Large cohorts help to define the true nature of left-ventricular noncompaction

The Cardiovascular Genetics and Genomics group is delighted to share the results of a new, multi-centre study on left-ventricular noncompaction (LVNC), published in Genetics in Medicine here.
LVNC is characterized by a partially noncompacted heart muscle, with a classically “spongy” appearance at imaging and diagnostic criteria based on the relative amount of compacted and noncompacted myocardium, as displayed in the figure below. Such diagnostic criteria though — when applied using the newer imaging technologies — can lead to the detection of LVNC in up to 15% of the population, highlighting a high risk of over-diagnosis.


Hearts in 4-chamber view of a patient affected with LVNC (left) and a healthy individual (right), obtained with cardiac magnetic resonance imaging. The non-compacted portion of the myocardial layer in the patient with LVNC is highlighted with a red line.

The true nature of LVNC as a clinical entity is subject to much debate as to LVNC represents a separate disease or whether it is to be considered a secondary trait that can manifest in presence of an underlying cardiomyopathy. Such discussion is still far from reaching a definitive answer, as reflected by the two different classifications assigned by the European Society of Cardiology and the American Heart Association, which label LVNC as an “unclassified cardiomyopathy” and a “primary genetic cardiomyopathy”, respectively.

In this study, developed in collaboration with Amsterdam UMC, the University of Florence, the Aswan Heart Centre and the Partners’ Healthcare Laboratory for Molecular Medicine, we have:

  • collected genetic data from over 800 patients referred for genetic testing for LVNC with LVNC (575 from previously published cohorts and 264 unpublished) and compared the frequencies of rare (carried by fewer than 1 in 10000 individuals) coding variants in 70 genes with those observed in a publicly available reference population (Genome Aggregation Database [gnomAD], ~125,000 individuals): this comparison showed which genetic variant classes are significantly associated with LVNC.
  • compared results with those obtained utilizing the same strategy on cohorts of patients with dilated cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM), in previous studies performed by our group: this analysis’ main purpose was to assess the genetic overlap between LVNC and these two other cardiomyopathies, so to estimate the proportion of cases in which non-compaction may be a secondary trait of underlying DCM or HCM, rather than a distinct process.
  • Analyzed cardiac phenotypes of population individuals carrying specific LVNC-associated variants, so to assess the effect of such variants in the population and investigate whether their presence correlates with sub-clinical phenotypes: this genotype-phenotype analysis serves to understand if genetics can help discern between physiological and pathological noncompaction.


One figure from the paper, recapitulating the main results: 14 variant classes (in 11 genes) are associated with LVNC, having been found significantly more often in LVNC patients compared with population individuals (% indicated by the blue bars vs % indicated by the gray bars). Such variant classes are estimated to explain approximately 37% of cases. The disease association for the majority of them (11, in 9 genes) is shared with DCM (first group from the left, orange bars), HCM (second group, green bars) or both (third group). The associations with variants in the HCN4 and RYR2 genes are shared between LVNC and arrhythmogenic conditions, while the associations with truncating variants in MYH7, PRDM16 and ACTN2 are unique to LVNC. TV=truncating variants, TM=transmembrane, hotspot=residues 634-638 in RBM20.

As shown in the figure, the vast majority of LVNC-associated variant classes are also causative of HCM and/or DCM, suggesting how in a majority of cases LVNC is probably to be considered a secondary trait manifesting in patients affected with an underlying primary cardiomyopathy. However, there are certain cases (estimated in approximately 1 in 20 in the analyzed cohorts) in which LVNC is observed in presence of genetic variants that are uniquely LVNC-associated, suggesting how LVNC can also present as an isolated, distinct disease entity.
The association shared between LVNC and arrhythmogenic conditions with variants in the ion channel genes HCN4 and RYR2 (associated with sinus node disease and catecholaminergic polymorphic ventricular tachycardia, respectively) highlights how a subset of LVNC patients may be at risk of life-threatening arrhythmogenic events. Systematic screening for variants in these genes in LVNC patients may aid the early identification of such cases.
Of the variant classes associated uniquely with LVNC, truncating variants in MYH7 are of particular interest, as heterozygous “loss-of-function” variants have not previously been associated with cardiomyopathies (in contrast with non-truncating variants in the same gene, which are among the most important genetic causes of HCM and DCM). In this study, besides the significant association of these variants with LVNC, we also found how individuals in the general population (including the UK Biobank) carrying this type of variants have a significantly higher degree of ventricular noncompaction compared with non-carriers.
Taken together, these results contribute to the characterization of LVNC and suggest how:

  • in the majority of cases – LVNC is a morphological phenotype manifesting in presence of an underlying cardiomyopathy.
  • in approximately 5% of the cases, LVNC appears to be caused by a distinct genetic aetiology which suggests it can also be a distinct disease entity.
  • in addition, results obtained on phenotyped population cohorts confirm a concrete danger of LVNC over-diagnosis, suggesting how diagnostic criteria for LVNC should not be based solely on imaging, but conjugated e.g. with genetic screening to help identify individuals at risk of developing left-ventricular dysfunction.

Written by Francesco Mazzarotto

Seven new publications out today describe the gnomAD resource and applications

We are delighted that the flagship publications describing the Genome Aggregation database (nomAD) have appeared today in Nature, Nature Medicine, and Nature Communications.Members of the CV Genetics & Genomics team have been long-term contributors to the gnomAD resource (and its predecessor, the exome aggregation consortium ExAC), and we have hugely enjoyed being part of this collaboration.Today’s package of publications includes two papers led by members of our team. “Characterising the loss-of-function impact of 5′ untranslated region variants in 15,708 individuals” has already been described in a previous blog post when we made our paper open access on the bioRxiv preprint server a year ago (see here), and leverages the power of gnomAD to annotate a particular group of variants that can cause severe rare genetic diseases. Rather than disrupting the sequence of the gene itself, these variants lie just upstream of the gene and interfere with regulatory signals that control translation of genes into proteins. This study was a home-grown effort, spear-headed by Nicky Whiffin while working at Imperial. The second study was also co-led by Nicky, and arose from a visit to spend some time working with Daniel MacArthur’s team at the Broad Institute. “The effect of LRRK2 loss-of-function variants in humans looks at a specific gene, LRRK2, suggested as a therapeutic target for Parkinson’s disease. There have been some concerns about the safety of targeting LRRK2 with drugs. However, Nicky and colleagues, working with data from gnomAD, 23andMe, and UK Biobank, found that individuals who carried genetic variants that inactivate one copy of this gene did not have any apparent adverse consequences. This provides reassurance that therapeutic inhibition of LRRK2 is likely to be well-tolerated and safe.You can read Nicky’s reflections on her work and the gnomAD package on her blog on her new webpage, which she has recently launched as she prepares to move to Oxford to set up a new lab at the Wellcome Centre for Human Genetics.

Post by James Ware

New Study Led by the Cardiovascular Genetics and Genomics Group Brings Together the Largest Published Dilated Cardiomyopathy Cohort to Clarify its Genetic Architecture

The years preceding the availability of large, publicly accessible cohorts of genetically characterized individuals witnessed a large number of studies implicating genes in diseases such as dilated cardiomyopathy (DCM), without sufficiently robust evidence according to today’s standards.

Projects such as the 1000 Genome Project, the Exome Sequencing Project and the Exome Aggregation Consortium (ExAC) database showed how, collectively, rare variants were considerably more common than had been expected in the population, implying that rarity is necessary but not sufficient to deem variants pathogenic under a dominant monogenic model assuming high penetrance, and casting doubts on many previously proposed gene-disease associations.

Here, we re-evaluate 56 genes in which rare variants have previously been reported to be associated with DCM. We have gathered data from more than 2,500 patients with DCM, and compare genetic variation in these individuals with control populations to provide a much needed re-assessment of the genetic architecture of monogenic DCM.

The study has been published two days ago by the journal Circulation
(https://www.ahajournals.org/doi/abs/10.1161/CIRCULATIONAHA.119.037661).

In summary:
– We analysed rare (carried by fewer than 1/10,000 individuals) genetic variants in 56 genes previously implicated in DCM.
– We first compared genetic data from two primary cohorts of ~1,000 DCM cases and ~1,000 healthy controls (with normal cardiac MRI scans) uniformly sequenced and processed thorough identical pipelines, to minimize technical bias.
– We also analysed a secondary cohort of ~1,500 DCM cases to encompass different patient profiles, and compared them to a population reference cohort (~60,000 ExAC individuals) as an additional population control cohort to maximise statistical power.
– ExAC is not a control cohort as such, and comprises whole-exome sequencing data. These two key characteristics, representing potential sources of bias, were addressed applying several quality control steps to minimize artefactual stratification.
– Across three separate comparisons, we detected robust statistical evidence for association with DCM for only 12 of the 56 genes (TTN, MYH7, LMNA, TNNT2, DSP, BAG3, TPM1, TNNC1, VCL, NEXN, ACTC1 and PLN).
– We computed the corresponding Etiological Fraction for each of the DCM-associated variant classes, which is an estimate of the probability that a variant (of the same class) is the cause of disease when it is found in a patient. Some variant classes, including all those predicted to result in a truncated protein, have Etiological Fraction values >0.95, indicating very high probabilities of pathogenicity even before evaluating additional evidence.
– Specific variant classes in VCL and TPM1 were observed significantly more often than expected in patients younger than 18 years of age, suggesting that these variants may primarily cause early-onset DCM.


One of the figures from the paper: the vertical axis displays an estimate of what proportion of DCM cases is due to variants in each gene. As an example, truncating variants in TTN explain approximately 10% of DCM. The horizontal axis displays the prior probability that a variant detected in a patient is the cause of disease, which corresponds to a measure of how interpretable variants are in the diagnostic setting.

These results do not rule out the other 44 analysed genes from playing a role in DCM, but they suggest that their contribution will be negligible or confined to very specific variant classes. This data will be instrumental to DCM genetic testing, both in the composition of diagnostic gene panels and in the interpretation of variants detected in patients, and to community gene curation efforts such as the one carried out by the ClinGen consortium, evaluating also other lines of evidence.

Post by Francesco Mazzarotto

New study identifies genetic link between cancer treatment and cardiomyopathy

We are excited to share the recent press release and BBC coverage of our study into genetic factors contributing to cancer drug related cardiomyopathy.

In this study, published in the journal Circulation, we analysed genes of more than 200 cancer patients from the UK, Spain and the US who had been diagnosed with a type of heart condition called cancer-therapy induced cardiomyopathy, or CCM. We found patients who developed the heart condition were more likely to carry genetic faults linked to cardiomyopathy – and in particular that patients were more likely to carry a faulty version of a gene called titin.

Characterising the loss-of-function impact of 5′ untranslated region variants in 15,708 individuals

We are excited to share our work using gnomAD to characterise 5’UTR variants that create or disrupt upstream open reading frames (uORFs) and explore their role in disease: https://www.biorxiv.org/content/10.1101/543504v2

We show that these variants are under strong negative selection (indicative of being deleterious), and identify a subset (that form ORFs overlapping the coding sequence) with signals of selection equivalent to coding missense variants.

We find increased signals of selection when these variants occur in the 5’UTRs of curated haploinsufficient, LoF intolerant and known dominant LoF developmental disease genes, supporting a loss-of-function effect of these variants on translation.

We identify specific genes where uORF perturbation appears to be an important disease mechanism (e.g. NF1 and IRF6), and report a novel uORF frameshift variant in NF2 that segregates with disease in two families with neurofibromatosis.

Our approach illustrates the power of using large population databases and grouping non-coding bases by functional effect, to identify subsets of variants that are highly deleterious. Although the strength of selection at the level of UTRs is equivalent to synonymous variants, we see a much stronger signal at these specific uORF-perturbing sites.

Finally, we have created a VEP plugin that annotates 5’UTR variants for uORF-perturbing effects:
https://github.com/ImperialCardioGenetics/uORFs/tree/master/5primeUTRannotator

post by Nicky Whiffin

Abnormalities in the Titin gene linked to potentially dangerous heart rhythm abnormalities

In our previous posts (June 2017 and October 2017) we explained two of our studies looking at the role of truncating variants in the Titin gene (TTNtv) in dilated cardiomyopathy (DCM). In the first of these, we found that having a TTNtv meant that a DCM patient was more likely to have had a past history of abnormal heart rhythms when they were first diagnosed with DCM. But we did not have enough data to definitively assess whether this was due to rhythms in the upper (atrial) chamber of the heart, the bottom (ventricular) chamber, or both. We also didn’t know whether this meant that DCM patients with a TTNtv were at higher long-term risk of potentially dangerous or life-threatening heart rhythm abnormalities. In our second study, we followed up a large group of patients with DCM but didn’t find an increased rate of major arrhythmic events in patients with a TTNtv. As a whole this group had relatively mild symptoms of DCM at the time of the study (the majority being in NHYA class I/II heart failure), with moderately impaired ventricular function – which puts them in a comparatively low-risk group for arrhythmias. Perhaps as a result, the overall rate of arrhythmic events in the study was low, making it harder to see subtle differences between those with and without a TTNtv.

To investigate further, we therefore started a new study on patients with DCM who also had an implantable cardiac device (ICD or CRTD). Not only are patients with these devices more likely to have an arrhythmia (they are typically given to the patients because they have worse heart function or symptoms) but, crucially, they monitor the patient’s heart rhythm 24-hours a day, 7-days a week. This allowed us to get complete coverage of each patient’s heart rhythm data over a number of years.

Our study, which was published earlier this month in JAMA: Network Open, found that TTNtv are associated with a significantly higher risk of both ventricular and atrial arrhythmia in DCM patients with an ICD or CRTD.

We had a strict definition of arrhythmia, for example only counting a ventricular arrhythmia if it was both fast (>200 beats per minute) and prolonged. These are the rhythms considered to pose the most danger to patients.

Importantly, we looked at the other major factors known to increase the risk of arrhythmia – age, male sex, level of left heart dysfunction and presence of heart fibrosis (scarring) on MRI – and even after controlling for these statistically TTNtvs were a strong predictor of arrhythmia.

Heart fibrosis, or scarring, on MRI is worth delving into a little more deeply. Fibrosis shows up as white patches on the heart in an MRI scan after a contrast agent (gadolinium) has been given via a vein. The presence of this fibrosis has long been known to be a strong predictor of arrhythmia, it’s likely that the scarred area interferes with the way electrical signals travel through the heart. DCM patients with a TTNtv don’t seem to have more fibrosis than other DCM patients – we found that in a previous study and confirmed it again here. However, a remarkable finding in our study was the apparent additive nature between fibrosis and a TTNtv in terms of arrhythmia risk. Of those patients that had both fibrosis and
a TTNtv, 62% experienced a ventricular arrhythmia, compared to only 5% of those with neither fibrosis or a TTNTv.

So how might TTNtvs lead to more arrhythmia in DCM?

This study wasn’t designed to answer this question and much more work needs to be done. But one avenue worth investigating is the fact that DCM patients with a TTNtv seem to have thinner heart muscle walls than those without a TTNtv, as we discussed in one of our previous blogs from October 2017. Thinner walls lead to increased ‘strain’ on the heart muscle cells, which can potentially alter their electrical properties.

Can knowing if a patient has a TTNtv help decide whether or not they should have an ICD? This was a relatively small study and the robustness of the findings certainly needs to be tested in much larger cohorts. Such studies are under way (see our blog post from October 25th 2018) and it is hoped that in the future TTNtv may be one of several factors that can help in risk-stratifying patients with DCM.

You can read our published study here: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2736938

Post by Ben Corden

Research gets personal for the heart hive team

Many researchers study a particular disease because they have a personal connection to someone who has been affected. For Dr Nicky Whiffin of Cardiovascular Genetics and Genomics group, it happened in reverse.

"I had been researching cardiomyopathies (diseases that affect the heart muscle) for a couple of years when my mum suddenly became very ill. Even walking up the stairs was a struggle, she had to pause halfway to catch her breath. ..." Read more about Nicky's story.

What is The Heart Hive?

In partnership with Cardiomyopathy UK, we are looking for ways to help people with heart muscle disease participate in research to move science forward and improve healthcare and quality of life. Patients often want to take part in research, but opportunities are concentrated around universities and major hospitals or involve long distance travel preventing many patients from getting involved.

The heart hive will make research accessible to everyone. Patients can upload their own health information to our secure web portal and decide which projects can use it. Genetic information can be collected via saliva kits sent out through the post. Check out http://thehearthive.org for more information.

New link found between alcohol, genes and heart failure

Scientists have revealed a new link between alcohol, heart health and our genes.

The researchers investigated faulty versions of a gene called titin which are carried by one in 100 people or 600,000 people in the UK.

Titin is crucial for maintaining the elasticity of the heart muscle, and faulty versions are linked to a type of heart failure called dilated cardiomyopathy.

Now new research suggests the faulty gene may interact with alcohol to accelerate heart failure in some patients with the gene, even if they only drink moderate amounts of alcohol.

Click here for more.

Largest ever study of fatal heart condition which affects 260,000 people in UK

The largest ever study of the heart condition dilated cardiomyopathy (DCM) is underway.

There is fresh hope for hundreds of thousands of people in the UK with the potentially deadly heart condition dilated cardiomyopathy (DCM), as the largest ever study of the poorly-understood disease begins led by Professor Stuart Cook of the National Heart and Lung Institute. The research is funded by an investment of over 2 million pounds by the British Heart Foundation (BHF).

Click here for more.