Detecting transthyretin amyloid cardiomyopathy (ATTR-CM) using machine learning: an evaluation of the performance of an algorithm in a UK setting

OBJECTIVE

The aim of this study was to evaluate the potential real-world application of a machine learning (ML) algorithm, developed and trained on heart failure (HF) cohorts in the USA, to detect patients with undiagnosed wild type cardiac amyloidosis (ATTRwt) in the UK.

DESIGN

In this retrospective observational study, anonymised, linked primary and secondary care data (Clinical Practice Research Datalink GOLD and Hospital Episode Statistics, respectively, were used to identify patients diagnosed with HF between 2009 and 2018 in the UK. International Classification of Diseases (ICD)-10 clinical modification codes were matched to equivalent Read (primary care) and ICD-10 WHO (secondary care) diagnosis codes used in the UK. In the absence of specific Read or ICD-10 WHO codes for ATTRwt, two proxy case definitions (definitive and possible cases) based on the degree of confidence that the contributing codes defined true ATTRwt cases were created using ML.

PRIMARY OUTCOME MEASURE

Algorithm performance was evaluated primarily using the area under the receiver operating curve (AUROC) by comparing the actual versus algorithm predicted case definitions at varying sensitivities and specificities.

RESULTS

The algorithm demonstrated strongest predictive ability when a combination of primary care and secondary care data were used (AUROC: 0.84 in definitive cohort and 0.86 in possible cohort). For primary care or secondary care data alone, performance ranged from 0.68 to 0.78.

CONCLUSION

The ML algorithm, despite being developed in a US population, was effective at identifying patients that may have ATTRwt in a UK setting. Its potential use in research and clinical care to aid identification of patients with undiagnosed ATTRwt, possibly enabling earlier diagnosis in the disease pathway, should be investigated.

Epidemiology of cardiomyopathies and incident heart failure in a population-based cohort study

AIMS

The population prevalence of cardiomyopathies and the natural history of symptomatic heart failure (HF) and arrhythmia across cardiomyopathy phenotypes is poorly understood. Study aims were to estimate the population-diagnosed prevalence of cardiomyopathies and describe the temporal relationship between a diagnosis of cardiomyopathy with HF and arrhythmia.

METHODS

People with cardiomyopathy (n=4116) were identified from linked electronic health records (~9 million individuals; 2000-2018) and categorised into hypertrophic cardiomyopathy (HCM), dilated cardiomyopathy (DCM), arrhythmogenic right ventricular cardiomyopathy (ARVC), restrictive cardiomyopathy (RCM) and cardiac amyloidosis (CA). Cardiomyopathy point prevalence, rates of symptomatic HF and arrhythmia and timing relative to a diagnosis of cardiomyopathy were determined.

RESULTS

In 2018, DCM was the most common cardiomyopathy. DCM and HCM were twice as common among men, with the reverse trend for ARVC. Between 2010 and 2018, prevalence increased for ARVC by 180% and HCM by 9%. At diagnosis, more patients with CA (66%), DCM (56%) and RCM (62%) had pre-existing HF compared with ARVC (29%) and HCM (27%). Among those free of HF at diagnosis of cardiomyopathy, annualised HF incidence was greatest in CA and DCM. Diagnoses of all cardiomyopathies clustered around the time of HF onset.

CONCLUSIONS

The recorded prevalence of all cardiomyopathies increased over the past decade. Recognition of CA is generally preceded by HF, whereas individuals with ARVC or HCM more often developed HF after their cardiomyopathy diagnosis suggesting a more indolent course or better asymptomatic recognition. The clustering of HF and cardiomyopathy diagnoses suggests opportunities for presymptomatic or earlier diagnosis.

Polysaccharide co-polymerase WzzB/WzzE chimeras reveal transmembrane 2 region of WzzB is important for interaction with WzyB

The ability of bacteria to synthesise complex polysaccharide chains at a controlled number of repeating units has wide implications for a range of biological activities that include: symbiosis, biofilm formation and immune system avoidance. Complex polysaccharide chains such as the O antigen (Oag) component of lipopolysaccharide and the enterobacterial common antigen (ECA) are synthesised by the most common polysaccharide synthesis pathway used in bacteria, known as the Wzy-dependent pathway. The Oag and ECA are polymerized into chains via the inner membrane proteins WzyB and WzyE, respectively, while the respective co-polymerases WzzB and WzzE modulate the number of repeat units in the chains or "the modal length" of the polysaccharide via a hypothesised interaction. Our data shows for the first time "cross-talk" between Oag and ECA synthesis in that WzzE is able to partially regulate Oag modal length via a potential interaction with WzyB. To investigate this, one or both of the transmembrane regions (TM1 and TM2) of WzzE and WzzB were swapped creating six chimera proteins. Several chimeric proteins showed significant increases Oag modal length control, while others reduced control. Additionally, co-purification experiments show an interaction between WzyB and WzzB for the first time without the use of a chemical cross-linker, and a novel interaction between WzyB and WzzE. These results suggest the TM2 region of Wzz proteins plays a critical role in Oag and ECA modal length control, presumably via the interaction with respective Wzy proteins, thus providing insight into the complex mechanism underlying the control of polysaccharide biosynthesis.ImportanceBacteria synthesise complex polysaccharide chains at a controlled number of repeating units, this has wide implications for a range of bacterial activities involved in virulence. Examples of complex polysaccharide chains include, the O antigen (Oag) component of lipopolysaccharide and the enterobacterial common antigen (ECA), both of these examples are predominantly synthesised by their own independent Wzy-dependent pathway. Our data show for the first time "cross-talk" between Oag and ECA synthesis and identifies novel physical protein-protein interactions between proteins in these systems. These findings further the understanding of how the system functions to control polysaccharide chain length which has great implications for novel biotechnologies and/or the combat of bacterial diseases.

Whole exome sequencing identifies genetic variants in inherited thrombocytopenia with secondary qualitative function defects

Inherited thrombocytopenias are a heterogeneous group of disorders characterized by abnormally low platelet counts which can be associated with abnormal bleeding. Next-generation sequencing has previously been employed in these disorders for the confirmation of suspected genetic abnormalities, and more recently in the discovery of novel disease-causing genes. However its full potential has not yet been exploited. Over the past 6 years we have sequenced the exomes from 55 patients, including 37 index cases and 18 additional family members, all of whom were recruited to the UK Genotyping and Phenotyping of Platelets study. All patients had inherited or sustained thrombocytopenia of unknown etiology with platelet counts varying from 11×10⁹/L to 186×10⁹/L. Of the 51 patients phenotypically tested, 37 (73%), had an additional secondary qualitative platelet defect. Using whole exome sequencing analysis we have identified "pathogenic" or "likely pathogenic" variants in 46% (17/37) of our index patients with thrombocytopenia. In addition, we report variants of uncertain significance in 12 index cases, including novel candidate genetic variants in previously unreported genes in four index cases. These results demonstrate that whole exome sequencing is an efficient method for elucidating potential pathogenic genetic variants in inherited thrombocytopenia. Whole exome sequencing also has the added benefit of discovering potentially pathogenic genetic variants for further study in novel genes not previously implicated in inherited thrombocytopenia.

Resistance of Dynamin-related Protein 1 Oligomers to Disassembly Impairs Mitophagy, Resulting in Myocardial Inflammation and Heart Failure

We have reported previously that a missense mutation in the mitochondrial fission gene Dynamin-related protein 1 (Drp1) underlies the Python mouse model of monogenic dilated cardiomyopathy. The aim of this study was to investigate the consequences of the C452F mutation on Drp1 protein function and to define the cellular sequelae leading to heart failure in the Python monogenic dilated cardiomyopathy model. We found that the C452F mutation increased Drp1 GTPase activity. The mutation also conferred resistance to oligomer disassembly by guanine nucleotides and high ionic strength solutions. In a mouse embryonic fibroblast model, Drp1 C452F cells exhibited abnormal mitochondrial morphology and defective mitophagy. Mitochondria in C452F mouse embryonic fibroblasts were depolarized and had reduced calcium uptake with impaired ATP production by oxidative phosphorylation. In the Python heart, we found a corresponding progressive decline in oxidative phosphorylation with age and activation of sterile inflammation. As a corollary, enhancing autophagy by exposure to a prolonged low-protein diet improved cardiac function in Python mice. In conclusion, failure of Drp1 disassembly impairs mitophagy, leading to a downstream cascade of mitochondrial depolarization, aberrant calcium handling, impaired ATP synthesis, and activation of sterile myocardial inflammation, resulting in heart failure.

A mutation in the mitochondrial fission gene Dnm1l leads to cardiomyopathy

Mutations in a number of genes have been linked to inherited dilated cardiomyopathy (DCM). However, such mutations account for only a small proportion of the clinical cases emphasising the need for alternative discovery approaches to uncovering novel pathogenic mutations in hitherto unidentified pathways. Accordingly, as part of a large-scale N-ethyl-N-nitrosourea mutagenesis screen, we identified a mouse mutant, Python, which develops DCM. We demonstrate that the Python phenotype is attributable to a dominant fully penetrant mutation in the dynamin-1-like (Dnm1l) gene, which has been shown to be critical for mitochondrial fission. The C452F mutation is in a highly conserved region of the M domain of Dnm1l that alters protein interactions in a yeast two-hybrid system, suggesting that the mutation might alter intramolecular interactions within the Dnm1l monomer. Heterozygous Python fibroblasts exhibit abnormal mitochondria and peroxisomes. Homozygosity for the mutation results in the death of embryos midway though gestation. Heterozygous Python hearts show reduced levels of mitochondria enzyme complexes and suffer from cardiac ATP depletion. The resulting energy deficiency may contribute to cardiomyopathy. This is the first demonstration that a defect in a gene involved in mitochondrial remodelling can result in cardiomyopathy, showing that the function of this gene is needed for the maintenance of normal cellular function in a relatively tissue-specific manner. This disease model attests to the importance of mitochondrial remodelling in the heart; similar defects might underlie human heart muscle disease.

Heart disease is very common. Some cases of heart disease are strongly influenced by lifestyle and diet, whereas others have a strong genetic component. A certain form of heart failure, known as dilated cardiomyopathy (DCM) quite often runs in families suggesting that a defective gene or genes underlie this disease. We describe a new mouse mutant called “Python” which suffers from a heart disease similar to DCM. We were able to pinpoint the defective gene responsible for the disease. This gene is normally involved in the division of mitochondria, the “power plants” of the cell that generate one of the main energy supplies for the cell. This is a unique model that implicates a new gene and mechanism of disease for further investigation.

The role of CACNA1S in predisposition to malignant hyperthermia

BACKGROUND

Malignant hyperthermia (MH) is an inherited pharmacogenetic disorder of skeletal muscle, characterised by an elevated calcium release from the skeletal muscle sarcoplasmic reticulum. The dihydropyridine receptor (DHPR) plays an essential role in excitation-contraction coupling and calcium homeostasis in skeletal muscle. This study focuses on the gene CACNA1S which encodes the alpha1 subunit of the DHPR, in order to establish whether CACNA1S plays a major role in MH susceptibility in the UK.

METHODS

We investigate the CACNA1S locus in detail in 50 independent MH patients, the largest study to date, to identify novel variants that may predispose to disease and also to characterise the haplotype structure across CACNA1S.

RESULTS

We present CACNA1S cDNA sequencing data from 50 MH patients in whom RYR1 mutations have been excluded, and subsequent mutation screening analysis. Furthermore we present haplotype analysis of unphased CACNA1S SNPs to (1) assess CACNA1S haplotype frequency differences between susceptible MH cases and a European control group and (2) analyse population-based association via clustering of CACNA1S haplotypes based on disease risk.

CONCLUSION

The study identified a single potentially pathogenic change in CACNA1S (p.Arg174Trp), and highlights that the haplotype structure across CACNA1S is diverse, with a high degree of variability.