Genotypic divergence in mouse oocyte transcriptomes: possible pathways to hybrid vigor impacting fertility and embryogenesis

Abstract

What causes hybrid vigor phenotypes in mammalian oocytes and preimplantation embryos? Answering this question should provide new insight into determinants of oocyte and embryo quality and infertility. Hybrid vigor could arise through a variety of mechanisms, many of which must operate through posttranscriptional mechanisms affecting oocyte mRNA accumulation, stability, translation, and degradation. The differential regulation of such mRNAs may impact essential pathways and functions within the oocyte. We conducted in-depth transcriptome comparisons of immature and mature oocytes of C57BL/6J and DBA/2J inbred strains and C57BL/6J × DBA/2J F1 (BDF1) hybrid oocytes with RNA sequencing, combined with novel computational methods of analysis. We observed extensive differences in mRNA expression and regulation between parental inbred strains and between inbred and hybrid genotypes, including mRNAs encoding proposed markers of oocyte quality. Unique BDF1 oocyte characteristics arise through a combination of additive dominance and incomplete dominance features in the transcriptome, with a lesser degree of transgressive mRNA expression. Special features of the BDF1 transcriptome most prominently relate to histone expression, mitochondrial function, and oxidative phosphorylation. The study reveals the major underlying mechanisms that contribute to superior properties of hybrid oocytes in a mouse model.

INTRODUCTION

The oocyte has arguably the most colossal task given to any cell type. After a prolonged meiotic arrest that can last up to decades in some species, the oocyte must quickly respond to a hormonal stimulus, grow considerably, finish building the maternal endowment of proteins and mRNAs, resume meiosis, segregate chromosomes, halt meiosis to await fertilization, re-enter the cell cycle after fertilization, reprogram the embryonic genome, and support early embryogenesis. The oocyte becomes transcriptionally inactive before meiosis resumes and germinal vesicle breakdown (GVBD) occurs (27). Thereafter, it is of utmost importance that the maternal mRNA endowment be correctly regulated through accumulation, timely polyadenylation and translation, and degradation. The precise regulation of the oocyte mRNA pool ensures that the correct proteins are produced in the correct sequence to ensure high oocyte quality and subsequent high embryonic developmental potential (12, 45, 57, 58).

Previous studies identified factors that drive early developmental events and attempted to identify molecular markers of oocyte quality that support oocyte function and early embryogenesis (2, 33, 44, 62, 70). However, the results of studies of oocyte quality markers need to be considered in the context of natural variation among fertile genotypes. Finding that a quantitative difference in expression of a perceived marker of oocyte quality is within the range of natural genetic variation in expression typical for oocytes of fertile individuals of different genotypes would call into question the value of such a marker or at least indicate a need for greater understanding of the broader cellular context in which that factor operates. However, little if any attention has been given to genetic variability in the oocyte transcriptome. The extent of this variation, how fertility is achieved in the face of such variation, and how variations in expression of one gene may be offset by variations in expression of other genes are not known. Adding to the complexity, the degree to which maternal mRNA populations are differentially regulated between fertile genotypes, and how this occurs, is not known. Addressing these questions will provide a new, in-depth understanding of the molecular mechanism that generate healthy oocytes, and to what degree oocytes of different genotypes take different pathways to fertility. Additionally, because germ cells are important determinants of reproductive compatibility, understanding genetic variation in oocyte mRNA expression and regulation is relevant to understanding key aspects of speciation and evolution.

Mouse inbred strains and hybrids provide powerful experimental tools for exploring genetic variation across oocyte transcriptomes to address the above questions. Fertile females of defined genotypes and consistent availability permit reproducible, long-term experimental study of how genetic variation relates to oocyte quality. For example, the C57BL/6J (B6) and DBA/2J (D2) strains are both fertile (although not to an equal degree, as C57BL/6J has higher fertility that DBA/2J; www.informatics.jax.org), as are their F1 hybrid offspring, B6D2F1 (BDF1). However, past studies noted differences in oocyte phenotype between the three genotypes. BDF1 hybrids display hybrid vigor for several aspects of oocyte function, including a superior ability to reprogram somatic nuclei after somatic cell nuclear transfer (11, 19, 61), lower rates of embryonic fragmentation and two-cell arrest, ooplasm granularity, size of the embryonic pronuclei, and other morphological attributes (16, 17, 24, 48, 49, 51). The parental strains also differ in many of these characteristics (23, 26), and previous studies have undertaken genetic mapping of these traits (3, 25).

An F1 genotype can lead to dominance effects, incomplete dominance, or transgressive gene expression. Additive effects of parental dominance at different genes and transgressive gene expression are of particular interest for understanding hybrid vigor (or hybrid suppression), but even incomplete dominance for particular combinations of genes could contribute to unique F1 hybrid phenotypes. Such effects have not been examined in mammalian oocytes. The goal of this study was to determine the effects of parental strain and F1 hybrid genotype on oocytes, and particularly to understand the potential basis for hybrid vigor in mouse oocytes, as a way to understand better the mechanisms that determine oocyte quality. This study is the first to combine oocytes of different stages (germinal vesicle and meiotic metaphase II) and different genotypes (two parental inbred strains and F1 hybrid) into a single RNA sequencing (RNA-Seq) analysis capable of identifying molecular characteristics that distinguish F1 hybrids. Our approach to analysis combined QIAGEN Ingenuity Pathway Analysis (IPA) and a powerful approach to discern the biological impact of an F1 hybrid genotype by comparison to “predicted” F1 hybrid gene expression data synthesized from parental strain data. Four remarkable findings emerged from the analysis: 1) there are extensive differences in transcriptomes between genotypes, with roughly one quarter or more of the expressed genes showing differences between the two parental strains at both stages, 2) the three genotypes modulate their mRNA populations differently during maturation, 3) the superior BDF1 oocyte properties arise through additive dominance and incomplete dominance effects, and to a much lesser extent, transgressive effects, and 4) the BDF1 hybrid genotype is associated with differences in expression or activity of certain upstream regulators and canonical pathways, most prominently related to histone expression, mitochondrial function, and oxidative phosphorylation.

MATERIALS AND METHODS

Ethics approval and consent to participate.

All studies were approved by the Michigan State University Institutional Animal Care and Use Committee, consistent with National Institutes of Health (NIH) Guide for the Care of Use of Laboratory Animal, and with the Association for Assessment and Accreditation of Laboratory Animal Care accreditation.

Oocyte isolation.

This data set consists of germinal vesicle (GV) and second meiotic metaphase (MII) stage oocytes from B6, D2, and BDF1 mice. Oocytes from individual mothers were pooled for each biological replicate (n = 9–20 oocytes/sample). A total of 32 samples were analyzed, comprising at least five for each of the stages/genotypes analyzed. Females were obtained from Jackson Laboratories at 7 wk age and used from 8 to 12 wk age. Oocytes were collected as described (54). Notably, GV oocyte samples were collected from mice that received only equine chorionic gonadotropin (eCG) to initiate meiotic resumption, whereas MII oocytes were collected from mice that received eCG followed by human chorionic gonadotropin to initiate ovulation. Briefly, GV or MII-stage oocytes were collected into room-temperature HEPES-buffered M2 medium. During GV-stage oocyte collection, M2 medium was supplemented with 0.225M 3-isobutyl-1-methylxanthine (Sigma, I7018) to prevent GVBD during the isolation procedure. Any abnormal or dead GV or MII-stage oocytes were excluded. Only fully grown GV oocytes that possessed visible intact nucleus were included. Using light microscopy, we could not distinguish GV oocyte samples as nonsurrounded nucleolus or surrounded nucleolus, and hence, both of these types of GV-stage oocytes would be included in the samples. Immediately after collection the zonae pellucidae were removed with acidified Tyrode’s buffer (60) for ~30 s followed by immediate washing through M2 medium for 1 min. This treatment eliminated cumulus cell processes or other adherent material as a possible source of RNA contamination. Some MII-stage oocytes had polar bodies remaining attached.

Oocyte RNA sequencing.

RNA was extracted from pools of GV or MII-stage oocytes with the ARCTURUS PicoPure RNA Isolation kit (ThermoFisher #12204-01), following the manufacturer protocol. Oocytes were lysed in 20 μL PicoPure extraction buffer, followed by heat treatment at 40°C for 30 min. Lysates were processed immediately or stored at −80°C. During the RNA isolation, DNase treatment was applied for 15 min at room temperature (Qiagen #79254). Isolated RNA was eluted with 11 μl of the PicoPure elution buffer and immediately used or stored at −80°C. Five microliters of each eluate were processed and amplified with SPIA technology using the Ovation RNA-Seq System V2 (NUGEN #7102-32). The cDNA libraries were purified using a Qiagen MinElute Cleanup Kit (Qiagen #28204), quantified on a Nanodrop (ThermoFisher NanoDrop Lite) or Biodrop (uLite), and stored at −80°C. For each library, ~1 μg of DNA was diluted into Tris-EDTA buffer (TE) for a total volume of 130 μL and fragmented to an average length of 300 bp on a Covaris shearer. We mixed 20 μL of each fragmented library with 14 μL water, 4 μL 10× S1 nuclease buffer, and 2 μL S1 nuclease for a 40 μL total reaction volume, and incubated it for 30 min at room temperature (Promega #M5761) (18). Starting with an additional bead purification, SPIA libraries were processed through the Ovation Ultralow System V2 1–16 (Nugen #0344-32) with amplification for 12 cycles. After the last bead purification, amplified-barcoded libraries were eluted in 30 μL of 1× TE.

Libraries were sequenced on an Illumina HiSeq 2500 system or an Illumina HiSeq 4000 system. Libraries were loaded at 65% of standard concentration with PhiX DNA added up to 10% of the input (HiSeq 2500) or with PhiX loaded to 1% with no decrease in total loading (HiSeq 4000). Samples for all genotypes and stages were represented on both platforms. Sequencing was performed with 50 nt single-end reads. The number of PF (passed-filter) reads ranged from 25.5 to 53.6 M, the fraction of Q30 bases from 90.7 to 95.4% and average Q from 36.9 to 38.9 (Supplemental Table S1, doi: 10.6084/m9.figshare.9555554; https://figshare.com/s/e1c03beddde3656dfbde). Sequencing data are available in Gene Expression Omnibus (GSE114158: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE114158). Results of the two sequencers were comparable, and any batch effect was accounted for in later processing as described below. Small increase of quality parameters for HiSeq 4000 was expected due to change in technology.

Reads were aligned to the mouse genome (build GRCm38.p4) with HISAT2 (22) and the option to avoid alignment of reads to pseudogenes. Reads aligned to ribosomal RNA (rRNA) or rRNA-like genes were removed, as were the duplicates caused by the sequencing technology, which were defined as one read in a pair of identical reads found within the distance of 100 units (optical duplicates in libraries sequenced on HiSeq 2500) or 2,500 units (“ExAmp” duplicates in libraries sequenced on HiSeq 4000) on the same tile of a sequencing lane. A total of 18.7–39.7 M reads per library were successfully aligned to unique non-rRNA gene transcript sequences. Reads aligned to genes were quantified with featureCounts (part of Subread 1.5.1 package) (32). DeSeq2 (35) was used for differential expression analyses between the genotypes (within the GV or MII stage), or between two stages (within a genotype). Because samples were processed and sequenced in two batches on two different sequencing systems, the batch identifier was provided to DESeq2 as a factor, and there were at least two samples in each batch for all six genotype/stage combinations. Initially, differentially expressed genes (DEGs) were defined as those with q-value (false discovery rate) ≤ 0.05. Due to the large number of DEGs identified at this threshold, a q-value of ≤ 0.01 was used in subsequent IPA. A comprehensive table of DEGs is provided (Gene Expression Omnibus GSE114158: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE114158). We note that mRNAs expressed at or above a threshold of 0.02 fragments per kilobase of transcript per million mapped reads (FPKM) were retained for analysis. One FPKM corresponds to ~50 mRNA copies in MII-stage oocytes, assuming ~17 M transcripts per oocyte (4) and 3.2 × 10⁵ FPKM average per MII-stage library, so a threshold of one mRNA copy per cell is 0.02 FPKM. We recognize that this threshold is lower than used in other studies, e.g., 0.3 suggested in Ramskold et al. (50). However, such higher thresholds were derived for RNA-Seq of various tissue and cell types with different mRNA content per cell. Therefore, we prefer not to limit the analysis at this early stage by imposing an arbitrarily chosen threshold at a higher value. We also recognize that the proportion of DEGs reported could be lower with a higher threshold.

To assess the similarity of samples’ expression profiles, raw read counts were normalized (DeSEQ2) and log-transformed, then corrected for batch effects of sequencing platform (limma R package (52). The obtained set of expression values was further filtered by excluding low-expression genes (mean across samples below 5). Distances between pairs of samples s₁, s₂ were then calculated as 1 − CC_Spearman(s₁, s₂). Multidimensional scaling was then performed using sklearn.manifold.MDS python module.

Predicted BDF1 versus actual BDF1.

For analysis of dominance effects, direct comparisons between the three genotypes initially employed a thresholding approach. As an alternative method to avoid thresholding, we devised a novel strategy of comparing the actual BDF1 transcriptomes to values calculated to be the intermediate values between the two parental genotypes (“predicted” BDF1). For each of the two stages, the expected expression profiles for BDF1 oocytes were predicted as the average of the expression profiles for B6 and D2 oocytes and compared with the actual BDF1 oocyte expression profiles. Since samples from B6 and D2 can be paired in different ways, a randomization procedure was used: 1) Samples from B6 and D2 were randomly paired (only within batch, keeping samples from two batches apart) and their expression profiles were averaged. 2) The combined B6/D2 samples were then compared with BDF1 samples using DESeq2. 3) Steps 1 and 2 were repeated 10 times and a gene was considered differentially expressed between the actual and predicted BDF1 transcriptomes if such outcome was obtained in at least 9 out of 10 repetitions.

IPA.

IPA was used to analyze the biological relevance of DEGs. Analysis tools applied from IPA included Canonical Pathway (CP) and Upstream Regulator (UR) analyses. For CP analysis, IPA calculates overlap P values, taking into account the number of DEGs and the number of molecules in the knowledge database associated with that pathway, and the number of DEGs and the number of molecules in the knowledge database. For UR analysis, results are based on the number of DEGs regulated by a given UR. In addition to overlap P values, z-scores are calculated for CPs and URs. The z-score reflects activation (z > 0) or inhibition (z < 0) of CPs and URs; it is based on the number of associated DEGs for which the direction of regulation (up- or down-) is consistent with activation or with inhibition. Because P(|z| > 1.96) = 0.05 for normal N (0,1) distribution, we consider CPs and URs with z > 1.96 to be significantly activated, and those with z < −1.96 to be significantly decreased. We note that characterization of URs as “affected,” “activated,” or “inhibited” does not necessarily imply a change in the expression of that UR itself but, rather, may reflect a change in the expression profiles of its downstream effectors.

As the initial IPA analyses provided a large number of URs or CPs, we truncated the output by retaining two categories of results: category 1 URs and CPs having both |z|>1.96 and P < .05 (−log₁₀ P > 1.3) for at least one comparison, category 2 URs and CPs with P < 0.05 for at least one comparison, but no significant z-score for the same comparison. Within these two categories, chemical URs had at least one comparison with both |z|>1.96 and P < 0.05, and URs expressed within the cell had FPKM >0.02 in at least one genotype/stage. The threshold of 0.02 is based on the reported number of ~17 M transcripts in MII-stage oocytes (4), and the sum of FPKM values for all transcripts is 3.2 × 10⁵ (the mean for MII-stage samples in our data sets); 1 FPKM corresponds to ~50 mRNA transcripts in our MII-stage data set. Secreted factors as URs (P < 0.05) were initially included. We removed all biologic drugs.

We note that some UR-effector relationships and intra-CP relationships are assigned in IPA on the basis of protein-protein interactions or other posttranscriptional events. Because the DEG data that were entered into IPA were derived by mRNA expression analysis, the interpretation of the IPA results incorporated inspection of UR and CP member gene expression patterns, and their functional classes (transcription factor, kinase, enzyme, transporter, etc.). Where necessary, mRNA expression data for these genes and related gene family members were assessed to clarify the biological significance of IPA results.

RESULTS

The RNA-Seq data set presented here is the largest of its kind representing three mouse genotypes and two stages of oocyte development. The overall quality of the data set is high with an average depth of 28.5 M aligned exonic reads per sample (Supplemental Table S1). The average number of detected mRNAs was ~18,600 for GV stage and ~16,400 for MII stage. Across all three genotypes we detected a total of 20,755 mRNAs and 16,525 mRNAs respectively at the GV and MII stages respectively (DEG Table GSE114158). We also included a multidimensional scaling plot that shows a clear clustering of samples by mouse genotype or oocyte developmental stage (Fig. 1). This plot also indicates a high degree of reproducibility between sample types.

Fig. 1.

Multidimensional scaling (MDS) maps samples' high-dimension expression profiles into points in 2-dimensional space. Distances between points correspond to the similarity of samples’ expression profiles. Germinal vesicle (GV) and metaphase of meiosis II (MII) samples are clearly separated, as are B6, BDF1, and D2 samples at each stage. Furthermore, points for BDF1 samples are positioned in the middle between B6 and D2 samples.

Extensive gene expression differences between B6 and D2 oocytes and attendant effects on pathways and functions.

Comparisons between the parental strains revealed a striking number of DEGs at both stages (4,583, 23% and 5,825, 33% DEGs at the GV and MII stages, respectively, at significance threshold q ≤ 0.05) (Fig. 2, G, G′). Even at the significance threshold of q ≤ 0.01, the fraction of DEGs between parental strains was large (16 and 23% for GV and MII stages, respectively) (DEG Table, GSE114158). Many of these differences were only observed between parental strains at one stage (922 GV-specific differences, and 1,683 MII-specific differences, at q ≤ 0.05; DEG Table, GSE114158).

Fig. 2.

Interstrain comparisons at GV and MII stages. Top: differentially expressed genes (DEGs) were first identified for each of the three genotype comparisons (B6 vs. BDF1, D2 vs. BDF1, and B6 vs. D2) at each stage (GV or MII) (P < 0.05). Then, these DEGs lists were compared with each other to determine the number of DEGs shared between comparisons (within each stage). Of particular interested are B (GV) and B′ (MII) include DEGs where BDF1 mice are different from both parental strains at each stage. Bottom: summary comparison of DEGs lists between stages. N and N′ groups denote DEGs observed at MII or GV that not observed at the other stage (i.e., stage-specific DEGs).

If genetic differences in the expression and regulation of some mRNAs was balanced by a difference in expression of other mRNAs in the same pathway, the net outcome would be no overall change in cellular process or function. Without such compensation in gene function, however, oocytes of different genotypes could manifest markedly different phenotypes. To investigate these possibilities, we applied IPA to assess the impact of interstrain DEGs on URs and CPs, with particular focus on those pathways for which a clear directional difference was evidenced by a significant z-score. IPA revealed hundreds of significant differences in URs and CPs between B6 and D2 oocytes, indicating that many differentially regulated genes in either strain are not compensated by differences in other genes in the opposite genotype; i.e., many significant differences likely impacting cellular processes are evident between genotypes. We observed 23 URs with significant z-scores, indicating significant activation (n = 14) or inhibition (n = 9), in D2 compared with B6 in GV-stage oocytes and 22 with significant z-scores (11 inhibited and 11 activated) at the MII stage. Eight of these URs were different between D2 and B6 at both stages, and the direction of their z-scores did not change during maturation. A single CP (sumoylation, z = −2.18 in D2 compared with B6) displayed a significant z-score at the GV stage. Many more CPs were affected at the MII stage, three of which were significantly increased (PPARα/RXRα activation, PTEN signaling, and RhoGDI signaling) and 44 of which were significantly decreased (including NGF and thrombin signaling) in D2 compared with B6 MII-stage oocytes. Cumulatively, these data indicate that although B6 and D2 mice are both fertile, there are significant differences in pathways and processes employed to generate viable oocytes.

Core set of maturational changes in the transcriptome.

To determine whether genotype also affects how oocytes regulate mRNA abundance during maturation, we identified mRNAs that change in abundance during maturation for each strain (i.e., GV vs. MII expression within strain). We then examined similarities and differences in maturational changes between genotypes. We observed >2,600 mRNAs that were regulated differently (Fig. 3, E vs. F; P < 0.05) between B6 and D2 oocytes, indicating a significant genetic difference in the regulation of many mRNAs during maturation. However, we observed >3,000 mRNAs (at q ≤ 0.05), of which 1,917 are at q ≤ 0.01, that shared similar changes in abundance during maturation across all three genotypes (two parental strains and BDF1 oocytes) (Fig. 3, A). This indicates that there is a core set of maturation-related changes in mRNA abundance. Applying IPA to this core set of 1,917 (at q ≤ 0.01) maturation-related DEGs, we found that the three nonchemical URs with the greatest activation in MII-stage oocytes across all three genotypes were VEGFA, INSR, and IGF1R. The top three nonchemical URs with the largest inhibition in MII-stage oocytes across all three genotypes were RICTOR, KDM5A, and alpha-catenin. Of the many CPs significantly altered during maturation, three displayed significant z-scores across genotypes, including EIF2 signaling (activated), and sirtuin signaling and RhoGDI signaling (inhibited). The core set of DEGS and these associated URs and CPs reflect features of oocyte maturation shared across genotype, indicating functions that are likely critical to this process and possibly for fertility.

Fig. 3.

Changes associated with oocyte maturation from GV to MII stage. DEGs which changed significantly during oocyte maturation (GV to MII stage) were identified for each of the three genotypes (B6, D2, BDF1), for significant threshold P < 0.05 (P < 0.01 shown below P < 0.05 numbers in square brackets). Then, these DEG lists were compared with each other to determine the number of maturation DEGs shared between genotypes. Top: DEGs organized according to their maturation overlap groups. Group A includes DEGs that changed during oocyte maturation across all three strains. Group G includes maturation set DEGs that are specific to BDF1s. Bottom: table that shows the directionality of the change in mRNA abundance observed during maturation and also the number of DEGs at the P < 0.01 significance level at each strain.

BDF1 oocytes display unique features of transcriptome modulation during maturation.

Analysis of BDF1 oocyte maturation revealed unique aspects of maturation that were distinct from either of the parental strains. BDF1 oocytes modulated the relative abundances of more mRNAs than either parental strain. Therefore, the overall fractions of modulated/detected mRNAs were highest within BDF1 oocytes at 35% compared with the two parental strains (29% in B6 and 30% in D2) (see DEG Table GSE114158). This resulted in nearly 1,000 more DEGs during BDF1 maturation (6,766 changes, at q ≤ 0.05) compared with either of the parental strains (5,698 and 3,747 changes for B6 and D2, at q ≤ 0.05, respectively) (Fig. 3). Additionally, BDF1 oocytes had more genotype-specific DEGs (n = 1,615) than either B6 or D2 oocytes (n = 1,401 and n = 1,241) (Fig. 3, E–G).

To further investigate genetic variation in mRNA stability and degradation, and how this may be uniquely regulated in BDF1 oocytes, we compared the number of mRNAs that significantly increased or decreased in abundance during maturation within each genotype (Fig. 3). We observed that BDF1 oocytes showed a higher ratio of mRNA increases to decreases (0.77) compared with B6 (0.68) and D2 (0.41), indicating a larger number of mRNAs with greater relative stability in BDF1 oocyte maturation than the two parental strains (Fig. 3). We also observed that the mRNAs displaying the largest magnitudes of changes in F1 oocytes underwent greater diminishment in abundance than in parental strains, indicating that the BDF1s may be more efficient at both selective mRNA degradation and selective mRNA stabilization (Table 1).

Table 1.

mRNAs that decrease over 100-fold only during BDF1 oocyte maturation (GV to MII stage)

mRNA	Fold-Change
Mmp14	−204.079
Syk	−172.346
Adamts1	−172.195
Il11ra1	−170.127
Ctso	−162.909
Trim2	−161.823
Mid2	−156.510
Mfge8	−151.587
Tgfbr2	−149.415
Ccdc137	−133.473
Sardh	−124.430
Grk6	−122.667
Pou6f1	−117.470
Rgs11	−116.354
Lmod1	−104.147
Tmem132a	−101.691

GV, germinal vesicle; MII, metaphase of meiosis II. List restricted to annotated genes significant at P < 0.01.

We applied UR and CP analysis to all of the mRNAs that changed in abundance during BDF1 oocyte maturation (Fig. 3, A, C, D, G). Of the 941 URs identified by this analysis, 24% were specific to BDF1 oocyte maturation (we chose this approach rather than IPA on just set G to increase sensitivity and better capture the impact of BDF1-specific changes that could work in conjunction with shared changes). Ten of these URs had significant z-scores (Table 2). Of the 183 CPs significantly altered during BDF1 maturation, 32% were specific to BDF1 oocyte maturation. One of these CPs (cell cycle: G2/M DNA damage checkpoint regulation) displayed a significant activation z-score. These results highlight specific cellular processes uniquely altered during maturation in BDF1 oocytes.

Table 2.

Upstream regulators with activation/inhibition states unique to BDF1 oocyte maturation (GV vs. MII)

	z-Score	−log10(P)
Upstream Regulator	F1-all (Fig. 3, A, C, D, G)	F1-all (Fig. 3, A, C, D, G)
BNIP3L	3.00	1.55
CD44	2.73	1.54
INHBB	2.21	1.58
S1PR3	2.21	1.32
ETV4	2.08	1.41
EGR1	2.02	1.67
LASP1	2.00	2.99
MTM1	−1.99	1.40
E2f	−2.06	1.33
ACOX1	−2.12	1.49

List restricted to upstream regulators (URs) with significant z score. Significant z score (z > 1.96 or z < −1.96) indicates activated or inhibited in BDF1 MII oocytes. Cellular upstream regulators restricted to those with FPKM > 0.02 in at least one sample. Chemical reagents, drugs, and toxicants were removed.

Overall, this approach revealed three clear results. First, BDF1 oocytes significantly alter the expression of >1,000 more mRNAs than either of the parental strains during maturation. Second, the vast changes in the BDF1 oocyte transcriptome appear to be due to enhanced rates of degradation and stabilization of particular subsets of mRNAs compared with the parental strains. Last, the activation state of many URs and CPs are altered only in BDF1 oocytes, indicating multiple pathways and networks differ in BDF1 oocytes.

Mechanisms underlying differences in mRNA expression and regulation that distinguish F1 oocytes from parental strain oocytes.

The discovery that B6, D2, and BDF1 oocytes display thousands of differences in regulation of mRNAs during maturation prompted us to ask to what degree gene expression and gene regulation differ between BDF1 and parental strains within oocyte stage, and how this could lead to a hybrid vigor phenotype. Differences between BDF1 and parental oocyte phenotypes could arise through several mechanisms, including: 1) transgressive gene expression (BDF1 gene expression levels are higher or lower than the expression range of the parental strains), 2) dominance in gene expression (F1 resembles one parental strain or the other), or 3) incomplete dominance (i.e., BDF1 displays an intermediate level of gene expression) (Fig. 4). Moreover, any or all of these mechanisms could affect gene expression levels within stage (i.e., GV or MII) or could arise because of differential regulation during oocyte maturation.

Fig. 4.

Overview of effects of hybrid genotype in oocyte transcriptomes. Summary of BDF1 gene expression phenotypes. BDF1 gene expression could transgress across the gene expression ranges set by the high or low expressing parental strain, TGE (transgressive gene expression). BDF1 expression could resemble one of the parental strains (dominance high; dominance low). Or, BDF1 expression could display incomplete dominance (i.e., an intermediate level of expression between the two parental strains). The number of DEGs that display incomplete dominance in GV- and MII-stage oocytes was calculated by taking the total number of DEGs identified for the respective stage (GV n = 4,825; MII n = 6,452) and subtracting all DEGs that display TGE or dominance for the respective stage.

One potential way for genetic hybridization to impact BDF1 oocyte characteristics would be through transgressive gene expression (TGE), wherein the BDF1 expression level of a gene is significantly higher (TGE-H) or lower (TGE-L) than the levels seen for the two parental strains (i.e., outside the range defined by the two parents) (Fig. 4). However, we observed only a single gene displaying TGE at the GV stage (4933406M09Rik), and a total of just 25 genes displaying TGE in MII-stage oocytes (22 TGE-L and 3 TGE-H; Supplemental Table S2). Among the most highly repressed TGE-L genes in BDF1 oocytes were Fads2, Qrsl1, Ebf1 and Sim1.

The limited number of TGE genes relative to the vast number observed BDF1 DEGs suggested other mechanisms underlie the BDF1 hybrid phenotype. The most prevalent pattern of BDF1 gene expression was incomplete dominance. Supporting this, BDF1 oocytes displayed fewer DEGs relative to either parental strain, ranging from 32 to 39% (q ≤ 0.05) of the number of DEGs identified between the two parental strains at the GV or MII stage, respectively (DEG Table GSE114158). There was a total of 4,456 DEGs at the GV stage and 5,921 DEGs at the MII stage that display an incomplete dominance pattern of expression in BDF1 oocytes where gene expression levels significantly vary between the parental strains but not between the BDF1 oocytes and either parental strain (Fig. 4). Only 465 of the 4,456 DEGs at the GV stage were differentially expressed in BDF1 oocytes at the MII stage. Together, these data indicate that the most prevalent pattern of BDF1 gene expression is incomplete dominance.

Dominance in gene expression was also observed, wherein mRNAs display similar expression values in BDF1 and one parental strain, but a significant difference from the other parental strain (Fig. 4). We initially observed dominance effects in examining DEGs and by using IPA above. For a more comprehensive evaluation of dominance at the level of individual genes, we identified subsets of interstrain DEGs (B6 vs. D2) that displayed dominance in BDF1 oocytes. We assigned dominance to a DEG when BDF1 and parental strain A expression were both significantly different from parental strain B, and BDF1 and parental strain A expression levels were within 20% (fraction of the parental expression range) of each other. We used a dominance ratio (DR) to determine these relationships. For gene g, we defined dominance ratio DR(g) = [expr_F1(g) – expr_B(g)]/[expr_A(g) – expr_B(g)], where A and B were parental strains. We call the strain A dominant for gene g if: DR(g) ≥ 0.8 and g was differentially expressed in comparisons A vs. B and BDF1 vs. B. We used the specific terms “dominance-high” for expr_A(g) > expr_B(g), and “dominance-low” for expr_A(g) < expr_B(g). This analysis revealed 368 genes displaying dominance at the GV stage and 526 genes displaying dominance at the MI stage, for a total of 894 genes displaying dominance in mRNA expression in BDF1 oocytes (Supplemental Tables S3 and S4; Table 3). Only 47 genes displayed similar dominance patterns at both stages (Table 3).

Table 3.

Genes that display parental dominance in BDF1 oocytes at GV and MII stages

Dominance Type	GV Stage	MII Stage	Both Stages	List of Genes across Both Stages for Dominance Pattern
B6 Dom High	68	113	9	Fam19a2, Haus4, Pusl1, Rbm34, Rnf168, Tlr9, Ublcp1, Zfp568, Zkscan4
D2 Dom High	83	90	6	1700024I08Rik, Gm32293, Gm9316, Krt73, Slc39a8, Zfp566
B6 Dom Low	80	145	19	1810013L24Rik, Arl14epl, Cdc5l, Gm35315, Gm5039, Hdx, Hist1h1e, Lclat1, LOC101056073, Lrrc8d, Lysmd3, Mterf3, Phc3, Prr23a3, Scn3a, Sf3b6, Stc1, Wac, Zfp24
D2 Dom Low	83	126	20	Atxn7l1, C5ar2, Far2, Fer1l6, Fxyd6, Gm36876, Gna14, Grm4, Lrrc31, Muc20, Olfr206, Pappa2, Pnp, Scnn1b, Sec14l3, Slc15a5, Syde1, Trim30a, Trpa1, Ttll10

BDF1 DEGs contribute to differences in biological processes.

The next question asked was to what degree BDF1 oocytes display differences in pathways and functions compared with parental genotypes. As described above, we performed IPA analysis on all of the interstrain DEGs detected (Fig. 2). Here, we examined IPA results for affected CPs and URs to visualize effects of TGE, dominance, and incomplete dominance in BDF1 oocytes. Transgressive effects would be visible in IPA as URs and CPs where BDF1 oocytes are significantly different (P < 0.05) from both parental strains but parental strains are not significantly different from each other. Dominance effects would be evident in IPA by URs and CPs where B6 is significantly different from D2 (P < 0.05 and |z| > 1.96) and only one parental strain is also significantly different from BDF1 (P < 0.05). Incomplete dominance effects would be evident in IPA by URs and CPs where B6 is significantly different from D2 (P < 0.05 and |z| > 1.96) but neither strain is significantly different from BDF1. We performed this analysis for both the GV and the MII stages.

Within the IPA UR and CP results, we observed once again transgressive, dominance, and incomplete dominance effects in BDF1 oocytes. Incomplete dominance effects were the most prevalent pattern for BDF1 URs with 19 at the GV and 16 at the MII stages, respectively (Fig. 5). There were also several URs that displayed incomplete dominance at the GV (n = 3) and MII (n = 7) stages. Transgressive effects for URs in BDF1 oocytes were limited to the cellular regulator RICTOR, which also had a significant negative z-score (−2.22), indicating inhibition in BDF1 oocytes compared with D2. Only MII-stage oocytes showed a considerable number of affected CPs between the genotypes. There were no transgressive CPs. Incomplete dominance and dominance effects on CPs were the most prevalent, with significant z-scores (n = 26 and 14, respectively; Fig. 6) and dominance having 14 CPs with significant z-scores (Fig. 6).

Fig. 5.

Upstream regulators (UR) identified by Ingenuity Pathway Analysis D2/B6 GV- and MII-stage oocytes. Incomplete Dominance URs are when D2 is different from B6 (P < 0.05 and z-score), but neither parental strain is different from BDF1. Dominance URs are when D2 is different from B6 (significant P value and z-score), and either D2 or B6 is also different from F1 (P < 0.05). Chemical reagents, drugs, and toxicants were removed. List limited to URs with significant z-score. z > 1.96 or z < −1.96 indicated activated and inhibited, respectively in the ratio (i.e., z-score of 2 indicates activated in D2 oocytes compared with B6 oocytes). UR list is restricted to those with FPKM > 0.02 in at least one sample.

Fig. 6.

Canonical pathways (CP) identified by IPA analysis D2/B6 MII-stage oocytes. Incomplete Dominance CPs are when D2 is different from B6 (P < 0.05 and z-score), but neither parental strain is different from BDF1. Dominance CPs are when D2 is different from B6 (significant P value and z-score), and either D2 or B6 is also different from BDF1 (P < 0.05). List is limited to CPs with significant z-score. z > 1.96 or z < A−1.96 indicated activated and inhibited, respectively in the ratio (i.e., z-score of 2 indicates activated in D2 oocytes compared with B6 oocytes). Sig., signaling; compl, complex; reg, regulation; med, mediated; card., cardiac.

BDF1 oocytes display a combination of B6-like and D2-like gene expression patterns, i.e., additive dominance.

Recognizing that BDF1 oocytes manifest both high and low B6-like and D2-like dominance effects on mRNA expression concurrently at each stage, and that the effects on phenotype would therefore be the additive output of these differences, we created “additive dominance” (AddDom) gene lists for each stage (B6-dominant and D2-dominant DEGS combined). We then applied IPA to the B6- and D2-dominance DEG lists separately and to the AddDom lists to determine the extent to which URs and CPs associated with dominance effects in BDF1 oocytes were driven by simple effects of either parental genetic contribution acting alone, or by additive dominance effects.

At the GV stage, there were many URs significantly associated with B6- and D2-dominance DEGs sets, but 37 additional URs emerged only when we analyzed the AddDom list. Fifteen of these 37 URs encompassed at least five associated DEGs (Table 4). Thirty-six URs only rose to significance at the MII stage using the AddDom list, of which 14 had five or more associated DEGs (Table 4). The top three of these URs with the most associated DEGs were the cytokine IL4, and two transcription regulators (CREB1, EP300) (Table 4). Thirty-eight CPs only rose to significance using the AddDom list analysis at the GV stage (Table 5). Among these were NRF2-mediated oxidative stress response and role of NFAT in cardiac hypertrophy (Table 5). In MII-stage oocytes, five CPs only rose to significance with the AddDom list, four with at least five associated DEGs, including Gαq signaling, mTOR signaling, and CXCR4 signaling (Table 5).

Table 4.

Upstream regulators at GV and MII that are only significant in the additive dominance category

GV			MII
Upstream Regulator	−log10(P)	Associated DEGs	Upstream Regulator	−log10(P)	Associated DEGs
UCHL1	2.05	4	CD28	1.79	13
EGR2	1.82	6	FOXO3	1.73	12
HIST1H1T	1.57	4	IL5	1.68	11
Hist1h1a	1.57	4	CREB1	1.57	18
ADORA2A	1.53	5	MYOC	1.56	5
E2F4	1.52	7	FOXO4	1.47	5
STAT5a/b	1.50	4	CXCL12	1.46	8
CAV1	1.49	5	APC	1.42	6
CNR1	1.45	5	EP300	1.39	14
SOD2	1.44	4	INSR	1.37	13
IL6ST	1.43	3	IL4	1.33	24
HES1	1.41	3	MYCN	1.32	10
IL32	1.41	3
HIF1A	1.41	10
HTT	1.33	15

Chemical reagents, drugs, and toxicants were removed. Cellular upstream regulators restricted to those with FPKM > 0.02 in at least one sample. No UR listed had significant −log₁₀(P) values in either the B6 or D2 dominance lists alone. DEG, differentially expressed gene.

^*Activated in additive dominance category.

Table 5.

Canonical pathways at GV and MII that are only significant in the additive dominance category

Ingenuity Canonical Pathway	−log10(P)	Associated DEGs
GV
Insulin receptor signaling	2.01	5
Glioblastoma multiforme signaling	1.74	6
14-3-3-mediated signaling	1.57	5
PKCα¸ signaling in T lymphocytes	1.54	5
Th1 pathway	1.52	5
Gα12/13 signaling	1.52	5
NRF2-mediated oxidative stress response	1.41	6
Role of NFAT in cardiac hypertrophy	1.41	6
Th2 pathway	1.35	5
MII
Gαq signaling	1.93	8
mTOR signaling	1.42	8
CXCR4 signaling	1.42	7
Cholecystokinin/gastrin-mediated signaling	1.37	5

Canonical Pathways shown only if 5 target DEGs were identified.

Analysis of dominance and incomplete dominance in BDF1 oocytes according to a nonthresholding approach.

The above analyses of dominance and additive dominance required the application of specific expression value thresholds to define dominance DEGs. The use of such thresholds creates a degree of imprecision because including/excluding genes near the threshold can affect outcome. To overcome the need for such thresholding, we devised an alternate approach to identify genes displaying dominance. This process began by generating groups of five “predicted BDF1” RNA-Seq expression libraries for each stage by averaging the gene expression levels between randomly selected B6 and D2 parent libraries (Fig. 4). This was repeated multiple times using different random pairings of parental libraries to produce the five predicted BDF1 transcriptomes. The averages of the two parental expression values represented the theoretical expression values for genes showing incomplete dominance (i.e., intermediate expression values) in BDF1 oocytes if both parental strains contributed equally to the level of gene expression within the BDF1 oocytes. Comparing the predicted and actual transcriptomes yielded a list of DEGs that departed significantly (q < 0.05) from the “predicted” expression values, and thus displaying dominance effects.

At the GV stage, we observed 32 DEGs comparing actual and predicted BDF1 transcriptomes (Table 6, Supplemental Table S5). Only 19 of these were identified in the analysis above using the thresholding approach with two displaying dominance and 17 displaying transgressive gene expression pattern. IPA analysis of all 32 DEGs showed significant effects on 14 upstream URs. Two of these [estrogen receptor (ESR) and V-Ha-Ras Harvey rat sarcoma viral oncogene homolog (HRAS)] had at least three affected target DEGs, but neither remained significant at the MII stage. Two CPs (protein kinase A signaling and synaptic long-term depression) had at least three affected DEGs, but neither of these remained significant at the MII stage.

Table 6.

DEGs between actual and predicted BDF1 transcriptomes

	GV	MII
Actual F1 > Predicted F1	9	49
Actual F1 < Predicted F1	23	158

Top Affected DEGs between Actual/Predicted MII-stage Oocytes
Gene	Fold-Change
Kcnip3	8879210.66
Gm40235	129.79
Mmp14	−80.06
Zfp622	−70.42
Bvht	69.60
Fads2	−50.74
Atxn7l3	−48.27
Arl9	−40.06
Pcdhb16	−32.07
Rtp4	−31.78
Qrsl1	−30.65
Jade1	−30.19
Calcb	−28.07
Mfge8	−27.72
Eid1	−25.16
Hexa	−23.69
Lrrc8c	−22.93

Fold-change restricted to DEGs with >20-fold change.

At the MII stage, the analysis revealed 199 DEGs differing between actual and predicted BDF1 transcriptomes, including 16 that had a >20-fold difference between the actual vs predicted (Table 6, Supplemental Table S6). Of these 199 DEGs, only 67 were previously identified in the analysis by the thresholding approach, including 22 in the transgressive gene category. IPA analysis of the 199 DEGs revealed 38 significantly affected URs that had least three associated DEGs, six of which displayed significant z-scores (Table 7). This included three URs inhibited in the actual BDF1 compared with the predicted BDF1 transcriptome (RICTOR, FFAR3, and ACOX1) and three activated URs (IGF1R, INSR, and MAP3K1) (Table 7). We observed seven altered CPs (with three associated DEGs) including the sirtuin signaling pathway that was decreased in actual BDF1 MII oocytes compared with the predicted BDF1 MII transcriptome. One UR (INSR) and one CP (sirtuin signaling) were the only IPA results that had similar significant z-scores from using both methods of analysis.

Table 7.

Upstream regulators with different activation/inhibition state in actual BDF1 compared with predicted BDF1 MII oocytes

Upstream Regulator	Molecule Type	Predicted Activation State	z-Score	−log10(P)	Target DEGs
RICTOR	other	inhibited	−3.000	3.77	↓Atp5O, ↓Cox4i1†, ↓Cox5b*, ↓Rpl41*, ↓Rplp1, ↓Rps19, ↓Rpsa, ↓Uba522†, ↓Uqcrq†
IGF1R	transmembrane receptor	activated	2.236	1.30	↓Atp5O, ↑Bcl2, ↓Cox4i1†, ↓Cox5b*, ↓Sfmbt2*
INSR	kinase	activated	2.213	1.54	↓Acadvl, ↓Atp5O, ↓Cdc5l*, ↓Cox4i1†, ↓Cox5b*, ↓Mbd1*, ↓Sfmbt2*
FFAR3	G protein-coupled receptor	inhibited	−2.000	2.54	↓Fads2†, ↓Foxp2*, ↓Pitpmn3, ↓Trpc4
ACOX1	enzyme	inhibited	−1.982	1.64	↓Acadvl, ↓Cd63, ↓Mfge8, ↓Pigp
MAP2K1	kinase	activated	1.980	1.98	↑Bcl2, ↓Gli2†, ↓Itgb4*, ↓Mmp14, ↓Ubc*

Chemical reagents, drugs, and toxicants were removed. Significant z score (z > 1.96 or z < −1.96) indicates activated or inhibited in actual F1 MII oocytes.

^*Displays dominant expression in F1 oocytes.

^†Displays TGE expression in F1 oocytes.

Although IPA indicated an effect on sirtuin signaling using both methods of analysis, we noted that none of the associated sirtuin mRNAs implicated by IPA for this CP (sirt1–7) were identified as DEGs, suggesting that the underlying reason for the IPA result for sirtuin signaling rested at the level of differentially expressed downstream genes associated with the pathway. Several of these downstream genes were in the histone 1H gene cluster (Hist1h1c, Histh1e, Hist1h1d). The expression levels of mRNAs for 21 of 55 members of this gene cluster were significantly downregulated during maturation only in BDF1 oocytes (Table 8). The mRNA for one member (Hist1h1t) increased in abundance during maturation.

Table 8.

Hist1h mRNAs only significantly altered in BDF1 oocytes during meiotic maturation

	GV vs. MII (fold-change)
Gene	B6	D2	F1
Hist1h1d	−1.83	−1.17	−2.53
Hist1h1t	1.16	1.40	25.40
Hist1h2aa	1.30	−1.32	−1.61
Hist1h2ac	−1.10	−1.58	−2.08
Hist1h2ae	−1.25	−1.59	−2.35
Hist1h2ag	−1.18	−1.61	−2.22
Hist1h2ah	−1.20	−1.58	−2.36
Hist1h2an	−1.17	−1.61	−2.45
Hist1h2bb	1.24	−1.21	−1.37
Hist1h2bc	−1.15	−1.10	−1.27
Hist1h2be	1.11	−1.22	−1.50
Hist1h2bf	1.19	−1.21	−1.41
Hist1h2bg	1.18	−1.07	−1.35
Hist1h2bh	1.22	−1.03	−1.31
Hist1h2bl	1.10	−1.21	−1.49
Hist1h2bm	−1.02	−1.35	−1.41
Hist1h2bn	1.25	−1.08	−1.46
Hist1h2bp	1.10	−1.17	−1.37
Hist1h2bq	1.16	−1.21	−1.41
Hist1h2br	1.18	−1.16	−1.36
Hist1h4d	−1.16	−1.22	−1.74

DISCUSSION

The main discoveries of these studies are fourfold. First, there is a large amount of difference in oocyte transcriptomes between the two inbred parental strains. Second, the different genotypes regulate maternal mRNAs differently during maturation. Third, a key aspect of superior BDF1 hybrid phenotype is attributable to additive dominance effects, with additional input by genes displaying incomplete dominance, and a much lesser degree of effect of transgressive gene expression (affecting <30 mRNAs). Fourth, the IPA analysis reveals effects of the BDF1 hybrid genotype on a discrete set of CPs and URs. Many of these effects only became apparent using a novel computational approach for assessing dominance effects in F1 hybrids.

Extensive differences between parental strains and between parental strains and BDF1 hybrids.

The vast differences between the transcriptomes of the three genotypes is surprising, given that all three genotypes are fertile. The transcriptomes differed more between the two parental strains than between either parental strain or the BDF1 hybrid. BDF1 oocytes displayed an intermediate level of expression for many genes. Despite this, BDF1 oocytes displayed many unique affected processes that distinguish them from the parental strain oocytes. Although the oocytes of the three genotypes display many differences in phenotypic characteristics, this vast difference in mRNA expression might not have been predicted a priori. We observed differences in the expression of prominently expressed mRNAs and mRNAs for genes known to be important in oogenesis and meiosis, such as genes encoding proteins with roles in chromosome alignment [AURKB (56)], sister-chromatid separation [CDC20 (20)], microtubule-organizing center stretching [KIF11 (5)], zygotic genome activation [YAP1 (67)], and microtubule-organizing center formation [CEP192 (28)].

These differences are accompanied by differences in level of activation/increase or inhibition/decrease in URs and CPs between parental strains. This includes SUMO3, a contributor to GVBD and spindle formation in the oocyte (6, 10), and VEGF, which supports preovulatory angiogenesis (31), and the pathways for sumoylation, signaling by RHO family GTPases, and mTOR signaling.

We also observed extensive differences in mRNA regulation during maturation. While the number of mRNAs undergoing changes in abundance were somewhat similar, BDF1 hybrid oocytes displayed more changes during maturation (35%) than B6 (29%) or D2 (30%) oocytes. Additionally, the distribution of changes between relative increases and decreases in mRNA abundance during maturation varied with strain, with >60% of the DEGs declining in abundance in maturing B6 and BDF1 oocytes, but <40% declining in D2 oocytes. This indicates profound genotype effects on maternal mRNA translation and degradation, further emphasizing the large variation possible between fertile oocytes.

Such large genetic differences in mRNA regulation raise new concerns about the use of single or small numbers of molecular markers of oocyte quality, by indicating that putative markers may vary in expression but have little effect on phenotype due to variations in expression of other genes. We compared our DEG lists to lists of putative oocyte quality markers reported in other studies, and found that a substantial fraction of these [36% bovine (42a), 46% rhesus monkey (30), and 50% human (13)] are differentially expressed between the mouse genotypes in this study. Additionally, for 25% of the primate and 36% of the human oocyte-quality markers the fold-change differences between mouse genotypes exceeded the fold-change differences between the different quality oocytes. This indicates that the genetic variation in expression of these genes between healthy oocytes could limit their utility as markers of oocyte quality. Consequently, knowledge of genetic modifiers and compensatory mechanisms that confer fertility even in the face of dramatic variation in particular transcript levels and the value of putative markers needs to be assessed across genotypes. Moreover, studies that address impacts of environmental factors, maternal health and nutrition, and other oocyte-extrinsic parameters must address genetic variability before being widely generalized.

Additive dominance as a major origin of BDF1 hybrid phenotype.

Additive dominance appears to be a major means by which the BDF1 oocytes may acquire unique phenotypic characteristics. The potential mechanism (i.e., affected cellular processes) responsible for additive dominance effects has not been evaluated previously in oocytes. Doing so here required combining individual B6 vs. BDF1 and D2 vs. BDF1 dominance DEG lists into a single DEG list and subjecting that list to IPA analysis. This approach revealed a number of significant effects on CPs and revealed changes in activation/inhibition states of several URs. There were many URs and CPs that are only reached significance in the AddDom category (Tables 4 and 5). These included URs with roles in regulating the oocyte cortex [UCHL1 (41)], oocyte maturation, [CNR1 (34)], oxidative stress protection [SOD2 (68)], and initiation of oocyte growth [FOXO3 (21)]. These also included CPs with roles in maintenance of quality in aged oocytes [insulin receptor signaling (59), maintenance of meiotic arrest [14-3-3-mediated signaling (38)], and first meiotic spindle migration [mTOR signaling (29)].

Further evidence of additive dominance effects was provided by the comparison of actual BDF1 transcriptome to “predicted” BDF1 transcriptome expression values. While there was some overlap between the additive dominance IPA results (UR and CP results with significant z-scores) comparing transcriptomes directly and the IPA dominance results obtained using the actual BDF1 versus predicted BDF1 comparison, the latter analysis yielded additional significantly affected CPs and URs, some of which possessed a significant z-score and others that did not. This novel computational approach, which avoids the need to apply expression value thresholds to define dominance effects, may be broadly applicable in understanding hybrid genotype effects.

We note that fewer DEGs were identified as dominance DEGs in the dominance (n = 896 genes) or the actual versus predicted (n = 239 genes) analyses than incomplete dominance DEGs (10,377 genes). However, this small number of DEGs collectively appears to make a significant contribution to the phenotypic characteristics of BDF1 oocytes as we identified URs and CPs that were significantly altered in BDF1 oocytes. One CP, sirtuin signaling, was identified in both analyses. Because expression of the relevant Sirt mRNAs was not substantially altered in BDF1 oocytes, effects on this CP is likely be mediated by differences in the expression of downstream target DEGs. Among these were several genes located in the histone 1H gene cluster on mouse chromosome 13. Further analysis showed that BDF1 oocytes selectively degrade nearly half the histone 1H cluster mRNAs during meiotic maturation. Histone variants have important roles in embryonic development, stem cell formation, cell plasticity, and reprogramming (53, 55, 63). The differential regulation of the histone 1H cluster mRNAs in BDF1 oocytes during maturation may support a unique chromatin architecture in BDF1 oocytes, and/or reflect differential chromatin regulation after fertilization, either of which could be a key contributor to the hybrid phenotype.

Several mechanisms may contribute to dominance effects in BDF1 hybrids, including interstrain differences in genomic imprinting, genetic variation in promoter or enhancer strength impacting transcription rate, and genetic polymorphisms affecting mRNA stability. Two imprinted genes displayed dominance at the GV stage (Igf2 and Plagl1) and two at the MII stage (Smfbt2 and Zrsr1), but all four are regulated in a manner that suggests differential control of mRNA stability rather than a genome effect on imprinting. Of the 528 genes displaying dominance at the MII stage, only 47 did so at the GV stage as well. This suggests that the vast majority of dominance effects displayed at the MII stage are related to mRNA stability and thus may be driven by differences in mRNA sequence, perhaps acting in concert with genetic variants controlling mRNA translation and degradation.

Only one gene at the GV stage and <30 genes at the MII stage displayed transgressive expression according to a thresholding method for defining such expression. Interestingly, the vast majority of TGE genes were expressed at their lowest abundances in BDF1 oocytes. Although the small number of TGE genes identified indicated that TGE is not a major contributor to the special BDF1 oocyte characteristics, four TGE genes (Coc4l1, Uba52, Fads2, and Gli1) were DEGs that contributed to the activated/inhibited URs in the comparison between actual and predicted BDF1 transcriptomes (Table 7). In the same comparison, five DEGs that contributed to the these activated/inhibited URs were dominant DEGs (Cox5b, Rpl41, Roxp2, Itgb4, and Ubc). Taken together, these results indicate that these few TGE genes may cooperate with the dominance genes to establish the BDF1 oocyte phenotype.

Unique BDF1 phenotypes might also arise through incomplete dominance of combinations of genes. This was evident in the overall number of DEGs between strains and more readily apparent in the comparison of actual versus predicted (blended values) BDF1 oocyte transcriptomes, in which expression values for just 32 and 199 genes departed from the “predicted” values at the GV and MII stages, respectively. This extensive blending of gene expression values may contribute to hybrid vigor, raising the possibility that the expression values observed in the inbred parental strains for many genes may not be advantageous to overall fertility.

UPs and CPs associated with BDF1 characteristic phenotype.

Using multiple approaches, we observed URs and CPs that were uniquely affected in BDF1 oocytes compared with parental strain oocytes, which may contribute to BDF1 hybrid vigor. Several of the URs uniquely affected in BDF1 oocytes have well-known roles in oocyte maturation [UCHL1 (41), CNR1 (34), HES1 (37)]. Other URs uniquely affected in BDF1 oocytes have roles in ovulation [IL6ST (39)], are oocyte secreted factors [INHBB (7)], or may arise from a combination of somatic cell and oocyte expression [CD44 (65)]. Additionally, BDF1 oocytes may differentially regulate the expression of transcription factors with roles in early embryos [E2F (43), EGR1 (14), and STAT5A/B (42)]. Other affected URs (ETV4, LASP1, EGR2, ADORA2A, IL32, FFAR3, MTM1) have not been previously associated with oocyte function but emerged here as new candidates for controlling oocyte characteristics. Other URs uniquely affected in BDF1 oocytes are related to diverse cellular functions important to oocyte quality such as oxidative phosphorylation, redox state, oxidative stress, mitochondrial turnover, and apoptosis [BNIP3L (1), S1PR3 (15), SOD2 (46), HIF1A (64)]. Interestingly, BDF1 MII-stage oocytes showed gene expression patterns consistent with activation of IGF1R and INSR, two proteins that are not necessary for oocyte maturation in mice (47) but that may contribute to survival under conditions of stress (36). Greater activation states of pathways involving INSR and IGF1R may contribute to resiliency in hybrid oocytes. One other interesting UR, ACOX1, is an enzyme that regulates β-oxidation and causes sterility when lost (9) but remains poorly studied in the oocyte. Maturation changes specific to BDF1 oocytes suggest that ACOX1 becomes inhibited during oocyte maturation. ACOX1 is also observed in the actual vs. predicted data set where it appears to be more inhibited in the actual BDF1 oocytes at MII than would be predicated by the parental expressions. An interesting future direction would be to better characterize the role of ACOX1 in during oocyte maturation to determine if it contributes to BDF1 hybrid vigor.

Along with these affected URs, there are a number of important CPs uniquely affected in the BDF1 oocytes. These include INSR signaling, 14-3-3 signaling, and mTOR. All of these pathways are connected to INSR1 and IGF1R, identified above as affected URs (40, 66). The analysis also indicated altered responses of BDF1 oocytes to cumulus cells via CXCR4 (69). Another CP (sirtuin signaling) likely to be key in oocyte health emerged in the comparison between actual and predicted BDF1 transcriptomes as being decreased in activity.

Overall Conclusions

To our knowledge, this is the first RNA-Seq study comparing oocytes from different maturational stages from three fertile genotypes of mice. The major findings of this study are fourfold. First, there are numerous DEGs between each mouse genotype at both oocyte stages. Second, the three genotypes differ in mRNA handling during maturation. Interestingly, BDF1 oocytes have more changes in mRNA abundances during maturation than either parental strain. Oocytes from the three genotypes become more different from each other during maturation. Third, additive dominance is a major source of unique BDF1 oocyte characteristics, working in conjunction with a limited number of genes showing transgressive RNA expression, and with a large degree of incomplete dominance. Last, these differences in mRNA expression are associated with changes in the predicted activation states of important URs and CPs related to oocyte health and function, impacting prominent functions such as histone expression, mitochondrial function, and oxidative phosphorylation. These results highlight the complexity of the oocyte, the distinct paths that the genetically different oocytes may take during oocyte maturation, and how BDF1 oocytes optimize this system to develop superior phenotypic characteristics. Such profound effects of genotype on the oocyte transcriptome highlight a need for considerable caution in efforts to identify molecular markers of oocyte quality and apply them on a broad basis in clinical or applied practices.

GRANTS

This work was supported in part by grants from the Eunice Kennedy Shiver National Institute of Child Health and Human Development of the National Institutes of Health under the award numbers T32HD-087166 and RO1HD-075903, by MSU AgBioResearch, and by Michigan State University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

K.E.L. conceived and designed research; A.L.S. performed experiments; A.L.S., U.M., and K.E.L. analyzed data; A.L.S., U.M., and K.E.L. interpreted results of experiments; A.L.S. prepared figures; A.L.S. and K.E.L. drafted manuscript; A.L.S., U.M., and K.E.L. edited and revised manuscript; A.L.S., U.M., and K.E.L. approved final version of manuscript.