Lipid Biomarkers for Atopic Dermatitis

2019-06-10T18:43:17Z (GMT) by Jackeline Franco

Atopic dermatitis (AD) is a common pruritic skin disease in people and domestic animals that can be severely debilitating and stressful to the patient and the caregiver. The diagnosis of AD requires time consuming and expensive procedures, and treatment is often lifelong at considerable cost. Alterations in the lipid composition of the epidermis are a hallmark of the disease, and these may represent changes caused by the inflammation and defects in the lipid barrier. Liquid chromatography tandem mass spectrometry (LC-MS/MS) and, more recently, untargeted profiling using high-resolution time-of-flight instruments have been used to quantify the lipid composition in skin and other tissues, but these approaches are highly demanding in sample preparation and instrument time. In addition, these methods either detect only a limited number of lipids at the time or the identification of detected mass-to-charge ratio (m/z) is problematic when untargeted profiling is used. New lipidomic approaches that generate lipid profiles in a faster and more efficient manner can lead to a better understanding of these lipid changes.

The mass spectrometry analytical strategy used in this study, multiple reaction monitoring (MRM)-profiling, rapidly identifies discriminant lipids of the epidermis by flow injection. MRM-profiling is a small molecule accelerated discovery workflow performed in two parts using a triple quadrupole mass spectrometer with electrospray ionization as the ion source. Briefly, the first step consists of discovery experiments based on neutral loss and precursor ion scans to detect lipids in pooled samples by targeting class-specific chemical motifs such as polar heads of phospholipids or sphingoid bases of ceramides. The second step of the MRM-profiling is the screening of individual samples for the transitions detected in the discovery phase.

We first developed the experimental approach of the MRM-profiling methodology using epidermal samples of mice with AD-like inflammatory skin disease (chronic proliferative dermatitis, cpdm). Subsequently, we investigated lipid changes as the disease in mice progressed from minimal to severe. In order to select the most relevant ions, we utilized a two-tiered filter/wrapper feature-selection strategy. First, we built linear models linking the presence of every lipid monitored to disease stage information. The top 10 lipids, ranked based on η2 effect size, were used to build a predictive elastic-net (E-net) regression model linking the lipid ions detected by MRM-profiling with disease progression. The developed model accurately identified disease stages based on the variations in relative amounts of lipid ions corresponding to phosphatidylcholines, cholesterol esters, and glycerolipids-containing and eicosapentaenoic acid fatty acyl residues. Finally, we investigated the lipid profile of the epidermis in dogs with canine AD using the previously developed methodology. Epidermis from client owned patients and healthy controls were collected. Patients were sampled from affected and unaffected skin avoiding areas with secondary infections and the canine atopic dermatitis extent and severity index (CADESI-4) was recorded. The monitored lipids substantially separated the samples of healthy dogs from atopic dogs and distinguished the affected from the unaffected skin of patients. Samples were grouped into two cohorts for low-score and high-score CADESI-4, the first principal component was able to differentiate the control group from the low and high-score group. Differences in the lipid composition associated with low and high score CADESI-4 were significantly different only after separating the samples by sex of the dogs, demonstrating sexual dimorphism in the lipid changes associated with disease. The compositional data was feature extracted using the CADESI-4 to build linear models that identified oleic acid-containing triacylglycerides, long-chain acylcarnitines and sphingolipids as highly predictive lipids and were subsequently used to construct a predictive E-net regression. The lipid fingerprint obtained from the MRM-profiling was highly correlated (R2=0.89) with the classification of the standardized CADESI-4 score.

This research showed that changes in the lipid composition of the epidermis can be detected by MRM-profiling in atopic dogs even when the skin looks clinically healthy and that sex is a modifying factor in the lipid profile of canine atopic dermatitis (CAD). We expect that this research leads to a better understanding of the lipid changes in the epidermis during the onset of AD and as the chronic inflammatory process develops. The high prediction rate given by the lipid biomarkers for disease progression identified here by the machine learning strategy provides a potential molecular assessment tool for the diagnosis and monitoring of atopic dermatitis and the patient response to treatment.