Dynamics of the Striped Bass (Morone saxatilis) Ovary Proteome Reveal a Complex Network of the Translasome

Dynamics of the Striped Bass (Morone saxatilis) Ovary ProteomeReveal a Complex Network of the TranslasomeBenjamin J. Reading,*,† Valerie N. Williams,† Robert W. Chapman,‡ Taufika Islam Williams,§

and Craig V. Sullivan†

†Department of Biology and §Mass Spectrometry Facility, Department of Chemistry, North Carolina State University, Raleigh, NorthCarolina, United States‡South Carolina Department of Natural Resources, Charleston, South Carolina, United States

*S Supporting Information

ABSTRACT: We evaluated changes in the striped bass(Morone saxatilis) ovary proteome during the annualreproductive cycle using label-free quantitative mass spectrom-etry and a novel machine learning analysis based on K-meansclustering and support vector machines. Modulated modularityclustering was used to group co-variable proteins intoexpression modules and Gene Ontology (GO) biologicalprocess and KEGG pathway enrichment analyses wereconducted for proteins within those modules. We discoveredthat components of the ribosome along with translationinitiation and elongation factors generally decrease as the annual ovarian cycle progresses toward ovulation, concomitant with aslight increase in components of the 26S-proteasome. Co-variation within more than one expression module of componentsfrom these two multi-protein complexes suggests that they are not only co-regulated, but that co-regulation occurs through morethan one sub-network. These components also co-vary with subunits of the TCP-1 chaperonin system and enzymes ofintermediary metabolic pathways, suggesting that protein folding and cellular bioenergetic state play important roles in proteinsynthesis and degradation. We provide further evidence to suggest that protein synthesis and degradation are intimately linked,and our results support function of a proteasome−ribosome supercomplex known as the translasome.

KEYWORDS: ribosome, proteasome, translasome, ovary, reproduction, mass-spectrometry, support vector machines, teleost, fish

■ INTRODUCTION

Cellular protein homeostasis depends on the balance ofcontinuous protein synthesis and degradation. Rates of proteinsynthesis and degradation also define protein turnover, whichunderlies adaptation of eukaryotic cells and organisms tochanging developmental states and physiological environments.Damaged proteins must be recycled and mistranslated proteinseliminated before they can act in dysfunctional manners.Complexes responsible for protein synthesis include theribosome and associated translation initiation and elongationfactors. The ribosome is a multi-protein complex comprised ofa 60S-large and 40S-small subunit, which catalyzes proteintranslation through polymerization of amino acids using mRNAas template.1 Most intracellular proteins are degraded byanother multi-protein complex, the 26S-proteasome.2 The 26S-proteasome comprises two subcomplexes: the 20S-catalyticcore and the 19S-regulatory particle. Targeted destruction ofproteins by the 26S-proteasome requires covalent linkage ofubiquitin and this process is mediated by a series of activating,conjugating, and ligating enzymes. These systems responsiblefor eukaryotic protein degradation are collectively termed theubiquitin-proteasome (UPS), and together they regulate the

stability of proteins involved in a wide range of cellularprocesses.If the cell produces more proteins than the UPS can

effectively turnover, then apoptosis occurs.3 Likewise, if proteinsynthesis cannot keep up with degradation, then the cell willeventually become devoid of protein over time. Therefore, aregulatory mechanism linking protein synthesis and degrada-tion must exist if homeostasis is maintained,4 and severalstudies have already established elementary linkages betweenthese processes.1,5−10 On average, 30% of eukaryotic cellularproteins are mistranslated by the ribosome or improperlyfolded during synthesis, and these defective products are co-translationally degraded by the 26S-proteasome.2,11 A super-complex termed the translasome resulting from co-localizationof ribosomes and proteasomes has been hypothesized toperform this co-translational degradation.12 However, acomprehensive listing of all interacting components of thetranslasome remains to be reported.Growing oocytes are in transcriptional stasis, acting as

storehouses of specific maternal RNAs, proteins, and other

Received: November 1, 2012Published: February 15, 2013

Article

pubs.acs.org/jpr

© 2013 American Chemical Society 1691 dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−1699

molecules required for fertilization competency and initiation ofzygotic development,13,14 and therefore are appropriate modelsfor study of protein turnover. In the present study, wecharacterize interactions of translasome components in thestriped bass (Morone saxatilis) ovary proteome using a label-free quantitative mass spectrometry based approach and ourrecently reported ovary transcriptome as a reference database.15

Additionally, since complex relationships are not maximallycaptured using traditional linear statistics,16 we report a novelanalysis of proteomics data based on machine learning.

■ METHODS

Sample Collection and Preparation

Female striped bass were reared in outdoor tanks at the NorthCarolina State University Pamlico Aquaculture Field Labo-ratory.17 Females were anesthetized with Finquel MS-222(Argent Chemical Laboratories, Redmond, WA), and wholeovary tissues were collected by dissection or by biopsy using aplastic cannula inserted through the urogenital pore.18 Tissueswere collected at four time points (N = 3 fish per point, allvalues are given as mean ± standard error of the mean): August(body weight 1.90 ± 0.35 kg; total length 512 ± 10.4 mm),November (3.80 ± 0.08 kg; 625 ± 3.45 mm), February (4.40 ±0.54 kg; 626 ± 27 mm), and April (4.54 ± 0.79 kg; 663 ± 43mm). As the striped bass is a group synchronous, single clutch,iteroparous spawner, the most advanced clutch of oocytesrepresented one of four stages of oocyte growth during thesetime points: early secondary growth (ESG), mid-vitellogenic(MVG), late-vitellogenic (LVG), or post-vitellogenic (PVG).The stage of ovarian development was initially judged from theseason, the appearance of biopsy samples under a dissectingstereomicroscope fitted with a calibrated ocular micrometer,and the maximum diameter of oocytes in the biopsy samples(ESG 310 ± 12.6 μm; MVG, 503 ± 34.7 μm; LVG, 831 ± 160μm; PVG, 986 ± 19.1 μm). The accuracy of the initialassignment of ovaries to stages was confirmed by histologicalexamination and oocyte staging following Berlinsky andSpecker.19 Tissues were fixed in a solution of 4% form-aldehyde/1% glutaraldehyde solution in 0.1 M phosphate buffer(pH 7.2−7.4), dehydrated in an ethanol series, embedded inparaffin, sectioned at 4 μm, and routinely stained withhematoxylin and eosin at the North Carolina State UniversityCollege of Veterinary Medicine Histology Laboratory (Raleigh,NC).Ovary tissues were frozen in liquid nitrogen and stored at

−80 °C before being homogenized 1:4 (w/v) in Milli-Qultrapure water (EMD Millipore Corporation, Billerica, MA).Extracts were centrifuged for 10 min at 13,000 rpm and 4 °C.The supernatant was diluted to 0.50 μg protein/μL with Milli-Q water. Ten microliters of each sample was added to 15.5 μLof 50 mM ammonium bicarbonate and 1.5 μL 100 mMdithiothreitol and incubated at 95 °C for 5 min. Upon cooling,3 μL of 100 mM iodoacetic acid was added, and samples wereincubated at room temperature in the dark for 20 min. Onemicroliter of 0.1 μg/μL porcine trypsin (Sigma, St. Louis, MO)was added, and samples were incubated at 37 °C for 3 h. Anadditional 1 μL of 0.1 μg/μL trypsin was added to each tubeand incubated at 30 °C overnight. Formic acid (1.5 μL of a 5%aqueous solution) was added to each sample to quench thetrypsin, and digests were dried to residues in a Savant speedvacuum (Thermo, San Jose, CA) for 30 min before submission

to the North Carolina State University Mass SpectrometryFacility (Raleigh, NC).

Mass Spectrometry

Digests were reconstituted in 100 μL of liquid chromatography(LC) mobile phase A [H2O/acetonitrile/formic acid (90/10/0.2 vol %)]. All samples were subjected to ultrafiltration for 30min at 15,000 rpm. Reversed phase HPLC separation andtandem mass spectrometry detection (nanoLC−MS/MS) wasperformed using an Eksigent (Dublin, CA) nanoLC-1D+system with autosampler coupled to a hybrid Thermo FisherLTQ Orbitrap XL mass spectrometer (Thermo Scientific, SanJose, CA). The nanoLC−MS/MS was operated with acontinuous vented column configuration for inline trap andelute.20 The analytical column was a self-packed 75 μm internaldiameter (i.d.) fused silica PicoFrit capillary (New Objective,Woburn, MA) with 15 cm of Magic C18AQ stationary phase(Michrom BioResources, Auburn, CA) in a methanolic slurry.The trap and dummy columns were self-packed 75 μm i.d.fused silica IntegraFrit capillaries (New Objective, Woburn,MA) with 5 and 20 cm of Magic C18AQ stationary phase(Michrom Bioresources, CA), respectively. The LC solventsused were mobile phase A and mobile phase B [H2O/acetonitrile/formic acid (10/90/0.2 vol %)]. Blank runs(injections of mobile phase A) were performed after everysample run to minimize carryover. Sample and blank injectionswere 2 μL on column. Each of the three biological replicatesfrom the four time points had three technical nanoLC−MS/MSruns. Analytical separations were performed on the nanoflowpump at 350 nL/min, initially maintaining 2% mobile phase B.The mass spectrometry (MS) method consisted of nine events:a precursor scan followed by eight data dependent tandem MSscans of the first−eighth most abundant peaks in the ion trap. Ahigh resolving power precursor scan of the eluted peptides wasobtained using the Orbitrap (60,000 resolution) with the eightmost abundant ions selected for MS/MS in the ion trapthrough dynamic exclusion. This method aims at good coverageof low and high abundance peptides. The instrument wasexternally tuned and calibrated according to the manufacturerand polycyclodimethylsiloxane (PCM; MH+ = 445.120024)from ambient air was employed as the lock mass for internalcalibration.21

Protein Identifications

NanoLC−MS/MS data were processed by MASCOT (MatrixScience, Boston, MA). Protein sequences for human keratins,porcine trypsin, and the striped bass ovary transcriptome(GenBank: SRX007394)15 translated in all six open readingframes with OrfPredictor22 were manually combined into oneFASTA file for MASCOT batch search (Supplementary File S1,Supporting Information). Vitellogenin gene transcripts are notexpressed in the striped bass ovary,15,23 thus vitellogenin-derived yolk proteins were not queried. The following variableand fixed amino acid modifications were allowed: variablemethionine (M) oxidation, asparagine (N) and glutamate (Q)deamidation, and fixed cysteine (C) carbamindomethylation.MASCOT search parameters were as follows: maximum missedcleavages 2, peptide charge 1+, 2+ and 3+, peptide tolerance ±5ppm, and MS/MS tolerance ±0.6 Da. Only those proteins witha probability >0.95 were reported.

Protein Quantifications and Analysis

ProteoIQ (NuSep, Bogart, GA) recently made available amethod of label-free quantitation of MS data. Briefly, spectral

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991692

Figure 1. Hierarchical clustering heat map of proteins expressed in striped bass ovary. The left panel shows the entire heat map of 355 proteins, andthe right panels show select regions indicated by the brackets that are enlarged. Proteins are listed by contig number, name, and approved geneabbreviation if known. “NA” indicates proteins of unknown orthology. Contig open reading frames are given in parenthetical brackets. Ovary stageswith different letters (a, b, or c) have significantly different proteomes by K-means clustering and support vector machines. Components of theubiquitin-proteasome, protein synthetic machineries (ribosome and TCP-1), and intermediary enzymes are indicated to the right of the heat map inred, green, and blue text, respectively. A complete version of this figure, with protein names, is included as Supplementary File S4 (SupportingInformation).

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991693

counts for identified proteins were normalized to total spectralcounts from MASCOT for each replicate. Shared peptides wereapportioned among protein groups24 and normalized to proteinlength (NSAF).25,26 These normalized spectral counts (N-SC)for each of the three technical replicates per biological samplewere exported from ProteoIQ and transformed to account forzero values [log10(y + 1), where y = N-SC].We performed K-means clustering as an unsupervised

learning tool to map protein expression [log10(y + 1)transformed N-SC values] to different stages of the annualreproductive cycle (ESG, MVG, LVG, and PVG) using WEKAversion 3.6.7 (http://www.cs.waikato.ac.nz/ml/weka/). The K-means clusters n objects into k partitions based on attributes, inthis case protein expression. We then evaluated the precision ofclustering into 2, 3, and 4 clusters using WEKA sequentialminimal optimization algorithm support vector machines(SVM) classifier.27 Hold-out estimates of classifier performancewere conducted using a stratified cross-validation with n = 10folds, where one fold was used for testing and n − 1 folds of therandomly reordered data set were used for training.

Graphic Representation of the Data

Approved gene abbreviations for all proteins were manuallycollected from the NCBI or GeneCards.28 Average N-SC valuesfor identified proteins were exported from ProteoIQ as baselinelog2 scale transformations. These values were normalized to themean of each protein expressed across ovary stages and Cluster3.0 was used to create a centroid linkage hierarchical clusteringheat map (Spearman Rank Correlation)29,30 that was visualizedusing Java Treeview.31 The log10(y + 1) transformed N-SC

values were subjected to a one-way ANOVA (α = 0.05) usingJMP Pro 9 (SAS Institute, Cary, NC), and residuals were inputfor modulated modularity clustering (MMC) performed usingPearson correlation coefficient.32 Relevance association net-works were generated from the MMC modules using a locallywritten MATLAB pipeline to select those with |r| ≥ 0.7, whichwere then fed into Cytoscape (www.cytoscape.org). Inter-actions between MMC modules 1−29 were computed from theaverage associations among modules and used to generaterelevance association networks as described above (|r| ≥ 0.3).The DAVID Functional Classification Tool33 was used togroup proteins on the basis of functional similarity withinMMC modules. Default parameters for DAVID were used andtherefore only those MMC modules with 4 or more proteinmembers were considered, as this is the minimum number ofmembers required to detect Gene Ontology (GO) enrichment.The KEGG Orthology System34 was used to further explore themolecular systems of the proteasome, ribosome, andintermediary pathways.

■ RESULTSRepresentative images of the histological sections of the stripedbass ovaries at the ESG, MVG, LVG, and PVG stage ofdevelopment are provided as Supplementary File S2 (Support-ing Information). Maximum oocyte diameters significantlydiffered between all ovarian stages except for LVG and PVG(analysis of variance F = 13.79, P = 0.002; followed byNewman−Keuls multiple range test, P ≤ 0.05).Data from the nanoLC−MS/MS and ProteoIQ are provided

as Supplementary File S3 (Supporting Information). A total of

Figure 2. (A) Modulated modularity clustering (MMC) heat map of 355 correlated striped bass ovary proteins (29 modules). (B) Relevancenetworks of ovary proteins correlated within MMC modules (|r| ≥ 0.7). Color-coding of networks corresponds to the horizontal bar shown underthe heat map in panel A. Only those relevance networks with 3 or more co-variable proteins are depicted. A ribosomal protein subcluster withinmodule 25 is indicated by the black arrow. (C) Interactions between MMC modules 1−29 (|r| ≥ 0.3).

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991694

355 individual proteins were identified in ovary of striped bass,and only 11 (∼3%) were unknown, unique sequences(Supplementary File S4, Supporting Information). A total of318, 307, 279, and 259 different proteins were detected in theESG, MVG, LVG, and PVG samples, respectively. Therefore,between 73% and 90% of the total individual proteinsindentified were expressed during each of the ovary stages. Atotal of 202 different proteins (57% of the total 355 proteins)had N-SC values >0 in all 4 ovary stages. Venn diagramsdepicting different proteins expressed by ovary stage areprovided as Supplementary File S5 (Supporting Information).Validation of K-means clustering by SVM indicates thatgrouping the expression data into 3 clusters (ESG, MVG, andLVG + PV) gives the best correct classification percentage(83.3%). Significant differences between the ovary proteomeoccur from PVG to ESG, ESG to MVG, and from MVG toLVG stages; there is no difference in the proteome from LVGto PVG stages (Figure 1). This clustering matches that ofmaximum oocyte diameter by stage, reinforcing the notion thatPVG is a translationally quiescent stage during which femaleswith fully grown oocytes await environmental conditionsappropriate to commence final maturation.Proteins were organized on the basis of co-variable

expression into 29 modules using MMC, and these aredepicted in Figure 2A (see Supplementary File S6, SupportingInformation). Relevance association networks within modulesare depicted with edges between protein nodes determined bycorrelations exceeding a threshold value (|r| ≥ 0.7) (Figure 2B).Relevance association networks between modules are similarlydepicted with edges representing average correlations where |r|≥ 0.3 (Figure 2C). These are undirected networks, andtherefore causality is unknown.We used DAVID to assess for each MMC module the degree

to which GO biological processes and pathways are over-represented. Four MMC modules (12, 25, 26, and 28) showedsignificant enrichment for proteasome, ribosome, and protein

chaperone (Table 1). The following KEGG pathways wereenriched for all proteins in the data set (1.0 × 10−3 cutoff):ribosome (1.8 × 10−59), proteasome (5.5 × 10−25), pentosepathway (1.8 × 10−5), and glycolysis/gluconeogenesis (2.0 ×10−4).

■ DISCUSSIONWe recently reported the first transcriptome database forspecies of the genus Morone15 and the present study providesthe first proteomic characterization using this referencedatabase. Proteome resources are currently available for othercommercially important finfishes, including Atlantic salmon(Salmo salar),35,36 channel catfish (Ictalurus punctatus),37

rainbow trout (Oncorhynchus mykiss),38−40 Senegalese sole(Solea senegalensis),41 yellow perch (Perca f lavescens),42 gilt-head seabream (Sparus aurata),43,44 and European seabass(Dicentrarchus labrax).45 Zebrafish (Danio rerio) is the only fishspecies with a published ovary proteome including over 1000proteins.13,14,44,46 Many of these proteins were similarlyidentified in striped bass ovary and include metabolic enzymes,chaperones, and regulators of protein synthesis and degrada-tion.MMC modules with the greatest intra-associations are

located in the upper left of the heat map (Figure 2A), andsuch association deceases from left to right down the diagonal.Most groupings with the greatest intra-association (modules 1−21) contain only 2−5 protein members, consisting mostly ofdyads and triads (Supplementary File S6, SupportingInformation). These small networks may represent feed-forward loops, feedback loops, or bifans, which are commonnetwork motifs that carry out key regulatory functions withinlarger biological networks.47,48 As examples, the followingproteins within these MMC modules have known regulatoryfunctions: Nop58 (module 8); Lsm14b (module 11); Piwi11(module 13); Nop56 (module 14); Snrpd2 (module 15); Ncl,Lsm14b and Pa2g4 (module 16); Ddx3y (module 17); Ddx4

Table 1. Enrichment of Striped Bass Ovary Proteins by Gene Ontology (GO) Class within Modulated Modularity Clustering(MMC) Modules 12, 25, 26, and 28 Using DAVID

MMC module GO Class and protein members (contig) P-value enrichment score

12 Ribosome 4.10 × 10−5 3.06Rps28 (11087), Rps16 (00157), Rpl31 (00051), Rpl38 (11185), Rpl22 (00840)

25 Ribosome 3.7 × 10−52 24.44Rpl13 (10709), Rps25 (00696), Rpl3 (09795), Rpl22l1 (02249), Rpl14 (11115), Rps3a (09669),Rps12 (10932), Rpl17 (02265), Rpl22 (03561), Rps14 (01136), Rps8 (09963), Rps18 (00105),Rpl15 (10800), Rps24 (11078), Rpl21 (09314), Rps11 (10659), Rps6 (01972), Rpl9 (10830),Rps4x (09952), Rpl7a (02265), Rps2 (10955), Rpl12 (09345), Rpl30 (09645), Rps23 (09464),Rpl7 (10144), Rplp0 (10309)Proteasome 6.4 × 10−11 3.15Psmb1 (01203), Psma3 (01407), Psma4 (09310), Psma6 (09001), Psmd1 (01972)

26 Ribosome 4.1 × 10−12 5.07Rpl26 (11079), Rpl10a (10894), Rplp2 (10044), Rpl8 (00527), Rps10 (00848), Rps19 (10401),Rpl37a (00866)Proteasome 6.4 × 10−11 3.12Psmd3 (10524), Psmc6 (00671), Psmd7 (09369), Psmb5 (03293), Psmb2 (02024)Chaperonin 5.4 × 10−7 2.77Cct7 (03096), Hsp90ab1 (10309), Tcp1 (00465), Cct4 (09681)

28 Proteasome 2.7 × 10−20 5.86Psmc2 (10820), Psmc1 (00084), Psmb3 (10149), Vcp (10277), Psmd11 (05603), Psmc6 (03192),Psmd2 (09229), Psma1 (10845), Psmc3 (10514), Psma2 (00069)Ribosome 1.9 × 10−10 4.80Rplp2 (01948), Rplp1 (00477), Rps13 (10658), Rps7 (10658), Rps5 (11189)Chaperonin 5.4 × 10−7 3.97Hspa8 (09917), Cct5 (09515), Cct2 (00164), Cct6a (01964)

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991695

(module 18); and Zar1 (module 19). Furthermore, the highdegree of interdependent association between these modulessuggests they may perform regulatory roles (Figure 2C).Sparsely populated MMC modules such as 3, 6, 11, 13, 21, and22 appear as network hubs with more edges (7 to 8) thanmembers (2 to 5 proteins). Identifying the presence of networkmotifs does not provide information to speculate directionality,since even 4-node bifans can have widely varying responses.49

The network motifs identified in the present study, however,will serve as the basis for additional experimentation aimed atunderstanding the intricacies of these interactions.In contrast, highly intraconnected MMC modules (23, 25,

26, and 28) are not interdependent, indicating that regulationof these units is divorced from other aspects of the proteinnetworks. Modules 25, 26, and 28 collectively contain 187proteins, representing over half of the proteome characterizedin this study (Supplementary File S6, Supporting Information).Therefore, number of protein members within a module doesnot appear to positively correlate with putative regulatoryimportance evaluated as a measure of network interdepend-ence.We used DAVID to evaluate GO class enrichment and

discovered that three cellular pathways co-vary: (1) proteinsynthesis machinery, including the ribosome and associatedcomponents, such as translation initiation and elongationfactors; (2) proteasome; and (3) chaperonin system. ThreeMMC modules (25, 26, and 28) are significantly enriched forproteasome (including regulation of ubiquitin ligase activity)and ribosome (including translational elongation), and two (26and 28) are significantly enriched for protein chaperone system(including chaperonin TCP-1) (Table 1). Proteins in modules25, 26, and 28 account for 64% (9 of the 14 components) and47% (9 of the 19 components) of the 20S-catalytic core and19S-regulatory particle of the proteasome, respectively and 50%(16 of the 32 components) and 39% (18 of the 46components) of the ribosomal 40S-small and 60S-largesubunits, respectively. An interesting feature of MMC module25 is a subcluster of 30 proteins that includes 25 ribosomalcomponents (Figure 2B), indicating a definitive role of proteinsynthesis. Six of the 8 proteins that form the hetero-oligomericTCP-1 are present in MMC modules 26 (Tcp1, Cct4, andCct7) and 28 (Cct2, Cct5, and Cct6a), and this complex acts asa molecular chaperone to fold nascent proteins.50 DAVID didnot detect GO enrichment for intermediary pathways withinMMC modules 25, 26, and 28; however, a small number of keymetabolic enzymes were shown to also co-vary with proteinsynthetic and degradation machineries (Table 2).Co-variation of different components representing the 26S-

proteasome and ribosome within three MMC modules suggestsnot only that these multi-protein complexes are co-regulatedbut that such co-regulation may occur in more than onemanner. This phenomenon may not be accidental, sinceseparating large supernetworks into smaller subnetworkpartitions reduces complexity and has previously beenidentified in regulation of the different ribosomal subunits.1,48

Therefore, the ribosome and 26S-proteasome may communi-cate through multiple subnetworks and this may be the case forother cellular multi-protein complexes as well.As the ovary progresses from recrudescence (ESG) through

the annual reproductive cycle (to PVG), an apparent reductionin protein synthetic capacity is observed (Figure 1). The TCP-1components vary in their expression as the ovary progressestoward ovulation. The Cct5, Cct6a, and Cct7 increase, whereas

Cct4 decreases and Tcp1 and Cct2 remain stable. A slightincrease in some protein translational components is observedduring MVG, concomitant with active vitellogenesis.19,23

Reduction in synthesis of ribosomal proteins by late stageoocytes is characterized in the mouse,51 and cytoplasmic latticeshave been shown to store ribosomal components withinoocytes during a period of selective translational repressionprior to ovulation.52,53 Due to our particular samplepreparation, any such stored ribosomes would have beendiscarded along with other insoluble cell membranes followingtissue homogenization. Therefore, the observed decrease inribosomal proteins during LVG and PVG in striped bass ovarymay reflect degradation or translocation to cytoplasmic lattices.Since such structures have not yet been described in fishoocytes, future study will be required to validate this possibility.In contrast, components of the protein degradation

machineries are either slightly upregulated or remainunchanged as the ovary progresses through the annualreproductive cycle (Figure 1). This indicates that capacity forprotein degradation remains intact even during periods ofapparent translational quiescence (PVG). Reduction inribosomes with concurrent increase in 20S-proteasomes isobserved during stress1 and translation elongation factors arelinked to UPS activity.9,11 A decrease in rate of proteintranslation during stress allows cells time to repair damage viathe UPS, since they need not dedicate efforts towardmonitoring nascent polypeptides.3 Additionally, maintainingthe 26S-proteasome during nutrient deprivation provides thecell with a means to cannibalize extant proteins for energy.8

The role(s) of elongation factors during such perturbations isless clear; however they may act to inhibit prematuredegradation of polypeptides through their interactions withthe UPS and substrates thereof. Six translation elongationfactors were identified including Eef1g, Eef1a1, Eef1d, Eef1b2,Eef2, and Eef1a2. The Eef1a1 and Eef2 are assigned to MMCmodule 25, and Eef1d is assigned to MMC module 28 (Figure2). Expression of some of these translation elongation factorsdecreases toward ovulation, whereas others remain unchanged(Figure 1). Other studies have provided evidence to suggestthat translation initiation factors also associate with compo-nents of the proteasome.12 This poses an interesting, additionalregulatory component to the system, since translation initiation

Table 2. Enzymes from Intermediary Pathways Assigned toModulated Modularity Clustering (MMC) Modules 25, 26,and 28

intermediary pathway enzyme (contig) MMC module

pyruvate metabolism Pkm2 (03943, 02192) 26, 28Glo1 (09544) 26Aldh7a1 (00630, 03089) 25, 28Mdh1 (03576, 02200) 26, 28Dld (04396) 26

glycolysis and gluconeogenesis Gapdh (10005) 26Tpi1 (00660) 26Aldob (01458) 28Pkm2 (03943, 02192) 26, 28Dld (04396) 26Aldh7a1 (00630, 03089) 25, 28Adh5 (09434) 26

tricarboxylic acid cycle Dld (04396) 26Mdh2 (00220) 26Idh2 (02438, 04323) 28

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991696

factor 3f (EIF3e) also acts as a deubiquitinase.54 Fourtranslation initiation factors were identified in striped bassovary, including Eif4h, Eif4e, Eif4a1, and Eif3b, and all fourdecrease in abundance toward ovulation. The Eif3b is assignedto MMC module 25 (Figure 2).The bioenergetic cost of proteins is not based on just

synthesis, but degradation as well, especially for proteins withshort half-lives, since ribosomes and proteasomes bothconsume ATP. Therefore, the rate of protein turnover isintrinsically dependent on bioenergetic affordability, anddelegate metabolic enzymes from glycolysis, gluconeogenesis,pyruvate cycle, and tricarboxylic acid cycle (TCA) wereidentified in MMC modules 25, 26, and 28 (Table 2). Amultipurpose regulatory role has already been reported forGapdh, an enzyme that participates in not only glycolysis butalso transcriptional activation.55 In addition to generating ATP,glycolysis provides biosynthetic intermediates that can be usedfor amino acid and nucleic acid synthesis and thus is animportant cycle for anabolic processes such as proteinsynthesis.56 Pkm2 governs the fate of glucose in this regard.Mdh and Idh catalyze two of the TCA steps that generatereducing power (i.e., NADH), and each enzyme has isoformsthat are expressed in the mitochondrion and cytosol.Additionally, Idh catalyzes the rate-limiting step of the TCA.

■ CONCLUSIONS

We used a novel analytical approach based on isotope-freequantitative MS/MS, machine learning, and MMC to show adirect and complex linkage between the cellular proteinsynthesis and degradation machineries and major bioenergeticmetabolic pathways in the ovary. Although we are not the firstto suggest the existence of a cellular translasome, we providethe first substantial index of proteins that potentially interact insuch a manner. We report the structures of protein networks;however, our observations require additional experimentalvalidations. Future studies aimed at uncovering the direction-ality of these interactions will allow us to understand how thesecombinations of proteins contribute to ovarian development,gamete quality, or pathology in striped bass and othervertebrate species. Since changes in cellular translationalcapacity, bioenergetics, and ribosome biogenesis rates areassociated with various cancers, our findings and methodologiesare relevant to human medicine as well.

■ ASSOCIATED CONTENT

*S Supporting Information

Additional experimental data and figures. This material isavailable free of charge via the Internet at http://pubs.acs.org

■ AUTHOR INFORMATION

Corresponding Author

*Tel: (919) 515-3830. Fax: (919) 515-2698. E-mail: [email protected].

Author Contributions

The manuscript was written through contributions of allauthors. All authors have given approval to the final version ofthe manuscript.

Notes

The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

We thank Andy S. McGinty and Michael S. Hopper (NCSUPamlico Aquaculture Field Laboratory) for care and main-tenance of the striped bass. This is an NAGRP AquacultureGenome (NRSP-8) Project and C.V.S. is the striped bassNRSP-8 species coordinator. This work was supported by bythe Center of Excellence in Oceans and Human Health CoEECenter for Marine Genomics at Hollings Marine Laboratoryand by research grants R/MG-1019 and R/12-SSS from theNorth Carolina Sea Grant Program and the National Oceanicand Atmospheric Administration, by special grant NC09211from the U.S. Department of Agriculture National Institute ofFood and Agriculture, and by the North Carolina AgriculturalFoundation, Inc. This manuscript is contribution number 703of the Marine Resources Division of the South CarolinaDepartment of Natural Resources.

■ ABBREVIATIONS

UPS, ubiquitin-proteasome system; ESG, early secondarygrowth; MVG, mid-vitellogenic growth; LVG, late-vitellogenicgrowth; PVG, post-vitellogenic; LC, liquid chromatography;MS, mass spectrometry; MS/MS, tandem mass spectrometry;nanoLC−MS/MS, reversed phase HPLC separation andtandem mass spectrometry; N-SC, normalized spectral count;SVM, support vector machines; MMC, modulated modularityclustering; TCA, tricarboxylic acid cycle

■ REFERENCES(1) Sprinzak, E.; Cokus, S. J.; Yeates, T. O.; Eisenberg, D.; Pellegrini,M. Detecting coordinated regulation of multi-protein complexes usinglogic analysis of gene expression. BMC Syst. Biol. 2009, DOI: 10.1186/1752-0509-3-115.(2) Goldberg, A. L. Protein degradation and protection againstmisfolded or damaged proteins. Nature 2003, 426 (6968), 895−9.(3) Jiang, H. Y.; Wek, R. C. Phosphorylation of the alpha-subunit ofthe eukaryotic initiation factor-2 (eIF2alpha) reduces protein synthesisand enhances apoptosis in response to proteasome inhibition. J. Biol.Chem. 2005, 280 (14), 14189−202.(4) Rothman, S. How is the balance between protein synthesis anddegradation achieved? Theor. Biol. Med. Modell. 2010, 7, 25DOI: 10.1186/1742-4682-7-25.(5) Andersen, J. S.; Lam, Y. W.; Leung, A. K.; Ong, S. E.; Lyon, C. E.;Lamond, A. I.; Mann, M. Nucleolar proteome dynamics. Nature 2005,433 (7021), 77−83.(6) Murata, T.; Shimotohno, K. Ubiquitination and proteasome-dependent degradation of human eukaryotic translation initiationfactor 4E. J. Biol. Chem. 2006, 281 (30), 20788−800.(7) Stavreva, D. A.; Kawasaki, M.; Dundr, M.; Koberna, K.; Muller,W. G.; Tsujimura-Takahashi, T.; Komatsu, W.; Hayano, T.; Isobe, T.;Raska, I.; Misteli, T.; Takahashi, N.; McNally, J. G. Potential roles forubiquitin and proteasome during ribosome biogenesis. Mol. Cell. Biol.2006, 26 (16), 5131−45.(8) Vabulas, R. M. Proteasome function and protein biosynthesis.Curr. Opin. Clin. Nutr. Metab. Care 2007, 10 (1), 24−31.(9) Li, X.; Matilainen, O.; Jin, C.; Glover-Cutter, K. M.; Holmberg,C. I.; Blackwell, T. K. Specific SKN-1/Nrf stress responses toperturbations in translation elongation and proteasome activity. PLoSGenet. 2011, 7 (6), e1002119 DOI: 10.1371/journal.pgen.1002119.(10) Ding, Q.; Dimayuga, E.; Markesbery, W. R.; Keller, J. N.Proteasome inhibition induces reversible impairments in proteinsynthesis. FASEB J. 2006, 20 (8), 1055−63.(11) Chuang, S. M.; Chen, L.; Lambertson, D.; Anand, M.; Kinzy, T.G.; Madura, K. Proteasome-mediated degradation of cotranslationallydamaged proteins involves translation elongation factor 1A. Mol. Cell.Biol. 2005, 25 (1), 403−13.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991697

(12) Sha, Z.; Brill, L. M.; Cabrera, R.; Kleifeld, O.; Scheliga, J. S.;Glickman, M. H.; Chang, E. C.; Wolf, D. A. The eIF3 interactomereveals the translasome, a supercomplex linking protein synthesis anddegradation machineries. Mol. Cell 2009, 36 (1), 141−52.(13) Knoll-Gellida, A.; Andre, M.; Gattegno, T.; Forgue, J.; Admon,A.; Babin, P. J. Molecular phenotype of zebrafish ovarian follicle byserial analysis of gene expression and proteomic profiling, andcomparison with the transcriptomes of other animals. BMC Genomics2006, 7, 46 DOI: 10.1186/1471-2164-7-46.(14) Knoll-Gellida, A.; Babin, P. J. In The Fish Oocyte: From BasicStudies to Biotechnological Applications; Babin, P. J., Cerda, J., Lubzens,E., Eds.; Springer: New York, NY, 2007; pp 77−97.(15) Reading, B. J.; Chapman, R. W.; Schaff, J. E.; Scholl, E. H.;Opperman, C. H.; Sullivan, C. V. An ovary transcriptome for allmaturational stages of the striped bass (Morone saxatilis), a highlyadvanced perciform fish. BMC Res. Notes 2012, 5, 111 DOI: 10.1186/1756-0500-5-111.(16) Chapman, R. W.; Mancia, A.; Beal, M.; Veloso, A.; Rathburn, C.;Blair, A.; Sanger, D.; Holland, A. F.; Warr, G. W.; Didonato, G. Atranscriptomic analysis of land-use impacts on the oyster, Crassostreavirginica, in the South Atlantic bight. Mol. Ecol. 2009, 18 (11), 2415−25.(17) Hodson, R. G.; Sullivan, C. V. Induced maturation andspawning of domestic and wild striped bass Morone saxatilis(Walbaum), broodstock with implanted GnRH analogue and injectedhCG. Aquacult. Res, 1993, 24 (3), 389−98.(18) Rees, R. A.; Harrell, R. M. In Culture and Propagation of StripedBass and its Hybrids; Harrell, R. M., Kerby, J. H., Minton, R. V., Eds.;American Fisheries Society: Bethesda, MD, 1990; pp 43−72.(19) Berlinsky, D. L.; Specker, J. L. Changes in gonadal hormonesduring oocyte development in the striped bass. Morone saxatilis. FishPhysiol. Biochem. 1991, 9 (1), 51−62.(20) Andrews, G. L.; Shuford, C. M.; Burnett, J. C.; Hawkridge, A.M.; Muddiman, D. C. Coupling of a vented column with splitlessNanoRPLC-ESI-MS for the improved separation and detection ofbrain natriuretic peptide-32 and its proteolytic peptides. J. Chromatogr.B 2009, 877 (10), 948−54.(21) Olsen, J. V.; de Godoy, L. M. F.; Li, G.; Macek, B.; Mortensen,P.; Pesch, R.; Makarov, A.; Lange, O.; Horning, S.; Mann, M. Parts permillion mass accuracy on an orbitrap mass spectrometer via lock massinjection into a C-trap. Mol. Cell. Proteomics 2005, 4, 2010−21.(22) Min, X. J.; Butler, G.; Storms, R.; Tsang, A. OrfPredictor:predicting protein-coding regions in EST-derived sequences. NucleicAcids Res. 2005, 33 (Web Server issue), W677−80, DOI: 10.1093/nar/gki394.(23) Reading, B. J.; Hiramatsu, N.; Sawaguchi, S.; Matsubara, T.;Hara, A.; Lively, M. O.; Sullivan, C. V. Conserved and variantmolecular and functional features of multiple egg yolk precursorproteins (vitellogenins) in white perch (Morone americana) and otherteleosts. Mar. Biotechnol. 2009, 11 (2), 169−87.(24) Zhang, Z.; Wen, Z.; Washburn, M. P.; Florens, L. Refinementsto label free proteome quantitation: how to deal with shared peptidesshared by multiple proteins. Anal. Chem. 2010, 82 (6), 2272−81.(25) Zybailov, B. L.; Mosley, A. L.; Sardiu, M. E.; Coleman, M. K.;Florens, L.; Washburn, M. P. Statistical analysis of membraneproteome expression changes in Saccharomyces cerevisiae. J. ProteomeRes. 2006, 5 (9), 2339−47.(26) Zybailov, B. L.; Florens, L.; Washburn, M. P. Quantitativeshotgun proteomics using a protease with broad specificity andnormalized spectral abundance factors. Mol. Biosyst. 2007, 3 (5), 354−60.(27) Chapman, R. W.; Mancia, A.; Beal, M.; Veloso, A.; Rathburn, C.;Blair, A.; Holland, A. F.; Warr, G. W.; Didinato, G.; Sokolova, I. M.;Wirth, E. F.; Duffy, E.; Sanger, D. The transcriptomic responses of theeastern oyster, Crassostrea virginica, to environmental conditions. Mol.Ecol. 2011, 20 (7), 1431−49.(28) Rebhan, M.; Chalifa-Caspi, V.; Prilusky, J.; Lancet, D.GeneCards: a novel functional genomics compendium with automated

data mining and query reformulation support. Bioinformatics 1998, 14(8), 656−64.(29) de Hoon, M. J. L.; Imoto, S.; Nolan, J.; Miyano, S. Open sourceclustering software. Bioinformatics 2004, 20 (9), 1453−4.(30) Eisen, M. B.; Spellman, P. T.; Brown, P. O.; Botstein, D. Clusteranalysis and display of genome-wide expression patterns. Proc. Natl.Acad. Sci. U.S.A. 1998, 95 (25), 14863−8.(31) Saldanha, A. J. Java Treeview–extensible visualization ofmicroarray data. Bioinformatics 2004, 20 (17), 3246−8.(32) Stone, E. A.; Ayroles, J. F. Modulated modularity clustering asan exploratory tool for functional genomic inference. PLoS Genet.2009, 5 (5), e1000479 DOI: 10.1371/journal.pgen.1000479.(33) Huang, D. W.; Sherman, B. T.; Lempicki, R. A. Systematic andintegrative analysis of large gene lists using DAVID bioinformaticsresources. Nat. Protoc. 2009, 4 (1), 44−57.(34) Kanehisa, M.; Goto, S.; Sato, Y.; Furumichi, M.; Tanabe, M.KEGG for integration and interpretation of large-scale moleculardatasets. Nucleic Acids Res. 2012, 40 (Database issue), D109−14,DOI: 10.1093/nar/gkr988.(35) Provan, F.; Bjornstad, A.; Pampanin, D. M.; Lyng, E.;Fontanillas, R.; Andersen, O. K.; Koppe, W.; Bamber, S. Massspectrometric profiling-a diagnostic tool in fish? Mar. Environ. Res.2006, 62 (Suppl), S105−8.(36) Liu, X.; Afonso, L.; Altman, E.; Johnson, S.; Brown, L.; Li, J. O-acetylation of sialic acids in N-glycans of Atlantic salmon (Salmo salar)serum is altered by handling stress. Proteomics 2008, 8 (14), 2849−57.(37) Booth, N. J.; Bilodeau-Bourgeois, L. Proteomic analysis of headkidney tissue from high and low susceptibility families of channelcatfish following challenge with Edwardsiella ictaluri. Fish ShellfishImmunol. 2009, 26 (1), 193−6.(38) Martin, S. A.; Vilhelmsson, O.; Medale, F.; Watt, P.; Kaushik, S.;Houlihan, D. F. Proteomic sensitivity to dietary manipulations inrainbow trout. Biochim. Biophys. Acta 2003, 165 (1−2), 17−29.(39) Rime, H.; Guitton, N.; Pineau, C.; Bonnet, E.; Bobe, J.; Jalabert,B. Postovulatory ageing and egg quality: A proteomic analysis ofrainbow trout coelomic fluid. Repr. Biol. Endocr 2004, 2, 26DOI: 10.1186/1477-7827-2-26.(40) Wulff, T.; Hoffmann, E. S.; Roepstorff, P.; Jessen, F.Comparison of two anoxia models in rainbow trout cells by a 2-DEand MS/MS-based proteome approach. Proteomics 2008, 8 (10),2035−44.(41) Forne, I.; Agulleiro, M. J.; Asensio, E.; Abian, J.; Cerda, J. 2-DDIGE analysis of Senegalese sole (Solea senegalensis) testis proteome inwild-caught and hormone-treated F1 fish. Proteomics 2009, 9 (8),2171−81.(42) Reddish, J. M.; St-Pierre, N.; Nichols, A.; Green-Church, K.;Wick, M. Proteomic analysis of proteins associated with body massand length in yellow perch, Perca f lavescens. Proteomics 2008, 8 (11),2333−43.(43) Zilli, L.; Schiavone, R.; Storelli, C.; Vilella, S. Molecularmechanisms determining sperm motility initiation in two sparids(Sparus aurata and Lithognathus mormyrus). Biol. Reprod. 2008, 79 (2),356−66.(44) Ziv, T.; Gattegno, T.; Chapovetsky, V.; Wolf, H.; Barnea, E.;Lubzens, E.; Admon, A. Comparative proteomics of the developingfish (zebrafish and gilthead seabream) oocytes. Comp. Biochem.Physiol., D 2008, 3 (1), 12−35.(45) Zilli, L.; Schiavone, R.; Zonno, V.; Rossano, R.; Storelli, C.;Vilella, S. Effect of cryopreservation on sea bass sperm proteins. Biol.Reprod. 2005, 72 (5), 1262−7.(46) Groh, K. J.; Nesatyy, V. J.; Segner, H.; Eggen, R. I.; Suter, M. J.Global proteomics analysis of testis and ovary in adult zebrafish (Daniorerio). Fish Physiol. Biochem. 2011, 37 (3), 619−47.(47) Milo, R.; Shen-Orr, S.; Itzkovitz, S.; Kashtan, N.; Chklovskii, D.;Alon, U. Network motifs: simple building blocks of complex networks.Science 2002, 298 (5594), 824−7.(48) Alon, U. Network motifs: theory and experimental approaches.Nat. Rev. Genet. 2007, 8 (6), 450−61.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991698

(49) Ingram, P. J. Network motifs: structure does not determinefunction. BMC Genomics 2006, 7, 108 DOI: 10.1186/1471-2164-7-108.(50) Frydman, J.; Nimmesgern, E.; Erdjument-Bromage, H.; Wall, J.S.; Tempst, P.; Hartl, F. U. Function in protein folding of TRiC, acytosolic ring complex containing TCP-1 and structurally relatedsubunits. EMBO J. 1992, 11 (13), 4767−78.(51) Taylor, K. D.; Piko, L. Expression of ribosomal protein genes inmouse oocytes and early embryos. Mol. Reprod. Dev. 1992, 31 (3),182−8.(52) Yurttas, P.; Vitale, A. M.; Fitzhenry, R. J.; Cohen-Gould, L.; Wu,W.; Gossen, J. A.; Coonrod, S. A. Role for PADI6 and the cytoplasmiclattices in ribosomal storage in oocytes and translational control in theearly mouse embryo. Development 2008, 135 (15), 2627−36.(53) Clarke, H. J. Post-transcriptional control of gene expressionduring mouse oogenesis. Results Probl. Cell Differ 2012, 55, 1−21.(54) Moretti, J.; Chastagner, P.; Gastaldello, S.; Heuss, S. F.; Dirac, A.M.; Bernards, R.; Masucci, M. G.; Israel, A.; Brou, C. The translationinitiation factor 3f (eIF3f) exhibits a deubiquitinase activity regulatingNotch activation. PLoS Biol. 2010, 8 (11), e1000545 DOI: 10.1371/journal.pbio.1000545.(55) Tarze, A.; Deniaud, A.; Le Bras, M.; Maillier, E.; Molle, D.;Larochette, N.; Zamzami, N.; Jan, G.; Kroemer, G.; Brenner, C.GAPDH, a novel regulator of the pro-apoptotic mitochondrialmembrane permeabilization. Oncogene 2007, 26 (18), 2606−20.(56) Barger, J. F.; Plas, D. R. Balancing biosynthesis andbioenergetics: metabolic programs in oncogenesis. Endocr. Relat.Cancer 2010, 17 (4), R287−304, DOI: 10.1677/ERC-10-0106.

Journal of Proteome Research Article

dx.doi.org/10.1021/pr3010293 | J. Proteome Res. 2013, 12, 1691−16991699