Molecular insights into myocilin and its glaucoma-causing misfolded olfactomedin domain variants

Conspectus

Numerous human disorders arise due to the inability of a particular protein to adopt its correct three-dimensional structure in the context of the cell, leading to aggregation. A new addition to the list of such protein conformational disorders is the inherited subtype of glaucoma. Different and rare coding mutations in myocilin, found in families throughout the world, are causal for early-onset ocular hypertension, a key glaucoma risk factor. Myocilin is expressed at high levels in the trabecular meshwork (TM) extracellular matrix. The TM is the anatomical region of the eye that regulates intraocular pressure and its dysfunction is associated with most forms of glaucoma. Disease variants, distributed across the 30 kDa olfactomedin domain (mOLF), cause myocilin to be sequestered intracellularly instead of being secreted to the TM extracellular matrix. The working hypothesis is that the intracellular aggregates cause a toxic gain of function: TM cell death is thought to lead to TM matrix dysfunction, hastening elevated intraocular pressure and subsequent vision loss.

Our lab has provided molecular underpinnings for myocilin structure and misfolding, placing myocilin-associated glaucoma within the context of amyloid diseases like Alzheimer and diabetes. We have dissected complexities of the modular wild-type (WT) myocilin structure and associated misfolded states. Our data support the model that full-length WT myocilin adopts a Y-shaped dimer-of-dimers conferred by two different coiled-coil regions, generating new hypotheses regarding its mysterious function. The mOLF β-propellers are paired at each tip of the Y. Disease-associated variants aggregate because mOLF are less stable, leading to facile aggregation under physiological conditions (37 °C, pH 7.2). Mutant myocilin aggregates exhibit numerous characteristics of amyloid in vitro and in cells, and aggregation proceeds from a partially-folded state accessed preferentially by disease variants at physiological conditions. Interestingly, destabilization is not a universal consequence of mutation. We identified counterintuitive, stabilizing point variants that adopt a non-native structure and do not aggregate; however, these variants have not been identified in glaucoma patients. An ongoing effort is differentiating pathogenic from benign variants. This effort is relevant to interpreting data from large-scale sequencing projects where clinical and family history data are not available. Finally, our work suggests avenues to develop disease-modifying precision medicines for myocilin-associated glaucoma.

Introduction

The inability of a polypeptide chain to adopt its native three-dimensional structure in a cellular context is associated with a wide range of human diseases. Although a consequence of such protein misfolding can be protein degradation and subsequent loss of function, such as in the case of Gaucher disease,⁵ an increasing number of diseases are associated with a toxic gain of function (GOF) caused by protein aggregation.⁶ One feature prevalent among GOF misfolding diseases is the formation of a templated aggregate called amyloid, which can occur with proteins that are intrinsically disordered or from those that adopt a native 3D structure,⁶ and can arise due to aberrant cellular environmental conditions as well as genetic mutations.⁷ Amyloid fibrils and other aggregated species can be toxic. Some disease-associated amyloids are thought to exert their cytotoxicity extracellularly⁸ whereas other aggregates accumulating within cells wreak havoc on cellular degradation systems, including ER-associated degradation (ERAD) and autophagy machinery, leading to pathogenic cytotoxicity.^9,10

Our lab has been studying a new addition to the growing list of protein conformational diseases, myocilin-associated primary open angle glaucoma (POAG). Autosomal-dominant inherited mutations in myocilin¹¹ account for 2 – 4% of the ~60 million cases of clinical POAG worldwide¹², leading to early onset of the key glaucoma risk factor, namely, elevated intraocular pressure (IOP) and subsequent vision loss.¹³ Myocilin is a component of the trabecular meshwork (TM) extracellular matrix, a tissue near the iris that is involved in regulating eye pressure (Fig. 1a).¹⁴ TM tissue dysfunction is associated with most forms of glaucoma.¹⁵ IOP arises as a result of the balance between producing and draining aqueous humor, a clear, colorless fluid that delivers nutrients to the tissues in the anterior eye. An imbalance of aqueous humor fluid outflow through the TM compared to its production in the ciliary body leads to elevated IOP, which is reduced with clinical treatments¹⁶. Despite its relevance to TM tissue and more than twenty years of research, the explicit physiological function(s) of wild-type (WT) myocilin in the eye remain(s) unknown. However, there is a general consensus on the overarching GOF pathogenic mechanism of mutant myocilin: disease variants of myocilin are sequestered intracellularly instead of being secreted to the TM extracellular space, leading to cell stress, cell death, and hastening of clinical glaucoma symptoms (Fig. 1b).¹⁴

Fig. 1. Overview of the eye and pathogenic role of myocilin.

(a) IOP maintenance is the balance between production and drainage of aqueous humor in the anterior eye segment. Production of aqueous humor occurs in the ciliary body, which then bathes anterior segment tissues before filtering through the trabecular meshwork and draining through Schlemm’s canal. Reproduced with permission from ref.¹⁷. Copyright 2016 American Chemical Society. (b) Myocilin is a secreted protein and component of the trabecular meshwork (left). Mutant myocilin accumulates in the ER alongside other proteins like the molecular chaperone glucose regulated protein 94 (Grp94) and thus is not secreted. This leads to cytotoxicity and cell death (right). Adapted from¹⁸. Copyright 2018 American Chemical Society.

More than 90% of the disease-relevant myocilin mutations are localized to its C-terminal olfactomedin domain (mOLF, Fig. 2).^13,19 Familial and population case-control genetic studies have documented nearly 100 mOLF variants in glaucoma patients as well as individuals without the disease,¹⁹ and many more variants of unknown pathogenicity are documented in large-scale sequencing databases (e.g. genome aggregation database, gnomAD²⁰, https://gnomad.broadinstitute.org/gene/ENSG00000034971?dataset=gnomad_r2_1). In line with a GOF mechanism, myocilin knock-out mice²¹ and individuals with truncation mutations that prevent myocilin translation,^22,23 do not develop glaucoma. As we elaborate below, upon introduction of disease-associated mutations, mOLF is thermally destabilized, increasing its propensity toward amyloid aggregation, conditions that can be induced in WT mOLF with mild environmental stressors that are relevant to glaucoma. Although misfolded proteins are typically targeted for degradation by the ER chaperone machinery, clearance of mOLF-resident myocilin disease variants appears to be hindered by their aberrant interactions with the molecular chaperone glucose regulated protein 94 (Grp94) (Fig. 1b).^24–30

Fig. 2. Myocilin structure.

(A) Gene structure with structural domains highlighted. (B) Overall myocilin structure is modeled as a Y-shaped dimer of dimers, conferred by its N-terminal CC and LZ regions. Over 90% of glaucoma-associated mutations are distributed across the 5-bladed β-propeller mOLF domain, and the consequence of missense mutations is destabilization (Fig. 3). Several unique features and variants discussed in this review are labeled. P1 and P3 are stretches within mOLF found to form amyloid with nanoscale morphologies similar to the full-length protein (Fig. 4).

Here we review our molecular biophysical data supporting the toxic-GOF hypothesis for mutant myocilin pathogenesis, its misfolding in the cellular context of full-length myocilin, and emerging therapeutic directions. We also demonstrate that predicting the effect of a given mutation is a challenge. Counter to our intuition, we found certain variants to stabilize mOLF. With large-scale sequencing data increasingly available and precision medicines on the horizon, the ability to predict the effect of a given coding mutation on myocilin protein behavior is of increasing clinical importance.

Structure of WT myocilin

Myocilin is a modular, multi-domain protein composed of a signal sequence followed by a coiled-coil (CC) region, a leucine zipper (LZ) domain, and ~30-kDa mOLF.³ To date, a structure of intact full-length myocilin is not yet available but we have amassed evidence for an overall Y-shaped dimer-of-dimers structural arrangement (Fig. 2) from a divide-and-conquer domain approach combining X-ray crystallography, synchrotron solution size-exclusion chromatography-small angle X-ray scattering (SEC-SAXS) experiments, and biochemical characterization. The SEC-SAXS envelope of a construct encompassing N-terminal residues 33 to 111 (Fig. 2) indicates a rod shape. Chemical crosslinking indicates the largest native oligomer is a tetramer and analysis of the role of the three cysteine residues located prior to the start of CC and capping LZ demonstrates that homotypic disulfide bonds stabilize only dimers in the native state. The SEC-SAXS envelope of a somewhat larger construct spanning residues 69–185, which encompassing both a small portion of CC and all of LZ, indicates a V shape, and a robust tetramer is generated with chemical crosslinking. Seven heptad repeats of LZ were crystallographically characterized from the mouse myocilin homolog of LZ, revealing a dimer capped by a disulfide bond. Connecting the LZ to mOLF is a ~60 residue linker. This portion of myocilin is predicted to be intrinsically disordered, and has thus far belied detailed characterization, mainly due to issues with degradation (unpublished observations). Finally, the 30 kDa C-terminal mOLF domain is a well-folded five-bladed β-propeller, distinct among structurally characterized five-bladed propellers (Fig. 2b). Each blade is composed of four antiparallel β-strands. A short α-helix is nestled at the side, between the most N-terminal blade and the final blade containing the N- and C-terminal molecular clasp.² The central hydrophilic cavity is capped by a long well-ordered loop consists of residues 360–379. Several residues within this loop appear to stabilize a curious internal cation-π interaction involving K423 and Y371, which in turn is within hydrogen bonding distance of D380, a key metal ion ligand. Indeed, two bound ions are coordinated within the hydrophilic cavity. The first is a heptacoordinate calcium ion ligated by the sidechains of Asp380, Asn428, and Asp478, plus two solvents and main chain carbonyl oxygens of Ala429 and Ile477. Prior metal analysis of mOLF confirmed stoichiometric levels of Ca²⁺ and established Asp380 as a ligand³¹. The second ion, modeled as sodium is coordinated by Asp380, Asp478, one solvent, and the carbonyl oxygen of Leu381. The true identity of this metal ion remains unknown; modeling was based on ligand and geometry considerations. The electrostatic surface potential indicates both positively charged and acidic patches on both the top and bottom face.² mOLF is isolated as a monomer by size-exclusion chromatography and only low levels of dimer are captured by chemical crosslinking³², suggesting that intrinsic dimerization is transient or weak. In sum, CC confers the molecular tetramer and LZ serves as molecular spacers: the tetrameric CC diverges into two parallel LZ dimer to form the Y, to which two pairs of the C-terminal mOLF are flexibly connected via the disordered linker.

mOLF misfolding and amyloid aggregation

The first indications that a toxic GOF accounts for the accelerated time course for glaucoma-associated IOP elevation came from early studies demonstrating that glaucoma-causing missense mutant myocilins are sequestered intracellularly^24,25,27 and aggregate into detergent-insoluble species³⁵. Notably, only missense disease variants within the mOLF domain share this misfolding consequence. Missense mutations outside of the mOLF domain, e.g. in CC, LZ or the unstructured linker leading to mOLF, do not exhibit an aggregation phenotype, although a change in tertiary structure³ and loss of adhesion³⁶ have been documented by us and others for selected variants. In addition, the effect of documented premature termination on glaucoma progression is still a topic of debate.^37,38

Ours were the first studies to explore the biophysical basis for the toxic gain GOF pathogenic mechanism by which missense mutations within the myocilin mOLF domain leads to misfolding (Fig. 3). Initially, we compared the thermal stability of WT (melting temperature (T_m) = ~52 °C, depending somewhat on buffer^1,31,39) with those of recombinantly-expressed and purified mOLF variants, showing that disease-associated variants are destabilized by ~5–12 °C (Fig. 3a).^1,31,33,40 Focusing on representative mild, moderate and severe variants A427T, D380A, and I499F, respectively, we observed that these variants adopt a partially folded state: tertiary structure differences were apparent from comparison of near-UV circular dichroism (CD) spectra acquired under ambient conditions (Fig. 3b), and less compact structures were also documented³³. Despite these differences, far-UV CD spectra are indistinguishable from those of WT, indicating that secondary structural elements remain largely unchanged. Based on these data, we proposed that the partially folded state(s) adopted by disease variants are populated under physiological conditions, although without accompanying structures, it is not known to what extent the partially folded states are similar. Still, these variants readily aggregate into a thioflavin T (ThT) positive fibrillar species with morphological hallmarks of amyloid at 36 °C and a neutral pH buffer containing 200 mM NaCl (Fig. 3c and see Fig. 4).

Fig. 3. mOLF destabilization and initial fibril formation.

(a) Thermal stability of mOLF variants. Red: severely destabilized variants; Blue: moderately destabilized variants; Green: weakly destabilized variants; Grey/black: WT-like stability. Reproduced with permission from ref¹. Copyright 2011 American Chemical Society. (b) near-UV spectra comparing WT OLF tertiary structure to that of representative disease variants A427T (mild disease), D380A (moderate disease), and I499F (severe disease), plus K398R, a neutral polymorphism variant. (c) Thioflavin T (ThT) fluorescence measured over time for variants shown in (b) upon incubation at 36 °C. Panels (b) and (c) are reproduced with permission from ref³³. Copyright 2014 Elsevier. (d) ThT fluorescence measured for WT mOLF upon incubation at 37 °C with gentle rocking. Reproduced with permission from ref³⁴. Copyright 2012 Elsevier.

Fig. 4. AFM images of mOLF-derived fibrils.

(a) Two end-point morphologies for WT mOLF accessed via different experimental conditions. (b) End-point fibril morphologies for variants A427T, D380A, I499F representing a range of disease severity (mild-to-severe) and displaying similarities with WT fibrils shown in (a). (c) End-point fibrils of two peptide stretches computationally predicted to form amyloid, again displaying similarities with WT fibrils in (a). Scalebar: 300 μm. Reprinted with permission from ref³³. Copyright 2014 Elsevier.

Compared to disease-associated mOLF variants, the well-folded WT mOLF β-propeller is more resistant to misfolding, but its intrinsic misfolding propensity can be accessed upon certain experimental conditions. While incubation of WT mOLF at 37 °C in neutral pH buffer for two weeks does not lead to aggregation, WT mOLF can be converted into a fibrillar morphology at 37 °C within two days, with the inclusion of gentle rocking (Fig. 3d).³⁴ The fibrillization lag phase can be bypassed if mOLF is incubated without rocking at 42 °C instead of 37 °C (e.g., see Fig. 5d),³³ upon the addition of low levels of detergent, oxidant, reductant, or acid, as well as seeding with pre-formed fibrils.³⁴ These data suggest that with select mild stressors, WT mOLF can access a low-lying excited state that share features with the partially-folded ground states of disease variants, leading to a fibrillar endpoint.

Fig. 5. Cellular properties of myocilin misfolding and role of Grp94.

(a) Thioflavin staining of CHO cells expressing myocilin(P370L), a severe glaucoma variant. Reproduced with permission from ref³⁴. Copyright 2012 Elsevier. (b) Model for pathogenesis of mutant myocilin misfolding and proposed role of Grp94. Adapted with permission from ref³². Copyright 2019 Nature Research. (c) Only depletion of Grp94 results in facile mutant myocilin(I477N) degradation. See also Ref²⁶. (d, e) Effect of Grp94 and its pharmacological inhibition on mOLF aggregation in vitro. Grp94 accelerates mOLF aggregation and results in coaggregation; inhibition with 4-Br-BnIm alters aggregation kinetics (d) and rescues coaggregation (e). In panel (d), S= soluble, W= wash, P=pellet. Panels (c) and (d) reprinted with permission from ref⁴¹. Copyright 2014 Oxford University Press. Panel (e) reprinted with permission from ref¹⁸. Copyright 2018 American Chemical Society.

Fibrils of WT mOLF and disease-causing variants exhibit two distinct morphologies observed by atomic force microscopy (AFM): long, straight fibrils or ones that appear as micron-length circles when deposited onto the mica substrate (Fig. 4a). Straight fibrils are observed for WT mOLF fibrillized at 37 °C with rocking, WT mOLF fibrillized at 42 °C at pH 7.2 and low ionic strength, as well as for mOLF variants A427T and I499F fibrillized at 36 °C at pH 7.2 with 200 mM NaCl (Fig. 4b). The circular fibrils, which enclose apparent oligomers, were observed for WT mOLF incubated at 42 °C and pH 7.2 with 200 mM NaCl as well as for D380A fibrillized as for the other two variants (Fig. 4a,b).³³ Additional support for the amyloid fibril features of these samples include FTIR spectral signatures,³³ and TEM, among other experiments.³⁴

Analysis of the mOLF sequence by various amyloid-prediction programs revealed three consensus peptide sequences with high propensity for amyloid of which two, G₃₂₆AVVYSGSLYFQ (P1) and V₄₂₆ANAFIICGTLYTVSSY (P3), experimentally mimicked the straight and circular fibrils, respectively (Fig. 4a, c).³³ Additional studies of these peptides in silico and by solid-state NMR reveal that P1 forms a largely homogeneous fibril species and P3 a more heterogeneous fibril mixture.⁴² Based on these results, we propose that P1 and/or P3 comprise the core amyloid region of mOLF fibrils, a prediction we hope to confirm experimentally.

The accumulation of mutant myocilin fibrillar aggregates^33,34, ER stress and cytotoxicity^25,30 indicate that clearance of misfolded myocilin by ERAD machinery is insufficient. Fibrils formed by the representative severe full-length mutant myocilin P370L expressed in CHO cells are robustly stained with ThT, whereas WT-expressing cells exhibited only background staining because myocilin is secreted (Fig. 1b, 5a).³⁴ We have further clarified that the failure of the ERAD cellular system to degrade mutant myocilin aggregates is at least partly due to the inability of the molecular chaperone Grp94 to triage mutant myocilin for degradation (Fig. 5b).²⁶ Grp94 is the ER-resident heat shock protein 90 (Hsp90) paralog, which acts in the late stages of the protein folding process. We showed Grp94 co-aggregates with mutant myocilin in cells⁴¹ and in a myocilin glaucoma mouse model.⁴³ When Grp94 is inactivated pharmacologically (not shown) or depleted with siRNA (Fig. 5c), mutant myocilin is degraded via autophagy.^26,41 Co-aggregation and pharmacological rescue of Grp94 can be recapitulated in vitro with purified proteins. The interaction between Grp94 and mOLF results in accelerated aggregation compared to mOLF-only conditions (Fig. 5d). The far-N-terminal unstructured residues of Grp94 contribute to the accelerated aggregation kinetics but removing these residues from Grp94 still results in co-aggregation. Only pharmacological inhibition, which should stabilize Grp94 in its closed state, reverts both aggregation kinetics and rescues co-aggregation (Fig. 5d,e).³² In support of the proposal that selective inhibition of Grp94 is a viable therapeutic approach for myocilin-associated glaucoma are preliminary in vivo studies in which the Grp94 inhibitor 4-Br-BnIm was administered to a mouse model of myocilin glaucoma, resulting in a reduction of IOP.⁴³ Our lab has used a medium-throughput screen to identify additional scaffolds for molecules that could rescue Grp94 from co-aggregation with mutant myocilin and promote its degradation,¹⁸ with a plan to extend to larger compound libraries.

Finally, we have compared human and mouse mOLFs. We were motivated by the weak phenotypes observed for mutant myocilin mice^44,45 and analogous studies relevant to other amyloid-forming proteins^46,47 where species variations lead to different outcomes. WT myocilin mOLF from mouse, which shares high sequence identity (87%) and whose structure is indistinguishable from human mOLF, has a slightly lower Tm (48 °C), resulting in faster aggregation.⁴⁸ We suggest that faster aggregation may reduce the lifetime of on-pathway intermediates, which are thought to be toxic in other amyloid diseases.^49–52 These observations further underscore the complexity of biophysical principles underlying amyloid formation and their clarification in model systems.

Stabilized mOLF variants

It is now well documented that disease-associated myocilin mOLF variants are destabilized and aggregation-prone compared to WT, but recently we discovered that not all variants are neutral or detrimental to stability. Some are notably stabilizing, even when physicochemical intuition suggests otherwise. When sequence conservation is mapped onto the mOLF protein structure, the interior residues are conserved, whereas surface residues are highly variable. In line with this observation, replacing the conserved calcium-coordinating residue Asp380 with Ser, Ala, or Asn results in the loss of metal binding and reduced mOLF stability, even though to date D380A is the only documented disease variant in the clinic. We reasoned the same would be true for substitutions of Asp478, another relatively conserved calcium ligand, albeit not clinically linked to glaucoma. Namely, we predicted that removing a stabilizing ionic interaction would reduce protein stability. Unexpectedly, substitution of Asp478 with Ser, Ala or Asn yields a more stable protein but with a non-native structure indistinguishable from disease variants by near-UV CD (Fig. 6a)⁴. The crystal structure of the D478S variant further shows a partially-folded structure (Fig. 6b). Counterintuitive to the observed increase in thermal stability but as expected from biochemical characterization, no Ca²⁺ ion was bound. A new metal ion center is formed by the remaining ligands, modeled as a Na⁺ site in the structure. Unexpectedly, we observed dramatic changes in secondary structure, including completely disordered side helix and neighboring loops within adjacent Blade A (Fig. 6b). A follow-up study to probe the relationship between the side helix and the environment of the central ion indicates crosstalk between the two structural elements⁵³, perhaps relevant to its still-unknown function. Moreover, D478S is also a corrector variant for cellular secretion, as it can be paired with glaucoma variants (e.g. D380A, P370L, or Y437H) to partially rescue thermal stabilities and cellular secretion. In fact, the double mutant D380A/D478S is more stable than WT and is able to resist unfolding at 37 °C and neutral pH conditions. Finally, we used the computational protein design server PROSS to improve stability, while retaining the native metal sites and avoiding introduction of known disease mutations. This led to the identification of a 21-variant mOLF with a remarkably high Tm, ~70 °C, nearly 20 °C more stable than WT.⁵⁴ The crystal structure of the 21-variant is essentially superimposable with that of WT mOLF (Fig. 6c). Taken together, the mOLF sequence has considerable capacity for additional stability, with or without retaining the metal center, suggesting at least in principle that pathogenic misfolding could be averted.

Fig. 6. Stabilized mOLF variants can adopt structures that compare favorably to WT or partially-folded disease variants.

(a) Near-UV CD spectra of destabilized disease variant D380A, stabilized variants D478N and D478S, and corrector double variants (D380A/D478N and D380A/D478S), which look similar to one another and deviate in tertiary structure from WT mOLF. Reproduced with permission from ref⁴. Copyright 2019 American Society for Biochemistry and Molecular Biology. (b) Superposition of WT and D447S mOLF. Despite higher thermal stability, D478S lacks calcium, has an unraveled the side helix and disordered neighboring loops (highlighted in yellow), plus a shift in blade A. (c) Superposition of WT mOLF with computationally-predicted variant containing 21 mutations, which retains the metal sites and native structure.

Concluding remarks

We have developed an overall molecular picture of myocilin, both WT and the effect of numerous variants. We have demonstrated a complex relationship among mutations, structure, aggregation, and cellular secretion. Disease-associated missense variants in mOLF adopt partially-folded structures with compromised thermal stability that renders them aggregation-prone clients of Grp94 and unable to be secreted or degraded. Near-native structure without accompanying compromised stability in mOLF does not appear to be pathogenic, likely because aggregation is inhibited.

Our structural model of WT myocilin has yielded some broad conjectures about its function. We have been attracted to the notion that myocilin function is tied to or affected by changes in the redox state of aqueous humor, in line with the knowledge that there are many potential sources of oxidative stress in the TM⁵⁵. Redox changes might shuffle inter-molecular disulfide bonds in the CC domain to purposely increase the oligomeric state of myocilin³, although to what end result is not clear. We have also considered a function related to mechanical contractility of the TM^56,57. Two findings support the notion that LZ can withstand mechanical deformation: LZ reversibly unfolds and refolds, and molecular dynamics simulations of LZ revealed an alternative structure with a bend and shift in interactions among residues near the LZ dimer C-terminus (Ser159-Gln170)³. Perhaps mechanical stretch of LZ triggers proteolysis in the linker^58,59 to release mOLF and transduce new downstream interactions or signaling events, as seen for the OLF domain of gliomedin⁶⁰. In addition to LZ, mOLF might be mechanosensitive: the helix unwinding observed in the mOLF calcium-depleted D478 variants suggests a similar conformational change could occur in a response to calcium flux. We can further envision a conformationally-selective interacting partner for mOLF, perhaps one with a leucine rich repeat domain akin to the scenario reported for OLF domain of latrophilin 3, which facilitates cell-cell communication^61,62. Finally, enzymatic activity for mOLF cannot be ruled out. The central cavity has a bridged dinuclear metal center and there is a proximal cavity occupied by glycerol in our structure². This cavity could bind a biological metabolite that could either trigger a conformational change in mOLF or result in a chemical transformation that in turn could trigger or propagate a signaling cascade. Unfortunately, without a strong knock-out phenotype in humans^22,63 and mice²¹, there is no clear path to cracking the mystery of myocilin function and it remains a major outstanding question in the field.

Our studies show that predicting how a given mutation will affect structure and stability is not straightforward, so a combination of clinical and molecular studies in the laboratory are needed to build a case for whether any given variant is pathogenic or benign. The main link to pathogenicity for a myocilin mutation comes from clinical data supporting early onset glaucoma and establishing a pattern of heritability within a family, but an increasing number of myocilin variants do not have accompanying clinical data because they are identified through large-scale genome screening efforts like gnomAD²⁰. In addition, myocilin is not a susceptibility gene for sporadic cases of glaucoma,^64,65 so the more common a variant is, the less likely it is statistically to be pathogenic. An open direction for future studies involves variants that result in premature termination, some of which are rather prevalent (approaching 1%) in the general population. With incomplete translation, mOLF will not adopt a native fold, but the complement of chaperones recruited will be different from those variants that adopt a near native state, changing the possible mode(s) of pathogenicity, e.g. to a disease modifier⁶⁶.

Myocilin-associated glaucoma is a newcomer to the list of disease-associated proteins that can form amyloid deposits. While mOLF aggregation falls into the broad (bio)chemical definition of amyloid, much is yet to be clarified. Currently, our ability to validate the amyloid hypothesis for myocilin glaucoma in a translational or clinical setting is limited. Rapidly-dividing model human cell lines such as CHO and HEK293 transiently transfected with mutant myocilins are generally not considered robust models for neuronal-like hTM cells⁶⁷, which cannot be transfected for analogous experiments, nor propagated long enough in the lab for gene editing techniques. Histochemical analysis of TM tissue from unique patients with P370L⁶⁸, T377M⁶⁸ or Y437H⁶⁹ myocilin variants show punctate myocilin staining and significant hTM cell loss; even if cells could be isolated from these exceedingly rare patient samples, culturing these primary cells to detect intracellular myocilin amyloid would likely be a significant challenge due to cytotoxicity of the mutant myocilin. In addition, the presence of fibrous proteins in the TM tissue, e.g. elastin and fibronectin, complicate the use of accessible amyloid reagents like ThT to visualize myocilin aggregates in any given complex sample such as a tissue from a trabeculectomy, a surgerical procedure used to reduce IOP. One option to circumvent these issues would be to generate myocilin antibodies selective for the amyloid conformation which then could be used to differentiate fibrillar myocilin from amorphous aggregates and aggregates or fibrils of other proteins, as pursued for example, for Alzheimer disease⁷⁰. In sum, much work is left to be done to evaluate whether myocilin-associated glaucoma might be considered a clinical amyloidosis (e.g. immunoglobulin light chain amyloidosis), a disease where protein misfolding plays a pathogenic role (e.g. tau, Aβ)⁷¹, or something new.

Given precedents from other diseases,^72–74 molecular insights into the pathophysiology of myocilin-associated glaucoma as well as associated therapeutic directions should be transferrable to glaucoma more broadly. WT myocilin is found at relatively high levels in the TM tissue^75,76, histopathology shows that myocilin accumulates in most forms of glaucoma⁷⁷, and the TM is generally diseased in glaucoma.¹⁵ In addition, we showed WT mOLF can misfold under known environmental stressors (oxidants, reductants, acidic pH) that lead to glaucoma^33,34. These observations suggest that glaucoma might be broadly considered a proteinopathy e.g. by the contribution of WT or mutant myocilin misfolding to TM dysfunction. We envision that WT myocilin aggregation could physically impair the ultrafiltration properties of the TM, thus contributing to TM dysfunction and IOP elevation. Significant additional studies are needed to investigate this speculation.

Our studies have unveiled three therapeutic avenues, all of which would constitute a disease-modifying treatment that significantly diverges from the current IOP management treatments and surgeries. First, given that myocilin is apparently non-essential,^21–23 an intervention that promotes mutant myocilin degradation would likely increase TM cell viability. While CRISPR-based gene editing is one direction,⁷⁸ our data suggest that Grp94 inhibition, a non-permanent molecular approach, may be attractive, particularly in low resource settings. Second, it may be possible to inhibit aggregation, likely via a small molecule stabilizer. In this pharmacological chaperone approach,⁷⁹ mutant myocilin will be stabilized and pass inspection by chaperones, preventing accumulation and enabling secretion. Since TM cells are highly phagocytotic,^80,81 once mutant myocilin is in the TM, it can be taken up for degradation, or perhaps can serve in its still-unknown functional capacity. This approach is particularly attractive as it could be applied to WT myocilin in the event a role for it in sporadic glaucoma could be confirmed. Third, we can envision clearing the toxic myocilin aggregate, as is being pursued (with varying levels of promise) for Alzheimer disease^70,82 and transthyretin amyloidosis⁸³, among others⁸⁴. With early genotyping of patients to identify pathogenic mutations, treatment could start before IOP elevation and irreversible damage occurs to the retina. Such disease-modifying precision medicine treatments would save the sight of several million individuals worldwide.

Biographical Information

Professor Raquel L. Lieberman obtained her BSc in Chemistry from MIT in 1998 and PhD in Chemistry from Northwestern University in 2005. After postdoctoral training in the area of structural biology and misfolding of neurodegenerative diseases at Brigham & Women’s Hospital/Harvard Medical School and Brandeis University from 2005–2007, Dr. Lieberman joined the faculty of the School of Chemistry & Biochemistry at Georgia Tech and rose through the ranks. The Lieberman lab’s research focuses on biophysical and structural characterization of proteins involved in misfolding disorders and ameliorating the misfolding phenotype by using chemical biology approaches. The long-term goal of the research program is to convert basic-research mechanistic discoveries into disease-modifying therapies as well as enhance the diversity of the STEM workforce.

Minh Thu (Alice) Ma moved from Vietnam to the USA during high school and received her B.S in Biochemistry in 2018 from Stetson University in Florida. Ms. Ma is currently a third-year PhD student in the School of Chemistry & Biochemistry at Georgia Tech. Her thesis research in Lieberman lab, supported by an NIH T32 Vision Sciences Training grant, deals with myocilin dynamics and conformationally-selective antibody development.