Purpose

Timely knowledge of the bacterial or viral etiology of infection is necessary for appropriate treatment, including empirical antibiotic therapy. Because the clinical symptoms of severe viral and bacterial infections are often similar, clinicians need reliable diagnostic biomarkers in order to make treatment decisions. Currently, C-reactive protein (CRP), white blood cell (WBC) counts, and neutrophil counts are the most frequently used biomarkers in daily clinical practice [1]. CRP is considered to be the most sensitive biomarker of bacterial infection in both outpatient and inpatient settings [2]. However, high CRP levels, increased WBC counts, or raised neutrophil numbers cannot always differentiate between bacterial, fungal, and severe viral infections [3].

Because of these well-known limitations, other serum biomarkers of bacterial infection, including the serum levels of procalcitonin (PCT), interleukin (IL)-6, and IL-8, have been recommended for use in parallel with old biomarkers [4–6]. Moreover, many other potential biomarkers have been demonstrated in different models of experimental infection and in human bacterial sepsis, including IL-1β, IL-12, tumor necrosis factor α (TNF-α), interferon (IFN)-γ, cortisol, heparin-binding protein (HBP), soluble CD14 (sCD14), and expression markers on the surface of circulating monocytes, including CD14, HLA-DR, toll-like receptor (TLR)2, and TLR4 [7–10]. In addition, some of the above-mentioned biomarkers were also evaluated for the early differentiation of bacterial etiology and focus of the infection [11, 12].

Therefore, the aim of our study was to evaluate these novel potential biomarkers, which were simultaneously analyzed along with routine biomarkers in order to compare their sensitivity and specificity for the differentiation between viral and bacterial infections. Furthermore, the biomarkers were also tested based on the focus of the bacterial infection.

Patients and methods

Patients

This prospective study was conducted after obtaining approval from the Ethics Committee of the University Hospital Bulovka (IRB00002721—Fakultní nemocnice Na Bulovce IRB #1—Biomedical) from April 2007 to September 2009. Adult patients (18–80 years old) presenting fever (axillary temperature ≥38°C) and a clinical diagnosis of infection (see below) were eligible for inclusion in this study if they were admitted to the standard wards of the Department of Infectious Diseases. Primary exclusion criteria consisted of antibiotic therapy prior to hospitalization, transfer from another institution, autoimmune diseases, corticosteroid treatment, malignancy, and HIV infection. All patients gave written informed consent prior to their inclusion in the study, and paired samples of whole peripheral blood and serum were obtained upon the admission to the study.

The diagnosis of bacterial infection was made clinically based on the findings of focal infection. Thus, bacterial pneumonia was confirmed by X-ray showing lobar infiltrate; urosepsis was confirmed with the detection of pathogenic bacteria in the blood, together with signs of urinary tract infection and pyelonephritis, with increased number of WBC count in the urine, ultrasonography findings, and/or elevated CRP; erysipelas and cellulitis was confirmed by characteristic skin signs (i.e., erythematous lesions that are clearly demarcated from the normal skin); invasive meningococcal disease (IMD) was confirmed with petechial rash, fever, and detection of the etiologic agent; toxic shock syndrome and sepsis were confirmed based on the accepted clinical criteria [13–15].

The particular bacterial etiology of the illness was confirmed by cultivation of the pathogenic bacteria from a sterile body fluid (i.e., blood and urine), detection of pneumococcal antigen in urine, detection of bacterial DNA in blood or cerebrospinal fluid (CSF), and serology. Furthermore, all of the patients from this cohort were treated with antibiotics, and a significant decrease in inflammatory markers and clinical improvement were documented. The demographic and clinical data of the 54 enrolled patients with bacterial infection are presented in Table 1. Bacterial etiology was laboratory confirmed in 29 patients, which represented 53.9% of the cases enrolled in this cohort. In these patients, the following etiological agents were detected by cultivation: Escherichia coli 16 times (urine seven times, urine and blood eight times, and blood once), Streptococcus pneumoniae four times, Haemophilus parainfluenzae once, and Staphylococcus hominis once (all detected in the blood only). Pneumococcal urinary antigen assay was positive in two patients (in both cases, S. pneumoniae was also detected in the blood), and polymerase chain reaction (PCR) detected the DNA of Neisseria meningitidis in the blood or CSF from two patients. In addition, a significant rise in antibodies against Chlamydophila pneumoniae indicated an acute infection in three patients, and antibodies against Legionella pneumophila indicated acute infection in one patient.

Table 1 Demographic and clinical data of patients with bacterial infection
Full size table

For comparison, 27 patients with common, acute viral infections that included tick-borne encephalitis (TBE), enteroviral meningitis, viral hepatitis A (VH A), varicella, and parvovirosis were enrolled. In these patients, characteristic clinical and laboratory findings were present (e.g., meningeal irritation and lymphocytic pleocytosis in CSF in TBE or enteroviral meningitis, jaundice, and elevated liver function test in VH A or typical skin findings of clear vesicles on the head and trunk in varicella). The clinical diagnosis was subsequently confirmed with serology, demonstrating specific IgM antibodies. In addition, the etiological diagnosis of enteroviral meningitis was made based on the detection of the viral RNA in the inflammatory CSF. Furthermore, CRP serum levels in this cohort were either within the normal range or mildly elevated, and the patients’ clinical conditions improved without antibiotic treatment. The demographic and clinical characteristics of the 27 patients with viral infection are shown in Table 2.

Table 2 Demographic and clinical data of patients with viral infection
Full size table

For the assessment of the influence of the focus of bacterial infection on selected biomarkers, the data of 21 patients with community-acquired bacterial pneumonia (CABP) were compared with the data obtained from 21 cases with pyelonephritis or urosepsis. The clinical and laboratory data of these patients are shown in Table 3.

Table 3 Comparison of demographic, clinical, and laboratory data of 42 patients with focal bacterial infection
Full size table

WBC and differential blood counts

WBC and differential blood counts were determined using a Coulter STKS clinical analyzer (Coulter Electronics Inc., Miami, FL, USA).

Serum levels of soluble mediators

Serum concentrations of CRP were measured on an AU 2700 analyzer (Olympus) using immunoturbidimetry (CRP Latex™, Olympus) with a normal range of 0–8 mg/l. PCT levels were obtained using the enzyme-linked fluorescent assay (ELFA) technique. This assay combines a one-step immunoassay sandwich method with a final automatic fluorescent detection by a VIDAS instrument (VIDAS B.R.A.H.M.S. PCT, bioMérieux, Lyon, France). The measurement range of this instrument was set to 0.05–200 ng/ml.

Serum concentrations of interleukin-1β (IL-1β), IL-6, IL-8, IL-10, IL-12, and TNF-α were measured using a CBA kit (BD™ Cytometric Bead Array—Human Inflammatory cytokine kit) and a three-color FACSCalibur™ flow cytometer (all BD Biosciences, San Jose, CA, USA). The detection limit of the methods for the cytokines was 20 pg/ml. Enzyme-linked immunosorbent assays (ELISAs) were used to measure the IFN-γ (BD Pharmingen, CA, USA), sCD14 (R&D, CA, USA), and HBP levels [16]. The analysis of cortisol was performed on an Architect i2000 immunochemistry analyzer (Abbott, Chicago, IL, USA) using the fluorescence polarization immunoassay FPIA (Abbott), with a normal range of 100–536 nmol/l.

Monocyte surface marker expression

Flow cytometric quantitative analysis of TLR2, TLR4, HLA-DR, and CD14 expression was performed on circulating monocytes using the monoclonal antibodies anti-TLR2, anti-TLR4 (eBiosciences, San Diego, CA, USA), anti-CD14, and anti-HLA-DR (BD Biosciences, San Jose, CA, USA), and the analysis was calculated as a relative mean fluorescent intensity.

Statistical analyses

Statistical analyses were performed using SPSS software™ by a certified biomedical statistician. Data are presented as the medians (interquartile range). Levels that were undetectable were assigned a value equal to the lower limit of detection for the assay. The differences in the analyzed parameters between groups were tested by the Mann–Whitney U-test. The analyses consisted of two-tailed tests with an α-level <0.05. Receiver operating characteristic (ROC) curves were drawn for selected variables as a measure of discriminating power between bacterial and viral infections. The ROC curve shows (on the x-axis) a false-positive rate and (on the y-axis) the sensitivity of a test. The areas under the curves (AUCs) were also evaluated. The correlations among the parameters were tested using Spearman correlation tests.

Results

The groups of patients with viral and bacterial infections did not significantly differ in their mean age and the duration of fever; however, the length of the hospital stay was significantly longer (P = 0.027) in the case of bacterial infection. The results of laboratory analyses and parameter differences between the groups of patients with bacterial and viral infections, including the AUC values, are presented in Table 4. The highest sensitivity and specificity for the diagnosis of bacterial infection demonstrated increased serum levels of PCT and neutrophil blood counts, reduced numbers of circulating lymphocytes, elevated serum concentrations of IL-6, HBP, Cort, and sCD14, as well as increased CD14 expression on blood monocytes; the ROC curves of these markers are presented in Fig. 1. A high sensitivity and specificity for the discrimination of bacterial infection was also found for CD14 expression and IL-8 levels. In addition, serum concentrations of IFN-γ, TNF-α, IL-1β, IL-10, and IL-12 and the monocyte expression of HLA-DR, TLR2, and TLR4 revealed the lack of a difference between the patients with bacterial infections versus the patients with viral infections.

Table 4 Comparison of laboratory parameters in patients with bacterial and viral infections
Full size table
Fig. 1
figure 1

Receiver operating characteristic (ROC) curves of selected biomarkers with high diagnostic value for discrimination between bacterial and viral infection

Full size image

Regarding the focus of the bacterial infection, the serum concentration of IFN-γ was significantly higher in patients with CABP in comparison to patients with pyelonephritis or urosepsis. Similarly, Cort was significantly higher in CABP cases when compared to pyelonephritis or urosepsis. Conversely, IL-8 was higher in patients with pyelonephritis and urosepsis in comparison to CABP cases. The data obtained from both groups of patients and their statistical comparisons are presented in Table 3.

Also, correlations among the parameters were evaluated in the cohorts with bacterial and viral infections. The strongest (r > 0.6) and most significant correlations (P < 0.001) in the cohort with bacterial infection were the following: PCT with IL-6 and IL-8 (r = 0.724; r = 0.633, respectively), IL-6 with IL-8 (r = 0.833), and TLR4 expression with CD14 and TLR2 (r = 0.688 in both cases). The only correlation observed in the cohort with viral infection was between TLR2 and TLR4 expression (r = 0.851, P < 0.001).

Discussion

Several biomarkers were tested with the aim of identifying optimal parameters for diagnostic use in routine clinical practice. We tested the biomarkers in febrile patients that were hospitalized in the Department of Infectious Diseases for a severe course of bacterial or viral infection. These biomarkers included soluble serum molecules and surface markers expressed on circulating monocytes.

PCT was found to have the highest sensitivity and specificity for bacterial infections in this study. This result supports findings from previous studies, where elevated PCT values indicated systemic bacterial infections [17, 18]. It has also been shown that PCT is produced shortly after stimulation, within 6 h, in comparison to 24 h for CRP. Thus, PCT may provide a helpful tool for the rapid detection of bacterial infections, which has already been demonstrated in children with IMD examined for fever in emergency care departments [19]. Furthermore, the PCT-based diagnostic algorithm supports the cautious use of antibiotics and has a favorable effect on clinical outcomes [20].

The second most reliable soluble biomarker that can discriminate viral and bacterial infections evaluated in our study was HBP. Elevated serum levels of HBP, which is stored in leukocytes and rapidly released after their stimulation, were already demonstrated in IMD, leptospirosis, and tropical malaria, as well as in severe sepsis and septic shock [8, 21]. However, the difference of HBP serum levels in association with the etiology of infection is a novel observation, which may indicate a role of this protein during less severe infectious processes.

From the other soluble biomarkers, we observed a good sensitivity and specificity for IL-6, Cort, and sCD14. IL-6 is a proinflammatory cytokine produced by monocytes and macrophages activated during bacterial infection, although it exerts its function more broadly due to its anti-inflammatory properties by activating inhibitory mechanisms of the acute-phase response. It has been documented that IL-6 is elevated in blood earlier than CRP during the course of bacterial infection [22]. Noor et al. [23] compared the usefulness of IL-6 with CRP in neonatal sepsis and discovered that IL-6 was the earliest marker of neonatal infection. However, we did not observe IL-6 in more than one-third of our patients with bacterial infection (data not shown), suggesting the limitations of IL-6 in less severe bacterial infections. The finding of higher Cort concentrations in bacterial infections in comparison to viral infections is in accordance with our earlier observations in patients with meningitis; cortisol levels in CSF were significantly higher in patients with bacterial meningitis in comparison to patients with viral meningitis. It is worth noting that cortisol levels of 46.1 nmol/l in CSF had 82% sensitivity and 100% specificity for the diagnosis of bacterial meningitis [24]. The importance of S-Cort observed in this study for the distinction between bacterial and viral infections has not yet been described.

It is interesting that both the CD14 expression on circulating monocytes and the serum level of sCD14 proved to be good biomarkers for the differentiation between bacterial and viral infections. Similar to membrane-bound CD14 (mCD14), sCD14 plays a crucial role in host defense against Gram-negative and Gram-positive pathogens, as sCD14 functions as a pattern recognition receptor of the innate immune response [25]. High levels of both mCD14 and sCD14 have been associated with a poor outcome of sepsis. Aalto et al. [26] measured mCD14 and sCD14 in patients with severe community-acquired infections and found a significant positive correlation between S-CRP and sCD14 levels, providing evidence that sCD14 may serve as an acute-phase reactant.

Regarding the differences in the observed results based on the focus of the infection, the highest serum levels of IFN-γ and Cort in CABP in comparison to pyelonephritis and urosepsis might be of interest. However, it is worth noting that the highest serum IFN-γ concentrations were observed in patients with pneumonia due to C. pneumoniae or CABP of unknown origin (data not shown). The high IFN-γ serum levels in CABP of chlamydial etiology may indicate an activation of intracellular killing mechanisms that are important for the elimination of the bacteria [27]. Next, the patients with the highest Cort levels had bacteremic pneumococcal pneumonia, reflecting a more severe course of infection compared to non-bacteremic pneumonia. In addition, higher IL-8 levels observed in pyelonephritis and urosepsis in comparison to CABP are suggestive of the Gram-negative etiology of infection and stimulatory effects of endotoxin [28].

It is not surprising that we observed a high discriminating power (AUC 0.852) of an increased number of neutrophils during bacterial infection in our study. This finding only stresses the well-known fact that the number of circulating neutrophils is a much better marker of bacterial infection than the number of WBCs. However, the observed decreased number of circulating lymphocytes predicting the bacterial origin of infection (AUC 0.841) is interesting because circulating lymphocytes are not frequently used by clinicians. The reason for the poorly estimated value of low lymphocyte numbers in bacterial infections could be that most authors describe lymphocytopenia in association with severe sepsis and septic shock [29].

Though TLRs are rapidly responding pattern-recognition receptors, we did not observe a significant modulation of TLR2 and TLR4 expression by distinct infections; therefore, the expression levels of these receptors do not seem to be useful for distinguishing between bacterial and viral infections. The same trend holds true for mCD14, which is modulated by experimental endotoxemia, and HLA-DR, which is altered in patients with a poor outcome of severe sepsis and septic shock [30]. Similarly, the early biomarkers that are rapidly produced after the stimulation of monocytes (i.e., IL-1β, IL-12, TNF-α, and IFN-γ) did not differ between viral and bacterial infections, which could be explained by the fact that the above-mentioned cytokines have rapid kinetics and their production is down-regulated shortly after stimulation [31].

We are aware of certain limitations to our study. First, the final etiological diagnosis was based not only on the results of microbiology methods, but also on the clinical decision supported by the findings of focal infection, elevated CRP serum levels (in the case of pyelonephritis), the clinical course, and the effects of empirical antibiotic therapy. Although this study design respects common clinical practices, it might lead to a bias. Thus, the clinical diagnosis was finally reviewed by an expert (i.e., an infectious disease consultant) who was not involved in the patients’ care. Second, we could not reliably compare CRP with the other biomarkers because this parameter was used for the diagnosis of pyelonephritis. Third, because we had to use quite stringent enrollment criteria in order to eliminate factors that could influence selected biomarkers and because the cohorts of patients were relatively small, the results have limited applicability. Therefore, further clinical studies with larger cohorts of patients should be performed to confirm the reported findings.

Conclusion

Our results strongly support the use of the procalcitonin (PCT) serum level as a routine biomarker of bacterial infection. Furthermore, some of the novel potential biomarkers (i.e., heparin-binding protein [HBP], interleukin-6 [IL-6], cortisol [Cort], interferon [IFN]-γ, and IL-8) demonstrated interesting differences based on the etiology and localization of infection, suggesting potential for their future clinical use.