In vivo correction of a Menkes disease model using antisense oligonucleotides

Abstract

Although the molecular basis of many inherited metabolic diseases has been defined, the availability of effective therapies in such disorders remains problematic. Menkes disease is a fatal neurodegenerative disorder due to loss-of-function mutations in the ATP7A gene encoding a copper-transporting P-type Atpase. To develop therapeutic approaches in affected patients, we have identified a zebrafish model of Menkes disease termed calamity that results from splicing defects in the zebrafish orthologue of the ATP7A gene. Embryonic-recessive lethal mutants have impaired copper homeostasis that results in absent melanin pigmentation, impaired notochord formation, and hindbrain neurodegeneration. In this current study, we have attempted to rescue these striking phenotypic alterations by using a series of antisense morpholino oligonucleotides directed against the splice-site junctions of two mutant calamity alleles. Our findings reveal a robust and complete correction of the copper-deficient defects of calamity in association with the generation of the WT Menkes protein in all rescued mutants. Interestingly, a quantitative analysis of atp7a-specific transcripts suggests that competitive translational regulation may account for the synthesis of WT protein in these embryos. This in vivo correction of Menkes disease through the rescue of aberrant splicing may provide therapeutic options in this fatal disease and illustrates the potential for zebrafish models of human genetic disease in the development of treatments based on the principles of interactions of synthetic oligonucleotide analogues with mRNA.

Menkes disease is an inherited metabolic disease due to loss-of-function mutations in the ATP7A gene encoding a P-type Atpase required for copper absorption and homeostasis (1, 2). Patients with Menkes disease have absent melanin pigmentation, impaired extracellular matrix formation, and neurodegeneration, resulting in intractable seizures, hypotonia, severe failure to thrive, and death in early infancy (3, 4). These pleiotropic phenotypes result from the loss of activity of essential cuproenzymes within distinct subcellular compartments. Treatment with exogenous copper is largely ineffective (5). Absent the development of directed pharmacotherapy, the restoration of normal gene function remains the most viable treatment for affected patients. Although partial correction of a murine model of Menkes disease was demonstrated by using a human Menkes transgene (6), methods of safely introducing WT cDNAs into humans have remained elusive.

We recently used chemical genetics to obtain calamity, a zebrafish model of Menkes disease (7). Embryonic mutants reveal a striking phenotype with absent melanin pigment, impaired notochord formation, and neurodegeneration. Because several calamity alleles arise from splicing defects in the zebrafish orthologue of ATP7A, we reasoned that if normal splicing could be restored, sufficient WT message might be generated to rescue these defects. Furthermore, because previous studies of copper nutrition in the zebrafish embryo revealed a hierarchy of temporal and dosage-dependent phenotypes (7), therapeutic windows might allow for long-term correction after transient restoration of WT expression. Such an approach has broad applicability because many metabolic diseases arise from mutations that interfere with normal splicing of the pre-mRNA message (8). In such cases, the threshold of the WT transcript necessary to restore functional protein may require only a small shift in favor of the WT message to ameliorate the disease.

RNA targeting has recently emerged as a potential alternative to more conventional approaches in gene therapy (9). Antisense oligonucleotides that bind to pre-mRNA and sterically alter processing have been used successfully in cell culture to demonstrate the utility of this approach (10–13). More recently, in murine models of Duchenne muscular dystrophy, systemic delivery of morpholino oligonucleotides has been shown to effect partial restoration of protein function on a cellular level in mutant mice (14–16). A critical factor in such strategies for the treatment of human metabolic disorders is the availability of robust disease models that permit rapid screening for effective phenotypic rescue. In this current study, we have used such an approach to rescue the phenotype of calamity, demonstrating the applicability to complex organisms of several recent cell culture studies using this technology in a number of human genetic diseases (13, 16, 17).

Results and Discussion

Recently, our laboratory has described the zebrafish mutant calamity (allele vu69) as a model of Menkes disease that recapitulates key aspects of the human phenotype (Fig. 1A) (7). The defect in calamity^vu69 is caused by the creation of a new 3′ splice site, resulting in a 7-bp insertion into the mRNA (Fig. 1F). The subsequent frame shift creates an early stop codon and the loss of >95% of the protein product (Fig. 1 D and E). In the course of continued screening, we have identified a second allele of calamity, designated as gw246. This allele causes an identical phenotype to the first and is noncomplementary in a standard genetic cross (Fig. 1 B and C). Cloning of the mRNA in this mutant revealed a 12-bp in-frame deletion at the 3′ end of exon 20, which removes 4 aa located in the critical and highly conserved ATPase domain of Atp7a (Fig. 1F). This deletion at the junction of two exons is due to the preferential use of a new 5′ splice site. Interestingly, genomic sequence analysis revealed the mutation of neither the WT 5′ splice site (splice donor) nor the creation of a new GT splice donor pair. Rather, the mutation was 5 bp downstream from the new mutant splice donor. Comparison of the mutant sequence to the previously established zebrafish splice donor consensus sequence indicated that a new splice consensus site had been formed (18). Given that the mutant displays such a severe phenotype, we inferred that this new consensus sequence creates a dominant splice donor to the near exclusion of the WT donor. High-resolution separation of a small PCR product from mutant cDNA demonstrates that this is indeed the case (Fig. 2B, lane 1). In this mutant, any WT splicing that does occur retains a single base change that causes a nonconservative amino acid change from Leu to Arg at position 1316 (Fig. 1F). Immunoblots with an antibody raised against a C-terminal peptide from zebrafish Atp7a revealed near-complete loss of protein product in both vu69 and gw246 (Fig. 1E). This lack of protein product in the gw246 allele suggests that there may be an early processing defect that causes rapid and early degradation of the protein.

Fig. 1.

Comparison of phenotype, genotype, and protein expression for two alleles of calamity. (A) calamity^vu69, characterized in ref. 7, displays the copper-deficient phenotype of loss of pigmentation, notochord abnormalities (arrowheads), and an enlarged ventricle and is due to a loss of function of atp7a. (B) calamity^gw246, a second allele of atp7a, has the same phenotype. (C) cal^gw246 fails to complement cal^vu69, confirming allelism. (D) Amino acid sequence changes caused by the mutations. cal^vu69 results in a frameshift and truncation of the C terminus of the protein. cal^gw246 contains an in-frame deletion in the highly conserved ATPase domain of Atp7a. (E) Immunoblot analysis using an antibody made to a C-terminal peptide from zebrafish Atp7a shows complete loss of WT protein in cal^vu69, cal^gw246, and in the compound heterozygote. (F) The mRNA and genomic sequences of cal^vu69 (Upper) and cal^gw246 (Lower), highlighting the mutation and subsequent changes in amino acid sequence due to missplicing of exons through the aberrant use of mutant splice sites (red boxes). Exon boundaries in the mRNA are illustrated by a change in color from magenta to blue. In the genomic sequence, introns are in lowercase and the mutations are the bases with the white background. The green boxes delineate the WT splice donor/acceptor pair and red boxes the mutant pair. For cal^gw246, a small percentage of transcripts splice at the WT splice junction retaining the nonsynonymous mutation, a Leu to Arg amino acid substitution (*).

Fig. 2.

A series of morpholinos causes alterations of splicing in cal^gw246. (A) Schematic and sequence information of the morpholino design for exploring alterations in splicing caused by variable morpholino placement with respect to WT and mutant splice donors. Exon sequence is uppercase. The mutation is red. Lines indicate the binding site of each listed morpholino. (B) Injection of the series of morpholinos in A into cal^gw246. Splice forms were amplified by PCR of cDNA using the primers illustrated in A. The WT and three alternate splice forms were seen on polyacrylamide gel stained with ethidium bromide. Each was sequenced to identify exact locations of splicing that are illustrated in A. All cryptic splice forms were located within exon 20.

The presence of the WT splice sites in both alleles of calamity immediately caused us to consider the use of morpholinos directed to specific sequences near the mutant splice sites in a bid to force WT splicing events. There were no data available indicating the precise location that would be best suited for accomplishing this goal. At the same time, the nature of both alleles provided an ideal system for investigating this very question. In the case of gw246, we had two competing splice donors sufficiently far apart that we might interfere with one or the other, but close enough that we also might block both.

Design of a series of morpholinos that would walk along the exon–intron boundary sequence was based on a comparison with the known consensus sequences (Fig. 2A). The basic 5′ splice site consensus sequence is small and well defined, making targeting straightforward for the gw246 mutant (18). Wherever a morpholino sequence overlapped a mutation, the mutant sequence was used at that base pair. The first series of morpholinos was injected at a dose of 1.44 pmol per embryo (12 ng per embryo) into intercrosses of gw246 heterozygous carriers. At 48 h after fertilization, embryos were examined for phenotypic effects, and mutants were collected for mRNA extraction and RT-PCR. The effects on splicing are shown in Fig. 2B. Across the series of morpholinos, several splice forms were seen. We confirmed that each band is a splice form of atp7a via direct sequencing of the RT-PCR product. Several patterns emerged from these experiments. First, and most interesting, the blocking of the mutant splice donor did indeed restore a significant amount of WT splice product (lanes e2 and e3). However, this finding was not accompanied by a rescue of the phenotype. We surmised that this result was due to the retention of a single base change caused by the mutation in the final, properly spliced transcript that results in the nonconservative L1316R substitution. Such a large change in polarity and charge in a highly conserved region of this ATPase might be expected to disrupt function. Consistent with this hypothesis, when we introduced this mutation into the WT cDNA, it resulted in an ATPase that fails to deliver copper to tyrosinase in transiently transfected ATP7A-deficient fibroblasts, indicating that this mutation results in an inactive transporter [supporting information (SI) Fig. 5]. Alternatively, the mutation may render the protein unstable, leading to early degradation and subsequent loss of function.

Second, the injection of the series of morpholinos revealed several other weak cryptic splice sites all contained within exon 20, immediately 5′ to the mutation (Fig. 2). This is most clearly seen when the morpholino overlies both the WT and mutant sites (lanes e4 and e5). There was no evidence of exon skipping induced by any of the morpholinos (data not shown). Bands corresponding to the inclusion of the entire intron in the final product were seen, but we could not totally rule out the possibility of genomic contamination of the cDNA as the source of these bands (data not shown).

Third, the morpholino that accomplished the most robust rescue of mRNA splicing (lane e3) overlay the mutant splice site on either side and remained at least 6 bp 5′ to the WT GT pair. Moving only four bases in either direction resulted in either an increased use of the mutant site or an increase in the upstream cryptic sites due to blockage of the WT and mutant sites (lanes e4 and e5).

Last, when the morpholino target sequence was located 3′ to the mutant splice donor, the mutant again became the dominant splice form (lane e7). WT embryos injected with this same morpholino fully displayed the calamity phenotype (SI Fig. 6B). Combined with the RT-PCR data, this indicates that when the WT 5′ splice site is blocked by a morpholino, the next most robust cryptic site is the one used in the mutant. Injection of WT embryos also resulted in the same pattern of splicing, illustrating why the mutation is so easily capable of coopting splicing to the cryptic site in the mutant (SI Fig. 6A).

These specific observations on a trial-and-error approach allow us to propose a set of generalities regarding morpholino targeting to an exon–intron junction. First, morpholinos targeted within 30 bp of a 5′ splice site will have an effect on splicing even if it is subtle. Second, in some cases, it appears that the magnitude of the effect of the morpholino depends not only on its distance from the splice site, but also on the general strength of the splice consensus (cf. lanes e1 and e4) and surrounding cryptic sites. Third, optimal blocking occurs in two forms: (i) the morpholino is entirely exonic except for a 2-bp overlap with the GT pair (lanes e2 and e5), or (ii) the morpholino extends at least 6 bp on either side of the exon–intron junction. Fourth, morpholinos can reveal unappreciated enhancers and suppressors of specific splicing events located outside of the consensus sequence (lane e10) and can be used to change the hierarchy of cryptic splicing events. Validation of these rules will require separate experiments in a different mutant with a similar defect and distinct phenotype; however, they do establish a framework that can guide further testing.

Taken together, these results demonstrate the utility of using morpholinos to correct aberrant splicing in an embryonic vertebrate in cases where WT splice donors are not affected. Morpholino e3 strongly rescues the mRNA. Unfortunately, we did not see a corresponding rescue of the phenotype. That this mutant protein was incapable of rescue indicates one of the potential caveats of this system: that the retention of a mutation in the rescued mRNA can just as strongly affect the phenotype of the organism as the misspliced sequence. Promisingly, however, it has been estimated that the probability that a random amino acid change would cause protein inactivation is ≈34% (19). Conversely, this would mean that just under two-thirds of random amino acid changes will result in functional protein sequence. Together with alternate approaches, such as forced exon skipping, the probability of rescuing any given splice mutation with morpholinos remains fairly high.

Unlike the gw246 allele, the rescue of vu69 would not cause inclusion of the mutated base, but would fully restore the WT sequence without alteration. Also unlike gw246, the mutation alters the splice acceptor. The length and placement of consensus splicing sequences around the splice acceptor are much less well defined and more degenerate, making targeting predictions much more cumbersome (18). Given the proximity of the vu69 splice acceptor to the polypyrimidine tract, we reasoned that directly overlapping the mutant splice acceptor with a morpholino would interfere with WT splicing. Similarly, we assumed that both mutant and WT splice acceptors would use the same polypyrimidine tract, making it difficult to preferentially target one splice acceptor by direct targeting of the polypyrimidine tract. However, the inherent degeneracy of the U2AF recognition sequence might permit a morpholino placed nearby to sterically interfere and alter the exact location of binding of the U2AF to the polypyrimidine tract (20). No structural data existed to aid in the design of such a morpholino. Therefore, we took a similar approach using a series of morpholinos that “walk” across the splice acceptor region, but we expanded our target area to include sequences up to 73 bp 5′ to the mutant acceptor (Fig. 3A). Briefly, morpholinos i1 and i2 both caused a more severe phenotype in calamity mutants and heterozygotes, implying an interference with WT splice events (data not shown). Morpholinos i5 to i7 had no effect on phenotype. Morpholinos i3, i3.5, and i4 caused robust rescue of the phenotype of mutant calamity embryos (Fig. 3 B–G). Melanin pigmentation was restored, and the notochord did not contain any discernible defects, compared with calamity-mutant embryos. The rescue was seen in 100% of the mutant embryos, with all injected mutants containing at least a few melanocytes and 95% having >20. No injected mutant embryo had a discernable notochord defect when scored at 48 h after fertilization (Fig. 3H). The rescue effect was still clearly visible at 6 days after fertilization, when ≈75% of the uninjected calamity embryos were either dead or dysmorphic (Fig. 3I), compared with nearly normal-looking rescued embryos (Fig. 3J). Lowering the dose of the morpholino resulted in similar pigmentation, with intermediate notochord defects consistent with previous observations on the developmental hierarchy of copper metabolism that revealed a critical threshold of lysyl oxidase activity necessary for normal notochord development (data not shown and refs. 7 and 21). Sequencing of genomic DNA confirmed that the rescued embryos were indeed mutant (data not shown).

Fig. 3.

Morpholinos rescue the phenotype of cal^vu69 embryos. (A) Schematic and sequence information for the design of morpholinos directed toward the rescue of the mutant splicing in cal^vu69. The WT exon is uppercase. The mutation is highlighted in red, and the WT splice acceptor is in green. Each morpholino-binding site is indicated by a line located 5′ to the polypyrimidine tract. Morpholinos that cause phenotypic rescue are highlighted in pink. Morpholino i1 served as a control to demonstrate that blocking of proper splicing at this exon–exon boundary would result in the calamity phenotype (data not shown). (B and C) Side-by-side comparisons of zebrafish 48 h after fertilization. (Top) WT sibling from cal^vu69 intercross clutch. (Bottom) cal^vu69/vu69-mutant embryo. (Middle) cal^vu69/vu69 embryo rescued by injection of morpholino i3. (D and E) Close-up view of cal^vu69/vu69-uninjected embryo showing lack of melanin pigmentation over head region and notochord abnormalities (arrowhead). (F and G) Close-up view of morpholino i3-rescued cal^vu69/vu69 embryo showing restoration of melanin pigmentation and a notochord absent of defects. (H) Quantification of the extent of rescue. A single group of embryos derived from several clutches, half injected with rescuing morpholino (red bars; uninjected, blue bars), was scored for the presence of any or a significant number of melanocytes (>20) and the presence of notochord defects. The scale of the graph reflects the expected 25% ratio of homozygous mutants derived from the heterozygous intercross. (I and J) Six days after fertilization, embryos were either uninjected (I) or injected (J) with rescuing morpholino i3.5. Rescued embryos maintain near-normal body morphology, whereas unrescued embryos swell, possibly because of decreased extracellular matrix integrity due to the loss of lysyl oxidase activity.

Because the mutant embryos had clear phenotypic rescue, we wanted to confirm that this was the result of restored protein expression. Immunoblot analysis of injected and rescued mutants with a zebrafish Atp7a-specific antibody shows the restoration of full-length WT protein in the rescued embryos, compared with no visible protein in uninjected mutants (Fig. 4A). To our surprise, however, this did not correspond with an increase in WT splice products as measured via quantitative (q)RT-PCR. Only three consistent patterns emerged from these experiments. The first is the presence of a small amount of amplifiable WT transcript in all of the mutants, consistent with the presence of a WT band when endpoint RT-PCR products are run on a polyacrylamide gel (Fig. 4C and SI Fig. 7 B and C). The second is that there is decreased total atp7a transcript in the mutants probably due to nonsense-mediated decay of mutant transcripts (SI Fig. 7D). The third is that there is a significant reduction of mutant transcript in the injected embryos. The most likely explanation for this reduction is the shifting of this transcript to a third splice form, which is corroborated by the RT-PCR gel analysis showing a third splice form containing an 800-bp intronic insertion due to the use of a cryptic 3′ splice site within the intron (Fig. 4C). Sequence analysis of the alternate splice form indicates a frame shift and early stop codon, which is analogous to the original mutation. It is important to note that, because the antibody used for immunoblot recognizes the C terminus of the WT zebrafish protein, the identification of the full-length protein will only occur when WT transcripts are present.

Fig. 4.

Rescue morpholinos cause restoration of full-length protein without detectable changes in mRNA. (A) Phenotypic rescue by morpholinos i3 and i3.5 is accompanied by the restoration of the WT protein. Forty-eight hours after fertilization, lysates from embryos of the indicated genotype were run on a SDS/6% PAGE gel and blotted for zebrafish Atp7a by using an antibody generated to the zebrafish protein. β-catenin was used as a loading control. Morpholino injection restored protein to 30–45% of WT levels in this experiment. (B) qRT-PCR using primers specific for WT and mutant atp7a transcript were used to measure the amounts of each in single embryos. Normalization of WT transcript was to cal^+/+ embryos and mutant transcript to cal^−/− embryos. Error bars indicate one standard deviation around the mean. (C) Polyacrylamide gel of RT-PCR products reveals a band containing 800 bp of intronic sequence caused by morpholino injection. This band was sequenced and represents use of a cryptic 3′ splice site located within the intron. Also seen are mutant and WT transcripts in uninjected and rescued embryos. A nonspecific band was present in both uninjected and injected mutant embryos (*).

Both the reduction in mutant transcript and the lack of any increase in WT transcript levels as measured by qRT-PCR were consistent across three embryo ages 12, 24, and 48 h after fertilization (data not shown), indicating that there is not an early correction of splicing that decreases as the morpholino is diluted during development. The lack of a measurable substantial increase in WT transcript in these studies could simply reflect the limits of quantitation for the threshold necessary to result in rescue. If so, then it suggests that only small changes in WT mRNA can result in a significant phenotypic improvement on an organismal level, at least in the case of Menkes disease. Alternatively, the data suggest that the increase in protein levels could result from the removal of inhibition on translation of the WT transcript. Such inhibition could arise from the competition for translation from the mutant transcript, a concept that would be consistent with observations on translational competition in Drosophila (22) and is supported by our finding that knockdown of the mutant transcript best correlates with rescue of the calamity phenotype through restoration of protein levels (Fig. 4). Although this proposed mechanism for the increase in functional protein is not directly testable under these conditions, if plausible, this significantly broadens the applicability of the morpholino approach because it demonstrates that loss of one particular mutant form may be sufficient to cause rescue of protein function and phenotype, with only minimal increases in the amount of WT mRNA.

Taken together, the data reported here have several implications of direct relevance to the development of effective therapeutics in Menkes disease. Experimental (7, 21) and clinical observations in affected patients (23) reveal that the maintenance of CNS function in this disease is prioritized under circumstances of limited copper availability or decreased ATP7A function. This hierarchy of copper delivery suggests that the prevention of the neurodegenerative features in Menkes disease may be possible with therapeutic interventions that result in a modest increase in ATP7A within a specific developmental window. The in vivo rescue observed in these studies directly supports these pathophysiological concepts and raises the potential for in utero interventions in this disease. The data in this article demonstrate the utility of zebrafish for rapid screens to identify specific antisense sequences effective in phenotypic rescue. Although morpholino delivery remains problematic, given the robustness of the assay reported here and the increasing number of zebrafish models of human genetic disease (24), these findings support further study of this technology in the development of effective therapeutics for inherited metabolic disease.

Methods

Animal Maintenance.

Fish stock maintenance was performed according to institutionally approved procedures. Embryos for experiments were obtained through in vitro fertilization and incubated at 28.5°C. The WT fish were of the AB strain. The mutant calamity has been described (7). Mutagenesis of WT fish was performed by using the chemical mutagen, ethyl nitrosourea. Mutants were screened for the calamity phenotype comprising lack of pigmentation and notochord defects. Putative mutants were crossed to calamity to verify allelism. These were then out-crossed to AB fish and grown to adulthood. Regions of genomic DNA containing putative mutations were sequenced in both AB and WIK strains to rule out the possibility of strain-dependent polymorphisms.

Morpholino Synthesis and Injection.

Morpholinos were synthesized by GeneTools, LLC. Morpholinos were dissolved to 2 mM in water and stored at −80°C. Injection of morpholinos was performed on a microinjection apparatus (Harvard Instruments). A constant volume (1.44 nl) of morpholinos was injected at different dilutions of the stock (1:2 to 1:10) to optimize the dose that gave the strongest phenotypic effect while minimizing off-target effects.

Molecular Cloning and PCR.

Total RNA was extracted from embryos by using the TRIzol reagent (Invitrogen). Random hexamer-primed cDNA was synthesized by using SuperScriptIII according to the manufacturer's protocol (Invitrogen). PCR was performed by using Phusion DNA polymerase (Finnzymes). High-resolution separation of products was performed on 10% TBE-polyacrylamide gels. All sequences were verified by direct sequencing. The L1316R clone of zebrafish atp7a was generated from the WT sequence by using the QuickChangeII site-directed mutagenesis kit (Stratagene). The altered clone was sequenced, and other PCR errors were excluded.

qRT-PCR.

qRT-PCR was performed by using primers specific to either the mutant or WT form of the Menkes transcript. This was accomplished by designing primers that were complementary to the exon–exon junction affected in the mutant (SI Fig. 7A). The following primer sequences were used: common forward primer, ATGATGAGCTCCGGACAGAC; calamity-specific reverse, GGAATGATCTTTTCCACCTGAG; WT-specific reverse, GGAATGATCTTTTCCACTGTCG; total atp7a primers, TGGAGCTTGTGGTCAGAGG and AGGGCAACTGAAGCGTAGAG; and ornithine decarboxylase primers, ATCTGGATCTCCGTTTTGCT and CCGTTTTACGCAGTGAAGTG. Primer efficiencies were calculated in triplicate for each primer set by using either WT or mutant cDNA, respectively. Random hexamer-primed cDNA was synthesized by using SuperScript III (Invitrogen) according to the manufacturer's protocol. qRT-PCR was carried out on an iCycler (BioRad) with iQ Sybr green SuperMix (BioRad). Calculations were performed by using the Pfaffl method, and each bar in the figures is the average of three to five individually prepared embryo RNA samples. Relative to WT, all cal embryos had a 50% decrease in total atp7a transcript (data not shown).

Immunoblots.

Forty-eight hours after fertilization, 20–30 zebrafish embryos were gently homogenized in sucrose homogenization buffer [250 mM sucrose and 5 mM Tris (pH 7.4)] plus protease inhibitors (Roche) and then spun at 350,000 × g in a table-top Beckman ultracentrifuge for 30 min. The pellet was then resuspended in 50 μl of RIPA buffer. Alternatively, 48 h after fertilization, embryos were manually deyolked and homogenized directly in RIPA buffer. Equal amounts of protein were loaded on 6% polyacrylamide gels. After transfer to nitrocellulose, immunoblotting was performed by using a rabbit polyclonal antibody raised to a C-terminal peptide of zebrafish Atp7a. Anti-rabbit secondary antibody and ECL reagents were purchased from Pierce and used according to the manufacturer's recommendations.