DERA is the human deoxyribose phosphate aldolase and is involved in stress response

(1)

Article

Reference

DERA is the human deoxyribose phosphate aldolase and is involved in stress response

SALLERON, Lisa, et al.

Abstract

Deoxyribose-phosphate aldolase (EC 4.1.2.4), which converts 2-deoxy-d-ribose-5-phosphate into glyceraldehyde-3-phosphate and acetaldehyde, belongs to the core metabolism of living organisms. It was previously shown that human cells harbor deoxyribose phosphate aldolase activity but the protein responsible of this activity has never been formally identified. This study provides the first experimental evidence that DERA, which is mainly expressed in lung, liver and colon, is the human deoxyribose phosphate aldolase. Among human cell lines, the highest DERA mRNA level and deoxyribose phosphate aldolase activity were observed in liver-derived Huh-7 cells. DERA was shown to interact with the known stress granule component YBX1 and to be recruited to stress granules after oxidative or mitochondrial stress. In addition, cells in which DERA expression was down-regulated using shRNA formed fewer stress granules and were more prone to apoptosis after clotrimazole stress, suggesting the importance of DERA for stress granule formation. Furthermore, the expression of DERA was shown to permit cells in which mitochondrial ATP production [...]

SALLERON, Lisa, et al . DERA is the human deoxyribose phosphate aldolase and is involved in stress response. Biochimica et Biophysica Acta - Molecular Cell Research , 2014, vol.

1843, no. 12, p. 2913-25

DOI : 10.1016/j.bbamcr.2014.09.007 PMID : 25229427

Available at:

http://archive-ouverte.unige.ch/unige:41750

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

DERA is the human deoxyribose phosphate aldolase and is involved in stress response

Lisa Salleron

^a,

⁎ , Giovanni Magistrelli

^b

, Camille Mary

^a

, Nicolas Fischer

^b

, Amos Bairoch

^a,c

, Lydie Lane

^a,c,

⁎⁎

aDepartment of Human Protein Sciences, Faculty of Medicine, University of Geneva, Geneva, Switzerland

bNovImmune SA, Plan-les-Ouates, Switzerland

cCALIPHO GroupSIB-Swiss Institute of Bioinformatics, Geneva, Switzerland

a b s t r a c t a r t i c l e i n f o

Article history:

Received 6 May 2014

Received in revised form 13 August 2014 Accepted 7 September 2014

Available online 16 September 2014

Keywords:

Deoxyribose phosphate aldolase Stress granule

Metabolic stress

Deoxynucleoside metabolism Apoptosis

Deoxyribose-phosphate aldolase (EC 4.1.2.4), which converts 2-deoxy-D-ribose-5-phosphate into glyceraldehyde- 3-phosphate and acetaldehyde, belongs to the core metabolism of living organisms. It was previously shown that human cells harbor deoxyribose phosphate aldolase activity but the protein responsible of this activity has never been formally identiﬁed. This study provides theﬁrst experimental evidence that DERA, which is mainly expressed in lung, liver and colon, is the human deoxyribose phosphate aldolase. Among human cell lines, the highest DERA mRNA level and deoxyribose phosphate aldolase activity were observed in liver-derived Huh-7 cells. DERA was shown to interact with the known stress granule component YBX1 and to be recruited to stress granules after oxidative or mitochondrial stress. In addition, cells in which DERA expression was down-regulated using shRNA formed fewer stress granules and were more prone to apoptosis after clotrimazole stress, suggesting the importance of DERA for stress granule formation. Furthermore, the expression of DERA was shown to permit cells in which mitochondrial ATP production was abolished to make use of extracellular deoxyinosine to maintain ATP levels.

This study unraveled a previously undescribed pathway which may allow cells with high deoxyribose-phosphate aldolase activity, such as liver cells, to minimize or delay stress-induced damage by producing energy through deoxynucleoside degradation.

1. Introduction

Deoxyribose phosphate aldolase (EC 4.1.2.4) catalyzes the conver- sion of 2-deoxy-D-ribose-5-phosphate (DR5P) into glyceraldehyde-3- phosphate (G3P) and acetaldehyde[1]. DeoC is a Schiff-forming class I aldolase, which is responsible for deoxyribose phosphate aldolase activity in bacteria[2,3]. It has been widely used as a biocatalyst to perform stereo-controlled aldol condensation because of its ability to use various aldehydes as substrates[4]. The resolution of the crystal structures of Escherichia coli(E. coli) deoC (PDB: 1JCJ and 1JCL)[5,6]and of its orthologs inThermus thermophilus(PDB: 1UB3 and 1J2W)[7]and in

the archaeAeropyrum pernix(PDB: 1N7K)[8]permitted identification of the lysine residues implicated in the Schiff base formation and engi- neering of DeoC mutants with increased specificity for substrates useful in industrial applications. Deoxyribose phosphate aldolase activity has been detected in mammalian samples including liver[9–12], human erythrocytes[13], and a variety of rat and human cells[12,14–16], but the protein responsible for this activity in mammals has never been formally identified. The mammalian enzyme shares some of its characteristics with the bacterial ones, such as sensitivity to carboxylic acids and to reducer agents[9,13].

Along with DeoA (thymidine phosphorylase, EC 2.4.2.4), DeoB (phosphopentomutase, EC 5.4.2.7), and DeoD (purine nucleoside phosphorylase, EC 2.4.2.1),E. coliDeoC is part of thedeooperon which allows bacteria to use extracellular deoxynucleosides as energy sources[17]:

DeoD or DeoA cleaves the glycosidic bond of the deoxyribonucleoside and liberates a deoxyribose-1-phosphate (DR1P) that is further converted into DR5P by DeoB. Finally, DeoC converts DR5P into G3P and acetaldehyde, which enter glycolysis and the Krebs cycle, respectively [1]. In mammalian cells a pathway involving TYMP (thymidine phosphorylase)[18,19], PGM2 (phosphopentomutase)[20], and PNP (purine nucleoside phosphorylase)[21,22]as functional equivalents of DeoA, DeoB and DeoD, respectively, has been described (Fig. 1A). Known as Abbreviations: CHX, Cycloheximide; DR5P, Deoxyribose-5-phosphate; DR1P,

Deoxyribose-1-phosphate;E. coli,Escherichia coli; G3P, Glyceraldehyde-3-phosphate;

FCCP, Carbonyl cyanide-4-(triﬂuoromethoxy)phenylhydrazone; PB, Processing bodies;

shRNA, Short hairpin RNA; SG, Stress granules

⁎ Correspondence to: L. Salleron, CALIPHO group, Department of Human Protein Sciences, University of Geneva, CMU, Michel-Servet 1, 1211 Geneva 4, Switzerland.

Tel.: +41 22 379 54 78.

⁎⁎Correspondence to: L. Lane, CALIPHO group, SIB-Swiss Institute of Bioinformatics, Department of Human Protein Sciences, University of Geneva, CMU, Michel-Servet 1, 1211 Geneva 4, Switzerland. Tel.: +41 22 379 58 41.

E-mail addresses:lisa.salleron@unige.ch(L. Salleron),lydie.lane@isb-sib.ch(L. Lane).

Contents lists available atScienceDirect

Biochimica et Biophysica Acta

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b b a m c r

(3)

the nucleoside salvage pathway, it allows cells to maintain a balanced pool of deoxynucleosides. However, it is still unclear whether in mammalian cells DR5P can be further converted into energy derivatives[23].

In this study, we show that DERA is the mammalian deoxyribose phosphate aldolase, and that this enzyme confers on cells the ability to use deoxynucleosides to replenish ATP cellular levels upon stress.

We report that DERA interacts with YBX1, a protein involved in DNA repair, and that, similar to YBX1[24], it is recruited to stress granules (SG) upon different kinds of stress. SG are cytoplasmic bodies which form when the cell is exposed to adverse conditions in order to prevent apoptosis and preserve cell viability[25]. We show also that DERA is necessary for SG formation and prevention of apoptosis after stress.

Fig. 1.The expression proﬁle of DERA, the DeoC ortholog, correlates with deoxyribose phosphate aldolase activity in human cells. (A) Enzymes implicated in the catabolism of the deoxyribose moiety of thymidine inE. coliand human. (B) Alignment ofE. coliDeoC (P0A6L0),T. kodakaraensisDeoC (Q877I0) and human DERA (Q9Y315). The lysines essential for DeoC activity are shown in bold. (C, D) DERA expression values measured by qRT-PCR on human tissues (representative experiment) (C) and cell lines (means ± s.d., n = 3) (D). (E) Deoxyribose phosphate aldolase activity values in cell lines (means ± s.d., n = 3).

(4)

2. Material and methods 2.1. Sequence analysis

The putative human ortholog for the DERA enzymes fromE. coli (P0A6L0) andThermococcus kodakaraensis(Q877I0) was searched by BLASTP analysis in the UniProtKB database version 15.14[26]. Multiple sequence alignments were performed using Clustal Omega with the default parameters[27].

2.2. Cell culture

All cells were grown in an incubator at 37 °C in a 5% CO2atmosphere.

HEK293, Huh-7, HepG2, Caco-2 and HeLa Kyoto cells were grown in DMEM (Life Technologies) supplemented with 10% fetal bovine serum (Sigma-Aldrich). Jurkat cells were grown in RPMI 1640 (Life Technologies) supplemented with 10% fetal bovine serum. Treatments with clotrimazole, sodium arsenite, cycloheximide (CHX) (Sigma- Aldrich) and carbonyl cyanide-4-(triﬂuoromethoxy)phenylhydrazone (FCCP) (Enzo Life Sciences) were done at 37 °C for the indicated time.

Sodium arsenite was used at 0.12 mM unless otherwise noted. Clotri- mazole and FCCP were dissolved in DMSO and used at 20μM and 10μM respectively. CHX was dissolved in DMEM without serum (Life Technologies) and used at 20μg/ml.

2.3. qRT-PCR

cDNA from cells was prepared from 500 ng RNA extracted with RNeasy kit (Qiagen), using superscript II reverse transcriptase (Life technologies). Tissue cDNAs were from Clontech. qRT-PCR reac- tions were performed in triplicates using 5′-CAAGGCTGCAGGCTGT AATA-3′and 5′-CATCTTCCACAGCCAATCTG-3′as primers and the Power SYBR Green PCR Master mix (Applied Biosystems). Raw cycle threshold values were obtained with SDS 2.2.2 (Applied Biosystems). Normalization by TBP expression was performed using GeNorm[28].

2.4. Plasmids and transfection

siRNA against YBX1 and silencer select negative control (40 nM, Life Technologies) were transfected using lipofectamine RNAiMAX (Life Technologies) in serum-free media 24 h before treatment with stress in- ducers. Plasmid transfections were performed using XtremeGene-HP (Roche). pDEST-mycYBX1 was from Addgene (plasmid 19878)[29].

DERA cDNA was ampliﬁed from HeLa cells by PCR using 5′-CACCTCCG CGCACAATCGGGGCAC-3′and 5′-CACCTCCGCGCACAATCGGGGCAC-3′ as primers, cloned in pENTR-D-TOPO, then in pDEST15 or in pcDNA3.1/nv5 using the Gateway cloning kit (Life Technologies).

DERA-6xHis sequence was obtained by PCR performed on pENTR-D- TOPO-DERA vector using CGT-GAA-AAG-CTT-ATG-TCC-GCG-CAC-AAT- CGG-GGC-ACC-GA and TAG-CAC-TAG-TTC-ATT-AGT-GGT-GGT-GGT- GGT-GGT-GGT-GGT-GGC as primers, and inserted in pEAK8 [30]

digested with HindIII and EcoRI. K218A and K254A DERA-6xHis mutants were obtained by overlap extension PCR[31]. The expression of DERA-6XHis and its mutants in PEAK cells was checked by Western blot using anti-His (1:2000; Abd serotec) and anti-β-catenin (1:2000;

Sigma-Aldrich) antibodies. Transient DERA downregulation was done by transfection of the pRNAi-H1-puro plasmid containing the shRNA sequence AAA-ACC-ATC-ATG-TGA-CTG-GAA-GAT-ATT-GGA-TCC-AAT- ATC-TTC-CAG-TCA-CAT-GAT-GG targeting DERA coding sequence.

HeLa cell clones stably expressing DERA shRNA were selected using 1μg/ml puromycin. The lacZ shRNA pRNAi vector (Biosettia) was used as control.

2.5. Imaging

Cells wereﬁxed with 4% paraformaldehyde, permeabilized and blocked with 3% BSA in PBS with 0.25% Triton X-100, and incubated with primary antibodies (rabbit anti-DERA (1:50, Epitomics), rabbit anti-YBX1 (1:50, HPA040304, gift from M. Uhlen), goat anti-G3BP1 (1:50, Santa Cruz), goat anti-DCP1A (1:50, Santa Cruz), goat anti-c- Myc (1:200, AbD Serotec), or mouse anti-PABPC1 (1:50, Santa Cruz).

For immunoﬂuorescence experiments, secondary antibodies were don- key anti-IgG coupled to Alexa Fluor (1:150, Life Technologies). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (mouting medium, Prolong, Life Technologies). Confocalﬂuorescence images were obtained using an Axio Imager M2 microscope with a 63× oil objective and the ZEN software (Carl Zeiss). Images were compiled using Adobe Photoshop software (Adobe Systems). Scale bars are 10μm in images and 2μm in inserts.

2.6. Stress granule counting

SG were counted using the“analyze particles”tool of the ImageJ software (NIH), with a SG area set to 0.7–78μm²[32], and all countings were manually checked. A minimum of 100 cells from duplicate coverslips were counted for each experiment.

2.7. In situ proximity ligation assay

The in situ proximity ligation assay was performed on paraformaldehyde-ﬁxed cells according to manufacturer's instruction (Olink Bioscience) using either anti-DERA (1:50) and anti c-Myc (1:200) or anti-v5 (1:200) (Life Technologies) and anti-YBX1 (1:50) as primary antibodies, and secondary antibodies coupled to Texas Red labeled oligonucleotides (1:40, Olink Bioscience). The secondary antibodies are linked with complementary labeled oligonucleotides that interact when they are in close proximity (less than 40 nm). Each interacting pair is ampliﬁed with a polymerase and visualized as a distinct spot.

2.8. Deoxyribose phosphate aldolase assay

Deoxyribose phosphate aldolase activity was assayed by spectro- photometry using the alcohol dehydrogenase (ADH)-coupled method [33] on cells lysed in RIPA buffer (150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0), 2% Triton X-100, with protease and phosphatase inhibitors (Complete; Roche). Assay was performed at 37 °C in phosphate citrate buffer pH 6.8 with 4 mg of total proteins, 2 mM DR5P (gift from P.-A. Kaminski), 500 × 10⁻³U Saccharomyces cerevisiae ADH (Sigma-Aldrich) and 0.4 M NADH (Sigma-Aldrich). One unit of enzyme activity catalyzes the formation of 1μmol of product/min.

2.9. ATP measurements

Cellular ATP concentrations were determined according to manufacturer's instructions (ATP Bioluminescence Assay Kit HS II, Roche Applied Science) following cell incubation for 2 h with 10μM FCCP in DMEM without glucose (Life Technologies), in presence of 8 mM inosine (Sigma-Aldrich) or 8 mM deoxyinosine (Sigma-Aldrich).

Luminescence was monitored in a plate-reader luminometer (Fluostar Optima, BMG Labtechnologies).

2.10. Detection of apoptosis by annexin V staining

Cells were stained with annexin V-APC (Allophycocyanin) and propidium iodide (PI) in binding buffer (PBS, 2 mM CaCl2) according to manufacturer's instructions (BD Biosciences), and analyzed using an Accury C6ﬂow cytometer (BD Biosciences). We considered both

(5)

cells in early apoptosis that were annexin V positive and PI negative, and cells in late apoptosis that were annexin V positive and PI positive.

2.11. Statistics

Data are shown as means ± standard deviation (s.d.) of results from at least three independent experiments, or ± standard error of the mean (s.e.m.) of results from more than three independent experiments. Statistical analysis was done by performing Student'sttest (star marks in theﬁgure legends: *pb0.05, **pb0.01, ***pb0.001).

2.12. GST pull-down

Competent BL21(DE3)pLysS cells (Agilent) were transformed with pDEST15-DERA or pGEX3t4-GST (a gift from P. Viollier). Protein expression was induced with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) for 16 h at 18 °C. Bacteria were lysed in extraction buffer (50 mM Tris pH 7.8, 300 mM NaCl, 5 mM MgCl2, 10 mM 2-mercaptoethanol, 2%

Triton X-100, with protease inhibitors (Complete, Roche) and 1 mg/ml lysozyme) and sonicated. After centrifugation, the supernatants containing the recombinant proteins were incubated with glutathione– sepharose (Sigma-Aldrich). GST or GST-DERA beads were washed and incubated with HeLa proteins (1.5 mg) extracted with RIPA buffer.

When indicated, lysates were treated with 500 units DNAse I or 4μg RNAse A for 1 h at 37 °C. Bound proteins were eluted with Laemmli buffer (2% SDS, 10% glycerol, 5% 2-mercaptoethanol, 0.125 M Tris HCl pH 6.8), and analyzed by Western blot using anti- YBX1 (1:1000). For mass spectrometry analysis, bound proteins were concentrated in a stacking SDS-PAGE gel, stained with Coomassie, excised and digested with trypsin. Peptides were submitted to liquid chromatography–tandem mass spectrometry (LC-MS/MS) using a LTQ Orbitrap velos (Thermo Electron) equipped with a NanoAcquity system (Waters). Peptide identification was performed using Mascot (Matrix Sciences) onHomo sapiensUniProtKB/Swiss- Prot sequences (release 15.10). Scaffold_4.2.0 (Proteome Software Inc., Portland, OR) was used to validate identifications. Peptide and protein identifications were accepted if they could be established at greater than 95.0% probability by Peptide Prophet and Protein Prophet, respectively[34]. Proteins with less than two unique tryptic peptides were discarded.

2.13. His pull-down

HeLa cells co-transfected with pDERA and pDEST-c-Myc-YBX1 were lysed in RIPA buffer with 10 mM 2-mercaptoethanol, 2% Triton X100 and incubated with Ni-NTA afﬁnity matrix (Qiagen) for 2 h at 4 °C.

Bound proteins were eluted with Laemmli buffer, and analyzed by Western blot using anti c-Myc (1:1000).

2.14. Co-immunoprecipitation

HeLa cells co-transfected with pDERA and pDEST-MycYBX1 were lysed in RIPA buffer with 2% TX-100, for 15 min at 4 °C then cell de- bris were removed by centrifugation. The supernatant containing the solubilized proteins (5 mg) was incubated for 1 h at 4 °C with anti c- Myc (30 μg) antibody. Subsequently, protein G agarose (GE Healthcare) was added (50μl of 50% slurry), and the incubation was continued for an additional 1 h at 4 °C. The protein G agarose was washed three times with PBS with 0.2% Tween and bound proteins were then eluted with Laemmli buffer. Samples were analyzed by SDS-PAGE and immunoblotting using anti-YBX1 (1/1000) and anti-DERA (1/500).

3. Results

3.1. DERA is the human deoxyribose phosphate aldolase.

By BLAST analysis, DERA appeared to be the ortholog of bacterial and archaeal DeoC, sharing 38% identity withE. coliDeoC and 22% identity withT. kodakaraensisDeoC. Furthermore, the two lysines essential for E. coliDeoC activity [6] are conserved in DERA (K218 and K254) (Fig. 1B). By qRT-PCR, the highest expression of DERA was observed in liver, lung and colon among human tissues (Fig. 1C), and in hepatocellu- lar carcinoma Huh-7 cells and colorectal adenocarcinoma Caco-2 cells among cell lines of various origins (Fig. 1D). As expected, the highest deoxyribose phosphate aldolase activity was observed in hepatocarcinoma Huh-7 (0.47 × 10⁻³ ± 0.05 U/mg), Jurkat T lymphocyte-derived cells and hepatoblastoma HepG2 cells showed moderate activities (0.18 × 10⁻³± 0.06 U/mg and 0.15 × 10⁻³± 0.03 U/mg, respectively) (Fig. 1E).

In order to conﬁrm that DERA can act as a deoxyribose phosphate aldolase, DERA-6xHis or its putative catalytic mutants (K218A; K254A and K218A/K254A) were overexpressed in HeLa cells. To this end, DERA-6xHis cDNA was cloned into pEAK8, to yield a bicistronic vector (pDERA) which permits high DERA-6xHis expression and easy detection of transfected cells which express GFP (Fig. 2A). pDERA-K218A, pDERA- K254A and pDERA-K218A/K254A constructs were generated by overlap extension PCR. As expected, HeLa cells transfected with increasing concentrations of pDERA showed increasing levels of DERA-6xHis (Fig. 2B) and increasing deoxyribose phosphate aldolase activity levels (0.150 × 10⁻³± 0.001 U/mg with 10μg pDERA; 0.284 × 10⁻³± 0.02 U/mg with 25μg pDERA and 0.304 × 10⁻³± 0.03 U/mg with 30μg pDERA), although the levels obtained by DERA-6xHis overexpression do not reach those found endogenously in hepatocarcinoma Huh-7 cells (0.470 × 10⁻³± 0.05 U/mg) (Fig. 1E). In contrast, HeLa cells transfected with 10μg pDERA mutants did not show any increase in deoxyribose phosphate aldolase activity compared to cells transfected with 10μg empty pEAK8 vector (0.091 × 10⁻³± 0.017 U/mg for empty vector versus 0.113 × 10⁻³± 0.01 U/mg for pDERA-K218A/K254A; 0.123 × 10

−3± 0.005 U/mg for pDERA-K218A; 0.105 × 10⁻³± 0.003 U/mg for pDERA-K254A) (Fig. 2C), which conﬁrms that K218 and K254 are required for DERA activity.

Next, to confirm that the deoxyribose phosphate aldolase activity measured in untransfected cells was contributed by endogenous DERA, we monitored deoxyribose phosphate aldolase activity in HeLa cell lines that stably downregulate DERA by expressing short hairpin RNA (shRNA) targeting its coding sequence. By immunofluorescence with a specific DERA antibody and by qRT-PCR, we checked that these cells display lower levels of DERA protein and mRNA than control cells expressing shRNA-LacZ (Fig. 2D and E). DERA shRNA cells had 75% lower aldolase activity than control shRNA cells (Fig. 2F), which confirmed that DERA is the human deoxyribose phosphate aldolase.

3.2. DERA interacts with YBX1

In order to elucidate DERA's physiological function, we attempted to ﬁnd its potential interacting partners by pull down experiments.

Recombinant GST-DERA produced in bacteria (Fig. 3A) was used as bait to perform pull down experiments on HeLa cell lysates and GST was used as a control for specificity. Fifty-six proteins were identified by mass spectrometry (MS) in the GST-DERA but not in the GST pull- down eluates. Among those, we excluded proteins that bound to other unrelated GST fusion proteins, and abundant contaminant proteins usually recovered using this method, including keratins and keratin- associated proteins. Finally, we obtained a list of 15 proteins which specifically bind to GST-DERA (Supplementary Table 1). Based on the fact that DR5P, the substrate of DERA, can come from deoxynucleotide catabolism or from base excision repair of damaged DNA, we

(6)

selected YBX1, known to interact with DNA repair enzymes[35–37]and identiﬁed by three unique peptides with 24% sequence coverage (Fig. 3B), for further validation.

We confirmed the YBX1 MS identification result by Western blot analysis using anti-YBX1 antibody (Fig. 3C). Since YBX1 is involved in transcription, translation, alternative splicing regulation and DNA Fig. 2.DERA is responsible for deoxyribose phosphate aldolase activity in mammalian cells. (A) Schematic representation of the different elements of the bi-cistronic pEAK8-DERA (pDERA): elongation factor 1 alpha (EF-1) promoter, DERA-6xHis, internal ribosome entry site (IRES) and GFP. (B, C) HeLa cells were transiently transfected with various amounts of pDERA or with 10μg of pDERA mutants. (B) Western blot analysis confirmed expression of the different constructs. (C) The bar graph shows means ± s.d. of deoxyribose phosphate aldolase activity values measured in three independent experiments. (D) HeLa cells stably expressing control and DERA shRNA werefixed and immunostained for DERA. Nuclei were visualized by DAPI. Scale bars are 10μm. (E) DERA downregulation in HeLa cells stably expressing DERA shRNA was checked by qRT-PCR. The bar graph shows means ± s.d. of values obtained in three independent experiments. (F) Deoxyribose phosphate aldolase activity is severely decreased in DERA shRNA cells. The bar graph shows means ± s.d. of deoxyribose phosphate aldolase activity values measured in three independent experiments.

(7)

Fig. 3.DERA interacts with YBX1 in stressed and unstressed cells. (A) Coomassie Blue-stained SDS-PAGE gel showing GST-DERA and GST immobilized on glutathione sepharose. (B) GST- DERA and GST immobilized on glutathione sepharose were used as baits for pull-down experiments on HeLa protein lysates and partners were identified by LC-MS/MS analysis of the eluates. YBX1 was identified by three unique peptides covering 24% of YBX1 sequence, from four spectra validated manually. (C) GST-DERA or GST beads (20μg) were incubated with HeLa lysates treated or not with DNAse I or RNAse A. The presence of YBX1 in eluates and inputs (10%) was analyzed by Western blot with anti-YBX1. (D). HeLa cells transfected with c-Myc-YBX1 (left panel) or v5-DERA (right panel) were either untreated or treated with 20μM clotrimazole (Clo) or 0.5 mM arsenite (As), then subjected toin situproximity ligation assay using anti-DERA and anti-c-Myc (left panel) or anti-v5 and anti-YBX1 antibodies (right panel). Untransfected cells were used as negative control. Each dot represents the two proteins in close proximity. Scale bars are 10μm. (E) HeLa cells were transfected with pDERA or pDERA-K218A/K254A and c-Myc-YBX1 and treated with clotrimazole. Cell lysates were incubated with nickel affinity columns to pull DERA-6xHis down. The presence of c-Myc-YBX1 and DERA-6xHis was analyzed by Western blot in pull down eluates (PD) and input samples (10%) using anti-c-Myc and anti-His. (F) c-Myc-YBX1 and DERA-6xHis were co-immunoprecipitated from HeLa cell lysates transfected with pDERA and c-Myc-YBX1. As controls we used untransfected cells or cells transfected with c-Myc-YBX1 alone.

(8)

repair, and has the ability to bind DNA and RNA[38,39], we checked that YBX1 was able to bind GST-DERA even when lysates were treated with DNAse or RNAse, thus excluding that DNA or RNA could mediate this interaction (Fig. 3C).

YBX1 is a known component of SG[24,40]. SG are cytoplasmic bodies formed upon translational arrest in response to environmental stresses such as sodium arsenite (oxidative stress) or clotrimazole (mitochondrial stress)[25]. To visualize the interaction between YBX1 and DERA in unstressed conditions and in conditions promoting SG formation, we performedin situproximity ligation assays [41]on c-Myc-YBX1-transfected HeLa cells using anti-c-Myc and anti-DERA antibodies and on v5-DERA-transfected cells using anti-v5 and anti- YBX1 antibodies. Endogenous DERA was found in proximity (b40 nm) to c-Myc-YBX1, both in unstressed cells and in stressed cells treated with sodium arsenite or clotrimazole (Fig. 3D, left panel). Similarly, endogenous YBX1 was found in close proximity to v5-DERA in the same conditions. In contrast, negligible non-speciﬁc binding of the probes was observed in non-transfected cells. The interaction between YBX1 and DERA in clotrimazole-stressed HeLa cells was conﬁrmed by His-pull down experiments in cells overexpressing c-Myc-YBX1 and DERA-6xHis (Fig. 3E). c-Myc-YBX1 was able to bind DERA-K218A/

K254A-6xHis as well (Fig. 3E), indicating that DERA catalytic residues are not involved in this interaction. To further conﬁrm the interaction between DERA and YBX1, we carried out co-immunoprecipitation experiments on untreated HeLa cells transfected with c-Myc-YBX1 and pDERA. DERA-6X-His was successfully co-immunoprecipitated with c-Myc-YBX1 using an anti-c-Myc antibody (Fig. 3F).

3.3. DERA is localized in stress granules but not in processing bodies

Because the interaction between DERA and YBX1 takes place both in unstressed and stressed cells, we investigated DERA localization by immunoﬂuorescence in both conditions. DERA is localized both in the nucleus and the cytoplasm in unstressed HeLa cells (Figs. 2D and4A), consistently with its suggested role in nucleotide metabolism. This localization is compatible with that of YBX1, which is cytosolic in resting cells (Fig. 4A), but can shuttle between cytoplasm and nucleus in speciﬁc conditions such as UV treatment[42], or upon interaction with the splicing factor SRp30c[43].

Upon exposure to arsenite or clotrimazole, DERA was detected in the same cytoplasmic granules as c-Myc-YBX1. These granules were identiﬁed as SG using G3BP1[44]and PABPC1[45]as markers. DERA also localized to H2O2-induced SG together with G3BP1 and PABP (data not shown). We checked that c-Myc-YBX1 has the same localization as endogenous YBX1 in each condition (Fig. 4). Similar results were obtained in Jurkat, Huh-7 and HepG2 cells (data not shown).

SG assembly requires polysome disassembly and the subsequent release of mRNAs, which are recruited by RNA-binding proteins which form the core of SG. As a consequence, cycloheximide (CHX), which prevents polysome disassembly, impairs SG formation[46]. CHX added at the same time as arsenite or clotrimazole totally prevented the formation of DERA, c-Myc-YBX1, G3BP1 and PABPC1 foci (Fig. 4B). Therefore, polysome disruption is required for DERA incorporation into SG. CHX has also been shown to disassemble pre-formed SG[46]. After clotrimazole stress, CHX completely disassembled DERA and G3BP1-containing SG, and partially disassembled c-Myc-YBX1- and PABPC1-containing SG. After arsenite stress, the effect of CHX was milder, since we observed only a partial dissolution of DERA-containing SG and no change in the other markers (Fig. 4C). These results suggest that DERA, G3BP1, c-Myc-YBX1- and PABPC1 have different SG exit kinetics, and that DERA may be theﬁrst marker to exit SG.

Processing bodies (PB) are cytoplasmic granules, distinct from SG, where mRNA decay takes place. They contain proteins involved in mRNA decapping, such as DCP1A, and proteins involved in mRNA degradation[47]. PB are always present in the cell, but their size and number grow upon arsenite treatment[48], and many proteins and

mRNAs contained in SG can shuttle to PB during stress [25,49].

However, using DCP1A as a marker, we observed that, upon arsenite treatment, DERA and YBX1 are not incorporated into PB (Fig. 4C).

Altogether, these results show that, in our conditions, endogenous DERA and YBX1 are associated with SG but not with PB.

Because DERA contains neither a typical RNA-binding motif, nor a glycine or glutamine rich C-terminal tail known to be involved in SG targeting[40], we studied if DERA interaction with YBX1 could be important for its localization to SG. We analyzed DERA localization after arsenite stress in HeLa cells in which YBX1 levels were downregu- lated using siRNA. The diminution in YBX1 protein level was checked by Western blot (Fig. 5A). SG containing PABPC1, G3BP1 and DERA could still be observed in these conditions (Fig. 5B), suggesting that YBX1 is not required for the formation of SG in these conditions, nor for DERA targeting to SG. However, the low levels of YBX1 remaining after siRNA treatment might be sufﬁcient to ensure proper SG formation and DERA targeting to SG.

3.4. DERA is important for stress granule formation

In order to know if DERA plays a role in SG assembly, we analyzed the effect of DERA depletion on SG formation in HeLa cells. We used transient rather than stable shRNA interference to make sure that the observed phenotype was not due to clonal effects of the stable cell lines. In HeLa cells expressing DERA shRNA, we observed a four-fold decrease in DERA mRNA measured by qRT-PCR (Fig. 6A), and a concomitant diminution in the number of G3BP1-containing and YBX1-containing SG formed during clotrimazole stress compared to cells expressing control LacZ shRNA (Fig. 6B and C). Fewer G3BP1- and YBX1-containing SG were also formed during arsenite stress (Fig. 6D).

In contrast, DCP1A-containing PB observed upon arsenite stress do not seem to be affected by DERA depletion (Fig. 6E). These results highlight the importance of DERA in SG formation, but not in PB formation.

SG are known to be a major adaptive response of cells to stress, by preventing them from undergoing apoptosis. For example, sequestra- tion of activated ROCK1 in SG prevents its association with TRAF3IP3, and, thus, prevents apoptosis[50]. Therefore, we tested whether cells downregulating DERA were more prone to apoptosis during clotrimazole stress. Indeed, we observed a 35% increase in apoptosis visualized with annexin V staining in cells expressing low level of DERA compared to cells transfected with control shRNA (Fig. 6F).

3.5. DERA activity may allow cells to use deoxyinosine as an energy source

Arsenite and clotrimazole trigger SG assembly through activation of kinases causing the phosphorylation of the translation initiation factor eIF2α, a component of the eIF2-GTP-Met-tRNA^Metcomplex. On the other hand, drugs that deplete intracellular energy stores, such as the mitochondrial uncoupler FCCP, cause SG assembly independent of eIF2αphosphorylation, by reducing the ATP/ADP ratio, which leads to a decrease in the GTP/GDP ratio and hence to a reduced formation of the eIF2-GTP-Met-tRNA^Metcomplex[25]. As described above, DERA is recruited to arsenite- or clotrimazole-induced SG (Fig. 4). To know if DERA is also recruited to SG upon energy depletion, we applied a combination of glucose deprivation and FCCP to HeLa cells. In these conditions, ATP levels drop (Fig. 7C) and cells form SG containing YBX1, PABPC1 and G3BP1 (Fig. 7A). Importantly, DERA is also recruited to SG in this context (Fig. 7A).

In bacteria, DeoB and DeoC allow cells to use the pentose moiety of deoxynucleosides present in their growth medium as an energy source [17,51]. In human, extracellular deoxynucleosides can come from nutri- tional sources, or be released from cells into the circulation. Mammalian PGM2[20], PNP[21]and DERA are functional homologues of bacterial DeoB, DeoD and DeoC. However, it is still not known if the phosphorylated pentose released by phosphorolytic cleavage of deoxynucleosides in human can be catabolized by DERA and used as a source of energy in

(9)

Fig. 4.DERA is recruited to stress granules (SG) but not to processing bodies (PB). (A–C) Left panels: c-Myc-YBX1-transfected HeLa cells were exposed to 0.12 mM arsenite (As) or 20μM clotrimazole (Clo) with or without 20μg/ml cycloheximide (CHX),ﬁxed and immunostained for DERA (red) and c-Myc-YBX1, G3BP1 or PABPC1 (green). Right panels: Untransfected HeLa cells were exposed to 0.12 mM arsenite (As) or 20μM clotrimazole (Clo) with or without 20μg/ml cycloheximide (CHX),ﬁxed and immunostained for YBX1 (red) and G3BP1 (green).

(A) DERA co-localized with SG components after arsenite or clotrimazole stress. (B) CHX added at the same time as clotrimazole or arsenite blocks SG formation. (C) DERA-containing SG are totally disassembled when CHX is added after clotrimazole and partially disassembled when CHX is added after arsenite. (D) Untransfected HeLa cells were exposed to 0.12 mM arsenite,ﬁxed and immunostained for DCP1A (green) and DERA or YBX1 (red). DERA and YBX1 granules cluster around DCP1A-containing PB in As-treated cells. Scale bars are 10μm in images and 2μm in inserts. Merge color is yellow.

(10)

addition to or as an alternative to glucose, like in bacteria[23]. Our hypothesis was that mammalian cells could use deoxyinosine as an energy source, through the action of PNP, PGM2 and DERA.

Deoxyinosine, which can be produced endogenously by deamination of deoxyadenosine by the adenosine deaminase enzyme (ADA), would be cleaved by PNP into hypoxanthine and DR1P which would be subsequently isomerized by PGM2 to DR5P[23]. Then, DR5P would be used by DERA to produce G3P that would enter the glycolysis pathway, and acetaldehyde that would enter the Krebs cycle so as to produce ATP.

Therefore, we tested whether DERA could allow ATP production from extracellular deoxyinosine in conditions of energy deprivation. We started our investigation with Huh-7 cells, which express high levels of DERA. After 2 h incubation with FCCP in the absence of glucose, ATP levels were less than 10% of those of control cells. The addition of extracellular inosine (from 0.5 mM to 8 mM) prevented the ATP drop, prob- ably through the action of PNP, which produces ribose-5-phosphate that can enter the pentose phosphate pathway. More importantly, deoxyinosine (from 0.5 mM to 8 mM) also prevented the ATP drop (Fig. 7B), indicating that Huh-7 cells are able to catabolize deoxyinosine into energy derivatives.

We repeated the experiment in HeLa cells, which express low levels of DERA. In these cells, the addition of extracellular inosine prevented the ATP drop that occurs in the presence of FCCP, but not the addition of deoxyinosine at a concentration which signiﬁcantly helps in ATP preservation in Huh-7 cells (8 mM) (Fig. 7C). We hypothesized that this was due to insufﬁcient endogenous deoxyribose phosphate aldolase activity, as previously suggested for other cell lines[16,23].

Thus, we tested if increasing the deoxyribose phosphate aldolase activity in HeLa cells would allow them to use deoxyinosine to overcome the stress induced by energy deprivation. To this end, we used HeLa cells transfected with wild type or mutant pDERA. HeLa cells overexpressing wild-type DERA restored their ATP levels in the presence of 8 mM deoxyinosine, whereas cells overexpressing DERA catalytic mutant did not (Fig. 7C). Then, we tested if the number of SG correlates with ATP depletion in cells transfected with wild type or mutant pDERA and incubated with extracellular inosine or deoxyinosine. In the presence of extracellular inosine, fewer SG are formed in cells transfected with wild type or mutant pDERA (Fig. 7D) conﬁrming that these cells can use inosine as a source of energy. In the presence of extracellular deoxyinosine, cells transfected with wild type pDERA formed fewer SG, whereas cells transfected with mutant pDERA formed the same number of SG (Fig. 7D), conﬁrming that only cells overexpressing a catalytically active deoxyribose phosphate aldolase were able to use deoxyinosine as a source of energy upon stress.

4. Discussion

In this study, we measured a signiﬁcant deoxyribose phosphate aldolase activity in various human cell lines, with the highest levels found in hepatocarcinoma Huh-7 cells, hepatoblastoma HepG2 cells

and Jurkat T lymphocyte-derived cells. In contrast to previous reports stating that deoxyribose phosphate aldolase activity levels vary among samples due to extractability and stability issues[1,11], we noted low enzyme activity variability among our samples. Also, even if the conditions we used to measure DERA activity were different from that of other studies, the activity levels we measured are in the range of what was previously reported for mammalian samples[12,13,16]. Although bacterial and mammalian deoxyribose phosphate aldolase were shown to be both activated by carboxylic acids and reducing agents [9,11], the speciﬁc activities reported for the mammalian enzymes are much lower than the bacterial ones. Indeed, the highest mammalian activity (6 × 10⁻³U/mg), measured in rat liver[12], is far below the lowest activity in bacteria (0.97 U/mg), measured inBacillus cereus [52]. This may explain why the mammalian enzyme had attracted such poor attention.

Using both up- and downregulation of DERA expression in HeLa cells, we demonstrated that DERA is responsible for the human deoxyribose phosphate aldolase activity, and identiﬁed two lysines implicated in its catalytic activity, conﬁrming that it is a class I aldolase[6].

Among human tissues, the highest DERA expression levels were found in the liver, where most of the early studies on the mammalian deoxyribose phosphate aldolase were performed[10,53,54]. Liver cells contain a high deoxynucleotide/nucleotide ratio compared to other cell types[55]. They incorporate alimentary extracellular nucleosides which can further be phosphorylated and reduced to deoxynucleotides [56,57]. It is also the main site of nucleobase recycling during the degradation of deoxynucleotides. Degradation of deoxynucleotides generates DR5P and is regulated by the need for free bases by other cells and by the abundance of nucleosides provided by food absorption[55]. Nota- bly, DNPH1, which produces DR5P directly from deoxynucleotides, is also highly expressed in the liver[58,59]. Therefore, one can expect that the liver would contain high levels of DERA's substrate. DERA is also well expressed in lung and in colon, where (deoxy)nucleosides originating from inhalated or symbiotic microorganisms can accumu- late and be reduced to deoxynucleotides, from which catabolism generates DR5P[60]. Accordingly, the highest DERA expression levels in cell lines were observed in liver Huh-7 and colon Caco-2 cells. Our panel did not include lung cancer cell lines, where we would have expected high levels of DERA expression as well.

We found that deoxyribose phosphate aldolase activity levels corre- late with DERA expression level in most cell lines. Indeed, we found a high deoxyribose phosphate aldolase activity in Huh-7 cells, intermedi- ate activities in HepG2 and Jurkat cells, and low activities in HeLa and HEK293 cells. The only discrepancy that we noticed was for Caco-2 cells, which display a low deoxyribose phosphate aldolase activity, despite high DERA mRNA levels. This result is in accordance with another study reporting that colon carcinoma LoVo cells do not display any detectable activity [23]. The unexpectedly low deoxyribose phosphate aldolase activity observed in colon cancer cell lines may be due to speciﬁc post-translational events, such as phosphorylation.

Fig. 5.YBX1 knockdown does not impair DERA recruitment to stress granules in HeLa cells. HeLa cells were transiently transfected with control siRNA or YBX1 siRNA for 24 h. (A) YBX1 expression was checked by Western blot using anti-YBX1 and anti-α-tubulin as loading control. (B) Cells were treated with 0.5 mM arsenite for 1 h,ﬁxed and immunostained for DERA (red) and G3BP1 or PABPC1 (green). Merge color is yellow. Scale bars are 10μm.

(11)

Mutagenesis of the bacterial enzyme showed that mutation S239E de- creases its activity[61]. The corresponding position in human DERA, Thr292, was shown to be a possible target for Polo-like kinase 1 (PLK1)[62], which is overexpressed in a majority of cancer cells.

Furthermore, we cannot exclude that DERA protein amounts could be modulated at the levels of protein translation and/or degradation.

Such mechanisms have been recently highlighted to explain the fre- quently observed lack of correlation between the cellular concentration of a protein and its mRNA abundance[63]. Our preliminary studies

using the TargetScan tool (http://www.targetscan.org/) predict that the 3′-UTR of DERA could be targeted by hsa-miR-4766-3p, but the presence of this miRNA in colon cells still needs to be investigated.

We highlighted that cells with high DERA activity can use deoxynucleosides as a source of energy. Indeed, Huh-7 cells endogenously expressing high DERA levels and HeLa cells overexpressing DERA can use extracellular deoxyinosine to recover the ATP pool depleted by FCCP in the absence of glucose. In contrast, cells with low DERA levels, such as wild-type HeLa cells, can convert inosine into ATP Fig. 6.DERA knockdown prevents assembly of stress granules in HeLa cells. HeLa cells were transiently transfected with control shRNA and shRNA targeting DERA coding sequence for 48 h. (A) DERA expression levels measured by qRT-PCR (n = 2). (B, C) HeLa cells were transiently transfected with control shRNA or DERA shRNA, and treated for 1 h with clotrimazole (Clo),ﬁxed and immunostained for DERA, G3BP1 and YBX1. Scale bars are 10μm. The bar graph (C) shows means ± s.e.m. of the number of G3BP1 and YBX1-containing SG analyzed in a minimum of 100 cells from duplicate coverslips. The number of YBX1 and G3BP1-containing stress granules per cell is decreased in cells down-regulating DERA. (D, E) HeLa cells were transiently transfected with control shRNA or DERA shRNA, and treated for 1 h with 0.5 mM arsenite (As),ﬁxed and immunostained for DERA or YBX1 (red) and G3BP1 or DCP1A (green), as indicated. Merge color is yellow. Scale bars are 10μm in images and 2μm in inserts. DERA depletion reduces the number of SG (D) but does not affect DCP1A localization (E) after arsenite treatment. (F) DERA depletion increases the percentage of apoptotic cells stained with annexin V after 1 h clotrimazole treatment. The bar graph shows means ± s.e.m. in eight independent experiments.

(12)

but cannot convert deoxyinosine into ATP, in accordance with what was previously shown in other cells[16]. Interestingly, the liver where we found a high expression of DERA has a high requirement for protein synthesis[64]and therefore needs high levels of energy that DERA may help supporting.

We observed that endogenous DERA is both cytoplasmic and nuclear, as expected from its role in deoxynucleotide metabolism.

More surprisingly, DERA was found to be recruited to SG upon energy starvation, oxidative stress or mitochondrial stress. SG assembly is a major defensive mechanism that the cell puts in place to avoid apoptosis and increase its viability[50]. For example, in erythroid cells that form spontaneous SG, a 60% increase in apoptosis occurs when SG formation is impaired by down-regulation of the core SG protein G3BP1[65]. In addition to mRNA and core mRNA-binding proteins, SG sequester pro- apoptotic proteins such as ROCK1[50]as well as small ribosomal subunits to prevent translation reinitiating[25], and contain proteins that allow quick cell recovery once the stress is over and SG are disassembled, such as HNRNPA1[66], CASC3[67]or DDX3X[68]. Also, DERA may help cells to recover energy after stress, by catabolizing the DR5P released through deoxynucleotide degradation. In line with this hypothesis, we showed that DERA downregulation in HeLa cells decreased the number of clotrimazole- and arsenite-induced SG, and signiﬁcantly increased the percentage of apoptotic cells following clotrimazole stress.

We showed that DERA was able to bind YBX1 in unstressed conditions and under SG-promoting stresses. While YBX1 is known to be targeted to SG via its‘cold-shock’RNA-binding domain[69], DERA has no putative SG-targeting domain and how it is targeted to SG remains unclear. Our results using siRNA against YBX1 indicate that YBX1 is

dispensable for arsenite-induced SG formation and for DERA localization in SG, but we cannot exclude that the little amount of YBX1 left in the cells after siRNA treatment would be sufficient to mediate both processes. Indeed, YBX1 knockdown reduces the number of SG induced by tRNA cleavage, which suggests that it has a role in their assembly[70]. It is still unclear whether DERA acts in SG assembly through its enzymatic activity, or through its interaction with YBX1. To test these two possibil- ities, one should try to rescue the SG formation defects observed in cells downregulating DERA by overexpressing either wild type DERA or its catalytically inactive mutant, which can still bind YBX1. However, this experiment is difficult to perform since we observed that the extracellular addition of DNA plasmids to cells downregulating DERA causes cell death, as it was shown for the addition of exogenous DNA to DeoC- deficient bacteria[17].

Beside its localization in SG upon stress, YBX1 is a cytoplasmic protein that can shuttle to the nucleus upon speciﬁc conditions, and bind to apurinic DNA sites[71], which are among the most frequent hallmarks of DNA damage. It has been shown that stresses that induce SG, such as arsenite, induce high levels of apurinic/apyrimidic DNA damage products[72], and that SG assembly helps in the establishment of the DNA damage response[73]. In addition, proteins involved in DNA repair, such as poly [ADP-ribose] polymerases, localize in SG and are essential for SG assembly[74,75], although their mechanism of action is still unclear. Since DERA was found both in SG and in the nucleus, its possible role in DNA repair in association with YBX1 would merit further investigation.

Looking for proteins influencing SG assembly, Ohn and colleagues identified 131 proteins involved in various biological processes Fig. 7.Cells with high deoxyribose phosphate aldolase activity can use deoxyinosine to preserve ATP levels during metabolic stress. Cells were either left in glucose-containing media (untreated), or grown in a medium without glucose in the presence or absence of FCCP, and incubated with deoxyinosine or inosine for 2 h. (A) HeLa cells were treated as indicated, fixed and immunostained for DERA or YBX1 (red) and PABPC1 (green). Scale bars are 10μm in images and 2μm in inserts. Merge color is yellow. HeLa cells form SG containing DERA, PABPC1 and YBX1 upon energy starvation. (B) ATP levels in Huh-7 cells grown with different concentrations of deoxyinosine or inosine (means ± s.d. values, n = 3). (C) ATP levels in HeLa cells transfected with empty vector, pDERA or pDERA-K218A/K254A (means ± s.d. values, n = 3). (D) Number of G3BP1-containing SG per cell after transfection of HeLa cells with pDERA (black bars) or pDERA-K218A/K254A (dashed bars) and the indicated treatment (means ± s.e.m. of nN100 cells from duplicate coverslips). DERA overexpression allows cells to overcome clotrimazole-induced stress.

(13)

including transcription, protein modiﬁcation, DNA repair or fructose metabolism[75]. Since then, several studies have expanded that list, but the precise role of each protein in SG assembly is not fully under- stood. Our study adds DERA to this list and shows that it interacts with YBX1, which is involved both in DNA repair and SG formation, thus providing unexpected links between deoxynucleotide metabolism, DNA repair and SG assembly pathways.

Supplementary data to this article can be found online athttp://dx.

doi.org/10.1016/j.bbamcr.2014.09.007.

Acknowledgments

We thank R. Fish (University of Geneva) for providing HEK293, Huh-7 and HepG2 cells, D. Massey-Harroche (IBDM, France) for providing Caco-2 cells, C. Arrieumelou (University of Basel) for providing HeLa Kyoto cells, O. Hartley (University of Geneva) for providing Jurkat cells and P. Malinge (NovImmune SA) for technical help. We are grateful to P.-A. Kaminski (Institut Pasteur) for providing DR5P, Pr. M. Uhlen for the anti-YBX1 antibody and P. Viollier for the pGEX3t4 plasmid. We thank C-L Kirkpatrick, P. Duek and D. Belin for critically reading the manuscript, and the whole CALIPHO team for helpful discussions and technical help. This work was supported by the Faculty of Medicine of the University of Geneva.

References

[1] E. Racker, Enzymatic synthesis and breakdown of desoxyribose phosphate, J. Biol.

Chem. 196 (1952) 347–365.

[2] P. Valentin-Hansen, K. Hammer-Jespersen, R.S. Buxton, Evidence for the existence of three promoters for the deo operon ofEscherichia coliK12 in vitro, J. Mol. Biol. 133 (1979) 1–17.

[3] P. Valentin-Hansen, F. Boëtius, K. Hammer-Jespersen, I. Svendsen, The primary structure ofEscherichia coliK12 2-deoxyribose 5-phosphate aldolase. Nucleotide sequence of the deoC gene and the amino acid sequence of the enzyme, Eur. J.

Biochem. 125 (1982) 561–566.

[4] C.F. Barbas, Y.F. Wang, C.H. Wong, Deoxyribose-5-phosphate aldolase as a synthetic catalyst, J. Am. Chem. Soc. 112 (1990) 2013–2014,http://dx.doi.org/10.1021/

ja00161a064.

[5] E.A. Stura, S. Ghosh, E. Garcia-Junceda, L. Chen, C.H. Wong, I.A. Wilson, Crystallization and preliminary crystallographic data for class I deoxyribose-5-phosphate aldolase fromEscherichia coli: an application of reverse screening, Proteins 22 (1995) 67–72,http://dx.doi.org/10.1002/prot.340220110.

[6] A. Heine, G. DeSantis, J.G. Luz, M. Mitchell, C.H. Wong, I.A. Wilson, Observation of covalent intermediates in an enzyme mechanism at atomic resolution, Science 294 (2001) 369–374,http://dx.doi.org/10.1126/science.1063601.

[7] N.K. Lokanath, I. Shiromizu, N. Ohshima, Y. Nodake, M. Sugahara, S. Yokoyama, et al., Structure of aldolase fromThermus thermophilusHB8 showing the contribution of oligomeric state to thermostability, Acta Crystallogr. D Biol. Crystallogr. 60 (2004) 1816–1823,http://dx.doi.org/10.1107/S0907444904020190.

[8] H. Sakuraba, H. Tsuge, I. Shimoya, R. Kawakami, S. Goda, Y. Kawarabayasi, et al., The ﬁrst crystal structure of archaeal aldolase. Unique tetrameric structure of 2-deoxy-d- ribose-5-phosphate aldolase from the hyperthermophilic archaeaAeropyrum pernix, J. Biol. Chem. 278 (2003) 10799–10806,http://dx.doi.org/10.1074/jbc.M212449200.

[9] N.S. Jiang, D.P. Groth, Polycarboxylic acid activation of rat liver deoxyribose phosphate aldolase, J. Biol. Chem. 237 (1962) 3339–3341.

[10]H.G. Roscoe, W.L. Nelson, Deoxyribose Phosphate Aldolase From Rat Liver, J. Biol.

Chem. 239 (1964) 8–11.

[11] D.P. Groth, Deoxyribose 5-phosphate aldolase. II. Puriﬁcation and properties of the rat liver enzyme, J. Biol. Chem. 242 (1967) 155–159.

[12]F. Sgarrella, F.P. Poddie, M.A. Meloni, L. Sciola, P. Pippia, M.G. Tozzi, Channelling of deoxyribose moiety of exogenous DNA into carbohydrate metabolism: role of deoxyriboaldolase, Comp. Biochem. Physiol. B Biochem. Mol. Biol. 117 (1997) 253–257.

[13]J.A. Jedziniak, F.J. Lionetti, Puriﬁcation and properties of deoxyriboaldolase from human erythrocytes, Biochim. Biophys. Acta 212 (1970) 478–487.

[14] D.W. Lincoln, P. Hoffee, Deoxyribose 5-phosphate aldolase: an enzyme that peaks in the G2 phase of rat hepatoma cells, Arch. Biochem. Biophys. 193 (1979) 392–397.

[15] A. Mattana Carta, M. Camici, S. Allegrini, M.G. Tozzi, F. Sgarrella, Catabolism of exogenous deoxyinosine in cultured epithelial amniotic cells, Biochim. Biophys.

Acta 1528 (2001) 74–80.

[16] F. Balestri, M. Giannecchini, F. Sgarrella, M.C. Carta, M.G. Tozzi, M. Camici, Purine and pyrimidine nucleosides preserve human astrocytoma cell adenylate energy charge under ischemic conditions, Neurochem. Int. 50 (2007) 517–523,http://dx.doi.org/

10.1016/j.neuint.2006.10.005.

[17] M.S. Lomax, G.R. Greenberg, Characteristics of the deo operon: role in thymine utilization and sensitivity to deoxyribonucleosides, J. Bacteriol. 96 (1968) 501–514.

[18] M. Friedken, D. Roberts, The enzymatic synthesis of nucleosides. I. Thymidine phosphorylase in mammalian tissue, J. Biol. Chem. 207 (1954) 245–256.

[19] M. Friedken, D. Roberts, The enzymatic synthesis of nucleosides. II. Thymidine and related pyrimidine nucleosides, J. Biol. Chem. 207 (1954) 257–266.

[20] P. Maliekal, T. Sokolova, D. Vertommen, M. Veiga-da-Cunha, E. Van Schaftingen, Molecular identiﬁcation of mammalian phosphopentomutase and glucose-1,6- bisphosphate synthase, two members of the alpha-D-phosphohexomutase family, J. Biol. Chem. 282 (2007) 31844–31851,http://dx.doi.org/10.1074/jbc.M706818200.

[21] A. Bzowska, E. Kulikowska, D. Shugar, Purine nucleoside phosphorylases: properties, functions, and clinical aspects, Pharmacol. Ther. 88 (2000) 349–425.

[22] B.K. Kim, S. Cha, R.E. Parks, Purine nucleoside phosphorylase from human erythrocytes. I. Puriﬁcation and properties, J. Biol. Chem. 243 (1968) 1763–1770.

[23] M. Giannecchini, M. Matteucci, R. Pesi, F. Sgarrella, M.G. Tozzi, M. Camici, Uptake and utilization of nucleosides for energy repletion, Int. J. Biochem. Cell Biol. 37 (2005) 797–808,http://dx.doi.org/10.1016/j.biocel.2004.10.005.

[24] H. Onishi, Y. Kino, T. Morita, E. Futai, N. Sasagawa, S. Ishiura, MBNL1 associates with YB-1 in cytoplasmic stress granules, J. Neurosci. Res. 86 (2008) 1994–2002, http://dx.doi.org/10.1002/jnr.21655.

[25] N. Kedersha, P. Anderson, Regulation of translation by stress granules and processing bodies, Prog. Mol. Biol. Transl. Sci. 90 (2009) 155–185,http://dx.doi.org/10.

1016/S1877-1173(09)90004-7.

[26]UniProt Consortium, The Universal Protein Resource (UniProt) in 2010, Nucleic Acids Res. 38 (2010) D142–D148.

[27] F. Sievers, A. Wilm, D. Dineen, T.J. Gibson, K. Karplus, W. Li, et al., Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega, Mol. Syst. Biol. 7 (2011) 539,http://dx.doi.org/10.1038/msb.2011.75.

[28]J. Vandesompele, K. De Preter, F. Pattyn, B. Poppe, N. Van Roy, A. De Paepe, et al., Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes, Genome Biol. 3 (2002) (RESEARCH0034).

[29] M. Landthaler, D. Gaidatzis, A. Rothballer, P.Y. Chen, S.J. Soll, L. Dinic, et al., Molecular characterization of human Argonaute-containing ribonucleoprotein complexes and their bound target mRNAs, RNA 14 (2008) 2580–2596,http://dx.doi.org/10.1261/

rna.1351608.

[30] G. Magistrelli, P. Malinge, R. Lissilaa, S. Fagète, F. Guilhot, V. Moine, et al., Rapid, simple and high yield production of recombinant proteins in mammalian cells using a versatile episomal system, Protein Expr. Purif. 72 (2010) 209–216, http://dx.doi.org/10.1016/j.pep.2010.04.007.

[31] S.N. Ho, H.D. Hunt, R.M. Horton, J.K. Pullen, L.R. Pease, Site-directed mutagenesis by overlap extension using the polymerase chain reaction, Gene 77 (1989) 51–59.

[32] P. Anderson, N. Kedersha, Stress granules, Curr. Biol. 19 (2009) R397–R398, http://dx.doi.org/10.1016/j.cub.2009.03.013.

[33] P. Hoffee, 2-Deoxyribose gene-enzyme complex in salmonella typhimurium I. Isola- tion and enzymatic characterization of 2-deoxyribose-negative mutants, J. Bacteriol.

95 (1968) 449.

[34] A. Keller, A.I. Nesvizhskii, E. Kolker, R. Aebersold, Empirical statistical model to estimate the accuracy of peptide identiﬁcations made by MS/MS and database search, Anal. Chem. 74 (2002) 5383–5392.

[35] R. Chattopadhyay, S. Das, A.K. Maiti, I. Boldogh, J. Xie, T.K. Hazra, et al., Regulatory role of human AP-endonuclease (APE1/Ref-1) in YB-1-mediated activation of the multidrug resistance gene MDR1, Mol. Cell Biol. 28 (2008) 7066–7080,http://dx.

doi.org/10.1128/MCB.00244-08.

[36] D.R. Marenstein, M.T. Ocampo, M.K. Chan, A. Altamirano, A.K. Basu, R.J. Boorstein, et al., Stimulation of human endonuclease III by Y box-binding protein 1 (DNA- binding protein B). Interaction between a base excision repair enzyme and a transcription factor, J. Biol. Chem. 276 (2001) 21242–21249,http://dx.doi.org/10.1074/

jbc.M101594200.

[37] S. Das, R. Chattopadhyay, K.K. Bhakat, I. Boldogh, K. Kohno, R. Prasad, et al., Stimula- tion of NEIL2-mediated oxidized base excision repair via YB-1 interaction during oxidative stress, J. Biol. Chem. 282 (2007) 28474–28484,http://dx.doi.org/10.

1074/jbc.M704672200.

[38]H.A. Chansky, M. Hu, D.D. Hickstein, L. Yang, Oncogenic TLS/ERG and EWS/Fli-1 fusion proteins inhibit RNA splicing mediated by YB-1 protein, Cancer Res. 61 (2001) 3586–3590.

[39] M.A. Skabkin, V. Evdokimova, A.A. Thomas, L.P. Ovchinnikov, The major messenger ribonucleoprotein particle protein p50 (YB-1) promotes nucleic acid strand annealing, J. Biol. Chem. 276 (2001) 44841–44847,http://dx.doi.org/10.1074/jbc.M107581200.

[40] N. Kedersha, P. Anderson, Mammalian stress granules and processing bodies, Methods Enzymol. 431 (2007) 61–81,http://dx.doi.org/10.1016/S0076-6879(07)31005-7.

[41] S. Fredriksson, M. Gullberg, J. Jarvius, C. Olsson, K. Pietras, S.M. Gústafsdóttir, et al., Protein detection using proximity-dependent DNA ligation assays, Nat. Biotechnol.

20 (2002) 473–477,http://dx.doi.org/10.1038/nbt0502-473.

[42] K. Koike, T. Uchiumi, T. Ohga, S. Toh, M. Wada, K. Kohno, et al., Nuclear translocation of the Y-box binding protein by ultraviolet irradiation, FEBS Lett. 417 (1997) 390–394.

[43] U. Raffetseder, B. Frye, T. Rauen, K. Jürchott, H.-D. Royer, P.L. Jansen, et al., Splicing factor SRp30c interaction with Y-box protein-1 confers nuclear YB-1 shuttling and alternative splice site selection, J. Biol. Chem. 278 (2003) 18241–18248,http://dx.

doi.org/10.1074/jbc.M212518200.

[44] H. Tourrière, K. Chebli, L. Zekri, B. Courselaud, J.M. Blanchard, E. Bertrand, et al., bba, J. Cell Biol. 160 (2003) 823–831,http://dx.doi.org/10.1083/jcb.200212128.

[45]N.L. Kedersha, M. Gupta, W. Li, I. Miller, P. Anderson, RNA-binding proteins TIA-1 and TIAR link the phosphorylation of eIF-2 alpha to the assembly of mammalian stress granules, J. Cell Biol. 147 (1999) 1431–1442.

[46] N. Kedersha, M.R. Cho, W. Li, P.W. Yacono, S. Chen, N. Gilks, et al., Dynamic shuttling of TIA-1 accompanies the recruitment of mRNA to mammalian stress granules, J.

Cell Biol. 151 (2000) 1257–1268.

[47] N. Cougot, S. Babajko, B. Séraphin, Cytoplasmic foci are sites of mRNA decay in human cells, J. Cell Biol. 165 (2004) 31–40,http://dx.doi.org/10.1083/jcb.

200309008.

(14)

[48] S. Souquere, S. Mollet, M. Kress, F. Dautry, G. Pierron, D. Weil, Unravelling the ultra- structure of stress granules and associated P-bodies in human cells, J. Cell Sci. 122 (2009) 3619–3626,http://dx.doi.org/10.1242/jcs.054437.

[49] J.R. Buchan, R. Parker, Eukaryotic stress granules: the ins and outs of translation, Mol. Cell 36 (2009) 932–941,http://dx.doi.org/10.1016/j.molcel.2009.11.020.

[50] N.-P. Tsai, L.-N. Wei, RhoA/ROCK1 signaling regulates stress granule formation and apoptosis, Cell. Signal. 22 (2010) 668–675,http://dx.doi.org/10.1016/j.cellsig.2009.

12.001.

[51] R.S. Buxton, Genetic analysis of thymidine-resistant and low-thymine-requiring mutants ofEscherichia coliK-12 induced by bacteriophage Mu-1, J. Bacteriol. 121 (1975) 475–484.

[52] F. Sgarrella, A. Del Corso, M.G. Tozzi, M. Camici, Deoxyribose 5-phosphate aldolase of Bacillus cereus: puriﬁcation and properties, Biochim. Biophys. Acta 1118 (1992) 130–133.

[53]G.E. Boxer, C.E. Shonk, Deoxyribose5-phosphate metabolism by normal liver and malignant hepatoma, J. Biol. Chem. 233 (1958) 535–540.

[54] D.P. Groth, N. Jiang, The role of deoxyribose 5-phosphate aldolase in the synthesis of deoxyribonucleotide in mammalian cells, Biochem. Biophys. Res. Commun. 22 (1966) 62–68.

[55] J.-M. Fustin, M. Doi, H. Yamada, R. Komatsu, S. Shimba, H. Okamura, Rhythmic nucleotide synthesis in the liver: temporal segregation of metabolites, Cell Rep. 1 (2012) 341–349,http://dx.doi.org/10.1016/j.celrep.2012.03.001.

[56] R.J. Torres, J.G. Puig, H.A. Jinnah, Update on the phenotypic spectrum of Lesch-Nyhan disease and its attenuated variants, Curr. Rheumatol. Rep. 14 (2012) 189–194, http://dx.doi.org/10.1007/s11926-011-0231-5.

[57] J.D. Carver, W. Allan Walker, The role of nucleotides in human nutrition, J. Nutr.

Biochem. 6 (1995) 58–72,http://dx.doi.org/10.1016/0955-2863(94)00019-I.

[58] S. Shin, D.G. Bosc, J.N. Ingle, T.C. Spelsberg, R. Janknecht, Rcl is a novel ETV1/ER81 target gene upregulated in breast tumors, J. Cell. Biochem. 105 (2008) 866–874, http://dx.doi.org/10.1002/jcb.21884.

[59] Y.K. Ghiorghi, K.I. Zeller, C.V. Dang, P.A. Kaminski, The c-Myc target gene Rcl (C6orf108) encodes a novel enzyme, deoxynucleoside 5′-monophosphate N-glycosidase, J. Biol. Chem. 282 (2007) 8150–8156,http://dx.doi.org/10.1074/jbc.

M610648200.

[60] M. Picher, Mechanisms regulating airway nucleotides, Subcell. Biochem. 55 (2011) 17–49,http://dx.doi.org/10.1007/978-94-007-1217-1_2.

[61] G. DeSantis, J. Liu, D.P. Clark, A. Heine, I.A. Wilson, C.-H. Wong, Structure-based mutagenesis approaches toward expanding the substrate speciﬁcity of D-2-deoxyribose-5- phosphate aldolase, Bioorg. Med. Chem. 11 (2003) 43–52.

[62] F.S. Oppermann, K. Grundner-Culemann, C. Kumar, O.J. Gruss, P.V. Jallepalli, H. Daub, Combination of chemical genetics and phosphoproteomics for kinase signaling analysis enables conﬁdent identiﬁcation of cellular downstream targets, Mol. Cell.

Proteomics 11 (2012),http://dx.doi.org/10.1074/mcp.O111.012351(O111.012351).

[63] C. Vogel, E.M. Marcotte, Insights into the regulation of protein abundance from proteomic and transcriptomic analyses, Nat. Rev. Genet. 13 (2012) 227–232, http://dx.doi.org/10.1038/nrg3185.

[64]S.J. James, B.J. Miller, A.G. Basnakian, I.P. Pogribny, M. Pogribna, L. Muskhelishvili, Apoptosis and proliferation under conditions of deoxynucleotide pool imbalance in liver of folate/methyl deﬁcient rats, Carcinogenesis 18 (1997) 287–293.

[65] L. Ghisolﬁ, S. Dutt, M.E. McConkey, B.L. Ebert, P. Anderson, Stress granules contribute toα-globin homeostasis in differentiating erythroid cells, Biochem. Biophys. Res.

Commun. 420 (2012) 768–774,http://dx.doi.org/10.1016/j.bbrc.2012.03.070.

[66] S. Guil, J.C. Long, J.F. Cáceres, hnRNP A1 relocalization to the stress granules reﬂects a role in the stress response, Mol. Cell Biol. 26 (2006) 5744–5758,http://dx.doi.org/

10.1128/MCB.00224-06.

[67] A. Baguet, S. Degot, N. Cougot, E. Bertrand, M.-P. Chenard, C. Wendling, et al., The exon-junction-complex-component metastatic lymph node 51 functions in stress- granule assembly, J. Cell Sci. 120 (2007) 2774–2784,http://dx.doi.org/10.1242/jcs.

009225.

[68] J.-W. Shih, W.-T. Wang, T.-Y. Tsai, C.-Y. Kuo, H.-K. Li, Y.-H.Wu. Lee, Critical roles of RNA helicase DDX3 and its interactions with eIF4E/PABP1 in stress granule assembly and stress response, Biochem. J. 441 (2012) 119–129,http://dx.doi.org/10.

1042/BJ20110739.

[69] W.-H. Yang, D.B. Bloch, Probing the mRNA processing body using protein macroarrays and“autoantigenomics”, RNA 13 (2007) 704–712,http://dx.doi.org/

10.1261/rna.411907.

[70] P. Ivanov, M.M. Emara, J. Villen, S.P. Gygi, P. Anderson, Angiogenin-induced tRNA fragments inhibit translation initiation, Mol. Cell 43 (2011) 613–623,http://dx.doi.

org/10.1016/j.molcel.2011.06.022.

[71] S.L. Hasegawa, P.W. Doetsch, K.K. Hamilton, A.M. Martin, S.A. Okenquist, J. Lenz, et al.

, DNA binding properties of YB-1 and dbpA: binding to double-stranded, single- stranded, and abasic site containing DNAs, Nucleic Acids Res. 19 (1991) 4915–4920.

[72] T. Schwerdtle, I. Walter, I. Mackiw, A. Hartwig, Induction of oxidative DNA damage by arsenite and its trivalent and pentavalent methylated metabolites in cultured human cells and isolated DNA, Carcinogenesis 24 (2003) 967–974.

[73] J. Pothof, N.S. Verkaik, W. van IJcken, E.A.C. Wiemer, V.T.B. Ta, G.T.J. van der Horst, et al., MicroRNA-mediated gene silencing modulates the UV-induced DNA- damage response, EMBO J. 28 (2009) 2090–2099, http://dx.doi.org/10.1038/

emboj.2009.156.

[74] A.K.L. Leung, S. Vyas, J.E. Rood, A. Bhutkar, P.A. Sharp, P. Chang, Poly(ADP-ribose) regulates stress responses and microRNA activity in the cytoplasm, Mol. Cell 42 (2011) 489–499,http://dx.doi.org/10.1016/j.molcel.2011.04.015.

[75] T. Ohn, N. Kedersha, T. Hickman, S. Tisdale, P. Anderson, A functional RNAi screen links O-GlcNAc modiﬁcation of ribosomal proteins to stress granule and processing body assembly, Nat. Cell Biol. 10 (2008) 1224–1231,http://dx.doi.org/10.1038/

ncb1783.