Frequency of Mutations in the Gene Encoding the α Subunit of Rod cGMP-Phosphodiesterase in Autosomal Recessive Retinitis Pigmentosa

Abstract

purpose. To determine the mutation spectrum of the PDE6A gene encoding theα subunit of rod cyclic guanosine monophosphate (cGMP)-phosphodiesterase and the proportion of patients with recessive retinitis pigmentosa (RP) due to mutations in this gene.

methods. The single-strand conformation polymorphism (SSCP) technique and a direct genomic sequencing technique were used to screen all 22 exons of this gene for mutations in 164 unrelated patients with recessive or isolate RP. Variant DNA fragments revealed by SSCP analysis were subsequently sequenced. Selected alleles that altered the coding region or intron splice sites were evaluated further through segregation analysis in the families of the index cases.

results. Four new families were identified with five novel mutations in this gene that cosegregated with disease. Combining the data presented here with those published earlier by the authors, eight different mutations in six families have been discovered to be pathogenic. Two of the mutations are nonsense, five are missense, and one affects a canonical splice-donor site.

conclusions. The PDE6A gene appears to account for roughly 3% to 4% of families with recessive RP in North America. A compilation of the pathogenic mutations in PDE6A and those reported in the homologous gene PDE6B encoding the β subunit of rod cGMP-phosphodiesterase shows that the cGMP-binding and catalytic domains are frequently affected.

Retinitis pigmentosa (RP) is a hereditary degenerative disease of the retina leading to blindness. There are both syndromic and nonsyndromic forms, and the inheritance can be dominant, recessive, X-linked, maternal, or digenic. It is genetically heterogeneous with more than 50 loci implicated through gene identifications or linkage studies. Two of the identified genes encode the active (α and β) subunits of cyclic guanosine monophosphate(cGMP)-phosphodiesterase, a component in the rod phototransduction cascade. ¹ The two subunits are similar both in size (860 and 854 residues, respectively) and in sequence. ^{2 3 4} Mutations in the PDE6B gene (chromosome 4p16.3) ^{4 5} encoding the β subunit of this enzyme account for approximately 4% of cases of recessive RP. ^{6 7 8} With regard to the PDE6A gene (chromosome 5q31.2 to 34) ² encoding the α subunit, our group ⁹ and Meins et al. ³ have described mutations in only three families with recessive RP. The report from our group was a partial evaluation of some of the exons of this gene. ⁹ Here, we describe a comprehensive evaluation of all 22 exons in a subset of the same group of patients. 

Methods

This study, which involved human subjects, conformed to the tenants of the Declaration of Helsinki. The diagnostic criteria used at this center for RP and for a recessive inheritance pattern have been described previously. ^{10 11} We excluded at the outset patients with RP known to be caused by pathogenic mutations in any other RP gene that had been analyzed in our laboratory. All the patients in this set had been included in our previously published evaluation of some of the exons of the PDE6A gene. ⁹ That report involved 173 patients with recessive RP, 9 of whom were excluded from this study. Patients were excluded because a pathogenic mutation in another RP gene has since been discovered, because a review of records uncovered that two index cases from the same family were both inadvertently included previously as unrelated patients, or because there was insufficient leukocyte DNA available in the laboratory to complete this study. Some patients previously misclassified as having recessive RP were recategorized as having isolate RP, because they had no affected siblings, and they were not the offspring of a consanguineous marriage. One hundred sixty-four patients were in this study, comprising 146 patients with recessive RP and 18 newly recategorized patients with isolate RP. Most of the patients resided in the United States or Canada. Control subjects for this study were without symptoms of retinal degeneration or a previous family history of retinal degeneration; they were also from the United States and Canada and, as a group, were of ethnic composition comparable to the affected patients in this study. 

The laboratory analysis was performed on DNA purified from blood samples from these patients. Mutations were discovered using the single-strand conformation polymorphism (SSCP) technique. ¹² Primer sequences and reaction conditions for amplifying each of 22 exons were identical with those reported previously by Huang et al., ⁹ except for exons 6 and 13, for which the following pairs of primers were used, respectively (sense; antisense): 5′-ATTTTTCTCTCTTTTGCCAG-3′; 5′-TGTCTTTTGACAGGTGAAAC-3′; 5′-TATTCCAACCCTCATGAGAC-3′; 5′-TACCATGTAGAGTCTGCATG-3′. This analysis included the entire coding sequence and flanking intron sequences containing all splice donor and acceptor sites except the splice acceptor site of intron 1. DNA samples with variant bands observed by SSCP were subsequently sequenced. The investigation of the segregation of alleles was through analysis by SSCP or by direct sequencing of index patients and all available relatives. 

Computations of the likelihood that a sequence variant might create a splice-acceptor or splice-donor site were according to the method of Reese et al. ¹³ and were performed by that group’s Web site (available at http://www-hgc.lbl.gov/projects/splice.html). 

Results

We discovered five novel mutations that were judged to be responsible for recessive RP. Using published recommendations for naming mutations, ¹⁴ these mutations are Arg102His (CGC to CAC), Arg102Ser (CGC to AGC), IVS6+1G→A, Gln569Lys (CAG to AAG), and Ser573Pro (TCC to CCC) (Fig. 1) . Each of these changes was found in affected subjects who were either homozygotes or compound heterozygotes for mutations in this category. In addition, these mutations cosegregated with RP in the respective families (Fig. 2) . None of them was found among unrelated normal control subjects (70 control subjects were evaluated for the mutations Arg102His and Arg102Ser in exon 1, 92 for the mutation at the splice-donor site of intron 6, and 79 for the mutations Gln569Lys and Ser573Pro in exon 13.) 

Six missense changes of uncertain pathogenicity were identified: Asn216Ser, Val277Ala, Pro293Leu, Val391Met, Lys827Gln, and Gly850Val (Table 1) . None of these appeared to be a polymorphism, because the minor allele frequencies were less than 1% among the set of surveyed patients. Each was found heterozygously in only one or two unrelated patients. No abnormality was detected in the PDE6A sequence of the other allele in any of these patients. The Val277Ala change was not found in the only affected sibling of the index patient, and it is therefore not likely to be the cause of recessive RP in that family. The Pro293Leu change was present heterozygously in the only affected child of parents who were first cousins; because a homozygous mutation would be expected in the affected offspring from consanguineous parents, the allele is also unlikely to be the cause of recessive RP. However, these results do not rule out the possibility that the Val277Ala and Pro293Leu changes are pathogenic, only that they are not responsible for RP in these families. The Asn216Ser and Lys827Gln changes were present heterozygously in the affected siblings of the index cases with recessive RP, but in each case there was only one affected sibling; therefore, this cosegregation could be explained by chance alone. The family of the index patient with the Val391Met change was unavailable for segregation analysis. The segregation analysis of the Gly850Val change was uninformative: the change was present heterozygously in the proband (an isolate case of RP) and in the proband’s mother, but not in the father and the only sibling. The Gly850Val allele was found heterozygously in one patient in the survey of Meins et al. ³ (It was labeled Gly849Val in that article; it is renumbered as Gly850Val here because of a revision in the cDNA sequence found in Table 3 of Meins et al. ³ that changes the specificity and numbering of codons after 845.) 

We encountered 17 variant sequences that were judged unlikely to affect the sequence of the encoded protein (Table 1) . Ten of these were silent changes in the coding region, affecting codons 28, 34, 111, 155, 362, 597, 746, 779, 800, and 808. The silent changes affecting codons 111, 155, 597, 779, 800, and 808 were probably polymorphisms, because the minor allele frequency of each was above 1% (summing data from the recessive and isolate patients). The minor alleles for the polymorphisms at codons 111 and 155 were always found together in heterozygotes and homozygotes, indicating that the changes are syntenic. The changes affecting codons 111, 155, 779, 800, and 808 were also encountered by Meins et al. ³ Eight changes were within introns at some distance from the canonical splice-donor or -acceptor sites. Of these, five were probably polymorphisms (see Table 1 for allele frequencies), and one (IVS18+21A→C) was also reported by Meins et al. ³ Four of the sequence variants (Phe597Phe, IVS18+21, Phe779Phe, and Asp800Asp) were in linkage disequilibrium. Three patients (2 with recessive RP and 1 isolate case) were homozygous for the minor allele at all four sites, and 10 patients (7 with recessive RP and 3 isolate cases) were heterozygous at all four sites. Only four patients, all with recessive RP, had some but not all of these four changes: two patients with the minor allele at codon 779 (one homozygously and one heterozygously) were homozygous for the common allele at the other three sites, and two patients were heterozygous for only the Phe597Phe variant. None of these latter four changes was thought likely to create an intron splice-acceptor or -donor site except for the IVS18+21 change, for which the less common allele created a sequence that was suggestive of a new splice-acceptor site (score = 0.18, at which the false-positive rate is approximately 5%). However, a segregation analysis was performed on the family of one patient who was homozygous for the rare allele. The allele did not cosegregate with disease, because an affected sibling was found to be heterozygous for this change. Furthermore, this polymorphism was also found among normal control subjects with an allele frequency not statistically different from that found in the patients analyzed (χ² = 0.503; P = 0.48). The less common allele was therefore considered unlikely to be pathogenic. 

Discussion

Summing the results from this study and an earlier one performed by our group, ⁹ eight pathogenic mutations in the PDE6A gene in 6 families were discovered among 164 families with recessive or isolate RP. From these results it can be estimated that approximately 3% to 4% of cases of recessive RP are caused by mutations in PDE6A. This value is approximate because of the small number of ascertained cases, because the SSCP screening technique misses approximately 10% of point mutations, ¹⁵ because the SSCP technique usually does not detect large gene deletions or rearrangements, and because the upstream and downstream sequences of the gene and most of the intron sequence were not evaluated. The estimate of prevalence does not change substantially if the 13 patients with recessive RP who were excluded from this analysis are taken into account. (Those patients were excluded because they had mutations in other RP genes discovered before the onset of this study or during its course. Four of the excluded patients had recessive RP due to mutations in PDE6B, ⁶ three had mutations in the gene encoding the α subunit of the cGMP-gated channel, ¹⁶ one had a homozygous mutation in the rhodopsin gene, ¹⁷ two had mutations in the TULP1 gene, ¹⁸ and three had mutations in the RPE65 gene. ¹⁹ ) The estimated prevalence of 3% to 4% for PDE6A mutations among families with recessive RP is close to the 4% prevalence estimated for PDE6B by summing data from three groups (4 of 92 families, ⁶ 3 of 19 families, ⁷ and 2 of 101 families. ⁸ ) 

Of the two active subunits of rod cGMP-phosphodiesterase, the β subunit is more often studied. Naturally arising, recessive defects in this gene cause retinal degeneration in mice (the rd strain), ^{20 21 22} in Irish setter dogs (rcd-1), ^{23 24 25 26} and in humans. ^{6 7 8 27 28 29 30 31 32 33} A dominant mutation in the PDE6B gene has been discovered in some Danish families with congenital stationary night blindness. ^{8 34 35} Mutations in the PDE6A gene encoding the α subunit cause recessive retinal degeneration in Welsh Cardigan Corgi dogs ³⁶ and recessive RP in humans. ^{3 9}  

The amino acids normally specified by codons 102, 569, and 573 affected by the novel missense mutations described in this article are conserved among the following phosphodiesterase subunits: the rod α subunit of mouse, ³⁷ cow, ^{2 38} and dog ³⁹ ; the rod β subunit of human, ⁴ mouse, ³⁷ cow, ⁴⁰ and dog ²⁴ ; and the cone α′ subunit of human, ⁴¹ cow, ⁴² and chicken. ⁴³ Specifically, the positions equivalent to codons 102 and 569 are always Arg and Gln, respectively, whereas the position equivalent to codon 573 is either Ser or Thr, two residues with similar side groups. In contrast, most of the missense changes of uncertain pathogenicity affected residues that have not been highly conserved in evolution: the residue in the location of Asn216 is Ser in bovine cone phosphodiesterase α′, ⁴⁰ Val277 is Ile in human and bovine cone phosphodiesterase α′, ^{41 42} Pro293 is Ser in human rod phosphodiesterase β, ⁴ and Lys827 and Gly850 are in a poorly conserved region near the carboxyl-terminus of the protein. The position equivalent to codon 391 is Val in all vertebrate rod and cone cGMP-phosphodiesterase α, β, and α′ subunits sequenced to date. However, the Val391Met change would not substantially alter the side group at position 391, because both Val and Met have nonpolar side groups of similar size. These comparisons to photoreceptor phosphodiesterases from other vertebrates are weak evidence that the missense changes Asn216Ser, Val277Ala, Pro293Leu, Val391Met, Lys827Gln, and Gly850Val are not pathogenic. However, the evidence from these comparisons and from the pedigree analyses mentioned earlier is insufficient to be conclusive. 

The mechanisms by which PDE6A and PDE6B mutations lead to RP are probably similar. Both the α and β subunits of cGMP-phosphodiesterase are necessary for the enzyme’s function. Mice and dogs with recessive mutations in the PDE6B gene have abnormally high concentrations of cGMP in their retinas before the severe loss of their photoreceptors. ^{44 45} These elevated levels of cGMP arise presumably because of an absent activity of cGMP-phosphodiesterase in the mutant photoreceptor cells. The cGMP levels are sufficiently elevated to be toxic to photoreceptors, although the specific biochemical pathways mediating the toxicity are unknown. One attractive hypothesis is that the high cGMP concentration causes an increase in the proportion of open cGMP-gated channels in the rod outer segment membrane. Only a small percentage of these channels are open in normal rod photoreceptors in the dark-adapted state when the cGMP concentration in the outer segment is physiologically at its highest level. ⁴⁶ The cGMP levels in mutant photoreceptors without functional phosphodiesterase are much higher than these physiologic concentrations. The high cGMP levels should result in a higher-than-normal proportion of open channels and a presumably toxic increase in the influx of sodium and calcium ions into the cytoplasm. 

Most of the patients in this study had been evaluated for mutations in the PDE6B gene. ⁶ We considered the possibility that some patients may have RP because of double heterozygosity for mutations in both the PDE6A and PDE6B genes. However, none of our patients with PDE6A alleles categorized as pathogenic also had a missense change in the PDE6B gene. This was also true for the patients with the rare variant missense changes in PDE6A of uncertain pathogenicity. These negative results leave as an unanswered question what the phenotype of a double (PDE6A plus PDE6B) heterozygote would be. Because recessive mutations of both genes are known in dogs, ^{24 36} the procedure to determine the double-heterozygote phenotype in that species should be straightforward, through appropriate breeding. 

Figure 3 depicts the locations of the known pathogenic mutations in the PDE6A and PDE6B genes causing retinal degeneration in humans, mice, and dogs. Mutations causing recessive RP were included in this figure only if they have been reported in affected people who are homozygotes or compound heterozygotes. The figure also shows the location of the missense mutation found in some Danish families with dominantly inherited congenital stationary night blindness. ^{8 34 35} Among the mutations causing recessive RP, some are highly likely to be null alleles, because they are splice-site mutations or they are frameshift or nonsense mutations that would truncate the carboxyl-terminus of the protein and eliminate most or all the catalytic domains. The novel splice site mutation (IVS6+1G→A) found in this study is included in this set of highly likely null alleles. Of the 12 missense mutations in these genes, 9 affect residues within the cGMP-binding and catalytic domains or the isoprenylation site, including the 4 novel missense changes in the PDE6A gene (Arg102His, Arg102Ser, Gln569Lys, and Ser573Pro) described in this study. The PDE6B mutation Leu854Val ⁸ affecting the isoprenylation site at the C terminus of the protein suggests that this posttranslational modification of the β subunit also is essential to the enzyme’s function in photoreceptors. The fact that many pathogenic mutations affect the cGMP-binding and catalytic domains is in accord with the hypothesis that retinal degeneration results from interference with phosphodiesterase activity and not with some other function of this enzyme. However, when all pathogenic missense changes in both genes are included, there is no statistically significant clustering of the mutations in these domains, considering that these domains represent approximately 65% of the primary sequence of the protein. 

 Reprint requests: Thaddeus P. Dryja, Massachusetts Eye and Ear Infirmary, 243 Charles Street, Boston, MA 02114.

 Supported by grants from the National Eye Institute: EY08683 and EY00169; the Foundation Fighting Blindness; and the Massachusetts Lions Eye Research Fund; and by private donations to the Taylor Smith Laboratory and the Ocular Molecular Genetics Institute. TPD is a Research to Prevent Blindness Senior Scientific Investigator.

 Submitted for publication October 8, 1998; revised February 18, 1999; accepted March 10, 1999.

 Proprietary interest category: N.

 

Figure 1.

 

DNA sequence of mutations in the PDE6A gene that are highly likely to be pathogenic. Above each panel are the identification numbers of the proband carrying the respective mutation and a control subject with the wild-type sequence. All sequences are oriented in the sense (5′ to 3′) direction from bottom to top. (A) Sequence of exon 1 showing the heterozygous mutation Arg102His in patient 003-040 from family 5965. (B) Sequence of exon 1 showing the heterozygous mutation Arg102Ser in patient 003-043 from family 6877. (C) Sequence of the exon 6/intron 6 boundary showing the mutation IVS6+1G→A affecting the canonical splice donor site of intron 6 found homozygously in patient 003-080 from family 6201. (D) Sequence of exon 13 showing the mutation Gln569Lys, which is homozygous in patient 003-010 from family 6736 and heterozygous in patient 003-040 from family 5965. (E) Sequence of exon 13 showing the heterozygous mutation Ser573Pro in patient 003-043 from family 6877.

Figure 2.

 

Schematic pedigrees of newly identified families with mutations in the PDE6A gene causing recessive RP. Filled symbols indicated affected family members. Below each symbol is each person’s genotype at the PDE6A locus. + , the wild-type sequence at the relevant position. The identification number of each family is at the upper left of each schematic pedigree.

Table 1.

 

Polymorphisms, Rare Silent Variants, and Rare Variants of Uncertain Pathogenicity

Sequence Variation	Exon or Intron	No. of Respective Alleles (Minor Allele Frequency)
				Recessive RP	Isolate RP	Normal Controls
Tyr* 28Tyr* (TAC vs. TAT)	1	290:2	36:0
Ser* 34Ser* (TCC vs. TCT)	1	291:1	35:1	140:0
Arg+111Arg+ (AGG vs. CGG)	1	256:36 (0.123)	29:7 (0.194)	126:14 (0.100)
Asn* 155Asn* (AAC vs. AAT)	1	256:36 (0.123)	29:7 (0.194)	126:14 (0.100)
IVS2 − 68 T vs. C	IVS2	288:4 (0.014)	36:0 (0.000)	ND
Asn* 216Ser* (AAT vs. AGT)	3	291:1	35:1	ND
IVS3− 4 insertion T	IVS3	291:1	35:1	ND
Valo277Alao (GTG vs. GCG)	4	291:1	36:0	ND
IVS4− 63 (A vs. G)	IVS4	271:21 (0.072)	31:5 (0.139)	ND
Proo293Leuo (CCG vs. CTG)	5	291:1	35:1	ND
IVS6+ 11 C vs. T	IVS6	290:2	36:0	184:0
Alao362Alao (GCG vs. GCA)	8	291:1	36:0	ND
Valo391Meto (GTG vs. ATG)	9	291:1	36:0	ND
IVS10− 34 A vs. G	IVS10	285:7 (0.024)	35:1 (0.028)	ND
IVS10− 26 insertion T	IVS10	291:1	36:0	ND
Pheo597Pheo (TTC vs. TTT)	14	279:13 (0.045)	31:5 (0.139)	ND
IVS15− 21 C vs. G	IVS15	288:4 (0.014)	35:1 (0.028)	ND
IVS18+ 21 A vs. C	IVS18	281:11 (0.038)	31:5 (0.139)	184:6 (0.032)
Gly* 746Gly* (GGT vs. GGG)	19	291:1	36:0	ND
Pheo779Pheo (TTT vs. TTC)	20	278:14 (0.047)	31:5 (0.139)	ND
Asp−800Asp− (GAC vs. GAT)	21	281:11 (0.037)	31:5 (0.139)	ND
Glu−808Glu− (GAG vs. GAA)	21	287:5 (0.017)	36:0 (0.000)	ND
Lys+827Gln* (AAG vs. CAG)	21	291:1	36:0	ND
Gly* 850Valo (GGT vs. GTT)	22	292:0	35:1	ND

 +, Positively charged amino acid residue; −, negatively charged amino acid residue; o, nonpolar amino acid residue;

*  , polar amino acid residue; ND, not done; IVS, intron. For the variants listed in the left column, the three columns on the right provide the number of alleles with the more common vs. the less common sequence found among 146 patients with recessive RP, 18 patients with isolate RP, and 70–95 normal controls. Minor allele frequencies are also provided in parentheses for polymorphisms (i.e., those variations with a minor allele frequency greater than 0.01.)

Figure 3.

 

Location of pathogenic mutations in the genes encoding the α and β subunits of rod cGMP-phosphodiesterase causing retinal degeneration in humans, mice, and dogs, including the mutations reported in this article. The cGMP-binding domains, catalytic domain, and the isoprenylation motif are shaded in the bar representing the protein. ^{40 47} The protein segment encoded by each exon is indicated with clear boxes in the bottom bar. The mutations in the PDE6A gene are from this article (Arg102His, Arg102Ser, IVS6+1G→A, Gln569Lys, and Ser573Pro), from Huang et al. ⁹ (Ser344Arg, Trp561Ter, Tyr583Ter), and from Meins et al. ³ (Thr706([1-bp del]); only those mutations that were found in affected patients who were homozygotes or compound heterozygotes are shown. The Asn616(1-bp del) mutation in the canine homologue of PDE6A was reported by Petersen–Jones et al. ³⁶ PDE6B mutations found in patients with RP who were either homozygotes or compound heterozygotes are Leu80(71-bp ins), ²⁸ IVS2−1G→T, ^{29 32} Cys270Ter, ⁷ Gln298Ter, ^{6 8 27} Pro496(1-bp del), ^{6 27} Leu527Pro, ^{6 8} Arg531Ter, ^{6 27} Ile535Asn, ³³ Arg552Gln, ³¹ His557Tyr, ^{6 27} Gly576Asp, ²⁹ His620(1-bp del), ²⁹ Leu699Arg, ³⁰ Lys706Ter, ^{6 29} IVS18+1G→A, ⁶ and Leu854Val. ⁸ The missense change His258Asn in PDE6B was found in some Danish families with dominantly inherited congenital stationary night blindness (CSNB). ^{8 34 35} The murine Tyr347Ter mutation causes retinal degeneration in rd mice, ²¹ and the Trp807Ter mutation causes retinal degeneration in Irish setter dogs. ^{24 25 26}

The authors thank Terri McGee, Carol Weigel–DiFranco, and Peggy Rodriguez for helpful comments and assistance.