Non-invasive prenatal diagnosis of single gene disorders with enhanced relative haplotype dosage analysis for diagnostic implementation

Abstract

Non-invasive prenatal diagnosis of single-gene disorders (SGD-NIPD) has been widely accepted, but is mostly limited to the exclusion of either paternal or de novo mutations. Indeed, it is still difficult to infer the inheritance of the maternal allele from cell-free DNA (cfDNA) analysis. Based on the study of maternal haplotype imbalance in cfDNA, relative haplotype dosage (RHDO) was developed to address this challenge. Although RHDO has been shown to be reliable, robust control of statistical error and explicit delineation of critical parameters for assessing the quality of the analysis have not been fully addressed. We present here a universal and adaptable enhanced-RHDO (eRHDO) procedure through an automated bioinformatics pipeline with a didactic visualization of the results, aiming to be applied for any SGD-NIPD in routine care. A training cohort of 43 families carrying CFTR, NF1, DMD, or F8 mutations allowed the characterization and optimal setting of several adjustable data variables, such as minimum sequencing depth, type 1 and type 2 statistical errors, as well as the quality assessment of intermediate steps and final results by block score and concordance score. Validation was successfully performed on a test cohort of 56 pregnancies. Finally, computer simulations were used to estimate the effect of fetal-fraction, sequencing depth and number of informative SNPs on the quality of results. Our workflow proved to be robust, as we obtained conclusive and correctly inferred fetal genotypes in 94.9% of cases, with no false-negative or false-positive results. By standardizing data generation and analysis, we fully describe a turnkey protocol for laboratories wishing to offer eRHDO-based non-invasive prenatal diagnosis for single-gene disorders as an alternative to conventional prenatal diagnosis.

Conflict of interest statement

The authors have declared that no competing interests exist.

Copyright: © 2023 Pacault et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Figures

Fig 1. Workflow and processing steps for samples obtained from the training and test cohorts.

The training cohort included couples at risk for SGD who were offered invasive prenatal testing and for whom parental haplotypes were determined using fetal genomic DNA from the current pregnancy. Test cohort included pregnant women at risk for SGD who were offered invasive prenatal testing and for whom a proband’s genomic DNA was available for haplotype reconstruction. The training cohort study allowed to evaluate and optimize the technical settings. The best combination of settings was then used to explore the test cohort. Finally, the diagnostic performance of our SGD-NIPD workflow with eRHDO was evaluated and compared to gold standard as defined by fetal status obtained by invasive sampling. Quality controls measures are described to minimize errors and to aid in technical validation and biological interpretation of the result.

Fig 2. Noninvasive fetal genomic analysis from maternal plasma DNA.

In the test cohort, the proband’s gDNA is used to identify each parental haplotypes. HapI is the maternal at-risk haplotype; HapII is the maternal non-at-risk haplotype, HapIII is the paternal at-risk haplotype and HapIV is the paternal non-at-risk haplotype. A family-specific pile-up was generated from cfDNA sequencing data, by counting the number of reads for each base. Each SNP of the pile-up can be categorized based on parental inheritance. SNP1 (mother AA and father BB) and SNP2 (mother AA and father AA) allow the basic parameters for maternal plasma DNA sequencing to be established, including fetal fraction and sequencing error rate estimations. SNP3 (mother AA and father AB) allow qualitative paternal haplotype transmission determination. SNP4 (mother AB and father AA) allow quantitative maternal haplotype transmission determination, thanks the tracking of fetal inheritance of a haplotype block close to the mutation carried by the mother. SNP5 were not analyzed in this study.

Fig 3

Example of a graphical result for autosomal inheritance in a family at risk of transmitting cystic fibrosis (family 2), subdivided into paternal (A, B) and maternal (C) inheritance. In this example, SNP3A/D/E/H correspond to SNP3 located on the “non-at-risk” paternal haplotype (A), while SNP3B/C/F/G correspond to SNP3 located on the “at-risk” paternal haplotype (B). “Non-at-risk” and “at-risk” haplotypes are shown in green and red, respectively. (a) For each result, the input data are summarized as follows: family ID, tested locus and parental inheritance, gDNA/cfDNA minimum sequencing depth, transmission mode (ARI = autosomal recessive inheritance, ADI = autosomal dominant inheritance, RXI = recessive X-linked inheritance) and number of SNP3. (b) Qualitative detection of fetal-specific SNPs with an allelic frequency (VAF) higher than our threshold visualized in (c) (AU = arbitrary units). SNP3A/D/E/H from the “non-at-risk haplotype” were not detected in cfDNA (top), whereas the majority of SNP3B/C/F/G from the “at risk” haplotype were qualitatively detected above the background threshold (bottom). (d) Fetal genotype from sequencing of fetal gDNA obtained by invasive sampling (e) Plots of inter-SNP distance. Each peak indicates a longer genomic distance between two consecutive tested positions. (f) Location of the parental mutation, symbolized by a vertical dashed line. (g) Input data and quality parameters of the SPRT analysis, namely number of SNP4, mean number of SNPs 4α or 4β per block (na and nb), concordance between conclusive haplotype blocks in forward and reverse orientation (pc) and proportion of nonconclusive haplotype blocks (in gray) among all blocks (nc), and type I and type II statistical errors used in the SPRT test (a/b). (h) and (i) SPRT analysis in forward and reverse orientation, respectively. (j) Visualization of haplotype blocks, divided into 4α (top) and 4β (bottom) analyses, with number of conclusive blocks. (k) Distribution of SNPs 4α (blue) and 4β (purple) along the genomic region.

Fig 4. Examples of results with different quality scores.

(A) Family 2_42 at risk of transmitting Duchenne/Becker muscular dystrophy. A high number of tested SNP4 (n = 1270) combined with a deep sequencing depth (mean = 211X) allowed the detection of a large number of blocks (59 haplotype blocks in forward direction and 59 haplotype blocks in reverse direction) with a high concordance in favor of the transmission of HapI (Sc = 1.00), despite a relatively low fetal fraction (f = 4.8%). Sb was estimated to be 0.89. (B) Family 2_5 at risk of transmitting cystic fibrosis. This family presented with a reasonable number of SNP4 (n = 396) and fetal fraction (f = 7.5%), but only a few haplotype blocks could be reconstructed (n = 39 blocks in forward + reverse directions) with a relatively high number of SNP4 per block (na = 17, nb = 28.1), resulting in a lower block score (Sb = 0.62). Note that the high concordance between forward and reverse analysis (Sc = 0.93), as well as the position of the maternal variant along the CFTR locus (black dashed line), still allow to conclude in favor or the transmission of HapII with high confidence. (C) Family 2_24 at risk for maternal transmission of neurofibromatosis type I. Only 26 haplotype blocks could be defined with a high mean number of SNPs per block (na = 40.5, nb = 29.6), resulting in longer blocks and therefore a lower block score (Sb = 0.54). Since 2/26 blocks were incorrectly classified as HapII instead of HapI, suggesting a SPRT error, the concordance score was estimated to be 0.84. Although the maternal variant is located in a concordant, HapI region, the small number of haplotype blocks, the haplotype change at the 3’ end of the locus, and the unclassified blocks at both ends prevent us from concluding on the fetal status for NF1. In a diagnostic setting, the analysis could be repeated on a subsequent sample. Sb: Bloc Score; Sc: Concordance score; na: mean number of SNP4a per haplotype bloc; nb: mean number of SNP4b per haplotype bloc; f: fetal fraction.