AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 1 of 64
Plasma Sphingolipids associated with COPD Phenotypes
Author list
Name
email
Institutional Affiliation
Russell P. Bowler
[email protected] National Jewish Health, Denver, Colorado
Sean Jacobson
[email protected] National Jewish Health, Denver, Colorado
Charmion Cruickshank
[email protected] National Jewish Health, Denver, Colorado
Grant J Hughes
[email protected] Department of Biostatistics and Informatics, University of Colorado Denver
Charlotte Siska
[email protected] Computational Bioscience Program, University of Colorado Denver
Daniel S.Ory,
[email protected] Diabetic Cardiovascular Disease Center and Department of Medicine, Washington University
Irina Petrache
[email protected] Indiana University School of Medicine, Richard Roudebush VA Medical Center, Indianapolis, IN
Jean E Schaffer
[email protected] Diabetic Cardiovascular Disease Center and Department of Medicine, Washington University
Nichole Reisdorph
[email protected] National Jewish Health, Denver, Colorado
Katerina Kechris
[email protected] Department of Biostatistics and Informatics, University of Colorado Denver
Corresponding author(s) Russell Bowler, M.D., Ph.D., 1400 Jackson Street, Room K715a Denver, Colorado 80206 Phone: 303 398 1639 Fax: 303 270 2249 Email:
[email protected] Irina Petrache, MD 980 W. Walnut Street, Walther Hall – R3, C400 Indianapolis, IN 46260 Phone: 317-278-0428 email:
[email protected] 1
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 2 of 64
Author contributions: RB, NR, KK designed the study; CC, DO, JS, NS generated data; SJ, CC, GH, CS, KK analyzed data; RB, IP, NR, KK interpreted data and wrote the manuscript.
Funding Support: NHLBI (U01 HL089856, U01 HL089897), NCRR/NIH (UL1 RR025780, S10 RR023703), NIH/NCATS Colorado CTSI (UL1 TR000154) P20 HL-113444, P20 HL-113445 Additional information: This article has a supplement.
Running title: Plasma Sphingolipids in COPD.
Descriptor: 9.9 COPD: General
Manuscript word count: (3437 words)
2
Copyright © 2014 by the American Thoracic Society
Page 3 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Abstract
Rationale: Chronic obstructive pulmonary disease (COPD) occurs in a minority of smokers and is characterized by intermittent exacerbations and clinical subphenotypes such as emphysema and chronic bronchitis. Although sphingolipids as a class are implicated in the pathogenesis of COPD, the particular sphingolipid species associated with COPD subphenotypes remain unknown.
Objectives: To use mass spectrometry to determine which plasma sphingolipids are associated with subphenotypes of COPD.
Methods: 129 current and former smokers from the COPDGene cohort had 69 distinct sphingolipid species detected in plasma by targeted mass spectrometry. Of these, 23 were also measured in 131 plasma samples (117 independent subjects) using an untargeted platform and in an independent laboratory. Regression analysis with adjustment for clinical covariates, correction for false discovery rate (FDR), and meta-analysis were used to test associations between COPD subphenotypes and sphingolipids. Peripheral blood mononuclear cells were used to test associations between sphingolipid gene expression and plasma sphingolipids.
Results: Of the measured plasma sphingolipids, five sphingomyelins were associated with emphysema; four trihexosylceramides and three dihexosylceramides were associated with COPD exacerbations. Three sphingolipids were strongly associated with sphingolipid gene expression and fifteen sphingolipid gene/metabolite pairs were differentially regulated between COPD cases and controls.
3
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 4 of 64
Conclusions: There is evidence of systemic dysregulation of sphingolipid metabolism in patients with COPD. Subphenotyping suggests that sphingomyelins are strongly associated with emphysema and glycosphingolipids are associated with COPD exacerbations.
Total words in abstract (maximum 250: 224)
Keywords: Ceramides; Sphingomyelins; Emphysema; Metabolomics; Genomics.
4
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 5 of 64
Introduction Chronic obstructive pulmonary disease (COPD) is the third leading cause of death in the United States (1). Although smoking is the major risk factor for COPD, most smokers do not develop COPD and those that do have variable clinical phenotypes such as airflow obstruction, emphysema, chronic bronchitis, and frequent COPD exacerbations (2). There is little known about what predisposes certain smokers to develop a particular COPD phenotype. Recently, there have been reports suggesting that alteration in sphingolipid metabolism may be linked to disease susceptibility (3-6) although large-scale human studies of specific sphingolipids have been lacking. Sphingolipids contain a backbone of sphingoid bases and, depending on the O-linked R group, are sub classified into sphingosines, sphingomyelins, ceramides, or glycosphingolipids. In mammals, there are more than 100 species of sphingolipids and a multitude of enzymes that regulate sphingolipid metabolism. Sphingolipids are found primarily in cell membranes, but also in biofluids such as plasma. Having pleiotropic effects, specific sphingolipids have been implicated in diverse biologic processes (7) that include: cell death, proliferation, and differentiation, autophagy, senescence, migration, and efferocytosis. Sphingolipids have been implicated in several lung diseases, including COPD (3,8,9). The majority of sphingolipid lung research has focused on ceramides, the central intermediate metabolites in the sphingolipid metabolism and key pro-apoptotic signaling molecules, and on sphingosine-1-phosphate (S1P), a downstream metabolite of ceramide and key pro-survival and immune regulator. Smokers (10) and COPD patients (11) have increased ceramides in their lungs and ceramides have been implicated in tobacco smoke-induced pulmonary vascular cell apoptosis (12) and clearance of apoptotic cells by alveolar macrophages (13). S1P receptor 5 expression is reduced the lungs of COPD patients (14) and S1P analogues inhibit emphysema development in animal models (15). Recently, in a small cohort, sphingolipids were found elevated in the sputum in COPD (6). These findings suggest that sphingolipid metabolism is dysregulated by smoking and COPD and that these derangements may be pathogenic and could serve as biomarkers (10). However, previous studies 5
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
have been limited by small sample sizes, limited clinical phenotyping, and limited focus on a particular subclass of sphingolipids. To overcome these limitations and further investigate sphingolipids for their association with clinical COPD phenotypes such as emphysema, chronic bronchitis, and frequent exacerbations, we used two mass spectrometry based approaches to quantitate 69 plasma sphingolipids in more than 250 smokers with and without COPD from the COPDGene study. Furthermore, we integrated plasma metabolomics analyses with recently published microarray profiling of this cohort (4) to discover relationships that might elucidate mechanisms of sphingolipid dysregulation in COPD.
6
Copyright © 2014 by the American Thoracic Society
Page 6 of 64
Page 7 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Methods Study population The institutional review boards of participating institutions approved this study. All subjects were from the COPDGene cohort, which is a NIH-sponsored multi-centered study of the genetic epidemiology of COPD (16). From this cohort we selected a subset of 129 subjects (Table 1) for targeted sphingolipid analysis in order to obtain a wide range of COPD phenotypes (defined below). Additionally, 131 subjects who participated in a microarray study of gene expression in peripheral blood mononuclear cells (GEO accession number GSE 42057) (4) were selected for untargeted sphingolipid analysis. Additional details are provided in the online supplement.
Clinical data collection and definitions of COPD phenotypes. Each subject was classified by four different clinical phenotypes: airflow obstruction (COPD), chronic bronchitis, emphysema, and exacerbations (Supplemental Table 1). COPD was defined using GOLD criteria (17). Emphysema was measured using quantitative high resolution CT (HRCT) as described (18). Exacerbations were defined by acutely worse cough, sputum, and dyspnea in those with and without COPD. Only moderate exacerbations (treated by corticosteroids and/or antibiotics) or severe exacerbations (causing hospitalization) were counted. Chronic bronchitis was defined as cough that produces sputum daily for three consecutive months, for at least two consecutive years.
Sphingolipid measurements were performed independently in two separate laboratories using two separate protocols (see online methods for more details). A targeted, quantitative, mass spectrometry panel (Washington University) included 69 sphingolipids (Supplemental Table 2). Sphingomyelins, dihydrosphingomyelins, ceramides, and dihydroceramides were extracted using a modified Bligh-Dyer extraction method, in the presence of internal standards (IS). Sphingoid bases, ceramide-1-phosphate, monohexosylsphingosine, monohexosylceramides, dihexosylceramides, trihexosylceramides, monohydroxylated monohexosylceramides, 7
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 8 of 64
monohydroxylated dihexosylceramides, sulfatides, and gangliosides were extracted after protein precipitation with methanol, followed by supernatant collection, drying, and reconstituting with 1:1 methanol/water, in the presence of internal standards. A second untargeted protocol (National Jewish Health) was performed as detailed elsewhere (19).
Statistical analysis: Differences in demographic characteristics of study subjects were analyzed using a t-test for continuous variables and a Chi squared test for categorical variables. Regression modeling and covariates are described further in the online methods. Since the sphingolipid levels were highly correlated within class (Supplemental Figure 1), we also computed the first principal component (PC) of each sphingolipid classes (Supplemental Table 3 & 4) using prcomp function in R. Replication between the targeted and untargeted platforms was determined using the Stouffer-Liptak Zscore method, which converts the p-values from the two studies to normal quantiles and averages them to obtain a combined p-value (20,21). Each of the 23 sphingolipids that overlapped between the two studies was tested and consistency in the direction of the effect on the phenotype was taken into account.
8
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 9 of 64
Results Study subjects and baseline characteristics Demographics, physiology, quantitative HRCT measurements and patient-reported outcomes for each group are listed in Table 1 and Supplemental Table 1. Except for slightly more subjects with emphysema in untargeted cohort, there were no statistically significant differences in the baseline characteristics between the targeted and untargeted cohorts.
Targeted identification of plasma sphingolipids Our previous results suggested that sphingolipids were candidate biomarkers for COPD (4); we therefore performed targeted measurement of multiple sphingolipid classes. These included: sphingomyelins (SM d18:1), dihydrosphingomyelins (SM d18:0), ceramides (Cer d18:1), dihydroceramides (Cer d18:0), sphingoid bases, ceramide-1-phosphate, monohexosylsphingosine, monohexosylceramides, dihexosylceramides, trihexosylceramides, monohydroxylated monohexosylceramides, monohydroxylated dihexosylceramides, sulfatides, and gangliosides. After filtering out species that exhibited no or very low peaks, overlapped with other peaks, exhibited multiple peaks with retention times in close proximity, or had large coefficients of variance, 69 sphingolipid species were used for quantitative comparisons (Supplemental Table 2). Multiple sphingolipids were associated with clinical covariates such as age, gender, BMI and current smoking (Supplemental Table 3). Three sphingolipid species showed a negative correlation with age (correlation test p-value < .01), seven species showed a negative correlation with BMI (correlation p-value < .01), and 32 species showed higher levels in female subjects (t-test p-value < .01). Although not significant, select species showed trends for alterations depending on current smoking status. Because of these associations, we adjusted for covariates in all our models. Because strong correlations were noted within classes of sphingolipids (Supplementary Figure 1), sphingolipid species were further grouped into ten classes (Supplemental Table 4), which further reveled 9
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 10 of 64
robust interclass collinearities. For instance, plasma sphingomyelins were strongly correlated with ceramides (P < 0.001) and both sphingomyelins and ceramides were inversely correlated with S1P (P < 0.001). Because of the high intraclass correlations for plasma sphingolipids, each class of sphingolipids was also represented in analyses by the first principal component from the intraclass principal component analysis.
Plasma sphingolipids association with different phenotypes of COPD COPD patients exhibit heterogeneity in clinical presentation, with different morbidities and prognoses for each phenotype; therefore, we investigated if specific sphingolipids or sphingolipid classes are significantly associated with individual phenotypes. The most statistically significant negative (inverse) association between specific COPD phenotypes and sphingolipid classes was for emphysema and sphingomyelins; the most statistically significant positive (direct) association was between COPD exacerbations and trihexosylceramides (Figure 1).
Individual sphingolipids associated with emphysema We identified 26 sphingolipids that were significantly associated with emphysema after adjustment for covariates, with 8 remaining significant after adjustment for false discovery rate (FDR) (Supplemental Table 5). The most highly represented group was that of sphingomyelins (ten species associated with emphysema, including five after adjustment for false discovery), all of which had negative associations with emphysema. The three plasma sphingolipids most significantly associated with emphysema were ceramide d18.1.N16.0 (P = 0.0006), ganglioside GM3 d18.1.N16.0 (P = 0.0004), and sphingomyelin d18.1.N16.0 (P = 0.0009), all being inversely associated with emphysema (Figure 2), followed by monohexosylceramide d18.1.N16.0 (p=0.002; Supplemental Table 5). When stratified by subjects with less severe airflow obstruction (i.e. milder COPD), the receiver operating characteristic (ROC) curves demonstrated that these top sphingolipids improved the ability to diagnose moderate to severe emphysema beyond just clinical and physiologic covariates (Supplemental Table 6). 10
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 11 of 64
Individual sphingolipids associated with COPD exacerbations After adjustment for covariates and correction for FDR, we identified eleven sphingolipids that were associated with severe COPD exacerbations including four trihexosylceramides, three dihexosylceramides, sulfatide d18.1.N16.0, ganglioside GD1.d18.1.N16.0, sphigomyelin SM.d18.1.N14.1 and S1P (Supplemental Table 7). All associations were positive except for S1P and SM.d18.1.N14.1 and the strongest signal was the positive association of severe COPD exacerbations with plasma trihexosylceramides d18.1.N18.0 (adjusted p = 1.6 X 10-6). Most of these associations were also identified with COPD exacerbations of lesser severity (Supplemental Table 8). ROC analysis demonstrates that these sphingolipids improved the ability to diagnose severe exacerbations beyond just clinical and physiologic covariates (Supplemental Table 9). Interestingly, for less severe exacerbations, only in subjects with less airflow obstruction, Cer.d18.1.N18.0 showed enhanced association with exacerbations beyond just clinical and physiologic covariates (Figure 3 and Supplemental Table 10).
Sphingolipids associated with other COPD phenotypes (airflow obstruction and chronic bronchitis) There were no plasma sphingolipid species significantly associated with chronic bronchitis (Supplementary Table 11), FEV1 (data not shown), FEV1/FVC (Supplementary Table 11), or bronchodilator responsiveness (data not shown) after correcting for FDR; however, several metabolomic trends noted in these subgroups were used in combination with genomic data for phenotype-metabolomic interpretations.
Replication of associations between COPD phenotypes and sphingolipids For replication, we utilized an independently generated metabolomics dataset (detected via untargeted mass spectrometry using orthogonal methods in an independent laboratory) from plasma of 131 subjects in COPDGene. Between the two datasets, we identified 23 overlapping sphingolipids (eight sphingomyelins, six ceramides, three gangliosides, two sulfogalactosylceramides, dihydroceramide, trihexosylceramide, S1P, and 11
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 12 of 64
sphingosine). For overlapping sphingolipids, we performed meta-analysis of the two datasets accounting for the direction of association. Seven of 23 sphingolipids showed negative associations with emphysema, of which five sphingomyelins (Table 2) exhibited the strongest signal for replication between the two platforms (FDR < 0.10). Four sphingolipids demonstrated associations with moderate and severe exacerbations (Table 3). Except for S1P, all were positively associated with exacerbations (FDR < 0.10).
Relationship between plasma sphingolipids and gene expression in smokers with and without COPD All subjects in the untargeted metabolomics data set had microarray data generated from peripheral blood mononuclear cells collected from the same plasma analyzed for metabolomics. Thus, we were able to explore relationships between plasma sphingolipid metabolites and potentially related sphingolipid pathway genes (Supplemental Table 13). The most significant correlations were between N-(tetradecanoyl)-sphing-4enine-1-(2-aminoethylphosphonate) and GM2 ganglioside activator (ρ = -0.39; P = 0.00001; FDR = 0.04), between SM(d18:1/12:0) and lysosomal sialidase (ρ = -0.36; P = 0.00007; FDR = 0.12) (Supplemental Figure 2), and between sphingosine and Ceroid-Lipofuscinosis, Neuronal 8 (ρ = -0.34; P = 0.0002; FDR = 0.03). Following interrogation of correlations between sphingolipid metabolite and gene expression and clinical phenotype, subjects who reported moderate or severe COPD exacerbations were most likely to have discordant correlations for gene-metabolite pairs (Table 4). For example, for the gene-metabolite pair C9orf47 and NeuGcalpha2-8NeuGcalpha2-3Galbeta1-4GlcNAcbeta1-3Galbeta1-4Glcbeta-Cer(d18:1/24:1(15Z)) had correlations -0.19 and 0.64 in subjects without or with severe COPD exacerbations, respectively (P = 0.00001; FDR = 0.01) (Figure 4).
12
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 13 of 64
Discussion This study of more than 250 subjects from the COPDGene cohort is the largest and the only combined targeted and nontargeted metabolomic study of plasma sphingolipid species in smokers with and without COPD; it is the first to link specific sphingolipids with distinct COPD phenotypes. We identified novel sphingolipid subclasses such as gangliosides and sphingomyelins as potential biomarkers inversely associated with severe emphysema and trihysoxylceramide as a potential biomarker positively associated with frequent COPD exacerbations. All these associations were noted even after adjusting for covariates of age, smoking status, airflow obstruction, gender and BMI; most were replicated in an independent cohort using an orthogonal mass spectrometry approach in a different laboratory. Sphingomyelins are integral components of the plasma membrane, typically enriched in lipid microdomains of caveolae and clathrin-coated pits (23). Sphingomyelins are produced from ceramides via sphingomyelin synthase, or they can generate ceramides through hydrolysis by sphingomyelinases (Figure 5). In addition to rapidly generating ceramides as second messengers, sphingomyelins may themselves play a role in cell migration, apoptosis, autophagy, and cell survival/proliferation (23). We identified ten specific sphingomyelins (e.g. SM.d18.1.N16.0) that were inversely associated with emphysema, whereas others (e.g. SM.d18.1.N18.0) were not. These species-specific effects could only be studied with mass spectrometry; however, we also noted strong correlations among the whole subclass of sphingomyelins. A recent report (5) found no statistically significant association between HRCT measured-emphysema severity at baseline and sphingomyelin, measured with less accurate spectrophotometric assays (24). However, low baseline plasma sphingomyelin was associated with worse COPD (lower FEV1), whereas high levels were associated with more rapid progression of emphysema. Our data support the finding that worse COPD is associated with lower plasma sphingomyelins (5), but the apparent paradoxical observation of higher plasma sphingomyelins with progression of emphysema cannot be validated by our experimental design. Such observation might be explained if low sphingomyelin levels reflect low lung mass and progression of emphysema occurs in lungs 13
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 14 of 64
that are not yet severely emphysematous. This could be explored by careful kinetics of specific sphingomyelins accompanied by concomitant HRCT measurements of the lung. Plasma ceramides may be generated by activated plasma sphingomyelinases (25) (Figure 5), and could signal outside-in (26) and/or be further metabolized to sphingosine (by plasma ceramidases). To the extent that plasma levels reflect cellular metabolism, the decrease in both sphingomyelins and ceramides with subsequent increases in ceramide metabolites suggests that low sphingomyelin levels are not due to a blockage in sphingomyelin synthase (which would result in ceramide accumulation), but due to increased catabolism coupled with decreased availability of its building block precursor, ceramide (Figure 5). In such conditions, supplementation of sphingomyelin may fuel ceramide production and catabolism, by a “revved up” system. Plasma sphingomyelins might also reflect the burden or quality of circulating (shed) cellular plasma membranes, such as apoptotic microparticles or exosomes. Clearly, these unbiased metabolomic results raise important mechanistic questions that can be explored in models of COPD. Ceramides (especially Cer.d18.1.N16.0), gangliosides (especially GM3.d18.1.N16.0) and monohexosylceramides (especially monohexcer.d18.1.N16.0) also showed an inverse correlation with emphysema. Although ceramides have been known to play a role in the pathogenesis of emphysema for several years (10,27), the involvement of gangliosides and monheoxylceramides has not been described. An upregulation of ceramide production has been typically associated with models of emphysema, by causing endothelial and epithelial cell apoptosis (3) or by inhibiting alveolar macrophage clearance of apoptotic cells (13). In this cross-sectional sampling, the inverse correlation of plasma ceramides with emphysema appears surprising. However, if our global metabolomics assessment of plasma sphingolipids reflects pulmonary cellular events, worse emphysema may be linked to increased activity of both sphingomyelinase and recycling (from gangliosides) pathways, followed by ceramide consumption/degradation, rather than accumulation (Figure 5), possibly via plasma ceramidases. Since plasma S1P was not increased and tended to be negatively correlated with emphysema and COPD exacerbations, unlike sphingosine, it is possible that the sphingolipid metabolism is accelerated towards ceramide production and degradation, but inhibited at the level of 14
Copyright © 2014 by the American Thoracic Society
Page 15 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
sphingosine kinase activity (Figure 5). The reduced production of S1P may be associated with decreased cellular survival and proliferation and increased apoptosis of alveolar cells (15). Much less is known about the significance of decreased gangliosides (or glycoceramides) in the lung, reports showing association with diabetes or hearing loss (28). While our earlier genomic analysis hinted at this involvement (4), this is the first report of decreased monosialodihexosylganglioside GM3.d18.1.N16.0 as putative biomarker of emphysema. Similarly, the role of monohexosylceramides in the lung is unknown, but they were shown to bind to surfactant protein D (29) and to decrease upon activation of acid sphingomyelinase and acid β-glucosidase1, as a sign of ceramide production via recycling from gangliosides during inflammation (38). These results may embolden future investigations of lung sphingolipid metabolomic signatures and the functional role of the gangliosides and monohexosylceramides in COPD. The strongest positive association with COPD exacerbations was for trihexosylceramides (trihexcer.d18.1.N22.0 and trihexcer.d18.1.N24.0), a lipid class with unknown lung effects; one case report of Fabry’s disease (characterized by overabundance of trihexosylceramides) showed association with COPD (30). Other sphingolipids such as ceramide (d18.1.N18.0) and S1P were inversely associated with COPD exacerbations. This, along with negative correlations of plasma sphingomyelins and ceramides with S1P, suggest a rapid flux of sphingosine-to-ceramide-to glycosylated ceramides metabolism, with potential disruption of ceramide-to-S1P metabolism (Figure 5) in those with COPD exacerbations. Alternatively, whether tissues or cells may differentially contribute to the abundance of plasma metabolites in these individuals could be explored by comparing plasma with lung tissue metabolomics signatures. Since these samples were collected at a time remote from an actual exacerbation, metabolomic signatures are unlikely to reflect corticosteroid or antibiotic treatment, or acute infections. Rather they may indicate specific changes associated with developing COPD exacerbations. For example, shunting ceramide metabolism towards trihexosylceramides may render the lungs more susceptible to inflammation, due to defects in lymphocyte function, as learned from a mouse model of Fabry’s disease (39).
15
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 16 of 64
To further translate biomarker findings to disease pathogenesis, we explored alterations of genemetabolite associations with disease phenotypes. We identified a negative association of sphingosine levels with the expression of CYR61 only in subjects with severe COPD. Although the significance of this finding remains to be tested, the extracellular matrix protein CCN1/Cyr61 is activated by S1P via receptor 2 (31), which also activates Rho Kinase to cause lung damage during cigarette smoke exposure (32,33). Other disrupted gene-metabolite associations of interest in those with severe COPD exacerbations were between ceramide and alkaline ceramidase, which degrades unsaturated long chain ceramides (34); ORMDL1, which regulates de novo ceramide biosynthesis, recently involved in asthma (35); and 3-ketodihydrosphingosine reductase, which reduces 3-ketodihydrosphingosine to dihydrosphingosine (36). Another intriguing association was that between circulating ceramide and COL4A3BP, also known as CERTL, a gene encoding for ceramide transporter protein Goodpasture Antigen-Binding Protein, which mediates the non-vesicular intracellular transport of ceramides (37). Superimposing potential sphingolipid biomarkers with genetic signatures helped map the directionality of sphingolipid metabolism in individuals with COPD exacerbations (Figure 5). Using similar approaches may strengthen the associations of sphingolipid metabolism and chronic bronchitis or obstructive ventilatory defects in COPD, which were weak at the plasma metabolite levels, but in conjunction with genomic findings, suggested a potential metabolic flux towards higher S1P and de novo pathway activation, respectively (Supplementary Figure 3). In summary, this study demonstrates that specific sphingolipids such ceramides, sphingomyelins, and gangliosides may be biomarker candidates of COPD phenotypes such emphysema and frequent COPD exacerbations. Our study’s strengths include the large number of highly phenotyped subjects, the multiple specific sphingolipids measurements, and independent replication. Limitations of our study include lack of healthy nonsmoking subjects, of longitudinal studies, and of lung tissue, bronchoalveolar lavage fluid, or sputum analyses. The latter would have allowed us to probe if plasma findings parallel those in the lung and to validate a small study that suggested sputum sphingolipids are associated with COPD (6). Our results in a much larger cohort of COPD patients corroborate those reported in the sputum, in particular that there are 16
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 17 of 64
abnormalities in sphingolipid metabolism and glycosphingolipid changes in COPD. The complementary nature of these studies conducted in different compartments may help interpret our results in plasma, which indicate an overall accelerated metabolism/catabolism of ceramide at systemic level, in contrast to the increased ceramides and dihydroceramidees recovered in the sputum in COPD. This may be due to active transfer and utilization of plasma sphingolipids in the lung with release of ceramides in sputum; alternatively it may be due to different dynamics of sphingolipid metabolism in circulating cells and endothelium compared to airway macrophages or epithelial cells that may contribute to sputum sphingolipid content. Our results strengthen the evidence implicating ceramide in emphysema and suggest a previously unsuspected role for ceramide metabolism in COPD exacerbations. Furthermore, our data suggest that additional subclasses of sphingolipids such as sphingomyelins, gangliosides, monohexosylceramides, and trihexosylceramides should be investigated, to validate their role as biomarkers and as potential contributors to COPD pathogenesis.
17
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 18 of 64
Acknowledgments: We thank Christina Schnell, Taylor Thorn, Cori Fratelli for assistance with recruiting subjects and preparing samples and Kristina Watson for administrative support.
18
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 19 of 64
References 1.
2. 3. 4.
5.
6.
7. 8. 9. 10.
11.
12.
13.
14.
15.
16.
17.
Murphy, S.L., Xu, J. and Kochanek, K.D. (2013) Deaths: final data for 2010. National vital statistics reports : from the Centers for Disease Control and Prevention, National Center for Health Statistics, National Vital Statistics System, 61, 1-117. Friedlander, A.L., Lynch, D., Dyar, L.A. and Bowler, R.P. (2007) Phenotypes of chronic obstructive pulmonary disease. Copd, 4, 355-384. Petrache, I., Petrusca, D.N., Bowler, R.P. and Kamocki, K. (2011) Involvement of ceramide in cell death responses in the pulmonary circulation. Proc Am Thorac Soc, 8, 492-496. Bahr, T.M., Hughes, G.J., Armstrong, M., Reisdorph, R., Coldren, C.D., Edwards, M.G., Schnell, C., Kedl, R., LaFlamme, D.J., Reisdorph, N. et al. (2013) Peripheral blood mononuclear cell gene expression in chronic obstructive pulmonary disease. Am J Respir Cell Mol Biol, 49, 316-323. Ahmed, F.S., Jiang, X.C., Schwartz, J.E., Hoffman, E.A., Yeboah, J., Shea, S., Burkart, K.M. and Barr, R.G. (2014) Plasma sphingomyelin and longitudinal change in percent emphysema on CT. The MESA lung study. Biomarkers, 19, 207-213. Telenga, E.D., Hoffmann, R.F., Ruben, t.K., Hoonhorst, S.J., Willemse, B.W., van Oosterhout, A.J., Heijink, I.H., van den Berge, M., Jorge, L., Sandra, P. et al. (2014) Untargeted lipidomic analysis in chronic obstructive pulmonary disease. Uncovering sphingolipids. Am J Respir Crit Care Med, 190, 155-164. Hannun, Y.A. and Obeid, L.M. (2008) Principles of bioactive lipid signalling: lessons from sphingolipids. Nature reviews. Molecular cell biology, 9, 139-150. Uhlig, S. and Gulbins, E. (2008) Sphingolipids in the lungs. Am J Respir Crit Care Med, 178, 11001114. Yang, Y. and Uhlig, S. (2011) The role of sphingolipids in respiratory disease. Ther Adv Respir Dis, 5, 325-344. Petrache, I., Natarajan, V., Zhen, L., Medler, T.R., Richter, A.T., Cho, C., Hubbard, W.C., Berdyshev, E.V. and Tuder, R.M. (2005) Ceramide upregulation causes pulmonary cell apoptosis and emphysema-like disease in mice. Nat Med, 11, 491-498. Scarpa, M.C., Baraldo, S., Marian, E., Turato, G., Calabrese, F., Saetta, M. and Maestrelli, P. (2013) Ceramide expression and cell homeostasis in chronic obstructive pulmonary disease. Respiration, 85, 342-349. Petrusca, D.N., Van Demark, M., Gu, Y., Justice, M.J., Rogozea, A., Hubbard, W.C. and Petrache, I. (2014) Smoking exposure induces human lung endothelial cell adaptation to apoptotic stress. Am J Respir Cell Mol Biol, 50, 513-525. Petrusca, D.N., Gu, Y., Adamowicz, J.J., Rush, N.I., Hubbard, W.C., Smith, P.A., Berdyshev, E.V., Birukov, K.G., Lee, C.H., Tuder, R.M. et al. (2010) Sphingolipid-mediated inhibition of apoptotic cell clearance by alveolar macrophages. J Biol Chem, 285, 40322-40332. Cordts, F., Pitson, S., Tabeling, C., Gibbins, I., Moffat, D.F., Jersmann, H., Hodge, S. and Haberberger, R.V. (2011) Expression profile of the sphingosine kinase signalling system in the lung of patients with chronic obstructive pulmonary disease. Life Sci, 89, 806-811. Diab, K.J., Adamowicz, J.J., Kamocki, K., Rush, N.I., Garrison, J., Gu, Y., Schweitzer, K.S., Skobeleva, A., Rajashekhar, G., Hubbard, W.C. et al. (2010) Stimulation of sphingosine 1-phosphate signaling as an alveolar cell survival strategy in emphysema. Am J Respir Crit Care Med, 181, 344-352. Regan, E.A., Hokanson, J.E., Murphy, J.R., Make, B., Lynch, D.A., Beaty, T.H., Curran-Everett, D., Silverman, E.K. and Crapo, J.D. (2010) Genetic epidemiology of COPD (COPDGene) study design. COPD, 7, 32-43. Pauwels, R.A., Buist, A.S., Ma, P., Jenkins, C.R. and Hurd, S.S. (2001) Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: National 19
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
18.
19.
20. 21. 22.
23.
24. 25.
26.
27. 28. 29.
30.
31.
32.
33.
34.
Page 20 of 64
Heart, Lung, and Blood Institute and World Health Organization Global Initiative for Chronic Obstructive Lung Disease (GOLD): executive summary. Respiratory care, 46, 798-825. Carolan, B.J., Kim, Y.I., Williams, A.A., Kechris, K., Lutz, S., Reisdorph, N. and Bowler, R.P. (2013) The association of adiponectin with computed tomography phenotypes in chronic obstructive pulmonary disease. Am J Respir Crit Care Med, 188, 561-566. Yang, Y., Cruickshank, C., Armstrong, M., Mahaffey, S., Reisdorph, R. and Reisdorph, N. (2013) New sample preparation approach for mass spectrometry-based profiling of plasma results in improved coverage of metabolome. Journal of chromatography. A, 1300, 217-226. Stouffer, S.A. (1949) The American soldier. Princeton University Press, Princeton,. Liptak, T. (1958) On the combination of independent tests. Magyar Tud. Akad. Mat. Kutato Int. Kozl, 3. Abdul-Salam, V.B., Wharton, J., Cupitt, J., Berryman, M., Edwards, R.J. and Wilkins, M.R. (2010) Proteomic analysis of lung tissues from patients with pulmonary arterial hypertension. Circulation, 122, 2058-2067. Taniguchi, M. and Okazaki, T. (2014) The role of sphingomyelin and sphingomyelin synthases in cell death, proliferation and migration-from cell and animal models to human disorders. Biochim Biophys Acta, 1841, 692-703. Hojjati, M.R. and Jiang, X.C. (2006) Rapid, specific, and sensitive measurements of plasma sphingomyelin and phosphatidylcholine. J Lipid Res, 47, 673-676. Levy, M., Khan, E., Careaga, M. and Goldkorn, T. (2009) Neutral sphingomyelinase 2 is activated by cigarette smoke to augment ceramide-induced apoptosis in lung cell death. Am J Physiol Lung Cell Mol Physiol, 297, L125-133. Medler, T.R., Petrusca, D.N., Lee, P.J., Hubbard, W.C., Berdyshev, E.V., Skirball, J., Kamocki, K., Schuchman, E., Tuder, R.M. and Petrache, I. (2008) Apoptotic sphingolipid signaling by ceramides in lung endothelial cells. Am J Respir Cell Mol Biol, 38, 639-646. Petrache, I. and Petrusca, D. (2013), Handbook of experimental pharmacology. Springer-Verlag GmbH, Vol. 216. Inokuchi, J. (2011) Physiopathological function of hematoside (GM3 ganglioside). Proceedings of the Japan Academy. Series B, Physical and biological sciences, 87, 179-198. Kuroki, Y., Gasa, S., Ogasawara, Y., Shiratori, M., Makita, A. and Akino, T. (1992) Binding specificity of lung surfactant protein SP-D for glucosylceramide. Biochem Biophys Res Commun, 187, 963969. Wang, R.Y., Abe, J.T., Cohen, A.H. and Wilcox, W.R. (2008) Enzyme replacement therapy stabilizes obstructive pulmonary Fabry disease associated with respiratory globotriaosylceramide storage. J Inherit Metab Dis, 31 Suppl 2, S369-374. Young, N. and Van Brocklyn, J.R. (2007) Roles of sphingosine-1-phosphate (S1P) receptors in malignant behavior of glioma cells. Differential effects of S1P2 on cell migration and invasiveness. Exp Cell Res, 313, 1615-1627. Richens, T.R., Linderman, D.J., Horstmann, S.A., Lambert, C., Xiao, Y.Q., Keith, R.L., Boe, D.M., Morimoto, K., Bowler, R.P., Day, B.J. et al. (2009) Cigarette smoke impairs clearance of apoptotic cells through oxidant-dependent activation of RhoA. Am J Respir Crit Care Med, 179, 1011-1021. Schweitzer, K.S., Hatoum, H., Brown, M.B., Gupta, M., Justice, M.J., Beteck, B., Van Demark, M., Gu, Y., Presson, R.G., Jr., Hubbard, W.C. et al. (2011) Mechanisms of lung endothelial barrier disruption induced by cigarette smoke: role of oxidative stress and ceramides. Am J Physiol Lung Cell Mol Physiol, 301, L836-846. Hu, W., Xu, R., Sun, W., Szulc, Z.M., Bielawski, J., Obeid, L.M. and Mao, C. (2010) Alkaline ceramidase 3 (ACER3) hydrolyzes unsaturated long-chain ceramides, and its down-regulation inhibits both cell proliferation and apoptosis. J Biol Chem, 285, 7964-7976. 20
Copyright © 2014 by the American Thoracic Society
Page 21 of 64
35.
36.
37.
38.
39.
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Michel, S., Busato, F., Genuneit, J., Pekkanen, J., Dalphin, J.C., Riedler, J., Mazaleyrat, N., Weber, J., Karvonen, A.M., Hirvonen, M.R. et al. (2013) Farm exposure and time trends in early childhood may influence DNA methylation in genes related to asthma and allergy. Allergy, 68, 355-364. Kihara, A. and Igarashi, Y. (2004) FVT-1 is a mammalian 3-ketodihydrosphingosine reductase with an active site that faces the cytosolic side of the endoplasmic reticulum membrane. J Biol Chem, 279, 49243-49250. Kudo, N., Kumagai, K., Tomishige, N., Yamaji, T., Wakatsuki, S., Nishijima, M., Hanada, K. and Kato, R. (2008) Structural basis for specific lipid recognition by CERT responsible for nonvesicular trafficking of ceramide. Proc Natl Acad Sci U S A, 105, 488-493. Kitatani, K., Sheldon, K., Rajagopalan, V., Anelli, V., Jenkins, R.W., Sun, Y., Grabowski, G.A., Obeid, L.M. and Hannun, Y.A. (2009) Involvement of acid beta-glucosidase 1 in the salvage pathway of ceramide formation. J Biol Chem, 284, 12972-12978. Pereira, C.S., Azevedo, O., Maia, M.L., Dias, A.F., Sa-Miranda, C. and Macedo, M.F. (2013) Invariant natural killer T cells are phenotypically and functionally altered in Fabry disease. Molecular genetics and metabolism, 108, 241-248.
21
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 22 of 64
Table 1: Characteristics of Subjects.
Age Gender (% Male) Current Smokers† (%) Pack Years Body Mass Index FEV1% FEV1/FVC Emphysema (%) Chronic Bronchitis (%) Exacerbations/year (moderate) Exacerbations/year (severe)
Targeted Cohort (N = 129) 63 (58 - 71) 57 23 46 (34 - 58) 27 (24 – 31) 64 (40 – 92) 0.56 (0.39 – 0.72) 11.5 ± 11.4 19% 0.7 ± 1.3 0.2 ± 0.7
Untargeted Cohort (N = 131) 64 (57 - 70) 56 23 42 (30 - 64) 27 (24 - 32) 66 (42 - 89) 0.61 (0.44 – 0.74) 9.18 ± 10.4 18% 0.6 ± 1 0.1 ± 0.3
P-value 0.96 0.79 0.95 0.71 0.58 0.8 0.15 0.09 0.95 0.3 0.17
Median (interquartile ratio) or mean ± standard deviation: Emphysema was defined as percent of lung attenuation voxels below -950 HU; chronic bronchitis was defined by daily productive cough for at least 3 months in the previous 2 consecutive years; exacerbations were defined as moderate (treated with either antibiotics or corticosteroids) or severe (leading to hospitalization) in the previous year.
22
Copyright © 2014 by the American Thoracic Society
Page 23 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Table 2: Meta-analysis of sphingolipids associated with emphysema
Sphingolipid Ganglioside GM3 (d18:1/16:0) sphingomyelin(d18:0/24:1(15Z)) sphingomyelin(d18:1/14:0) sphingomyelin(d18:1/16:0) sphingomyelin(d18:1/16:1) sphingomyelin(d18:1/24:1(15Z)) sphingomyelin(d18:2/14:0) * FDR < 0.10
Targeted Cohort ß slope
P-value
Untargete d Cohort ß slope
-0.7654 -0.4534 -0.4335 -0.8964 -0.5198 -0.6062 -0.3520
0.0004 0.0057 0.0040 0.0010 0.0054 0.0009 0.0237
-0.1277 -0.0623 -0.1070 -0.1867 -0.1459 -0.2182 -0.0673
23
Copyright © 2014 by the American Thoracic Society
Meta-analysis Pvalue 0.3768 0.5349 0.3650 0.4258 0.3621 0.0829 0.5444
P-value 0.0018* 0.0167* 0.0075* 0.0037* 0.0090* 0.0004* 0.0425
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 24 of 64
Table 3: Meta-analysis of sphingolipids associated with exacerbations Targeted Cohort Sphingolipid Sphingosine 1-phosphate Trihexosylceramide (d18:1/16:0) 3-O-Sulfogalactosylceramide (d18:1/16:0) Galabiosylceramide (d18:1/24:1(15Z)) * FDR < 0.10
ß slope -0.7051 0.8804 0.8262 0.6077
P 0.1838 0.0693 0.0507 0.1741
Untargeted Cohort ß slope -0.6928 0.9839 0.5191 0.4177
24
Copyright © 2014 by the American Thoracic Society
P 0.0004 0.0059 0.0792 0.0641
Meta-Analysis p-value 0.0006* 0.0012* 0.0087* 0.0232
Page 25 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Table 4: Differential Metabolite-Gene Expression Correlations for Sphingolipids by COPD Phenotype ρ without phenotype
ρ with Phenotype
p-value
FDR
-0.17
0.50
0.00005
0.06
-0.15
0.54
0.00001
0.04
-0.05
0.63
0.00006
0.05
C9orf47
-0.19
0.64
0.00001
0.01
Cer d18:1 N16:0 *1-3Galalpha1-3Galalpha1-3Galalpha1-4Galbeta14Glcbeta-Cer(d18:1/24:1(15Z))†
ACER3
0.11
0.73
0.00001
0.01
C9orf47
-0.23
0.56
0.00009
0.06
Cer d18:1 N16:0 Cer(d18:1/24:1(15Z))† Cer(d16:1/23:0)†
VAPA COL4A3BP COL4A3BP
-0.03
0.62
0.00014
0.08
-0.07
0.60
0.00019
0.08
-0.12
0.56
0.00040
0.15
Cer d18:1 N16:0
KDSR
-0.10
0.55
0.00055
0.15
Cer d18:1 N16:0 SM(d18:0/24:1(15Z))† 1st PC ceramides
ORMDL1 CYR61 COL4A3BP
-0.14
0.54
0.00051
0.15
-0.18
0.53
0.00048
0.15
0.09
0.54
0.00242
0.16
1st PC sphingosines
ST8SIA3
Emphysema (upper lobe)
-0.47
0.19
0.00039
0.06
1st PC sphingosines
CYR61
Airflow obstruction
0.28
-0.39
0.00011
0.02
Sphingolipid N-(tetradecanoyl)-sphing-4-enine-1-(2aminoethylphosphonate)† SM(d18:1/16:1) CerP(d18:1/24:1(15Z)) † NeuGcalpha2-3Galbeta1-4GlcNAcbeta1-3Galbeta14Glcbeta-Cer(d18:1/24:1(15Z))†
†
Gene
COPD Subphenotype Moderate or severe exacerbation
KDSR GM2A ST8SIA2
Severe exacerbation
Discovered in the untargeted profiling.
25
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 26 of 64
Figure legends
Figure 1. Heat map showing sphingolipids associated with specific COPD phenotypes. Each class was represented by the first principle component. Emphysema was defined as percent of lung attenuation voxels below -950 HU; chronic bronchitis was defined by daily productive cough for at least 3 months in the previous 2 consecutive years; exacerbations (severe) were the number of hospital admissions for a COPD admission in the previous year. Each phenotype was modeled using regression with covariates (age, gender, smoking status, body mass index (BMI), and airflow obstruction (FEV1) where appropriate), detailed in the online methods. The color legend shows shading based on the signed -10 log of the P value with negative (-) associations shown in red and positive (+) associations shown in green for each sphingolipid class. Non-statistically significant associations are shown in orange and yellow.
Figure 2: Plasma ceramides are inversely associated with emphysema severity. Ceramide d18:1/N16:0), sphingomyelin d18:1/N24:1), and sphingomyelin (d18:1/N16:0) are representative examples (FDR < 0.02 in Supplemental Table 5). Shown are box plots of median, quartiles, and outliers. 33%, 91%, 97%, and 100% of subjects with no, mild, moderate, or severe emphysema had airflow obstruction (post-bronchodilator FEV1/FVC < 0.7).
Figure 3: Receiver operating curve analysis and area under the curve (AUC) for sphingolipids significantly associated with moderate COPD exacerbations. The first model uses only the demographic and clinical covariates (Covariates only) including covariates age, gender, smoking status, body mass index (BMI), and airflow obstruction (FEV1). The second model includes the same 26
Copyright © 2014 by the American Thoracic Society
Page 27 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
covariates and adds plasma levels of Cer.d18.1.N18.0, which was the sphingolipid that most improved ROC characteristics.
Figure 4. Disruption of sphingolipid metabolite-gene correlations by COPD phenotype. Scatterplot of NeuGcalpha2-3Galbeta1-4GlcNAcbeta1-3Galbeta1-4Glcbeta-Cer(d18:1/24:1(15Z)) and C9orf47 gene expression. Subjects without severe COPD exacerbations had a correlation of -0.19 whereas subjects with a history severe COPD exacerbations had a correlation of 0.64 (P = 0.00001; FDR = 0.01).
Figure 5: Sphingolipids metabolo-genomic pathway patterns associated with COPD subphenotypes. A. Detailed sphingolipid metabolic pathway with the major ceramide synthetic pathways noted as sphingomyelinase-, de novo-, and recycling pathways (abbreviations list in the text). Individual or classes of metabolites measured in the study are identified as precursors and/or products relative to synthetic or catabolic reactions with directionality marked by black arrows. Catalytic enzymes are noted next to the middle of the reactions arrow. Circled are enzymes whose genes were significantly related with COPD phenotypes in genomic analyses. In color are classes (arrowheads) or species (arrows) of metabolites that were associated with respective (color-coded) COPD phenotypes. Solid arrows indicate: 16-18 carbon chain-; dotted arrows 24-26 carbon chain-; single dash arrows: 14 carbon chain- sphingolipids. The direction of the arrowhead or arrow marks the direction of association (i.e. up: increased; down: decreased). Closed headed arrows represent changes with significance levels P < 0.05. B-C. Superimposed on the sphingolipid metabolic pathways are deducted directions of sphingolipid metabolic flux, based on the results of the metabolo-genomic study, by disease phenotype, as noted.
27
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Copyright © 2014 by the American Thoracic Society
Page 28 of 64
Page 29 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Copyright © 2014 by the American Thoracic Society
Page 30 of 64
0.6 0.4 0.2
Covariates Only, AUC = 0.8521 Cer.d18.1.N18.0, AUC = 0.935
0.0
True Positive Rate
0.8
1.0
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
0.0
0.2
0.4
0.6
False Positive Rate
Copyright © 2014 by the American Thoracic Society
0.8
1.0
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
4.5 4.0 3.5
gene expression
5.0
Page 31 of 64
control exacerbation 16.0
16.5
17.0
17.5
18.0
metabolite abundance
Copyright © 2014 by the American Thoracic Society
18.5
19.0
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Copyright © 2014 by the American Thoracic Society
Page 32 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 33 of 64
Online Methods Methods Study population The institutional review boards of participating institutions approved this study. All subjects were from the COPDGene cohort, which is a NIH-sponsored multi-centered study of the genetic epidemiology of COPD (1). The study enrolled 10,192 non-Hispanic White and Black individuals, aged 45-80 years old with at least a 10 pack-year history of smoking, who had not had an exacerbation of COPD for at least the previous 30 days. Additional information on the COPDGene study and the collection of clinical data has been described previously (1). 20% of COPDGene subjects had fresh frozen plasma collected using a p100 tube (BD) at two COPDGene sites (National Jewish Health and University of Iowa). From this cohort we selected a subset of 129 subjects (Table 1) for targeted sphingolipid analysis in order to obtain a wide range of COPD phenotypes (defined below). An additional 131 independent subjects were selected for untargeted sphingolipid analysis based on their participation in a microarray study of gene expression in peripheral blood mononuclear cells (GEO accession number GSE 42057) (2). Fourteen subjects were studied on both platforms.
Clinical data collection and definitions of COPD phenotypes Each subject was classified for 4 different clinical phenotypes: airflow obstruction (COPD), chronic bronchitis, emphysema, and exacerbations (Supplemental Table 1). COPD was defined as post bronchodilator ratio of forced expiratory volume in the first second (FEV1) to forced vital capacity (FVC) < 0.70. Current or ex-smokers without spirometric evidence of airflow obstruction (FEV1/FVC ≥ 0.70) were classified as controls (3). Emphysema was measured using quantitative high resolution CT as previously described (4). In brief, emphysema was quantified as the percent of lung attenuation lung voxels (LAA) ≤-950 HU on the inspiratory images for the whole lung. Emphysema was further categorized as none (LAA ≤ 5%), mild (5% < LAA ≤ 10%), moderate (10% < LAA ≤ 20%), or severe (20% < LAA). Exacerbations (flare ups) of COPD or bronchitis were defined by acutely worse cough, sputum, and dyspnea. Subjects could report 0,1,2,3,4,5, or ≥ 6 exacerbations year prior to enrollment. Only moderate exacerbations (treated by corticosteroids and/or
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 34 of 64
antibiotics) or severe exacerbations (causing an admission to the hospital) were counted. Chronic Bronchitis was defined as having a cough that produces sputum that is present daily for 3 months in a row, with those periods occurring at least 2 consecutive years.
Targeted sphingolipid measurements in plasma Sphingomyelins, dihydrosphingomyelins, ceramides and dihydroceramides A modified Bligh-Dyer extraction method was used in the presence of internal standards (IS). Sample analysis was performed in four batches using a Shimadzu 10A HPLC system coupled to a TSQ Quantum Ultra triple quadrupole mass spectrometer operated in SRM mode under ESI(+). A Thermo Betasil C18 column was employed to achieve the optimal sensitivity and separation. Data processing was conducted with Xcalibur software (Thermo) to obtain the peak area ratios of analytes to the corresponding internal standards (Supplemental Table 2). A pooled lipid extract from study samples was used as a quality control (QC) sample, and was injected between every 6 study samples to verify the instrument consistency. The peak area ratios were then normalized to the average of QC peak area ratios in the same batch, so that the results for different batches were brought to the same baseline.
Sphingoid bases, ceramide-1-phosphate, monohexosylsphingosine, monohexosylceramides, dihexosylceramides, trihexosylceramides, monohydroxylated monohexosylceramides, monohydroxylated dihexosylceramides, sulfatides, and gangliosides Plasma protein was precipitated with methanol, followed by supernatant collection, drying, and reconstituting with 1:1 methanol/water in the presence of internal standards. Duplicate sample analysis (1st analysis and 2nd analysis) was performed in four batches with an online trapping API-4000 system, (consists of a API-4000 mass spectrometer, a Leap PAL autosampler, an Agilent 1100 HPLC, and a Shimadzu HPLC) operated in multiple reaction monitoring (MRM) mode under ESI(+) for all species except sulfatides and gangliosides ESI(-). An Agilent Eclipse XRB-C18 column was used for sphingoid bases and ceramide-1phosphate and a Varian Metasil C18 column was used for analysis of monohexosylsphingosine, mono-, di-, tri-
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 35 of 64
hexosylceramides as well as their monohydroxylated ceramides. An Agilent Zobax Eclipse Plus C18 column was used for analysis of sulfatides and gangliosides. A Thermo Betasil C18 trapping column was also used for all analytes. Data processing was conducted with Analyst software (Applied Sciences) to obtain the peak area ratios of analytes to the corresponding internal standards. A pooled lipid extract from study samples was used as a quality control (QC) sample, and was injected between every 10 study samples to verify the instrumental consistency. All duplicated sample runs were averaged. The values for each sphingolipid were peak area ratios that were normalized to internal standard and then normalized to QC standards. Values were relative peak area ratios and unitless.
Untargeted Sphingolipid analysis Methods have been detailed elsewhere (5). Briefly, plasma samples, stored at -80oC, were thawed and spiked with internal standards. Protein precipitation with methanol was performed, followed by liquid-liquid extraction of the supernatant using methyl-tert-butyl ether (MTBE). The lipid fraction was analyzed on an Agilent 6210 Time-of-Flight mass spectrometer (TOF-MS) coupled to an ultra-high performance liquid chromatography (UHPLC) system (Agilent 1290 Series). Analysis was performed in positive electrospray ionization (+ESI) mode. An Agilent Zorbax Rapid Resolution HD SB-C18 analytical column (2.1x100mm, 1.8µm) and Agilent Zorbax SB-C18 guard column (2.1x5mm, 1.8µm) were used for chromatographic separation. Gradient elution with mobile phase A (water with 0.1% formic acid) and mobile phase B (isopropyl alcohol/acetonitrile/water (60/36/4) with 0.1% formic acid) was as follows: 0-0.5 minutes 30%B, 0.5-7.42 minutes 30-70%B, 7.42-9.9 minutes 70-100% B, 9.9-10 minutes 100-30%B, 10-14.6 minutes 30%B with a flow rate of 0.7mL/minute. Equal volumes of all samples were pooled for a quality control (QC) sample and prepared with all other samples. Samples were randomized and injected in triplicate. QCs (used to monitor instrument reproducibility) and blanks (for carryover reduction) were run after every five sample injections. Mass spectral peaks were extracted using Mass Hunter Qualitative Analysis software (Agilent), and imported into Mass Profiler Professional (Agilent) for mass and retention time alignment using a 15ppm mass Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 36 of 64
cut-off and 0.2min retention time window. Sphingolipids were annotated using a sphingolipid-specific in-house database comprising HMDB, Lipid Maps, Metlin and KEGG with a mass error window of ≤ 10ppm. Results were exported to Mass Hunter Quantitative Analysis software (Agilent) to obtain peak area values for each annotated sphingolipid.
Microarray data For the subjects in the untargeted data, we previously profiled PBMC gene expression (2). See Table 1 and Supplemental Table 1 for demographic and clinical phenotypes for this cohort, which is the same used for the untargeted metabolomic study. In the gene expression data set, we identified sphingolipid related genes using the AmiGO website (http://amigo.geneontology.org/amigo) with gene ontology search terms "sphingolipid" and "ceramide”, which returned 24 and 46 terms respectively. Of the 70 gene ontology terms, 41 were most related to metabolism (e.g., ceramide transporter activity) and contained 212 unique genes identified in UniProt. Of those, 105 genes occurred in the list of 54676 probe sets and correspond to 257 transcripts (probe sets). For the untargeted sphingolipid data set, there were 33 sphingolipids measured; the 23 sphingolipids used in the meta-analysis above was a subset of these that occurred in both the untargeted and targeted data sets. To integrate the genomic and metabolomic data, we examined correlations between transcripts and sphingolipids overall and by disease status. Since correlation analysis can be sensitive to extreme outliers, transcripts and sphingolipids were filtered using the Grubb’s outlier test (p-value 20%, n= 28); control (lower lobe emphysema 20%, n= 31); control (n=98) and severe exacerbation (>=1 Copyright © 2014 by the American Thoracic Society
Page 37 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
event, n=19) and control (n=84) and exacerbation frequency (≥1 moderate or severe exacerbation during follow up, n=34).
Statistical analysis Hypothesis testing: Differences in demographic characteristics of study subjects were analyzed using a t test for continuous variables and a Chi squared test for categorical variables. Emphysema severity and FEV1/FVC were modeled as a continuous percent variable with the ß- distribution (betareg package in R; (6)). Emphysema was also modeled as an ordinal variable using ordinal logistic regression with emphysema classified as none, mild, moderate, and severe (polr function in R). FEV1% was modeled with linear regression. Exacerbations were modeled as a negative binomial (glm.nb function in R). Each sphingolipid or its principle component (see below) was modeled separately and covariates (age, gender, smoking status, body mass index (BMI), and airflow obstruction (FEV1)) were included. The slope relating each sphingolipid to the phenotype was evaluated for significance using t-tests or Z-tests for the beta and negative binomial models with FDR < 0.05.
Principle component analysis: Since the sphingolipid levels were highly correlated within class, we also computed the first principal component (PC) of each sphingolipid classes (Supplemental Table 3 & 4) using prcomp function in R. After performing PCA, we found that the loadings of the first principal components were the same sign (- or +) for each sphingolipid within each class. However, some loadings had all negative weights. To facilitate interpretation of our findings for these negative cases, the weights were multiplied by -1 so that an increase in the PC was associated with increased levels of the sphingolipids. This operation is possible since PCA is invariant to the signs of the components.
Receiver Operating Curve (ROC) Analysis: The outcomes were dichotomized to fit a logistic model: subjects with zero exacerbations were compared to subjects with one or more exacerbations; and mild or no emphysema subjects were compared to moderate to severe emphysema subjects (i.e., percent emphysema ≥ 10%).
Copyright © 2014 by the American Thoracic Society
Using
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 38 of 64
different thresholds in the model predictions, the ROC curves are a plot of the True Positive Rate as a function of False Positive Rate and represent the ability of the fitted logistic model to predict cases versus controls.
Meta-Analysis: Replication between the targeted and untargeted platforms was determined using the StoufferLiptak Z-score method, which converts the p-values from the two studies to normal quantiles and averages them to obtain a combined p-value (7,8). Each of the 23 sphingolipids that overlapped between the two studies were tested and consistency in the direction of the effect on the phenotype was taken into account.
Copyright © 2014 by the American Thoracic Society
Page 39 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
References for online supplement 1.
Regan, E.A., Hokanson, J.E., Murphy, J.R., Make, B., Lynch, D.A., Beaty, T.H., Curran-Everett, D., Silverman, E.K. and Crapo, J.D. (2010) Genetic epidemiology of COPD (COPDGene) study design. COPD, 7, 32-43.
2.
Bahr, T.M., Hughes, G.J., Armstrong, M., Reisdorph, R., Coldren, C.D., Edwards, M.G., Schnell, C., Kedl, R., LaFlamme, D.J., Reisdorph, N. et al. (2013) Peripheral blood mononuclear cell gene expression in chronic obstructive pulmonary disease. Am J Respir Cell Mol Biol, 49, 316-323.
3.
Pauwels, R.A., Buist, A.S., Ma, P., Jenkins, C.R. and Hurd, S.S. (2001) Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: National Heart, Lung, and Blood Institute and World Health Organization Global Initiative for Chronic Obstructive Lung Disease (GOLD): executive summary. Respiratory care, 46, 798-825.
4.
Carolan, B.J., Kim, Y.I., Williams, A.A., Kechris, K., Lutz, S., Reisdorph, N. and Bowler, R.P. (2013) The association of adiponectin with computed tomography phenotypes in chronic obstructive pulmonary disease. Am J Respir Crit Care Med, 188, 561-566.
5.
Yang, Y., Cruickshank, C., Armstrong, M., Mahaffey, S., Reisdorph, R. and Reisdorph, N. (2013) New sample preparation approach for mass spectrometry-based profiling of plasma results in improved coverage of metabolome. Journal of chromatography. A, 1300, 217-226.
6.
Cribari-Neto, F. and Zeileis, A. (2010) Beta Regression in R. Journal of statistical software, 34, 1–24.
7.
Stouffer, S.A. (1949) The American soldier. Princeton University Press, Princeton,.
8.
Liptak, T. (1958) On the combination of independent tests. Magyar Tud. Akad. Mat. Kutato Int. Kozl, 3.
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 40 of 64
Supplemental Tables and Figures Supplemental Table 1a: Clinical subphenotypes for each cohort Targeted Untargeted Cohort Cohort (N = 131) (N = 129) COPD None 29% 32% GOLD 1 11% 8% GOLD 2 25% 27% GOLD 3 24% 21% GOLD 4 12% 12% Chronic Bronchitis 19% 18% Emphysema None 41% 57% Mild 9% 6% Moderate 29% 20% Severe 21% 17% Exacerbations Moderate or Severe 0.70 ± 1.27 0.55 ± 1.02 Severe 0.20 ± 0.70 0.11 ± 0.33
Supplemental Table 1b: Additional characteristics of targeted cohort Control COPD Age 60 (53 - 70) 58 (63 - 70) Gender (% Male) 41 57 Current Smokers† (%) 27 22 Pack Years 40 (29 - 52) 46 (34- 60) Body Mass Index 27 (24 – 30) 27 (24 - 31) FEV1% 100 (93 – 105) 59 (37 - 74) FEV1/FVC 0.79 (0.74 – 0.82) 0.52 (0.37 – 0.74) Emphysema (%) 1.5 ± 2.2. 11.9 ± 11.8 Chronic Bronchitis (%) 5% 20% Exacerbations/year (moderate) 0.3 ± 1 0.73 ± 1.3 Exacerbations/year (severe) 0±0 0.23 ± 0.74
P-value 0.56 0.37 0.73 0.61 0.88 0.95 0.01 0.49 0.065 0.45 0.30 0.17
P-value 0.19 0.094 0.53 0.044 0.58 < 0.0001 < 0.0001 < 0.0001 0.0071 0.024 0.0014
Shown are percentages, medians (interquartile ranges) except for mean and standard deviation for exacerbations.
Copyright © 2014 by the American Thoracic Society
Page 41 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 2: Internal standards used for various lipid classes Internal Standards
Lipid Classes
C17 sphingomyelin (d18:1 N17:0) C17 dihydrosphingomyelin (d18:0 N17:0) C17 ceramide (d18:1 N17:0) C17 sphingosine (d17:1) C17 sphingosine (d17:1)-1-phosphate
Sphingomyelins dihydrosphingomyelins ceramides sphingosine and sphinganine sphingosine-1-phosphate (S-1-P) and sphinganine-1-phosphate (S’-1-P) ceramide (d18:1 N 16:0)-1-phosphate monohexosylsphingosine, monohexosylceramides, and monohydroxylated monohexosylceramides Dihexosylceramides, monohydroxylated dihexosylceramides, and trihexosylceramides sulfatides gangliosides
C12 ceramide (d18:1 N12:0)-1-phosphate deuterated glucosylceramide (d18:1 N18:0 d35) deuterated lactosylceramide (d18:1 N16:0 d3) deuterated sulfatide (d18:1 N18:0 d3) deuterated ganglioside GM1(d18:1 N18:0 d3)
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 42 of 64
Supplemental Table 3: Associations between sphingolipids and clinical covariates P Values
Correlations
Age
BMI
gender
Current Smoking
SM.d18.1.N14.1
0.369089335
0.482900721
1.52490828595084e-10
SM.d18.1.N14.0
0.501172453
0.988511852
SM.d18.1.N16.1
0.335513818
0.906090213
SM.d18.1.N16.0
0.077469807
0.0736016
SM.d18.1.N18.1
0.485359679
SM.d18.1.N18.0
0.578045779
SM.d18.1.N20.0
0.541388093
0.93274209
3.06409992874895e-07
0.288183528
SM.d18.1.N24.1
0.075389633
0.202033736
2.83740672083777e-05
0.776960411
SM.d18.1.N22.0
0.082566022
0.957576718
3.90247271419952e-05
0.668832101
-0.153432214
SM.d18.1.N25.1
0.386169892
0.066503472
0.000267464
0.866303217
-0.076934147
T Statistics
Age
BMI
gender
0.326609041
-0.079730441
0.062322901
7.173062188
Current Smoking 0.990974665
4.57580720811407e-06
0.339199384
-0.05975183
0.0012802
4.840258177
0.963306233
1.1592575775062e-07
0.369677617
-0.085469965
-0.010488619
5.697019386
0.904167851
2.68373824197941e-05
0.720192317
-0.156014037
-0.158066598
4.384016501
-0.360004262
0.625049213
1.90058727583179e-07
0.297800938
-0.061973916
0.04343066
5.625999778
1.049183912
0.614631188
5.22491336502058e-05
0.352128602
-0.049425463
0.044741677
4.232278278
0.936757194
-0.054258974
-0.007503528
5.438591724
1.072815525
-0.157107275
-0.113066004
4.39190997
-0.284529521
0.004729547
4.282867918
0.430180646
-0.16207167
3.792669251
0.16903069 1.091283356
SM.d18.1.N23.0
0.250623387
0.963462704
1.13966693966193e-06
0.28014028
-0.101877609
0.00407286
5.164206149
SM.d18.1.N24.0
0.034506422
0.884226168
0.000261145
0.931915264
-0.186316825
-0.012946064
3.776835944
-0.08584639
SM.d18.0.N16.0
0.238479704
0.879043817
3.30715554694656e-05
0.697128326
-0.104519929
-0.01352991
4.320847816
-0.391522848
SM.d18.0.N17.0
0.198267213
0.538537275
0.037371892
0.186498982
-0.114010349
-0.054641426
2.105914966
1.337204559
Cer.d18.1.N16.0
0.83282197
0.885956333
0.005668312
0.756438565
-0.018765902
-0.012751268
2.81461228
-0.311620023
Cer.d18.1.N18.0
0.846774185
0.551335904
0.009610928
0.749192913
-0.017179436
0.052932116
2.637760688
0.321275992
Cer.d18.1.N22.0
0.477190979
0.48298968
0.051635793
0.637182855
-0.063136951
-0.062310259
1.96553912
0.473386579
Cer.d18.1.N23.0
0.730676397
0.615721431
0.005542179
0.418627944
-0.030598127
-0.044604032
2.827437083
0.814117363
Cer.d18.1.N24.0
0.456478926
0.572083586
0.011903811
0.983605702
-0.066135313
-0.050201454
2.556988772
0.020624785
Cer.d18.1.N25.0
0.591790256
0.437439745
2.61336756315812e-05
0.765533412
-0.047650579
-0.068958875
4.439300214
0.299776456
Cer.d18.1.N26.0
0.877680949
0.580411011
8.90624347613412e-05
0.274810745
0.013683546
-0.049118657
4.106416695
1.100623906
sphingosine
0.301323258
0.75994083
0.440144861
0.510047216
-0.091704069
-0.027162342
-0.774725841
0.661139417
sphinganine
0.468750889
0.852566144
0.658857318
0.773784886
-0.064349995
-0.016522462
-0.44253225
-0.289001982
S.1.P
0.00692518
0.056916995
0.123912533
0.056605333
-0.236674645
-0.168074635
-1.549044655
-1.95481029
S..1.P
0.067290358
0.097601778
0.577279644
0.997790725
-0.161611034
-0.146492778
-0.558820573
-0.002777055
Cer.1.P.d18.1.N16.0
0.710521243
0.001618164
0.500250948
0.024446968
0.032990689
-0.274880324
0.676568616
2.309021327
monohexsph
0.157295297
0.137712074
0.000397464
0.051433363
-0.125243926
-0.131402677
3.679787253
-2.007445189
monohexcer.d18.1.N16.0
0.04062281
0.216767477
0.001885156
0.782085985
-0.180533201
-0.109489023
3.195810324
-0.277958058
monohexcer.d18.1.N18.0
0.822241905
0.17723826
0.037514387
0.90311024
0.019972854
-0.11953695
2.107936011
0.122207584
monohexcer.d18.1.N20.0
0.731461618
0.027091909
0.259245898
0.542878063
0.030505369
-0.194626049
1.133684021
0.611756898
monohexcer.d18.1.N22.0
0.658890795
0.272803918
0.833380298
0.666323633
-0.039233983
-0.097270346
0.210850901
-0.433416799
monohexcer.d18.1.N24.0
0.398311105
0.049884263
0.368845113
0.577417763
-0.074991659
-0.173032034
0.902245219
-0.560848047
monohexcer.d18.1.N24.1
0.472840363
0.035337837
0.170963394
0.458792269
-0.063760769
-0.185482267
1.379180565
-0.74556853
dihexcer.d18.1.N16.0
0.53126141
0.08509567
0.757453942
0.542213136
-0.055622106
-0.152197833
0.309552026
0.613565493
dihexcer.d18.1.N18.0
0.101864617
0.153376566
0.167179925
0.901199505
0.144680597
-0.126428063
-1.389681056
0.124852129
dihexcer.d18.1.N20.0
0.33437824
0.01792126
0.273660016
0.048462592
0.085670279
-0.208161542
1.100331843
2.005232809
dihexcer.d18.1.N22.0
0.202350257
0.037504021
0.335944379
0.065554078
0.112987226
-0.183383149
0.966577367
1.871257604
dihexcer.d18.1.N24.0
0.374722206
0.058274614
0.715847956
0.100697981
0.078799715
-0.1671764
0.364873069
1.660728657
dihexcer.d18.1.N24.1
0.110755561
0.009800659
0.688443999
0.207754254
0.141084808
-0.226640358
-0.40197494
1.272252539
trihexcer.d18.1.N16.0
0.453117492
0.00029782
0.017146774
0.508423685
-0.066628955
-0.313461479
2.420456002
0.665762724
trihexcer.d18.1.N18.0
0.122833146
0.274382322
0.633008127
0.316381234
0.136544645
-0.096952347
-0.478850145
1.010507255
trihexcer.d18.1.N22.0
0.38178452
0.031315192
0.649235214
0.331145647
0.07764479
-0.189686726
0.456107538
0.982138253
trihexcer.d18.1.N24.0
0.516055904
0.101326113
0.648796076
0.208554092
0.057693646
-0.144906185
0.456563212
1.273570765
OH.monohexcer.d18.1.N16.0
0.033365039
0.001484823
5.25670296104681e-05
0.169797486
-0.187490619
-0.276976303
4.226353144
-1.393949306 -0.795412652
OH.monohexcer.d18.1.N18.0
0.522832146
0.39651302
0.065350462
0.429880114
0.056766706
0.07527708
-1.859420604
OH.monohexcer.d18.1.N20.0
0.013566142
0.531933455
0.010660494
0.244433013
0.21686156
0.055531245
2.605242277
1.178445845
OH.monohexcer.d18.1.N22.0
0.634014341
0.901362302
0.147431239
0.894791439
0.042309924
0.011019286
1.458096559
0.132839269
OH.monohexcer.d18.1.N24.0
0.103827248
0.742761832
0.010613774
0.256922869
0.143866273
-0.029174048
2.606341119
1.144040904
OH.dihexcer.d18.1.N16.0
0.976488901
0.07052868
0.052127777
0.908632229
0.002620282
-0.159760228
1.961622892
-0.115367354
OH.dihexcer.d18.1.N18.0
0.248472581
0.156644371
0.002679127
0.640421039
-0.102339032
-0.125439075
3.069635147
-0.469920707
sulfa.d18.1.N16.0
0.158061084
0.042217104
0.003181802
0.405451495
-0.125015114
-0.179146573
3.023144048
0.839975294
sulfa.d18.1.N20.0
0.018036538
0.793894339
0.02482685
0.428743472
-0.20795739
-0.023225353
2.274159561
-0.799048236
sulfa.d18.1.N22.0
0.014335419
0.159134796
1.76032375526658e-05
0.556725682
-0.215162887
-0.124695689
4.474128435
-0.592059212
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 43 of 64 GM1.d18.1.N16.0
0.193014097
0.009872519
0.002181394
0.318669343
-0.115349319
-0.226424864
3.131386418
-1.00729403
GM1.d18.1.N18.0
0.026452968
0.048453486
0.159889699
0.205314686
-0.195429646
-0.174110729
1.414200463
1.281319304
GM1.d18.1.N18.1
0.786671816
0.924346065
0.016861183
0.670747314
-0.024058803
-0.00844285
2.425254498
-0.427916968
GM2.d18.1.N16.0
0.015585571
0.004758061
0.007715445
0.642445403
-0.212564746
-0.247087894
2.716299536
-0.467233664
GM2.d18.1.N18.0
0.001116636
0.012885007
0.006935034
0.694468698
-0.283808639
-0.218437305
2.747763418
-0.39533954
GM3.d18.1.N16.0
0.395221108
0.012850917
0.001196628
0.685927212
-0.07548263
-0.218518067
3.332773235
-0.406744439
GM3.d18.1.N18.0
0.369669108
0.120698745
0.25624326
0.509096475
-0.079634245
-0.137321025
1.141488836
0.663553473
GM3.d18.1.N18.1
0.378860585
0.031271368
0.045213834
0.733999906
-0.078121362
-0.189734991
2.026475619
0.341772839
GM3.d18.1.N20.0
0.372945165
0.30265217
0.000115826
0.892328035
-0.079092407
-0.091453393
4.020798126
-0.136113502
GM3.d18.1.N22.0
0.478172741
0.19723499
0.033602435
0.244478982
-0.062996607
-0.114271408
2.154134263
-1.175146768
GD1.d18.1.N16.0
0.971499524
0.10841858
0.010053697
0.735900812
-0.003176555
-0.142007407
2.624630248
0.33852482
GD1.d18.1.N18.0
0.00265816
0.000531785
0.551844355
0.6258375
-0.262452574
-0.300837914
0.596629513
-0.491283167
GD1.d18.1.N20.0
0.019316175
0.02668718
0.005226199
0.427860248
-0.205763906
-0.195133173
2.854889195
-0.801362121
GD3.d18.1.N18.0
0.746795825
0.018745803
0.218953584
0.253153143
-0.028700366
-0.206725591
1.235758932
-1.15671005
GQ1.d18.1.N18.0
0.44073011
0.658104133
6.43170288783615e-06
0.845917177
-0.068466024
-0.039330526
4.829214112
-0.195408519
GQ1.d18.1.N20.0
0.709329212
0.54686086
0.000136855
0.696395799
0.033132917
-0.053527543
3.961080786
0.392581448
Yellow indicates P < 0.05
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplement Table 4a: Sphingolipid Classes from Targeted Dataset Class sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins sphingomyelins ceramides ceramides ceramides ceramides ceramides ceramides ceramides sphingosine sphinganine sphingosine sphinganine ceramide-1-phosphate monohexosylceramides monohexosylceramides monohexosylceramides monohexosylceramides monohexosylceramides monohexosylceramides monohexosylceramides dihydroceramides dihydroceramides dihydroceramides dihydroceramides dihydroceramides dihydroceramides trihexosylceramides trihexosylceramides trihexosylceramides trihexosylceramides monohydroxylated ceramides monohydroxylated ceramides monohydroxylated ceramides monohydroxylated ceramides monohydroxylated ceramides monohydroxylated ceramides monohydroxylated
Metabolite sphingomyelin.d18.1.N14.1 sphingomyelin.d18.1.N14.0 sphingomyelin.d18.1.N16.1 sphingomyelin.d18.1.N16.0 sphingomyelin.d18.1.N18.1 sphingomyelin.d18.1.N18.0 sphingomyelin.d18.1.N20.0 sphingomyelin.d18.1.N24.1 sphingomyelin.d18.1.N22.0 sphingomyelin.d18.1.N25.1 sphingomyelin.d18.1.N23.0 sphingomyelin.d18.1.N24.0 sphingomyelin.d18.0.N16.0 sphingomyelin.d18.0.N17.0 ceramide.d18.1.N16.0 ceramide.d18.1.N18.0 ceramide.d18.1.N22.0 ceramide.d18.1.N23.0 ceramide.d18.1.N24.0 ceramide.d18.1.N25.0 ceramide.d18.1.N26.0 sphingosine sphinganine sphingosine-1-phosphate sphinganine-1-phosphate ceramide-1-phosphate d18.1.N16.0 monohexosylsphingosine monohexosylceramide.d18.1.N16.0 monohexosylceramide.d18.1.N18.0 monohexosylceramide.d18.1.N20.0 monohexosylceramide.d18.1.N22.0 monohexosylceramide.d18.1.N24.0 monohexosylceramide.d18.1.N24.1 dihexceramide.d18.1.N16.0 dihexceramide.d18.1.N18.0 dihexceramide.d18.1.N20.0 dihexceramide.d18.1.N22.0 dihexceramide.d18.1.N24.0 dihexceramide.d18.1.N24.1 trihexosylceramide.d18.1.N16.0 trihexosylceramide.d18.1.N18.0 trihexosylceramide.d18.1.N22.0 trihexosylceramide.d18.1.N24.0 OH.monohexceramide.d18.1.N16.0 OH.monohexceramide.d18.1.N18.0 OH.monohexceramide.d18.1.N20.0 OH.monohexceramide.d18.1.N22.0 OH.monohexceramide.d18.1.N24.0 OH.dihexceramide.d18.1.N16.0 OH.dihexceramide.d18.1.N18.0
Copyright © 2014 by the American Thoracic Society
Page 44 of 64
Page 45 of 64 ceramides sulfatides sulfatides sulfatides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides gangliosides
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
sulfatide.d18.1.N16.0 sulfatide.d18.1.N20.0 sulfatide.d18.1.N22.0 GM1.d18.1.N16.0 GM1.d18.1.N18.0 GM1.d18.1.N18.1 GM2.d18.1.N16.0 GM2.d18.1.N18.0 GM3.d18.1.N16.0 GM3.d18.1.N18.0 GM3.d18.1.N18.1 GM3.d18.1.N20.0 GM3.d18.1.N22.0 GD1.d18.1.N16.0 GD1.d18.1.N18.0 GD1.d18.1.N20.0 GD3.d18.1.N18.0 GQ1.d18.1.N18.0 GQ1.d18.1.N20.0
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplement Table 4b: Sphingolipid classes used for untargeted dataset Class Metabolite sphingosines (4E,8E,9Me-d19:2)sphingosine sphingosines N-Lignoceroylsphingosine sphingosines N-Palmitoylsphingosine sphingosines Sphingosine sphingosines Sphingosine sphingosines Sphingosine sphingosines Sphingosine-1-phosphocholine sphingomyelins SM(d16:1/17:0) sphingomyelins SM(d18:0/16:1(9Z)) sphingomyelins SM(d18:0/18:1(11Z)) sphingomyelins SM(d18:0/24:1(15Z)) sphingomyelins SM(d18:1/12:0) sphingomyelins SM(d18:1/14:0) sphingomyelins SM(d18:1/16:0) sphingomyelins SM(d18:1/16:1) sphingomyelins SM(d18:1/20:1) sphingomyelins SM(d18:1/23:0) sphingomyelins SM(d18:1/24:1(15Z)) sphingomyelins SM(d18:2/14:0) sphingomyelins SM(d18:2/15:0) sphingomyelins SM(d18:2/23:0) ceramides Cer(d16:1/23:0) ceramides Cer(d18:1/22:1(13Z)) ceramides Cer(d18:1/23:0) ceramides Cer(d18:1/24:1(15Z)) ceramides Cer(d18:2/23:0) galacto-beta-ceramides *1-3Galalpha1-3Galalpha1-3Galalpha1-4Galbeta1-4Glcbeta-Cer(d18:1/24:1(15Z)) galacto-beta-ceramides Fucalpha1-2Galalpha1-3Galbeta1-4Glcbeta-Cer(d18:1/18:0) galacto-beta-ceramides GlcAbeta-Cer(d18:1/18:0) galacto-beta-ceramides NeuAcalpha2-3Galbeta-Cer(d18:1/20:0) galacto-beta-ceramides NeuGcalpha2-3Galbeta1-4GlcNAcbeta1-3Galbeta1-4Glcbeta-Cer(d18:1/24:1(15Z)) ceramide phosphate CerP(d18:1/24:1(15Z))
Copyright © 2014 by the American Thoracic Society
Page 46 of 64
Page 47 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 5: Sphingolipids associated with emphysema from targeted dataset Sphingolipid
ß
SE
P
P (FDR)
Cer.1.P.d18.1.N16.0
-0.268358669
0.209095221 0.199342158 0.299013237
Cer.d18.1.N16.0
-0.587234104
0.171878623
Cer.d18.1.N18.0
-0.187219117
0.108712979 0.085043914 0.163000836
Cer.d18.1.N22.0
0.00063417 0.016427536
-0.12810837 0.110832068 0.247730985 0.352139261
Cer.d18.1.N23.0
-0.119869688
0.106488811 0.260311477 0.352186116
Cer.d18.1.N24.0
-0.176904977
0.089140069 0.047191995 0.130626252
Cer.d18.1.N25.0
-0.207948725
0.10549082 0.048695345 0.130626252
Cer.d18.1.N26.0
-0.249160338
0.105509474 0.018201304
dihexcer.d18.1.N16.0
-0.451408235
0.215390391 0.036102799 0.130626252
dihexcer.d18.1.N18.0
-0.013216556
0.220977646 0.952307382 0.963849124
dihexcer.d18.1.N20.0
-0.133101417
0.228435103
dihexcer.d18.1.N22.0
-0.199223003
0.189637388 0.293466661
dihexcer.d18.1.N24.0
-0.332316325
0.168974279 0.049221486 0.130626252
dihexcer.d18.1.N24.1
-0.376614337
GD1.d18.1.N16.0
-0.268456228
0.151861357 0.077099057 0.156465733
GD1.d18.1.N18.0
-0.107458669
0.184077573 0.559376146 0.644135723
GD1.d18.1.N20.0
0.21234966
0.1046575
0.56011802 0.644135723 0.38206037
0.07613637 0.156465733
0.217878554 0.112128482 0.052002246 0.132330751
GD3.d18.1.N18.0
-0.025812238
0.101384166 0.799033362 0.842896298
GM1.d18.1.N16.0
-0.254298947
0.204113395 0.212811781 0.312425806
GM1.d18.1.N18.0
-0.196497617
0.138020167 0.154536891 0.253102267
GM1.d18.1.N18.1
-0.037788998
0.08148698 0.642832046 0.704054145
GM2.d18.1.N16.0
-0.053079867
GM2.d18.1.N18.0
-0.085048737
0.130304008 0.513953115
GM3.d18.1.N16.0
-0.765447586
0.216859931 0.000416052 0.016427536
GM3.d18.1.N18.0
-0.492576325
0.221217008 0.025969624 0.116902375
GM3.d18.1.N18.1
-0.270725258
0.18716289 0.148045791 0.249150233
GM3.d18.1.N20.0
-0.367613572
0.193559942 0.057535109 0.132330751
GM3.d18.1.N22.0
-0.247326971
0.14967801 0.098454361 0.178772392
GQ1.d18.1.N18.0
-0.426885514
0.223155617 0.055754467 0.132330751
GQ1.d18.1.N20.0
-0.216014908
0.188996374 0.253056051 0.352139261
monohexcer.d18.1.N16.0
0.07615105 0.485781547 0.598552263
-0.47385556 0.155169851
0.62215377
0.00225973 0.031184276
monohexcer.d18.1.N18.0
-0.281621984
0.183177667 0.124189189
monohexcer.d18.1.N20.0
-0.207997277
0.161062046 0.196561244 0.299013237
0.21422635
monohexcer.d18.1.N22.0
-0.157341348
0.138276681 0.255173377 0.352139261
monohexcer.d18.1.N24.0
-0.263784007
0.152332667 0.083339238 0.163000836
monohexcer.d18.1.N24.1
-0.342051651
0.166632631 0.040098858 0.130626252
monohexsph
-0.614435537
0.308625071 0.046494062 0.130626252
OH.dihexcer.d18.1.N16.0
0.111352836 0.179327955 0.534635669 0.636032089
OH.dihexcer.d18.1.N18.0
-0.2516346 0.188995256 0.183046438 0.287050096
OH.monohexcer.d18.1.N16.0
-0.340264846
0.153968639 0.027107797 0.116902375
OH.monohexcer.d18.1.N18.0
-0.742965941
0.408879774 0.069205358 0.154037733
OH.monohexcer.d18.1.N20.0
0.005411705 0.119400635 0.963849124 0.963849124
OH.monohexcer.d18.1.N22.0
-0.021805815
0.146392237 0.881589362 0.907905462
OH.monohexcer.d18.1.N24.0
0.029703764
0.12110719 0.806248633 0.842896298
Sphinganine-1-phosphate
0.177784948 0.125843526 0.157730398 0.253102267
Sphingosine-1-phosphate (S1P)
-0.113641519
0.222672188 0.609803864 0.678652688
SM.d18.0.N16.0
-0.487380621
0.274222924 0.075516364 0.156465733
SM.d18.1.N14.0
-0.433480208
0.150673175 0.004015306 0.046176023
SM.d18.1.N14.1
-0.352047072
0.155621496
SM.d18.1.N16.0
-0.896404045
0.271288498 0.000952321 0.016427536
SM.d18.1.N16.1
-0.519792993
0.186879577 0.005411953
SM.d18.1.N18.0
0.023685 0.116733214 0.0491139
-0.62134225 0.325527609 0.056297749 0.132330751
SM.d18.1.N18.1
-0.368548326
0.22181764 0.096614591 0.178772392
SM.d18.1.N20.0
-0.447569594
0.186191524 0.016225073 0.101775461
SM.d18.1.N22.0
-0.422594384
0.175242935 0.015888035 0.101775461
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
SM.d18.1.N23.0
-0.374665236
0.173344291
SM.d18.1.N24.0
-0.453420189
0.163992689 0.005694365
SM.d18.1.N24.1
-0.606206008
0.182735907 0.000908632 0.016427536
SM.d18.1.N25.1 sphinganine sphingosine
Page 48 of 64
0.03066495 0.117548975 0.0491139
-0.33635917 0.168707799 0.046180681 0.130626252 -0.088776578
0.17115283 0.603971571 0.678652688
0.044852617 0.125457361
0.72070792 0.777013226
sulfa.d18.1.N16.0
-0.260339996
0.161721998 0.107441629 0.190089036
sulfa.d18.1.N20.0
-0.311502812
0.137694674
sulfa.d18.1.N22.0
-0.420944626
0.193671156 0.029742563 0.117548975
trihexcer.d18.1.N16.0
-0.569061881
0.221519549 0.010202148 0.078216467
trihexcer.d18.1.N18.0
-0.397842639
0.192209876 0.038467842 0.130626252
trihexcer.d18.1.N22.0
-0.207245275
trihexcer.d18.1.N24.0
-0.146224827
0.0236806 0.116733214
0.19624703 0.290948634
0.38206037
0.172125938 0.395590716 0.496286534
Emphysema modeling (ß-regression) included covariates (see methods for details); green indicates P < 0.01 with false discovery rate (FDR) adjustment < 0.05 and yellow indicates all others with P < 0.05
Copyright © 2014 by the American Thoracic Society
Page 49 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 6: Area under the curve from receiver operating characteristic curve analysis for moderate to severe emphysema.
Emphysema
All Subjects
Less Severe COPD Subjects (GOLD 0,1,2)
Covariates (cov) only
0.90
0.82
SM.d18.1.N14.0 + cov
0.90
0.83
SM.d18.1.N16.1 + cov
0.90
0.83
SM.d18.1.N16.0 + cov
0.90
0.84
SM.d18.1.N24.1 + cov
0.90
0.84
SM.d18.1.N24.0 + cov
0.91
0.86
Cer.d18.1.N16.0 + cov
0.91
0.84
monohexcer.d18.1.N16.0 + cov
0.90
0.83
GM3.d18.1.N16.0 + cov
0.91
0.84
All sphingolipids above + cov
0.93
0.88
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 50 of 64
Supplemental Table 7: Sphingolipids associated with severe (hospitalization) COPD Exacerbations Sphingolipid Cer.1.P.d18.1.N16.0
ß
SE
P
P (FDR)
3.167285212
1.570994651
0.043789327
0.177733152
Cer.d18.1.N16.0
1.09758561
0.667208287
0.099961164
0.328443824
Cer.d18.1.N18.0
-0.009695998
0.505353273
0.984692268
0.984692268
Cer.d18.1.N22.0
-0.257807437
0.624250015
0.679615791
0.843743546
Cer.d18.1.N23.0
-0.711560187
0.820854138
0.386022348
0.739876167
Cer.d18.1.N24.0
-0.067752091
0.394688637
0.863705155
0.918706636
Cer.d18.1.N25.0
-0.046144021
0.453570612
0.918966995
0.960738222
Cer.d18.1.N26.0
0.217241983
0.375571046
0.562973085
0.826492401
dihexcer.d18.1.N16.0
-0.057861702
1.159147464
0.96018816
0.974308574
dihexcer.d18.1.N18.0
1.863626957
0.799386247
0.019736373
0.104754595
dihexcer.d18.1.N20.0
2.353175503
0.752325656
0.001760791
0.01735637
dihexcer.d18.1.N22.0
1.707374916
0.507968849
0.000776101
0.008925165
dihexcer.d18.1.N24.0
1.022383852
0.345039865
0.003045684
0.022501416
dihexcer.d18.1.N24.1
1.098040707
0.765033407
0.151205999
0.386415332
GD1.d18.1.N16.0
1.878559785
0.624983179
0.002649036
0.022501416
GD1.d18.1.N18.0
1.775746996
1.100231961
0.106532826
0.334125681
GD1.d18.1.N20.0
0.938530928
0.503212435
0.062170336
0.238319623
GD3.d18.1.N18.0
1.521069035
0.628318981
0.01548406
0.089033344
GM1.d18.1.N16.0
0.537459845
1.164184887
0.644323774
0.843743546
GM1.d18.1.N18.0
0.227229342
0.583588284
0.697005538
0.843743546
GM1.d18.1.N18.1
0.368053093
0.478590724
0.441872439
0.743638983
GM2.d18.1.N16.0
0.764683923
0.950771747
0.421236973
0.743638983
GM2.d18.1.N18.0
-0.323966019
0.718528523
0.652080173
0.843743546
GM3.d18.1.N16.0
0.246305107
0.891012136
0.782215557
0.884801204
GM3.d18.1.N18.0
1.178739613
0.969933189
0.224259629
0.499158529
GM3.d18.1.N18.1
0.367168931
0.858390552
0.668839613
0.843743546
GM3.d18.1.N20.0
0.935184914
0.743736007
0.20860335
0.49633211
GM3.d18.1.N22.0
0.505909438
0.650290224
0.436583841
0.743638983
GQ1.d18.1.N18.0
0.56754364
0.899319957
0.527987634
0.791981451
GQ1.d18.1.N20.0
-0.382233241
0.960939493
0.690799473
0.843743546
monohexcer.d18.1.N16.0
-1.504409737
0.996906496
0.131278832
0.377426641
monohexcer.d18.1.N18.0
0.270590137
0.846924449
0.749349383
0.884801204
monohexcer.d18.1.N20.0
-0.204387753
0.810359744
0.800872147
0.891293196
monohexcer.d18.1.N22.0
-0.43265403
0.61471621
0.481540352
0.772759415
monohexcer.d18.1.N24.0
-1.098875012
0.821743972
0.181141885
0.44638536
monohexcer.d18.1.N24.1
-0.552272692
0.853739317
0.517705211
0.791981451
monohexsph
-1.754073837
1.842347958
0.34105333
0.67236228
1.535903615
0.849018173
0.070445562
0.255828618
OH.dihexcer.d18.1.N16.0 OH.dihexcer.d18.1.N18.0
0.942439025
1.116846802
0.39875931
0.743632227
OH.monohexcer.d18.1.N16.0
-1.445632488
0.976790181
0.138877886
0.383302964
OH.monohexcer.d18.1.N18.0
1.153455522
2.145099011
0.590772661
0.843743546
OH.monohexcer.d18.1.N20.0
-1.132208958
0.666540684
0.089388084
0.308388891
OH.monohexcer.d18.1.N22.0
-0.540122887
0.670324857
0.420379585
0.743638983
OH.monohexcer.d18.1.N24.0
0.749374888
0.73505072
0.30797167
0.643940765
Sphinganine-1-phosphate
-1.427179331
0.980319113
0.145439238
0.385973361
Sphingosine-1-phosphate (S1P)
-4.241148157
1.441590245
0.003261075
0.022501416
SM.d18.0.N16.0
2.810510366
1.240601385
0.023485575
0.115750335
SM.d18.0.N17.0
9.011429139
4.155566855
0.030119157
0.138548124
SM.d18.1.N14.0
-0.802170826
0.716492215
0.262892502
0.566861956
SM.d18.1.N14.1
-2.048914567
0.765695646
0.00745319
0.04675183
SM.d18.1.N16.0
0.818705144
1.163310023
0.481574708
0.772759415
SM.d18.1.N16.1
0.133965528
0.790623428
0.86544828
0.918706636
SM.d18.1.N18.0
0.73465774
1.409778891
0.60228626
0.843743546
Copyright © 2014 by the American Thoracic Society
Page 51 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
SM.d18.1.N18.1
1.512275297
0.971395923
0.11951648
0.35854944
SM.d18.1.N20.0
-0.261110322
0.937509302
0.780617118
0.884801204
SM.d18.1.N22.0
0.915962553
0.919288975
0.319064805
0.647513869
SM.d18.1.N23.0
-0.20886808
0.744220204
0.7789761
0.884801204
SM.d18.1.N24.0
-0.127016535
0.744370874
0.864509919
0.918706636
SM.d18.1.N24.1
-0.054199631
0.741352487
0.941719281
0.969830304
SM.d18.1.N25.1
-0.348351944
0.805117187
0.665252821
0.843743546
sphinganine
2.145526154
1.018429021
0.035143444
0.151556104
sphingosine
1.30447274
2.684399218
0.627005538
0.843743546
sulfa.d18.1.N16.0
4.365353784
1.058826627
sulfa.d18.1.N20.0
-0.377053167
0.588337336
sulfa.d18.1.N22.0
0.986768513
0.80750671
trihexcer.d18.1.N16.0
2.421679925
0.474207903
3.27678477610e-07
1.13049e-05
trihexcer.d18.1.N18.0
3.441030675
0.615560298
2.26966374069e-08
1.56606e-06
trihexcer.d18.1.N22.0
3.25645855
0.804369207
5.15551983384e-05
0.000711462
trihexcer.d18.1.N24.0
2.648267264
0.650191719
4.64002676581e-05
0.000711462
3.7425879318325e-05 0.521601164 0.221709822
0.000711462 0.791981451 0.499158529
Exacerbation modeling (negative binomial regression) included covariates (see methods for details); green indicates P < 0.01 with false discovery rate (FDR) adjustment < 0.05 and yellow indicates all others with P < 0.05
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 8: Sphingolipids associated with moderate and severe COPD Exacerbations Sphingolipid
ß
SE
P
P (FDR)
Cer.1.P.d18.1.N16.0
0.686686191
0.601537839
0.253640654
0.561719854
Cer.d18.1.N16.0
0.050193699
0.434522539
0.908037245
0.921390734
Cer.d18.1.N18.0
-0.997509829
0.31573857
0.001581538
0.036375379
Cer.d18.1.N22.0
-0.299566946
0.284387922
0.292169449
0.568323159
Cer.d18.1.N23.0
-0.201713067
0.253961219
0.427039893
0.648567918
Cer.d18.1.N24.0
-0.125670798
0.206479307
0.542765549
0.723031668
Cer.d18.1.N25.0
-0.143638303
0.229482193
0.531365103
0.723031668
Cer.d18.1.N26.0
-0.129878649
0.236160571
0.582347443
0.730581337
dihexcer.d18.1.N16.0
0.167830324
0.533559031
0.7531043
0.866069945
dihexcer.d18.1.N18.0
1.525704152
0.506253908
0.002580715
0.044517338
dihexcer.d18.1.N20.0
0.762751658
0.462403961
0.099038017
0.480866162
dihexcer.d18.1.N22.0
0.869383318
0.389708257
0.025690615
0.177265246
dihexcer.d18.1.N24.0
0.593181554
0.311099497
0.056556056
0.32519732
dihexcer.d18.1.N24.1
0.606779978
0.447334641
0.174961077
0.498366346
GD1.d18.1.N16.0
0.988980923
0.372710858
0.007966642
0.109939658
GD1.d18.1.N18.0
-0.122277262
0.53286115
0.818501282
0.889888623
GD1.d18.1.N20.0
0.255596803
0.275776296
0.354016578
0.579918023
GD3.d18.1.N18.0
-0.053923811
0.274961025
0.844520642
0.889888623
GM1.d18.1.N16.0
0.5922162
0.660334441
0.369802797
0.579918023
GM1.d18.1.N18.0
-0.213975209
0.356506165
0.548372564
0.723031668
GM1.d18.1.N18.1
0.124240003
0.273373476
0.649490686
0.78195126
GM2.d18.1.N16.0
0.543399862
0.218154857
0.012742338
0.125603048
GM2.d18.1.N18.0
0.323677602
0.357348571
0.365054659
0.579918023
GM3.d18.1.N16.0
-0.509823364
0.510845096
0.318279398
0.573009805
GM3.d18.1.N18.0
-0.548048514
0.506169615
0.278925225
0.566054133
GM3.d18.1.N18.1
-0.105298478
0.664659495
0.874122155
0.900215354
GM3.d18.1.N20.0
0.097496717
0.515910766
0.850108644
0.889888623
GM3.d18.1.N22.0
0.088167204
0.322241027
0.784387124
0.887257566
GQ1.d18.1.N18.0
-0.659423364
0.476826955
0.166683075
0.498366346
GQ1.d18.1.N20.0
-0.523462955
0.501433521
0.296516431
0.568323159
monohexcer.d18.1.N16.0
-3.57019E-05
0.388557925
0.999926688
0.999926688
monohexcer.d18.1.N18.0
-0.225477097
0.465532176
0.628142698
0.773961539
monohexcer.d18.1.N20.0
0.682964166
0.420738924
0.104536122
0.480866162
monohexcer.d18.1.N22.0
-0.177603156
0.400318576
0.657292363
0.78195126
monohexcer.d18.1.N24.0
0.407809376
0.406814223
0.316128133
0.573009805
monohexcer.d18.1.N24.1
-0.230456177
0.40010707
0.564624466
0.723031668
monohexsph
0.812248807
0.823342599
0.323875107
0.573009805
OH.dihexcer.d18.1.N16.0
0.632121974
0.531137644
0.233995945
0.538190673
OH.dihexcer.d18.1.N18.0
-0.692314823
0.494754176
0.161719826
0.498366346
OH.monohexcer.d18.1.N16.0
-0.068699793
0.366222478
0.851197814
0.889888623
OH.monohexcer.d18.1.N18.0
0.259319585
1.229552901
0.832960939
0.889888623
OH.monohexcer.d18.1.N20.0
0.361093065
0.27414842
0.187790217
0.498366346
OH.monohexcer.d18.1.N22.0
0.36443039
0.301092288
0.226140366
0.538058111
OH.monohexcer.d18.1.N24.0
0.574545474
0.244729692
0.01889106
0.162935391
Sphinganine-1-phosphate
-0.436125798
0.358678591
0.224013991
0.538058111
Sphingosine-1-phosphate (S1P)
-0.709026888
0.530551245
0.181419672
0.498366346
SM.d18.0.N16.0
0.396441388
0.690456575
0.565850871
0.723031668
SM.d18.1.N14.0
-0.583921361
0.381677161
0.126045717
0.498366346
SM.d18.1.N14.1
-0.459687249
0.408541229
0.260507758
0.561719854
SM.d18.1.N16.0
-0.411077263
0.669272575
0.539073254
0.723031668
SM.d18.1.N16.1
-0.283072211
0.457053911
0.535691762
0.723031668
SM.d18.1.N18.0
-1.191487467
0.797581775
0.135208296
0.498366346
SM.d18.1.N18.1
-0.773640026
0.528378215
0.143145181
0.498366346
SM.d18.1.N20.0
-0.304471086
0.472847373
0.519633887
0.723031668
Copyright © 2014 by the American Thoracic Society
Page 52 of 64
Page 53 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
SM.d18.1.N22.0
-0.699913716
0.476009891
0.141459836
0.498366346
SM.d18.1.N23.0
-0.573179104
0.407608902
0.159665133
0.498366346
SM.d18.1.N24.0
-0.48899893
0.448569284
0.275655787
0.566054133
SM.d18.1.N24.1
-1.040447541
0.460107602
0.02373985
0.177265246
SM.d18.1.N25.1
-0.515684962
0.380127931
0.174905489
0.498366346
sphinganine
0.513907242
0.54237291
0.343375689
0.579918023
sphingosine
-0.29225067
0.310445092
0.346503775
0.579918023
sulfa.d18.1.N16.0
0.826161197
0.422883867
0.050744266
0.318304941
sulfa.d18.1.N20.0
-0.108917802
0.287547534
0.704849827
0.824315899
sulfa.d18.1.N22.0
0.563297626
0.445546247
0.206127599
0.526770532
trihexcer.d18.1.N16.0
0.88044359
0.484757639
0.069331232
0.367988845
trihexcer.d18.1.N18.0
1.28135432
0.494935582
0.009627424
0.110715378
trihexcer.d18.1.N22.0
1.83590452
0.469162151
9.10978E-05
0.003142873
trihexcer.d18.1.N24.0
1.743391415
0.434593576
6.03228E-05
0.003142873
Exacerbation modeling (negative binomial regression) included covariates (see methods for details); green indicates P < 0.01 with false discovery rate (FDR) adjustment < 0.05 and yellow indicates all others with P < 0.05
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 54 of 64
Supplemental Table 9: Area under the curve from receiver operating characteristic curve analysis for severe exacerbations. *There were not enough cases of severe exacerbations to stratify by GOLD stage.
Severe Exacerbations*
All Subjects
Covariates (cov) only
0.90
S.1.P + cov
0.90
dihexcer.d18.1.N20.0 + cov
0.90
dihexcer.d18.1.N22.0 + cov
0.90
dihexcer.d18.1.N24.0 + cov
0.90
trihexcer.d18.1.N16.0 + cov
0.91
trihexcer.d18.1.N18.0 + cov
0.89
trihexcer.d18.1.N22.0 + cov
0.89
trihexcer.d18.1.N24.0 + cov
0.89
sulfa.d18.1.N16.0 + cov
0.91
GD1.d18.1.N16.0 + cov
0.90
All sphingolipids above + cov
0.94
Copyright © 2014 by the American Thoracic Society
Page 55 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 10: Area under the curve from receiver operating characteristic curve analysis for moderate and severe exacerbations. Moderate and Severe Exacerbations
All Subjects
Less Severe COPD Subjects (GOLD 0,1,2)
Covariates (cov) only
0.81
0.85
dihexcer.d18.1.N18.0 + cov
0.81
0.85
trihexcer.d18.1.N22.0 + cov
0.82
0.86
trihexcer.d18.1.N24.0 + cov
0.81
0.86
SM.d18.1.N24.1 + cov
0.81
0.89
Cer.d18.1.N18.0 + cov
0.82
0.94
All sphingolipids above & cov
0.83
0.94
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Table 11: Sphingolipids associated with chronic bronchitis Sphingolipid
Cer.1.P.d18.1.N16.0 Cer.d18.1.N16.0 Cer.d18.1.N18.0 Cer.d18.1.N22.0 Cer.d18.1.N23.0 Cer.d18.1.N24.0 Cer.d18.1.N25.0 Cer.d18.1.N26.0 dihexcer.d18.1.N16.0 dihexcer.d18.1.N18.0 dihexcer.d18.1.N20.0 dihexcer.d18.1.N22.0 dihexcer.d18.1.N24.0 dihexcer.d18.1.N24.1 GD1.d18.1.N16.0 GD1.d18.1.N18.0 GD1.d18.1.N20.0 GD3.d18.1.N18.0 GM1.d18.1.N16.0 GM1.d18.1.N18.0 GM1.d18.1.N18.1 GM2.d18.1.N16.0 GM2.d18.1.N18.0 GM3.d18.1.N16.0 GM3.d18.1.N18.0 GM3.d18.1.N18.1 GM3.d18.1.N20.0 GM3.d18.1.N22.0 GQ1.d18.1.N18.0 GQ1.d18.1.N20.0 monohexcer.d18.1.N16.0 monohexcer.d18.1.N18.0 monohexcer.d18.1.N20.0 monohexcer.d18.1.N22.0 monohexcer.d18.1.N24.0 monohexcer.d18.1.N24.1 monohexsph OH.dihexcer.d18.1.N16.0 OH.dihexcer.d18.1.N18.0 OH.monohexcer.d18.1.N16.0 OH.monohexcer.d18.1.N18.0 OH.monohexcer.d18.1.N20.0 OH.monohexcer.d18.1.N22.0 OH.monohexcer.d18.1.N24.0 S..1.P S.1.P SM.d18.0.N16.0 SM.d18.0.N17.0 SM.d18.1.N14.0 SM.d18.1.N14.1 SM.d18.1.N16.0 SM.d18.1.N16.1 SM.d18.1.N18.0 SM.d18.1.N18.1
ß -0.253684345 -1.317209568 -0.69012277 -0.739078503 -1.513658452 -1.034478121 -1.149658895 -0.54693321 0.533416633 1.400232721 -0.016006927 -0.348744542 0.07255756 0.862557595 -0.689924472 -0.719084328 -0.240456463 0.332301934 -0.010230441 -2.655615842 -1.301490436 -0.883318854 -0.964739471 -0.917808448 1.004686431 -0.507790771 -0.79809677 -0.588879879 -0.845377439 -1.231177289 -0.787149677 0.185057343 -0.449412634 0.781469102 -0.093993229 0.713276436 -4.084127683 -2.008876014 -2.133358126 -1.101704921 5.48625778 0.049245715 -0.42418899 -0.404094387 1.493205908 2.793664716 -1.854548047 -5.583604673 -1.55775226 -2.437483936 -1.521001511 -1.700772915 -0.423109469 -0.85802368
SE
1.026899411 0.956298671 0.668220504 0.729328986 0.854301952 0.738686688 0.724169969 0.594119431 1.005510778 0.903937602 1.061382044 0.893780641 0.714870634 0.969117146 0.745775199 0.80589004 0.515009914 0.440802381 0.869576102 0.939932549 0.532480129 0.494944547 0.703577838 1.03716991 1.068436115 0.837621315 0.948521164 0.717157522 1.087228403 0.922193548 0.787488846 0.867907071 0.780517262 0.65650527 0.694139945 0.765169182 1.675540944 1.008960119 0.941353808 0.789933719 2.148968294 0.511693994 0.717840336 0.584574276 0.664227387 1.064400405 1.318423005 4.523068832 0.979005836 1.165914504 1.403389454 1.14848041 1.569868313 1.145887169
P
0.804877935 0.168387392 0.301708268 0.310884518 0.076426547 0.161384947 0.112387238 0.357270853 0.595770021 0.121372844 0.987967391 0.696395829 0.919155567 0.373441876 0.354908428 0.372239704 0.640573739 0.450935129 0.990613217 0.004723254 0.014517187 0.074312837 0.17031548 0.376201834 0.347046445 0.54436235 0.400117354 0.411571967 0.436832723 0.181858991 0.317518985 0.831153395 0.564758346 0.233910022 0.892288015 0.351242815 0.014789296 0.046476919 0.023435102 0.163112297 0.010680781 0.923329521 0.554571431 0.489400473 0.024574007 0.008674138 0.159533826 0.21702684 0.111574185 0.03656216 0.27845095 0.138635944 0.787530259 0.45398651
P (FDR)
0.895751251 0.481669972 0.58479615 0.58479615 0.405648596 0.481669972 0.481669972 0.58479615 0.708760887 0.481669972 0.990613217 0.800855204 0.950891596 0.58479615 0.58479615 0.58479615 0.749145559 0.602405176 0.990613217 0.161998713 0.161998713 0.405648596 0.481669972 0.58479615 0.58479615 0.682927312 0.587406329 0.591634703 0.602405176 0.481669972 0.58479615 0.910310861 0.68365484 0.520638435 0.947198047 0.58479615 0.161998713 0.29153704 0.18840072 0.481669972 0.161998713 0.950891596 0.683311227 0.630393318 0.18840072 0.161998713 0.481669972 0.499914813 0.481669972 0.252278906 0.58479615 0.481669972 0.890812916 0.602405176
Copyright © 2014 by the American Thoracic Society
Page 56 of 64
Page 57 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
SM.d18.1.N20.0 -1.035001744 0.965876395 0.283914342 0.58479615 SM.d18.1.N22.0 -1.115092759 0.903938287 0.217354267 0.499914813 SM.d18.1.N23.0 -1.253303696 0.937767111 0.18139358 0.481669972 SM.d18.1.N24.0 -0.685429213 0.808053933 0.396300504 0.587406329 SM.d18.1.N24.1 -0.891921916 1.000747806 0.372792154 0.58479615 SM.d18.1.N25.1 -0.117896168 0.816746237 0.88522521 0.947198047 sphinganine 0.92208621 0.701158093 0.188479554 0.481669972 sphingosine 0.854505253 0.500239124 0.087600432 0.431744985 sulfa.d18.1.N16.0 -0.664175748 0.758762938 0.381388794 0.58479615 sulfa.d18.1.N20.0 -1.918376097 0.799617489 0.016434652 0.161998713 sulfa.d18.1.N22.0 -2.651405897 1.066805558 0.012941706 0.161998713 trihexcer.d18.1.N16.0 -0.812106257 1.058142386 0.442794417 0.602405176 trihexcer.d18.1.N18.0 -0.624404564 0.911555054 0.493351292 0.630393318 trihexcer.d18.1.N22.0 -1.268702603 0.963812768 0.188060889 0.481669972 Exacerbation modeling (negative binomial regression) included covariates (see methods for details); green indicates P < 0.01 with false discovery rate (FDR) adjustment < 0.05 and yellow indicates all others with P < 0.05
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 58 of 64
Supplemental Table 12: Sphingolipids associated with FEV1/FVC post-bronchodilator Sphingolipid
ß
SE
P
Cer.1.P.d18.1.N16.0
-0.101296342
0.210842709
0.63091769
0.79151492
Cer.d18.1.N16.0
0.01146349
0.18119318
0.949554152
0.979910154
Cer.d18.1.N18.0
-0.128009399
0.115969539
0.269671593
0.664547855
Cer.d18.1.N22.0
-0.001918583
0.115180625
0.986710116
0.998670752
Cer.d18.1.N23.0
-0.070624358
0.115705439
0.541609701
0.788766376
Cer.d18.1.N24.0
0.02896388
0.095617322
0.761954965
0.891099875
Cer.d18.1.N25.0
-0.055979894
0.112029898
0.617295424
0.788766376
Cer.d18.1.N26.0
-0.120629272
0.116828797
0.301823765
0.671801284
dihexcer.d18.1.N16.0
-0.390868101
0.228604708
0.087303049
0.585412059
dihexcer.d18.1.N18.0
-0.205742992
0.225164431
0.360850576
0.70686584
dihexcer.d18.1.N20.0
-0.158717319
0.232865061
0.495501441
0.788766376
dihexcer.d18.1.N22.0
-0.151207959
0.198344532
0.445851123
0.788766376
dihexcer.d18.1.N24.0
-0.19159016
0.183307383
0.295937391
0.671801284
dihexcer.d18.1.N24.1
-0.229686322
0.219603372
0.29560057
0.671801284
GD1.d18.1.N16.0
0.056433417
0.153603098
0.713322427
0.848607715
GD1.d18.1.N18.0
0.172896933
0.17820124
0.331929778
0.70686584
GD1.d18.1.N20.0
-0.152899782
0.119428321
0.200452298
0.627834194
GD3.d18.1.N18.0
0.010317161
0.108360075
0.924146571
0.979910154
GM1.d18.1.N16.0
-0.378436388
0.202410262
0.061532409
0.585412059
GM1.d18.1.N18.0
0.022770386
0.147002363
0.876901653
0.979910154
GM1.d18.1.N18.1
-0.101526645
0.086101773
0.238339658
0.627834194
GM2.d18.1.N16.0
0.00732513
0.077089444
0.924298006
0.979910154
GM2.d18.1.N18.0
0.070773176
0.129825064
0.585655429
0.788766376
GM3.d18.1.N16.0
-0.131795564
0.236729924
0.577709281
0.788766376
GM3.d18.1.N18.0
-0.018550958
0.233530009
0.936684922
0.979910154
GM3.d18.1.N18.1
-0.351200463
0.19393772
0.070157414
0.585412059
GM3.d18.1.N20.0
0.107572474
0.211707686
0.61137063
0.788766376
GM3.d18.1.N22.0
0.069437207
0.159987073
0.664275652
0.818482499
GQ1.d18.1.N18.0
0.134963722
0.242164649
0.577307348
0.788766376
GQ1.d18.1.N20.0
0.133246648
0.203327201
0.51225457
0.788766376
monohexcer.d18.1.N16.0
-0.2476466
0.171624447
0.149032744
0.627834194
monohexcer.d18.1.N18.0
-0.314667883
0.192323606
0.101810793
0.585412059
monohexcer.d18.1.N20.0
-0.282605422
0.172334889
0.10103388
0.585412059
monohexcer.d18.1.N22.0
-0.189406591
0.154430025
0.220015043
0.627834194
monohexcer.d18.1.N24.0
-0.367136597
0.163224438
0.024495025
0.585412059
monohexcer.d18.1.N24.1
-0.25604972
0.174712268
0.142770677
0.627834194
monohexsph
-0.12279479
0.318797536
0.700103235
0.847493389
OH.dihexcer.d18.1.N16.0
-0.222641677
0.183428259
0.224831487
0.627834194
OH.dihexcer.d18.1.N18.0
-0.422681048
0.199399182
0.034025174
0.585412059
OH.monohexcer.d18.1.N16.0
-0.112617737
0.16468165
0.494069158
0.788766376
OH.monohexcer.d18.1.N18.0
0.397888887
0.437200022
0.36277804
0.70686584
OH.monohexcer.d18.1.N20.0
-0.228812118
0.127415938
0.072528109
0.585412059
OH.monohexcer.d18.1.N22.0
-0.289460678
0.15551804
0.062706406
0.585412059
OH.monohexcer.d18.1.N24.0
-0.214031548
0.128427627
0.095603165
0.585412059
S..1.P
-0.201254176
0.154748799
0.193422212
0.627834194
S.1.P
0.013820461
0.2272561
0.951506961
0.979910154
SM.d18.0.N16.0
-0.237333575
0.302018226
0.431969819
0.784366251
Copyright © 2014 by the American Thoracic Society
P (FDR)
Page 59 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
SM.d18.0.N17.0
0.570050362
0.990359062
0.564885796
0.788766376
SM.d18.1.N14.0
0.015272316
0.163624398
0.925635325
0.979910154
SM.d18.1.N14.1
-0.094914727
0.164472692
0.563882023
0.788766376
SM.d18.1.N16.0
-0.283663169
0.296391463
0.338539091
0.70686584
SM.d18.1.N16.1
-0.159449366
0.198973497
0.422923367
0.784366251
SM.d18.1.N18.0
-0.479178234
0.346354254
0.166514152
0.627834194
SM.d18.1.N18.1
-0.554146411
0.227099822
0.014683173
0.585412059
SM.d18.1.N20.0
0.10606536
0.208298573
0.610612964
0.788766376
SM.d18.1.N22.0
-0.044634242
0.194629744
0.818613224
0.941405207
SM.d18.1.N23.0
-0.229612041
0.191328344
0.230102748
0.627834194
SM.d18.1.N24.0
-0.000308339
0.185081265
0.998670752
0.998670752
SM.d18.1.N24.1
-0.106206302
0.20083184
0.596922003
0.788766376
SM.d18.1.N25.1
-0.263709809
0.17779564
0.138016328
0.627834194
sphinganine
-0.230435303
0.180443734
0.201585316
0.627834194
sphingosine
-0.284649078
0.151819019
0.060803221
0.585412059
sulfa.d18.1.N16.0
-0.196541194
0.169297727
0.24567425
0.627834194
sulfa.d18.1.N20.0
-0.179187087
0.141882163
0.206614727
0.627834194
sulfa.d18.1.N22.0
0.151593282
0.205176937
0.460003111
0.788766376
trihexcer.d18.1.N16.0
-0.319338612
0.234296627
0.172892856
0.627834194
trihexcer.d18.1.N18.0
-0.180414721
0.200745362
0.368799569
0.70686584
trihexcer.d18.1.N22.0
-0.1308613
0.190616659
0.492388062
0.788766376
trihexcer.d18.1.N24.0 -0.208881009 0.171650256 0.223642682 0.627834194 Exacerbation modeling (negative binomial regression) included covariates (see methods for details); green indicates P < 0.01 with false discovery rate (FDR) adjustment < 0.05 and yellow indicates all others with P < 0.05
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 60 of 64
Supplemental Table 13: Plasma sphingolipid metabolites that correlate with PBMC gene expression Metabolite
Gene
ρ
P
FDR
N-(tetradecanoyl)-sphing-4-enine-1-(2-aminoethylphosphonate)
GM2A
-0.39
1.28E-05
0.043317393
SM(d18:1/12:0)
NEU1
-0.36
6.82E-05
0.1153031
CLN8
-0.34
2.08E-04
0.02697507
st
Sphingosines (1 principle component)
Copyright © 2014 by the American Thoracic Society
Page 61 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Figure 1. Correlations among measured sphingolipid species. Pairwise correlations between each sphingolipid species were calculated with green representing positive correlations and red representing negative correlations. Boxes indicate grouping of sphingolipid analysis used for principle component analysis.
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
A
Page 62 of 64
B
Supplemental Figure 2. Correlation between plasma sphingolipid metabolites and PBMC gene expression among all subjects in the untargeted cohort. (A) Correlations between SM(d18:1/12:0) and Neu1. (B) Correlations between N(tetradecanoyl)-sphing-4-enine-1-(2-aminoethylphosphonate) and GM2A. Lipid structure from LipidMaps (http://www.lipidmaps.org/) accessed on September 17, 2014.
Copyright © 2014 by the American Thoracic Society
Page 63 of 64
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Supplemental Figure 3: Sphingolipids metabolo-genomic pathway patterns associated with COPD subphenotypes (A) Chronic Bronchitis and (B) Airflow Limitation (FEV1 and FEV1/FVCDetailed sphingolipid metabolic pathway with the major ceramide synthetic pathways noted as sphingomyelinase-, de novo-, and recycling pathways (abbreviations list in the text). Individual or classes of metabolites measured in the study are identified as precursors and/or products relative to
Copyright © 2014 by the American Thoracic Society
AJRCCM Articles in Press. Published on 10-December-2014 as 10.1164/rccm.201410-1771OC
Page 64 of 64
synthetic or catabolic reactions with directionality marked by black arrows. Catalytic enzymes are noted next to the middle of the reactions arrow. Circled are enzymes whose genes were significantly related with COPD phenotypes in genomic analyses. In color are classes (arrowheads) or species (arrows) of metabolites that were associated with respective (color-coded) COPD phenotypes. Solid arrows indicate: 16-18 carbon chain-; dotted arrows 24-26 carbon chain-; single dash arrows: 14 carbon chain- sphingolipids. The direction of the arrowhead or arrow marks the direction of association (i.e. up: increased; down: decreased). Closed headed arrows represent changes with significance levels P < 0.05. Open headed arrows represent changes with significance levels P < 0.10. Superimposed on the sphingolipid metabolic pathways are deducted directions of sphingolipid metabolic flux, based on the results of the metabolo-genomic study, by disease phenotype, as noted.
Copyright © 2014 by the American Thoracic Society