Transgenic zebrafish models reveal distinct molecular mechanisms for cataract-linked αA-crystallin mutants

Mutations in the small heat shock proteins α-crystallins have been linked to autosomal dominant cataracts in humans. Extensive studies in vitro have revealed a spectrum of alterations to the structure and function of these proteins including shifts in the size of the oligomer, modulation of subunit exchange and modification of their affinity to client proteins. Although mouse models of these mutants were instrumental in identifying changes in cellular proliferation and lens development, a direct comparative analysis of their effects on lens proteostasis has not been performed. Here, we have transgenically expressed cataract-linked mutants of αA- and αB-crystallin in the zebrafish lens to dissect the underlying molecular changes that contribute to the loss of lens optical properties. Zebrafish lines expressing these mutants displayed a range of morphological lens defects. Phenotype penetrance and severity were dependent on the mutation even in fish lines lacking endogenous α-crystallin. The mechanistic origins of these differences were investigated by the transgenic co-expression of a destabilized human γD-crystallin mutant. We found that the R49C but not the R116C mutant of αA-crystallin drove aggregation of γD-crystallin, although both mutants have similar affinity to client proteins in vitro. Our working model attributes these differences to the propensity of R49C, located in the buried N-terminal domain of αA-crystallin, to disulfide crosslinking as previously demonstrated in vitro. Our findings complement and extend previous work in mouse models and emphasize the need of investigating chaperone/client protein interactions in appropriate cellular context.


Introduction
In its most common form, age-related cataract is an opacity of the lens characterized by the formation of protein aggregates that scatter light [1]. Protein aggregation is driven by the progressive insolubilization of lens proteins as a consequence of age-dependent changes to their sequences and structures [2,3]. Terminally differentiated lens fiber cells, devoid of cellular machineries that repair and turn proteins over, rely on two small heat shock proteins (sHSPs), αA-and αB-crystallin, which function as molecular chaperones by sequestering thermodynamically destabilized a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 that the more deleterious effects of R49C are the result of its driving the aggregation of lens proteins as predicted from the thermodynamic coupling model [11,41].

Zebrafish transgenesis
To establish the transgenic zebrafish expressing rat (Rno) Cryaa gene (Rno.Cryaa) specifically in the lens, Tg(cryaa:Rno.Cryaa,myl7:Cerulean) was constructed by inserting mutated Rno. Cryaa cDNA (by site-mutagenesis; R49C and R116C) downstream of zebrafish cryaa promoter (1.2 kb) [43] in the pT2HBLR vector that was also contains myl7 promoter-driven Cerulean as the selection marker (i.e. in the heart). Tol2 mediated transgenesis were performed as previously descried [32] for the Rno.Cryaa (R49C and R116C) transgenes. The same protocol was followed for generating the mouse cryab R120G variant (Mmu.Cryab_R120G) transgene. Potential founder lines were selected by detecting cardiac Cerulean expression at 4dpf. At least two founder lines (F0) for each construct were screened and out-crossed to established stable F1 generations. Each F1 line was propagated and raised into F2 and F3 generations. The lens defects data collected were from F3 or F4 embryos, and similar penetrance of lens defects was observed in individual stable lines that express the same α-crystallin constructs.

Cell death assays
Embryos were fixed overnight at 4˚C in 4% paraformaldehyde in PBS, dehydrated with 100% MeOH and stored in -20˚C. The procedures of TUNEL staining were carried out following the manufacturer's suggested protocol (In Situ Cell Death Detection Kit, TMR red; Sigma-Aldrich #12156792910).

Microscopy and image processing
Lenses of live embryos in 0.3x Daneau water with PTU/tricane were analyzed by bright field microscopy (Zeiss Axiovert 200) at 4dpf and graded into three classes depending on the severity of lens defects as defined in our previous study [32]. Fluorescence images were taken with Zeiss AxioZoom.V16 microscope. Differential interference contrast (DIC) were performed by Zeiss LSM510 inverted confocal microscope with λem = 488 on 4dpf embryos which were embedded in 2.5% methylcellulose.

Statistics
The percentages of lens phenotype from multiple crosses (>2) for each experimental condition were averaged and reported as means ± standard error (SE). For those only have two data points, we reported the standard deviation (SD). Differences among comparing groups were analyzed by Student's t-test in Excel. Statistical significance was accepted when p < 0.05 (and marked as asterisk). Box plots were generated by BoxPlotR online software.

Transgenic expression of αA-crystallin variants induce embryonic lens defects in zebrafish
To investigate the in vivo interactions between cataract-linked αA-crystallin mutants and lens chaperones in the context of the native environment of the lens fiber cells, we generated transgenic zebrafish lines that express the αA-crystallin mutants R49C and R116C (Tg[cryaa:Rno. Cryaa_R49C] and Tg[cryaa:Rno.Cryaa_R116C]; hereafter referred to as"αA-R49C" and "αA-116C" for simplicity). Using zebrafish transgenesis protocols established previously [32], the two αA-crystallin variants were specifically expressed in the lens under the control of the zebrafish cryaa promoter with myl7 promoter-driven Cerulean fluorescent protein in the heart as a convenient selection marker for transgenic animals.
Expression of both αA-crystallin variants led to various degrees of embryonic lens abnormalities that were readily visible starting at 3dpf, without affecting the overall morphology of the embryos (Fig 1A and 1B). The nature of these defects was similar to those previously described for α-crystallin deficient [32,37] or γD-crystallin mutant transgenic lines [33]. Defective lenses exhibited phenotypic features that were either spherical, shiny crystal-like droplets spread sporadically across the lens (minor defects), or frequent droplets covering a large fraction of the lens as well as occasional large irregular protuberances located in the center of the lens (major defects) (Fig 1B), all of which could lead to opacity and changes in light scattering as previously reported [33]. For each αA-crystallin mutant, we screened a large number of embryos (see Fig 1C) and scored their lens phenotype based on the severity ( Fig  1C). While the expression of both mutants led to a considerable fraction of embryos displaying lens defects, the data suggest a more deleterious consequence for the R49C mutation both in the penetrance and severity of the phenotype.

Genetic dosage of αA-crystallin determines the level of lens defects induced by the mutants
The R49C and R116C mutations of human αA-crystallin cause congenital cataracts in an autosomal dominant fashion, while only 40~50% of the embryos show lens defects in the transgenic fish lines expressing these mutant forms of αA-crystallin. The gene dosage of αAcrystallin has been shown to impact the penetrance and/or severity of lens phenotypes, as demonstrated earlier in cryaa zebrafish mutants [32] as well as the Cryaa-R49C knock-in mouse model [26], where the lenses from Cryaa-R49C homozygotes (equivalent of losing two copies of wild-type Cryaa) showed more severe lens opacity.
To investigate if the relatively moderate penetrance of lens defects was due to the expression of the endogenous zebrafish cryaa gene (i.e. wild-type) in these transgenic lines, we generated heterozygote αA-crystallin mutants. For this purpose, the αA-R49C and αA-R116C transgenic lines were crossed into cryaa null mutants [32] to create genetic conditions more closely mimicking the human autosomal dominant CRYAA mutations. The absence of one endogenous cryaa allele (i.e. cryaa +/-), led to an increase in the percentage of lens defects from both αAcrystallin mutant transgenic lines, albeit not to 100% (Fig 1B). Absence of the other endogenous cryaa allele by backcrossing to generate αA-crystallin mutant transgenic lines under the cryaa null background (cryaa -/-) further increased the percentage of embryos with lens defects approximating 100% penetrance for αA-R49C and near 80% for αA-R116C.
Because cell death has been implicated in cataract phenotypes, including radiation-induced cataracts, diabetic retinopathy-associated cataracts [44][45][46] as well as age-related cataract [47], we tested for apoptotic cell death in the transgenic lines. TUNEL assay (Fig 2) did not show evidence of elevated apoptosis in the embryonic lens carrying αA-crystallin mutant transgenes when compared to WT. Therefore, cell death by apoptosis does not seem to play a major role in the embryonic lens defects in the αA-crystallin zebrafish transgenes, a result which differs from mouse αA-crystallin knock-in and transgenic models [24,26] where increase of cell death were reported in the postnatal lens.

αA-R49C but not R116C increases the phenotype penetrance of a destabilized γ-crystallin mutant
The model of Koteiche and Mchaourab predicts that R49C and R116C would act to bind native lens proteins driving unfolding and aggregation [11,48] and leading to formation of large particulates that scatter light. To test this model, we co-expressed each of the mutants with the human γD-crystallin variant I4F [33]. The I4F substitution reduces the free energy of unfolding reflecting a lower stability of the N-terminal domain [39]. In addition, the mutant exhibits a slow aggregation susceptibility in vitro [40]. Transgenic expression of γD-crystallin I4F mutant by itself resulted in moderate frequency (~30%) of major lens defects (γD-I4F; Fig  3), consistent with our previous study [33]. We focused on the major defects because of their direct correlation to aggregation of γD-crystallin I4F (see below).

Expression of αA-R49C increases the susceptibility of destabilized γDcrystallin to aggregation
In a previous study, we demonstrated that the penetrance of lens defects due to the expression of destabilized γD-crystallin mutant proteins strongly correlated with their susceptibility to form aggregates in the zebrafish lens [33]. Thus, the enhanced lens defects observed in the double transgenic embryos (Fig 3; αA-R49C; γD-I4F) could result from increased aggregation of γD-I4F when the αA-R49C transgene was co-expressed.
We directly tested this model utilizing a similar experimental approach with the Gal4/UAS expression system to facilitate mosaic analysis (Fig 4A). γD-crystallin mutant was tagged by Distinct mechanisms of congenital cataract-linked α-crystallin mutants red fluorescent protein and the aggregation was detected by the observation of fluorescent punctates as demonstrated previously [33]. By comparing the percentage of fluorescent punctate formation (i.e. mCherry-tagged γD-I4F) between transgenic carriers and non-transgenic embryos, we observed a significant difference between αA-R49C and αA-R116C transgenes in inducing γD-crystallin aggregation (Fig 4B). In the presence of the αA-R49C transgene, the percentage of embryos with fluorescent punctates, which reflects γD-crystallin aggregation, was higher by about 15~20% (Left panel); while no apparent difference was observed regardless of the expression of αA-R116C (Right panel). As expected, the phenotype penetrance of αA-R49C is increased in the absence of endogenous αA-crystallin.
A higher percentage of embryos displaying γD-crystallin aggregation (manifested by fluorescent punctate formation) was observed for the non-transgenic embryos from the αA-R116C crosses compared to the ones from the αA-R49C crosses (Fig 4B). A number of potential reasons could contribute to this difference. First, the stochastic nature of mosaic analysis, where the mCherry-tagged γD-crystallin cDNA constructs were randomly distributed in the lens fiber cells by microinjection, probably led to different expression/distribution of γD-crystallin, even though the same nominal amount of DNA was injected into each 1-cell zygote. Consistent with this conjecture, a wide-range of positive F0 carriers was reported in the context of zebrafish transgenesis [49]. Second and of critical importance is the underappreciated phenomenon of cross-generational effects of genetic modifiers [50], which was also suggested to account for variations in lens defects severity in the progeny of cryaa +/− adult lines in a previous study [32]. The adult transgenic αA-R116C and αA-R49C fish lines used here were neither phenotypically screened/selected before raised to adulthood, nor restricted for any particular sex (i.e. random usage of male and female fishes). Thus, the zygotes from αA-R116C and αA-R49C crosses may possess intrinsically different proteostasis capacity inherited from their parents, which would manifest as a difference in the basal level of protein aggregation.

Expression of murine αB-crystallin R120G mutant induces zebrafish lens defects but does not promote γD-crystallin aggregation
αB-crystallin is a molecular chaperone involved in the cell response to stress in various tissues (see reviews in [51,52]). Loss-of-function studies demonstrated that αB-crystallin is critical in maintaining lens proteostasis [37], and several missense mutations in human αB-crystallin gene are linked to hereditary cataracts, including P20S [53] and D140N [54]. The most-studied mutation, R120G, resides within the α-crystallin domain and is associated with desmin-related myopathy [20,55].
Because αB-crystallin R120 is homologous to αA-crystallin R116 and the two mutations (R120G and R116C) have similar properties in vitro, we generated a transgenic line expressing the αB-crystallin R120G mutant specifically in the lens (αB-R120G). As expected αB-R120G induced lens defects with a moderate penetrance but the severity of lens phenotypes was mostly of the minor class (Fig 5A). Reducing endogenous αB-crystallin levels by crossing αB-R120G into zebrafish αB-crystallin null mutants (cryaba -/-; cryabb -/-) [37] led to significant enhancement of lens defects in both penetrance and severity (Fig 5A) compared to αB-crystallin null mutants alone [37]. This is similar to the increase in penetrance and severity of lens defects in the absence of endogenous αA-crystallin.
If the nature of the substitution-i.e. glycine versus cysteine is not critical for the interaction with γD-crystallin, then the results of αA-R116C in Fig 4 predicts that αB-R120G would have a minimal effect on the propensity of γD-crystallin aggregation. Indeed, transgenic expression of αB-R120G did not significantly increase the percentage of embryos displaying γD-I4F aggregation when compared to non-transgenic siblings (Fig 5B). Together with the results from the αA-crystallin transgenic lines, this finding suggests that the increase in the penetrance of γDcrystallin aggregation phenotype induced by R49C co-expression is directly related to the position of R49 in the primary sequence. Distinct mechanisms of congenital cataract-linked α-crystallin mutants

Discussion
The work reported here takes advantage of the power of zebrafish as a model organism to carry out the first direct comparison between cataract-linked mutants of αA-crystallin in a similar genetic background. R49C and R116C are two human autosomal-dominant mutations that have been associated with a wide range of effects on the structure and function of αA-crystallin. Similar to mouse models, expression of these mutants in the zebrafish lens leads to lens defects that have been previously demonstrated to cause increased light scattering [32,33]. While the absence of one copy of the WT protein (i.e. endogenous allele) accentuates the penetrance and severity of the phenotype, the R49C mutation was consistently more deleterious than the R116C mutation. We linked the differential effects of the two mutants to the R49C being more effective in inducing protein aggregation. These mutants have similar chaperone properties as deduced from in vitro binding studies to the model client protein T4L [11]. Thus, our findings highlight the importance of in vivo investigation in validating mechanistic models.
How does co-expression of αA-R49C drive γD-I4F aggregation? Our thermodynamic coupling model [11] predicts that an increase in the apparent affinity of the chaperone to the unfolded state of the client protein would drive unfolding of the latter by simple mass action laws, i.e. the folding equilibrium is shifted towards the unfolded state as a result of it being depleted by binding to the chaperone. In the double transgenic line, the interaction with the γD-crystallin mutant is expected to be exclusive to the R49C mutant because of its higher affinity to γD-crystallin [39]. Therefore, if the level of expression of the two proteins is similar, a reasonable assumption given that both are transgenically driven by the same cryaa promoter with a single genomic insertion [33], we predict that the aggregation of γD-I4F involves the formation of 1 to 1 complexes with the R49C. Experiments are underway to test this conclusion.
In addition to enhanced binding affinity, R49C, but not R116C, has been shown to undergo disulfide cross-linking at the mutation site [41] suggesting that the location of the cysteine is critical. In studies of binding to the client protein βB1-crystallin, formation of an αA-R49C dimer involving a mixed α/β-crystallin intermediate was observed [41]. This dimer was strictly dependent on the labeling of βB1 at a single cysteine by a disulfide-linked fluorescent probe, suggesting that the formation of chaperone/client protein complex is an intermediate step [41]. Because human γ-crystallins contain multiple cysteines, we predict that the formation of the complex with R49C is stabilized by disulfide cross-links. Previous studies have shown that the R49C mutation in human γD-crystallin increases the propensity of intermolecular disulfide bond formation [56,57]. Importantly, in the homozygote knock-in R49C mouse line, the mutant αA-crystallin was reported to be disulfide cross-linked although the construct contained a second cysteine, which could lead to different types of covalent dimers. Consistent with our conclusion that disulfide cross-linking is the primary factor in the deleterious effect of the R49C mutation, the αB-crystallin R120G mutant did not promote the aggregation of γD-I4F.
Finally, a critical parameter typically missing in vitro is molecular crowding, which in lens fiber cells is predicted to substantially alter the thermodynamics of molecular interactions. Not only does the excluded volume effect, due to crowding, stabilize the folded states of client proteins but it is also predicted to affect the subunit exchange of the α-crystallins [58,59]. Equilibrium dissociation of α-crystallins is a mechanism of chaperone activation [23]. In this context, it is notable that the two mutations (R49C, and R116C) have profoundly different effects on the organization of the oligomeric structure [11]. Our finding of distinctive phenotypic effects of the two mutants in a model organism emphasizes the necessity of challenging mechanistic conclusions obtained in vitro and highlights the role of cellular context in shaping chaperone interactions with native client proteins.