What did we learn from multiple omics studies in asthma?

2.1 Genomewide association studies (GWAS) and candidate-gene studies failed to provide a “magic bullet” gene list that provoke asthma or explain treatment response

Genomics studies genetic variants of organisms in order to find associations and functional relevance of these variants with phenotypes. Since high-throughput technologies are becoming relatively cheap, efficient and widespread, there have been many discoveries in the field of genomics. These findings are predominantly correlational due to many factors involved in the phenotypic manifestations and our limited knowledge on cellular and molecular processes. Moreover, most of the studies are observational and in some cases underpowered.16, 17 Still, genetic association studies help gain insights into the underlying molecular mechanisms of the diseases (or subphenotypes).

Candidate-gene (CG) studies and genomewide association studies (GWASs) are the two main approaches to study genome-phenotype associations. GWAS represents an unbiased approach for finding single nucleotide polymorphism (SNP) associations with specific phenotypes and may result in novel genes to consider for a phenotype. In contrast, CG studies use a priori information from different sources to identify the association.18 GWAS and CG can organically be combined: the list of SNPs found in GWAS can serve as a prior in CG study. GWAS acquired popularity in asthma research over the last decade as it helped to overcome the lack of reproducibility of previous CG studies.19, 20 However, for complex human diseases, GWASs often result in nonintersecting outcomes and small effect size (typical OR is 1.1-1.520, 21). Some of the major critiques to GWAS are inability to take into account gene-gene interaction (ie epistatic effects), and the requirement of large patient numbers, which is not always possible. A recent GWAS with the largest number of patients to date featuring approximately 5000 patients (5135 cases and 25 675 nonasthmatics for stage 1; 5414 cases and 21 471 nonasthmatics for stage 2) found 21 signals (out of 24 in total) intersecting with previous studies.22

For complex human diseases like asthma, the expectation to find just one or several genes for disease susceptibility seems biologically and mathematically unrealistic23 (see Figure 1). Genome studies often explain only a fraction of heritability of complex human diseases.24-26 Despite the efforts to create a well-reproducible list of SNPs/genes for asthma susceptibility throughout the years, this goal was not reached.

Even though the results of GWAS in asthma often do not intersect, one prominent result demonstrated exception from the rule: the 17q12-21 locus. 17q12 was initially found to be associated with asthma in the first GWAS on asthma,27 was reproduced many times,22, 28-33 and further enlarged to 17q12-21. The expression of three genes (ORMDL3, GSDMB, PGAP3) was found to be connected to this locus,34 but their exact roles in asthma just started to be elucidated.35

A clinically relevant phenotype being investigated in GWAS studies is the response to medication: short-acting beta agonists (SABA), long-acting beta agonists (LABA), inhaled corticosteroids (ICS) or leukotriene modifiers (LTM). Most of the reported genetic variants do not reach the significance threshold and rarely intersect in their reported variants.36 Bronchodilator response in GWAS (measured in % of change in lung function) after SABA is estimated to be 10%-29% heritable,37, 38 with genes including ADRB2 (beta-2 agonists receptor).39 ADRB2 codes for a receptor which is the pharmacological target for short- and long-acting beta agonists (SABA and LABA, respectively). ADRB2 has mixed results across studies, some claiming that variants of ADRB2 are responsible for poor LABA response, some claiming no significant difference can be found.39-41 ADRB2 has a more stable association with LABA response in paediatric asthma,42 and the currently ongoing PUFFIN clinical trial43 aims to investigate whether prospective genotyping of ADRB2 variation can be used for a better treatment outcome.43 For ICS response, three loci within the genes GLCCI1, CRHR1 and FCER2 were replicated in at least one study.44 For leukotriene receptor agonist response, ALOX5 and MRP1 were found to be associated, and ALOX5 was replicated positively in independent populations.45 For a further detailed overview on GWAS studies in asthma, one may consult following references.36, 45-49

We emphasize that at the moment both genotyping of 17q12-21 and ADRB2 are not used in clinical practice; however, research on these continues, and this might lead to clinical implementation in the future. More information on these two discoveries can be found in the “Major milestones discoveries” panel.

To sum up, various genetic variants were identified to be associated with asthma and treatment response. The fact that associations are often not replicated might be due to low patient numbers, different inclusion criteria, differences in asthma definition, differences in disease severity, intrinsic heterogeneity and different outcomes measured. Classical asthma definition on the basis of clinical phenotypes is still being actively used in clinical trials. However, one could seriously question the present clinical gold standard. So, an ouroboros-like problem arises: heterogeneity in cohorts hampers delineation of the molecular subphenotypes, and as molecular subphenotypes are unknown in advance, we cannot improve inclusion criteria for enhancing subphenotype signals. These unresolved issues intercept us from using genomics in routine asthma care.

2.2 Transcriptomics provides a list of plausible biomarkers for asthma subphenotypes

Transcriptome is defined by all RNA transcripts in the cell at a given time and under a specific condition. Transcriptomic approaches are applied widely in asthma. Amongst available methods (RNA-seq, microarrays), microarrays were mostly utilized in asthma research due to low cost. Microarrays allow to detect a fraction of all mRNA transcripts that is predefined by probes; RNA-seq is a newer technology, and it can collect transcripts in an unbiased way with high accuracy.

Asthma subphenotypes on immune pathways (Th2-high/Th2-low), previously identified on the basis of cell counts in sputum,50 was also established in a seminal microarray study on bronchial brushings.51 As expected, Th2-high subphenotype is connected to atopic profile and airway remodelling and showed better corticosteroids response compared to Th2-low.

The U-BIOPRED52 consortium collected a large-scale data set on asthma patients. Analysis of transcriptomics in collected blood identified a vast amount (1693) of differentially expressed genes for asthma in comparison with control (nonasthmatics).53 Analysis of sputum transcriptomics from U-BIOPRED showed different pathobiological profiles in four clinical clusters (see “Major Discoveries” panel). Furthermore, it appeared that clinical phenotypes such as adult-onset asthma or asthma with fixed airflow limitation are characterized with distinct transcriptomic profiles.54, 55 However, such differential gene expression varied between sample sites (sputum, nasal brush, endobronchial brush and endobronchial biopsy), highlighting the role of local biology in asthma.

Unbiased sputum transcriptomics delineated three distinct gene expression profiles that were only partly associated with type 2 inflammation,56 suggesting that gene expression profiling provides complementary information to traditional biological concepts on asthma phenotyping. Interestingly, the results from sputum transcriptomics caused renaming Th2-high profile to type 2 gene mean (T2GM), characterized by three genes (IL-4, IL-5 and IL-13).57 Another study provided a confirmation for this in SARP cohort.58, 59 They also demonstrated that in severe asthma corticosteroids do not significantly decrease the T2GM profile in sputum; they suggested to add extra medication like inhibitors of type 2 cytokines for these patients. These findings demonstrate the potential of omics studies in personalized medicine approach.

Overall, transcriptomics provided confirmation for the necessity to subphenotype asthma. It has been demonstrated that transcriptomic level is not completely reflected on proteomics level.60, 61 This may signal that neither of these technologies provide highly accurate results or that more factors should be considered in our search of “missing heritability” of asthma and explaining disease processes.

2.3 Epigenomics may hold keys to a missing genotype-phenotype link

A combination of environmental factors including allergens (ie pet exposure), smoking and city pollution have been confirmed to influence asthma onset.62-64 This implies that genetic variants cannot serve as a diagnostic/prognostic tool by themselves; therefore, the combination of environmental exposures and genetic predisposition is important to study. An intricate question of how the environmental effect changes the genetics' manifestation can be answered by epigenetics.

Epigenetics is broadly defined as a nuclear information that complements genomics and can alter gene expression. Epigenomics includes information about DNA methylation, histone modifications and various forms of noncoding RNA. Epigenetics, especially methylation, can be inherited across generations. Methylation is the most popular type of epigenetics studied nowadays. Epigenetics depends on the exposome (all internal and external exposures throughout life65) and can be highly fluctuating; one may consider epigenetics as one of the adaptive cellular mechanisms, immune system in particular, in response to a changing environment. Epigenome-wide association studies (EWASs), similar to GWAS, investigate association of epigenomics with asthma, its severity and drug response.

Epigenomics shows a high potential for prenatal and perinatal asthma routes. Skewness of immune system to Th2 pathways has been one of the hypotheses for early-onset asthma,66 and studies established that epigenetics controls differentiation of T cells.67, 68 A meta-EWAS with 8 cohorts (668 cases) found many differentially methylated sites associated with asthma-related immune response.69 Also, early-onset asthma was connected to prenatal and perinatal exposome such as maternal smoking and diet.70, 71 In a large-scale EWAS, methylation in 14 CpG sites (connected to activation of eosinophils and cytotoxic T cells) was associated with childhood asthma.72

Microbiome can be considered as another exposure factor on epigenomic level as it is now believed to drastically influence a host organisms' health.73-80 The understanding of this influence is still in its infancy, but it shows a big potential. Huang et al81 demonstrated significant association between bacteria abundance and their diversity in airways with bronchial hyperresponsiveness. Also, specific microbiota in severe asthma patients were associated with clinical disease features.82 The fact that obesity sometimes co-occurs with asthma is another example of a possible association between asthma and the microbiome.83-86 Obesity is potentially connected to a different gut microbiome composition in comparison with normal weight subjects,87, 88 which might lead to a systemic changes and occurrence or aggravation of asthma symptoms.84-86

Epigenomics acquired some scepticism in asthma research, mainly due to the expected limited effect, lack of reproducibility, similar to genomics.89-92 EWASs are also criticized for merely reflecting cell counts, since epigenetics highly depends on a tissue type. In addition, methylation assays currently cover only a small part of all CpG islands (1.6%-2.9%93). Finally, the reported differentially methylation sites often do not show any connectivity to asthma-related genes.94

A question of tissue and cell type relevance has major importance for epigenomics, since epigenomics signatures greatly vary in different tissue types.95 Until now, most of the studies in asthma analysed epigenome in peripheral blood cells, while some used nasal epithelium and lung tissue.96 Signals from eosinophils and regulatory T cells (Treg) were found to be associated with asthma in EWAS.72, 97, 98 For allergic asthma, immune blood cells were found to be more relevant than for the nonallergic counterpart.96 Nasal epithelium has been found to be a good proxy for bronchial tissue in a recent study.99, 100 A novel approach—single-cell omics—is rising in popularity now and would likely help clarify cell types importance.101, 102 Epigenomic alterations at single-cell level will likely enable studying the cell-specific signal connected to disease such as asthma.103-106

Overall, epigenomics may be a missing link between the genotype and phenotype. To use epigenomics widely, better technology should be developed and sample collections and analyses have to be standardized.107, 108 Translational perspectives on epigenomics include diagnostics and prognostics, subphenotypes determination, prevention and treatment.

3.1 Promise of proteomics: subphenotype asthma patients

Proteomics is a research area that focusses on the large-scale study of proteins produced by cells, tissues and organisms. Analytical techniques used in this field of research identify, qualify and quantify proteins and their functions.109-113 Proteins play fundamental roles in fulfilment of biological functions in cells. Therefore, their expression levels can be considered as momentary reflections of the cellular state. Despite the valuable contribution of genomics and transcriptomics studies to asthma research, it should be noted that a single gene or mRNA strand can generate several different protein isoforms as a result of alternative splicing of RNA and post-translational modification of synthesized proteins. Accordingly, even though the human genome comprises approximately 30,000 genes, cells contain multiples of this amount in proteins.114-121

In respiratory research, proteomics can be used to identify biomarkers that determine disease in an early state, allow patient risk stratification, determination of disease progression, personalization of therapy regimens and/or assess therapeutic response.122, 123 Currently, FeNO is recommended in some guidelines as an indirect measurement of eosinophilic inflammation in patients124; however, classical asthma diagnosis according to the GINA guidelines1 does not consider any markers of inflammation or proteins. Proteomics may offer a more sensitive and specific diagnosis based on molecular and disease-specific protein pathways.

Different types of samples are suitable for proteomics analytical techniques. In respiratory research, proteomics have previously been applied on different biological samples, such as serum, circulating cells, bronchoalveolar lavage fluid (BALF), nasal lavage fluid (NLF), induced sputum, exhaled breath condensate (EBC), epithelial lining fluid (ELF) and, albeit sporadically, biopsies (see Table 1). However, data from a single sampling location most likely will provide insufficient information about the complete respiratory proteome. For example, BALF provides an impression of the proteome status of the epithelial lining fluid, but does not give complete insight into the changes of the proteome that occur in the walls of the airways.122

Whereas in genomics and transcriptomics studies, we can sequence and measure expression level comprehensively, proteomic studies are hardly performed at a similar scale as result of the diverse biochemical properties amongst proteins. Different methods of detection are used in proteomics, which can roughly be divided into immunoassays (Western blot, immunohistochemistry and ELISA) and mass spectrometry (MS). The former appeared earlier and is suitable for detection of small portion of proteins,125 whereas the latter aims to collect fractions of the whole proteome.

Before detection can occur, a targeted and complex separation of the protein mixtures is required.126-131 Initially, proteomic analyses were performed using two-dimensional gel electrophoresis.132-135 Nonetheless, these methods are generally unsuitable for the analysis of diverse complex mixtures of proteins and are strongly influenced by differences in technical procedures.136-139 Hence, liquid chromatography (LC)-based separation methods have been developed, which are usually followed by mass spectrometry.140 Both techniques are combined in shotgun proteomics, wherein generated spectra of peptide fragments are used for the identification of proteins with the aid of spectral databases.122, 125 Another popular high-throughput technique is based on microarrays—an emerging tool in proteomics that will make it possible to analyse large quantities of different proteins.128, 129, 141, 142

Proteomics techniques are increasingly used in asthma research for mapping the respiratory proteome. Large-scale registers for asthma patients, in which clinical data and biological materials from large numbers of asthma patients have been collected, have delivered valuable findings for the asthma research field using proteomics. The SARP research group was able to classify patients on asthma severity based on protein expression levels of 18 cytokines in BALF samples.143 Moreover, this research group was able to demonstrate differences in inflammatory mediators in sputum between granulocytic inflammatory profiles of asthmatic patients and asthma severity.144-146 More recently, researchers from U-BIOPRED were able to form four clusters based on clinical characteristics of asthma patients. Subsequently, significant differences in the expression levels of ten proteins in sputum could be demonstrated amongst previously established clinical clusters.147 Furthermore, current-smoking and ex-smoking severe asthmatic patients could be distinguished from nonsmoking patients in U-BIOPRED at the sputum proteomic level. In addition, significant differences were demonstrated in gene expression profiles between bronchial epithelial cells from current-smoking severe asthmatics and nonsmoking severe asthma patients as determined by pathway analysis, gene set variation analysis and protein–protein interaction analysis.148 Moreover, multiple molecular subphenotypes of eosinophil- and neutrophil-mediated asthma have very recently been identified on the basis of sputum proteomic signatures.149

3.2 Metabolites are potential surrogates of current and (previous) cell processes. Breathomics tries to bridge the gap between bench and bedside

Metabolomics focusses on analysing low-molecular biochemical compounds (molecular weight < 1500 Da) that originate from the metabolism. Metabolites are strongly implicated in homeostasis and disorders where they help maintain the redox balance, participate in oxidative stress, cellular signalling, apoptosis and inflammatory processes, etc The metabolic state is the result of both gene expression and environmental factors and can therefore be informative for disorders of multifactorial nature, such as asthma.150, 151 The composition of metabolites and/or their fragments can possibly offer a pathophysiological reflection of the current state of the disorder. As inflammatory mediators usually have a short half-life, they are rapidly degraded to various metabolites. Potentially, the analysis of these metabolites allows determination of previous cellular responses involved in inflammatory processes. If a metabolic composition is specific to a particular disease, it could potentially be used as a biomarker or to broaden knowledge on the pathophysiology of a disorder.152-156

The development of metabolomics occurred approximately simultaneously with the development of the proteomic research field. Despite the enormous difference in size between the compounds of interest in these research disciplines, both fields share many of their analytical techniques: sample separation based on gas or liquid chromatography and analysis via MS Similar to the proteome, the metabolome is dynamic and is highly influenced by factors such as BMI, treatment, diet and smoking status.157-159 In addition to mass spectrometry, nuclear magnetic resonance (NMR) spectroscopy is regularly applied in metabolomic analyses. However, this analytical method features much lower sensitivity and specificity compared to MS-based methods. Consequently, high concentrations of analytes would be required for valid analysis.152, 160-162 Metabolomic studies predominantly experience the same issues as proteomic studies, such as very limited sample sizes, lack of standardization for sample collection, handling and storage as well as a very wide dynamic range found in samples.122, 163

Most metabolomic studies to date have focussed on distinguishing asthma patients from healthy controls, COPD patients and amongst each other on the basis of phenotypes such as asthma severity. The main metabolites found in these studies are involved in tricarboxylic acid metabolism, hypermethylation, phospholipid regulation, hypoxia, oxidative stress and immune reactions.150, 152, 164-178 However, most studies were severely limited by low sample size, diagnostic heterogeneity and very limited number of metabolites that was studied. Furthermore, results are very rarely independently replicated and, due to a lack of standardization of the analytical methodology, technical inconsistencies between studies complicate comparisons of outcomes.

An emerging derivative of metabolomics is breathomics, wherein either gas chromatography/mass spectrometry (GC/MS) or electronic noses measure volatile organic compounds (VOC) in exhaled air.179 The first technology is based on the principles of standard analytical chemistry, whereas the second is based on cross-reactive sensor technology with the subsequent application of powerful pattern recognition algorithms.180 The first results with eNoses are promising; recent studies have demonstrated clusters of asthma and COPD patients with specific inflammatory phenotypes (eosinophilic or neutrophilic) on the basis of specific VOC profiles in exhaled air.181-185 With proper validation and proven clinical utility, molecular profiling of VOCs may provide a noninvasive alternative to blood and sputum measurements. Moreover, the speed of assessments, the small size of devices and the possibility to link some eNoses to existing spirometry equipment make it highly applicable in clinical practice.186 However, it should be emphasized that the detection method does not allow to delineate the contributing individual compounds in exhaled air. For gaining mechanistic insight, simultaneous analysis of breath samples by mass spectrometry is required.187