Startle Disease: New Molecular Insights into an Old Neurological Disorder

Startle disease (SD; hyperekplexia, stiff baby syndrome, OMIM 149400) is a rare neurological disease known since the 1960s. This rare neuromotor disorder affects spinal inhibitory motor circuits, causing increased muscle tone and exaggerated startle reflexes. However, not every excessive startle reaction is caused by SD. Three main characteristics determine SD: severe muscle stiffness in neonates, enhanced startle reflexes in response to unexpected acoustic or tactile stimuli, and generalized stiffness following the startle response without any possible voluntary movement (Tijssen and others 2002). However, the severity of symptoms can vary considerably among affected individuals, ranging from generalized muscle stiffness to unprotected falls and injury or even seizures and/or fatal apnea episodes (Mine and others 2015).

The major causes of SD are genetic defects in GLRA1 and GLRB, encoding the postsynaptic glycine receptor (GlyR) α1 and β subunits, or SLC6A5, encoding the glycine transporter 2 (GlyT2; Figs. 1, 2; Rees and others 2006). Importantly, symptomatic pharmacotherapy with benzodiazepines (e.g., clonazepam) is effective in patients with mutations in GlyR subunits or GlyT2 (Thomas and others 2010).

Mutations in GlyRs or GlyT2 reduce glycinergic neurotransmission, thus affecting the balance between excitation and inhibition in brainstem and spinal cord circuits. A reduction in the activity of glycinergic interneurons in the spinal cord results in increased muscle contraction and stiffness. Moreover, the massive and uncontrolled startle reaction observed in SD is due to an enhanced startle reflex mediated by brainstem nuclei (Schaefer and others 2012). Glycinergic dysfunction in brainstem nuclei is also thought to be responsible for neonatal apnea episodes or in rare cases of sudden infant death (Bode and Lynch 2014).

GlyRs enable fast synaptic inhibition in the central nervous system and belong to the superfamily of cys-loop receptors. Cys-loop receptors form pentameric ion channels that share a common disulfide bridge in the ligand-binding extracellular domain (ECD). The ECD is followed by four transmembrane (TM) domains connected by loop structures (TM1-2, TM2-3, and TM3-4) and a short extracellular C-terminus (Schaefer and others 2018a). The large intracellular domain (ICD) exhibits the highest variability among GlyRs and determines subunit-specific properties, such as phosphorylation, ubiquitination, and interaction with intracellular binding partners (Langlhofer and others 2020). GlyRs are activated by glycine, β-alanine, and taurine and antagonized by strychnine, an alkaloid from Nux vomica (Schaefer and others 2012). GlyRs can form homomeric (α1-4) or heteromeric (αβ) receptors. Cryo-EM (electron microscopy) structures suggest a stoichiometry of native GlyRs with 4α:1β (Fig. 3; Yu and others 2021; Zhu and Gouaux 2021).

GlyT2 belongs to the family of neurotransmitter sodium symporters and regulates the concentration of synaptic glycine via reuptake to presynaptic terminals. GlyT2 is a 12-TM protein with intracellular N- and C-termini (Kristensen and others 2011). Alterations in transport mechanisms upon GlyT2 mutations in SD have been studied by homology modeling and molecular dynamic simulations using structures of the bacterial leucine transporter, the dopamine transporter, the human serotonin transporter, and the recently solved X-ray structure of the human GlyT1 transporter (Carta and others 2012; Shahsavar and others 2021).

Dysfunctional glycinergic synapses in the spinal cord and brainstem underlie SD pathology at the molecular level. Excited ventral spinal cord motoneurons fire action potentials to the neuromuscular endplate with signal propagation via cholinergic pathways to the muscle. This process is controlled by feedback inhibition via inhibitory interneurons (Schaefer and others 2012). Our classical view of glycinergic synapses is based on presynaptic packaging of glycine into vesicles mediated by the vesicular inhibitory amino acid transporter (VIAAT/VGAT), GlyT2 as a pivotal transporter for glycine reuptake to the presynaptic compartment, and the postsynaptic chloride-permeable GlyR anchored by the scaffolding protein gephyrin to the cytoskeleton (Fig. 1; Chapdelaine and others 2021). However, this model has been extended by the existence of homomeric presynaptic GlyRs, which contribute to SD pathology; the presynaptic Asc-1 (alanine-serine-cysteine transporter 1), which is important for glycine and serine transport; the presence of extrasynaptic GlyRs, also coupled to gephyrin; and a possibly modulatory role of the interaction partner syndapin I in SD (Figs. 1 and 2, Table 1; Chapdelaine and others 2021; Langlhofer and others 2020; Safory and others 2015; Xiong and others 2014).

Although known since the 1960s, the genetic basis for SD was defined with the discovery of mutations in GLRA1 (Shiang and others 1993) encoding the inhibitory GlyRα1 subunit. Since then, >50 mutations in GLRA1 have been reported, making mutations in this gene the most common cause of SD, followed by SLC6A5 and GLRB. The first mutations in GLRB were discovered by Rees and others (2002), while mutations in SLC6A5 were first reported in 2006 (Rees and others 2006), with further cohort studies reported in 2012 (Carta and others 2012; Gimenez and others 2012; Fig. 2, Table 1). At the molecular level, mutations have been classified into dominant, recessive, and gain of function (Bode and Lynch 2014).

Novel GlyRα1 subunit mutations discovered in patients or mice with SD in the last 5 y include D70N, Q177K, L224X, R316X, P366L, A384P, and I401N (Fig. 3A–C, Table 2). Among them, D70N/R316X in a compound heterozygous form and I401N as a single mutation or in a compound heterozygous form with L224X have not been functionally characterized (Zhan and others 2020; Zhang and others 2020).

The murine GlyRα1 mutation Q177K and the human variant P366L have been extensively studied with single alterations that for themselves were unable to explain the SD phenotype. Both GlyRα1 mutants are examples of the complexity of SD since they cause small but significant impairments in glycinergic neurotransmission. These GlyRα1 variants are detailed in the next section (Langlhofer and others 2020; Schaefer and others 2017; Schaefer and others 2018a).

The first monozygotic twins with SD homozygous for an intronic GLRB splice site mutation (IVS5+5G→A) were recently identified (Estevez-Fraga and others 2020). This mutation had previously been described demonstrating that the intronic IVS5 mutation decreased splicing efficiency of GLRB exon 5 (Rees and others 2002). The homozygous carriers of IVS5+5G→A therefore represent the first human correlate of the spontaneous mouse mutant spastic. In spastic, an intronic LINE1 element insertion resulted in aberrant splicing and significantly reduced GlyRβ levels (Schaefer and others 2012). Homozygous GLRB (R190X) was identified in a 28-d-old infant with massive stiffness episodes and a nonhabituating nose-tap response. At the age of 4.5 y, intermittent bouts of tonic spasms still required enhanced doses of clonazepam for treatment (Table 3; Gupta and others 2020).

New GlyT2 mutations have been reported in a novel compound heterozygous case exhibiting a decline in severity of symptoms and frequency with age (S477Ffs9X/S477P; Dafsari and others 2019). A novel case with a P429L missense mutation included severe neuromotor deficits (Kitzenmaier and others 2019), as supported by functional studies demonstrating that GlyT2^P429L was expressed at the cell surface but nonfunctional in terms of glycine uptake.

To explain SD symptoms, unaffected expression and small functional changes are insufficient. New knowledge on the following has enhanced our current knowledge of SD pathomechanisms at the molecular level: mutants causing spontaneous opening, prolonged open times with subsequent temporal increases in intracellular chloride levels, changes in protein-protein interactions with identification of novel GlyR interactors, and the importance of presynaptic homomeric GlyRs.

Until 2015, no X-ray or cryo-EM structures of the GlyR were available. First insights into the organization of cys-loop receptor ECDs came from the X-ray structure of the homologous acetylcholine-binding protein isolated from Lymnaea stagnalis (Brejc and others 2001). The cryo-EM structures of the zebrafish GlyRα1 and the X-ray structure of human GlyRα3 revolutionized our current knowledge on GlyR receptor states, such as the open/closed and desensitized/partially desensitized configurations. In addition, residues involved in structural transitions following ligand binding to ion channel opening as well as residues participating in neurosteroid, picrotoxin, and anesthetic binding have been identified (Fig. 2; Du and others 2015; Huang and others 2015). These structures provided significant impact on additional interpretations of observed physiological changes for SD mutants and their contributions to the disease phenotype. The strychnine-bound GlyR structure revealed electrostatic attraction of Q226E in TM1 to R271 in TM2 of the adjacent subunit, causing a tilt of TM2 away from the pore axis and explaining the constitutive open channel observed for this SD mutation. The ivermectin/glycine-bound GlyR illuminated the role of SD mutations localized at the ECD–TM domain interface and how SD mutations such as R218Q in pre-M1 or Y279C in the TM2-3 loop disturb their interactions with residues in the cys-loop followed by transitional block upon agonist binding. The glycine-bound open GlyR structure clarified the modified interactions between adjacent pore-lining mutated residues in SD (e.g., Q266H and R271Q/L) and their influence on ion channel properties. Moreover, these structures enabled predictions of altered stacking interactions within TM and between TM regions for novel GlyRβ SD mutations and their functional consequences (Piro and others 2021). The correct GlyR stoichiometry has been a long debate. The recently solved cryo-EM structures of heteromeric GlyRs suggest a 4α:1β stoichiometry (Yu and others 2021; Zhu and Gouaux 2021) and will shed light on our interpretations and understanding of postsynaptic changes at glycinergic synapses under healthy and SD conditions.

However, almost all structures lack the large ICD where some SD mutations are localized. The ICD was replaced by a short peptide sequence present in bacterial cys-loop receptor homologs GLIC and ELIC. Kumar and others (2020) used the full-length human GlyRα1 for structural investigation, but the ICD was unstructured except for small domains following TM3 and before TM4, which showed α-helical arrangements. Similar observations have been made following the analysis of the cryo-EM structure of heteromeric GlyRs (Zhu and Gouaux 2021). The largely disordered ICD might also represent a consequence of the lack of intracellular postsynaptic proteins during structural investigations. Even if unfolded or disordered, the ICD structure is essential to understand intracellular receptor modifications induced by SD mutations at the ECD.

The structure of GlyT2 has still not been resolved. The X-ray structure of the GlyT1 transporter was very recently solved with a resolution of 3.4 Å (Shahsavar and others 2021). GlyT1 has a 50% homology to GlyT2 at the nucleotide and amino acid level. As such, the solved GlyT1 structure will now allow more precise interpretations of GlyT2 SD mutation pathomechanisms.

Another obvious gap is the missing X-ray structure of full-length gephyrin. Researchers have focused on the structure of the gephyrin E-domain, which is responsible for the interaction with defined motifs in the ICD of GlyRβ or GABA_A receptor subunits α1, α2, and α3 (Maric and others 2017). Protein-protein interactions involve short linear motifs. The use of alternative approaches, such as temperature-related intensity change, allowed the detection of low- and high-affinity protein-peptide interactions between gephyrin and GlyRβ validated by binding to recombinant and native GlyRs (Maric and others 2017; Schulte and others 2021). Although such methods have limitations, the offered high throughput of temperature-related intensity change allows a rapid detection of protein-binding profiles. Such studies that elucidated the protein interactions with different affinities between gephyrin and the GlyR enhanced our knowledge on synaptic exchange rates, receptor fields, and dynamics and are essential to further understand cooperative influences of extracellular GlyR mutations to the binding of intracellular protein partners, which may differ under disease conditions such as SD.

Recent findings on GlyRα1 and β SD mutations, although located in the ECD or TM domains, have pointed toward alterations of GlyR–accessory protein interactions, which in turn affect glycinergic function. How extracellular or TM mutations influence intracellular protein-protein interactions remains elusive, but in our view, this is likely to be linked to changes in GlyR conformation. Moreover, our current knowledge on receptor-lipid interactions and their relevance for stabilization of receptor states and receptor gating transitions is still limited. Finally, the use of three-dimensional (3D) cell culture models might reflect a suitable alternative to study disease mechanisms of rare diseases and their molecular pathologies.

The GlyR ICD lacks cryo-EM densities (Figs. 2 and 3; Zhu and Gouaux 2021). Similar findings have been observed for the closely related heteromeric GABA_AR (α1β3γ2L) demonstrating a disordered TM3-4 loop, which likely reflects a lack of interacting postsynaptic proteins (Laverty and others 2019). Therefore, the solved GlyR structures obtained to date are unable to provide hints for structural transitions transmitted from an ECD mutation toward intracellular binding partners. Thus, there is still an important gap between structural knowledge and findings from in vitro molecular, cellular, and protein biochemical approaches as well as physiological assessments.

Protein-protein interaction as the tight association between GlyRβ (residues 378–426) and gephyrin and its necessity for synaptic localization is well characterized (Schrader and others 2004). The binding between gephyrin and GlyRβ is bimodal with a high- and a low-affinity binding sites. The residues responsible for the high-affinity gephyrin-GlyRβ interaction have been mapped to C-terminal residues of the GlyRβ binding core (residues 394–413), while the low-affinity binding site is formed by N-terminal residues within the binding core. High-affinity gephyrin binding is involved in GlyR confinement at synaptic and extrasynaptic sites. Low-affinity binding rather contributes to GlyR confinement by fine-tuning synaptic clusters (Grunewald and others 2018). High- and low-affinity binding of gephyrin to GlyRs is in line with the dynamic organization of different subpopulations at inhibitory synapses, a tight interacting receptor-scaffold domain and a more loosely bound receptor scaffold population, of reciprocally stabilized GlyR and gephyrin proteins (Chapdelaine and others 2021). By using high-resolution microscopy techniques, such as single-molecule localization microscopy and correlative light and electron microscopic analysis that enable high spatial resolution, the GlyR-scaffold occupancy and receptor densities at glycinergic synapses have been investigated. A constant packing of 2000 GlyRs µm^-2 at spinal cord synapses throughout adulthood has been estimated. This number of GlyRs per synapse did not change in analysis of heterozygous oscillator mice that lacked 50% of the GlyRα1 subunit. A significant decrease was found in synapse size in heterozygous oscillator animals, arguing that under disease conditions the morphology and size of glycinergic synapses might play a key role in governing glycinergic postsynaptic strength in spinal cord circuits (Maynard and others 2021).

Furthermore, the protein-protein interaction between GlyRβ and syndapin I as well as between GlyRα1 and syndapin I but with lower affinity seems to be important for glycinergic synapse organization (Del Pino and others 2014; Langlhofer and others 2020). Similar to GlyRβ decoupled from gephyrin, syndapin 1 knockout increases GlyRβ mobility. Yet, gephyrin decoupling from GlyRs increases GlyR-syndapin interactions and thus regulates fine-tuning of synaptic efficacy at inhibitory synapses (Troeger and others 2021). Thus, gephyrin and syndapin I seem to cooperate in receptor scaffolding functions, regulating the size and density of GlyR clusters. It also seems clear from recent GlyR SD mutations that the impairment of this postsynaptic fine-tuning underlies or at least contributes to SD clinical phenotypes.

Besides protein-protein interaction, protein-lipid interaction might interfere and/or stabilize receptor states. Lipid densities have been detected close to TM domains on the extra- and intracellular sides of the bilayer (Fig. 2; Laverty and others 2019; Zhu and Gouaux 2021). Molecular dynamic simulations for the GlyR have shown that phosphatidylcholine interacts with receptor regions involved in structural transitions during the gating process, such as loop C, pre-TM1, β8-β9 loop, β1-β2 loop, cys-loop, and TM2-3 linker. In contrast, phosphatidylserine has been found closer to the intracellular side bound to TM3 and TM4 and coincides with the PIP2 binding site (phosphatidylinositol-4,5-biphosphate; Damgen and Biggin 2021). As lipids might interfere with receptor gating, the identification of protein-lipid interaction sites may become more important in the future as a prerequisite to design lipid-like allosteric drugs.

Three-dimensional cell culture models to study disease mechanism such as SD may become an important tool in the future. Three-dimensional cell culture models can be set up from various neuronal subtypes and can be easily targeted by viral infections to introduce SD mutations of interest. So far, SD mutations are rarely studied in a neuronal context and then only in 2-dimensional (2D) monolayers. Between in vitro 2D cell cultures and in vivo experiments, there is a large gap that might get filled by investigations of 3D cell culture disease models. The importance of the third dimension has been pointed out in a 3D spinal cord neuronal model comparing synapse development in 2D with 3D (Fig. 2; Fischhaber and others 2021). This model suggests that synaptic maturation and network formation in 3D cultures are faster. A critical mass of interacting neurons is, however, essential to detect disease phenotypes and study protein-protein interactions in the 3D context. However, there are still several issues to overcome until such 3D disease models will represent suitable tools and an alternative for animal experiments. In 3D, reproducible hydrogel compositions resembling the native extracellular matrix need to be defined that allow neuronal differentiation and possibly redifferentiation of neurons from patient-specific stem cells.

This review provides an update on novel SD mutations within the last 5 y. Increasing knowledge at the structural level for homomeric and heteromeric GlyRs and new mouse models have enhanced top-down structure-function approaches for SD mutations. For future investigations, main directions can be defined as follows: the identification of mechanisms that underlie alterations of intracellular interactions with GlyR accessory proteins by extra- or intracellular SD mutations, the study of GlyR-protein/lipid interactions and how they are affected by SD mutations, the precise contribution of presynaptic GlyRs to SD phenotypes, the exploration of pharmacotherapies suitable for overcoming SD mutations resulting in trafficking defects for GlyRs or GlyTs, and the establishment of suitable 3D cell culture models to further ascertain our molecular understanding of SD mutations.

The authors thank Anna Wenzl from the design department at the University Hospital Würzburg for her assistance in figure design.