Aminoglycoside-induced premature termination codon readthrough of COL4A5 nonsense mutations that cause Alport syndrome

Alport syndrome (AS) is characterized by glomerular basement membrane (GBM) abnormalities leading to progressive glomerulosclerosis. Mutations in the COL4A3, COL4A4 or COL4A5 genes encoding type IV collagen α3α4α5 cause AS. Truncated α3, α4, and α5 chains lacking an intact COOH-terminal noncollagenous domain due to a premature termination codon (PTC) cannot assemble into heterotrimers or incorporate into the GBM. Therefore, achieving full-length protein expression is a potential therapy for AS caused by truncating nonsense mutations. Small molecule-based PTC readthrough (PTC-RT) therapy has been well studied in other genetic diseases, but whether PTC-RT is applicable to AS is unexplored. To investigate the feasibility of PTC-RT therapy in AS, we made a cDNA to express COL4A5 fused to a C-terminal NanoLuc luciferase (NLuc) to monitor full-length translation. Full-length COL4A5-NLuc produces luminescence, but mutants truncated due to a PTC do not. To screen for COL4A5 nonsense mutants susceptible to PTC-RT, we introduced 49 individual nonsense mutations found in AS patients into the COL4A5-NLuc cDNA. Luciferase assays revealed that 11 mutations (C29X, S36X, E130X, C1521X, R1563X, C1567X, W1594X, S1632X, R1683X, C1684X and K1689X) were susceptible to PTC-RT induced by G418, which is known to have high readthrough activity. Moreover, we found that some next-generation “designer” PTC-RT drugs induced RT, and RT enhancer compounds increased the efficacy of PTC-RT in a G418-susceptible PTC mutant. These results suggest that PTC-RT therapy is a feasible approach for some patients with AS. Our luciferase-based COL4A5 translation reporter system will contribute to further development of PTC-RT therapies in a personalized medicine approach to treating AS.

All three chains are necessary to form a functional type IV collagen 345 network. The type chains assemble inside cells into 345 heterotrimers (protomers), which are secreted into the extracellular space (6), where they polymerize to build a basement membrane with other components such as laminins, nidogen and heparan sulfate proteoglycans (7).
The lack or reduction of type IV collagen 345 in AS eventually leads to GBM abnormalities including thinning, thickening, and splitting. Current standard-of-care therapy uses renin-angiotensin system (RAS) inhibitors such as angiotensin-converting enzyme (ACE) inhibitors or Angiotensin II receptor blockers (ARBs). Though they delay progression to kidney failure, they do not cure AS (11)(12)(13). In contrast, development of methods to fix the pathogenic GBM abnormalities-compositional, structural, and functional-could cure AS or overcome the limitations of current treatments.
One of the potential barriers to treatment of AS using a GBM repair approach is the requirement that the abnormal GBM composed of collagen IV 112 be able to incorporate 345. Genetic rescue experiments in a Col4a3-null AS mouse model has shown that postnatal induction of COL4A3 production in podocytes, the glomerular cells that normally synthesize the collagen IV 345 network, enables 345 trimer synthesis, secretion, and incorporation into the Alport GBM (14). This study shows that restoration of the normal type IV collagen 345 network in the Alport GBM, by either cell-, protein-, chemical-or gene-based approach, is a feasible approach towards a cure.
In the present study, we focused on chemical-induced restoration COL4A5 expression in COL4A5 nonsense mutation types of AS. Nonsense mutations resulting in premature termination codons (PTCs) account for about 6% of AS cases (15). Type IV collagen chains have a Cterminal NC1 domain that is essential for assembly of heterotrimers inside cells and for network formation in the GBM. Truncated 3, 4, 5 chains without an intact NC1 domain due to PTCs cannot form trimers or polymerize in the GBM (16). Therefore, achieving full-length protein expression is a potential therapy for AS due to nonsense mutations.
Small molecule-based PTC readthrough (PTC-RT) therapy has been well studied in other genetic diseases such as cystic fibrosis (17,18), Duchenne muscular dystrophy (19,20), and inherited skin disorders (21,22). G418, an aminoglycoside class antibiotic, is the most studied PTC-RT drug. Aminoglycosides bind to prokaryotic ribosomes and inhibit protein synthesis in gram-negative bacteria. In addition to their affinity for prokaryotic ribosomes, aminoglycosides are known to bind to eukaryotic ribosomes and induce PTC-RT of nonsense mutants by enabling a near-cognate aminoacyl-tRNA to recognize PTCs (23). Since the discovery of aminoglycosideinduced PTC-RT, it has been shown that PTC-RT can be induced by various aminoglycoside compounds (17,24,25). Although aminoglycoside-mediated PTC-RT has the advantages of being well-studied and highly efficient, high doses can cause nephrotoxicity and ototoxicity. To overcome these limitations, structurally designed aminoglycosides with reduced nephrotoxicity and ototoxicity that maintain readthrough activity have been developed (24,26). In addition, nonaminoglycoside PTC-RT compounds have been identified by high throughput library screening 5 (19,27) and some have been chemically modified to improve their activity (28-30). In addition, several enhancer compounds have been found that enhance the activity of PTC-RT compounds, allowing the use of reduced doses of aminoglycosides and thus reduced toxicity (31,32).
With these technological advances, PTC-RT-based therapy has become a more realistic option. However, it is unexplored whether nonsense readthrough therapy is applicable to AS.
Here, we tested the feasibility of PTC-RT therapy for nonsense mutant AS. We generated a NanoLuc-based translation reporter system to evaluate which COL4A5 nonsense mutations are susceptible to PTC-RT therapy. 49 nonsense mutations reported in patients with AS were tested, and 11 of them were highly sensitive to aminoglycoside-mediated PTC-RT. Moreover, we found that designer aminoglycoside and non-aminoglycoside PTC-RT drugs with reduced nephrotoxicity and ototoxicity induced full-length COL4A5 protein synthesis in PTC-RTsensitive mutants. Also, PTC-RT enhancer compounds potentiated aminoglycoside-mediated PTC-RT. These results contribute important basic knowledge for the feasibility of PTC-RT therapy in AS.

Development of a NanoLuc-based COL4A5 translation reporter system
The efficacy of aminoglycoside-induced PTC-RT varies greatly from mutation to mutation (22).
At least 76 COL4A5 nonsense mutations have been reported in patients with X-linked AS, accounting for 6% of all known X-linked AS mutations (15). Therefore, determining which nonsense mutations are susceptible to PTC-RT is a crucial first step towards clinical application.
To evaluate the sensitivity of COL4A5 nonsense mutations to PTC-RT in a high-throughput system, we generated a COL4A5-NanoLuc reporter plasmid by in-frame fusion of NanoLuc to the COOH-terminus of COL4A5 (Fig. 1A). Introduction of a PTC into the COL4A5-NanoLuc cDNA leads to synthesis of truncated protein without the COOH-terminal NanoLuc, so luminescence is not produced. In this reporter system, it is assumed that a small molecule compound such as an aminoglycoside will promote PTC-RT, leading to synthesis of a full-length protein and production of luminescence.
First, we introduced a pathogenic nonsense mutation, R1563X, into the COL4A5-NanoLuc cDNA to see if we could detect PTC-RT in the presence of G418, which is known to have high readthrough activity and is considered the gold standard PTC-RT drug in vitro. G418 increased luminescence in HEK293 cells expressing COL4A5-R1563X-NanoLuc to a level that was 20-30% of that of WT (Fig. 1B). Moreover, to show that the PTC-RT was not artifactually related to the presence of the NanoLuc mRNA in the transcript, we investigated whether G418 could induce PTC-RT of the isolated COL4A5-R1563X mRNA, using COL4A5-R1563X-NanoLuc ( Fig. 1C). Immunoblot analysis showed G418 induced PTC-RT in HEK293 cells expressing either COL4A5-R1563X-NanoLuc or COL4A5-R1563X NanoLuc (Fig. 1D). These results show that the COL4A5-NanoLuc reporter cDNA was sensitive and quantitative enough for monitoring translation of full-length COL4A5 protein in a multi-well plate format.

Screening for PTC-RT-susceptible COL4A5 nonsense mutations
To screen COL4A5 nonsense mutations reported in patients with X-linked AS for susceptibility to PTC-RT, we introduced 49 individual mutations into the COL4A5-NanoLuc cDNA reporter by site-directed mutagenesis. Aminoglycoside-induced PTC-RT is known to be influenced by the type of PTC (UGA>UAG>UAA) (33). Therefore, we comprehensively selected COL4A5 UGA, UAG, and UAA nonsense mutations for evaluation. Nonsense mutant COL4A5-NanoLuc plasmids were individually transfected into HEK293 cells, and the cells were treated for 24 h with different concentrations of G418. G418 induced significant PTC-RT of 40 of the 49 nonsense mutations ( Fig. 2A-C). Many of them were statistically significant, but some did not have high PTC-RT rates. Of the types of mutations that responded to G418, UGA PTCs showed the highest readthrough rates, which is in line with previous studies. 11 of 49 COL4A5 nonsense mutants (C29X, S36X, E130X, C1521X, R1563X, C1567X, W1594X, S1632X, R1683X, C1684X, K1689X) showed more than 5-fold induction of PTC-RT. The amount of luminescence produced from these G418-susceptible mutants was about 10-30% of the WT level (Fig. 2D).

Gentamicin, an aminoglycoside approved for clinical use, induces PTC-RT in G418susceptible mutants
Although G418 is one of the most potent readthrough inducers, it is highly toxic and cannot be used clinically. Gentamicin is a clinically approved aminoglycoside class antibiotic. Therefore, we investigated whether gentamicin induced PTC-RT of the G418-susceptible mutants.

The efficacy of aminoglycoside-mediated PTC-RT is dose-and treatment time-dependent
To investigate whether aminoglycoside-induced PTC-RT is treatment time-dependent, we performed long-term treatment experiments using low doses of aminoglycosides on cells expressing COL4A5-R1563X and -R1683X mutants that were highly responsive to G418 and gentamicin. For long-term treatment, we generated stable COL4A5-R1563X/R1683X-NanoLuc cDNA-expressing cells by lentivirus transduction. The degree of PTC-RT was increased with low-dose G418 treatment in a time-dependent manner (Fig. 4A, B). The low concentrations of G418 (10 and 30 µg/mL) slightly increased readthrough, and the moderate concentration (100 µg/mL) dramatically increased full-length protein synthesis, depending on treatment time. The longer treatment with low concentrations of gentamicin (30 and 100 µg/mL) did not increase the efficacy of PTC-RT, but readthrough was increased at the moderate (300 µg/mL) and high (1000 µg/mL) concentrations (Fig. 4C, D).

Designer aminoglycoside and non-aminoglycoside readthrough drugs induce PTC-RT in the highly susceptible mutant COL4A5-R1563X
The potential for successful PTC-RT therapy with aminoglycosides is dependent on their degrees of activity, nephrotoxicity, and ototoxicity (34). Aminoglycoside toxicity is attributed to a structural site different from that responsible for PTC-RT activity (24,35). Therefore, chemical modification has been used to reduce the toxicity of aminoglycosides, with the aim of reducing toxicity while maintaining readthrough activity. Several aminoglycoside derivatives have been developed and are called designer aminoglycosides (36). The use of PTC-RT compounds with non-aminoglycoside structures is also a strategy to reduce toxicity. We tested a set of next-

Designer aminoglycoside and non-aminoglycoside PTC-RT drugs are ineffective for the non-G418-susceptible mutant COL4A5-G5X
In addition to the G418-susceptible COL4A5-R1563X mutant, we also tested whether any next generation PTC-RT drugs induced PTC-RT for the G418 non-susceptible COL4A5-G5X mutant These results indicate that PTC-RT was not strongly induced in the G418-nonsusceptible COL4A5-G5X mutant by either ELX-02, which has the same mechanism as G418, or DAP, which exerts its effects via a different mechanism.

Enhancer drugs improve the efficiency of aminoglycoside-induced PTC-RT
Several PTC-RT enhancer compounds have been developed to reduce aminoglycoside-induced toxicity by lowering the dose required for sufficient PTC-RT. The PTC-RT enhancer CDX5 was identified in a yeast cell-based assay in the presence of the aminoglycoside paromomycin. The effect of CDX5 was also significant in mammalian cells (31). A more recent study showed that the anti-malarial drug mefloquine potentiated G418-mediated PTC-RT in mammalian cells (32).

Functionality of the possible PTC-RT products of G418-susceptible mutants
So far, we have evaluated COL4A5 nonsense mutants in terms of their susceptibility to PTC-RT, but whether the resulting protein product is functional or not is also important for therapeutic applications. PTC-RT drugs suppress PTC by facilitating the insertion of near-cognate aminoacyl-tRNAs into the ribosomal-A site during protein translation. Therefore, the readthrough product is often a full-length protein with an incorrect amino acid at the PTC. If such a substitution impairs the function of the protein, it may be difficult to rescue the mutant phenotype even if a full-length protein is produced.
To begin to investigate whether the PTC-RT products from G418-susceptible mutants are functional, we utilized a split-NanoLuc-based collagen IV 345 heterotrimer formation assay.
This platform assays 345 heterotrimer formation by measuring the luminescence produced by the proximity of NanoLuc fragments fused to the ends of COL4A3 and COL4A5 that are brought together during the formation of COL4A3/4/5 heterotrimers (Fig. 8A) (40). Most pathogenic COL4A5 missense mutations affect 345 heterotrimer formation and prevent production of functional collagen IV 345, which causes Alport syndrome. Therefore, assessing whether PTC-RT products can assemble into 345 heterotrimers is important for evaluating the feasibility of PTC-RT therapy.
It is known that during G418-induced PTC-RT, Arg, Trp and Cys are inserted for UGA, Tyr and Gln are inserted for UAA, and Gln is inserted for UAA (41). The potential readthrough products from G418-susceptible COL4A5 mutants are shown in Table 1. In several cases, it is possible that PTC-RT will result in production of some wild-type protein. To investigate the function of the mutant readthrough products, all possible missense mutant substitutions were generated by site-directed mutagenesis and assayed using the C-terminal tagged split NanoLucbased 345 heterotrimer assay. The luminescence reflecting heterotrimer formation was significantly decreased intracellularly and extracellularly in some readthrough products from C1521X, R1563X, C1567X, W1594X, R1683X and C1684X. On the other hand, all readthrough products from C29X, S36X, E130X, S1632X and K1689X retained the ability to form 345 heterotrimers (Fig. 8B, C). It should be noted that for Arg (R) codons mutated to UGA, more 12 than half of the product is the wild-type R (42), so a higher percentage of functional full-length proteins are produced than for Cys (C) to UGA and Trp (W) to UGA mutants. For C29X, S36X, and E130X, the mutations are located close to the N-terminus; thus, in addition to the C-terminal tag assays (Fig. 8A, left), we also evaluated the function of appropriate PTC-RT products using the N-terminal tag system (Fig. 8A, right). The extent of heterotrimer formation for C29Xderived products was reduced by half, whereas S36X-and E130X-derived products retained their functions (Fig. 8D). These results indicate that inducing PTC-RT is a valid approach for a subset of COL4A5 nonsense mutations.

DISCUSSION
The goal of this study was to determine the applicability of PTC-RT therapy in Alport syndrome caused by nonsense mutation. Since the susceptibility to PTC-RT varies greatly among mutations (43,44), the first step would be to determine which mutations are susceptible. Because the type IV collagen genes are relatively large with many reported nonsense mutations (15,45), we thought it would be essential to evaluate a simple reporter system with high throughput to cover most of them. In addition, previous studies have shown that PTC-RT activity is affected by the sequences surrounding the nonsense mutations (33,46). Therefore, we constructed a reporter using the full-length COL4A5 cDNA instead of a short cDNA containing the PTC. Since the fulllength COL4A5 cDNA itself is about 5 kb, we used NanoLuc as a reporter to construct the fusion gene because of its small size (513 bp) and high sensitivity (47,48). The COL4A5-NanoLuc reporter cDNA developed in this study allowed us to identify which of 49 tested COL4A5 nonsense mutations susceptible to PTC-RT. Also, this study showed the efficacy of next generation PTC-RT drugs and potentiator compounds that enhance PTC-RT activity.
Through comprehensive mutation screening, we found that UGA-type COL4A5 nonsense mutations were more susceptible to aminoglycoside-mediated PTC-RT than other stop codons.
This result is consistent with previous reports (43). However, as previously reported, not all UGA PTCs showed high susceptibility (49), and it was reconfirmed that susceptibility varied depending on the mutation. Susceptibility is known to be affected by the surrounding nucleotide sequence. To attempt to determine whether susceptibility is based on the flanking nucleotide sequence, we aligned the cDNA sequences around PTCs that were G418-susceptible and compared them to those flanking the non-susceptible PTCs (Fig. S2). However, no overt differences between susceptible and non-susceptible sequences were observed. This suggests that susceptibility is defined by factors other than the peripheral sequence, such as the location of the mutation in the gene. Although more detailed comparative studies are needed, these results highlight the importance of screening for PTC-RT using full-length cDNAs rather than just short cDNA reporters carrying sequence peripheral to the PTC.
The limitation of this reporter system is that the presence of nonsense-mediated mRNA decay (NMD) cannot be taken into account because the cDNA is already spliced and does not have the exon-exon junctions that are required for mRNA surveillance (50). The mRNAs produced from genes with nonsense mutations are partially degraded by NMD (51). Therefore, promotion of basal-readthrough by suppression of NMD is one of the therapeutic strategies for nonsense mutations, but we have not been able to investigate this. However, since the induction of PTC-RT by NMD inhibition alone is not expected to be very high (52), and many aminoglycosides have an activity that inhibits NMD itself, this limitation is not likely a serious problem, but it should be taken into account to accurately determine PTC-RT activity in vivo. To overcome this limitation, CRISPR/Cas9-mediated genome editing could be used to create cells with point mutations in endogenous genes (53), which would allow the evaluation of PTC-RT activity under the same conditions as in vivo. However, the throughput of this method would be very low, and it is not suitable for evaluating a large number of mutants as in the present study.
In addition to defining the readthrough susceptibility of each mutation, it is also important to determine whether the possible PTC-RT products are functional (54). To investigate whether readthrough products from G418-susceptible mutants are functional, we used split-NanoLucbased type IV collagen 345 heterotrimer formation assays. This assay allowed us to identify which full-length but missense mutant proteins are functional. Some mutations had high readthrough activity but may lose function when a different amino acid from the original is inserted. Many of the mutations are located in the COL4A5 C-terminal NC1 domain, which is essential for 5 to form a functional triple-helical structure with other -chains (16). Therefore, structural changes in the protein due to missense mutations are likely to interfere with the formation of the correct NC1 complex. However, the results showed that the wild-type S1632 and K1689 residues are not required to form the NC1 complex (Fig. 8 B, C). Also, the C29, S36 and E130 residues in the N-terminal 7S domain could be replaced without a total inhibition of heterotrimer assembly (Fig. 8 B-E). The results suggested that C29 substitution did not impact NC1 complex formation, but partially affected assembly at the N-terminus. This is consistent with the importance of the Cys residue in the 7S domain for disulfide bond formation to other chains (55). Using two different assay systems, the PTC-RT reporter assay and the 345 heterotrimer assay, we identified several mutations that seem truly susceptible to PTC-RT therapy.  (Fig. 5). In addition, from the viewpoint specific to Alport syndrome, type IV collagen 345 incorporated into the GBM should be stable for a long period of time (56). This means that once enough type IV collagen 345 is induced by PTC-RT and incorporated into the GBM, continuous treatment should not be necessary, though intermittent treatments would likely be required. This suggests that Alport syndrome is especially suitable for PTC-RT therapy.
In summary, the present study proposes PTC-RT as a novel personalized therapeutic approach for Alport syndrome based on the susceptibility of specific pathogenic nonsense mutations to readthrough. Forms of AS caused by nonsense mutations, which are classified as truncating mutations, are typically more severe than the non-truncating forms, which are usually caused by

Plasmids
To generate the COL4A5 with C-terminal NanoLuc fusion expression vector, full-length human  Table S1. The introduced mutations were verified by Sanger sequencing. To produce lentivirus, 293T packaging cells were seeded at 5.5-6.0x10 5 cells per wells in DMEM in 6-well tissue culture plates. Seeded cells were incubated at 37 ºC, 5% CO2 for ~20 h.

Cell Lysis, Gel Electrophoresis, and Immunoblotting
Transfected HEK293 cells were washed twice with ice-cold PBS and lysed in RIPA buffer (0.05 M Tris-HCl [pH 7.5], 0.15 M NaCl, 1% v/v Nonidet P-40, 1% w/v Na deoxycholate, and 1% protease inhibitor cocktail). The cell lysates were centrifuged at 14,000 g for 15 min at 4 ºC, and clear supernatants were collected. The protein concentration was determined using a bicinchoninic acid kit (Thermo), and equal amounts of protein lysates were loaded and separated by SDS PAGE, immunoblotted with anti-COL4A5 NC1 antibody (H52, Chondrex) and antivinculin antibody (7F9, Santa Cruz), and visualized using SuperSignal West Pico Chemiluminescent Substrate (Thermo).
After 48 h, culture media were changed to culture media containing the test compounds. At 72 h after transfection, Nano-Glo Dual Luciferase Reporter Assay reagent (Promega) was added, and the luciferase activity in the cell lysates was measured using a GloMax Navigator system (Promega). All luciferase assays were conducted in LumiNunc 96-well white plates (Invitrogen). NanoLuc luciferase was normalized by constitutively expressed firefly luciferase.

Statistical analysis
Statistical parameters are reported in the Fig. Legends. Immunoblot experiments were performed in triplicate using 3 independent transfections. Luciferase assays were performed in quadruplicate using 4 independent cell cultures. The significance of differences between two groups was assessed using Student's unpaired two-tailed t-tests. For three-group comparisons, we used analysis of variance (ANOVA) with Tukey-Kramer post-hoc or Dunnett's tests.
Differences with P values of less than 0.05 were considered statistically significant.

Data availability
All data are presented in the main manuscript and supporting information.
Luminescence was measured in cell lysates (B) and culture media (C) from HEK293T cells expressing C-terminal tagged COL4A5-Lg (WT or the indicated mutants), COL4A3-Sm, and non-tagged COL4A4. The HSV-TK-Luc2 (firefly) plasmid was included for normalization.