Discussion
A major gap in knowledge is the limited understanding of why PLWH are more susceptible to developing COPD independent of smoking status.1 ,2 ,5 ,10 Currently, there is no biomarker to identify risk or lend insight into the mechanisms of developing rapid decline in lung function and subsequent COPD in PLWH. The pulmonary substudy of START offers a unique opportunity to identify biomarkers of HIV-COPD prior to the development of an AIDS-defining illness. This is important since pulmonary infection itself can lead to a loss of lung function.24 ,25 In this exploratory study using an untargeted metabolomic approach, we identified 172 unique plasma analytes that associated with HIV-COPD compared with participants with HIV without COPD after controlling for multiple confounders such as smoking status, race and sex.
We performed LC-MS/MS to identify metabolites of interest and found that 17 of these metabolites were lipids. Among these lipids were diacylglycerol and several sphingolipids. Sphingolipids serve as a structural component of the plasma membrane lipid bilayer and also participate in cell recognition and signalling. In individuals with COPD, increased concentrations of sphingolipids have been measured in sputum. Telenga et al26 identified 168 sphingolipids and 36 phosphatidylethanolamine lipids in the sputum that were significantly higher in smokers with COPD compared with smokers without COPD. These lipids correlated with reduced lung function, sputum inflammation and smoking history. In our study, we found a complex expression of sphingolipids with both elevated and decreased plasma concentrations of various sphingolipids. This complex plasma expression of sphingolipids was observed by Bowler et al, where they used both targeted and untargeted metabolomic platforms that identified 23 sphingolipids common to both platforms in COPD. Some of these lipids correlated to COPD phenotype with five associating with emphysema and seven associating with COPD exacerbations.26 ,27 In addition, sphingolipid expression was complex depending on lipid type and COPD phenotype. Our data support these findings in individuals with HIV, as we found that sphingolipids concentrations were altered in the setting of HIV-COPD when controlling for smoking status.
Among the lipids, we also identified ceramide. Although ceramide concentrations were higher in HIV-COPD, in this relatively small exploratory study, the difference did not quite meet statistical significance (q-value=0.11). Ceramide consists of a sphingolipid backbone (sphingosine) plus a fatty acid. It has been implicated in the pathogenesis of COPD. Petrache et al28 found that ceramide acts as a second messenger and a crucial mediator of apoptosis and thus alveolar destruction in a murine model of emphysema. In human studies, elevated concentrations of ceramide have been found in sputum and lung tissue.26 ,28 ,29 Bowler et al27 further characterised plasma lipids in various COPD phenotypes and found that ceramides had a negative association with emphysema as measured by quantitative high-resolution CT (HRCT); however, trihexosylceramides had a positive association with frequent exacerbators. We also found an elevation of diacylglycerol, a protein kinase C activator that cooperates with ceramide in inducing apoptosis,30 but we did not have additional studies, such as HRCT, for further phenotyping.
Lipid dysregulation is a known complication of AIDS and is a well-known side effect of protease inhibitors. Chetwynd et al31 found that ART resulted in reductions in certain urinary metabolites, including sphingamines. Our study is significant as all of the START participants were ART-naïve and prior to the onset of AIDS. Therefore, the differences in lipid profiles between those with and without COPD cannot be attributed to ART. In addition, lipid abnormalities have been linked to markers of inflammation in HIV compared with healthy controls; however, all of the participants were on HAART and it is unclear whether the lipid abnormalities are due to HIV infection versus ART.32 Pertinent to our study, Scarpelini et al33 identified metabolites that associated with rapid progression of HIV infection and these included sphingomyelin metabolism. It is possible that lipid dysregulation in HIV infection is a marker of worsened disease, including the development of COPD.
In addition to lipids, we identified an analyte consistent with Trp that trended towards an association with HIV-COPD in our metabolite profiling. Trp and its main catabolic enzyme, IDO, have been associated with both COPD and HIV infection.18 ,20 Therefore, we used targeted SRM to measure Trp and its major metabolite, Kyn, to determine if the Trp catabolism is a biomarker for HIV-COPD. Consistent with previous reports, we found decreased Trp and increased Kyn concentrations in individuals with HIV. As a reflection of IDO activity, the Kyn/Trp ratio was significantly elevated in the setting of HIV, but not in COPD participants. Although not statistically significant, we did find that Kyn concentrations were higher than one would predict in those with both COPD and HIV (data not shown). One possibility is that the effects of HIV on IDO activation are much larger than those of COPD, therefore requiring a larger cohort to determine if IDO activation plays a role in HIV-COPD.
There are several limitations to this study. While HIV-positive controls were matched on age, sex, race, region and smoking status to our cases, our non-HIV controls were difficult to match on age and race as our HIV-COPD cases were much younger and over half were black. In this study, we found that metabolites associated with HIV-COPD were not associated with COPD in HIV-negative controls. This may be due to confounders, such as age and race; however, COPD in HIV-negative individuals is most likely a very different phenotype and disease. Therefore, diagnosis and treatment may need to be very different for HIV-COPD and these metabolites have the potential to identify those with HIV who are at risk of developing COPD. This study was relatively small and may have been underpowered to determine all significant biomarkers and metabolomics pathways of HIV-COPD. Future studies would benefit from larger sample sizes and longitudinal study designs such as that recently completed and now possible in the START Pulmonary Substudy.