Structural insights into SUMO E1-E2 interactions in Arabidopsis uncovers a distinctive platform for securing SUMO conjugation specificity across evolution

SUMOylation of proteins involves the concerted action of the E1-activating enzyme, E2-conjugating enzyme and E3-ligases. An essential discrimination step in the SUMOylation pathway corresponds to the initial interaction between E1 ubiquitinfold domain (UFD) and E2 enzymes. Although E2 orthologs possess high sequence identity, the E2 binding region of the UFD domains have diverged across evolution. Moreover, in reciprocal in vitro conjugation reactions Arabidopsis E1 and E2 SCE1 fail to interact efficiently with cognate human E2 Ubc9 and E1 partners, respectively. To gain more insights into the properties of this interface in evolutionary distant organisms, we solved the crystal structure of SUMO E2 SCE1 and its complex with E1 UFD in Arabidopsis. In addition to a few common structural determinants, the interface between the E1 UFD and E2 in Arabidopsis is distinct compared with human and yeast, in particular by the presence of a longer α-helix in the Arabidopsis UFD domain. Despite the variability of E1 UFD domains in these surfaces, they establish specific interactions with highly conserved surfaces of their cognate E2 enzymes. Functional analysis of the different E2 interface residues between human and Arabidopsis revealed Val37 (Met36 in human), as a determinant that provides specificity in the E1-E2 recognition in plants.


Introduction
Small ubiquitin-related modifier (SUMO) protein is one of the most extensively studied ubiquitin-like proteins (Ubls) present in eukaryotes. Yeast and invertebrates have a single SUMO protein named Smt3, while vertebrates have several SUMO proteins. Mammalians such as human express four SUMO proteins. SUMO2 and SUMO3 share 97% sequence identity and cannot be distinguished by antibodies [1,2]. However, SUMO1 is quite different from SUMO2/3 and only shares ~47% sequence identity with SUMO1 and SUMO2 [2]. In plants, SUMO conjugation machinery displays high complexity according to isoform number and variability [3,4].
Although only few studies have addressed the biological relevance of this molecular complexity [5][6][7], the diverse biological function of SUMOylation in plant development and plant responses to environmental stresses could benefit from the existence of isoforms dedicated to SUMOylate proteins under specific physiological conditions. Accordingly, increasing evidences suggest that a deep knowledge of plant SUMOylation could provide novel strategies for increasing crop productivity [4,8,9].
SUMO is conjugated to target proteins through a three-step enzyme cascade as ubiquitin via SUMO E1 activating enzyme, E2 conjugating enzyme, and E3 ligase [1,2]. In the model proposed for the SUMO/Ubl conjugation pathway, the E2 conjugating enzyme interacts with the E1 activating enzyme by means of the UFD domain [15][16][17][18]. The structure of the UFD-E2 interaction is known for ubiquitin, Nedd8, and SUMO systems, showing a direct binding of E2 to the E1 UFD domain [17][18][19][20][21][22][23]. The UFD domain is located in the C-terminal end and connected to the E1 active adenylation domain through a flexible hinge. During the thioester transfer between E1 and E2, the UFD domain undergoes a rotation to bring the catalytic cysteine residues of E1 and E2 into proximity for thioester transfer [17,18,[24][25][26].
Moreover, crystal structure of ubiquitin E1-E2 complex reveals a direct interaction of E1 and E2 through their catalytic cysteine domains, which occurs after UFD-E2 binding and a significant rotation of the UFD domain, providing structure insights for the E1 E2 thioester transfer [18,24]. This interaction was also proposed in the SUMO pathway by NMR analyses. However, UFD-E2 interactions (Kd= 1.2μM) display much higher affinity than the catalytic cysteine domain interactions (Kd= 87μM), consistent with a major role of E1 UFD domain as the initial E2 binding platform [27].
Structural and functional data indicate that the UFD-E2 interactions represent the first contacts between the E1-activating enzyme and the E2-conjugating enzyme, providing an important discrimination step for the pathway [28,29]. UFD-E2 interactions have been revealed essential to provide specificity between the different systems of ubiquitin-like modifiers (i.e. SUMO, Nedd8, ubiquitin…). In the SUMO system, this interaction also provides specificity across species, despite the high homology between the E2-conjugating enzymes [30]. The UFD domains show little protein sequence homology across distant species, and it is even lower when considering just the E2 binding region or LHEB2 domain (Low Homology region involved in E2 Binding 2) [31]. However, the opposite is found in close related species, in which the LHEB2 domain display higher conservation than the UFD domain highlighting its crucial role [22,31].
Previous structural studies of the SUMO UFD-E2 complex in yeast and human revealed two alternative UFD interaction interfaces [20][21][22], in which particular contacts with a highly conserved E2 structure take place in each species. Despite the 65% sequence identity between human and Arabidopsis SUMO E2 conjugating enzymes, human Ubc9 cannot complement Arabidopsis SCE1 in our in vitro conjugating assays using RanGAP1 as a model substrate, and vice versa. As a consequence of the amino acid sequence divergence found between Arabidopsis, human and yeast UFD domains, the study of Arabidopsis E1-E2 interactions cannot rely only on predictive models, and a structural analysis is required to reveal the particular contacts between Arabidopsis E1 and E2 enzymes.
In order to elucidate the structural determinants for such exquisite evolutionary specificity in the interface of the UFD-E2 interaction, we have solved the crystal structure of the Arabidopsis E2-conjugating enzyme (AtSCE1), alone at a high resolution (1.2 Å) and in complex with the Arabidopsis UFD domain (AtUFD) of the E1-conjugating enzyme at 1.8 Å resolution. We have compared the Arabidopsis UFD-E2 domain complex interface with the human and yeast systems and performed biochemical and mutagenesis analysis to understand the role of the interface in the specificity of the SUMO conjugation pathway. Our results indicate that Arabidopsis UFD provides a novel molecular strategy for E2 selection, not interchangeable with its human and yeast orthologs.

Cloning, protein expression and purification
Expression constructs were generated by a standard PCR-based cloning method.

Protein complex preparation and crystallization
AtSCE1 and UFD or UFDC complexes were made by mixing equimolar amounts of proteins and purified by gel filtration chromatography using a Superdex75 column.
AtSCE1 and UFD or UFDC were co-eluted in a single peak in buffer containing 100 mM NaCl, HEPES pH 7.0, 1 mM β-mercaptoethanol, and confirmed by SDS-PAGE.
Purified protein complex was concentrated to 15 g/L using an Amicon Ultra-10K ultrafiltration device (Millipore) prior to crystallization. Crystals were grown at 18 °C by the sitting-drop vapor diffusion method by mixing the protein with an equal volume of reservoir solution containing 0.1 M Bis-tris pH 5.5, 25% w/v PEG3350 for AtSCE1, or containing 15% PEG6000, 5% glycerol for UFDC-AtSCE1 complex. AtSCE1 crystals appeared after one week, while UFDC-AtSCE1 crystals appeared after several months. Big crystals were soaked in mother liquor supplemented with gradually increasing concentration of 5%, 10%, 20% (v/v) glycerol and flash frozen in liquid nitrogen.

Data collection, structure determination and refinement
Diffraction data were collected to 1.20 Å resolution for AtSCE1 and 1.8 Å for UFDC-AtSCE1 complex at ALBA synchrotron in Barcelona (BL13-XALOC beamline) [32]. Data were processed with XDS [33] and scaled, reduced, and further analyzed using CCP4 [34]. More details are shown in Table 1. The structure of AtSCE1 was determined by molecular replacement method using the full length human Ubc9 (PDB code 1U9B) as a search model in PHASER [35]. The structure of UFDC-AtSCE1 complex was determined by molecular replacement method using the solved AtSCE1 as a search model in PHASER [35]. Initial electron density was manually improved to build up the final model using Coot [36], and the refinement was performed using Phenix [37]. Refinement statistics are shown in Table1. instead. Reactions were incubated at 37 °C and aliquots were mixed with loading buffer for analysis by SDS-PAGE, followed by staining with coomassie brilliant blue.

In vitro conjugation reaction and pull-down assays
Ni 2+ Sepharose 6 Fast Flow beads (GE healthcare) were washed with assay buffer containing 100 mM NaCl, 20 mM Bis-tris propane pH 7.5, 0.1% Tween 20, 1 mM DTT and incubated with 30 μg His-AtUFDC for 1 hour. The beads were then incubated with the same buffer containing 2 BSA for 1 hour to block the unspecific binding, followed by washing three times with the assay buffer. 10 μg of the untagged E2 was mixed in 300 μl of assay buffer and added in the beads, followed by incubation for 1 hour. The beads were then washed three times and elute with assay buffer containing 500 mM imidazole. Elution aliquots were mixed with loading buffer and analyzed by SDS-PAGE with coomassie brilliant blue staining.

Temperature and urea induced unfolding assays
The tryptophan fluorescence spectra of temperature induced unfolding measurements were taken at excitation wavelength 280 nM and emission 340 nM in PBS buffer containing 0.15 g/L proteins. Data collection was at temperature from 25 °C to 80 °C with 1 °C steps and 1 minute between steps for equilibration in triplicates. Fraction unfolded was obtained by plotting the normalized fluorescence versus the corresponding temperatures. The urea induced unfolding were performed in PBS buffer by mixing 0.15 g/L proteins with increasing concentrations of urea (0 -9M) at 25 °C overnight to reach equilibrium. Tryptophan emission spectra of urea induced unfolding were obtained by setting the excitation wavelength at 280 nM and collecting emission in the 300-400 nM range These spectra were quantified as the center of spectral mass (ν) according to equation ν = Σν i F i /ΣF i where F i stands for the fluorescence emission at a given wavelength (ν i ) and the summation is carried out over the range of appreciable values of F [38,39]. Fraction unfolded was obtained by plotting the normalized center of spectral mass versus the corresponding urea concentrations. All measurements were conducted using Jasco FP-8200 spectrofluorimeter. Curves were generated using GraphPad Prism 7.0.

E1 and E2 enzymes are specific between human and Arabidopsis
An intriguing question in the field deals with the lack of compatibility between enzymes of the SUMO conjugation pathway from different species, even though the high sequence identity present between E2 enzymes orthologs. We have conducted in vitro SUMO conjugation reactions swapping the SUMOylation components of Arabidopsis and human, including SUMO proteins, E1, and E2. Human RanGAP1, which contains the SUMO consensus motif Ψ-K-X-D/E [28,40], can be sufficiently conjugated by the SUMO machinery from both species, validating it as an in vitro universal SUMO substrate in E3-independent SUMO conjugation assays [41]. In conjugation assays containing human or Arabidopsis E1 and E2, the purified human RanGAP1 is quantitatively SUMOylated and migrates at ~32kD, consistent with the molecular weight of single SUMO modification (Fig. 1A,B).
The results of SUMO swapping between human and Arabidopsis show differences in conjugation rate and SUMO ortholog specificity. Under our experimental conditions, the human E1 and E2 facilitate SUMO conjugation to RanGAP1 more efficiently than the Arabidopsis E1 and E2 pair. This higher efficiency is observed independently on the SUMO1 ortholog (hSUMO1, hSUMO2, AtSUMO1 or AtSUMO2) present in the assay. The Arabidopsis E1 and E2 pair catalyze the conjugation of AtSUMO1 and AtSUMO2 to RanGAP1 to a similar extend, consistently with results obtained using the plant substrate AtCAT3 [13], although with lower efficiency than the human E1 and E2. This reduced activity of the Arabidopsis E1 and E2 is more dramatic when the human SUMO1 or SUMO2 are included in the assay.
( Fig. 1A,B). These results are consistent with a described role of Arabidopsis E1 in SUMO isoform selection [13], which has not been reported for human E1 up to date.
The present results suggest the existence of molecular determinants in Arabidopsis E1 that contribute to confer higher selectivity in SUMO recognition than the human E1 ortholog. Future structural analysis will be required to uncover these molecular determinants present in the Arabidopsis system, which we speculate could be the result of the higher divergence between SUMO isoforms present in plants.
Nonetheless, these results show that SUMOs can be interchanged between systems at different levels, indicating that probably the observed complete incompatibility between species raises from the E2 interaction to E1 and/or to the SUMO-thioester transfer between E1 and E2 (Fig. 1C).
E2-conjugating enzymes display a 65% sequence identity between human and Arabidopsis. When similar SUMO conjugation reactions were performed combining human and Arabidopsis E1 and E2 enzymes, SUMOylated RanGAP1 could not be detected in trace amounts in the Arabidopsis E1 and human E2 combination after 120 min reaction (Fig. 1D). These results indicate despite that the formation of the E1-SUMO thioester could be achieved by combining human or plant SUMOs (Fig. 1A,C), the whole SUMO conjugation process relies on compatible interactions between E1 and E2 enzymes, which are necessary for the SUMO-thioester transfer from the E1 to the E2-conjugating enzyme. This organism-dependent specificity between human and Arabidopsis E1 and E2 interaction is consistent with results obtained between human and P. falciparum systems [21,30].

Complex formation and crystallization
To further define the E1-E2 interaction interface between human and Arabidopsis by protein crystallography, we have purified the recombinant Arabidopsis E2- Complex formation between AtSCE1 and AtUFDC or AtUFD was conducted by mixing equimolar amounts of both components and running a size-exclusion chromatography ( Fig. 2A,B). The fractions of the complex were pooled and concentrated to set up crystallization screens. Interestingly, only the complex including the C-terminal extension of the UFD domain (AtUFDC) produced good diffraction quality crystals after several months at 18°C. The role of this C-terminal extension of the SAE2 E1 large subunit has not been completely clarified, although it might contribute to subcellular compartment localization both in human and Arabidopsis, as it contains a nuclear localization signal [7,42], but it does not seem relevant for the catalytic activity despite the presence of a SIM (SUMO interaction motif) observed in the active form of the SUMO E1 crystal structure [15,43].

Structure of Arabidopsis E2 alone and in complex with the E1 UFD domain
The structure of the Arabidopsis E2-conjugating enzyme, AtSCE1, was solved at 1.2 Å resolution (Fig. 3A) ( Table 1), which allows the visualization of structural details in the final electron density maps, such as the presence of residues with alternate conformations. Intriguingly, the final electron density reveals a double conformation of a 6 residues helical stretch in the C-terminal region of AtSCE1, between Tyr135 and Asp141. Despite this particular feature, the Arabidopsis AtSCE1 structure is highly similar to the structure of human Ubc9 (65% identity and rmsd=0.88 Å for 155 aligned residues) (Fig. 3D). The major differences in the backbone trace with human Ubc9 (PDB code 1A3S) [44] corresponds to the β1-β2 loop, and to a small loop at the end of the β4 strand, produced by a 3 residue substitution (P-Q-G in A.thaliana, E-P-P in human) (Fig. 3D).
We have also solved the crystal structure of AtSCE1 in complex with the UFDC domain of the Arabidopsis E1-activating enzyme at 1,8 Å resolution (Fig. 3C) (Table   1), however the UFDC C-terminal extension was not observed in the final electron density maps. The structure of the complex was solved by molecular replacement using our previous AtSCE1 structure as a search model. As detailed later, the interface region in the AtSCE1 E2-conjugating enzyme comprises contacts in the α1 helix and in the β1-β2 loop with the β-grasp fold of the E1 UFD domain (Fig. 3C), as expected from previous structures of the complex in human and yeast [20][21][22]. The structure of AtSCE1 in the complex is nearly identical to its isolated form (rmsd=0.35 Å for 155 aligned residues), with the most notable difference being the slight variation in the β1-β2 loop conformation (Fig. 3B). Compared with a prior complex structure in human (5FQ2), AtSCE1 displays high degree of similarity with Ubc9 (rmsd=0.47 Å for 155 aligned residues), with a slight shift in the β1-β2 loop to accommodate the AtUFD surface, but superimposition of the E2 moieties results in a shift in the orientation of their cognate UFD domains (Fig. 3E). Moreover, alignment of the UFD domains between human and Arabidopsis (30% identity and rmsd=1.30 Å for 93 aligned residues) also displays a notable rotation of SCE1 and Ubc9 across the β-grasp fold (Fig. 3F). Similar shifted orientations between the two moieties in the complex are also observed when compared with the yeast UFD-Ubc9 complex (Figs. 3G, H, I).

Structural analysis of the interface between Arabidopsis E1 and E2.
The binding surface in AtSCE1 resembles the structures of the complex in human and yeast, and it is formed by contacts between the α1 helix and the β1-β2 loop of AtSCE1. Major specific contacts include Val37, Leu39 and Met40 from the β1-β2 loop and Arg7, Arg14, Lys15, Arg18 and Lys19 from the α1 helix basic patch (Fig.   4A). Differences between human and Arabidopsis in the E2 contact residues are few, only the presence of Arg7 (leucine in human), Lys28 (valine in human) and Val37 (Met36 in human) (Fig. 4B). Moreover, Glu67, which is not located in the SCE1 α1helix/β1-β2 binding module, forms a new hydrogen bond not present in human (Fig.   4). Previous mutational analysis of human Ubc9 contact residues, such as the basic patch in α1 helix, revealed its major role for the interaction to the UFD E1 domain, and was proven to be essential for the transfer of SUMO between E1 and E2 enzymes in the conjugation reaction [45]. Interestingly, despite their high conservation between human and Arabidopsis, E2-conjugating enzymes are not interchangeable in our in vitro conjugation assays (Fig. 1D).
The structure of the AtUFD domain of Arabidopsis E1-activating enzyme displays a similar fold as the human or yeast counterparts (20% identity and rmsd=2.17 Å for 91 aligned residues between Arabidopsis and yeast UFD). However, a major structural difference in AtUFD corresponds to the presence of a longer α2-helix (Fig.   5A), resulting from a sequence insertion occurring in Arabidopsis (Fig. 5C). Residues from this novel longer α2 helix together with residues emanating from central β-sheet of the AtUFD (β4-β2-β3) directly participate in the interface with the E2 enzyme (Fig.   5B). We have previously defined the interacting region between E2 and UFD E1 domain as Low Homology region involved in E2 Binding 2 (LHEB2) [31], because it displays a lower homology in comparison to the rest of the UFD domain: i.e. UFD sequence identity between human and Arabidopsis is 31%, but is reduced to 11% when considering only the LHEB2 binding region. The LHEB2 can be divided in two regions, each establishing interactions with either the α1 helix or the β1-β2 loop of AtSCE1. Interestingly, a unique contact, not present in human or yeast, is observed in Arabidopsis between AtUFD Asn474 and AtSCE1 Glu67, a region far from the α1helix/β1-β2 loop module of the E2-conjugating enzyme (Fig. 5B).

The first contact region in the AtUFD domain is composed by Glu479, Asp483
and Glu516, which contact the positive charged groups of the AtSCE1 α1-helix (Arg14, Lys15, Arg18 and Lys19) and by Leu469 and Leu518, which are buried in a hydrophobic patch formed by the aliphatic side chain atoms of the aforementioned AtSCE1 α1 helix basic residues (Fig. 5B). Interestingly, Phe522 occupies this region in human, whereas in yeast it is occupied by Tyr489 (Fig. 5C,D), all of them buried in a similar aliphatic pocket wedged between basic residues of the AtSCE1 α1 helix [22].
Other contacts present in the AtSCE1 α1 helix include the unique Arg7 (leucine in human), and Ala11, which interact with AtUFD Met471 (glutamine or leucine in human and yeast, respectively) (Fig. 5B,D).
The second region of AtUFD E1 domain includes contacts with the β1-β2 loop surface of AtSCE1, and is composed by interactions from the β2-β3 loop and the longer α2 helix of AtUFD. Leu476, Asp485, Asn491, Tyr492 and Pro504 side chains are the major contacts within the AtUFD domain (Fig. 5B,C). In human and yeast this region is formed by a different type of contacts (Fig. 5C,D), only few similar to Arabidopsis: the central Leu476 (isoleucine and leucine in human and yeast, respectively) and the backbone hydrogen bond contacts between Asn474 and Leu476, with AtSCE1 Val37 and Leu39 (Fig. 5D).
The analysis of the E1 UFD-E2 interfaces between Arabidopsis and human suggest that all contact residue substitutions in the UFD domain across species are accompanied by the compensatory residue substitutions to interact with the E2 enzyme, suggesting a convergent co-evolution of the UFD domains to interact with the conserved residues of the E2 surface. Finally, the structural analysis of our UFD-E2 complex in Arabidopsis in the present work confirms the impaired E2 binding interaction of two predicted AtUFD mutants (L476A, D485A) published previously [31].

SUMO conjugation assays with mutants of the Arabidopsis SCE1
Since SUMO modifier swapping was not a major issue for in vitro conjugation assays between human and Arabidopsis, we attribute the specificity of the SUMO conjugation reaction to the E1-E2 interaction (Fig. 1). This interaction has been proposed to be an essential recognition step for the recruitment of E2 on the E1conjugation enzyme surface. Our structural analysis indicates that the E2 surface in contact with the UFD domain is highly conserved between Arabidopsis and human ( Fig. 4A,B), with only few modifications: Arg7-Gly8 (Leu-Ser in human); Lys28 (Val in human), and Val37 (Met in human) (Fig. 4A,B). To investigate the structural determinants for the specificity between Arabidopsis and human E2 enzymes, we have produced 5 different mutants in Arabidopsis AtSCE1 interface with the UFD domain: R7L; R7L/G8S; R7L/G8S/K28V; R7L/G8S/K28V/V37M; V37M. All residues in AtSCE1 have been replaced by the corresponding residues in human Ubc9 (Fig. 4B).
Additionally, we have also assessed the role of the distinctive contact in Arabidopsis outside the α1-helix/β1-β2 loop module, by producing the AtSCE1 E67S point mutant.
In vitro conjugation assays were conducted using the Arabidopsis enzymes with RanGAP1 as a model substrate. Replacement of AtSCE1 Arg7 and Gly8 (by Leu and Ser in human, respectively), present at the beginning of the α1-helix, did not show any defect on the SUMO conjugation and displayed similar activities as the wild-type form (Fig. 6A). Regarding to AtSCE1 Lys28 (Val in human) and Val37 (Met in human), both present in the β1-β2 loop region, only Val37 had a strong effect on the conjugation activity (Fig. 6A). Such loss of the activity is probably caused by an E1 binding defect, since the R7L/G8S/K28V/V37M and V37M mutants also showed a reduced UFD binding in pull-down assays (Fig. 6B). Finally, disruption of the unique contact described in A.thaliana, Glu67, only displayed a slight reduction in the conjugation activity (Fig. 6A,B).
To further assess the relevance of the Val37 point mutant, protein stability assays were conducted with AtSCE1 wild type and the R7L/G8S/K28V/V37M tetrapoint mutant. Temperature-induced protein unfolding assays, indicate a similar thermal stability for the wild type AtSCE1 (Tm 48.4 °C) and R7L/G8S/K28V/V37M mutant (Tm 49.1 °C) (Fig. 6C). Also, urea-induced protein unfolding assays with wild type and R7L/G8S/K28V/V37M mutant displayed similar half transition concentration values of 2.6 M and 2.4 M urea, respectively (Fig. 6C). These results support the correct structural integrity for the AtSCE1 R7L/G8S/K28V/V37M mutant, probably attributing the SUMO conjugation impairment to a defect in the binding between UFD E1 and E2 enzymes caused by the Val37 substitution for methionine, as confirmed by the correct activity of the AtSCE1 R7L/G8S/K28V triple-mutant (Fig. 6A).
It is worth mentioning here that the Arabidopsis AtSCE1 R7L/G8S/K28V/V37M tetra-mutant, in which most of the contacts with the UFD domain have been replaced for those in human Ubc9, do not display any SUMO conjugation when used as an E2 with human SUMO1 or -2, human E1 and RanGAP1 (data not shown). These results indicate that the different specificities in SUMO conjugation between human and Arabidopsis not only rely on the highly conserved E1 UFD contact surface. Thus, despite the strong conservation between human and Arabidopsis E2 enzymes, surface residues outside the UFD interface in the E2 enzyme are also critical for SUMO specificity across species. Probably these residues outside the UFD interface are required for the interaction with other regions of the E1 enzyme, such as the E1 Cys-domain, as observed in the crystal structures of the ubiquitin E1-E2 complexes captured during the ubiquitin trans-thiolation transfer [18,24].
Interestingly, the different angle orientation of the SUMO E2 enzyme in complex with human, yeast and Arabidopsis SUMO UFD E1 domain probably contributes to provide specificity with its cognate E1 enzyme, thus it might elude steric clashes with other regions of the E1 enzyme during the catalytic trans-thiolation reaction. Our results are also consistent with the SUMO E2 specificity analysis conducted between human and Plasmodium falciparum [21,30], in which the SUMO conjugation reaction was only efficient only after the substitution of half of the E2 enzyme sequence between both species.

Conclusion
Our study provides the first detailed atomic structures of the Arabidopsis SUMO E2-conjugating enzyme, AtSCE1, alone and in complex with the UFD domain of the Arabidopsis SUMO E1-activating enzyme. This interaction between E1 and E2 enzymes represents the initial step for the recruitment of the E2 enzyme to the SUMO conjugation pathway. The atomic details in the E1-E2 interface provided by our structure sheds light to the specificity displayed in the SUMO conjugation pathway across species, in which at least three different E1 UFD binding platforms, in human, yeast and plants, have evolved to interact with a highly conserved cognate E2 enzyme. Our study suggests that this evolution is based on the interaction between the E2-conjugating enzyme and the E1 UFD domain, in which modifications on the UFD domain surface occur to compensate mutations that affect its interaction with the conserved surface of the E2 enzyme. Also, despite this high similarity in the E2 conjugating enzymes in the E1 binding interface (more than 65% identity between human and Arabidopsis), they cannot be efficiently interchanged in our in vitro conjugation assays, indicating that these substitutions are critical for efficient cognate E1-E2 species-dependent recognition.