Abolition of Peroxiredoxin-5 Mitochondrial Targeting during Canid Evolution

In human, the subcellular targeting of peroxiredoxin-5 (PRDX5), a thioredoxin peroxidase, is dependent on the use of multiple alternative transcription start sites and two alternative in-frame translation initiation sites, which determine whether or not the region encoding a mitochondrial targeting sequence (MTS) is translated. In the present study, the abolition of PRDX5 mitochondrial targeting in dog is highlighted and the molecular mechanism underlying the loss of mitochondrial PRDX5 during evolution is examined. Here, we show that the absence of mitochondrial PRDX5 is generalized among the extant canids and that the first events leading to PRDX5 MTS abolition in canids involve a mutation in the more 5′ translation initiation codon as well as the appearance of a STOP codon. Furthermore, we found that PRDX5 MTS functionality is maintained in giant panda and northern elephant seal, which are phylogenetically closely related to canids. Also, the functional consequences of the restoration of mitochondrial PRDX5 in dog Madin-Darby canine kidney (MDCK) cells were investigated. The restoration of PRDX5 mitochondrial targeting in MDCK cells, instead of protecting, provokes deleterious effects following peroxide exposure independently of its peroxidase activity, indicating that mitochondrial PRDX5 gains cytotoxic properties under acute oxidative stress in MDCK cells. Altogether our results show that, although mitochondrial PRDX5 cytoprotective function against oxidative stress has been clearly demonstrated in human and rodents, PRDX5 targeting to mitochondria has been evolutionary lost in canids. Moreover, restoration of mitochondrial PRDX5 in dog MDCK cells, instead of conferring protection against peroxide exposure, makes them more vulnerable.


Introduction
Due to the presence of the electron transport chain, mitochondria are particularly exposed to oxidative attacks caused by reactive oxygen species (ROS) and reactive nitrogen species (RNS). To scavenge ROS/RNS, the mitochondrial matrix is equipped with antioxidant enzymes. Among them, manganese superoxide dismutase (MnSOD or SOD2) is responsible for the dismutation of the superoxide radical (O 2 N-) to hydrogen peroxide (H 2 O 2 ). H 2 O 2 is subsequently reduced to water by mitochondrial peroxidases, i.e. glutathione peroxidase 1 (GPX1), glutathione peroxidase 4 (GPX4), peroxiredoxin-3 (PRDX3) and peroxiredoxin-5 (PRDX5) [1].
In human, the complex subcellular distribution of PRDX5 is a consequence of the existence of two protein isoforms. The latter are encoded by a single gene containing alternative transcription start sites and two alternative in-frame translation initiation sites ( Figure 1). Translation at the more 39 start codon produces a short PRDX5 form (S-PRDX5) which is found in the cytosol, the peroxisomes and the nucleus [11]. Peroxisomal targeting is linked to the presence of a weak C-terminal peroxisomal targeting signal (PTS1) [2,7]. Translation at the more 59 start codon produces a long PRDX5 form (L-PRDX5) containing an additional Nterminal mitochondrial targeting sequence (MTS) or presequence [2,11]. This MTS is cleaved after mitochondrial import, releasing the mature mitochondrial PRDX5 which is identical to S-PRDX5. Moreover, mitochondrial targeting of PRDX5 seems to be a conserved feature in animal kingdom. So far, the mitochondrial localization of the protein was experimentally verified in human, rat, mouse, hamster, bovine and fruitfly [2,[12][13][14][15][16]. Furthermore, presequences were also identified in orthologous PRDX5 of invertebrate species [16][17][18][19] (see also Figure 2). Some of these invertebrate species belong to basal metazoan phyla, i.e. porifera and placozoa. However, we reported recently that mitochondrial PRDX5 is absent in some mammalian species, such as in pig [19]. Moreover, according to sequence analysis, we also predicted that mitochondrial PRDX5 would be absent in dog ( Figure 2) [11]. While the present work was in progress, the absence of dog L-PRDX5 was also evidenced by [20].
Here, we confirm that dog PRDX5 is not addressed to mitochondria and we examined the molecular basis responsible for the abolition of mitochondrial targeting. Furthermore, we show that the loss of mitochondrial PRDX5 is generalized throughout the Canidae family. Finally, the functional consequences of the restoration of mitochondrial PRDX5 in dog MDCK cells were also investigated.

Mitochondrial PRDX5 is Absent in Domestic Dog
The absence of mitochondrial PRDX5 in domestic dog was demonstrated by subcellular fractionation of a liver homogenate. Marker analysis on post-nuclear (E), cytosolic (S), mitochondrial (Mito) and peroxisomal (Perox) fractions was performed by Western blotting ( Figure 3A). The results indicated that the cytosolic, mitochondrial and peroxisomal fractions are enriched in Hsp90 (cytosolic marker), COX4 (mitochondrial marker) and catalase (peroxisomal marker) respectively. Evaluation of PRDX5 content by Western blotting revealed the presence of PRDX5 in cytosolic and peroxisomal fractions but PRDX5 was absent from the mitochondrial fraction.
To confirm the absence of mitochondrial PRDX5 in domestic dog, PRDX5 expression was examined using immunofluorescence in the MDCK (dog) cell line. PRDX5 was detected in the cytosol but no colocalization with the mitochondrial marker was observed ( Figure 3B-D). PRDX5 detection by immunofluorescence also evidenced the peroxisomal localization of the protein, as illustrated by colocalization with peroxisomal catalase ( Figure 3E-G).
Analysis of domestic dog PRDX5 sequences from GenBank databases (GeneID: 476032) revealed the presence of a single translation initiation codon. This ATG codon corresponds to the second translation initiation site of human PRDX5 that produces the short form of the protein (S-PRDX5). The first translation initiation codon that enables the synthesis of the long form of PRDX5 (L-PRDX5) was not present in the genomic sequence of domestic dog.

L-PRDX5 is Absent in Other Canids
To date, giant panda (Ursidae) is the phylogenetically closest related species to domestic dog (Canidae) for which PRDX5 mRNA sequence is available in Genbank databases. These two species belong both to the Caniformia phylum ( Figure 4A). Analysis of the 59 region of giant panda PRDX5 mRNA (GenBank ID: XM_002916664) revealed the existence of two in-frame translation initiation codons (AUG+1 and AUG+172). These two codons are presumed to lead to the translation of both L-PRDX5 and S-PRDX5. Moreover, subcellular prediction programs predicted a mitochondrial distribution for giant panda L-PRDX5 protein (TargetP: 0.804 and MitoProt II: 0.859 compared with human PRDX5 TargetP: 0.830 and MitoProt II: 0.832; [21,22]). Since the translation initiation codon of L-PRDX5 appears to be maintained in giant panda but not in domestic dog, the loss of this translation initiation site was likely to have occurred in additional species which are phylogenetically positioned between domestic dog and giant panda, like the other canids ( Figure 4A). Canidae family contains three major phylogenetic groups: the fox-like canids, the South American canids and the wolf-like canids ( Figure 4B). Additionally, several canids do not belong to any of the three main clades, such as the gray fox, which was shown to be more primitive [23]. To assess the presence of a sequence encoding an MTS in PRDX5 of other canids, the PRDX5 59 region of gray fox, raccoon dog (fox-like canid), red fox (fox-like canid), maned wolf (South American canid) and arctic wolf (wolf-like canid) was sequenced. Moreover, PRDX5 59 region was also sequenced in northern elephant seal (Phocidae) and American black bear (Ursidae). To obtain the sequence of PRDX5 59 region in these species, the sequence was first PCR-amplified using genomic DNA as template ( Figure 4C). The primers were designated to map on regions sharing 100% identity between giant panda and domestic dog sequences. PCR products were cloned in pCR2.1 vector and were subsequently sequenced. Genbank accession numbers are presented in Table 1. The alignment of the 59 PRDX5 sequences revealed that the ATG+1 of giant panda, northern elephant seal and American black bear PRDX5 corresponds to a GTG codon in all the studied canids ( Figure 4D). Moreover, in gray fox and fox-like canids, a STOP codon (position290) is present in the same reading frame as the ATG+1. Furthermore, in the South American and wolf-like canids PRDX5, the GTG-171 codon is not in the same reading frame as the ATG+1 codon. In the nucleotidic alignment, this shift in the reading frame corresponds to a gap between nucleotides264 and263. It is also worth noting that two slightly different sequences, corresponding to two different alleles, were found in gray fox and American black bear PRDX5 59 region (Table 1).

A functional MTS is Present in Giant Panda and Northern Elephant Seal PRDX5
Given that giant panda and northern elephant seal PRDX5 present two in-frame ATG codons, it was interesting to assess whether the region comprised between these codons encodes a functional MTS. To this end, two pcDNA3.1 plasmids containing the sequence encoding the 58 N-terminal residues of giant panda or northern elephant seal PRDX5 fused with the green fluorescent protein (GFP) were generated ( Figure 5A

Human L-PRDX5 Makes MDCK Cells More Vulnerable to Acute Oxidative Stress
Since PRDX5 is absent in dog mitochondria, it was interesting to assess the functional consequences of the restoration of mitochondrial localization of the protein in MDCK (dog) cells. pEF-BOS was used as mammalian expression vector ( Figure 6A). Figure 6B presents the proteins encoded by the constructs used for MDCK transfection. pEF-BOS-mito-PRDX5 and pEF-BOS-cyto-PRDX5 vectors encode human L-PRDX5 (Mito-hum) and S-PRDX5 (Cyto-hum) respectively and were described previously [13]. pEF-BOS-mito-PRDX5-C47A and pEF-BOS-cyto-PRDX5-C47S vectors encode the enzymatically inactive human L-PRDX5 (Mito-hum-C47A) and S-PRDX5 (Cyto-hum-C47S) respectively. The two latter were used in experiments which are described in the following section.
After transfection of MDCK cells with the empty expression vector (control cells), with pEF-BOS-mito-PRDX5 or with pEF-BOS-cyto-PRDX5 vectors, stable clones were selected. The clones were named MDCK control, MDCK mito-hum or MDCK cytohum, according to the name of the proteins encoded by the expression vectors. Overexpression of PRDX5 was quantified by immunoblotting ( Figure 6C). Immunoblots of total cell lysates revealed a 15-to 30-fold overexpression compared to the MDCK control cells, and a 1.5-to 3-fold higher PRDX5 content compared to human SH-SY5Y cells ( Figure 6D). The latter cells contain endogenous mitochondrial PRDX5 [24]. The intracellular localization of PRDX5 was verified by immunofluorescence ( Figure 6E). In the MDCK mito-hum cells, PRDX5 was largely located in mitochondria, while MDCK cyto-hum cells strongly expressed PRDX5 in the cytosol. Incorporation of puromycin resistance gene in the genome of MDCK control cells was verified by PCR on genomic DNA (data not shown).   [23]. Canidae family contains three major phylogenetic groups, which are indicated in pink for the fox-like canids, green for the South American canids and blue for the wolf-like canids. Additionally, several canids do not belong to any of the three main clades, such as the gray fox (yellow), which was shown to be more primitive. Adapted from [23].

(C) Schematic
The expression of TXN2, the known mitochondrial physiological reductant of PRDX5, was also assessed to confirm that MDCK mitochondria have the proper system to reduce efficiently human PRDX5. Immunofluorescence detection of TXN2 evidenced that the enzyme is indeed present in MDCK mitochondria ( Figure S1). Moreover, comparison of endogenous TXN2 expression levels between MDCK and human SH-SY5Y cells using Western blotting, revealed that expression levels are comparable in the two cell lines ( Figure S1).
The MDCK clones were exposed to acute oxidative stress with increasing concentrations of H 2 O 2 or t-BHP ( Figure 7). In basal conditions, the MDCK cells expressing human PRDX5 exhibited no significantly different LDH release compared to control cells. Cell death increased with H 2 O 2 and t-BHP exposure. Surprisingly, cell death was significantly more important in MDCK mito-hum cells compared to control cells at 15 mM H 2 O 2 and 70-80 mM t-BHP. On the contrary, cell death was significantly decreased in MDCK cyto-hum cells compared to control cells at high concentrations of H 2 O 2 and t-BHP.

Increased Cell Death of MDCK Mito-hum Cells During Acute Oxidative Stress is not Linked to PRDX5 Peroxidatic Activity
Since human L-PRDX5 provoked increased cell mortality during acute peroxide-induced oxidative stress in MDCK cells, we assessed whether this deleterious effect was linked to PRDX5 peroxidatic activity. This latter depends on the presence of a peroxidatic cysteine (Cp, see Figure 1). Therefore, MDCK mitohum-C47A cells were generated, expressing human L-PRDX5 with the Cp replaced by an alanine (C47A mutation; Figure 6B). This mutation inactivates the peroxidatic activity of the enzyme.
MDCK cyto-hum-C47S cells were also generated, expressing the enzymatically inactive human S-PRDX5 with the C47S mutation. After transfection and puromycin resistance selection, stable clones were selected and overexpression was quantified by immunoblotting ( Figure 6C). Immunoblots of total cell lysates revealed a 10-to 30-fold overexpression compared to MDCK control cells, and a 1.5-to 3-fold higher PRDX5 content compared to human SH-SY5Y cells ( Figure 6D). The intracellular localization of PRDX5 was verified in MDCK clones by immunofluorescence ( Figure 6E). In the MDCK mito-hum-C47A cells, PRDX5 was largely located in the mitochondria, while MDCK cyto-hum-C47S cells strongly expressed PRDX5 in the cytosol.
MDCK cell lines were exposed to acute oxidative stress induced by H 2 O 2 or t-BHP exposure (Figure 7). Cell death was significantly more important in MDCK mito-hum-C47A cells compared to control cells at 15 mM H 2 O 2 and 70-80 mM t-BHP ( Figure 7A and 7C). On the other hand, transient overexpression of a nonspecific mitochondrial protein (GFP) in MDCK mitochondria did not make dog cells more sensitive to 7.5-15 mM H 2 O 2 or 60-80 mM t-BHP ( Figure S2) and expression of Cyto-hum-C47S protein had no effect on cell viability following H 2 O 2 or t-BHP exposure.

Discussion
In this work, we show that mitochondrial PRDX5 (L-PRDX5) is absent in domestic dog although cytosolic/peroxisomal PRDX5 (S-PRDX5) is conserved. Sequence analysis of the PRDX5 59 region revealed that the abolition of L-PRDX5 in domestic dog results from the loss of one nucleotide as well as the mutation of the first ATG initiation codon. To determine whether the abolition of mitochondrial PRDX5 occurred only in dog, we analyzed the PRDX5 59 region of several species which are phylogenetically (see Figure 4) closely related to dog and Canidae: giant panda (Ursidae), American black bear (Ursidae) and northern elephant seal (Phocidae). We found that, unlike in domestic dog, the translation initiation codon allowing the synthesis of L-PRDX5 is maintained in these three species. Moreover, we experimentally confirmed that the 58 N-terminal residues of giant panda and northern elephant seal L-PRDX5 are able to target the GFP reporter protein into mitochondria. This indicates that the N-terminal region of giant panda and northern elephant seal L-PRDX5 is a functional MTS. Since PRDX5 MTS is abolished in domestic dog but neither in giant panda nor in northern elephant seal, we investigated the existence of an MTS in PRDX5 in other canids. It appeared that the first ATG codon is mutated in a GTG codon in PRDX5 in all canids that were examined. From these results, we conclude that the loss of mitochondrial PRDX5 is generalized among the extant canids. Moreover, in gray fox, raccoon dog and red fox PRDX5, a STOP codon at position290 is present in the same reading frame as the diagram of HSPC152/TRMT112 and PRDX5 genes and cloning strategy. Exons I from both genes are detailed. 59 UTRs are indicated by open boxes. Gray box highlights the sequence encoding the potential PRDX5 MTS. Conserved translation start sites are indicated in black. The potential translation start site enabling the translation of L-PRDX5 is indicated in red. P1 and P2 indicate the positions to which the PCR primers (green arrows) map. The primers were designated to map on regions sharing 100% identity between giant panda and domestic dog sequences. (D) ClustalX 2.0.7. nucleotidic sequence alignment of PRDX5 59 region from nine Caniformia. Asterisks (*) indicate identities. Sequences encoding a potential MTS are highlighted in gray. Translation start sites enabling the translation of L-PRDX5 and S-PRDX5 are highlighted in red and black respectively. STOP codons being in-frame with the ATG+1 codon are underlined. Gaps are boxed. Species names are written in brown for Phocidae, gray for Ursidae, yellow for gray fox, red for the fox-like canids, green for the South American canids and blue for the wolf-like canids. Only one of the two alleles    translation initiation codon of S-PRDX5 (TCG/TAG mutation). Thus, in these more remote canids, two factors impede L-PRDX5 synthesis: the absence of L-PRDX5 translation initiation codon and the presence of a STOP codon. Taken together, sequence analyses suggest that the 59 region of PRDX5 in canids corresponds to a sequence that was able to encode a presequence ancestrally. The first events leading to PRDX5 MTS abolition in canids would have involved the mutation of the first translation initiation codon (ATG/GTG mutation) as well as the apparition of a STOP codon (TCG/TAG mutation). However, from our sequence analyses, we cannot conclude which of these two events occurred first during evolution. Furthermore, in South American canids (maned wolf) and wolf-like canids (domestic dog and arctic wolf), the ATG/ GTG mutation would have been maintained but the STOP codon would further have evolved to a CAG codon. Also, in the ancestral PRDX5 MTS from these species, a single nucleotide would have been lost. Finally, our sequence analyses indicate that the abolition of PRDX5 MTS would have occurred after the divergence of Canidae from the other Caniformia members. The Canidae family diverged from other Caniformia about 50 million years ago (mya), but extant canids have diverged from a common ancestor within the last 10 million years [25]. Therefore, the loss of mitochondrial PRDX5 in canids would have occurred between 50 and 10 mya. In a recent report, we have shown that domestic pig has also lost mitochondrial PRDX5 according to a similar molecular mechanism, namely by the loss of the translation initiation codon of L-PRDX5 [19]. Since mitochondrial PRDX5 is maintained in bovine, we concluded that PRDX5 MTS was lost during Cetartiodactyla evolution. Therefore, the loss of L-PRDX5 in canids and domestic pig results from two distinct events during mammalian evolution, i.e. one having occurred during Caniformia diversification, the other taking place during Cetartiodactyla evolution. Given that PRDX5 distribution was analyzed in a limited number of species so far, we cannot exclude that its subcellular distribution would have changed at more occasions during the evolution of mammals. Accordingly, it would be interesting to study PRDX5 subcellular distribution in additional phylogenetic groups. To our knowledge, few enzymes have been reported to have changed their subcellular distribution during mammalian evolution. One example concerns the intermediary metabolic enzyme alanine:glyoxylate aminotransferase (AGT), whose subcellular localization was reported to change during mammalian evolution [26,27]. Interestingly, the molecular mechanism responsible for the loss of mitochondrial targeting in AGT and PRDX5 is very similar.
The abolition of PRDX5 MTS in canids shows that mitochondrial PRDX5 is not essential in canid species. Some of these species have encountered high evolutionary success, like the raccoon dog or the red fox. It appears thus that the absence of mitochondrial PRDX5 does not impair species' ability to meet a huge evolutionary success. The abolition of mitochondrial PRDX5 in canids is nevertheless surprising. Indeed, an MTS is conserved in various other vertebrate and invertebrate species [2,[12][13][14][15][16][17][18][19], even belonging to basal metazoan phyla, like placozoa and porifera (see Figure 2). Moreover, the cytoprotective function of mitochondrial PRDX5 was described in several studies. For example, it has been shown that overexpression of L-PRDX5 confers resistance to peroxide or drug-induced DNA damage and cell death [13,[28][29][30][31]. The abolition of L-PRDX5 is unexpected since mitochondria are particularly exposed to oxidative attacks caused by ROS and RNS [1] but we also confirm here that PRDX3 is conserved in dog mitochondria (Supporting Information). Since mitochondrial PRDX5 is absent in dog, it was interesting to investigate the functional consequences of the restoration of L-PRDX5 in dog cells. In order to evaluate the ability of mitochondrial PRDX5 to afford antioxidant protection in dog during oxidative stress, we generated MDCK (dog) cell lines expressing human mitochondrial PRDX5. Moreover, to compare PRDX5 ability to exert its cytoprotective effect according to subcellular localization, we also generated MDCK expressing human cytosolic/peroxisomal PRDX5. Surprisingly, PRDX5 expression in MDCK mitochondria resulted in an increased cell mortality in cells exposed to high peroxide concentrations, while PRDX5 expression in MDCK cytosol/peroxisomes afforded a greater resistance to acute oxidative stress. As mentioned before, L-PRDX5 differs from S-PRDX5 by the presence of an additional N-terminal MTS. However, this MTS is cleaved after mitochondrial import, releasing a functionally active 17 kDa mature mitochondrial PRDX5 which is identical to S-PRDX5 [6]. In MDCK mito-hum cells, no PRDX5 signal at higher molecular weight was detected by Western blotting, indicating that the MTS was removed efficiently. Therefore, it appears that mitochondrial and cytosolic/peroxisomal human PRDX5, although being identical, do not equally protect MDCK cells, depending on their subcellular localization. Moreover, PRDX5 does not only appear to be an ineffective antioxidant protective enzyme in MDCK mitochondria-which would have resulted in an unchanged cell viability during oxidative stress-but this protein seems also to gain cytotoxic properties under acute oxidative stress, resulting in an increased cell mortality. Furthermore, given that MDCK mitohum cells exhibit an increased cell death during oxidative stress but not under basal conditions, one would have hypothesized that the sensitization was linked to PRDX5 peroxidatic activity. However, MDCK cells expressing enzymatically inactive mitochondrial PRDX5 exhibited also a higher cell mortality following oxidative stress, indicating that human PRDX5 provokes deleterious effects in MDCK mitochondria independently of its enzymatic activity. Thus, it appears that the protein, rather than its peroxidatic activity, is inappropriate in the MDCK mitochondrial matrix during acute oxidative stress. At this stage, it is difficult to establish which molecular mechanism is responsible for the peroxide-induced sensitization of the MDCK mito-hum cells. We could hypothesize that PRDX5 cytotoxic property is linked to a redox-dependent interaction with another MDCK mitochondrial protein, impairing the function of the binding partner. Interactions of PRDXs with various proteins have been described. Such interactions have been shown to modulate the functions of the binding partners [32]. However, so far, little is known about PRDX5 interactions with mitochondrial proteins. Moreover, since expression of human PRDX5 in MDCK mitochondria led to high overexpression (Fig. 6C), a non-specific protein effect could have been considered but overexpression of GFP in MDCK mitochondria did not increase cell sensitivity to H 2 O 2 and t-BHP ( Figure  S2). It must be also noted that, in a previous study, we used the same expression vectors to assess the cytoprotective function of human L-PRDX5 and S-PRDX5 in Chinese hamster ovary (CHO) cells [13]. In these cells, which contain endogenous mitochondrial PRDX5, both L-PRDX5 and S-PRDX5 afforded antioxidant protection during acute oxidative stress. It appears thus that human L-PRDX5 triggers deleterious effects during acute oxidative stress in cells lacking endogenous mitochondrial PRDX5, while the same protein confers antioxidant resistance in cells possessing endogenous mitochondrial PRDX5. This led us to suggest that, despite the fact that PRDX5 cytoprotective function has been clearly demonstrated by several studies in human and rodents, the conservation of mitochondrial PRDX5 would not confer necessarily an advantage in all mammalian species. Also, although the loss of L-PRDX5 in canids could have occurred independently of this phenomenon, it is tempting to speculate that the cytotoxic property of L-PRDX5 in mitochondria is at the origin of the abolition of mitochondrial PRDX5 in canids.
Interestingly, analysis of PRDX5 sequences from different animal species reveals that the non-conservation of PRDX5 mitochondrial targeting represents only a fraction of PRDX5 subcellular variability that is found through animal kingdom (see Figure 2). Indeed, in addition to the lack of PRDX5 MTS in some mammals, the PTS1, responsible for peroxisomal localization, appears also to be abolished in some animal species. Moreover, the presence of the second translation initiation codon, permitting the translation of S-PRDX5, appears to be absent in some species, like in bay scallop. In these species, only L-PRDX5 is likely to be translated. Taken together, PRDX5 sequence analyses suggest that PRDX5 subcellular relocalizations occurred at many occasions during animal evolution. Whether this variability permits an adaptation to specific needs, metabolic pathways or life environments remains a matter of speculation.
To conclude, the present study shows that mitochondrial PRDX5 has been lost during canid evolution. Our results suggest that the abolition of mitochondrial PRDX5 in the Canidae family would have occurred between 50 and 10 mya and involves the mutation of the first translation initiation codon of PRDX5 as well as the appearance of a STOP codon in the ancestral MTS. Finally, our work shows also that the restoration of mitochondrial PRDX5 in MDCK cells, even devoid of its peroxidatic activity, triggers deleterious effects during acute oxidative stress conditions.

Dog Liver Fractionation
Fresh liver was collected from beagle dog under the guidelines of the European Convention on the Protection of Vertebrate Animals [34]. Dog liver was homogenized and centrifuged to obtain nuclear fractions (N) and postnuclear supernatants (E). Efraction was further fractionated by differential centrifugation into heavy mitochondrial (M), light mitochondrial (L), microsomal (P) and cytosolic (S) fractions. Subfractions of the L-fraction were obtained by centrifugation through a Percoll gradient. Marker enzymes and protein content were measured in each fraction. Based on glutamate dehydrogenase (GDH, mitochondrial marker) and catalase (peroxisomal marker) activities, Percoll fractions 2 and 10 were chosen to represent mitochondrial (Mito) and peroxisomal (Perox) fractions respectively. Experimental procedures have been described previously [35,36].

Sequencing the 59 Region of PRDX5 from Caniformia Species
The species examined in this experiment are the following: arctic wolf (Canis lupus arctos), American black bear (Ursus americanus), domestic dog (Canis lupus familiaris), giant panda (Ailuropoda melanoleuca), gray fox (Urocyon cinereoargenteus), maned wolf (Chrysocyon brachyurus), northern elephant seal (Mirounga angustirostris), raccoon dog (Nyctereutes procyonides) and red fox (Vulpes vulpes). Arctic wolf blood sample was kindly provided during a routine blood checkup by Dr. JC Bertho, zoological park ''Le Monde Sauvage'', Aywaille, Belgium. American black bear heart sample has been used in [38] and was generously given by Pr. MAG Van Der Heyden (University Medical Center Utrecht, The Netherlands). Gray fox and maned wolf DNA samples, described in [23], were kindly provided by Pr. RK Wayne (University of California, USA). Northern elephant seal blood sample has been used in [39] and was provided by C Louis and Pr. C Debier (Université catholique de Louvain, Belgium). Tissue samples of raccoon dog and red fox were taken at necropsy of animals that have been found dead accidentally respectively in Finland and in Belgium. No animal was killed for the purpose of this study.
Genomic DNA was extracted from blood or muscle using the DNeasy Blood and tissue Kit (Qiagen) following manufacturer's instructions. Primers 59-CCAGCGGTCACCTTGATTGGGGC-39 (forward) and 59-ATGCGAGCTCAGCAGGTTGTGG-39 (reverse) were used to amplify the genomic region comprised between the first exon of PRDX5 and the first exon of HSPC152/ TRMT112, which is in close proximity of PRDX5 and positioned in opposite direction ( Fig. 1 and [11]). The primers were designated to map on regions sharing 100% identity between giant panda and dog sequences. The PCR reaction was performed using the Taq DNA polymerase (Eurogentec) in the presence of 5% (v/v) dimethyl sulfoxide (DMSO). The PCR products were cloned into pCR2.1 vector and sequenced. At least eight clones, deriving from three independent PCR reactions, were sequenced for each species. The sequences of the 59 region of PRDX5 from caniformia species have been submitted to GenBank (see Table 1).

Expression of Giant Panda and Northern Elephant seal PRDX5 MTS in Fusion with Green Fluorescent Protein (GFP) in MDCK Cells
Based on Genbank database sequence XM_002916664, the sequence encoding giant panda PRDX5 MTS was synthesized and cloned into pEX-A vector by Eurofins MWG Operon. The sequence coding for giant panda PRDX5 MTS was subsequently PCR-amplified using forward primer 59-GGGATCGGCAT-GAAGCTTTTTCCGCACCG-39 and reverse primer 59-CCATGGCTGAGGCGGCGCTTCTGAAACC-39. The sequence encoding the MTS of northern elephant seal PRDX5 was PCR-amplified from a pCR2.1 vector containing northern elephant seal PRDX5 59 flanking region (see point 4.6.) using forward primer 59-GGTTTGGGCATGCCGTTCGTTCAGC-TACG-39 and reverse primer 59-CCATGGCTGAGGCGGCG CTTCTGAAACC-39. PCR products were cloned in the pcDNA3.1/CT-GFP-TOPO vector (Invitrogen) and the resulting plasmids were sequenced. MDCK cells were transiently transfected using Amaxa Nucleofection system (Lonza AG) according to manufacturer's instructions. 72 h after transfection, cells were incubated with Mitotracker Red prior to fixation. The coverslips were mounted in Vectashield containing 50 mg/ml DAPI and analyzed by fluorescence microscopy.
MDCK cells were transfected with plasmids using Amaxa Nucleofection system according to manufacturer's instructions. Two days after transfection, cells were selected 16 days with 2 mg/ ml puromycin (Sigma-Aldrich). Subsequently, stably transfected cell lines were obtained by isolation of individual clones by a dilution technique. At least three clones per cell line were obtained and characterized.
Overexpression of PRDX5 was confirmed by immunofluorescence and immunoblotting ( Figure 6). Incorporation of puromycin resistance gene in the genome of MDCK control cells was verified by PCR on genomic DNA. The latter was extracted from 5610 6 cells using the DNeasy Blood and Tissue kit (Qiagen) following manufacturer's instructions. The puromycin N-acetyltransferase gene was amplified by using forward primer 59-GACCGAGTA-CAAGCCCACGGTGCG-39 and reverse primer 59-CGA-GACGCC GACGGTGGCCAGGAACCACG-39.
Cell Treatment with Hydrogen Peroxide and Tert-butyl Hydroperoxide (t-BHP) and Lactate Dehydrogenase (LDH) Assay MDCK cells were seeded at a density of 3610 4 cells/cm 2 into 24-well plates. After 15-18 h culture, stress was induced during one hour in DMEM without phenol red containing 4.5 g/l Dglucose and supplemented with 1% FBS, 100 U/ml penicillin, 100 U/ml streptomycin, 1% non-essential amino acids and 2 mM GlutaMAX. Cytotoxicity was evaluated by the lactate dehydrogenase (LDH) release assay (Cytotoxicity detection Kit, Roche). Fifty ml of supernatants were used for LDH assay. The total LDH release (100% cell death) was determined for each well after lysis of cells with 2% Triton X-100 detergent. Cell mortality was calculated according to this scale.

Statistical Analysis
Cell death values were calculated from triplicates within an experiment. Each experiment was repeated three times for one clone, allowing the calculation of means and SEM. Statistical analysis was performed using two-way ANOVA with the Bonferroni post hoc test. Significance is designated as *p,0.05, **p,0.01, ***p,0.001. At least three clones were tested for each cell line. Figure S1 (A) Subcellular localization of PRDX3 and TXN2 was assessed by immunofluorescence detection of endogenous proteins in dog MDCK cells. Mitotracker staining and immunofluorescence were performed as described in section 4.5, using 1:200 rabbit anti-human PRDX3 [24] and 1:200 rabbit antihuman TXN2 [41] polyclonal antibodies. Nuclei were counterstained with DAPI. Expression levels of PRDX3 (B) and TXN2 (C) were quantitated and compared in MDCK and SH-SY5Y cells.

Supporting Information
PRDX3 and TXN2 were detected by Western blotting in soluble proteins from whole cell extracts. Indicated amounts of human recombinant 6xHis-tagged PRDX3 [24] and TXN2 [42] were also blotted for quantitation. Western blotting was performed as described under section 4.3. Blots were probed with 1:4000 rabbit anti-human PRDX3 and 1:4000 rabbit anti-human TXN2. PRDX3 (D) and TXN2 (E) expression levels were quantified by comparison with recombinant protein standards. PRDX3 and TXN2 expression levels were expressed as the percentage of total protein content. Values are means 6 SEM of triplicates.