An ‘invisible’ ubiquitin conformation is required for efficient phosphorylation by PINK1

The Ser/Thr protein kinase PINK1 phosphorylates the well-folded, globular protein ubiquitin (Ub) at a relatively protected site, Ser65. We had previously shown that Ser65-phosphorylation results in a conformational change, in which Ub adopts a dynamic equilibrium between the known, common Ub conformation and a distinct, second conformation in which the last β-strand is retracted to extend the Ser65 loop and shorten the C-terminal tail. We here show using Chemical Exchange Saturation Transfer (CEST) NMR experiments, that a similar, C-terminally retracted (Ub-CR) conformation exists in wild-type Ub. Ub point mutations in the moving β5-strand and in neighbouring strands shift the Ub/Ub-CR equilibrium. This enabled functional studies of the two states, and we show that the Ub-CR conformation binds to the PINK1 kinase domain through its extended Ser65 loop and is a superior PINK1 substrate. Together our data suggest that PINK1 utilises a lowly populated yet more suitable Ub-CR conformation of Ub for efficient phosphorylation. Our findings could be relevant for many kinases that phosphorylate residues in folded proteins or domains.

3 PhosphoUb is generated on mitochondria by the Ser/Thr protein kinase PINK1 (Kazlauskaite et al, 2014;Koyano et al, 2014;Kane et al, 2014;Ordureau et al, 2014;Wauer et al, 2015b), which is stabilised on the cytosolic face of mitochondria upon membrane depolarisation (Narendra et al, 2010). PINK1 phosphorylates Ub attached to outer mitochondrial membrane proteins, and this recruits and allosterically activates the E3 ligase Parkin (Wauer et al, 2015a;Kazlauskaite et al, 2015;Sauvé et al, 2015;Kumar et al, 2015). PINK1 also phosphorylates Parkin in its Ub-like (Ubl) domain, which is required for full Parkin activation and leads to strong, localised mitochondrial ubiquitination (Kondapalli et al, 2012;Wauer et al, 2015a;Ordureau et al, 2014). PINK1/Parkin action attracts adaptor proteins and recruits the mitophagy machinery, leading to clearance of the damaged organelle (Lazarou et al, 2015;Heo et al, 2015). The pathophysiological importance of PINK1/Parkin mediated mitophagy is underlined by the fact that mutations in PINK1 and Parkin are linked to autosomal recessive juvenile Parkinson's disease (AR-JP), a neurodegenerative condition arising from loss of dopaminergic neurons in the substantia nigra (Corti et al, 2011;Pickrell & Youle, 2015).
The generation of phosphoUb by PINK1 is mechanistically poorly understood. PINK1 is an unusual Ser/Thr kinase, highly divergent from other kinases in the kinome (Manning et al, 2002). In part this is due to several large insertions in the kinase Nlobe, which complicate structural modelling (Trempe & Fon, 2013). Also its substrate, Ub, is a non-classical kinase target since its 76 amino acids form a globular, highly robust and stable β-grasp fold, in which Ser65 is markedly protected. Ub Ser65 resides in the loop preceding the β5-strand, and its side chain hydroxyl group engages in a backbone hydrogen bond with Gln62. In addition, nearby side chains of Phe4 and Phe45 further stabilise the Ser65-containing loop (Fig. EV1A). Ub Ser65 is structurally identical to Ser65 in the Parkin Ubl domain, but the two substrates lack similarity at the sequence level and a PINK1 phosphorylation consensus motif is not obvious (Kazlauskaite et al, 2014). The Ser65 position and interactions within a wellfolded, globular domain make this residue an unlikely phosphorylation site for PINK1 or indeed any kinase.
Ub is highly similar in the >100 Ub crystal structures in the protein data bank (Perica & Chothia, 2010), and its biophysical properties and availability have made it a popular model system for protein folding studies (Jackson, 2006) and nuclear magnetic resonance (NMR) techniques (Torchia, 2015;Lange et al, 2008;Fushman et al, 2004). NMR studies in particular have shown that despite its compact fold and high intrinsic stability, Ub is very dynamic and contains several regions of local conformational flexibility (Lange et al, 2008). These include a mobile four-residue Cterminal tail, as well as a flexible β-hairpin structure, the β1/β2-loop, that alters the interaction profile of Ub (Lange et al, 2008;Hospenthal et al, 2013;Phillips & Corn, 2015). Importantly, we previously discovered that Ser65-phosphorylation resulted in a further, dramatic conformational change in Ub (Wauer et al, 2015b).
The Ser65 loop and the last β5 strand were previously not known to be conformationally dynamic, yet phosphorylation led to an equilibrium between two phosphoUb conformations (Fig. 1A). The first state resembles the 'common' Ub conformation observed all reported crystal structures to date (hereafter simply Ub conformation). This phosphoUb conformation was confirmed in a crystal structure (Wauer et al, 2015b) and more recently by an NMR structure (Dong et al, 2017).
More striking was a second conformation, in which the entire last β-strand slipped by two amino acids, extending the Ser65 loop, and simultaneously shortening the Ub Cterminal tail (hereafter referred to as the Ub-CR conformation for C-terminally retracted) (Wauer et al, 2015b;Dong et al, 2017). This change is facilitated by a Leurepeat pattern in the β5-strand: Leu67, Leu69 and Leu71 occupy complementary Leu-pockets in the Ub core, whereas Leu73 is mostly solvent exposed. In the Ub-CR conformation observed in phosphoUb, Leu73 occupies the Leu71 pocket, Leu71 occupies the Leu69 pocket, and Leu69 occupies the Leu67 pocket, resulting in Leu67 residing in a more exposed position that was formerly occupied by Ser65 (Fig. 1A). Experimentally, the phosphoUb-CR conformation was supported by large (>1.5 ppm) chemical shift perturbations and by determination of the hydrogen bonding patterns for the β-sheet, using long range HNCO-based NMR analysis (Wauer et al, 2015b). A recent NMR structure of the phosphoUb-CR conformation confirmed our findings (Dong et al, 2017). We speculated that the Ub-CR conformation is a consequence of Ub phosphorylation, enabling proteins to specifically recognize phosphoUb (Wauer et al, 2015b).
We here show that the Ub-CR conformation can indeed be detected in unphosphorylated Ub, when analysing NMR exchange timescales accessible by Chemical Exchange Saturation Transfer (CEST) experiments (Fig. 1B). This previously unrecognised equilibrium between a common Ub and a Ub-CR conformation in wild-type (wt) Ub can be shifted in either direction through point mutations in unphosphorylated Ub. Crystal structures as well as biophysical and NMR measurements enable in-depth characterisation of the Ub-CR conformation, and reveal its functional relevance during Ser65 phosphorylation. The Ub-CR conformation of Ub, with its mobile Ser65 loop, forms a more stable complex with PINK1 as assessed by NMR binding studies. More importantly, Ub in the Ub-CR conformation is a superior substrate for PINK1, and we provide evidence that the Ub-CR conformation is required for efficient PINK1 phosphorylation. Together, our work suggests that PINK1 utilises a lowly populated, 'invisible' form of Ub in which the buried phosphorylation site becomes exposed.

Identification of a Ub-CR conformation in wild-type Ub
We had previously shown that phosphoUb resides in a conformational equilibrium in which the β5-strand (aa 67-71) slides in its binding groove (Fig. 1A). The immediate question arising, even during the review of our earlier paper (Wauer et al, 2015b), was whether a similar Ub-CR conformation may exist in unphosphorylated, wt Ub.
Dynamic aspects of Ub have been under excessive scrutiny, in particular by NMR, and numerous studies have collectively covered most motional timescales from fast ps internal motions up to µs-ms conformational exchange processes using RDC analysis (Lange et al, 2008;Torchia, 2015). Our initial detection of the phosphoUb / phosphoUb-CR transition was enabled by a near-equal population of both states, and ZZ-exchange experiments indicated a slow exchange (2 s -1 ) between these conformations (Fig. 1B).
Given the timescales of motion probed in previous Ub studies, we hypothesised that a very lowly populated, slow exchanging Ub-CR conformation in wt Ub could have 6 been systematically missed in previous experiments. Furthermore, we assumed that raising the temperature of the experiment would lower the energy barrier between the two states and potentially allow an increase in population of the Ub-CR species.
The detection of lowly populated, also known as 'dark' or 'invisible', conformational states can be enabled by CEST experiments (Vallurupalli et al, 2012;Kay, 2016). In CEST, samples are pulsed at specific frequencies to saturate individual resonances in the 'invisible' state. During an exchange period, magnetisation is transferred between conformers resulting in an amplification of the signals from previously Together, CEST experiments revealed the existence of a previously undetected Ub conformation in wt Ub, which by chemical shift analysis resembles the phosphoUb-CR conformation reported earlier.

Stabilisation of the Ub-CR conformation
With the occurrence of the wt Ub-CR conformation confirmed, we set out to stabilise it for further study. For this, we mutated Leu67 to Ser, hoping that occupation of the Leu67 pocket by Ser encourages β5-strand slippage to fill this hydrophobic pocket with Leu69 instead ( Fig. 2A). Indeed, 1 H-15 N BEST-TROSY 2D spectra (bTROSY) of 15 N-labelled Ub L67S showed 73 peaks implying a single Ub conformation (Appendix Fig. S2). The chemical shift pattern did not match wt Ub, but more closely resembled the pattern seen for the phosphoUb-CR conformation. This can be assessed using well-dispersed reporter resonances, such as Lys11 (Fig. 2B), while a global comparison of the full spectra can be drawn from chemical shift perturbation heat maps (Fig. 2C). Hence, Ub L67S predominantly adopts the Ub-CR conformation despite lacking phosphorylated Ser65.

Mimicking the Ub-CR conformer in Ser65 phosphoUb
We also wanted to study the phosphoUb-CR conformer in more detail, and hence we phosphorylated Ub L67S with Pediculus humanus (Ph)PINK1 (Wauer et al, 2015b;Woodroof et al, 2011). Strikingly, phosphorylation transformed the simple Ub-CR bTROSY spectrum to a complicated spectrum with the occurrence of many additional peaks (data not shown). Phos-tag gels and mass-spectrometry (MS) showed that PhPINK1 phosphorylates Ub L67S at multiple sites, on Ser65, and on Thr66 or on the introduced Ser67 ( Fig. EV2A and B). A mixture of phosphorylated species explains the complexity of the observed NMR spectrum. PINK1 shows exquisite preference for Ser65 in wt Ub, and only phosphorylates Thr66 at very high enzyme concentrations and late time points (Wauer et al, 2015b). Hence, the doubly phosphorylated species are a result of the Ub-CR conformation induced by the L67S mutation. These data indicated that the Ub-CR conformation has profound effects on PINK1-mediated Ub phosphorylation, but suggested that this mutation was limited in its usefulness for the study of phosphoUb-CR.
To overcome this and to generate exclusively Ser65-phosphorylated Ub in the Ub-CR conformation, Leu67 was mutated to Asn, and Thr66 was mutated to Val (termed hereafter Ub TVLN mutant) (Fig. 2D). Ub TVLN was phosphorylated only once, on Ser65 ( Fig. 2E), showed a clean, single-species bTROSY spectrum highly similar to Ub L67S in the unphosphorylated form, and when phosphorylated was highly similar to the phosphoUb-CR conformation ( Fig. 2F-H, Appendix Fig. S3A and B) (Wauer et al, 2015b). Together, this showed that phosphoUb TVLN is an excellent mimic for the Ub-CR conformer of the PINK1-catalysed phosphoUb species.

Crystal structures of Ub in the Ub-CR conformation
The identification of Ub mutants stably in the Ub-CR conformation allowed us to obtain high-resolution crystal structures of Ub L67S (1.63 Å) and phosphoUb TVLN (1.6 Å) ( Table 1, Fig. 3, Fig. EV3). Both structures confirmed that the β5-strand is retracted by two amino acids, and Ser/Asn67, Leu69, Leu71 and Leu73 adopt near identical conformations as compared to Ser65, Leu67, Leu69 and Leu71, respectively, seen in previous Ub structures ( Fig. 3A-D). Hydrogen bonding patterns observed in the crystal structures matched the experimentally determined hydrogen 8 bonding pattern for phosphoUb-CR conformation (Wauer et al, 2015b), and the phosphoUb TVLN structure is similar to a recently reported NMR structure of phosphoUb-CR ( Fig. EV3E and F).
The structures reveal important consequences of the Ub-CR conformation. The Ser65-containing loop (aa 62-66) protrudes from and lacks defined contacts with the Ub core, is flexible judging by B-factor analysis, and in Ub L67S adopts distinct conformations in the two molecules in the asymmetric unit (Fig. 3A, Fig. EV3D).
Likewise, in phosphoUb TVLN, the Ser65-containing loop is extended and seemingly mobile, and the phosphate group is exposed and does not contact the Ub core. A further important feature of the Ub-CR conformation is the disruption of Ub interaction interfaces, the most important of which is the Ile44 hydrophobic patch, which also utilizes Leu8 in the flexible β1/β2-hairpin, and Val70 and His68 of Ub β5strand (Komander & Rape, 2012). In the Ub-CR conformation, the Ile44 hydrophobic patch is disrupted due to dislocation of β5-residues Val70 and His68 (Fig. 3E). In contrast, a second interaction site, the Ile36 hydrophobic patch (Hospenthal et al, 2013) is intact (Fig. 3F, Fig. EV3C). Finally, retraction of β5-strand by two residues reduces the reach and conformational flexibility of the important Ub C-terminal tail.

Affecting Ub/Ub-CR conformational equilibrium
Mutating the first hydrophobic residue of the β5 strand, Leu67, favours the Ub-CR conformation, since Leu69 and Leu71 can occupy alternative positions easily. We reasoned that mutating Leu71 to a larger residue, which cannot occupy the Leu69 position, might stabilise it in the common Ub conformation, and disfavour the Ub-CR conformation after phosphorylation (Fig. 4A). Indeed, this was the case; Ub L71Y displays a common Ub spectrum without phosphorylation, and a spectrum highly similar to the common phosphoUb species after phosphorylation ( Fig. 4B-D, Appendix Fig. S4A and B). Hence, Ub L71Y is a mutation in which the Ub-CR conformation is disfavoured.
Thus far the introduced mutations change residues on the slipping β5-strand. We wondered whether residues in the vicinity, e.g. from the neighbouring β1-strand, could also shift the observed equilibrium. A good candidate was Phe4 with its solvent exposed side chain (Fig. 4E), which would be anticipated to have only subtle effects on Ub conformation per se. Indeed, Ub F4A displayed a wild-type-like bTROSY spectrum ( Fig. 4H and F, Appendix Fig. S5A). However, strikingly, phosphorylation of Ub F4A results in a spectrum where the most intense peaks are in a position associated with the phosphoUb-CR conformation and only a minor (~12% by peak intensity) population of peaks match the common Ub conformation ( Fig. 4H and G,   Appendix Fig. S5B). This reverses the population ratio observed in the wt phosphoUb spectrum. This demonstrates that while the mutant resides in the common Ub conformation without phosphorylation, it almost completely shifts to a Ub-CR conformation upon phosphorylation. Hence, residues contacting and stabilising the slipping β5-strand are able to affect the conformational equilibrium.

Comparative stability studies of Ub mutants
The fascinating and unexpected conformational plasticity of Ub with regards to β5strand slippage was further confirmed in comparative studies. We had previously shown decreased thermal stability of phosphoUb, which we speculated was due to the Ub/Ub-CR equilibrium (Wauer et al, 2015b). Indeed, differential scanning calorimetry (DSC) experiments revealed that Ub-CR mutants Ub L67S and Ub TVLN mutant display a T m of ~83 ºC, with or without phosphorylation ( Fig. EV4) (compared to 97 ºC for wt Ub, and 87 ºC for the phosphoUb equilibrium, (Wauer et al, 2015b)).
In comparison, Ub F4A displays an intermediate stability (Tm 89 ºC) consistent with NMR findings. Importantly, Ub L71Y is as stable as wt Ub (Tm 96 ºC) indicating that the mutation does not induce unfolding, but merely stabilises the common Ub conformation. Hence, the Ub-CR conformation is less thermostable as compared to the common Ub conformation, and this explains lower stability of phosphoUb.

Additional NMR evidence for a wt Ub-CR conformation
As already discussed above, the Ub-CR conformation results in a retracted Cterminus and an extended Ser65-containing loop, conferring new dynamic properties to the structure at these locations. As Ub mutants can modulate the newly found Ub / Ub-CR equilibrium, this prompted us to search for additional evidence for a Ub-CR conformation beyond CEST (Fig. 1, Fig. EV1, Appendix Fig. S1). For this, we used fast (ms timescale) hydrogen solvent exchange experiments known as clean chemical exchange (CLEANEX) and 15 N{ 1 H} heteronuclear NOE (hetNOE) experiments (that consider ps timescales of motion, Fig. 1B).
CLEANEX experiments measure the ability of backbone amide protons to exchange with the solvent, thus reporting on the relative solvent exposure of each residue.
CLEANEX experiments for each Ub variant revealed a similar set of solvent accessible residues for wt Ub, Ub L71Y and Ub F4A (Appendix Fig. S6A and B), but considerably more solvent exchange especially in the Ser65-loop region in Ub TVLN. This is consistent with the structural data. Interestingly, residues of the nearby Leu8-loop report on the conformational preferences of each Ub mutant through their population averaged rates of solvent exchange (Fig. 5A). Note the loop in the TVLN mutant has the greatest degree of solvent accessibility, with the F4A mutant and wt Ub rates being greater than the L71Y Ub-CR inhibited mutant. This correlates with the overall stability seen in the T m measurements (Fig. EV4).
Further, indirect evidence for a contribution of the Ub-CR conformation in wt Ub was provided by hetNOE analysis, which assesses fast ps flexibility of individual residues (Appendix Fig. S7A). Probing the mobility of Arg74 with hetNOE experiments reveals an increased rigidity in Ub TVLN compared to wt Ub. In contrast the hetNOE values for Arg74 in the L71Y spectrum indicate increased flexibility as compared to the wt Ub spectrum ( Fig. 5B and C). Given the above considerations one could infer that wt Ub has a contribution from its lowly populated Ub-CR conformation to the bulk dynamic properties. This would indicate that L71Y is the most 'pure' common Ub conformation of the mutants studied. While it is possible that Ub mutations contribute to some of the observed effects in CLEANEX and hetNOE analysis, the observed differences in stability and dynamics clearly also correlate with accessibility of the Ub-CR conformation.

The Ub-CR conformation affects ubiquitination reactions
Our identification of Ub mutants adopting the Ub-CR conformation facilitated experiments to test the biochemical impact of this species on Ub assembly enzymes.
As discussed above, retraction of the Ub C-terminal tail reduces flexibility of this key feature of Ub, but also disjoints the Ile44 hydrophobic patch, which is important for many enzymes of the assembly cascade.
Nonetheless, we found that Ub TVLN, which adopts the Ub-CR conformation in solution, was readily charged by E1 onto E2 enzymes, including UBE2D3, UBE2L3, UBE2S, UBE2N and UBE2R1 (Fig. 6A), which is perhaps surprising in light of recent findings that that the hydrophobic patch is important for E1-mediated E2 charging (Singh et al, 2017). However, the E1 reaction is known to be relatively permissive and can also accommodate conformation changing Ub mutations such as Ub L73P (Békés et al, 2013).

The Ub-CR conformation stably binds PhPINK1
While a Ub-CR-inducing mutation had the anticipated inhibitory effects on the ubiquitination cascade, we still wondered whether this conformation had physiological roles. A number of observations pointed towards potential importance in PINK1-mediated Ub phosphorylation. As discussed above, Ser65 in wt Ub is poorly accessible, but becomes more exposed in the Ub-CR conformation.
Moreover, phosphorylation of Ub L67S and Ub TVLN mutants was markedly accelerated as compared to wt Ub as shown by qualitative Phos-tag gels (Fig. 2E).
We hence tested how PINK1 interacted with its substrates, and performed bTROSY experiments with unlabelled PhPINK1 (aa 115-575) and 15 N-labelled wt Ub, Ub mutants, and Parkin Ubl domain (aa 1-76), respectively, in the absence of MgATP ( Fig. 7). In the presence of PhPINK1 all peaks were exchange broadened to some extent due to the formation of a weakly associated 62 kDa complex. A subset of Ub/Ubl peaks, which were additionally broadened, revealed the residues that interact with PhPINK1 (Fig. 7, left column). The same samples were also used in CLEANEX experiments, whereby the binding of PhPINK1 to Ub or Ubl masks the interacting residues on the substrate (Fig. 7, right column).
Remarkably, the footprint of PhPINK1 on its substrates varied (Fig. 7, middle column). In wt Ub, as well as Ub F4A and Ub L71Y, broadened residues correspond to the C-terminal tail and the Ile44 patch, but strikingly did not include residues from the Ser65-containing loop. In these three samples, a similar degree of overall line broadening suggests similar (weak) binding. CLEANEX experiments of substrates without PhPINK1 (light colours) and with PhPINK1 (dark colours) reveal the Ile44 patch interaction of these substrates, but also confirm relative protection of the Ser65-loop, in which only Lys63 is sufficiently solvent exposed to be observed.
In contrast, the Ub TVLN mutant as well as the Parkin Ubl form larger interfaces involving the entire β5-strand, and importantly, all residues from the Ser65- Together, these experiments indicated that the significantly faster rate of Ub phosphorylation seen in the Ub-CR mutants such as Ub TVLN can be explained with enhanced binding of PINK1 to Ub-CR, which can form additional interactions via the Ser65-loop.

Ub conformations affect PINK1 activity
The fact that Ub mutations stabilise the Ub-CR conformation in the absence of phosphorylation (Figs. 2-5), and the discovery that wt Ub dwells in a Ub / Ub-CR equilibrium under physiological conditions ( Fig. 1 and 5), opened the fascinating 13 possibility that the Ub-CR conformation is used or even required for PINK1-mediated Ub phosphorylation.
To test this, we compared phosphorylation rates by treating Ub, Ub mutants and Parkin Ubl samples with PhPINK1 / MgATP, in qualitative experiments using Phostag gels (Fig. EV5A), or in semi-quantitative, real-time experiments using 15 Nlabelled Ub/Ubl substrates in NMR experiments (Fig. 8). Direct assessment of individual peak disappearance/appearance over time from unphosphorylated to fully phosphorylated samples enabled generation of phosphorylation rate curves, revealing strikingly different rates (Fig. 8). The fastest rates were observed for Ub TVLN and Parkin Ubl, and these were almost indistinguishable from each other, even at lower enzyme concentrations (Fig. EV5B). 50% of the substrate was phosphorylated after ~2 min. The Ub F4A sample was also quite fast, being halfphosphorylated after ~5 min. In contrast, it took ~90 min and ~275 min to phosphorylate 50% of wt Ub or Ub L71Y, respectively, under identical conditions. Hence, phosphorylation of the Ub-CR mutant Ub TVLN is ~45-140 fold faster as compared to variants where this conformation is much less populated (wt Ub) or disfavoured (Ub L71Y). This suggested that PINK1 not only prefers the Ub-CR conformation, but that it requires it for efficient phosphorylation of a Ub molecule.

DISCUSSION
Ubiquitin is a most fascinating molecule. Despite being the focus of three decades of biochemical, biophysical and structural research, we here uncover a new Ub conformation, in which the C-terminal β5-strand of Ub is retracted by two residues, which extends the upstream Ser65 containing loop, and shortens the otherwise extended Ub C-terminus. We show that Ub adopts the Ub-CR conformation under ambient conditions, and although this conformation is lowly populated, our data suggest that it is functionally relevant.
Our previous work showed that a very similar Ub-CR conformation was readily induced in Ser65-phosphoUb (Wauer et al, 2015b) which was recently confirmed by an NMR solution structure (Dong et al, 2017). We hence set out to to identify stable versions of phosphoUb-CR for further study. To our surprise, we identified point mutations that readily adopted the Ub-CR conformation even without 14 phosphorylation, enabling us to shift the equilibrium. Still, we initially failed to detect a Ub-CR conformation in wt Ub, which was clearly a minute conformation not observable within the limits of datasets extending to µs dynamics of Ub (Lange et al, 2008). Based on the assumption that a potential wt Ub to Ub-CR exchange would occur at slower rates (as shown by the exchange rate of phosphoUb and phosphoUb-CR) and that the Ub-CR conformation would only be a lowly populated state referred to in the literature as a 'dark' or 'invisible' conformation (Baldwin & Kay, 2009;Kay, 2016), CEST experiments would be the analytical method of choice. Importantly, we also show a physiological role for a Ub-CR conformation. Ub is a well-folded, stable protein, and as such is an unlikely kinase substrate. Many protein kinases prefer or require disordered target sequences for phosphorylation. The wellordered Ser65-containing loop in Ub does not fit this criterion, but the more mobile loop provided in the Ub-CR conformation enables efficient binding and phosphorylation. Hence, Ub-CR mutants are superior PINK1 substrates. Considering wt Ub, it is tempting to speculate that PINK1 stabilises the Ub-CR conformation, or indeed, that a Ub conformational change may impose a rate-limiting step for phosphorylation. So far, we have not observed this with wt Ub, but the timescales of binding experiments (µs) vs. conformational change (s) presents a challenge to directly detect a precatalytic state with wt Ub. It is exciting that we may be able to mimic this precatalytic state with the Ub TVLN mutant, and this may be useful for future structural studies on PINK1.

15
Our findings likely have pathophysiological relevance. PINK1 mutations result in AR-JP, and our results reveal that its key substrate, Ub, needs to be in a particular conformation to enable efficient phosphorylation. It is easy to imagine that conditions or binding partners that stabilise Ub in a common conformation (e.g. Ile44-patch binding domains in mitochondrial associated proteins) may impede PINK1 activity and imbalance the system. In this context, it will also be interesting to test whether different chain contexts modulate the observed Ub/Ub-CR equilibrium and affect the rate at which chains can be phosphorylated. A further open question relates to the Parkin Ubl domain, for which there is no evidence at current of a similar, C-terminally retracted Ubl conformation.
We had previously shown that the Ub-CR conformation is present also in phosphoUb (Wauer et al, 2015b), but strikingly, the only known phosphoUb receptor, Parkin, recognizes the common phosphoUb conformation and does not utilise the more distinctive phosphoUb-CR conformation (Wauer et al, 2015a). It is possible that alternative receptors for phosphoUb-CR exist, but it is also imaginable that this form exists predominantly to facilitate phosphorylation in the first place. While the Ub-CR conformation explains how Ser65 can be phosphorylated by PINK1, questions remain how other sites on Ub, such as well-ordered Thr12 and Thr14 on the β2strand, can be phosphorylated.
More globally, our data explains how an inaccessible phosphorylation site in a folded protein can be targeted via exploitation of an 'invisible' conformation. Hence, our work is likely relevant for other kinases that target folded domains or proteins.

Molecular biology
Ub constructs were cloned into pET17b vectors, and site directed mutagenesis was carried out using the QuikChange protocol with Phusion polymerase (NEB).

Protein purification
All Ub mutants were expressed in Rosetta 2 (DE3) pLacI cells, and purified following the protocol of (Pickart & Raasi, 2005). In short, unlabelled proteins were expressed in 2xTY medium, protein expression was induced at OD 600 of 0.6 -0. Reactions were quenched at the given time points with EDTA-free LDS sample buffer.

Mass-spectrometry analysis
LC-MS analysis was carried out on an Agilent 1200 Series chromatography system coupled to an Agilent 6130 Quadrupole mass spectrometer. Samples were eluted from a phenomenex Jupiter column (5 µm, 300 Å, C4 column ,150 x 2.0 mm) using an acetonitrile gradient + 0.2% (v/v) formic acid. Protein was ionized using an ESI source (3 kV ionization voltage) and spectra were analysed in positive ion mode with a mass range between 400-2,000 m/z. Averaged spectra were deconvoluted using the manufacturer's software and plotted using GraphPad Prism (version 7).
Reaction progress at 25°C was monitored using LC-MS, and once there were no changes in recorded spectra, the reaction mixture was dialysed against water, using a 3.5 kDa cut-off Dialysis cassette (Thermo scientific). The dialysate was applied to an anion exchange (MonoQ 5/50 GL, GE Life Sciences) column. PhosphoUb was eluted by 50 mM Tris (pH 7.4) and further purified by SEC (Superdex 75, GE Life Sciences) into NMR-buffer. PhosphoUb TVLN for crystallography was purified by SEC in 25 mM Tris (pH 7.4).

Crystallisation, data collection and structure determination
Ub L67S was crystallised at 12.5 mg/ml by sitting-drop vapour diffusion against 3 M (NH 4 ) 2 SO 4 , 0.1 M MES (pH 6.0) using a 2:1 protein to reservoir ratio at 18 ˚C. A single crystal was harvested and vitrified in liquid nitrogen.
PhosphoUb TVLN was crystallised at 11.2 mg/ml by sitting-drop vapour diffusion Data collection and refinement statistics can be found in Table 1.

Stability measurements
Samples were dialysed into NMR buffer (18 mM Na 2 HPO 4 , 7 mM NaH 2 PO 4 , 100 mM NaCl (pH = 7.2) using 3.5 kDa MW cut off dialysis cassettes (Thermo Scientific) and subsequently diluted to 50 µM. Differential scanning calorimetry (DSC) was performed using a VP-Capillary DSC instrument (Malvern Instruments). Samples were scanned at a heating rate of 90 °C / h in mid-feedback mode. Data were corrected for instrumental baseline using average buffer scans recorded immediately before and after Ub runs and plotted. After concentration normalisation, the intrinsic protein baseline between pre-and post-transitional levels was corrected using the progress function in the Origin software supplied with the instrument. Corrected endotherms were fitted to a non-two state model allowing T m , ∆H calorimetric and ∆H van't Hoff to vary independently.
For E2 charging and E2-mediated assembly HsUBE1 was used at 0.2 µM, Ub was used at 20 µM and E2s were used a 4 µM. For E3-mediated assembly HsUBE1 was used at 0.2 µM, Ub was used at 20 µM, E2s were used a 2 µM and GST-cIAP1, GST-TRAF6, HUWE1, pParkin were used at 5 µM, while HOIP RBR LDD was used at 1 µM. Samples were taken at indicated time points, the reactions quenched with 19 LDS sample buffer with reducing agent unless otherwise indicated, resolved on 4-12% SDS Gradient gels (NuPage) and stained with Instant Blue SafeStain. A representative example of an experiment done at least in duplicate is shown.

General acquisition parameters
NMR acquisition was carried out at 25 ˚C unless otherwise stated on either Bruker Avance III 600 MHz, Bruker Avance II+ 700 MHz or Bruker Avance III HD 800 MHz spectrometers equipped with a cryogenic triple resonance TCI probes. Topspin

Backbone chemical shift assignments
De novo assignments or reassignments (L67S, TVLN, pTVLN, F4A, pF4A) NMR acquisition was carried out at 25 ˚C on Bruker Avance III 600 MHz spectrometer equipped with a cryogenic triple resonance TCI probe. Backbone chemical shift assignments were completed using Bruker triple resonance pulse sequences. HNCACB spectra were collected with 512*32*55 complex points in the 1 H, 15 N, 13 C dimensions respectively. CBCA(CO)NH, HNCO and HN(CA)CO spectra were collected with 512*32*48 complex points in the 1 H, 15 N, 13 C dimensions respectively. All experiments were collected using Non Uniform Sampling (NUS) at a rate of 50% of complex points in the 1 H, 15 N, 13 C dimensions respectively and

N { 1 H }-heteronuclear NOE measurements
15 N { 1 H }-heteronuclear NOE (hetNOE) measurements were carried out using standard Bruker pulse programs, applying a 120° 1 H pulse train with a 5 ms interpulse delay for a total of 5 s interleaved on-or off-resonance saturation. The hetNOE values were calculated from peak intensities according the equation I on /I off .

CLEANEX experiment on Ub (Fitting)
All CLEANEX experiments were collected at 800 MHz with a 3 s acquisition delay and mixing times of 5.

PhPINK1 -Ub/Ubl binding experiments
Binding experiments were performed by recording BEST-TROSY and CLEANEX (with 166.4 ms mixing time) with 65 µM Ub/Parkin Ubl constructs with and without equimolar amounts of PhPINK1. For the BEST-TROSY experiments, the peak heights of the datasets with PhPINK1 were normalised against the respective peaks without PhPINK1 for each Ub / Parkin Ubl construct and plotted accordingly. For the CLEANEX experiments, the absolute peak heights with and without PhPINK1 were plotted side by side. Data were plotted with GraphPad Prism (version 7).

Acknowledgements
We would like to thank Minmin Yu and beam-line staff at Diamond Light Source, beam lines I-04. Access to DLS was supported in part by the EU FP7 infrastructure grant BIOSTRUCT-X (contract no. 283570

Accession numbers
Coordinates and structure factors for have been deposited with the protein data bank accession codes 5OXI (Ub L67S), 5OXH (phosphoUb TLVN).