Frequency and spectrum of mutations in human sperm measured using duplex sequencing correlate with trio-based de novo mutation analyses

Abstract

De novo mutations (DNMs) are drivers of genetic disorders. However, the study of DNMs is hampered by technological limitations preventing accurate quantification of ultra-rare mutations. Duplex Sequencing (DS) theoretically has < 1 error/billion base-pairs (bp). To determine the DS utility to quantify and characterize DNMs, we analyzed DNA from blood and spermatozoa from six healthy, 18-year-old Swedish men using the TwinStrand DS mutagenesis panel (48 kb spanning 20 genic and intergenic loci). The mean single nucleotide variant mutation frequency (MF) was 1.2 × 10^{- 7} per bp in blood and 2.5 × 10^{- 8} per bp in sperm, with the most common base substitution being C > T. Blood MF and substitution spectrum were similar to those reported in blood cells with an orthogonal method. The sperm MF was in the same order of magnitude and had a strikingly similar spectrum to DNMs from publicly available whole genome sequencing data from human pedigrees (1.2 × 10^{- 8} per bp). DS revealed much larger numbers of insertions and deletions in sperm over blood, driven by an abundance of putative extra-chromosomal circular DNAs. The study indicates the strong potential of DS to characterize human DNMs to inform factors that contribute to disease susceptibility and heritable genetic risks.

Keywords: De novo mutations; Duplex sequencing; Extrachromosomal circular DNA; Mutation frequency; Mutational spectrum; Sperm DNA mutations.

Fig. 1

Proportions of single base substitutions (a) or short (≤ 20 bp) insertions (Ins) and deletions (Del) as well as multi-nucleotide variants MNVs () in blood and sperm. The number of mutations for each type of substitution is shown above the bars. The spectrum of SNVs and the spectrum of the small insertions, deletions and MNVs differed between blood and sperm (both p < 0.001).

Fig. 2

Frequencies of combined indels and SVs > 20 bp in length by tissue. For consistency with Table 1, only variants with VAF < 1% are included. (A) Frequencies by type, as determined by the standard data analysis pipeline. (B) Frequencies by amended type, with candidate circles plotted separately from insertions. Numbers above bars represent unique counts.

Fig. 3

(A) Schematic of (i) a reference allele ABCD, (ii) a duplex consensus read pair, (iii) split alignment of the R1 consensus read supporting a novel junction between the D and A (D-A junction), and two possible alternate alleles containing D-A junctions. Both (iv) excision and circularization of ABCD and (v) a chromosomal tandem duplication (TD) of ABCD create a D-A junction. (B) Histogram of allele length for insertion calls ≥ 20 bp identified in sperm DNA, colored by whether they contain a D-A junction. The vertical dashed line is at the median fragment size of the libraries (233 bp). (C) Considering a subset of events with allele length smaller than the median fragment size and containing a D-A junction (teal events left of vertical dashed line in B, n = 62), (i) chromosomal TDs should have fragment sizes that mirror the distribution of the whole library (teal and grey distributions, respectively), with fragment size independent of allele length. DNA circles in this subset should have fragment sizes less than or equal to the allele size, and thus, less than the median fragment size (diagonal hashes). (ii) All observed fragments were smaller than the median fragment size, significantly different from the expected distribution for chromosomal TDs (binomial p-value = 2.17 × 10^− 19).

Fig. 4

Proportions of single-base substitutions in the blood of the six men (present study) compared to proportions found in human granulocytes and pediatric blood cells, with no statistically significant difference between the spectrum in blood of our six men, and in the granulocytes (p = 0.86), and the pediatric blood cells (p = 0.49).

Fig. 5

Proportion of different single base substitutions in sperm (present study) and de novo mutations identified in children from sequencing human pedigrees (p for difference = 0.93). Data on de novo mutations are compiled from previously published studies^,,–.

Fig. 6

In both blood and sperm, the reconstructed mutational profiles with the highest similarity to the ones we found in blood and sperm were derived primarily from COSMIC mutational signatures associated with clock-like (i.e., age-dependent) processes. (A) The trinucleotide mutational profiles from mutations in blood showed the highest cosine similarity to a combination of SBS40, SBS1, SBS11 and SBS5. (B) The trinucleotide mutational profiles from mutations in sperm showed the highest cosine similarity to a combination of SBS5 and SBS1.