Yatsenko AN, Shroyer NF, Lewis RA, Lupski JR 2001 Late-onset Stargardt disease is associated with missense mutations that map outside known functional regions of ABCR (ABCA4). Hum Genet 108:346^355

DISCUSSION

LaVail: I have a question regarding the CRB1 localization. You mentioned that some people have found antibody artefacts in immunocytochemistry, and I’d agree that this is common. Do you see any other expression or localization there? If that is an artefact, where is the CRB1 actually localized?

Cremers: The artefact might be on the cone plasma membrane. We believe that CRB1 is expressed at the inner limiting membrane functioning as an anchor for cytoplasmic proteins and connecting photoreceptor cells with Mˇller cells. The cone has an envelope which contains among other proteins proteoglycans and similar molecules. This might be related to the non-speci¢c reactions that other people are seeing with some antibodies to the same structure.

Gal: I have a more general question. I have the feeling that nowadays it is ‘in’ to call any RP gene an LCA gene. Nonetheless, by so doing we dilute this clinical diagnosis for, in the end, we include all recessive RP cases with early onset. Notably, in the papers ¢rst describing the two large RP12-famililes (van Soest et al 1994 for the Dutch family and Leutelt et al 1995 for the Pakistani family), the patients’ phenotype was called autosomal recessive RP with para-arteriolar preservation of the RPE (PPRPE), and the diagnosis LCA was not considered for either of them. In the patients you studied, several had the unique phenotype of PPRPE. This is a hard criterion. On the other hand, you had some soft criteria, such as the early onset and severe course of the disease. Would it not be more sensible to classify these patients’ disease according to the clearly unique phenotype (arRP with PPRPE) instead of the ill-de¢ned age of onset and severity (LCA)?

Cremers: When we started a new project on genotype^phenotype analysis on LCA we looked at what had been published on these criteria. We have established a protocol to delineate the di¡erent groups. We took some of the classical criteria and grouped our patients. Our protocol is not fully accurate yet, but it would be a good idea to develop a widely accepted classi¢cation. In my own patient database I have a lot of patients that I can not classify well enough. In response to the comment of Dr Gal, patients with ARRP and PPRPE are not positioned in the LCA group, but some patients with LCA, based on the fact that they had no vision at birth or lost vision before the age of 1 year, did show PPRPE.

Aguirre: You mentioned about limiting light exposure in these patients. Do you write letters to them recommending this, or to their physicians?

Cremers: Currently the patients themselves get general information on our molecular ¢ndings. As yet, we do not yet advise them speci¢cally to avoid excessive light, but this might be important.

Aguirre: Many patients are now being advised for evaluation not to have the standard documentation with the six-¢eld fundus photographs, but instead to have eight overlapping ¢elds using the wide ¢eld camera. I have heard comments that this is more uncomfortable because the light is brighter, but they get a wider ¢eld of view. From what you are saying, this is a concern, because the light is much brighter than the light we are exposed to in the environment.

Bok: You showed a cartoon of CRB1 where it was depicted as a single-pass integral membrane protein with a very large extracellular mass. This suggests that it is either binding to the ECM or to itself. Do you have any gene disruption data yet to tell us about the cytoarchitecture of the retina?

Cremers: Studies of this kind are not performed in my group so I can’t give details.

Bok: On the basis of the localization studies you described, presumably you don’t think CRB1 is part of the zonula adherens, but is actually distal to that, in the plasma membrane of the inner segment.

Cremers: Yes, maybe just distal to the zonula adherens.

Bok: Do you predict that there would be some breakdown in the cytoarchitecture caused by a perturbation in cell^cell interactions?

Cremers: Yes. That is also seen in Drosophila CRB1 mutants, where in development, at a certain point the ommatidia elongate at a tremendous rate, with a 908 kink. In mosaic Crumbs mutant £ies, these structures are not completely developed.

Bok: What about the rhabdomeres?

Cremers: It is the stalk region that is not growing well enough to produce this full structure.

Bok: You mentioned that homozygous CRB1 mutation is embryonic lethal in

Drosophila.

Cremers: Yes, because it has a very important function in embryonic epithelia. Normally the cells are closely connected, but if CRB1 is missing their organization is disrupted.

McInnes: You called it a homologue, but from everything you have said it may be more accurate to say that it is the orthologue. Structural di¡erences like the ones you showed are very common between £ies and mammals, especially in the size of the protein.

Cremers: This might be true. However, in some regions CRB2 is more homologous to the Drosophila protein than CRB1 is and actually might be the true orthologue. At this point it is di⁄cult to suggest which of the two is the actual orthologue.

Bolz: You showed some proteins that interact with CRB1. Have you looked at the expression pattern of the corresponding genes, or have you investigated them in patients?

Cremers: We have done some RT-PCR analysis. Many of the putatively interacting proteins show high but not exclusive expression in the retina.

Bolz: Is there any overlap of the loci with a known RP locus? Cremers: Not as far as we know.

Daiger: I am interested in how you know that an amino acid substitution in the ABCA4 gene is in fact pathogenic. Could you tell us about the background variation in this gene in a normal population? If you sequenced this gene in 100 individuals without disease, how many distinctly di¡erent amino acid substitutions would you ¢nd? How many of them are polymorphic in the classic sense that the allele frequency of the less frequent allele is 1% or greater? How many distinctly di¡erent protein haplotypes would be detected in that population of 200 chromosomes? Finally, in relation to this, what is your de¢nition of a mutation in this context?

Cremers: A Japanese group recently analysed the ABCA4 gene in normal individuals (Iida et al 2002). Also, data are known for a German control population which was tested for the presence of known ABCA4 mutations. Pathologic mutations are found in *5% of healthy individuals.

Daiger: How do you de¢ne ‘mutations’ and ‘polymorphisms’ in this context? Cremers: About 50 mutations have been tested in a rather crude invitrofunctional assay by looking at whether the protein is expressed and is transported to the cell membrane, and whether there is ATPase activity. In these tests most of these

mutations have shown de¢ciencies in the protein.

Daiger: Isn’t it true that some of the things that you are calling ‘polymorphisms’ in the laboratory also show activity diminution? What leads you to say whether or not they are pathogenic?

Cremers: The 2588G4C allele I described always is found together with another mutation. Both show a de¢ciency in this arti¢cial test. In humans with Stargardt disease we ¢nd the 2588G4C allele alone in a few cases without the what we think is a polymorphic variant, but not vice versa. From this interpretation it is almost certain that it is the 2588G4C mild allele that is pathological.

Daiger: So you use the term ‘polymorphic’ to describe a variant that is nonpathogenic, independent of its frequency, and you use ‘mutation’ for a variant that is pathogenic, again independent of its frequency.

Cremers: Yes.

McInnes: I don’t think we should reinvent genetic terminology in this way. A polymorphism is any allele found in 1% or more of the population.

Daiger: I am not trying to hone in on the accuracy of nomenclature. Nomenclature in this ¢eld has changed radically in the last ¢ve years in a way that

has horri¢ed population geneticists. But what is important is that we are clear about what we are talking about. How many amino acid substitutions are there in Caucasian populations which by any criteria are pathogenic, and how many are non-pathogenic? This is what has worried me most about the ¢eld of ABCA4: the di⁄culty of de¢ning what is truly a pathogenic amino acid substitution.

Swaroop: I think we should refrain from saying anything is a mutation unless we know it is pathological. In the absence of this evidence of a causative change we use the term ‘sequence variant’.

Cremers: I agree. But what about the IVS38-10T4C polymorphism that I indicated? If this is found in 25 alleles of 500 Stargardt disease patients and is not found in 500 controls, what more evidence do we need that this variant is in linkage disequilibrium with the actual (unknown) pathologic DNA change? To not use this as a diagnostic marker would be unrealistic.

McInnes: We have to remain orthodox in our language. A mutation is any change in a DNA sequence: it has nothing to do with pathogenicity although we all tend to use it that way. The trouble is that the loose terminology will lead to loose thinking.

Bird: As a clinician it is di⁄cult listening to others’ nomenclature. It is generally believed in ophthalmology that Stargardt disease is found in about 1 in 6500 of the population. This means a carrier rate of about one in 40. This would ¢t, in general, with the risk to the child of an a¡ected individual being about 1 in 80, which ¢ts with our experience. But I then ¢nd it di⁄cult when the implication is that there is a disease-causing mutation that is much more common than 2.5% of the population. If the carrier rate is 1 in 40, and a sequence change that is thought to be disease causing is more common than that, I have a hard time understanding it, if you assume one mutation per allele.

Daiger: The cystic ¢brosis allele frequency among Caucasians is 3^5%, and the sickle cell allele frequency is 5^7% among African Americans. It is not unheard of for recessive diseases to have carrier frequencies in the 5^10% range.

Cremers: Most of the mutations are mild alleles, otherwise there would be many more Stargardt cases. What needs to be done with this chip is to test a few thousand of these cases and also a couple of thousand controls to see what the relationship is. This might also resolve the question of whether ABCA4 mutations play a role in AMD.

Bhattacharya: As long as the controls have been tested very thoroughly for any signs of AMD.

Daiger: I can’t resist adding that one of the most di⁄cult statistical problems is comparing a common disease such as AMD with a gene that has common variants in it. With regards to the cost of analysis, I don’t think we disagree much on the numbers. If you limit the study of autosomal dominant RP to ‘common’ mutations, and de¢ne 50 or so of those, you could reduce these to the gene chip

technology. The use of the chip may be only $75, but this doesn’t include preparation of DNA, reports, personnel costs and so on.

Cremers: These costs have already been paid for. The laboratory and sta¡ are already there.

Daiger: When you give a quote of $75 to test 380 variants, are you including overhead and personnel costs in that?

Cremers: Yes, the Estonian company has done this for this sum which is a research price. In most situations we already have the diagnostic set-up. There is a parallel group that does the diagnostics more-or-less independently from my group and they already have the infrastructure for this analysis.

Daiger: My point is that someone had to pay for that in the ¢rst place.

Zack: I have a more general comment in terms of the regulatory issues of CLIA certi¢cation. We have had problems here in Baltimore with university regulations that we can’t tell the patient anything unless the results are coming from a CLIAcerti¢ed lab. This means that if we have a result we can’t tell them. When a patient wants something tested, how do you deal with this?

Daiger: There is a practical answer. In addition to getting your own lab CLIA certi¢ed, which isn’t as di⁄cult as it sounds, there are certain reference laboratories that are CLIA certi¢ed. If you ¢nd a mutation, the patient then sends this lab $500 and a blood sample and they will con¢rm the existence of that mutation. This may not be completely satisfactory, but it is one way of doing this. You can tell the lab where to look for the mutation.

Cremers: This is the situation we are faced with. We have a DNA diagnostic division. They take the blood sample, we obtain half the DNA sample to do a complex mutation analysis and then we give the result back to them. Alternatively, if the blood sample was sent directly to us, another blood sample is drawn and directly sent to the DNA diagnostics facility. They repeat the analysis and send out the result.

Daiger: Ted Dryja, how do you handle patients?

Dryja: I am nervous about this issue because we don’t have a general rule. I feel uncomfortable informing patients of results from our research lab because it is not operated with the quality controls that are required for a clinical testing lab. I estimate that we probably have an error rate of 1%, but many patients will take the results from a research lab and assume they are absolutely correct. What is our legal liability here? In reality, most of the time we don’t inform the patients. They may have given blood 5^10 years ago; we have samples from thousands of patients; it would take an enormous amount of e¡ort to just mail out results to patients each year.

Daiger: We have o⁄cial approval from our Human Subjects Committee to initiate a process to go back to the individuals and inform them that we have speci¢c genotype information on them relevant to their retinal disease. We ask

them whether they want this information released to a third party, who has to be a ‘knowledgeable healthcare professional’. With that permission from the patient, one of our standard forms goes out to the clinician. It is cumbersome, but our human subjects committee has been comfortable with this for the last few years.

References

Iida A, Saito S, Sekine A et al 2002 Catalog of 605 single-nucleotide polymorphisms (SNPs) among 13 genes encoding human ATP-binding cassette transporters: ABCA4, ABCA7, ABCA8, ABCD1, ABCD3, ABCD4, ABCE1, ABCF1, ABCG1, ABCG2, ABCG4, ABCG5, and ABCG8. J Hum Genet 47:285^310

Leutelt J, Oehlmann R, Younus F et al 1995 Autosomal recessive retinitis pigmentosa locus maps on chromosome 1q in a large consanguineous family from Pakistan. Clin Genet 47:122^124

van Soest S, Ingeborgh van den Born L, Gal A et al 1994 Assignment of a gene for autosomal recessive retinitis pigmentosa (RP12) to chromosome 1q31-q32.1 in an inbred and genetically heterogeneous disease population. Genomics 22:499^504