Blood sampling.

Within 1 h after installation of the venous catheter, subjects had approximately 2 ml of blood removed with replacement of volume using heparinized saline. Patients were then allowed to go to sleep between 10:30 and 11:00 p.m. while being monitored. Three additional blood samples were taken at approximately 1:00 a.m., 3:00 a.m., and 5:00 a.m. Subjects then awoke between 7:00 and 7:30 a.m., after which a final blood sample was taken. Whole heparinized blood samples were kept at room temperature, and PBMCs were generated, counted, and plated in ELISPOT plates within 6 h of collection. Samples in PAXgene tubes (Qiagen Inc., Valencia, CA) were collected at the time of the venipuncture and stored frozen (−20°C) until further processing (RNA isolation for quantitative reverse transcription-PCR [qRT-PCR]).

Analysis of plasma cytokines using Luminex.

Immediately after collection, an aliquot of blood was centrifuged, and then plasma was removed and snap-frozen on dry ice with subsequent storage at −80°C until time of assay. Samples were then thawed, diluted 1:1 with assay buffer, and assayed with Millipore's Beadlyte human multicytokine detection system 2 (Billerica, MA) according to the instructions in the kit, with some modifications. This kit simultaneously measured the proinflammatory cytokines IL-1β, IL-6, IL-8, and tumor necrosis factor alpha (TNF-α) and the anti-inflammatory cytokines IL-4 and IL-10. There were two modifications to the assay protocol. To stop the reaction and remove the unbound luminescence, the wells in the 96-well plates were rinsed twice with assay buffer instead of using the stop solution. In addition, the standard solution was serially diluted to obtain a standard curve with concentrations of 0, 0.49, 0.98, 1.95, 3.90, 7.80, 15.60, 31.25, and 62.5 pg/ml. Data reduction was performed using the MasterPlex-QT software (Miraibio, Alameda, CA); the standard curves were fit with a linear model using those standards that encompassed the values for each cytokine obtained from the samples in each assay. The r² values for the linear fits were always above 0.95 and very often above 0.99. In each 96-well plate, the standards and a quality control pool were run in triplicate, and 33 samples were run in duplicate. Additional details about this assay have been published elsewhere (28).

Measurement of ELISPOT data.

Because the ELISPOT approach measures cytokine production within a maximum period of 48 h (9), it establishes frequencies of Th1 and Th2 cells as present in vivo, allowing induction of memory cells to produce cytokines but not their priming, proliferation, or in vitro differentiation. We (S. K. Schwander) have been involved in developing a sensitive IL-10 ELISPOT assay (9). Individual antigen-specific cytokine-producing cells can be detected with high sensitivity (18) directly ex vivo within a large pool (1/300,000 cells) of pharmacologically untreated cells. In the case of nonspecific stimuli that were used in this assay, proportions of responder cells are expected to be considerably higher and represent maximally inducible and functional cell numbers.

Preservation of nocturnal whole blood samples.

Extended pilot experiments were performed to assess optimal short-term storage conditions for whole blood samples for ELISPOT-based cytokine assessments. Nonspecific ELISPOT assay results were minimized by keeping whole blood samples in upright standing Vacutainer tubes (BD, Franklin Lakes, NJ) at room temperature until PBMCs were generated. To ensure comparability of the assays between samples drawn at around midnight and samples drawn in the early morning hours, all whole blood samples were kept for 4 to 6 h at room temperature until preparation of their respective PBMCs. Plating of PBMCs into the ELISPOT assay plates started at 9:00 a.m. with the sample from the earliest venipuncture time point and ended at 1:00 p.m with the sample from the latest (7:00 a.m.) venipuncture time point.

Preparation of PBMCs.

PBMCs were prepared using standard procedures (2). Briefly, heparinized venous whole blood diluted 1:1 with supplemented RPMI medium (BioWhittaker, Walkersville, MD) was centrifuged (1,200 rpm, 45 min, 21°C) over a Ficoll-Paque density gradient. PBMCs were removed from the interface, washed twice in RPMI, and resuspended in complete culture medium (RPMI with 2% pooled human AB serum). PBMCs were counted in a hemacytometer, and PBMCs numbers were adjusted to various concentrations. Viability of PBMCs was 100% based on trypan blue exclusion.

Constitutive and stimulant-induced IL-1β, IL-4, IL-6, IL-8, IL-10, and TNF-α production based on ELISPOT assays.

To assess constitutive or maximally stimulated cytokine production, PBMCs were cultured in complete culture medium only or in the presence of lipopolysaccharide (LPS; 1 μg/ml, final concentration) or phytohemagglutinin (PHA; 5 μg/ml, final concentration), respectively. Because no definitive etiopathogenic agent is known for chronic fatigue syndrome, it was not possible to assess antigen-specific cytokine induction. Aliquots of 100 μl of the stimuli in complete culture medium were added to the wells. Aliquots (100 μl) of the PBMC suspensions were then transferred into each well of the ELISPOT plates by using large-orifice pipette tips. ELISPOT plates (Multiscreen_HTS high protein binding 96-well plates; Millipore, Bedford, MA) were coated with primary monoclonal antibodies against IL-1β, IL-4, IL-6, IL-8, IL-10, and TNF-α. IL-1β, IL-6, and IL-8 ELISPOT assays were done using commercial kits (Cell Sciences, Canton, MA) according to the manufacturer's instructions. IL-4- and IL-10-producing as well as TNF-α-producing cells were assessed using in-house ELISPOT assays and capture and biotinylated detection antibodies from BD Pharmingen, San Diego, CA (IL-4 and IL-10) and Endogen Pierce, Rockford, IL (TNF-α), respectively. Constitutive and stimulant-induced cytokine production levels by the PBMCs were assessed in triplicate wells. The optimal concentrations of PBMCs to be used per well differed for the various cytokine ELISPOT assays and had been identified in pilot experiments. Final PBMC concentrations per well were as follows: IL-1β, 100,000; IL-4, 200,000; IL-6, 2,000; IL-8, 2,000; IL-10, 200,000; TNF-α, 20,000 and 40,000 (dual setup).

Similarly, optimal short-term culture incubation periods varied by cytokine and had been identified in pilot experiments or were according to manufacturers' specifications: IL-1β, 20 h; IL-4, 48 h; IL-6, 24 h; IL-8, 24 h; IL-10, 48 h; TNF-α, 4 h.

Following the short-term cultures, PBMCs were washed off the plates with phosphate-buffered saline (PBS; three times) and PBS-0.05% Tween (three times), and detection antibodies were added overnight at 4°C. On the next day plates were washed and streptavidin/horseradish peroxidase (HRP) was added for 2 h. Plates were then washed again and spots visualized using color reagents (kits) or with peroxidase-conjugated streptavidin (Dako, Glostrup, Denmark) and the chromogen 1% 3-amino-9-ethylcarbazole (AEC; Pierce, Rockford, IL) in the in-house assays. Following an additional washing step, plates were dried overnight.

Measurement of cytokine production.

Plates were then scanned, and spot-forming units (cytokine spots) were enumerated using an automated ELISPOT reader (Immunospot series I analyzer, software version 3.0; Cellular Technology, Ltd., Cleveland, OH) that permits single-cell analysis of size, chromatic density, and shape of the cytokine spots by employing objective definition criteria for comparison of separate wells. Frequencies of cytokine-producing cells were calculated by averaging the number of spots from triplicate wells per condition.

Measurement of cytokine gene expression. (i) cDNA synthesis.

RNA was isolated from PAXgene tubes according to the manufacturer's guidelines. cDNA synthesis was performed using the following method. For each reaction mixture, 10 μl of RNA was added to the following reagents in 0.2-ml PCR tubes: 2.0 μl reaction buffer (10×), 0.8 μl of deoxynucleoside triphosphate mix, 2.0 μl random primers, and 1.0 μl reverse transcriptase. The volume was brought to 20 μl by adding 4.2 μl of PCR-grade water. The samples were incubated at 25°C for 10 min and the at 37°C for 120 min, followed by 85°C for 5 s. All samples were then held at 4°C and stored at −70°C until further use.

(ii) Real-time PCR assay.

The TaqMan universal PCR was performed in 96-well optical plates using an ABI 7500 sequence detection system (Applied Biosystems, Foster City, CA). Table ​Table22 provides a list of genes that were evaluated: human glyceraldehyde 3-phosphate dehydrogenase (GAPDH), IL-10, IL-1B, TNF, and IL-4. Each run contained, in duplicate, one cDNA sample, a no-template control (NTC), and one untreated and one treated cell line control sample. Each 25-μl reaction mixture contained 12.5 μl 2× master mix (Applied Biosystems, Foster City, CA), 1 μl primer/probe mix, 1 μl cDNA, 10.5 μl PCR-grade water. The samples were incubated at 50°C for 2 min, followed by 95°C for 10 min, and then 40 cycles at 95°C for 15 s, 60°C for 1 min.

Data preprocessing.

Based on the fact that cytokines display burst-like changes in level over time (10) and our preliminary observation that every data set, for every cytokine and from every assay, had some extreme outlier values, producing non-Gaussian distributions, data preprocessing was necessary. We considered linear and second-order polynomial detrenders and selected the former for two reasons: first because our earlier work found no evidence of higher-order trends in 24-h data, and next because nonlinear detrending did not adequately identify outliers. Thus, we developed a filtering method based on LOWESS, a locally weighted regression, to eliminate extreme values from the data (6). The following are the details of our procedure: we repeatedly computed a linear least-squares regression of a given time series {x_i}, i = 1,…, m (m = 5 for this study) using weights {δ_i}, i = 1,…, m calculated from the weight function B(x). The weight function B(x) is a bisquare function defined by B(x) = (1 − x²)² for x < 1 and B(x) = 0 for x ≥ 1.

Then, the weight δ_i of each data point was decided by δ_i = B(e_k/6s), where e_i = ŷ_i − y_i is residual between the current fitted value ŷ_i and given data point y_i at each iteration and s is the median of e_i . The weights {δ_i} were updated at each iteration step. We repeated the above weighted regression until the chi-square statistic marked the minimum value. After estimating the proper trend, the 95% confidence areas of the trend were also determined. Then we defined outliers as the data which exceeded these areas and excluded them from the data for the later analyses. Secondly, in order to improve the normality of the data, the data were log transformed after adding a value of 1 to each datum. To improve understanding of the results, however, we have prepared the tables using the actual numerical data (see Tables ​Tables33 to ​to5,5, below).

For almost all data, one or two iterations of weighted regression were enough to obtain the minimum chi-square value. Of the 1,380 samples evaluated on Luminex and ELISPOT, 2 and 24, respectively, were excluded as outliers; of the 920 samples evaluated by qRT-PCR, 9 were excluded. We detected no temporal pattern for outliers in any of the data sets. If more than two of five data points were missed or eliminated, we excluded the data for the analysis. Thus, the number of subjects whose samples were analyzed (i.e., those listed in Tables ​Tables33 to ​to5)5) is not the same as the total number of subjects who entered the study. From the preprocessed data, we calculated a mean cytokine value for each cytokine of each subject. This provided us final data sets comprised of the mean cytokine values for six different cytokines × n subjects for the following analyses.

Correlations across different assays.

We first examined correlations of expression levels among the three cytokine detection assay types for each cytokine by using Spearman's correlation coefficients. The only significant correlation was an inverse one for IL-10 between the results from the Luminex and unstimulated ELISPOT assays. However, this effect was principally due to the group difference in IL-10 between patients and controls, with patients having higher levels than controls in the Luminex assay and lower levels than controls in the ELISPOT assay. We also evaluated the correlations by using Pearson's correlation coefficients, an independent component analysis (ICA), and the alternating conditional expectations (ACE) method (3); however, these analyses revealed no significant correlations. Thus, expression levels of cytokines in the various assays were independent of each other.

Luminex assay: group differences in cytokine production.

Table ​Table33 summarizes mean cytokine levels for each study group. There were no significant differences between CFS patients either with or without FM as a single group versus control subjects for any of the cytokines. To further evaluate the data, we stratified the CFS data into two subgroups (CFS with FM [n = 15] and CFS without FM [n = 11]) and compared the group differences among groups (Table ​(Table3).3). The expression level of IL-10 for the CFS without FM subgroup was significantly higher than that of healthy control subjects or the group with CFS and FM. In addition, CFS patients with FM tended to have decreased IL-1β (P = 0.068) compared with control subjects.

Exploration for correlations between cytokine data and clinical variables was negative except for a tendency toward lower IL-8 values with higher CFS severity score (r = −0.44; P = 0.058). Scores for patients with CFS alone did not differ in severity from those with CFS plus FM (Table ​(Table1).1). There was a significant correlation between total sleep period and IL-10 (r = −0.33; P = 0.042) in patients.

The mixed model analysis reiterated our group analysis by showing a significant difference in only IL-10 levels. Figure ​Figure11 shows the distribution of IL-10 data over time. CFS patients without FM showed higher IL-10 values than controls at 3:00 and 5:00 a.m. and higher values than for subjects with CFS and FM at 5:00 a.m.

Open in a separate window

FIG. 1.

Group-averaged IL-10 levels (in pg/ml) from the Luminex assay for control, CFS with FM, and CFS without FM groups. The error bars indicate standard errors of means. *, significant difference from control; †, significant difference between CFS with FM and CFS without FM. Note that the statistical tests were conducted with logarithmic-transformed data.

ELISPOT assay: group differences in cytokines under basal conditions.

Table ​Table44 summarizes mean cytokine levels for all groups from the ELISPOT assay. There were no significant group differences among any of the cytokines for patients compared to controls under either stimulated or unstimulated conditions. As had been the case with plasma protein, IL-1β tended to be lower in CFS subjects than controls (P = 0.099). We found no significant correlation between mean cytokine levels in the ELISPOT assay and clinical variables for the patients. For the mixed model analysis, there were no significant differences in any cytokines under either stimulated or unstimulated conditions.

This work was supported by NIH AI-54478 and NIH HL-077462. Additional support to extend the work to IL-6 and IL-8 was provided by the American Fibromyalgia Syndrome Association.