Oxidative stress and Nrf2 expression in peripheral blood mononuclear cells derived from COPD patients: an observational longitudinal study

Sample size was estimated based on the behavior of FEV₁ decline (ml/year) in Global Initiative for Chronic Obstructive Lung Disease (GOLD) 1 and GOLD 2 stage COPD patients and of Nrf2 mRNA as previously reported [14, 16]. Adopting a level of significance of p = 0.05, a power of 80% and a minimum correlation of 0.40, we estimated the minimum sample size to be approximately 32 subjects [17]. At beginning of the study we enrolled 60 consecutive, mild-moderate (GOLD 1, n = 29; GOLD 2, n = 31;) COPD patients referring to Respiratory Medicine Outpatient Clinic of our Institution. The GOLD guideline was used to make the diagnosis and to grade COPD severity [18]. The other group (n = 71) comprised age-sex-matched no-COPD subjects randomly selected from the general population [19]. Principal necessary conditions for the enrollment of both groups were lack of infectious or acute/chronic inflammatory diseases, malignancy, absence of acute/chronic renal failure and hepatic failure. No COPD subjects were using supplemental oxygen, oral glucocorticoids, and antibiotics. Inhaled corticosteroid and bronchodilator agents were administered following guideline [18, 19]. A clinical evaluation, blood sampling tests and a spirometry was performed at baseline and after a mean follow-up of at least 40 months.

Forced expiratory volume in 1st second (FEV₁) and FEV₁/forced vital capacity (FVC) were measured using a water-sealed spirometer (Biomedin, Padua, Italy). Lung function values were expressed as a percentage of predicted values, and the lower limit of normal low limit of normality for the FEV₁/FVC was calculated according to Quanjer [20].

Venous blood samples were obtained from each subject after 12 h fasting and drawn into pyrogen-free blood collection tubes. Several aliquots of plasma were placed into sterile 1 mL screw-capped polypropylene vials containing the phenolic antioxidant 2,6-di-tert-butyl-4-methylphenol (10 mM, Sigma-Aldrich Co., St Louis, MO, USA) to avoid lipid peroxidation and stored at − 80 °C. The samples were frozen and thawed only once. PBMCs were isolated as previously described [21]. High sensitivity C-reactive protein (CRP) was evaluated using a commercially available high-sensitivity turbidimetric method (Syncron-PCR; Beckman Coulter, Brea, CA, USA). Plasma 8-isoprostane (8-iso) was measured by means of Cayman’s 8-iso ELISA kit following the manufacturer’s indications. Total glutathione (GSH) was measured by means of Abcam’s GSH/GSSG ratio detection kit.

Total RNA was isolated with RNEasy Mini Kit (Qiagen, Hilden, Germany). The concentration and quality of RNA were evaluated using the RNA 6000 Nano LabChip Kit (Agilent 2100 Bioer, Agilent Technologies Inc., Santa Clara, CA, USA). Reverse transcription of total RNA was carried out using IScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA) according to the manufacturer’s recommendations. The relative mRNA expression levels of Nrf2, HO-1 and GCLC were performed in triplicate using the QuantiTect Primer Assay and QuantiTect SYBR Green PCR Kit (Qiagen) on the MyiQ Thermal Cycler (Bio-Rad). QuantiTect Hs-ACTB Assay (Qiagen) was used as normalizer. Normalized gene expression levels are given as the ratio between the mean value for the target gene and that for the beta-actin in each sample.

Continuous variables are expressed as mean ± SD values. Differences between the groups were analyzed by two-tailed paired and unpaired Student’s t-test. Categorical variables were compared using the Chi squared test. Pearson’s correlations were used to test the relationship between the variables. A hierarchical stepwise multiple linear regression model was used to evaluate the joint effect of independent variables percent variation (Δ) on ΔFEV₁ (% predicted, dependent variable). ΔNrf2 mRNA, ΔHO-1 mRNA, ΔGCLC mRNA, Δ8-iso and ΔGSH were the independent variables considered. In the model, other baseline key factors related to change in FEV₁ including baseline FEV₁, age, sex, smoking status, smoking history (pack year), BMI and follow-up time between FEV₁ measurements were taken into consideration. To exclude model overfitting, the estimated regression equation of SPSS to calculate the predicted value of R-squared was used. Missing data were handled via complete case analysis as previously suggested [22]. We also assessed the association between the dependent variable FEV₁ and the independent variables Nrf2 mRNA, HO-1 mRNA, GCLC mRNA, 8-iso and GSH, a 2-level random intercept linear model was fitted to the data, with level 1 units (measurements/visits) nested into level 2 units (subjects). The between-subject variability was modeled as a random effect, i.e. as a random-intercept term at the subject level. The Huber/White sandwich estimate of variance was used at the highest level (subjects) in the mixed model. The model also adjusted for age, sex, smoking status, smoking history (pack year), BMI and follow-up time. A p value less than 0.05 was considered significant. Statistical analysis of the data was conducted using SPSS version 20.0 (IBM Corporation, Armonk, NY, USA).

At baseline 131 subjects (60 mild-moderate COPD and 71 no-COPD) were selected. In agreement with the inclusion criteria, basal FEV₁ (actual and % of predicted) and FEV₁/FVC (actual and % of predicted) resulted significantly lower in COPD than in no-COPD subjects (data not shown). After a period of at least 40 months (mean 49.7 ± 6.9 months) of the 131 subjects enrolled in the first phase of the study, 7 COPD and 15 no-COPD did not reply, 3 COPD and 13 no-COPD declined invitation, 7 COPD and 6 no-COPD refused for supervening medical reasons. Furthermore 10 COPD subjects were meanwhile dead. Figure 1 shows a detailed diagram of participant flow and group allocation. Of the 61 missing data, 38 were not related to medical issues or death, and reasonably lost at random, while 23 were missed not at random. Lung function, anthropometric, clinical and metabolic parameters of drop-out subjects were similar to those of the subjects participating to the follow-up. Therefore, 70 subjects (37 no-COPD and 33 COPD patients) participated to both baseline and follow-up phases. GOLD stage and lung function for both groups are reported in Table 1. Not unexpectedly, GOLD stage change was different in no-COPD compared to COPD group after the follow-up period. In particular only few subjects resulted slightly obstructed (GOLD 1) in no-COPD group at follow-up. In the COPD group the number of patients classified as GOLD 1 and GOLD 2 at baseline was significantly reduced after the follow-up period; therefore, some subjects were classified as GOLD 3 and GOLD 4 after the period of follow-up. Accordingly, FEV₁ and FEV₁/FVC at follow-up were significantly lower than at baseline in COPD patients, whereas no differences were observed in no-COPD group. As shown in Fig. 2, lung function decline (expressed as annual FEV₁ variation between baseline and follow-up values) was much greater in COPD patients than in no-COPD subjects (p < 0.01). Medical history, clinical and metabolic characteristics of both groups of subjects at baseline and at follow-up are shown in Table 1. Concerning cardiovascular risk factors hypertension, diabetes and dyslipidemia were similar in COPD and no-COPD subjects at baseline. On the contrary, the two groups differed for smoking habit: in particular the number of no smokers was significantly higher in no-COPD than in COPD subjects, whereas the number of past-smokers was higher in COPD group. As shown in Table 1, the number of current smokers was small in both groups: after the period of follow-up, it reduced from 5 to 2 in no-COPD and from 6 to 4 in COPD. Accordingly, there was an increment of the number of past smokers in both groups. Moreover basal and anthropometric characteristics, lipid profile, plasma glucose, systolic and diastolic blood pressure were similar in both groups both at baseline and at follow-up.

Plasma CRP concentrations were higher (p < 0.01) in COPD than in no-COPD subjects at baseline. After the follow-up, CRP persisted to be higher in COPD patients than in controls (p < 0.01) and the values were even higher than the values detected at baseline (p < 0.01), (Fig. 3a). White blood cells (WBCs), even if in the normal range, were significantly higher in COPD than in no-COPD at baseline (p < 0.01). After the follow-up there was a further significant increment of WBCs in COPD when compared with the baseline values (p < 0.01) (Fig. 3b).

As for 8-iso plasma concentrations the results of this study show that at baseline 8-iso values were significantly more elevated in COPD than in no-COPD subjects (p < 0.001). After the follow-up, 8-iso concentrations continued to be higher in COPD than in no-COPD and moreover 8-iso values resulted further increased than those observed at baseline (p < 0.01), (Fig. 3c).

Circulating GSH concentrations were similar in COPD and in no-COPD subjects at baseline. After the period of follow-up, GSH values were significantly reduced in COPD compared to both no-COPD subjects (p < 0.01) and to the corresponding values measured at baseline (p < 0.01), (Fig. 3d).

We also evaluated the mRNA expression of Nrf2 and of its correlated genes HO-1 and GCLC in both groups of subjects. As shown in Fig. 4(a-c), at baseline the expression of Nrf2, HO-1 and GCLC was significantly (p < 0.01) up-regulated in COPD compared to no-COPD group. After the follow-up there was a significant down-regulation of Nrf2, HO-1 and GCLC expression in COPD patients (p < 0.01), while in no-COPD subjects their expression did not change.

At baseline FEV₁ was inversely correlated with Nrf2 mRNA (r = − 0.441, p < 0.001) (Fig. 5a) and HO-1 mRNA (r = − 0.358, p < 0.002) (Fig. 5b) in all the subjects. After the period of follow-up, Nrf2 and HO-1 mRNA were no longer correlated with FEV₁.

ΔFEV₁ detected after the follow-up in COPD patients was directly correlated with ΔNrf2 (r = 0.826, p < 0.001) (Fig. 6a), ΔHO-1 (r = 0.820, p < 0.001), (Fig. 6b), ΔGCLC (r = 0.840, p < 0.001), (Fig. 6c). Moreover ΔFEV₁ was also directly correlated with ΔGSH (r = 0.595, p < 0.01, data not shown) and inversely correlated with Δ8-iso (r = − 0.587, p < 0.01, data not shown). Furthermore among the further baseline key factors which may influence ΔFEV₁ (baseline FEV₁, age, sex, smoking status, pack year, BMI and follow-up time between FEV₁ measurement) only pack year resulted inversely correlated with ΔFEV₁ (r = − 0.39. p < 0.03, data not shown). Finally no correlation was found between ΔFEV₁, ΔCRP and ΔWBCs.

We then performed a model of hierarchical stepwise multiple linear regression in order to evaluate the combined effect of independent variables on FEV₁ decline taking into account some baseline key factors (baseline FEV₁, age, sex, smoking status, pack year, BMI and follow-up time) which may be related to FEV₁. Our results show that Δ8-iso, ΔGSH plasma concentrations and baseline pack year were no longer correlated with ΔFEV₁, whereas ΔNrf2, ΔHO-1 and ΔGCLC were found to be significant predictors of ΔFEV₁. On the whole R² was 0.895 and R² predicted 0.881 (Table 2) and baseline FEV₁, age, sex, smoking status, pack-year, BMI and follow-up time accounted for 15,6% of ΔFEV₁ variance.

Finally the association between the absolute values of the independent variables of interest and FEV₁ was investigated by means of a linear mixed model (Table 3). After adjusting for potential confounders, the absolute value of Nrf2 mRNA was significantly associated with FEV₁, in particular for an increase in one unit of Nrf2 mRNA, FEV₁ increased by about 8.31% (95%CI, 4.72;11.91%). The absolute values of HO-1 mRNA, GCLC mRNA, 8-iso and GSH plasma concentrations were not associated with FEV₁.