Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/yexcr

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3

http://dx.doi.org/10.10014-4827/& 2014 E

nCorresponding autE-mail address: v

Research Article

MxA interacts with and is modified by theSUMOylation machinery

Carlos Eduardo Brantis-de-Carvalhoa,b, Ghizlane Maarifib,Paulo Eduardo Gonçalves Boldrina, Cleslei Fernando Zanellia,Sébastien Nisoleb, Mounira K. Chelbi-Alixb, Sandro Roberto Valentinia,naDepartment of Biological Sciences, School of Pharmaceutical Sciences, Univ Estadual Paulista - UNESP,Araraquara 14801-902, SP, BrazilbINSERM UMR-S 1124, Université Paris Descartes, Paris 75006, France

a r t i c l e i n f o r m a t i o n

Article Chronology:

Received 2 June 2014

Received in revised form22 October 2014Accepted 23 October 2014Available online 31 October 2014

Keywords:

MX1

MxAYeast two-hybridSUMOylationEIL loopSUMOUbc9 and antiviral activity

016/j.yexcr.2014.10.020lsevier Inc. All rights reser

hor. Fax: þ55 16 3301 6940alentsr@fcfar.unesp.br (S.R

a b s t r a c t

Mx proteins are evolutionarily conserved dynamin-like large GTPases involved in viral resistancetriggered by types I and III interferons. The human MxA is a cytoplasmic protein that confers

resistance to a large number of viruses. The MxA protein is also known to self-assembly into highmolecular weight homo-oligomers. Using a yeast two-hybrid screen, we identified 27 MxAbinding partners, some of which are related to the SUMOylation machinery. The interaction ofMxA with Small-Ubiquitin MOdifier 1 (SUMO1) and Ubiquitin conjugating enzyme 9 (Ubc9) wasconfirmed by co-immunoprecipitation and co-localization by confocal microscopy. We identifiedone SUMO conjugation site at lysine 48 and two putative SUMO interacting motifs (SIMa andSIMb). We showed that MxA interacts with the EIL loop of SUMO1 in a SIM-independent mannervia its CID–GED domain. The yeast two-hybrid mapping also revealed that Ubc9 binds to the MxAGTPase domain. Mutation in the putative SIMa and SIMb, which are located in the GTPase bindingdomain, reduced MxA antiviral activity. In addition, we showed that MxA can be conjugated toSUMO2 or SUMO3 at lysine 48 and that the SUMOylation-deficient mutant of MxA (MxAK48R)

retained its capacity to oligomerize and to inhibit Vesicular Stomatitis Virus (VSV) and InfluenzaA Virus replication, suggesting that MxA SUMOylation is not essential for its antiviral activity.

& 2014 Elsevier Inc. All rights reserved.

Introduction

Viral resistance can be induced by treatment of cells with interferons(IFNs) [1]. This resistance is caused by the expression of a largenumber of IFN-stimulated genes including MX1 [2]. MX1 was firstdiscovered due to the resistant phenotype of some inbred mice to theinfection by Influenza A Virus (FLUAV) [3,4]. These mice weredescribed to carry a functional copy of the MX1 gene, which wasnamed Myxovirus resistance gene 1 due to its antiviral activity against

ved.

.

. Valentini).

viruses of the Orthomyxoviridae family. Most of the vertebrates,including humans, rodents, fish and birds, and some invertebratesare known to have one or more copies of the MX gene [5]. Humansharbor two MX (MX1 and MX2) genes in the chromosome 21 [6],which are induced only in response to type I or type III IFNs viaactivation of the JAK/STAT pathway [7–10].MX1 and MX2 gene products are MxA and MxB, respectively.

MxA is a cytoplasmic protein [11] while MxB can be found in thecytoplasm and in the nucleus [12]. Mx proteins are classified as

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3152

large dynamin-like GTPases due to their similarities to dynamins[5]. MxA and dynamins share structural and functional aspectsincluding domain organization, GTPase activity and homo-oligomerization capacity [13].Based on sequence alignment with dynamins, MxA was initially

considered to be organized in three domains: a N-terminal GTP-binding domain (GTPase), a central interactive domain (CID) and aC-terminal GTPase effector domain (GED) [13]. Recently, crystal-lography studies have revealed that MxA is in fact composed oftwo domains connected to each other by a central region. The firstdomain still corresponds to the GTPase domain, which binds andhydrolyzes GTP. However, according to this new nomenclature,the two domains previously called CID and GED are now involvedin the formation of a single tertiary structure named stalk, whichcoordinates the formation of MxA homo-oligomers. The connec-tive central region, called Bundle-Signaling Element (BSE), standsbetween the GTPase and stalk domains. The BSE is generated bysequences present in the N-terminal, C-terminal and the middleportion of MxA, between the GTPase and stalk domains. The BSEexerts an important function in the formation of MxA homo-oligomers [14]. The integrity of both GTPase and stalk domains isnecessary for the correct assembly of MxA homo-oligomers [14–17]. MxA homo-oligomerization is required for its GTPase activity[18,19]. However, it is not clear whether GTPase activity isrequired or if GTP-binding alone is sufficient to confer antiviralactivity [5,20]. Accordingly, the L612K monomeric MxA mutantlacks detectable GTPase activity but retains its GTP-bindingcapacity and antiviral activity against Thogoto virus (THOV) andVesicular Stomatitis Virus (VSV) [21].MxA acts as a mediator of the IFN-induced antiviral state,

indeed its expression confers resistance to a large number of RNAviruses including species belonging to the Orthomyxoviridae andRhabdoviridae families [5,11,22,23].The SUMOylation machinery modifies many cellular proteins

involved in the antiviral response as well as viral proteins. Someviruses have developed mechanisms to evade host response bytargeting components of the SUMOylation machinery [24–26].SUMO (Small-Ubiquitin MOdifier) proteins are members of the

ubiquitin-like family of modifiers [27]. Mammals have four SUMOproteins named SUMO1 through SUMO4. SUMO1, SUMO2 andSUMO3 have a widespread expression, whereas SUMO4 expres-sion seems to be restricted to kidney, lymph nodes and spleen[28]. SUMO interacts with proteins by two mechanisms: (1) it canbe covalently attached to proteins that contain a consensus motifof SUMOylation (ΨKxE/D), where a glycine residue of SUMO islinked to the lysine residue of the target protein and/or (2) it caninteract in a non-covalent manner to proteins containing a SUMO-interacting motif (SIM), composed of an hydrophobic core(ΨΨxΨ). While SUMO1–3 paralogs are covalently attached toproteins under normal conditions, SUMO4 seems to be attachedto its targets only under stress conditions [27,29]. SUMO2–3, butnot SUMO1, can lead to the formation of poly-SUMO2/3 chains[32–33]. SUMOylated proteins are found mainly in the nucleus,but they can also be found in the cytoplasm [30,31].Protein SUMOylation is mainly involved in the modification of

protein activities, localization or stability [27,32,33]. Moreover,SUMOylation contributes to the regulation of gene transcription,cell apoptosis, intracellular stress response and cell cycle progres-sion [33]. These processes can be associated with the presence ofSUMOylated proteins in PML NBs (ProMyelocytic Leukemia

protein Nuclear Bodies), subnuclear structures where SUMO andSUMOylated proteins are found [34]. Almost 40% of PML partnershave been confirmed to be SUMOylated, suggesting that PML NBsare enriched sites for SUMOylated proteins and may function asnuclear SUMOylation hotspots [34].

Here we found by a yeast two-hybrid system that MxA interactswith 27 nuclear and cytoplasmic putative partners, includingSUMO1, Ubc9, PML NB-associated proteins and proteins implicatedin cell cycle control. MxA interaction with SUMO1 and Ubc9 hasbeen validated by co-immunoprecipitation and co-localization. Inaddition, we showed that MxA contains one site of SUMOylation(K48) used for the covalent conjugation to SUMO2 or SUMO3. Weidentified two putative SIM sequences, SIMa (V260V261D262V263) andSIMb (V171P172D173L174T175), on MxA primary structure that are notinvolved in the interaction with SUMO1.

We have determined that the MxA CID–GED (stalk) domain isinvolved in the interaction with SUMO1 in a SIM-independentmanner. Furthermore, the interaction between SUMO1 and MxArequired the E67-interacting loop (EIL) but not the SUMO-interacting groove (SIG). In addition, the MxA oligomerizationcapacity was important for its interaction with SUMO1 and Ubc9.

Finally, the MxA SUMOylation-deficient mutant (MxAK48R) stilloligomerizes and inhibits VSV and Influenza A Virus replication,suggesting that MxA covalent modification by SUMO is notessential for its antiviral activity. The SIMa (MxAV260A–V261A–

D262A–V263A) and SIMb (MxAV171A–P172A–L174A) mutants, which arelocated in the GTPase domain, have a reduced antiviral activity.

Materials and methods

Strains, plasmids, chemicals and antibodies

Plasmids and primers used in this study are listed inSupplementary Tables 1 and 2. All chemicals were obtained fromSigma, United States Biological or Thermo Scientific unless other-wise indicated.

The rabbit anti-MxA polyclonal antibody was produced for thisstudy. The efficacy of the polyclonal anti-MxA (α-pMxA) wasdetermined by immunofluorescence and Western blot as shownin Supplementary Fig. S1. The mouse monoclonal anti-MxA anti-body (α-mMxA) was provided by Dr. Otto Haller. The rabbit anti-VSV polyclonal antibody was provided by Dr. Danielle Blondel.The Rabbit anti-PML (sc-5621), rabbit anti-SUMO1 (Sc9060),rabbit anti-LexA and rabbit anti-Actin clone C-11 (sc-1615) anti-bodies were from Santa Cruz Biotechnology. The mouse anti-6xHis, the mouse anti-Ubc9 (AM1261a) and the rabbit anti-SUMO2/3 antibodies were from Clontech, Abgent and Invitrogen,respectively.

Cell culture

Human HeLa cells and murine NIH3T3 cells were grown at 37 1Cin DMEM supplemented with 10% FCS. NIH3T3 harboring theempty vector or the plasmid coding for MxA, kindly provided byDr. J. Pavlovic, were kept in medium supplemented with 0.5 mg/mL of neomycin [11].

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 153

Viral stocks

Stock of VSV (Mudd-Summer strain, Indiana serotype) (109 PFU/ml) and of Influenza A Virus (WSN strain) (4�108 PFU/ml) wasprepared from supernatants of BSR and NIH 3T3 infected cells,respectively.

Western blot analysis

Cells were washed and re-suspended in PBS, lysed in hot Laemmlisample buffer and boiled for 5 min. Approximately 30 μg ofproteins was resolved on a 10% SDS-PAGE gel and transferred toa nitrocellulose membrane. The membrane was blocked with 5%skimmed milk in PBS for 2 h and incubated overnight at 4 1C withthe primary antibody of interest. The membrane was then washedextensively in PBS-Tween and incubated for 1 h with the appro-priate peroxidase-coupled secondary antibody. All blots wererevealed by chemiluminescence (ECL, Amershan).

Yeast two-hybrid system

The yeast two-hybrid screen was performed using the Sacchar-omyces cerevisiae L40 (SVL86—MATa trp1 his3 leu2 ade2 LYS2::lexAop(4x)-HIS3 URA3::lexAop(8x)-lacZ) strain that harbors theHIS3 and lacZ reporter genes [35]. The strain was transformedwith the bait plasmid pBTM116-KAN encoding an in-frame fusionof the lexA DNA binding protein with the coding sequence of MX1.Expression of the bait protein (lexA-MxA) in L40 cells wasconfirmed by Western blot using an anti-lexA antibody. L40 cellscontaining the MxA bait plasmid were transformed with plasmidsof a Human Fetal Brain cDNA library fused to the Gal4 activationdomain (Clontech), constructed in the pACT2 vector. Transfor-mants were plated onto selective medium (-Leu/Trp/His) andincubated for 3 days at 30 1C. The Hisþ transformants werefurther screened for the formation of blue colonies in the β-galactosidase lift assay with 5-bromo-4-chloro-3-indolyl β-D-galactoside (X-gal) as a complementary test for interaction [36].The library plasmids from initially selected clones were isolated,amplified in E. coli TOP10, retransformed into L40 cells, and bothassays repeated. Plasmid DNA was isolated from positive clonesand sequenced to identify the genes encoding the interactingproteins.

The mapping and mutant interaction by yeast two-hybridassays were performed in a similar way. Briefly, the L40 strainof S. cerevisiae was co-transformed with one of the MX1pBTM116-KAN plasmids, harboring truncated or mutated formsof MxA, and one of the pACT2 plasmid constructs encoding wildtype or mutants of SUMO1/2/3 and Ubc9. The interactions wereassayed by the expression of the reporter genes.

Co-immunoprecipitation assay

Cells were lysed two days post-transfection in 50 mM Tris, 0.1%Nonidet P-40, 5 mM MgCl2, and 0.5 mM dithiothreitol, pH 7.5.After removal of cell debris (15 min at 12,000g), supernatantswere used for co-immunoprecipitation. First, lysates of MxA and6xHis-SUMO or MxA and 6xHis-Ubc9-expressing cells weremixed with a mouse monoclonal anti-6xHis antibody and withprotein A-Sepharose beads (Qiagen). After 1 h incubation, theimmunocomplex bound to protein A-Sepharose beads was

washed with lysis buffer. The immunocomplex was then incu-bated for 2 h at 4 1C in the presence of 150 mM NaCl. Followingintensive washing with lysis buffer containing 150 mM NaCl, thebound proteins were resuspended in SDS sample buffer. Co-precipitated proteins were detected by Western blot using α-pMxA, anti-Ubc9, anti-SUMO1 or anti-6xHis antibodies.

SUMOylation assay by purification of 6xHis-tagged SUMOconjugates

Two days post-transfection, cells were lysed in 4 mL of 6 Mguanidinium-HCl, 0.1 M Na2HPO4/NaH2PO4, and 0.01 M Tris–HCl,pH 8.0, plus 5 mM imidazole and 10 mM β-mercaptoethanol per75-cm2

flask. After sonication to reduce viscosity, the lysates weremixed with 50 μL of Ni2þ-nitrilo-triacetic acid (NTA)-agarosebeads (Qiagen) prewashed with lysis buffer and incubated for2 h at room temperature. The beads were successively washedwith the following: 6 M guanidinium-HCl, 0.1 M Na2HPO4/NaH2PO4, 0.01 M Tris–HCl, pH 8.0, 10 mM β-mercaptoethanol;8 M urea, 0.1 M Na2HPO4/NaH2PO4, 0.01 M Tris–HCl, pH 8.0,10 mM β-mercaptoethanol, 6 M guanidinium-HCl, 0.1 MNa2HPO4/NaH2PO4, 0.01 M Tris–HCl, pH 6.3, 10 mM β-mercaptoethanol (buffer A); buffer A plus 0.2% Triton X-100;and buffer A plus 0.1% Triton X-100. After the last wash withbuffer A, the beads were treated with 200 mM imidazole. Theeluates were subjected to Western blotting with α-pMxA, anti-6xHis or anti-PML antibodies.

Immunofluorescence microscopy

For the co-localization assay, cells transiently expressing MxAwere grown on a cover slip, fixed in 4% paraformaldehyde for15 min at 4 1C and permeabilized for 5 min with 0.1% Triton X-100in phosphate-buffered saline (PBS). Then, cells were prepared fordouble-immunofluorescence staining with rabbit or mouse anti-MxA antibody and rabbit anti-SUMO1 or mouse anti-Ubc9 anti-body. Cells were then washed twice and incubated for 1 h withthe appropriate Alexa Fluor-conjugated secondary antibody(Alexa Fluor 488 and 594—Molecular Probe, Inc.). The cells weremounted onto glass slides by using Immu-Mount (Shandon)containing 4,6-diamidino-2-phenylindole (DAPI) to stain nucleiand analyzed by confocal laser microscopy using a Zeiss LSM 710microscope.

Results

Searching for MxA binding partners

We have previously described that the gene encoding MxA issilenced by methylation in head and neck tumors [37]. In order tofurther investigate the MxA mechanisms of action, we performeda yeast two-hybrid screen to look for new putative bindingpartners of the human MxA protein. The coding region of thehuman MX1 gene was amplified by PCR from pCMV6-XL5-MX1plasmid (Origene) and cloned into the EcoRI site of the two-hybridvector pBTM116-KAN, generating a plasmid encoding the fusionprotein LexA-MxA (pBTM116-KAN-MX1). This plasmid wasinserted into the Saccharomyces cerevisiae L40 strain and theproduction of the fusion LexA-MxA was confirmed by Western

Table 1 – MxA putative binding partners revealed by the yeast two-hybrid screen.

Clone Gene symbol Gene ID Gene name

1 DAXX 1616 Death-domain associated protein2 RelA (p65) 5970 V-rel avian reticuloendotheliosis viral oncogene homolog A3 SUMO1 7341 Small ubiquitin-like modifier 110 RUSC2 9852 RUN and SH3 domain containing 211 MxA (MX1) 4599 Myxovirus (influenza virus) resistance 119 ZBTB16 (PLZF) 7704 Zinc finger and BTB domain containing 1621 Ubc9 (UBE2I) 7329 Ubiquitin-conjugating enzyme E2I38 KIF26B 55083 Kinesin family member 26B41 ZNF251 90987 Zinc finger protein 25148 FLASH (CASP8AP2) 9994 Caspase 8 associated protein 252 ZNF623 9831 Zinc finger protein 62354 PSD3 23362 Pleckstrin and Sec7 domain containing 356 ZCCHC12 170261 Zinc finger, CCHC domain containing 1260 PIAS1 8554 Protein inhibitor of activated STAT 169 CBX4 (Pc2) 8535 Chromobox homolog 476 KLHL35 283212 Kelch-like family member 3577 C7orf25 79020 Chromosome 7 open reading frame 2586 CHD3 1107 Chromodomain helicase DNA binding protein 390 ADRM1 11047 Adhesion regulating molecule 198 BRD7 29117 Bromodomain containing 7101 TDG 6996 Thymine-DNA glycosylase105 PLRG1 5356 Pleiotropic regulator 1109 TUBA1A 7846 Tubulin, alpha 1a120 HMGXB4 10042 HMG box domain containing 4133 SLC25A3 5250 Solute carrier family 25 (mitochondrial carrier; phosphate carrier), member 3137 LRRC4B 94030 Leucine rich repeat containing 4B138 GMEB1 10691 Glucocorticoid modulatory element binding protein 1

Putative MxA partners in common with Mx1 [38] are shown in bold.

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3154

blot (data not shown). To test for MxA self-activation in the yeasttwo-hybrid system, the L40 yeast strain harboring the plasmidpBTM116-KAN-MX1 was also transformed with the vector pACT2and then assayed for the activation of HIS3 and lacZ reportergenes. MxA was not able to activate by itself either reporter genes(data not shown). For the screen, the L40 yeast strain harboringthe pBTM116-KAN-MX1 was transformed with a human fetalbrain cDNA library (Clontech) constructed in the pACT2 vectorand plated onto SC -leu,-trp,-his selective medium. Clones wereselected and analyzed as described in Material and methods.This screen for new MxA binding partners by yeast two-hybrid

revealed a total of 27 putative ligands shown in Table 1. Amongthese ligands are nine proteins already described as mouse Mx1partners, also identified by yeast two-hybrid screen [38]. Theseputative ligands can be grouped into four different biologicalprocesses by gene ontology: cell cycle control, transcriptionregulation, apoptosis and DNA damage response. Additionalligands are PML NB partners and proteins related to the SUMOy-lation process. Moreover, some of the proteins listed in Table 1 areknown to be SUMOylated (FLASH, DAXX, PIAS1, CBX4, TDG, Ubc9and PLZF) and two of them (PIAS1 and CBX4) have SUMO E3-ligase activity [39–47].

MxA binds to SUMO1 and Ubc9

To confirm MxA interaction with two essential components of theSUMOylation machinery, SUMO1 and Ubc9, we used yeast two-hybrid and co-immunopreciptation assays. Although the paralogs

SUMO2 and SUMO3 were not revealed in the two-hybrid screen,their interaction with MxA was also investigated.

For the yeast two-hybrid assay, we amplified by PCR the codingregion of SUMO1/2/3 and Ubc9 from the plasmids pcDNA3.1-6xHis-SUMO1, pcDNA3.1-6xHis-SUMO2, pcDNA3.1-6xHis-SUMO3and pLTR-6xHis-Ubc9. The products of amplification were clonedinto the EcoRI/BamHI sites of the pACT2 vector and the expressionof SUMO1/2/3 and Ubc9 was confirmed by Western blot. All theseproteins did not self-activate the reporter genes of the two-hybridsystem (data not shown). The L40 strain containing the pBTM116-KAN-MX1 was transformed independently with one of these fourplasmids and tested for the expression of the reporter genes. MxAwas confirmed to interact with Ubc9 and SUMO1, but not withSUMO2 and SUMO3 (Fig. 1A).

The interaction of MxA with SUMO1 and Ubc9 was also assayedby co-immunoprecipitation using protein extracts from HeLa cellsco-transfected with pcDNA3.1-MX1 and pcDNA3.1-6xHis-SUMO1or pLTR-6XHis-Ubc9. As shown in Fig. 1B, MxA physically inter-acted with Ubc9 and SUMO1. As SUMO1 and Ubc9 act together inprotein SUMOylation [48], we also tested the interaction betweenSUMO1 and MxA in the presence of Ubc9. In this assay, HeLa cellswere co-transfected with the plasmids expressing MxA, SUMO1and Ubc9. Surprisingly, the presence of SUMO1 negatively inter-fered with the interaction between MxA and Ubc9 (Fig. 1B).Interaction of MxA with SUMO2 and SUMO3 was also assayedby co-immunoprecipitation but no interaction was revealed (datanot shown).

Additionally, confocal microscopy was used to investigate the co-localization of MxA and SUMO1 or Ubc9 in HeLa cells transiently

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 155

expressing MxA. The double immunofluorescence stainingrevealed that MxA partially co-localized with SUMO1 and Ubc9in the cytoplasm (Fig. 1C), corroborating the interaction resultsobtained by yeast two-hybrid and co-immunoprecipitation.

MxA binds to the EIL loop of SUMO1 via the CID–GEDdomains

To map the MxA binding site to SUMO1, the software SUMOsp[49] was initially used to predict SUMO-Interacting Motif (SIM) inthe MxA sequence. This analysis led to the prediction of a SIMa(VVDV, V260–V263) and a SIMb (VPDLT, V171–T175) in the MxAGTPase domain. These sites were mutated by site-directed muta-genesis using the pBTM116-KAN-MX1 plasmid as template, gen-erating two mutated forms of MxA (MxASIMa—V260A, V261A,D262A, V263A- and MxASIMb—V171A, P172A, L174A). Additionally,we also obtained two truncated forms of MxA in the same vector,encoding only the GTPase domain (1-340 aa – MxAGTPase) or thecombination of both CID and GED domains (341-662 aa – MxACID–

GED). To determine the impact of MxA oligomerization on its

interaction with SUMO1 and Ubc9, the monomeric mutant ofMxA (MxAL612K) was also obtained by site-directed mutagenesisusing the pBTM116-KAN-MX1 plasmid as template. A schematicrepresentation of the mutated and truncated forms of MxA isshown in Fig. 2A. The expression of these MxA proteins wasconfirmed by Western blot (Fig. 2B).As shown in Fig. 2C, the mutants MxASIMa and MxASIMb

unexpectedly conserved their capacity to interact with SUMO1by yeast two-hybrid, demonstrating that these motifs are notrequired for this interaction. Moreover, the MxA interaction withSUMO1 was abolished in the MxAGTPase truncation whereas it waspreserved in the MxACID–GED construct. Conversely, the MxAinteraction with Ubc9 was maintained with the MxAGTPase trunca-tion, but not with the MxACID–GED construct. As both thesetruncated forms of MxA are unable to oligomerize, their interac-tions with SUMO1 or Ubc9 are not as robust as the interactionobserved with wild type MxA. This is in agreement with theresults obtained with the monomeric mutant MxAL612K, whichshows a decrease in the interaction with SUMO1 and Ubc9. Takentogether, these results suggest that the interaction between MxAand SUMO1 or MxA and Ubc9 occurs via the CID–GED (stalk)domains or the GTPase domain, respectively, and is affected bythe loss of the oligomerization capacity.To complement the mapping analysis, we tested the impact of

mutations in the well-known SIG region of SUMO1 frequently usedin its interaction with other proteins. In this case, the SUMO1 genewas mutated in the SIG region to generate the SUMO1K37AK39Amutant, which is unable to interact with any SIM. Additionally, weinvestigated the participation of the recently described E67-interacting loop (EIL) of SUMO1 [50,51] in the interaction withMxA. Therefore, the SUMO1 gene was mutated in the EIL loop togenerate the SUMO1E67A mutant, which causes the loss of

Fig. 1 – MxA interacts with SUMO1 and Ubc9. (A) Yeast two-hybridassay. Transformants of the S. cerevisiae L40 strain containing theplasmid pBTM116-MX1WT and one of the pACT2 constructs(pACT2-6xHis-SUMO1, pACT2-6xHis-SUMO2, pACT2-6xHis-SUMO3or pACT2-6xHis-Ubc9) were cultivated on SC -LT and SC -LTHcontaining 0.5mM of 3AT, and tested for β-galactosidase activity.L40/pBTM116-TIF51AþpACT-Dys1 (þþ), L40/pBTM116-TIF51AþpACT-Lia1(þ), L40/pBTM116-TIF51AþpACT (�) were usedas controls for two-hybrid interactions [63]. (B) Co-immunoprecipitation assay. Total extracts from HeLa cellstransfected (þ) or not (�) with the plasmids encoding theindicated proteins were incubated with protein G-Sepharose beadsand anti-6xHis antibody. The co-immunoprecipitated MxA wasdetected with α-pMxA antibody (upper panel). The presence ofMxA, SUMO1 and Ubc9 in the cellular lysates (Input) wasconfirmed by Western blot (lower panel). As the anti-Ubc9antibody also detected the endogenous Ubc9 protein, the anti-6xHis antibody was used to confirm the expression of theexogenous his-tagged Ubc9 in the immunoprecipitated complex(middle panel). (C) Confocal microscopy. HeLa cells transientlyexpressing MxA were stained for MxA (red – Alexa Fluor 594) andendogenous Ubc9 (green – Alexa Fluor 488) or endogenousSUMO1 (green – Alexa Fluor 488) by the use of α-pMxA andmouseanti-Ubc9 or α-mMxA and rabbit anti-SUMO1. Cell nucleus wasstained with DAPI. Merge images were obtained by the use ofImageJ software.

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3156

interaction between SUMO1 and DPP8/9. Also, the genes encodingSUMO2 and SUMO3 were mutated in the region corresponding tothe EIL loop in SUMO1, generating SUMO2D71H and SUMO3D70Hmutants. Schematic representations of mutated forms of SUMO1/2/

3 are shown in Fig. 3A. All SUMO1/2/3 mutants were tested forprotein production by Western blot (Fig. 3B).

The results presented in Fig. 3C show that MxA still interactswith SUMO1K37AK39A, suggesting that the SIG region of SUMO1 is

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 157

not required for its interaction with MxA. As the SIG region isinvolved in the interaction with SIM sequences, this resultcorroborates the finding that the SIM sequences present in theGTPase domain of MxA are not required for its interaction withSUMO1 (Fig. 2C). Interestingly, the mutation SUMO1E67A drama-tically decreased the interaction with MxA, indicating that the EILloop is the binding site of SUMO1 to MxA. As expected, themutations SUMO2D71H and SUMO3D70H promoted the interactionof these paralogs with MxA, reinforcing the implication of the EILloop of SUMO1 in the interaction with MxA.

MxA is SUMOylated at the lysine K48 residue

To determine whether MxA undergoes SUMOylation, we per-formed an Ni2þ-NTA purification of 6xHis-tagged SUMO conju-gates assay [52]. Purified SUMO-conjugates from HeLa cellstransiently expressing MxA and 6xHis-SUMO1, 6xHis-SUMO2 or6xHis-SUMO3 revealed that MxA is modified by SUMO2 orSUMO3 (MxASUMO) (Fig. 4A). The SUMO1 modification of MxAby Ni2þ-NTA was not detected (Fig. 4A); this is in line with thenotion that SUMO1 conjugation is more difficult to detect thanSUMO2/3 conjugation [53].

The software SUMOsp was then used to predict SUMOylationsites on MxA and the lysines K48 and K259 were highlighted aspossible SUMOylation acceptor sites. These two lysines arelocated in highly conserved regions of MxA. The lysine K48 islocated in the N-terminal alpha1 helix that is part of the so-calledBundle-Signalling Element connecting the G domain to the stalk,while the lysine K259 is present at the GTPase domain, near toone of the GTP-binding motifs. Both of them encompass theinverted type of SUMOylation consensus motif (ICM – E/DxKΨ).

To investigate the participation of lysine 48 (EEKV, E46–V49)and lysine 259 (EDKV, E257–V260) in MxA SUMOylation, themutants MxAK48R and MxAK259R were generated in the pcDNA3.1-MX1 plasmid by site-directed mutagenesis. The position of thesemutations on the MxA structure is shown in Fig. 2A. Western blotanalysis revealed that MxAK259R expression was very low com-pared to that of MxAWT or MxAK48R, probably due to proteininstability (Fig. 4B). Therefore, the participation of the lysine K259in MxA SUMOylation could not be investigated. In the SUMOyla-tion assay, we used extracts from HeLa cells co-transfected withplasmids encoding 6xHis-SUMO3 and MxAWT or MxAK48R. Asseen in Fig. 4C, the K48 lysine of MxA is the acceptor site ofSUMOylation as no SUMOylated bands were revealed in Ni2þ-NTApurification of extracts from cells expressing MxAK48R.

Next, we asked whether endogenous MxA could be conjugatedto SUMO. To do this, cells transfected with the empty vector orHis-SUMO3 were treated with 1000 U/mL of IFNα for 16 h to

Fig. 2 – Mapping the binding sites of MxA to SUMO1 or Ubc9. (A) Schemof MxA are indicated. The sequences necessary for GTP binding (GXXXCin the GTPase domain and the different portions of the Bundle-Signalinall MxA mutants used in this study are represented (see also SupplemTotal protein extracts from transformants of the L40 strain containingMX1L612K, MX1GTPase and MX1CID–GED) were tested for MxA productionthe pBTM116-KAN vector were used as controls. (C) Yeast two-hybrid mcontaining one of the pBTM116-KAN constructs (MX1WT, MX1SIMb, MXconstructs (SUMO1 or Ubc9) were cultivated on SC -LT and SC -LTH conpBTM116-TIF51AþpACT-Dys1 (þþ), L40/pBTM116-TIF51AþpACT-Lia1(þ)hybrid interactions [63].

induce MxA expression and cell extracts purified on Ni2þ-NTA-agarose beads were analyzed by Western blot. Endogenous MxAwas found to be conjugated to SUMO3 in Ni2þ-NTA purifiedextracts from cells expressing SUMO3 treated with IFNα (Fig. 4D).

SUMOylation-deficient mutant MxAK48R oligomerizes andretains the antiviral activity

To test the oligomerization capacity of MxA and its mutants, weperformed Western blot analysis, under native conditions, ofprotein extracts from HeLa cells transiently expressing MxAWT,MxAL612K, MxASIMa, MxASIMb or MxAK48R. This analysis revealedthat MxASIMa, MxASIMb and MxAK48R mutants retain their oligo-merization capacity (Fig. 5A). We found, as previously described[21], that MxAL612K mutant loses the capacity to produce oligo-mers. Immunofluorescence analysis showed that MxASIMa, MxA-

SIMb or MxAK48R mutants are still able to form granular dots in thecytoplasm, whereas the monomeric mutant MxAL612K is diffuse inthe cytoplasm (Fig. 5B).Overexpression of MxA protein was shown to confer resistance

to VSV [11]. We ought to determine the effects of these mutationson MxA antiviral activity by infecting HeLa cells transientlyexpressing MxAWT, MxAL612K, MxAK48R, MxASIMa or MxASIMb withVSV at a multiplicity of infection (MOI) of 1 for 8 h. Antiviralactivity was assayed by double immunofluorescence performedusing α-mMxA and rabbit anti-VSV antibodies. As expected, cellsexpressing MxAWT or MxAL612K inhibited VSV protein synthesis(Fig. 5C) whereas cells expressing the MxASIMa or MxASIMb mutantexpressed viral proteins. Thus, mutation of the putative SIMs isaccompanied by a loss of MxA anti-VSV activity. Strikingly, cellsexpressing the mutant MxAK48R prevented the accumulation ofVSV proteins, demonstrating that MxA covalent SUMOylation isnot required for its antiviral activity against VSV.The SIMa and SIMb predicted here are located close to the

Switch II region of the MxA GTPase domain (Fig. 2A). Anymutation in these motifs could affect GTPase activity and mayexplain the loss of anti-VSV activity. Also we have shown by theyeast two hybrid system that these putative SIMs are not involvedin the interaction of MxA with SUMO1 (Fig. 2C). Therefore, thenext experiments were performed only with MxAK48R.To further confirm that the effect of SUMOylation deficient MxA

mutant MxAK48R on VSV, NIH 3T3 cells transfected with plasmidsencoding for MxAWT or MxAK48R were infected with VSV at a MOIof 1 and Western blot analysis of VSV protein synthesis was carriedout 8 h post-infection (Fig. 6A). The expression level of MxA andMxAK48R was similar in transfected cells (Fig. 6A). VSV G, N and Mproteins were highly expressed in control cells transfected with theempty vector and, as previously reported [11], their synthesis was

atics of mutated and truncated forms of MxA. The three domainsGKS, DXXG and TKXG), the Switch regions (Switch I and Switch II)g Elements (BSE) of MxA are also indicated. Finally, the locations ofentary Fig. S2). (B) Evaluation of the yeast two-hydrid constructs.one of the pBTM116-KAN constructs (MX1WT, MX1SIMb, MX1SIMa,

by Western blot using anti-LexA antibody. L40 and L40 containingapping of MxA interacting sites. Transformants of the L40 strain

1SIMa, MX1GTPase, MX1CID–GED or MX1L612K) and one of the pACT2taining 0.5mM of 3AT, and tested for β-galactosidase activity. L40/, L40/pBTM116-TIF51AþpACT (�) were used as controls for two-

Fig. 3 – Mapping the binding site of SUMO1 to MxA. (A)Schematics of mutated forms of SUMO1/2/3. The SUMO-Interacting Groove (SIG) and the E67-interacting loop (EIL) ofthe three SUMO paralogs are represented in the scheme. Themutations generated in SUMO1/2/3 are also indicated. (B)Evaluation of the yeast two-hydrid constructs. Total proteinextracts from transformants of the L40 strain containing oneof the pACT2 constructs (SUMO1WT, SUMO1K37A/K39A,SUMO1E67A, SUMO2WT, SUMO2D71H, SUMO3WT, SUMO3D70H andUbc9) were tested for SUMO1/2/3 or Ubc9 production byWestern blot using anti-HA antibody. L40 and L40 containingthe pACT2 vector were used as controls. (C) Yeast two-hybridmapping of SUMO interacting sites. Transformants of the L40strain containing the pBTM116-KAN-MX1WT construct and oneof the pACT2 constructs (SUMO1WT, SUMO1K37A/K39A,SUMO1E67A, SUMO2WT, SUMO2D71H, SUMO3WT or SUMO3D70H)were cultivated on SC -LT and SC -LTH containing 0.5 mM of3AT, and tested for β-galactosidase activity. L40/pBTM116-TIF51AþpACT-Dys1 (þþ), L40/pBTM116-TIF51AþpACT-Lia1(þ),L40/pBTM116-TIF51AþpACT (�) were used as controls for two-hybrid interactions [63].

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3158

reduced in MxAWT expressing cells. The level of VSV proteinsdetected in cells expressing MxAK48R was much lower than incontrol cells (Fig. 6A). In MxAWT- and MxAK48R-expressing cells, VSV

N protein was not detected whereas VSV M was expressed atsimilar low level. VSV G was expressed at a higher level in MxAK48R-expressing cells compared to cells expressing MxAWT, but this levelwas much lower than that in control cells.

We also determined VSV titers in the supernatants of infectedcells. VSV production was reduced in cells expressing MxAWT orMxAK48R compared to control cells transfected with the emptyvector (Fig. 6B). More precisely, VSV production was slightlyhigher in cells expressing MxAK48R than in MxAWT-expressingcells, but much lower than in control cells (Fig. 6B). These resultsperfectly corroborate the Western blot data.

Next, we tested the effect of MxAK48R on Influenza A Virusreplication. For this, NIH 3T3 cells transfected with plasmidsencoding for MxAWT or MxAK48R were infected with influenzavirus at a MOI of 1 for 8 h. The viral titers measured in thesupernatants revealed that Influenza A viral production wasinhibited in MxAK48R- and in MxAWT-expressing cells, thusdemonstrating that MxAK48R mutant retained its antiviral activityagainst Influenza A Virus (Fig. 6C).

Taken together, these results show that the SUMOylationdeficient MxA mutant MxAK48R is still able to confer viralresistance.

Discussion

In this study, we reported the interaction of MxA with compo-nents of the SUMOylation machinery, including SUMO1 and Ubc9,and SUMO modified proteins. Out of these 27 MxA ligands, ninewere already described as mouse Mx1 ligands by the yeast two-hybrid system [38]. Mouse Mx1 was shown to interact with 14proteins including SUMO1 and the E1-Activating enzyme theUBA1/AOS2 heterodimer, but none of these interactions werevalidated by co-immunoprecipitation. Moreover, mouse Mx1partially associates with SUMO1 in PML NBs [54,55]. Accordingly,some of the identified MxA ligands were already described asPML NB-associated proteins (Daxx, FLASH, PIAS1, TDG, ZCCHC12,Ubc9, SUMO1 and PLZF). Surprisingly, MxA is a cytoplasmicprotein and interacts with many PML NB-associated partners.Remarkably, both MxA and PML, the organizer of NBs, areimplicated in antiviral defense and confer resistance to VSV andInfluenza A Virus [5,56–58].

The physical interaction of MxA with SUMO1 and Ubc9 wasconfirmed by co-immunoprecipitation and SUMO1 negativelyinterfered in the interaction of MxA with Ubc9. We describedthat the binding site of MxA to SUMO1 involves the E67-Interaction loop on SUMO1. As SUMO1 also interacts with Ubc9through the EIL loop [50,59], this could explain the impact ofSUMO1 on the interaction between MxA and Ubc9. The findingthat the interaction of MxA with SUMO1 involves the EIL loop, butnot the SIG region, may also explain the lack of MxA interactionwith SUMO2 and SUMO3 in yeast two-hybrid.

However, we demonstrated using Ni2þ-NTA purification thatoverexpressed MxA or IFN-induced MxA is conjugated to SUMO2or SUMO3. Also, we identified the SUMOylation site for MxAprotein as the lysine at the position K48, since its mutationabolished MxA SUMOylation.

Human MxA and MxB and mouse Mx1 proteins have highsequence identity in the region comprehending the lysine 48 ofMxA, but modification by SUMO of the corresponding residues,

Fig. 4 – Investigation of MxA SUMOylation and determination of its SUMO acceptor site. (A) MxA is modified by SUMO. Proteinextracts from HeLa cells co-transfected with plasmid encoding MxA and 6xHis-SUMO1, 6xHis-SUMO2 or 6xHis-SUMO3 weresubmitted to affinity purification with Ni2þ-NTA resin under denaturant conditions. Plasmids used for co-transfection areindicated above the picture (þ) or (�). SUMOylated MxA (MxASUMO) or SUMOylated PMLIII (PMLIIISUMO3) are indicated. Thepresence of MxA, PMLIII and SUMO1/2/3 in the cellular lysates (Input) was confirmed by Western blot (lower panels). (B) Evaluationof the expression of MxA mutants. The expression of MxA mutants was tested by Western blot of protein extracts from HeLa cellstransiently expressing MxAWT, MxAK48R or MxAK259R. Actin was used as loading control. (C) MxAK48R is not modified by SUMO3.Protein extracts from HeLa cells transiently expressing 6xHis-SUMO3 and MxAWT or MxAK48R were also submitted to theSUMOylation assay. SUMOylated MxA (MxASUMO3) is indicated. (D) Endogenous MxA is conjugated to SUMO. HeLa cells transfectedwith the empty vector or with His-SUMO3 were treated with 1000 U/mL of IFNα for 16 h. Cell extracts were purified on Ni2þ-NTA-agarose beads. The purified extracts were analyzed by Western blot using anti-MxA and anti-His antibodies. The input is shownfor MxA.

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 159

K96 in MxB and K14 in Mx1, was not identified yet. These residuesare predicted to be SUMO acceptor sites. It would be interesting totest whether MxB and Mx1 interact with the EIL loop of SUMO1and undergo SUMOylation.

In addition, we have identified two putative SIMs in MxA, SIMaand SIMb, located in the GTP binding domain. MxASIMa andMxASIMb mutation did not alter the localization and the oligomer-ization capacity of MxA. Also, we found by yeast two-hybrid thatboth SIMa and SIMb MxA mutants were still able to interact withSUMO1, suggesting that they are not involved in the interaction ofMxA with SUMO1. These results corroborate the interaction

between MxA and SUMO1 via the EIL loop as the consensussequences of SUMO interaction motifs, SIMa and SIMb, are onlyrequired for the interaction with the SIG region of SUMO [27,29].Surprisingly, the interaction between MxA and SUMO1 occursthrough the CID–GED domains while the GTPase domain estab-lishes the interaction with Ubc9. However, we cannot rule out thepossibility that the putative MxA SIMs are not involved in theinteraction of MxA with SUMO2-3.MxA oligomerization plays an important role in the GTPase activity

[17]. In this study, we determined that the oligomerization of MxA isalso important for the interaction between MxA and SUMO1 or Ubc9.

Fig. 5 – Evaluation of the oligomerization capacity and the antiviral activity of MxA mutants. (A) Investigation of theoligomerization capacity of MxA mutants. Protein extracts of HeLa cells transiently expressing MxAWT, MxAK48R, MxASIMa, MxASIMb

or MxAL612K were analyzed by Western blot under native conditions using the anti-MxA antibody (α-pMxA). In this analysis,cellular extract was normalized by the MxA protein levels and resolved in a 4–12% Bis-Tris precast gel (Life technologies). (B)Distribution pattern of MxA mutants. HeLa cells transfected with plasmids encoding MxAWT, MxAK48R, MxASIMa, MxASIMb orMxAL612K were analyzed by immunofluorescence using the anti-MxA antibody (α-pMxA). Nucleus was stained with DAPI. (C) Anti-VSV activity of MxA mutants. HeLa cells transiently expressing MxAWT, MxAL612K, MxAK48R, MxASIMa or MxASIMb were infected withVSV at a MOI of 1 for 8 h and double stained using mouse anti-MxA (α-mMxA) and rabbit anti-VSV antibodies. MxA and VSVproteins were detected using secondary antibodies labeled with Alexa Fluor 594 (red) or Alexa Fluor 488 (green), respectively.Nucleus was stained with DAPI.

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3160

Fig. 6 – SUMOylation deficient MxA mutant confers resistance to VSV and Influenza A Virus. (A) NIH3T3 cells were transfected withempty vector, MxAWT or MxAK48R. Two days later, cells were not infected (�), infected with VSV, or influenza virus at MOI of 1 for8 h. Cell extracts were analyzed by Western blotting using anti-VSV, anti-MxA and anti-Actin antibodies. (B, C) Supernatants fromVSV-(B) or Influenza A Virus-(C) infected cells were removed and the virus titers were determined on NIH3T3 cells by the 50% tissueculture infective methods (TCID50).

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 161

In addition to MxA and Ubc9, three other proteins weredescribed to interact with the EIL loop of SUMO1, DPP8, DPP9and Dyn1 [50,51,60]. For these proteins, the interaction with thisloop seems to have a negative effect on their functions. In the caseof Dyn1, the interaction with the EIL loop was also described toabrogate its oligomerization and, consequently, its lipid bindingcapacity [60]. Interestingly, Dyn-1 and MxA, two constituents ofthe large dynamin GTPase family, use similar domains to interactwith SUMO1, GED for dynamin and CID–GED for MxA. Thequestion whether the interaction between the EIL loop of SUMO1and the CID–GED domains of MxA could abrogate the oligomer-ization capacity of MxA is open to investigation.

The CID–GED domains of MxA and Dyn-1 were described to beinvolved in the formation of the stalk structure [14]. Thisstructure is basically formed by alpha helixes and unstructuredloops and is the main coordinator of the oligomer formation. TheL4 loop of the MxA stalk is the main region of viral particlesrecognition [15]. Therefore, the interaction of the EIL loop ofSUMO1 with the stalk could lead to the inhibition of MxAoligomerization and its GTPase activity, which would probablyinterfere with its antiviral activity.

Herein, we have also demonstrated that MxA is SUMOylatedand that the lysine 48 (K48) is the acceptor site for SUMO. The

lysine K48 on MxA sequence is present in one of the Bundle-Signaling Elements (BSE). It was shown that these elements areseparated in the primary sequence but fold together to form anintermediary domain, between the GTPase and CID–GED (stalk)domains, in the MxA tertiary structure. This BSE domain has twohinge regions involved in the induction of inter- and intra-molecular conformational changes, essential for the correctassembly of the oligomer and antiviral activity [17]. In our study,the SUMOylation-deficient mutant MxAK48R retained its oligo-merization capacity and did not alter MxA localization.MxA is known to confer resistance to viruses whose replication

takes place in the cytoplasm or in the nucleus such as VSV and theInfluenza A Virus, respectively. MxA was shown to inhibit VSVmRNA synthesis, probably by affecting the elongation of the viralRNA chain [61]. The interaction between MxA and VSV proteinshas not been demonstrated yet. We have shown here that cellsexpressing the MxA SUMOylation-deficient mutant MxAK48R stillinhibit VSV protein synthesis and viral production, suggestingthat MxA covalent SUMOylation is not required for conferringresistance to VSV. Also, MxAK48R retains its anti-Influenza A Virusproperty. At the opposite, immunofluorescence studies revealedthat cells expressing the MxASIMa and the MxASIMb mutants doexpress VSV antigens. Since we showed that neither SIMa nor

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3162

SIMb is involved in MxA interaction with SUMO1, binding of MxAto SUMO2/3 may be important for its anti-VSV activity. Alterna-tively, the lack of anti-VSV activity of MxASIMa and MxASIMb

mutants could be independent of SUMO and due to theirlocalization within the GTPase domain of MxA, near to the GTP-binding sites. Further analysis is needed to demonstrate whetherthese mutants have lost the GTP-binding capacity and/or GTPaseactivity. It has been shown that the mouse Mx1K49A mutantharboring a mutation in the first GTP-binding motif preventsGTP binding, GTPase activity and, consequently, the antiviralactivity [20,62].Finally, our results show that MxA interacts with and is

conjugated to SUMO and that the MxA SUMOylation-deficientmutant MxAK48R is still able to confer resistance to VSV andInfluenza A Virus. However, further studies are needed to deter-mine the implication of MxA SUMOylation in the resistance toother viruses.

Author contributions

Conceived and designed the experiments: CEBDC, SRV, CFZ, SN,MKCA. Performed the experiments: CEBDC, GM and PEGB. Ana-lyzed the data: CEBDC, SRV and MKCA. Contributed reagents/materials/analysis tools: SRV and MKCA. Wrote the paper: CEBDC,SRV, SN and MKCA.

Competing financial interests

The authors declare no competing financial interests.

Acknowledgments

We are grateful to members of the Valentini and Chelbi-Alix labsfor helpful discussion. This work was supported by Grants to SRVfrom FAPESP (2003/09497-3 and 2010/50044-6) and PADC-FCF(2011/13-I) and to MKCA from the Agence Nationale de laRecherche (ANR 11BSV3002803). GM is funded by the ANR.CEBDC was a recipient of PhD fellowships from FAPESP andCAPES-PDSE.

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.yexcr.2014.10.020.

r e f e r e n c e s

[1] M.K. Chelbi-Alix, J. Wietzerbin, Interferon, a growing cytokinefamily: 50 years of interferon research, Biochimie 89 (2007)713–718.

[2] A.J. Sadler, B.R.G. Williams, Interferon-inducible antiviral effec-tors, Nat. Rev. Immunol. 8 (2008) 559–568.

[3] H. Arnheiter, S. Skuntz, M. Noteborn, S. Chang, E. Meier,Transgenic mice with intracellular immunity to influenza virus,Cell 62 (1990) 51–61.

[4] P. Staeheli, O. Haller, W. Boll, J. Lindenmann, C. Weissmann, Mxprotein: constitutive expression in 3T3 cells transformed withcloned Mx cDNA confers selective resistance to influenza virus,Cell 44 (1986) 147–158.

[5] J. Verhelst, P. Hulpiau, X. Saelens, Mx proteins: antiviral gate-keepers that restrain the uninvited, Microbiol. Mol. Biol. Rev. 77(2013) 551–566.

[6] M. Aebi, J. Fah, N. Hurt, C.E. Samuel, D. Thomis, L. Bazzigher, et al.,cDNA structures and regulation of two interferon-inducedhuman Mx proteins, Mol. Cell. Biol. 9 (1989) 5062–5072.

[7] O. Haller, H. Arnheiter, J. Lindenmann, I. Gresser, Host geneinfluences sensitivity to interferon action selectively for influenzavirus, Nature 283 (1980) 660–662.

[8] K.C. Chang, E. Hansen, L. Foroni, J. Lida, G. Goldspink, Molecularand functional analysis of the virus- and interferon-induciblehuman MxA promoter, Arch. Virol. 117 (1991) 1–15.

[9] K. Nakade, H. Handa, K. Nagata, Promoter structure of the MxAgene that confers resistance to influenza virus, FEBS Lett. 418(1997) 315–318.

[10] D. Holzinger, C. Jorns, S. Stertz, S. Boisson-Dupuis, R. Thimme,M. Weidmann, et al., Induction of MxA gene expression byInfluenza A Virus requires type I or type III interferon signaling,J. Virol. 81 (2007) 7776–7785.

[11] T. Zürcher, J. Pavlovic, P. Staeheli, Mouse Mx2 protein inhibitsvesicular stomatitis virus but not influenza virus, Virology 187(1992) 796–800.

[12] K. Melén, P. Keskinen, T. Ronni, T. Sareneva, K. Lounatmaa,I. Julkunen, Human MxB protein, an interferon-alpha-inducibleGTPase, contains a nuclear targeting signal and is localized in theheterochromatin region beneath the nuclear envelope, J. Biol.Chem. 271 (1996) 23478–23486.

[13] O. Haller, G. Kochs, Interferon-induced Mx proteins: dynamin-like GTPases with antiviral activity, Traffic 3 (2002) 710–717.

[14] S. Gao, A. von der Malsburg, S. Paeschke, J. Behlke, O. Haller,G. Kochs, et al., Structural basis of oligomerization in the stalkregion of dynamin-like MxA, Nature 465 (2010) 502–506.

[15] O. Haller, S. Gao, A. von der Malsburg, O. Daumke, G. Kochs,Dynamin-like MxA GTPase: structural insights into oligomeriza-tion and implications for antiviral activity, J. Biol. Chem. 285(2010) 28419–28424.

[16] O. Daumke, S. Gao, A. von der Malsburg, O. Haller, G. Kochs,Structure of the MxA stalk elucidates the assembly of ring-likeunits of an antiviral module, Small GTPases 1 (2010) 62–64.

[17] S. Gao, A. von der Malsburg, A. Dick, K. Faelber, G.F. Schröder,O. Haller, et al., Structure of myxovirus resistance protein areveals intra- and intermolecular domain interactions requiredfor the antiviral function, Immunity 35 (2011) 514–525.

[18] G. Kochs, M. Trost, C. Janzen, O. Haller, MxA GTPase: oligomer-ization and GTP-dependent interaction with viral RNP targetstructures, Methods 15 (1998) 255–263.

[19] G. Kochs, M. Haener, U. Aebi, O. Haller, Self-assembly of humanMxA GTPase into highly ordered dynamin-like oligomers, J. Biol.Chem. 277 (2002) 14172–14176.

[20] F. Pitossi, A. Blank, A. Schröder, A. Schwarz, P. Hussi, M. Schwemmle,et al., A functional GTP-binding motif is necessary for antiviralactivity of Mx proteins, J. Virol. 67 (1993) 6726–6732.

[21] C. Janzen, G. Kochs, O. Haller, A monomeric GTPase-negative MxAmutant with antiviral activity, J. Virol. 74 (2000) 8202–8206.

[22] O. Haller, P. Staeheli, G. Kochs, Interferon-induced Mx proteins inantiviral host defense, Biochimie 89 (2007) 812–818.

[23] P.S. Mitchell, M. Emerman, H.S. Malik, An evolutionary perspec-tive on the broad antiviral specificity of MxA, Curr. Opin.Microbiol. 16 (2013) 493–499.

[24] R.D. Everett, C. Boutell, B.G. Hale, Interplay between viruses and hostsumoylation pathways, Nat. Rev. Microbiol. 11 (2013) 400–411.

[25] D. Mattoscio, C.V. Segré, S. Chiocca, Viral manipulation of cellularprotein conjugation pathways: the SUMO lesson, World J. Virol. 2(2013) 79–90.

E X P E R I M E N T A L C E L L R E S E A R C H 3 3 0 ( 2 0 1 5 ) 1 5 1 – 1 6 3 163

[26] V.G. Wilson, SUMOylation at the host–pathogen interface, Bio-molecules 2 (2012) 203–227.

[27] E. Meulmeester, F. Melchior, Cell biology: SUMO, Nature 452(2008) 709–711.

[28] D. Guo, M. Li, Y. Zhang, P. Yang, S. Eckenrode, D. Hopkins, et al., Afunctional variant of SUMO4, a new IκBα modifier, is associatedwith type 1 diabetes, Nat. Genet. 36 (2004) 837–841.

[29] W. Wei, P. Yang, J. Pang, S. Zhang, Y. Wang, M.-H. Wang, et al., Astress-dependent SUMO4 SUMOylation of its substrate proteins,Biochem. Biophys. Res. Commun. 375 (2008) 454–459.

[30] R. Geiss-Friedlander, F. Melchior, Concepts in SUMOylation: adecade on, Nat. Rev. Mol. Cell Biol. 8 (2007) 947–956.

[31] J. Zhao, SUMOylation regulates diverse biological processes, Cell.Mol. Life Sci. 64 (2007) 3017–3033.

[32] R.J. Dohmen, SUMO protein modification, Biochim. Biophys. Acta1695 (2004) 113–131.

[33] R.T. Hay, Decoding the SUMO signal, Biochem. Soc. Trans. 41(2013) 463–473.

[34] E. Van Damme, K. Laukens, T.H. Dang, X. Van Ostade, A manuallycurated network of the PML nuclear body interactome reveals animportant role for PML-NBs in SUMOylation dynamics, Int. J. Biol.Sci. 6 (2010) 51–67.

[35] S.M. Hollenberg, R. Sternglanz, P.F. Cheng, H. Weintraub, Identi-fication of a new family of tissue-specific basic helix–loop–helixproteins with a two-hybrid system, Mol. Cell. Biol. 15 (1995)3813–3822.

[36] L. Van Aelst, Two-hybrid analysis of Ras–Raf interactions, Meth-ods Mol. Biol. 84 (1998) 201–222.

[37] M.F. Calmon, R.V. Rodrigues, C.M. Kaneto, R.P. Moura, S.D. Silva,L.D. Mota, et al., Epigenetic silencing of CRABP2 and MX1 in headand neck tumors, Neoplasia 11 (2009) 1329–1339.

[38] O.G. Engelhardt, E. Ullrich, G. Kochs, O. Haller, Interferon-inducedantiviral Mx1 GTPase is associated with components of theSUMO-1 system and promyelocytic leukemia protein nuclearbodies, Exp. Cell Res. 271 (2001) 286–295.

[39] A.H. Alm-Kristiansen, I.L. Norman, V. Matre, O.S. Gabrielsen,SUMO modification regulates the transcriptional activity ofFLASH, Biochem. Biophys. Res. Commun. 387 (2009) 494–499.

[40] M.-S. Jang, S.-W. Ryu, E. Kim, Modification of Daxx by smallubiquitin-related modifier-1, Biochem. Biophys. Res. Commun.295 (2002) 495–500.

[41] N. Kotaja, U. Karvonen, O.A. Janne, J.J. Palvimo, PIAS proteinsmodulate transcription factors by functioning as SUMO-1 ligases,Mol. Cell. Biol. 22 (2002) 5222–5234.

[42] D. Wotton, J.C. Merrill, Pc2 and SUMOylation, Biochem. Soc.Trans. 35 (2007) 1401–1404.

[43] C. Smet-Nocca, J.-M. Wieruszeski, H. Léger, S. Eilebrecht,A. Benecke, SUMO-1 regulates the conformational dynamics ofthymine-DNA Glycosylase regulatory domain and competes withits DNA binding activity, BMC Biochem. 12 (2011) 4.

[44] P. Knipscheer, A. Flotho, H. Klug, J.V. Olsen, W.J. van Dijk, A. Fish,et al., Ubc9 sumoylation regulates SUMO target discrimination,Mol. Cell 31 (2008) 371–382.

[45] T.-T. Chao, C.-C. Chang, H.-M. Shih, SUMO modification modu-lates the transrepression activity of PLZF, Biochem. Biophys. Res.Commun. 358 (2007) 475–482.

[46] T. Kahyo, T. Nishida, H. Yasuda, Involvement of PIAS1 in thesumoylation of tumor suppressor p53, Mol. Cell 8 (2001) 713–718.

[47] M.H. Kagey, T.A. Melhuish, D. Wotton, The polycomb protein Pc2is a SUMO E3, Cell 113 (2003) 127–137.

[48] S.E. Schwarz, K. Matuschewski, D. Liakopoulos, M. Scheffner,S. Jentsch, The ubiquitin-like proteins SMT3 and SUMO-1 areconjugated by the UBC9 E2 enzyme, Proc. Natl. Acad. Sci. USA 95(1998) 560–564.

[49] Y. Xue, F. Zhou, C. Fu, Y. Xu, X. Yao, SUMOsp: a web serverfor sumoylation site prediction, Nucleic Acids Res. 34 (2006)W254–W257.

[50] E. Pilla, U. Moller, G. Sauer, F. Mattiroli, F. Melchior, R. Geiss-Friedlander, A novel SUMO1-specific interacting motif in dipep-tidyl peptidase 9 (DPP9) that is important for enzymatic reg-ulation, J. Biol. Chem. 287 (2012) 44320–44329.

[51] E. Pilla, M. Kilisch, C. Lenz, H. Urlaub, R. Geiss-Friedlander, TheSUMO1-E67 interacting loop peptide is an allosteric inhibitor ofthe dipeptidyl peptidases 8 and 9, J. Biol. Chem. 288 (2013)32787–32796.

[52] F. Galisson, L. Mahrouche, M. Courcelles, E. Bonneil, S. Meloche,M.K. Chelbi-Alix, et al., A novel proteomics approach to identifySUMOylated proteins and their modification sites in human cells(M110.004796), Mol. Cell. Proteomics . (2011)http://dx.doi.org/10.1074/mcp.M110.004796.

[53] H. Saitoh, J. Hinchey, Functional heterogeneity of small ubiquitin-related protein modifiers SUMO-1 versus SUMO-2/3, J. Biol.Chem. 275 (2000) 6252–6258.

[54] M.K. Chelbi-Alix, L. Pelicano, F. Quignon, M.H. Koken, L. Ventur-ini, M. Stadler, et al., Induction of the PML protein by interferonsin normal and APL cells, Leukemia 9 (1995) 2027–2033.

[55] O.G. Engelhardt, Mx1 GTPase accumulates in distinct nucleardomains and inhibits influenza A virus in cells that lackpromyelocytic leukaemia protein nuclear bodies, J. Gen. Virol. 85(2004) 2315–2326.

[56] R.D. Everett, M.K. Chelbi-Alix, PML and PML nuclear bodies:implications in antiviral defence, Biochimie 89 (2007) 819–830.

[57] F. El Asmi, M.A. Maroui, J. Dutrieux, D. Blondel, S. Nisole, M.K.Chelbi-Alix, Implication of PMLIV in both intrinsic and innateimmunity, PLoS Pathog. 10 (2014) e1003975, http://dx.doi.org/10.1371/journal.ppat.1003975.

[58] M.K. Chelbi-Alix, F. Quignon, L. Pelicano, M.H. Koken, H. de Thé,Resistance to virus infection conferred by the interferon-inducedpromyelocytic leukemia protein, J. Virol. 72 (1998) 1043–1051.

[59] P. Knipscheer, W.J. van Dijk, J.V. Olsen, M. Mann, T.K. Sixma,Noncovalent interaction between Ubc9 and SUMO promotesSUMO chain formation, EMBO J. 26 (2007) 2797–2807.

[60] R.K. Mishra, S.S. Jatiani, A. Kumar, V.R. Simhadri, R.V. Hosur,R. Mittal, Dynamin interacts with members of the SUMOylationmachinery, J. Biol. Chem. 279 (2004) 31445–31454.

[61] P. Staeheli, J. Pavlovic, Inhibition of Vesicular Stomatitis VirusmRNA synthesis by human MxA protein, J. Virol. 65 (1991) 4498–4501.

[62] K. Melén, T. Ronni, T. Lotta, I. Julkunen, Enzymatic characteriza-tion of interferon-induced antiviral GTPases murine Mx1 andhuman MxA proteins, J. Biol. Chem. 269 (1994) 2009–2015.

[63] G.M. Thompson, V.S.P. Cano, S.R. Valentini, Mapping eIF5Abinding sites for Dys1 and Lia1: in vivo evidence for regulation ofeIF5A hypusination, FEBS Lett. 555 (2003) 464–468.