Proof-of-principle rapid noninvasive prenatal diagnosis of autosomal recessive founder mutations

Study description.

Eight pregnant AJ couples, of which one or both partners were heteroallelic carriers of GBA N370S, were enrolled in the study (Figure 1). Although families 1 and 4 were at risk of giving birth to a homozygote N370S child (unlike families 2, 3, 5, 6, 7, and 8 in which one parent of the fetus did not carry any mutation in GBA), we tested all couples strictly for proof-of-principle purposes. Plasma samples were collected from female participants at the time points indicated in Figure 1 for DNA extraction and targeted high-throughput sequencing of GBA-flanking SNPs. To enhance diagnostic accuracy, we elected to defer direct mutation sequencing in favor of a more specific and sensitive linkage-based analytical regimen. This methodology strengthens diagnostic confidence with increasing fetal haplotype size (measured by the number of SNPs in the inferred fetal haplotype) (Supplemental Tables 1–3; supplemental material available online with this article; doi:10.1172/JCI79322DS1). To accomplish this goal, we first sequenced GBA-flanking SNPs (up to ±250 kb distance from GBA) of the parents and their first-degree relatives in families 1 and 2, so as to construct parental haplotypes. However, these family-based haplotypes were of limited size (Supplemental Table 4). Therefore, we sought a larger haplotype sequence to aid fetal diagnosis by mapping a consensus N370S founder region surrounding the GBA gene.

Fine mapping of the consensus AJ N370S founder haplotype region.

To fine map the N370S founder region, we sequenced 7 unrelated homoallelic AJ mutation carriers on our targeted GBA-flanking SNP panel. Six of these homoallelic patients with type I Gaucher disease were homozygotic for all 490 SNPs on our initial sequencing panel. The seventh sample shared the same haplotype within and 3′ to GBA, but a heterozygous region was clearly identified 144,388 nucleotides 5′ to the gene and beyond (at rs2306124, dbSNP 138). Hence, this sample was used to demarcate a preliminary consensus founder haplotype (Figure 2A). To further clarify the N370S sequence, we then crossed the preliminary version with linkage-based N370S haplotypes from the families under investigation in this study in addition to 3 other unrelated heteroallelic AJ N370S mutation carrier duos (2 first-degree relatives, each of whom carries the same mutation). This analysis identified a recombined region only 17,858 nucleotides upstream of GBA (at rs148168407, dbSNP 138) (Figure 2B), but remarkably, not a single recombination event was identified in the entire 219-kb region downstream of GBA in any of the 20 N370S chromosomes analyzed. Not coincidentally, this 3′ conserved region has been previously characterized as a nonrecombination hot spot by the HapMap Consortium (18). Moreover, previous studies have identified a conserved AJ N370S founder haplotype that extends even further downstream of GBA (12, 19). Nevertheless, although it is likely that the founder haplotype is longer than we initially mapped, our SNP-sequencing panel still successfully linked a sizable amount of GBA-flanking SNPs (153 altogether) to a consensus AJ haplotype sequence (Figure 2C and Supplemental Table 5). The next question was to determine whether this population-based haplotype could be used as a diagnostic tool for NIPD.

Preliminary NIPD of an autosomal recessive founder mutation.

For pilot testing, families 1 and 2 offered 3 different avenues with which to assess the utility of the consensus N370S haplotype. This was because both parents (of the fetus) in family 1 were N370S carriers in addition to the mother in family 2 (Figure 1). For family 1, the N370S carrier father was completely homozygous for the entire consensus N370S sequence. This precluded the use of the N370S haplotype for NIPD of his allele. Nonetheless, the familial mutation-linked haplotype of the father in family 1 did facilitate the identification of his WT allele in the fetus (Figure 3A and Supplemental Tables 6 and 7). Regarding the maternal alleles in families 1 and 2, the consensus N370S haplotype proved to be quite valuable. The family-based maternal N370S-linked haplotype was clearly identified in family 1 plasma DNA. This fetal haplotype was completely concordant with the consensus N370S haplotype (Figure 3B and Supplemental Tables 7 and 8). For family 2, the family-based maternal N370S haplotype could not reliably discern which allele was transmitted to the fetus because the fetal haplotype was determined on differing SNP positions. On the other hand, the longer consensus N370S haplotype clearly matched the inferred maternal haplotype in the fetus, indicating inheritance of the N370S allele (Figure 3C and Supplemental Tables 7 and 9). Thus, in this case, the consensus N370S sequence was crucial to the diagnosis of the maternal allele in the family 2 fetus.

Extended fine mapping of the consensus AJ N370S founder haplotype region.

Although initial testing of families 1 and 2 showed promising results regarding the utility of the consensus N370S haplotype for incorporation into NIPD, it was clear that for expanded N370S testing in a clinical setting a more sophisticated sequencing panel would be required to facilitate setup of a universal assay for noninvasive prenatal Gaucher disease testing. Our concerns with the initial 490-SNP sequencing panel were 4-fold. As evidenced by HapMap (18) and deCode (20) data, meiotic recombination is quite infrequent in the immediate human GBA-flanking locus (±250 kb), which was the small target of our pilot sequencing panel. In this genomic context, homozygosity of an N370S mutation carrier parent, such as in the family 1 father, would be expected to occur commonly because DNA is rearranged at a reduced rate in the peri-GBA locus. Along these lines, low recombination rates translate into low genotypic complexity, which, in turn, leads to limited availability of linkage-informative SNPs, which are crucial to fetal haplotyping. Thus, small family-based haplotypes, which generally handicap fetal haplotyping, such as that of the family 1 father (3 SNPs) and that of the family 2 mother (11 SNPs; Supplemental Table 4), would be predicted to represent the majority as opposed to the minority of cases. Another reason to consider looking beyond a distance of 250 kb from GBA would be to complete fine mapping of the 3′ boundary of the consensus N370S sequence, which proved so beneficial for fetal typing of family 1 and 2 maternal N370S-paired alleles. Finally, N370S aside, the implementation of a larger targeted sequencing panel should hypothetically be used to diagnose any mutation in GBA via familial linkage analysis, regardless of whether the mutation is a founder allele or not. For all these aforementioned reasons, we designed a newer and much improved targeted deep-sequencing panel to sequence 10 times the amount of GBA-flanking SNPs (~5,000 SNPs) across an 8-fold-sized genomic region (GBA ± 2 Mb) before moving forward with NIPD for other families in the study.

The first priority, in terms of test implementation, was to use the new expanded sequencing panel to complete fine mapping of the founder N370S haplotype. As mentioned above, the original sequencing panel successfully demarcated a 5′ boundary for the consensus sequence that was approximately 17 kb upstream of GBA and at least 219 kb downstream. When repeating the same exercise (as that described in Figure 2) using the large sequencing panel, the 5′ boundary for the consensus haplotype mapped approximately 28 kb upstream of GBA (at SNP rs914615, dbSNP141). This 11 kb discrepancy between fine-map boundaries is quite remarkable, given that, due to technical reasons, the newer panel did not incorporate many of the SNPs sequenced previously with the older panel. Nevertheless, the preliminary 5′ cutoff, based only on N370S homozygotes, strikingly mapped to the exact same SNP position (SNP rs2306124) 144,388 nucleotides 5′ to GBA in both panels. Thus, given this concordance between old and new sequencing panels regarding the 5′ N370S consensus sequence, it was especially edifying that the 3′ boundary of the haplotype fine mapped to a position that is roughly 650 kb downstream of GBA (at SNP rs1055184, dbSNP 141) by the new and improved panel. This expanded N370S-linked 670-kb-sized sequence (composed of 301 SNPs; Supplemental Table 10) would already seem quite large for what is considered to be an ancient founder allele. Yet, remarkably, previous studies have shown that the N370S haplotype should, in fact, extend much further downstream from GBA, up to a full Mb from the gene (12, 19). Indeed, after careful scrutiny of the new sequencing data, we found that, among 16 sequenced N370S chromosomes from 8 N370S homozygotes, 15 chromosomes shared a near-consensus haplotype that extended 1.1 Mb 3′ to GBA (Figure 4A). Therefore, we postulated that, if a 250-kb consensus sequence could be used to haplotype fetal alleles in families 1 and 2 (as in Figure 3), then a 1.1-Mb sequence might prove even more useful for typing of N370S chromosomes in most mutation carrier families in general.

To make effective use of the expanded near-consensus N370S haplotype without allowing haplotype errors to corrupt downstream fetal analysis, we carefully inspected each N370S chromosome in all mutation carrier parents in the study (families 1 through 8) using the large sequencing panel. We found that, in some cases, we could detect recombination in the true parent-specific N370S-linked sequence with respect to the founder mutation near-consensus haplotype (Figure 4B). These discrepancies were applied on an allele-specific basis toward refinement of the consensus sequence, so that it could be used for the analysis of fetal haplotypes, which were not phased by conventional family-based linkage analysis (Figure 4C). In such scenarios, the parent-specific near-consensus N370S haplotype was appropriated for the resolution of unphased fetal haplotypes to increase confidence in the final NIPD test result (Figure 4D). For example, if an N370S carrier mother and her immediate family members (whose samples were used for standard linkage analysis) were all heterozygous for the same SNP loci, the mother’s genotypes were considered informative, even though linkage could not set phase on her N370S-linked haplotype. In this case, it was possible that the mother’s unphased SNPs were located within the fine-mapped mother-specific consensus N370S haplotype (as in Figure 4B) and genotyped in her fetus (as in Figure 4C). When this occurs, the correct haplotype can be identified in the fetus, even though conventional family-based linkage analysis fails. More examples of this new approach to NIPD will be illustrated below.

NIPD of GBA N370S using an improved targeted sequencing panel.

Having setup the framework with which to embark on streamlined NIPD for the N370S founder mutation, we returned to families 1 and 2 and retested the same samples using the expanded sequencing panel. One of the primary issues with the previous analysis involving these families was the small size of linkage-based haplotypes in the family 1 paternal N370S allele and the family 2 maternal N370S allele (Supplemental Table 4). As expected, the large sequencing panel clearly solved this issue for families 1 and 2 (and, essentially, all families in this study). Ranging from 113 to 336 phased SNPs, all parental family-based haplotypes in the current investigation were of substantial size and content to enable scoring of fetal haplotypes with generally high confidence (Supplemental Table 11). Interestingly, the family 1 father turned out to be homozygous for the entire N370S consensus and near-consensus sequence by the new panel analysis. Nonetheless, his linkage-based N370S haplotype facilitated highly unambiguous identification of his WT allele in the fetus (Figure 5A and Supplemental Table 12). This test result was clearly of much higher quality, in terms of fetal haplotype size (14 phased SNPs), in comparison with the previous test (2 phased SNPs) involving the same samples (Figure 3A). Regarding the family 1 maternal allele, there was little doubt from the previous panel whether the mother had transmitted her N370S allele to the fetus (9 phased SNPs in fetus; Figure 3B and Supplemental Table 8). In the newer panel, it was even more obvious that the mother had transmitted her N370S allele to the fetus based on clear matches between the fetal haplotype and the mother’s family-based N370S allele as well as her consensus and near-consensus N370S sequences (17 phased SNPs altogether; Figure 5B and Supplemental Table 13).

With family 2, the value and importance of the consensus N370S haplotype grew manifold after reanalysis on the larger sequencing panel. In the previous assessment, only 1 of 3 SNPs in the fetal haplotype was phased to the family-based N370S allele (Figure 3C). In the newer evaluation, the fetal haplotype was much larger, but only 5 SNPs were phased to the family-based maternal N370S haplotype, one of which was isolated on the 3′ side of GBA (1 Mb distance from the mutation). To strengthen the certainty of this test result, the fetal haplotype was compared to the parent-specific consensus and near-consensus N370S haplotype. This comparison yielded another 5 phased SNPs located 3′ to the mutation, which, together with family-based fetal alleles, led to the correct diagnosis of the N370S mutation in the family 2 fetus (based on 10 phased SNPs altogether; Figure 5C and Supplemental Table 14).

The principles set forth in these preliminary tests were subsequently put into practice for fetal allele identification involving families 3 through 8 (Figure 5, D–J, and Supplemental Tables 15–21). Of particular importance is the fact that 4 of 6 N370S-paired alleles in these families were typed noninvasively with the aid of the N370S consensus and/or near-consensus sequence (Figure 5, F–H, and J). These results thereby confirmed our assumption that the founder N370S haplotype is a valuable tool for incorporation into standard NIPD protocol. Another point to consider, which supports the use of larger sequencing panels in NIPD in general, is the fact that, even when a nonfounder GBA mutation was tested (such as that of the family 4 paternal allele; Figure 5E and Supplemental Table 16), the extended sequencing panel facilitated construction of a well-defined fetal haplotype nonetheless.

To summarize, the outcomes of this proof-of-concept study are presented in Table 1. All noninvasive test results were validated with conventional prenatal or postnatal diagnostics.

Sample collection.

Pregnant AJ couples, carrying mutation(s) in the GBA gene, were recruited at the SZMC Gaucher Clinic. Peripheral blood samples were collected from each couple, relevant mutation carrier family members, 8 unrelated AJ GBA N370S homozygotes, and 3 unrelated AJ GBA N370S heterozygote duos.

Targeted NGS.

Genomic DNA was prepared from blood samples, and plasma DNA was isolated from each pregnant female participant followed by targeted NGS of 490 SNPs and/or approximately 5,000 GBA-flanking SNPs on an Illumina MiSeq or NextSeq 500 instrument, respectively. Family-specific and N370S-specific haplotypes were constructed from NGS data and used to identify fetal alleles in each plasma-derived cell-free fetal DNA sample. All fetal diagnoses were validated with conventional prenatal or postnatal diagnostics.

Statistics.

Accuracy of identified fetal haplotypes was estimated by modeling assay-specific sensitivity/specificity as a function of inferred fetal haplotype size and genomic site–specific meiotic recombination rates (based on ref. 20 and Supplemental Tables 1–3).

Study approval.

Ethical approval for the study, including usage of materials from human subjects, was obtained from the SZMC institutional review board, and written informed consent was obtained from all study participants.