The Centers for Disease Control and Prevention (CDC) Human Subjects committee approved the study protocol, which met the ethical standards of the Helsinki Declaration, and all subjects gave written informed consent. This study included the 81 subjects (34 CFS and 47 NF) from a population-based follow-up study of CFS in Georgia, USA who had completed the TSST as part of a three-day study in the Emory General Clinical Research Center. CFS cases met the 1994 international research definition of CFS as evaluated by standardized questionnaires, including the Multidimensional Fatigue Inventory, the SF-36^® Health Survey, and the CDC Symptom Inventory.26 There were no statistical differences between CFS and NF groups in the mean age (CFS, 45 ± 9 years; NF, 46 ± 9 years; P = 0.75), sex (CFS, 82% female; NF, 77% female; P = 0.53), race (CFS, 76% Caucasians; NF, 85% Caucasians; P = 0.23) or body mass index (CFS, 28 ± 5; NF, 26 ± 5; P = 0.075).

The TSST was performed to assess stress-induced regulation of neuroendocrine, autonomic and immune responses to challenge. The TSST was consistently started at the same time of day to ensure a similar diurnal response between subjects. The test consists of a preparatory and anticipation phase (beginning at 1:15 pm) and a subsequent 10-minute public speaking and 10-minute mental arithmetic task in front of three trained staff members (TSST panel, 1:30 pm to 1:50 pm). An indwelling catheter was placed at 7:30 am for blood draws. Blood was collected in Tempus™ tubes (Applied Biosystems, CA) for microarray analysis at 8:00 am and at 1:00 pm baseline, as well as immediately prior to the TSST panel at 1:30 pm, immediately following the TSST panel at 1:50 pm, and at subsequent fifteen minute intervals until 3:05 pm. Blood was also collected in 8 mL Cell Preparation Tubes (CPT™) for PBMC isolation (BD Biosciences, CA) at 10:00 am (3.5 hours prior to TSST) and at 3:05 pm (1.5 hours post TSST). Cell counts were determined in blood collected in Ethylenediaminetetraacetic acid (EDTA) tubes corresponding to all Tempus™ tube blood collection times except 8:00 am.

Blood drawn into CPT™ was processed within 1.5–5 hours to isolate PBMCs according to the manufacturer’s instructions. PBMCs were frozen in RPMI 1640 media (Invitrogen, CA) at 5 × 10⁶ cells/mL, and stored in liquid nitrogen until use. Deoxyribonucleic acid (DNA) and RNA were isolated from aliquots of the stored PBMC. PBMC RNA was isolated using Trizol (Sigma Aldrich, MO) and PBMC DNA was extracted using the Roche DNA Isolation Kit for Mammalian Blood (Roche Applied Science, IN) following the manufacturer’s protocol.

Tempus™ tubes were frozen at −20 °C until extraction (<one month). Whole blood RNA was extracted from Tempus™ tube blood using the 5 PRIME Perfect Pure RNA Cultured Cell Kit (Fisher Scientific, PA). For all samples, RNA quality and quantity were assessed using Agilent 2100 Bioanalyzer RNA Nano Chips (Agilent Technologies, CA) and a Nanodrop 1000 spectrophotometer (Thermo Scientific, DE).

EDTA tubes were submitted to Quest Diagnostics (Atlanta, GA) on the day of collection for determination of Complete Blood Count with differential and flow cytometric determination of T, B, and NK cell counts and percentages.

Microarray analysis was carried out as previously described,27 using whole blood RNA. One microgram of RNA was labeled using the Exon WT Sense Target Labeling Assay (Affymetrix, CA) and after hybridization to the Affymetrix Human Exon 1.0 ST array, chips were scanned using the Affymetrix GeneChip Scanner 3000. Array analysis was performed using Affymetrix^® Expression Console™ (v 1.1) at the transcript level using core-level probe sets. For this analysis, only PRF1 expression was used from this microarray data set.

PBMC RNA (500 ng) was DNase I treated in a 10 μL volume using the MessageClean^® Kit (GenHunter, TN) and then reverse transcribed in the same tubes using 20 μL reactions with Superscript™ III (Invitrogen, CA) and a combination of oligo(dT) and random hexanucleotide primers (2.5 μM each). LightCycler PCR (20 μL) was performed using the SybrGreen 480 Master Mix (Roche Applied Sciences) that contained 2 μL of 1:20 dilution of complementary DNA (cDNA) and 0.5 μM of each primer. Thermal cycling conditions were as follows: 1 cycle of 94 °C for 5 minutes (min), 50 cycles of 94 °C 15 seconds (s), 62 °C 15 s, and 72 °C 15 s. All reactions were carried out in duplicate with previously described peptidylprolyl isomerase B (PPIB),28 and PRF1 primers (forward 5′AGG AGC TGG GCA GAA GGA CAA GA 3′, reverse 5′ CAC CAT AGA GGG CTC AAG GGA AGG 3′, product 88 bp).9 PCR efficiencies of PRF1 and PPIB reactions were 1.96 and 1.97 respectively. Relative quantitation was done using the 2^−ΔΔCT method using the equation 2^-((sample PRF1 Ct – PPIB Ct)-(calibrator PRF1 Ct – PPIB Ct)) where the calibrator was a 1:100 dilution of HeLa cell cDNA (prepared as above for PBMC cDNA), included in each plate.

PBMC DNA (200 ng/reaction) was bisulfite treated using the Epitect Bisulfite Kit (Qiagen, CA) according to the manufacturer’s instructions. Bisulfite-pyrosequencing was conducted as previously described to examine methylation levels at seven CpG sites in the MSR of the PRF1 promoter (Fig. 1; sites −876, −776, −744, −720, −691, −670 and −650 base pairs upstream of the transcription start site).29 Three amplicons, D, E and F, and a total of five sequencing primers were used to cover all seven CpG sites. We used a touchdown PCR that consisted of one cycle of 94 °C for 5 min for the initial denaturation step followed by 5 cycles each of denaturation at 94 °C for 30 s. This process involved varying annealing temperatures for 30 s, and an extension at 72 °C for 30 s. Annealing temperatures for the touchdown portion were as follows: 65 °C for 5 cycles, 62 °C for 5 cycles, 59 °C for 5 cycles, 56 °C for 5 cycles and 52 °C for 5 cycles. And a further 25 cycles of the following: 94 °C for 30 s, 50 °C for 30 s and 72 °C for 30 s. PCR was terminated after a final cycle at 72 °C for 7 min. The PyroGold Kit was used in conjunction with the PSQ 96MA instrument (Qiagen), and each pyrosequencing reaction used 20 μL of PCR product. All reactions were carried out in duplicate.

Bioinformatic analysis of transcription factor binding sites (TFBS) in the MSR was carried out using the Genomatix Matinspector database (Genomatix Software GmbH (http://www.genomatix.de).30 The identified TFBS are as follows: HIF-1 ancillary sequence family, previously identified as AP-2 (HAS/AP-2),31 E-Twenty Six family (ETS), E2A basic helix loop helix family (E2A), Glucocorticoid Response Element (GRE), Signal Transduction and Activator of Transcription (STAT),32 and p53 transcription factor family (P53; Fig. 1).

Statistical analysis was carried out using either SAS version 9.3 or SAS enterprise guide version 4.3. Normality was tested using the D’Agostino and Pearson Omnibus Normality Test, and parametric tests including two-tailed t-tests and Pearson correlation coefficients were used for analysis.

As shown in Figure 2A, there was a significant increase in the overall percentage of NK cells in both CFS (P < 0.0001) and NF (P < 0.0001) subjects during the TSST, with no significant difference between groups. Similar to NK cells, other cell type percentages (neutrophils, T-cells, and B-cells) also increased in response to TSST and did not differ with respect to disease status (data not shown).

Microarray analysis of whole blood (Fig. 2B) showed a significant increase in PRF1 expression between the introduction/preparation phase and immediately prior to the TSST oral presentation (1:00 pm–1:30 pm) of about 1.6-fold in both NF (P < 0.0001) and CFS (P = 0.0003). PRF1 expression continued to increase in NF subjects during the TSST oral presentation (1:30–1:50 pm) whereas expression decreased slightly in CFS subjects, resulting in a significant difference between NF and CFS subjects (P = 0.04) at 1:50 pm. Over the 45 minutes following the TSST, PRF1 expression declined and reached baseline levels in both groups (sampling time 2:35 pm). PRF1 expression in NF subjects continued to decrease, but increased in CFS subjects, resulting in 1.3-fold higher expression in CFS than NF at this time, P = 0.06 (Fig. 2B). PRF1 expression at 10:00 am and 3:05 pm in PBMCs determined by qRT-PCR (Fig. 2C) were generally consistent with whole blood microarray results. PRF1 expression in PBMCs did not differ between CFS and NF at 10:00 am, but at 3:05 pm, PRF1 expression was higher in CFS than NF (1.4 fold, P = 0.02).

Site-specific CpG methylation in the MSR at 10 am ranged from 38% to 79% and increased at all 7 CpG sites by 2.1% to 4% after the TSST (at 3:05 pm) for the study population as a whole (P < 0.0001 to P = 0.01, Fig. 3A). However, in CFS subjects (Fig. 3B) the increase in methylation levels after TSST were significant at only two CpG sites (−776 and −744) whereas in NF controls (Fig. 3C) the increase was significant at four CpG sites (−876, −744, −691, and −670). There were no significant overall differences between CFS and NF in terms of site-specific methylation or average methylation of all 7 CpG sites before or after the TSST.

To examine the impact of PRF1 methylation on PRF1 expression, we examined correlations between average methylation of the 7 CpG sites and gene expression for each participant and time point in whole blood and PBMCs. Average baseline methylation (10:00 am) was negatively correlated with PRF1 expression in whole blood at all 8 time points (Table 1) in both CFS and NF, statistically significant for all except one NF time point (2:35 pm). Average methylation at 3:05 pm was also negatively correlated with whole blood PRF1 expression at 3:05 pm (Table 1) in both NF and CFS subjects, although correlation was not significant in CFS. Similar negative correlations between PRF1 methylation and PRF1 expression in PBMCs were observed in both NF and CFS subjects (Table 1).

We used linear regression to quantify the relationship between PRF1 average methylation at 10 am and whole blood PRF1 expression at all time-points (Fig. 4). In CFS subjects, this analysis predicted a 3% increase in PRF1 expression (95% confidence interval 2.5% to 3.7%) with every 1% decrease in PRF1 methylation (y = −0.044x + 12; the antilog₂ of 0.044 = 1.03). A similar inverse relationship was found with NF subjects, however 31% and 14% of the variance in PRF1 expression was explained by PRF1 promoter methylation in CFS (R² = 0.3133) and NF (R² = 0.1425) respectively.

We used the median split of the average methylation level of the 7 CpG sites at 10:00 am to categorize subjects into high (≥65%) and low (<65%) methylation. Whole blood PRF1 expression was significantly higher in the low methylation group at all TSST time-points (P-value from 0.0002 to 0.05; Fig. 5A). When the high or low methylation group was stratified by illness, a significant difference was noted between CFS and NF only at the 1:50 pm time-point (immediately following TSST). The CFS high-methylation group had a significantly lower PRF1 expression than the NF high-methylation group (P = 0.005, Fig. 5B).