Retained chromosomal integrity following CRISPR-Cas9-based mutational correction in human embryos

Abstract

Human germline gene correction by targeted nucleases holds great promise for reducing mutation transmission. However, recent studies have reported concerning observations in CRISPR-Cas9-targeted human embryos, including mosaicism and loss of heterozygosity (LOH). The latter has been associated with either gene conversion or (partial) chromosome loss events. In this study, we aimed to correct a heterozygous basepair substitution in PLCZ1, related to infertility. In 36% of the targeted embryos that originated from mutant sperm, only wild-type alleles were observed. By performing genome-wide double-digest restriction site-associated DNA sequencing, integrity of the targeted chromosome (i.e., no deletions larger than 3 Mb or chromosome loss) was confirmed in all seven targeted GENType-analyzed embryos (mutant editing and absence of mutation), while short-range LOH events (shorter than 10 Mb) were clearly observed by single-nucleotide polymorphism assessment in two of these embryos. These results fuel the currently ongoing discussion on double-strand break repair in early human embryos, making a case for the occurrence of gene conversion events or partial template-based homology-directed repair.

Graphical abstract

Heindryckx and colleagues describe that human germline mutational correction by CRISPR-Cas9 results in embryos showing additional mutagenesis as well as solely wild-type alleles. These embryos displayed integrity of the targeted chromosome, while short-range loss of heterozygosity could be detected, which could represent gene conversion events or partial template-based homology-directed repair.

Introduction

Recent advancements in site-specific nucleases have greatly facilitated targeted genome editing.¹ To date, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) is one of the most promising versatile editing techniques available. The Cas9 DNA endonuclease introduces a double-strand break (DSB) at a specific locus, which can be resolved by several cell-endogenous repair mechanisms, including the non-homologous end joining (NHEJ) and homology-directed repair (HDR) pathways.^2,3 During NHEJ, the DSB is repaired by DNA strand ligation at the cleavage site. However, this may lead to insertion or deletion of base pairs (so-called insertions or deletions [indels]), which can result in shifts in the DNA reading frame and potential loss of gene function. On the other hand, HDR relies on a DNA template to repair the DSB, which can be exploited for precise gene editing.

Successful CRISPR-Cas9 application in human embryos has already been shown.^4,5,6,7 However, even though moderate targeting efficiencies were observed, they were often accompanied by mosaicism (different cells within one embryo harboring unedited and/or differentially edited alleles at the targeted locus). Mosaicism can appear in embryos because of continuous Cas9 cleavage coupled to erroneous repair attempts at later embryonic stages.^8,9,10 Mosaicism may be reduced, or perhaps even averted, by administering the CRISPR-Cas9 components simultaneously with the sperm into metaphase II (MII) oocytes instead of zygote-stage editing.⁶

Efforts on CRISPR-Cas9 editing have focused on correction of sperm mutations. The research group of Ma et al.⁶ reported efficient correction of a heterozygous 4-bp deletion in MYBPC3 by delivering the CRISPR-Cas9 components simultaneously with a repair template harboring the wild-type (WT) sequence. Remarkably, the mutant (MUT) alleles were not corrected using the exogenous repair template. Instead, they proposed a mechanism where the maternal WT allele guided the DSB repair attempts, which was designated as inter-homolog homologous recombination (IH-HR) or gene conversion.⁶ These findings were criticized by two research groups^11,12 that questioned whether technical issues could have led to detection of solely the untargeted maternal WT allele. While a more profound evaluation of the initially reported embryos indeed pointed to gene conversion,¹³ two subsequent independent studies targeting different genomic loci demonstrated partial or complete loss of the targeted chromosome following CRISPR-Cas9-mediated DSB induction.^7,14 Gene conversion events and (partial) chromosome removal could lead to loss of heterozygosity (LOH) at the targeted chromosome, which is loss of the genetic contribution of one of the parents to the cell. This can occur at different magnitudes. A summary of the genetic events potentially taking place in human embryos upon DSB induction is shown in Figure S1.

In this study, we attempted to correct an infertility-related point mutation (c.136-1G>C) in PLCZ1. Worldwide, up to 15% of couples are confronted with subfertility.¹⁵ Intracytoplasmic sperm injection (ICSI) is frequently used to overcome different forms of infertility. Still, complete fertilization failure occurs in 1%–5% of ICSI cycles. The main cause is the inability of the sperm to activate the oocytes, prompting resumption of meiosis. This process is dependent on the sperm-specific phospholipase C zeta (PLCζ) protein, encoded by PLCZ1, triggering calcium oscillations.^{16,17,18,19,20,21} Even though assisted oocyte activation (AOA), which artificially induces calcium oscillations, could overcome fertilization failure caused by these PLCZ1 defects, children born from this procedure may still inherit the infertility-causing mutation, resulting in the need for future medically artificial reproductive technologies. In this proof-of-concept study, the correction of a heterozygous base pair mutation (c.136-1G>C) in PLCZ1 by CRISPR-Cas9 was investigated to improve current knowledge of the DNA repair mechanisms taking place in human embryos. Genomic integrity of the samples was investigated by use of GENType,²² an improved/enhanced double-digest restriction site-associated DNA sequencing (ddRAD-seq) protocol.

Results

Allele-specific editing of the c.136-1G>C mutation in patient-specific induced pluripotent stem cells

The potential to efficiently correct single-base-pair substitution mutations, which is the most common type of protein-altering mutations, has remained understudied in human embryos. Therefore, we recruited one male patient with a heterozygous c.136-1G>C base pair substitution in PLCZ1. The PLCZ1 gene is located on the short arm of chromosome 12 (p12.3). The mutation is present in intron 3 at the border of exon 4 and represents a splice-site variant (Ensembl; rs1401353772). Two guide RNA (gRNA) molecules (gRNA1 and gRNA2) were designed (Figure 1A), containing the MUT allele in their spacer site. We employed a single-stranded oligodeoxynucleotide (ssODN) as a repair template, harboring the WT allele as well as a synonymous variant to track template usage.

Figure 1.

Correction of the c.136-1G>C PLCZ1 mutation in iPSCs

(A) Schematic of the designed CRISPR-Cas9 components. The mutant (MUT) DNA sequence (lowercase, intronic region; uppercase, exonic region) is shown in yellow, with the c.136-1G>C mutation marked in red. On the MUT sequence, the two gRNA target sequences are displayed, with their respective PAM sites and expected cleavage sites (dashed line) in the MUT DNA sequence, and below the ssODN molecule, where the WT “g” base (blue) replaces the MUT “c” base (red). The introduced synonymous variant to theoretically track template use is depicted in green. MUT allele correction can occur by two mechanisms: gene conversion (left; repair of the double strand break by homologous recombination with an interallelic homologous sequence, for instance, the maternal WT allele) or template-based homology-directed repair (right). (B) Bright-field image of induced pluripotent stem cells (iPSCs) colonies. Scale bar, 1,000 μm. (C) Representative line views following karyotyping by low-pass whole-genome sequencing in an iPSCs whole-well DNA extract. (D) Immunofluorescence analysis for DAPI (blue), OCT4 (green), and NANOG (red) in iPSCs colonies demonstrated that the iPSCs were pluripotent before and after the CRISPR-Cas9 nucleofection with gRNA2 and the template. Scale bars, 20 μm. (E) Workflow for nucleofection and analysis of iPSCs, which contain a heterozygous c.136-1G>C mutation in PLCZ1. The DNA was extracted from the whole well (i.e. 6 well-plate) or individual colonies. (F) Calculation of the allele types observed in the paternal allele for the different CRISPR-Cas9 component compositions for whole-well-extracted DNA and individual colony DNA by next-generation sequencing (NGS). (G) Calculation of the template use rate comparing the two gRNA designs in whole-well-extracted DNA and individual colony DNA by NGS. For the bulk analysis, a two-sided independent t test showed no significant difference (p > 0.05) between both gRNA designs for repair template use. (F and G) The bar charts represents the mean and standard deviation of 3 replicates.

To assess the efficiency and specificity of the designed CRISPR-Cas9 components, we first derived induced pluripotent stem cells (iPSCs) from the patient’s renal cells, which showed a primed pluripotency state and a normal karyotype (Figures 1B–1D). gRNA1 or gRNA2, complexed with Cas9, were nucleofected into the iPSCs either with or without the repair template (Figure 1E and Table S1). Targeted next-generation sequencing (NGS) analysis, performed on whole-well DNA extracts and individual colonies, revealed three types of alleles: the untargeted MUT allele; the MUT allele edited by NHEJ, leading to indel mutations (i.e., mutant editing); and a WT allele, which could either represent the original WT allele or the corrected MUT allele. For the whole-well data, the rates of the three allele types were similar (with an average repair rate of 18%) when comparing the two gRNA designs either with or without addition of the template (Figure 1F). To calculate the occurrence rate of the allele types with more certainty, compared with whole-well analysis, individual colonies were analyzed. For these individual colonies, a repair rate (i.e., WT rate) of 43% (26 of 60) was detected. Furthermore, from the colonies categorized as mutant editing (52%, 31 of 60), two did not contain the MUT allele anymore but instead displayed another indel mutation. This could be caused by failed repair attempts for which the donor ssODN got corrupted²³ (Figure S2) or by NHEJ associated with (multiple rounds of) DNA resection nucleotide addition,²⁴ where the PLCZ1 patient mutation could have vanished by resection, but additional indels appeared by nucleotide addition.

To assess template use during repair of the targeted MUT allele, we further analyzed the WT reads that exceeded the theoretical cutoff of 50% (representing the WT maternal allele; materials and methods). We observed that, for the bulk analysis, in 44% and 30% of those reads, template usage could be tracked because of the presence of the synonymous variant, for gRNA1 and gRNA2, respectively. This difference was not statistically significant (p = 0.331; Figure 1G). In individual colonies (n = 26; Table S2), 35% of the colonies displayed template use. This suggests that, in iPSCs, the template is not always required for repair.

To exclude (partial) chromosome loss of the targeted chromosome, analysis of 4 heterozygous single-nucleotide polymorphisms (SNPs) spread over a region of 1,600 bp around the target site was performed (Table S3). In all analyzed colonies (n = 33), heterozygosity was detected, pointing to gene conversion events instead of chromosome loss events in colonies displaying solely the WT allele, in which repair template use could not be detected (n = 12). Still, template-driven repair without incorporation of the synonymous variant could not be fully excluded. Finally, because gRNA1 and gRNA2 showed similar performance in terms of efficiency, and because the cleavage site of gRNA2 lies central to both intended base pair changes (i.e., MUT allele and synonymous variant), we chose to perform further human embryo gene editing experiments with gRNA2.

CRISPR-Cas9 utilization in human embryos does not influence developmental potential

The patient’s sperm (heterozygous, thus 50% chance of containing the mutation) was injected into donated spare oocytes (materials and methods) either without (i.e., in vitro controls; n = 41) or with (i.e., CRISPR-Cas9 experiments; n = 75) gRNA2, complexed with Cas9, and the repair template, followed by double exposure to ionomycin (i.e., AOA)²¹ to induce fertilization (Figure 2A; Tables S4 and S5). The fertilized embryos were cultured for 3–5 days in vitro. We compared the experimental development of both groups with the clinical data from injection of the patient’s sperm into his partner’s in vivo-matured oocytes (n = 37). We observed no significant difference in developmental rates (p = 0.7454) between the groups (Figure 2B). In summary, the CRISPR-Cas9 injections did not hamper the embryonic developmental potential.

Figure 2.

Development of M-phase oocytes injected with heterozygous mutated sperm and CRISPR-Cas9 components

(A) For the CRISPR-Cas9 experiments, the sperm was injected simultaneously with the CRISPR-Cas9 components in donated spare oocytes (e.g. in vitro matured [IVM] metaphase II [MII] oocytes; materials and methods). The injected oocytes underwent assisted oocyte activation (AOA), resulting in fertilized embryos that were kept in vitro in culture for 3–5 days. Because of the heterozygous nature of the mutation, it was expected that half of the injected sperm cells were mutant (MUT). Gene targeting of the paternal MUT allele could theoretically lead to different outcomes, presuming that the maternal wild-type (WT) allele will be left untargeted. Solely the WT allele could be present because of three editing outcomes. First, homology-directed repair (HDR) using the exogenously delivered repair template could have taken place, which theoretically could be tracked by the presence of an additional synonymous mutation. Second, gene conversion using the maternal allele could have occurred (leading to loss of heterozygosity [LOH]). Third, the targeted chromosome could have been (partially) lost (leading to LOH). Next to detection of solely WT alleles, additional indel mutations could also be observed on the MUT allele because of non-homologous end joining (NHEJ). Finally, the paternal c.136-1G>C PLCZ1 mutation could also be left untargeted. Moreover, a mosaic embryo could also display a combination of different edited alleles at the targeted locus for different blastomeres. (B) The developmental data of the in vivo control (MII), in vitro control (IVM) or smooth endoplasmic reticulum aggregates (SERa) oocytes; materials and methods), and a CRISPR-Cas9-injected group (IVM/SERa) for 2 pronuclei (2PN) appearance, and day 3 and day 5 survival are shown in a Kaplan-Meier plot and a table. A log rank (Mantel-Cox) test was performed. No significant difference (p > 0.05) in developmental rates was seen.

Specific and moderately efficient targeting of the MUT c.136-1G>C allele

Because of the heterozygous nature of the mutation, we hypothesized that half of the sperm cells were MUT. This was further validated by NGS analysis of a bulk sperm sample, showing that 48.5% (48.5% ± 3.9%, n = 4, average coverage 650 reads) of the reads contained the c.136-1G>C mutation. To assess whether the generated embryos originated from WT or MUT sperm, we carried out a short tandem repeat (STR) analysis, investigating a 10- to 18-Mb genomic region around the target site (Table S6). Similar to the previously described mutation rate in the bulk sperm sample, the STR data of the embryos demonstrated that 47% (27 of 58) of the embryos originated from MUT sperm (Figure 3A; Table S4). Through targeted NGS assessment of the PLCZ1 gene, it could be concluded that the designed CRISPR-Cas9 components seemed to almost specifically target the MUT allele because only 3% (1 of 31) of the embryos originating from WT sperm (Figure 3B; Table S4, embryo 1), showed indels in only a small percentage (6.81%) of the reads, which may also be caused by PCR or sequencing artifacts.

Figure 3.

Analysis of Repli-G whole-genome amplified human embryo samples targeted for correction of the c.136-1G>C mutation

(A) The percentages of embryos originating from either wild-type (WT) or mutant (MUT) sperm based on the short-tandem repeat (STR) data. (B) A schematic representation of the allele types for embryos originating from WT (n = 31) or MUT sperm (n = 27), as identified by next-generation sequencing (NGS). (C) A schematic of the allele types for the individual blastomeres originating from WT or MUT sperm by NGS. The allele types per embryo are shown, with each count representing one blastomere. (D) The percentages of specific allele types for the individual blastomeres from the embryos originating from MUT sperm. (E) Frequency of observation of loss of heterozygosity (LOH) events (based on the STR analysis) for the different allele types.

When considering the embryos originating from the MUT sperm, NGS analysis of the targeted region revealed three main categories of allele types, comparable with those observed in the iPSCs: 22% (6 of 27) of the embryos showed the untargeted paternal MUT allele, 52% (14 of 27) showed the paternal MUT allele harboring indel mutations, and 26% (7 of 27) showed a WT allele, which may represent the corrected paternal allele (hereafter referred to as “absence of mutation”) (Figure 3B). The synonymous variant was detected in only one of these seven “absence of mutation” embryos (Table S4, embryo 56). In this specific blastocyst, the synonymous variant was, however, present in only 1% (7 of 1,116) of the examined sequencing reads, which suggests that the template was employed in only a very limited number of cells. It is, however, also possible that this finding represents a PCR or sequencing artifact. In summary, a quarter of the embryos originating from MUT sperm showed only the WT allele, which could represent correction events, while half of the embryos displayed signs of DSB repair by NHEJ.

Even though whole-embryo extracts were classified in one of the three main categories of allele types (untargeted paternal MUT allele, paternal MUT allele with indel mutations, and WT allele), signs of mosaicism could be observed in the whole-embryo samples. To make a further distinction, embryos were classified in the category “paternal MUT allele with indel mutations” when indel mutations were observed at a rate equal to or higher than 5%, which might lead to overestimation of the occurrence of these events.

To more reliably map mosaicism, we next analyzed individual blastomeres from 11 embryos, obtained on day 3 (Table S5; Figure 3C). For the five embryos originating from MUT sperm, 6% (2 of 32) of the blastomeres showed the untargeted MUT allele, 19% (6 of 32) additional indel mutations, and 75% (24 of 32) the absence of the MUT allele (Figure 3D). Four of these five embryos were mosaic. The synonymous variant incorporated in the template was never present, which could suggest that the template is hardly employed in human embryos or that template-driven repair without incorporation of the synonymous variant occurred.

Based on these findings and previous CRISPR-Cas9-based human embryo editing research, the absence of the MUT allele in embryos that originated from a MUT sperm cell could be explained by either gene conversion^6,13,25 (i.e., repair of the DSB by recombination with an interallelic homologous sequence, for instance, the maternal WT allele) or by segmental/complete loss of the paternal chromosome.⁷ To preliminarily assess the occurrence of these events in our current dataset, we employed the generated STR data that were used to determine sperm origin earlier (Table S6; Figure 3E). Based on the presence of solely one STR marker peak at one or multiple STR loci close the mutation site, gene conversion could be suspected, while one STR marker peak at all investigated STR loci could point to larger gene conversion events or chromosome loss. In 43% (3 of 7) of the whole-embryo samples originating from MUT sperm, where the targeted mutation was absent, signs of LOH events were observed. More specifically, in one embryo (Table S6, embryo 56), multiple STR markers adjacent to the target site only displayed one peak, while at farther distance, both parental peaks were identified. In two other embryos (Table S6, embryos 52 and 53), one STR marker peak was predominantly observed across the fully analyzed region. Similar observations were made in the individual blastomeres analyzed. In 17% (3 of 18) of the blastomeres showing absence of mutation, one STR marker peak was predominantly observed across the fully analyzed region (n = 2; Table S6, cells 7E and 8G) or the complete downstream tract (n = 1; Table S6, cell 11C). In the cases where maternal and paternal markers were still present across the fully analyzed region, we hypothesize that gene conversion events smaller than the resolution of our assay could have occurred, which should be detected with other methods, such as SNP analysis. Interestingly, in 3% (whole embryos, 1 of 30) or 3% (individual blastomeres, 1 of 33) of the seemingly untargeted embryos originating from WT sperm, signs of LOH were detected, which seems to suggest that CRISPR-Cas9 targeting also introduced a DSB in these embryos, resulting in LOH events. However, the lack of maternal DNA samples for some samples, together with occasional PCR artifacts for the STR analysis, makes it difficult to draw clear conclusions.

In summary, the STR data hinted at the presence of gene conversion and/or chromosome loss events and, therefore, a more in-depth, genome-wide analysis with a higher resolution is required to draw further conclusions.

Whole genome-wide SNP analysis of whole-genome amplification (WGA) samples reveals retained chromosomal integrity at the targeted locus

To enable generation of sufficient amounts of DNA for genetic analysis of single-cell and limited-cell samples, WGA was carried out on all embryo samples in this study. Initially we used the Repli-G (Advanced DNA) Single Cell Kit, in accordance with previous studies on human embryo gene editing.^6,7,13,25 However, a major drawback of WGA is the occurrence of allelic drop-out (ADO), in which one of the alleles of a heterozygous DNA sequence is insufficiently amplified, which can lead to grave misinterpretation of the data.²⁶ Depending on the WGA method used, the impact of ADO can differ.^22,26,27 Based on findings from the technical validation performed in the framework of the GENType pipeline,²² we concluded that the PicoPLEX Single Cell WGA Kit V3 outperformed the Repli-G kit in terms of copy number profile noise, especially with low amounts of input DNA.

We therefore created, in total, 31 new whole-embryo samples. Their analysis revealed (Figure 4A; Table S7) that 50% (9 of 18) of the embryos originating from MUT sperm showed only the WT allele. The other half (9 of 18) of the embryos originating from MUT sperm displayed indel mutations (i.e., mutant editing). Remarkably, 3 of 9 (embryos 18’, 19’, and 21’) of these mutant editing embryos did not show the targeted variant but instead displayed only new indel mutations, which could be the result of failed repair events (Figure S2) or NHEJ on the MUT allele, as described previously for two mutant editing colonies. The low indel rates in some of the whole-embryo samples suggest the occurrence of mosaicism. To further investigate mosaicism, three additional embryos were created. Single cells were obtained at the eight-cell stage. Two embryos originated from MUT sperm, which were both mosaic, and 60% (9 of 15) of the blastomeres possessed only the WT allele (Figure 4B). In three blastomeres (2’C, 3’B, and 3’G), the MUT allele with an indel mutation was detected in almost all the reads. Therefore, a new category was introduced, called “homo-indels”. Detection of only one indel was suggestive of the presence of LOH.

Figure 4.

Analysis of PicoPLEX whole-genome amplified human embryo samples targeted for correction of the c.136-1G>C mutation

(A) A schematic representation of the allele types identified by next-generation sequencing (NGS) in the embryo samples whole-genome amplified with PicoPLEX, originating from wild-type (WT;n = 13) or mutant (MUT;n = 18) sperm. (B) A schematic representation of the allele types identified in the individual blastomeres whole-genome amplified with PicoPLEX, originating from WT (n = 1) or MUT (n = 2) sperm by NGS. The allele types per embryo are shown, with each count representing one blastomere.

The GENType pipeline enables genome-wide SNP-based haplotyping, assessing embryo and related (parental and sibling) DNA samples. In the case of this study, we included the DNA from the patient (i.e., sperm donor), DNA from the sperm donor’s mother (which carries the pathogenic mutation), DNA from the donated oocytes (obtained from the cumulus cells corresponding to the donated oocyte), and the embryonic whole-genome-amplified DNA. For each informative SNP present in the embryo, its origin can therefore be traced back to the inherited chromosome, enabling haplotyping (Figures 5A and 5B). Additionally, the copy number profiles for each chromosome can be investigated. As a quality measure, only samples with an ADO rate lower than 50% were further analyzed. For the retained whole-embryo samples, an average ADO rate of 23% (and 2% of allelic drop-in) was detected. In total, the GENType results of 13 whole-embryo samples were evaluated. Six embryos originated from WT sperm (embryos 1’, 2’, 3’, 4’, 8’, and 13’), while seven originated from MUT sperm (embryos 14’, 15’, 16’, 20’, 27’, 30’, and 31’). From the latter samples, four displayed additional mutagenesis of the MUT alleles (embryos 14’, 15’, 16’, and 20’), while three solely showed the WT allele (embryos 27’, 30’, and 31’), pointing to repair events. All copy number variation (CNV) profiles showed retained integrity of the targeted chromosome 12, which carries the targeted allele (Figures 5C and S3). The B-allele frequency profiles did not show any signs of LOH events (resolution of ∼10 Mb) in any sample (Figures 5D and S4). The individual blastomere analysis displayed an average ADO rate of 53%. Nevertheless, we could still conclude that chromosome 12 was retained and that no large-scale LOH events (larger than 10 Mb) were present on the targeted chromosome. In summary, our data demonstrated that CRISPR-Cas9-mediated genome editing in human embryos is not predominantly linked with (partial) chromosome loss.

Figure 5.

GENType pipeline summary, performed on the PicoPLEX whole genome amplified embryo 31′ sample, showing absence of mutation

(A) A pedigree was created based on the whole-genome-wide assessed single-nucleotide polymorphisms (SNPs), which enabled us to trace back the parental origin of each allele. The mutant (MUT) allele of the sperm can be traced back to the mother of the sperm donor and is colored in light blue, while the wild-type (WT) paternal allele is colored in light green. The maternal WT alleles are dark green or pink. In this example, the created embryo originates from the MUT allele. (B) A close up of one of the 78,399 analyzed SNPs from embryo 31’. (C) The copy number variant profiling graph, generated through Vivar,²⁸ from embryo 31’. (D) The B-allele frequency profile²² for chromosome 12 from embryo 31’.

Although ddRAD-seq has genome-wide coverage, the resolution is limited, and first reported SNPs lie, on average, ∼100 kb upstream and ∼75 kb downstream of the mutation site in whole-embryo samples. To complement these data with information more closely to the mutation site, we performed a manual SNP assay in which we investigated 5 SNPs in a region of 4,160 bp upstream and 6 SNPs in a region of 4,354 bp downstream of the mutation site (Figure S5A; Table S8). A sample was categorized as “no LOH events” when both parental markers were present in all SNPs or when at least the SNPs closest to the mutation site (up- and downstream) displayed both paternal markers. In a first phase, we analyzed all 31 PicoPLEX-amplified samples. LOH events were detected in 56% of the “absence of mutation” whole-embryo (5 of 9) and 44% (4 of 9) of the individual blastomeres samples. Remarkably, seemingly untargeted WT embryos displayed LOH events in 8% (1 of 13) of the whole-embryo and 67% (4 of 6) of the individual blastomeres samples. In the 3 homo-indel blastomeres, LOH events were detected downstream of the target site. However, higher ADO rates make it impossible to fully state whether the homozygous markers observed represent true LOH events or are due to ADO. Interestingly, the STR analysis (Table S9) displayed no LOH events for the whole-embryo samples. Together, this suggests that, if gene conversion took place, then the conversion tract would be of a restricted size. Additionally, for samples that showed solely the WT allele but only heterozygous markers in the SNP assay, we hypothesize that gene conversion events smaller than the resolution of our assay (i.e., 476 bp and −977 bp) could have occurred to repair the DSB and remove the c.136-1G>C mutation. However, template-driven repair without incorporation of the synonymous variant could also have appeared.

In a second phase, we focused only on the 13 PicoPLEX-amplified whole-embryo samples with sufficient WGA quality. Predominantly heterozygous SNPs were detected up- and downstream of the mutation site in 83% (5 of 6) of the untargeted WT, 75% (3 of 4) of the mutant editing, and 100% (3 of 3) of the “absence of mutation” whole-embryo samples (Figure S5B).

To exclude the presence of large deletions, we performed long-range PCR with two primer pairs (Figure S6) from 1,410 bp (long-range PCR1) and 1,367 bp (long-range PCR2). The long-range PCR2 primers did not result in conclusive results for the individual blastomeres, which could be explained by the fact that PicoPLEX generates WGA products ranging from 100 bp to 2 kb, which limits use of long-range PCR. Therefore we excluded the long-range PCR2 results from these samples. Overall, for the investigated whole-embryo samples (i.e., 13 untargeted WT, 7 indel mutations, and 8 “absence of mutation” embryos) and the individual blastomeres, no signs of large deletions were detected.

In summary, because of the presence of heterozygous SNPs in the “absence of mutation” embryos (embryos 27’, 30’, and 31’) up- and downstream, the absence of large deletions up to 1.4 kb, the low ADO rate (average 28%), and the normal copy number of chromosome 12, we conclude that, for our GENType-analyzed samples, short-range gene conversion could have occurred and, therefore, that the pathogenic mutation has been corrected.

Discussion

Human embryo gene editing studies have been accompanied by technical inconsistencies and ethical controversies. In this study, we aimed to contribute to the current knowledge, targeting an infertility-related heterozygous mutation in PLCZ1 with CRISPR-Cas9. Only a moderate (i.e., 36%; 16 of 45) “absence of mutation” rate was observed when whole embryos were analyzed, which is in line with previous studies.^6,7,25 By contrast, our single-blastomere data showed a more efficient (i.e., 70%; 33 of 47) “absence of mutation” rate. This illustrates that the level of “absence of mutation” may be underestimated when analyzed at the whole-embryo level. This is due to the data interpretation of whole-embryos, where the presence of (even a low percentage of) reads with indels results in classification of the embryo as additionally mutated (i.e., indel mutations) even though corrected blastomeres could be present. Additionally, detection of a “T” insertion or “T” deletion in a run of T bases (embryos 47, 51, 18’, and 21’) could represent a small sequencing artifact. This could therefore result in overestimation of NHEJ-induced defects in samples, which is impossible to distinguish from genuine small indels in repetitive sequences.

Ma et al.⁶ previously suggested that M-phase injection practically eliminates mosaicism compared with S-phase injection. However, in our study, single-cell analysis demonstrated that mosaicism is still present in approximately 86% (6 of 7) of the investigated embryos obtained after M-phase injection with MUT sperm. These findings are in line with the single-cell sequencing data of a recent study of Zuccaro et al.⁷ The occurrence of mosaicism impedes potential future clinical implementation of this technology, where the success of editing should be reliably assessed by preimplantation genetic testing using single blastomeres or trophectoderm biopsies. Additionally, the need for good-quality oocytes for M-phase injections to reduce the level of mosaicism can be an obstacle for patients with an advanced maternal age.²⁹ Next to a lower oocyte amount, the decreased gene expression and mRNA level in advanced age oocytes could result in DNA repair deficiency and negatively impact embryonic development, especially in the presence of DNA damage caused by CRISPR-Cas9.^30,31,32

To enable creation of sufficient amounts of DNA, WGA is necessary, which is inherently linked with the occurrence of ADO.²⁶ Previous studies have reported that the ADO rate is variable for different sample input types and is dependent on the WGA technique (with PicoPLEX surpassing the Repli-G kit), on the number of cells included (with single cells displaying more ADO than multiple cells), and on the target locus.^22,25,26,27 In this study, the occurrence of ADO at the mutation site could lead to inability to detect the c.136-1G>C allele, potentially leading to misinterpretation of mutational correction. Because of the presence of ADO after WGA, it is challenging to identify whether the identified short-range LOH events in “absence of mutation” embryos are due to gene conversion events or due to ADO. Recently, Liang et al.²⁵ derived embryonic stem cells of edited embryos to investigate the impact of WGA-induced ADO. They detected a lower amount of LOH in stem cells compared with individual blastomeres analysis, demonstrating the limiting factor of ADO on embryo assessment. However, the fact that LOH could still be detected in the embryonic stem cells, which did not require WGA, suggests that ADO, by WGA, is not the sole explanation for the occurrence of LOH. Additionally, Borgström et al.,²⁶ who thoroughly analyzed the genetic impact of WGA in a human cell line, showed that ADO is not restricted to any specific genomic region and is instead spread over the entire chromosome. Therefore, it is unlikely that ADO was the sole reason for the disappearance of the MUT allele and that all short-range LOH events are solely due to ADO. Instead, a large proportion of the observed LOH events should be due to short-range gene conversion. Finally, when average genome-wide ADO rates are low for a sample, as was the case for the 13 selected PicoPLEX-amplified samples, assessment of the allele type at the mutation site and the most adjacent SNPs should give a strong indication of the repair events that had occurred, which, in the case of the “absence of mutation” embryos, is gene conversion.

Nonetheless, the only way to reliably prove gene conversion is selection of a (preferably homozygous) mutation site with multiple heterozygous SNP(s) being present within the range of a couple of hundred base pairs up-and downstream of the mutation site. Such a selection would align perfectly with the range of amplification products generated by, for instance, PicoPLEX WGA (average 500 bp) and enables analysis of mutation site and informative SNPs within the range of one PCR and Sanger sequencing product, which should be aimed for in future experiments to fully rule out the insecurity associated with ADO.

It is important to note that another possible explanation for the observation of these short-range LOH patterns is the presence of large deletions, which has already been reported by different research groups; for instance, in mouse embryos¹² and embryonic cells.³³ In contrast, in human embryo experiments performed by Ma et al.,¹³ no large deletions were detected after gene editing, as assessed by performing long-range PCR. Liang et al.²⁵ detected large deletions in 13% of the individual blastomeres when targeting both WT alleles, but this amount decreased to 3% (i.e., all from one embryo and calculated from mosaic embryos) when targeting only a heterozygous allele. Apart from the CNV analysis performed in this study, which could identify deletions down to 3 Mb, we performed two long-range PCR reactions (i.e., 1,410 bp and 1,367 bp). No large deletions in the amplified fragments were detected. The nature of generated WGA products made it theoretically impossible to perform long-range PCR for a larger region. In more detail, PicoPLEX generates WGA products with an average size ranging between 100 bp and 2 kb, which limits the use of long-range PCR.

A short-range LOH event was identified in one embryo originating from WT sperm, which could point to either ADO or gene conversion events. This indicates that the CRISPR-Cas9 system might not specifically target the MUT allele, although it should be mentioned that no indel mutations have been observed in the PicoPLEX-amplified embryos originating from WT sperm. For future therapeutic use, introduction of additional mutations at the mutated paternal allele as well as at the WT maternal allele should be avoided. Targeting of sites with high sequence similarity is a well-known limitation of CRISPR-Cas9 editing and is in agreement with the observation of various groups that single variations in the 5′ end of the protospacer sequence are well tolerated.^34,35 Unintentional targeting of the WT allele because of the single-base-pair difference between the WT and MUT allele would, however, greatly complicate potential future clinical use of CRISPR-Cas9 to correct single-base-pair mutations, which is the most common type of disease-causing mutation. It is important to note that our study is the first study to thoroughly evaluate LOH in the seemingly untargeted embryos.

Mutational correction should preferably entail minimal alterations to the organism’s genome. Recent publications^6,7,13,25 have sparked a debate on the mechanisms induced upon DSB generation in human embryos. Unexpectedly, it appeared that exogenous repair templates are hardly employed to correct DSBs, which was in agreement with the lack of synonymous variant incorporation in our embryos. It is important to remark that it is theoretically possible that the template was utilized without incorporation of the synonymous variant because of the unidirectional nature of DNA synthesis. Therefore, for future studies, the most optimal setup would include an ssODN that contains a synonymous variant up- and downstream of the theoretical cleavage site. This would ensure that, independent of the direction of DNA repair synthesis, a synonymous variant would be incorporated. Nevertheless, previous zebrafish germline²³ and cell editing³⁶ experiments have demonstrated that variants up- and downstream of the cleavage site could be incorporated. In our setup, a downstream synonymous variant was selected that disrupted the protospacer adjacent motif (PAM) sequence (for gRNA2) to reduce the chance of re-editing in repaired alleles. In general, it is well reported that changes closer than 10 bp to the cleavage site are introduced at a high efficiency.^37,38 For the gRNA utilized for the germline experiments (i.e., gRNA2), the synonymous variant is only 6 bp away, which implies a high variant incorporation efficiency. Additionally, template use was observed in the iPSCs, which provides further proof that template-driven repair could be absent during germline editing. For four of the nine “absence of mutation” embryos displaying LOH patterns around the cleavage site with heterozygous SNPs farther away, the presence of template-driven repair could be excluded.

As mentioned before, seemingly repaired alleles have been suggested previously to be the result of either gene conversion^6,13,25 or chromosome loss events.⁷ In contrast to Zuccaro et al.,⁷ no chromosome loss was detected in our GENType-analyzed samples with absolute certainty. Instead, the embryos exhibited short-range LOH or potential non-detectable LOH, taking into account the resolution of our analysis, as discussed already above. This could point to the relative frequent occurrence of gene conversion events, which is in line with the report of Ma et al.⁶ The clear lack of chromosome loss following Cas9-based DSB induction, in contrast to Zuccaro et al.,⁷ could be explained by the chromosomal location of the target site and the chromosomal context (e.g., open or closed chromatin state). A second discrepancy between our and previous studies is the use of slightly different concentrations of CRISPR-Cas9 components, although the concentrations used in our study are situated between the concentrations used by Ma et al.,⁶ Fogarty et al.,³⁹ and Stamatiadis et al.^40,41

Overall, these results underscore that removal of the mutation in human embryos is possible, but caution is warranted when assessing the outcome of CRISPR-Cas9 editing in human embryos. Even though we did not observe loss of the targeted chromosome in our GENType-analyzed samples, we still observed short-range LOH events. Even in these cases, a potential detrimental outcome could be expression of recessive alleles, which has also been seen in cancer and could also result in epigenetic imprinting disorders.⁴²

To minimize the number of created human embryos, we employed patient-specific iPSCs to evaluate the potency of the designed CRISPR-Cas9 components. Our results show that cells can be used as a valuable approach, providing a first glance into gene editing. Remarkably, in agreement with our embryo results, no loss of the targeted chromosome seems to be present after gene editing. However, the fact that the exogenously delivered repair template was used, which was also seen (to a greater degree) by Ma et al.,⁶ limits its full predictive value for human germline editing.

In summary, this is the first study aiming to overcome an infertility-related mutation, which shows the potential of CRISPR-Cas9 gene editing in combination with other assisted reproduction techniques. Additionally, our results demonstrate the non-robust nature of CRISPR-Cas9, exemplified by the occurrence of complex genetic outcomes, such as mosaicism and LOH, which impedes potential future clinical use of CRISPR-Cas9 in its current form. However, we can exclude the hypothesis that DSB induction in human embryos is inherently linked with (partial) chromosome loss, at least in our GENType-analyzed samples. Still, the occurrence of DSB-induced gene conversion and chromosome loss requires further investigation with the appropriate analysis techniques in a higher number of embryos.

Materials and methods

iPSCs

Ethics approval

The iPSC experiments were approved by an independent Commission for Medical Ethics affiliated with the University Hospital of Ghent and the University of Ghent (EC 2014/1155). Written consent was obtained from the patient with the heterozygous base pair substitution in PLCZ1 (c.136-1G>C) for his urine donation.

CRISPR-Cas9 complex preparation

Two specific gRNA designs (i.e., gRNA1 and gRNA2) and one design of the template were made (Table S10; Figure 1A). To identify the mode of correction, a synonymous single base variant was present in the repair DNA template, which was an ssODN. In case the template was utilized, two single-base substitutions would theoretically be present: “c” to “g” on c.136-1 and the synonymous variant “g” to “a.” In case gene conversion occurred, only one substitution was expected to be present; namely, “c” to “g” on c.136-1.

The CRISPR components (i.e., Alt-R crispr RNA (crRNA), trans-activating crRNA (tracrRNA), Streptococcus pyogenes Cas9 nuclease V3, and ssODN template) were obtained from IDT (Belgium). The gRNA was prepared with equimolar concentrations of crRNA and tracRNA. The CRISPR-Cas9 mixture contained 202 pmol gRNA, 88 pmol Cas9 protein, and 100 pmol ssODN template.

Creation of iPSCs

Urine was donated by the patient. The iPSCs were created from renal cells extracted from a urine sample following the protocol from Zhou et al.⁴³ The biggest adaptation was use of the CytoTune-iPS 2.0 Sendai Reprogramming Kit (Thermo Fisher Scientific, MA, USA) for transfection. The iPSCs were seeded on inactivated mouse embryonic fibroblasts (MEFs). The iPSC medium (all elements from Thermo Fisher Scientific) consisted of 78% Dulbecco's Modified Eagle Medium (DMEM)/F-12 GlutmaMAX, 20% KnockOut serum replacement, 1% 10 mM Minimum Essential Medium (MEM) non-essential amino acid solution, 0.1% 55 mM β-mercaptoethanol, 1% penicillin-streptomycin, and 0.04% 10 μg/mL basic fibroblast growth factor (bFGF, Peprotech, UK). These iPSCs were in a primed state of pluripotency after reprogramming. After a few passages, the iPSC medium was changed to another culture medium consisting of 80% KnockOut DMEM, 20% KnockOut serum replacement, 2 mM L-Glutamine, 0.1 mM MEM non-essential amino acid solution, 100 U μg/mL penicillin-streptomycin, and 0.1 mM β-mercaptoethanol supplemented with 4 ng/mL bFGF.

Characterization of iPSCs

DNA was extracted with GenElute Mammalian Genomic DNA Miniprep (Sigma-Aldrich, MA, USA) from a whole well and PCR-amplified with PLCZ1 primers (Table S10). The PCR fragments were analyzed with a fragment analyzer (5300 Fragment Analyzer, Agilent, CA, USA), which utilized qualitative 915/930 double-stranded DNA (dsDNA) kits (Agilent Technologies), and through targeted NGS (Miseq, Illumina, CA, USA) (see below) to detect the heterozygous presence of the c.136-1G>C mutation.

The pluripotency of the iPSCs was checked before and after nucleofection through immunostaining. Coverslips with iPSCs seeded on inactivated MEFs were fixed with 4% paraformaldehyde (PFA) for 20 min at room temperature. After fixation, the coverslips were washed in PBS. The cells were permeabilized for 8 min in PBS with 0.1% Triton X-100, followed by 1 h of blocking at room temperature in blocking solution (PBS + 0.1% Tween + 10% fetal calf serum [FCS] + 1% BSA). The primary antibodies for Oct4 (1:100, sc-5279, Santa Cruz Biotechnology, TX, USA) and Nanog (1:100, 500-P236, Peprotech) were applied overnight at 4°C. The next day, the coverslips were washed in PBS. After washing, the coverslips were stained with Alexa Fluor 594 and Alexa Fluor 488 (1:500, Abcam, UK) for 1 h at room temperature. After a wash step, the cells were exposed to DAPI (1:50, Thermo Fisher Scientific) for 10 min, followed by a last wash step. The stained coverslips were visualized with a laser-scanning confocal microscope (Carl Zeiss, LSM900).

Nucleofection of iPSCs

The nucleofection protocol is based on the article of Jacobs et al.,⁴⁴ which employed the P3 Primary Cell 4D-Nucleofector X Kit (Lonza, Switzerland) and the 4D Nucleofector (Lonza). Around 75,000–100,000 cells were collected per reaction. The P3 Primary Cell Nucleofector solution and supplement were mixed together with the constructed ribonucleoprotein (RNP) complex and the Alt-R Cas9 Electroporation Enhancer (IDT). The cells were added to this mixture, which was then transferred to the Nucleocuvette. The nucleofection was performed on program DN100 with the 4D Nucleofector. Afterward, the cells were resuspended in pre-warmed culture medium supplemented with 10 μM RHO/ROCK pathway inhibitor (ROCKi, Y27632, Enzo Life Sciences, NY, USA) and 15 μM Alt-R HDR Enhancer or 0.69 mM Alt-R HDR Enhancer V2 (IDT). These cells were left in the incubator (37°C, 5% O₂, 5% CO₂) for 48 h. After 2 days, the medium was refreshed, but ROCKi or the enhancer was not added.

Genetic analysis of edited iPSCs

The cells were cultured in a 6-well plate after nucleofection. For whole-well extraction, all colonies from one well were pooled together in one sample after reaching 80% confluency. DNA was extracted with GenElute Mammalian Genomic DNA Miniprep (Sigma-Aldrich). For each condition, 3 biological replicates were tested. For one condition (sgRNA2 and ssODN), we analyzed individual colonies. These colonies, which did contain clear borders, were manually isolated. DNA was extracted with the Arcturus PicoPure DNA Extraction Kit (Thermo Fisher Scientific). The extracted DNA was PCR amplified with PLCZ1-specific primers (Table S10). The amplified PCR fragments, with an amplicon length of approximately 500 bp surrounding the mutation site, were analyzed with a fragment analyzer (Agilent) and through targeted NGS (see below) to evaluate the occurrence of different allele types and the presence of the synonymous variant.

For the whole-well data, the additional WT percentage was calculated by subtracting the theoretical expected 50% WT reads originating from the WT allele (i.e., 50% cutoff). It is important to note that these are the theoretical numbers because bulk sequencing makes it impossible to calculate the true excess of WT alleles, which can be assessed by analyzing individual colonies.

For each whole-well sample, we calculated the template rate by dividing the WT reads with the synonymous variant with the total amount of WT reads that exceeded the theoretical 50% cutoff. In more detail, if 60% of the reads contained the WT allele, then we calculated that 10% of these reads were additional WT reads (i.e., above the 50% cutoff). Additionally, for this sample, 5% of the reads contained the WT allele with the synonymous variant but without indels or the c.136-1G>C patient mutation. To calculate the template rate, we divided the 5% (WT reads with synonymous variant) by the 10% (additional WT reads), resulting in a template rate of 50%. In Figure 1F, the theoretical allele type percentages of the paternal allele were displayed by multiplying the average additional WT, untargeted MUT, and mutant editing rate (i.e., indel mutations) by two.

For the individual colony data, MUT/indel/allele rates in less than 15% were seen as background signal.⁴⁵ Additionally, colonies that displayed signs of more than two allele types were removed from the analysis. This could be due to collection of cells of another colony during manual isolation or due to formation of a colony from more than one cell.

The LOH events around the cleavage mutations site were analyzed through a SNP assay (see below).

Germline editing: M-phase injections

Ethics approval

The human germline editing experiments were approved by an independent Commission for Medical Ethics affiliated with the University Hospital of Ghent and the University of Ghent (EC 2018/0908 and BC-08272) and by the Belgian Federal Ethical Committee (Adv_077_UZGent and Adv_086_UZGent). Written consent was provided and signed by the female patients who donated their spare oocytes: immature oocytes and in vivo-matured oocytes containingaggregates of smooth endoplasmic reticulum (SERa). Written consent was obtained from the recruited male patient with the heterozygous base pair substitution in PLCZ1 (c.136-1G>C) for his sperm. CRISPR-Cas9 was not used in a clinical context but only in an experimental context. Clinical data and diagnostic tests obtained in the Ghent University Hospital were collected from the patient before the start of this research and are summarized in Table S11.

CRISPR-Cas9 complex preparation

The gRNA2 and template design described for the iPSC experiments were employed for the germline experiments. The gRNA was prepared with equimolar concentrations of crRNA and tracrRNA. The final CRISPR-Cas9 mixture contained 50 ng/μL gRNA, 100 ng/μL Cas9 protein, and 100 ng/μL ssODN template.

Human oocyte collection

All oocytes were collected from patients undergoing infertility treatment at the Ghent University Hospital. For most patients, an agonist treatment was chosen for controlled ovarian stimulation, followed by triggering of ovulation by administration of human chorionic gonadotrophin (hCG). After oocyte collection, the oocytes were enzymatically and chemically denuded. Only oocytes that were immature (namely, germinal vesicles [GVs] or metaphase I [MI]) or in vivo-matured oocytes showing SER aggregates were donated for research after informed consent. Immature oocytes were in vitro matured into MII oocytes in an in-house in vitro maturation (IVM) medium for GV oocytes or in Sydney in vitro fertilisation (IVF) cleavage medium (CC, Cook Medical, IN, USA) for MI oocytes. In vivo-matured SERa oocytes were collected in CC medium. Only IVM oocytes that extracted their polar body within 35 h of culture were used in this study. All dishes were covered with mineral oil and cultured in an incubator (37°C, 6% CO₂, 5% O₂, 89% N₂).

Sperm handling

The human sperm was donated by one patient with a heterozygous base pair substitution in PLCZ1 (c.136-1G>C). Freshly donated sperm was mixed with cryoprotectant (SpermFreeze, FertiPro, Belgium) and vitrified on the day of donation. The sperm was normally thawed at 37°C just before injecting the oocytes. The content of the straw was diluted in Sydney IVF gamete buffer (GB; Cook Medical). This mixture was centrifuged for 10 min at 1600 rpm. After removal of the supernatant, the sperm pellet was diluted in a small amount of GB.

ICSI and AOA procedure

Sperm cells were immobilized and injected simultaneously with the CRISPR components into the cytoplasm of MII oocytes with an Olympus IX71 microscope. After injection, the oocytes were cultured in CC medium. After a minimum of 25 min, the injected oocytes were exposed to 10 μM ionomycin (Sigma-Aldrich), which was dissolved in CC medium, for 10 min. After exposure, extensive washing in CC medium took place, followed by 25-min culture in CC medium. After this washing step, a second 10-min exposure to ionomycin occurred, which was followed by extensive washing in CC medium. The oocytes were then cultured in CC medium. After 3 days of culture, the embryos were transferred to Sydney IVF blastocyst medium (Cook Medical). After 5 days of culture, the embryos were scored using the Gardner grading system. The oocytes were kept in the incubator (37°C, 6% CO₂, 5% O₂) for culture and after exposure or injections. All manipulation steps took place under mineral oil.

Embryo development follow-up

Because of the fact that we did not work with in vivo-matured oocytes donated for research but were dependent on oocytes donated during the fertility treatment, a lower development rate was expected. To get the most similar situation with the in vivo group, the development number of the GV matured oocytes (which have been known to have a higher rate of cleavage arrest) were excluded. We also only utilized the developmental rates of the embryos that we kept in culture until natural cleavage arrest or until day 5. We did not incorporate the developmental rates of the artificially arrested embryos that we collected on day 3 in the initial stages of the experiments (to ensure genetic stability) or for the individual blastomere experiments (see below).

The CRISPR group was the experimental group that was discussed under “ICSI and AOA procedure.” The in vitro control group underwent the same procedure as the CRISPR group, but no CRISPR components were injected. The in vivo control group numbers were the development numbers of the patient sperm with his partner’s oocytes after AOA (in the clinic). These numbers were expected to be the highest because of the use of in vivo-matured oocytes and the extra addition of 0.1 mol/L CaCl₂ when injecting sperm, which was not included in the other groups because of the added stickiness.

To get insight into the development rate, three rates were calculated. (1) The fertilization rate is the amount of zygotes showing two pronuclei (2PN; normally fertilized zygote) divided by the amount of oocytes surviving the injection. (2) The day 3 development rate is the amount of normally fertilized zygotes that were 4 cells or more on day 3. (3) The day 5 development rate was the amount of normally fertilized zygotes that became blastocysts on day 5.

DNA extraction for genetic analysis

Only embryos that were 4 cells or more on day 3 were analyzed. The DNA was extracted, and whole-genome amplified with a Repli-G Single Cell Kit or Repli-G Advanced DNA Single Cell Kit (QIAGEN, Germany) or PicoPLEX Single Cell WGA Kit V3 (Takara, Japan), as described by manufacturer’s instructions, with the exception of the Repli-G Single Cell Kit, for which the incubation period was generally shortened to 2 h for the whole embryo. The PicoPLEX products were purified with Ampure XP beads (Agencourt Beckman Coulter, CA, USA) to remove residual primers. To collect maternal (oocyte donor) DNA for genetic analysis, we collected the first polar body of the oocytes (before ICSI) and/or the non-matured oocytes and/or the cumulus cells from the same female patient. From these samples, DNA was extracted and underwent WGA with the Repli-G (Advanced DNA) Single Cell Kit or PicoPLEX Single Cell WGA Kit V3. This DNA was stored at −20°C. We want to highlight that we used oocytes donated from multiple different patients, and therefore we needed the maternal DNA from each “donor.” For the paternal DNA (the mutated sperm donor), we used the DNA extracted from the non-manipulated iPSCs or from the renal cells (see iPSC part). For the DNA extraction of sperm, we used the QIAamp DNA Mini Kit (QIAGEN), following the protocol mentioned in the article by Wu et al.⁴⁶ An overview of the specific analyses per sample is given in Figure S7.

Whole-embryo analysis of edited embryos

The paternal DNA was characterized using the analysis techniques explained below. When present, the oocyte donor DNA was also characterized. The results of the only embryo originating from WT sperm showing indels was excluded from the editing rate and STR assay calculations because of the lack of replicates. WGA embryo or individual blastomeres samples that contained a too-low DNA yield were not utilized for further analysis. When the analyses were inconclusive, the samples were not included in the calculations.

The mutation has a heterozygous nature (half of the sperm is WT, and half contains the mutation). Therefore, a characterization was needed to know whether theMUT- or WT-carrying sperm gave rise to the embryo. This identification happened through STR analysis. Primers were designed upstream and downstream of the mutational site of the DNA sequence, spanning a 10-Mb (Table S12, standard STR markers) or 18-Mb (extended STR markers) region. Based on the STR marker data we calculated the amount of paternal markers, which showed the MUT or WT chromosome marker. Because of the occasional PCR errors related to STR assays, multiple STR loci were analyzed per sample to strengthen the conclusions. STR results were compared with maternal DNA when available. When both markers were present but one/both of the markers was/were the same as the mother DNA, we counted them as the majority type of marker. Because not every STR analysis gave a clear and conclusive result in each sample, we implemented the following thresholds on the percentage of STR markers with only one peak to get an idea of the LOH rate: 0%–50% was seen as “no LOH,” and 50%–100% was seen as “complete LOH.” Some special cases were seen as “short-range LOH” (even in the case of <50% single markers) when a specific distance around the cleavage site showed LOH for at least two consecutive markers.

Through targeted NGS (Miseq),^45,47 which relies on a DNA preamplification step by PCR with region-specific primers (Table S10) to amplify a specific region of interest in the genome, we studied a region of approximately 500 bp surrounding the mutation site. Before NGS, the PCR fragments were analyzed with a fragment analyzer (Agilent Technologies), which utilized qualitative 915/930 dsDNA kits (Agilent Technologies), We first performed targeted NGS on the sperm DNA to confirm the heterozygous nature of the sperm. For the embryos, we determined whether the embryo contained only the WT alleles or also the MUT allele (c.136-1G>C) and the presence of the synonymous variant related to the template. We also checked whether additional indels were present. As a threshold for the indel and MUT (c.136-1G>C) rate, we used 5%, which we saw as the maximum background indel rate in non-targeted embryos.⁴⁸ These background errors can occur because of PCR and sequencing errors but also because of WGA.

In a limited number of cases for the embryo (n = 2) and iPSC results (whole well, n = 1; individual colonies, n = 4), the predominant indel contained the mutation site (e.g., gtgagcccaccagtcatttttttag[taCgacaatgacag]gctgaaacaa, with the capital C representing the patient mutation), which made it impossible to determine whether this indel was related to the MUT allele. For clarity and because of the low occurrence rate of these predominant indels, these samples were still classified as indel mutations.

A closer look around the mutation site was undertaken through a SNP assay. After designing multiple primer couples that covered a region of approximately 9,000 bp (Table S13), the DNA from the mutated sperm donor, his parents, and some oocyte “donors” were analyzed. After screening for SNPs present on the DNA of the sperm donor, only the primer pairs that gave rise to SNPs that were informative (Table S14) for the origin of the sperm (MUT or WT) and/or for the occurrence of LOH were tested in the targeted embryos. PCR products were created for each SNP primer pair, which were then sequenced through Sanger sequencing and NGS. A sample was categorized as “no LOH events” when both parental markers were present in all SNPs or when at least the SNPs closest to the mutation site (up- and downstream) displayed both paternal markers. The presence of LOH patterns farther away could potentially be due to ADO.

The samples created with the PicoPLEX kit were analyzed with the GENType pipeline, for which detailed instructions can be found back in De Witte et al.²² The genome-wide haplotypes, SNP profiles, B-allele frequencies (BAFs; a normalized measure of the allelic intensity ratio of the two parental alleles) and copy number profiles were generated by Hopla.²² The CNV profiles were determined by read depth analysis with ViVar.²⁸ Samples with an ADO rate higher than 50% were removed because of insufficient quality. The haplotypes generated with the GENType pipeline confirmed the STR-based WT/MUT classification.

To exclude the presence of large deletions, the PicoPLEX samples were investigated with 1,410-bp (1) and 1,367-bp (2) long-range PCR primers. Two specific primer pairs (Table S10) were designed to analyze a region upstream (Figure S6A) and downstream (Figure S6B) of the mutation site, with the mutation site and two informative SNPs in the amplified sequence. For the upstream design (long-range PCR 1), standard PCR was performed with Kapa2G Robust HotStart ReadyMix (Merck, NJ, USA), while for the downstream design (long-range PCR2) TaKaRa LA Taq DNA polymerase was utilized, following the manufacturer’s instructions. The PCR products were put on the fragment analyzer to check whether the expected size was present.

Single-cell analysis of edited embryos

On day 3 of in vitro development, embryos were dissociated into single cells. The embryos were briefly exposed to pre-warmed Acidic Tyrode solution (Sigma-Aldrich) while pipetting up and down to remove the zona pellucida. Afterward the cells were washed in CC, followed by dissociation in TrypleE (Thermo Fisher Scientific) for a maximum of 5 min, which happened on a glass bottom dish (WillCo-Wells, the Netherlands). The single cells were then individually collected in REPLI-G advanced single-cell storage buffer or in CC medium. The DNA from these single cells was then extracted and underwent WGA with the Repli-G Advanced DNA Single Cell Kit or PicoPLEX Single Cell WGA Kit V3. The PCR fragments and DNA were analyzed the same way as the whole embryos (see above). Mutations with a frequency less than 15% were generally seen as background signal.⁴⁵ The classification of homo-indels (i.e., the presence of the patient mutation with a MUT allele in almost all reads) was added for the PicoPLEX samples (2’C, 3’B and 3’G). Theoretically, blastomere 8C of the Repli-G samples could have been classified as “homo-indel,” but because of the missing SNP analysis data, classification of homo-indels was only utilized for the PicoPLEX samples.

Statistics

Statistics were executed with SPSS Statistics (IBM) or GraphPad Prism. A two-sided independent t test was performed for comparisons in Figure 1G. A Kaplan-Meier plot was employed to reflect the developmental rates in Figure 2B. A p value less than 0.05 was considered statistically significant.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. Clinical data of our patients will only be shared in a way that ensures that the patient cannot be identified.