Introduction
The incidence of sepsis is increasing, and there is still a high mortality associated with severe sepsis [
1‐
4]. A wide spectrum of host responses [
4‐
6] contribute to the considerable clinical heterogeneity, as well as to the repeated failures of clinical trials with inflammatory modulators [
7‐
9]. The presence of a prominent immune activation, with a “cytokine storm”, and even a “genomic storm” as shown in response to low-dose bacterial endotoxin [
10], can be a
sine qua non sepsis factor, occurring early and acting anti-microbially for the benefit of the host. Recently, consequences of the administration of immunosuppressive therapy became the subject of much attention [
11‐
14]. Nowadays, these approaches are considered to be a major cause of sepsis mortality.
It is a debated question whether or not signs suggesting immunosuppression can be viewed to form a state of compensatory anti-inflammatory response syndrome (CARS) or not [
8,
9,
11,
14]. A perplexing sign observed early and in most sepsis patients is lymphopenia, mediated at least in part by apoptosis [
15]. Whether this lymphopenia should be regarded as part of a compensatory immunosuppression, thus contributing to organ dysfunction and opportunistic infections commonly seen in later stages of the disease, is unknown. Experimental studies performed in mice suggest that inhibition of the sepsis-induced lymphocyte apoptosis specifically increases survival [
16]. It has also been shown that T lymphocytes repopulate their respective compartments after sepsis via tightly regulated mechanisms [
17]. Neither is it known whether clinically significant immunosuppression is a common phenomenon, or only occurring in the most advanced patients, thus rarely indicating a need for immune-restoring therapy. The answers to such questions are important because they decide whether sepsis patients should be monitored with immune biomarkers, and whether there is a need to develop appropriate immunomodulating therapeutics.
In this study, we made an attempt to broadly document the incidence and nature of immune alterations in sepsis patients with different clinical severity and causative microorganisms (19 gram-negative and 13 gram-positive patients) and compared this to patients with generalised virosis or healthy controls. We report that a marked inter-individual variation among sepsis patients indicates that sepsis care may benefit from a personalised approach, including a detailed assessment of immune status.
Materials and methods
Patients
Samples from all of the sepsis and virosis patients were obtained within 4 days after admission to hospital. The majority of samples were taken within 2 days [Gram-positive sepsis patients (7/10) and Gram-negative sepsis patients (7/12)] and those taken later than 2 days were especially controlled to not bias the conclusions in our study. For shock patients, the samples were obtained within 2 days. All patients were prospectively included, to cover a spectrum of illness severity including shock, and to have several microbial causative agents represented. Swedish national guideline criteria for sepsis diagnosis were adhered to, being similar to those of Bone et al. [
18]. Sepsis was, thus, defined as the presence of a suspected or microbiologically proven infection, together with a systemic inflammatory response syndrome (SIRS), with SIRS defined by at least two of the following parameters: hypothermia (≤36 °C) or hyperthermia (≥38 °C); tachycardia (≥90/min); tachypnoea (≥20 breaths/min) and/or arterial PCO
2 32 mmHg or lower and/or mechanical ventilation; and leukocytosis (≥12,000/μl) or leukopaenia (≤4,000/μl) and/or a left-shifted white blood cell differential count of 10 % or higher. Septic shock was defined as sepsis-induced hypotension persisting despite adequate fluid administration. There were 19 patients with sepsis, three patients with severe sepsis and ten patients with septic shock (Table
1). Since the patients with severe sepsis were so few, we chose to define them in the group of sepsis but not septic shock. Standardised antibiotic treatment according to Malmö University Hospital’s guidelines was given until the cultures were finalised. The following subjects, with the indicated microbial agents recovered by culture or diagnosed by polymerase chain reaction (PCR) (using Swedish national QC-approved methods) were included: septic shock (
n = 10, 3 females and 7 males), with blood culture isolate of
Escherichia coli (4 patients),
Klebsiella oxytoca (1 patient),
Staphylococcus aureus (2 patients),
Streptococcus pyogenes (1 patient), and in two patients, no microbiological agent was isolated from blood but with suspected Gram-positive (1 patient) and Gram-negative (1 patient) etiology (these patients responded quickly to either Gram-positive or Gram-negative antibiotic treatment and also had clear symptoms from either the urinary tract or the lungs); Gram-positive sepsis (
n = 10, 3 females and 7 males), with blood culture isolate of
S.
aureus (3 patients),
S.
pneumoniae (4 patients), anaerobic cocci (1 patient) and the 9th and 10th patients with a probable pneumococcus etiology; Gram-negative sepsis (
n = 12, 8 females and 4 males), with blood culture isolate of
E.
coli (3 patients),
K.
oxytoca (1 patient),
Citrobacter koseri (1 patient) and another five patients with significant quantity of
E.
coli in urine (Table
1). The virosis cases (
n = 11, 8 females and 3 males) were one influenza A, four influenza B, two influenza H1N1 (one coinfected with influenza A), one hepatitis A, one acute hepatitis B and two HSV2 patients. Healthy controls were also included (
n = 13, 9 females and 4 males). Ethical permit was obtained from the local ethical committee at Lund University (Dnr 288/2007) and an informed consent was given from the participating patients or their relatives if the patient was not in a condition to provide an informed consent him/herself.
Table 1
Clinical characteristics of the sepsis patients (n = 32)
Age, median years (range) | 74 (51–87) | 68.5 (33–89) | 51 (20–79) | 46 (26–64) |
Males | 7 | 7 | 4 | 4 |
28-day mortality | 3 | 1 | | |
Length of ICU stay |
Days median (range) | 6 (1–49) | 0 | 0 | n/a |
Length of hospitalization |
Days median (range) | 23 (3–53) | 7 (3–55) | 4 (3–46) | n/a |
Ventilation |
Days median (range) | 6 (0–49) | 0 | 0 | n/a |
Nosocomial infection |
Percentage (number) | 50 (n = 5) | 0 | 0 | n/a |
WBC count at enrollment |
Cells (×106)/ml median (range) | 20 (1.5–60.4) | 18 (8.6–32.3) | 8.7 (5.1–32.4) | 5.7 (4.6–11) |
Lymphocyte count at enrollment |
Cells (×106)/ml median (range) | 1 (0.1–2.4) | 1 (0.3–2.1) | 1.2 (0.5–2.4) | 1.8 (0.8–3.7) |
Comorbidities |
Diabetes | 3 | 3 | 3 | n/a |
Renal failure | 1 | 1 | 0 | n/a |
Cardiovascular disease | 5 | 6 | 4 | n/a |
Pulmonary disease | 6 | 3 | 0 | n/a |
Metastatic cancer | 1 | 0 | 1 | n/a |
Blood or urine culture finding |
Escherichia coli
| 4 | n/a | 10 | n/a |
Citrobacter koseri
| 0 | n/a | 1 | n/a |
Staphylococcus aureus (1 TSSS+) | 2 | 3 | n/a | n/a |
Streptococcus pyogenes
| 1 | 0 | n/a | n/a |
Klebsiella oxytoca
| 1 | n/a | 0 | n/a |
Klebsiella pneumoniae
| 0 | n/a | 1 | n/a |
Streptococcus pneumoniae
| 0 | 4 | n/a | n/a |
Gram-positive anaerobic cocci | 0 | 1 | n/a | n/a |
No bacteria isolated from blood/urine | 2 | 2 | 0 | n/a |
Blood clinical routine methods
Venous blood was drawn in EDTA tubes and the time between sample collection and analysis was less than 6 h. The leukocyte concentration and differential count were determined using an LH750 machine (Beckman Coulter, Hialeah, FL, USA) using Swedish national QC-approved clinical diagnostic methodology.
PCR method for Tregs
The analysis of FOXP3
+ T regulatory cells (Tregs) was performed by DNA methylation of the
FOXP3 CpG-rich gene promoter, as previously described [
19]. A limited number of whole blood samples was available for this purpose.
Flow cytometry
In this study, we chose to use the following markers; CD3, a marker for all T cells; CD8, a marker for cytotoxic T cells; CD11c, expressed on and used as a marker for dendritic cells, monocytes, macrophages, neutrophils and some B cells; CD14, expressed on all monocytes; CD40, the receptor for T cell-CD40L expressed on antigen-presenting cells (APCs) and indicates the level of activation, but is also expressed on a small subpopulation of CD3
+ T cells [
20]; CD163, a scavenger receptor expressed on monocytes and macrophages and often used to represent activated anti-inflammatory cells [
21]; HLA-DR, the human major histocompatibility complex class II used as an activation marker on monocytes, but also as a marker of immune suppression when downregulated. HLA-DR can be expressed in activated T cells during severe infections [
22].
The stainings were performed in a routine flow cytometry laboratory at Skåne University Hospital Malmö, with internal controls for isotype or staining variations. Whole blood (50 μl) in each of the four tubes was incubated at room temperature with antibodies conjugated with the fluorochromes fluorescein isothiocyanate (FITC), phycoerythrin (PE), PE-Texas Red, allophycocyanin (APC) or Alexa-Fluor 647 and PE-cyanine7 (PE-Cy7) to permit up to five-colour analysis. The following antibodies were used: negative controls IgG1-PE or -FITC (clone 679.1Mc7; Beckman Coulter), CD45-ECD (clone J33; Beckman Coulter), CD3-APC, -PE, -ECD or -PCy5 (clone UCHT1; DakoCytomation), HLA-DR-FITC (clone L243; Becton Dickinson), CD8-ECD (clone SFCI21Thy2D3; Beckman Coulter), CD14-PE-Cy7 (clone RM052; Beckman Coulter), CD40-PE (clone MAB89; Immunotech Beckman Coulter), CD64-FITC (clone 22; Immunotech Beckman Coulter), CD11c-PE (clone BU15; Beckman Coulter). The combination of antibodies used were as follows for each tube: negative control-FITC/negative control-PE/CD45-PE-Texas Red/CD3-APC, HLA-DR-FITC/CD3-PE/CD8-PE-Texas Red/cholera toxin B-Alexa Fluor 647/CD14-PE-Cy7, HLA-DR-FITC/CD40-PE/CD3-PE-Texas Red/cholera toxin B-Alexa Fluor 647/CD14-PE-Cy7, CD64-FITC/CD11c-PE/CD45-PE-Texas Red/cholera toxin B-Alexa Fluor 647/CD14-PE-Cy7. The results for cholera toxin B and CD64 were scored only as part of a separate granulocyte study. Background autofluorescence and non-specific binding of mouse Ig were monitored for two of the colours (FITC and PE) with isotypic non-specific mouse Ig, separately for lymphocytes and monocytes; the FITC-labelled non-specific mouse IgG was also used for setting the cut-off for HLA-DR-positive reaction in T cells. Monocytes were defined as CD3−CD14+ cells with light-scattering typical for monocytes (when HLA-DR and CD40 were scored) or as CD14+ cells with CD45 intensity and light-scattering typical for monocytes (when CD11c was scored). The technical variation is low, as illustrated by the small variation observed for ten separate determinations of monocyte and lymphocyte HLA-DR (ten 50-μl blood fractions from one patient were labelled with anti-HLA-DR-FITC; data not shown). A lyse-no-wash protocol, or the automated Beckman Coulter TQprep machine, was used. In-house solutions for lysis and fixation were used. At least 1,000 events were analysed in an FC500 Beckman Coulter flow cytometer using both lasers. Acquisition and analysis were made using the CXP software (Beckman Coulter). Flow cytometry analysis of monocyte CD163 was performed as part of a separate study, on a subgroup of the shock patients (n = 6; four with Gram-positive and two with Gram-negative septic shock), using a FACSCalibur and the antibody conjugates CD14-FITC (clone M5E2), CD163-PE (clone GHI/61) and HLA-DR-APC (clone G46-6) from Becton Dickinson. An immune response index, intended to function as an indicator of overall immunoactivation or immunosuppression, was calculated for each patient. This index is based simply on those immune status parameters used in this study. It is, therefore, based on the results for three-monocyte (expression of HLA-DR, CD11c and CD40) and two-lymphocyte (frequency of activated CD3+ and CD8+ T cells) parameters. One point is awarded for a variation of one standard deviation from the mean of the healthy control result, up to a maximum of three standard deviations per parameter; for example, −0.8 SD for HLA-DR, −0.7 SD for CD11c, +3.1 SD for CD40, +5.6 SD for CD3+ T and +6.5 SD for CD8+ T give an index of +1.5, indicating overall immunoactivation.
Cytokine analysis
Plasma was fresh-frozen at −20 °C and then analysed for IL-6, IL-1β, IL-18, TNF-α and TIMP-1 by enzyme-linked immunosorbent assay (ELISA). The minimum detectable concentrations were: IL-6 0.70 pg/ml, IL-1β 10 pg/ml, IL-18 12.5 pg/ml, TNF-α 10 pg/ml and TIMP-1 0.08 ng/ml. All ELISA kits were from R&D Systems.
Statistical analysis
The non-parametric Wilcoxon’s two-sample rank test (the Mann–Whitney test) was used, and a p-value <0.05 was regarded as significant. Pearson’s correlation coefficient was calculated in order to measure the correlation (linear dependence) between two variables. The calculation of the immune response index was based on the standard deviation because the healthy control data were considered to reflect a normal distribution.
Discussion
Sepsis is considered to be a complex disorder, and it has been suggested that it is probably too heterogeneous to treat as one disease [
4]. One obvious reason is the variety of eliciting microbial agents carrying distinct sets of pathogenic factors. Many host factors influence the clinical picture, such as genetic determinants and co-morbidities influencing immune status. There can be variation even within the individual patient, as exemplified by an early appearing SIRS eliciting CARS [
4,
11,
25]. Even though viruses are not conventionally judged to cause sepsis, we decided to use virosis as the control subgroup due to the systemic effects elicited by the immune response in these patients.
Monocyte cytokine production and the expression of surface markers are well known to become depressed in many sepsis patients [
4‐
6,
11‐
13,
26,
27]. For instance, a weak production of pro-inflammatory cytokines by ex vivo-stimulated monocytes and a low level of cell surface CD14 expression were noted to correlate with sepsis severity [
28]. However, a more recent report also demonstrated a high level of immunosuppressive CD163
+ monocytes already upon admission, and with no difference 7 days later, and with a lack of correlation between the frequency of this monocyte subset and clinical outcome, illustrating the complexity of the immune response in sepsis [
29].
There was a wide variation in age between the study groups, with the oldest patients found in the group of septic shock. There is clear evidence that the incidence of sepsis increases slowly throughout most of adulthood, and the reasons for this is multifactoral [
30]. It is generally agreed that the senescence in the immune system affects the adaptive immune response, which leads to major defects in the cell-mediated immunity and the humoral immune response, whereas the innate immune response is largely spared during life [
31]. In this study, there were more males than females developing septic shock. It has been reported that the gender-related differences in the immune response could be part of the explanation for this behaviour [
32]. A predominance of anti-inflammatory mediators in women may have a protective effect in terms of developing severe sepsis and septic shock [
32].
Most of our patients had one or several abnormal findings among monocytes and/or T lymphocytes deviating strongly from that of all the 13 healthy controls. Signs of immunosuppression were most evident in septic shock, regarding both the number of patients and the size of the alteration. However, this was sporadically seen also in all the other diagnostic groups studied, including the non-septical virosis group. Indeed, it should be noted that the levels of antigen cell surface expression are overlapping in a lot of cases. Therefore, conclusions regarding the immune status are impossible to draw. Still, we believe that our data indicate that the suppression of some immune functions most probably occurs in a majority of patients with sepsis, as well as in other severe infections. However, our main finding is the marked inter-individual variation regarding the number and intensity of signs of immunoreactivity. A major former study on immune changes during acute sepsis in humans suffers from a bias in terms of the fact that only samples from deceased patients were analysed [
12]. The present study does not have that bias and suggests that, at least for the limited panel of markers used here, large inter-individual differences are evident. We emphasise the complexity of the homeostatic network operating in a sepsis patient that eventually decides whether or not the net result will be a functional clinically overt immunodeficiency [
33,
34]. These considerations are in line with a report appearing during the preparation of this manuscript [
35].
Monocytes are known to be exceptionally plastic in phenotype and function, with a great capacity to adapt quickly to environmental stimuli [
23,
36]. Monocyte HLA-DR, CD11c, CD40 and CD14 are typically present on all monocytes, and they were selected because a high expression level has been reported to be a measure of immune activation. To the best of our knowledge, only for one of them, i.e. HLA-DR, has low expression been associated with immunosuppression [
37]. However, we observe a clearly reduced cell surface intensity of HLA-DR, CD11c, and CD14 in comparison to the healthy control group in primarily septic shock and Gram-positive patients, in conjunction with a high level of immunosuppressive CD163
+ monocytes. This also correlated to an increased proportion of the immunosuppressive HLA-DR
−/low (%) monocyte population, in the septic shock and Gram-positive patient groups. In general, it was interesting to note that, in sharp contrast to the Gram-positive patient group, patients with Gram-negative sepsis had a trend towards a pro-inflammatory activated monocyte pattern. Whether this difference is due to different immune response patterns being elicited by the various pathogens or is due to the clinical severity of the disease is worth discussing. Importantly however, monocyte CD40 expression showed a marked variation in all cases and sepsis groups, including shock, thus indicating an active CD40:CD40L immune activation also in patients with a typical immunosuppressive monocyte phenotype regarding HLA-DR, CD11c and CD14.
T cells are not counted as antigen-presenting cells, and, thus, are considered to lack HLA-DR; nevertheless, it is common knowledge that, upon activation, both CD4
+ and CD8
+ T cells express this molecule on their surface [
22,
38]. We considered HLA-DR to be a suitable T cell activation marker for the present study, not only because of its strong fluorescence intensity, but also with regard to its in vivo expression (whereas, in our experience, the upregulation of e.g. CD25, i.e. the IL2R, can only rarely be documented in patients while being easily detected in T cells activated ex vivo) and to its kinetics, peaking relatively late (whereas e.g. CD69 appears already 2 h after stimulation). On average, none of the sepsis patient subgroups analysed had significantly altered proportions of activated T cells; however, especially in the shock group, there were cases with marked elevation or reduction, respectively, of activated T cells. Interestingly, the alternative activation marker for T cells, CD40 [
16,
39], was increased in all sepsis groups, suggesting further studies on this population in sepsis patients. Also, in the context of Tregs, this indicates that, although sepsis is a disease that, at the time points in this study (1–4 days after onset), would be expected to affect cells of the innate immune system primarily, we do actually see an activation of adaptive immunity.
Our observation of altered in vivo concentrations of cytokines in sepsis patients is in agreement with previous data [
35], and provides evidence that our cellular results reflect functional derangements. Patient serum cytokine levels with respect to the type of pathogen might be worth analysing in more detail in the future, as would IL-10 levels in general. There are certainly additional factors to be considered when the relevance for the clinical situation of our findings are assessed, for example, whether a cellular change in the blood compartment results from the recruitment of cells from major reservoirs such as spleen or from an altered cellular state or even selective apoptosis of certain populations. A recent post-mortem study on sepsis suggests that blood findings can mirror tissue status, because flow cytometric analysis of both spleen and lung cells showed the expansion of suppressor cell populations [
12]. A critical issue is which immunosuppressive characteristics indicate that the patient is in a net state of immunosuppression conferring increased susceptibility to microorganisms, and, thus, should be treated accordingly. Whether an isolated finding of an increase in an immunosuppressive monocyte marker such as HLA-DR, CD163 or programmed cell death protein 1 (PD-1) [
12,
40,
41] merit immunoactivating treatment is worth debating. The principal issue is: what amount of measurable phenotypic or functional cellular depression is needed in order for a clinically relevant immunodeficiency to occur. In conclusion, our findings of a pronounced inter-individual variation in the analysed monocyte and lymphocyte markers form a strong argument that, when immunomodulatory treatment is considered in a sepsis patient, it should be personalised and guided by a detailed immune status assessment.