Chronic obstructive pulmonary disease (COPD) is a complex disease, the pathogenesis of which remains incompletely understood. Colonization with Pneumocystis jirovecii may play a role in COPD pathogenesis; however, the mechanisms by which such colonization contributes to COPD are unknown. The objective of this study was to determine lung gene expression profiles associated with Pneumocystis colonization in patients with COPD to identify potential key pathways involved in disease pathogenesis. Using COPD lung tissue samples made available through the Lung Tissue Research Consortium (LTRC), Pneumocystis colonization status was determined by nested PCR. Microarray gene expression profiles were performed for each sample and the profiles of colonized and non-colonized samples compared. Overall, 18 participants (8.5%) were Pneumocystis-colonized. Pneumocystis colonization was associated with fold increase in expression of four closely related genes: INF-γ and the three chemokine ligands CXCL9, CXCL10, and CXCL11. These ligands are chemoattractants for the common cognate receptor CXCR3, which is predominantly expressed on activated Th1 T-lymphocytes. Although these ligand–receptor pairs have previously been implicated in COPD pathogenesis, few initiators of ligand expression and subsequent lymphocyte trafficking have been identified: our findings implicate Pneumocystis as a potential trigger. The finding of upregulation of these inflammatory genes in the setting of Pneumocystis colonization sheds light on infectious-immune relationships in COPD.
chronic obstructive pulmonary disease
chemokine (C-X-C motif) ligand
diffusing capacity for carbon monoxide
false discovery rate
forced expiratory volume in 1 s
forced vital capacity
Global Initiative for Chronic Obstructive Lung Disease
Lung Tissue Research Consortium
quantitative real-time polymerase chain reaction
Recent advances in the characterization of COPD have established that this disease is remarkably varied in its presentation and progression, and that our understanding of its pathogenesis is incomplete [1-3]. Most smokers do not develop COPD, suggesting that cigarette smoking alone cannot completely account for disease burden. Because unique disease contributors may be amenable to novel interventions, identification of independent triggers and molecular pathways that potentiate the development and progression of COPD is essential.
One cofactor that may accelerate COPD is infection. Bacterial infections of the lower airways have been associated with exacerbations and severity of COPD . Additionally, several studies have shown that carriage of microorganisms (colonization) is common among persons with COPD, and that presence of respiratory tract colonizers is associated with airway and alveolar inflammation [5-8].
Pneumocystis jirovecii, a frequent lower respiratory pathogen in immunocompromised hosts, colonizes the lungs of immunocompetent persons, particularly those with underlying airway disease, and may be related to COPD [9-12]. A study of immunocompetent smokers with COPD demonstrated Pneumocystis carriage prevalence of 19.1% overall and 36.7% among persons with very severe COPD , higher than seen in individuals with other end-stage lung diseases including idiopathic pulmonary fibrosis or cystic fibrosis [10, 12, 14]. The presence of Pneumocystis colonization is associated with increasing severity of COPD according to the GOLD classification, independent of smoking history . These findings suggest an association between COPD and Pneumocystis colonization.
Animal models also provide support for a causal relationship between Pneumocystis colonization and COPD. Immunosuppressed macaques  and immunocompetent, smoke-exposed mice  develop airflow obstruction and emphysema following colonization with Pneumocystis. Colonized immunosuppressed macaques more frequently have bronchial-associated lymphoid tissue and have greater amounts of lung inflammatory cytokines than do non-colonized controls . Pneumocystis-colonized mice had increased bronchoalveolar lavage lymphocytes and alveolar macrophages . These findings suggest that pulmonary inflammation mediated by colonization may play a role in development of COPD.
While it appears that Pneumocystis colonization is a unique risk factor in the development and worsening of COPD, the mechanisms of pathogenesis are unclear. In order to identify important pathways involved in the pathogenesis of COPD in association with Pneumocystis colonization, we investigated whole lung gene expression in participants with COPD and compared the profiles of Pneumocystis-colonized and non-colonized samples. Using microarray, we analyzed distinct gene expression patterns associated with Pneumocystis colonization in order to identify specific mechanisms of COPD development that could lead to targeted, personalized therapies in at-risk subjects.
MATERIALS AND METHODS
Subjects were participants in the National Heart Lung and Blood Institute-sponsored LTRC. LTRC protocols were approved by the institutional review boards of each participating site (including the Mayo Clinic, Rochester; University of Michigan; University of Pittsburgh and University of Colorado). This study was also approved by the University of Pittsburgh institutional review board. Study protocols conform to the provisions of the Declaration of Helsinki. All participants completed written informed consent.
Lung Tissue Research Consortium participants who met criteria for COPD based either on irreversible airflow obstruction on spirometry (as defined by post-bronchodilator FEV1 to FVC ratio of <70%) or on evidence of radiographic or pathologic emphysema as determined by standardized assessment of chest CT scans or lung tissue samples. Lung tissue samples were obtained from individuals undergoing lung explant, resection, or biopsy. Tissue samples from patients with pulmonary malignancy were obtained at a margin of at least 5 cm from the malignant lesion. One cubic centimeter tissue blocks were flash frozen and stored at −80 °C. Relevant clinical and questionnaire data, quantitative CT analyses, and pulmonary function testing were obtained from the LTRC database. Spirometry and DLCO were performed according to LTRC protocols and percent predicted values were calculated using Hankinson or Neas equations, respectively [17, 18]. Persons with physiologic obstruction were categorized by the GOLD classification . Subjects were excluded if they had been treated with systemic immunosuppressants other than corticosteroids. Quantitative CT analyses were performed using previously described protocols [20, 21].
Total RNA was extracted using the QIAcube system (QIAgen, Valencia, CA, USA) with an mRNEasy kit. A Nanodrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) was used to measure RNA concentrations and a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA) to assess RNA quality. An RNA Integrity Number greater than 7.0 was used as the criterion for acceptable quality. The first 92 samples (chosen randomly) were labeled and hybridized to Agilent V2 Human Whole Genome microarrays (Agilent Technologies) as previously described ; the remaining samples along with 10 repeat subjects (to account for potential batch effects) were hybridized to Agilent V3 Human Whole Genome microarrays. The arrays were scanned using the Agilent Microarray Scanner and the data extracted using Agilent Feature Extraction software version 9.5.3. These array data can be downloaded from the Lung Genomics Research Consortium website at www.lung-genomics.org and from the LTRC website (www.ltrcpublic.com); they have also been deposited in the Gene Expression Omnibus at the National Center for Biotechnology Information (Accession Number GSE47460).
Between the two platforms, common probes represented 15,261 individual genes. Common probes were first selected in both platforms by matching their probe IDs. Duplicated probes in each array were collapsed to their mean values. For sample-wise normalization, the cyclic loess method implemented in R Bioconductor was used, as previously described . Finally, among probes having the same gene symbols, only the probe with the largest interquantile range was considered. Even though samples were repeated frequently as a hedge against batch effects, only one array per subject was included in the analysis of differential gene expression.
Direct comparisons between groups of subjects was performed using the BRB ArrayTools software package v. 4.1 (available at linus.nci.nih.gov/BRB-ArrayTools.html) using Student's t-test with P < 0.001 as the threshold for significance. Unidimensional clustering was performed using the ScoreGenes 1.0 software package .
Determination of Pneumocystis colonization
DNA was isolated from 1 cm lung tissue blocks using a standard technique (tissue blocks were the same blocks from which the RNA was extracted). Lung tissue was first homogenized in ATL tissue lysis buffer followed by DNA purification with a QIAamp blood DNA Mini Kit (QIAgen). Nested PCR was performed at the Pneumocystis mitochondrial large subunit ribosomal RNA gene, as previously described . Negative and positive controls were included, and the absence of PCR inhibitors was confirmed by PCR for beta-globin . Positive nested PCR products were purified and confirmed to be P. jirovecii by sequencing reactions. A subject was considered Pneumocystis-colonized if Pneumocystis DNA was detected and determined by sequencing to be human Pneumocystis.
Quantitative real time polymerase chain reaction
Quantitative real time PCR was performed for selected genes of interest to validate microarray findings of significant differential expression. Samples positive for P. jirovecii were matched 1:2 with Pneumocystis-negative control samples provided two such controls were available. Matching criteria included systemic steroid use, inhaled steroid use, smoking history, pack years (within 10 years when possible), and age (within 5 years when possible). RNA was extracted and quantified as described above and reverse transcribed to cDNA using a high capacity RNA-to-cDNA kit (Applied Biosystems, Carlsbad, CA, USA). qRT-PCR reactions were performed in duplicate on a StepOnePlus system (Applied Biosystems) using TaqMan gene expression assays (Applied Biosystems) and following the protocols and standard cycling parameters recommended by the manufacturer. Genes of interest were normalized to the reference gene GAPDH. Gene expression data were analyzed with REST 2009 software (QIAgen).
Stata 12 (StataCorp LP, College Station, TX, USA) was used for analysis and significance set at P ≤ 0.05. Relevant clinical characteristics, including age, smoking history (qualitative and quantitative), corticosteroid use, macrolide use, and cancer diagnosis were compared between the Pneumocystis-colonized and Pneumocystis-negative groups. Colonization status was correlated with severity of airflow obstruction, DLCO and quantitative percent emphysema. Univariate analyses were employed to compare clinical characteristics between groups, using either the two-tailed t-test or Mann–Whitney rank sum for continuous variables or χ2 or Fisher's exact test for categorical variables. Smoking status and age were identified a priori as potential confounders and controlled for using multivariable regression. Because missing data precluded adjustments based on pack years, history of ever-smoking (>100 cigarettes) was used as the covariate in these analyses.
Characteristics of the cohort
There were 211 subjects with physiologic, radiographic, or pathologic findings consistent with COPD; their characteristics are presented in Table 1. The median age was 66 years and the majority of participants (54.5%) were male. Of participants for whom smoking data was available (n = 207), almost all (95.7%) were current or previous smokers; the median pack years smoked was 45. Sixty percent of participants had a lung cancer diagnosis, the overwhelming majority of which (98.4%) were non-small cell lung cancer. Although not all participants had spirometric evidence of airflow obstruction (participants were included if they either had spirometry consistent with airflow obstruction or radiographic/histologic evidence of emphysema), the median post-bronchodilator FEV1 percent predicted was 59% and median post-bronchodilator FEV1/FVC ratio 53%. The median GOLD stage for smokers was II (22 GOLD 0, 21 GOLD I, 87 GOLD II, 29 GOLD III, and 49 GOLD IV). Median percent predicted DLco was 55%. The median quantitative emphysema score was 7.1%.
|Overall (n = 211a)||No colonization (n = 193a)||Pneumocystis-colonized (n = 18a)|
|Male, n (%)||115 (55)||104 (54)||11 (61)|
|White, n (%)||205 (97)||188 (97)||17 (94)|
|Age (years), mean (SD)||65.1 (10.0)||65.1 (10.2)||65.7 (8.7)|
|Smoking history (pack years), median (range)||45 (0–212)||45 (0–212)||54 (0–182)|
|Smoking history, n = 207|
|Current smoker, n (%)||12 (6)||12 (6)||0 (0)|
|Ever smoker (>100 cigarettes), n (%)||198 (95)||181 (96)||17 (94)|
|Using inhaled steroidsb, n = 208, n (%)||97 (45)||85 (45)||12 (67)|
|Using oral steroids‡, n = 208, n (%)||38 (18)||34 (18)||4 (22)|
|Using oral macrolideb, n = 208, n (%)||1 (<1)||1 (<1)||0 (0)|
|Cancer diagnosis, n = 208, n (%)||127 (60)||116 (60)||11 (61)|
According to nested PCR, 18 participants (8.5%) were colonized with Pneumocystis. There were no statistically significant differences in relevant baseline characteristics (including age, smoking history, or steroid use) between the Pneumocystis-colonized and non-colonized groups (Table 1). There was no correlation between Pneumocystis positivity and tissue location (upper lobe 7.6% colonized, lower lobe 8.5% colonized; P = 0.16).
Association of Pneumocystis with pulmonary function testing and quantitative emphysema scoring
Post-bronchodilator spirometry and DLCO values were available for 175 subjects. The post-bronchodilator FEV1 was significantly lower in the Pneumocystis-colonized than non-colonized group (median 49% predicted vs. 60% predicted, respectively, P = 0.04), as was the post-bronchodilator FEV1/FVC ratio (47% vs. 54%, respectively, P = 0.014). There were no significant differences between groups in DLCO or quantitative emphysema scores. When adjusted for smoking, the FEV1/FVC ratio remained significantly different between the colonized and non-colonized groups (Table 2).
|Overall (n = 211a)||No colonization (n = 193a)||Pneumocystis-colonized (n = 18a)||Rank sum P-value||Adjusted associations coef. (95% CI); P-value|
|FEV1, n = 175, median % predicted (range)b||58 (14–123)||60 (14–123)||49 (17–83)||0.048||−13.5 (−27.2 to 0.3); 0.056|
|FEV1/FVC, n = 175, median % (range)b||53 (17–88)||54 (17–88)||47 (23–64)||0.01||−0.1 (−0.2 to 0); 0.02|
|GOLD, n = 175, median (range)||2 (0–4)||2 (0–4)||3 (0–4)||0.4||—|
|DLco, n = 191, median % predicted (range)||55 (13–115)||56 (13–115)||47 (17–96)||0.4||−4.6 (−14.1 to 8.6); 0.4|
|Emphysema score, n = 196, median % (range)||7.1 (0.01–62.0)||6.6 (0.01–62.0)||16.1 (0.15–51.4)||0.1||5.8 (−1.8 to 13.3); 0.1c|
Gene expression analysis
Gene expression profiling revealed 50 genes with differential expression at or below a threshold P-value of 0.001 (Fig. 1). Of these 50 genes, 9 were below a FDR threshold of 0.05, with fold changes between colonized and non-colonized groups of 1.98–3.20 (Table 3). Several of the resultant genes are closely associated with the inflammatory Th1 pathway. The three genes of interest with the highest differential expression, CXCL9 (interferon-gamma induced monokine), CXCL10 (interferon-inducible 10 kDa protein), and CXCL11 (interferon-inducible T cell chemoattractant), encode the ligands for the chemokine receptor CXCR3. CXCR3 is a 7-transmembrane g-protein coupled cell surface receptor preferentially expressed on activated Th1 cells and is a master regulator of INF-γ, which was also significantly differentially expressed. qRT-PCR was performed on a subset of the samples (Pneumocystis-positive samples matched to Pneumocystis negative samples, as described above) to validate differential expression of the gene products CXCL9, CXCL10, CXCL11, and INF-γ. Of 36 samples initially identified as acceptable matches (12 Pneumocystis-positives, 24 controls), 30 (11 Pneumocystis-positives, 19 controls) were successfully transcribed to cDNA (eight matched 1:2; three matched 1:1). The success of the matching on key clinical characteristics was confirmed (Table 4). The three chemokine ligands were confirmed to be significantly upregulated in Pneumocystis-positive samples, and there was a trend toward upregulation of INF-γ (Table 4).
|Human gene||Human gene name||FDR||Fold ratio|
|CXCL9||Chemokine (C-X-C motif) ligand 9||0.00153||3.204959|
|CXCL10||Chemokine (C-X-C motif) ligand 10||0.00992||3.079771|
|CXCL11||Chemokine (C-X-C motif) ligand 11||0.0193||2.931513|
|KCNJ10||Potassium inwardly-rectifying channel, subfamily J, Member 10||0.0487||2.257885|
|IDO1||Indolamine 2,3-dioxygenase 1||0.0487||2.231945|
|GBP5||Guanylate binding protein 5||0.0487||2.038483|
|FAM26F||Family with sequence similarity 26, member F||0.0487||1.976551|
|Clinical characteristics of selected Pneumocystis-positives and matched controls|
|Pneumocystis-colonized (n = 11)||Non-colonized (n = 19)||P-valuea|
|Age, mean (SD)||67.1 (10.3)||65.5 (12.2)||0.7|
|Pack-years smoked, med (IQR)||48 (10–60)||50 (25–60)||0.8|
|Ever smokers, n (%)||10 (90.2%)||18 (94.7%)||1.0|
|Using systemic steroids, n (%)||3 (27.3%)||4 (21.1%)||1.0|
|Using inhaled steroids, n (%)||7 (63.6%)||13 (68.4%)||0.8|
|Cancer diagnosis, n (%)||8 (72.7%)||14 (73.7%)||1.0|
|qRT-PCR gene expression ratios|
|Human gene||Expression ratio (95% CI)b||P-value|
|CXCL9 (11 Pc, 19 control)||7.47 (0.04–443.13)||0.007|
|CXCL10 (10 Pc, 18 control)||5.09 (0.21–98.39)||0.002|
|CXCL11 (10 Pc, 18 control)||7.51 (0.22–333.51)||0.002|
|INF-γ (8 Pc, 13 control)||27.26 (0.00–2.58 × 106)||0.08|
In a well-characterized cohort of subjects with COPD, we performed microarray analysis of lung tissue to identify biologically relevant genes that were significantly differentially expressed among persons with and without Pneumocystis colonization. The strongest resulting genes, CXCL9, CXCL10 and CXCL11, encode related components of the inflammatory Th1 pathway and have previously been implicated as probable contributors to COPD pathogenesis via recruitment of activated T lymphocytes to the peripheral airways [26-28]. Our study thus identifies Pneumocystis colonization as a potential inflammatory trigger that initiates expression in the lung of these potent pro-inflammatory chemokines.
We found Pneumocystis colonization in 8.5% of individuals, a lower prevalence than reported in previous studies of COPD [9-14]. Two prior investigations including hospitalized persons with COPD reported colonization frequencies of 41–43% [9, 10]. Although it is likely that efficacy of Pneumocystis detection varies based on processing methods, sample type, and sample location (assays of sputum or tracheal aspirates, representing proximal airways, may be more sensitive for detection than assays of peripheral lung), Morris et al., who used similar methods and samples as in the current study, found a 19% prevalence of colonization among smokers with GOLD 0-IV COPD . The higher prevalence in previous studies may be attributable to differences in risk factors such as more severe COPD, more acute illness, or stronger smoking history. Differences in prevalence of colonization may also result from differences in amount or location of lung tissue samples. Given the nature of LTRC tissue collection, we had only one lung sample per subject (approximately 1 mL in volume) available for analysis. A recent study of lung samples from immunocompetent adults reported a 65% Pneumocystis colonization frequency when 3% of the right upper lobe was examined, suggesting that higher volume samples may increase detection . Additionally, we have previously shown that rate of detection of Pneumocystis colonization varies depending on location of origin of sample, a higher frequency of colonization being found when lower lobes are sampled . Given that only 20% of tissue specimens were obtained from the lower lobe, it is possible that we under detected Pneumocystis colonization in this cohort; however, there was no apparent relationship between location of origin of tissue block and Pneumocystis colonization in this study.
Because the mechanism by which Pneumocystis carriage influences the presentation of COPD is not known, we included both major phenotypes of COPD (irreversible airflow obstruction or emphysema) in our analysis. While prior human studies have determined an association between Pneumocystis colonization and airflow obstruction [13, 31], its relationship with markers of emphysema has not previously been explored. Animal models examining the effects of Pneumocystis colonization on lung pathology have demonstrated histologic evidence of both airway-centric bronchial-associated lymphoid tissue formation and emphysema [15, 16], implying that airways, alveoli, or both may be adversely affected by colonization. In the current study, the Pneumocystis-colonized group had significantly more severe airflow obstruction, an association that persisted when adjusted for smoking status. In contrast, comparisons of DLCO and radiographic emphysema showed no significant differences attributable to Pneumocystis colonization. While these results appear to suggest that Pneumocystis primarily contributes to COPD via mechanisms of airflow obstruction (vs. other latent infections, which have been associated with an emphysematous phenotype ), these findings must be interpreted with caution, given that the current sample was likely underpowered to detect a significant difference in radiographic emphysema between groups.
This study identified the principal upregulated genes associated with Pneumocystis colonization as the related CXC chemokine genes CXCL9-11. These chemokines are the three dominant ligands for the CXCR3 receptor, a 7-transmembrane g-protein coupled cell surface receptor that is preferentially expressed on activated T lymphocytes with Th1 polarity. As such, these ligands are highly chemotactic for activated Th1-type T cells, and are likely to play a role in COPD development.
Although IFN-γ and the CXCR3 ligands CXCL9 and CXCL10 are highly expressed in the bronchoalveolar lavage fluid of mice with PCP , a similar inflammatory environment in Pneumocystis colonization in humans has not previously been described. Studies by McAllister et al. have demonstrated that the CXCR3 chemokine CXCL10 is involved in host defense against Pneumocystis in both wild type and T-cell depleted mice, primarily through recruitment of CD8+ T cells to the lungs . Qin et al. evaluated the immune response in lung tissue in normal versus simian immunodeficiency virus-infected macaques, several of which developed PCP . The simian immunodeficiency virus-infected animals with PCP demonstrated markedly increased lung tissue CXCL9-11 expression when analyzed by immunohistochemistry. In the PCP animals only, the CXCL9-11 expressing cells (primarily macrophages) formed aggregates in the interstitium and alveoli; CXCR3-expressing lymphocytes were increased in the same regions. Additionally, IFN-γ mRNA was reportedly significantly overexpressed in lung tissue from animals with PCP compared with all other groups. These findings suggest that Pneumocystis infection directly induces the Th1 inflammatory pathway. The findings of the current study dovetail with those of these animal models of PCP and are the first to provide evidence that colonization with human Pneumocystis may induce activation of the potentially injurious Th1 inflammatory pathway via expression of ligands of CXCR3+ immune cells and IFN-γ.
The relationship between CXCL9-11 and COPD has been demonstrated in several recent studies. Sauty et al. were the first to demonstrate that human bronchial epithelial cells can be induced to express the three CXCR3 ligands in response to IFN-γ stimulation in vitro . Interestingly, expression of these ligands was not suppressed by administration of the corticosteroid dexamethasone, suggesting a basis for corticosteroid resistance in COPD. Saetta et al. examined peripheral tissue samples from persons with COPD, smokers, and non-smokers and characterized the chemokine ligand and receptor expression patterns in the pulmonary constitutive cells and immune cells, respectively . These authors demonstrated increases in CXCR3-expressing lymphocytes in the peripheral airways, epithelium, and submucosa of the COPD group, along with increased epithelial and submucosal expression of CXCL10. They noted that CXCR3 co-localized with CD3+/CD8+ (but not CD3+/CD4+) and with IFN-γ, in keeping with its Th1 association. The number of CXCR3+ cells present in the pulmonary epithelium and submucosa also correlated with a decrement in spirometric findings, supporting the contention that CXCL10/CXCR3-related lymphocyte recruitment contributes to COPD. Subsequently, a study by Grumelli et al. evaluated chemokine expression patterns of peripheral lung lymphocytes from persons with emphysema versus non-COPD controls . They found greater surface expression of CXCR3, increased production of the cognate ligands CXCL9-11, and increased IFN-γ production by the lymphocytes from emphysematous lungs than by those from non-COPD controls. Additionally, this group found that alveolar macrophages stimulated in the presence of CXCL9 and CXCL10 produce MMP-12 in a CXCR3-dependent fashion, suggesting a mechanism for the development of emphysema. Of note, in our previous study evaluating airway obstruction and amount of sputum MMP in HIV-infected persons, we found that MMP-12 was significantly increased in association with colonization . A more recent investigation by Costa et al. examined induced sputum samples from COPD patients and controls to clarify the relationships between each of these three chemokines and airway inflammation and obstruction; the amounts of all three chemokines were significantly greater in COPD patients than in non-smoking controls . Our current findings build on the relationships established by previous investigators by describing an association between expression of CXCL9-11 and IFN-γ and a biologically plausible trigger for the amplification of this inflammatory pathway.
This investigation sheds light on the relationships among Pneumocystis colonization, host adaptive immunity and COPD; however, it has several limitations. First, because we obtained relevant patient variables from an established prospective database, data on trimethoprim–sulfamethoxazole use were not available. It is therefore possible that we did not capture instances of Pneumocystis colonization that had been cleared by recent antibiotic administration. Additionally, because tissue was acquired at one discrete timepoint, we could not assess the chronicity of carriage. Future investigations may address the persistence of Pneumocystis detection in relationship to outcomes in COPD. Furthermore, the sampling method used for Pneumocystis in the study (using small volumes of lung parenchyma rather than bronchoalveolar lavage fluids or induced sputum, which represent larger portions of the lungs and may therefore be more sensitive for Pneumocystis detection) may have underdetected colonization and thus affected the analysis; however, underdetection would have been more likely to bias the analysis toward the finding of no difference in gene expression between colonized and non-colonized groups. An additional possible confounder in this particular population is the high prevalence of lung cancer (60% overall), which may alter the chemokine milieu in the lung. However, given that cancer diagnoses were equally distributed between the colonized and non-colonized groups, this likely had little effect on the analysis. Nonetheless, it is possible that there is an interaction between colonization and malignancy and that gene expression patterns in response to colonizers may differ in lungs without cancer. We could not explore this possibility in this cohort. Finally, because the microarray analysis used only one array per subject, there is potential for unstable results; prior evaluations of microarray have found that assays of samples from the same subjects may not result in reproducible data for differential expression of genes of interest . We attempted to mitigate the risk of inaccuracy in this study by choosing a moderately stringent FDR (5%) and by focusing our attention on genes with a fold change greater than 2 that were associated with a biologically plausible pathway. Additionally, the microarray findings for genes of particular interest were validated by the use of qRT-PCR.
Because the study design precluded us from addressing causal relationships, an alternative mechanism may account for our findings; namely, that the increases in CXCL9-11 are a function of an undetected factor that also influences the propensity to Pneumocystis colonization. However, in vitro evidence that Pneumocystis cysts directly induce expression of IFN-γ and CXCL9-11 mRNA in lung cells from immunocompetent macaques suggests that Pneumocystis could be causal in such inflammation . It is also possible that persons who are predisposed to colonization with Pneumocystis also harbor other bacteria, either colonizers or causing infections, which could contribute to changes in chemokine ligand expression. Emerging data from investigations of the lung microbiome suggest that complex communities of bacteria, viruses, and fungi are important in pulmonary disease states, including COPD [28, 34, 36]. Detectable Pneumocystis in COPD may be a marker of transformations in the entire microbial community; such analyses were beyond the scope of this study but warrant further investigation. A final limitation is that CXC9-11 expression cannot be directly localized to the lung epithelial cells. Although these chemokines are produced by lung epithelial cells [28, 34, 36], they are also expressed by neutrophils, lymphocytes, and alveolar macrophages, any or all of which may have been present in the biopsy specimens. It is possible that Pneumocystis colonization affects both the resident pulmonary immune cells and the pulmonary parenchyma; these associations will be interesting to investigate in future studies.
In summary, we have demonstrated that Pneumocystis colonization in persons with COPD is associated with increased transcription of three key ligands (CXCL9, CXCL10, and CXCL11) of CXCR3+ T lymphocytes, which are central actors in the Th1 inflammatory pathway. Although this mechanism has been implicated in the effective immune response to Pneumocystis pneumonia, the host response to colonization has not previously been described. Our data strongly suggest that colonization with Pneumocystis induces expression of CXCL3 ligands, which are known to attract effector T-cells to the lungs and also to induce alveolar macrophages to produce a metalloproteinase that is key in emphysema, MMP12 . These mechanisms, if persistent, may contribute to the pathogenesis and progression of COPD among persons who are Pneumocystis-colonized.
This study was supported by NIH 1F32HL114426 (M.E.F.); P50 HL084948; and N01 HR46163R01 (F.C.S.); RO1HL073745, R01HL095397, R01LM009657 (N.K.); R03 HL095370 (A.M.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This study utilized biological specimens and data provided by the LTRC supported by the National Heart, Lung, and Blood Institute.
The authors have no financial arrangements or competing interests to disclose.