A SMALL FAMILY OF CHLOROPLAST ATYPICAL THIOREDOXINS

The reduction and the formation of regulatory disulfide bonds serve as a key signaling element in chloroplasts. Members of the thioredoxins superfamily of oxidoreductases play a major role in these processes. We have characterized a small family of plant specific thioredoxins in Arabidopsis thaliana that are rich in cysteine and histidine residues, and are typified by a variable non-canonical redox active site. We found that the redox midpoint potential of three selected family members is significantly less reducing than that of the classic thioredoxins. Assays of subcellular localization demonstrated that all proteins are localized to the chloroplast. Selected members showed high activity, contingent on a dithiol electron donor, towards the chloroplast 2-Cys peroxiredoxin A, and poor activity towards the chloroplast NADP-malate dehydrogenase. The expression profile of the family members suggests that they have distinct roles. The intermediate redox midpoint potential value of the atypical thioredoxins might imply adaptability to function in modulating the redox state of chloroplast proteins with regulatory disulfides.


INTRODUCTION
Recent studies suggest that redox regulation is a major mechanism in cellular signal transduction and in control of gene expression in species as diverse as humans, plants and bacteria (Danon, 2002;Toledano et al., 2004;Buchanan and Balmer, 2005;Holmgren et al., 2005). Members of the thioredoxin (Trx) superfamily of thiol-disulfide oxidoreductases play key roles in cellular redox regulation in almost all life forms (Debarbieux and Beckwith, 2000;Holmgren et al., 2005;Carvalho et al., 2006). Trxs share a similar fold (Trx-fold) and a common active site motif comprised of two vicinal cysteines separated by two variable amino acids, the CXXC motif (Martin, 1995). Despite these common features, members of the Trx superfamily have diverse functions ranging from protein disulfide reductases to disulfide isomerases or to disulfide transferases. The versatility in function is reflected in part by the broad range of redox midpoint potentials measured for the Trxs. Redox midpoint potentials range from -300 mV for the reductive type Trx-m (Hirasawa et al., 1999) to -120 mV for the oxidative type Escherichia coli DsbA (Zapun et al., 1993;Aslund et al., 1997). Protein disulfide isomerases (PDIs), the bacterial DsbD, and glutaredoxins have intermediate redox midpoint potentials, from -175 to -241 mV (Lundstrom andHolmgren, 1993,Collet, 2002 #10;Aslund et al., 1997).
The observed differences in the redox midpoint potential between different Trxs reflect to some extent divergence in the stabilization of the negative charge of the N-terminal catalytic thiolate, and thus in the pKa, of the redox active site. One of the factors suggested to have an important role in determining the pKa of the catalytic cysteine and the redox midpoint potential of the protein is the amino acid identity of the central dipeptide, located between the two cysteines of the active site. The classic reductive-type Trxs share a canonical active site motif, the C(G/P)PC motif, while PDIs and the oxidative DsbA have the conserved CPHC and a CGHC motifs, respectively. Mutating the central dipeptide was shown to change the redox midpoint potential of the protein, as well as its activity and its interactions with other proteins (Krause and Holmgren, 1991;Lundstrom et al., 1992;Chivers et al., 1996;Huber-Wunderlich and Glockshuber, 1998;Mossner et al., 1998;Quan et al., 2007). Additional amino acids outside of the active site were also suggested to contribute to the redox midpoint potential or the activity of different Trxs. Among them are several charged residues in the vicinity of the active site, and a www.plantphysiol.org on August 29, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved.
neighboring conserved tryptophane that precedes the active site. The exact contribution of each of these factors to the overall activity of Trxs is not yet clear, and might differ between different members of the superfamily (Debarbieux and Beckwith, 2000).
In plants, the redox reactions of the photosynthetic machinery of the chloroplast need to adjust to both rapid and gradual environmental changes, such as photon flux, CO 2 availability or water potential, and the associated free radicals production. Four families of chloroplastlocalized classic Trxs containing a canonical active site, Trx-f, Trx-m, Trx-x and Trx-y were characterized. The f-type and m-type Trxs were shown to be reduced by the Ferredoxin-Thioredoxin system, which receives electrons from the light-capturing reactions of photosynthesis. The classic chloroplast Trxs are implicated in the light-dependent activation by reduction of different enzymes, including several metabolic enzymes such as NADP-malate dehydrogenase (MDH) and Fructose1,6-bisphosphatase, as well as in protection against oxidative stress (Schurmann and Buchanan, 2008). Additional roles, such as in oxidative-type reactions, are inferred indirectly by studies demonstrating that protein disulfide bonds are formed correctly in the chloroplast (Staub et al., 2000;Bally et al., 2008;Wittenberg and Danon, 2008).
Further roles of atypical Trxs, containing a non-canonical redox active site or additional non-Trx domains, are just starting to emerge. For example, the recently characterized thylakoid membrane anchored HCF164 protein has an atypical CEVC active site, and a redox midpoint potential value of -224 mV (Motohashi and Hisabori, 2006). It was found that HCF164 is essential for the assembly of the chloroplast cytochrome b 6 f (Lennartz et al., 2001).
As studying atypical Trxs is expected to shed light on additional functions of Trxs in plants, we initiated a search in A. thaliana genome for Trxs with a non-canonical active site. We characterized a small family of proteins, which we denoted AtACHTs (Atypical Cysteine Histidine rich Thioredoxins). Notably, we found that the redox midpoint potential of three selected proteins, AtACHT1, AtACHT2a and AtACHT4a, is significantly higher than that of the classic A. thaliana Trxf1 (AtTrx-f1). Assays of the subcellular localization of the AtACHTs suggest that they are localized to the chloroplast. Examination of AtACHT1 and AtACHT4a showed that they are partitioned between the soluble stroma and the thylakoid membranes.
AtACHT1, AtACHT2a and AtACHT4a showed high preference in their activity towards the

Characteristics of the AtACHT family of proteins
To begin to look at Trxs with possible distinct activities from these of the well-characterized reductive type chloroplast Trxs, we searched A. thaliana genome for small Trx-like genes with a non-canonical CXXC active site. The first protein with a non-canonical active site that emerged from a NCBI BLASTP search using a query of AtTrx-f1 sequence was AtACHT1. A follow up analysis with AtACHT1 sequence identified a family of five genes denoted AtACHT1-5 (Fig. 1).
The proteins are homologues of the previously reported Lilium longiflorum Trx-like sequence (Meyer et al., 1999). AtACHT2a and AtACHT2b are predicted splice variants that differ only in their C terminus. Whereas, the predicted splice variants AtACHT4a and AtACHT4b differ only in a small portion of their N terminus. Three family members, AtACHT3, AtACHT4a-4b and AtACHT5, do not contain the highly conserved tryptophan that precedes the redox active site.
All family members include an N-terminal leader predicted to function as an organelle-targeting signal by the computer program TargetP (Emanuelsson et al., 2000). The proteins are typified by several conserved cysteines and histidines in addition to the active-site cysteines. They also contain a C-terminal tail with varying lengths with no expected common fold or function ( Fig.   1). Reverse-transcriptase polymerase chain reaction authenticated that all AtACHTs are actively transcribed (Data not shown). Secondary structure calculation, using the computer program JNet (Cuff and Barton, 1999), suggests that the core of all five proteins contains the structure elements that confer the Trx fold (Martin, 1995;Capitani et al., 2000) (Fig. 1).
To examine whether the proteins are active thiol-oxidoreductases, leaderless forms of three selected proteins, AtACHT1, AtACHT2a and AtACHT4a were expressed in Escherichia coli, purified to homogeneity, and their catalytic activity was compared to that of the classic AtTrx-f1 in the standard insulin turbidity assay (Holmgren, 1979). We found that the catalytic activity of AtACHT1 was slightly higher, and that of AtACHT2a was slightly lower, than that of AtTrx-f1 (Fig. 2). These results suggest that the AtACHT proteins preserve the catalytic activity that is common to members of the Trx oxidoreductase superfamily. Notably, the activity of the tryptophan-less AtACHT4a in parallel reactions was poor (Fig. 2).

The AtACHT proteins have a higher redox midpoint potentials than AtTrx-f1
It was shown that the central dipeptide sequence of the active site could influence the redox midpoint potential as well as the protein function (Krause and Holmgren, 1991;Lundstrom et al., 1992;Huber-Wunderlich and Glockshuber, 1998;Quan et al., 2007). Thus, the non-canonical active site motifs of the AtACHTs might suggest that they have redox properties that differ from these of the classic reductive-type Trxs. We therefore estimated their redox midpoint potential by the monobromobimane (mBBr) method (Hirasawa et al., 1999). We found the redox midpoint potential of a control Trx, AtTrx-f1, to be -290 mV at pH 7.0 (data not shown) in agreement with its published value (Hirasawa et al., 1999;Collin et al., 2003). In comparison, the selected AtACHT proteins, AtACHT1, AtACHT2a, and AtACHT4a, displayed significantly higher estimated redox midpoint potential values at pH 7.0, -237 mV ± 3, -239 mV ± 2.5, and -240 mV ± 1.5, correspondingly (Fig. 3). This finding infers that the AtACHT proteins might have roles differing from those of the reductive type classic Trxs.

The AtACHTs are targeted to chloroplasts
The presence of a leader peptide in the N-terminal sequence of all AtACHTs indicated that they might be localized to subcellular organelles. Thus, to study the localization of the AtACHTs, we compared the accumulation of each of the transiently expressed AtACHTs, fused with GFP at its C-terminus (AtACHT1-5:GFP), to that of chloroplast-localized GFP (Chpst:GFP) and to the autofluorescence of chlorophyll (CHLR), to mitochondrion-localized GFP (Mito:GFP), to an ER-localized GFP (ER:GFP), and to cytoplasm localized GFP (Cyto:GFP) in A. thaliana protoplasts using confocal laser microscopy (Fig. 4A). To avoid mislocalization due to overaccumulation of expressed proteins we imaged only protoplasts displaying the earliest signal of GFP fluorescence, as previously done in our lab (Levitan et al., 2005). Interestingly, the fluorescence images of all the AtACHTs resembled that of the chloroplast-localized GFP and overlapped the chlorophyll autofluorescence. In contrast, none of the AtACHTs fluorescence paralleled that of GFP localized to the other plant organelles, suggesting that all assayed AtACHTs are targeted to the chloroplast. We chose two proteins, AtACHT1 and AtACHT4a, for further studies. First, in order to authenticate the protein localization, transgenic plants expressing AtACHT1 or AtACHT4a, each fused with YFP at their C-terminus, were generated and imaged using confocal laser microscopy.
The fluorescence images corroborated that both AtACHT1 and AtACHT4a are targeted to chloroplasts in planta and that they do not accumulate to detectible levels in other subcellular organelles (Fig. 4B). Second, to further study the subchloroplast localization of the proteins, we purified intact chloroplasts from transgenic plants, expressing the AtACHTs with HA 3 affinity tag at their C-terminus, and analyzed their soluble and membranal fractions by immunoblot analysis. The AtACHT1 and AtACHT4a proteins were found partitioned in both the stromal fraction, containing the large subunit of RuBisCO (LSU), and the thylakoid membranes, containing the D1 protein ( Fig. 4C), suggesting that they have a role in both compartments.

The AtACHTs show preference towards 2-Cys Prx A
The plastidial localization of the AtACHT proteins prompted us to examine their activity towards known targets of chloroplast Trxs, such as 2-Cys Prx A and MDH. The 2-Cys Prxs detoxify peroxides by their reduction. Several Trx-like proteins were shown to reduce 2-Cys Prx  (Fig. 5A). In contrast, at similar protein concentration the activity of AtTrx-f1 was close to background activity, and low activity was observed at four-fold higher protein concentration (data not shown). The higher redox potential of the ACHTs (around -240 mV, Fig.3) suggests that these proteins might be alternatively reduced by GSH. However, when the Prx activity assay was repeated in the presence of GSH rather than DTT, the proteins were not able to efficiently increase the peroxidase activity of At-2-Cys PrxA (Fig. 5B), suggesting that a dithiol rather than a monothiol is the preferred electron donor of the AtACHTs. The Trxs did not display any peroxidase activity in the absence of At-2-Cys PrxA (data not shown). The activation of MDH, which catalyzes the reduction of oxaloacetate into malate using NADPH as a cofactor, is dependent on the light-regulated reduction of two disulfide bonds by Trx (Schurmann and Buchanan, 2008). Though originally thought to be specifically lightregulated through Trx-m, Trx-f was later shown to be the most efficient activator of sorghum MDH (Collin et al., 2003;Collin et al., 2004). Activation assays were performed by incubating recombinant mature A. thaliana NADP-MDH (AtMDH) with either AtTrx-f1 or selected AtACHTs in the presence of DTT, and measuring the AtMDH enzymatic activity as oxidation of NADPH as described before (Jacquot and Buchanan, 1981). While 1 μM of AtTrx-f1 was sufficient to efficiently activate AtMDH under our experimental conditions, the same concentration of the AtACHTs did not appreciably activate AtMDH (Fig. 5C). By increasing the AtACHTs concentration to 20 μM a poor activation of AtMDH was observed (data not shown).
Thus, the AtACHTs showed high preference in their activity towards At-2-Cys PrxA than their activity towards AtMDH. suggesting that its expression is specific to pollen tissue (Pina et al., 2005). A more detailed study of these two genes is required to determine whether they have specialized roles in reproductive tissues. We further studied the expression of the two splice variants, AtACHT2a and AtACHT2b, by semiquantitative RT-PCR analysis and found differences in their tissue specificity (Fig. 6D). AtACHT2a transcripts accumulate to high levels in green tissues whereas AtACHT2b transcript amounts are even in all analyzed plant tissues, suggesting that AtACHT2a role is unique to chloroplasts whereas AtACHT2b might be required for all types of plastids.

The ACHT family is unique to plants
The plastid localization of AtACHTs suggested plant specific function. Thus, in order to determine whether this family is unique to plants, a BLAST search was made against all available sequences in NCBI and in specific databases of selected organisms. We found homologues from the green algae Chlamydomonas reinhardtii and Ostreococcus tauri, the moss Physcomitrella patens, and the higher plants Oryza sativa and Zea mays. Notably, the genomes of non-plant species as well as cyanobacteria or other prokaryotes do not seem to contain ACHT homologues, suggesting that similarly to the chloroplast Trx-f (Schurmann and Buchanan, 2008) they are plant specific and of eukaryotic origin. A phylogenetic analysis that included one Trx domain proteins with an atypical or classic vicinal dithiol active site showed that the AtACHTs and their homologues are clustered together on a unique node, whereas the other atypical Trxs are more closely related to the classic Trxs (Fig. 7). The finding of an AtACHT homologue in Ostreococcus (Fig. 7), one of the most ancient branches of the green lineage (Derelle et al., 2006), suggests that this family originated early before the split of higher plants from unicellular plants.
The phylogenetic analysis implies that the AtACHTs are subdivided into two classes ( Fig. 7). Class 1 contains AtACHTs 1, 2a, and 2b, whereas Class 2 includes AtACHTs 3, 4a, 4b, and 5. All Class 2 AtACHTs and their homologues lack the conserved tryptophan preceding the redox active site and, except for one Oryza protein, have an identical active site motif, CGGC.
All of them are typified by a large C terminal extension. Class 1 AtACHTs have C(G/A)SC active site motif and a shorter C-terminus relative to Class 2 proteins. Both Class 1 and 2 proteins contain additional cysteines which are conserved in all proteins and histidines which are conserved within each class (Fig. 1). The relevance of the conserved cysteines and histidines to possibly newly evolved activities of these proteins will have to be addressed in future studies.
The high degree of conservation of this protein family along the plant evolution course, their absence from non-plant organisms, and their chloroplast localization imply that they have a preserved role in plants. Their unique sequence elements differentiate them from the classic Trxs, and their higher redox potential raises the possibility of specialized redox function in plant chloroplasts.

DISCUSSION
The AtACHTs constitute a small family of chloroplast Trx-like proteins that display a redox midpoint potential that is significantly less reducing compared to the classic Trxs (Fig. 3), a redox active site with a different central dipeptide than the canonical sequence, and several additional conserved cysteine and histidine residues outside of the active site (Fig. 1). Previous experiments in which the central dipeptide sequence of one Trx family member was swapped with that of second Trx resulted in a profound effect on the activity of the Trx. Often, a shift in the redox properties of the mutated Trx towards those of the second Trx was observed (Krause and Holmgren, 1991;Lundstrom et al., 1992;Chivers et al., 1996;Huber-Wunderlich and Glockshuber, 1998;Mossner et al., ;Quan et al., 2007). Notably, the two Classes of AtACHTs differ in their central dipeptide; the Class 1 dipeptide is C(G/A)SC, and the Class 2 is CGGC ( Fig. 7). Yet, in spite of the difference in the central dipeptide, the redox midpoint potentials of selected members from both Class 1 and 2 AtACHTs were found to be almost identical (Fig. 3), suggesting that other residues in addition to the redox active site dipeptide might affect the redox midpoint potential as well. This assumption is further supported by the finding that the redox midpoint potential of different Trxs, containing the same canonical central dipeptide, may vary between -300 mV and -275 mV (Collin et al., 2003;Collin et al., 2004). Hence, the AtACHTs have evolved a redox active site with variable central dipeptide while maintaining a similar redox midpoint potential value that has been conceivably preserved by compensatory changes in other residues than the central dipeptide.
Both prokaryotic and eukaryotic type Trxs contain a highly conserved tryptophan residue preceding the active site. In A. thaliana fifty-seven out of the sixty-four Trx-related gene products contain this tryptophan (Meyer et al., 2007). Three out of the seven A. thaliana gene products missing the conserved tryptophan residue are the Class 2 AtACHTs (Fig. 1) activity of the Trx towards in vitro substrates but not its redox midpoint potential (Krause and Holmgren, 1991;Krimm et al., 1998;Menchise et al., 2001). Here, the tryptophan-less AtACHT4a was less active than the tryptophan-containing AtACHTs 1 and 2a in the standard insulin reduction assay (Fig. 2). But, similarly to the tryptophan-less CDSP32 and NTRC Trxlike proteins, as well as the tryptophan-containing AtACHT1 and 2a, AtACHT4a showed high activity in assays containing the chloroplast enzyme At-2-Cys PrxA (Fig. 5). These findings suggest that in spite of the importance of the tryptophan to the activity of a Trx, its presence or absence from a naturally occurring Trx is not a good predictor of its activity against specific substrates.
A plastidial function for the AtACHTs is implicated by their localization (Fig. 4) and by their phylogeny (Fig. 7), which is unique to the plastid-containing viridiplantae. What might be, then, the unique function of the AtACHTs among the twenty or so Trx-like proteins shown to reside in plastids? The first clue might be found in the expression pattern of the AtACHTs.
AtACHT2a and AtACHT4a-4b express mainly in photosynthetic tissue (Fig. 6), suggesting that they might function in parallel to the classic Trxs, such as Trx-f and Trx-m. In contrast, AtACHT1 and AtACHT2b transcripts accumulate quite evenly in green and non-green tissues, suggesting a more generalized role that is not limited to the chloroplast (Fig. 6). The expression of the AtACHT5 and AtACHT3 transcripts appears to be unique. AtACHT5 transcript is enriched in floral tissues and accumulates to very high levels in the double mutant of regulators of floral meristem development (Sup/ap1) (Fig. 6). Whereas, the AtACHT3 transcript is highly specific to pollen (Fig. 6), suggesting a more specialized function for these two proteins.
A second clue might be found in the activity in vitro of the AtACHTs, which relatively to AtTrx-f1, reacted with high efficiency with At-2-Cys PrxA and poorly with AtMDH (Fig. 5).
The canonical site-containing AtTrx-x and AtTrx-y were also found to be inefficient activators of sorghum MDH and to react efficiently with peroxiredoxins in vitro (Collin et al., 2003;Collin et al., 2004). Based on their in vitro activities it was suggested that Trx-x and Trx-y might function specifically in resisting oxidative stress, rather than in enzyme regulation (Collin et al., 2003). The similar in vitro activity profile of the AtACHTs might hint at similar roles. Furthermore, the similar partitioning between the stroma and the thylakoids of the AtACHTs (Fig. 4) and 2-Cys PrxA (Dietz et al., 2006) suggests that the two activities might be linked.
Another clue to the role of AtACHTs might be found in their redox midpoint potential which is of an intermediate value (Fig. 3) between the two extremes of the oxidative DsbA (Zapun et al., 1993;Aslund et al., 1997) and the reductive Trx-m (Hirasawa et al., 1999), and is almost identical to the -241 mV value of the soluble Trx-like γ-domain of the membrane bound bacterial DsbD (Collet et al., 2002). Notably, DsbD role is to provide electrons to redox pathways supporting transient reduction for the isomerization of misformed disulfides (Nakamoto and Bardwell, 2004;Porat et al., 2004). Hence the intermediate value of the redox midpoint potential of the AtACHTs might suggest adaptability to function in reversible reactions, such as those modulating the redox state of regulatory proteins (Buchanan, 1980;Trebitsh et al., 2000). Yet, it is important to note that while the midpoint redox potential value of a protein likely reflects its adaptation to oxidative or reductive type reactions, its in vivo role is not necessarily correlated with its redox potential (Ortenberg and Beckwith, 2003) and might depend on additional factors, such as dimerization, localization, or interaction with accessory proteins. Hence, further genetic and molecular studies are required in order to elucidate the exact redox function of the individual AtACHTs.

Identification and cloning of the AtACHT proteins from Arabidopsis thaliana
The protein sequence of AtTrx-f1 was used to search the A. thaliana database for Trx-like proteins containing non-canonical active site using BLASTP program at NCBI (National Center for Biotechnology Information, http://www.ncbi.nlm.nih.gov). Total RNA was extracted using Tri-reagent (MRC, Cincinnati, OH, USA). The RNA was subjected to reverse transcription using Superscript II (Invitrogen, Rhenium Ltd, Jerusalem, Israel) and oligo-dT. PCR reaction were done with gene specific primers.

Generation of protein multiple alignment and phylogenetic trees
Protein multiple alignments were generated using the ClustalX program (Thompson et al., 1997) and viewed through Jalview (Clamp et al., 2004). Secondary structure elements were predicted www.plantphysiol.org on August 29, 2017 -Published by Downloaded from Copyright © 2008 American Society of Plant Biologists. All rights reserved.
To identify alternative splicing variants, the genome sequence of A. thaliana was analyzed for gene structure positions (exons and introns borders). The borders between expressed sequences and introns were compared to the cDNA sequences of AtACHT2a and AtACHT2b, which were isolated by RT-PCR. The existence of the splice variants was predicted also by TAIR (The Arabidopsis Information Resource; www.arabidopsis.org).

Expression and purification of recombinant proteins
The cDNAs encoding AtACHTs, AtMDH (At5g58330) and At 2-Cys PrxA (At3g11630) without their putative transit peptide (primer sequences are listed in Table S1), as predicted by TargetP (Emanuelsson et al., 2000), were cloned into the expression vector pQE30 (Qiagen, Eldan Ltd., Israel), resulting in N-terminal His-tagged proteins. M15 cells were transformed with the plasmids and protein expression induction was performed according to the manufacturer instructions. The proteins were purified using HiTrap nickel affinity column (HiTrap Chelating HP, Amersham Pharmacia, Piscataway, NJ, USA) according to the manufacturer instructions, and dialyzed overnight against 2000 volumes of Dialysis Buffer, 1x phosphate saline buffer, 20% glycerol, 2 mM EDTA, and 5 mM β-mercaptoethanol for the AtACHTs, or the dialysis buffer without EDTA and β-mercaptoethanol for AtMDH and At 2-Cys PrxA. The proteins were purified to homogeneity of above 95% as judged by Coomassie Brilliant Blue gel staining.

Insulin turbidity assay
Recombinant proteins (5 μM) were used in the insulin reduction assay, as previously described (Holmgren, 1979). The turbidity of the reduced insulin chains was recorded using Synergy HT microplate reader at 650 nm every 30 sec for 2 hours.

Determination of redox midpoint potential
The redox potential was evaluated using the thiol-labeling reagent monobromobimane (mBBr), as previously described (Hirasawa et al., 1999). Different ratios of reduced:oxidized DTT or glutathione in tricine buffer pH 7.0 were used for AtTrx-f1 and AtACHT proteins respectively.
The redox potentials were calculated by curve fitting to the Nernst equation using GraphPad Prism version 4.0, and redox midpoint potential values of -330 mV for DTT (Hirasawa et al., 1999) and -240 mV for glutathione (Aslund et al., 1997). Experiments were repeated 3 times and the obtained redox midpoint potential values were independent of equilibrium time.

Determination of subcellular localization
Each AtACHT ORF was ligated through SmaI sites, into a puc18-GFP5 containing vector, yielding a fusion protein upstream and in frame to GFP5 ORF, under the control of the cauliflower mosaic virus 35S promoter. Control fusion proteins were prepared as well and included the small subunit of RuBisCO (kindly provided by Yoram Eyal, Agricultural Research Organization, Volcani Center), the chitinase ER marker (kindly provided by Jean-Marc Neuhaus, University of Neuchatel, Switzerland), the mitochondrial AtTrx-o1 (At2g35010) and the mature form of the GFP protein. The transient expression assays were performed as described before (Sheen, 2002). Protoplasts were viewed 16 hours after transformation.
For transgenic plants generation, AtACHT1 and AtACHT4a were ligated upstream and in frame to HA 3 or YFP tags. The fragments were cloned into pART 7 vector, under the control of the 35S promoter and 35S terminator. As a control, the YFP ORF was ligated as well. Not I digested fragments were ligated into the Binary vector pBART, which confer glufosinate (BASTA) resistance in plants. The binary plasmids were introduced into Agrobacterium tumefaciens through electroporation. Plant transformation was made by the floral dip method (Clough and Bent, 1998).
Fluorescence images were obtained as described before (Levitan et al., 2005). Because transient expression under the 35S promoter may over produce the expressed protein, resulting in its mislocalization, we imaged only viable protoplasts displaying the earliest signal of GFP fluorescence, and performed western blots to confirm the expression of the full-length fusion proteins (data not shown).

NADP malate dehydrogenase (MDH) activity assay.
DTT dependant activation of AtMDH by 1 μM of AtTrx-f1 or of the different AtACHTs was carried out with 1.5 μM AtMDH and 10 mM DTT in 100 mM TRIS-HCl pH 7.9. At fixed times, 20 μl aliquots of the activation mixture were used to determine the activity of the AtMDH by following the initial rate of consumption of NADPH at room temperature, in a 1 ml assay mixture containing 140 μM NADPH and 750 μM oxaloacetate in 100 mM TRIS-HCl pH 7.9.
The oxidation of NADPH was followed spectroscopically at 340 nm. One unit of MDH activity is defined as the amount oxidizing 1 μmol NADPH / min., and corresponds to an absorbance change of 6.22.

Gene expression analysis
The MPSS (http://mpss.udel.edu/at/java.html) and the Genevestigator database (www.genevestigator.ethz.ch/at) were used to evaluate differential gene expression. Validation of the main results was done using semiquantitative RT-PCR (sqRT-PCR). Total RNA was extracted and equal amounts of RNA (2 μg) were subjected to reverse transcription using Superscript II (Invitrogen) and oligo-dT. PCR reactions were done with gene specific primers (as listed in Table S1). As a control, the Actin and Tubulin genes were amplified as well. Selected bands were excised, purified from the gel and sequenced to authenticate their identity.

Figure 2. The activity of AtACHT proteins in the standard insulin assay
AtACHT1 (triangles), AtACHT2a (diamonds) and AtACHT4a (rectangular) were tested for their ability to catalyze the reduction of insulin disulfides. The precipitation rate of the reduced insulin by the different AtACHTs was compared to reactions catalyzed by AtTrx-f1 as positive control (circles) or DTT alone as negative control (asterisk).

Figure 3. The AtACHT proteins have higher redox midpoint potentials in comparison to canonical Trxs
AtACHT1, AtACHT2a and AtACHT4a proteins were incubated at defined redox potentials (Eh) values at pH 7.0, established by mixing various ratios of reduced and oxidized glutathione. The fluorescent thiollabeling reagent mBBr was added, and its fluorescence was measured. Estimated redox midpoint potential, standard error and R 2 values are presented. stromal large subunit of RuBisCO (LSU) and the membranal D1 protein were blotted for quality assessment of the fractionation procedure.