Expression of PPARγ and Paraoxonase 2 Correlated with Pseudomonas aeruginosa Infection in Cystic Fibrosis

The Pseudomonas aeruginosa quorum sensing signal molecule N-3-oxododecanoyl-l-homoserine lactone (3OC12HSL) can inhibit function of the mammalian anti-inflammatory transcription factor peroxisome proliferator activated receptor (PPAR)γ, and can be degraded by human paraoxonase (PON)2. Because 3OC12HSL is detected in lungs of cystic fibrosis (CF) patients infected with P. aeruginosa, we investigated the relationship between P. aeruginosa infection and gene expression of PPARγ and PON2 in bronchoalveolar lavage fluid (BALF) of children with CF. Total RNA was extracted from cell pellets of BALF from 43 children aged 6 months–5 years and analyzed by reverse transcription–quantitative real time PCR for gene expression of PPARγ, PON2, and P. aeruginosa lasI, the 3OC12HSL synthase. Patients with culture-confirmed P. aeruginosa infection had significantly lower gene expression of PPARγ and PON2 than patients without P. aeruginosa infection. All samples that were culture-positive for P. aeruginosa were also positive for lasI expression. There was no significant difference in PPARγ or PON2 expression between patients without culture-detectable infection and those with non-Pseudomonal bacterial infection, so reduced expression was specifically associated with P. aeruginosa infection. Expression of both PPARγ and PON2 was inversely correlated with neutrophil counts in BALF, but showed no correlation with other variables evaluated. Thus, lower PPARγ and PON2 gene expression in the BALF of children with CF is associated specifically with P. aeruginosa infection and neutrophilia. We cannot differentiate whether this is a cause or the effect of P. aeruginosa infection, but propose that the level of expression of these genes may be a marker for susceptibility to early acquisition of P. aeruginosa in children with CF.


Introduction
Individuals with cystic fibrosis (CF) are particularly susceptible to infection with the opportunistic pathogen Pseudomonas aeruginosa, and ultimately over 80% of adults with CF patients are infected with this organism [1]. It is widely accepted that P. aeruginosa present in the lungs of CF patients, particularly during chronic infection, exist in the form of biofilms which protect the bacteria from both antibiotics and the host's immune defenses. The reasons why CF patients are predisposed to P. aeruginosa infection are not clear, but it is known that infection is associated with more rapid decline in lung function [2,3].
A hallmark of CF is a hyperinflammatory response to infection [4,5], particularly with P. aeruginosa [6,7,8,9]. The mammalian transcription factor, peroxisome proliferator activated receptor (PPAR)c is a master negative regulator of inflammation, modulating signaling through NFkB and MAP kinases. It is expressed in respiratory epithelium [10] and has been reported to be expressed in immune cells such as macrophages and neutrophils [11,12,13,14]. The expression and function of PPARc have been reported to be low in human CF respiratory epithelial cell lines [10] and in cystic fibrosis transmembrane conductance regulator (cftr) knockout mice [15,16]. This deficiency may contribute to the hyperinflammatory response in CF. To date this has not been confirmed in ex vivo samples from CF patients. However, PPARc agonists have been reported to ameliorate intestinal symptoms in cftr knockout mice, and their potential use as therapy in chronic inflammatory disease has been widely discussed [17].
Biofilm formation and maturation of P. aeruginosa, together with the expression of a range of virulence factors, is regulated by quorum sensing signals that coordinate bacterial gene expression on a population-wide basis [18,19]. These signal molecules have consistently been detected in nanomolar amounts in sputum and lung tissue from CF patients infected with P. aeruginosa [20] but localised concentrations are suggested to be in the micromolar range [21]. It has also been demonstrated that one of the major quorum sensing signal molecules of P.aeruginosa, N-3-oxododecanoyl-L-homoserine lactone (3OC 12 HSL), can cross the mammalian cell membrane [22] and that it induces hyperinflammatory responses in CF airway epithelial cell lines [6]. We [23] and others [24] have demonstrated that 3OC 12 HSL can bind to and modulate the function of PPARc. We hypothesized that inhibition by 3OC 12 HSL of the function of an already low level of PPARc expressed in CF could further exacerbate hyperinflammation in cystic fibrosis. To assess the viability of this hypothesis and to evaluate the level of PPARc expressed in the lungs of children with CF, we examined PPARc gene expression in cells from bronchoalveolar lavage fluid (BALF), comparing children with CF with and without demonstrated P. aeruginosa infection.
Another factor that could affect the efficiency of P. aeruginosa biofilm formation and virulence factor expression in the lung is the presence of mammalian lactonases, in particular paraoxonase 2 (PON2). PON2 is an intracellular enzyme that has been demonstrated to efficiently degrade 3OC 12 HSL [25], a process that can have a quorum quenching effect. Mouse tracheal epithelial cells deficient in PON2 have impaired ability to inactivate 3OC 12 HSL [26], and PON2 expression has been reported to be inhibited by 3OC 12 HSL [27]. We therefore also investigated the gene expression of PON2 in the cells from BALF of the children with CF.
We found that the expression of PPARc and PON2 in BALF cells was significantly lower in patients infected with P. aeruginosa and was inversely correlated with total neutrophils in the BALF. Our results suggest that low PPARc and PON2 expression is specifically associated with P. aeruginosa infection and neutrophilia in BALF.

Ethics Statement
Children with CF aged up to 5 years undergoing surveillance BAL between January 2009 and April 2011 and whose parents had given informed written consent for participation in a study of early lung disease at The Sydney Children's Hospital, Randwick, Australia, contributed an aliquot of BALF for the current analyses. This study was approved by the South Eastern Sydney Area Health Service Human Research Ethics Committee (Approval no. 02/098) and registered at the Australian and New Zealand Clinical Trial Register (ACTRN12611000945921).

Patients
Children with CF had been identified through newborn screening or meconium ileus presentation and the diagnosis confirmed by sweat chloride analysis (chloride concentration .60 mmol/mL) and/or cftr mutation analysis. Demographic and clinical variables were obtained from the patient's medical records and are summarized in Table 1.

BAL and Sample Processing
Flexible bronchoscopy with BAL was performed as previously described [28]. In brief, BAL was performed under general anesthesia. Suctioning through the bronchoscope was avoided until the tip had passed beyond the carina to avoid upper airway contamination. BAL was sequentially performed in three lobes, right upper lobe, right middle lobe and lingual, using a single aliquot (1 mL/kg, minimum 10 mL, maximum 20 mL) of sterile saline to each. BAL fluid samples were pooled.
For gene expression analysis, the cells contained in 0.5 mL of BALF were pelleted by centrifugation and stored in 300 ml of RNAlaterH (Ambion, Mulgrave VIC, Australia) at 220uC until required. For a small number of patients, two samples taken at different times were available. Only the earliest sample from these patients was included in the main analysis.

Detection of Bacteria in BALF
The presence of bacteria and fungi in BALF were evaluated by standard microbiological culture methods at Sydney Children's Hospital. Airway infection for all microbes was defined by Sydney Children's Hospital as pathogen growth .10 5 colony-forming units per mL (cfu/mL) of BALF [29,30].

RNA Extraction and cDNA Synthesis
Total RNA was extracted from BALF cells using TRIzolH reagent (Invitrogen, Mulgrave, VIC, Australia). RNAlaterH was removed from samples, and cells resuspended in 750 ml of TRIzolH before being transferred to a 2 ml tube containing 1 ml of 0.1 mm zirconia/silica beads (Daintree Scientific, St Helens, TAS, Australia.). A further 650 ml of TRIzolH was added before cells were homogenized for 2 min on maximum speed using a Mini-Beadbeater (Daintree Scientific.). Beads and cellular debris were pelleted by centrifugation at 10 0006g for 10 min at 4uC. Approximately 1 ml of supernatant was removed to a fresh tube and total RNA extracted as per the TRIzolH protocol with the addition of 5 mg of glycogen (Invitrogen) to the upper phase to aid RNA precipitation. Isolated RNA was resuspended in 25 ml of DEPC-treated water (Ambion) and DNAse treated using a TURBO DNA-free kit TM (Ambion) according to the manufacturer's instructions. The concentration and purity of recovered RNA was quantified using a NanoDrop TM 1000 spectrophotometer (Thermo Scientific, Scoresby, VIC, Australia).
Up to 200 ng of total RNA was included in a cDNA synthesis reaction using a Transcriptor First Strand cDNA Synthesis kit (Roche, Dee Why, NSW Australia) according to the manufacturer's instructions. The reaction included both anchored oligo d(T) 18 and random hexamer primers to ensure reverse transcription of both mammalian and bacterial transcripts. cDNA was stored at 220uC until required. Four of these samples also had P. aeruginosa and four had H. influenzae. % Three of these samples also had S. aureus; one had P. aeruginosa. PPARc primers detected transcripts for PPARc isoforms 1 and 2. Primers used for the detection of P. aeruginosa 16S rRNA [31] and lasI [32] have been published elsewhere. Human gene expression relative to the housekeeping gene b-actin was calculated using 2 2DCp . The presence of P. aeruginosa 16S rRNA in BAL fluid samples was defined as positive by both Cp and melt curve analysis, with level of infection defined as high (Cp, 20), medium (Cp$20 and ,30) and low (Cp$30 and #40). Negative for P. aeruginosa by PCR was defined as the absence of specific PCR product. Detection of lasI expression was defined as the presence of specific PCR product and Cp#40 cycles.

Statistical Analysis
All statistical analyses were performed using Graphpad Prism v4.03 (Graphpad Software, La Jolla, CA, USA). Microbiological culture results were coded positive or negative. Genotype was coded as F508del homozygous, F508del heterozygous or other. Results of RT-qPCR for P. aeruginosa were coded as negative, low, medium or high and lasI was coded as negative/positive. Age at BAL was expressed in days, and gene expression of PPARc and PON2 calculated relative to that of the reference gene, b-actin. Because PPARc and PON2 gene expression levels were not normally distributed, all analysis was performed using nonparametric statistics: Mann-Whitney U test for comparisons between two groups, Kruskal-Wallis analysis with Dunn's multiple comparison test for comparisons between multiple groups and Spearman correlation for assessing correlations between continuous variables. Fisher's exact test was used to evaluate associations between categorical variables.

Results
Gene Expression of PPARc in BALF Cells of Children with CF is Lower in those Who have P. aeruginosa Infection Cells from BALF samples from all subjects were examined for PPARc gene expression by RT-qPCR, for the presence of P. aeruginosa by culture and RT-qPCR, and for the presence of any other pathogen by culture only. ''Culture positive'' and airway infection were defined as detection of .10 5 cfu/mL in BALF [29,30]. Figure 1 shows that PPARc gene expression was significantly lower (approximately threefold, P = 0.003, Mann-Whitney U test) in BALF from children with P. aeruginosa infection than in those without P. aeruginosa infection.
We also assessed the presence and level of P. aeruginosa and the expression of lasI in BALF by RT-qPCR. Semi-quantitation of P. aeruginosa 16S RNA was performed based on Cp as described in the Materials and Methods; lasI expression was classified as positive or negative. All samples that were culture-positive for P. aeruginosa were also positive for lasI expression by RT-qPCR. All other samples were negative for lasI, including those that were culture negative but where low or medium levels of P. aeruginosa were detected by RT-qPCR. As shown in Table 2, there was an excellent correlation between detection of high or medium levels of P. aeruginosa 16S rRNA expression in BALF by RT-qPCR and culture detection of .10 5 cfu/mL P. aeruginosa (P,0.0001, Fisher's exact test), with RT-qPCR showing a sensitivity of 86% and specificity of 97% relative to culture.
We also evaluated PPARc gene expression in cells from BALF based on P. aeruginosa RT-qPCR results (Figure 2), which confirmed that detection of medium or high levels of P. aeruginosa was correlated with low gene expression of PPARc.
Gene Expression of PPARc is Low in Children with P. aeruginosa Infection but not in those Infected with Other Pathogens Most patients had more than one pathogen detected in BALF, so we were unable to meaningfully analyze the effects on PPARc gene expression of infection with each individual type of microbe because of possible interactions between them. For example, there were three patients with only P. aeruginosa  influenzae. Therefore, to determine whether low PPARc gene expression was merely the result of infection in general, or was specifically associated with P. aeruginosa infection, we used culture results to divide patient samples into four groups: those without detected pathogens (n = 13), those with P. aeruginosa alone or in combination with other pathogens (n = 7), those with H. influenzae and/or with S. aureus (non-Pseudomonal bacteria group, n = 15) and those with Aspergillus fumigatus with or without non-Pseudomonal bacteria (n = 8). The results shown in Figure 3 demonstrate that only the presence of P. aeruginosa was associated with significantly lower PPARc gene expression.

Gene Expression of PON2 is Low in Patients with P. aeruginosa Infection but not with other Bacterial Infections
Because the human lactonase PON2 has been shown to efficiently inactivate and degrade the P. aeruginosa signaling molecule 3OC 12 HSL that can modulate PPARc activity, we were interested in evaluating whether the gene expression of PON2 was also correlated with P. aeruginosa infection. We hypothesized that if PON2 expression was also low in patients with P. aeruginosa infection, then this could contribute to their susceptibility to P. aeruginosa infection because they would have a reduced ability to degrade and inactivate 3OC 12 HSL. The results shown in Figure 4 confirm that PON2 gene expression is approximately twofold lower in those children with P. aeruginosa infection than in those infected with non-Pseudomonal bacteria or without detectable infection. There was one sample in the no infection group for   which PON2 expression data was unavailable, leaving 12 patients in that group.

Correlation of Other Variables with P. aeruginosa Infection and with PPARc and PON2 Gene Expression
The results described above indicate an association between P. aeruginosa infection and low gene expression of PPARc and PON2. To analyze this further, we evaluated associations of a range of variables with the presence of .10 5 cfu/mL P. aeruginosa in BALF (''Pseudomonas positive'') ( Table 3) or PPARc and PON2 gene expression (Table 4). To evaluate the association of PPARc and PON2 gene expression with inflammation, we used the total leukocyte and total neutrophil counts in BALF. As the severity of CF disease has been reported to differ between males and females [33,34]and will also depend on the type of cftr mutation present in the patients, we evaluated associations with sex and genotype. The effect of patient age was also evaluated, because the incidence of P. aeruginosa in CF increases with age, and in some animal models gene expression of PPARc has been reported to change with age [35,36]. Lastly, because PPARc function is important in regulating insulin sensitivity [37] and polymorphisms in PON2 have been associated with diabetes [38], we evaluated associations with fasting blood glucose levels. We were unable to obtain glucose tolerance test results for many patients, although this is a more reliable measure of CF-related diabetes [39]. We also considered the use of azithromycin, because this antibiotic has been reported to modulate quorum sensing and biofilm formation in P. aeruginosa [40], but as only one patient was being treated with azithromycin at the time of BAL, this variable was not included in the analysis. Neutrophil and total leukocyte counts were significantly higher in P. aeruginosa-positive samples (P = 0.006 and P = 0.009 respectively) compared with negative samples. The results showed no significant differences between P. aeruginosa-positive and P. aeruginosa-negative samples for any of the other variables.
When evaluating associations between each variable and PPARc expression, the only correlations observed were an inverse correlation between PPARc expression and total leukocytes (P = 0.0016), and an extremely strong inverse correlation for PPARc expression and total neutrophils (P,0.0001); i.e. low PPARc gene expression is associated with high total leukocyte and neutrophil counts. For PON2, there was an inverse correlation of PON2 gene expression with total neutrophils (P = 0.018), but no other significant correlations.

PPARc and PON2 Gene Expression in Different BALF Samples from Individual Patients
To attempt to evaluate whether P. aeruginosa infection was associated with changes in PPARc and PON2 gene expression in individual patients, we seperately examined two consecutive BALF samples from six patients who either acquired P. aeruginosa infection between the first and second sample (n = 3), or cleared P. aeruginosa infection between the first and second sample (n = 3). There was no consistent pattern of change in PPARc or PON2 gene expression. Results from another group of 16 patients for whom we had at least two BALF samples but who did not change their P. aeruginosa infection status (either stayed negative or stayed positive) also showed no significant change in median PPARc and PON2 gene expression between samples.

Discussion
Previous studies in human cell lines and cftr knockout mice have suggested that expression of PPARc is reduced in CF [10,15,16], although this has not been directly demonstrated in individuals with CF, and no previous study has suggested an association of PPARc expression with P. aeruginosa infection. There is no information in the literature on the expression of PON2 directly in CF. Thus, this study is the first to examine expression of the genes encoding PPARc and PON2 in cells from BALF of CF children with and without P. aeruginosa infection.
The results of this study show that in children with CF, low expression of PPARc and PON2 genes in BALF cells is associated with P. aeruginosa infection but not with the presence of other pathogens. Patients with P. aeruginosa infection had significantly higher neutrophil and leukocyte counts compared to those without, which has been previously reported [41]. The findings also demonstrate that PPARc gene expression, and to a lesser extent PON2 gene expression, shows a strong inverse correlation with neutrophil counts, indicating that low PPARc and PON2 gene expression is associated with high levels of inflammation. It is possible that the observed correlation is due to sampling from different cell populations and we were unable to determine the source of detected PPARc transcripts in our study. However PPARc has been reported to be expressed in bronchial epithelial cells [10,42], alveolar macrophages [14] and neutrophils [11,12,13].
These findings were consistent whether P. aeruginosa infection was detected by culture or by RT-qPCR. While microbiological culture remains the gold standard detection method, our results indicate that detection by RT-qPCR shows an excellent correlation with culture results. Furthermore, detection of low levels of P. aeruginosa in patients classified on culture results as uninfected suggests that RT-qPCR may detect lower levels of bacteria than culture, although the clinical relevance of this is uncertain. However, the findings suggest that RT-qPCR could be a useful confirmatory technique for P. aeruginosa infections.
Our results also showed that expression of the bacterial 3OC 12 HSL synthase lasI gene was detectable in BALF of all patients with culture-defined P. aeruginosa infection, but in none of the other patients. This confirms that at least a subpopulation of the P. aeruginosa present in BALF in all patients with culturedefined infection with P. aeruginosa were able to produce 3OC 12 HSL, even in those patients who had been infected for up to three years. This contrasts with reports that isolates of P. aeruginosa from chronically-infected CF patients accumulate mutations and lose quorum sensing activity [43,44]. However, these studies reported quorum sensing gene mutations in clinical isolates rather than population-wide gene expression directly in the CF lung. It is likely that mixed populations of P. aeruginosa exist in the CF lung, some of which may harbor mutations in genes involved in quorum sensing [45,46]. In addition, it has been reported that quorum sensing activity in isolates is not lost until the late stages of chronic infection [44]. It is important to emphasize that our study reports data on gene expression, not protein activity, of PPARc and PON2. However, a positive regulatory loop where increased functional protein leads to increased gene expression, and inhibition of protein function leads to decreased mRNA expression, has been demonstrated for PPARc [47] and other PPAR family members PPARa [48] and PPARd [49]. Thus, there is good evidence that gene expression and protein function of PPARc are closely correlated. Our finding that expression of the PPARc gene is strongly inversely correlated with the presence of an inflammatory neutrophil infiltrate is also indirect evidence that PPARc transcriptional activity is likely to be reduced in P. aeruginosa-positive patients, because PPARc protein is a transcription factor that is known to downregulate inflammatory gene expression [37]. Although information about the association between PON2 gene expression and protein function is limited, it is known that substrate inhibition of gene expression and probably protein levels occurs [27]. Thus, while it remains necessary to confirm the levels of functional PPARc and PON2 protein in these patients, evidence suggests that these too are likely to be reduced.
This was primarily a cross-sectional study and although we had paired BALF samples from five patients who acquired or cleared P. aeruginosa infection, changes in gene expression between these samples were inconsistent. The changes in gene expression between these patients did not differ from changes seen in paired samples from 10 patients who did not change P. aeruginosa infection status. Because of the limited number of samples, we were unable to conclusively show that expression of PPARc and PON2 genes either changed or remained the same in individual patients upon infection with P. aeruginosa. Thus our data did not allow us to determine whether low PPARc predisposes patients to early acquisition of P. aeruginosa. Our hypothesis that 3OC 12 HSL inhibits PPARc function in CF lungs remains plausible, because we showed expression of lasI, the 3OC 12 HSL synthase, in the BALF of all patients with culture-defined P. aeruginosa infection. However, confirmation of this hypothesis requires further study, including analysis of expression of PPARc protein levels and function, that is not possible with the limited material available from BALF. The alternative possibility, that low PPARc expression predisposes CF patients to early acquisition of P. aeruginosa, also remains plausible, and could provide a useful screening test that might allow preemptive treatment in those children at higher risk. This requires further studies of sequential samples from individual patients, to determine whether those who have low PPARc expression acquire P. aeruginosa earlier in life or at a higher frequency than patients with higher PPARc expression.
Our observation that lower expression of the PON2 gene is associated with P. aeruginosa infection in CF patients is also of importance. PON2 is a member of the paraoxonase family of human enzymes that have lactonase activity [25] and can inactivate and degrade 3OC 12 HSL [25,50,51]. Thus, reduced PON2 activity could facilitate chronic infection and biofilm formation by P. aeruginosa by allowing higher levels of functional 3OC 12 HSL to accumulate in the CF lung. Indeed, PON2deficient mouse tracheal epithelial cells allow increased P. aeruginosa quorum sensing activity [50] and PON2 expression has been shown to be downregulated by 3OC 12 HSL [27]. Further, PON2 has antioxidant activity and can protect cells against the effects of the P. aeruginosa virulence factor pyocyanin [27]. Importantly, PON2 has activity against a range of homoserine lactone signaling molecules, including that produced by Burkholderia cepacia complex, but as none of our patients had detectable infection with this microorganism, we were unable to evaluate any associations.
Overall, our study represents the first demonstration of two host factors that are specifically associated with early childhood infection with P. aeruginosa in individuals with CF. While our results do not allow us to determine whether this association is cause or effect, they provide a useful starting point for designing new therapeutic strategies. For example, treating CF patients who have low PPARc expression with a pharmacological PPARc agonist such as a glitazone may allow augmentation of PPARc activity and provide some protection against P. aeruginosa infection.