Toward the development and refinement of non-invasive preimplantation genetic testing
The Embgenix ESM Screen Kit was used to perform copy number variant (CNV) analysis on cfDNA from IVF embryo spent medium (ESM) samples from three independent labs and results were compared against conventional PGT‑A from corresponding trophectoderm (TE) biopsy samples
Total chromosome concordance between ESM- and TE-biopsy-based results ranged from ~57–80% for each batch of samples, varying between labs and depending on factors such as day of sample collection
Higher concordance was observed for samples collected on Day 6/7 as compared to samples collected on Day 5 (74% vs. 60%); higher concordance was also observed for embryos not subjected to assisted hatching (75% vs. 70%)
Sex chromosome concordance ranged from ~80–94% across batches; all discordant sex chromosome calls were called female by ESM analysis and male by PGT‑A and overrepresentation of female calls was observed among euploid ESM calls, which is consistent with the effects of maternal DNA contamination
ESM assay results are promising, but further refinements are needed to mitigate the effects of maternal DNA contamination and yield an assay suitable for non-invasive PGT‑A (niPGT‑A)
Over the past decade preimplantation genetic testing for aneuploidy (PGT‑A) via NGS-based analysis of TE-biopsy samples has become a well-established method for assessing the copy number status of IVF embryos and informing prioritization of embryos for implantation (ESHRE PGT-SR/PGT‑A Working Group et al., 2020). TE biopsy-based PGT‑A has helped improve the likelihood of successful pregnancy particularly for in vitro fertilization (IVF) patients of advanced age (Sanders et al., 2021), but the biopsy process carries a few potential limitations and drawbacks; it is a laborious procedure that requires specialized training and equipment and adds significantly to the overall cost of IVF; it is an invasive approach that carries a small (albeit remote) risk of damage to the embryo and has been associated with increased incidence of nuclear DNA loss (Domingo-Muelas et al., 2023); it has been associated with increased rates of preeclampsia (Zhang et al., 2019) and hypertensive disorders (Makhijani et al., 2021); and it samples a subset of cells that give rise to the placenta and may not represent the copy-number status of the embryo as a whole.
Given the limitations and challenges associated with TE biopsy-based PGT‑A, recent efforts have focused on the development of non- or less-invasive approaches (niPGT‑A) that analyze cfDNA released from IVF embryos into culture medium and/or blastocoel fluid in lieu of TE biopsy (Rubio et al., 2019; Cinnioglu et al., 2023). While these efforts have given rise to niPGT‑A screening assays that are now applied in the lab, the assays face their own set of challenges, including insufficient or low-quality cfDNA in the ESM input, maternal DNA contamination—which can obscure the copy number status reflected in the cfDNA released from the embryo—and the impact of upstream process variability on assay concordance with conventional PGT‑A.
To support further research efforts toward robust approaches for niPGT‑A, Takara Bio developed the Embgenix ESM Screen Kit and Embgenix Analysis Software, which enable NGS-based detection of aneuploidies and mosaicisms from cfDNA in IVF embryo culture medium and/or blastocoel fluid. Also included in the kit are reagents for qPCR-based quantitation of cfDNA prior to sample processing. Via collaboration with three independent labs, the Embgenix assay was used to analyze ESM samples from IVF embryos processed under varying conditions and results were compared against PGT‑A calls obtained from corresponding TE biopsy samples from each embryo. Concordance rates between the Embgenix ESM and PGT‑A results varied between labs and sample processing parameters, as presented in the results below.
Methods
Samples and methods
157 embryo spent media (ESM) samples were collected from three different labs in four batches for analysis. Fertilized embryos were cultured until Day 4, when each embryo was washed and transferred to an individual fresh media droplet ranging in volume from 10–75 μl. For embryos in the “assisted hatching” groups, assisted hatching was carried out on either Day 4 (Labs A and B) or Day 3 (Lab C) prior to washing. Embryos included in this study developed into mature blastocysts between Days 5–7. Before TE biopsy, media samples were collected into nuclease-free PCR tubes and stored immediately at –20°C or –80°C. For each sample, care was taken to collect the entire media droplet while minimizing oil carryover. TE biopsies were analyzed according to each site’s established protocol. ESM samples were shipped to Takara Bio on either dry ice or ice packs for analysis using the Embgenix ESM Screen Kit. Assessment of ESM assay performance indicated that sample shipment on dry ice was not associated with higher concordance as compared to shipment on ice packs (data not shown).
Site
Number of samples
Assisted/Non-assisted hatching
Day(s) of collection
Average input volume (μL)
Average input amount (pg)
A
78
Assisted/Non-assisted
6–7
13.8
14.5
B - Batch 1
23
Non-assisted
5–7
9.2
17.3
B - Batch 2
28
Assisted/Non-Assisted
5–7
7.6
29.4
C
28
Assisted
6
22.1
26.7
Table 1. Overview of ESM samples processed, by site. Samples were obtained in four different batches from three different laboratories. Day of sample collection, embryo handling conditions, volume of ESM sample, and quantity of cfDNA processed per sample varied between labs as indicated.
The Embgenix ESM Screen Kit was used to assess the concentration and fragmentation of cfDNA in the media samples and to generate NGS libraries from each sample for sequencing. Libraries were sequenced on an Illumina MiSeq system with the MiSeq Reagent Kit v3 (150-cycle) using 2 x 75 bp paired-end reads and resulting sequencing data were analyzed using Embgenix Analysis Software.
Site
Reads per sample
% of reads ≥ Q30
Cluster density (k/mm2)
A
1609568
95.9
1135
B - Batch 1
1918516
96.4
1276
B - Batch 2
1976525
96.0
1240
C
1281495
95.6
1213
Table 2. Sequencing performance for Illumina libraries generated from ESM samples, by site. Sequencing libraries were prepared from the ESM samples using the Embgenix ESM Screen Kit and sequenced on an Illumina MiSeq System. Sequencing metrics for each batch of samples were averaged across multiple sequencing runs.
Samples that failed automated QC by Embgenix Analysis Software due to insufficient informative reads or noisy results (assessed using the DLRS metric) were excluded from further analysis. Additionally, samples that yielded five or more CNV calls (aneuploidy or mosaicism) or an NME (Noise Metric Euploid) >1000 (considered chaotic results) were manually excluded. NME is a complementary noise metric (calculated in a different manner than DLRS or MAPD) that can be used to assess assay noise from experiment to experiment and across varying conditions such as sampling procedures, sequencing parameters, etc. Lastly, samples for which PGT‑A analysis of the corresponding TE biopsy yielded chaotic results were also excluded.
Figure 2. Example CNV plots generated with Embgenix Analysis Software for samples that were included or excluded in assessing concordance with conventional PGT‑A. Panel A. Result for sample that was included in the concordance assessment and yielded an aneuploid call that was concordant with the PGT‑A result. Panel B. Result for sample that was included and yielded a mosaic call that was concordant with PGT‑A. Panel C. Result for sample that was included and yielded a euploid call that was concordant with PGT‑A. Panel D. Result for sample that was excluded because assessment with Embgenix Analysis Software yielded a DLRS score >0.23. Panel E. Result for sample that was excluded due to calling of ≥5 CNVs. Panel F. Result for sample that was excluded because it yielded an NME score >1000.
Comparison of cfDNA amounts between ESM samples indicated that cfDNA inputs were higher for samples that passed QC and were included in the analysis (18.8 µg average input) as compared to samples that failed QC and were excluded (7.8 µg average input).
Figure 3. Comparison of cfDNA input amounts between ESM samples that were excluded or included in analysis of assay concordance. ESM samples were binned according to their exclusion or inclusion in the downstream analysis and average cfDNA concentrations were calculated.
Classification of results
For each ESM sample included in the analysis, Embgenix Analysis Software provided a sample call interpretation to broadly describe the CNV status of the sample, using the following seven categories listed below in order of priority from high to low (i.e., a sample predicted to include multiple different CNVs was classified according to the CNV with the highest priority):
Sample CNV classfication
Prediction
Aneuploid
Contains at least one full whole chromosomal gain or loss
Segmental aneuploid
Contains at least one full segmental gain or loss
Mosaic high
Contains at least one whole chromosomal mosaicism at ≥50% frequency
Mosaic low
Contains least one whole chromosomal mosaicism at ≥30% frequency
Mosaic segmental high
Contains least one segmental mosaicism at ≥50% frequency
Mosaic segmental low
Contains least one segmental mosaicism at ≥30% frequency
Euploid
Does not contain any CNVs
Table 3. CNV classifications applied by Embgenix Analysis Software. Classification categories are listed from highest to lowest priority. Samples interpreted to have multiple different CNVs are classified according to the CNV with the highest priority.
For the PGT‑A analysis of the corresponding TE biopsy samples, results were categorized as follows (with unknown thresholds) depending on the site:
Lab A: Abnormal, Mosaic, or Normal
Labs B and C: Aneuploid, Mosaic, or Euploid
In addition to the CNV analysis, sex chromosome counts were assessed for each ESM sample and for a subset of the TE biopsy samples.
Determination of assay concordance
Results from the Embgenix ESM analysis were compared to corresponding PGT‑A results provided by each collaborating lab to assess concordance of total chromosome and sex chromosome calls between the respective approaches across labs and in the context of factors such as day of ESM sample collection, assisted vs. non-assisted hatching, and maternal DNA contamination.
Because different sample call classification systems were used for the ESM and TE biopsy (PGT‑A) assays with unknown thresholds in the latter cases, sample call categories were collapsed, and results were reclassified as follows to allow for assessment of total chromosome concordance between assays:
ESM results
Approach 1: all aneuploid and mosaic calls reclassified as Abnormal, all euploid calls reclassified as Normal
Approach 2: all aneuploid and high mosaic calls reclassified as Abnormal, all euploid and low mosaic calls reclassified as Normal
PGT‑A results
All aneuploid/abnormal and mosaic calls reclassified as Abnormal, all euploid calls reclassified as Normal.
The two different reclassification results for the ESM results yielded similar results vs. PGT‑A in terms of total chromosome concordance (71% overall concordance in each case), so only the results generated using Approach 1 are presented below.
Results
Assay concordance across labs
ESM assay results were compared against PGT-A results from corresponding TE biopsy samples for assessment of total chromosome and sex chromosome concordance across different collaborating labs and sample handling protocols. Total chromosome concordance was 71% overall, and ranged from 57–80% between batches. Concordance of sex chromosome calls was 84% overall, ranging from 80–94% between batches.
Site
Total chr concordance
Sex chr concordance
Sensitivity
Specificity
PPV
NPV
Overall
71% (n=120)
84% (n=117)
78%
65%
67%
75%
A
73% (n=66)
82% (n=66)
77%
67%
77%
67%
B—Batch 1
57% (n=23)
85% (n=20)
86%
44%
40%
88%
B—Batch 2
75% (n=16)
94% (n=16)
50%
90%
75%
75%
C
80% (n=15)
80% (n=15)
100%
67%
67%
100%
Table 4. Assessment of ESM assay performance relative to conventional PGT-A across collaborating labs. ESM and PGT-A results that both yielded an Abnormal (aneuploid/abnormal/mosaic) or Normal (euploid) call were classified as concordant (“total chr concordance”) as were results that both yielded a male or female call (“sex chr concordance”). Sensitivity = (number of concordant abnormal calls)/(number of concordant abnormal calls + number of discordant normal ESM calls); specificity = (number of concordant normal calls)/(number of concordant normal calls + number of discordant abnormal ESM calls); positive predictive value (PPV) = (number of concordant abnormal calls)/(number of concordant abnormal calls + number of discordant abnormal ESM calls); negative predictive value (NPV) = (number of concordant normal calls)/(number of concordant normal calls + number of discordant normal ESM calls).
Assay concordance by day of sample collection
ESM and PGT‑A assay results from all labs were binned into two categories according to the day of ESM sample collection (Day 5 vs. Day 6/7). As compared to analysis of ESM samples collected on Day 5, ESM sample collection on Day 6/7 was associated with higher concordance of total chromosome calls (74% vs. 60%) and higher concordance of sex chromosome calls (89% vs. 64%). ESM sample collection on Day 6/7 was also associated with higher inputs of cfDNA as compared to sample collection on Day 5 (21.8 pg vs. 7.4 pg average inputs, respectively).
Day of ESM sample collection
Average ESM DNA conc. (pg/μl)
Average ESM DNA input (pg)
Total chr concordance
Sex chr concordance
Sensitivity
Specificity
PPV
NPV
Overall
2.2
18.8
71% (n=120)
84% (n=117)
78%
65%
67%
75%
Day 6/7
2.6
21.8
74% (n=95)
89% (n=92)
85%
63%
68%
82%
Day 5
0.6
7.4
60% (n=25)
64% (n=25)
50%
69%
60%
60%
Table 5. Assessment of ESM assay performance based on timing of sample collection. Concordance of ESM assay and PGT‑A results was assessed for ESM samples collected at varying time points following in vitro fertilization. ESM and PGT‑A results that both yielded an abnormal (aneuploid/mosaic) or normal (euploid) call were classified as concordant (“total chr concordance”) as were results that both yielded a male or female call (“sex chr concordance”). Average concentrations and amounts of cfDNA in the ESM samples were also determined.
Assay concordance across embryo culturing conditions
Results for ESM samples collected on Day 6/7 were further stratified depending on whether the corresponding embryos were subjected to assisted hatching. ESM assay concordance with PGT‑A was higher for samples from non-assisted hatching embryos as compared to embryos subjected to assisted hatching: 75% vs. 70% total chromosome concordance, 97% vs. 82% sex chromosome concordance.
Culture condition (day 6/7 samples)
Number of samples
Total chr concordance
Sex chr concordance
Sensitivity
Specificity
PPV
NPV
Overall
90
72%
89%
84%
61%
67%
80%
Assisted
50
70%
82%
88%
54%
64%
82%
Non-assisted
40
75%
97%
80%
70%
73%
78%
Table 6. Assessment of ESM assay performance based on embryo culturing conditions. Concordance of ESM and PGT‑A results was compared for ESM samples collected on Day 6 or 7 depending on whether embryos were subjected to assisted hatching. ESM and PGT‑A results that both yielded an abnormal (aneuploid/mosaic) or normal (euploid) call were classified as concordant (“total chr concordance”) as were results that both yielded a male or female call (“sex chr concordance”).
Assessment of maternal DNA contamination effects
The presence of maternal DNA in ESM samples is recognized as a major challenge facing the development of non-invasive PGT‑A assays, as it can obscure the detection of aneuploidies or mosaicisms from embryo DNA and/or lead to miscalling of sex chromosomes (Vera-Rodriguez et al., 2018). Consistent with these possibilities, assessment of the ESM data revealed that out of 19 instances where the ESM assay yielded a discordant sex chromosome call vs. PGT‑A, 100% involved scenarios where the ESM call was female and the PGT‑A call was male. Similarly, out of 52 samples that were called euploid by ESM analysis, 37 (71%) were called as female, an overrepresentation of female calls as compared to the ~50% frequency of females in the population at large.
Figure 4. Discordant ESM assay sex chromosome calls due to maternal DNA contamination. Presence of maternal DNA in ESM samples may be associated with a bias towards female sex chromosome calls discordant with PGT‑A results.
Discussion
Assessment of Embgenix ESM assay concordance with conventional PGT‑A yielded results comparable to previous niPGT‑A studies (Liu et al., 2017; Rubio et al., 2019; Chen et al., 2021). The variability in concordance observed between laboratories is not surprising given that ESM assay performance was shown to be dependent on embryo culturing and sample processing parameters that vary according to each lab’s established protocols. A noteworthy limitation in this study was the lack of consistency among the PGT‑A assays employed by each site (e.g., varying thresholds for mosaic and aneuploid CNV calls). To overcome this challenge and enable comparison across labs, ESM and PGT‑A CNV calls were reclassified using two different binary frameworks that ultimately yielded similar results.
Consistent with previous studies (Hammond et al., 2017; Rubio et al., 2019), a positive correlation was observed between day of ESM sample collection, cfDNA amount, and assay concordance. In addition to increasing the likelihood that sufficient cfDNA input material is present to minimize assay noise, satisfy assay QC criteria, and enable accurate CNV calling (the Embgenix ESM protocol specifies a minimum input of 2 pg), ESM sample collection at later timepoints also mitigates the effects of maternal DNA contamination by allowing for a higher ratio of embryonic to maternal DNA in the input (Lane et al., 2017; Vera-Rodriguez et al., 2018). While optimizing the timing of ESM sample collection is an important step towards developing a robust approach for niPGT‑A, it also poses a practical challenge given that a corresponding adjustment in the timing of embryo implantation (e.g., from Day 5 to Day 6) may require development and validation of new protocols.
In contrast with the relationship observed between day of ESM sample collection and assay concordance, the impact of assisted hatching was less pronounced, though higher concordance was observed for embryos not subjected to assisted hatching. This result is consistent with previous findings (Ho et al., 2018) which demonstrated slightly higher concordance for non-assisted hatching embryos and dispelled the notion that assisted hatching promotes release and greater abundance of cfDNA. These results are noteworthy because avoidance of assisted hatching aligns more closely with the core objectives of non-invasive screening approaches.
Beyond the technical challenges associated with obtaining NGS libraries of sufficient quality for CNV analysis from highly variable picogram quantities of cfDNA, the prospect that ~90% of ESM samples carry some level of maternal DNA contamination (Vera-Rodriguez et al., 2018) may be the biggest hurdle preventing wider application of niPGT‑A assays. While direct detection and quantitation of maternal DNA was beyond the scope of this study, the effect of maternal DNA contamination was observed in the skew towards female sex chromosome calls in the ESM results. Embryo handling and sample processing conditions (e.g., timing of ESM collection, removal of cumulus cells) can help mitigate the impacts of maternal DNA contamination, but they may not be sufficient to yield an niPGT‑A assay suitable for broad clinical application. To this end, methods for identifying samples impacted by maternal DNA contamination might be required. Such methods could allow for exclusion of impacted samples from further consideration or, ideally, exclusion of data attributed to maternal DNA.
In summary, the results of this multi-center study establish the Embgenix ESM Screen Kit as a promising starting point for niPGT‑A assay development. Fulfillment of this objective will require further tightening and validation of embryo culturing and sample processing procedures and perhaps an overlaying technology for discrimination of maternal DNA contamination. Given the tremendous potential of a non-invasive screening approach to extend the accessibility of PGT‑A and its benefits to a wider population, it is clearly a goal worth pursuing further.
References
Cinnioglu, C. et al. A systematic review of noninvasive preimplantation genetic testing for aneuploidy. Fertility and Sterility120 (2), 235–39 (2023).
Coonen, E. et al. ESHRE PGT Consortium good practice recommendations for the detection of structural and numerical chromosomal aberrations. Human Reproduction Open2020 (3), (2020).
Domingo-Muelas, A. Human embryo live imaging reveals nuclear DNA shedding during blastocyst expansion and biopsy. Cell186 (15), 3166-3181.e18 (2023).
Hammond, E. et al. Characterizing nuclear and mitochondrial DNA in spent embryo culture media: genetic contamination identified. Fertility and Sterility107 (1), 220-228.e5 (2016).
Ho, J. et al. Pushing the limits of detection: investigation of cell-free DNA for aneuploidy screening in embryos. Fertility and Sterility110 (3), 467-475.e2 (2018).
Lane, M. et al. Ability to detect aneuploidy from cell free DNA collected from media is dependent on the stage of development of the embryo. Fertility and Sterility108 (3), e61 (2017).
Liu, WQ et al. Non-invasive pre-implantation aneuploidy screening and diagnosis of beta thalassemia IVSII654 mutation using spent embryo culture medium. Annals of Medicine49 (4), 319–28 (2017).
Makhijani, R. et al. Impact of trophectoderm biopsy on obstetric and perinatal outcomes following frozen-thawed embryo transfer cycles. Human Reproduction36 (2), 340–48 (2021).
Rubio, C. Embryonic Cell-free DNA versus trophectoderm biopsy for aneuploidy testing: concordance rate and clinical implications. Fertility and Sterility112 (3), 510–19 (2019).
Sanders, K. et al. Analysis of IVF live birth outcomes with and without preimplantation genetic testing for aneuploidy (PGT‑A): UK Human Fertilisation and Embryology Authority data collection 2016–2018. Journal of Assisted Reproduction and Genetics 38 (12), 3277–85 (2021).
Vera-Rodriguez, M. et al. Origin and composition of cell-free DNA in spent medium from human embryo culture during preimplantation development. Human Reproduction33 (4), 745–56 (2018).
Zhang, W. et al. Maternal and neonatal outcomes associated with trophectoderm biopsy. Fertility and Sterility112 (2), 283-290.e2 (2019).
