ClinGen guidance for use of the PP1/BS4 co-segregation and PP4 phenotype specificity criteria for sequence variant pathogenicity classification

Abstract

The 2015 American College of Medical Genetics and Genomics and the Association for Molecular Pathology variant classification publication established a standard employed internationally to guide laboratories in variant assessment. Those recommendations included both pathogenic (PP1) and benign (BS4) criteria for evaluating the inheritance patterns of variants, but details of how to apply those criteria at appropriate evidence levels were sparse. Several publications have since attempted to provide additional guidance, but anecdotally, this issue is still challenging. Additionally, it is not clear that those prior efforts fully distinguished disease-gene identification considerations from variant pathogenicity considerations nor did they address autosomal-recessive and X-linked inheritance. Here, we have taken a mixed inductive and deductive approach to this problem using real diseases as examples. We have developed a practical heuristic for genetic co-segregation evidence and have also determined that the specific phenotype criterion (PP4) is inseparably coupled to the co-segregation criterion. We have also determined that negative evidence at one locus constitutes positive evidence for other loci for disorders with locus heterogeneity. Finally, we provide a points-based system for evaluating phenotype and co-segregation as evidence types to support or refute a locus and show how that can be integrated into the Bayesian framework now used for variant classification and consistent with the 2015 guidelines.

Figure 1

Flow diagram showing steps of evidence assessment for the phenotype-specificity (PP4) and co-segregation (PP1/BS4) criteria

Figure 2

Cystinosis as an example for locus homogeneity and autosomal-recessive inheritance (A) A simplex family affected with a child with the cystinosis phenotype and molecular genetic testing results for the CTNS gene. The affected offspring has typical cystinosis, which is associated with a nearly 96% yield of detecting two causative variants in trans using common sequencing methods that target exons and flanking introns. There are no other genes known to be associated with this phenotype. In this family, CTNS was sequenced, and two novel missense variants were found in the proband and one of them in each parent. We estimate that there is at least a 95% likelihood that those are the causative variant—other possible mechanisms are implausible. Back-calculating the Bayesian points from the posterior probability of 95% in Table 2 shows that each allele would warrant +7.0 points. However, we limited this evidence type to +5.0 points (see text) for each allele, and because there was only one variant on each allele, each of these variants accrues the full +5.0 points. Were there to be additional affected children born to these parents, co-segregation pathogenicity evidence (PP1) would not be garnered because co-segregation cannot be considered evidence in traits with locus homogeneity. (B) A second simplex family with the same scenario as in Figure 2A except that the paternal allele harbors two missense variants. In this case, the data suggest that the two paternally inherited variants in cis are equally likely to be the causative variant on that allele. Therefore, each of them has a 48% likelihood of being pathogenic. Referring to Table 2, that leads to assigning +2.5 points to each variant on that allele. The missense variant on the maternal allele would garner +5.0 points, as in Figure 2A. One should not divide the points arithmetically—that would be mathematically incorrect. One needs to refer to Table 2 each time, as further examples will show.

Figure 3

Cockayne syndrome as an example of locus heterogeneity and autosomal-recessive inheritance (A) In this hypothetical example, a family with two children with the typical Cockayne syndrome phenotype have been sequenced for the more commonly mutated of the two genes associated with this trait, ERCC6. Both children are homozygous for a variant that is heterozygous in both parents. The novel variant is assigned +4.0 points, because the prior probability is 70%, given that 70% of patients with typical Cockayne syndrome have causative variants in ERCC6, and from Table 2, that corresponds to +4.0 points. However, because the evidence ceiling is +5.0 points, there is an affected sibling, and this trait exhibits locus heterogeneity, co-segregation evidence can be considered. Phase is established by individual II-1 and there is one co-segregating homozygous individual. Referring to Table 3, for a disorder with autosomal-recessive inheritance and homozygous genotype, each meiosis for an affected individual garners +2.0 points for the allele. These +2.0 points are then added to the +4.0 points, yielding +6.0 total points, which is reduced to +5.0 points because of the evidence cap. (B) In this hypothetical example, a family with two children with the typical Cockayne syndrome phenotype have been sequenced for the more common of the two genes associated with this phenotype, ERCC6. Both children are compound heterozygotes for novel variants that are heterozygous in one of the parents. Both variants are assigned +4.0 points, because the prior probability is 70%, given that 70% of patients with typical Cockayne syndrome have causative variants in ERCC6, and from Table 2, that corresponds to +4.0 points. However, because the evidence ceiling is +5.0 points, there is a co-segregation in a sibling, and this trait exhibits locus heterogeneity, co-segregation evidence can be considered. Phase is established by individual II-1. Referring to Table 3, for a disorder with autosomal-recessive inheritance and compound heterozygosity, each co-segregating affected individual garners +2.0 point for each allele, which is added to the +4.0 points (PP4), summing to +6.0, which is reduced to +5.0 because of the cap. Note that less evidence is garnered here than in the example in Figure 3A, but the evidence was capped in that example. (C) In another example using Cockayne syndrome, in this case the affected proband is compound heterozygous for a variant present in one each of his parents, and his unaffected sibling did not inherit the variants. As in Figure 3A, the PP4 evidence was +4.0 points. However, the co-segregation evidence is quite different than in the Figure 3A example because unaffected individuals provide less evidence of co-segregation than do affected individuals in autosomal-recessive inheritance. Referring to the second row of Table 3, one can see that one unaffected individual (who has an affected family member who is compound heterozygous) with consistent co-segregation garners +0.4 points for each allele, which is then rounded down to 0.0 points, which does not contribute to the evidence for pathogenicity of this variant. Were the family to have five unaffected children instead of one, that would garner +2.0 point (+0.4 × 5) for each variant. (D) A family with one child affected with the typical Cockayne syndrome phenotype and testing results for the only two genes associated with this disease, namely ERCC6 and ERCC8. Note that there is an observation of non-segregation with ERCC8 in the unaffected sibling—she has the same genotype as the affected. This essentially excludes ERCC8 and the associated variant from a causative role for this disorder in this family. In this case, the problem effectively transforms into a locus homogeneity problem with near 100% mutation sensitivity. In such a case, one would award +9.0 PP4 points plus +0.4 co-segregation points (for the unaffected), which is rounded down to +9.0 and then capped at +5.0 points.

Figure 4

Examples of autosomal-dominant inheritance; Marfan syndrome for locus homogeneity and tuberous sclerosis for locus heterogeneity (A) An example of the use of a single meiosis to set phase in a disorder with autosomal-dominant inheritance. Using the approach described here, because the yield is 91%, the likelihood for each variant is 45.5%, which would correspond to +2.5 PP4 phenotype-specificity points (Table 2) would be assigned to each variant in the child had the parents not been tested. However, one can set phase by observing that the p.Gly940Val variant is co-segregating with the phenotype. Thus, the only remaining candidate variant in the child is p.Gly940Val, and therefore the full 91% yield now pertains, which would correspond to +6.0 points (Table 2) but is capped at +5.0 PP4 phenotype-specificity points, which can be assigned to that variant. (B) An example of autosomal-dominant inheritance of tuberous sclerosis (TS). In this family of three affected individuals, a single variant has been identified in both TSC1 and TSC2. Because 26% of TS is associated with TSC1 and 69% with TSC2, the p.Leu19Gln TSC1 variant is assigned +1.5 points (rounding down to 25% in Table 2), and the p.Lys34Thr TSC2 variant is assigned +4.0 points (rounding down to 65% in Table 2) for phenotype-specificity points (PP4). The TSC1 variant can be phased using individuals I-1 and II-1. The TSC1 variant also garners +2.0 points for the co-segregations (II-2 and II-3) for a total of +3.5 points. However, the TSC2 variant shows non-segregations in both II-2 and II-3, effectively excluding that variant (and that gene) from being causative in this family. Therefore, the PP4 points need to be reevaluated. Now, the expected yield for TSC1 would be more than 95%, and thus +7.0 points (points from 95% on Table 2) would be awarded to the TSC1 variant, which would also garner +2.0 additional points for the co-segregations but be capped at +5.0 points irrespective of those segregations.

Figure 5

Example of autosomal-dominant inheritance with low diagnostic yield—dilated cardiomyopathy (DCM) For this disease, diagnostic yield for MYH7 does not reach 20%, so PP4 cannot be applied. One can eliminate the need to set phase by observing no variation on the second allele. (A) The only candidate variant in the proband (II-1) is p.Arg1500Trp, and +1.0 point from Table 3 can be awarded for the single co-segregation from the affected father (I-1). (B) In this family there are at least two co-segregations (II-2 [obligate] and III-1), which garners +2.0 points from Table 3. One could also justify counting individual I-2 given that a positive genotype can be inferred from the genotype-negative I-1 individual, adding another point for a total of +3.0 points. The absence of phenotype in individual III-2 who harbors the variant is not considered given reduced penetrance and age-related onset of DCM.

Figure 6

An example of co-segregation in a three-generation family with X-linked-recessive inheritance The affected males have typical TARP syndrome (talipes equinovarus, atrial septal defect, Robin sequence [micrognathia, cleft palate, and glossoptosis]) and persistent left superior vena cava, (MIM: 311900), which is only known to be associated with the RBM10 gene. The variant identified in this family is p.Cys219Phe. The diagnostic yield is not known for this phenotype because it is so rare, so PP4 points are not awarded. Phase is established in an affected male in disorders with X-linked inheritance without the need for observation of any meioses. In this family there are four co-segregations (II-1, II-2, III-1, and III-2), which garners +4.0 points from Table 3. Note that individual II-2 garners +1.0 point because her status as a carrier is affirmed by having an affected brother and son.