Compromised global embryonic transcriptome associated with advanced maternal age

Surplus, cryopreserved, transferrable quality (grade ≥ 3BB) human blastocysts were fertilized using intracytoplasmic sperm injection and sequentially cultured under low oxygen conditions prior to being donated to research with Institutional Review Board approval and patient consent: young, oocyte donor control with no male factor infertility (DC < 30 years; n = 12 from 6 different patients), and advanced maternal age with no male factor or other significant female factor infertility (AMA ≥ 42 years old; n = 12 from 12 different patients). Blastocysts were identified by PGT-A to be void of autosomal chromosomal aneuploidies and were warmed following previously published vitrification protocols [24]. All fathers were ≤ 48 years of age and therefore not considered to be advanced paternal age.

RNA was isolated from individual blastocysts using the PicoPure™ RNA Isolation Kit (Applied Biosystems, Foster City, CA) with minor modifications to the manufacturer’s protocol. Briefly, blastocysts were lysed at 42 °C for 30 min in 10 μl of extraction buffer. One volume of 70% EtOH was mixed with each sample prior to loading onto a pre-conditioned purification column. Each sample was on-column deoxyribonuclease treated at room temperature for 15 min (Qiagen, Germantown, MD). After several washes, RNA was eluted in 20 μl of elution buffer.

The entirety of the purified RNA from each blastocyst (n = 12, 6 from each group) was utilized to prepare sequencing libraries using the SMARTer® Ultra® Low Kit (Clontech Laboratories, Fremont, CA) following manufacturer’s instructions and then sequenced on the HiSEQ 4000 (Illumina, San Diego, CA) as single pass 50 bp reads. Derived sequences were processed by performing quality checks and normalization on each read and then mapping them to the human genome (build GRCh38) using Trimmomatic v0.36 [25], GMAP-GSNAP v2014-12-17 [26], SAMtools v1.5 [27], and Cufflinks v2.2.1 [28]. Transcripts with no reads across all samples, as well as transcripts in the bottom quintile based on mean expression across all samples, were excluded from analysis. A log transformation was performed prior to statistical analysis which included a 2-sample, 2-sided, independent Student’s t test (significance at Q < 0.05). The false discovery rate was then adjusted using the Benjamini-Hochberg procedure and the expression log ratio was used for each differentially expressed transcript. Pathway analysis was performed on differentially expressed transcripts using Ingenuity Pathway Analysis (Qiagen). The density of gene start site positions was quantified by performing a kernel density estimation (KDE) for each chromosome using the distplot function from the Seaborn v0.8.1 package for Python (www.​python.​org). Bandwidth for the KDE was set manually for chromosome 20 and scaled linearly for all other chromosomes based on chromosome length. To determine whether the KDE for the differentially expressed transcripts was a common occurrence, a 95% percentile interval was constructed. A total of 2699 gene start sites were randomly selected (11 of which were start sites for identical genes) and a KDE was performed. This process was repeated 1000 times to obtain 1000 distributions. For each point where the KDE was calculated, values were obtained for the 97.5th percentile and the 2.5th percentile. These values were then used to construct an interval that would contain 95% of the randomly generated distributions.

Sequencing validation on genes of interest was completed with isolated RNA from additional blastocysts (n = 12, 6 from each group). Reverse transcription was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and cDNA was diluted (1:4) in 1X Tris-EDTA buffer prior to performing quantitative reverse transcription PCR (RT-qPCR) on the QuantStudio 5 Real-Time PCR System (Applied Biosystems). Three microliters of diluted cDNA was combined with 5 μM primer mix and Power SYBR™ Green PCR Master Mix (Applied Biosystems) in a 15 μl final volume and amplified under the following thermal cycling conditions: 95 °C for 10 min followed by 40 cycles at 95 °C for 15 s and 60 °C for 1 min, and a melt curve stage at 95 °C for 15 s, 60 °C for 1 min, and 95 °C for 15 s. Each sample was run in duplicate for 11 genes of interest (ALK, TNFRSF10A, TSPAN9, CCND3, GNAS, LTBP3, MAPK8IP1, NDRG1, SREBF1, EPSTI1, TLE2) and analyzed compared to the expression of three, stable, housekeeping genes (GAPDH, PPIA, RPL19). Statistical analysis was performed using REST 2009 software (Qiagen) which uses PCR efficiencies and mean crossing point deviation between the sample and control groups to test for significance by a Pair Wise Fixed Reallocation Randomisation Test© (significance at P < 0.05) [29]. The most consistent housekeeping gene was selected for normalization (PPIA).

All participants provided written informed consent and this study was approved by the Western Institutional Review Board (protocol no. 20140458).

RNA sequencing data was collected and an average of 46.9 million reads were acquired for each blastocyst. Quality filters were applied to all samples to remove reads with poor sequencing quality resulting in 91% of the reads passing these filters and proceeding for alignment to the human reference genome (Supplementary Table 1). One sample from the DC group was eliminated from further analysis due to poor quality. Expression intensity was calculated using RPKM method and only transcripts with a mean expression of 0.05 or greater were considered for analysis, resulting in 26,489 Ensembl IDs expressed between the two sample sets. A FDR adjusted p value was used to determine significance and reduce false positives. Additional quality control was performed to identify and remove any outliers and overly abundant transcripts to reduce normalization artifact. In total, 2688 (10%) of the expressed transcripts were considered significantly differentially transcribed in AMA blastocysts compared to DC with 2551 of these having decreased expression (95%) and 137 displaying increased transcription (5%) (Supplementary Table 2; Q < 0.05). When considering all transcripts analyzed (including those that were non-significant), 70% displayed decreased expression in the AMA sample set, revealing an overall global decrease in transcription in the blastocysts from this group. Volcano plot analysis depicts this trend with the vast majority of significantly differentially expressed transcripts having a negative fold change in AMA blastocysts (Fig. 1).

Principle component analysis (PCA) was performed to reveal global differences between the sample groups. Two distinct sets were identified using PCA, with the AMA blastocysts grouping together but separately from the grouped DC blastocysts, indicating uniformity and low biological variability among samples within each group (Fig. 2). Unsupervised hierarchical clustering analysis also distinguished the two groups, with each having uniquely different transcription patterns and branching separately (Fig. 3). The uniformity observed in the AMA blastocyst transcriptome reflects a strong phenotype for this sample group. Examination of the gene density on individual chromosomes revealed a significantly higher number (> 30%; Q < 0.05) of differentially expressed transcripts localized at the telomeric regions for 9 chromosomes (4, 9, 11, 16, 17, 19, 20, 21, 22) (Supplementary Figure 1). Over a quarter (28%) of the differentially expressed transcripts in this study were found within 10 Mb of the telomeric regions at both chromosome ends.

Canonical pathway analysis (IPA®) of the 2688 significantly altered transcripts revealed hits for G protein–coupled receptor signaling, calcium signaling, AMPK signaling, and gap junction signaling, among others (Table 1). Expanded pathway analysis of these statistically significant transcripts revealed a set of interacting molecular signaling networks and upstream regulators important for embryo development which included SP1, ESR1, HDAC1, DMNT3B, TP53, MAPK3, VEGFA, and CREBBP (Fig. 4). A Regulator Effects analytic (IPA®) was performed to identify the top 5 upstream regulators driving the observed global decreased transcription in AMA blastocysts. Altered pathways that are predicted to work with other molecules and lead to changes in downstream RNA expression included pathways involved in invasion, transmission and activation of cells, cell movement, and organismal death in blastocysts from women of advanced maternal age (Supplementary Figure 2).

Genes chosen for validation utilizing RT-qPCR focused on the transcription ligand-dependent activation of ESR1/SP/CREBBP pathway as it was observed to be a highly significant upstream regulator important for a variety of biological processes including cell cycle regulation, proliferation, and apoptosis. In total, 62 genes regulated by the ESR1 network were found to be significantly decreased in AMA blastocysts, as well as 56 genes regulated by SP1, and 41 genes regulated by CREBBP (P < 0.05). ESR1-regulated genes chosen for validation included ALK, TNFRSF10A, and TSPAN9. SP1-regulated genes examined for validation were CCND3, GNAS, LTBP3, MAPK8IP1, NDRG1, and SREBF1. Finally, CREBBP-regulated genes tested included EPSTI1 and TLE2. All genes displayed a trend towards reduced expression in additional AMA blastocysts tested by RT-qPCR, compared to DC, confirming our observations from the RNA sequencing analysis (Fig. 5). TNFRSF10A and TSPAN9, which are regulated by ESR1, were significantly reduced in the blastocysts from older women (fold change = 0.63 and 0.23 respectively; P < 0.05; Fig. 5). MAPK8IP1 which is regulated by SP1 was also significantly reduced in AMA blastocysts (fold change = 0.17; P < 0.05; Fig. 5). EPSTI1, regulated by CREBBP, was the final gene that displayed a significant reduction in gene expression among the AMA group (fold change = 0.14; P < 0.05; Fig. 5).