Extensive destruction of transcripts occurs in oocytes during meiotic maturation yet there is only limited information on which transcripts are degraded or relatively stable (Bachvarova et al., 1985
; Bettegowda et al., 2006
; Lequarre et al., 2004
; Lonergan et al., 2003
; Paynton et al., 1988
; Stutz et al., 1998
; Zheng et al., 2005a
; Zheng et al., 2005b
). A global perspective of the relative changes in oocyte transcript profiles during the process of maturation has been lacking. Here, using microarray and bioinformatics approaches for analysis of gene ontology and pathways, we assessed changes in transcript profiles in GV and MII mouse oocytes, and investigated the biological themes associated with the corresponding global changes. A modified RNA linear amplification procedure in which the Full Spectrum™ MultiStart Primers for in vitro transcription, a primer set uniquely designed to initiate cDNA synthesis at multiple points along the mRNA with reduced bias of the length of 3’-poly(A) tail, replaced the commonly used T7-Oligo(dT) primers whose priming efficiency is affected by the 3’-poly(A) tail. This modified amplification protocol was used to reduce potential profiling errors derived from the active mRNA polyadenylation/deadenylation that occurs in oocytes during maturation.
Global changes in oocyte transcripts during meiotic maturation in vivo
The Affymetrix Mouse Expression 430v2.0 GeneChip contains 45,037 probe sets that represent 14,484 full-length genes, 9,450 non-ESTs and 21,103 ESTs, according to the manufacturer. In the present study, 21,646 probe sets were detected as present in either GV or MII oocytes. The mean relative changes of these transcripts during meiotic maturation from GV to MII are shown in . Of all the present transcripts, 3,002 probe sets were identified to be significantly changed using the criteria of FWER p-value < 0.05, and 9,260 probe sets were identified to be relatively stable using a separate criteria for defining stable transcripts as described in Materials and Methods. Of particularly note, 2957 probe sets were significantly downregulated, and they represent the overwhelming majority (98.50%) of the changed transcripts. Moreover, 61.12% of the total changed transcripts (1838 probe sets) were downregulated more than 4 fold in MII oocytes. However, 1.5% of the changed transcripts (54 probe sets) appeared to be up-regulated in MII oocytes, but real-time PCR analysis suggested that this increase is artifactual (see below section for further discussion of this issue).
Fig. 1 Distribution of transcripts at various magnitudes of difference in expression levels between MII and GV oocytes. Transcripts, as represented by Affymetrix probe sets, were categorized by fold-changes in expression levels between MII and GV oocytes. Only (more ...)
It is important to note that FWER, rather than FDR (false discovery rate), was utilized here to identify transcripts whose expression levels become changed significantly during GV-to-MII transition. Of the available methods for testing the significance of differential expression of probe sets, FWER best controls for the possibility of making one or more type I error -rejecting null hypothesis when the null hypothesis is true- in the entire experiment. While this method is quite stringent, it allows for selecting specific transcripts with very high statistical significance and thus reducing false positive rates. This conservative strategy for identification of significantly changed transcripts is essential in the present study. This is because of the difference in mRNA abundance between a GV- and a MII-oocyte, and the problem is that the normalization method employed here cannot eliminate all the artifacts inherent in the equal loading of the microarrays (see below for more discussion of this issue).
Although FWER is appropriate for identification of drastically changed transcripts with low false positive rate in the significantly changed category, its high stringency will probably leave some differentially expressed transcripts in the not significantly changed category (Cui and Churchill, 2003
). Therefore, it is inappropriate to simply define the complement of the significantly changed transcripts as the stable ones. A separate strategy was therefore required to identify the stable transcripts while reducing the incidence of false positives. The criteria used in the present study for the stable transcripts met this need, and ensured the identification of the 9,260 stable transcripts with both high accuracy (their log2
ratios close to zero) and high precision (low technical and biological variation) (Woo et al., 2004
). Thus, of the 21,646 probe sets that were detected in oocytes, 3,002 were determined to be significantly changed and 9,260 unchanged using the two methods of statistical evaluation. This means that 7,384 did not fall into either category with certainty and were not used in the pathway analyses.
Validation of the microarray data
Thirty-six transcripts were selected from the array expression profile for further analysis by real-time RT-PCR. These 36 transcripts were categorized into 5 groups consisting of: (A) transcripts normally expressed in the expanded cumulus oophorus but not in oocytes, i.e., Has2, Ptgs2, Ptx3 and Tnfaip6; (B) transcripts polyadenylated during oocyte maturation, i.e., Ccndbp1, G6pdx, Hprt1, Mos, Plat and Spin; (C) transcripts expressed only or highly in mouse oocytes, i.e., Bmp15, Fgf8, Gdf9, H1foo, Nalp5 and Zar1; (D) transcripts apparently up-regulated in MII oocytes as shown by the present microarray data, i.e., Clybl, Ina and Ly6e; (E) transcripts that were significantly downregulated as shown by the present array data and are linked to various specific biological functions and cellular processes, i.e., Calm2, Ccnh, Dedd, Dnajc2, Exosc8, Gja1, Lmna, Ndufv1, Paip2, Polr2b, Psmc2, Rpl19, Rps9, Rps17, Statip1, Tacc3 and Xrcc5.
As shown in , for all the transcripts subjected to validation, none were found to be up-regulated in MII oocytes by real-time PCR analysis. Of particular note, transcripts included in group D were marginally downregulated in MII oocytes as revealed by real-time PCR analysis even though they appeared to be at higher levels than in GV-stage oocytes by microarray analysis (). This result of real-time PCR analysis indicates that no transcripts were up-regulated in MII oocytes, and suggests that the detection of the upregulation of these 3 transcripts shown in , as well as the other 43 transcripts suggested to be up-regulated by the microarray data (), is an artifact. The actual levels of those up-regulated transcripts in MII oocytes were slightly below levels present in GV- oocytes. Since the amplification system used in this study minimizes the impact of polyadenylation, the cause of this artificial increase must have a different etiology. Because total mRNA in mouse oocytes decreases by at least 30% during maturation (Paynton et al., 1988
), and because equal amounts of total RNA were used as input into the amplification and equivalent amounts of resulting cRNA products were applied to the arrays, some stable transcripts present at MII were probably applied to the arrays in amounts greater than those present at the GV-stage. Technically, there is no ideal way to avoid this problem by differential loading since correction for one group of transcripts would simply distort another. Although we used a normalization method to correct most of these artifacts (see Materials and Methods), some noise inevitably persists.
Fig. 2 Real-time RT-PCR analysis of transcripts selected from microarray expression profiles. Five categories (A–E) of transcripts were selected for real-time RT-PCR analysis. (A): Transcripts that are normally expressed in expanded cumulus oophorus (more ...)
Group A transcripts are expressed in expanded cumulus cells and not oocytes (Fulop et al., 2003
; Joyce et al., 2001
; Salustri et al., 1999
; Salustri et al., 2004
; Varani et al., 2002
), and are therefore good indicators of potential cumulus cell contamination of oocyte RNA preparations. An apparent upregulation of these transcripts would be detected in MII oocytes if there were any cumulus cells inadvertently included in the oocyte samples. As shown in , none of the four cumulus cell-specific transcripts, Tnfaip6
, or Ptgs2
, was found at higher levels in MII oocytes by either microarray or real-time RT-PCR analysis. Moreover, the expression levels of all the cumulus cell-specific transcripts that were detected in oocytes by either method were extremely low as shown by their RMA expression measures or Ct values (data not shown). Thus, there was no contamination of oocyte RNA preparations with cumulus cells that could confound interpretation of the microarray results.
Transcripts in group B are normally polyadenylated during oocyte maturation (Gebauer et al., 1994
; Huarte et al., 1987
; Oh et al., 2000
; Paynton and Bachvarova, 1994
; Sheets et al., 1994
), and could appear to be up-regulated in MII oocytes when poly(A)-biased mRNA quantification approaches are used to compare their expression levels in GV and MII oocytes (Wang et al., 2004
). In contrast, no apparent up-regulation should be detected for any of these transcripts if the method used for mRNA quantification is unaffected by poly(A) tails as expected for the Full Spectrum™ MultiStart Primers for in vitro transcription. Indeed, as shown in , none of these transcripts was found at levels higher in the MII oocytes than GV-stage by either microarray or by RT-PCR analysis. This observation provides strong validation for the use of MultiStart Primers for in vitro transcription to reduce the influence of polyadenylation. However, real-time PCR but not microarray detected that all the transcripts in this group were significantly down-regulated in MII oocytes. This could be due to the differences in sensitivity or lower limit of detection of these two methods used for mRNA quantification, or to the overloading of the MII sample to the microarrays as described above. Nevertheless, it is consistent that none of these polyadenylated transcripts was up-regulated in MII oocytes. Two other transcripts reported to be present at higher levels in MII oocytes than GV-stage oocytes are Napa
(RA81) (Mann et al., 1995
) and Gnai2
) (Rambhatla et al., 1995
). The authors suggested that this apparent increase could be artifactual due to polyadenylation (Rambhatla et al., 1995
). Our microarray analysis shows no change in the levels of these transcripts at MII compared with the GV-stage supporting the use of the Full Spectrum™ MultiStart Primers for amplification to avoid potential complications of polyadenylation.
Transcripts in group C encode proteins that are oocyte-specific or highly expressed in oocytes and play crucial roles in promoting the normal development and functions of oocytes and/or follicles by functioning as paracrine factors, i.e., Gdf9
(Buratini et al., 2005
; Dong et al., 1996
; Eppig, 2001
; Matzuk et al., 2002
; Su et al., 2004
; Valve et al., 1997
; Yan et al., 2001
), promoting oocyte-to-embryo transition as well as preimplantation development, i.e., Zar1
(also called Mater
) (Tong et al., 2000
; Wu et al., 2003b
), or potentially regulating chromatin remodeling during oocyte/and or embryo development, i.e., H1foo
(Gao et al., 2004
; Tanaka et al., 2005
; Teranishi et al., 2004
). Little if anything, however, is known about the changes in the relative levels of these transcripts during mouse oocyte maturation. As shown in , microarray analysis detected the significant down-regulation of Fgf8
in MII oocytes and this down-regulation was validated by real-time PCR. Although the microarray study indicated that the rest of the 5 transcripts in group C were not changed in MII oocytes, real-time RT-PCR analysis revealed that they were actually all significantly down-regulated. This discrepancy could be due to the same causes discussed in the preceding section. Nevertheless, real-time PCR analysis further confirmed no up-regulation of transcripts in MII oocytes.
Transcripts included in group E are representative of those involved in specific biological function/pathway and/or cellular process, such as protein synthesis and metabolism (e.g., Rps17, Rps9, Psmc2) and oxidative phosphorylation (e.g., Ndufv1). For all the transcripts in group E, a similar pattern of change was consistently revealed by both microarray and real-time PCR (). For example, Rps17 was down-regulated as demonstrated by both microarray and real-time RT-PCR, 31.6 and 45.4 fold respectively.
In sum, real-time RT-PCR validated (1) the specificity of our array data allowing the exclusion of potential cumulus cell contamination from confounding results; (2) the use of the modified mRNA linear amplification procedure in preventing profiling errors caused by polyadenylation; (3) no up-regulation of any transcripts in MII oocytes; and (4) the overall trends of message stability or loss indicated by microarray analysis.
Biological pathways and functions associated with the transcripts lost during oocyte maturation
To determine whether there are biological themes and specificity underlying the degraded oocyte transcripts during meiotic maturation, we carried out a series of pathway analyses using both IPA and GenMAPP/MAPFinder software. IPA is a web-based software application that enables the modeling and analysis of biological systems using microarray data, while GenMAPP/MAPPFinder is a stand-alone computer program designed for viewing and analyzing gene expression data in the context of gene ontology and biological pathways. Both programs are commonly used and demonstrated to be powerful tools for identifying the biological themes of gene expression data (Calvano et al., 2005
; Dahlquist et al., 2002
; Doniger et al., 2003
Canonical pathways analysis by IPA identified the pathways that were significantly associated with the degraded transcripts. Degraded transcripts as defined by FWER p
< 0.05 were used for the analysis. Fisher’s exact test was used to calculate a p
-value (< 0.05) determining the probability that the association between the transcripts in the dataset and the canonical pathway could be explained by chance alone. As shown in , of the total of 103 metabolic and signaling pathways in the IPA canonical pathway library, 17 are significantly associated with the transcripts that were lost in MII oocytes. GenMAPP/MAPPFinder also identified most of these pathways (see Tables S1
). Within these 17 canonical pathways, 13 are involved in metabolic processes. Interestingly, most of these processes are closely related. For example, 59 transcripts associated with oxidative phosphorylation were downregulated in MII oocytes. The same process, termed “electron transport chain” in GenMAPP, was identified to be among the most significantly affected MAPPs as well (see Table S2
). As illustrated in , oxidative phosphorylation is the major pathway for conversion of the energy of NADH oxidation into phosphate-bond energy in ATP. This process takes place in mitochondria and is catalyzed by five large respiratory enzyme complexes residing in the mitochondrial inner membrane. Most of the degraded transcripts that were involved in this pathway encode these enzyme complexes, indicating the potential for a reduced rate of ATP production in mature oocytes if the encoded proteins are rapidly turned over. The second pathway most impacted by the loss of transcripts was ubiquinone biosynthesis where 25 transcripts encoding the enzymes that convert ubiquinol into ubiquinone were involved. Interestingly, ubiquinone is an essential electron carrier in the oxidative phosphorylation process, thus the two canonical pathways having the highest significance rating by IPA are closely related. Other canonical pathways, such as pyruvate metabolism and citrate cycle, are also closely connected to the oxidative phosphorylation pathway. For example, oocytes efficiently use pyruvate but not glucose to produce energy essential for oocyte maturation (Biggers et al., 1967
; Downs et al., 2002
). The energy produced by pyruvate metabolism and citrate cycle is partially in a form of high-energy electrons that are transiently held by NADH. NADH is then used in the oxidative phosphorylation pathway in the mitochondrial inner membrane and the high-energy electrons are eventually used for ATP synthesis. Therefore, these two pathways are also closely related to the oxidative phosphorylation process, and the loss of transcripts involved in these two pathways could result in the lower rate of ATP synthesis. Because a high rate of oxidative phosphorylation and ATP production is always associated with cellular processes that utilizes energy, such as macromolecular synthesis (e.g., protein synthesis), protein phosphorylation, ion-transporting, etc., the loss of transcripts responsible for these processes in MII oocytes could reflect the relatively quiescent status of MII oocytes in consuming energy. Indeed, it has been found that oocyte maturation was associated with increased pyruvate consumption, hence energy utilization; and MII oocytes consumed less pyruvate than oocytes undergoing meiotic maturation (Downs et al., 2002
). Therefore, it is possible that loss of relevant transcripts participates in decreased energy production and utilization apparent in MII oocytes if the amount of the encoded proteins is also reduced.
Fig. 3 Canonical pathways associated with the degraded transcripts during oocyte maturation. (A): All the pathways identified by IPA that are associated with the degraded transcripts. A larger value on the y-axis indicates a higher degree of significance, i.e., (more ...)
The degradation of transcripts involved in energy production process may also be associated with the relief of ATP consumption in maintenance of meiotic arrest at the GV-stage. A constant high level of cAMP maintains meiotic arrest in fully-grown oocytes, and this high level of cAMP is produced from ATP by an oocyte-specific G-protein coupled adenyl cyclase (Eppig et al., 2004
; Horner et al., 2003
; Mehlmann et al., 2002
; Mehlmann et al., 2004
). Therefore, the requirement for ATP in maintenance of meiotic arrest may explain why transcripts involved in oxidative phosphorylation were highly expressed in GV-oocyte as shown here and elsewhere (Zeng et al., 2004
). Taken together, dynamic changes in the rates of oxidative phosphorylation, energy production, and consumption are tightly correlated with the meiotic status of the oocyte, and the completion of the first meiotic division is correlated with the loss of transcripts involved in these energy producing and using processes.
To further understand the biological and molecular functions represented by transcripts degraded during the GV to MII transition, an IPA Functional Analysis was carried out. This analysis identified the biological functions that were significantly associated with the set of lost transcripts. Degraded transcripts, as defined by FWER p
< 0.05, and were associated with the biological functions in IPA Knowledge Base, were considered for the analysis. There were 26 molecular functions identified by this analysis to be significantly (P
< 0.05) associated with the lost transcripts. The top 15 functions are shown in . Interestingly, most of the functions identified by IPA analysis were also independently identified by GenMAPP/MAPPFinder (Tables S1
), and were well in agreement with those aforementioned Canonical Pathways. For example, the degraded transcripts involved in oxidative phosphorylation and ubiquinone biosynthesis as identified by the IPA Canonical Pathways Analysis suggests a low rate of macromolecule biogenesis in MII oocytes, and this is consistent with the identification of protein synthesis by the IPA Functional Analysis as shown here. In addition, the identification of pyruvate metabolism and citrate cycle by IPA Canonical Pathways Analysis indicates low rate of energy production in MII oocytes, and this is in agreement with the identification of energy production as a biological function in the IPA Functional Analysis. Most interestingly, similar processes were identified simultaneously by both the IPA Canonical Pathways Analysis and Functional Analysis. For example, the nucleotide excision repair pathway was identified by the IPA Canonical Pathway analysis, and the similar process termed DNA Replication, Recombination, and Repair was identified in the IPA Functional Analysis.
Fig. 4 Biological functions associated with the degraded transcripts during oocyte maturation. (A): Biological functions identified by IPA that were associated with the degraded transcripts. The number of degraded transcripts involved in each function is indicated (more ...)
IPA analysis identified protein synthesis as the function that is most significantly associated (P = 1.1 x 10−12
) with the degraded transcripts. There were 109 degraded transcripts involved in this function and they were associated with 7 sub-categories of functions (Table S3
). Of the degraded transcripts participating in protein synthesis, a large number of them encode cytoplasmic and mitochondrial ribosomal proteins. This function was also identified by GenMAPP (Table S1
) and by MAPPFinder as the most significantly affected (Table S2
). Of the 78 transcripts that encode cytoplasmic ribosomal proteins in the large and small ribosomal subunits, 68 of them were significantly degraded in MII oocytes (). Coordinate with the loss of mRNAs encoding the ribosomal proteins, there is a decline of about 60 pg/oocyte in rRNA (Paynton et al., 1988
). Presumably, the degradation of rRNA and transcripts encoding ribosomal proteins results in loss of ribosomes and reduction in the potential for message translation. In fact, the absolute rate of protein synthesis decreases almost 30% during oocyte maturation (Schultz and Wassarman, 1977
). Coordinated with this, transcripts encoding translation initiation factors are lost; 15 transcripts encoding translation initiation and translation elongation factors were significantly degraded during oocyte maturation (Table S3
). While the capacity for protein synthesis becomes reduced during oocyte maturation, so does protein degradation since many transcripts encoding the components of the ubiquitin-proteasome degradation system (12 transcripts in total) were lost.
There are high levels of expression of transcripts involved in the regulation of the meiotic cell cycle and DNA repair during oocyte growth and development (Pan et al., 2005
). It is shown here that many of the transcripts associated with these functions are degraded during oocytes maturation (. and Tables S1
Taken together, results presented here show that transcripts participating in specific biological pathways and functions essential for supporting oocyte meiotic arrest, resumption, and progression are highly represented among transcripts degraded during oocyte maturation. Of particular note are transcripts associated with protein synthesis and energy production and utilization. Interestingly, transcripts associated with protein synthesis are dramatically up-regulated during zygotic gene activation at the two-cell stage (Zeng and Schultz, 2005
). The explanation for this apparent inefficiency is as lost as the transcripts.
Biological functions and pathways associated with transcripts that are relatively stable during oocyte maturation
Despite the substantial loss of transcripts during oocyte maturation, a great number of transcripts are relatively stable. Using the statistical strategy described in the section of material and methods, 9,260 transcripts were categorized as stable. To understand the biological functions and processes represented by these retained transcripts in MII oocytes, this set of transcripts was subjected to IPA and GenMAPP/MAPPFinder analyses. IPA analysis revealed that 30 canonical pathways were significantly associated with the relatively stable transcripts in MII oocytes (the top 15 is shown in ). Twenty-five of these (83.33%) were associated with signal transduction. This was consistent with GenMAPP/MAPPFinder analysis where cell communication, signal transduction, and signal transducer activity were the most significant MAPPs and gene ontology (GO) terms associated with stable transcripts (Table S4
). It is of interest to note that transcripts associated with the same GO categories were reported to be gradually degraded in mouse embryos starting from the 1-cell stage and continuing after zygotic gene activation at 2-cell stage (Zeng et al., 2004
Fig. 5 Canonical pathways associated with the stable transcripts during oocyte maturation. (A): Pathways identified by IPA that are associated with the stable transcripts. The number of stable transcripts involved in each pathway is shown above the bars. There (more ...)
IPA analysis revealed that transcripts associated with the canonical pathways of ERK/MAPK and PI3/AKT signaling were highly represented in the stable group. These two signaling pathways are strongly implicated in the regulation of oocyte meiosis (Choi et al., 1996
; Colledge et al., 1994
; Hashimoto et al., 1994
; Kalous et al., 2006
; Phillips et al., 2002
; Su et al., 2002b
; Tomek and Smiljakovic, 2005
; Verlhac et al., 1996
; Vigneron et al., 2004
). The ERK/MAPK signaling pathway is crucial for the maintenance of MII arrest in oocytes (Colledge et al., 1994
; Hashimoto et al., 1994
; Phillips et al., 2002
; Su et al., 2002a
). Of the 121 transcripts involved in the ERK/MAPK signaling pathway, 68 (56.19%) was identified as stable. The display of transcripts involved in MAPK signaling pathway by GenMAPP was shown in . Therefore, transcripts involved in signaling pathways essential for maintaining meiotic arrest at MII are stable during the GV to MII transition.
To compare transcript profiles between GV- and MII-oocyte, one must consider the influence of mRNA polyadenylation and non-equivalency problems prevalent during GV-to-MII transition. The modified RNA linear amplification system and the data normalization method used in the present microarray study appear to be an effective way in preventing and correcting these problems. In fact, use of this system may be advantageous when conducting any transcriptome comparisons where polyadenylation issues could hamper interpretation, a situation that could be quite common.
Data presented here indicate that destruction of transcripts during the GV to MII transition in oocytes is a selective rather than promiscuous process. Transcripts associated with meiotic arrest at the GV-stage and the progression of oocyte maturation such as oxidative phosphorylation, energy production, and protein synthesis and metabolism were dramatically degraded. In contrast, transcripts encoding participants in signaling pathways essential for maintaining the unique characteristics of the MII-arrested oocyte, such as those involving protein kinase pathways, were most prominent among those retained. Aberrant degradation or maintenance of certain class of transcripts during oocyte maturation could be deleterious to oocyte quality, impacting developmental competence. Although transcriptional profiling of GV-stage mouse oocytes with decreased developmental competence did not reveal dramatic differences from oocytes with high developmental competence (Pan et al., 2005
), differences may become more profound at MII.