Noninvasive Fetal Genotyping by Droplet Digital PCR to Identify Maternally Inherited Monogenic Diabetes Variants

Abstract

Background: Babies of women with heterozygous pathogenic glucokinase (GCK) variants causing mild fasting hyperglycemia are at risk of macrosomia if they do not inherit the variant. Conversely, babies who inherit a pathogenic hepatocyte nuclear factor 4α (HNF4A) diabetes variant are at increased risk of high birth weight. Noninvasive fetal genotyping for maternal pathogenic variants would inform pregnancy management.

Methods: Droplet digital PCR was used to quantify reference and variant alleles in cell-free DNA extracted from blood from 38 pregnant women heterozygous for a GCK or HNF4A variant and to determine fetal fraction by measurement of informative maternal and paternal variants. Droplet numbers positive for the reference/alternate allele together with the fetal fraction were used in a Bayesian analysis to derive probability for the fetal genotype. The babies' genotypes were ascertained postnatally by Sanger sequencing.

Results: Droplet digital PCR assays for GCK or HNF4A variants were validated for testing in all 38 pregnancies. Fetal fraction of ≥2% was demonstrated in at least 1 cell-free DNA sample from 33 pregnancies. A threshold of ≥0.95 for calling homozygous reference genotypes and ≤0.05 for heterozygous fetal genotypes allowed correct genotype calls for all 33 pregnancies with no false-positive results. In 30 of 33 pregnancies, a result was obtained from a single blood sample.

Conclusions: This assay can be used to identify pregnancies at risk of macrosomia due to maternal monogenic diabetes variants.

Keywords: Bayesian analysis; droplet digital PCR; monogenic diabetes; non-invasive prenatal diagnosis.

Fig. 1. Allele abundance and fetal fraction in cfDNA.

Fractional abundance of the GCK or HNF4A reference allele for each variant (y-axis) is plotted against fetal fraction (x-axis) for 55 cfDNA samples from 38 pregnancies. Data points from pregnancies in which fetal genotype was homozygous reference or heterozygous are shown as circles or squares, respectively; in both cases, filled symbols show samples meeting quality thresholds for ddPCR analysis, whereas open symbols show samples with fetal fraction <2% (threshold indicated by vertical broken line) or with <10 positive droplets for the paternal allele in fetal fraction assays (7 and 1 samples, respectively). Error bars show 95% Poisson confidence intervals, as calculated by QuantaSoft software. The horizontal broken line indicates 50% fractional abundance, along which samples derived from pregnancies with heterozygous fetal genotype would be expected to lie.

Fig. 2. Probability of fetal genotype for homozygosity.

Probability scores for fetal homozygosity, derived from the custom Bayesian model, shown for 33 pregnancies in which at least 1 sample passed ddPCR quality thresholds. Pregnancies in which fetal genotype was homozygous reference or heterozygous are shown as circles or squares, respectively; in both cases, filled symbols show samples in which the model determined fetal genotype with ≥ 95% probability (i.e. P ≥ 0.95 for homozygous reference, P ≤ 0.05 for heterozygosity; n = 30), whereas open symbols show samples with intermediate probability scores (n = 3).

Fig. 3. Summary of workflow and results.