Time-course of Toll-like receptor 2 expression, as a predictor of recurrence in patients with bacterial infectious diseases

Abstract

The clinical course of bacterial infectious diseases is often variable, especially in elderly patients. Thus, new biological markers have been sought to predict the disease outcome. Recent studies have revealed that Toll-like receptor (TLR) 2 and/or TLR4 on circulating monocytes are significantly up-regulated in bacterial infections. However, the lack of reliable quantification methods hampers extensive study on the modulation of these molecules in response to the patient's clinical condition. In this study, we developed a new quantitative flow cytometric analysis system for TLR2. We then carried out a longitudinal study on TLR2 expression levels on monocytes from patients suffering from bacterial infectious diseases during and after antibiotic treatment. The clinical outcome divided 37 patients into ‘cure’ (n = 24) and ‘recurrence’ (n = 13) groups. A significant difference between the two groups was recognized in the TLR2 levels just after antibiotic treatment (antibody-binding sites/cell, 4395 ± 784 versus 5794 ± 1484, P < 0·001). The risk of recurrence was associated significantly with TLR2 (P < 0·001), but not C-reactive protein (P = 0·351) levels assayed during the first remission. Furthermore, antibiotic effectiveness was associated inversely with TLR2 levels during antibiotic administration (P < 0·001). Taken together, TLR2 expression levels on monocytes provide critical information for planning treatment against bacterial infectious diseases.

Introduction

The innate immune system is a universal and ancient type of host defence mechanism against infections, which distinguishes infectious ‘non-self’ from non-infectious-‘self’. Toll-like receptors (TLRs) play a major role in pathogen recognition and initiation of both innate and acquired immune responses [1,2]. Currently, more than 10 kinds of TLRs have been identified, and each is devoted to recognizing a distinct set of molecular patterns not found in normal vertebrates. In humans, 10 TLRs (TLR1–10) have been identified. TLR4 plays essential roles in cellular responses to Gram-negative bacterial lipopolysaccharide (LPS) [3–5]. TLR2 and its co-receptors TLR1 and TLR6 recognize a wide range of bacterial products, such as lipoteichoic acid, lipoarabinomannan and diacylated lipopeptide [6–8]. Bacterial flagellar protein flagellin is a specific ligand of TLR5 [9]. TLR3, TLR7/8 and TLR9 recognize viral double-stranded RNA, small interfering RNA and unmethylated CpG DNA, respectively [10–12]. TLR10 is highly expressed in B cells and lung, and weakly expressed in plasmacytoid dendritic cells and lymphoid tissues such as spleen, lymph nodes, thymus and tonsils [13,14]. However, specific ligands for TLR10 are unknown. Regarding cellular localization, TLR1, TLR2, TLR4, TLR5 and TLR6 are located on the cell surface. Meanwhile, a subset of TLRs (TLR3, TLR7, TLR8 and TLR9), which sense viral and bacterial nucleic acids, are in endosomal compartments [15].

Recently, modulation of TLRs expression levels on the cell surface has been the focus from a clinical point of view. Several studies have reported that TLR2 and/or TLR4 on circulating monocytes and granulocytes are up-regulated significantly in various disorders including bacterial and viral infections, compared with those in healthy volunteers [16–26]. On the other hand, it is well known that pre-exposure to endotoxin renders monocytes and granulocytes unresponsive to stimulation with a secondary endotoxin challenge, the phenomenon called ‘endotoxin tolerance’. Some reports have demonstrated that endotoxin tolerance may be attributed to the modulation of TLRs expression levels [27–30].

From these findings, we hypothesized that TLRs expression levels on monocytes or granulocytes might be modulated transitionally according to clinical conditions during the course of infectious diseases. To test this hypothesis, it is essential to generate data on TLR expression levels obtained at different time-points in individual patients and to compare the data. However, conventional flow cytometric methods cannot be applied to this purpose due to assay-to-assay variation caused by daily fluctuation in the performance of flow cytometers and anti-TLR monoclonal antibodies (MoAbs). Therefore, it was necessary to develop a new reliable quantification system for TLRs.

In this study, we developed a new quantification system for TLR2, and investigated longitudinally in order to understand the relationship between the clinical course and TLR2 levels on monocytes from patients suffering from infectious diseases. This is the first follow-up study on the modulation of TLR expression levels on monocytes. The results showed several important characteristics of TLR2 expression during the course of bacterial infectious diseases.

Materials and methods

Patient population and blood collection

Volunteers were confirmed at the time of sampling that they did not have any infectious diseases for at least a month and they did not have a fever within a week from the time the blood sample was drawn. Simultaneously, white blood cell counts (WBC), C-reactive protein (CRP) levels and other inflammatory/biochemical parameters were checked, and volunteers with abnormal values were excluded. Patients who were hospitalized between May 2005 and July 2006, and diagnosed as suffering from various bacterial infectious diseases such as sepsis, pneumonia, enteritis and pyelitis, were enrolled into this study (Table 1). Informed consent was obtained from all patients and volunteers, which was in accordance with protocol approved by the Kagoshima University Ethics Committee. Ten ml of peripheral blood were taken from each subject with a heparinized blood collecting tube.

Table 1.

Patient population.

				Infecting organism
Diagnosis	No. of patients	Age range (mean)	Male/female	G(+)	G(–)	Mixed	Unclear
Pneumonia	22	31–98 (73)	10/12	15	1	5	1
Aspiration pneumonia	1	(78)	1/0	0	1	0	0
Enteritis	9	62–90 (78)	5/4	0	8	1	0
Pyelitis	3	58–78 (68)	3/0	3	0	0	0
Sepsis	3	55–95 (74)	1/2	2	1	0	0
Endocarditis	1	(40)	0/1	1	0	0	0
Pleuritis	2	92–93 (93)	1/1	2	0	0	0
Cholecystitis	1	(85)	0/1	0	1	0	0
Peridontitis	1	(42)	1/0	1	0	0	0
Tonsillitis	1	(30)	1/0	1	0	0	0
Subcutaneous infection	1	(84)	0/1	0	0	1	0
Total	45	30–98 (74)	23/22	25	12	7	1

G(+): Gram-positive bacteria, G(–): Gram-negative bacteria. Mixed: mixed infection with both Gram-positive and -negative bacteria.

Generation of recombinant TLR2

The extracellular region of TLR2 (corresponding to amino acid M¹ to T⁵⁸⁸) was cloned by reverse transcription–polymerase chain reaction (RT–PCR) using poly(A)⁺ RNA from CD14⁺ peripheral blood mononuclear cells (PBMC) from a healthy volunteer as a template. A sense primer, containing a HindIII site and Kozak consensus sequence (5′-TTTAAAAGCTTGCCGCCATGCCACATACTTTGTGGATGGTG), and an anti-sense primer, containing a 6xHis sequence and NotI site (5′-AAAAAGCGGCCGCTAGTGATGGTGATGGTGATGTGTCCTGTGACATTCCGA), were used to amplify the cDNA. The PCR product was digested with HindIII and NotI, and inserted into the HindIII/NotI sites of the pRc/CMV vector (Invitrogen, Carlsbad, CA, USA). The resulting expression vector was introduced stably into 293 cells. A subclone of the transfected 293 cells, which released the recombinant protein into the culture medium at the highest concentration, was selected by a limiting dilution method, and the resulting clone was expanded in 293F medium (Invitrogen). The recombinant TLR2 protein, which is modified with a His tag at its C-terminus, was purified to more than 95% homogeneity from the culture supernatant by Ni-NTA (Qiagen, Valencia, CA, USA), Mono Q (Amersham Biosciences, GE Healthcare, UK) and TALON (Clontech, Mountain View, CA, USA) affinity column chromatography.

Generation of TLR2-coupled standard beads

The TLR2 recombinant protein was coupled to 6 µm polystyrene amino-beads (Polyscience, Warrington, PA, USA) at different concentrations (0·1–3 mg/ml) by using glutaraldehyde as a cross-linking reagent. The coupling reaction was stopped by the addition of ethanolamine and the beads were blocked in PBS containing 0·5% bovine serum albumin (BSA) and 0·05% NaN₃ (blocking buffer). A series of standard beads that covered the expression range of TLR2 on monocytes were selected (Fig. 1a–c), and the average number of antibody-binding sites per bead was determined for each standard bead preparation by Scatchard plot analysis using ¹²⁵I-labelled anti-TLR2 antibody (clone T2·1; eBIOSCIENSE, San Diego, CA, USA). The beads were stored in aliquots at − 80°C. No significant loss in the number of antibody binding sites per bead and the affinities of the antibody was seen during the storage period (more than 6 months) as judged by repeated Scatchard plot analysis. In every assay, an aliquot of the beads was thawed and used as standard beads.

Fig. 1.

Quantification of Toll-like receptor (TLR) 2 expression levels on monocytes by flow cytometry. Analysis steps with a patient are shown. Recombinant TLR2-coupled standard beads (left panels) and peripheral blood mononuclear cells (PBMC) (right panels) were stained with phycoerythrin (PE)-labelled anti-TLR2 or control IgG2a, PE-labelled anti-CD14, anti-TLR2 or control IgG2a, respectively. The stained standard beads were gated according to the forward/side-scatter pattern as shown in panel a [region 1(R1)], and were analysed for their mean fluorescence intensity (MFI) values in TLR2 (b) and control (c) staining. The net MFI values of each standard bead were obtained by subtracting the MFI value in control staining [M1 in (c)] from MFI values in TLR2 staining [M1–M4 in (b)]. The standard curve was generated by plotting the net MFI values against the known numbers of antibody binding sites per bead (d). For PBMC, CD14-positive cells were defined first as shown in (f) [region 1(R1)] by using the CD14-stained sample. Next, these R1-gated cells are merged [black dots in (e)] into forward/side-scatter plots of total PBMC [grey dots in (e)] in order to set a gate for monocytes [region 2 (R2) in (e)]. Using this monocytes gate setting [R2 in (e)], MFI values for TLR2 [M1 in (g)] and control [M1 in (h)] were determined.

Analysis of TLR2 expression levels by flow cytometry

Monoclonal antibodies used in the flow cytometric analysis were phycoerythrin (PE)-conjugated anti-TLR2 MoAb (clone T2·1; eBIOSCIENSE), PE-labelled anti-CD14 MoAb (clone M5E2; eBIOSCIENSE) and PE-labelled control mouse IgG2a (eBM2a; eBIOSCIENSE). PBMCs were isolated from heparinized peripheral blood by density gradient centrifugation using Ficoll-Paque plus (Amersham). In this study, we employed a single-colour flow cytometric analysis to avoid problems of interference between fluorescence dyes. To assess the expression levels of TLR2 on monocytes, PBMCs were divided into three tubes and stained in parallel with PE-labelled anti-TLR2 MoAb, anti-CD14 MoAb or control mouse IgG2a in blocking buffer for 30 min at room temperature. The CD14-stained sample was used to ensure monocytes gating. The stained cells were analysed on a fluorescence activated cell sorter (FACS)Calibur flow cytometer using CellQuest software (Becton-Dickinson Biosciences, San Jose, CA, USA). For each donor, monocytes were first gated according to the forward/side-scatter properties and CD14 staining, which was merged into the forward/side-scatter plots, by using the CD14-stained PBMC sample (Fig. 1e,f). Subsequently, the same gate setting for monocytes was applied to the analysis of TLR2- and control-stained PBMC samples (Fig. 1g,h). Net mean fluorescence intensity (MFI) value for TLR2 was obtained by subtracting the MFI value from the control staining from the MFI value from the TLR2 staining. In order to quantify the TLR2 molecules on the monocyte surface, a mixture of TLR2-coupled standard beads was stained in parallel with PBMCs under the same experimental conditions. A calibration curve was generated in each assay by plotting net MFI values of the standard beads against their values of antibody-binding sites per bead, which had been determined by Scatchard plot analysis as shown in Fig. 1d. Using this calibration curve, the net MFI value of monocytes was converted to the number of antibody binding sites per cell.

Statistical analysis

Comparison of two groups was performed using Student's t-test if the distributions were normal and homoscedastic. If the distributions were not normal and/or homoscedastic, the Mann–Whitney U-test was applied. Multiple comparisons of more than two data sets were performed by the Steel–Dwass test. Fisher's exact test (extended) was applied to the test of dependence in Tables 2 and 3.

Table 2.

Recurrence rates of bacterial infectious disease in patient groups classified according to C-reactive protein (CRP) levels when antibiotics are discontinued.

	CRP level (mg/dl)
	< 0·5	0·5–1·0	> 1·0
% Recurrence rate	26·1%	50·0%	50·0%
(n/N)	(6/23)	(5/10)	(2/4)

n = relapsed patient number, N = relevant patient number.

Table 3.

Recurrence rates of bacterial infectious disease in patient groups classified according to Toll-like receptor (TLR) 2 expression levels when antibiotics are discontinued.

	TLR2 level (sites/cell)
	< 4395	4395–5179	5180–5964	> 5964
% Recurrence rate	6·7%	27·3%	66·7%	100%
(n/N)	(1/15)	(3/11)	(4/6)	(5/5)

n = relapsed patient number, N = relevant patient number.

Results

Quantification of TLR2 molecules on monocytes using recombinant TLR2-coupled standard beads

In a preliminary study, we examined expression levels of TLR1, TLR2 and TLR4 on monocytes from patients with bacterial infectious diseases by conventional flow cytometry using MFI values as an indication of TLR expression levels. The results demonstrated that patients generally showed higher TLR2 levels on their monocytes compared to healthy donors, which has also been shown in other studies [16,17,22]. In contrast, expression levels of TLR1 and TLR4 were very low among individuals and definitive results were not obtained in our preliminary experiments. These results prompted us to examine in detail TLR2 expression levels on monocytes in various disease settings. However, in conventional flow cytometric analysis, it is generally difficult to monitor precisely the expression levels of an antigen because of high interassay variation. Consequently, to assess TLR2 expression levels quantitatively, we generated a set of standard beads to calibrate the flow cytometric analysis, in which recombinant TLR2 protein was coupled at different densities on the bead surface. In this system, the number of TLR2 molecules on a cell was approximated by measuring the number of anti-TLR2 antibody binding sites on the cell. The average number of antibody-binding sites in each standard bead was determined from data in two to four independent Scatchard plot analysis experiments. They were estimated as 364 ± 79 sites for low beads, 1229 ± 30 sites for low–medium beads, 3320 ± 1154 sites for high–medium beads and 14067 ± 1884 sites for high beads. Scatchard plot analysis also showed that dissociation constant (Kd) values for the recombinant TLR2 ranged from 2·1 to 16·6 × 10⁻¹⁰ M. In the parallel experiments, those for native TLR2 on pooled CD14⁺ monocytes from 10 normal volunteers, HL-60 cell line and THP-1 cell line were 2·1, 2·8 and 16·7 × 10⁻¹⁰ M, respectively. Thus, the affinity of the T2·1 MoAb was essentially the same for both the recombinant and native TLR2 molecules. In the assay, these standard beads were run in separate tubes (n = 2) under the same conditions as the PBMC samples. More than 10 000 events were collected and a calibration curve was generated from net MFI values of the standard beads in each assay. Representative results are shown in Fig. 1a–d. Net MFI values of monocytes were then converted to the number of antibody-binding sites per cell by using the calibration curve (Fig. 1e–h). As an example, when a net MFI value of a monocyte population was 87·9 fluorescence units, the number of antibody-binding sites of the cell was calculated as 4193 sites/cell.

In this assay system, interassay variation was within acceptable levels [coefficient of variance (CV) < 6·8%] when tested with the same PBMC samples from different donors (n = 3, five assays), which had been stored in aliquots in liquid nitrogen. Interassay variation was also maintained at low levels (CV < 6·5%), even when the amplifier settings of the flow cytometer (530–630 V) and concentrations of PE-labelled anti-TLR2 antibody (1–3 µg/ml) were changed (data not shown). These results demonstrated the high reliability of this assay system.

Modulation of TLR2 expression level on the monocyte surface in patients with bacterial infectious disease

We collected peripheral blood samples from 62 normal volunteers (31 male and 31 female), ranging from 30 to 94 years (mean = 60), and determined their TLR2 expression levels on monocytes by flow cytometry analysis using recombinant TLR2-coupled standard beads. In Fig. 2a, measurements from male and female volunteers are shown and are divided into three age groups: 30–49 years, 50–69 years and greater than 69 years. Statistical analysis showed that there was neither gender difference nor age-related difference in TLR2 expression level (Fig. 2a; for all pairs, P > 0·200). TLR2 expression levels in these normal volunteers were distributed from 2354 sites/cell to 6000 sites/cell, with a medium value of 4488 sites/cell (Fig. 2a). Overall the distribution of TLR2 levels in 62 volunteers appeared normal, and the mean, the mean ± 1 standard deviation (s.d.) and the mean ± 2 s.d. values of TLR2 levels were 4395 sites/cell, 5179 sites/cell and 5964 sites/cell, respectively. Based on these results, we set the upper limit of the normal TLR2 level at 5964 sites/cell (mean ± 2 s.d.) for this study.

Fig. 2.

Comparison of Toll-like receptor (TLR) 2 expression levels between volunteers and patients. (a) TLR2 levels in 31 male and 31 female volunteers are shown (mean ± s.d.) and divided into three age groups: 30–49 years, 50–69 years and greater than 69 years; n.s.: no significant difference was found for each paired combination. The Steel–Dwass test was applied for the statistical analysis. (b) Distribution of TLR2 sites/cell in 62 volunteers and 37 patients with bacterial infectious diseases is shown. For each patient, the greatest TLR2 value during the period of disease is displayed. Student's t-test was applied to this statistical analysis. Circles represent individual TLR2 expression values. Horizontal bars represent the mean value per group. Numbers under the circles indicate the mean ± s.d. value per group.

Next, we examined periodically and longitudinally the TLR2 expression levels on circulating monocytes in 45 patients (23 male and 22 female) who suffered from various bacterial infectious diseases and were hospitalized between March 2005 and July 2006 (Table 1). The mean patient age was 74 years (range: 30–97). We followed TLR2 expression levels of these patients from the initiation of medical intervention to discharge, if possible, over a period up to 3 weeks maximum after antibiotic treatment. As a result, we found that TLR2 levels changed dramatically on a daily basis in a patient-dependent fashion. Most patients showed their peak levels in the acute phase or in the recurrence phase during the antibiotic dosing period, while in some patients the greatest values were seen in the first remission when the antibiotics were discontinued (for example, see below, Fig. 3a). Thus, the peak values of TLR2 levels might be missed for some patients. Therefore, we picked the greatest value of each patient for the sake of convenience and compared it with values from healthy volunteers. Our preliminary study showed that the TLR2 expression levels of volunteers were stable, maintaining their proper levels if they remained in good health. As shown in Fig. 2b, TLR2 expression levels on monocytes were significantly higher in the patient group than in the volunteer group (P < 0·001). Unexpectedly, however, we also found that the greatest TLR2 values in each patient did not necessarily coincide with the peak values of WBC and CRP during bacterial infection. This was confirmed by correlation analysis with the WBC and CRP levels that were measured at the same time-points as the greatest observed TLR2 values. As shown in Fig. 3, TLR2 levels did not correlate with WBC (r = − 0·0125), while they were correlated only poorly with CRP levels (r = 0·318). These results suggested that the TLR2 level on monocytes might serve as a new biological parameter independent of WBC and CRP levels during the course of infection. This finding led to the question of under what clinical conditions does the TLR2 expression level on the monocyte surface increase.

Fig. 3.

Correlation of Toll-like receptor (TLR) 2 levels with white blood cell (WBC) and C-reactive protein (CRP) levels. (a) Correlation between the greatest TLR2 value in each patient (x-axis) and WBC value at the same time-point (y-axis) is shown. (b) Correlation between the greatest TLR2 value in each patient (x-axis) and CRP value at the same time-point (y-axis) is shown. Number of patients (n) and index of correlation (r) are indicated in the correlation charts. Dots represent individual cases.

Follow-up of patients with bacterial infectious disease

Of the 45 patients, 37 patients (21 male and 16 female, age 30–95) could be followed until the convalescent stage. As for the remaining eight patients, some died from a severe form of bacterial infection (n = 3) or complications during treatment (n = 2), while other patients (n = 3) left the hospital for various reasons.

All these 37 patients had been in remission once and the antibiotic administration had been discontinued. TLR2 expression level on the surface of circulating monocytes surface was assayed once a week during antibiotic administration and if possible, subsequently over 3 weeks after the antibiotic dosing period. According to the clinical courses, we divided the patients into two groups: ‘cure’ group [n = 24, 15 male and nine female: age 30–92 years (mean = 65)] and ‘recurrence’ group [n = 13, six male and seven female: age 78–95 years (mean = 88)]. As shown in Fig. 4a, TLR2 levels on monocytes had variable fluctuations among these patients. We set the upper limit of normal TLR2 expression level at 5964 sites/cell, under which 97·5% of volunteers were included (Fig. 2b), and analysed the characteristics of the transition patterns of TLR2 levels. In the cure group TLR2 levels declined rapidly to the normal range, or were already within a normal range when they entered this study, and thereafter they were maintained at a normal range for up to 3 weeks after antibiotic administration (Fig. 4a). In the recurrence group, TLR2 levels were relatively stable with minor fluctuations during and after the dosing period, and a substantial proportion of the recurrence group surpassed the upper limit even at the time of remission. For these 13 relapsed patients, antibiotic agents were readministered as quickly as possible when the recurrences were diagnosed.

Fig. 4.

Transition of Toll-like receptor (TLR) 2 levels during the period of disease in the cure group and recurrence group. (a) Thirty-seven patients that were followed during and after antibiotic treatment were divided into two groups according to the outcomes ‘cure’ and ‘recurrence’. The respective transitions of TLR2 levels are shown. The upper limit from the volunteers (5964 sites/cell) is shown by a solid line. An antibiotic agent was readministered in the ‘recurrence’ group as quickly as possible once the relapsed infection was diagnosed (readministration: dashed line). (b) The comparison between the ‘cure’ group and ‘recurrence (Re)’ group was performed for white blood cell (WBC), C-reactive protein (CRP) and TLR2 levels at the time of antibiotic discontinuation. Circles represent individual TLR2 expression values. Horizontal bars represent the mean value per group. Numbers under the circles indicate mean ± s.d. values per group. In the centre graph, CRP levels are indicated on a logarithmic scale. Student's t-test was applied to statistical analysis of WBC and TLR2 levels. The Mann–Whitney U-test was applied to statistical analysis of CRP levels.

Figure 4b shows the distribution of values for WBC, CRP and TLR2 at the time of discontinuing the antibiotics. At this time-point, a significant difference between the two groups was not visible in the WBC. A slightly significant difference was found in CRP levels between the cure and recurrence groups (P = 0·031). There was a patient in each group who had continuously high levels of CRP caused by their underlying diseases, metastatic liver cancer and collagen disease, independent of bacterial infection diseases (Fig. 4b, centre panel). In contrast to these conventional inflammatory parameters, TLR2 expression levels were higher in the recurrence group than in the cure group (P < 0·001). Additionally, we picked 11 volunteers older than 77 years [four male and seven female, 78–94 years (mean ± s.d., 84 ± 5)] from the volunteer group in order to age-match the recurrence group. Statistical analysis showed significant differences in TLR2 levels between the recurrence group and the age-matched volunteer group (5731 ± 864 versus 4880 ± 386; P < 0·006).

Recurrence rate according to the criteria of TLR2 level just after the dosing period

In this study, the standard level of WBC was set as less than or equal to 9700/µl in men and 9300/µl in women, and that of CRP was less than 0·5 mg/dl. In the above 37 patients, WBC was in the normal range for all patients while the CRP level surpassed the normal range in some patients at the point of antibiotic discontinuation.

To test its predictable value, we divided 37 patients into three groups according to CRP levels, and the association between CRP levels and risk of recurrence was evaluated. As shown in Table 2, six patients (26·1%) of 23 with the standard level of CRP relapsed. Five patients (50·0%) of 10 relapsed where the CRP level was 0·5–1·0 mg/dl. Two patients (50·0%) of four with a CRP level of greater than 1·0 mg/dl relapsed. The overall recurrence rate of the total was 35·1% (13/37). Fisher's exact test (extended) was applied for the statistical analyses, and it showed that the recurrence rate was not dependent upon CRP levels (P = 0·351). The CRP level under normalizing WBC just after the antibiotics dosing period was insufficient to predict the complete recovery from bacterial infectious diseases.

On the other hand, the TLR2 level was segmented with the mean value (4395 sites/cell), mean ± 1 s.d. value (5179 sites/cell) and mean ± 2 s.d. value (5964 sites/cell) of volunteers (Table 3). In the case where the TLR2 level was less than 4395 sites/cell, the recurrence rate was 6·7% (1/15). Patients who had a TLR2 level greater than 5964 sites/cell had a high risk (100%) of recurrence. When TLR2 levels were in ranges of 4395–5179 sites/cell and 5180–5964 sites/cell, recurrence rates were 27·3% (3/10) and 66·7% (4/6), respectively. Fisher's exact test (extended) showed that TLR2 levels were associated significantly with the risk of recurrence (P < 0·001). Consequently, it was suggested that the TLR2 expression level just after the antibiotic dosing period has greater predictive value for the outcome of bacterial infectious disease, complete cure or recurrence than the CRP level at the same time-point.

Examination of TLR2 levels according to the effectiveness level of an administering antibiotic agent

As shown in Fig. 2b a substantial number of patients, who were all receiving antibiotics, showed the normal range of TLR2 levels even at their greatest values. Thus, we next focused on the relationship between TLR2 levels and efficacy of the applied treatment. We checked the TLR2 levels of 39 cases who were receiving antibiotic agents and subsequently followed the clinical course for 5 days or greater when there was no exchange of an antibiotic agent. In Fig. 5, TLR2 levels assayed on day 2 or later after the initiation of antibiotics are plotted according to the effectiveness of the antibiotic agents for the cases. In this study, we counted data from three patients as six independent cases, because when the antibiotic treatment was changed a marked improvement in the clinical conditions occurred in these three patients.

Fig. 5.

Toll-like receptor (TLR) 2 levels in respective groups divided according to the effectiveness level of the antibiotics. Thirty-nine cases were divided into three groups: ‘marked effect’ (n = 21), ‘intermediate’ (n = 9) and ‘ineffective’ (n = 9), according to the effectiveness of the antibiotics for the patient. TLR2 levels of the respective groups are shown. Circles represent individual TLR2 expression values. Horizontal bars represent the mean value per group. Numbers under circles indicate mean ± s.d. values per group. The Steel–Dwass test was applied to this statistical analysis.

The effectiveness was defined as described below, and 39 cases were categorized into three corresponding groups. The first group (n = 21) consisted of cases where the antibiotic agents had a marked effect on the bacterial infection. The cases in the first group went into remission dramatically until day 3 after the initiation of an antibiotic agent, diagnosed from the normalization of WBC, the dramatic decline of CRP levels, the lowering of fever and other clinical symptoms. In the second categorized group (n = 9), WBC and CRP levels had been declining gradually throughout the week and fever had fluctuated, but eventually went down within a week. Finally, the cases classified in the third group (n = 9) had suffered from refractory infectious diseases due to drug-resistant bacteria or underlying diseases. Diagnosed comprehensively from the course of the clinical state including the transition of WBC, CRP levels and fever, the antibiotic agents were ineffective in these nine cases. Thus, the first-line agents were exchanged for another antibiotic at day 5 or later after initiation of antibiotic administration.

TLR 2 levels of groups categorized into the ‘marked effect’ group and ‘intermediate effect’ group were distributed as shown in Fig. 5 (4434 ± 1118 sites/cell and 5767 ± 634 sites/cell, respectively). The TLR2 levels in the ‘ineffective’ group were distributed in the highest range (7068 ± 850 sites/cell). As a result of retrospective analysis, significant differences of TLR2 levels were recognized in each pairing of the three groups (‘marked effect’ versus ‘ineffective’, P < 0·001; ‘marked effect’ versus ‘intermediate effect’, P = 0·006; ‘intermediate effect’ versus ‘ineffective’, P = 0·019).

Characterization of TLR2 levels in various non-infectious inflammatory diseases

We assayed TLR2 expression levels on the monocyte surface in 12 patients with various non-infectious inflammatory diseases, such as collagen diseases, surgical injury and ischaemic diseases (Table 4). In patients with these non-infectious diseases the levels of WBC and CRP were often elevated, and fever of an intermittent type was also observed. In cases where bacterial infectious disease develops in combination with a non-infectious inflammatory disease, it may be hard to detect rapidly the bacterial infectious disease. Thus, administration of antibiotics for these cases is likely to be delayed. In this study, it could be determined retrospectively whether the bacterial infectious disease was consolidated in 12 patients, considering various clinical conditions, such as the alteration of WBC and CRP levels during diagnostic antibiotic administration and subjective and objective symptoms.

Table 4.

Patients with non-infectious inflammatory diseases.

Case	Date of blood sampling	Diagnosis	Age	Sex	TLR2	WBC	CRP	AST	ALT	INF
1	13·12·2005	Hypoxic liver damage	84	F	3750	10 500	4·0	653	1413	(–)
	26·12·2005	(Pneumonia)			7228	5 100	0·1	36	33	(+)
2	14·6.2006	Cerebral infarction	92	F	5469	6 500	1·5	30	24	(–)
3	3·3.2006	AMI	50	M	4241	9 600	0·1	50	20	(–)
4	4·5.2006	Metastatic liver cancer	85	F	5458	8 900	14·9	34	25	(–)
	24·52006	(Cholecystitis)			7313	14 900	13·6	101	38	(+)
5	7·6.2005	Colon cancer	86	M	3162	8 400	2·4	34	30	(–)
6	7·6.2005	SLE	78	F	3471	4 800	2·4	32	28	(–)
7	15·3.2006	Collagen disease	83	M	4540	6 000	6·1	36	32	(–)
8	26·12·2006	Rheumatic arthritis	74	F	4374	5 300	0·1	22	16	(–)
9	8·3.2006	(Appendicitis)			6095	5 500	1·1	28	18	(+)
	9·3.2006*	Surgical operation	85	F
	29·3.2006	Surgical injury			4909	5 600	1·9	31	20	(–)
10	22·11·2005	Surgical injury	65	M	3834	13 000	3·3	23	13	(–)
11	19·4.2006	Alcoholic liver disease	73	M	5225	6 900	0·1	87	58	(–)
12	13·2.2006	Chronic hepatitis	79	F	4977	4 800	0·2	89	107	(–)

SLE: systematic lupus erythematosus; AMI: acute myocardial infarction; INF: infection; AST: aspartate aminotransferase (units/l); ALT: alanine aminotransferase (units/l); TLR2 = sites/cell; white blood cells = cells/µl; C-reactive protein = mg/dl.

^*Indicates the date of surgical operation. Parentheses indicates a bacterial infectious disease.

As shown in Table 4, when these patients had no infectious disease, TLR2 levels remained less than the upper limit (5964 sites/cell) of the level seen in volunteers (Fig. 2b), even though WBC and CRP levels were abnormal in these patients. Antibiotic agents had not been administered within a week before the TLR2 assay in the conditions without infectious diseases. When these patients had bacterial infectious diseases [infection (+), cases 2, 4 and 9 in Table 4], again the TLR2 levels surpassed the upper limit during the infection. Particularly in the clinical course of case 9, the up-regulation of the TLR2 level was not observed postoperatively (TLR2 level = 4909 sites/cell), although elevation of the CRP level, caused by surgical injury, lasted until day 7 after the antibiotic dosing period. In these 12 patients, modulation of TLR2 expression level on monocytes seemed not to fluctuate markedly along with the elevation of WBC and CRP levels caused by non-infectious diseases.

Discussion

In this study, we have shown that a high level of TLR2 expression on monocytes enables us to predict a relapse of the bacterial infectious disease within 3 weeks after the antibiotic dosing period. These data suggest that assessing TLR2 expression levels in addition to WBC and CRP levels will be critical in deciding the antibiotic dosage schedule (Fig. 4 and Table 3). In addition, TLR2 expression levels on the monocyte surface represented the effectiveness of ongoing antibiotic treatment (Fig. 5). Only checking the TLR2 level on day 2 or later after the initiation of antibiotics might provide the choice criterion of antibiotics without continuing the administration of an ineffective agent and prolonging illness. Taken together, we show for the first time how TLR2 expression level on monocytes is modulated during the period of antibiotic treatment for bacterial infectious diseases, and how it differs from other conventional inflammatory markers, such as WBC and CRP.

We hypothesized that TLRs might be useful as markers in monitoring the state of bacterial infections, considering previous studies about TLRs modulations in various disorders [16–26]. To this end, we developed a new methodology that significantly reduced assay-to-assay variation, and enabled us to compare one sample with another sample assayed at different time-points. The measurement values are represented as the number of sites per cell that were recognized by an anti-TLR2 monoclonal antibody. Throughout the clinical course, we confirmed that the transition of medical conditions was reflected more feasibly in TLR2 levels determined by our system than those obtained by conventional systems (data not shown).

We have had no reliable clinical marker to assess the opportunity of discontinuing a dosing antibiotic agent in the convalescence stage. In order to make a decision for the antibiotic dosing period, information of the clinical condition and patient background, species of bacteria causing infection and original site of infection, as well as examination of the level of inflammatory markers such as WBC and CRP, is required. The judgement depends in large measure, therefore, on the individual experience of the physician. When the dosing period culminates too soon, the recurrence of bacterial infection will increase in frequency. In contrast, when it continues for too long and beyond necessity, it might drive iatrogenically to reproduce drug-resistant bacteria. Assaying the TLR2 expression level on the monocyte surface may provide the criteria of defining an antibiotic dosing period. For more practical applications, we should investigate in more detail bacterial infectious diseases based on infected internal organs, and amass more cases with various kinds of underlying diseases. Thinking deductively from the outcomes of some patients who were at similar TLR2 levels in the range of 4395–5964 sites/cell, it seems that the recurrence of bacterial enteritis might be less than those of other bacterial infectious diseases. Furthermore, it is conceivable that various underlying diseases have an obvious effect on the recurrence of bacterial infectious diseases. In fact, we experienced a case where the patient suffered from pneumonia in association with bronchiectasis (age 22 years, female, data not shown). She had always suffered from coughing up phlegm and had had pneumonia several years previously. In the remission of pneumonia, her TLR2 level declined once in response to antibiotic administration, but a month later her TLR2 level had increased again. It was difficult to diagnose if she had relapsed, thus this case was an exception. In this case, long-term low-dose macrolide therapy with 400 mg/day clarithromycin was performed, and the up-regulated TLR2 level was restored to normal the following month.

In this study we also examined TLR2 levels from blood samples from patients with non-infectious inflammatory diseases, such as ischaemic diseases, hypoxic organ damage, surgical injury, malignant tumours and collagen diseases. WBC and CRP levels are frequently in the abnormal range in these disorders themselves. However, we found that the TLR2 levels remained below the upper limit of the controls in 12 patients with these disorders, even when CRP and WBC levels were abnormal (Table 4). The TLR2 expression levels were up-regulated in three patients only when the bacterial infection was consolidated. Although we cannot generalize with this small patient sample size, TLR2 expression levels on the monocyte surface might be characterized as a more specific marker for infectious disease than conventional parameters. Further studies with various disorders are required in order to confirm this observation.

It remains unclear what molecular mechanisms work in the regulation of TLR2 expression on monocytes. Furthermore, it is interesting to see why the expression of TLRs is modulated so promptly in response to the patient's clinical condition. In our ongoing ex vivo study, we observed that some cytokines could alter the TLR2 expression levels on monocytes. However, we have not known what factors modulate in vivo TLR2 expression levels on monocytes. Future studies are required to examine this question further.

Taken together, our developed quantification system revealed some characteristics of TLR2 expression on monocytes in various disorders. TLR2 expression level was modulated promptly in response to the clinical conditions through the period of bacterial infectious diseases. TLR2 expression level on circulating monocytes may be a useful and important marker for understanding medical conditions and for defining the antibiotic dosing period and any change in antibiotic treatment.