br Results br Discussion In
Discussion In this paper we present insights into the observed specificities of inhibitors targeting the ubiquitin-activating and related E1 Taxifolin via crystal structures of the specific NEDD8-E1 inhibitor MLN4924, the dual NEDD8/Ub E1 inhibitor ABPA3, and the selective Ub-E1 inhibitor MLN7243 bound to yeast Uba1, and augment these structures with biochemical and computational studies. Although our crystal structures quite accurately depict the bound state of ubiquitin⋅inhibitor adduct to Uba1, due to the complex mechanism of inhibition of these compounds it is difficult to describe all of their specificity determinants, especially when the structures of the transition states of the enzyme bound to inhibitors are unavailable where the catalytic cysteine domain is expected to be in close proximity to the adenylation domain (Olsen et al., 2010, An and Statsyuk, 2015b). The purified Ub⋅MLN4924 adduct was shown to exhibit a more than 100-fold faster dissociation rate for UBA1 in comparison with the Ub⋅compound I adduct (Chen et al., 2011), thus suggesting that inhibitor potency is the sum of the rate of adduct formation and the affinity between the adduct and E1. Therefore, it is clear that a high binding affinity of the final Ub⋅inhibitor adduct to the E1 enzyme will undoubtedly contribute significantly to its specificity. Differences in the ATP-binding pockets of different E1 enzymes can thus be exploited by complementing the unique pharmacophores presented by each of these enzymes (Figures 4D–4H). For example, R554 in Uba1 can be targeted by extending MLN7243 with an H-bond acceptor, or even a negatively charged group, thus rendering the inhibitor more specific for UBA1. The E1 ATP-binding pockets also differ in their secondary structure elements, which influence the available E1 surface interacting with the inhibitors (Figure S4). Whereas in Uba1 the loop between β-strand 19 and α-helix 26 is interrupted by a 310 helix and features three β and one γ turn, this loop is shorter in UBA3 (PDB: 3GZN) and UBA2 (PDB: 1Y8Q). Atg7 (PDB: 3VH4) has the longest insertion in this region, which folds into the longer α-helix 6. UBA5 (PDB: 3H8V) (Bacik et al., 2010) represents the second longest insertion with α-helix 4 being framed by β turns. In the SUMO and NEDD8 E1 enzymes there is an additional contact with the inhibitor provided by the UFD (represented by residues 474–478 and 448–451, respectively) to this pocket. This domain defines the specificity for the E1-E2 transthioesterification and exhibits conformational flexibility (Huang et al., 2007). In the closed conformation of the E1 enzymes, where it is in proximity to the adenylation domain, the UFD may influence inhibitor binding. Furthermore, in the superposition, M97 of the SUMO E1 and M405 of Atg7 pose clashes with the N10 substituents in MLN4924 and ABPA3 as well as the SCF3-substituted phenyl group of MLN7243, hence suggesting why these compounds are less potent against these two enzymes (Brownell et al., 2010, An and Statsyuk, 2013, An and Statsyuk, 2015a). Both MLN7243 and MLN4924 are currently in phase 1 and phase 1b clinical trials, respectively, for the treatment of advanced solid tumors (Sarantopoulos et al., 2016, Shah et al., 2016), thus underlining their therapeutic potential. Our structural and computational studies represent a starting point to design more potent and selective E1 inhibitors. While UBA1 inhibition has been proposed to overcome the limitation of proteasomal inhibitors against solid tumors, both SUMO and NEDD8 E1s have already been shown to be potential targets for cancer therapy (Kessler et al., 2012). In addition, the specific inhibition of individual E1s can further improve our understanding of the diverse roles played by Ub/Ubl post-translational modifications in eukaryotes. In summary, our structures can be used to design additional inhibitors featuring nucleobase and ribose isosteres displaying improved pharmacokinetic properties (St Jean and Fotsch, 2012), potencies, and selectivities. The MLN4924⋅NEDD8 adduct has been shown to be unable to inhibit UBA1; thus, by combining adenosyl sulfamates with the C-terminal tails of Ub/Ubl modifiers, one can impart high specificity for individual E1 enzymes to these inhibitors. As the inhibitor-resistant mutants reported so far belong to either the ATP-binding pocket or the Ubl-binding site, developing inhibitors that encompass both these sites will offer a significant advantage to prevent the evolution of resistance mutants. The reported inhibitor-resistant mutants should be taken into account to develop second-generation inhibitors, which may turn out to be very effective in combination therapies. While on the one hand very specific E1 inhibitors will greatly improve our understanding of the Ub/Ubl post-translational modifications, they might be disadvantageous from a therapeutic perspective, as treatment with a highly specific inhibitor targeting a single E1 enzyme will presumably lead to the more rapid emergence of resistant mutations. To overcome this limitation pan-specific E1 inhibitors such as ABPA3, which target the two E1 enzymes UBA1 and UBA3, may be more advantageous, given that both UBA1 and UBA3 converge on the same pathway, namely ubiquitylation. Such compounds could be more difficult to overcome via drug-resistance mutations. The structures presented here together with those of MLN4924⋅NEDD8 bound to the NEDD8 E1, SUMO⋅AMP mimicking intermediates bound to SUMO E1 (Olsen et al., 2010), ATP-UBA5 (Bacik et al., 2010, Oweis et al., 2016), and yeast Atg7CTD-Atg8-MgATP (Noda et al., 2011) provide a platform to develop potent adenosyl sulfamate inhibitors targeting individual E1s, which will be highly attractive compounds for basic research and drug discovery purposes.