Article Text


Basic science review series
A straightforward guide to the sarcomeric basis of cardiomyopathies
  1. Luís R Lopes1,2,3,
  2. Perry M Elliott1
  1. 1Institute of Cardiovascular Science, University College London, London, UK
  2. 2Centro de Cardiologia da Universidade de Lisboa, University of Lisbon, Lisbon, Portugal
  3. 3Cardiology Department, Hospital Garcia de Orta, Almada, Portugal
  1. Correspondence to Professor Perry M Elliott, The Heart Hospital, 16-18 Westmoreland Street, London W1G 8PH, UK; perry.elliott{at}


The sarcomere is the principal contractile unit of striated muscle. Mutations in genes encoding sarcomeric proteins are responsible for a range of diseases including hypertrophic, dilated and restrictive cardiomyopathies and ventricular non-compaction. The downstream molecular pathways leading to these heterogeneous phenotypes include changes in acto-myosin cross-bridge kinetics, altered mechanosensation, disturbed calcium sensitivity, de-regulated signalling pathways, inefficient energetics, myocardial ischaemia and fibrosis. The elucidation of the genetic causes of cardiomyopathy has helped in understanding the structure and function of the sarcomere and a more detailed knowledge of the sarcomere and its associated proteins has suggested additional gene candidates. The new hope is that these advances will stimulate the discovery of disease-modifying drugs.

View Full Text

Statistics from

Normal sarcomere structure

Each cardiac myocyte is composed of a parallel arrangement of myofibrils that run longitudinally along the cell and are transversely subdivided into contractile units called sarcomeres.1 The sarcomere is a specialised cytoskeletal structure which constitutes the fundamental motor unit of the cardiomyocyte. Each sarcomere is comprised of thin and thick myofilaments.2 The thick filament is composed of around 300 molecules of myosin, each made up of 2 protein units of β- or α-myosin heavy chain and 4 myosin light chain molecules: two essential and two regulatory2 (figure 1). The thin filament is composed of repeating actin molecules, closely associated with the regulatory troponin complex (troponin T (TnT), troponin I (TnI) and troponin C (TnC)) which exerts its function in coordination with α-tropomyosin.3 The protein cardiac myosin-binding protein C also contributes to the regulation of actin–myosin interaction and cross-bridge kinetics. The regular arrangement of the myofilaments delineates a number of discrete topographic zones in each sarcomere: the M-band in the middle of the sarcomere; the A-band which is made up of multiple and parallel thick filaments and includes a central region, the H-zone, devoid of cross-bridges and consisting only of myosin tails; the I-band in which there are only aligned thin filaments with no overlap with thick filaments; and, finally, the Z-disc which constitutes the lateral boundary of the sarcomere2 (figure 2).

Figure 1

Schematic representation of the main eight sarcomere proteins and interactions in the sarcomere. Adapted from2 with permission. cMyBP-C, cardiac myosin-binding protein C, domains are numbered C0 to C10, starting at the N terminus of cMyBP-C; m, regulatory motif between domains C1 and C2 that contains PKA phosphorylation sites; PA, a proline/ alanine-rich linker sequence between C0 and C1, ELC, essential myosin light chain; LMM, myosin heavy chain's light meromyosin; RLC, regulatory myosin light chain; Tm, tropomyosin; TnC, troponin C; TnI, troponin I; TnT, troponin T.

Figure 2

Schematic representation of the discrete topographic zones in each sarcomere. The M-band constitutes the middle of the sarcomere; the A-band is made up of multiple and parallel thick filaments and includes a central region, the H-zone, devoid of cross-bridges and consisting only of myosin tails; the I-band in which there are only aligned thin filaments, with no overlap with thick filaments; and finally, the Z-disc which constitutes the lateral boundary of the sarcomere.

Myosin heavy chain is divided into three main domains: the globular head (the ‘motor domain’), which includes the ATP-binding site and the actin-binding site; the neck, an α-helical domain where the light chains bind and which is further subdivided in a converter region and a lever arm involved in the amplification of mechanical energy required for the power stroke of myosin along actin; and the tail or rod region, an helical coiled-coil protein that intertwines with the tail of another heavy chain molecule to form the thick filament.4 There are two highly homologous isoforms of the myosin heavy chain, encoded sequentially by 2 genes located in tandem on chromosome 14: MYH6 encodes α-myosin heavy chain and MYH7 encodes β-myosin heavy chain. MYH7 is the dominantly expressed gene in adult human ventricular cardiomyocytes (around 70%–80% of β-myosin heavy chain mRNA), while MYH6 is expressed at higher levels in the developing heart, particularly in the atria.w1 w2 4

Myosin light chains are part of the EF-hand family of calcium-binding proteins. Their function is believed to be essentially mechanical, providing support to the heavy chain neck and possibly regulating the interaction of myosin head with actin and consequently the force and velocity of contraction. The essential light chain, encoded by MYL3, is located closer to the myosin head and the regulatory light chain, encoded by MYL2, binds to the neck closer to the rod domain.2

Myosin-binding protein C (MyBPC) is made up of 11 domains (8 immunoglobulin-like and 3 fibronectin-like) numbered C0–C10; it also includes a proline/alanine-rich linker sequence between C0 and C1 and a regulatory motif between domains C1 and C2 that contains protein kinase A (PKA) phosphorylation sites.4 MyBPC interacts with the thick filament through 3 domains at its C terminus, which bind to both myosin and titin. The ratio is believed to be of 2–4 MyBPC molecules per myosin. At the N-terminus, MyBPC is suggested to exert a regulatory role through its interaction with the myosin neck and actin.4

The cardiac thin filament is a multiprotein complex made of 7 actin monomers, 1 troponin complex (TnI, TnT and TnC) and 1 α-tropomyosin coiled-coil dimer.3 The overall structure of the troponins is similar in that they are composed by helical domains separated by linker sequences of varying lengths that confer a high degree of flexibility to the molecule.3 α-tropomyosin is an α-helical coiled-coil dimer, which runs along the length of the actin filament and plays a key role in regulating contraction. α-cardiac muscle actin is one of the isoforms of actin, a group of highly conserved proteins involved in cell motility. Globular actin polymerises to form filament actin, which is structurally organised as a two-stranded helix.w3

Mechanism of contraction: the cross-bridge cycle

The fundamental principle governing muscle contraction is that of the sliding filament theory, in which the sliding of actin past myosin generates muscle tension. The interaction between myosin and actin in which the globular head of the myosin molecule bends towards and then binds to actin, contracts, releases actin, and then initiates a new cycle is known as myosin-actin cycling. The links between the myosin head and actin are called cross-bridges; the contraction of the myosin S1 region which requires the hydrolysis of ATP to release energy is called the power stroke.

Electrical activation of the heart and contraction are coupled through the intracellular movement of calcium. Depolarisation of the cardiomyocyte cell membrane during the action potential activates the L-type voltage-dependent calcium channels in the T tubule; the subsequent influx of calcium into the cell then leads to the opening of ryanodine-receptor channels in the adjacent sarcoplasmic reticulum with a rapid increase in cytosolic calcium.w4 Calcium binds to TnC, inducing an allosteric conformational change in TnI and TnT that is transmitted to tropomyosin. This causes a transition from a ‘blocked’ to a ‘closed’ state, which then exposes the myosin-binding sites of actin and allows the cross-bridges to occur (‘open’ state).w4 3 The myosin heavy chain head, with ADP and inorganic phosphate bound to its nucleotide-binding pocket, then interacts with the exposed actin-binding sites. This is followed by the release of ADP and inorganic phosphate, which occurs simultaneously with the power stroke, resulting in force development and shortening of the sarcomere and the I-band and the approximation of the Z-discs. Following this, ATP binds to the nucleotide-binding pocket of the myosin heavy chain head, which detaches from actin and myosin then hydrolyses ATP into ADP and inorganic phosphate again,w4 restarting the cycle.


Cardiomyopathies are myocardial disorders that are not solely explained by coronary artery disease or abnormal loading conditions. They are classified according to ventricular function and morphology into four main subtypes: hypertrophic cardiomyopathy (HCM); dilated cardiomyopathy (DCM); restrictive cardiomyopathy (RCM); and arrhythmogenic right ventricular cardiomyopathy (figure 3). Diseases that do fit into these categories are termed unclassified cardiomyopathies and include left ventricular non-compaction (LVNC), endocardial fibroelastosis and Tako-Tsubo cardiomyopathy. All cardiomyopathies are caused by genetic and non-genetic disease.

Figure 3

Schematic showing the different main cardiomyopathy phenotypes.

Mutations in proteins of the cardiac sarcomere comprise the main cause of HCM. Approximately 40%–50% of patients carry rare and potentially pathogenic variations in eight main sarcomeric genes.5 With the exception of the myosin light chains, genetic variation in all of these proteins can also cause DCM, although a larger percentage of cases remain genetically unsolved.6 A proportion of RCM is also caused by mutations in some of these genes.w5 Table 1 summarises the sarcomeric and associated proteins, encoding genes, interactions, associated disease phenotypes and their respective prevalence.

Table 1

Summary of the sarcomeric and associated proteins, encoding genes, interactions, associated disease phenotypes and respective prevalence

Genetic mechanisms of disease

Two different basic pathogenic mechanisms are thought to account for disease associated with mutations in cardiac sarcomere proteins. Missense single nucleotide variants (a nucleotide change and hence codon change that results in an aminoacid being substituted by another aminoacid in the protein) predominantly lead to a dominant negative effect (described as ‘poison peptide’) in which the mutated protein is not destroyed but rather integrates into the sarcomere, leading to the disease phenotype. This is thought to be characteristic of MYH7 variants.7 Alternatively, nonsense single nucleotide variants or small frameshift insertion–deletions can introduce a premature stop codon and hence result in haploinsufficiency due to nonsense mRNA mediated decay or proteolysis of a truncated protein.w6 8 This mechanism is believed to be typical of the majority of pathogenic MYBPC3 mutations. The various effects of individual variants on fibre contractile velocity, force and calcium sensitivity have been proposed as an explanation for the existence of dramatically different phenotypes arising from genetic variation in the same molecule.w7 4 A paradigm has been proposed whereby mutations that increase motor activity and power output lead to HCM, while those that diminish motor function and decrease power output lead to DCM.w7 4

Role of the non-contractile sarcomere proteins in disease

Mutations in the non-contractile elements of the sarcomere may also result in a cardiomyopathy phenotype. Some of the most important are as follows.


Titin is a giant macromolecule that consists almost entirely of hundreds of immunoglobulin and fibronectin-like domains, which interact with sarcomeric proteins at the Z-disc, the myosin filament and the M-band. Other specific domains provide the elastic spring connection to myosin and actin filaments in the I-band9 w8 10 w9 (figures 1, 2 and 4). Titin spans an entire half-sarcomere, bound at the N-terminus to the Z-disc and at the C-terminus to myosin and myosin-binding protein C. It has a major role in determining the mechanical properties of the heart through its effects on passive tension during myocardial stretch and restoring forces during early ventricular filling and is an important biomechanical sensor and organisational element within the sarcomere.10

Figure 4

Schematic representation of the sarcomere, focusing on the Z-disc and the M-line proteins and its interactions. Adapted fromw51 with permission.

The protein is encoded by the TTN gene on chromosome 2, which undergoes complex differential splicing, resulting in isoforms with variable elastic properties.11

Isoform switches with increased expression of longer and more compliant isoforms (increasing the ratio N2BA:N2B) are reported in heart failure associated with reduced and preserved EF.11

Titin has a serine/threonine kinase domain in the transition of the M to the A-band.12 Some of its targets include ubiquitin ligases that act on pathways relevant to cell survival and autophagy: Nbr, p62 and Murf.

Variation in TTN was linked to DCM for the first time in 2002;w10 13 10 years later, truncating titin mutations were reported as the most common genetic cause of DCM.w11 Although more debatable, there are also published reports of an association with HCM,w12 RCMw13 and arrhythmogenic cardiomyopathy.w14

Z-disc and associated proteins

The Z-disc corresponds to the lateral borders of the sarcomere. Besides its mechanical function, this structure is composed of proteins that have signalling functions related to the transcriptional regulation of muscle growth and mechanotransduction.1 ,14 Mutations in Z-disc associated proteins (figure 4) are predominantly associated with DCM rather than HCM:

  • α-actinin (encoded by ACTN2) has an important role in actin localisation and forms the principal cross-links at the Z-disc between actin filaments of opposite polarity originating from contiguous sarcomeres. α-actinin mutations have been described in both HCM and DCM.w15 w16

  • Muscle LIM protein (MLP, encoded by CSRP3) is a LIM-domain protein that interacts with α actinin, calcineurin and telethonin. It is also localised in the nucleus and has been implicated in various pathways, including mechanosensation and mechanotransduction, calcium metabolism, myofibrillogenesis and actin polymerisation. Genetic variation is associated with HCM and DCM.w16 w17

  • Telethonin (encoded by TCAP) binds the N-terminal domain of two adjacent titin molecules and interacts with various proteins such as ankyrin repeat protein 2, T-tubular system components and ion channels (e.g. Nav1.5). It is also implicated in regulation of myocardial hypertrophy by interacting with calsarcin-1 (myozenin-2) and myostatin. Variation in this gene has been observed in HCM and DCM.w18

  • Integrin-linked kinase interacts with integrin and with MLP at the Z-disc. Mutations in this protein have been linked with DCM.w19

  • Calsarcin 1 or myozenin 2 (encoded by MYOZ2) is involved in the transcriptional regulation of muscle growth. Variation in this gene has been proposed as a cause of HCM.w20

  • Four-and-a-half-LIM-domains-1 (encoded by the X-chromosome gene FHL1) localises to the sarcomeric Z-disc/I-band area and the nucleus. This protein provides structural maintenance of the sarcomere and transcriptional regulation. Both HCM and DCM phenotypes have been described, in isolation or associated with neuromuscular disease (e.g. Emery–Dreifuss muscular dystrophy).w21

  • Cypher/ZASP or LIM domain-binding protein 3 (encoded by LDB3) has an amino-terminal PDZ domain that interacts with α-actinin; this protein also interacts with protein kinase C at the C-terminal. Mutations in LDB3 are reported in DCM and LVNCw22 and also HCM,w23 although evidence for the last is debated.

  • Actinin-binding LIM Protein (ALP) or PDZ/LIM domain protein 3 (encoded by PDLIM3) belongs to the ALP family of proteins, characterised by an N-terminal PDZ domain and a C-terminal LIM domain. It has been proposed as a candidate gene for HCM, but no convincing evidence was found.w24

  • Bcl2-associated athanogene (BAG3) is an Hsp70/HSC70-binding co-chaperone. A recent study15 using genome-wide association and whole-exome sequencing identified rare variants in BAG3 as a cause for familial DCM.

  • Desmin (encoded by DES) is an intermediate filament, which connects to desmosomes and interacts with the Z-disc. Pathological variation in this gene leads to a number of phenotypes, including muscular disease, RCM with conduction abnormalitiesw25 and DCM.w26 An arrhythmogenic cardiomyopathy phenotype has also been described.w27 w28

  • α-B crystalline (encoded by CRYAB) is a member of the small heat shock protein (HSP20) family and interacts with desmin, vimentin and actin and is localised to the Z-disc. A mutation has been described as a cause of DCM.w29

  • Nebulette (NEBL) interacts with actin and inserts into the Z-disc. Mutations have been reported as a cause of DCM.w30

  • Myopalladin (MYPN) interacts with several proteins including α-actinin, titin, nebulette and cardiac ankyrin repeat protein. Variation in this gene has been described in association with DCM, HCM and RCM.w31

  • Nexilin (NEXN) is another structural Z-disc protein whose role seems to be the maintenance of the integrity of the Z-discs against the tension generated within the sarcomere. It interacts with actin and mutations have been associated with HCMw32 and DCM.w33

  • Ankyrin repeat domain 1 (ANKRD1)-encoded cardiac adriamycin responsive protein, or cardiac ankyrin repeat protein, is a protein localised in the I-band that interacts with the N2A domain of titin and the N-terminal region of myopalladin. It can also be found in the nucleus, where it functions as a transcription cofactor in the embryonic and fetal heart and in advanced heart failure. Genetic variation has been associated with HCMw12 and DCM.w34 w35

The M-band and associated proteins

The M-band is located in the centre of the A-band (figures 2 and 4) and is important in the even distribution of tension along the sarcomere. The two main proteins of the M-band are myomesin and M-protein. Both myomesin and M-protein interact with other sarcomere proteins such as titin and myosin and M-band associated structural proteins such as obscurin and regulatory proteins, e.g. Nrb1 or FHL. A mutation in myomesin has been associated with the development of HCM.w36

Obscurin (encoded by OBSCN) is a giant protein. Alternative splicing of OBSCN results in different isoforms. Obscurin interacts at the M-band with ankyrin-1 (involved in sarcolemma–cytoskeleton interactions) and interacts with the sarcoplasmic reticulum.1 ,16 A single report suggested that an obscurin mutation could be a cause of HCM, but this remains to be confirmed.w37

Translating genomic information into therapy

It is striking that while the first sarcomeric protein gene mutations were reported over 20 years ago, the treatment of patients with cardiomyopathy is still largely palliative. Fortunately, greater understanding of the biological consequences of sarcomere mutations is leading to new ideas on the treatment of genetic heart muscle diseases (figure 5).

Cross-bridge kinetics

Much research has focused on an exploration of the consequences of sarcomeric protein gene mutations on myofilament contraction, mostly using recombinant proteins or cellular/molecular studies from engineered animal models and less often, patient tissue from myectomies or explanted hearts. Based on these findings, it has been hypothesised that small molecule inhibitors or activators of acto-myosin cross-bridge formation might be able to correct one of the fundamental pathophysiological mechanisms in cardiomyopathy (figure 5A),1618 acknowledging that other mechanisms may also account for low force generation, including lower myofibril density and higher cardiomyocyte area, aspects of cardiomyocyte remodelling that vary between mutated genes.19

Figure 5

Mechanisms of disease in hypertrophic cardiomyopathy and possible drug targets. Adapted fromw51 with permission.

Calcium sensitivity and cycling

Based on a paradigm in which the differential effects of mutations on calcium homeostasis explain phenotypic variation,w7 4 manipulation of myofilament calcium sensitivity is being explored as a potential therapy (figure 5B). A recent animal model of a tropomyosin mutation demonstrated that correction of increased calcium sensitivity, by coexpressing a pseudo-phosporylated TnI, prevented the development of the phenotype, including diastolic dysfunction.20

Deranged calcium cycling is a feature of heart failure of various aetiologies and the consequence of changed expression or post-translational modifications of sarcoplasmic/endoplasmic reticulum calcium ATPase 2 (SERCA2) and the ryanodine receptor 2.21 Altered calcium cycling has been proposed as a fundamental aspect of HCM at the cellular level, possibly as a consequence of calcium being retained at the sarcomere.22 ,23 w38 Altered calcium cycling in HCM and other cardiomyopathies may also be explained by reduced uptake into the sarcoplasmic reticulum by the SERCA2 pump due to a cellular energetic defect.w39 The clinical consequences of altered calcium dynamics are early and delayed after-depolarisations, arrhythmia and diastolic dysfunction.w40 w41

Animal models have shown that de-sensitisation to calcium corrects the arrhythmogenic phenotypew41 and in a recent study using induced pluripotent stem cells as a model for HCM,24 pharmacological inhibition of calcium and sodium entry attenuated the arrhythmogenic cellular phenotype. In two rodent models of HCM, an α- myosin heavy chain missense mutant mouse and a transgenic mouse model expressing a cardiac TnT mutation, inhibition of plasma sarcolemmal L-type calcium channel by diltiazem prevented fibrosis and improved LV diastolic function.22 w42 Another recent paper focused on altered electro-mechanical coupling using cardiomyocytes from HCM patients submitted to myectomy.25 In a comparison with cardiomyocytes from non-hypertrophic non-failing tissue, late sodium current inhibition with ranolazine improved a cellular phenotype provoked by increased calmodulin kinase II activity and increased phosphorylation of its targets resulting in a reduction in early and delayed after-depolarisations. A clinical trial of ranolazine in HCM is in progress (figure 5B).

Signalling pathways and protein degradation pathways

As previously described, many of the sarcomere, Z-disc and M-band proteins, as well as titin, are linked with cardiomyocyte survival and growth pathways, via nuclear factor of activated T cells transcription factor, mitogen-activated protein kinases and other protein kinases and transcription/co-transcription factors. This places the contractile and associated proteins as intermediaries between these downstream pathways and the mechanical processes of muscular tension, contractility and energy consumption.1 ,16 ,21 Possible targets in these pathways include protein kinases, enzymes that catalyse the phosphorylation of various substrates (the most frequent post-translational modification of proteins) and that act as key mediators of cell activity via signal transduction pathways vital for metabolism, cell cycle, motility, survival/apoptosis and cell–cell communication.26 Protein phosphatases have the opposite role: by dephosphorylating the same substrates as the kinases, they contribute to the regulation of the same pathways, as the activity of a signal transduction protein depends on the ratio between phosphorylated and dephosphorylated state.w43 There is also growing interest in protein degradation pathways including the ubiquitine–proteosome system and degradation by lysosomes via autophagy.w44

Cardiomyocyte–fibroblast cross-talk

Interstitial fibrosis is one of the defining histological characteristics of cardiomyopathyw45 and is thought to contribute to arrhythmia and LV diastolic and systolic dysfunction. In a murine model of HCM, sarcomere protein gene mutations activated proliferative and profibrotic signals in fibroblasts to produce pathological remodelling.27 Fibroblasts also demonstrated increased expression of TGF-β and other pro-fibrotic proteins, like periostin;27 administration of neutralising antibodies or losartan attenuated fibroblast proliferation and fibrosis, and in the case of losartan, hypertrophy in mutation-positive pre-hypertrophic mice was prevented (figure 5C).

Cardiomyocyte energetics

The myocardium depends on oxygen for high-energy phosphate production by oxidative phosphorylation (figure 5A). In the normal heart, ATP is produced primarily by the metabolism of free fatty acids (FFAs) and carbohydrates, with FFAs accounting for approximately 70% of total ATP production. Importantly, FFAs are less efficient as a source of myocardial energy as they require approximately 10% more oxygen than glucose in order to produce an equivalent amount of ATP.w46 Evidence from animal and human studies suggest that HCM (and perhaps other cardiomyopathies) is characterised by a reduction in the concentration of high-energy phosphates in the myocardium.28 This might be partly explained by myocardial ischaemia caused by microvascular dysfunction, but may also be a direct consequence of sarcomere protein gene mutations on myocardial contractile efficiency.w47 A trial of perhexiline that acts by inhibiting carnitine palmitoyl transferase (CPT-1/2), involved in mitochondrial uptake of long-chain fatty acids, versus placebo, showed an improvement of diastolic function and increased exercise capacity.29 A number of other studies are evaluating the effect of drugs that stimulate glucose oxidation and reduce fatty acid oxidation and thereby improve myocardial efficiency and lower oxygen demand in patients with HCM and DCM.

Gene therapy

Gene therapy refers to the delivery of genetic material, in the form of DNA (the gene in itself), non-coding RNA or oligonucleotides, into cells for the purpose of modulating gene expression.w48 Challenges to its clinical application in cardiac disease relate mainly to the delivery vector (adeno-associated viruses are currently the most promising due to their capacity to infect non-dividing cells and generation of a less intense immune responsew48) and the delivery technique where intracoronary injection is the most likely to be applied clinically,w48 compared with intramyocardial injection which is commonly applied in animal models.

One landmark study showing the potential clinical application of gene therapy in cardiovascular disease was a phase II clinical trial (CUPID, calcium upregulation by percutaneous administration of gene therapy in cardiac disease) in class III/IV heart failure patients. This trial showed that patients who were given high-dose intracoronary sarcoendoplasmic reticulum calcium-ATPase 2a, known to be expressed at lower levels in heart failure, had a decrease in symptoms, improved LV function after 12 months and a lower number of events at 3 years follow-up.w49

Genetic therapy in inherited heart muscle disease faces additional challenges related to the mechanisms of disease. In addition to a certain level of normally functioning transcript/protein, the effect of poison peptides—sometimes responsible for an autosomal dominant disease—has to be suppressed. Although tested in humans with other genetic diseases such as Duchenne muscular dystrophy,30 proof-of-concept animal studies demonstrating the possibility of genetic therapy in inherited cardiomyopathy have only very recently been published. These include an exon skipping strategy in a knock-in mouse model carrying a MYBPC3 mutation30 and RNA interference to reduce expression of mutant missense Myh6 (R403Q) transcripts, which resulted in silencing of the phenotype.w50 If successfully translated to humans, these approaches would allow the correction of the primary genetic cause of cardiomyopathy, instead of targeting secondary downstream mediators—as discussed in previous sections—or merely treating symptoms, as in current clinical practice.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors LRL and PME drafted the manuscript. PME revised it critically for important intellectual content. Both authors approved the final version.

  • Funding LRL was supported by a grant from the Gulbenkian Doctoral Programme for Advanced Medical Education, sponsored by Fundação Calouste Gulbenkian, Fundação Champalimaud, Ministério da Saúde and Fundação para a Ciência e Tecnologia, Portugal.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.