Insights on the role of thyroid hormone transport in neurosensory organs and implication for the Allan–Herndon–Dudley syndrome

Abstract

Thyroid hormones play an important role during the development and functioning of the different sensory systems. In order to exert their actions, thyroid hormones need to access their target cells through transmembrane transporter proteins, among which the monocarboxylate transporter 8 (MCT8) stands out for its pathophysiological relevance. Mutations in the gene encoding for MCT8 lead to the Allan–Herndon–Dudley syndrome (AHDS), a rare disease characterised by severe neuromotor and cognitive impairments. The impact of MCT8 deficiency in the neurosensory capacity of AHDS patients is less clear, with only a few patients displaying visual and auditory impairments. In this review we aim to gather data from different animal models regarding thyroid hormone transport and action in the different neurosensory systems that could aid to identify potential neurosensorial alterations in MCT8-deficient patients.

Introduction

Thyroid hormones (THs), T3 (3,5,3′-triiodo-l-thyronine), and T4 (3,5,3′,5′-tetraiodo-l-thyronine) are essential for the correct development and functioning of most organs. The thyroid gland mostly secretes T4, which is metabolised into the genomically active form T3 by the action of the enzymes deiodinase type 1 (DIO1) and type 2 (DIO2). On the other hand, the DIO3 deiodinates T4 and T3 into their inactive forms rT3 (reverse 3,3′,5′-triiodo-l-thyronine) and T2 (3,5-diiodo-l-thyronine), respectively, to terminate TH action. Once they reach their target tissue, T3 and T4 enter into their target cells through specific plasma membrane transporters. There are several transporter proteins with the capacity of transporting TH, including l-type amino acid transporters such as LAT1 and LAT2, the organic anion transporter1 C1 (OATP1C1), and the monocarboxylate transporters 8 (MCT8) and 10 (MCT10). Among these, MCT8 stands out for its physiological relevance (1). In the canonical pathway for TH action, T3 modulates gene transcription by binding to TH receptors (TRs), that include TRα1 (encoded by THRA), TRβ1, and TRβ2 (both encoded by TRHB). These nuclear receptors act as ligand modulated transcription factors that regulate the expression of different target genes. This is the pathway by which TH mediates most of its actions (2).

The Allan–Herndon–Dudley syndrome (AHDS), or MCT8 deficiency, is a rare X-linked disease caused by inactivating mutations in the SLC16A2 gene, which encodes for MCT8 (3, 4, 5). Patients present severe neurological disorders that include delayed neurological development, severe intellectual disability, and central hypotonia with spastic paraplegia, among others. These alterations arise as a consequence of brain hypothyroidism (6, 7, 8), most probably caused by the lack of a functional MCT8 in the brain barriers impairing TH transport into the brain (9, 10, 11). In addition, AHDS is characterised by high serum concentrations of T3, low T4, and rT3, and normal or slightly high levels of thyroid-stimulating hormone (TSH), indicating that there are also alterations in TH synthesis and metabolism (12). The restricted TH entry to the brain and the subsequent dysregulation of the hypothalamus–pituitary–thyroid gland axis, in addition to impairments in TH metabolism might be responsible for an increase in circulating T3 levels (1), producing a state of peripheral hyperthyroidism. Thereby, AHDS encompasses with both a neurological and an endocrine component.

Recent discoveries suggest that the symptomatic spectrum of AHDS patients might be wider than initially expected. Some of the patients diagnosed in the last years present less severe clinical symptoms including mild intellectual disability and mild or non-existent extrapyramidal symptoms, and they even develop the ability to articulate simple sentences or stand by themselves (13, 14). This complex and heterogeneous symptomatology makes the diagnosis of AHDS even more challenging and suggests that different patients might benefit from different therapeutic approaches. In view of this, it might be of relevance to gather information on other previously overlooked symptoms that may serve as additional biomarkers to address the phenotypical complexity of the AHDS. In view of the limited data available regarding neurosensorial alterations in AHDS patients or on the role of TH in sensory pathways in humans, in the present review we have gathered data obtained from different animal models regarding TH transport on neurosensory pathways. To better contextualise the relevance of MCT8 in these sensory pathways, other regulators of TH availability and action are also discussed.

Audition

The development of auditory performance is critically dependent on TH signalling. Hearing loss in humans has been associated with iodine deficiency and congenital hypothyroidism (15, 16). Audition takes place in the cochlea, which is the hearing organ located at the inner ear. This fluid-filled, spiral-shaped tube consists of three compartments: scala tympani, scala media and scala vestibuli. Separating the scala media and the scala tympani is the basilar membrane (BM) upon which lays the auditory sensory epithelium: the organ of Corti. The organ of Corti comprises a single row of inner hair cells (IHCs), which are the primary sensory receptors and an additional three rows of outer hair cells (OHCs). Overlying the organ of Corti is the tectorial membrane (TM), an extracellular matrix structure to which the stereocilia of the hair cells attach. Sound waves generate movements in the cochlear fluids, producing vibrations on the BM that are translated into radial displacements of the hair cells, culminating in the activation of mechanoelectrical transducer ion channels (17).

In rodents, the crucial period for TH action during the ontogeny of the main auditory cell types spans from late embryonic to neonatal stages (18, 19). Auditory defects and impaired cochlear differentiation have been observed as a result of developmental hypothyroidism (18, 20), as well as a result of deletions in TRs (21, 22, 23), deiodinases (24, 25) and TH transporters (26).

The genes encoding the TRα1 and TRβ TH receptors, Thra and Thrb, are expressed at the cochlea. In the rat, Thra has been found distributed throughout the entire cochlea, while Thrb has been predominantly observed at the organ of Corti (27, 28). Studies in Thrb-deficient mice indicated that the hearing impairments in these animals reside within the cochlea (29). Even though no apparent structural defects were initially observed in the cochlea of Thrb-deficient mice (29), it was later proposed that malformations of the TM, especially during early postnatal development, lead to reduced mechanical performance of the cochlea and deafness (21, 30). While mice with TRα deletions exhibit normal hearing (31), TRα1 in Thrb-deficient mice may substitute for the absence of TRβ (32), suggesting a role for TRα in audition. Interestingly, recent studies have found that mice with frameshift mutations in Thra are prone to age-related hearing loss and display some defects in OHC (33). According to this, deletions in both TRβ and TRα1 lead to delayed sensory epithelium differentiation, malformation of TM, impairment of electromechanical transduction in OHCs, and a low endocochlear potential (21). Of interest, the Thra and Thrb genes have also been found at the developing ossicles of the mouse middle ear, which allow sound transmission from the outer to the inner ear (34), suggesting a role for TH in the development of this structure. Supporting this idea, mice expressing a dominant-negative TRα1 protein (Thra^+/PV mice) present deafness and a range of middle ear abnormalities including defective ossicles (34).

Activity of deiodinases is also involved in auditory function. DIO2-deficient mice present hearing impairments (24, 35) that, interestingly, resemble those present in TRβ/TRα1-deficient mice. These alterations include delayed differentiation of the cochlear inner sulcus and sensory epithelium, and malformations in the TM (24). Due to the similarity in the hearing phenotype between TRβ/TRα1-deficient and DIO2-deficient mice, it has been proposed that DIO2 controls T3 availability that activates TRs in the cochlea (24). DIO3-deficient mice also present auditory deficits. Interestingly, expression of Dio3 in the immature cochlea has been found to overlap with Thrb expression and precedes Dio2 expression. In contrast to DIO2-deficient mice, DIO3-deficient animals display accelerated cochlear differentiation, all of which suggest that DIO3 prevents premature stimulation of TRβ (25).

To identify the role of TH transporters in auditory function, the expression of main TH transporters was explored during cochlear development between embryonic day (E) 18 and postnatal day (P) 15 (36). LAT1 was the main TH transporter identified in cochlear blood vessels at the spiral ganglion, spiral limbus, spiral ligament, and stria vascularis, suggesting a prominent role for LAT1 in the uptake of TH from the blood. LAT1 was also observed in the cell membrane of hair cells, both in IHC and OHCs. Mct8 expression was prominently detected in the spiral ganglion, columnar cells in the greater epithelial ridge, and stria vascularis, while showing comparatively lower levels of expression in the spiral ligament, spiral limbus, and lesser epithelial ridge. Mct8 expression peaked from P3 to P5, although was still present at the spiral ganglion, faintly at blood vessels of the stria vascularis, and the spiral ligament until at least P15. In the spiral ganglion and greater epithelial ridge, the Mct8 expression was found overlapping with Thrb expression in the sensory epithelium. Expression of Oatp1c1, which in the rodent brain is usually found in microvessels, was observed in fibrocytes – a supporting cell type involved in regulating endolymph electrolyte homeostasis and immune response (37) – of the spiral limbus and spiral ligament close to blood vessels. Interestingly, this expression pattern was overlapping with Dio2 expression, as previously described (38). Mct10 was identified in a highly restricted pattern mainly at the outer sulcus epithelium. Further studies evidenced that, while MCT8- or MCT10-deficient mice do not present hearing impairments, double MCT8/MCT10-deficient mice displayed hearing loss. Sharlin et al. (26) found degeneration of hair cells, retarded cochlear remodelling and malformations in the TM in MCT8/MCT10-deficient mice, similar to findings in TRβ/TRα1-deficient mice. The findings suggest that, in addition to the systemic availability of TH, cochlear development and hearing also depend on the uptake or efflux of TH by transmembrane transporters at target cells. The TH transporter LAT2 seems to have a relevant role in the adult inner ear, as its expression – mainly present in fibrocytes of the spiral ligament and limbus – increases from P60 to P360 in mice (39). Consequently, LAT2-deficient mice present important alterations in several cochlear structures and age-related hearing loss (39). However, whether these effects are due to impaired TH action or defective transport of neutral amino acids is yet to be determined.

Based on these findings, a model for the location of TH transporters in the inner ear is proposed in Fig. 1. Bearing in mind the limitations associated to determining the precise localisation of TH transporters as discussed below, we consider that the next challenge would be to assess if TH uptake from the circulation is mediated by a TH transporter other than MCT8, such as LAT1. It would be worth studying if T4 is then transported into the fibrocytes by MCT8, converted into T3 by DIO2 activity, and exported by MCT8 or another TH transporter such as LAT2, and whether T3 enters the target cells in the sensory epithelium by MCT8 and/or LAT1. Another relevant issue to address is whether MCT8 mediates uptake or efflux of TH (40, 41) during cochlear development. The fact that both Dio3 and Mct8 expression precede Dio2 expression could indicate that, at least at early stages, MCT8 cooperates with DIO3 to prevent TH from prematurely binding to TRβ. Lastly, even though TH transporters are known to be involved in bone development (42), the role of TH transport has not been studied in the context of middle ear development.

Figure 1.

Proposed location of thyroid hormone (TH) transporters in the cochlea based on current findings. Findings from Sharlin et al. (36) identified Lat1 expression in cochlear blood vessels at the spiral ganglion (not shown), spiral limbus (not shown), spiral ligament, and stria vascularis between P1 and P15; as well as in inner hair cells (IHCs) and outer hair cells (OHCs) at P3. Mct8 expression was found in the greater epithelial ridge (Ger) and lesser epithelial ridge (Ler) at P1-P5; at the fibrocytes of the spiral ligament at P1-P15; at the tympanic border cells (tbc) a P7-P15; at the spiral ganglion neurons (not shown) at P1-P15; and faintly identified at blood vessels of the stria vascularis and the spiral ligament (not shown) at P7-P15. Mct10 expression was found mainly restricted to the outer sulcus epithelium (ose) at P3-P7. Oatp1c1 expression was observed at fibrocytes of the spiral ligament and the spiral limbus (not shown) and at the tbc at P7-P15.

Our current understanding of the impact of MCT8 deficiency on the hearing capacity of AHDS patients remains limited and incomplete. Hearing loss is not usually a feature of AHDS but has been reported in some patients (4, 43). The relevance of the current findings to human inner ear development should be explored. Another aspect to consider is whether AHDS patients maintain their hearing capacity throughout their entire lifespan or whether they are prone to hearing loss due to deficient TH signalling.

Vision

TH action is also essential for retinal development and maturation (44). The retina is a light-sensitive tissue layer that contains specialised cells that transform light into electrical signals, ultimately transmitting visual information to the brain. Rod and cone photoreceptors are specialised sensory neurons that initiate phototransduction in the retina. Rods, the most abundant cell type in the retina, mediate dark-adapted vision. Conversely, cones account for just a small fraction of the overall photoreceptor population and mediate colour vision during bright-light conditions. In rodents, cones express different opsin photopigments that are either sensitive to middle (M, green) or short (S, blue) wavelengths (45).

Impaired retinal development and function have been observed in different animal models as a result of developmental hypothyroidism (46), adult-onset hypothyroidism (47), as well as a consequence of deletions in TRs (48), deiodinases (49, 50), and TH transporters (51, 52). Developmental hypothyroidism has been shown to alter the layering of the retina and differentiation of retinal cells (46), while adult hypothyroidism affects opsin expression patterns (47).

Expression pattern studies of TRs in chick revealed that Thrα1 is expressed in all retinal layers, Thrβ1 is present at the outer and inner nuclear layers during development and that Thrβ2 expression is restricted to the outer nuclear layer of the retina containing the developing photoreceptors (53). TRβ2-deficient mice present a decrease in M-cones and an increase in S-opsin cones, indicating that binding of TH to TRβ2 promotes M-cone identity (48). Expression pattern studies of deiodinases, also performed in mice, identified a peak in Dio3 expression preceding an increase in the expression levels of Dio2 during retinal development, suggesting opposite patterns of developmental expression for Dio3 and Dio2 (50). Dio2 has been detected in Müller glia (54), in cone photoreceptors in contrast to rods (55), and at the inner nuclear layer (50), with its expression peaking at juvenile stages between P11 and P24 in mice (55). DIO3-deficient mice were shown to lose both M-cones and S-cones, while rod photoreceptors remained intact (49). Interestingly, cone development and differentiation were preserved in mice lacking both DIO3 and DIO2, indicating counterbalancing roles for these deiodinases in regulating TH action (50). All of this could indicate that the binding of T3 to TRβ2 would be essential to promote the correct development of cone photoreceptors and that DIO3 could protect from excessive TH action during this process.

There is more limited understanding regarding the transport of TH into the retina. The blood–retina barrier (BRB) consists of two distinct layers: the outer BRB, found at the retinal pigment epithelial (RPE) cell layer, controls the transfer of substances and nutrients from the choroid to the sub-retinal space. In contrast, the inner BRB, similar to the blood–brain barrier, is situated in the inner retinal blood vessels and consists of the endothelial cells lining these vessels. To summarise experimental evidences found in the literature so far, a model for the location of TH transporters in the retina is proposed in Fig. 2. OATP1C1 has been observed at the abluminal and luminal membrane of the inner retinal blood vessels of the inner nuclear layer and inner and outer plexiform layer of the rat retina. OATP1C1 was also observed at the basolateral membrane of the rat RPE (56). MCT8 expression in the early postnatal mouse brain was observed in the interface between the RPE cell layers and the photoreceptors, suggesting that other TH transporters, such as OATP1C1, could regulate the access of TH from the blood to the RPE. Moreover, MCT8 was also identified at cells of the inner nuclear layers and the ganglion cell layer. Contrary to photoreceptors, cells in the inner nuclear layers and the ganglion cell layer are not connected to the RPE, and the access of TH to these cells would be modulated by the inner BRB, where MCT8 expression was not found (57). Henning and Szafranski (57) also found higher MCT8 expression in juvenile mouse retinae in comparison to adults, suggesting a role for MCT8 in the postnatal maturation of the retina. In chicken, MCT8 has been found in the retina from embryonic day 6 (E6) throughout all embryonic development, suggesting a role of MCT8 in the regulation of TH during retinal development (58). Vancamp et al. showed that knocking down MCT8 in chick retinal precursor cells during early development leads to reduced proliferation of retinal precursor cells and decreased thickness of the retina; however, a balanced proportion of the different generated cell types suggested no defects in differentiation. Interestingly, while the proportion of rods was unaffected, more S-cones were found at the expense of M-cones (51), resembling findings in TRβ2-deficient mice, although with a milder phenotype. Recent transcriptomic profiling analysis by Rozenblat et al. (52) in mct8^−/− zebrafish larvae revealed altered expression of vision-related genes, including downregulation of opn1mw2 (encoding a medium-wavelength sensitive opsin) (52), consistent with findings in MCT8 knockdown chicks. In addition, these authors found a reduction in the number of saccades and pursuits as well as alterations in the activity of pretectal neurons, suggesting that Mct8 deficiency might alter the development and function of these neurons and lead to impairments in conjugate eye movements.

Figure 2.

Proposed location of thyroid hormone (TH) transporters in the retina based on current findings. Akanuma et al. (56) found OATP1C1 at both at the inner and outer blood–retina barrier, at the retinal pigment epithelium and the inner retinal blood vessels. Henning et al. (57) found MCT8 present at the outer blood–retina barrier, at the interface between the retinal pigment epithelium and the photoreceptors. MCT8 was also been found in the inner nuclear and ganglion cell layers, but not expressed at the inner blood–retina barrier. AC (amacrine cell), BC (bipolar cell), GC (ganglion cell), HC (horizontal cell), MG (Müller glia).

Eye disorders have been documented in some AHDS patients including anisocoria (59, 60) and even blindness (5, 60). Notably, one MCT8-deficient patient developed severe visual impairments during his clinical follow-up (61). However, due to the limited number of documented cases of MCT8 deficiency associated with visual impairments, it is not possible to confirm whether these impairments are direct causes of mutations in MCT8. The data available regarding the role of MCT8 in the retina are very scarce and many questions are still unanswered. For example, even though AHDS patients do not present severe visual impairments, could there be an increased susceptibility to conditions such as colour blindness and impairments in conjugate eye movements? Gathering further insights on the role of MCT8 in the retina will shed more light regarding the potential visual alterations in AHDS.

Olfaction and taste

TH also plays a role in the development and functioning of olfactory neurons and taste buds (62, 63). Adult-onset hypothyroidism has been shown to disrupt the differentiation of olfactory neurons in the olfactory bulb (OB), leading to decreased thickness of the olfactory epithelium and anosmia (64). TH deficiency also seems to alter taste preference behaviour in adult rats (65).

The olfactory system is formed by olfactory sensory neurons (OSNs) which axons project to the OB, initially through the olfactory nerve layer (ONL) before entering the glomeruli. Within this structure, OSN axons establish synapses with mitral (located in the mitral cell layer, MCL) and tufted cells (located in the external plexiform layer, EPL). Every mitral and tufted cell extends a long axon along the olfactory tract. The olfactory tract distributes these axonal fibres to the cortex of the pyriform lobe, which is the cortical destination for the olfactory pathway (66).

In agreement with the identified role for TH in olfactory neurons, Thrb expression has been identified in immature olfactory neurons of chum salmon (67); however, to our knowledge the effects of deletions of TR in olfactory performance have not been assessed. Dio3 is highly expressed during development in the mouse OB (68) and DIO3-deficient female mice exhibit impaired olfactory function (69). Even though astrocytes are found in all layers of the OB (70), Dio2 expression has been observed more prominently in the EPL of the rat OB (71). MCT8 has been found to be highly expressed in the anterior olfactory nucleus and in the granular cell layer of the mouse OB (72). Studies exploring the olfactory behaviour in mice lacking MCT8 and OATP1C1 have identified alterations in short-term olfactory memory although this has been associated with reduced adult neurogenesis in the subventricular zone and not with alterations in the OB itself (73).

It is known that THs play an important role in the maturation and specialisation of taste buds (62), but the mechanisms for TH action in these cells have not been explored. Taste buds are composed of type I cells, which are glial-like cells that may transduce salty taste; type II cells that express GPCR receptors and are possibly involved in mediating sweet, umami and bitter tastes; and type III presynaptic cells that are possibly involved transducing sour taste. Type II cells are thought to communicate with type III cells through P2Y adenosine receptors, and ultimately, type III cells release serotonin to afferent neurons to facilitate taste signal transmission to the brain (74). To our knowledge, there are no studies identifying the location or the role of TRs, deiodinases or TH transporters in taste buds. However, the effect of thyroid function in taste has been studied in the context of the burning mouth syndrome, where patients show low levels of TH and high levels of TSH, that lead to hypogeusia and burning sensations (75).

There are no reports of alterations in the smelling capacity or taste in AHDS patients. However, since alterations in these sensory capacities, while significant in their own, may have received less attention compared to the more incapacitating neuromotor and cognitive impairments experienced by these patients, could these might have gone unnoticed?

Touch

There is very little information regarding the role of TH in mechanoreception and nociception. In one study, Yonkers and Ribera (76) found that, in zebrafish, T4 administration rapidly modulated voltage-gated sodium currents in the mechanoreceptive Rohon–Beard sensory cells, indicating that T4 might be regulating mechanoreception by non-genomic actions.

TH action might also be involved in nociception. It has been described that the sensitivity to nociceptive inhibitory agents, such as adenosine, is increased in response to hypothyroidism, as TH modulate the development, expression and function of adenosine A1 receptors in rats (77). On the other hand, hyperthyroidism in rats has been shown to modify the hydrolysis of AMP to adenosine within synaptosomes of the spinal cord, resulting in either analgesic or hyperalgesic responses depending on the timing of hyperthyroidism onset (78). In addition, TH deficiency during rat intrauterine development seems to reduce the thermal nociceptive threshold and to increase thermal sensitivity, without affecting mechanical nociceptive responses (79).

In view of the findings described previously, and since MCT8 deficiency leads to low T4 circulating levels and brain content, could mechanoreception potentially be affected in AHDS patients? In addition, could defective TH signalling lead to changes in pain perception in these patients?

Conclusion

Ng et al. (80) pointed out that the senses are our window to the world, our connection to our environment, and the foundation of our communication with one another. THs regulate all senses, however, how these might be affected in AHDS patients has not been thoroughly explored. Understanding the relevance of MCT8-mediated TH transport in sensory systems could help to address the phenotypical complexity of the AHDS. In this review, we have gathered data regarding TH action in hearing, vision, smell, taste and touch in different animal models that could suggest how MCT8 deficiency might impact these senses.

One of the main limitations in this area of research is that, while the location and expression of some TH receptors and transporters has been described in the sensory systems, the precision of these findings is limited. For instance, most of the results found in the literature reported the use of antibody-based techniques, which might be affected by the low expression of the proteins and/or the lack of reliable antibodies. Others have used in situ hybridisation approaches with low cellular resolution. In consequence, it is not possible to reach definitive and unequivocal conclusions about the exact locations of TH receptors and transporters in the different biological and sensory systems studied. Of interest, very few studies have focused on addressing the mechanisms of TH transport in the sensory systems in the recent years. Future studies using cell-specific and increased sensitive technologies, such us sorting and/or single cell analyses, may help to answer these relevant questions.

In particular, there is not enough data regarding the role of MCT8 in sensory inputs. With this limited data, it is not possible to discern whether the few sensorial defects described in these patients are direct consequence of MCT8 mutations or not. Another issue to consider is that, should these senses be affected, the manifestation of such effects could vary across patients. In addition, the complex TH profile of MCT8-deficient patients characterised by brain hypothyroidism and peripheral hyperthyroidism raises questions about the potential impact of T3 excess on the development of sensory functions, such as touch which relies not only on the central nervous system but also on the peripheral nervous system.

In short, the role of MCT8 in sensory inputs remains an open and interesting question worth exploring. In view of the available data, could MCT8-deficient patients be prone to hearing loss, colour blindness, impairments in conjugate eye movements, dysosmia and dysgeusia, or impaired pain perception? Addressing these potential impairments could ultimately contribute to the well-being of patients.

Declaration of interest

The authors declare that the study was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

This study was supported by grants from the Spanish Ministry of Research and Innovation MCIN/AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR (Grant No IJC2020-043543-I) to SB-L, and MCIN/AEI/10.13039/501100011033 and ‘ERDF A way of making Europe’ (grant no. PID2020-113139RB-I00 to AG-F.), a grant from The Sherman Foundation, Australia (grant no. OTR02211 to SB-L and AG-F) and from the Consejo Superior de Investigaciones Científicas, Spain (grant no. 2020AEP044 to AG-F). MG-Y is a recipient of a fellowship from the Programa de Formación de Profesorado (FPU, FPU19/02006), Plan Estatal de I+D+I of the Ministerio de Universidades, Spain. VV-H is a recipient of a contract from MCIN/AEI/10.13039/501100011033 and by ‘ESF Investing in your future’, Spain (grant no. PRE2018-086185). AG-A is a recipient of a contract FPI-UAM from Universidad Autónoma de Madrid, Spain.

Author contribution statement

AG-A: writing – original draft; MG-Y: writing – original draft; VV-H: writing – original draft; AM-P: writing – review and editing; AG-F: writing – review and editing; SB-L: writing – review and editing, generation of figures, conceptualisation.