The biomedical use of silk as a suture dates back to antiquity. Fibroin is the structural element that determines the strength of silk and here we consider the safety of fibroin in its role in ophthalmology. The high mechanical strength of silk meant sufficiently thin threads could be made for eye microsurgery, but such usage was all but superseded by synthetic polymer sutures, primarily because silk in its entirety was more inflammatory. Significant immunological response can normally be avoided by careful manufacturing to provide high purity fibroin, and it has been utilised in this form for tissue engineering an array of fibre and film substrata deployed in research with cells of the eye. Films of fibroin can also be made transparent, which is a required property in the visual pathway. Transparent layers of corneal epithelial, stromal and endothelial cells have all been demonstrated with maintenance of phenotype, as have constructs supporting retinal cells. Fibroin has a lack of demonstrable infectious agent transfer, an ability to be sterilised and prepared with minimal contamination, long-term predictable degradation and low direct cytotoxicity. However, there remains a known ability to be involved in amyloid formation and potential amyloidosis which, without further examination, is enough to currently question whether fibroin should be employed in the eye given its innervation into the brain.
- experimental & laboratory
- treatment other
- treatment surgery
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Silks are natural proteins. Most silk proteins are elongated molecules, belonging to the group of fibrous proteins that also includes collagens. Fibrous proteins have their polypeptide chains arranged in parallel along a single axis. Silks all contain highly repetitive amino acid sequences with ordered secondary-structure regions. This homogeneity leads to outstanding mechanical properties and remarkable functional performance.
Silk has been used for millennia as suture,1 being accepted over the ages as a useful material that is generally safe to be used in the human body.2 This allowed silk to gain regulatory approval, essentially on real-world evidence. While this follows a risk-assessment path, whereby a lack of known previous problems supports further use, a biomaterial should be reviewed again as scientific knowledge increases, potentially expanding the range of safety issues. Here we carry out such a review on silk in ophthalmology to consider its safety in the eye.
The origin and application of silk fibroin in the eye
Silks are produced primarily by species of Euarthropoda, such as moths, spiders and centipedes. In general, the silk components are stored in the organism as liquids, which are then configured into fibres on extrusion. The number of natural silks is very large, as is the variation in their chemical composition, structure and properties. Human usage of silk has been contingent on the ease of its collection; spider silk fibres are stronger, but it is difficult to collect in large quantity and spiders cannot be domesticated. Some moths on the other hand, spin silk fibres to form a cocoon to protect the pupae and these cocoons can be collected in abundance. This type of silk has been valued for textile production for millennia and the wild silk moth was putatively domesticated to Bombyx mori to further ease supply.
As expected, the silk produced by the B. mori silkworm has been the most investigated. Each silk fibre consists of two core monofilaments made of the protein fibroin coated and glued together primarily by another protein, sericin. The strength lies in the fibroin monofilaments which have in this species a repeating primary sequence of Gly–Ala–Gly–Ala–Gly–Ser hexapeptide, with a three-step higher-order periodicity. This regularity determines the final strength, in conjunction with the proportions of the three molecular conformations dominating its macromolecular secondary structure of antiparallel pleated sheets (β-sheets), α-helices or random coils.
Silk usage as suture arises from the strength of the silk fibre and the ease with which it can be spun into a thread. The silk suture was valued because of its lack of memory, good workability and knot tying security. As early as 1883, Kuhnt3 was advocating the use of silk for corneal sutures and silk also became the most prominent suture in general surgery.
Following the seminal work of Minoura et al,4 fibroin has now also been considered for many further biomedical applications,5 including those in ophthalmology.6 These range from the cornea, where differentiated epithelial cells,7 including those in a limbal construct,8 stromal keratocytes9 and endothelial cells10 have been grown on fibroin, through to a construct of film and sponge stacks of fibroin that incorporated three cell types of corneal nerve, epithelium and stromal cells.11 In the retina, fibroin has been used to fabricate a substitute Bruch’s membrane.12
As well as growing cells on fibroin, silk-derived proteins have been used for ocular wound healing to enhance corneal epithelial cell growth media.13 Furthermore, silk has been used for ocular drug delivery, with a hydrogel for slow-release of bevacizumab in a rabbit model for the treatment of age-related macular degeneration and for the introduction of cellular growth factors.14 Silk ocular prostheses are also advocated for corneal reshaping to restore visual acuity.15
The safety of silk
Sterilisation and decontamination
Biomaterials need to be sterile, not only to prevent the transmission of infectious agents from the source, but also to remove contaminant microbes gained during processing. Silk can be sterilised by all three of the standard methods of ethylene oxide, autoclaving or gamma irradiation, although with differing but predictable effects on chemical and physical structure. Increasing the irradiation dose can be used to speed the degradation rate, while steam sterilisation can make stronger, stiffer films that change cell attachment.16
By using water-based solvents to reconstitute fibroin,17 processing contamination risk can be limited to simple salts that can be reduced to very low levels by extensive dialysis against water.18 Only traces of degumming agents such as calcium carbonate and solvents such as lithium bromide should remain.7 Such an ‘all green’ chemistry17 approach has been used with silk fibroin wound-healing products.19
Infectious risk from organisms or viruses
The non-mammalian origin of silk confers a distinct advantage. Biomaterials such as collagen require their animal sources to be screened to reduce opportunity for transmission of mammalian diseases. In contrast, moths are not mammalian and they do not feed on mammalian blood, so they do not pose a zoonosis risk. A small group of insect retroviruses may be similar to mammalian retroviruses,20 but there has been no documented disease in humans.
Silk is also a secretion that does not contain cellular elements that might sustain infection, indeed the European Economic Community does not classify silk to be an animal product. Taking all these aspects together, silk has a low infectious risk status, even before it is processed and sterilised. Processing can further reduce risk, as the fibroin is dissolved, often purified by dialysis, and then reconstituted.18 Dissolving customarily involves lengthy boiling in a strongly alkaline chaotropic solution, such as lithium bromide, or the use of a strong acid or ionic liquid, often combined with filtration or centrifugation. Such treatments are effective decontamination processes in their own right.
Silk sutures have been used extensively in ophthalmology, first as so-called ‘virgin silk’, where the sericin coat is all but left intact and the fibres twisted together to the required thickness, then as ‘black braided silk’. To produce black braided silk, the sericin coating is first removed by degumming, whereby the peptide bonds of the sericin are broken by acidic or alkaline hydrolytic or enzymatic treatment, usually along with heat,18 and the sericin washed off. The resulting fibrils of fibroin can then be spun into thread. Ophthalmic surgery is generally microsurgery, and monofilaments made of nylon or polypropylene were developed with enough strength for this purpose.
These synthetic alternatives displaced most of the use of silk, with adverse immunological reaction cited as the primary disadvantage of using virgin silk in the eye,21 although rabbit studies showed little difference between materials.22 Severe reaction from second cataract surgery suggested a possible role of prior sensitisation from a first encounter with virgin silk, with less dramatic reaction with black braided suture.23 Silk suture usually elicits only a minimal acute inflammatory reaction which involves infiltrative migration of polymorphonuclear leucocytes, which is followed by gradual encapsulation of the suture by fibrous connective tissue. A fibroin film placed into a corneal stromal pocket in the rabbit was well tolerated, with after 6 months only a few lymphocytes and some activated keratocytes.24 These data suggest that purified silk fibroin may produce a lesser response compared with the virgin silk suture. A murine model of dry eye suggested that a silk fibroin solution even has anti-inflammatory effects.25
Sensitisation to virgin silk, and therefore sericin, has been repeatedly reported, including in ophthalmology.23 However, despite the past widespread use of silk suture throughout the body, there have been few reports of delayed hypersensitivity.26 27 Now that silk usage is less common, generally only older patients will have prior sensitisation risk and there remains the option to carry out a preoperative allergenic skin test.
The conundrum of fibroin and sericin combined
Clinically, less corneal inflammation occurred when using black braided rather than virgin silk, which has a higher sericin content23 and it was taken that sericin was inflammatory.5 However, sericin did not activate murine macrophages in vitro by itself, although it did have a synergistic effect with bacterial lipopolysaccharide.28 There was also low immunogenicity in vivo with mice, and Jiao et al,29 reviewing previous studies, concluded they were mostly only suggestive of adverse reaction, and that subsequent improper referencinwas also low immunog gave a misperception regarding sericin’s biosafety. The method of removing sericin from silk may change adsorption of exogenous proteins onto fibroin hampering cell adhesion.30 Hence, we are left with an incomplete understanding of immunogenicity, but it appears that fibroin will benefit from reduced sericin concentration if cell adhesion is adequate. Quality control techniques to assess residual levels of sericin are now available30 and overall, close attention to processing is required to ensure that sericin does not confound the biological response to fibroin.
The widespread use of silk suture in ophthalmology is testament that it is not grossly cytotoxic to many cells. Cell death from direct contact does not normally occur, although necrosis has been reported, which is probably secondary to an immunological reaction.23 With the ocular surface, epithelial cells rapidly grow down the apertures formed by silk suture, indicating that cells in close apposition retain viability.
The interaction of silk proteins in vivo throughout the body has been reviewed31; summarising that there are excellent bioresponses in vivo with low immunogenicity and an ability to be remodelled and replaced by native tissue. However, one group using fibroin nanoparticles, reported that they cause cellular and mitochondrial dysfunction in cultured fibroblasts, blood cells and umbilical vein cells.32
Table 1 describes the major characteristics of the generic host response to silk and silk fibroin and relates these to the variables that could influence host response.
Neoplasia, carcinogenicity and teratogenicity and genotoxicity
Silk fibroin hydrogels have been shown to suppress tumour formation in chick chorioallantoic membrane.33 In vivo, in mice with lung cancer, tumour growth was suppressed with fibroin.34 These types of results, taken together with the extensive use of silk suture in the eye without the association of widespread resulting carcinoma, indicate that fibroin may be beneficial in this regard, rather than pathogenic.
In the early 1900s, silk suture was regularly used in human uterine surgery35 without documented adverse effect. Studies of the effects of intrauterine silk thread on the fertility of female rats found no evidence of teratogenicity and no effect on the oestrous cycle, concluding it is safe for uterine use.36 Yan et al37 performed a study for acute toxicity, genotoxicity and effects on the reproductive system after implantation of fibroin in nerve guides; no effect was found with any of the parameters measured.
Amyloid is a term describing insoluble protein aggregates of specific structure. Although no specific amyloidogenic peptide sequence pattern has been identified, there does appears to be a common core structure of polypeptide chains, generally known as the ‘cross-β’ structure.38 Amyloids occur naturally, and in many roles do not result in disease, instead achieving a diverse range of advantageous biological functions,39 for example in moths, where silk is produced as a protective case in larval development. In a similar way amyloid β (Aβ) peptide protects against fungal and bacterial infections in mammals.40 It can promote recovery from injury, including repairing leaks in the blood–brain barrier41 and may act as a modulator of synaptic plasticity with implications in learning and memory.
Conversely, it has been noted that Aβ may rapidly increase in response to a physiological challenge and often diminishes on recovery.41 It is also known that amyloids are involved in various human pathologies42 including Alzheimer’s disease (AD) and Parkinson’s disease where accumulation of abnormally misfolded proteins, a process known as amyloidosis, results in neurodegeneration. There is a balance between physiological and pathological effects of Aβ, and in disease the levels of Aβ are elevated from pnM to nM or μM levels.43 Amyloids can be toxic to cells in a variety of ways including: disruption of cytoskeleton, physical damage of cellular membranes, DNA damage, oxidative stress, mitochondrial disfunction, apoptosis and adverse intracellular calcium signalling. Another dangerous feature of amyloids is that there is some form of pathogenic spread from region to region in the organ/body.44
The cascade theory of Hardy and Higgins45 proposed with AD that amyloid starts a cascade of events that then leads to amyloidosis and cellular damage. It is believed that neither the monomeric protein, nor the deposited fibrils, exert neurotoxicity per se. It is intermediate low molecular mass oligomers and protofibrils that are considered to be the likely neurotoxic species.46 Although the cascade theory has been reconsidered and criticised,47 there is still acknowledgement that a nucleus can initiate amyloid formation and a possible pathogenetic role of Aβ cannot be completely ruled out.48 Regardless of any further mechanism, amyloid originating from nuclei is a concern and both fibroin and sericin have been shown to be seeds.49 These proteins also bound to Aβ leading to aggregation, suggesting a potential role in the propagation of Aβ amyloidosis. Transmission of amyloid-β protein pathology from cadaveric pituitary growth hormone to intracerebrally inoculated mice has been demonstrated,50 with the formation of a mutant, humanised amyloid precursor protein. The same report also documents iatrogenic transfer of Aβ pathology to humans.
In humans, 36 proteins/peptides are already known to generate amyloid deposits and disease.51 Each distinct form of amyloidosis is uniquely characterised by the chemical identity of the amyloid fibril protein that deposits in tissues to give rise to the disease. At least one, Enfuvirtide, is synthetic, and not a native human cellular protein, that binds with the HIV envelope protein gp41. Its presentation is also iatrogenic,52 highlighting how amyloidosis risk may arise from medical treatment even with non-human peptides or proteins.
Amyloid and the eye
In primates, amyloid is associated with glaucoma; in a glaucoma model made by deliberate damage to the retina, AD-like pathologies are also established in the lateral geniculate nucleus.53 Such an association is concerning, as changes to the eye may therefore result in adverse change in the conduit to the brain. Studies to identify ocular markers as an Alzheimer’s diagnostic have shown Aβ plaques in the eyes of those with the disease, indicating the close association between sites.54 Indeed, in Alzheimer’s-transgenic mice, retinal Aβ deposits preceded brain Aβ deposition.55
In mice, subretinal injection of Aβ caused several adverse events leading to retinal degeneration. Understanding the role of Aβ may be important in diseases of retinal degeneration, such as age-related macular degeneration.56
The prion diseases of Creutzfeldt-Jakob disease, Kuru, fatal familial insomnia and Gerstmann-Straussler-Scheinker syndrome diseases involve a conformation isoform of the normal cell surface glycoprotein PrPC into the pathogenic protein PrPSc. The architecture of mammalian prions appears fundamental to their lethal pathogenicity, being self-propagating assemblies of misfolded host-encoded protein.57 Infection can be transmitted not only by ingestion of tissues, but also by iatrogenic transfer.
There is evidence in mice that Aβ aggregates are prions by the demonstration of widespread cerebral β-amyloidosis following inoculation of purified Aβ aggregates derived from brain, or even synthetic Aβ aggregates,58 such that a unifying process of prions has been proposed.59 Furthermore, recent study in humans discovered similar prion-like propagation of Aβ aggregates.50
Role of inflammation in amyloidosis
Many amyloidosis-associated processes are accompanied or triggered by inflammation.60 In murine models with persistent inflammation, an amyloid-enhancing factor (AEF) protein has been identified that originates 1–2 days before amyloid A (AA).61 Specific fibrils from synthetic peptides with no amyloid relationship can induce AA amyloidosis during inflammation in an animal model.62 In addition, innocuous proteins can build up toxic amyloids under certain conditions.63 An amyloid-related fibril can also act as an AEF where a characteristic property is their β-sheet organisation.64 Taken together these show a relationship of β-sheet protein organisation and host inflammation in amyloidosis.
Association of fibroin β-sheet with amyloid and amyloidosis
Fibroin β-sheet formation of silk has similar structural characteristics to amyloid.65 Dissolved fibroin has been reported to accelerate amyloid accumulation in mice61 with Lundmark et al66 indicating that silk fibroin can act as a cross-seed nucleus and that there is, ‘Transmissibility of systemic amyloidosis by a prion-like mechanism’. The threat from a nucleation-polymerisation model67 is that fibroin placed into humans might seed deleterious amyloid formation. Such risk may prevent devices that contain fibroin gaining regulatory approval for new usage, particularly like those in ophthalmology where there is a close link between the eye via the optic nerve to the brain; the retina is effectively an extension of the central nervous system.68
Silk fibroin gel with small fibrils and β-sheet content injected into mice resulted in a dose-dependent AA formation in the spleen, but only if the fibroin had first been sonicated and there was an inflammatory stimulus by use of silver nitrate. In the mouse, 1 µg is sufficient to induce AEF.61 If animal studies are applicable to humans, using silk fibroin with a patient who has chronic inflammatory disease enhances the risk of amyloidosis.69 Contrary to this view, not only is there no direct evidence of such occurrence in man, but there is a growing body of experimental evidence showing reduced risk. The cytotoxicity of amyloid peptides is related more to their nanoassembly, rather than the β-sheet content.70 Furthermore, spider fibroin with the amino acid sequence of amyloid β-sheets had a marked cytotoxic effect on cultured human neural cells, but silk fibroin β-sheets did not.71 They attributed this to lack of surface charge on the silk peptides, even though they have the same β-sheet content as amyloid. Fibroin has also been shown to be a relatively poor seed in Aβ40 and Aβ42 aggregation.49
A direct study of the risk of fibroin for amyloidosis by Tsukawak et al,72 concluded that silk occasionally promotes amyloidogenesis, but has a low potential for amyloidosis. Amyloid deposits were rarely observed in mice injected with silk fibroin in solution, but with certain methods of production, heavy amyloid deposit resulted from fibroin fabric placed in the abdomen; the production method is an important determinant of its biocompatibility.
Biodegradation of silk
The routine use of silk as suture has resulted in it being recognised in regulation,73 classifying it as non-absorbable because it is retained for more than 60 days. Despite this nomenclature, silk is eminently absorbable, as fibroin as a protein can be biodegraded by the major protein degradation pathways, via peptides to amino acids.74 Fibroin is inherently slow to degrade; the long sequences inhibit substrate unfolding, decreasing the efficiency of the proteasomes and compact regions sterically delay degradation with the most crystalline regions degrading last.2
Proteases are ubiquitous in tears, anterior chamber and vitreous, but their concentration varies, not only with site, but also with age, race and disease state. The physical presentation also affects degradation rate with, for example, silk fibres being attacked at a slower rate than films.74 Proteolytic action can also be delayed if the site encapsulates the silk with fibrous connective tissue. Thus, the degradation rate of fibroin fibres placed in an inflamed cornea may differ markedly from silk film embedded in retina.
Silk fibres during biodegradation show an increase of surface roughness and crystallinity before fragmentation occurs. Again, depending on site, traces of fibroin may remain for years. The ability and the speed of degradation are not the only considerations for a successful biomaterial as it must continue to function as it degrades. Cracking, void formation74 and fragmentation can also be important, as this can dislodge cells or allow the movement of the biomaterial to unsuitable areas. For example, in the eye, fragments of silk might lodge in the iris angle. Although models of degradation have been used in vitro, it is only through in vivo modelling that more accurate simulation can be demonstrated. Nevertheless, animal models may differ from the human eye, and clinical trials are the ultimate test of suitability.
Biocompatibility of silk
The aim of suture is purely to provide mechanical strength; ideally it is chemically and biologically inert and does not interact with the host. Silk fibroin for tissue engineering on the other hand has a need for specific and direct interactions with tissue components.75 Silk in tissue engineering is subject to sterile inflammation76 that can present different conditions to a conventional pathogenic response. As reviewed by Altman et al,5 there are many reports stating silk fibroin is biocompatible which are routinely based on the long history of use without direct pathology and often supported by the short-term growth of a single cell type from the planned site of use. Short-term growth of a monoculture in the laboratory only demonstrates the material is not cytotoxic to those cells. Biocompatibility requires us to show that the material does not induce a negative biological response when implanted into a tissue composed of many cell types with a complex arrangement of blood supply and nerves.
Biocompatibility varies with the application.76 It is not a property of the material alone, but it is the biomaterial in a host system. For this reason biocompatibility keeps getting redefined.77 Thus, the biocompatibility of a fibroin film placed on the surface of an avascular cornea may markedly differ from the same product placed in an inflamed, vascularised cornea, even though it is the same tissue. Numerous differences may be involved: different enzymes may be present, immune reaction may be primed, mechanobiology may change cell growth, and so on.
Non-B. mori silk
There is a vast range of fibroin types, with individual species producing fibroin with very different characteristics from different types of gland.78 Thus, knowledge of the safety of fibroin from B. mori alone is insufficient to assess all fibroins. Importantly, where the fibroin is not a direct moth secretion, then there can be potential additional hazards carried over from the source and the manufacturing process. Here fibroin has been divided into those arising from moth and non-moth sources.
The fibroin taken directly from moths other than B. mori has been investigated, particularly those of the Antheraea genus such as mylitta79 and pernyi,80 which, unlike B. mori, have an RGD sequence that binds to integrin receptors in cell membranes to improve cell adhesion. The processing techniques used are similar to those used for B. mori, hence only the resultant moth-protein characteristics may cause an additional hazard. However, to increase cellular attachment and growth, or to select for other requirements, blends of fibroins from different species, coatings10 or fibroin mixed with cell attachment or growth factors have been used.14 In these cases the safety may differ from the individual components, and must be assessed on a case-by-case basis.
The use of non-moth fibroin has been limited by the quantity of supply. Sources such as ants have been little studied, but the unsurpassed strength of spider silk has meant that alternative methods of manufacture have been justified to increase supply. Gene insertion into B. mori to code chimeric spider fibroin has been one method.81 In this case, the lack of cell transfer from moth to fibroin will again limit the resultant hazard to the protein characteristics, although a poorly designed gene insertion vector may also propagate its own protein. Other types of manufacture can further increase risk. Methods already used involve gene insertion into bacteria, yeast, tobacco or potatoes, which could produce foreign gene proteins from those organisms, as may the use of transgenic animal sources, such as goats, where methods have been established to harvest the protein from the milk.82 However, the highest risk arises from a cultured mammalian cell source, where cell proteins and disease agents have a more direct route to potentially contaminate any silk product. Additional risk from non-B. mori fibroin can be reduced by preventing all but the desired protein to be coded. This can be achieved by following these procedures: prior screening of animals, cell lines and organisms for disease; careful selection of the method of gene insertion; and precise control of polymerisation by the use of non-regenerable flanking restriction enzyme sites during gene control.83 The resultant protein can then be purified by methods such as immobilised metal affinity chromatography.83 In these ways the outcome can be a defined product.
Not only is silk suture regulated and approved by the United States Pharmacopeia,73 but a fibroin scaffold84 (SERI surgical scaffold, Allergan, USA) has been used for abdominal wall repair. There have also been clinical trials of fibroin for tympanic membrane85 and breast reconstruction86 with at least three devices already with approval for clinical use.19
Most authorities have harmonised their regulation to the ISO 10993 standards.87 If there is similarity to a predicate, clinical trials may not even be required for device approval and a less onerous route to approval permitted, for example 510(k) in the USA. Thus, levels of evidence of silk fibroin biocompatibility may relate primarily to the specific function of the cell therapy in order to gain regulatory approval because it will be assumed that fibroin has been proven safe because of real-world results over many years.
A safety review according to ISO 10993 for acute toxicity, genotoxicity, toxicological effects on the reproductive system and local effects after implantation of fibroin as nerve guides showed no effect on any of the parameters measured.37 Similar studies are required for silk fibroin deployed in the eye. The issue remains however, that although such a study with the eye may also suggest absence of an adverse effect, it may not address, for example, if there is a risk from fibroin amyloid in a human host, over long periods, with associated inflammation. If a material has been approved in a medical device such as suture, this does not mean that the material is safe to use in a tissue engineering role, where there is different biological activity. ISO 10993 may currently be a poor mechanism to assess this.88
Silk is a well-established biomaterial; as a suture it has been taken to cause little harm. It can have an inflammatory response, but this can be minimised by precise manufacture to produce pure fibroin. Fibroin meets many of the criteria for a competent biomaterial, such as lack of direct cytotoxicity, and its long usage in the human body has supported regulatory approval. However, lack of direct cytotoxicity does not guarantee biocompatibility. Williams,89 in his definition of biocompatibility, importantly includes that a biomaterial should perform its desired function, ‘without eliciting any undesirable local or systemic effects in the recipient’.
β-amyloid is a molecule with unclear nature, behaviour and pathological roles that is an important participant in the pathogenesis of many neurodegenerative diseases. Currently, despite centuries of use, there appears to be little evidence that human amyloidosis has been caused by silk, but conversely there has also been insufficient specific long-term investigation. A low potential for amyloidosis72 needs to be risk-assessed with long-term studies for specific sites involving simultaneous inflammation.69 First, to do no harm, remains a basic tenet of medical treatment and potential harm needs to be fully assessed before fibroin is used as a biomaterial in the eye.
Contributors PWM conceived and planned the study, researched the field and wrote the manuscript draft. IK reviewed the manuscript and wrote sections of the manuscript. MJA supervised the study and reviewed and amended the manuscript.
Funding This work was supported by funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement no. 637460) and from Science Foundation Ireland (15/ERC/3269).
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
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.