Novel Proteins, Putative Membrane Transporters, and an ... · Novel Proteins, Putative Membrane Transporters, and an Integrated Metabolic Network Are Revealed by Quantitative Proteomic
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Novel Proteins, Putative Membrane Transporters, and anIntegrated Metabolic Network Are Revealed byQuantitative Proteomic Analysis of Arabidopsis CellCulture Peroxisomes1[W][OA]
Holger Eubel2, Etienne H. Meyer2, Nicolas L. Taylor2, John D. Bussell2, Nicholas O’Toole,Joshua L. Heazlewood, Ian Castleden, Ian D. Small, Steven M. Smith, and A. Harvey Millar*
Australian Research Council Centre of Excellence in Plant Energy Biology, M316 (H.E., E.H.M., N.L.T., J.D.B.,J.L.H., I.D.S., S.M.S., A.H.M.), and Centre of Excellence for Computational Systems Biology (N.O., I.C., I.D.S.),University of Western Australia, Crawley, Western Australia 6009, Australia
Peroxisomes play key roles in energy metabolism, cell signaling, and plant development. A better understanding of theseimportant functions will be achieved with a more complete definition of the peroxisome proteome. The isolation of peroxisomesand their separation from mitochondria and other major membrane systems have been significant challenges in the Arabidopsis(Arabidopsis thaliana) model system. In this study, we present new data on the Arabidopsis peroxisome proteome obtained usingtwo new technical advances that have not previously been applied to studies of plant peroxisomes. First, we followed densitygradient centrifugation with free-flow electrophoresis to improve the separation of peroxisomes from mitochondria. Second, weused quantitative proteomics to identify proteins enriched in the peroxisome fractions relative to mitochondrial fractions.We provide evidence for peroxisomal localization of 89 proteins, 36 of which have not previously been identified in other analysesof Arabidopsis peroxisomes. Chimeric green fluorescent protein constructs of 35 proteins have been used to confirm their lo-calization in peroxisomes or to identify endoplasmic reticulum contaminants. The distribution of many of these peroxisomalproteins between soluble, membrane-associated, and integral membrane locations has also been determined. This core per-oxisomal proteome from nonphotosynthetic cultured cells contains a proportion of proteins that cannot be predicted to beperoxisomal due to the lack of recognizable peroxisomal targeting sequence 1 (PTS1) or PTS2 signals. Proteins identified are likelyto be components in peroxisome biogenesis, b-oxidation for fatty acid degradation and hormone biosynthesis, photorespiration,and metabolite transport. A considerable number of the proteins found in peroxisomes have no known function, and potentialroles of these proteins in peroxisomal metabolism are discussed. This is aided by a metabolic network analysis that reveals a tightintegration of functions and highlights specific metabolite nodes that most probably represent entry and exit metabolites thatcould require transport across the peroxisomal membrane.
Within the plant cell, energy metabolism is mainlydistributed among three distinct organelles: plastids,mitochondria, and peroxisomes. Although the pro-teomes of both plastids and mitochondria have beeninvestigated extensively, comparatively little system-
atic analysis of the protein content of plant peroxisomeshas been undertaken. The main obstacle for proteomicsof plant peroxisomes is the availability of purifiedorganelles from model plants that are also amenableto mass spectrometry (MS)-based identification bymatching to protein sequence data. Whereas the prep-aration of peroxisomes in sufficient amounts and purityfrom spinach (Spinacia oleracea), cucumber (Cucumissativus), pea (Pisum sativum), and soybean (Glycine max)for proteomic purposes is possible (Schwitzguebel andSiegenthaler, 1984; Corpas et al., 1994; Lopez-Huertaset al., 1999; Arai et al., 2008), the purification of perox-isomes from Arabidopsis (Arabidopsis thaliana) hasproved to be extremely difficult due to the low yieldof intact organelles and contamination with other cellorganelles. This complicates data analysis and com-promises confidence in the subcellular localization ofthe identified proteins. So far, three studies in Arabi-dopsis have been reported, using greening (Fukao et al.,2002) or etiolated (Fukao et al., 2003) cotyledons ormature plant leaves (Reumann et al., 2007), each usingdifferent purification methods. In these studies, 42putatively peroxisomal proteins were identified from
1 This work was supported by grants from the Australian Re-search Council (ARC) through the Centres of Excellence Program(grant no. CE0561495), by the Western Australian State Governmentvia its Centres of Excellence program, and by a University of WesternAustralia Research Grant to J.D.B. H.E., N.L.T., and J.L.H. aresupported as ARC Australian Postdoctoral Fellows, A.H.M. as anARC Australian Professorial Fellow, and S.M.S. as an ARC Feder-ation Fellow.
2 These authors contributed equally to the article.* Corresponding author; e-mail [email protected] author responsible for distribution of materials integral to the
findings presented in this article in accordance with the policydescribed in the Instructions for Authors (www.plantphysiol.org) is:A. Harvey Millar ([email protected]).
[W] The online version of this article contains Web-only data.[OA] Open Access articles can be viewed online without a sub-
Plant Physiology, December 2008, Vol. 148, pp. 1809–1829, www.plantphysiol.org � 2008 American Society of Plant Biologists 1809 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
cotyledons and 78 from leaves, but the overlap betweenthe sets from both tissues was only 11 proteins.
The protein composition of peroxisomes from differ-ent tissues is likely to vary significantly as the functionof these organelles changes. Therefore, a full under-standing of peroxisomal function requires experimen-tal analysis of these organelles from a variety of plantorgans during different developmental stages. Peroxi-somes in seedlings of oilseed plants such as Arabidop-sis are mainly involved in the breakdown of fatty acidsderived from storage triacylglycerols via b-oxidationduring germination prior to the initiation of photosyn-thesis (Graham and Eastmond, 2002). Most of theacetyl-CoA generated by fatty acid b-oxidation is fedinto the glyoxylate cycle to produce succinate, whichmay then be exported out of the organelles or used as aprecursor for other metabolites and processes such asgluconeogenesis (Eastmond and Graham, 2001). Leafperoxisomes too perform b-oxidation; however, thisusually happens at a lower rate and is also involved inthe production of signaling compounds and hormonessuch as jasmonic acid (JA) and in the conversion ofindole-3-butyric acid (IBA) into indole-3-acetic acid. Amajor role of peroxisomes in leaf tissue is in photores-piration by oxidation of glycolate derived from theoxygenase reaction of Rubisco to make substrates formitochondria and the reduction of Ser to glycerate forthe return of carbon intermediates to the Calvin cycle(Raghavendra et al., 1998). Peroxisomes in senescingtissue are multifunctional organelles involved in thedegradation of cellular constituents, including fattyacids and the remobilization of nitrogen into ureides(Vicentini and Matile, 1993). The transition from oneform to another is mediated by a change in the proteincontent of existing organelles rather than by a degra-dation and de novo synthesis of organelles (Hayashiet al., 2000). Apart from the above-mentioned func-tions, plant peroxisomes are also involved in nitrogenmetabolism in root nodule cells, amino acid and ureidemetabolism, and the degradation of hydrogen peroxideproduced during a number of their catalytic functions(Hayashi and Nishimura, 2006).
The size of the peroxisomal proteome is unknown,but it is probably substantially smaller than the thou-sands of proteins found in the endosymbiont-derivedmitochondria and chloroplasts. Peroxisomes lack ge-netic material and therefore do not require proteinsinvolved in genome replication, transcription, matura-tion of transcripts, or translation. However, due to thediversity of the plant-specific roles of these organelles,the proteome of the plant peroxisome may well belarger than that of its mammalian or fungal counter-parts (Emanuelsson et al., 2003). All peroxisomal pro-teins are imported posttranslationally into the organelleand therefore require some form of targeting recogni-tion sequence or secondary structure.
Matrix proteins can be directed to the peroxisomes byone of two types of peroxisomal targeting signals(PTSs). PTS1 signals consist of three amino acids atthe C terminus of a peroxisomal protein. Although a
considerable amount of variation in the PTS1 sequenceexists, it usually consists of a small amino acid residue,followed by a basic one, and then a hydrophobicresidue, and it is not cleaved off after import. SKL is atypical PTS1 sequence. PTS2 sequences are composedof nine amino acids located at the N terminus ofperoxisomal proteins and are removed after importinto the organelle. RLx5HL and RIx5HL are typical PTS2sequences. Searches for peroxisome targeting signalswithin the protein-coding regions of the Arabidopsisgenome have identified 256 to 280 proteins containingputative PTS signals (Kamada et al., 2003; Reumannet al., 2004). However, not all matrix proteins foundexperimentally in peroxisomes contain these knownPTS signals. Recently, the presence of a novel PTS1 in anenoyl-CoA hydratase involved in the b-oxidation ofcis-unsaturated fatty acids was described (Ser-Ser-Leu;Goepfert et al., 2006), which was recently confirmed bya study of the leaf peroxisomal proteome (Reumannet al., 2007). Other peroxisomal proteins seem to lackconventional PTS sequences at the N or C terminusbut possess internal sequences serving as targeting sig-nals. The most prominent example is catalase, whichpossesses an internal, PTS1-like targeting sequence(Kamigaki et al., 2003) but is not recognized for importby the normal PTS1 mechanism (Oshima et al., 2008).
Peroxisomal membrane proteins (PMPs) do notpossess PTS1 or PTS2 sequences. Instead, they con-tain a stretch of positively charged amino acids that isusually flanked by transmembrane domains. Some-times, this sequence is referred to as a membrane PTS(mPTS). However, it is not as conserved as conven-tional PTS1 and PTS2 sequences, and the definition ofa consensus sequence for membrane targeting ofperoxisomal proteins is difficult (Trelease, 2002). Ingeneral, two import pathways for proteins destinedfor the peroxisomal membrane are discussed. In thefirst model, proteins are synthesized in the cytosoland subsequently directly inserted into the peroxi-somal membrane and are usually said to have a mPTStype 1 (mPTS1). Alternatively, proteins can be syn-thesized on rough endoplasmic reticulum (ER) andinserted cotranslationally into the ER membrane.Vesicles containing these proteins then bud from theER and fuse with the peroxisomal membrane. Inaddition to the mPTS1 sequence, these proteins alsocontain an ER sorting signal, and the combination ofboth the mPTS1 and the ER signal is referred to asmPTS2. The finding that peroxisomal membraneproteins such as ascorbate peroxidase (APX) andthe peroxins PEX10 and PEX16 are transferred tothe peroxisomes via the ER led to the formulation ofthe ‘‘ER semiautonomous peroxisome maturationand replication’’ model (for review, see Mullen andTrelease, 2006). It uses the largely contradictorymodels of autonomous organelles and purely ER-derived peroxisomes and combines them. Accordingto the semiautonomous maturation model, peroxi-somes can be derived by budding from the ER butalso by fission of existing organelles.
Eubel et al.
1810 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
In a bid to better experimentally define the solubleand membrane proteome of peroxisomes in Arabidop-sis, we have combined conventional centrifugation-based organelle isolation techniques with free-flowelectrophoresis (FFE), which has already been success-fully applied to the isolation of other organelles inplants (Bardy et al., 1998; Eubel et al., 2007) and perox-isomes in mammals (Volkl et al., 1997). As this tech-nique employs surface charge rather than the size ordensity of particles as a separating parameter, it repre-sents a true additional dimension in the purificationprocess and can lead to organellar fractions of greaterpurity. Peroxisomal proteins were analyzed by MS andnonperoxisomal proteins were excluded by quantita-tive comparison between highly purified peroxisomalsamples and other cellular fractions. Where needed,organellar localization was also confirmed by chimericfluorescent protein visualization. A metabolic mappingapproach was adopted to gauge the completeness ofthis peroxisomal proteome and, therefore, our under-standing of the biochemical processes taking placewithin it and the biogenesis of this organelle.
RESULTS
FFE Separation of Organelles to PurifyArabidopsis Peroxisomes
An organellar fraction consisting mainly of mitochon-dria and peroxisomes was obtained from disruptedArabidopsis protoplasts using differential centrifuga-tion and density gradient purification. FFE was thenemployed to separate peroxisomes from the bulk ofmitochondrial material. We have demonstrated previ-ously that FFE is able to increase the purity of mito-chondrial isolates by separating them from plastids andperoxisomes according to differences in the surfacecharge of each organelle, yielding mitochondria thathave seven times less contamination (Eubel et al., 2007).Under the conditions applied, mitochondria, peroxi-somes, and other cellular material migrated close to-gether in FFE, but with different degrees of overlap(Eubel et al., 2007). The electrophoretic mobility ofperoxisomes was lower than that of mitochondria at600 V (Fig. 1A). In order to optimize FFE for theisolation of Arabidopsis peroxisomes, a higher voltage(800 V) was applied to increase the separation betweenthe mitochondrial and peroxisomal fraction peaks (Fig.1B). The distribution of mitochondria and peroxisomeswas monitored using immunodetection of marker pro-teins. At 600 V, the mitochondrial fraction markerprotein, HSP70, peaked in fraction 44, while the perox-isomal marker, KAT2, peaked in fraction 48, but thedistributions were overlapping. By expanding the dis-tributions at 800 V, both mitochondrial and peroxi-somal signals moved further toward anodic fractions,revealing two mitochondrial subpopulations with dif-ferent electrophoretic mobilities peaking between frac-tions 21 to 23 and fractions 29 to 31. Meanwhile, thepeak fractions enriched in peroxisomes were nearly 10
fractions away, in fractions 39 to 45, and devoid ofvisible mitochondrial contamination. While this highervoltage did tend to lead to greater losses of KAT2 signalinto the anodic mitochondrial fractions than was ap-parent at 600 V (Fig. 1A), the displacement of the bulk ofmitochondria from peroxisomes was optimal for thepreparation of peroxisomes. Using this approach, per-oxisomal and mitochondrial samples with only mini-mal amounts of contamination were obtained byselecting the central three fractions of each distributionat 800 V to ensure the best compromise between yieldand purity. In order to produce a highly enrichedmitochondrial sample for comparative functional as-says against the peroxisomal fraction, the mitochondriawere then subjected to another round of FFE at 600 V(i.e. repeating the separation in Fig. 1A) to separatethem from contaminating plastids, which migratedinto the left border of the separation chamber andtherefore were not resolved well from the mitochondriaat 800 V (data not shown).
Measurement of catalase activity as a marker forperoxisomes gave an oxygen production rate of 60mmol min21 in pre-FFE organelle pellets (SupplementalTable S1A), which was approximately 20% of totalcellular extract activity. Total catalase activity was ap-proximately 15 mmol min21 in the FFE-purified perox-isome fraction, which is approximately 23% of theactivity in pre-FFE organelle pellets and approximately4% of total cellular extract activity. This degree ofrecovery is similar to that shown in preparations ofleaf peroxisomes from Arabidopsis when followinghydroxypyruvate reductase activity (Reumann et al.,2007). To quantify the difference between mitochon-drial and peroxisome fractions, catalase and succinatedehydrogenase activities were measured as markerenzymes (Supplemental Table S1A). The specific activ-ity of catalase was much higher in the putative perox-isomal fraction than in the pre-FFE or mitochondrialsample, indicating clear enrichment of peroxisomes inthis fraction. At the same time, the specific activity ofsuccinate dehydrogenase in the putative peroxisomalfraction was only approximately 8% of that measuredin the mitochondrial sample.
Coomassie Brilliant Blue-stained SDS gel lanes of apre-FFE organellar sample compared with post-FFEsamples of pooled peroxisomes and mitochondriafractions are shown in Figure 1C. While the bandingpattern of the pre-FFE sample was very similar to thatof the mitochondrial fraction, the putative peroxisomesdisplayed a distinct pattern with very little resem-blance to the other two samples. Therefore, we con-cluded that the pre-FFE fraction contained mostlymitochondria and only a limited proportion of otherorganelles, whereas the putative peroxisomal fractionwas largely free of mitochondrial proteins. Abundantprotein bands present in the mitochondrial and perox-isomal samples were excised from one-dimensional(1D) gels for identification by MS (Fig. 1C, annotationson gel lanes). The primary protein identification foreach spot is summarized in Supplemental Table S1B,
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1811 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
supporting these as purified mitochondrial and perox-isomal fractions derived from FFE. Subsequent quan-titative proteomic analysis (see below) allowed thequantification of a series of classical peroxisomal andmitochondrial markers between the two fractions. Thisshows an average ratio of 0.14 for mitochondrial pro-teins in the peroxisomal compared with the mitochon-drial samples and approximately 70-fold enrichmentfor peroxisomal proteins in the peroxisomal comparedwith mitochondrial samples (Fig. 1D). In general agree-ment, the succinate dehydrogenase specific activity inthe peroxisomes was one-twelfth of that in the mito-chondrial sample (18 6 1 versus 231 6 45 nmol O2min21 mg21 protein), and the catalase specific activitywas approximately 30-fold higher in the peroxisomesample (13.7 6 1.7 versus 513 6 181 nmol O2 min21
mg21 protein; Supplemental Table S1A). Based on thesespecific activity measurements and protein ratios, weestimated the approximate peroxisome purity as 85%to 90%. This is based on mitochondria being the largestcontaminant in the peroxisome preparations and thismitochondrial contamination being 8% to 14% (basedon Fig. 1D and Supplemental Table S1A). Also, thecatalase specific activity in the mitochondrial fraction isonly 3% of that in the peroxisome fraction, so the at leastapproximately 30-fold increase in catalase and otherperoxisome markers indicates that approximately 90%of the protein in the peroxisome fraction should be ofperoxisomal origin.
Whole Peroxisome Protein Profiling
An in-depth analysis of the proteins in each fractionwas obtained by two-dimensional differential in gelelectrophoresis isoelectric focusing (DIGE 2D IEF/SDS-PAGE) using three independent peroxisomesand mitochondria preparations. Figure 2 shows thefalse-colored spot intensity maps for both organellefractions (top) as well as a superimposition of the twospot maps for one of the three gel sets. The amount ofoverlap between the two samples is low, and twoclearly distinct patterns can be observed. While themitochondrial pattern is focused around pI valuesbetween 5 and 8 and resembles that of a previouslypublished study using the same material (Millar et al.,2001), the distribution of the peroxisomal pattern isheavily skewed toward the basic side of the pI spec-trum. This shift to basic pI can also be observed on 2Dgels of leaf peroxisome proteins (Reumann et al., 2007).Many peroxisome protein spots of high abundancehave very high pI values, and a considerable numberdid not even reach their pI on the pH 3 to 10 nonlinearimmobilized pH gradient strips but migrated right tothe cathode. The basic nature of many peroxisome pro-teins may be related to the typically alkaline nature (pH8–8.5) of the peroxisome lumen (van Roermund et al.,2004). Quantification of mitochondrial and peroxisomal
Figure 1. Quantification of protein content from 1D SDS-PAGE andimmunoblots of every second of the central 30 fractions collected afterFFE separation at 600 V (A) and 800 V (B). Relative protein quantity isdisplayed as the percentage of the fraction with the highest abundance(top); distribution of marker proteins for mitochondria (mtHSP70) andperoxisomes (3-ketoacyl-CoA thiolase; KAT2) is shown below. C, 1DSDS-PAGE of 40 mg of pre-FFE organelle protein sample and pooledprotein fractions of peroxisomes and mitochondria stained withCoomassie Brilliant Blue. Bands indicated on gels (A1–A5 and B1–B4) were in-gel digested and analyzed by MS (Supplemental TableS1B). Molecular masses in kD are shown at left of gel lanes. D,Quantification of classical mitochondrial and peroxisomal markers infractions shown in C by 2D DIGE analysis (Supplemental Table S2).Each protein’s AGI accession number and description are shown along
with the average ratio of the quantitation between peroxisomal andmitochondrial samples (n 5 3, P , 0.05).
Eubel et al.
1812 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
spot intensities from all three gel sets was performedusing the DeCyder software package (GE Healthcare).In total, 1,019 matched spots were found consistentlyacross the three sets. Spots with P # 0.05 and an averageratio of 2 or greater (mitochondria-peroxisomes) weredesignated as mitochondrial, and those with an averageratio of 2 or greater (peroxisomes-mitochondria) weredesignated as peroxisomal. On average, the peroxisome-mitochondria ratio was approximately 37 for selectedputative peroxisomal protein spots, while the averagemitochondria-peroxisome ratio was approximately5 for selected putative mitochondrial protein spots.Spots from a preparative gel were matched withthose on the DIGE gels. Prominent peroxisomal andmitochondrial spots matched to the DIGE gels (Sup-plemental Fig. S1) were excised for tandem massspectrometry (MS/MS) identification. In total, 136unique proteins were identified from this gel (Sup-plemental Table S2).
As a complement to the gel-based approaches,reverse-phase HPLC separation coupled with MS/
MS was used to gain additional depth in the wholeperoxisome proteome. Four 60-min Intelligent Data-Dependent Acquisition (IDDA) runs of the same sam-ple were conducted in series, with peptides identifiedin each run subsequently excluded from MS/MS acqui-sition in the next run. This analysis was performed onthe same peroxisomal sample used for the 2D gelscontaining 8% to 10% mitochondria (high purity) andalso on a second sample containing twice the amountof mitochondrial contamination (21%; low purity) asdeduced from the respiratory assays. In cases in whicha protein was identified in both samples, the compar-ison of (1) the emPAI values (which is the exponen-tially modified ratio of the observed peptides over thenumber of peptides that can theoretically be observed)and (2) the difference in a protein’s MOWSE scorebetween the samples was taken as a semiquantitativeindicator for localization of a protein. Proteins with ahigher emPAI value or a higher MOWSE score in thehigh-purity sample are putative peroxisomal proteins,and those that produced higher values in the low-purity sample are putative contaminants. In cases inwhich a protein was identified in only the high-puritysample, this has been considered to be an indication ofa higher abundance of the protein in that sample. Atotal of 135 proteins were identified in the high-puritysample and 209 in the low-purity sample, with anoverlap of 93 between both fractions (SupplementalTable S4). DIGE and the MS-derived quantitative data(emPAI and MOWSE difference measures) form animportant part of our final assessment of the peroxi-somal proteome (see below). Sequence information forproteins identified by only a single but significantpeptide is given in Supplemental Data Set S2.
Peroxisomal Membrane Protein Profiling
Three suborganellar fractions were prepared fromwhole peroxisomes by repeated freeze/thaw cycles,followed by centrifugation in order to separate solubleand membrane proteins. Isolation of integral mem-brane proteins was achieved by sodium bicarbonatestripping of an aliquot of the membrane proteins. Atotal of 40 mg of protein from the soluble and completemembrane fractions was separated by 1D SDS-PAGE,whereas only 5 mg could be loaded from the integralmembrane fraction due to the limited amount ofmaterial available. The gel reveals distinct differencesin the protein banding patterns between the peroxi-somal subfractions (Fig. 3). Based on the proteinquantitation (data not shown), we estimate that theintegral proteins account for about 5% of the wholemembrane fraction protein content. Spots from thebands indicated in Figure 3 were excised, digested,and analyzed by liquid chromatography (LC)-MS/MS. From 65 gel bands, a total of 94 unique proteinswere identified using this approach (SupplementalTable S3A). Double SDS-PAGE (dSDS-PAGE) separa-tion of 50 mg of the membrane fraction and 5 mg of theintegral membrane fraction was also performed (Sup-
Figure 2. DIGE 2D IEF/SDS-PAGE of peroxisomal (labeled with Cy5;shown in green) and mitochondrial (labeled with Cy3; shown in red)protein composition. Top, Gel images as derived from the Typhoon Trio(GE Healthcare) fluorescence scanner, analyzed with the DeCydersoftware package (GE Healthcare). Bottom, Fluorescent images wereelectronically overlaid using ImageQuant TL software (GE Healthcare).Yellow spots represent proteins of similar abundance in both samples,with green spots showing an increased abundance in the peroxisomalfraction and red spots indicating a higher abundance in the mitochon-drial fraction. Protein spots from identical preparative gels correspond-ing to these fluorescent proteins were excised, digested with trypsin,and unambiguously identified by MS (Supplemental Fig. S1). Thefluorescent ratios of each identified protein spot are shown in Supple-mental Table S2 and used in Table I.
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1813 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
plemental Fig. S2). The dSDS gel system has an in-creased resolution compared with conventional SDS-PAGE because proteins are separated not only by sizebut also, to a limited extent, by their hydrophobicity(Meyer et al., 2008). Therefore, it is well suited for theseparation of membrane proteins. A set of 49 spotsfrom both gels were cut out, and the identities of 68proteins were derived from tandem MS analysis (Sup-plemental Table S3B). Unfortunately, a DIGE approachwas not feasible for quantification of the membraneproteins in these two approaches. This is due to thelimited resolution inherent to SDS-PAGE gels com-pared with IEF/SDS-PAGE, which often leads to theidentification of more than one protein in a band.Therefore, even if the quantitative analysis of a 1DDIGE gel could indicate a difference in protein abun-dance, it would be impossible to determine the proteinresponsible for the change. To complement the 1D and2D SDS gels, a carbonate-stripped membrane sampleof the high-purity peroxisomal fraction was also ana-lyzed by IDDA-MS/MS for the detection of furtherintegral membrane proteins. This analysis resulted inthe identification of 89 proteins (Supplemental TableS5). We also used a phosphopeptide enrichment strat-egy from organelle peptide preparations to identifyphosphopeptides from peroxisome proteins; the onesignificant match was a Ser phosphopeptide for one ofthe membrane proteins, PMP38 (At2g39970), at Ser-155 (Supplemental Data Set S3).
Confirmation of Localization byGFP-Targeting Experiments
Several hundred unique proteins were identified inthe analyses noted above. While the quantitativeproteomics data provided evidence to validate orinvalidate peroxisomal location in many cases, aselection of proteins for which this analysis was notclear was further verified in vivo on the basis of thetransient expression of fluorescent fusion proteins.For this purpose, proteins carrying a GFP5 insertion10 to 13 amino acids from their C terminus, to allowtargeting by N- and C-terminal sequences and theinfluence of the mature protein sequence on targeting(Tian et al., 2004), were compared with red fluores-cent protein (RFP) fused to the 10 C-terminal aminoacids of the PTS1-containing pumpkin (Cucurbitamaxima) malate synthase. Four proteins (At1g54340,At3g12800, At4g05530, and At4g14430) were used ascontrols. These were found by our MS analysis andhad each previously been documented to be in leafperoxisomes by GFP and MS (Reumann et al., 2007;Table I). As our GFP data were completely convergentwith these localizations, a range of additional pro-teins were selected for analysis (Supplemental DataSet S1). This list does not include all of the proteinswith ambiguous localization data; rather, it merelyfocuses on those for which clarification, in our view,would be most beneficial.
Figure 3. SDS-PAGE separation of40 mg of peroxisomal membraneprotein (A), 5 mg of integral mem-brane protein (B), and 40 mg ofsoluble protein (C) fractions. Mo-lecular masses in kD are shown atleft of each lane, and band numbersextracted for in-gel digestion andprotein identification are shown atright of each lane. Proteins identi-fied are shown in Supplemental Ta-ble S3 and summarized in Table I.
Eubel et al.
1814 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
For four of these proteins (At5g11520, At1g65520,At1g49670, and At5g27520), the GFP localization con-firmed peroxisomal location, and no previous GFP datahave been reported, to our knowledge. Three of themhave recognizable PTS1 and PTS2 sequences, whileAt5g27520 lacks a recognizable PTS and appears to bea six-transmembrane domain carrier family protein(Fig. 4A). Accordingly, the fluorescence of the At5g27520-GFP construct appears ring like, surrounding the matrix-targeted RFP-SRL.
Five targets had been previously reported to localizeGFP to other cell compartments (At3g02360, cytosol[Reumann et al., 2007]; At5g42020, ER [Kim et al., 2001];At4g29130, mitochondria [Damari-Weissler et al.,2007]; At5g58070, vacuole [Jaquinod et al., 2007]; andAt4g16210, unclear [Cutler et al., 2000]). However,these earlier reports had used terminal GFP fusions,which frequently localize differentially depending onthe terminus to which GFP is fused (Simpson et al.,2001). For example, At3g02360 is localized solely to thecytosol when the whole protein is fused to the C ter-minus of enhanced yellow fluorescent protein (Reumannet al., 2007) but is found to be present in peroxisomes byinternal GFP fusion of the whole protein (Fig. 4B). Theother reported locations were confirmed, except forAt5g58070, which did not allow for a confident positiveassignment of location based on the internal fusion(data not shown).
A set of 14 other proteins analyzed by internal GFPfusion to supplement quantitative proteomic data(At1g02930, At1g07920, At1g44575, At1g44820,At2g16060, At1g77120, At2g17265, At3g52190,
At4g00390, At4g29130, At5g19760, At5g43190,At5g43940, and At5g46710) could not be assigned toperoxisomes by GFP and are most likely mitochon-drial, cytosolic, plastidic, nuclear, or vacuolar contam-inants (Supplemental Data Set S1).
Identification of ER Proteins
Within our list of identified proteins, 14 appeared to bemajor ER proteins (Table II). These proteins were con-sidered to be of special interest, as connections betweenthe ER and the peroxisomes in relation to peroxisomalprotein import have been reported previously (Elgersmaet al., 1997; Flynn et al., 2005; Karnik and Trelease, 2005).One of the proteins detected was calreticulin. The dis-tribution of calreticulin across the 800-V separationsused for peroxisomal purification reveals enrichment ofthe antibody signal in the same lanes as the peroxisomes(Fig. 5A), indicating either the comigration of ER com-ponents with peroxisomes during FFE or the presence ofcalreticulin in these organelles. Analysis of the DIGE 2DIEF/SDS-PAGE data revealed that many of the ERproteins reported in Table II, including three isoformsof calreticulin, were enriched in the peroxisome samplerelative to the mitochondrial sample (Fig. 5B), consis-tent with the enrichment in peroxisomal fractions seenwith the calreticulin antibodies (Fig. 5A).
Eleven of the 14 proteins suspected to be ER proteins(At1g08450, At1g09210, At1g21750, At1g56340,At2g47470, At4g15955, At4g24190, At5g28540,At5g42020, At5g60640, and At5g61790) were analyzedby internal GFP fusions. Besides their obvious ERlocation, none of them could be positively localized toperoxisomes (Table II; Supplemental Data Set S1).However, it became apparent from the analysis of thefluorescent images that some sort of interaction be-tween the ER and the peroxisomes might exist. Fre-quently, peroxisomes appear to be heavily embeddedin the ER, with the fluorescence intensity peaking at theER-peroxisome border (e.g. calreticulin, BIP, and pro-tein disulfide isomerase; Fig. 5C), but so far it is unclearwhether this is a result of a higher ER density in thisarea or a higher concentration of the fusion protein.
Assembling the Data to Define a Peroxisomal Proteome
To ensure maximum depth in protein analysis, mul-tiple strategies employing gel and nongel separation ofproteins and peptides were used in our analysis.Quantitative or semiquantitative measurement of theabundance of proteins was obtained using DIGE fluo-rescence measurements from highly purified peroxi-some versus purified mitochondrial samples and theratio of MOWSE scores, or emPAI (Ishihama et al.,2005), in the nongel LC-MS/MS experiments of high-purity peroxisomal versus low-purity peroxisomalsamples. These provided data to discriminate peroxi-somal proteins from potential contaminants. Alto-gether, 250 unique proteins were identified in thecourse of this study. All of them were assessed by a
Figure 4. Fluorescence images of subcellular localization of selectedproteins found in peroxisome preparations by transient expression ofinternal GFP fusions. A, Localization of membrane carrier proteinAt5g27520 in an Arabidopsis cultured cell. B, Localization of6-phosphogluconate dehydrogenase (At3g02360) in an Arabidopsiscultured cell. GFP images of the indicated proteins are overlaid with aRFP peroxisome marker as outlined in ‘‘Materials and Methods.’’
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1815 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
series of factors to define the final proteome set. Theseinclude the quantitative data generated in this study,our and other GFP data, as well as previous reports ofsubcellular localization from other proteome studies.Additionally, all proteins identified from gels and theliquid phase analysis were examined for the presenceor absence of known PTS1 and PTS2 sequences accord-ing to the AraPerox database (Reumann et al., 2004). Inlight of these data, we have classified 89 of theseproteins as of peroxisomal origin (Table I) and 14 asER proteins that copurify with peroxisomes (Table II).
Subfractionation of the organelles to soluble, peripheralmembrane and integral membrane fractions providedsuborganelle location information.
Hydrophobic proteins are notoriously underrep-resented in proteome studies using 2D IEF-SDSapproaches. To determine our success rate in the iden-tification of hydrophobic proteins, we used the ARA-MEMNON database (Schwacke et al., 2003; release 5.0)to predict transmembrane domains in the set of 89peroxisomal proteins; 16 proteins were predicted tohave one or two transmembrane domains, two proteinsmost probably had three to four transmembrane do-mains, and four proteins had more than four trans-membrane domains. The latter group (At2g39970,At4g04470, At4g39850, and At5g27520) mainly con-sisted of the proteins listed as transporters or integralmembrane proteins in Table I.
Developing a Model of Peroxisome Metabolism
In order to test if our proteins form an integrated setof functions and to predict if many peroxisomal en-zymes are missing from our list, we created a meta-bolic network using data from the AraCyc databaseand visualized it with the Cytoscape software package(Version 2.6.0). This was based on the Enzyme Com-mission (EC) number of all identified proteins asannotated in AraCyc, which links Arabidopsis Ge-nome Initiative (AGI) numbers with EC reactions. Atotal of 44 of the proteins in our peroxisome set fromTable I were assigned EC numbers, making a nonre-dundant set of 28 enzyme nodes and 64 metabolites.Focusing on the metabolites that represent the sub-strates, reactants, and products of the network enablesus to see points of contact between the differentfunctional categories of proteins and also to predictthe need for transporters in the peroxisomal mem-brane by analysis of the metabolite end points of thenetwork. In the network, colored nodes (roundedsquares) represent enzymes in different functionalcategories, metabolites are shown as small gray circles,while the reaction is shown as connecting lines be-tween the enzymes and metabolite nodes (Fig. 6).
The result is a structure derived from our proteomediscovery strategy that appears as a well-connectedsingle metabolic entity, showing that the differentmetabolic pathways within this peroxisome proteomeare interlinked and that there are no large gaps createdby lack of identification of critical enzymes in a se-quence. About one-third of the metabolites do notrepresent starting or end points of a pathway but areconsidered to be intermediates of peroxisome metab-olism by this network. The terminal metabolite nodesof the network are potential substrates to be trans-ported in or out across the peroxisomal membrane viatransporters or pores. Many of these metabolites havebeen reported to be transported, to diffuse freelyacross peroxisomal membranes, or to be incapable ofcrossing the peroxisomal membrane, and these arehighlighted in Figure 6.
Figure 5. Enrichment of ER proteins in peroxisome fractions in vitro,and fluorescence-based evidence for association of ER and peroxi-somes in planta. A, Immunoblots of every second of the central 30fractions collected after FFE separation at 800 V showing distribution ofmarker proteins for peroxisomes (3-ketoacyl-CoA thiolase; KAT2) andER (calreticulin). B, Quantification of a range of ER proteins inperoxisome fractions by 2D DIGE analysis (Supplemental Table S2).Each protein’s AGI accession number and description are shown alongwith the average ratio of the quantitation between peroxisomal andmitochondrial samples (n 5 3, P , 0.05). C, Fluorescence images ofsubcellular localization of selected ER proteins found in peroxisomepreparations by transient expression of internal GFP fusions: CRT3(At1g08450), BiP-1 (At5g28540), and DIL2 (At2g47070). GFP imagesof the indicated proteins are overlaid with a RFP peroxisome marker asoutlined in ‘‘Materials and Methods.’’
Eubel et al.
1816 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
The network shows key features of established per-oxisome biochemistry. It shows citrate synthase andmalate dehydrogenase as key points of contact betweenb-oxidation and organic acid/amino acid metabolismthrough CoA/acetyl-CoA and NAD/NADH pools. Italso shows that antioxidant defense enzymes operate todetoxify hydrogen peroxide produced by a series ofoxidases. The metabolites oxygen, hydrogen peroxide,water, and CoA are the most heavily connected nodes,but removal of these ‘‘currency metabolites’’ (Huss and
Holme, 2007) does not break the highly interconnectedstructure of this metabolic network.
DISCUSSION
Of the 89 peroxisomal proteins identified here, 54have been identified previously by the peroxisomeproteome studies of other tissue types by Fukao et al.(2002, 2003) and Reumann et al. (2007). These proteins
Table II. Nonredundant list of probable ER proteins found in peroxisomal samples but removed on the basis of our GFP data (SupplementalData Set S1) and previous proteomics reports on ER samples
The identities of proteins were determined by MS/MS; the predicted molecular mass (MW) of the match is shown along with the best MOWSE score(P , 0.05 when score . 37), the number of peptides and sequence coverage (Seq. Cov.), and the number of times this identification was made (No.Exp.) across the series of experiments performed. Localization of proteins within peroxisomes can be deduced from subfractionation of organelles(Mem 5 membrane fraction, Sol 5 soluble fraction, Whole 5 whole organelles). Quantitative assessments presented in detail in Supplemental TablesS2, S4, and S5 are presented in the DIGE column (P 5 higher in peroxisome sample, M 5 higher in mitochondria sample, P/M 5 both localizationspossible) and the MS High versus Low column (based on the MOWSE score in high-purity compared with low-purity sample, with the sample with thegreater score displayed in the column). New GFP localization data are shown in Our GFP (presented as images in Supplemental Data Set S1).Literature reports of these proteins in subcellular locations based on GFP data (Cyt, cytosol; Mito, mitochondria; Nuc, nucleus; P-Mem, plasmamembrane; Plast, plastid; Vac, vacuole), MS data, or in protein name annotations are shown in the last three columns; these data were sourced fromthe SUBA database (www.suba.bcs.uwa.edu.au).
AGI No. Description MW Score PeptidesSeq.
Cov.
No.
Exp.
1D
SDS
1D
SDS
Sol
1D
SDS
Mem
dSDS2D
IEF/SDS
LC-MS
Whole
LC-MS
MemDIGE
MS High
versus
Low
Our
GFPGFP MS Annotation
At5g60640 Protein
disulfide
isomerase
(PDIL1-4)
66,316 47 1 3 1 – – – – – x – – Low ER – ER, Mito,
Plast, Vac
–
At5g61790 Calnexin 1
(CNX1)
60,448 151 7 11 2 – – – x – x – – High ER – ER, Mito,
Vac
ER
At5g27330 Unknown
protein
71,967 37 1 1 1 – – – – – – x – – – – ER –
At5g28540 Luminal
binding
protein
(BIP1)
73,600 891 31 39 3 – x – – x x – P Low ER – Nuc, Plast,
Vac
ER
At5g42020 Luminal
binding
protein
(BIP2)
73,500 396 14 16 2 – – – – x x – P Low ER ER ER, Nuc,
Vac
ER
At4g15955 Epoxide
hydrolase-
related
(ATsEH)
19,960 133 5 24 1 – – – – x – – P/M – ER – – –
At4g24190 Heat shock
protein
90 (SHD)
94,146 677 15 20 4 x – x – x x – P Low ER – ER, Nuc,
Plast, Mito
–
At2g47470 Protein
disulfide
isomerase
(PDIL2-1)
39,472 723 16 40 2 – – – – x x – – High ER – ER ER
At1g08450 Calreticulin
3 (CRT3)
49,813 134 6 14 1 – – – – x – – P – ER – P-Mem ER
At1g09210 Calreticulin
2 (CRT2)
48,127 318 14 23 1 – – – – x – – P – ER – ER, Mito,
P-Mem
ER
At1g21750 Protein
disulfide
isomerase
(PDIL1-1)
55,567 208 6 13 1 – – – – x – – P – ER – ER, Plast ER
At1g56340 Calreticulin
(CRT1))
48,497 199 7 20 3 – – – x x x – P Low ER – ER, Plast,
Mito
ER
At1g77510 Protein
disulfide
isomerase
(PDIL1-2)
56,329 313 12 25 1 – – – – x – – P – – – ER, Plast ER
At2g29960 Peptidyl-
prolyl
cis-trans
isomerase
(CYP5)
21,520 69 5 17 1 – – – – – x – – High – ER ER Cyt
Eubel et al.
1820 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
participate in all of the classical peroxisomal pathways,such as fatty acid and JA b-oxidation, reactive oxygenspecies detoxification, carbon and nitrogen metabo-lism, and organelle biogenesis (Table I). Interestingly,all of the proteins involved in the peroxisomal steps ofthe photorespiratory pathway have been identifiedhere in cell cultures, even though they are not greenand were grown in the dark for 7 d prior to peroxisomepreparation. A total of 35 proteins are reported here inperoxisomes, to our knowledge, for the first time (TableI). These include a series of isoforms of proteins par-ticipating in well-characterized peroxisomal pathwaysthat are predicted to possess different substrate spec-ificities. Interestingly, these isoforms are often morehydrophobic than their counterparts previously char-acterized by MS, based on comparison of their pre-dicted number of transmembrane domains (Table I).Examples are a member of the 4-coumarate-CoA ligase(4CL)-like proteins (At1g20480), PEX11c, and PEX11e,
indicating that the broadness of the experimental ap-proach helps to identify such proteins. This is alsosupported by the identification of several proteins withunknown function that are reported here, to our knowl-edge, for the first time and that include a number ofputative membrane carrier proteins with four or moretransmembrane domains.
In the following paragraphs, we discuss differentclasses of this newly identified set of 35 proteins; wealso discuss classes of reported peroxisome proteinsnot found here, the close association of peroxisomeswith the ER, and the challenge of defining the widerperoxisome proteome in plants.
Acyl-Activating Enzymes and Acyl-CoA Oxidases
Our data reveal a variety of acyl-activating enzymes(AAEs) in cell culture peroxisomes that provide thepoint of entry for many substrates into the b-oxidation
Figure 6. Visualization of peroxisome metabolic functions based on the proteins identified in this study. Colored nodes (roundedsquares) represent enzymes in different functional categories as shown in the key, metabolites are shown as small gray circles, andreactions are shown as connecting lines between the enzyme and metabolite nodes. The transportability of metabolites across theperoxisome membrane is shown as indicated in the key based on published reports covering plant and nonplant species.
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1821 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
pathway. These proteins are known to possess a certaindegree of substrate specificity. In addition to the two4CL family proteins identified by Reumann et al. (2007;At1g20510 and At4g05160), an additional member ofthis protein family has been identified (At1g20480),which may not be expressed in leaves (Koo et al., 2006).In our study, this protein was identified by five differentapproaches (Table I), strongly suggesting a significantabundance of this protein. At1g20510 and At4g05160are expected to participate in JA biosynthesis, asAt1g20510 preferentially activates OPC 8:0 (Koo et al.,2006) while At4g05160 prefers OPC 6:0. They are alsoreported to accept medium- to long-chain fatty acids assubstrates. These results have been confirmed in acomparison of substrate specificity of 4CL-like proteins(Kienow et al., 2008). None of the tested substratesseemed to suit At1g20480, whose function in the con-text of substrate activation for b-oxidation still remainsunclear. Other new AAEs include At5g16340, whichshares 92% sequence identity with the previouslyfound At5g16370, AAE17 (At5g23050), At1g20560,and LACS6 (At3g05970), a long-chain fatty acid CoAligase (Shockey et al., 2002) targeted to peroxisomes(Fulda et al., 2002). For some of these enzymes, nosubstrate has been identified and their predicted func-tions have only been deduced by amino acid sequencecomparisons. Many of these enzymes with knownfunctions have partially overlapping substrate speci-ficities, according to the literature (Kienow et al., 2008).
Acyl-CoA oxidases (ACXs) perform the next step inb-oxidation after acyl activation, and four of the sixknown isoforms of this enzyme in Arabidopsis havebeen identified in this study (At4g16760, ACX1;At1g06290, ACX3; At3g51840, ACX4; and At2g35690,ACX5). Only one of these was found by MS in pub-lished reports (Table I). Although these enzymes partlyoverlap in their substrate specificity, they appear toprefer different types of substrates, which they feedinto the b-oxidation process. ACX1 preferentially usesmedium- to long-chain fatty acids and seems to be themost important ACX for JA synthesis (Schilmiller et al.,2007). ACX3 and ACX4 prefer medium- or short-chainFAs. All three are reported to be involved in theb-oxidation of IBA (Adham et al., 2005), although todiffering degrees. Using the GENEVESTIGATOR da-tabase (Zimmermann et al., 2004), the expression pro-files of all six ACXs were compared in different plantorgans. ACX6 is expressed most strongly in root tip,ACX1 in sepal, ACX5 in stamen, and ACX2 and ACX3in endosperm but also strongly in seed, imbibed seed,and embryo, which is consistent with their role inmetabolizing lipids in energy metabolism in oilseeds.The transcript for ACX4 is lowly expressed in mosttissues, but we have detected the protein by five dif-ferent approaches, and it has also been found in leafperoxisomes (Reumann et al., 2007). Interestingly,ACX4 is the most divergent protein within the ACXfamily, as it is missing an oxidase domain (Adham et al.,2005). Expression patterns of ACX2, ACX3, and ACX4were found to cluster during development as well as in
their distribution between plant organs in our analysisof GENEVESTIGATOR data (Zimmermann et al.,2004). ACX2 and ACX6 (which may be a pseudogene,as no transcripts have been reported) were not found inthis study. In ACX2 knockout line seeds, long-chainfatty acids accumulate (Pinfield-Wells et al., 2005),indicating together with its expression pattern thatthis protein is mainly involved in the breakdown ofstorage lipids. Why ACX2 appears to be absent or inlow abundance in cell culture, while ACX3 is present, iscurrently unclear.
In contrast to the AAEs and ACXs, the overlap inidentification of all other core enzymes involved inb-oxidation is much higher between studies reportingperoxisome proteomes from different plant tissues.This may indicate that AAEs and ACXs regulate sub-strate flow into the more generic b-oxidation pathway.A comparison of protein abundance of these enzymesfrom different plant organs or different developmen-tal stages will probably shed more light on the differ-ent functions they fulfill in priming substrates forb-oxidation.
Transporters and Integral Membrane Proteins
Four proteins have been identified that fit clearly intothe category of integral membrane proteins: At2g39970(PMP38), At5g27520, At4g39850 (CTS/PXA/PED3),and At4g04470 (PMP22). All four proteins were foundonly in the peroxisomal membrane fractions, are pre-dicted to possess three or more transmembrane do-mains, and lack known PTS1 and PTS2 sequences fortargeting.
PMP38 (At2g39970) contains functional domainssimilar to those found in the mitochondrial carrierfamily (MCF). A homolog of PMP38 in pumpkin hasbeen localized in the peroxisomal membrane by im-munohistochemistry (Fukao et al., 2001); but in Arabi-dopsis, PMP38 has been, most likely erroneously,detected in the vacuole (Jaquinod et al., 2007). PMP38is a clear homolog of the yeast peroxisomal PMP47 thatwas shown many years ago to be a peroxisomal mem-ber of the yeast MCF (McCammon et al., 1990). Thesubstrate for PMP38 in plants is currently not clear, butpossible candidates are ADP/ATP, 2-oxoglutarate/ma-late, phosphate, or tricarboxylates. The yeast equiva-lent is clearly an ADP/ATP transporter but may alsofunction in DpH formation (Lasorsa et al., 2004). Theclosest Arabidopsis homolog to PMP38 is annotated asa plastidic folate transporter (At5g66380) discovered bycomplementation of folate-deficient hamster cells andbacteria. However, knockout of At5g66380 in Arabi-dopsis did not change folate concentration in the plas-tids, suggesting either a different function in plant cellsor the presence of alternative folate transporters inplastids (Bedhomme et al., 2005). The phosphorylationof PMP38 at Ser-155 (Supplemental Data Set S3) mayindicate that it could be regulated by a phosphoryla-tion/dephosphorylation event. Phosphoproteomics ofyeast mitochondria identified phosphorylation of the
Eubel et al.
1822 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
main ATP/ADP transporter (AAC1) at Ser-155/Thr-156 or Ser-157. Alignment of the two proteins revealedbroad but low-level sequence similarity, as both areMCF members, but the phosphorylation sites are innearly identical locations within the hydrophilic loopbetween the third and fourth transmembrane domainsof both proteins (Supplemental Data Set S3).
The product of At5g27520 is an unknown functionmember of the MCF and is most closely related insequence to mitochondrial ADP/ATP transporters andthe peroxisomal ADP/ATP transporter identified inyeast. The designation of this carrier as a peroxisomalcarrier is strengthened by its identification here in threeseparate experiments (Table I) and the absence of itsidentification in our focused work to identify mito-chondrial carriers of this type from mitochondria iso-lated from this same cell culture material (Millar andHeazlewood, 2003). We also tested At5g27520 for per-oxisomal localization using transient expression of aGFP fusion protein, confirming its peroxisomal loca-tion in vivo (Fig. 4A).
At4g39850 (CTS/PXA/PED3) possesses an ATP-binding cassette domain and is considered a fullATP-binding cassette transporter, containing 12 trans-membrane domains and acting as a monomer. There isstill some controversy as to its exact mechanistic func-tion and its corequirement for acyl-CoA synthetasescompared with the more clear-cut evidence for half-ATP transporters in acyl-CoA entry to peroxisomes inyeast (Visser et al., 2007). However, it appears from arange of genetic studies that CTS has a role in the entryof a broad range of fatty acid substrates into plantperoxisomes. CTS knockout mutants are resistant toIBA or 4-(2,4-dichlorophenoxy)butyric acid, indicat-ing that the transport of both compounds is facilitatedinto peroxisomes by this protein (Theodoulou et al.,2005). Seed lipid metabolism also seemed to be af-fected by the knockout of this protein, suggesting apotential role of CTS in the import of fatty acids intoperoxisomes (Footitt et al., 2002). Additionally, CTSmutants are JA-deficient and therefore most likely havea reduced import rate of 12-oxo-phytodienoic acid, aplastid-derived JA precursor molecule. However, CTSdoes not seem to be the only route by which 12-oxo-phytodienoic acid is transferred into the organelle;alternative routes via different, unidentified trans-porters or diffusion pathways might exist (Theodoulouet al., 2005).
At4g04470 (PMP22) is an integral membrane proteinthat possesses similarities to known mammalian andyeast peroxisomal proteins (Tugal et al., 1999). But de-spite a series of reports showing that its loss of functionis detrimental in mouse and yeast (Zwacka et al., 1994;Trott and Morano, 2004), relatively little is known aboutits biological function. The most likely roles proposedare as a nonselective membrane channel and/or as areactive oxygen species forming protein in the perox-isome membrane (Visser et al., 2007).
There has been a report recently of the molecularidentification and cloning of the peroxisomal channel
protein in the grass Bromus inermis (Wu et al., 2005).Such a channel has been predicted in a range of speciesto allow free movement of compounds of less than 300D across the membrane, but it has not previously beenidentified in any species (Visser et al., 2007). TheArabidopsis homolog of this protein is an OEP16-likeprotein, At4g16160, but we found no evidence for thisprotein in the membranes of our peroxisomes. Re-cently, GFP localization of this OEP16 homolog inArabidopsis indicated that it localizes to the plastid(Murcha et al., 2007), further discounting the likeli-hood that it is a peroxisomal membrane channel.
Overall, if these identified transporters represent thebulk of the transport functions on mature peroxisomes,then the metabolite pools and needs for exchange (Fig.6) could be understood by the following scenario.Mature peroxisomes may have a static population ofcofactors such as CoA and pyridine nucleotides such asNAD1 and NADP1, which explains their impermeabil-ity to these reagents. Import of fatty acids would bemediated by CTS, while the two MCF-type carrierscould provide a broad entry for organic and aminoacids, ATP/ADP, and inorganic phosphate, and somesmall molecules may diffuse. This would be consistentwith the known properties/substrates of other MCFsand the CTS and might provide for many of the knowntransport properties of plant peroxisomes.
PEX Proteins
Four of the five PEX11 family proteins that have apredicted role in peroxisome proliferation (Orth et al.,2007) in Arabidopsis have been found in this study(PEX11a, -c, -d, and -e). Three of these peroxins(PEX11c, -d, and -e) form a distinct group within theArabidopsis PEX11 genes, while PEX11a and PEX11beach represent a group on their own. While overex-pression of PEX11a and PEX11b resulted merely inelongation of peroxisomes, the overexpression ofPEX11c, -d, and -e led to the induction of a completedivision process. PEX11c, -d, and -e were also the onlyperoxins able to complement the yeast pex11 knockoutmutant (Orth et al., 2007). Only PEX11d was found byMS in the previous proteome study by Reumann et al.(2007) in leaf peroxisomes. As cell culture is a rapidlydividing tissue, peroxisomal biogenesis and prolifera-tion may be enhanced, which might explain the relativeabundance of the PEX11, PEX7, and PEX14 proteinsobserved here. PEX11b, which has not been detectedby us, was recently shown to be involved in light-dependent regulation of peroxisome proliferation (Desaiand Hu, 2008) during seedling morphogenesis andtherefore is most likely of low abundance in dark-growncell culture. Controversy persists about the precise mo-lecular role of PEX11, because despite genetic evidencefor roles in biogenesis and organelle proliferation (Orthet al., 2007), specific isoforms of these membrane pro-teins have also been implicated in metabolic functionsupstream of b-oxidation, most likely in fatty acid acti-
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1823 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
vation or fatty acid transport in yeast (Hettema andTabak, 2000; van Roermund et al., 2000).
Nucleotide Pools and Redox Shuttling in Peroxisomes
The relative importance of putative organic acid/amino acid shuttles to move reductant in or out of theperoxisome versus an internal balance of reducing andoxidizing reactions remains unclear. This is in large partdue to the lack of information about the composition ofperoxisomes and their reductive and oxidative catalyticactivities. A series of known NAD(H)- or NADP(H)-dependent peroxisomal enzymes were found (Table I)and are clustered in Figure 6 based on their common useof these peroxisomal nucleotide pools. The NAD1 clus-ter surrounds malate dehydrogenase, and the NADP1
cluster surrounds isocitrate dehydrogenase. In additionto the enzymes of known substrate and products shownin Figure 6, a series of reductases were found in Table Ithat likely link to these pools. For example, we founddienoyl-CoA reductase proteins. In yeast, this is knownto be an NADPH-dependent enzyme and has beenconvincingly shown to be supplied with NADPH byisocitrate dehydrogenase through genetic studies (vanRoermund et al., 1998). Other reductases thought to beinvolved in the b-oxidation of unsaturated substrates(Table I) were also found in leaf peroxisomes (Reumannet al., 2007), but their specificities for nucleotides areunknown. We also found three putative quinone reduc-tases grouped in the reducing metabolism section ofTable I, and one of these was also found in leaf perox-isomes (Reumann et al., 2007). All three have PTS1signals (Table I) and may link soluble nucleotide redoxpools with membrane redox pools of quinone or otherunknown reductants. These identifications broaden thepossibilities for redox pool regulation in peroxisomesfrom the classical malate/oxaloacetate and malate/Aspshuttles proposed, based on metabolic models of perox-isome function (Visser et al., 2007).
Proteins Previously Identified in Peroxisomes That Are
Missing in This Study
Interestingly, of the seven proteins associated withprotection against pathogen attack and herbivoresfound in leaf peroxisomes (Reumann et al., 2007),none was found in our analysis. The lack of these pro-teins in the cell culture peroxisomes is striking, as in allother areas of peroxisomal metabolism, including pho-torespiration, a large overlap could be observed. Leaf-specific expression of the corresponding genes isnot evident in GENEVESTIGATOR microarray data(Zimmermann et al., 2004). One of the putative de-fense proteins (OZI1; At4g00860) has previously beenfound in the analysis of the mitochondrial proteome(Heazlewood et al., 2004) and on blue-native/SDS gelsof Arabidopsis mitochondrial membranes. The latterdata suggest an association of OZI1 with the cyto-chrome c oxidase complex (complex IV) of the plantrespiratory chain (Millar et al., 2004); subsequently, the
protein was named COX-X2. This protein does notpossess a PTS, and targeting prediction programs donot suggest a peroxisomal location, so in our view thedata point toward it being a mitochondrial protein. Theother six putatively defense-related proteins claimed asperoxisomal (Reumann et al., 2007) have previouslybeen identified in an analysis of the leaf vacuole pro-teome (Carter et al., 2004), and most are predicted tohave secretory signals or ER signal peptides. Anothernotable absence from the peroxisomes analyzed hereare the glyoxylate cycle enzymes isocitrate lyase andmalate synthase. This clearly shows that these cellculture organelles cannot be considered to be catalyz-ing a glyoxylate cycle.
The vast majority of proteins that apparently arenonperoxisomal in our preparations have a mitochon-drial origin. Given that mitochondria are probably themost abundant organelles present in dark-grown cells,that they migrate very close to the peroxisomes duringFFE, and that they are widely considered the majorcontaminant in yeast and mammalian peroxisomeproteome analyses (Kikuchi et al., 2004; Marelli et al.,2004; Saleem et al., 2006), this is not surprising. Ourongoing investigation of the Arabidopsis mitochon-drial proteome (Millar et al., 2005), facilitated by tech-nical advances in MS and the increase in the purity ofmitochondrial isolations by FFE (Eubel et al., 2007),enables a very precise assignment of mitochondrialproteins in our overall nonredundant protein list. Thesubtraction of mitochondrial proteins is supported inthe vast majority of cases by the quantitative datagenerated in the course of this study, namely throughDIGE and comparative LC-MS/MS. Contaminationsby cellular compartments other than mitochondria aresignificantly harder to confidently exclude. This isespecially true in the case of the ER proteins identifiedhere, such as calreticulin, calnexin, HSP90, SEC12p,and the lumenal chaperones BIP1 and BIP2. While it isbeyond doubt that these proteins are primarily found inthe ER, it is harder to tell whether this group alsorepresents bona fide proteins in peroxisomal prepara-tions or if they are merely contaminating proteins dueto a copurification of ER vesicles with the peroxisomes.Similar dilemmas have been presented a number oftimes in the literature about separation of the ER andperoxisome proteomes in other species (Saleem et al.,2006, and references therein). The resolution to thesequestions may be tightly linked to peroxisomal biogen-esis and the import of peroxisomal proteins, especiallythose inserted into the membrane of these organelles.An involvement of the ER in the import of proteins intoperoxisomes has been proposed for PEX15 in yeast(Elgersma et al., 1997) and for PEX10 and PEX16 inArabidopsis cell suspension cultures (Flynn et al., 2005;Karnik and Trelease, 2005), although PEX10 has alsobeen reported to be targeted to peroxisomes without a
Eubel et al.
1824 Plant Physiol. Vol. 148, 2008 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from
passage through the ER (Sparkes et al., 2005). BIP hasbeen found in the same gradient fraction as APX inpumpkin (Nito et al., 2001). The involvement of the ERwith peroxisomal biogenesis and protein import hasbeen reviewed (Mullen and Trelease, 2006), and ricino-somes emerging from the ER were found to contain BiPand PDI protein (Schmid et al., 2001). Our own GFPlocalization studies did not support the presence of theanalyzed ER proteins inside peroxisomes in vivo (TableII; Fig. 4; Supplemental Data Set S1). The presence ofthese proteins in the peroxisomal samples might beexplained by the close interaction of these two com-partments observed in our GFP images (Fig. 5; Supple-mental Data Set S1) and the potential for isolatingperoxisomes tethered to ER vesicles.
On balance, the data are more in favor of contam-ination of the peroxisomes with a small amount of ERmaterial, either due to comigration in FFE (Fig. 5A) ordue to physical association between ER fragments andperoxisomes (Fig. 5C), but probably not due to thepresence of ER proteins within the peroxisome mem-brane or lumen.
The Complexity of Defining the FullPeroxisomal Proteome
Despite substantial efforts of three separate groupsworking on different Arabidopsis tissue types (Fukaoet al., 2002, 2003; Reumann et al., 2007; Table I), less than20% of the proteins with predicted PTS sequences(Reumann et al., 2004) have been experimentally con-firmed in peroxisomal preparations. In the process,each group has found a range of proteins that lack PTSsignals, suggesting that bioinformatics predictions willbe of limited help. The task of experimentally confirm-ing the complete peroxisomal proteome in Arabidopsismay still be some way off. While 11 proteins (photo-respiratory enzymes, catalases, and malate dehydro-genases) have been found in two or three studies(Table I), significant discrepancies exist between thelists of peroxisomal proteins from our analysis andthose from other Arabidopsis tissues. Reumann et al.(2007) claimed 29 proteins as peroxisomal that were notconfirmed in our study; only four have recognizablePTS sequences (At1g06460, At1g21770, At1g60550, andAt5g48880; Supplemental Table S6). Fukao et al. (2002,2003) claim an entirely separate set of 29 proteins thatwere not confirmed in this study (Supplemental TableS6); only one has a recognizable PTS, malate synthase(At5g03860), and most of the others have currentlyunclear functions in peroxisomes. This study claims 35proteins not found in any of the other studies of leaf orcotyledon peroxisomes (Table I); 21 contain PTS1 orPTS2, and most are involved in fatty acid oxidation. Inaddition, six are transmembrane proteins that wouldnot have a clear PTS. More quantitative data andassessment of targeting are necessary to resolve ifeach of these proteins lacking independent confirma-tion are peroxisomal in a range of plant tissue types.
Further advances in peroxisome purification are re-quired so that 10 to 100 mg of peroxisomes can beisolated from Arabidopsis tissues to allow suborganellefractionation, so effective for improving protein iden-tification in other organelles. However, many proteinshave now been found in two or three independentstudies (Table I). This gives a solid foundation fordetailed analysis of peroxisomal function in plants, as itis likely that many of the major metabolic pathways cannow be reconstructed (Fig. 6).
MATERIALS AND METHODS
Cell Culture Maintenance
An Arabidopsis (Arabidopsis thaliana) suspension cell culture has been
maintained and subcultured as stated elsewhere (Millar et al., 2001). Briefly, 20
mL of a 7-d-old cell culture grown in the light was transferred into 100 mL of
fresh medium. Starting material for the peroxisome preparation was grown in
the dark for 7 d, whereas material used for the maintenance of the culture was
incubated in the light for the same period of time.
Organelle Preparation
The preparation of the organelles was based on Eubel et al. (2007), with some
modifications. Approximately 200 g of cells was used for each preparation.
After enzymatic digestion of the cell wall, the resulting protoplasts were
disrupted by four strokes in a Potter-Elvehjem homogenizer. The homogeni-
zation medium contained bovine serum albumin and EDTA as general protease
inhibitors; in addition, Complete (Roche Applied Science) protease inhibitor
cocktail was used according to the manufacturer’s instructions to inhibit Ser,
Cys, and metalloproteases. The homogenate was centrifuged at 3,000g for 5 min
to remove cell debris. The supernatant was spun at 24,000g for 10 min to
concentrate the organelles. However, in order to avoid pelleting of the organ-
elles, the supernatant of the low-speed spin was layered on top of 5 mL of a
Percoll cushion (60% [v/v] Percoll, 10 mM MOPS-KOH, pH 7.2). The organelle-
containing interphase was taken and diluted 1:1 with wash buffer and loaded
onto discontinuous Percoll density gradients consisting of 5 mL of 50% (v/v)
Percoll and 25 mL of 25% (v/v) Percoll (bottom to top) in wash buffer. After
centrifugation, the mitochondrial/peroxisomal band was found between the
50% (v/v) and the 25% (v/v) Percoll phases. This band was extracted, and
the Percoll was removed by three repeated washes in FFE separation buffer. The
first two washes were performed using a 60% (v/v) Suc cushion (60% [v/v] Suc,
10 mM MOPS-KOH, pH 7.2) at the bottom of the tubes, again to prevent
pelleting of the organelles. The third wash was performed without any cushion,
and the organelles were pelleted. After resuspension in a small volume of FFE
separation buffer (2 mL), the organelle suspension was slowly homogenized in
a Potter-Elvehjem homogenizer in preparation for FFE.
FFE buffer composition and conditions were similar to those described
previously (Eubel et al., 2007) with the exception of the use of either 800 Vor 600
V in experiments as indicated. Using the higher voltage, mitochondria and
plastids were deflected to such an extent that the plastids were running into the
anode stabilization medium, whereas the mitochondria just stayed within the
borders of the separation medium. Although a certain amount of mixing
between those two organelles occurred, the distance between the mitochondria
and the peroxisomes increased, which led to a higher level of purity of the
peroxisomal fraction. After visual inspection of the 96-well plate, mitochon-
drial fractions (contaminated with plastids) and peroxisomal fractions were
pooled separately. The mitochondria were then subjected to a second round of
FFE at 600 Vin order to obtain purer material for the subsequent measurements.
Oxygen Electrode Measurements
Catalase activity was measured using a Clark-type oxygen electrode
(Hansatech) as outlined previously (Eubel et al., 2007). Succinate dehydro-
genase measurements were performed using a Clark-type oxygen electrode in
the presence of 10 mM succinate, 500 nmol of ADP, and 100 nmol of ATP. Fifty
micrograms of protein was used for each assay.
Arabidopsis Peroxisome Proteome
Plant Physiol. Vol. 148, 2008 1825 www.plantphysiol.orgon July 28, 2018 - Published by Downloaded from