SUMO polymeric chains are involved in nuclear foci formation and chromatin organization in Trypanosoma brucei procyclic forms

SUMOylation is a post-translational modification conserved in eukaryotic organisms that involves the covalent attachment of the small ubiquitin-like protein SUMO to internal lysine residues in target proteins. This tag usually alters the interaction surface of the modified protein and can be translated into changes in its biological activity, stability or subcellular localization, among other possible outputs. SUMO can be attached as a single moiety or as SUMO polymers in case there are internal acceptor sites in SUMO itself. These chains have been shown to be important for proteasomal degradation as well as for the formation of subnuclear structures such as the synaptonemal complex in Saccharomyces cerevisiae or promyelocytic leukemia nuclear bodies in mammals. In this work, we have examined SUMO chain formation in the protozoan parasite Trypanosoma brucei. Using a recently developed bacterial strain engineered to produce SUMOylated proteins we confirmed the ability of TbSUMO to form polymers and determined the type of linkage using site-directed mutational analysis. By generating transgenic procyclic parasites unable to form chains we demonstrated that although not essential for normal growth, SUMO polymerization determines the localization of the modified proteins in the nucleus. In addition, FISH analysis of telomeres showed a differential positioning depending on the polySUMOylation abilities of the cells. Thus, our observations suggest that TbSUMO chains might play a role in establishing interaction platforms contributing to chromatin organization.


Introduction
SUMOylation is a eukaryotic post-translational modification that involves the covalent attachment of a Small Ubiquitin-like Modifier (SUMO) to a diverse range of protein substrates. SUMO conjugation is analogous to ubiquitination, and requires the sequential action of activating (E1), conjugating (E2) and eventually ligating (E3) enzymes [1,2]. This enzymatic cascade catalyzes the formation of an isopeptide bond between the carboxyl group of the C-terminal Gly residue in SUMO and the ε-amino group of a lysine residue in the substrate protein.
The target lysine residue is often embedded within a canonical consensus motif, CKxE (where C is a hydrophobic residue and x any amino acid) [3] or in related variants encompassing an adjacent cluster of negative charges [4][5][6] recognized by the E2 enzyme [7]. In addition, SUMOylation can take place at non-consensus sites [8], but in these cases the process is usually assisted by E3 ligases. The main consequence of protein SUMOylation is an alteration of the interaction surface where SUMO can recruit effector proteins containing short stretches of hydrophobic residues known as SUMO-interacting motifs (SIMs) [9,10]. This modification can then be translated into changes in the subcellular localization, the subset of interacting partners, and the stability, among other outputs [1,2]. Protein SUMOylation can be reverted by the action of specific peptidases [11] contributing to a dynamic modification of the biochemical properties of the target proteins.
Proteomic studies have highlighted the role of SUMO in many central cellular processes, such as DNA replication and repair, chromatin organization, transcriptional control, RNA processing, transport, macromolecular assembly and signal transduction [12,13]. Accordingly, SUMO has been shown to be essential for the viability of many eukaryotic cells [14][15][16][17].
Trypanosomes, unicellular organisms which belong to a diverse group of eukaryotes of the excavate lineage [18], possess a single SUMO gene [16,19,20]. These protozoan parasites have a major impact on human and animal health. Trypanosoma cruzi is the causative agent of Chagas disease in South America while Trypanosoma brucei is the ethiological agent of African sleeping sickness in humans and nagana in cattle. T. brucei is a model organism amenable for genetic manipulation and knock-down experiments demonstrated that TbSUMO is essential for mitosis in both procyclic (PCF) and bloodstream (BSF) forms of the parasite, which are the replicative forms present in the tsetse fly vector and the infected mammal, respectively [16,17]. Moreover, site-specific proteomic studies of PCF identified a number of SUMOylated proteins involved in fundamental nuclear biological processes [21]. Interestingly, SUMO has been linked to T. brucei antigenic variation, a process where the major surface antigenic protein is replaced by a different variant with certain frequency as a strategy to elude the specific immune response of the host. TbSUMO was found to be enriched in a particular region of the nucleus of BSF parasites together with the E3 ligase Siz1 and the RNA polymerase I, specifically at the chromatin region that is actively transcribing the variant surface glycoprotein (VSG), suggesting that a highly SUMOylated focus provides an environment permissive for VSG transcription [22].
Although in most cases SUMO is conjugated as a monomer, this modifier also shares with ubiquitin the ability to form chains. Mammalian SUMO-2/-3 isoforms [23] as well as yeast SUMO (Smt3) [24,25] bear internal SUMOylation sites at their N-terminal regions that allow polymerization, and from the phenotype of SUMO chain mutants the contribution of the polymeric signal has started to be appreciated. In particular, structural roles for poly-SUMO chains in the formation of the synaptonemal complex [26] or promyelocytic leukemia nuclear bodies [27] have been described.
In this work, we have investigated SUMO chain formation in T. brucei. Using an in bacteria SUMOylation system, we confirmed the ability of TbSUMO to polymerize and identified the site of branching using site-directed mutagenesis. By generating transgenic PCF parasites unable to form chains we demonstrated that although not essential for normal growth, these polymers are involved in the assembly of nuclear foci, likely establishing interaction platforms in the nucleus, which might contribute to chromatin organization.

Lysine 27 is required for TbSUMO chain formation
To facilitate the biochemical validation of T. brucei SUMO targets, we have previously developed an in bacteria SUMOylation system consisting of the heterologous co-expression of the SUMOylation machinery of the parasite (TbE1a/E1b, TbE2, TbSUMO), together with a potential substrate, in E. coli [28]. When testing this system in the absence of any substrate, we detected a ladder of wild-type SUMO (TbSUMO) that was no longer visible when all eight lysine residues present in the protein were mutated to arginine (TbSUMOallKR) (Fig 1A). This observation led us to speculate that TbSUMO, similar to its human counterparts SUMO-2/-3 or yeast Smt3, is capable of forming poly-SUMO chains.
Inspection of the amino acid sequence within the unstructured N-terminal region ( Fig 1B) revealed that the lysine residue in position 27 (K27) could be the branching site, being predicted to be SUMOylated by different bioinformatic tools (GPS-SUMO [29], JASSA [30], SUMOAMVR [31], PCI-SUMO [32]). To establish the importance of K27 in the formation of poly-SUMO chains, this residue was changed to arginine (K27R), which preserves the charge on the molecule but cannot be engaged in the formation of an isopeptide bond. After in bacteria SUMOylation, reaction products were analysed by Western blot (Fig 1C). Whereas wild type TbSUMO formed multimers, these higher molecular weight forms were absent when the assays were performed with TbSUMOallKR or TbSUMO K27R. Furthermore, the pattern of TbSUMOallKR was reverted to that of wild type SUMO when K27 was restored (TbSU-MOallKR R27K). Thus, we conclude that poly-SUMO chains are mainly formed through K27 linkages.

TbSUMO chains bind to SUMO-interacting motifs placed in tandem
Having shown that TbSUMO is able to form chains, we next examined if they display the typical ability of binding to SUMO-interacting motifs (SIMs) placed in tandem. To test this, we used a probe originally developed for SUMO-2 [33] consisting of a fragment of the Ring Finger Protein 4 (RNF4) comprising four SIMs (SIMx4) and a mutated version in which Val, Leu and Ile residues within SIMs were changed to Ala (mutSIMx4) to disrupt SUMO-SIM interaction [10] (Fig 2A and 2B). Protein extracts expressing HA-TbSUMO monomer (monomer) or polySUMO(chains) were incubated separately with SIMx4, mutSIMx4 or Ni +2 -Sepharose beads without adding SIMx4 or mutSIMx4 probes (negative control), and their capacity to recognize both probes was tested ( Fig 2C and S2 Fig). After washing and elution, Western blot analysis with anti-HA antibodies was performed for three replicates. The results showed that SIMx4 probe was able to pull-down TbSUMO chains with higher ability (more than 3-fold increase) than the mutSIMx4 probe or Ni +2 -Sepharose beads alone (Fig 2C and S2B Fig). However, under identical experimental conditions the SUMO monomer bound with relatively similar capacity to any of the bait tested (SIMx4, mutSIMx4 or beads) since no statistical differences were observed in the amount of protein detected in the eluates (S2C and S2D Fig). Taking together, the previous results indicate that poly-SUMO chains produced in bacterial systems are capable of binding to SIMs placed in tandem.
After this biochemical validation we used the probe with cell-free extracts from T. brucei 427 PCF, and incubated equal parts in parallel with either purified SIMx4, mutSIMx4 or beads alone ( Fig 2D and S2E Fig). To preserve the modification, these experiments were performed with N-ethylmaleimide (NEM), a well-known inhibitor of SUMO deconjugating proteases and Western blot analysis with anti-TbSUMO antibodies was performed. Although quantification of proteins bands in the eluates for three replicates revealed no statistical differences coli BL21 DE3 cells were used to co-express the complete SUMOylation system using either a wild type TbSUMO variant (TbSUMO) or a mutant version with all lysine residues replaced by arginines (TbSUMOallKR). Cleared cell lysates were subjected to Ni +2 affinity chromatography and input (In) and eluates (El) were analyzed by 10% SDS-PAGE followed by Coomassie Blue staining. The positions of TbSUMO monomer (SUMO), dimers (SUMO) 2 and trimers (SUMO) 3 are indicated. Note that the E2 enzyme (upper band of the doublet in the lane that corresponds to the eluate from TbSUMOallKR strain) copurifies with SUMO (S1 Fig). (B) Comparison of the three dimensional structure of TbSUMO (2K8H, red) and human SUMO-2 (2N1W, green). Sequence alignment of Nterminal extensions reveals a potential conserved SUMOylation site in TbSUMO K27 marked with an asterisk. (C) In bacteria SUMOylation assays were performed with different HA-tagged TbSUMO variants. Cleared lysates were subjected to Ni +2 affinity chromatography and analysed by Western blot using anti-HA antibodies. Note that in addition to SUMO polymers, the bands marked with asterisks could also correspond to SUMOylated forms of other components of T. brucei SUMOylation machinery, such as TbE2 or TbE1a (S1 Fig). between poly-SUMO conjugates bound to SIMx4 in comparison to mutSIMx4 (S2F Fig), some high molecular weight bands were mildly enriched in the experiment pulled-down with SIMx4 (see arrows in Fig 2D). E. coli BL21 DE3 cells expressing SIMx4 or mutSIMx4 probes were lysed and cleared lysates were subjected to Ni +2 chromatography. Inputs, In; Eluates, El. (C) E. coli BL21 DE3 cells expressing HA-TbSUMO chains (Inputs) were incubated with the SIMx4 or the mutSIMx4 probes. Samples were pulled-down using Ni +2 -Sepharose beads. Proteins were visualized by Western blot using anti-HA antibodies. Coomassie stained gels of the inputs are shown. (D) Cell free extract from T. brucei 427 PCF was incubated with SIMx4 or mutSIMx4 probes, which were subsequently purified by Ni +2 chromatography. The capture of SUMOylated proteins was determined by Western blot analysis using anti-TbSUMO antibodies, while anti-α-tubulin was used as the loading control for inputs. https://doi.org/10.1371/journal.pone.0193528.g002

Generation of SUMO chain mutant parasites
To study the biological roles of SUMO chains in T. brucei, we generated mutant PCF parasites in which the endogenous SUMO gene was replaced by the open reading frame of TbSUMO where all eight lysine codons were mutated to code for arginine. As described above, the resulting protein can be conjugated to one or more sites of a target (mono or multi-SUMOylation), but lacks the ability to form chains (polySUMOylation) due to the absence of internal modification sites (Fig 3A). The SUMO variant expressed by these parasites is an unprocessed TbSUMO version with an N-terminal His-HA tag (HisHA-TbSUMOallKR). For comparison, a His-HA tagged TbSUMO keeping all internal lysine residues was also generated (HisHA-TbSUMO).
After two rounds of transfection and sequential replacement of both alleles by homologous recombination (T. brucei is diploid), parasites were cloned by limiting dilution and mutations were confirmed by DNA sequencing (S3 Fig). Unlike TbSUMO-deficient cells, which arrest in G2/M [16], HisHA-TbSUMOallKR strains are viable; indicating that the proteins modified with SUMOallKR can fulfill the essential functions for the parasite, at least under regular culture conditions ( Fig 3B).
Comparison of the SUMOylation pattern between His-HA-tagged TbSUMO variants able or unable to form chains by Western blot using anti-HA antibodies evidenced an increase in free (unconjugated) SUMO protein, and a decrease in the intensity of high molecular weight SUMO conjugates ( Fig 3C) likely reflecting the absence of polySUMOylated proteins in the TbSUMOallKR mutant.
To further investigate potential differences we evaluated the subcellular localization of TbSUMO in wild type, HisHA-TbSUMO and HisHA-TbSUMOallKR parasites. In wild type PCF as well as in HisHA-TbSUMO parasites, TbSUMO accumulates in discrete foci dispersed in the nucleus ( Fig 3D). This characteristic localization agrees with the fact that most SUMO targets participate in nuclear processes, as it has been reported for several organisms [8,13,34]. Site-specific proteomic analysis had identified a high proportion of TbSUMO targets involved in DNA replication and repair, chromatin remodeling and transcription, among others [21]. Intriguingly, the characteristic nuclear foci disappeared in chain mutant parasites, which now exhibit a more homogenous distribution in the nucleus as well as in the cytoplasm. These results show that nuclear foci formation relies on polySUMOylation of certain proteins and at the same time confirm the occurrence of SUMO chains in PCFs.

Bioinformatical detection of SIMs
We next executed a four-step bioinformatic pipeline to predict potential SIMs that could be recruited by poly-SUMO chains. To this end, 11567 trypanosome translated coding sequences were first run against the ScanProsite database (http://prosite.expasy.org/) to search for the three classes of previously published patterns of SIMs, resulting in 1302 protein hits. This output was next filtered to exclude sequences where SIMs reside within reported PFAM domains (214 proteins) or in predicted globular regions (703 proteins). The 385 sequences that survived these criteria were finally evaluated to identify SIMs located in disordered regions yielding a final list of 102 candidate proteins (S1 Table). Regarding the number of SIMs in each protein, 79 sequences have a single motif, 15 sequences have two SIMs, 6 sequences have 3 SIMs, Tb927.9.7690 -one hypothetical protein-contains 4 SIMs and Breast Cancer Type 2 susceptibility protein (BRCA2, Tb927.1.640) harbors 14 SIMs. Near 75% of the SIM candidates are predicted to have nuclear localization according to NucPred server [35] with a NucPred-score >0.5, including a subgroup of 40 proteins displaying the classical eukaryotic nuclear localization signal KRxR that is functional in kinetoplastid parasites. Furthermore, 18 hits were indeed identified as part of the nuclear proteome [36]. Globally, the identified set of SIM hits is enriched in nuclear proteins (chi-square test, p < 0.0002).
Among the 102 SIM-containing proteins, 43 are hypothetical proteins with unknown functions while 59 proteins have a predicted or experimentally validated function. This latter group was manually classified into functional categories (Fig 4) involving processes often associated with SUMOylation like mRNA metabolism, DNA structure, replication and repair. Gene Ontology enrichment showed that these SIM-containing proteins are related to microtubule cytoskeleton, nuclear lamina, cillium, axonema, nuclear periphery, SUMO activating enzyme complex, among others (S1 Table). Altogether, these results suggest the relevance of the SUMO-associated proteins in nuclear organization in T. brucei.

PolySUMOylation and chromatin organization
Since the group of T. brucei sequences with predicted SIMs is enriched in nuclear proteins, we sought to test if polySUMOylation could be determinant for the degree of chromatin compactation using the micrococcal nuclease (MNase) digestion assay. As shown in Fig 5A, samples from TbSUMO wild-type or chain mutants showed similar production of a ladder of mono-, di-and tri-nucleosomes, suggesting that SUMO chain formation does not impact on the global chromatin state.
Next, we examined the position of telomeres by fluorescence in situ hybridization (FISH) using a probe specific for telomeric repeats. In PCF parasites competent for SUMO chain formation, we observed the typical distribution of telomeres [37]: at the periphery of interphase nuclei (Fig 5B, 1N1K

Discussion
Previous studies have shown that, like ubiquitin, SUMO is able to form polymeric chains. These chains are involved in several processes, such as mitosis, meiosis and proteosomal degradation of proteins in different organisms [38,39]. In this work, using a bacterial system designed to express T. brucei SUMOylation machinery and by site-specific mutagenesis, we were able to determine K27 of TbSUMO as the main residue involved in chain formation.
Mutant PCF parasites unable to form SUMO chains were viable under in vitro culture conditions, suggesting that the essential functions of SUMO in T. brucei can be accomplished in TbSUMOallKR strain (TbSUMOallKR) compared to HisHA-TbSUMO (TbSUMO) and wild-type parasites. Wild-type and transgenic parasites were cultured up to one month without observing significant differences in growth rate. (C) Conjugating capacity of HisHA-TbSUMO (TbSUMO) and HisHA-TbSUMOallKR (TbSUMOallKR) parasites. Whole-cell extracts were boiled in Laemmli sample buffer immediately after harvesting, separated in a 7.5%-12.5% discontinuous acrylamide gel (3,5x10 7 cells/lane), and analysed by Western blot using anti-HA antibodies. An unspecific cross-reacting band of *50 kDa from anti-TcSUMO antibodies [22] was used as loading control. (D) Immunofluorescence analysis of wild-type, HisHA-TbSUMO (TbSUMO) and HisHA-TbSUMOallKR (TbSUMOallKR) parasites. Nuclear and kinetoplast DNA were visualized by DAPI staining (blue). Representative images of anti-TbSUMO (green), anti-HA (green) and TbSUMO/anti-HA-DAPI merged images are shown.
https://doi.org/10.1371/journal.pone.0193528.g003 absence of polySUMOylation. Something similar has been reported for S. cerevisiae where it has been observed that SUMO chains are not essential for vegetative growth but are required for mitosis [40], meiosis [25,41,42], DNA replication [43] and repair [44]. Notably, the allKR chain mutant parasites showed a less intense pattern of SUMOylation and loss of the TbSUMO nuclear foci normally observed in wild type parasites by indirect immunofluorescence. Taken together these results demonstrate for the first time that SUMO chains are indeed present in T. brucei PCF, presumably modifying different nuclear proteins as we were not able to detect these structures as free (unanchored) TbSUMO polymers (S5 Fig). As it has been described in other organisms, SUMO chains could be involved in establishing interaction platforms. For example, these chains are required for the assembly of the synaptonemal complex, a chromosomal structure that links the homologous chromosomes during meiosis. In addition, SUMO chains also participate in the formation of polycomplexes in S. cerevisiae [25] and particularly in the assembly and function of the complex that catalyzes the resection of double-strand breaks by homologous recombination during vegetative growth and meiosis [44,45]. In this work, we have identified SIMs in several DNA repair proteins including TbBRCA2, a protein implicated in DNA repair by homologous recombination, with 14 mapped SIMs. Therefore, TbSUMO chains might have a role in DNA repair in T. brucei as it has been described in yeast.
The characteristic nuclear TbSUMO pattern of PCF cells and the drastic localization change observed in TbSUMOallKR parasites (Fig 3D) suggest that polySUMOylation might be regulating the activity of nuclear proteins involved in the formation of subnuclear structures. Similarly, human SUMO-5 polymeric chains have been associated with the enlargement of promyelocytic leukemia nuclear bodies after its ectopic expression [46]. Furthermore, it has been reported that SUMO chains are required for maintenance of chromatin organization and transcriptional repression in S. cerevisiae [43].
In T. brucei BSF, a highly SUMOylated focus can be detected at the telomeric VSG expression site exerting a positive regulation [22]. In this work, we have shown that SUMOylation could be important for telomere positioning in PCF, more likely in the case of the abundant minichromosomes. We speculate that abnormal telomere positioning observed in HisHA-TbSUMOallKR parasites might be due to the disassembly of protein complexes formed through SUMO-SIM interactions; however, the identification of which factors are involved need to be addressed in future studies.
Considering that SUMOylation is an ubiquitous post-translational modification regulating a wide range of cellular processes, the effects observed in HisHA-TbSUMOallKR parasites might be consider as the joint action of a high number of nuclear SUMOylated targets. The processes identified in this work as regulated by polySUMOylation in T. brucei may constitute the first step to assess the role of these polymeric structures in individual protein targets.
The His tags of the TbE1a and TbE2 enzymes were removed by PCR using the following primers: TbE1a forward CCATGGCCAATGCGGACGAAAAAACG, TbE1a reverse AAGCTTCT ACGGGTTGCGCAGGTGCC, TbE2 forward CCATGGCCTCCGGGCTATCTTTAGC and TbE2 reverse GCGGCCGCTTATACCCGCTTCCGGTG. PCR products were cloned into pGEM-T Easy vector (Promega) and sequenced (Macrogen). The NcoI/HindIII (New England Biolabs) fragment of TbE1a and the NcoI/NotI (New England Biolabs) fragment of TbE2 were cloned into the vector pACyCDuet-1 and pCDFDuet-1, respectively.

Purification of TbSUMO chains and SIM probes
E. coli BL21 DE3 competent cells were transformed with plasmids containing the sequence of TbE1, TbE2 and TbSUMO, or with the SIMs probes. Cells were cultured in LB medium at 37˚C to an OD 600nm of 0.6, and then induced with 1 mM IPTG (Sigma) for 5 h at 37˚C with vigorous shaking. Cells were harvested by centrifugation and resuspended in lysis buffer and sonicated. Cleared lysates were obtained by centrifugation at 23000 x g for 30 min and subjected to Ni +2 chromatography. Briefly, samples were incubated with Ni +2 -resin (GE Healthcare, Pittsburgh, PA, USA) for 1 h at 4˚C with shaking, washed with 30 column volumes (CV) of 50 mM Tris HCl (pH 7.6), 150 mM NaCl, 0.1% Triton X-100, 1mM PMSF and 20 mM imidazole, and bound proteins were eluted with 5 CV of the same buffer but containing 500 mM imidazole. Samples were then resuspended in Laemmli sample buffer with 100 mM DTT (7:3) and boiled for 5 min. Finally, proteins were separated by SDS-PAGE and visualized by Coomassie Blue staining.

TbSUMO chains pull down in vitro
TbSUMO chains or monomer-expressing bacteria were lysed as described previously and then incubated with the SIMx4 probe or the mutSIMx4 probe in 50 mM Tris HCl (pH 7.6), 150 mM NaCl, 1% NP40, 1mM PMSF and 0.5% deoxycholate for 1.5 h at 4˚C. Samples were then incubated with Ni +2 -resin (GE Healthcare) containing 20 mM imidazole for 1 h at 4˚C with shaking. Washes were performed with 20 CV of 50 mM Tris HCl (pH 7.6), 250 mM NaCl, 1% NP40 and 1mM PMSF, and with 10 CV of the same buffer but containing 20 mM imidazole. Proteins were eluted with the same buffer but containing 100 mM imidazole (2 CV), 200 mM imidazole (2 CV) or 500 mM imidazole (2 CV), resuspended in Laemmli sample buffer with 100 mM DTT (7:3), boiled for 5 min and separated by SDS-PAGE followed by Western blot analysis using anti-HA antibodies. Purifications from bacteria expressing TbSUMO chains or monomer without the incubation with the probes were performed as control.

TbSUMO chains pull down in vivo
About 5 x 10 8 T. brucei brucei 427 PCF parasites [48] and 25 ml of induced cultures of SIMs probes were collected by centrifugation, resuspended in 1.25 ml of lysis buffer (500 mM NaCl, 50 mM Tris-HCl, 1 mM PMSF, 1% NP40, 20 mM N-ethylmaleimide (NEM) (Sigma)-pH 7.5) and sonicated up to loss of viscosity. For further purification lysates were cleared by centrifugation for 15 min at 23000 x g at 4˚C and supernatants were incubated with each other for 1.5 h at 4˚C with gentle stirring, adding 0.5% deoxycholate. Samples were then subjected to Ni +2 chromatography as described above. Purifications from parasites without incubation with SIMs probes were used as control.

Trypanosome culture
In this work we used the PCF of T. brucei brucei Lister 427 [48]. As describe previously, parasites were cultured at 28˚C in SDM-79 medium [49] (Life Technologies, Carlsbad, CA, USA) supplemented with 10% (vol/vol) heat-inactivated fetal calf serum (Natocor, Córdoba, Argentina) and 7.5 mg/l hemin. Parasites viability was evaluated by growth curves obtained by counting cell number daily [21].

Generation of TbSUMO-transfectant cell lines
For HisHA-TbSUMO strain, a HisHA-TbSUMO construct was synthesized (GenScript), containing the coding sequence for an 8His (24bp)-HA (27bp) tag, 249bp of the 5'end of TbSUMO open reading frame (ORF; Tb927.5.3210), a XhoI restriction site and 250bp of the 3'end of 5'untranslated region (UTR) of TbSUMO. For HisHA-TbSUMOallKR strain, the former construction was modified by replacing the 5'end of TbSUMO ORF by the complete TbSUMO ORF with all 8 lysine residues replaced by arginine residues and 200bp of the 5'end of 3'UTR of this gen (GenScript). These constructions were flanked by HindIII and BamHI restriction sites to be cloned in pEnT6P, an endogenous locus tagging vector containing a puromycin or hygromycin resistance marker cassette, for sequential replacement of both TbSUMO alleles [50]. The vectors were linearized and electroporated as described previously [21]. Transfected cells were cloned by limiting dilution and selected with 2 μg/ml of puromycin or 25 μg/ml of hygromycin (InvivoGen, San Diego, CA, USA) in 96-wells plates as described in http://tryps. rockefeller.edu/. To confirm the appropriated replacement of endogenous TbSUMO allele we performed a PCR using a sense primer corresponding to the upstream gene Tb927.5.3220 of TbSUMO in the genome (Up foward: GTGACTCGTTTGTACCTCAC) and an antisense primer corresponding to pEnT6P puromycin (Puro reverse: CGTGGGCTTGTACTCGGTC) or hygromycin (Hygro reverse: GCCTATTCCTTTGCCCTCGG). The absence of endogenous TbSUMO after the replacement of both alleles was confirmed by PCR using a sense primer corresponding to the 5'UTR of TbSUMO (5'UTR foward: GAGTTGGGTTCATTTCTGAGCC) and an antisense primer corresponding to the 3'UTR of TbSUMO (3'UTR reverse: AGTTG TGATGGACGAAGCAG).
To assess the ability of TbSUMO to form free chains in vivo we generated a TbSUMO variant unable to modified targets, by removing the di-glycine motif (ΔGG), but capable of being modified by other TbSUMO molecules, due to the presence of acceptor lysines. As negative control we generated a TbSUMO variant which was deficient in the di-glycine motif and the acceptor lysines. To generate TbSUMOΔGG-3xFlag and TbSUMOallKRΔGG-3xFlag variants we used plasmid constructions with the complete ORF of TbSUMO or TbSUMOallKR as template for PCR using the following primers: TbSUMO forward CATATGGACGAACACCACC, TbSUMO reverse CTCGAGTCACCCGCCACGCTGCTCCACC, TbSUMOΔGG forward GCGGG ATCCATGGACGAACCCACTCATAACTCCAACAACG, TbSUMOΔGG reverse GGCTCTAGATG TCTGCTCCACCATCGCATCAATC. Amplification products were first cloned into pGEM-T Easy vector (Promega) and sequenced (Macrogen), after which they were digested with BamHI (New England Biolabs) and XbaI (New England Biolabs) and cloned into the vector pRP TAG , modified to add a 3xFlag tag at the N-terminal or C-terminal end of the construction. The vectors were linearized and electroporated into PCF T. brucei brucei Lister 427-pLew13 as described above. Transfected parasites were induced with 5 μg/ml of doxycycline (Sigma) for 48 h.

Indirect immunofluorescence
PCF parasites were subjected to inmunofluorescence as previously described [21]. The anti-HA antibody (Roche) and the anti-TbSUMO antibody were used at a 1:500 dilution, while secondary antibodies were used at a 1:1000 dilution [polyclonal goat anti-rat Alexa Fluor 1 546 or polyclonal goat anti-mouse Alexa Fluor 1 488 (Jackson)]. Finally, images were obtained with a fluorescence microscope Nikon 80i LED and analyzed by ImageJ software.

In situ hybridization
This protocol was carried out as described by the Cross laboratory (http://tryps.rockefeller. edu/), with some modifications. Briefly, 1×10 6 parasites were washed with PBS and fixed in 4% paraformaldehyde for 10 min at room temperature. After two washes with PBS the cells were attached onto poly-lysine coated glass coverslips for 30 min and incubated with 25 mM NH 4 Cl for 15 min. Parasites were permeabilized with 0.2% saponin in PBS for 30 min at room temperature. The DNA was denatured by incubating permeabilized cells in 70% formamide/ 2×SSC for 5 min at 70˚C. The coverslips were washed briefly in 2×SSC and blocked for 3 h in hybridization mix (SSC 4x, Denhardt 5X, 8 μg/ml yeast tRNA, 8 μg/ml herring sperm DNA, 5% dextran sulfate, 60% formamide). Parasites were then probed with a Cy3-labeled (CCCTAA) 9 probe (Macrogen) at 50˚C, overnight. Coverslips were then washed with SSC 4x-40% formamide for 10 min, SSC 4x for 10 min, SSC 2x for 10 min and two times with SSC 1x for 5 min. Finally, coverslips were mounted using FluorSave reagent (Merck) containing 5 mg/ ml DAPI (Life Technologies). Images were obtained with a fluorescence microscope Nikon 80i LED and analyzed by ImageJ software.

Micrococcal nuclease digestion
Micrococcal assay was adapted from [51] and [52]. Parasites (5x10 7 ) were washed in a buffer containing 1 mM L-glutamate, 250 mM sucrose, 2.5 mM CaCl 2 and 1 mM PMSF, and then lysed with the same buffer but also containing 1% Triton X-100. Pellets were washed [10 mM HEPES, 35 mM NaCl, 500 μM MgCl 2 , 500 μM CaCl 2 , 1 mM PMSF, 5.2 mM β-mercaptoethanol (Sigma)] and incubated with increasing amounts of microccocal nuclease (New England Biolabs) for 5 min at 37˚C (0, 3, 7 U). Reactions were stopped by adding a stop solution (0.09 M EDTA, 0.72% SDS). Extractions were performed with chloroform and DNA was precipitated with isopropanol. Finally, the DNA was washed with 70% ethanol, resuspended in a buffer containg 50 mM Tris HCl pH 8 and 10 mM EDTA and incubated with 10 μg/ml of RNAse for 2 h. Samples were electrophoresed on horizontal 2% agarose gels and stained with ethidium bromide. Bands were quantified with Image Studio software and statistical analysis was performed using Student's t test.

Identification of good SIM candidates in Trypanosoma brucei genome
We performed an in silico search with four different evaluation criteria to priorizate good candidate proteins containing multiple SIMs that could be recruited by poly-SUMO chains. Briefly, we generated a FASTA-formatted version of the translated T. brucei coding sequences downloaded from TriTrypDB v9.0 [URL www.tritrypdb.org [53]] and used the ScanProsite server [URL http://www.expasy.org/tools/scanprosite [54]] to screen the entire proteome of species for the presence of the three classes of previously published SIMs: SIM-a (PILVM)-(ILVM)-x-(ILVM)-(DSE>)(3); SIM-b (PILVM)-(ILVM)-D-L-T; and SIM-r (DSE)(3)-(ILVM)-x-(ILVMF) (2). The hits reported in the previous step were further validated using PFAM server to search for homology domains [URL http://pfam.xfam.org [55]], and all SIM candidates located within a known domain were filtered. Next, we run IUPRED program [URL http://iupred.enzim.hu/ [56]] with the "structured regions" option to evaluate if SIM candidates reside in a region predicted as globular (much less likely to be a functional SUMO binder). Output hits that pass these filtering criteria were finally evaluated using the same software for disorder prediction (with the option "long disorder" and cut-off value above 0.5), since if the SIM candidate resides in a region predicted to be disordered increases its chances to be functional. GO enrichment analysis was performed using Tritrypdb Analyze Results tool (http://tritrypdb.org) with Fisher exact test filtering for false discovery rate (FDR) lower than 0.1.
Supporting information S1 Fig. SUMOylation of TbE1a and TbE2. E. coli BL21 DE3 cells expressing TbE1a-TbE1b, His-TbE2 and TbSUMO (complete system His-TbE2) or TbE2, His-TbE1a-TbE1b and TbSUMO (complete system His-TbE1a) were lysed and cleared lysates were subjected to Ni +2 chromatography. Proteins were separated by SDS-PAGE and visualized by Western blot using anti-His antibodies. Thus, it was possible to observe His-TbE2 (~30 kDa) and higher molecular weight bands corresponding to SUMOylated TbE2. Also, we visualized His-TbE1a (~40 kDa) and high molecular weight bands corresponding to SUMOylated TbE1a.