What makes a bad egg? Egg transcriptome reveals dysregulation of translational machinery and novel fertility genes important for fertilization

Abstract

Background: Egg quality can be defined as the egg ability to be fertilized and subsequently develop into a normal embryo. Previous research has shed light on factors that can influence egg quality. Large gaps however remain including a comprehensive view of what makes a bad egg. Initial development of the embryo relies on maternally-inherited molecules, such as transcripts, deposited in the egg during its formation. Bad egg quality is therefore susceptible to be associated with alteration or dysregulation of maternally-inherited transcripts. We performed transcriptome analysis on a large number (N = 136) of zebrafish egg clutches, each clutch being split to monitor developmental success and perform transcriptome analysis in parallel. We aimed at drawing a molecular portrait of the egg in order to characterize the relation between egg transcriptome and developmental success and to subsequently identify new candidate genes involved in fertility.

Results: We identified 66 transcript that were differentially abundant in eggs of contrasted phenotype (low or high developmental success). Statistical modeling using partial least squares regression and genetics algorithm demonstrated that gene signatures from transcriptomic data can be used to predict developmental success. The identity and function of differentially expressed genes indicate a major dysregulation of genes of the translational machinery in poor quality eggs. Two genes, otulina and slc29a1a, predominantly expressed in the ovary and dysregulated in poor quality eggs were further investigated using CRISPR/Cas9 mediated genome editing. Mutants of each gene revealed remarkable subfertility whereby the majority of their eggs were unfertilizable. The Wnt pathway appeared to be dysregulated in the otulina mutant-derived eggs.

Conclusions: Here we show that egg transcriptome contains molecular signatures, which can be used to predict developmental success. Our results also indicate that poor egg quality in zebrafish is associated with a dysregulation of (i) the translational machinery genes and (ii) novel fertility genes, otulina and slc29a1a, playing an important role for fertilization. Together, our observations highlight the diversity of the possible causes of egg quality defects and reveal mechanisms of maternal origin behind the lack of fertilization and early embryonic failures that can occur under normal reproduction conditions.

Keywords: Differentially expressed genes; Egg quality; Microarray; Prediction model; Transcriptome; Zebrafish.

Fig. 1

a: Heat map showing unsupervised clustering of the 66 differentially expressed genes (DEGs) between good and bad quality eggs from 32 clutches of fertilized zebrafish eggs. Yellow signal denotes upregulation, blue signal denotes downregulation, and black defines no change in expression. b: Gene ontology analysis using the DAVID online program of the 55 DEGs with known information. The enriched terms are shown on the y-axis and the fold enrichment is on the x-axis. Annotated terms are derived from UniProtKB keywords (pink circle); Gene Ontology BP (red circle), MF (blue circle), and CC (yellow circle); and KEGG pathways (green circle). Statistical significance is represented by the colored squares next to the enriched terms

Fig. 2

Validation of the microarray data by performance of quantitative real-time PCR (qPCR). Eight genes, including (a) rpf2, (b) spon1b, (c) tspan7b, (d) rps27.2, (e) stra13, (f) rtn4ip1, (g) U1, and (h) slc29a1a were subjected to qPCR using the primers listed in Additional file 2, whereby LSM couples member 14B (lsm14b), prefoldin subunit 2 (pfdn2), and ring finger protein 8 (rnf8) as well as 18S rRNA, beta-actin (bact), and elongation factor 1 alpha (EF1α) were used as internal controls. * p-value ≤0.05, ** p-value ≤0.01, *** p-value ≤0.001, **** p-value<< 0.001

Fig. 3

Tissue localization of otulina (a) and slc29a1a (b) based on qPCR assays. c: Expression level of otulina and slc29a1a in spawned eggs from mutant females mated with WT males as assessed by qPCR. 18S rRNA, beta-actin (bact), and elongation factor 1 alpha (EF1α) were used as internal controls, and experiments performed in triplicate. d: Developmental success in terms of survival rate of embryos at 24 h post-fertilization (hpf) from otulina- and slc29a1a-deficient mutant females mated with WT males. N = 4 r different females for otulina and N = 10 different females for slc29a1a, using eggs from at least three spawns for each individual female

Fig. 4

Representative images showing the development between 0 and 24 h post-fertilization (hpf) of F1 embryos from wildtype control (a-d), otulina-deficient (e-h), and slc29a1a-deficient (i-q) females. In the control eggs, the embryos were at 64-cell (A), oblong (b), shield (c), and 24-somite (d) stages according to Kimmel et al. [56]. Eggs from otulina and slc29a1a mutant females were non-developing and did not under any cell division (E-L). Some eggs from two slc29a1a mutant females were developing abnormally (M-Q). (a, e, i, m) = images taken at 2 hpf; (b, f, j, n) = images taken at 4 hpf; (c, g, k, o) = images taken at 6 hpf; (P) = image taken at 8 hpf; (d, h, l, q) = images taken at 24 hpf. The arrow demonstrates a partially cellularized blastodisc that was sitting atop an enlarged syncytium. Scale bars denote 500 μm. R: PCR genotyping for nucleoplasmin 2b (npm2b) and vasa:eGFP in spawned eggs from WT, otulina-, and slc29a1a-mutant females crossed with vasa:eGFP males to detect fertilization of the eggs. Std = 1 kb ladder; Con = WT female crossed with vasa:eGFP male

Fig. 5

Evaluation of the expression levels (arbitrary units) of wnt3a (a), tcf7 (b), lef1 (c), and dvl2 (d) in spawned eggs from otulina-deficient mutant females mated with WT males as assessed by qPCR. 18S rRNA, beta-actin (bact), and elongation factor 1 alpha (EF1α) were used as internal controls, and experiments performed in triplicate. N = 4 different females, at least three spawns from each female

Fig. 6

a The average 2-fold cross validation R² values obtained from the actual dataset were compared to the ones obtained from the pseudo-datasets with permuted survival rates. b: The frequency that each variable was selected in populations from the actual data and from the randomized data. The 95th and 99th percentiles of the distribution of frequencies in the randomized data were used to obtain sets of genes that were the most often selected