G‐CSF as a potential early biomarker for diagnosis of bloodstream infection

Abstract Background Cytokines play an important role in bacterial infection, and thus, we aim to find out cytokines that may be diagnostically significant in early stage of bacterial bloodstream infection. Methods Mice models infected with Staphylococcus aureus and Klebsiella pneumoniae were established. Then dynamic changes of nine serum cytokines were monitored within 48 hours after the infection. Cytokines with significant differences between the infected groups and control group were further analyzed. Clinical samples of patients who were suspected of bloodstream infection were collected. Then the diagnostic efficiency of screened cytokines was determined with receiver operating characteristic curve analysis. Results As for mice models infected by Staphylococcus aureus and Klebsiella pneumoniae, six cytokines including IL‐1β, IL‐6, IL‐12p70, G‐CSF, IFN‐γ, and TNF‐α were significantly different (P < .05) between two bacterial infected groups. As for clinical samples, three cytokines including IL‐6, IL‐12p70, and G‐CSF showed significant differences between infection group (Staphylococcus aureus and Klebsiella pneumonia group) and negative control group. With the area under curve of 0.7350 and 0.6431 for G‐CSF and IL‐6, respectively, these two cytokines were significantly different between Staphylococcus aureus and Klebsiella pneumoniae infection groups. Combination of G‐CSF and IL‐6 could improve the AUC to 0.8136. Conclusions G‐CSF cannot only identify bacterial bloodstream infection, but can also distinguish the infection of Staphylococcus aureus from Klebsiella pneumoniae. Further investigation should be performed concerning the diagnostic efficiency of G‐CSF in diagnosing different types of bacterial bloodstream infection.


Although Matrix Assisted Laser Desorption Ionization Time-of-Flight
Mass Spectrometry (MALDI-TOF MS) has been used in clinical laboratories to shorten the time required for microbial identification, it still requires the formation of bacterial colony and would also take more than 2 days to generate positive results. BSI progresses rapidly, which leads to the fact that clinicians prefer to apply empiric broad-spectrum antimicrobial agents. Unfortunately, this practice contributes to the occurrence of multidrug-resistant pathogens and increases the economic burden of patients simultaneously. 1,2 Therefore, we are in urgent need of fast and reliable diagnostic tools to provide precise rapid treatment and to enable de-escalation antimicrobial treatment. 4 Thus, we have to explore efficient biomarkers at the early stage of BSI.
When pathogen enters the bloodstream, the innate immune response is activated by recognition of pathogen-associated molecule pattern (PAMP) such as peptides, lipopolysaccharide, and teichoic acid on the surface of bacteria. 5 The pattern recognition receptor (PRR) of innate immune cells subsequently transmits the signal to the nuclear, promoting or inhibiting the release of inflammatory cytokines. Therefore, inflammatory cytokines are potential candidates as early diagnostic markers of bacterial infection. Up until now, several bacterial infection markers including white blood cell (WBC), C-reactive protein (CRP), interleukin-6 (IL-6), vascular endothelial growth factor (VEGF), macrophage inflammatory protein-1β (MIP-1β), procalcitonin (PCT), tumor necrosis factorα (TNFα), IL-2, IL-8, and IL-10 have been considered as the potential biomarkers for BSI diagnosis. 6,7 In order to screen for diagnostic markers of bloodstream infection for clinical application, 32 cytokines and chemokines related to the bacterial infection were chosen as potential parameters in clinical samples from patients. In addition, we aimed to further investigate the diagnostic biomarkers, which could distinguish gram-negative bacteria from gram-positive bacteria using mouse BSI model. Thus, mice infected with Staphylococcus aureus or Klebsiella pneumonia was used as the animal model to monitor the dynamic variation pattern of several infection-related cytokines. Then bacterial suspension was prepared and quantified by Maxwell turbidity method. The resuspended bacteria were first diluted to 3.9

| ME THODS AND MATERIAL
Maxwell's turbidity. Then we performed a 10-fold dilution and diluted six concentrations sequentially. The diluted bacterial suspension was injected into the mice through tail vein, 5 mice in each group, with 6 groups in total. The injection volume was 0.1 mL/10 g. Mice behaviors, weight changes, and death rate were observed and recorded daily for consecutive 7 days. The LD50 was calculated by Karber method.

| Establishment of mice model with bloodstream infection
To ensure that mice were alive under the infection condition during In brief, the mice were inoculated with the 1/2 LD50 diluent of standard strain of S aureus or K pneumoniae in a final volume of 0.1 mL/10 g via tail veins. The time of challenge was designated as time 0 of the experiment. Blood from mice eye was collected 1, 3, 6, 12, 24, and 48 hours after inoculation and placed at 4°C for 8 hours to isolate the bacteria.  Inclusion criteria: Patients with positive blood culture results, with PCT > 0.5 mg/mL or IL-6 > 5.9 mg/mL or CRP > 0.8 mg/mL were included as BSI group. While patients with negative blood culture, with PCT < 0.5 mg/mL, IL-6 < 5.9 mg/mL, and CRP < 0.8 mg/ mL were recruited as negative control. Exclusion criteria: patients infected with multiple types of bacteria were removed.

| Measurement of the serum cytokine levels, C-reactive protein (CRP), and procalcitonin (PCT)
Serum levels of cytokines including IL-1β, IL-5, IL-6, IL-7, IL-12p40, IL-12p70, G-CSF, IFNγ, and TNFα were detected using the Luminex ® xMAP™ System (Millipore Corporation, Germany). The assays were conducted in strict accordance with the operating instructions of the instrument. CRP and PCT were measured during the first 48 hours after onset of fever. Serum CRP was measured through nephelometric method (Siemens BNII, Cardiophase, Germany) and PCT was measured by Cobas800 (Roche, Switzerland) based on electrochemical luminescence technology.

| Statistical analysis
The Kolmogorov-Smirnov or Shapiro-Wilk tests were used to assess the normality of distribution of parameters. Continuous variables were described using the median (inter-quartile range [

| LD50 of Staphylococcus aureus and Klebsiella pneumonia
The number of dead mice and the survival rates for each concentration were recorded and calculated (Table S1). Then Karber method 8 was used to determine the LD50 of the S aureus and K pneumoniae: 8.1 × 10 8 CFU/mL and 1.11 × 10 8 CFU/mL, respectively.

| Clinical symptoms and changes of weight after infection in mice
About 1 hour after the injection of bacteria, the mice in the experimental group showed symptoms of piloerection, closed eyes, and decreased activity. Three hours after injection, these behaviors were more obvious, with their bodies curling up and defecated loose stool. Twenty-four hours after injection, mice had significantly decreased diet and activities. Forty-eight hours after injection, death of mice occurred while the remaining alive ones returned to normal condition. In addition, weight of mice in S aureus group and K pneumoniae group decreased by an average of 3.9 g and 4.7 g, respectively 24 hours after infection when compared with the original weight. The weight of mice in both infected groups maintained at low levels during the first 48 hours and then gradually recovered 72 hours after infection (Figure 1).

| Kinetic changes of nine cytokines in S aureus and K pneumoniae mouse infection models
Kinetic changes of nine cytokines in serum collected from the three groups mice between 0 to 48 hours after injection are shown in Figure 3. All cytokines started to increase 0.5 hour after injection and decreased back to their initial level at 48 hours after reaching the highest peaks. The rate and quantity of the cytokines increased in K pneumonia group were higher than that of S aureus group after infection except for IL-12p70. Among the 9 cytokines, the substantial change of G-CSF is the same as IL-6 between 0.5 and 6 hours. After 6 hours, IL-6 dropped rapidly whereas G-CSF still remains high.
There were no significant changes for the levels of cytokines in control group during the experiment.

| Comparison of six cytokines in serum samples between mice models infected with S aureus and K pneumoniae
We defined 0-6 hours as the early stage of the infection after bacterial injection in mice models. During this stage, the statistical differences were found for G-CSF between two infected groups at 1, 3, and 6 hours. For IL-6, the difference occurred at 0.5, 1, 3, and 6 hours. For TNFα, the difference was at 1 hours, while for IL-12p70, IFNγ, and IL-1β, it was at 6 hours (P < .05). However, for IL-5, IL-7 and IL-12p40,   Table 1. Commonly used diagnostic biomarkers in clinical laboratory including CRP and PCT were not statistically different between S aureus group and K pneumoniae group. As for S aureus group, mean level of CRP and PCT were 11.01 ± 4.86 mg/dL and 1.98 ± 0.56 ng/mL, respectively. As for K pneumoniae group, mean level of CRP and PCT were 10.36 ± 3.79 mg/dL and 2.34 ± 0.51 ng/ mL, respectively. While CRP and PCT in both gram-positive bacterial group and gram-negative bacterial group were significantly higher than those in negative control group (CRP: 0.56 ± 0.21 mg/ dL, PCT: 0.35 ± 0.11 ng/mL).

| The cutoff values and diagnostic efficiencies for the potential biomarkers for bacterial bloodstream infection
The results of six cytokines (IL-1β, IL-6, IL-12p70, G-CSF, IFNγ, and TNFα) were shown as mean ± standard deviation and the differences among the S aureus group, K pneumoniae group and negative control group were shown in

F I G U R E 3
The kinetic changes of nine cytokines in the serum of infected mice models among Staphylococcus aureus, Klebsiella pneumoniae and negative control groups. *P < .05 comparing S aureus group with negative control group. #P < .05 comparing K pneumoniae group with negative control group. §P < .05 comparing S aureus group with K pneumoniae group sensitivities, and specificities for these six cytokines in bacterial bloodstream infection were shown in Table 2.

| DISCUSS ION
In recent years, the morbidity of sepsis caused by bacteria infection has increased. 9 Lack of timely diagnosis and initial control, together   13,14 has also been considered as a biomarker for infection. Therefore, it is essential to investigate a novel panel of biomarkers for bacterial infection. 15,16 Among all these biomarkers for infection, interleukin 6 (IL-6) is thought to be more widely used in clinical laboratories. IL-6 is an interleukin that acts as both a pro-inflammatory cytokine and an antiinflammatory one, 17 which is secreted by T cells and macrophages Endocrine system disease 2 (2.9%) 1 (4.3%) -NS Gynecological disease 5 (7.5%) 1 (4.3%) -NS Heat shock 0 1 (4.3%) -NS Trauma 6 (9.0%) 1 (4.3%) -<.05 Note: Indicate a significant statistical difference between S aureus group and K pneumoniae group.

TA B L E 1 Clinical characteristics of clinical patients infected with
Staphylococcus aureus and Klebsiella pneumoniae and negative control group to stimulate immune response. Thus, it plays an important role in fighting infection and is often used clinically as an inflammatory biomarker. In our current study, the efficiency of IL-6 is consistent with the previous literature, 11 which could discriminate between infectious condition and non-infectious condition.
Interestingly, our results also showed that the changing tendency of G-CSF is not only similar to IL-6, but also increased more significantly than IL-6 in infected mice models.
G-CSF reached its peak rapidly 1 hour after infection and this increase lasted for 24 hours, then it rapidly descended to the initial level at 48 hours. The changes of G-CSF in S aureus infected group were more slightly when compared with K pneumoniae infected group (P < .01). This might be attributed to the more wild response of mice to S aureus infection than K pneumonia. G-CSF is a glycoprotein that stimulates bone marrow to produce granulocytes and stem cells, which are then released into the bloodstream. 18

| CON CLUS IONS
Based on the kinetic changes of these nine cytokines in animal models and the validation of six cytokines in clinical samples, G-CSF is con-