Stages of ovarian development
To identify differentially expressed genes during ovarian development in Senegalese sole, samples of ovarian tissue were collected from adult females sacrificed throughout the annual reproductive cycle [28
] or after hormonal treatment (Figure ). Thus, samples of ovaries at previtellogenesis (Figure and ), vitellogenesis (Figure and ), maturation (Figure and ), and undergoing follicular atresia (Figure and ), were used for transcriptome analysis. As many other fractional spawner teleosts, the Senegalese sole has a group-synchronous ovary in which follicles of all sizes up through vitellogenesis are present at any time, and populations (or clutches) of follicles are periodically recruited into maturation from a population of oocytes in late vitellogenic stages [30
]. Therefore, the increased frequency of vitellogenic, mature or atretic ovarian follicles in the ovary, as determined by histological analysis (Figure and ), defined the ovarian developmental stages used in the present study.
Figure 1 Developmental stage of the Senegalese sole ovaries used for microarray analysis. Representative light micrographs of histological sections of the ovary (n = 3 females) stained with hematoxylin-eosin (A, B, D, E, G, H, and J-L), and frequency of ovarian (more ...)
The previtellogenic ovary was formed by exclusively ovarian follicles with oocytes at the primary growth stage (oocyte/follicle diameter up to approximately 150 μm) in which vitellogenin incorporation and yolk formation did not yet start (Figure ). In the vitellogenic ovary, a population of follicles were recruited into vitellogenesis, and consequently the proportion of follicles at the primary growth stage decreased (Figure ). At this stage, follicles containing oocytes at the cortical alveolus stage (up to approximately 300 μm), characterized by the presence of nascent cortical alveoli within the ooplasm, were more abundant (Figure , inset). Vitellogenic oocytes surrounded by the zona radiata and the somatic follicular cells, granulosa and theca cells, increased in size (up to 500 μm in diameter at late vitellogenesis) and their cytoplasm was filled with yolk granules where vitellogenin-derived yolk proteins are stored (Figure ). As a result of this growing phase, the gonadosomatic index (GSI) of females increased by approximately 7-fold (Figure and ).
Maturing ovaries containing follicle-enclosed oocytes undergoing meiosis resumption, and ovaries carrying mature oocytes prior to ovulation, were collected 24-48 h after treatment of vitellogenic females with gonadotropin-releasing hormone agonist [D-Ala6
, NEt] (GnRHa) [24
]. In the mature ovary, a population of follicle-enclosed oocytes at late stages of vitellogenesis was further recruited into maturation (Figure ). In these oocytes, the germinal vesicle migrates towards the animal pole and yolk globules fuse one another (Figure ), eventually forming a large mass of yolk (Figure ). The mature oocyte reached 800-900 μm in diameter due to water uptake (hydration), resulting in a further 2-fold increase of the GSI (Figure and ).
Finally, atretic ovaries were collected from females showing spontaneously occurring ovarian follicle atresia during the spawning season, or induced after GnRHa treatment. In these ovaries, approximately up to 30-40% of the ovarian follicles showed different levels of atresia and maturing/mature oocytes were absent (Figure ). In early atretic follicles, vitellogenic oocytes shrank, the zona radiata folded, and follicles became irregularly shaped (Figure and ). The follicular cells were hypertrophied and the theca was poorly developed. Advanced follicular atresia was characterized by breakdown and resorption of the zona radiata, and the appearance of highly columnar follicular cells apparently showing an intense phagocytic activity as suggested by the presence of large vacuoles (Figure ). At this stage, accumulation of blood cells, erythrocytes and leukocytes in the follicle, as well as in the oocyte, was also noted (Figure ).
Figure 2 Photomicrographs of ovarian follicles at advanced atresia. Light micrographs of histological sections stained with hematoxylin-eosin. bc, blood cells; ep, epithelium; tc, theca cells; v, vacuole; gc, granulosa cells; e, erythrocytes; le, leukocytes. Bars, (more ...)
Differential gene expression in the four ovarian developmental stages was determined using a Senegalese sole-specific oligonucleotide microarray containing 60-mer probes representing 5,087 unique genes [27
]. This platform was previously designed from a Senegalese sole EST database derived from a multi-tissue normalized cDNA library from different adult tissues (including ovaries at different developmental stage) and larval and juvenile stages [27
]. Therefore, although this platform was not ovary-specific and most likely did not contain all the transcripts expressed in the sole ovary, its was useful to obtain a first insight into the overall changes of gene expression during ovarian development.
To determine the false discovery rate (FDR) in each of the differential gene expression experiments (vitellogenic vs
. previtellogenic ovaries, mature vs
. vitellogenic ovaries, and atretic vs
. vitellogenic/mature ovaries), an additional microarray experiment was performed by hybridizing differentially labelled (Cy3 and Cy5) aliquots of amplified RNA (aRNA) from the same sample (previtellogenic ovary). As expected, there were few differences between the Cy3 and Cy5 signals for most of the microarray spots in these experiments giving an estimated overall FDR of 3.0, 3.8 and 5.6% for vitellogenic vs
. previtellogenic ovaries, mature vs
. vitellogenic ovaries, and atretic vs
. vitellogenic/mature ovaries, respectively (see Additional file 1
Microarray data analysis indicated significant (p < 0.01) regulation of genes in vitellogenic (46 ESTs), mature (46 ESTs) and atretic ovaries (26 ESTs), which showed fold change (FC) values from 1.4 up to 5.1. These ESTs, and the corresponding GenBank accession numbers are listed in Tables , and . In Table , differential expressed genes in atretic ovaries relative to vitellogenic or mature ovaries are pooled together. Some of these ESTs (26%) could not be annotated, even after sequencing the respective clones from the 5' end, and are not included in these tables.
Transcripts regulated in vitellogenic ovary relative to previtellogenic ovary
Transcripts regulated in mature ovary relative to vitellogenic ovary
Transcripts regulated in atretic ovary relative to vitellogenic/mature ovary
Gene ontology annotation
To obtain a first assessment of the more important physiological processes occurring during ovarian development, gene ontology (GO) analysis was carried out using the BLAST2GO v1 program [31
]. Most of the annotated ESTs (93%) had GO assignments, and many of those had 3-6 assignments each (49%) and a significant proportion (34%) had 7 or more assignments.
Figure shows the differentially expressed genes in the three ovarian stages (vitellogenesis, maturation and follicular atresia) classified according to GO terms biological process (level 3), cellular component (level 5) and molecular function (level 3). During vitellogenesis, the majority of regulated ESTs were dedicated to metabolic process, oxidation reduction, regulation and anatomical structure development, in the biological process category. A similar distribution of GO terms was seen within the EST cluster regulated during maturation, although in this case transcripts related to cell cycle, localization of cell, cellular component organization, and system process, were also detected. During ovarian follicle atresia, most regulated genes fall in the cellular metabolism, establishment of localization, and cellular component organization attributes.
Figure 3 Gene ontology (GO) analysis of differentially expressed genes in the Senegalese sole ovary. Genes regulated during vitellogenesis (V), maturation (M) and atresia (A) were classified according to GO terms biological process (level 3), cellular component (more ...)
During vitellogenesis and maturation, most protein products were mainly inferred to be associated with mitochondria based on the cellular component category, although some also might be in the cytoskeleton (specially during vitellogenesis), the nucleus, and intracellularly in organelles. Interestingly, during atresia, the products of most of the up-regulated genes showed putative extracellular location, whereas the products of the down-regulated genes had membrane and nucleus locations.
Finally, classification using the molecular function category indicated that most of the gene products regulated during vitellogenesis and maturation were dedicated to binding and catalytic functions, including nucleotide binding, protein binding, ion binding, and transferase and hydrolase activities. However, products involved in transmembrane transporter activity only appeared during maturation. In the atretic ovary, the majority of products were associated with ion and lipid binding.
In the vitellogenic ovary, 34 and 12 transcripts were found to be up- and down-regulated, respectively, relative to previtellogenic ovaries; 35 had a significant hit in Swiss-Prot database (Table ). The most highly up-regulated transcripts corresponded to selenoprotein W2a (sepw2a), hypothetical 18K protein from Carassius auratus mitochondrion, muscle-specific beta 1 integrin binding protein 2 (mibp2), zona pellucida protein 3 (zp3), cytochrome c oxidase subunit I (cox1), cytochrome b (cytb), cytosolic heat shock protein 90 beta (hsp90b), NADH dehydrogenase subunit 3 (nd3) and 1 (nd1), and beta actin 1 (bactin1). The sequence similarity of clone pgsP0015N21 to tilapia (Oreochromis mossambicus) sepw2a was low (4E-04) possibly because its nucleotide sequence only covered the C terminus of tilapia sepw2a.
Other up-regulated genes in the vitellogenic ovary, but at lower levels, were additional components of the cytoskeleton, such as alpha actin (actc1l), keratin 8 (krt8), tropomyosin1-1 (tpm1-1), myosin (myh11), and transgelin (tagln), or of intracellular signaling pathways, such as Ras homolog member G (rhog), a novel protein similar to serum/glucocorticoid regulated kinase (sgk) also found in zebrafish, and inositol monophosphatase 3 (impa3). Proteolytic complexes and enzymes, such as protein arginine methyltransferase (prmt1), acetyl-coenzyme A acyltransferase 2 (acaa2), and creatine kinase (ckb), and the putative homeodomain transcription factor 1 (phtf1), transducer of ERBB2 (tob1a), neurexin 1a (nrxn1a) and thrombospondin 4b (thbs4b) were also up-regulated in vitellogenic ovaries.
Two transcripts similar to CDC-like kinase 2 (si:ch211-81a5.7) and CD53 cell surface glycoprotein (zgc:64051), as well as alanine-glyoxylate aminotransferase (agxt), were the most down-regulated genes during vitellogenesis. Other down-regulated genes included the actin-binding protein scinderin (scin), mitochondrial enzymes, such as NADH dehydrogenase (ubiquinone) 1 alpha subcomplex subunit 11 (ndufa11) and a dihydrodipicolinate synthase-like enzyme, proteolytic complexes and enzymes, as proteasome subunit beta type-9 precursor (psmb9a) and carboxypeptidase H (cph), a hypothetical protein-encoding gene also found in Xenopus laevis (LOC735233), and a coiled-coil domain containing 90B novel product (ccdc90b).
Microarray analysis detected 26 up-regulated and 20 down-regulated transcripts in maturing/mature ovaries relative to vitellogenic ovaries, and 32 transcripts could be annotated (Table ). The most highly up-regulated transcript corresponded to an EST showing sequence similarity to the amphioxus (Branchiostoma floridae) BRAFLDRAFT_128798 gene, which encodes an hypothetical protein with inferred cysteinyl-tRNA aminoacylation activity. However, the BLAST E-value for the similarity of Senegalese sole clone pgsP0012B12 to this protein was relatively low (2E-14), and therefore conclusive annotation will require the cloning of the sole full-length cDNA. Other highly up-regulated transcripts encoded Na+/K+-ATPase subunits, such as the beta subunit 1a (atp1b1a), alpha subunit 1 (atp1a1) and another isoform of the beta subunit (atpb), and alpha-2-macroglobulin (a2m).
Cytoskeletal proteins, myosin 10 (myo10) and type II keratin E3-like protein, and proteins involved in transcriptional and translational responses, such as makorin RING zinc finger protein 1a (mkrn1) and ribosomal protein L36 (rpl36), were other up-regulated transcripts in mature ovaries. Interestingly, a regulator of vesicular traffic, novel protein similar to vertebrate ADP-ribosylation factor 4 and extended synaptotagmin-2-A (e-syt2-a), was also up-regulated. Other transcripts were retinoic acid receptor responder protein 3 (rarres3), thioredoxin interacting protein (txnip), mitogen-activated protein kinase p38delta (mapk13), cytochrome c oxidase subunit I (cox1), UDP-glucose dehydrogenase (ugdh), and myeloid-associated differentiation marker homolog (myadm). Some transcripts that were up-regulated during vitellogenesis showed a further increase during maturation, such as tob1a and nd1.
During ovarian maturation, more genes appeared to be down-regulated than during vitellogenesis. Among those transcripts, we found type II Na/Pi cotransport system protein (slc34a2a) and enzymes such as succinate dehydrogenase complex subunit B (sdhb) and fructose-1,6-bisphosphatase (fbp1), involved in carbohydrate metabolism, and impa3, which was up-regulated during vitellogenesis. Transcript abundance was also reduced for some components and regulators of the cytoskeleton, such as myosin binding protein H (mybph), gelsolin (gsna), and centaurin delta 2-like (LOC100005008). Other transcripts were thrombin (f2), monocytic leukemia zinc finger protein (myst3), 14-3-3 protein epsilon (ywhae), an apolipoprotein L-like protein (LOC100150119), zona pellucida C2 (zpc2) and histone H2B (hist1h2be). These mRNAs are potentially implicated in proteolysis (f2), transcription and signal transduction regulation (myst3 and ywhae, respectively), lipid transport (LOC100150119), formation of the zona radiata (zpc2), and chromatin compaction (hist1h2be).
The comparison of ovaries undergoing follicular atresia vs
. vitellogenic and mature ovaries revealed the up- and down-regulation of 10 and 16 transcripts, respectively, and 18 transcripts could be annotated (Table ). One of these transcripts (GenBank accession number FF286365
), which was the most highly up-regulated, had sequence similarity to two unknown predicted proteins from gilthead sea bream (Sparus aurata
) and the puffer fish Tetraodon nigroviridis
(BLAST E-values of 9E-08 and 3E-04, respectively). This EST apparently encoded a full-length polypeptide which shared 25% identity with a protein named gastrula-specific embryonic protein 1 found in the orange-spotted grouper (Epinephelus coioides
). The corresponding cDNA clone (pgsP0015C05) was then sequenced in full-length, and the presence of conserved motifs in its deduced amino acid sequence was investigated. These analyses, together with a preliminary phylogenetic reconstruction, clearly indicated that sole FF286365 encoded an ortholog of apolipoprotein C-I (apoc1
) (Additional file 2
). The nucleotide and amino acid sequence of this cDNA was deposited in GenBank with accession number EU835856
Other transcripts also significantly up-regulated in atretic ovaries were leukocyte cell-derived chemotaxin 2 (lect
), thrombospondin (thbs
), heme-binding protein 2 (hebp2
), apolipoprotein A-I (apoa1
), S100-like calcium binding protein (s100
), and enolase (eno3
). The S100-encoding EST (pgsP0020M08) was a full-length cDNA which allowed further analysis of its deduced amino acid sequence. The analysis indicated that this transcript belongs to the S100a10 subgroup of the EF-Hand calcium-binding proteins superfamily (Additional file 3
Regarding down-regulated transcripts, BRAFLDRAFT_128798 and a2m showed the strongest repression in atretic ovaries, which interestingly were highly up-regulated in mature ovaries. Other reduced transcripts were potentially involved in the organization of Golgi complex, such as C-terminal binding protein 1 (ctbp1) and Golgi membrane protein 1 (golm1), or the telomeric region, such as a novel protein similar to vertebrate RAP1 interacting factor homolog (rif1), as well as in transcription and translation regulation, such as elongation factor 1 alpha (ef1a) and zinc finger protein 576 (znf576). The identity of sole FF288651 and FF282343 as ef1a and znf576, respectively, was however not conclusive since the BLAST E-values were low. Cytoplasmic FMR1 interacting protein 1 homolog (cyfip1) and elongation of very long chain fatty acids protein 1 (elovl1b), which may be involved in the control of cell projections and fatty acid biosynthesis, respectively, were also down-regulated.
Validation of microarray data by real-time qPCR
A number of differentially expressed ESTs (n
= 20) in vitellogenic, mature and atretic ovaries were further selected to verify the changes in expression by real-time quantitative RT-PCR (qPCR). The expression of all twenty genes followed the same pattern whether evaluated by microarray or qPCR (Figure ). Two genes, tob1a
, were however an exception. For tob1a
, a significant increase during maturation observed with the microarray could not be detected (p
= 0.78) by qPCR (Figure and ), whereas the significant down-regulation of LOC100090881
during atresia could not be confirmed (p
= 0.88) by qPCR (Figure ). All other genes showed in general a similar relative expression pattern by both microarray and qPCR, resulting in an overall success rate of 91% (2 inconsistencies out of 22 comparisons, since tob1a
were significantly regulated both during vitellogenesis and atresia by microarray analysis). For a2m
, however, the FC determined with the array (2.35) was about 10 times lower than that measured by qPCR (18.38), which is a known phenomenon observed in oligo-arrays when background subtraction is not performed (as in the present study) [32
]. Usually, a two-fold change is considered as the cut-off around which microarray and qRT-PCR data begin to loose correlation [33
]. Finally, few transcripts that did not show significant differences in expression levels with the microarray were also selected for qPCR. These analyses did not show significant changes in the expression level consistent with the array data (data not shown).
Figure 4 Real-time qPCR validation of differential expression in vitellogenic vs. previtellogenic ovaries (A), mature vs. vitellogenic ovaries (B), and atretic vs. vitellogenic/mature ovaries (C). Bar graphs represent relative fold change (FC) of selected transcripts (more ...)
Differential expression during follicular atresia
Some regulated transcripts in atretic ovaries relative to mature/vitellogenic ovaries, such as apoc1, apoa1, thbs, lect2, s100a10, a2m and BRAFLDRAFT_128798, were further analyzed by qPCR to investigate how broadly they might be expressed during ovarian development (Figure ). For apoc1, thbs, s100a10, a2m and BRAFLDRAFT_128798, these analyses were also carried out on manually isolated ovarian follicles at the stages of vitellogenesis, maturation and atresia. The results confirmed that apoc1, apoa1, thbs, lect2 and s1001a10 transcripts were significantly (p < 0.05) up-regulated in atretic ovaries, whereas a2m and BRAFLDRAFT_128798 transcripts were accumulated in mature ovaries and strongly down-regulated in atretic ovaries, thus demonstrating the same expression pattern as that observed with the microarray. The data also revealed that apoa1, thbs and lect2 showed relatively high relative expression levels in previtellogenic ovaries in addition to during atresia.
Figure 5 Expression profile of selected transcripts during Senegalese sole ovarian development. Histograms represent relative mean expression values ± SEM (n = 3 females) in ovaries (left panels) or in isolated ovarian follicles (right panels) of apolipoprotein (more ...)
Cellular localization of differentially expressed genes
To determine the cell type-specific expression of representative transcripts in the ovary, in situ hybridization was carried out on ovarian histological sections using specific antisense riboprobes. For these experiments, we selected transcripts that were up-regulated in vitellogenic and mature ovaries, zp3, tob1a, mapk13 and mkrn1 (Figure ), or in atretic ovaries, apoc1, s100a10, thbs and lect2 (Figure ). The zp3 hybridization signal was weakly detected in the cytoplasm of previtellogenic oocytes, whereas the signal increased in early cortical alveolus stage oocytes to subsequently diminished again at later stages (Figure and ). The staining was absent in vitellogenic oocytes as well as in the surrounding follicle cell layers. A similar localization pattern was observed for tob1a (Figure and ) and mkrn1 (Figure and ), although their hybridization signals remained visible, but much weaker, in the cytoplasm during vitellogenesis. A weak mkrn1 staining was also seen in follicular cells of vitellogenic follicles. mapk13 transcripts were exclusively localized in the surrounding follicular cells of late vitellogenic oocytes, whereas expression in ovarian follicles at other stages was not consistently detected (Figure and ). For all these transcripts, sense probes resulted in no signal (Figure and ).
Figure 6 In situ hybridization of zona protein 3 (zp3), transducer of ERBB2 (tob1a), mitogen-activated protein kinase p38delta (mpk13), and makorin RING zinc finger protein 1a (mkrn1) transcripts in the Senegalese sole ovary. Ovarian histological sections were (more ...)
Figure 7 In situ hybridization of apolipoprotein C-I (apoc1), S100A10 calcium binding protein (s100a10), thrombospondin (thbs), and leukocyte cell-derived chemotaxin 2 (lect2) transcripts in the Senegalese sole ovary. Ovarian histological sections were hybridized (more ...)
Regarding the transcripts up-regulated during ovarian atresia, apoc1-specific antisense probes showed an intense and specific staining in hypertrophied and vacuolized follicular cells of atretic follicles, which was increasing as follicular atresia progressed (Figure and ). The same staining pattern was found for s100a10 (Figure and ) and thbs (Figure and ). The lect2 transcripts were found in theca cells of atretic follicles (Figure inset) but a weaker and more diffuse staining was also detected in hypertrophied granulosa cells (Figure ). Primary growth oocytes, including cortical alveolus stage oocytes, also expressed thbs and lect2 mRNAs (Figure ), in agreement with their increased levels previously found in previtellogenic ovaries by qPCR (Figure and ). Sense probes for all of these transcripts were negative (Figure and ).