2.2.1. Matrix protein import

2.2.2. Membrane protein import.

Group I and Group II PMPs show distinct targeting sequences, known as membrane PTSs (mPTSs). Group II PMPs employ mPTS1, a stretch of positively charged residues in proximity to a transmembrane domain (TMD), whereas Group I PMPs use mPTS2, which bears an additional ER sorting signal that either overlaps or is adjacent to the mPTS1 (Mullen and Trelease, 2006; Van Ael and Fransen, 2006). In yeasts and humans, PEX19, a versatile peroxin with multiple binding domains that bind to a number of PMPs, acts as a cytosolic receptor for PMPs. It is believed that PEX19 binds to nascent PMPs, stabilizes them, and transports them to the peroxisome membrane, where it anchors by binding to its membrane-bound receptor, PEX3 (Fang et al., 2004; Shibata et al., 2004; Fujiki et al., 2006). It is unknown what triggers the subsequent insertion of PMPs into the peroxisome membrane. Additionally, PEX19-independent pathways for PMP targeting are present in mammals and hinted at for yeasts (Fransen et al., 2004; Otzen et al., 2004). Arabidopsis has two PEX19 isoforms, AtPEX19-1 and AtPEX19-2, with AtPEX19-1 existing as dimeric species in vivo and binding to AtPEX10 in pulldown assays (Hadden et al., 2006). RNAi lines of AtPEX19-1 or AtPEX19-2 contain enlarged peroxisomes, suggesting non-redundant functions of the two proteins. Knocking down either of the two AtPEX3 isoforms had no effect, whereas double knockdown AtPEX3 plants display peroxisome elongation. The phenotypes of these RNAi lines led to the conclusion that AtPEX3 and AtPEX19 influence peroxisome morphology but do not contribute to peroxisome protein import (Nito et al., 2007). Mechanisms by which plant peroxisomes recruit PMPs remain to be elucidated.

Fig. 2 depicts current models for PTS1 and PTS2 import pathways in plants. Supplemental  Table I (sd1_1.xls) summarizes the characterized Arabidopsis PEX proteins and their mutant phenotypes. Although many facets of peroxisome protein import, especially the components involved and their impacts on plant development and physiology, have been elucidated, our knowledge of this process is still hazy. Are there other unidentified peroxins in plants? Do functional homologs of PEX17 or PEX8 exist? What is the significance of having two isoforms for PEX19 and PEX3 and have they acquired plant-specific functions? Does AtPEX4 also function in receptor recycling? Are any of the other Arabidopsis peroxins involved in PexAD? These are some of the questions that merit further investigations.

2.2.3. Post-import process: heat shock proteins (Hsps) and endoproteinases

Most subcellular organelles host a battery of Hsps and molecular chaperones in their lumen to assist in the folding of newly imported proteins or refolding of denatured polypeptides (reviewed in (Leidhold and Voos, 2007). Peroxisomes are distinguished from other organelles by their capacity to import fully folded proteins, as indicated by the phenomena of piggybacking (Lee et al., 1997). This import of folded substrates seems to preclude the need for chaperone activity within the organelle. Although in vitro import assays have shown that cytosolic Hsp70 and Hsp90 chaperone activities are associated with and enhance peroxisomal protein import (Crookes and Olsen, 1998), little is known about protein (re-)folding within peroxisomes. Considering the high production rate of reactive oxygen species in peroxisomes, it is likely that peroxisomes have evolved a mechanism to safeguard matrix proteins from denaturation.

Although mammals and yeast do not seem to have any chaperones in their peroxisomes, two Arabidopsis small Hsps in the peroxisome matrix complemented the yeast mutant, hinting at the functional conservation of this type of Hsps as chaperones across kingdoms (Ma et al., 2006). Moreover, one of the sHsps is heat and oxidative stress inducible, further cementing the notion that peroxisomes can implement stress responses (Ma et al., 2006). Analysis of loss-of-function mutants of the sHsps will be instrumental in defining the physiological function of these proteins in greater details. Small Hsps lack ATP-hydrolyzing activity and need other Hsps for protein renaturation. Various Hsp70 isoforms and DnaJ homologues have been reported to be present in peroxisomes in other plant species (Preisig-Muller et al., 1994; Corpas and Trelease, 1997; Wimmer et al., 1997; Diefenbach and Kindl, 2000); their orthologs could conceivably be present in Arabidopsis. These proteins, in conjunction with the identified sHsps, may help to alleviate stress-induced protein aggregation and denaturation in the peroxisome matrix.

One prominent feature for PTS2 protein import is the cleavage of the N-terminal signal peptide after import. In mitochondria and chloroplasts, cleavage is mandatory for folding or further targeting to sub-structures (Adam and Clarke, 2002; Gakh et al., 2002). However, the activity of the peroxisomal protein malate dehydrogenase (MDH) was shown to be unaffected in the absence of PTS2 cleavage (Helm et al., 2007). PTS2 protein processing seems to be restricted to higher eukaryotes, as PTS2-containing proteins in yeasts do not undergo any cleavage. Arabidopsis encodes a significant number of PTS2 proteins, and in silico and proteomic analyses also indicate the presence of several proteases in Arabidopsis peroxisomes (Reumann et al., 2004; Reumann et al., 2007; Reumann et al., 2009). Recently, an Arabidopsis DEG protease (DEG15) was characterized as a serine endopeptidase and implicated in the processing of PTS2 peroxisomal proteins such as MDH, KAT2 and long-chain acyl-CoA synthetase isoform 6 (LACS6; Schuhmann et al., 2008). Studies in pea have found up to seven endoprotease activities associated with senescent peroxisomes, and these activities were speculated to be involved in turning over peroxisomal proteins during early senescence, in addition to their role in degrading proteins from multiple subcellular compartments during advanced stages of senescence (Distefano et al., 1997; Distefano et al., 1999).

An unanticipated role has been determined for the mammalian ortholog of AtDEGI5, Tysnd1, which not only cleaves PTS2 off of PTS2 proteins but also cleaves three PTS1 proteins involved in the β-oxidation pathway internally (Kurochkin et al., 2007). It was speculated that this processing facilitates the arrangement of the enzymes into multienzyme complexes, resulting in effective metabolic channeling and consequently higher enzymatic efficiencies (Wouters et al., 1998). Plant peroxisomes accommodate enzymes for various metabolic processes and are also purported to have similar multienzyme complex arrangements in their matrix (Heupel et al., 1991; Heupel and Heldt, 1994). Several prospective peroxisome proteases were identified in an Arabidopsis leaf proteomic experiment (Reumann et al., 2009). Validation and characterization of these putative peroxisomal proteases, as well as identification of processing-competent PTS1 proteins, may shed light on the precise nature of peroxisomal endoproteolytic activity and its possible contribution to efficient assembly of multienzyme complexes or other processes in the Arabidopsis peroxisome.

Figure 2.

Speculative model for matrix protein import.

(A) PTS1 import. PEX5 recognizes and binds PTS1-containing proteins in the cytosol. The receptor-PTS1 protein complex then traffics to the peroxisome where it associates with the docking complex on the peroxisome membrane. The docking complex in Arabidopsis probably comprises PEX14 and PEX13 and is believed to tether the receptor-protein complex to the peroxisome membrane through the interaction of PEX5 with PEX14. Subsequently, the PTS1 protein is dissociated from PEX5 and released into the peroxisomal matrix by an unknown mechanism. The export of the receptor to the cytosol might be facilitated by its possible ubiquitination (not shown) by the PEX22-anchored PEX4, a putative ubiquitin-conjugating enzyme, followed by ATP-driven dislocation mediated by the peroxisomal AAA ATPases PEX1 and PEX6. The RING complex, which is composed of the PEX2, PEX10 and PEX12 RING peroxins, plausibly plays a role in the import and export processes, although the exact function(s) of this complex are not well understood.

(B) PTS2 import. PEX7 recognizes and binds PTS2-containing proteins in the cytosol. The receptor-PTS2 protein complex binds coordinately with PEX5 and is ferried to the peroxisome docking complex. The subsequent steps of import are assumed to be similar to PTS1 import. The events facilitating the release of PEX7 and PTS2 protein are not well known.

2.3.1. Peroxisome division

2.3.1.2. Early steps: the PEROXIN11 (PEX11) protein family.

Seminal work carried out with yeast peroxisome mutants has provided critical insights into the molecular players involved in peroxisome division. Factors involved in peroxisome division have been classified into four families of peroxisome membrane proteins, namely, Pex11p, Pex25p/Pex27p, Pex28p/Pex29p, and Pex30p/Pex31p/Pex32p (Rottensteiner et al., 2003; Tam et al., 2003; Vizeacoumar et al., 2003; Vizeacoumar et al., 2004). Amongst these proteins, Pex11p was the first to be identified and is believed to mediate the initial step of peroxisome division, i.e., peroxisome elongation/tubulation (Fagarasanu et al., 2007). The S. cerevisiae Pex11p protein was shown to be a peripheral membrane protein located in the inner surface of the peroxisome membrane. The pex11 mutant embodies prototypical division mutant phenotypes, with loss-of-function mutation resulting in peroxisomes enlarged in size and reduced in number and overexpression conferring an increase in peroxisome abundance (Erdmann and Blobel, 1995; Marshall et al., 1995). Pex11p also comprises the only PMP family exclusively involved in peroxisome division that has known counterparts in mammals and plants. Phylogenetic analysis suggested that PEX11 genes have a single origin and evolved and diversified independently in plant, fungal, and animal lineages (Orth et al., 2007). PEX11 in mammals have three isoforms (PEX11α, -β, and -γ), each exerting varying contributions to peroxisome division (Schrader et al., 1998; Li et al., 2002b; Li et al., 2002a). In S. cerevisiae, Pex11p shares weak sequence similarity with two homologous proteins, Pex25p and Pex27p; partial redundancy was found among these three proteins (Rottensteiner et al., 2003).

The Arabidopsis PEX11 family has further expanded into five isoforms grouped into three subfamilies: AtPEX11a, AtPEX11b, and AtPEX11c-e, all of which are integral membrane proteins of the peroxisome (Lingard and Trelease, 2006; Orth et al., 2007). Protein expression and topology studies using Arabidopsis suspension cell cultures confirmed that all AtPEX11s are Group II PMPs that target directly to peroxisomes without trafficking via the ER. Furthermore, AtPEX11b-e, like mammalian PEX11 homologs, have both the N- and C-terminal tails exposed to the cytosol, whereas AtPEX11a has a distinctive C terminus that protrudes into the peroxisomal matrix (Lingard and Trelease, 2006). Overexpression of individual AtPEX11 isoforms variedly promoted peroxisome elongation and/or increase in the number of peroxisomes (Lingard and Trelease, 2006; Orth et al., 2007). Decreasing the expression of individual PEX11 or a subfamily of PEX11 genes using RNAi led to reduction in the total number of peroxisomes (Orth et al., 2007) or slightly enlarged peroxisomes (Nito et al., 2007). No physiological or growth phenotypes were observed in the RNAi plants, possibly reflecting functional redundancy among isoforms (Orth et al., 2007). Further, AtPEX11e and -c partially compensated for the lack of Pex11p in the yeast mutant (Orth et al., 2007). Taken together, all these data point toward a conserved and positive role for PEX11 in regulating peroxisome proliferation across kingdoms.

Figure 3.

Model for peroxisome proliferation and replication in Arabidopsis.

The PEX11 family of proteins (PEX11a-e) acts in the initial steps of peroxisome proliferation and replication, resulting in pronounced elongation or expansion of peroxisomes. Proteins overseeing the subsequent membrane constriction are not known. Fission of the constricted peroxisomes is enabled by the scission activities of dynamin-related proteins DRP3A and DRP3B, which are tethered to the peroxisome membrane by FIS1A and FIS1B. The divided peroxisomes are then transported to various parts of the cell by the indicated myosin proteins (XI-2, XI-I, XI-E, XI-K) along actin cables. The transcription factor HYH has been implicated in the phyA-mediated light regulation of peroxisome elongation via activation of the PEX11b gene. Hydrogen peroxide (H₂O₂), isoproturon, ozone and Jasmonates (JA) are other cues that induce or repress peroxisome proliferation through factors currently unknown, indicated by dotted arrow.

Among the three mammalian PEX11 proteins, PEX11β is essential for survival, leading to embryo lethality when its function is disruped (Li et al., 2002a). In contrast, none of the AtPEX11 family members are vital for survival. Drawing from the observations that null mutants for some Arabidopsis peroxisome proteins are embryo lethal (see 2.2.1), it is possible that peroxisome proliferation is essential to plant survival and hence plants have evolved functionally redundant PEX11 isoforms that perform in specific peroxisome subtypes and respond to diverse environmental cues. Evidence supporting these notions includes specific transcriptional induction of particular isoforms (Orth et al., 2007; Desai and Hu, 2008) and distinct peroxisome morphologies caused by over-expressing different family members (Lingard and Trelease, 2006; Orth et al., 2007). Alternatively, based on the facts that (i) only a subset of the AtPEX11 proteins complemented the yeast mutant, (ii) AtPEX11a carries a distinctive membrane topology, and (iii) a C-terminal dilysine motif is restricted to the PEX11c-e subfamily in Arabidopsis, it is speculated that PEX11 proteins may perform discrete biochemical functions (Hu, 2007).

Although PEX11 has been the best-characterized protein in peroxisome division and is known to be primarily responsible for initiating the process, there are gaps in our knowledge of its precise mode of action. Functions proposed for PEX11 include membrane modification through phospholipid binding, transport of metabolites, and recruitment of downstream effector proteins (Thoms and Erdmann, 2005; Fagarasanu et al., 2007). Some PEX11 proteins, including AtPEX11c-e, contain a C-terminal dilysine motif. This motif was shown in rat to bind to coat protein 1 (COP1) and recruit (ADP)-ribosylation factor (ARF1), suggesting that PEX11 and coatamers coordinately execute peroxisome division by promoting membrane vesiculation (Passreiter et al., 1998; Anton et al., 2000). However, this motif is dispensable for promoting peroxisome division in Arabidopsis cell cultures (Lingard and Trelease, 2006), indicating that species-specific factors may be involved in facilitating the function of PEX11.

2.3.2. Peroxisome movement and distribution

After multiplication, peroxisomes are separated and move to different locations within a cell, and partition into daughter cells in approximately equal numbers upon cell division or budding (Fagarasanu et al., 2007). In animal cells, peroxisomes primarily travel along microtubules, facilitated by microtubule-based motor molecules such as kinesin and dynein proteins (Rapp et al., 1996; Schrader et al., 1996b; Schrader et al., 1996a; Wiemer et al., 1997). In contrast, yeast and plant peroxisomes employ the actin-based cytoskeleton for intracellular trafficking and their movement is powered by actin-associated molecular motors named myosins (Hoepfner et al., 2001; Jedd and Chua, 2002; Mano et al., 2002; Mathur et al., 2002). Co-visualization studies using fluorescent protein-tagged peroxisomes and microfilaments revealed that plant peroxisomes associate with actin microfilaments in the cell, and that the application of the actin inhibitor latrunculin B arrests peroxisome movement (Jedd and Chua, 2002; Mano et al., 2002; Mathur et al., 2002). Individual peroxisome movement was found to be independent and highly dynamic with frequent changes in the state of motility, velocity and direction (Mathur et al., 2002). Velocity of the movement was variously reported as 2–10 µm s^-1 (Jedd and Chua, 2002; Mano et al., 2002), much higher than the known motility rates for mammalian peroxisomes on the microtubules, which were recorded to be 0.75–2 µm s^-1 and 0.2–0.5 µm s^-1 (Rapp et al., 1996; Wiemer et al., 1997).

The myosin motor proteins are classified into 24 classes; however, plant myosins are restricted to class VIII and XI (Reddy and Day, 2001; Jiang and Ramachandran, 2004). Myosins utilize ATP hydrolysis via their N-terminal ATP-binding domains to shuttle cargo molecules on actin tracks. These proteins share a conserved motor/head for actin and ATP binding, a neck region consisting of IQ motifs that bind to calmodulin or calmodulin-like light chains, a coiled-coil domain in the stalk, and a highly variable tail that binds to cargos and determines functional specificity (Foth et al., 2006). Several Arabidopsis class XI myosins, including XI-2, XI-I, XI-K, and XI-E, have been shown to partially associate with peroxisomes and play roles in the movement of peroxisomes (Fig. 3). Subcellular targeting can be accomplished using merely the globular tail domain, indicating that the cargo binding and targeting information is present in the globular tails (Hashimoto et al., 2005; Li and Nebenfuhr, 2007; Reisen and Hanson, 2007; Sparkes et al., 2008). A model was proposed for assembly of myosin XI-cargo complex on the actin filament, wherein coordinated dimerization, organelle targeting and actin-based processive movement by the myosin XI facilitates efficient regulation of the motor activity (Li and Nebenfuhr, 2008).

Studies from Arabidopsis and yeast have shown that multiple isoforms of myosins are normally associated with an organelle to facilitate its movement. Likewise, each myosin can participate in the movement of several types of organelles, because knockout mutants for a single myosin can slow down the velocity of several subcellular compartments and sometimes lead to developmental defects (Hashimoto et al., 2005; Fagarasanu et al., 2007; Li and Nebenfuhr, 2007; Reisen and Hanson, 2007; Peremyslov et al., 2008; Sparkes et al., 2008). How is each myosin recognized by a particular organelle? It turns out that in budding yeast, a Class V myosin named Myo2p, which is associated with at least six types of cargos such as the Golgi vesicles, vacuoles, secretory vesicles, mitochondria, mitotic spindle, and peroxisomes, bind to specific receptors on the respective organelles via distinct domains in the myosin globular tail (Fagarasanu et al., 2007; Li and Nebenfuhr, 2008). Inp2p, an integral membrane protein with a cytoplasmically exposed coiled-coil domain, is the peroxisomespecific receptor for Myo2p (Fagarasanu et al., 2005; Fagarasanu et al., 2006). It remains to be elucidated in plants whether such organelle-specific receptors exist and how exactly peroxisome transport and inheritance is achieved.

Peroxisomes are often found to be intimately associated with chloroplasts and mitochondria. Thus, chloroplast and mitochondrial components or their overall homeostasis might dictate peroxisome distribution in the cell. Peroxisomes in the Arabidopsis conditional mutants of PEX10 were found to have lost their physical contacts with chloroplasts, indicating that PEX10 might be a peroxisomal protein mediating the association of peroxisomes with chloroplasts (Schumann et al., 2007).

2.3.3. Environmental, metabolic, and nuclear regulation of peroxisome proliferation

A wide range of endogenous and exogenous cues are reported to have an impact on peroxisome abundance. Peroxisome proliferation, or induced division, refers specifically to increases in peroxisome abundance or volume in response to environmental and metabolic stimuli (Yan et al., 2005). Oleates and hypolipidemic ligands (e.g., clofibrate) induce peroxisome proliferation in yeast and mammals, respectively. Transcription factors responsible for implementing peroxisome proliferation in both yeasts (Oaf1p, Pip2p, and Adr1p) and mammals (PPARα and RXR) and the responsive promoter elements in the target genes have been identified (Desvergne and Wahli, 1999; Gurvitz and Rottensteiner, 2006). Reports of equivalent proteins in Arabidopsis are yet lacking, although introduction of Xenopus PPARα into tobacco plants was found to induce acyl-CoA oxidase (ACX) gene expression along with a concomitant increase in peroxisome number (Nila et al., 2006). There are some hints that plants synthesize analogs to hypolipidemic drugs, as extracts from bitter gourd, plant isoprenols (farnesol and geranylgeraniol), and soybean isoflavones were found to activate PPARα and -γ (Takahashi et al., 2002; Chao and Huang, 2003; Shay and Banz, 2005). Treatment of pea and Arabidopsis with clofibrate results in an increase in peroxisome number and an induction in the transcription of Arabidopsis PEX1, PEX14, and KAT2 genes (Palma et al., 1991; Castillo et al., 2008). Collectively these results can be interpreted to indicate some degree of conservation between pathways regulating metabolic and nuclear control of peroxisome proliferation in plants and mammals (Hu, 2007).

Various treatments such as ozone, high light, isoproturon, and jasmonate have been observed to cause change in peroxisome number in plants (de Felipe et al., 1988; Ferreira et al., 1989; Ulloa et al., 2002; Oksanen et al., 2003; Castillo et al., 2008). Recent observations also support the hypothesis that metabolic status of the peroxisome might modulate peroxisome morphology and division decisions as well (Titorenko and Rachubinski, 2004; Quo et al., 2007). Studies in plants showed that elevated levels of acyl-CoAs within the peroxisome matrix appear to drive peroxisome enlargement (reviewed in Baker et al., 2006). Another study conducted in Vigna nodules found retarded peroxisome development in the absence of uricase, an enzyme involved in the purine oxidation pathway (Lee et al., 1993). In addition, the size and morphology of plant peroxisomes change between developmental stages (Mano et al., 2002). Lastly, in Arabidopsis, exposure to H₂O₂ and UV light-induced hydroxyl radicals causes peroxisomes to produce dramatic extensions termed peroxules, followed by fission of the elongated peroxisomes (Sinclair et al., 2009). Thus, peroxisomes seem to be sensitive to a multitude of signals and rapidly incorporate and manifest these signals by integrative changes in their abundance and morphology. However, despite the numerous tabulated cues for peroxisome proliferation, the underlying molecular mechanisms are largely unknown in plants.

A recent work has partially dissected the mechanism by which light regulates plant peroxisome abundance (Desai and Hu, 2008). It was observed that light treatment results in an increase in AtPEX11b transcript levels, which parallels the increase in peroxisome abundance during dark-light transition in Arabidopsis cotyledons and hypocotyls. In agreement with this observation, AtPEX11b RNAi lines exhibit decreased peroxisome proliferation under light treatment. Subsequent screening of various light signaling mutants for changes in AtPEX11b expression resulted in the identification of the photoreceptor phyA and a bZIP transcription factor, HYH, to be mediators of the light-induced expression of AtPEX11b. Both phyA and hyh mutants exhibit reduced AtPEX11b expression in the light and contain drastically reduced peroxisome numbers, a phenotype that can be rescued by over-expression of AtPEX11b. The promoter of AtPEX11b contains several light responsive elements and interacts with the HYH protein, suggesting that AtPEX11b is a direct transcriptional target of HYH. The response to light appears preparative in nature, compelled by the need for efficient photosynthesis and photorespiration with resultant peroxisome proliferation during seedling photomorphogenesis. The authors proposed the existence of a novel branch of the phyA-mediated light signaling network, which consists of the downstream effector HYH and the AtPEX11b gene, and suggested that peroxisome proliferation is an outcome of HYH binding and activation of AtPEX11b (Desai and Hu, 2008). This model does not rule out possible regulation of the AtPEX11b gene by other signaling intermediates in the phyA network. Future work should define this circuitry with higher precision and identify additional nuclear factors supervising peroxisome proliferation in plants. It would also be interesting to investigate whether various cues for proliferation are channeled through discrete or overlapping pathways and whether they converge at a probable master regulator of peroxisome proliferation.

Signals for changes in peroxisome abundance can also be emitted from other organelles. In budding yeast, respiratory deficiency in mitochondria results in retrograde regulation of nuclear genes, turning on a suite of genes associated with peroxisome biogenesis and proliferation (Hallstrom and Moye-Rowley, 2000; Epstein et al., 2001; Butow and Avadhani, 2004). Such retrograde signaling pathways have not been identified for plant peroxisomes.

3.7.1. Catalase

Oxidative reactions are at the core of peroxisome metabolism. One common by-product generated by such reactions is hydrogen peroxide (H₂O₂), which is scavenged primarily by catalase in the peroxisome matrix. Catalase is one of the most abundant peroxisomal proteins and forms a tetramer containing four porphyrin heme (iron) groups. It is also one of the very few matrix proteins that carry internal PTS1-like tripeptides located approx. 10 amino acids upstream of the C terminus (Kamigaki et al., 2003). Despite being the key enzyme in H₂O₂ metabolism, catalase has a very high Km, which renders the enzyme an inefficient scavenger at low concentrations. Moreover, catalase is light sensitive, undergoing photoinactivation followed by degradation in high light conditions (Feierabend and Engel, 1986; Tel-Or et al., 1986; Volk and Feierabend, 1989; Feierabend et al., 1992; Hertwig et al., 1992). Initial studies on catalase, which were mostly conducted in tobacco and focused on the association of this enzyme with the photorespiratory cycle, found that elevated catalase levels can reduce photorespiratory carbon losses and increase net photosynthesis under conditions that favor photorespiration (Zelitch, 1989; Zelitch et al., 1991; Zelitch, 1992; Brisson et al., 1998). Depletion of CAT1 activity in tobacco results in necrotic lesions in high light, hypersensitive response to bacterial infection, and increased susceptibility to paraquat, salt and ozone (Willekens et al., 1997; Mittler et al., 1999).

Arabidopsis expresses three peroxisome-targeted catalase isoforms, among which CAT2 has been the best characterized for its paramount role in photorespiration. CAT2 is maximally expressed in leaves, induced by cold and light, and regulated by the circadian clock (McClung, 1997). Reduction of CAT2 activity increases plant sensitivity towards ozone and photorespiratory H₂O₂-induced cell death; microarray analysis of the mutant also revealed that photorespiration-generated H₂O₂ influences transcriptional programs in plants (Vandenabeele et al., 2004). Moreover, CAT2 antisense plants show reduced activities of peroxisomal enzymes such as ICL and MLS during seedling germination (Vandenabeele et al., 2004; Eastmond, 2007). In concurrence with this observation, ICL activity can be inactivated by H₂O₂, whereas catalase associates with ICL to prevent this inactivation (Vandenabeele et al., 2004; Nguyen and Donaldson, 2005; Yanik and Donaldson, 2005). Another study found the cat2 null mutant to be highly compromised in growth in ambient air, with decreased rosette biomass, redox perturbations, and increased oxidative stress responses (Queval et al., 2007). Additionally, the cat2 mutant has been used as a tool to investigate effects of increased H₂O₂ on ion homeostasis, whereby the mutant is pale green and hypersensitive to assorted stresses including H₂O₂, salt, norspermidine, and cold, but surprisingly tolerant to lithium. It was subsequently deduced that the mutant is defective in ethylene production and perception and confers lithium tolerance, thus revealing novel cross talks between oxidative stress, cation homeostasis, and ethylene (Bueso et al., 2007). Consistent with the predominant expression of AtCAT3 in non-photosynthetic vasculature, the cat3 mutant does not display any evident phenotypes (Zimmermann et al., 2006). Lastly, CAT1 expression is induced by ABA via the mitogen associated protein kinase (MARK) signaling module (Xing et al., 2008).

3.7.2. APX, MDAR, DHAR, and GR

Besides catalase, plant peroxisomes utilize the cooperative activities of ascorbate peroxidase (APX), monohydroascorbate reductase (MDAR), dehydroascorbate reductase (DHAR), and glutathione reductase (GR) as part of the ascorbate-glutathione cycle to purge the peroxisome-generated H₂O₂ (Yamaguchi et al., 1995; Bunkelmann and Trelease, 1996; Mullen et al., 1999; Karyotou and Donaldson, 2005; Lisenbee et al., 2005). The system operates with APX catalyzing the oxidation of ascorbate to monohydroascorbate and the simultaneous reduction of H₂O₂ to water (Asada, 1992). Next, the monohydroascorbate is converted back into ascorbate by MDAR in an NADH-dependent manner (Hossain and Asada, 1984; Asada and Takahashi, 1987). Ascorbate is also capable of reacting with either superoxide or H₂O₂ nonenzymatically to produce dehydroascorbate or monodehydroascorbate. Spontaneous disproportionation of monohydroascorbate also occurs, whereas dehydroascrobate can be reduced to ascorbate by DHAR using glutathione as a reductant (Hossain and Asada, 1984). The glutathione oxidized in this process is reduced by GR using the cofactor NADPH. Overall, the cycle has no net consumption of the antioxidants, ascorbate or glutathione, while reducing H₂O₂ and generating oxidized cofactor NAD⁺ and NADP⁺.

The actions of the membrane-associated APX/MDAR system have been proposed to be a preventive measure, barricading the seepage of H₂O₂ and avoiding oxidative damage to membrane lipids and integral proteins (Yamaguchi et al., 1995; Mullen and Trelease, 1996). Senescence-associated degradation of proteins, lipid peroxidation, and general disintegration of organellar structures are believed to be incurred due to the high concentration of H₂O₂ , which is accumulated partly from the loss of catalase and diminished activities of APX/MDAR (del Rio et al., 1998; Zimmermann et al., 2006). Furthermore, these scavenging activities for ROS are regulated. For example, plant APX and CAT activities are suppressed during plant pathogen interactions, thereby increasing the incidence of H₂O₂-triggered defense responses (Dorey et al., 1998; Mittler et al., 1998; Leterrier et al., 2005).

APXs have been characterized to show higher affinity for H₂O₂ than catalase (Mittler and Zilinskas, 1991; Ishikawa et al., 1998). Arabidopsis is predicted to have eight APXs, among them APX3 is localized at the cytosolic side of the peroxisome membrane (Jespersen et al., 1997; Panchuk et al., 2002; Narendra et al., 2006). Based on N- and C-terminal sequence similarity to APX3, APX5 was postulated to be peroxisomal (Panchuk et al., 2002) but has not yet been experimentally associated with the organelle. Numerous studies report varying degrees of benefits to plants from overexpressing peroxisomal APX proteins, such as enhanced tolerance to high temperature, increased protection against oxidative stress caused by inhibition of catalase, and higher seed production under drought conditions (Wang et al., 1999; Shi et al., 2001; Yan et al., 2003). However, an analysis of the Arabidopsis apx3T-DNA insertion mutant indicated that APX3 is dispensable for growth and development, although this lack of phenotype could also be attributed to compensatory activities by other enzymes such as catalase and other APXs in the peroxisome (Narendra et al., 2006).

Arabidopsis peroxisomes have been catalogued to possess two MDARs with possibly redundant roles in ascorbate recycling. MDAR1 is a PTS1-containing matrix enzyme, whereas MDAR4 is an integral membrane protein with its catalytic domain facing the matrix side (Lisenbee et al., 2005). The functional importance of MDAR4 has been elucidated using a combination of forward genetic screens and extensive biochemical characterizations (Eastmond, 2007). MDAR4 activity is essential in detoxifying H₂O₂, which could otherwise leak and cause oxidative damage to the peroxisome-neighboring oil bodies. Because the TAG lipase present in the oil bodies is inhibited by H₂O₂, the leakage directly affects lipid metabolism by causing an effective block of TAG hydrolysis and the subsequent fatty acid β-oxidation steps, thus rendering sucrose dependence to post-germinative growth of the plant. However, peroxisome functions in the mdar4 mutant are unaffected, implying that other antioxidant enzymes may compensate for MDAR inactivity within the peroxisome. The primary physiological role of MDAR4 was proposed to be preventing the escape of H₂O₂ beyond the peroxisomal membrane (Eastmond, 2007).

DHAR1 (At1g19570) was detected in Arabidopsis leaf peroxisomes by proteomics and confirmed to be peroxisome localized (Reumann et al., 2009). Proteomic experiments also identified GR1 (At3g24170) in Arabidopsis peroxisomes (Eubel et al., 2008; Reumann et al., 2009), and its peroxisomal targeting was confirmed in vivo (A. Kataya and S. Reumann, unpublished data). The oxidized cofactors generated during the operation of the ascorbate-glutathione cycle are most likely reduced by the activities of several dehydrogenases documented to be present in the peroxisomes, i.e., malate dehydrogenases, NADP isocitrate dehydrogenase (At1g54340), 6-phosphogluconate dehydrogenase (At3g02360), and 6-phosphoglucono-lactonase (At5g24400) (Nyathi and Baker, 2006; Reumann et al., 2007; Eubel et al., 2008; Reumann et al., 2009).

3.7.3. Other antioxidant enzymes

Glutathione S-transferases (GST), peroxiredoxins (Prx), and superoxide dismutase are other antioxidant enzymes reported to exist in plant peroxisomes (del Rio et al., 2006). GSTs are involved in detoxification by virtue of their ability to conjugate glutathione to various electrophilic metabolites, tagging them for vacuolar sequestration (Edwards et al., 2000). All three members of the theta (T) family of GST, namely, GSTT1, 2, and 3, localize to peroxisomes and preferentially act as glutathione peroxidases, reducing toxic hydroperoxides to monohydroxy alcohols (Reumann et al., 2007; Dixon et al., 2009). Glutathione peroxidase activity has been implicated in conferring oxidative stress tolerance, thus it is likely that these enzymes contribute actively to nullifying the threat posed by the high ROS flux and phytotoxic fatty acid hydroperoxides in peroxisomes (Cummins et al., 1999; Dixon et al., 2009). A Myb-like DNA-binding domain is located at the C termini of GSTT2 and GSTT3, masking a predicted PTS1; it was postulated that alternative splicing leads to the omission of the 3′ Myb domain and consequently peroxisome localization of the PTS1-containing protein (Dixon et al., 2009). Members of the phi (F) and tau (U) family were also detected by a proteome analysis of Arabidopsis leaf peroxisomes (Reumann et al., 2009).

Prxs are a group of H₂O₂-decomposing enzymes acting upon alkyl hydroperoxides and peroxinitrites. They are often associated with thioredoxins and cyclophilins, both of which have been detected in peroxisome proteome studies (Dietz et al., 2006; Reumann et al., 2009). Although reported in pea peroxisomes, to date none of the 10 Arabidopsis Prxs have been demonstrated to be peroxisomal (Horling and Dietz, 2002; Dietz et al., 2006). Finally, Arabidopsis peroxisomes also harbor a specific isoform of Cu-Zn superoxide dismutase (SOD3), which counters the threat posed by O₂-; further information about this protein can be found in the chapter devoted to oxidative stress in this book (Grene, 2002).

Taken together, Arabidopsis peroxisomes contain a comprehensive set of antioxidant enzymes that act coordinately to fortify the plant against H₂O₂-imposed oxidative stresses. A model summarizing the reactions catalyzed by this group of enzymes is depicted in Fig. 7.

3.11.1. Branched-chain amino acid metabolism

Peroxisomes in isolated mungbean hypocotyls could completely catabolize branched-chain amino acids (BCAA) such as valine, leucine, and isoleucine (Gerbling and Gerhardt, 1988, 1989). However, based on knowledge derived from mammalian systems and subsequent analysis in Arabidopsis, mitochondria harbor all enzymes required for catabolism of BCAAs to propionyl-CoA, with the fate of the propionyl-CoA unresolved [reviewed in (Graham and Eastmond, 2002)].

The basis for the peroxisomal involvement in BCAA metabolism came from the isolation of a peroxisomal β-hydroxyisobutyryl (HIBYL) -CoA hydrolase from the IBA-resistant mutant screen. The mutant, chy1, displays IBA resistance in root elongation and lateral root induction, and sucrose-dependent germination. The recombinant protein can hydrolyze HIBYL-CoA in vitro, and the chy1 mutant phenotype was complemented by the closest mammalian homolog, the mitochondrial CoA thioester hydrolase. These observations led to the proposition that CHY1 catabolizes valine in Arabidopsis and that the block in CHY1 activity causes accumulation of a toxic intermediate, methacrylyl-CoA, which would sequester free CoA and Cys-containing proteins, thus impairing β-oxidation activities (Zolman et al., 2001 b). A genetic screen for 2,4-DB-resistant (dbr) mutants identified the same gene (DBR5) and subsequent analysis demonstrated that a decrease in KAT activity, possibly caused by inhibition of the enzyme by the accumulated methacrylyl-CoA, was the cause for IBA resistance and sucrose dependence in the dbr5 mutant (Lange et al., 2004). ¹⁴C feeding assays showed that both dbr5 and chy1 mutants have significantly reduced valine catabolism (Lange et al., 2004).

A subsequent study found that the chy1 mutant was sensitive to increasing concentrations of isobutyrate and propionate but not valine. These results led to the suggestion that CHY1 likely functions in the metabolism of isobutyryl-CoA and propionyl-CoA, which are derived from branched-chain fatty acids such as phytanic acid, instead of valine (Lucas et al., 2007). However, results from this study seem difficult to reconcile with the reduction in valine catabolism observed previously in the dbr5 and chy1 mutants (Lange et al., 2004). A recent work has also implicated CHY1 in cold stress signaling, where the mutant showed altered expression of CSFs [C-repeat (CRT) dehydration responsive element (DRE)-binding factor], increased accumulation of ROS, higher sensitivity to freezing treatment post cold acclimation, and darkness-induced starvation (Dong et al., 2009). The exact function and biochemical role of CHY1 in the peroxisome still remains to be established.

Completion of the Arabidopsis genome sequencing led to the discovery of the monomeric sarcosine oxidase (SOX), which belongs to a family of flavin-containing enzymes whose homologs in bacteria and mammals catalyze oxidative reactions using secondary or tertiary amino acids as substrates (Reuber et al., 1997; Trickey et al., 1999). SOX not only acts on sarcosine, a glycine derivative and a key metabolite in mammals, but also utilizes L-pipecolate as substrate. Although there is no evidence that sarcosine occurs in plants, pipecolate is a known catabolite of lysine in plants (Galili et al., 2001; Farres et al., 2002; Rontein et al., 2002). The recombinant AtSOX catalyzes the oxidation of sarcosine and N-methyl amino acids, and the oxidation of L-pipecolate to Δ1-piperideine-6-carboxylate (P6C). In vitro import assays established the peroxisomal localization of this protein. Suppression of AtSOX through RNAi resulted in hyperaccumulation of pipecolate and reduction in α-aminoadipate, indicating that AtSOX metabolizes pipecolate in planta. It was proposed that AtSOX serves as a salvage enzyme, recycling pipecolate back to P6C to feed into the lysine catabolic pathway. Based on the activity of recombinant AtSOX against sarcosine and N-methyl amino acids, both of which are widespread in nature, and the fact that AtSOX could also detoxify antimetabolites, structural analogs of N-methyl aminoacids, it was speculated that this enzyme enables the plant to opportunistically utilize the two natural substrates (Goyer et al., 2004).

3.11.2. The ureide pathway

The ureide pathway funnels the nitrogen released from the purine bases into amino acid backbones and assumes additional importance in leguminous plants, which transport fixed nitrogen in the form of ureides from the nodules to the aerial portions of the plant (Todd et al., 2006). The pathway constitutes three enzymes that act sequentially in the conversion of uric acid to the ureide, S-allantoin (Fig. 9). Uricase oxidizes uric acid to produce H₂O₂ and 5-hydroxyisourate (5-HIU), which is then hydrolyzed by HIU hydrolase to 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline (OHCU). Then, OHCU is decarboxylated by OHCU decarboxylase to yield S-allantoin (Todd et al., 2006).

Uricase has been extensively studied as a peroxisomal enzyme in nodules of leguminous species and detected in Arabidopsis peroxisomes by proteomic experiments (Reumann et al., 2007; Eubel et al., 2008; Reumann et al., 2009). HIU hydrolase activity was first found to be associated with a soybean β-glucosidase-like protein in the peroxisome; however, none of its closest homologs in Arabidopsis are predicted to target to peroxisomes, suggesting that the peroxisomal isoform of the β-glucosidase-like protein might be restricted to leguminous species (Sarma et al., 1999; Raychaudhuri and Tipton, 2002). On the other hand, recombinant transthyretin-like proteins (TLPs or TTLs) from various non-plant species contain HIUase activity; the high sequence conservation among TLPs and 3D structure predictions suggested that the Arabidopsis TLP may possess similar activity (Hennebry et al., 2006; Ramazzina et al., 2006). Indeed, the Arabidopsis TLP, a bifunctional protein with a PTS2 sequence lodged between the HIU hydrolase and OHCU decarboxylase domains, was confirmed to be peroxisomal in a leaf peroxisome proteome analysis (Hennebry et al., 2006; Reumann et al., 2007). HIU hydrolase is catalytically active as a tetrameric protein, whereas OHCU decarboxylase processes its substrate as a protein dimer, prompting the hypothesis that the bifunctional Arabidopsis protein possibly undergoes post-import cleavage to facilitate efficient catalysis (Hennebry et al., 2006). However, the detection of full-length TLP in a 2-DE-based proteome study strongly argues against this idea, but instead favors a 1:1 stoichiometry of HIU hydrolase and OHCU decarboxylase in the plant peroxisomal HIU hydrolase/OHCU decarboxylase complex. S-allantoin is subsequently processed to ammonia and carbon dioxide in the ER (Werner et al., 2008). In addition to its function in the ureide pathway, Arabidopsis TLP may play other physiological roles. For example, two alternative splice variants of TLP lack the internal PTS2 and are likely to represent non-peroxisomal isoforms performing other functions (Reumann et al., 2007). In addition, full-length TLP was also shown to be plasma membrane-associated and interacting with the brassinosteroid receptor BRI1 (Nam and Li, 2004).

Figure 9.

Urate degradation

The oxidation of urate is catalyzed by uricase to generate 5-Hydroxyisourate (5-HIU). 5-HIU is subsequently hydrolysed by HIU hydrolase (HIUase) to yield 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline (OHCU), which undergoes decarboxylation by OHCU decarboxylase to release carbon dioxide and produce S-allantoin. The latter two activities are imparted by the bifunctional transthyretin-like protein (TLP) in Arabidopsis.

Thus, in all likelihood the ureide pathway is compartmentalized in the peroxisome (Fig. 9) and executed by the consecutive actions of uricase, HIU hydrolase, and OHCU carboxylase to release S-allantoin, whose subsequent processing occurs in the ER to liberate ammonia and carbon dioxide.

3.11.3. Salicylic acid biosynthesis

Salicylic acid (SA) is an important signaling molecule in inducing plant response to pathogens (Metraux, 2002). It is believed that SA is generated either through chorismate via the shikimate pathway or alternatively, via processing of cinnamic acid derived from phenylalanine (Sticher et al., 1997; Wildermuth et al., 2001). Because the processing of cinnamate to SA ultimately involves reduction by two carbon moieties, it has been suggested that this reduction occurs through β-oxidation in peroxisomes (Reumann, 2004). It was predicted that peroxisomal 4CL-like AAE isoforms might possibly be involved in the CoA activation of a small subset of structurally related compounds such as cinnamic and benzoic acids (Shockey et al., 2003). It is worth noting that, in an in vitro enzyme assay, three putative peroxisomal 4CL-like proteins activated cinnamic acid and derivatives thereof to their corresponding CoA esters (Kienow et al., 2008). Further in-depth characterization of these isoforms may help expanding our knowledge of SA biosynthesis. Currently, the specific contribution of the cinnamate-SA pathway in modulating SA-induced responses is unclear, thus warranting further work to establish its relevance in Arabidopsis (Metraux, 2002).

3.11.4. Isoprenoid mevalonic acid pathway

Isoprenoids are naturally occurring compounds found in all organisms and are widespread in plants in various forms, spanning from photosynthetic pigments, redox cofactors, hormones and sterols to a multitude of secondary metabolites (Seemann et al., 2006). All isoprenoids are synthesized via condensation of isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP; McGarvey and Croteau, 1995). In mammals, enzymes in the mevalonic acid (MVA) pathway are localized in peroxisomes (Kovacs et al., 2002; Kovacs et al., 2007). In plants, IPP/DMAPP biosynthesis occurs through the cytosolic MVA pathway or the plastidic methylerythritol phosphate pathway (MEP; McGarvey and Croteau, 1995; Rodriguez-Concepcion and Boronat, 2002). However, recent studies suggest that at least part of the MVA pathway may have been duplicated to localize in the peroxisome.

The first enzyme of the cytosolic MVA pathway, acetoacetyl-CoA thiolase (ACAT), seems to be dual located. A specific transcriptional variant of Arabidopsis ACAT isoform 1 (ACAT1.3) targets to peroxisomes in vivo as a C-terminal GFP fusion, whereas the best predicted peroxisomal candidate, ACAT1.2, which carries a PTS1-related tripeptide SAL>, remains cytosolic as an N-terminal fusion (Carrie et al., 2007). The second ACAT isoform, ACAT2, which lacks a PTS-related sequence, was identified from Arabidopsis leaf peroxisomes by proteomics (Reumann et al., 2007), but its localization has not been authenticated.

Sapir-Mir et al. (2008) analyzed the localization of the two known Arabidopsis IPP isomerases (IPI1 and IPI2), which catalyze the seventh reaction of the MVA biosynthesis in both cytosol and plastids by reversibly isomerizing IPP to DMAPP. The two IPP isomerases each have two transcriptional isoforms, with the full-length version containing N-terminal targeting signals for the plastid and mitochondria and the shorter isoform lacking the targeting signals. Both longer variants are dual targeted to plastids and mitochondria as C-terminal GFP fusions (IPI_long-GFP); the shorter variants either remain cytosolic as N-terminal fusions (GFP-IPI1_short) or target to peroxisomes with GFP inserted 10 residues upstream of the C-terminus (IPI_short-GFP-10aa). The C-terminal tripeptide (HKL>) of IPI1 and IPI2 has not been characterized as a PTS1 in any plant study nor is H known to be tolerated at pos. -3 in any plant PTS1 (see Fig. 10). It is possible that the enzyme adopts a default confirmation in the cytosol, in which the C terminus is internally hidden, and that the novel PTS1 HKL> becomes functionally active in peroxisome targeting only under specific yet unknown conditions.

Three other enzymes of the cytosolic mevalonate pathway, i.e., HMG-CoA synthase, mevalonate kinase, and mevalonate-PP decarboxylase, were proposed to carry PTS2-related nonapeptides (Sapir-Mir et al., 2008). However, one or even both of the two invariant residues of the plant PTS2 nonapeptides (Rx₆Hx) defined by Reumann (2004) are altered in all three enzymes, bringing their peroxisome targeting into question. Further work to determine the subcellular localization of these enzymes will be important before the association of the MVA pathway with plant peroxisomes can be firmly established.

4.3.1. Overview

Computational prediction uses mathematical algorithms (models) to predict organelle targeted proteins from genome sequences. Localization of the predicted proteins can then be verified in transient in vivo systems using fluorescent protein fusions. Computational prediction is particularly important in discovering low abundance peroxisomal proteins, and proteins whose genes are specifically expressed in organs inaccessible to peroxisomal proteomic analyses or induced only under specific developmental or environmental conditions.

Successful prediction algorithms need to meet two criteria, high prediction sensitivity and high prediction specificity. Proteins targeted to an organelle by a specific type of targeting signal such as PTS1 may be predictable if the primary structure of the signals is sufficiently specific and conserved. For peroxisomes, only matrix proteins carrying C-terminal PTS1 or N-terminal PTS2 peptides can be predicted. Proteins that cannot be predicted include those that are imported into the matrix either by internally located PTS1/2 peptides or by alternative targeting signals or pathways, proteins possibly imported by “piggy-backing”, and peripheral and integral membrane proteins (Saleem et al., 2006).

Although mechanisms of the PTS1 and PTS2 protein import pathways are largely conserved across evolutionary lineages (see 2.2), they vary in some details in each lineage. Biochemically, the three amino acid residues in PTS1 tripeptides are generally small-basic-hydrophobic, but significant differences in the ability to tolerate certain amino acid changes within PTS1s have been found between kingdoms (Kragler et al., 1998; Lamet-schwandtner et al., 1998). For example, histidine frequently occurs at position (pos.) -2 in PTS1s in fungi (Lametschwandtner et al., 1998), but it is rarely present at the same position in plant PTS1s (Reumann, 2004). Plant PTS1s also differ from those of fungi by frequently tolerating non-basic residues such as S and L at pos. -2 (Reumann et al., 2007; Reumann et al., 2009). Thus, PTS peptides need to be defined specifically for each taxon (e.g., higher plants, spermatophyta) before taxon-specific genome screens are conducted. Likewise, general prediction algorithms developed primarily for specific taxa should be applied to other organisms with caution (Emanuelsson et al., 2003; Neuberger et al., 2003b; Hawkins et al., 2007).

Although the C-terminal PTS1 tripeptide is considered the major determinant for the PTS1 protein import pathway, approx. seven to nine amino acid residues upstream of the tripeptide may also be pivotal in enhancing or reducing the efficiency of peroxisomal protein targeting (Neuberger et al., 2003b). This upstream region and the C-terminal PTS1 tripeptide are together referred to as the PTS1 domain. Therefore, PTS1 protein prediction is generally a two-step procedure whereby (i) the probability for the C-terminal tripeptide to represent a functional PTS1 (PTS1 tripeptide filter) and (N) the degree to which the upstream region (pos. -4 to -10 or -12) matches consensus PTS1 domains, are sequentially evaluated.

4.3.2. Plant PTS1 tripeptides

Comprehensive identification of PTS1 tripeptides in higher plants (spermatophyta) is necessary to ensure a high prediction and identification rate of plant peroxisomal proteins. Plant PTS1 tripeptides have been defined by both large-scale experimental and computational analyses.

Experimental studies include (i) in vivo subcellular targeting of reporter proteins fused to short peptides containing PTS1 (Hayashi et al., 1996; Hayashi et al., 1997; Mullen et al., 1997), and (ii) yeast two-hybrid analysis of the interaction between these small peptides with the PTS1 receptor PEX5 (Kragler et al., 1998). A number of PTS1 tripeptides were confirmed by independent studies, whereas other tripeptides (e.g., SKI>, SSL>, SHL>, and GRL>) yielded discrepant subcellular targeting results, possibly due to differences in analytical sensitivity, primary structure of the reporter protein, or structure of the spacer region upstream of PTS1 (Fig. 11, also see summary in Reumann, 2004).

For computational prediction, a training set of true positive examples is generally assembled by searching well-annotated protein databases for known PTS1 proteins, followed by a homology-based search for the orthologs of the identified proteins (Neuberger et al., 2003a; Hawkins et al., 2007). Due to the relatively small number of known PTS1 proteins, model organisms for peroxisome research, and fully sequenced genomes, previous sequence datasets for mammals, fungi, and plants were rather limited in size. Expressed sequence tag (EST) databases are a valuable resource in identifying orthologous PTS1 proteins from a wide variety of organisms. The usage of EST databases was first applied to Arabidopsis, which led to a 5-fold extension of the full-length plant PTS1 protein dataset and the identification of novel, non-canonical PTS1 tripeptides (Reumann, 2004).

Figure 11.

Sequence logos for plant proteins with major or minor PTS1 tripeptides.

Position-specific amino acid compositions of the C-terminal 10 amino acid residues of true peroxisomal PTS1-containing proteins, which carry either one of the nine major PTS1s (A) or one of the eleven minor PTS1s (B) as defined by Reumann (2004), are shown. The sequence logo was generated by WebLogo 3 ( weblogo.berkeley.edu/).

Despite valuable PTS1 predictions, the computational approach has been severely limited by the low number of known plant PTS1 proteins available. For example, only 13 orthologous groups could be used as queries for the identification of homologous plant ESTs (Reumann, 2004). Furthermore, plant homologs of most known PTS1 enzymes, such as those involved in photorespiration, ROS metabolism and fatty acid β-oxidation, usually carry well-known major PTS1s and thus hardly contribute to the identification of novel PTS1 tripeptides. In contrast, plant homologs of a few other enzymes such as sulfite oxidase preferentially carry non-canonical PTS1 tripeptides (e.g., SNL>, ANL>, SSM>, SNM>), which allowed the identification of a large number of previously unrecognized, minor PTS1s (Reumann, 2004). Why plant PTS1 proteins display such a great variety in PTS1 tripeptides remains largely speculative. However, limited experimental and computational data suggest that (i) low-abundance PTS1 proteins with reduced expression levels can tolerate less efficient PTS1 tripeptides for quantitative import into peroxisomes, and (ii) sequences upstream of the C-terminal tripeptides in some orthologous groups of PTS1 proteins are particularly enriched in conserved targeting-enhancing elements that may compensate for the lower peroxisome targeting efficiency of non-canonical, minor PTS1s (see below).

Proteome analysis of Arabidopsis peroxisomes has emerged as a particularly successful approach, both directly and indirectly (after retrieval of homologous ESTs), in identifying novel PTS1 proteins/tripeptides. Two recent proteome analyses of leaf peroxisomes, combined with in vivo protein targeting studies, together established five new tripeptides as novel functional plant PTS1s (Reumann et al., 2007; Reumann et al., 2009). In addition, the C-terminal tripeptide AKI> in monodehydroascorbate reductase was shown to be another likely PTS1 (Lisenbee et al., 2005), and SHL> in a peroxisomal protein kinase was characterized as a functional PTS1 (Ma and Reumann, 2008). In total, 35 tripeptides have been confirmed or are predicted from reliable datasets to represent functional plant PTS1 tripeptides (Fig. 10).

Six to seven amino acid residues appear to be allowed at each position in plant PTS1 tripeptides (pos. -3: S, A, P, C, G, T; pos. -2: R, K, N, S, M, H, L; pos. -1: L, M, I, V, F, Y; Fig. 10). Although the current list of functional PTS1s is certainly incomplete due to training dataset limitations, it seems that only a subset of all possible 252 amino acid combinations of the principally allowed amino acid residues are functional plant PTS1s. As a general rule, a plant PTS1 tripeptide is only functional in peroxisome targeting if it carries at least two of the six most abundant position-specific amino acid residues (i.e., S, A, R, K, L, M) in the form of [SA][RK]x>, [SA]y[LM]>, or z[RK][LM]> (Reumann, 2004). To date only one PTS1 tripeptide, SSI>, slightly diverges from this rule (Fig. 10).

4.3.3. Effects of upstream sequences on PTS1 targeting

Peroxisome targeting efficiency differs among distinct PTS1 tripeptides, several of which are sufficient to target reporter proteins to peroxisomes irrespective of upstream residues (Hayashi et al., 1997; Mullen et al., 1997; Kragler et al., 1998). These “stand-alone” signals, such as SKL> and SRL>, are the prototypical “strong” PTS1 tripeptides. The function of other so-called weak PTS1 tripeptides often depends on some basic amino acid residues located in close proximity to the C-terminal tripeptide (Fig. 11). Experimentally defined weak PTS1 tripeptides include ANL> and SHL> (Mullen et al., 1997; Ma and Reumann, 2008).

Given the lack of systematic experimental data and large sequence datasets, the current classification of strong and weak PTS1 tripeptides is only arbitrary. Plant PTS1s have been categorized into nine major and eleven minor PTS1 tripeptides based on their abundance in the dataset, which is thought to roughly correlate with the targeting strength of the tripeptides (Reumann, 2004). This classification is further supported by statistically significant differences in position-specific amino acid composition between plant proteins carrying major and minor PTS1s. Upstream domains in proteins terminating with a minor PTS1 are significantly enriched in basic residues, prolines, and/ or hydrophobic residues (A, L, V, I), whereas these regions in proteins with a major PTS1 are on average only weakly enriched in such residues (Fig. 11). Except for the well-documented function of basic residues (R, K), the predicted physiological function of these enriched residues as enhancing elements for targeting remains to be verified experimentally. The predicted inhibitory effect of residues underrepresented in PTS1 domains (e.g. E, D) has been largely untested. However, a striking example for such putative inhibitory elements comes from the study of an Arabidopsis protein kinase, AtPK1, which contains several acidic residues upstream from the C-terminal prototypical PTS1 (SKL>) that appeared to inhibit peroxisome targeting. Such upstream acidic residues can inhibit and even prevent peroxisome targeting despite the presence of strong PTS1 tripeptides, and thus also need to be considered in prediction algorithms (Ma and Reumann, 2008).

4.3.4. Definition of PTS2 peptides and domains

About one-third of plant peroxisomal matrix proteins appear to use the PTS2 import pathway (Reumann, 2004). The corresponding targeting signal PTS2, with RLx₅HL as the prototype, is a nonapeptide located near the N terminus and consisting of four conserved residues split into two parts by a stretch of five relatively flexible residues. More restrictive PTS2s such as R[ILQx₅HL] (Kato et al., 1996; Kato et al., 1998) or permissive PTS2s such as [RK]X₆[HQ][ALF] (Flynn et al., 1998) have been defined. Twelve functional PTS2 nonapeptides, all of which carry R at pos. 1 and H at pos. 8, were defined from a plant-specific EST training dataset of PTS2 proteins. Pos. 2 was found to be the most variable position of the nonapeptide, allowing L, I, Q, T, M, A, and V; pos. 9 is the second variable, allowing L, I, and F (Reumann, 2004). The up- and downstream regions of PTS2 and the internal x₅-region are enriched in basic amino acid residues, such as R and P, and scarce in acidic residues.

PTS2 nonapeptides were believed to be confined to the first 30 amino acid residues at the N terminus. This general rule has been challenged by the proteomic (followed by in vivo targeting) identification of the Arabidopsis transthyretin-like protein (TLP), a bifunctional enzyme involved in purine metabolism and carrying an internal PTS2 (RLx₅HL) in a linker region connecting the decarboxylase and hydrolase domains (Reumann et al., 2007). This atypical PTS2 appears to have evolved in TLP by fusion of a decarboxylase gene with a hydrolase gene containing the PTS2 (see 3.11.2). Thus, functional PTS2 nonapeptides are not restricted to N-terminal protein domains. Genome screens should be extended to proteins with internally located PTS2s, giving specific consideration to bifunctional proteins.

4.3.5. Prediction of PTS1/2 proteins in Arabidopsis

Moderate prediction accuracy is generally sufficient in predicting the localization of specific proteins. However, large-scale genomic screens require high prediction specificity to keep the number of false positive predictions to a minimum. In general, PTS1 tripeptide-based search algorithms or machine learning techniques have been applied to develop prediction tools (Emanuelsson et al., 2003; Neuberger et al., 2003a, b; Hawkins et al., 2007). Current PTS1 prediction servers include PEROXIP ( www.bioinfo.se/PeroxiP: Emanuelsson et al., 2003), the PTS1 PREDICTOR (mendel.imp.ac.at/mendeljsp/sat/pts1/PTS1 predictor.jsp; Neuberger et al., 2003a), and PProwler (pprowler.itee. uq.edu.au; Hawkins et al., 2007). Plant-specific predictions by machine learning techniques are still in their infancy because of small-sized and non-representative datasets (Emanuelsson et al., 2003; Neuberger et al., 2003a).

Public databases allow the scientific community to gain access to genome-wide predicted PTS1 proteins. Genetic and functional information about known peroxisomal proteins of Homo sapiens (85 proteins) and Saccharomyces (61 proteins) can be found in the database PeroxisomeDB ( www.peroxisomedb.org; Schluter et al., 2007). The knowledge of higher plant PTS1 and PTS2 peptides combined with improved ORF prediction and large-scale full-length cDNA sequencing for Arabidopsis, led to the prediction of approx. 320 PTS1 and 60 PTS2 proteins in the Arabidopsis genome (Kamada et al., 2003; Reumann et al., 2004); this information is summarized in the database AraPerox 1.2 ( www3.uis.no/araperoxV1). Many predicted PTS1 proteins have been verified as truly peroxisome targeted. The nearly 30 novel Arabidopsis PTS1 proteins identified since 2004 should significantly improve the size and quality (i.e. coverage of natural variability) of the EST sequence training dataset and assist with the development of plant-specific PTS1 prediction algorithms. Future challenges include the prediction of surface-exposed PTSs and dual/multiple subcellular protein targeting and transfer of prediction methodology established for PTS1 proteins to the more difficult prediction of PTS2 proteins.

Dual targeting of plant peroxisomal proteins has been addressed in two recent studies (Carrie et al., 2008; Carrie et al., 2009) but remains difficult to predict, if native folding of full-length proteins and competition of multiple targeting signals are to be considered. In recent years, complexity of the Arabidopsis proteome has further increased with the characterization of additional protein variants, many of which differ in the presence or absence of predicted PTSs. The existence of these variants indicates that subcellular targeting is often regulated at the transcriptional, translational, and even post-translational levels. Thus, dual or multiple protein localization is much more common than previously anticipated.

4.4.1. Methodology and dynamic range of protein identification

Most proteomic studies of plant peroxisomes have emerged only recently. There are several inherent problems for plant peroxisomal proteome analyses. First, genomic sequence information is lacking for established model plant species (e.g., pea, pumpkin, and spinach) used in peroxisome research, which prompted the need to establish peroxisome isolation protocols specifically for Arabidopsis. Second, Brassicaceae are known to contain the highest numbers and concentrations of organelle-destabilizing secondary metabolites, making peroxisomes particularly fragile during isolation. Third, tight physical interaction between peroxisomes, mitochondria, and chloroplasts in Brassicaceae reduces the purity of leaf peroxisome isolates. Here, we summarize and compare all six proteome studies of plant peroxisomes published to date, with respect to methodologies in organelle isolation, protein and peptide separation, dynamic range of protein identification, and peroxisome targeting verification.

Two pioneering proteomic studies in Arabidopsis were performed on peroxisomes from greening and etiolated cotyledons, respectively (Fukao et al., 2002; Fukao et al., 2003). Peroxisomes were purified by a single density gradient, and the purity was analyzed by immunoblotting of the gradient fractions. After reduction of sample complexity by 2D-gel electrophoresis (2DE), the proteins were visualized by silver staining and the tryptic peptides were analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS). Protein identification was solely based on peptide mass fingerprinting (PMF) data, which bears some risk in identifying false positives. Several enzymes involved in ROS metabolism were identified from both studies, whereas known proteins of photorespiration and fatty acid β-oxidation were only specifically identified in green or etiolated cotyledons. Twenty of the 29 proteins identified from green cotyledons and 13 of the 19 proteins identified from etiolated cotyledons were described as putative novel proteins of peroxisomes.

A later study focusing on peroxisomes from mature Arabidopsis leaves used a Percoll density gradient followed by sucrose density gradient centrifugation to enrich the organelles (Reumann et al., 2007). For comparison of peroxisome isolates, this work used the specific activity of a marker enzyme, hydroxypyruvate reductase (HPR), as an indication of organelle purity. Sample complexity was reduced by 2DE and proteins were identified by MALDI-TOF tandem MS, or by gel-free shotgun proteomics using LC-MS/MS. This study detected 36 established peroxisomal proteins, including many proteins with roles in so-called minor functions of leaf peroxisomes such as fatty acid β-oxidation, S and N metabolism, and JA biosynthesis. Forty-two proteins were considered putative novel proteins, indicating a higher level of sensitivity achieved in protein identification.

The dynamic range of protein identification was further increased in a study carried out by the Arabidopsis 2010 Peroxisome Project ( www.peroxisome.msu.edu) using mature leaves (Reumann et al., 2009). Peroxisome enrichment was further improved through post-preparative immunoblotting analysis and by application of a 1DE shotgun LC-MS/MS approach. Among the 150 peroxisome-associated proteins identified, 65 represent established peroxisomal proteins including nearly all known matrix proteins of plant peroxisomes and several membrane proteins such as PXA1, PMP38, and all five PEX11 isoforms. Using stringent selection criteria, 85 proteins were assigned as putatively novel, among which 30 proteins had also been identified by previous proteome analysis but not yet verified by independent methods, and 55 proteins were discovered for the first time in the proteome of plant peroxisomes.

Using Arabidopsis suspension-cultured cells, Eubel et al. (2008) purified peroxisomes with a free-flow electrophoresis method established earlier for plant mitochondrial enrichment (Eubel et al., 2007). This study applied two quantitative proteomic methodologies to identify true peroxisomal proteins: (i) differential in-gel electrophoresis (DIGE) of enriched peroxisomes and mitochondria, and (N) normalized spectral count analysis of shotgun proteome data from peroxisome fractions differing in their degree of purity. The identification of membrane proteins was optimized by sodium carbonate treatment of peroxisomes. In total, eighty-nine proteins were assigned as peroxisomal; among them about 20 proteins had not been associated with peroxisomes before. Phosphopeptide enrichment also allowed the detection of a peptide from PMP38 (At2g39970), which was phosphorylated at Ser-155.

Only a single peroxisome proteome study has been reported for plant species other than Arabidopsis, mostly due to difficulties in protein identification that generally requires de novo sequencing of tryptic peptides. Arai et al. (2008a) purified peroxisomes from etiolated soybean (Glycine max) cotyledons, using Percoll followed by iodixanol density gradient centrifugation. Immunoblotting was applied for purity monitoring and protocol optimization. Following protein separation by 2DE, protein identification by PMF was made possible by data searches against a soybean database (DFCI Soybean Gene Index) consisting of non-redundant soybean tentative consensus sequences created by soybean EST assemblies. Thirty proteins were assigned to peroxisomes, including 26 orthologs of established plant peroxisomal proteins, three orthologs of proteins already identified by former proteome studies of Arabidopsis peroxisomes, and one putative novel protein, i.e., a homolog of the mitochondrial VDAC family. In a follow-up study, Arai et al. (2008b) separated protein complexes by blue native/SDS-PAGE and identified one membrane protein (Gm PNC1) that was later characterized as a member of the peroxisomal adenine nucleotide carrier family.

4.4.2. Verification of peroxisome targeting by in vivo subcellular targeting analysis

Peroxisomes can be enriched over other organelles, but they cannot be isolated to absolute purity. For this reason, the distinction between true peroxisomal proteins and proteins originating from minor contaminations by other organelles is a serious concern for peroxisome proteome studies in all organisms. With an increasing dynamic range of protein identification, the annotation of unknown proteins becomes more challenging. Several methods have been applied to verify the postulated peroxisome association of unknown proteins detected in peroxisomal proteome studies: (i) analysis of predicted PTSs, (ii) analysis of PTS peptide conservation in orthologous proteins and homologous EST sequences (Reumann et al., 2007; Reumann et al., 2009), (iii) immunobiochemical analysis of density gradient fractions for protein co-migration with markers (Fukao et al., 2003), (iv) semi-quantitative proteome analysis of peroxisome isolates with different degrees of purity (Eubel et al., 2008), and (v) in vivo subcellular targeting analysis of fluorescent protein fusions. This last method has emerged as the most efficient way to provide conclusive evidence for peroxisomal targeting and is nowadays widely applied as an extension to proteome analysis of peroxisomes from fungi, mammals, and plants (Marelli et al., 2004; Reumann et al., 2007; Wiese et al., 2007; Arai et al., 2008a; Eubel et al., 2008; Reumann et al., 2009).

In the two initial proteomic studies, the number of putative novel proteins carrying known or PTS1/2-related peptides was low (Fukao et al., 2002; Fukao et al., 2003). In the green cotyledon peroxisomal analysis, for instance, two putative novel proteins carry C-terminal PTS1s (the protein kinase PK2, At4g31230, PKL>; and a putative beta glucosidase, At4g27820, SSL>) and two additional proteins contain PTS1-related peptides, SKD> and IRL> (Fukao et al., 2002). The protein kinase PK2 has been verified to carry a functional PTS1 domain that is sufficient to direct GFP to peroxisomes; however, the full-length protein remains cytosolic when expressed in onion epidermal cells, indicating that the kinase is likely dual-targeted to both the cytosol and peroxisomes in a transient and regulated manner (Ma and Reumann, 2008). Three proteins identified from the peroxisome proteome of etiolated cotyledons (Fukao et al., 2003) carry predicted PTS1 tripeptides (isocitrate dehydrogenase, IDH, At1g54340, SRL>; small thioesterase 3, sT3, At3g61200, SKL>; and glyoxysomal protein kinase 1, GPK1/PK7, At3g17420, AKI>). Both IDH and sT3 have been repeatedly identified in later proteome analyses of Arabidopsis peroxisomes and were confirmed to be peroxisome-targeted as fluorescent protein fusions (Reumann et al., 2007; Eubel et al., 2008; Reumann et al., 2009). Peroxisome targeting of glyoxysomal protein kinase 1 (GPK1/PK7) was supported in the original proteome study by immunobiochemical analysis of the density gradient fractions (Fukao et al., 2003). However, organelle-like but non-peroxisomal targeting of full-length EYFP-GPK1/PK7, EYFP-GPK1/PK7ΔAKI, and EYFP fused with the putative PTS1 domain of GPK1/PK7 raised some doubts on its localization in peroxisomes (Ma and Reumann, 2008).

Of the 42 putative novel proteins identified by the first proteome study of Arabidopsis peroxisomes from mature leaves, 28 carry known or PTS-related targeting signals, many of which are conserved among homologous ESTs. Eleven proteins were confirmed to be peroxisomal as GFP fusions in the same study (Reumann et al., 2007). Five additional proteins (BSMDR, At1g49670; ECHIa, At4g16210; ECHIc, At1g65520; SDRa, At4g05530; and NBP/HIT1, At4g16566) from this list have subsequently been verified as peroxisomal by in vivo targeting analyses (Eubel et al., 2008; Goepfert et al., 2008; Reumann et al., 2009; Wiszniewski et al., 2009).

Eubel et al. (2008) detected about 20 putative novel proteins of peroxisomes; one of them, which is homologous to the yeast peroxisomal adenine nucleotide transporter ANT1 (Palmieri et al., 2001) and encoded by At5g27520, was confirmed to be peroxisomal by in vivo subcellular targeting analysis in the same study. This protein and its paralog encoded by At3g05290 have been functionally characterized as plant peroxisomal adenine nucleotide carriers, with PNC1 assigned to At3g05290 and PNC2 to At5g27520 (Arai et al., 2008b; Linka et al., 2008). Five of the 20 putative novel proteins identified by this study, i.e., acyl-activating enzyme 1 (AAE1), zinc-binding dehydrogenase (ZnDH), malonyl-CoA decarboxylase (MCD), indigoidine synthase A (IndA), and copper amine oxidase (CuAO), were later conclusively demonstrated to be peroxisome targeted (Reumann et al., 2009).

The most comprehensive in vivo subcellular localization study of peroxisomal proteins was carried out in an in-depth proteome analysis of Arabidopsis leaf peroxisomes (Reumann et al., 2009). For medium throughput cloning of candidate genes, Gateway®-compatible vectors were created for fusing the coding region of the candidate cDNAs to the N (for PTS2-containing proteins) or C terminus (for PTS1-containing proteins) of YFP. Among the 55 putative novel peroxisomal proteins assigned in this study, nine proteins carry predicted plant PTS peptides and seven contain PTS1/2-related sequences such as STL>, SLL>, SQV>, SLM>, and RVx₅HF. Peroxisomal localization was confirmed for 13 proteins, including six with known PTSs, three carrying PTS-related peptides and four proteins that lack conventional targeting signals. This study also validated the peroxisomal targeting for five proteins that had independently been detected by Eubel et al. (2008). Although the 14 proteins that failed to show peroxisome targeting in the transient assay are possible contaminants from other cell compartments, several of them were expected to be truly associated with peroxisomes in plant cells under specific physiological conditions.

In the soybean peroxisome proteome study, all three PTS1 proteins (GmSDR, GmECHI, and GmHCDH) non-homologous to established plant peroxisomal proteins were verified as peroxisomal by in vivo subcellular targeting analysis (Arai et al., 2008a). Arabidopsis orthologs for these three proteins (AtSDRa, At4g05530; AtECHIa, At4g16210; AtHCDH, At3g15290) were also identified and proven to be peroxisomal targeted in vivo (Reumann et al., 2007; Eubel et al., 2008; Wiszniewski et al., 2009). Surprisingly, a specific homolog of the mitochondrial superfamily of voltage-dependent anion-selective channels (VDAC) was shown in vivo to be targeted exclusively to peroxisomes in both N- and C-terminal GFP orientations (Arai et al., 2008a). Functional characterization of the conductance properties of this VDAC homolog is required to determine whether it is similar to the well-characterized mitochondrial VDAC homologs or the porin-like channel characterized in spinach leaf peroxisomes and castor bean glyoxysomes (see 3.4).

4.4.3. Novel targeting signals

The identification of novel peroxisomal matrix proteins by proteome approaches is not only essential to discovering unexpected functions of plant peroxisomes, but also important in improving the prediction of PTS1/2 proteins from plant genome sequences. For several novel peroxisomal proteins containing PTS1 or PTS2, more than 50 homologous plant sequences can be retrieved from EST databases for each protein. Novel matrix proteins can thus substantially contribute to the extension of the training dataset needed for developing high accuracy prediction algorithms. To this end, in vivo subcellular targeting analyses should be performed for each novel protein to show that (i) the full-length protein is peroxisomal, (ii) removal of the corresponding PTS leads to mistargeting of the protein to other locations, and (iii) fusion of the predicted PTS domain to a fluorescent protein directs the protein to peroxisomes. Using these methods, several novel PTS1 tripeptides (SSL>, SSI>, ASL>, SLM>, and SKV>) were conclusively identified in extended proteome studies (Reumann et al., 2007; Reumann et al., 2009). These novel PTS1 s can be directly applied to genome-wide searches for novel PTS1 proteins, with the note of caution that peroxisome targeting by these presumably weak PTS1 tripeptides is likely to depend on the presence of auxiliary targeting-enhancing elements in the upstream region. Additionally, the nonapeptide RVx₅HF also represents a new PTS2 variant (Reumann et al., 2009; S. Quan and J. Hu, unpublished data).

With the increasing sensitivity of proteome studies, more peroxisomal proteins possessing PTS-related peptides or even lacking recognizable PTSs have been identified. This suggests that low-abundance proteins of plant peroxisomes mostly carry non-canonical and atypical PTS peptides and that yet unrecognized PTS1/2-independent targeting pathways exist for some peroxisomal matrix proteins.

4.4.4. Future challenges and opportunities

With elevated sensitivity in proteomics and MS methodology, quantification techniques become increasingly important to differentiate between peroxisomal proteins and low-abundance contaminants. Mass spectrometry is inherently poorly quantitative, but gel-based relative quantification techniques have been developed for quantification at the protein level (e.g., DIGE, differential in-gel electrophoresis) and successfully applied to plant peroxisomes (Eubel et al., 2008). In shotgun MS experiments, peptides can be quantified by spectral counts, i.e., integrated spectrum peak (Ong and Mann, 2005) or spectral abundance factors normalized to the protein size, NSAF (Paoletti et al., 2006). For the assignment of membrane proteins to specific organelles, the LOPIT (localization of organelle proteins by isotope tagging) method was developed for proteome studies of endomembranes (Dunkley et al., 2004; Dunkley et al., 2006); a related approach was also successfully applied to mouse peroxisomes (Wiese et al., 2007). However, these methods have not been employed to plant peroxisomes.

The identification of peroxisomal membrane proteins remains most challenging. Forty grams of yeast cells or Arabidopsis leaves, after labor-intensive enrichment procedures, generally yield less than 100 µg of purified peroxisomes. As such, the isolation of 50 µ9 of peroxisomal membrane proteins, which comprise less than 5% of total peroxisomal proteins, would require more than 10 high-purity organelle isolations. Moreover, membrane proteome analyses of peroxisomes require even higher peroxisome purity as compared to whole organelle proteome studies, because the membrane protein content of contaminating mitochondria and chloroplasts exceeds that of peroxisomes and leads to an apparent increase in non-peroxisomal proteins in the sample. Affinity-tagging of peroxisomal membrane proteins, a method used for S. cerevisiae (Marelli et al., 2004) and rat liver (Kikuchi et al., 2004) peroxisomes may help to alleviate this problem and even allow proteome characterization of peroxisomal sub-populations in plants.

In addition to reliable assignment of low-abundance proteins to peroxisomes by quantitative proteomics methods, it will soon be desirable to identify proteins transiently associated with peroxisomes and to monitor peroxisomal proteome alterations in Arabidopsis mutants. A widely applied alternative method for achieving higher sensitivity in protein identifications is chemical labelling of peptides derived from two different biological samples. One such example is isotope-coded affinity tagging (ICAT), which has been applied to Saccharomyces peroxisomes (Marelli et al., 2004). Its variant, referred to as isobaric Tag for Relative and Absolute Quantitation (iTRAQ), is increasingly applied. An alternative to chemical labelling is metabolic labelling of proteins with stable heavy isotopes. SILAC (stable isotope labelling by amino acids in cell culture) was particularly developed for cell cultures, in which ¹⁵N-L-lysine was provided to the cells. The relative protein abundance is determined by comparing lysine- or N-containing differentially labelled peptide pairs (Conrads et al., 2001). For proteome analysis of whole plants, hydroponic isotope labeling of entire plants (HILEP) seems to be a cost-effective method that enables metabolic labeling of whole and mature plants with a stable isotope such as ¹⁵N, which is applied in the form of nitrate as the sole nitrogen source (Bindschedler et al., 2008). The application of these quantitative MS methods will allow us to gain deeper insights into the dynamics of the peroxisome proteome and the characterization of peroxisomal subpopulations.