SMN dosage analysis and risk assessment for spinal muscular atrophy.

To the Editor: Feldkotter et al. (2002) recently reported a new method to determine, on the basis of real-time, quantitative PCR, copy numbers of SMN1 (MIM 600354) and SMN2 (MIM 601627). Their method allows a greater degree of automation and a faster turnaround time than do methods that have been described elsewhere (McAndrew et al. 1997; Chen et al. 1999; Wirth et al. 1999; Gerard et al. 2000; Scheffer et al. 2000; Ogino et al. 2001). Using their new method, they demonstrated that the copy number of SMN2—which is the centromeric homologue of SMN1, the disease gene for spinal muscular atrophy (SMA [MIM 253300 for type I; MIM 253550 for type II; and MIM 253400 for type III])—influences the severity of SMA in affected individuals with homozygous deletions of SMN1. They found that, the greater the copy number of SMN2 was, the greater the likelihood was of a milder SMA type. Because this correlation is not absolute, they used Bayesian-type analyses to determine the posterior probabilities of developing each SMA type, with both a homozygous deletion of SMN1 and a given copy number of SMN2. We discuss below several important ethical, prognostic, and technical issues raised in their article. In table 6, Feldkotter et al. report “Probabilities That an Unaffected Who Has Been Tested after Birth and Has Been Found to Carry a Homozygous Absence of SMN1 Will Develop Type I, II, or III SMA, on the Basis of Number of SMN2 Copies.” SMA is usually a childhood-onset disease, and testing of unaffected children is ethically problematic. We agree with the American Society of Human Genetics and the American College of Medical Genetics that “Timely medical benefit to the child should be the primary justification for genetic testing in children and adolescents” (American Society of Human Genetics Board of Directors and American College of Medical Genetics Board of Directors 1995, p. 1233). Since there are currently no effective treatments, presymptomatic or otherwise, for SMA, the timely medical benefit of the testing of unaffected children is unclear. For the purpose of predicting SMA type from the SMN2 copy number in unaffected children who lack SMN1, Feldkotter et al. perform Bayesian-type analyses by use of odds ratios, rather than conventional conditional probabilities. For the prior probabilities, they use the distribution of types of SMA among individuals affected with SMA: .51, for type I; .32, for type II; and .17, for type III. Even if one were to test unaffected children in this way, for this purpose, these prior probabilities would not be the correct ones to use for Bayesian or Bayesian-type analyses. If a child is asymptomatic at age 10 mo, for example, he or she is much less likely to have type I SMA than to have one of the other types (Zerres and Rudnik-Schoneborn 1995). One would have to incorporate the conditional probabilities of being asymptomatic at a particular age, for the hypothesis of each SMA type. The data on SMN2 copy number given by Feldkotter et al. could be used in prenatal testing, to predict SMA type. However, the prior probabilities that they use would be applicable only if the family history of SMA is of an unknown type. Although families with more than one type of SMA have been described—and are far from rare—knowing the type of SMA in an affected family member increases the prior probability of that type of SMA in a relative who is at risk of developing SMA. If the type of SMA in that affected family member is unknown, then the distribution of SMA types among all individuals with SMA would be relevant to the assignment of prior probabilities. On the basis of all reported data, Feldkotter et al. state that, because two SMN1 copies were found on 20/834 (2.4%) healthy chromosomes, “4.8% of normal individuals would be misinterpreted as noncarriers on the basis of the direct SMN1 test” (p. 365). Actually, these data imply that ∼4.8% of noncarriers would have three copies of SMN1 and that ∼2.4% of carriers with an SMN1 deletion on one chromosome 5 would have two SMN1 copies on the other chromosome 5. We have referred to the latter as the “2+0” genotype (Chen et al. 1999). Taking into account the ∼1.7% of carriers who have an intragenic mutation undetectable as an SMN1 exon 7 deletion, Feldkotter et al. state that this “reduces the sensitivity of the test to 93.5% for a person from the general population” (p. 365). Combining the ∼1.7% of carriers who have an intragenic mutation with the ∼2.4% (i.e., 0.024×[1-0.017]) of carriers who have the 2+0 genotype gives the overall sensitivity of SMN dosage analysis for the detection of SMA carriers in the general population as ∼95.9%. If an affected family member were known to have a homozygous deletion of SMN1, then the sensitivity of SMN dosage analysis for the detection of carriers among unaffected family members would be ∼97.6% (i.e., 0.959/[1-0.017]). This is because the probability of an intragenic-mutation carrier in this family is greatly decreased relative to the probability of a 2+0 carrier (Ogino et al., in press). Updating our combined data (McAndrew et al. 1997; Ogino et al. 2002, in press) gives 23 of 590 normal chromosomes 5 that have two copies of SMN1. Combining these data with those of Feldkotter et al. gives a total of 37 of 1,120 (3.3%) normal chromosomes 5 that have two copies of SMN1. We excluded other data in the literature (Wirth et al. 1999; Gerard et al. 2000; Scheffer et al. 2000), for reasons described elsewhere (Ogino et al., in press). On the basis of these numbers, ∼3.2% (i.e., 0.033×[1–0.017]) of carriers would have the 2+0 genotype. Therefore, the sensitivity of SMN dosage analysis for the detection of carriers in the general population would be ∼95.1%, and that for the detection of carriers in a family with an affected individual lacking SMN1 would be ∼96.7% (i.e., 0.951/[1-0.017]). Taking advantage of the single nucleotide differences between SMN1 and SMN2 in both exon 7 and intron 7, Feldkotter et al. used gene-specific primer pairs to amplify only SMN1 or only SMN2. The primer pairs for each gene were mismatched for the other gene at either the final or the penultimate nucleotide from the 3′ end. These mismatches corresponded to the sequence differences in exon 7 (forward primers) and intron 7 (reverse primers). Gene conversions between SMN1 and SMN2, which have been reviewed elsewhere (Burghes 1997), could potentially complicate this approach. If the SMN1 exon 7 sequence (C) were converted to the SMN2 exon 7 sequence (T) but the SMN1 intron 7 sequence remained the same, the converted gene would presumably function as an SMN2 gene in vivo. This is because the C→T transition in exon 7 of SMN2, although translationally silent, decreases the activity of an exonic splicing enhancer, so that less full-length protein is expressed (Lorson et al. 1999; Monani et al. 1999; Jong et al. 2000). By use of the gene-specific primers given by Feldkotter et al., the converted gene might have a different amplification efficiency from that of the normal SMN1 or SMN2 gene. Primers that are allele specific only for the functionally important polymorphism in exon 7 but not for the polymorphism in intron 7 might alleviate this problem.