Introduction

Existing approaches for determining egg quality focus on either measuring parameters that can predict the viability of fertilized eggs (e.g. blastomere morphology, Shields et al. 1997) or examining the abundance and composition of certain compounds (e.g. vitamins, Palace and Werner 2006; fatty acids, Salze et al. 2005) that may affect egg quality. However, how these egg quality measurements relate to the normal development of fertilized eggs is not well understood.

The ovulated egg is transcriptionally inactive, and rapid early postfertilization development is strongly dependent on stored proteins and transcripts (see review by Li et al. 2013). As such, these stored entities are likely to have a notable bearing on the ‘quality’ of the egg, which has been exploited in fish through transcriptomics (Ma et al. 2012; Lanes et al. 2013; Chapman et al. 2014) and proteomics studies (Rime et al. 2004; Crespel et al. 2008; Papakostas et al. 2010; Castets et al. 2012). The first study to evaluate both the transcriptome and the proteome of essentially the same sample of fully grown ovarian zebrafish follicles concluded that both omics approaches yielded very different information (Knoll-Gellida et al. 2006). Moreover, the transcriptome was not predictive of protein abundance in the proteome, highlighting the need for both proteome and transcriptome analyses. Presently, transcriptomics studies on ovulated eggs from oviparous vertebrates seem to be more numerous than comparable proteomics studies, warranting proteomics analyses to gain novel insights into egg quality during very early embryonic development.

A global proteomic approach aims at the identification, and often quantification, of entire protein complements in various biological systems and thus provides insights into the relationships of proteins within relevant processes such as metabolic and regulatory pathways (Cerda et al. 2008). Furthermore, as a global profiling approach, proteomics is complementary to (rather than repeating) mRNA expression profiling at a different level (c.f., Knoll-Gellida et al. 2006). Often mRNA levels do not fully represent the processes researchers aim to investigate. Discrepancies between corresponding mRNA and protein levels are apparent and may depend on various regulatory mechanisms such as translational regulation, protein stability and protein transport processes (Anderson and Anderson 1998; Martyniuk and Denslow 2009; Schwanhäusser et al. 2011).

Proteomic methods have only recently been applied to questions related to embryogenesis, toxicology, biomedicine, aquaculture, environmental effects and food safety in fish (Reviewed in Forné et al. 2010; Sanchez et al. 2011; Rodrigues et al. 2012; Lokman and Symonds 2014). Most of these studies illustrate the power of proteomic methods in describing the relevant processes that influence development and the differences between normal embryos and those that have been exposed to toxic agents. These studies were undertaken mostly in zebrafish (Danio rerio) and salmonids (reviewed in Forné et al. 2010). To date, most studies have used 2D gel electrophoresis-based approaches (Forné et al. 2010). However, gel-free protein quantification by mass spectrometry using stable isotope labelling techniques, such as iTRAQ, allows simultaneous analysis of up to eight samples that have been labelled differentially with unique isobaric tags. The iTRAQ approach facilitates not only the differential profiling of relative protein abundances but also simultaneous global protein identification and therefore provides information of regulated and nonregulated proteins. Such a global picture of protein expression is usually not achieved by gel-based approaches. (Zieske 2006). Indeed, iTRAQ was recently viewed as a sensitive technique that can be employed very successfully even when full genome sequence data are lacking (Martyniuk et al. 2012).

Of the studies mentioned above, only a few are directly related to oocyte and egg quality, and these are all gel-based studies. Differences were identified in the abundance of various proteins, albeit without multiple statistical testing, in oocytes (Yoshikuni et al. 2003), coelomic fluid (Rime et al. 2004), fertilized eggs (Crespel et al. 2008) and larvae (Castets et al. 2012), and the suitability of proteomics in identifying protein targets that are involved in processes affecting egg and larval quality in fish could be demonstrated. However, due to the complexity of this biological system, more comprehensive approaches are required to gain deeper insights into the underlying mechanisms.

In hapuku (Polyprion oxygeneios) aquaculture, fertilized eggs show variable quality with some batches containing mostly poorly developing embryos, as assessed by blastomere morphology scoring (Kohn and Symonds 2012). Elucidating the biochemical processes that lead to poor development could benefit the reproductive success of this species in captivity and may be a step forward in addressing the common constraint of poor egg quality in finfish aquaculture.

The main aims of the present study were to (1) describe the proteome of early-stage embryos of hapuku with regard to protein functions and biochemical pathways and to (2) correlate the abundance of these proteins with blastomere symmetry to find potential target proteins/pathways that are associated with egg quality.

Materials and Methods

Sampling

Hapuku broodstock were reared at the National Institute of Water and Atmospheric Research (NIWA) aquaculture facility in Bream Bay, Ruakaka, New Zealand. Hapuku are serial batch spawners, and each spawning female produces multiple batches of eggs during the spawning season in spring. Wild-caught broodstock of both sexes, most of which had spawned in previous years, were housed in communal tanks and allowed to spawn without intervention. In 2009, fertilized eggs were collected from two broodstock tanks (11 females and 7 males in one tank; 6 females and 9 males in another) and in 2011 from one broodstock tank (9 females and 8 males). As parentage analyses were not conducted, the contribution of the specific individuals to the batches analysed is not known. The protocols for fish rearing and sampling of fertilized eggs were described in Kohn and Symonds (2012).

Eight batches of fertilized eggs (at the 8-cell embryo stage) that exhibited differences in blastomere morphology scores (Kohn and Symonds 2012) were selected during the 2009 (n = 3 batches, 30 embryos/batch) and 2011 (n = 5 batches, ~100 embryos/batch) spawning seasons. The embryos were photographed, and the blastomere symmetry was given a quality score of between ‘0’, being the best quality with perfect symmetry, and ‘4’ representing the worst quality, for completely asymmetric blastomere patterning—in doing so, we obtained egg batches from hapuku (Fig. 1) that ranged between ‘moderately good’ (score = 1.50) to ‘bad’ (3.50); we refer to these batches in italics text (e.g. Table 1) as good (G), moderate (M) and bad (X), followed by the quality score (e.g. M2.45 represents a batch of moderate quality with a score of 2.45). All methods used followed those detailed in Kohn and Symonds (2012), except that only symmetry was scored as a measure of egg quality, as validated previously (Kohn and Symonds 2012). The collected eggs were snap-frozen in liquid nitrogen and stored at −70 °C prior to protein extraction.

Fig. 1
figure 1

Batches of fertilized eggs from hapuku, P. oxygeneios, displaying blastomeres with different degrees of symmetry. Eggs were scored as moderately good (a; quality score = G1.77), moderate (b; quality score = M2.45) or bad (c; quality score = X3.50). For details on egg quality scores, see text

Full size image
Table 1 Blastomere quality scores (‘quality’) of the eight batches of eggs from hapuku (P. oxygeneios) used for iTRAQ proteomic analysis
Full size table

Protein extraction

Approximately 500 mg of 8-cell-stage embryos from each batch were placed in liquid nitrogen and promptly crushed using a pestle (BIO PLAS, USA) and homogenized in ×4 (v/w) phosphate buffer (10 mM, pH 7.4) and ×1 (v/w) protease inhibitor (×10, Complete Mini, Roche, Germany). Samples were then homogenized further using an electric homogenizer (IKA T25B Ultra TURRAX, Malaysia) and finally sonicated (Ultra Tip, Wave Energy Systems, USA) for 3 × 30 s. Homogenized samples were centrifuged at 10,600g (25 min, 4 °C), and the supernatant, excluding the floating lipid fraction, was retained for iTRAQ analysis (see below). Protein contents of the supernatants were estimated using a bicinchoninic acid assay (Sigma-Aldrich, http://www.sigmaaldrich.com) on a plate reader (FLUOstar Omega, BMG Labtech, Germany).

iTRAQ analysis

The 8-plex iTRAQ method (Applied Biosystems, USA) used in the present study utilized eight different stable isotope-labelled isobaric tags that modify peptides (after proteins were trypsin-digested) at the N terminus and the lysine side chain, as well as the tyrosine side chain at a lower frequency. After differential labelling (50 μg protein/sample), all eight samples were pooled and analysed by liquid chromatography-coupled tandem mass spectrometry. Due to the isobaric nature of the tag, the same peptides from different samples behave as one species in liquid chromatography, ionization and mass measurement. Upon peptide fragmentation by collision-induced dissociation (CID)-based ion activation, however, tag-specific reporter fragment ions are released and recorded in the fragment ion spectrum. The relative intensities of the eight different reporter fragment ions are used for quantification and other peptide-specific fragment ions for identification (Ross et al. 2004). Accordingly, the iTRAQ approach only reveals relative protein quantities unless the absolute amount of proteins is known for one of the samples. In this study, we focused our attention on soluble proteins from 8-cell-stage embryos. Although the iTRAQ approach is suitable for the quantification of membrane proteins, our protein extraction was biased towards water-soluble proteins. In future studies, the screen for potential target proteins may be extended to membrane protein fractions.

Filter-aided sample preparation (FASP) and iTRAQ labelling

All samples were processed using a filter-aided sample preparation (FASP, Wiśniewski et al. 2009) procedure that was modified for compatibility with downstream iTRAQ labelling. In brief, supernatants from each homogenized sample were loaded into 0.5 ml ultrafiltration units (10 kDa cut-off) and buffer-exchanged into 8 M urea in 0.25 M triethylammonium bicarbonate (TEAB) by at least three consecutive centrifugation steps, each reducing the volume down to 20 µl. Samples were then reduced and alkylated on-filter using the reagents from the iTRAQ kit and buffer-exchanged to 0.5 M TEAB by two consecutive centrifugation steps. An aliquot of each sample was taken for protein measurement using the Bradford method. Proteins were digested with trypsin (1 µg trypsin/20 µg protein) on-filter over night at 37 °C. After the first digestion, an additional aliquot of trypsin was added to each sample and samples were incubated for another 4 h. The entire filter content was then transferred to a new tube, and peptides were labelled with the iTRAQ reagents (8-plex kit) according to the manufacturer’s instructions. Samples were pooled following the protein measurement.

Peptide fractionation and LC–MS/MS analysis

To increase the number of peptide identifications and quantification counts per protein in the complex protein extracts of 8-cell-stage embryos, we implemented a two-dimensional peptide prefractionation using off-line peptide isoelectric focusing followed by uHPLC inline coupled to tandem mass spectrometry, a procedure that is compatible with iTRAQ quantification (Chenau et al. 2008). Therefore, peptides of the pooled iTRAQ-labelled sample were subjected to off-gel isoelectric focusing using an Agilent OFFGEL Fractionator (Agilent Technologies) and separated into 24 fractions along a pH gradient from 4 to 10 according to the manufacturer’s instructions. Fractions were concentrated using a centrifugal vacuum concentrator and analysed in duplicate by nanoflow liquid chromatography-coupled LTQ Orbitrap tandem mass spectrometry (LC–Orbitrap MS/MS). One half of each concentrated fraction was loaded onto a trap column and separated on an in-house emitter tip column (75 um ID PicoTip fused silica tubing (New Objectives, USA) packed with C-18 material on a length of 12 cm) through an acetonitrile gradient from 5 to 35 % in aqueous 0.2 % formic acid for 45 min followed by an increase to 99 % acetonitrile for 10 min. The Orbitrap analyser was operated in full MS mode at a resolution (full width at half maximum—FWHM) of 60,000 at m/z 400 followed by eight data-dependent MS/MS scans. The first four MS/MS scans were performed as CID scans on the strongest four precursors for peptide identification followed by four high-energy collision-induced dissociation (HCD) scans of the same four precursors for reporter ion quantification. CID fragment ions were detected in the LTQ ion trap and HCD fragment ions in the Orbitrap analyser at a resolution FWHM of 15,000 at m/z 400. Dynamic exclusion was enabled, allowing two repeated fragmentation analyses on the same precursor during a 3-min time window.

Data analysis

We aimed to identify associations between protein species abundance and egg quality and therefore used regression analysis in order to evaluate how well protein abundance predicts egg quality. Statistical analyses were conducted using SPSS (version 19, IBM®) and R software (R development Core Team 2012). Raw spectra were processed through the Proteome Discoverer software (Thermo Scientific) and then submitted to the Mascot (Matrix Science LTD, UK) and SEQUEST (Eng et al. 1994) search engines for protein identification. Spectra were researched against the Teleostei subset of the NCBI nonredundant amino acid sequence database and its decoy version with reversed sequence entries to estimate the false discovery rate (FDR). The Mascot and SEQUEST search settings were identical and allowed for full tryptic peptides only with two missed cleavage sites and the following variable modifications: iTRAQ labels (N terminus, K, Y), oxidation (M), methylthio (C) and deamidation (N, Q). The score threshold for identification was adjusted to achieve an estimated FDR of <1 %. Only proteins that were identified by two or more peptide hits were accepted. The identified proteins were assigned to one of nine groups according to their function/pathway using the free software KEGG (http://www.genome.jp/kegg/pathway.html). Relative reporter ion intensities from the iTRAQ analysis were standardized against the sample labelled as ‘113’ (G1.50; Table 1). Thus, intensity values were expressed as a ratio. A mixed-effects model analysis implemented in the R package ‘lme4’ (Bates et al. 2011) was used to test for relationships between the relative abundances of the proteins and egg batch quality within each of the nine functional groups of proteins. More specifically, for each protein group, we ran separate random slope mixed models with the abundance as the response (dependent) variable, the egg quality as a fixed effect, the protein identity and the sample identity as random intercepts and the egg quality as a random slope for each protein. A simple linear regression analysis was then performed for each protein separately to identify trends in relative abundance of a given protein in relation to egg quality (with the abundance as the response and the egg quality as a predictor). Performing multiple tests can lead to the increased probability of reporting false-positive results (i.e. significant associations that appear by chance), and in such situation, significance levels need to be adjusted accordingly to the number of tests performed (Manly et al. 2004). A false discovery rate (FDR) correction was, therefore, applied to the p values from the mixed-model analysis and to the p values from the simple regression analysis, using the R package ‘fdrtool’ (Strimmer 2008) in order to control the number of false discoveries in multiple tests.

Results

Mascot/SEQUEST matched the suite of sequenced oligopeptides to 121 proteins with unique NCBI accession ID numbers; of these, 27 were vitellogenins (Vtgs) and 23 were ‘unknown’ or ‘unnamed’. These 23 proteins were identified manually by submitting their unique peptide sequences to NCBI and selecting the first protein hit associated with a teleost. All of the 98 identified proteins were assigned to one of nine groups (amino acid metabolism antioxidants/housekeeping, ATP metabolism, glycolysis, heat-shock proteins, Krebs cycle, other metabolism-related proteins, purine metabolism and pentose phosphate pathway) based on their function and/or relevance to a biochemical pathway. A tenth group included proteins that did not share functionality or pathway (Table 2).

Table 2 Proteins identified from hapuku (P. oxygeneios) 8-cell-stage embryo homogenates using proteomic iTRAQ analysis, presented by functional group
Full size table

The mixed-effects model fitted for each protein group showed a statistically significant relationship between decreasing relative protein abundance and decreasing egg quality for the heat-shock protein group (p = 0.046). No statistically significant differences were identified for any of the other groups (all p > 0.05). When FDR correction was applied to the mixed-model results, all p values became statistically nonsignificant.

The simple regression analysis performed for each of the 121 proteins separately revealed seven proteins that exhibited statistically significant associations (at p < 0.05) between egg quality and relative protein abundance (Fig. 2; for results from the other proteins, see Table 2). All of these proteins (vitellogenin, heat-shock protein-70, 78 kDa glucose-regulated protein, elongation factor-2, GTP-binding nuclear protein Ran, iduronate 2-sulfatase and 6-phosphogluconate dehydrogenase) demonstrated decreased relative abundance with decreasing egg quality. The strongest trend (slope estimate −0.239 ± 0.086 SE, R 2adj  = 0.492, Fig. 2) was observed for the enzyme iduronate 2-sulfatase (Ids) for which the abundance ratios ranged between 1.48 and 0.66 from the highest (G1.50, G1.70, G1.77) to lowest quality egg batches (X2.93, X3.31, X3.50), respectively (Table 2). When FDR correction was applied to the regression results, the trends observed in the seven proteins listed above were no longer statistically significant; thus, using FDR, the likelihood of obtaining individual tests with the observed probability values when conducting 121 such tests was p ≥ 0.59.

Fig. 2
figure 2

Relationships between blastomere symmetry score and relative abundances of a set of proteins from hapuku, P. oxygeneios. Relative abundances of proteins in eight batches of eggs were determined using iTRAQ, and abundances were correlated with the egg blastomere symmetry scores (‘egg quality’); only significant relationships (prior to FDR correction) are shown

Full size image

On subjecting proteins and their abundances to cluster analysis, several of the heat-shock proteins grouped together. However, the heat map did not clearly associate these proteins, or any other cluster, with egg quality (Fig. 3).

Fig. 3
figure 3

Heat map, representing the hierarchical clustering of peptide abundance data (n = 121) in early embryos (8-cell stage) of different quality (‘0’ is best, ‘4’ is worst) in hapuku (P. oxygeneios). The abundance of each peptide within a sample was normalized over that of the sample isobarically labelled with a tag of 113 Da (G1.50; peptides in this sample all received a relative abundance score of 2log(1) = 0, represented by the black colour). Relative peptide abundances for the remaining 7 samples are indicated on the colour scale above the heat map, such that a score of −1 (green) or +1 (red) represents a half-fold or twofold increase in relative peptide abundance, respectively. For details on egg quality scores, see text. (Color figure online)

Full size image

Discussion

The present study investigated the proteome of early-stage fertilized eggs of hapuku and the relationship between egg quality and the relative abundance of identified proteins. The functions and/or biochemical pathways corresponding to the proteins identified in this study depict common cellular processes that were reported in earlier studies that investigated protein contents in fish eggs and embryos (e.g. Papakostas et al. 2010; Castets et al. 2012). Unlike the above-mentioned studies, however, which used gel-based methods, the present study utilized the gel-free iTRAQ method. The use of iTRAQ allowed for simultaneous analysis of eight different samples, one of which was used as the reference sample.

We found that the relative abundance of heat-shock proteins (as a group) decreased with decreasing egg quality. Heat-shock proteins (HSP) in general, and Hsp70 specifically, are highly expressed, not only in response to stress, but also during normal development of fish, for example, in zebrafish (reviewed in Basu et al. 2002). Heat-shock proteins were also abundant in normal oocytes of gilthead seabream (Sparus aurata, Ziv et al. 2008). Castets et al. (2012) found several heat-shock proteins (including Hsp70) to have reduced relative abundances in poor quality eggs, whereas Chapman et al. (2014) identified hsp90 transcripts among a subset of 233 mRNA species whose abundance had predictive value for egg quality in striped bass. A comparable general trend was found in the present study, in which the heat-shock protein group, split between two clusters in the heat map, exhibited lower relative abundance in lower quality eggs. Furthermore, two of the seven proteins that exhibited a trend in relative abundance associated with egg quality belong to the HSP70 family [Hsp70 and 78 kDa glucose-regulated protein (Grp78)]. Grp78 functions as a chaperone involved in translocation and protein folding (Haas 1994) and has been previously identified in fish liver proteome (Malécot et al. 2009). Taken together, these results reinforce the existence of a link between the abundance of heat-shock proteins and egg quality in hapuku and, quite possibly, teleost fish in general.

Significantly, a negative trend (decrease in relative abundance with decreasing quality) was found in the present study for the enzyme iduronate 2-sulfatase (Ids). Activity of this enzyme during the early stages (64-cell stage) of zebrafish development was previously detected and peaked at the gastrula stage (Moro et al. 2010), indicating that Ids may be especially important during the early stages of embryo development; indeed, knock-down of the gene encoding this enzyme was strongly associated with severe head and trunk deformities in developing zebrafish larvae (Moro et al. 2010). These findings are of particular interest in the light of the results from the present study, since poor quality hapuku eggs were characterized by severe blastomere asymmetry (Kohn and Symonds 2012), which correlated with low abundances of Ids. As ids transcripts have been found in unfertilized zebrafish eggs (Moro et al. 2010), this mRNA species is likely to be of maternal origin and, as such, contributes to the ‘legacy’ of a female to its offspring (Gandolfi and Gandolfi 2001); accordingly, the production of eggs low in ids transcript or Ids protein abundance may well reflect a ‘poor legacy’ by a female, and early identification and purging of such females from the broodstock tank could improve hatchery production efficiency.

Notwithstanding the identification of a single Vtg peptide that showed a significant positive correlation between quality and relative Vtg abundance, the overall relative abundance of identified Vtgs showed no significant relationship with egg quality.

The yolk proteins that are derived from the cleavage of Vtgs were shown to comprise more than 90 % of total egg protein in zebrafish (D. rerio) embryos (Link et al. 2006). Because of their abundance, yolk proteins can mask the presence of less abundant proteins that may be important to embryo development (Link et al. 2006). Attempts to reduce the amount of Vtg in the samples, using ammonium sulphate precipitation and immunoprecipitation in the present study (results not shown), did not yield the desired results. Thus, similar to previous studies which investigated egg quality in fish using proteomics (Yoshikuni et al. 2003; Rime et al. 2004, Crespel et al. 2008, Castets et al. 2012), the analysis in the present study was carried out with yolk proteins included in the tested samples.

Ziv et al. (2008) found several proteins related to translation control and protein synthesis in maturing oocytes of zebrafish and gilthead seabream. In the present study, the relative abundance of elongation factor-2, a ubiquitous protein that is critical for protein synthesis (Robinson et al. 1974), was lower with decreasing egg quality. A similar trend was seen for the GTP-binding nuclear protein Ran, which is involved in protein import into the nucleus and is known to affect DNA replication, cell cycle and preservation of nuclear structure (Moore and Blobel 1993).

The citric acid cycle, glycolysis and pentose phosphate pathways are fundamental to biological systems and function in essential carbon metabolic processes and metabolism. Aconitase hydratase, an enzyme involved in the citric acid cycle, was found to be less abundant in lower quality eggs in Eurasian perch (Castets et al. 2012). In the present study, the relative abundance of 6-phosphogluconate dehydrogenase, which is an enzyme in the pentose phosphate pathway, also exhibited a negative trend, decreasing in relative abundance from the high- to low-quality groups. Similarly, transcript abundances of a suite of genes associated with metabolism-related pathways in eggs from wild Atlantic cod brood stock (‘high quality’) differed from those in farmed brood stock (Lanes et al. 2013), reinforcing the perhaps obvious notion that metabolic ‘state’ of the egg is likely to be important for its quality.

The likelihood of false positives (type I error) when examining a large data set is high, particularly when the differences between samples are small or associations are weak. Indeed, when FDR correction was applied to our results, these uncovered associations were no longer statistically significant. This may have been influenced by two factors. Firstly, the differences between the egg quality groups may not have been extreme enough for this type of analysis. We note that the highest quality batch score in the present study corresponded to a ‘moderate’ quality batch as described in Kohn and Symonds (2012). Therefore, the lack of strictly ‘good’ quality batches has made it more difficult to detect differences between batches with differing quality scores, which is also evident from the clustering analysis shown in the heat map. Secondly, recent reviews that have critically examined the efficiency of the iTRAQ method in quantifying proteins (Christoforou and Lilley 2012; Evans et al. 2012) have argued that iTRAQ analysis may underestimate the fold change in protein abundance. It has been shown that ratios of reporter ions, especially of low-abundance peptides, are biased towards ‘1’ based on the co-isolation and co-fragmentation of peptides from unregulated high-abundance proteins which have very similar m/z values to the targeted low-abundance peptides (Ow et al. 2009, Karp et al. 2010, and Mahoney et al. 2011). This results in mixed reporter ion ratios within the fragment ion spectrum. Thus, iTRAQ might have a limited sensitivity for the quantification of weak fold changes in low-abundance proteins. Furthermore, the contamination of specific peptide fragment spectra depends on the sample complexity and the sample matrix with which the peptides co-elute. Consequently, not every peptide fragment ion spectrum of the same protein shows contaminated reporter ion ratios which results in high coefficients of variation in low-abundance proteins. The aim of the here-presented iTRAQ study was to perform a global screen for potential target proteins or pathways that are associated with egg quality. Such targets, however, require further confirmation using orthogonal and more accurate methods which in general have a limited throughput and are not suitable for global quantitative profiling. Regardless, proteins with similar functions to those identified here have been implicated in other studies (which, incidentally, have not implemented FDR correction) to be important for embryo development. Therefore, the trends observed here can help elucidate the importance of these proteins with regard to the normal development of early-stage embryos.

Whilst our analyses identified a number of candidate proteins that correlated with egg quality, it is important to note that individual candidates are likely to have limited predictive value of egg quality, as illustrated by Chapman et al. (2014). These authors concluded from their analysis of transcriptome data on striped bass (Morone saxatilis) that individual candidate transcripts may only explain ~1 % of the variation in egg quality. They further contended that egg quality clearly is too complex a trait to be linked to a single candidate, but that as (relatively) few as 250 genes could explain as much as 90 % of the variation in egg quality. Therefore, strong associations between egg quality and candidate gene expression (mRNA, protein) may well prove useful as low-cost and quick turnover biomarkers in future.

In conclusion, our study reported on the proteome of hapuku embryos at the early 8-cell stage of development. Generically, heat-shock proteins decreased in abundance with egg quality. Regression analysis further identified seven proteins that exhibited a decreasing trend in relative abundance in relation to egg quality—of particular interest is the association between the decreasing relative abundance of Ids that was previously linked to impaired fish embryo development and decreasing egg quality. We contend that Ids and heat-shock proteins are obvious candidate biomarkers for egg quality in hapuku.