The study was approved by the Ethics Committee of the Reproductive and Genetic Hospital of CITIC-Xiangya.

We enrolled 2,253 couples with autosomal reciprocal translocations, whose blastocysts were tested using NGS, from October 2013 to December 2019 from the Reproductive and Genetic Hospital of CITIC-Xiangya. G-banding chromosome analysis of cultured lymphocytes from peripheral blood was performed at our hospital or other certified organizations. The karyotypes in this report are described according to the ISCN 2013 guidelines. All individuals had a history of infertility, recurrent spontaneous abortion, or pregnancy with chromosomal anomalies.

Preimplantation genetic testing for chromosomal structural rearrangement (PGT-SR) was performed at the Reproductive & Genetic Hospital of CITIC-Xiangya, as previously described. In brief, ovarian stimulation was conducted based on the patient’s status. Fertilization was assessed at 16–18 h post-intracytoplasmic sperm injection (ICSI) by visualization of two pronuclei and two polar bodies. All embryos were cultured in sequential media (G1 and G2; Vitrolife) to the blastocyst stage under 6% CO₂, 5% O₂, and 89% N₂ in a Cook mini-incubator. Trophectoderm (TE) biopsy was performed on day 5 or day 6, and approximately five cells were collected after zona pellucida dissection and hatching using a laser (ZILOS-tk, Hamilton Thorne). The blastocysts were then cryopreserved as previously described (Zhang et al., 2016).

Two NGS platforms were used in this study. For the initial 2,345 trophectoderm cells (from October 2013 to May 2015), whole-genome amplification (WGA) was performed using the WGA4 GenomePlex Single Cell Whole Genome Amplification kit (Sigma-Aldrich, St. Louis, MO, United States). The Illumina HiSeq2000 platform was used for NGS, and approximately 5–10 million single-end (SE) 50 bp reads were obtained for each sample. For the remaining 8,501 trophectoderm cells, WGA was performed using the REPLI-g Single Cell Kit (Qiagen, Hilden, Germany). The Illumina NextSeq platform was used for NGS, and a 10 million SE 75 bp reads were obtained for each sample. Copy number variations larger than 1–4 megabase (>1 Mb) could be detected (Tan et al., 2014; Zhou et al., 2018). The NGS-PGT platform results were confirmed using FISH or SNP arrays before clinical application, as described in our previous studies (Tan et al., 2013; Zhang et al., 2013).

One or two cryopreserved embryos could be transferred, and only euploid blastocysts were selected for transfer. No blastocyst transfer was performed with an endometrium of < 7 mm thickness.

The distribution of patient demographics and cytogenetic characteristics was analyzed using the Kolmogorov–Smirnov test. Continuous variables were expressed as mean ± standard deviation (SD). Categorical variables were described as the frequency and percentage. One-way analysis of variance (ANOVA) or Wilcoxon rank-sum test (alternate segregation group versus non-alternate segregation group) or Kruskal–Wallis H test (alternate segregation group versus adjacent-1 segregation group versus adjacent-2 segregation group versus 3:1 segregation group versus ND group) was used for comparison according to the distribution of continuous variables; chi-square or Fisher’s exact tests were used to compare differences between categorical variables.

To evaluate associations of the patient demographics and cytogenetic characteristics with the embryo separation patterns, we utilized a multivariate generalized linear mixed model (GLMM) with random intercepts to account for multiple embryos from the same woman. A binomial distribution with a logit link function (alternate (balanced) segregation group versus unbalanced segregation group) or a multinomial distribution with a logit link function (alternate segregation group versus adjacent-1 segregation group versus adjacent-2 segregation group versus 3:1 segregation group versus not determined (ND) group) was specified for the embryo segregation mode. Covariates with statistically significant differences in univariate GLMM were considered for inclusion in the multivariate GLMM. The following covariates were considered for inclusion in the univariate GLMM: carrier sex, age, and chromosomes involved in the translocation (chromosomes #1 and #2 corresponding to the small and large chromosomes, respectively, according to the guidelines of the International System for Human Cytogenetic Nomenclature (ISCN) 2016). In addition, the inclusion of an Acr-ch, size of the involved chromosomes, size of the translocated segments (TS1 and TS2), size of the centric segments (CS1 and CS2), ratio of the longest and shortest translocated segments (TSR), ratio of the longest and shortest centric segments (CSR), and ratio of the translocated segment over the chromosome arm (TAR) were analyzed. The location of breakpoints was identified by the PGT-SR results of the karyotype; the length of both centric and translocated segments (CS and TS) was measured according to the copy number variations of the unbalanced embryos. TSR, CSR, and TAR were then calculated separately.

A receiver operating characteristic (ROC) curve was used to discriminate the predictive values of the patient demographics and clinical variables in a multivariate GLMM for the embryo separation mode (balanced group versus unbalanced group).

The Cochran–Mantel–Haenszel (CMH) test was utilized to explore the interaction effect of carrier sex and the involvement of an Acr-ch on the likelihood of every single embryo.

All statistical analyses were performed using SPSS 25.0 software (IBM, Armonk, NY, United States), and a two-sided p-value < 0.05 was considered statistically significant.

This study included 1,028 male and 1,225 female autosomal reciprocal translocation carriers. There were no significant differences in demographic characteristics between female and male carriers. In total, 31,295 oocytes originating from 2,871 oocyte retrieval cycles underwent ICSI, and 12,328 (39.4%) embryos were developed to blastocysts on days 5 or 6. Among 12,328 TE-biopsied embryos, 10,876 (88.2%) were tested and 10,846 were diagnosed successfully (Table 1); the percentage of normal/balanced blastocysts was 34.3% (3,716/10,846). The clinical pregnancy rate and delivery rate in the male carrier group were significantly higher than those in the female carrier group (p < 0.05) (Table 1).

The segregation patterns were determined using PGT-SR results from translocation carrier individuals. Overall, among 10,864 diagnosed blastocysts, 2:2 segregation was observed in 9,751 (90.0%) blastocysts, of which alternate segregation was found as the most frequent segregation pattern (46.1%, 5,005/10,846), followed by adjacent-1 segregation (31.3%, 3,397/10,846) and adjacent-2 segregation (12.4%, 1,349/10,846). The details (karyotype and segregation pattern) are shown in Supplementary Table S1. The frequencies of 3:1 segregation and 4:0/other segregation patterns were 5.4% (585/10,846) and 4.7% (510/10,864), respectively. Among all 4:0/other segregation products, only four 4:0 products were identified, and most were classified as undetermined segregation products; therefore, we concentrated on the 2:2 and 3:1 segregation pattern.

As only alternate segregation could produce normal or balanced gametes, we first analyzed the factors that could affect the proportion of alternate segregation. The blastocysts were divided into balanced (alternate segregation) and unbalanced (non-alternate segregation) groups according to their segregation patterns. The results of the univariate and multivariate GLMMs are presented in Table 2. All variables with p < 0.05 were used for multivariate analysis, including the length of the involved chromosomes, TAR1, CS2, carrier sex, and involvement of an Acr-ch. The multivariate analysis results indicated that TAR, carrier sex, and involvement of an Acr-ch could significantly influence the proportion of alternate segregation. In particular, 1) increased TAR1 reduced the non-alternate segregation ratio (odds ratio (OR), 0.806 [95% confidence interval (CI), 0.696, 0.933]; p = 0.004), female sex (OR, 1.293 [95% CI, 1.187, 1.408]; p < 0.001), and involvement of an Acr-ch (OR, 1.208 [95% CI, 1.089, 1.339]; p < 0.001) enhanced the proportion of non-alternate segregation. Furthermore, a ROC curve was used to evaluate the predictive values of alternate segregation. The values of area under the curve (AUC), sensitivity, and specificity were 0.708, 84.4%, and 40.3%, respectively (Supplementary Figure S1).

The CMH test was employed to further assess the effects of a carrier’s sex and involvement of an Acr-ch on the likelihood of each single embryo in an unbalanced rearrangement. The results were summarized as follows (Table 3 and Figure 1A): 1) the proportion of blastocysts carrying an unbalanced rearrangement was higher in the female group than that in the male group, both in carriers with an Acr-ch [62% versus 51.1%, p < 0.001, OR = 1.561, 95% CI (1.374, 1.775)] and without an Acr-ch [53.8% versus 49.8%, p = 0.001, OR = 1.176, 95% CI (1.07, 1.293)]. 2) The risk of an embryo with an unbalanced rearrangement was higher in carriers with an Acr-ch than in those without an Acr-ch in the female group [62e % versus 53.8%, p < 0.001, OR = 1.403, 95% CI (1.26, 1.563)], rather than in the male group (51.1% versus 49.8%, p = 0.353). The data were further stratified according to the chromosomes involved; in the Acr-ch group, the proportion of alternate segregation in female carriers, involving chr-2, -4, -6, -9, -10, and -20, was significantly lower than that in the corresponding male carriers (p < 0.05). However, in the non-Acr-ch group, the proportion of alternate segregation in female carriers, involving chr-5, -8, -9, and -12 (p < 0.05), was significantly lower than that in the corresponding male carriers (Figure 1B and Supplementary Table S2).

We identified several factors that could influence the proportions of adjacent-1, adjacent-2, and 3:1 segregation patterns (Table 4). In particular, the proportion of adjacent-1 segregation was affected by the TAR1, and a greater TAR1 reduced the proportion of adjacent-1 segregation [OR = 0.316, 95% CI (0.174, 0.575), p < 0.001]. For making simplification, we divided the TAR1 into TAR1 ≤ 0.2 group and TAR1 > 0.2 group. The results indicated that the proportion of adjacent-1 segregation was significantly higher in TAR1 ≤ 0.2 group than in TAR1 > 0.2 group (36.2% versus 30.2%, p < 0.001) and both in female subgroup (36.8% versus 29.2%, p < 0.001) and male subgroup (35.4% versus 31.5%, p < 0.05) (Figure 1C and Supplementary Table S3). The proportion of adjacent-2 segregation was influenced by the TS1 and carrier’s sex, and longer TS1 and female sex increased the proportion of adjacent-2 segregation [OR = 1.014, 95% CI(1.003, 1.024), p = 0.009 and OR = 1.66, 95% CI(1.427, 1.931), p < 0.001, respectively] (Figure 1D and Supplementary Table S4); besides, the proportion of 3:1 segregation was affected by a carrier’s sex and involvement of an Acr-ch, wherein female sex and involvement of an Acr-ch increased the proportion of 3:1 segregation [OR = 2.643, 95% CI (2.147, 3.253), p < 0.001 and OR = 1.387, 95% CI (1.075, 1.789), p = 0.012, respectively] (Table 4; Figure 1E and Supplementary Table S5).