We ascertained a large Israeli Jewish family of Yemenite origin (family TB32,
Fig. 1A ). The family is not known to be consanguineous, but both parents originated from the same small village in Yemen. Both parents had normal vision and there was no known history of retinal degeneration in previous generations. Yet, five of their 14 offspring had severe retinal degeneration with early macular involvement diagnosed in their late teens to early 20s. We concluded that the most probable mode of inheritance in this family is autosomal recessive.
To identify the mutated gene underlying retinal degeneration in family TB32, we performed haplotype analysis with markers linked to all known autosomal recessive RP and CRD loci and genes (RetNet). A haplotype of three polymorphic marker alleles (
D2S2261,
D2S364, and
D2S2273) linked to the RP26 locus on the long arm of chromosome 2, region 31.3, cosegregated with retinal degeneration in this family
(Fig. 1A) .
CERKL, the mutated gene in the original RP26 family, was initially reported to comprise 13 exons.
4 8 Nine alternatively spliced isoforms of
CERKL have been submitted to GenBank.
8 Comparison of each of the nine
CERKL splice variants to the human genome sequence at the UCSC Genome Browser revealed that the gene actually comprises 15 exons, 14 of which are coding
(Fig. 2) . In most splice variants (8/9) the first exon is exon 1, which includes a translation initiation codon. However, one of the reported splice isoforms (variant i) starts with an alternative, noncoding first exon (0). In this isoform, the open reading frame starts within exon 4. An additional exon (4a) is included in only one of the splice variants (b). The different splice variants encode nine protein products, varying in length from 173 to 558 amino acids
(Fig. 2) .
To detect mutations in the CERKL gene, we determined the sequence of each of the 14 coding exons (exons 1–13), including exon–intron boundaries, in one of the affected individuals from family TB32. We identified two homozygous single-base changes, both of which were found to cosegregate with retinal degeneration in this family.
The first identified homozygous change is an A to C transversion located in exon 2, which leads to the substitution of alanine for aspartic acid at position 81 of the CERKL protein (variants a–h; p.D81A). Aspartic acid at position 81 is conserved in mouse, dog, and chimpanzee CERKL proteins; however, in rat Cerkl, there is a glycine at this position. Moreover, we screened a panel of 112 Yemenite Jewish control DNA samples and found a carrier frequency of 9% for the p.81A allele. We concluded that p.D81A is most probably a common polymorphism in the Yemenite Jewish population.
The second identified homozygous change is a G-to-A transition located at the conserved donor site of exon 1 (c.238+1G>A;
Fig. 3A ). This change is expected to lead to altered splicing of exon 1. Exon 1 is included in eight of nine
CERKL splice variants and harbors the translation initiation codon
(Fig. 2) . Due to the limited expression pattern of
CERKL, we could not evaluate the effect of c.238+1G>A on splicing in patient-derived RNA. Alternatively, we used an in vitro splicing assay approach. For this purpose, we created a minigene construct. This construct harbors
CERKL exons 1 and 2, flanked by 50 to 220 bp of intronic sequences, downstream of a CMV promoter
(Fig. 3B) . Constructs harboring either the wt or the c.238+1G>A allele were transfected into NIH3T3 cells, followed by RNA extraction and RT-PCR analysis with primers located in each of the exons. RNA derived from the wt construct was correctly spliced, whereas RNA derived from the mutant construct yielded two types of aberrantly spliced products. In one of these products, the entire intron was retained, whereas in the second product the intron was only partially retained, due to the use of a cryptic donor site, located 202 bp upstream of exon 2
(Figs. 3B 3C) . In silico analysis of this cryptic donor site (using Splice Site Prediction by Neural Network) yielded a relatively low score of 0.75 (in comparison to a score of 0.98 in the original donor site). Retention of intron 1 is expected to lead to premature translation termination, after the insertion of 115 incorrect amino acids, starting at position 80 of the CERKL protein. The in vitro splicing assay we performed demonstrates that an exon 1 donor splice site harboring the c.238 +1G>A change is not efficiently recognized by the mammalian splicing machinery. Although the exact effect of this splicing mutation on
CERKL transcripts in human retina is not known, the expected outcome is incorrect splicing, leading to an abnormal protein product.