M-MDSCs, and were already documented in few earlier studies in sepsis patients with compensatory anti-inflammatory response syndrome (CARS) (20,(22)(23)(24)(25). Gene profiling furhter delineated that these populations of monocytes/M-MDSC were similarly immunosuppressive in both breast cancer and sepsis patients, but not other infective diseases and healthy controls (21).Monocytes from early breast cancer group produced comparable level of IL-1β, IL-6, IL-8 and TNF as monocytes from metastatic group, indicating change of monocyte function early in disease.Furthermore, sepsis patients had significantly more total CD14 + cells, CD14 + CD16 -cells, CD14 ++ CD16 + intermediate monocytes and CD14 ++ CD16 ++ nonclassical monocytes comparing to both early and metastatic breast cancer patients and healthy controls, with increased CD16 + /CD16 -monocyte ratio.Authors concluded that these Mo/M-MDSCs were induced early during tumor growth and progression and that monocytes are affected by tumor much before their extravasation into tumor tissue.Based on observation of similar phenotypic and molecular findings in breast cancer and sepsis patients we could assume that sepsis progression could re-programme monocytes and granulocytes in the same way.
Re-programming process is not a rare event and could have physiological implications.Zhao et al. (26) demonstrated that human trophoblast cells efficiently change differentiation programme in monocytes, inducing their maturation toward dendritic cells (DC).Those trophoblast induced monocyte derived DCs display altered, hypo-stimulatory capacity to T lymphocytes and induce generation of inhibitory regulatory T lymphocytes.Sepsis itself induces numerous changes in monocyte functions.Shalova et al. (27) demonstrated that sepsis patients' monocytes exert significant up-regulation of genes associated with inflammation (IL-1b, IL-6, CCL3, CCL5), but also with tissue remodeling genes (VEGF, MMPs).Authors found that hypoxia inducible factor -1 (HIF-1a) was specifically up-regulated in sepsis patients' monocytes but not in the control ones.HIF-1a negatively regulated TLR monocyte activation, resulting in diminished pro-inflammatory response to endotoxin challenge, so called endotoxin tolerance.Although this study did not investigate MDSCs, authors concluded that HIF-1a is important regulator of monocyte reprogramming toward immunosuppressive functions in sepsis patients.
The main limitation of our study is sample size.Significant number of critically ill patients with secondary sepsis due to diffuse peritonitis had to be excluded because of malignant disease.Larger trial is essential for possible confirmation of our findings.We demonstrated that the Gram-positive infectious agents were powerful inducers of MDSCs generation in sepsis.Also, underlying causes of secondary sepsis might not seem to influence the MDSCs accumulation.

Fig. 2 -
Fig. 2 -Scattergram on log 10 scales of G-MDSCs and M-MDSCs absolute numbers versus Gram-positive sepsis in patient population on Day 1

Fig. 3 -
Fig. 3 -Scattergram on log 10 scales of G-MDSCs and M-MDSCs frequencies versus Gram-positive sepsis in patient population on Day 5

Table 3 . Univariate ORs of variables for predicting Gram-positive sepsis in patient population on Day 1 and Day 5
Significant differences are marked by *(P< 0.05)

Table 7 . Baseline characteristics of patient population according to nature of bacteremia on Day 1 and Day 5
MDSCs and Mo/M-MDSCs ratios were lowest in critically ill patients with Grampositive bacteremia both on Day 1 and Day 5. On the first day that difference did not reach statistical significance, but on Day 5 both ratios were statistically significantly lower in patients with Gram-positive bacteremia compared to patients with Gram-negative or polymicrobial blood culture (Ne/G-MDSCs: Chi-Square = 6.806, p< 0.05; Mo/M-MDSCs: SDstandard deviation; M -Median; min -Minimum; max -MaximumNe/G-