Human promyelocytic leukemia protein is targeted to distinct subnuclear domains in plant nuclei and colocalizes with nucleolar constituents in a SUMO‐dependent manner

Eukaryotic nuclei are subdivided into subnuclear structures. Among the most prominent of these structures are the nucleolus and the PML nuclear bodies (PML‐NBs). PML‐NBs are spherical multiprotein aggregates of varying size localized in the interchromosomal area. PML‐NB formation is dependent on the presence of the promyelocytic leukemia protein (PML) as well as on post‐translational modification of core components by covalent attachment of the small ubiquitin‐like modifier (SUMO). So far, PML‐NBs as well as PML have been described in mammalian cells only, whereas no orthologs are known in the plant kingdom. In order to investigate conserved mechanisms in PML targeting, we expressed human PML (hPML) fused to the GFP in Nicotiana benthamiana. Using confocal laser scanning microscopy and coimmunoprecipitation followed by mass spectrometric analysis, we found the fusion protein in association with nucleolar constituents. Importantly, mutants of hPML, which are no longer SUMOylated, showed altered localizations, implying SUMO‐dependent targeting of hPML in plants as has previously been shown for mammalian cells. Interestingly, in the presence of proteasome inhibitors, hPML could also be found in the nucleolus of mammalian cells suggesting conserved targeting mechanisms of PML across kingdoms. Finally, Solanum tuberosum COP1, a proposed PML‐like protein from plants, was fused to the red fluorescent protein (RFP) and coexpressed with hPML::eGFP. Microscopic analysis confirmed the localization of COP1::RFP in nuclear speckles. However, hPML::eGFP did not colocalize with COP1::RFP. Hence, we conclude that plants do not possess specialized PML‐NBs, but that their functions may be covered by other subnuclear structures like the nucleolus. Database Proteomics data have been deposited to the ProteomeXchange Consortium with the identifier PXD004254.

Although animal and plant nuclei are superficially similar, they differ in their compositional details. While it has long been known that eukaryotic nuclei in general contain further subnuclear structures like nucleoli or cajal bodies which are well known in both kingdoms, other structures seem to be less conserved [1,2]. One prominent example for a subnuclear domain characterized in metazoans but not in plants are structures termed PML nuclear bodies (PML-NBs). These dot-like structures of varying size, which are also known as nuclear domain 10 (ND10) or PML oncogenic domains (POD), are localized to the interchromosomal area and are characterized by a subset of interacting core proteins [3]. Among these is the eponymous tumor-suppressor protein, promyelocytic leukemia protein (PML), as well as Sp100 and Daxx [1,4,5]. Importantly, the integrity of PML-NBs is dependent on the presence of PML as well as its post-translational modification by small ubiquitin-like modifier (SUMO) proteins. Also, the ability of PML and other PML-NB constituents to bind SUMO noncovalently is essential for the efficient formation of PML-NBs [5][6][7][8]. The suggested functions conferred by this subnuclear multiprotein structure are manifold: Due to the large number of in part transiently associated cellular proteins, PML nuclear bodies were considered as sites of nuclear storage or as a dump for excess proteins [9]. The fact that PML is reacting toward the application of heat stress with redistribution into smaller aggregates, led to the belief that PML-NBs are also involved in a general heat stress response [10]. However, in recent years these subnuclear structures were successively assigned to several other, more active, functions. Since PML knockout mice exhibit lower genomic stability and an increased liability to tumor formation during exposure to carcinogens, a role in DNA damage repair as well as tumor suppression was proposed [11,12]. Interestingly, PML-NBs were also implicated in antiviral defense and thus also pose a target for viral effectors [9,[13][14][15].
To date, no clear plant ortholog of PML has been found and it remains uncertain whether PML-NBs exist in plants. However, work in Arabidopsis thaliana identified the constitutive photomorphogenic protein 1 (COP1) as a potential functional homolog of PML [16]. As the name suggests, the plant protein is an important negative regulator in plant photomorphogenesis. When grown in darkness, wild-type seedlings undergo etiolation [17], while cop1-knockout mutants do not undergo etiolation, but show a constitutive photomorphogenic developmental pattern [18,19]. The postulated functional analogy of COP1 and PML is based on similar localization patterns and domain architecture, since both A. thaliana COP1 (AtCOP1) and human PML (hPML) localize to the nucleus in a speckled manner and both contain an N-terminal RING-finger with an adjacent coiled-coil domain [16]. However, the primary sequences of COP1 and hPML do not show significant similarity, and COP1 is missing B-Box domains present in hPML, while hPML does not possess WD40 repeats found at the C terminus of COP1 [20]. Furthermore, Wang et al. [21] could identify a COP1 homolog in mammalian cells which could be shown to interact with Jun transcription factors. Interestingly, the mammalian COP1 homolog exhibited a light-dependent shift in localization, similar to the Arabidopsis protein [22,23]. Based on the described inconsistencies, a functional commonality of hPML and AtCOP1 seemed unlikely and therefore it has been suggested that PML is not evolutionarily conserved [20].
To address this unresolved question, we investigated the targeting and SUMOylation status of hPML in plant cells. In Nicotiana benthamiana, the protein could be shown to localize to interchromosomal areas and to colocalize to some extent with the nucleolus/cajal bodymarker fibrillarin 2 (FIB2) when heterologously expressed [24]. Of note, a nucleolar localization could also be found in mammalian cells upon stress treatment. Moreover, coimmunoprecipitation of hPML::eGFP allowed us to identify 29 associated plant nuclear or nucleolar proteins. Strikingly, some of these proteins are also known to reside in the hPML-interactome in human cells. Furthermore, hPML mutants exhibiting impaired SUMOylation [8] showed mistargeting of the protein, suggesting that SUMOylation is important for the nucleolar localization of hPML in plant cells. Finally, colocalization experiments with hPML and plant COP1 revealed both proteins to reside in distinct subnuclear domains, supporting the previous evaluation of COP1 to be not functionally redundant to hPML and suggesting that PML-NBs may not be conserved across kingdoms.

Results
Human PML is targeted to interchromosomal subnuclear domains and interacts with nuclear and nucleolar proteins in planta In order to address the still unsettled question whether PML-NBs are present in plant nuclei, we fused the coding region of hPML isoform VI with eGFP to generate the chimeric construct hPML::eGFP [25]. Using correspondingly transformed Agrobacterium tumefaciens, the protein was expressed in N. benthamiana leaves together with the nuclear envelope marker AtSUN1::RFP [26].
By means of confocal laser scanning microscopy, we found the fusion protein to form spherical structures of varying size in the nucleus (Fig. 1A-F). However, the observed structures did not resemble PML-NBs as present after expression of FLAG-tagged hPML isoform VI in mammalian cells (Compare Fig. 1J-L), since they were larger and appeared in fewer numbers. Staining of leaves expressing hPML::eGFP with 4 0 ,6-diamidino-2-phenylindole (DAPI) revealed the human protein to be targeted to interchromosomal areas ( Fig. 1G-I), and colocalization studies with FIB2::RFP [27] indicated partial nucleolar association ( Fig. 2A-C). A localization pattern similar to the typical dot-like pattern of PML-NBs in mammalian cells could not be observed. However, upon treatment of mammalian cells with the proteasome inhibitor MG132, hPML isoform VI was found to relocalize to the nucleolus, a finding which has previously been described for endogenous hPML in a variety    [28]. To further ascertain this result and to reveal hPML-associated plant proteins, we conducted a coimmunoprecipitation of hPML::eGFP followed by mass spectrometric analysis. In total, 69 proteins were identified being significantly (> 139 -10lg P) and more than five times enriched in samples containing hPML::eGFP (Table S1). Of these 69 proteins, 22 were annotated as chloroplastic or mitochondrial. Eight of the identified proteins were assigned to the cytoplasm, five were membrane proteins, four were annotated as extracellular or cell wall-associated, and one was annotated as peroxisomal. The remaining 29 proteins were annotated as nuclear proteins, of which 8 could be assigned to the nucleolus (Fig. 3). Foremost, the high number of nucleolar and nuclear proteins underlines the result obtained in the microscopical investigation and confirms hPML:: eGFP association with constituents of the plant nucleolus. Strikingly, several of the identified nuclear proteins (or proteins related to them) have previously been described to interact with hPML in mammalian cells (Table 1). This finding suggests that some interactions are conserved in mammalian and plant cells and point to general functions.

SUMOylation-negative hPML mutants show aberrant localization patterns
In mammalian cells, proper targeting of hPML is dependent on SUMOylation of the protein. To test whether targeting of hPML::eGFP in plant cells may also be SUMO-dependent, we generated eGFP fusion constructs of hPML mutants termed hPMLcs (containing two cysteine to serine amino acid exchanges in the protein's RING domain) and hPML3 m (in which the three known lysine residues modified by SUMO are changed to arginines). Both mutants have previously been shown to be SUMOylation-negative and not to be able to induce formation of proper PML-NBs [8]. In mammalian cells, FLAG-tagged hPMLcs show a dispersed localization throughout the nucleoplasm ( Fig. 4G-I). Interestingly, upon expression of hPMLcs::eGFP in N. benthamiana, we observed a localization pattern comparable to the situation observed in mammalian cells ( Fig. 4A-F). The hPMLcs::eGFP fusion protein was found in the nucleoplasm in a granular pattern, but not in the nucleolus. Expressed in mammalian cells, the SUMOylation mutant, hPML3 m, is still able to form subnuclear aggregates, but not functional PML-NBs [8]. In plant cells, hPML3 m::eGFP could also be localized in subnuclear aggregates similar to the wildtype protein. However, coexpression of hPML3 m:: eGFP and FIB2::RFP revealed that the protein does not colocalize with the nucleolar marker, FIB2 ( Fig. 2D-F). This could also be confirmed by statistical validation of the colocalization of hPML::eGFP, hPML3 m::eGFP, and FIB2::RFP. To this end, Zstack images of coexpressing cells were evaluated according to the classification by Zinchuk et al. [29].
Here, approximately 58% of all screened nuclei showed strong or moderate colocalization of hPML:: eGFP and FIB2::RFP, whereas hPML3 m::eGFP colocalized only in 5.9% of all observed nuclei to a moderate extent with the nucleolus marker, while the remainder exhibited only weak colocalization (Fig. 2G). Taken together, this indicates that the hPML RING domain represents an important factor for the subnuclear localization of the protein, not only in mammalian cells but also in N. benthamiana. Furthermore, the reduced ability of hPML3 m::eGFP to associate with FIB2::RFP suggests that SUMOylation of hPML is an important determinant for nucleolar targeting in plants and suggests an evolutionarily conserved function of SUMOylation in targeting of proteins to subnuclear domains.
hPML::eGFP is SUMOylated, while hPMLcs::eGFP and hPML3 m::eGFP exhibit impaired SUMOylation in planta To verify SUMOylation of hPML in planta, protein extracts of N. benthamiana leaves expressing either  hPML::eGFP, hPMLcs::eGFP, hPML3 m::eGFP, or free GFP were subjected to immunoprecipitation followed by western blot analysis. Target protein precipitation and correct expression of the chimeric gene constructs was determined by using a polyclonal anti-GFP antibody. As indicated in Fig. 5A, hPML::eGFP, hPMLcs::eGFP, and hPML3 m::eGFP proteins were efficiently precipitated and accumulated to comparable amounts at approximately 100 kDa. Interestingly, hPML::eGFP showed a faint additional band at ca. 130 kDa, which was absent in hPMLcs:: eGFP and hPML3 m::eGFP. Subsequently, potential SUMOylation of the precipitated proteins was tested with an individual blot using a polyclonal antibody directed against A. thaliana SUMO-1. For hPML:: eGFP, a distinct band could be detected at approximately 130 kDa (Fig. 5B), thus running at the same height as the faint band observed before. Hence, this higher molecular band most likely represents SUMOylated hPML::eGFP. In contrast to this, hPMLcs::eGFP as well as hPML3 m::eGFP did not yield bands migrating at this height. In good accordance with the study initially characterizing the mutant [8], this indicates an abolished SUMOylation state of both hPML mutants in N. benthamiana.

Human PML and plant COP1 are targeted to individual subnuclear structures
Since the suggested functional analogy of plant COP1 to hPML is still under debate, [16,20], we wanted to investigate whether both proteins would colocalize in the same subnuclear compartment in plant cells. Therefore, we coexpressed hPML::eGFP and Solanum tuberosum COP1::RFP in N. benthamiana leaves. As described above, hPML::eGFP showed a nucleolar localization, while COP1::RFP was present in independent subnuclear domains (Fig. 6). Although both signals did not overlap, both domains were intriguingly adjoining. Therefore, a relation of both proteins to some extent cannot be excluded. Their spatial proximity might even hint at a dynamic exchange of contents between both subnuclear structures, which, considering the dynamic nature of nuclear bodies, might not be surprising [2]. However, the largely missing colocalization of hPML:: eGFP and COP1::RFP is further evidence for diverging roles of the respective structures in the nucleus.

Discussion
In mammalian cells, the subnuclear structures known as PML-NBs are intensively studied and implicated in various cellular processes. In the plant kingdom, however, these structures are largely uncharacterized. In this study, we utilized hPML for heterologous protein expression experiments in order to test for the possible presence of PML-NBs in N. benthamiana cells. Upon expression of an hPML::eGFP construct, we could show targeting of the human protein to subnuclear interchromosomal regions associated with the nucleolus, but not to punctuate structures comparable to PML-NBs in the mammalian system [3]. Of note, association of hPML with the nucleolus could be shown to different degree in human cell lines: In senescent human mesenchymal stem cells and human skin fibroblasts, but not in immortalized cell lines, hPML may be found in patterns adjacent to nucleoli or nucleolar elements differing from regular PML-NBs [30]. Similarly, various forms of cellular stress like UV-or c-radiation as well as chemical inhibition of topoisomerase II, RNA polymerase I and the proteasome lead to a redistribution of hPML in favor of an association with nucleolar structures or constituents. Interestingly, a nucleolar targeting domain was identified at the C-terminal part of hPML isoform I, which was suggested to be auto-inhibited by the adjacent SUMOinteracting motif [28,31]. Although the hPML isoform VI used in this study does not contain this C-terminal part responsible for hPML isoform I nucleolar targeting, it still relocalized to the nucleolus upon treatment of mammalian cells with proteasome inhibitors [25]. This may indicate that additional protein interactions contribute to the nucleolar targeting of hPML. Nevertheless, our observations underline the dynamic nature of subnuclear structures and connect hPML functions with the nucleolus, indicating that the observed localization of hPML::eGFP in N. benthamiana is reasonable.
Coimmunoprecipitation of proteins associated with hPML::eGFP revealed a large number of nuclear and nucleolar proteins supporting the microscopical analysis. Among those, several proteins known from the hPML-interaction network in human cells could be identified. This indicates conserved protein-protein interactions between mammalian and plant cells. In this regard, the list of nuclear and nucleolar proteins contained five proteins annotated as replication factor C subunits and one protein (annotated as Holliday junction ATP-dependent DNA helicase RuvB) that shows high sequence similarity to Arabidopsis replication factor C 3 (At5G27740). Two of the former five proteins showed the highest similarity (based on BLAST analyses) to the identical

protein
(Gene locus identifier: At1G63160). Interestingly, an interaction of human replication factor C4 with the hPML/RARa fusion protein relevant in acute promyelocytic leukemia could be shown before [32]. Other plant nuclear proteins found in the immunoprecipitation include a DNA-directed RNA polymerase subunit as well as two histone-lysine N-methyltransferases. Of note, the interaction with the human DNA-directed RNA polymerase POLR2E was demonstrated by twohybrid experiments, while the interaction of hPML with the histone-lysine N-methyltransferase G9a could be shown via immunoprecipitation in HEK293 cells [33,34]. Notably, the Arabidopsis histone-lysine N-methyltransferase SDG8 could be identified in a screening for plant SUMO-interacting proteins [35]. The coimmunoprecipitation also yielded a protein annotated as snRK1-interacting protein 1, which shows high similarity to Arabidopsis RADIATION-SENSITIVE 52-1 (RAD52), a protein involved in DNA double-strand break repair [36]. Strikingly, an involvement of hPML in DNA double-strand repair as well as an interaction of hPML isoforms V and I with the RAD52-interactor RAD51 was shown in human cell lines [37,38]. Another interesting finding is the 65-kDa microtubule associated protein 6, since interaction of hPML with a human microtubuleinteracting protein (microtubule-associated protein 1 light chain 3 beta) was suggested to be important for cell growth control [39]. Some of the proteins identified in the present experiment may also give hints toward the intracellular transport of hPML, like the importin subunit alpha or the nuclear pore complex protein NUP85, whereas some other candidates on the list remain enigmatic and need further investigation. Considering the absence of structures resembling human PML-NBs in plant cells expressing hPML::eGFP but keeping the high number of coimmunoprecipitated nucleolar proteins in mind, it is tempting to speculate about the nucleolus as a site of hPML-associated functions in plants. This assumption is further fortified by the loss of nucleolar association of hPML in N. benthamiana, when its SUMOylation is impaired by mutating the target lysine residues [8,25]. Moreover, the dependency of the protein's SUMOylation status for its subnuclear localization could be one starting point to investigate a SUMO-mediated nucleolar-targeting mechanism in plants, a mechanism which is already known from human topoisomerase I as well as Werners helicase. Here, SUMOylation leads to redistribution from the nucleolus to the nucleoplasm [40][41][42]. In addition, large amounts of SUMO-1 are found at the nuclear envelope and in the nucleolus of mammalian cells [43]. Furthermore, the SUMOdeconjugating sentrin/SUMO-specific protease SENP5 as well as its close homolog SENP3 are concentrated in the nucleolus, where they are involved in cell division and rRNA processing, respectively [44][45][46][47]. Also, a proteomic study revealed a set of nucleolar SUMO target proteins, reflecting the tight interconnection of this post-translational modification and the nucleolus [48]. In plants, the important role of SUMO is also becoming apparent. SUMOylation has been implicated in processes ranging from stress responses, pathogen defense, and abscisic acid signaling to flower induction [49][50][51][52][53][54]. Importantly, the involvement of the plant nucleolus in SUMO-related processes is underlined by the identification of several SUMO-interacting proteins located in the nucleolus [35]. In regard of the drastic shift of hPML localization both in plants as well as in mammalian cells when mutating two cysteine residues in the RING domain, it is important to note that the RING domain was shown to be a mediator of protein-protein interactions, among others with the E2 SUMO-1-conjugating enzyme, UBC9 [55]. Taking into account that hPMLcs::eGFP was not SUMOylated anymore in planta, it seems reasonable that the RING domain is mediating post-translational modification of hPML with SUMO. Whether the observed mistargeting of hPMLcs::eGFP is an effect of impeded protein-protein interaction or a consequence of effective blocking of SUMOylation remains unclear.
Taken together, our data indicate that human PML is targeted to subnuclear domains in the plant nucleus, where it associates with nucleolar constituents dependent on the integrity of its RING domain and its SUMOylation status. It is able to interact with several factors also present in the hPML-interactome in human cells. Since we were not able to identify distinct hPML-containing subnuclear structures resembling PML-NBs in plants, we suggest that some PML-NB functions may be carried out by the plant nucleolus. Further studies will be needed to elucidate the specific role of this subnuclear compartment in processes related to mammalian PML-NBs and to unravel the role of SUMOylation in subnuclear protein targeting and/or function.

Plant materials
Nicotiana benthamiana plants were grown on soil in a greenhouse with a 16 h light/8 h darkness photoperiod and temperatures of 25 and 20°C, respectively.
The S. tuberosum COP1::RFP construct as well as the A. thaliana FIB2::RFP construct [27] was kindly provided by Prof. Michael Taliansky (The James Hutton Institute, Scotland UK).
Transient expression in primary human fibroblast, N. benthamiana, and confocal laser scanning microscopy Primary human foreskin fibroblasts (HFFs) with a small interfering RNA-mediated knockdown of PML were cultured as described previously [58]. For transient expression of PMLVI in HFFs, 3 9 10 5 cells were seeded into six-well dishes and transfected using the DNA transfection reagent FuGENE6 (Promega, Mannheim, Germany). FLAG-tagged PML variants were detected by indirect immunofluorescence analysis using the monoclonal antibody anti-FLAG 1804 (Sigma-Aldrich, Deisenhofen, Germany) and Alexa Fluor 488-conjugated secondary antibodies. DAPI staining was used to visualize the cell's nucleus. For transient expression approaches in plant cells, correspondingly transformed Agrobacterium tumefaciens (Strain C58C1) were pressure infiltrated as described [59,60]. Staining with DAPI and imaging of leaf samples using a Leica TCS SP5 II confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany) were conducted as described earlier [61]. Statistical analysis of colocalization was carried out using Pearson's correlation coefficient as calculated by the LAS AF software (Leica Microsystems) for Z-stack micrographs covering the whole nucleus.
For proteomic analysis, the same amount of plant material as stated earlier was harvested from plants expressing PML::eGFP or free GFP, and four biological replicates of each sample were freshly homogenized in 1 mL of ice-cold extraction buffer containing 100 mM Tris HCl pH 7.5, 300 mM NaCl, 10 mM DTT, 0.5 mM EDTA, 0.5% (v/v) NP-40, cOmplete ULTRA Tablets (Roche diagnostics). After 20 min of incubation on ice, extracts were centrifuged for 10 min at 15 900 g and 4°C. Supernatant was transferred into a new microcentrifugation tube and diluted with 1 mL of ice-cold dilution buffer (100 mM Tris HCl pH 7.5, 300 mM NaCl, 10 mM DTT, 0.5 mM EDTA, cOmplete ULTRA Tablets (Roche diagnostics)) supplemented with 25 lL GFP-Trap Ò magnetic bead slurry (Chromotek). Samples were kept at 4°C under constant mixing for 2 h, before beads were washed three times with cold dilution buffer with additional 0.1% NP-40 and once with dilution buffer devoid of detergent. Immunoprecipitated proteins were eluted in 100 lL of 200 mM glycine pH 2.5. After extraction of beads, supernatant containing proteins was transferred into new microcentrifugation tubes and pH was neutralized using 20 lL Tris HCl pH 8.0.

Protein polyacrylamide gel electrophoresis and western blot analysis
After extraction of magnetic beads, samples derived from immunoprecipitation were subjected to centrifugation for 3 min at 15 900 g. Supernatant was loaded on a Bis/Tris Gel containing 12% acrylamide and subjected to electrophoresis. Proteins were then transferred onto a nitrocellulose membrane, which was afterward blocked for 1 h in blocking buffer (20 mM Tris HCl, 500 mM NaCl, 0.1% (v/v) Tween 20, 5% (w/v) milk powder). Detection was achieved by the use of suitable primary antibodies (Polyclonal anti-AtSUMO-1 antibody: Abcam, Cambridge, UK; polyclonal anti-GFP antibody: Roche diagnostics) and secondary antibodies conjugated to horseradish peroxidase (Thermo Fisher Scientific) followed by enhanced chemiluminescence detection.

Tryptic digest
Tryptic digest of previously extracted proteins was carried out based on the modified filter-aided sample preparation (FASP) method [63]. In short, eluted proteins were collected on 10 kDa Vivacon 500 Hydrosart Ò membrane filters (Sartorius Stedim Biotech, G€ ottingen, Germany) by centrifugation at 15 900 g for 8 min, flow-through was discarded. Reduction was performed by the addition of 250 lL of buffer containing 8 M urea, 100 mM triethylammonium bicarbonate buffer (TEAB), 20 mM DTT, and incubation for 30 min at room temperature under constant mixing. After centrifugation and discarding of flowthrough, alkylation of reduced proteins was achieved by addition of 250 lL of 8 M urea, 100 mM TEAB, 40 mM chloroacetamide, and incubation for 30 min at room temperature in the dark under constant mixing. After centrifugation, a washing step with 8 M urea and 100 mM TEAB was performed, before filters were transferred to new microcentrifugation tubes. For tryptic digest, 250 lL of 1 M urea, 50 mM TEAB, and 1 lg trypsin was added and samples were incubated overnight at 37°C. Resulting peptides were retrieved by centrifugation and desalted on C18 stage tips. Prior to LC-MS/MS analysis, peptides were concentrated on a centrifugal evaporator and resuspended in 0.1% trifluoroacetic acid (TFA).

Nano-LC-MS/MS analysis
Peptides obtained by tryptic digestion were analyzed by an Orbitrap Fusion TM Tribrid TM mass spectrometer in connection with an UltiMate 3000 nano-UHPLC system (Thermo Fisher Scientific). The samples were loaded and separated on PepMap100 columns (300 lm i.d. 9 5 mm, 5 lm, 100 A and 75 lm i.d. 9 50 cm, 3 lm, 100 A, respectively) with a 3-35% gradient of 0.1% formic acid in acetonitrile for 160 min at a flow rate of 300 nLÁmin À1 at 35°C. Ionization of peptides was achieved using a nanospray Flex TM ion source (Thermo Fisher Scientific), precursor ions were analyzed at a scan range of 300-2000 m/z. The most intense peptides of each scan cycle were selected and fragmented by collision-induced dissociation, resulting MS/MS scans were analyzed in the ion trap. Raw data files were evaluated using PEAKS7 (Bioinformatics Solutions Inc., Waterloo, ON, Canada) and a N. benthamiana proteome database (Sol Genomics Network, retrieved on 06.03.2015 [64]). The amino acid sequence of hPML::eGFP was additionally added to this database. Peptide identification was based on two trypsin miscleavages maximum and 10 ppm mass tolerance for survey scans. Carbamidomethylation was set as fixed modification, whereas oxidation of methionine residues was regarded optional. Resulting peptides were only considered as identified with a false discovery rate (FDR) below 1%. For label-free quantification, two groups (free GFP control and PML::eGFP immunoprecipitation) of which each contained the respective four biological replicates were used. The samples were normalized by the total ion count (TIC) and only proteins with two or more unique peptides were included in the calculation. Finally, only proteins that were at least fivefold more abundant in the hPML::eGFP-samples (with a significance of 13 -10lgP or higher) were included in the list used in this study. All mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.pro teomexchange.org) via the PRIDE partner repository [65] with the dataset identifier PXD004254.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Table S1. Complete list of proteins identified by hPML coimmunoprecipitation.