Plasma sphingolipids associated with chronic obstructive pulmonary disease phenotypes.

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]


Charmion Cruickshank

[email protected]


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]


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


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


Page 3 of 64


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



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



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



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


Page 6 of 64

Page 7 of 64


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



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



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



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



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



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



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



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


Page 15 of 64


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



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



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



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



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



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


Page 21 of 64

35.

36.

37.

38.

39.


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



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


Page 23 of 64


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


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


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


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


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



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


Page 27 of 64


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




Page 28 of 64

Page 29 of 64



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


0.0

0.2

0.4

0.6

False Positive Rate


0.8

1.0


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


18.5

19.0



Page 32 of 64


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



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-



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


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


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%).


Using


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.


Page 39 of 64


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.



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.


Page 41 of 64


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)



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


0.822241905

0.17723826

0.037514387

0.90311024

0.019972854

-0.11953695

2.107936011

0.122207584


0.731461618

0.027091909

0.259245898

0.542878063

0.030505369

-0.194626049

1.133684021

0.611756898


0.658890795

0.272803918

0.833380298

0.666323633

-0.039233983

-0.097270346

0.210850901

-0.433416799


0.398311105

0.049884263

0.368845113

0.577417763

-0.074991659

-0.173032034

0.902245219

-0.560848047


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


0.101864617

0.153376566

0.167179925

0.901199505

0.144680597

-0.126428063

-1.389681056

0.124852129


0.33437824

0.01792126

0.273660016

0.048462592

0.085670279

-0.208161542

1.100331843

2.005232809


0.202350257

0.037504021

0.335944379

0.065554078

0.112987226

-0.183383149

0.966577367

1.871257604


0.374722206

0.058274614

0.715847956

0.100697981

0.078799715

-0.1671764

0.364873069

1.660728657


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


0.122833146

0.274382322

0.633008127

0.316381234

0.136544645

-0.096952347

-0.478850145

1.010507255


0.38178452

0.031315192

0.649235214

0.331145647

0.07764479

-0.189686726

0.456107538

0.982138253


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


0.522832146

0.39651302

0.065350462

0.429880114

0.056766706

0.07527708

-1.859420604


0.013566142

0.531933455

0.010660494

0.244433013

0.21686156

0.055531245

2.605242277

1.178445845


0.634014341

0.901362302

0.147431239

0.894791439

0.042309924

0.011019286

1.458096559

0.132839269


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


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



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



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


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


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



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))


Page 46 of 64

Page 47 of 64


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


-0.451408235

0.215390391 0.036102799 0.130626252


-0.013216556

0.220977646 0.952307382 0.963849124


-0.133101417

0.228435103


-0.199223003

0.189637388 0.293466661


-0.332316325

0.168974279 0.049221486 0.130626252


-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


0.07615105 0.485781547 0.598552263

-0.47385556 0.155169851

0.62215377

0.00225973 0.031184276


-0.281621984

0.183177667 0.124189189


-0.207997277

0.161062046 0.196561244 0.299013237

0.21422635


-0.157341348

0.138276681 0.255173377 0.352139261


-0.263784007

0.152332667 0.083339238 0.163000836


-0.342051651

0.166632631 0.040098858 0.130626252

monohexsph

-0.614435537

0.308625071 0.046494062 0.130626252


0.111352836 0.179327955 0.534635669 0.636032089


-0.2516346 0.188995256 0.183046438 0.287050096


-0.340264846

0.153968639 0.027107797 0.116902375


-0.742965941

0.408879774 0.069205358 0.154037733


0.005411705 0.119400635 0.963849124 0.963849124


-0.021805815

0.146392237 0.881589362 0.907905462


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



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


-0.569061881

0.221519549 0.010202148 0.078216467


-0.397842639

0.192209876 0.038467842 0.130626252


-0.207245275


-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


Page 49 of 64


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



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


-0.057861702

1.159147464

0.96018816

0.974308574


1.863626957

0.799386247

0.019736373

0.104754595


2.353175503

0.752325656

0.001760791

0.01735637


1.707374916

0.507968849

0.000776101

0.008925165


1.022383852

0.345039865

0.003045684

0.022501416


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


-1.504409737

0.996906496

0.131278832

0.377426641


0.270590137

0.846924449

0.749349383

0.884801204


-0.204387753

0.810359744

0.800872147

0.891293196


-0.43265403

0.61471621

0.481540352

0.772759415


-1.098875012

0.821743972

0.181141885

0.44638536


-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


-1.445632488

0.976790181

0.138877886

0.383302964


1.153455522

2.145099011

0.590772661

0.843743546


-1.132208958

0.666540684

0.089388084

0.308388891


-0.540122887

0.670324857

0.420379585

0.743638983


0.749374888

0.73505072

0.30797167

0.643940765


-1.427179331

0.980319113

0.145439238

0.385973361


-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


Page 51 of 64


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


2.421679925

0.474207903

3.27678477610e-07

1.13049e-05


3.441030675

0.615560298

2.26966374069e-08

1.56606e-06


3.25645855

0.804369207

5.15551983384e-05

0.000711462


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



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


0.167830324

0.533559031

0.7531043

0.866069945


1.525704152

0.506253908

0.002580715

0.044517338


0.762751658

0.462403961

0.099038017

0.480866162


0.869383318

0.389708257

0.025690615

0.177265246


0.593181554

0.311099497

0.056556056

0.32519732


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


-3.57019E-05

0.388557925

0.999926688

0.999926688


-0.225477097

0.465532176

0.628142698

0.773961539


0.682964166

0.420738924

0.104536122

0.480866162


-0.177603156

0.400318576

0.657292363

0.78195126


0.407809376

0.406814223

0.316128133

0.573009805


-0.230456177

0.40010707

0.564624466

0.723031668

monohexsph

0.812248807

0.823342599

0.323875107

0.573009805


0.632121974

0.531137644

0.233995945

0.538190673


-0.692314823

0.494754176

0.161719826

0.498366346


-0.068699793

0.366222478

0.851197814

0.889888623


0.259319585

1.229552901

0.832960939

0.889888623


0.361093065

0.27414842

0.187790217

0.498366346


0.36443039

0.301092288

0.226140366

0.538058111


0.574545474

0.244729692

0.01889106

0.162935391


-0.436125798

0.358678591

0.224013991

0.538058111


-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


Page 52 of 64

Page 53 of 64


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


0.88044359

0.484757639

0.069331232

0.367988845


1.28135432

0.494935582

0.009627424

0.110715378


1.83590452

0.469162151

9.10978E-05

0.003142873


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



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


0.90

S.1.P + cov

0.90

dihexcer.d18.1.N20.0 + cov

0.90


0.90


0.90

trihexcer.d18.1.N16.0 + cov

0.91


0.89


0.89


0.89

sulfa.d18.1.N16.0 + cov

0.91

GD1.d18.1.N16.0 + cov

0.90

All sphingolipids above + cov

0.94


Page 55 of 64


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)


0.81

0.85


0.81

0.85


0.82

0.86


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



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


Page 56 of 64

Page 57 of 64


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



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


-0.390868101

0.228604708

0.087303049

0.585412059


-0.205742992

0.225164431

0.360850576

0.70686584


-0.158717319

0.232865061

0.495501441

0.788766376


-0.151207959

0.198344532

0.445851123

0.788766376


-0.19159016

0.183307383

0.295937391

0.671801284


-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


-0.2476466

0.171624447

0.149032744

0.627834194


-0.314667883

0.192323606

0.101810793

0.585412059


-0.282605422

0.172334889

0.10103388

0.585412059


-0.189406591

0.154430025

0.220015043

0.627834194


-0.367136597

0.163224438

0.024495025

0.585412059


-0.25604972

0.174712268

0.142770677

0.627834194

monohexsph

-0.12279479

0.318797536

0.700103235

0.847493389


-0.222641677

0.183428259

0.224831487

0.627834194


-0.422681048

0.199399182

0.034025174

0.585412059


-0.112617737

0.16468165

0.494069158

0.788766376


0.397888887

0.437200022

0.36277804

0.70686584


-0.228812118

0.127415938

0.072528109

0.585412059


-0.289460678

0.15551804

0.062706406

0.585412059


-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


P (FDR)

Page 59 of 64


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


-0.319338612

0.234296627

0.172892856

0.627834194


-0.180414721

0.200745362

0.368799569

0.70686584


-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



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)


Page 61 of 64


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.



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.


Page 63 of 64


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



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.


Plasma sphingolipids in HIV-associated chronic obstructive pulmonary disease.

Untargeted lipidomic analysis in chronic obstructive pulmonary disease. Uncovering sphingolipids.

Osteoporosis Associated with Chronic Obstructive Pulmonary Disease.

Chronic obstructive pulmonary disease phenotypes and balance impairment.

Chronic obstructive pulmonary disease phenotypes. Past, present, and future.

Phenotypes and disease characterization in chronic obstructive pulmonary disease. Toward the extinction of phenotypes?

Risk factors associated with chronic obstructive pulmonary disease early readmission.

Chronic obstructive pulmonary disease severity is associated with severe pneumonia.

Lumbar hernia associated with chronic obstructive pulmonary disease (COPD).

Chronic obstructive pulmonary disease.

Comorbidity associated with referral to pulmonary rehabilitation in people hospitalized with chronic obstructive pulmonary disease.

HIV Associated Chronic Obstructive Pulmonary Disease in Nigeria.

Invasive pulmonary aspergillosis in patients with chronic obstructive pulmonary disease.

[Invasive pulmonary aspergillosis in patients with chronic obstructive pulmonary disease].

Asthma-Chronic Obstructive Pulmonary Disease Overlap Syndrome Associated with Risk of Pulmonary Embolism.

Pulmonary hypertension associated with chronic obstructive lung disease and idiopathic pulmonary fibrosis.

IREB2 and GALC are associated with pulmonary artery enlargement in chronic obstructive pulmonary disease.

Persistent and Newly Developed Chronic Bronchitis Are Associated with Worse Outcomes in Chronic Obstructive Pulmonary Disease.

Integrated disease management for adults with chronic obstructive pulmonary disease.

Coronary artery disease concomitant with chronic obstructive pulmonary disease.

Small Airway Disease in Patients with Chronic Obstructive Pulmonary Disease.

The search for distinct and clinically useful phenotypes in chronic obstructive pulmonary disease.

[Usefulness of the Chronic Obstructive Pulmonary Disease Assessment Test in chronic obstructive pulmonary disease with severe airflow limitation].

Frequency and characteristics of different clinical phenotypes of chronic obstructive pulmonary disease.