Modeling the enigma of complex disease etiology

The clinical conditions (asthma, diabetes and fetal alcohol syndrome) were selected for testing the model as their etiology involved at least two factors, a genetic susceptibility and an environmental driver. Additionally, from the breadth of conditions considered, we selected those where members of our team have extensive knowledge; where sufficient and recently published literature was available and where the complexity of the condition was broad enough to test the model.

Modeling of complex disease evolved within the Human Disease Ontology project, building on the DO Clinician group's work over the past five years, in which we systematically reviewed and revised the DO’s classification system for syndromes, genetic diseases and physical disorders. This work involved the development of a system to enable differential diagnosis of complex genetic diseases, by redefining etiology classifications for non-monogenic diseases, to enable precise etiology definitions. For example, previously in the DO Prader-Willi was defined as a chromosomal disorder and the revised classification captures multiple possible etiologies of a ‘loss of function variant’ in combination with either maternal_uniparental_disomy, paternal_variant, chromosomal_deletion or chromosomal_translocation. Through this work, to improve the DO classification of genetic causes of disease it became clear the DO classification strategy did not have a way to rigorously capture and convey non-genetic (social, environmental, and other) factors pertinent to disease etiology.

The development of the Complex Disease model evolved through months of discussions among the Human Disease Ontology’s Clinical team (co-authors of this manuscript), as we were charged with considering how the Human Disease Ontology could contribute to a more in depth understanding of the complex factors involved. We established a list of common diseases involving varied and complex etiologies involving nuanced etiologies, including lysosomal storage diseases, Parkinson’s disease, fetal alcohol syndrome, autism, amyloidosis, diabetes, and asthma. Based on the group’s interests, areas of expertise and the likelihood of each example to seriously challenge the scope of the model, we selected to explore the three examples of diabetes, asthma and fetal alcohol syndrome. We intentionally picked challenging case-studies to be sure the resulting model would be robust.

An evolving understanding of asthma

The first step of defining asthma as a complex disease model was to examine current asthma classifications in DO and across clinical vocabularies (ICD, SNOMED CT, OMIM) [26,27,28], and to research literature to integrate start-of-the-art knowledge on genetic susceptibilities, environmental drivers, severity, endotypes and how researchers and clinicians are defining asthma subtypes (Fig. 2).

The complexities revealed by this representation in the DO informed the development of a modularized, structured model. We hypothesized this method could be used to model a streamlined, integrative approach for subtyping complex human diseases by defining diseases sharing similar molecular variant types, genetic susceptibilities and/or environmental drivers from authoritative clinical, genetic, and phenotype resources to identify diseases with common underlying etiology. This approach will enable researchers and clinicians to explore common, rare, and complex disease drivers across genetic diseases, syndromes, and cancer and to formulate testable hypotheses to examine mechanisms of pathogenesis (Fig. 4).

Fig. 4figure 4

Asthma classification. a Asthma classification before refactoring; b Refactored asthma classification. Including endotypes, expansion of subtypes

Modeling the heterogeneity of asthma has more recently evolved to encompass distinct pathophysiological, mechanistic pathways (endotypes) and variable clinical presentations (phenotypes) thus shifting therapy paradigms leading to precision medicine approaches [29,30,31]. Asthma has a constellation of phenotypes that can be associated with endotypes to guide clinical management.

Asthma endotypes

Asthma endotypes may be broadly regarded as type 2 (T2) high or Non T2-low. The phenotypes of T2 (high) endotypes include atopic, late onset, and aspirin exacerbated respiratory illness (AERD) and have defined clinical characteristics, molecular mechanisms, biomarkers and natural history. For example, atopic asthma is seen early, is sensitive to steroids, is molecularly associated with allergic sensitization, is associated with biomarkers including high IgE, is readily identifiable and is associated with preservation of lung function to complete the characteristic phenotype. Contrarily, Non T2 low endotypes include phenotypes such as non-atopic individuals, smokers, individuals with obesity related illnesses and the elderly. For example, for smokers, clinical characteristics would include older adults with a molecular basis of oxidative stress, biomarkers of induced sputum neutrophils, and a clinical course with more frequent exacerbations and lower lung function. This strategy to associate molecular mechanisms to phenotype and asthma endotypes allows us to describe distinct pathophysiologic mechanisms at a cellular and molecular level with implications for treatment and prognosis [30]. Integrating endotypes into current disease etiology modeling will incorporate the multifactorial genetic, environmental and pathophysiological mechanisms of disease causation [32].

DiabetesHistory of diabetes nomenclature

Reuse of clinical terms complicates disease nomenclature, as exemplified by the usage of highly similar names such as diabetes insipidus (DI) and diabetes mellitus (DM) to represent distinct disease entities associated with excessive urine output.

Historically, the two conditions were differentiated based on the work of Thomas Willis (1670 s) followed by Johann Peter Frank (1794) [33, 34]. While DM had already been identified as a disease in ancient Egypt, Greece, and Asia, DI was described several thousand years later. Thomas Willis first noted the sweet taste of urine from polyuric subjects compared with healthy subjects, leading to the differentiation of DM from the rare DI. Johann Peter Frank’s description of polyuric patients with not sweet urine led to the terminology of DI. The historical milestones identifying the different forms of DI evolved over time, beginning with the observation by DeLange in 1935 that some patients with DI did not respond to pituitary extract and thus that DI was nephrogenic in origin rather than central [33]. Subsequently, in 1947, Williams and Henry introduced the term “nephrogenic diabetes insipidus” for the congenital syndrome characterized by polyuria and renal concentrating defect but unaffected by vasopressin. Recognizing this important history warrants caution, noting that current medical usage of the word ‘diabetes’ is generally assumed to refer to disorders of glucose regulation. While usage of the word ‘diabetes’ in literature may refer to either DM or DI. Given the nomenclature history of diabetes [35], we additionally reviewed and updated the classification of DI to provide an up-to-date disease classification of both DM and DI.

Previous diabetes mellitus nomenclature revisions have included updates from ‘type I diabetes mellitus’ or ‘insulin-dependent diabetes mellitus’ to ‘type 1 diabetes mellitus’ and from: ‘type 2 diabetes’, non-insulin-dependent diabetes mellitus’, type II diabetes mellitus’ to ‘type 2 diabetes mellitus’. Molecular subtypes as defined by the Online Mendelian Inheritance in Man (OMIM) [27] were added to the DO for type 1 diabetes mellitus (DM1). OMIM type 2 diabetes mellitus (DM2) subtypes, which define susceptibility phenotypes, will be added to the DO when additional evidence of genetic association is defined in OMIM.

Recent re-evaluation of diabetes as a complex disease resulted in a DM2 reclassification (Fig. 5). The review identified 14 molecular subtypes of MODY (maturity-onset diabetes of the young, DOID:0050524) and the reclassification of ‘latent autoimmune diabetes in adults’ (DOID:0080846) as a subtype of ‘type 1 diabetes mellitus’. Review of DI (DOID:9409), a subtype of ‘kidney disease’ (DOID:557), identified four subtypes: central diabetes insipidus, nephrogenic diabetes insipidus, gestational diabetes insipidus, and dipsogenic diabetes insipidus.

Fig. 5figure 5

Diabetes mellitus reclassification. Showing the reclassification of diabetes mellitus following the recent review

Fetal alcohol syndrome as a test for the complex disease model

History of Disease. Fetal alcohol spectrum disorder (FASD (DOID:0050696)) is the name given to a constellation of signs and symptoms associated with prenatal ethanol exposure. Fetal Alcohol Syndrome (FAS) (DOID:0050665) is the most severe manifestation of FASD.

FAS was first described in 1968 in 127 children born to alcoholics in France [36]. It was more widely recognized following an article published in the Lancet in 1973 that described common features of 8 children born to alcoholics of three different ethnicities [37].

These children were born to women who were chronic alcoholics throughout pregnancy, so they were constantly exposed in utero. The manifestations of FAS in each case were similar, leading to the establishment of criteria needed to make the diagnosis. The criteria are: (1) documentation of growth deficits (weight, length, head circumference), (2) documentation of the following three facial features: smooth philtrum, thin upper lip, short palpebral fissures, and (3) documentation of central nervous system abnormalities which can be structural, neurological or functional) [38]. Maternal alcohol use can be either confirmed or unknown. Because the three criteria may be fulfilled at different stages of development, FAS is most frequently diagnosed at school age when behavioral problems are noted by a teacher. A FAS diagnosis can be accompanied by a variety of other structural and function deficits in other organs, including the cardiovascular system, genitourinary system, sensory systems (most notably auditory) and the autonomic nervous system.

The term, “Fetal Alcohol Syndrome (FAS)” was coined in a 1973 article in the Lancet by Smith and Jones to describe common physical features of children born to alcoholic mothers. The first recognition of a pattern of specific deficits seen in children of women who consumed alcohol heavily in pregnancy was made by Lemoine et al. in 1968. However, as this observation was published in French and did not give a name to the pattern, it was not recognized until the 1973 publication in Lancet. Initially, it was thought to be due to malnutrition but is now recognized as an effect of ethanol exposure. Because the maternal drinking in these initial cases was continuous and ongoing during pregnancy, new questions of dose and timing of exposure arose. New research then began to fully describe FAS, determine its mechanism (still unknown), and to determine the effects of dose and timing of dose to the outcome. Over the course of many years of research, a plethora of terms was introduced to describe the range of outcomes from a range of ethanol exposure patterns during pregnancy. These terms initially included fetal alcohol effects and alcohol related birth defects. In 1996, the Institute of Medicine developed a diagnostic nomenclature that included the terms of FAS, partial FAS, alcohol-related birth defects (ARBD) and alcohol-related neurodevelopmental disorder (ARND). The category, “Fetal alcohol effects” subsequently was phased out.

The diagnostic criteria for FAS according to the Institute of Medicine (now the National Academy of Medicine) is the presence of characteristic facial features (short palpebral fissures, a flat elongated philtrum, and a thin upper vermillion lip border), growth impairment (< 10th percentile weight) and central nervous system deficits (head circumference < 10th%, poor suck, weak cry, mental retardation). The diagnostic term of partial FAS is applied when facial features are present along with either growth deficits or physical central nervous system (CNS) deficits (e.g. microcephaly) or characteristic neurobehavioral problems are present. ARBD is used when congenital structural defects (cardiac, kidney, auditory) are present along with a history of maternal alcohol consumption modified in 2005 to include the requirement that facial features of FAS be present. The diagnosis of ARND requires evidence of either physical CNS deficits or neurobehavioral deficits similar to those in FAS are present along with a history of maternal alcohol use in pregnancy.

The term, now in wide usage, of Fetal Alcohol Spectrum Disorder (FASD), includes FAS, partial FAS, ARBD, ARND, and other outcomes thought to be related to alcohol use during pregnancy. There are no established definitive diagnostic criteria for diagnosing FASD. Interestingly, there are several ICD-10 codes for ethanol use/exposure during pregnancy: P04.3 Newborn affected by maternal use of alcohol; Q86.0 Fetal alcohol syndrome (dysmorphic); 099.31 alcohol use complicating pregnancy, childbirth, and the puerperium, alcohol use complicating pregnancy, unspecified trimester; 035.4XX0 maternal care for damage to fetus from alcohol, not applicable or unspecified. FASD is not an ICD-10 code. The lack of inclusion in coding systems impedes diagnosis and treatment.

Ethanol, the environmental driver. It became apparent that the diagnosis of FAS was not capturing all infants affected by maternal ethanol exposure. Manifestations could vary depending on the pattern of drinking (a woman who drinks 5 drinks per week may drink all 5 on one day (binge episode) or 1 drink per day for 5 days of a week. Both result in a 5 drinks/week dose, however, binge drinking is more harmful to the fetus. Various disease names were applied depending on the predominant effects: alcohol related birth defects (DOID:0050668), alcohol related neurodevelopmental disorder (DOID:0050667), partial fetal alcohol syndrome (DOID:0050666), neurobehavioral disorder with prenatal alcohol exposure (DOID:00810520). The diagnosis of FASD has been complicated, but recent progress has been made in defining how these diagnoses are to be made. Hoyme et al. published a diagnostic criteria for each of the syndromes for the clinical diagnosis of FAS and FASDs [39]. They have all now been placed under the term Fetal Alcohol Spectrum Disorder. The DO was augmented following this review, by the addition of neurobehavioral disorder with prenatal alcohol exposure (DOID:00810520), the addition of age of onset annotations and the annotation of alcohol as an environmental driver. Age of onset was added utilizing the Relations Ontology term: 'existence starts during' and onset as defined in the Human Phenotype Ontology, 'Pediatric onset', where the onset of disease manifestations occurs before adulthood, defined as before the age of 15 years (HP:0410280). The causal relationship between alcohol and FASD was defined with the addition of a new Relation Ontology term, ‘has disease driver’ (RO:0007001) and term ‘alcohol’. In addition to the factors known to influence development of FASD, there is growing evidence of other environmental, host, ‘social’ determinants of health, genetic and epigenetic contributing factors that need further studies to elucidate these relationships [39].

Other environmental drivers. In some studies, maternal smoking is always associated with FAS. Recent data suggests that dietary factors may also influence the impact of alcohol: both polyunsaturated fatty acids and choline have been shown to both prevent the effects of alcohol on the developing fetus, as well as to repair some of the damage. In a study from the Ukraine, a randomized controlled trial of choline supplementation to children who had the diagnosis of FAS/FASD resulted in a modest improvement in outcome [40, 41].

Other Host Factors. The stage of development has different effects on the fetus (first trimester only, third trimester only, all three trimesters, or only before recognition of pregnancy). Exposure during the first trimester is required for the facial dysmorphology that is required for a diagnosis of FAS. However, the brain remains vulnerable throughout pregnancy. In a sheep model, effects of second trimester only and third trimester only binge exposure has been described to lead to different effects on outcome [42].

“Social” Determinants of Health. Drinking and heavy drinking varies by ethnicity with rates higher in whites and Native Americans than in Asian Americans and Hispanics [43]. Socioeconomic status also influences drinking behavior amongst women and the health impacts of drinking in a complex way [44]. For the outcome of FASD, several studies based on single cities demonstrate the complexity of the interaction between social determinants of health and FAS/FASD. Late recognition of pregnancy and higher dose of consumed alcohol appear to be two common risk factors [45,46,47,48].

Genetics. Genetic influences must consider both the genetic makeup of the parents as well as those of the fetus. Genes influencing the development of FAS/FASD have mostly focused on maternal alcohol dehydrogenase and acetaldehyde dehydrogenase [49]. Other implicated pathways in the fetus include the retinoic acid pathway, sonic hedgehog and cholesterol homeostasis, nitric oxide synthase I, and platelet derived growth factor/mTOR pathways [49]. Recently, a twin study showed convincing evidence that genetics plays a part in FAS/FASD. The study showed decreasing concordance with decreasing genetic relatedness [50].

Epigenetics. Multiple genes in many pathways have been found to be epigenetically regulated by fetal ethanol exposure [51, 52]. Examples of the impact of alcohol induced alteration of gene expression are on the cortical thinning present in FAS/FASD [53] and hypothalamic-pituitary axis [54].

In summary, the model of complex disease for FAS/FASD makes several points clear, that there are multiple opportunities to intervene that may affect the impact of ethanol exposure to reduce the burden of FAS/FASD. However, the simplest solution might be to prevent prenatal alcohol exposure in the first place.

留言 (0)

沒有登入
gif