B6D2F1 (C57BL/6 × DBA/2), C57BL/6, and ICR background mice were purchased from Beijing Vital River Laboratory Animal Technologies Co., Ltd., (Beijing, China). All mice were housed in the Laboratory Animal Center of the First Hospital of Jilin University. The light period was 8 am to 8 pm, and mice were provided ad libitum access to food and water.

Pregnant mare serum gonadotropin (PMSG) (7.5 IU; Ningbo Second Hormone Company, Ningbo, China) was injected into 6- to 10-week-old female B6D2F1 mice. After 48 h, 7.5 IU human chorionic gonadotropin (HCG) (Ningbo Second Hormone Company) was injected into the same mice. Then, 14 h after the HCG injection, mice were euthanized, and oocyte corona cumulus complexes (OCCOs) were collected from the ampulla of the fallopian tube. OCCOs were placed in an M2 medium (SIGMA, United States) containing 0.1% hyaluronidase to remove granular cells. The prepared eggs were placed in KSOMaa solution (Caisson Labs, Smithfield, UT, United States) and incubated for at least 8 h in an incubator (37.0°C, 5% CO₂). The eggs were placed in an incubator for no more than 3 h before injection.

Sexually mature male C57BL/6 mice were used to obtain round spermatids and spermatozoa. Round spermatids and mature sperm were collected from the testis and epididymis, respectively. Fluorescence-activated cell sorting (FACS) was used to distinguish round spermatids from other cell types. First, part of the seminiferous tubule was broken into pieces and sections using a 1 mL syringe. The suspensions were filtered through a 400-mesh screen (38 μm) and stained with Hoechst 33342 (Beyotime Biotechnology, Haimen, China) at 37.0°C for 10 min. The round spermatids were then isolated using a BD Influx flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, United States).

ICSI and ROSI procedures were consistent with previously described methods (Kimura and Yanagimachi, 1995b), with a few differences. We used a PiezoXpert device (Eppendorf, Hamburg, Germany). For ICSI, the inner diameter of the injection needle was approximately 9–10 μm, while for ROSI, it was approximately 6–7 μm. Breaking the zona pellucida and egg membrane require different piezo strengths (intensity = 5 and speed = 15 or intensity = 1 and speed = 1, respectively). For ICSI, the head of the spermatozoa was separated from the tail by applying 1 Piezo pulse (intensity = 5 and speed = 15) to the head-tail junction. The head was subsequently microinjected into an oocyte. For ROSI, eggs need assisted activation before ROSI (see the next section for details). Round spermatids were identified by flow cytometry. Because the inner diameter of the injection needle was small, when a round spermatid was drawn in, the nucleus was separated from the cytoplasm and subsequently microinjected. About ten eggs were injected per 15 min, and the injected oocytes were cultured in balanced KSOMaa solution for at least 8 h in an incubator (37.0°C, 5% CO₂).

Because the round spermatids cannot activate the egg by themselves, oocytes were first placed in a Ca²⁺-free CZB medium containing 10 mM SrCl₂ for 20 min before injection for manual activation.

Embryos generated using natural in vivo fertilization were used as the control group. Male C57BL/6 mice were mated with female B6D2F1 at a ratio of 1:2 in the afternoon. The next morning, the vaginal plug was examined. Mice with a vaginal plug were euthanized, and fertilized eggs were collected and cultured to the 2-cell stage.

Six hours after injection, embryos with a second polar body and male pronucleus formation were classified and further cultured in the ICSI and ROSI groups. The cleavage rate and blastocyst formation conditions were calculated 24 and 96 h after injection, respectively.

Embryo transfer was performed as previously described (Wang et al., 2020).

Fetuses with pronounced defects were excluded from transcriptome and DNA methylation sequencing analyses. In clinical practice, when fetuses with noticeable defects are identified, the pregnancy is terminated. Eight fetuses and their corresponding placentas identified as morphologically normal were selected for sequencing (No observations of male and female pronuclei formation). Four ICSI-fertilized fetuses and their corresponding placentas served as the ICSI group, and four natural in vivo fertilized fetuses and their corresponding placentas served as the control group. Sequencing was performed at Berry Genomics Corporation (Beijing, China).

RNA extraction and qRT-PCR were performed as previously described (Xu et al., 2017). qRT-PCR was performed using an Applied Biosystems QuantStudio 6 Flex (Foster City, CA, United States). Relative gene expression was analyzed using the 2^–ΔΔCt method with GAPDH as the internal control. All primers are listed in Supplementary Table 1.

Chi-squared tests were used to compare in vitro and in vivo developmental efficiency (IBM SPSS 23). Other statistical analyses were performed using GraphPad Prism (GraphPad). Student’s t-tests were used for pairwise comparisons of other parameters. Differences were considered significant at P < 0.05.

Given that ROSI embryos showed diverse pronucleus formation phenotypes, we chose only embryos with male and female pronuclei (2PN) for analyses. In our culture conditions, the cleavage rates of embryos derived upon fertilization with ROSI and ICSI were 100.00% (Supplementary Figure S1A). Unexpectedly, the blastocyst percentage was higher in the ROSI group than that in the ICSI group (88.79% vs. 75.19%, P < 0.01) (Supplementary Figure S1B), which may be explained by the smaller diameter of the injection needle used for ROSI (6–7 μm) than for ICSI (9–10 μm). Secondly, as the microdroplets for ICSI–but not ROSI–contain polyvinylpyrrolidone (PVP), a small amount of PVP will inevitably be injected into the cytoplasm during ICSI. The third reason may be that round sperm injection requires pre-oocyte activation, and mature sperm have the ability to activate eggs.

Next, we investigated the in vivo development of embryos derived upon using the two methods. Embryos were collected at E11.5, and the survival rate was determined. There was no significant difference in the E11.5 survival rate between the ROSI and ICSI groups (51.61% vs. 55.00%, P > 0.05) (Supplementary Figure S1C). The birth rate (i.e., the ability of embryos to develop to term) was lower for ROSI than that for ICSI (22.86% vs. 50.00%, P < 0.05) (Supplementary Figure S1D). The developmental data for ROSI- and ICSI-derived embryos are summarized in Table 1.

The developmental efficiency gap of the ROSI and ICSI groups increased from 3.39% at E11.5 (55.00% minus 51.61%) to 27.14% at the developmental maturity stage (50.00% minus 22.86%), indicating that some fetuses were lost after E11.5. This phenomenon prompted us to explore the mechanism underlying the difference in the fetus and placenta between embryos from round spermatids or mature sperm at E11.5. We randomly collected eight E11.5 whole fetuses and corresponding whole placentas from the ROSI group as experimental objects (Supplementary Figure S1E). The crown-rump length (CRL) of the fetuses is 6.23 ± 0.68 cm (6.44, 6.64, 6.36, 7.15, 6.09, 6.55, 5.68, and 4.91, respectively); and the maximum diameter of the placenta is 5.08 ± 0.50 cm (5.56, 4.91, 4.59, 5.70, 5.31, 5.42, 4.85, and 4.29, respectively). All fetuses and placenta have no apparent abnormalities. Four ICSI and in vivo fertilized E11.5 fetuses and corresponding placentas were used as control objects (data not shown). Transcriptome and methylation levels of the fetus and placenta were compared among the three groups.

RNA-sequencing was performed to analyze the fetal and placental transcriptomes. Unsupervised hierarchical clustering analysis was used to compare the transcriptomes of all 16 fetuses and 16 placentas (Figure 1A). As expected, the fetuses and placentas were clearly separated. However, there was no distinct clustering of the fetuses in control, ICSI, and ROSI groups. Similar results were obtained for the 16 placentas, indicating that the overall gene expression patterns were not clearly distinguishable among the three groups. DEGs in pairwise comparisons were further evaluated. In total, nine DEGs were identified between control and ROSI fetuses and 18 DEGs between the ICSI and ROSI fetuses, and seven DEGs were shared in both comparisons (Figure 1B). However, for placentas, 64 DEGs were identified between the control and ROSI groups, 29 DEGs between the ICSI and ROSI groups, and 10 DEGs were shared in the two comparisons (Figure 1C). Heatmap visualization revealed seven and ten overlapping DEGs among the groups in the fetuses (Figure 1D) and placentas (Figure 1E), respectively. Next, low-throughput qRT-PCR was performed for randomly selected DEGs to verify the RNA-seq results (Figures 1F–I). Especially, Maternally expressed gene 3 (Meg3), which is located on the murine Dlk1-Dio3 locus, was clearly differentially expressed in ROSI fetuses compared to control or ICSI groups. Seven genes in the Dlk1-Dio3 locus were detected with expression in E11.5 fetuses and placentas. Except for Meg3, Mirg and Rian, which expressed from maternal allele also were up-regulated in ROSI fetuses or placentas but less than 2-fold changes; on the contrary, Dio3, which expressed from the paternal allele, was down-regulated in ROSI placentas (Supplementary Figure S2).

Repetitive DNA sequences (RS) account for more than 50% of the mammalian genomic DNA (de Koning et al., 2011). However, little is known about their specific functions. The expression of RS in the three groups was assessed, indicating that several RS were differentially expressed in ROSI fetuses (Figure 2A), including 22 long interspersed nuclear elements (LINEs), four short interspersed nuclear elements (SINEs), and 25 long terminal repeats (LTRs). However, only five differentially expressed RS were identified in placentas, most of which were LTRs (Figure 2B). Notably, consistent RS results were obtained by qRT-PCR, thus validating the RNA-seq data (Figures 2C–F).

Given that imprinted genes play an essential role in embryogenesis (Ferguson-Smith and Bourc’his, 2018; Tucci et al., 2019), we performed unsupervised hierarchical clustering analysis of all imprinted genes in the three groups. We did not detect biased clustering, indicating that the expression of imprinted genes did not differ significantly among the fetal (Supplementary Figure S3A) or placental (Supplementary Figure S3B) groups. Next, several important imprinting genes, including H19 and Igf2, were selected for validation by qRT-PCR. The qRT-PCR results were consistent with the RNA-seq data, supporting the reliability of our results (Supplementary Figure S4).

Given that a high abortion rate is a major barrier to practical ROSI implementation (Tanaka et al., 2018), we further analyzed the expression of seven well-studied abortion-related genes in the three groups. Vegfa, Ccl2, and Il1b showed elevated expression in partial fetuses or placentas (Supplementary Figures S5A,B). Moreover, Ccl2 was highly expressed in the placenta (Supplementary Figure S5C), while Vegfa showed high expression in fetuses (Supplementary Figure S5D).

Given that epigenetic modification, such as DNA methylation, plays a vital role in the regulation of gene expression and disease (Greenberg and Bourc’his, 2019), we next evaluated the fetal and placental DNA methylation status. WGBS was used for DNA methylation level detection. The whole genome DNA methylation levels were evaluated with CpG sites covered by at least 10 reads. Notably, the DNA methylation status of all CpG sites were similar in the fetuses and placentas of all three groups (Figure 3A). In particular, more than 50% of the CpG sites in the fetuses were hypermethylated (>80%), while most of the CpG sites in the placentas showed intermediate methylation levels (20–80%) (Figure 3A). Furthermore, we found that in both fetuses and placentas, transcription start site (TSS) regions were hypomethylated, and transcription termination site (TTS) regions were hypermethylated (Figures 3B,C). Methylation patterns of the promoter regions of DEGs between ROSI and control or ICSI groups also were not detected with significant changes (Supplementary Figure S6). As described above, RS showed differential expression profiles among the three groups, which prompted us to assess whether this could be explained by a difference in the DNA methylation status. We divided RS into different types according to RepeatMasker annotation. Notably, we observed similar methylation patterns in LINEs (Figure 3D), SINEs (Figure 3E), and LTRs (Figure 3F) among the three groups. Further, RS elements were hypermethylated in the fetuses and hypomethylated in the placentas, suggesting that DNA methylation and other factors may act together to regulate RS expression.

Next, the DNA methylation levels of the paternal and maternal ICRs in sperm and oocytes were analyzed. Only confirmed ICRs in a previous publication (Wang et al., 2014) were used for the analysis. Our results revealed similar methylation profiles between the three groups (Figures 3G,H). The expression of imprinted genes is regulated through the methylation of ICRs, and most ICRs are hypermethylated in oocytes and hypomethylated in sperm (Smallwood et al., 2011). In line with the finding that there were no significant differences in the expression of imprinted genes between the three groups, we observed that the ICRs of the fetuses and placentas from the three groups were similarly methylated (Figures 3I,J). Especially, the Meg3 gene located in the Dlk1-Dio3 region was differentially expressed in RNA levels. In contrast, the Dlk1-Dio3 intergenic differentially methylated region (IG-DMR), which regulates the gene expression pattern of the whole region, did not show the different methylation status between the ICSI and ROSI fetuses. Maybe the factors beyond DNA methylation regulated the expression of Meg3.

The de novo identification of DMR was performed by metilene (Juhling et al., 2016), CpG site with more than 10% methylation difference between ICSI and ROSI groups were labeled as differentially methylated CpG site (DMC). More than five continuous labeled DMCs while with less than 500 bp maximum distance between 2 DMCs were recorded as one DMR. In fetal tissues, 119 hypermethylated DMRs and 272 hypomethylated DMRs were identified in the ROSI group compared with the ICSI group (Supplementary Figure S7A). In placental tissues, 665 hypermethylated DMRs and 1519 hypomethylated DMRs were identified in the ROSI group compared with the ICSI group (Supplementary Figure S7B). Checking the expression levels of genes overlapped with these DMRs and we did not detect consistently differentially expressed patterns with DMRs (Supplementary Figures S7C–F).

Given the similarities in the transcriptome and DNA methylation levels among the three groups, we investigated the differences between fetuses and placentas from ROSI. The CRL is a common indicator of fetal quality. Accordingly, we selected fetus no. 8 as an abnormally developed embryo because its CRL was less than 6 mm (the standard size at this period). Fetuses no. 1 and no. 2 with CRLs greater than 6 mm were selected as controls. Thirty-nine genes were down-regulated and twenty-seven genes were up-regulated in fetus no. 8 compared with fetuses no. 1 and no. 2. The down-regulated genes were enriched for functions in embryonic hindgut morphogenesis (Figure 4A). However, in the placental tissues of fetus no. 8, we observed 262 up-regulated genes, which were mainly involved in pathways related to vasoconstriction, cell growth regulation, and vasculogenesis (Figure 4B). The 91 down-regulated genes were primarily enriched for functions in the extracellular matrix organization, estradiol response, and fatty acid transport (Figure 4B). Moreover, 14 differentially expressed RS, including one LINE, two SINEs, and eight LTRs, were identified in fetus no. 8 (Supplementary Figure S8A). However, in placenta no. 8, only four RS, including one LINE and three LTRs, were differentially expressed (Supplementary Figure S8B). With respect to DMRs, with more stringent filtering that 20% methylation difference and 10 continuous DMCs in DMR definition by metilene, we identified 234 hypermethylated and 69 hypomethylated DMRs in fetus no. 8 (Supplementary Figure S9A). In addition, we identified 2185 hypermethylated DMRs and 203 hypomethylated DMRs in the placenta of fetus no. 8 (Supplementary Figure S9B).

We could not identify a relationship between gene expression and DNA methylation in the overall ROSI group. However, in fetus no. 8 and controls (CRL > 6 mm) in the ROSI group, we found a relationship between the methylation status of promoter regions and the expression of Fggy (Figure 5A) and Rec8 (Figure 5B). Additionally, a relationship was identified between the methylation status of promoter regions and the expression of Rec8 (Figure 5C) in placenta no. 8. The promoter regions of these genes were hypermethylated in fetus and placenta no. 8, and accordingly, the mRNA expression of these genes was reduced (Figures 5D–F).