Pathogenic mutations in retinitis pigmentosa 2 predominantly result in loss of RP2 protein stability in humans and zebrafish

Mutations in retinitis pigmentosa 2 (RP2) account for 10–20% of X-linked retinitis pigmentosa (RP) cases. The encoded RP2 protein is implicated in ciliary trafficking of myristoylated and prenylated proteins in photoreceptor cells. To date >70 mutations in RP2 have been identified. How these mutations disrupt the function of RP2 is not fully understood. Here we report a novel in-frame 12-bp deletion (c.357_368del, p.Pro120_Gly123del) in zebrafish rp2. The mutant zebrafish shows reduced rod phototransduction proteins and progressive retinal degeneration. Interestingly, the protein level of mutant Rp2 is almost undetectable, whereas its mRNA level is near normal, indicating a possible post-translational effect of the mutation. Consistent with this hypothesis, the equivalent 12-bp deletion in human RP2 markedly impairs RP2 protein stability and reduces its protein level. Furthermore, we found that a majority of the RP2 pathogenic mutations (including missense, single-residue deletion, and C-terminal truncation mutations) severely destabilize the RP2 protein. The destabilized RP2 mutant proteins are degraded via the proteasome pathway, resulting in dramatically decreased protein levels. The remaining non-destabilizing mutations T87I, R118H/R118G/R118L/R118C, E138G, and R211H/R211L are suggested to impair the interaction between RP2 and its protein partners (such as ARL3) or with as yet unknown partners. By utilizing a combination of in silico, in vitro, and in vivo approaches, our work comprehensively indicates that loss of RP2 protein structural stability is the predominating pathogenic consequence for most RP2 mutations. Our study also reveals a role of the C-terminal domain of RP2 in maintaining the overall protein stability.


Mutations in retinitis pigmentosa 2 (RP2) account for 10 -20% of X-linked retinitis pigmentosa (RP) cases. The encoded RP2 protein is implicated in ciliary trafficking of myristoylated and prenylated proteins in photoreceptor cells. To date
>70 mutations in RP2 have been identified. How these mutations disrupt the function of RP2 is not fully understood. Here we report a novel in-frame 12-bp deletion (c.357_368del, p.Pro120_Gly123del) in zebrafish rp2. The mutant zebrafish shows reduced rod phototransduction proteins and progressive retinal degeneration. Interestingly, the protein level of mutant Rp2 is almost undetectable, whereas its mRNA level is near normal, indicating a possible post-translational effect of the mutation. Consistent with this hypothesis, the equivalent 12-bp deletion in human RP2 markedly impairs RP2 protein stability and reduces its protein level. Furthermore, we found that a majority of the RP2 pathogenic mutations (including missense, singleresidue deletion, and C-terminal truncation mutations) severely destabilize the RP2 protein. The destabilized RP2 mutant proteins are degraded via the proteasome pathway, resulting in dramatically decreased protein levels. The remaining non-destabilizing mutations T87I, R118H/R118G/R118L/R118C, E138G, and R211H/R211L are suggested to impair the interaction between RP2 and its protein partners (such as ARL3) or with as yet unknown partners. By utilizing a combination of in silico, in vitro, and in vivo approaches, our work comprehensively indicates that loss of RP2 protein structural stability is the predominating pathogenic consequence for most RP2 mutations. Our study also reveals a role of the C-terminal domain of RP2 in maintaining the overall protein stability.
Retinitis pigmentosa (RP, 3 OMIM #268000) is a group of inherited retinal degenerative diseases characterized by progressive loss of photoreceptor function. RP affects ϳ1 in 4000 individuals worldwide (1). Typically, RP patients undergo night blindness, gradually worsening vision from the periphery to the center, and even complete blindness. Currently no therapy can effectively relieve the vision loss caused by RP.
Mutations in the retinitis pigmentosa 2 (RP2) gene alone represent the genetic causes of 10 -20% of X-linked retinitis pigmentosa (XLRP) families (2,3). Human RP2 encodes a ubiquitously expressed polypeptide of 350 amino acids and is well conserved across different vertebrate species including mouse (88% identity) and zebrafish (67% identity). Knock-out of the orthologs of RP2 in mice (4 -6) and zebrafish (7) indeed lead to progressive retinal degeneration, suggesting a crucial role of RP2 protein in maintaining the normal functions of the retina.
The RP2 protein contains two domains, an N-terminal tubulin folding cofactor C-like (TBCC) domain (residues ϳ42-192) and a C-terminal nucleoside diphosphate kinase-like (NDPK) domain (residues ϳ251-346) (8 -12). Two proteins, the ADPribosylation factor like GTPase 3 (ARL3) (12,13) and the G protein subunit ␤1 (GNB1, ␤ subunit of rod transducin) (14), are known to interact with the N-terminal TBCC domain of RP2. By acting as a GTPase-activating protein (GAP) specifically for ARL3, RP2 plays a role in the ciliary trafficking of myristoylated and prenylated proteins (5,(15)(16)(17). Binding of RP2 to GNB1 also facilitates the membrane association and traffic of GNB1 probably in an ARL3-independent manner (14). The pathogenic mutations R118H and E138G drastically reduce the affinity of RP2 to ARL3 (11,12), suggesting loss of this interaction might be correlated with the occurrence of RP. The C-terminal NDPK domain of RP2 shows sequence and structure similarity to the nucleotide diphosphate kinases. Nevertheless, due to the variations around the active site, it is unlikely that RP2 protein has kinase activity (11,18). The physiological functions of this domain remain unclear.
To date, Ͼ70 pathogenic mutations in RP2 have been identified (Human Gene Mutation Database) (19). More than 2 ⁄ 3 of these mutations are predicted to result in C-terminal truncated RP2 proteins or the complete loss of the RP2 gene. The most common truncation mutation, Arg120Ter, has been reported to cause loss of the entire RP2 protein, probably due to nonsense-mediated mRNA decay (20,21). The remaining mutations in RP2 are missense or small in-frame deletions. With the exception of R118H and E138G, little is known about the effects of these mutations on the structure and function of RP2 protein.
In the present study we obtained a novel rp2 mutant zebrafish line carrying a 12-bp in-frame deletion mutation (4-residue deletion at protein level) generated by the transcription activator-like effector nuclease (TALEN) technology. The retinal phenotype and molecular etiology of the novel rp2 mutant zebrafish was determined. Furthermore, we revealed that like the 12-bp deletion identified in zebrafish rp2, most of the missense and C-terminal truncation mutations identified in RP2 patients destabilized the RP2 protein and caused protein degradation through the proteasome pathway.

The del12 mutation causes loss of Rp2 protein in zebrafish
We previously reported the generation of rp2 knock-out zebrafish by TALEN technology that resulted in a frameshift mutation (c.359_363del, named del5) causing null allele of rp2 (7). In this study we identified a novel rp2 mutant zebrafish line carrying a 12-bp in-frame deletion (c.357_368del, named del12). The del12 mutation is predicted to cause deletion of four-conserved amino acid residues (Pro-120, Val-121, Lys-122, and Gly-123) located in the TBCC domain of Rp2 at the protein level. Homozygotes were screened out and confirmed by DNA sequencing (Fig. 1A). Like the rp2 knock-out zebrafish (7), the del12 homozygous mutants did not show any apparent developmental defects and grew and bred normally.
Intriguingly, using the anti-zebrafish Rp2 antibody described previously (7), we found that the Rp2 protein levels were reduced to ϳ10% in del12 homozygotes (Fig. 1B) and 60 -80% in del12 heterozygotes (supplemental Fig. S1) in Western blot analysis. Furthermore, we examined the localization of Rp2 in WT and del12 mutant retinas at the age of 3 months under dark-and light-adapted conditions. In WT retinas, Rp2 was mainly associated with the plasma membrane of photoreceptor cells from the synapses to the outer segments. But in the del12 mutant retinas, Rp2 signals were much weaker ( Fig. 1C and supplemental Fig. S2), which is in agreement with the Western blot results.
To test if the loss of Rp2 protein was due to the reduction of rp2 mRNA, quantitative reverse transcription PCR (RT-qPCR) was performed using WT and mutant (del5 and del12) zebrafish eyes. Unlike the del5 mutant mRNA levels, which were decreased to ϳ15% of WT group by nonsense-mediated decay, the del12 mutant mRNA levels were slightly reduced to Ͼ80% of WT group (Fig. 1D). Because the mild reduction of del12 mutant mRNA is unlikely to cause the near complete loss of Rp2 protein, we reasoned that there might be a post-translational mechanism degrading the mutant Rp2 protein.

Reduced rod phototransduction proteins and progressive retinal degeneration in del12 mutant zebrafish
As the del12 mutants showed very low levels of Rp2 protein, we speculated that the normal function of Rp2 might be abolished or at least impaired. To validate this we first chose several proteins that had been studied in the rp2 knock-out zebrafish (7) and detected their protein levels in WT and del12 mutant zebrafish by Western blot. The prenylated Pde6b and Grk1 and the myristoylated Gnat1, which participate in the phototransduction process of rod cells, are candidate trafficking substrates regulated by the RP2-ARL3 complex (22). All of them were down-regulated in the del12 mutants at the ages of 1 month ( Fig. 2A) and 2 months (Fig. 2B). The degrees of down-regulation of the three proteins in the del12 mutants were similar to those of the rp2 knock-out (del5) zebrafish. GNB1 and GNB3 are the ␤ subunit of the rod and cone transducin complex, respectively. RP2 interacts with GNB1 but not GNB3 (14). In the del12 mutant zebrafish, the protein levels of Gnb1 were reduced, whereas Gnb3 levels were unchanged. Additionally, we checked the expression of two other Rp2-interacting proteins, Arl3 and Nsf (23). Their protein levels were unaffected in both rp2 mutant zebrafish lines (Fig. 2, A and B and supplemental Fig. S1).
To assess the retinal phenotype of the del12 mutant zebrafish, we used the peanut lectin (PNA) and the anti-rhodopsin (4D2) antibody to label the outer segments of cones and rods, respectively. At the age of 6 months, the outer segments of rods in del12 mutant zebrafish showed a reduction in length as compared with WT zebrafish, whereas the outer segments of cones were mildly affected (Fig. 2C). Furthermore, we performed transmission electron microscopy assays on the retinas of 7-month-old WT and del12 mutant zebrafish. The del12 mutant zebrafish also showed shortened outer segments of photoreceptors when compared with WT (Fig. 2D). These results suggest that the del12 mutant zebrafish exhibits similar molecular defects and retinal phenotypes observed previously in the rp2 knock-out zebrafish (7).

The equivalent del12 mutation in human RP2 decreases its protein level and stability
The four residues affected by the del12 mutation are identical between human and zebrafish RP2 proteins. We constructed an equivalent del12 mutation by deleting the corresponding four residues (Pro-95, Val-96, Lys-97, and Gly-98) in human RP2 and then expressed the WT and del12 mutant RP2 protein with a C-terminal GFP or FLAG tag in ARPE-19 cells. Because the expression was controlled by the CMV promoter, without the 5Ј-and 3Ј-UTR of native RP2 mRNA, the interference of possible regulatory mechanisms acting on mRNA levels was excluded. We found that the protein levels of del12 mutant RP2 were reduced to ϳ30% (GFP-tagged) and 15% (FLAG-tagged) of WT (Fig. 3, A and B). We also detected the expression levels of del12 mutant RP2 in two other human cell lines (H1299 and HepG2) and obtained similar results (supplemental Fig. S3). These results suggest that the biological process causing the decrease of the del12 mutant protein is similar in human and zebrafish and present in different cell types.
As the Pro-95 residue was affected by the del12 mutation and a pathogenic mutation in the same codon (P95L) has been identified in RP patients (24), we then examined the expression level of P95L mutant RP2 in ARPE-19 cells. Similar to the del12 mutation, P95L reduced the protein levels of RP2 to ϳ30% (Fig.  3, C and D) when compared with WT or the R118H mutation, which does not affect the protein levels (25). Two other nearby pathogenic mutations, C86Y (24) and C108G (20), also showed similar effects on the protein levels of RP2 in the expression experiments (Fig. 3, C and D). These results suggest that the del12 mutation and the nearby C68Y, P95L, and C108G missense mutations might cause retinal degenerations through a shared mechanism that dramatically decreases the protein levels of RP2.
Crystal structure studies of RP2 revealed that the N-terminal domain of RP2 forms a parallel ␤-helix (residues 34 -192) with a tightly packed and mainly hydrophobic core (11). The Pro-95, Val-96, Lys-97, and Gly-98 affected by the del12 mutation make up part of the ␤-strand (residues 90 -103) that participate in the ␤-sheet formation within this ␤-helix domain. Analysis of the intraprotein interactions (26) of the four residues based on the human RP2 crystal structure reveals that these residues support critical ␤-helix stabilizing main-chain to main-chain hydrogen bonds and hydrophobic interactions that are required for core structural stability of the protein (Fig. 3E). For example, Val-96 is involved in main-chain to main-chain H-bonds with the residue Ala-113 in another ␤-strand. Therefore, deletion of these four residues might compromise A, the c.357_368del mutation (del12) in zebrafish rp2 was confirmed by Sanger sequencing at the genomic level. The chromatograms of WT and del12 zebrafish are shown. The TALE binding sequences are marked by red boxes. The deleted 12 bp is labeled by vertical lines. B, protein levels of Rp2 were detected in WT, del5 (knock-out) and del12 mutant zebrafish at the age of 3 months by Western blot using the anti-zebrafish Rp2 antibody. Tubulin served as a loading control. Quantitative results of Rp2 protein levels in del5 and del12 homozygotes from three independent experiments (n ϭ 6) are shown as the mean with S.D. in the lower panel. **, p Ͻ 0.01. C, immunostaining of Rp2 in dark-adapted retinas from WT and del12 mutant zebrafish at the age of 3 months. As compared with WT, the del12 mutants show much weaker fluorescent signals in the retina. OS, outer segment; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer, GCL, ganglion cell layer. Scale bar: 20 m. D, the corresponding rp2 mRNA levels in 3-month-old WT and rp2 mutant zebrafish were detected by RT-qPCR. Actb1 was used as an endogenous control. The results are shown as the mean with S.D. (n ϭ 3). **, p Ͻ 0.01. NS, not significant. the folding of the parallel ␤-helix and in turn the overall protein conformation, which would likely cause rapid degradation of the mutant protein.
To substantiate the prediction, we expressed and purified the WT, del12, P95L, and R118H forms of human RP2 protein with an N-terminal GST tag in Escherichia coli cells. Under the same conditions the WT and R118H forms of RP2 were mainly soluble and present in the supernatants of cell lysates, whereas the majority of del12 and P95L mutant RP2 formed inclusion bodies in the precipitates (supplemental Fig. S4). Differential scan-ning fluorometry (DSF) was used to determine the thermal stabilities of the purified RP2 proteins. The R118H group showed similar patterns of fluorescence to the WT group (Fig. 3F, blue and green lines). Meanwhile, the del12 and P95L groups exhibited high initial fluorescence signals at low temperatures (Fig.  3F, red and purple lines), suggesting the proteins were not well folded at all. These results indicate the del12 and P95L mutant RP2 proteins were unstable and more inclined to be unfolded and aggregated than WT RP2 protein.

Over half of the missense and small in-frame deletion mutations in RP2 are destabilizing
To date, Ͼ70 pathogenic mutations in RP2 have been reported in the literature, including 22 missense or small inframe deletion mutations (19). As the C86Y, P95L, and C108G mutations clearly decrease the protein levels of RP2, we next wanted to know if the remaining RP2 missense or small inframe deletion mutations showed similar effects. A total of 19 RP2 additional mutations (see  (24), S140F (29), L253R (30), and L253P (31) together with the del12 mutation were found to markedly decrease the protein levels of RP2 (Fig.  4, B and C), suggesting that they are likely destabilizing. On the contrary, 10 mutations, C3S (31), T87I (31), R118H, R118G (2), R118L (9), R118C (32), E138G (9), L188P (28), R211H (3), and R211L (33), showed no or mild effects on the expression levels of RP2 (Fig. 4, B and C). R282W, which is a non-pathogenic polymorphism (34), was used as a negative control in the study and showed normal RP2 protein levels. Repeated experiments in HepG2 cells showed similar results (supplemental Fig. S5). Using the del12 mutation, which showed protein levels of ϳ30% of WT in cultured cells (in vitro) but Ͻ10% in zebrafish (in vivo), as a reference, we predicted that patients carrying these mutations should have very low levels of RP2 protein (Fig.  4C). These results indicate that for more than half of the missense and small in-frame deletion mutations, loss of RP2 proteins might be the main cause of disease.
We further performed in silico stability analysis of these mutations using FoldX (35,36) to evaluate their effects on the three-dimensional crystal structures of RP2 protein (11,12) and provide a structural framework for interpreting pathogenicity. A total of 17 missense mutations in RP2 affecting 12 amino acid residue positions were analyzed (Fig. 5A) (note, the two N-terminal missense mutations G2V and C3S are present within a disordered region and not visible in the solved crystal structures). 10 of the 12 positions lie within the N-terminal TBCC domain, whereas one mutation position (Arg-211) lies in the linker region and one (Leu-253) lies in the C-terminal NDPK domain. From the FoldX in silico mutagenesis analysis of the unbound RP2 structure (Fig. 5B and supplemental Table S1), pathogenic mutations T87I, R118H/R118L/R118C, E138G, and R211H/R211L and polymorphism R282W do not destabilize the protein, as judged by ⌬⌬G values Ͻ1.6 kcal/mol (36). All of these positions are solvent-exposed (Fig. 4A, blue-or cyan-labeled), suggesting they play critical roles in protein-protein interaction, which are otherwise impaired in the mutated RP2 proteins. On the other hand, the buried C67Y, C86Y, P95L, C105W, C108G, S140F, L253R, and L253P (Fig. 4A, white labeled) are all predicted to severely destabilize the protein (⌬⌬G Ͼ 1.6 kcal/mol) and thus potentially cause degradation of the mutated RP2, which may be interpreted as decreased levels of the expressed mutant proteins (see Fig. 4, B and C). The FoldX analysis of the Arl3-bound RP2 crystal structure mirrors the results of the unbound structure (supplemental Table S1).
Two mutations, R118G and L188P, were predicted to be destabilizing in silico but showed mild effects on the protein levels of RP2 (Fig. 4, B and C). To investigate these aberrant results more carefully, we performed DSF analysis to directly measure the stabilities of these mutant RP2 proteins. As shown in supplemental Fig. S6, the R118G, E138G, and L188P mutations decreased the protein stabilities of RP2 by ϳ3°C, 5°C, and 11°C, respectively. The T87I, R118H, and R211H mutant proteins showed no significant differences when compared with WT. In general, the results of expression tests and DSF analysis fit well but not 100 % with the in silico prediction.

Truncation mutations causing imperfect C-terminal domain destabilize RP2 protein
Truncation mutations in RP2 often induce nonsense-mediated mRNA decay and cause loss of the entire protein, such as Arg-120* (37) and Tyr-151* (25). However, when the incorrect termination codon is located very close to the C-terminal end of RP2, the mutant mRNA might escape from nonsense-mediated decay, but still no RP2 protein could be detected (25,38). In that case the prematurely terminated RP2 protein with an incomplete C-terminal domain, but not the mutant RP2 mRNA, might be implicated. This prompted us that the C-terminal domain might play a role in stabilizing the RP2 protein. To verify this hypothesis, we constructed a series of C-terminaltruncated RP2 mutants (1-42, 1-192, 1-251, 1-300, 1-330 and 1-340) to mimic the truncation mutations (Fig. 6A) and examined their expression levels in ARPE-19 cells.
We found that only the 1-340 of C-terminal truncation showed comparable protein levels as WT (Fig. 6, B and C).  incomplete NDPK domain) forms of RP2 were unstable. These results suggest that an intact C-terminal domain (longer than 1-330) is essential for the overall stability of RP2. To date, the insertion-deletion mutation c.968delAinsTCC (K323Ifs*8), which results in a premature termination at codon 330, is the closest mutation to the C terminus of RP2 (27). According to our work and previous studies, patients carrying truncation mutations in RP2 are unlikely to have normal levels of RP2 protein.
Next, we also examined the N-terminal-truncated RP2 mutants , and 251-350) (Fig. 6A). The 192-350 truncation showed a mild decrease of the RP2 protein level in Loss of RP2 protein stability causes retinal degeneration APRIL 14, 2017 • VOLUME 292 • NUMBER 15 ARPE-19 cells (Fig. 6, D and E), suggesting that the entire C-terminal domain of RP2 is likely stable when expressed alone. The 251-350 truncation (missing the first 60 amino acids of 192-350 truncation) showed a markedly decreased protein level. This truncation is unstable probably due to the damage of the first ␤-strand (residues 234 -240, ␤13 in Fig. 4A) and ␣-helix  ⌬⌬G Յ 1.6 kcal/mol, no effect on structural stability (blue); ⌬⌬G Ͼ 1.6 kcal/mol, severely reduced structural stability (red). Negative values indicate enhanced stabilities.
(residues 246 -259, ␣3 in Fig. 4A) of the C-terminal structure of RP2. Taken together our results demonstrate that RP2 truncation mutations, which do not cause nonsense-mediated decay of mRNA, would still lead to destabilization and degradation of RP2 protein. In addition, the C-terminal NDPK domain of RP2, which probably has no enzymatic activity, might play an essential "structural" role in maintaining the global protein stability.

Destabilized RP2 proteins are degraded through the proteasome pathway
Protein destabilization can cause misfolded or even aggregated proteins, which is harmful and is cleared mainly through the proteasomal and lysosomal degradation pathways in eukaryotic cells (39). To determine the involved degradation pathway of the destabilized RP2 proteins, we treated the transfected cells with proteasomal and lysosomal inhibitors and then examined the protein levels of RP2 by Western blot. The proteasome inhibitors MG-132 and lactacystin both significantly restored the protein levels of del12 mutant RP2 (Fig. 7, A  and B). In contrast, the lysosomal inhibitors chloroquine and leupeptin showed no effect ( Fig. 7A and supplemental Fig. S7). To further confirm it, we repeated the experiments on those destabilizing pathogenic mutations and the N-terminal (42-350) and C-terminal (1-300) truncations. As expected, MG-132 treatment significantly increased the protein levels of P95L and 1-300 mutant forms of RP2 (Fig. 7, C and D) as well as the other destabilizing mutations (supplemental Fig. S8). These data suggest that the mutant RP2 proteins with significantly reduced stabilities are mainly degraded through the proteasome pathway.

Non-destabilizing mutations T87I and R211H do not affect the binding capacities of RP2 to ARL3 and GNB1
As for the remaining non-destabilizing mutations, R118H/ R118G/R118L/R118C and E138G, which affect the residues buried at the interaction interface of RP2 and ARL3 (12), were previously predicted and validated to abolish the interaction. Expectedly, our pulldown assays also revealed that the R118G and E138G mutation reduced the strength of interaction between RP2 and ARL3 to ϳ10 and 40%, respectively (Fig. 8A). However, the highly solvent-accessible T87I and R211H muta- tions and the partly solvent-accessible L188P mutation showed no impact in the same experiments. To further identify the pathogenic effects of these mutations, we tested another important RP2 interacting protein, GNB1. Still, no obvious alteration was detected (Fig. 8B). These results indicate that loss of interaction with other proteins such as NSF (N-eth-  ylmaleimide-sensitive factor, vesicle fusing ATPase) (23), PKD2 (polycystin 2, transient receptor potential cation channel) (40), or unknown new partners might play a causal role for developing retinitis pigmentosa in patients carrying T87I and R211H/R211L mutations.

Discussion
Patients with RP2 mutations usually show a rapid course of vision loss with decreased visual acuity at an early age (31). Several knockdown and knock-out animal models have been established to study the phenotypes and gene functions of RP2 (4 -7, 40 -42). In this study we report a novel rp2 mutant zebrafish line carrying an in-frame 12-bp deletion generated by TALEN technology. The protein levels of Rp2 in del12 homozygotes are extremely low, indicating a possible pathogenicity of the mutation. As expected, the rhodopsin kinase Grk1, the rod transducin subunits Gnat1 and Gnb1, and the ␤ subunit of rod cGMP phosphodiesterase (Pde6b), are decreased in the del12 mutant zebrafish. In terms of retinal phenotypes, the del12 mutants show a slowly progressing retinal degeneration, which is likewise similar to the rp2 knock-out zebrafish. As the mRNA level of del12 mutant rp2 is only slightly reduced, over-degradation of the mutant protein might be the most reasonable explanation for the loss of Rp2 protein. Indeed, our in vitro expression tests, protein stability assay, and intraprotein interaction analysis based on the crystal structure of human RP2 demonstrate that the del12 mutation is highly destabilizing and will cause rapid degradation of the mutant RP2 protein.
More intriguingly, our experimental results indicate that in addition to the del12 mutation, more than half of the missense mutations (including single residue deletions) substantially decrease the protein stabilities and expression levels of RP2. The two highest resolution crystal structures of RP2 in Arl3bound (resolution 1.9 Å) and unbound conformation (resolution 2.1 Å) were selected to perform in silico mutagenesis and stability calculations using the FoldX algorithm (35,36), as was previously undertaken for mutations in rhodopsin (that also causes retinitis pigmentosa) (43) and for mutations in PAX6, a transcription factor involved in eye development and disease, such as aniridia (44). The resulting quantitative estimation of energy changes on the stability of the protein correlates well in both crystal structures and with our experimental data. To our knowledge this is the first report based on strict experiments that reveal a large proportion of the missense and small in-frame deletion mutations in RP2 affect protein stabilities and protein levels.
How each of the mutations impact on the function of RP2 can be rationalized from our combined structure-based bioinformatics approach and experimental results. Mutations G2A, C3S, F5A, and S6del have been reported to impede the N-acylation of RP2 and interfere with its plasma membrane targeting previously (45,46). More specifically, the G2A, F5A, and S6del mutant protein are not attached to any membranes, whereas the C3S mutant protein is associated with intracellular, but not plasma membranes (45). Interestingly, in our study the G2V, F5del, and S6del, but not C3S, are destabilizing, indicating the membrane association might be indispensable for RP2 protein stabilization. In addition, the observation that N-terminal truncation of RP2 lacking the first 41 amino acids (42-350) also shows an extremely low expression level (Fig. 6, D and E) is in agreement with the hypothesis.
Mutations C67Y, C86Y, P95L, C105W, C108G, I137del, S140F, L253R, and L253P, which cause notably reduced RP2 protein levels, are classified as destabilizing mutations in this study. The C67Y, C86Y, P95L, C108G, I137del, and L253R mutations were previously predicted to destabilize RP2 protein based on their locations (11), which are consistent with our in silico stability calculations and experimental data. All of the involved amino acid residues are buried in the hydrophobic core of the structure, including the previously unstudied Cys-105 and Ser-140. It is noteworthy that the Leu-253 is the only mutated residue located in the C-terminal NDPK domain of RP2.
Interestingly, L188P is suggested to be a destabilizing mutation by our in silico stability calculations and a previous prediction (11). Although the L188P mutant RP2 shows 60 -70% expression levels relative to wild type RP2, its protein stability is obviously decreased as revealed by our DSF analysis (ϳ11°C reduction). This suggests that L188P is a destabilizing mutation substantially but not as severe as those destabilizing mutations mentioned above. In terms of structure, Leu-188 is located at the top of the final ␤-strand (residues 186 -188) of the N-terminal TBCC domain and is in van der Waals contact with Trp-171 and Trp-186; importantly, the leucine aliphatic side chain interacts with the hydrophobic component of Arg-228 located in the linker region of RP2. These interactions will in part be compromised in L188P mutant RP2 protein, and therefore, a local destabilization effect is possible even though the mutant protein is expressed relatively well. In addition, the loss of the interdomain interaction between Leu-188 and Arg-228 allows greater flexibility between the TBCC and linker-NDPK domains, which may affect potential protein-protein interactions that rely on more than one domain.
Mutations T87I, R118H/R118G/R118L/R118C, E138G, and R211H/R211L on the other hand are identified as non-destabilizing mutations in our study. Previously undertaken circular dichroism spectra of R118H and E138G mutant RP2 proteins confirmed that they have no impact on the secondary structure content and hence no effect on the overall folding of the mutant proteins (11). Our experiments suggest that R118G and E138G show very slight destabilizing effects (Ͻ5°C reductions in DSF analysis) and near normal expression levels. Probably, the mild destabilizing effects are insufficient to cause severe misfolding and degradation of the mutant RP2 proteins. According to the crystal structure, the side chains of these affected residues are highly surface-exposed and, therefore, might not contribute to the overall stability of RP2 protein.
For these non-destabilizing mutations, alteration of proteinprotein interaction is the likely cause of disease. It is already known that R118H and E138G weaken the binding capacity of RP2 to ARL3 and thus prevent the activation of ARL3 GTPase (11,12). Our GST pulldown assays also show that R118G and E138G reduce the interaction between RP2 and ARL3. The Arg-118 residue of RP2 has been suggested to act as an "arginine finger" within the common conserved active site found in GAPs (10), suggesting that the other two mutations (R118C and R118L) affecting the same residue are likely to abolish the GAP activity of RP2 as well. The T87I and R211H mutant forms of RP2 interact normally with ARL3 and GNB1 in our GST pulldown assays, indicating that impaired interaction(s) with other RP2 partners such as NSF (23) and polycystin 2 (PKD2) (40) might be the pathogenic effect of these mutations.
C-terminal-truncated mutations account for ϳ2/3 cases of RP2 patients. Our work indicates that except nonsense-mediated decay, destabilization and degradation of the C-terminaltruncated RP2 protein is another cause of pathogenicity for truncation mutations in RP2. According to the crystal structure of RP2 (11,12), the C-terminal domain (residues 229 -347) of RP2 forms a three-layered ␣ϩ␤ sandwich with an antiparallel ␤-sheet surrounded by 4 ␣-helices and a short 3 10 -helix on both sides. The 1-340 truncation retains nearly the entire ␣ϩ␤ sandwich structure and is stable, whereas the 1-330 truncation that lacked the last ␣-helix (residues 330 -345, ␣6 in Fig. 4A) is unstable, suggesting that a full conformation of the C-terminal domain is crucial for the overall stability of RP2 protein.
In conclusion we report a novel zebrafish line with an inframe 12-bp deletion in rp2. Loss of protein stability is the main consequence of the del12 mutation, which further leads to retinal degeneration. In addition, more than half of missense and single-residue deletion mutations and the C-terminal truncation mutations identified in RP2 patients are destabilizing. The destabilized RP2 proteins are degraded through the proteasome pathway. For the first time our study comprehensively characterizes the disease-causing mutations in RP2 and provides robust evidence to indicate that loss of RP2 protein stability but not alteration of its molecular function (except for T87I, R118H/R118G/R118L/R118C, E138G, and R211H/ R211L) is probably the major pathogenic cause for most RP2 mutations. Our study also reveals a role of the C-terminal domain of RP2 in maintaining the overall protein stability. Further studies are needed to elucidate how T87I and R211H/ R211L alter the function of RP2 and how loss of RP2 protein causes retinal degeneration.

Zebrafish maintenance and genotyping
Wild type and mutant zebrafish were maintained and bred as described previously (7). The del12 mutant zebrafish were screened out by the PCR-restriction fragment length polymorphism (PCR-RFLP) method. Briefly, a 574-bp DNA fragment was amplified (forward, 5Ј-tgcggtatcgatgctgaaatga-3Ј; reverse, 5Ј-ggcgcagcacaaaaagacgt-3Ј) using genomic DNA isolated from tail fin and digested with MspI (New England BioLabs) at 37°C for 4 h. The native MspI restriction site will be disrupted by the del12 mutation, which leads to an uncut band in DNA gel electrophoresis. The homozygotes were further confirmed by DNA sequencing. Zebrafish with similar body sizes at the same ages form different genotype groups were randomly selected and included in the study. All procedures involving zebrafish were approved by the Ethics Committee of Huazhong University of Science and Technology and were in accordance with the ARRIVE guidelines.

Site-directed mutagenesis and vector construction
Full-length human RP2 coding sequence was amplified from the cDNA of ARPE-19 cells. RP2 mutations were generated by overlap extension PCR-mediated site-directed mutagenesis. WT and mutant RP2 were cloned into the pEGFPN1 (Clontech) and pcDNA3.1 (Invitrogen) vectors for eukaryotic expression of C-terminal GFP and FLAG-tagged protein, respectively. For prokaryotic expression, the WT, del12, P95L, and R118H forms of RP2 were cloned into the pGEX-4T-1 vector (GE Healthcare). All plasmids were confirmed by DNA sequencing.

Cell culture, transfection, and inhibitor treatment
The ARPE-19 (ATCC catalog #CRL-2302), HepG2 (ATCC catalog #HB-8065), and H1299 (ATCC catalog #CRL-5803) cell lines were purchased from ATCC (American Type Culture Collection). DAPI staining of each cell line was performed to examine the status of mycoplasma contamination. Cells were cultured under recommended conditions. In brief cells were placed in DMEM medium (HepG2 and H1299) or DMEM:F-12 medium (ARPE-19) containing 10% fetal bovine serum and incubated at 37°C in 5% CO 2 in a humidified incubator. When grown to 80 -90% confluence, cells were re-plated at a split ratio of 1:3ϳ1:5. Cell transfection was performed using Lipofectamine 3000 (Invitrogen) according to the manual.

Protein purification, differential scanning fluorimetry, and GST pulldown
Expression of GST-tagged RP2 proteins were performed in BL21 strain following the handbook of GST Gene Fusion System (GE Healthcare). E. coli cells were induced with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside at 25°C for 4 h. After that, cells were harvested and lysed in PBS buffer containing 0.1% Triton X-100 and 1 mM PMSF by sonication. The lysates were centrifuged at 4°C, 12,000 rpm for 20 min to separate the soluble (supernatant) and insoluble (precipitate) components. Glutathione beads (Thermo Fisher Scientific) were added to the supernatants and rotated at 4°C for 4 h. GST fusion protein was eluted with elution buffer (50 mM Tris, pH 8.0, 20 mM reduced glutathione) by rotating at 4°C for 1 h. The supernatant, precipitate, and purified protein were examined by SDS-PAGE and Coomassie Blue staining. DSF (47) was performed to detect the thermal stabilities of these purified proteins using the Protein Thermal Shift™ Dye kit (Applied Biosystems) on the StepOnePlus™ real-time PCR System (Applied Biosystems). The data were analyzed by the Protein Thermal Shift Software 1.3 (Thermo Fisher Scientific). For GST-pulldown, cell lysates expressing ARL3-Q71L (GTP-bound form) or GNB1 were incubated with immobilized RP2 proteins on glutathione beads for 4 h at 4°C. After washing the beads with lysis buffer for 3-4 times, the bound proteins were eluted and analyzed by Western blot.

Western blot and immunofluorescence staining
Zebrafish were sacrificed by immersing in 0.02% MS222 (Sigma) solution for 15 min or until no opercular movement. For protein preparation, eyes were enucleated and sonicated in radioimmune precipitation assay lysis buffer containing protease inhibitor mixture (Sigma). Cultured cells were washed twice with cold PBS and lysed with the same radioimmune precipitation assay buffer. Protein concentrations were determined using the BCA Protein Assay kit (Beyotime, Shanghai, China). Western blot was performed as described previously (48). The membranes were developed using the ChemiDoc XRSϩ imaging system (Bio-Rad). Gray intensity analysis of bands was performed by the Quantity One 4.62 software. The immunostaining of zebrafish retina sections was performed as described previously (7). Fluorescence images were acquired on the FluoView™ FV1000 confocal laser-scanning microscope (Olympus). The following primary antibodies were used in this study: the anti-zebrafish Rp2 antibody (1:1000 for WB, 1:50 for immunofluorescence (IF)) (7), the anti-rhodopsin 4D2 antibody

RT-qPCR
RNAs were isolated using RNAiso Plus reagent (Takara) from at least three zebrafish eyes for each sample. Reverse transcription was performed using MMLV reverse transcriptase (Invitrogen) and oligo(dT) primer (Takara). The mRNA levels of WT and mutant rp2 were detected on the StepOnePlus™ real-time PCR System (Applied Biosystems) using AceQ™ qPCR SYBR Green Master Mix (Vazyme Biotech, Nanjing, China). The data were analyzed using StepOne software (version 2.3), and significance was determined by two-tailed Student's t test. Primer sequences were reported in our previous studies (7).

Structure-based modeling analysis of RP2 mutations
The X-ray crystal structures of both native (uncomplexed) human RP2 (PDB ID 2BX6, resolution 2.1 Å) (11) and RP2 in complex with Arl3 (PDB ID 3BH7, resolution 1.9 Å) (12) have been solved and were analyzed independently. The empirical forcefield FoldX version 4 (35,36) was used to estimate the free energy difference i.e. stability change upon mutagenesis from wild type (⌬⌬G) and also to perform interaction energy calculations upon mutation in the complexed RP2 structure. In brief, amino acid residues with missing side-chain atoms, Van der Waals clashes, with bad torsion angles, or with a high energy in the crystal structure were repaired by replacing the side chain conformations (rotamers) observed in the X-ray structures by lower energy rotamers. This was achieved using the RepairPDB option, which also optimized the hydrogen bond networks (35,36). After energy minimization, the BuildModel option was used to calculate the stability change using the following parameters (number of runs, 3; pH 7; temperature 298 K; ionic strength 0.05 M; VdW design, 2). Where mutations were present at or near the interface with Arl3/GDP, the potential change in interaction energy was determined using the "PSSM" option (number of runs: 3) that incorporates the BuildModel and Anal-yseComplex commands. All resulting energies are expressed in kcal/mol. The error margin of FoldX is ϳ0.5 kcal/mol (35) so changes in that range are considered insignificant. The prediction decision on whether the mutation destabilizes the structure is based upon Refs. 36, 43 and 49), where no effect on structural stability ⌬⌬G is Յ1.6 kcal/mol, and severely reduced structural stability ⌬⌬G is Ͼ1.6 kcal/mol. Intraprotein interactions of the four residues affected by the del12 mutation were analyzed using the Protein Interactions Calculator (PIC) server (26). ESPript version 3 (50) was used to annotate the alignment.

Statistical analysis
Two-sided Student's t test was performed with a 95% confidence level to determine the statistical significance between two groups using the GraphPad Prism 6 software. All data are presented as mean with S.D. * represents p Ͻ 0.05; ** represents p Ͻ 0.01.