The Parkinson Disease-linked LRRK2 Protein Mutation I2020T Stabilizes an Active State Conformation Leading to Increased Kinase Activity*

Background: LRRK2 has emerged as one of the most relevant players in Parkinson disease pathogenesis. Results: Enzyme kinetic and modeling studies elucidate the effect of the LRRK2 I2020T mutant on kinase activity. Conclusion: The I2020T mutant stabilizes the active conformation and leads to increased kinase activity. As a result, it binds to type II inhibitors competitively. Significance: This study may contribute to the development of new classes of inhibitors of LRRK2. The effect of leucine-rich repeat kinase 2 (LRRK2) mutation I2020T on its kinase activity has been controversial, with both increased and decreased effects being reported. We conducted steady-state and pre-steady-state kinetic studies on LRRKtide and its analog LRRKtideS. Their phosphorylation differs by the rate-limiting steps: product release is rate-limiting for LRRKtide and phosphoryl transfer is rate-limiting for LRRKtideS. As a result, we observed that the I2020T mutant is more active than wild type (WT) LRRK2 for LRRKtideS phosphorylation, whereas it is less active than WT for LRRKtide phosphorylation. Our pre-steady-state kinetic data suggest that (i) the I2020T mutant accelerates the rates of phosphoryl transfer of both reactions by 3–7-fold; (ii) this increase is masked by a rate-limiting product release step for LRRKtide phosphorylation; and (iii) the observed lower activity of the mutant for LRRKtide phosphorylation is a consequence of its instability: the concentration of the active form of the mutant is 3-fold lower than WT. The I2020T mutant has a dramatically low KATP and therefore leads to resistance to ATP competitive inhibitors. Two well known DFG-out or type II inhibitors are also weaker toward the mutant because they inhibit the mutant in an unexpected ATP competitive mechanism. The I2020 residue lies next to the DYG motif of the activation loop of the LRRK2 kinase domain. Our modeling and metadynamic simulations suggest that the I2020T mutant stabilizes the DYG-in active conformation and creates an unusual allosteric pocket that can bind type II inhibitors but in an ATP competitive fashion.

The effect of leucine-rich repeat kinase 2 (LRRK2) mutation I2020T on its kinase activity has been controversial, with both increased and decreased effects being reported. We conducted steady-state and pre-steady-state kinetic studies on LRRKtide and its analog LRRKtide S . Their phosphorylation differs by the rate-limiting steps: product release is rate-limiting for LRRKtide and phosphoryl transfer is rate-limiting for LRRKtide S . As a result, we observed that the I2020T mutant is more active than wild type (WT) LRRK2 for LRRKtide S phosphorylation, whereas it is less active than WT for LRRKtide phosphorylation. Our pre-steady-state kinetic data suggest that (i) the I2020T mutant accelerates the rates of phosphoryl transfer of both reactions by 3-7-fold; (ii) this increase is masked by a rate-limiting product release step for LRRKtide phosphorylation; and (iii) the observed lower activity of the mutant for LRRKtide phosphorylation is a consequence of its instability: the concentration of the active form of the mutant is 3-fold lower than WT. The I2020T mutant has a dramatically low K ATP and therefore leads to resistance to ATP competitive inhibitors. Two well known DFGout or type II inhibitors are also weaker toward the mutant because they inhibit the mutant in an unexpected ATP competitive mechanism. The I2020 residue lies next to the DYG motif of the activation loop of the LRRK2 kinase domain. Our modeling and metadynamic simulations suggest that the I2020T mutant stabilizes the DYG-in active conformation and creates an unusual allosteric pocket that can bind type II inhibitors but in an ATP competitive fashion.
Parkinson disease (PD), 2 characterized by tremor, rigidity, bradykinesia, and postural instability, is the second most common neurodegenerative disorder after Alzheimer disease. It affects over 1 million Americans and more than 60,000 patients are newly diagnosed each year. PD is caused by a loss of dopaminergic neurons in a part of the brain called substantia nigra. Once damaged, these neurons stop its normal function of producing dopamine, an essential neurotransmitter, and compromise the ability of the brain to control movement. Mutations in several genes have been genetically linked to PD in recent years (1). Among them, leucine-rich repeat kinase 2 (LRRK2) has emerged as one of the most relevant players in PD pathogenesis. At least 40 mutations in LRRK2 have been found in both familial and sporadic forms of PD, and have been associated with typical idiopathic, lateonset PD (2)(3)(4)(5)(6).
LRRK2 is a large and complex protein containing several independent domains, including a leucine-rich repeat (LRR) domain, a Roc domain followed by its associated COR domain, a kinase domain, and a C-terminal WD40 domain (2,7). LRRK2 is unusual in that it encodes two distinct but functionally linked enzymes: a protein kinase and a GTPase (7)(8)(9)(10)(11). Several mutations demonstrate increased kinase activity that is correlated with increased neurotoxicity in central neurons (7,12,13). However, most of the LRRK2 mutations do not manifest their effects by simply increasing kinase activity. Among them, the effect of the mutation I2020T on kinase activity is controversial. Depending on the phosphoryl acceptors, both decreased and increased activity for the I2020T mutant has been reported (14 -16). To better understand the effect of the I2020T mutant on kinase activity, in this study we (i) characterized phosphorylation of two peptide substrates LRRKtide and LRRKtide S , which were reported previously to have different rate-limiting steps (17); (ii) integrated steady-state and pre-steady-state kinetic studies and revealed the different effects of the I2020T mutant on the phosphoryl transfer and the product release steps; (iii) performed inhibition mechanism studies and showed that DFG-out inhibitors exhibit surprising variability in their inhibition mechanisms between wild type LRRK2 and the I2020T mutant; and (iv) conducted metadynamic simulation studies and provided structural support for the observed kinetics.
Kinetic Analysis of LRRK2-catalyzed LRRKtide and LRRKtide S Phosphorylation-The kinase assays for phosphorylation of LRRKtide (RLGRDKYKTLRQIRQ) or LRRKtide S (RLGRD-KYKSLRQIRQ) were conducted in buffer containing 20 mM HEPES (pH 7.4), 50 mM NaCl, 10 mM MgCl 2 , 1 mM DTT, 0.5 mg/ml of BSA, 1 mM ␤-Gly-PO 4 , LRRKtide, ATP, and [␥-33 P]ATP. ␤-Gly-PO 4 is a phosphatase inhibitor. Peptidic substrates and ATP were used at various concentrations as indicated under "Results" and the ratio of ATP to [␥-33 P]ATP was kept constant at all ATP concentrations (250 M ATP, 5 Ci of [␥-33 P]ATP). The reactions were conducted in duplicate, initiated by the addition of 6 nM LRRK2, and incubated at room temperature for 150 min. The reactions were stopped by the addition of 20 mM EDTA and the mixture was transferred to a multiscreen PH filtration plate (Millipore, Billerica, MA) and washed six times with 75 mM H 3 PO 4 . The plate was dried, filters were removed, and the samples were counted with a scintillation counter. Background reactions were conducted in the absence of LRRK2. In all cases, reaction progress curves for production of phospho-LRRKtide or phospho-LRRKtide S were linear over at last 240 min and allowed calculation of initial velocities.
Modeling of LRRK2 Kinase Domain and Docking-The LRRK2 kinase domain, between residues 1859 and 2138, was modeled using Modeler version 9.11 (18 -21). Modeling details have been described in detail previously (22). Briefly, the main variables in homology modeling are template selection and sequence alignment between the target and the template. B-raf kinase with 33% sequence identity to LRRK2 kinase domain was used as a template for homology modeling because this enzyme had the highest sequence similarity with LRRK2 around the active site region and the ATP binding hinge compared with other kinases. In addition, B-raf and LRRK2 inhibitors can cross-inhibit each other supporting our choice of template selection for modeling (details to be published elsewhere). Recently, the kinase domain of Roco4 from Dictyostelium discoidium (PDB code 4F0G), a closely related member of the LRRK2 family, was published (23) and we constructed a model of LRRK2 with this as a template and compared with our previous model. The comparison of the two LRRK2 models revealed that the overall C␣ atom root mean square deviation was less than 1.3 Å. The I2020T mutant was introduced using PyMol and the resultant structure was subjected to molecular dynamics (MD) thermalization using the OPLS.2005 force field.
Docking of ATP was carried out using Glide version 2.2 (24) against the DFG-in conformation of LRRK2 (WT and I2020T), which treats the receptor rigidly. Docking of type II inhibitors used in this study was carried out using the induced fit docking (25) protocol implemented in the Schrödinger suite (25). Briefly, in the first stage of induced fit docking 20 initial poses are generated using Glide with a softened potential to allow for clashes with the receptor. For each of the top 20 poses from the initial softened-potential docking step, a cycle of protein side chain prediction and full residue minimization is performed using Prime to generate 20 induced-fit receptor structures (26). All residues having at least one atom within 5 Å of any atom from the 20 ligand poses are refined. Finally, the ligand is redocked using Glide with default settings into each induced-fit receptor structure and a composite score that accounts for the protein/ligand interaction energy (GlideScore) and the total energy of the system (Prime energy) is used to rank the inducedfit structures.
Metadynamics Simulations-Metadynamics (27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37) simulations for the activation loop were carried out with Desmond version 3.0 using two collective variables (CVs). To test for robustness of the simulations and choice of CVs, we chose two independent pairs of CVs (one pair with distances and the other pair with torsions) and ran separate simulations. We chose the distance between the center of mass of a collection of atoms in the ␤-sheet (residues 1878 -1906) and the N-terminal C-helix (residues 1915-1928) as the first CV denoted (s). For the second CV denoted (z), we chose the center of mass of a subset of residues in the activation loop (Tyr-2018, Ala-2021, Cys-2024, Arg-2026, and Met-2027) and the center of mass for the ATP binding hinge comprised of residues (Met-1947, Leu-1949, and Lys-1952. The distance between the N-terminal ␤-sheet domain and C-helix in the N-terminal domain describes the opening and closing motion of LRRK2 around the hinge region. The center of mass of the activation loop is the primary degree of freedom of interest for this study because it directly involves the transition from active to inactive form. For measuring the dissociation rates for ATP a root mean square deviation of CV without including the H-atoms for ATP was chosen (s). In addition, a second angular CV was chosen using center of masses for two fixed water oxygen atoms and the center of mass for ATP (z).
The default equilibration protocol in Desmond was run before the metadynamics simulations, which relaxes the system in a gradual way with progressively weaker harmonic restraints and heating of the system from 0 to 300 K. The total length of each production simulation was 250 ns for the activation loop and 50 ns for ATP dissociation at a temperature of 300 K and pressure of 1 atm. A Gaussian biasing potential with a height of 0.01 kcal/mol was injected at 0.09-ps intervals throughout the simulation and the free-energy surface (FES) was reconstructed from these Gaussians. Multiple simulations were carried out using different initial random seeds. Convergence was observed in all cases and the FES constructed using these simulations were virtually identical. For the ATP dissociation experiment, the highest energy barrier was measured and used with Eyring's equation to calculate theoretical off-rates and half-lives.
Data Analysis, Basic Equations-Data were analyzed by nonlinear least-squares using either Sigma Plot or Grafit software packages. Standard kinetic mechanisms for two-substrate reactions and their rate equations are shown below. Ping K A and K B are substrate dissociation constants from EA and EB, respectively.
Rapid equilibrium random/steady-state ordered, For rapid equilibrium systems, K A , K B , ␣K A , and ␣K B are substrate dissociation constants from EA, EB, and EAB; for steadystate systems, K A is a substrate dissociation constant from EA, ␣K A and ␣K B are Michaelis constants. See Segel (38) for definitions of mechanisms, substrate dissociation constants, and ␣.

t-WT LRRK2-and the Mutation t-I2020T-catalyzed Phosphorylation of LRRKtide, Initial
Velocity Studies-In this study we used N-terminal truncated WT and the I2020T mutant, therefore they are referred to t-WT and t-I2020T, respectively. To determine the effect of the t-I2020T mutant on kinase activity, phosphorylation of LRRKtide (RLGRDKYKTLRQIRQ) was characterized. Initial velocities were measured as a function of LRRKtide concentration at several fixed concentrations of ATP for the t-I2020T mutant (data not shown). The complete data sets were subjected to global analysis by nonlinear least-squares fits to the three standard mechanisms (ping-pong, rapid equilibrium order, and random/steady-state ordered) using Equations 1-3. Statistically the data fits the random/steady-state ordered mechanism the best for the t-I2020T mutant, yielding the following estimates averaged from two independent experiments: k cat ϭ 4.4 Ϯ 0.1 min Ϫ1 , K ATP ϭ 1.6 Ϯ 0.4 M, K LRRKtide ϭ 110 Ϯ 23 M, and ␣ ϭ 7.3 Ϯ 1.6 for the t-I2020T mutant. The same kinetic mechanism for t-WT was reported previously (17). The kinetic parameters of the t-I2020T mutant and those of t-WT are summarized in Table 1. A direct comparison of the t-I2020T mutant with t-WT LRRK2 is also shown in Fig. 1A, where the initial velocities were measured as a function of [ATP] at a saturating concentration of LRRKtide. The data reveal that (i) the t-I2020T mutant has lower kinase activity than t-WT and (ii) ATP binds to the mutant much tighter than t-WT, with a K ATP value 40-fold lower than that of t-WT.
t-WT LRRK2-and the Mutation t-I2020T-catalyzed Phosphorylation of LRRKtide S , Initial Velocity Studies-We have previously shown that the rate-limiting step for LRRKtide phosphorylation catalyzed by t-WT or the t-G2019S mutant is the product release step, whereas peptide LRRKtide S (RLGRD-KYKSLRQIRQ, the phosphorylatable threonine of LRRKtide was replaced by a serine residue) is rate limited by the phosphoryl transfer step (17). To evaluate whether the t-I2020T mutant has a different effect on the phosphoryl transfer step, the initial velocity studies for LRRKtide S phosphorylation were conducted (data not shown). The data suggest a random/ steady-state ordered mechanism for the t-I2020T mutant (the same kinetic mechanism for t-WT was reported previously, Ref. 17). The kinetic parameter estimates for the t-I2020T mutant and those for t-WT are summarized in Table 1. A direct comparison of the t-I2020T mutant with t-WT LRRK2 is shown in Fig. 1B, where the initial velocities were measured as a function of [ATP] at a saturating concentration of LRRKtide S for t-WT and the t-I2020T mutant. Unlike LRRKtide, when LRRKtide S is the phosphoryl acceptor, the t-I2020T mutant is more active than t-WT.
pL-dependence of Steady-state Kinetic Parameters and SKIE-We have seen both decreased and increased kinase activity for the t-I2020T mutant on two very similar peptide substrates. We hypothesize that it is due to the difference in the rate-limiting steps of the reactions. Although we have shown that for t-WT and the t-G2019S mutant, LRRKtide is rate limited by a product release step and LRRKtide S is associated with  a slower phosphoryl transfer step, we need to make certain that those features remain for the t-I2020T mutant. Therefore, the steady-state kinetic parameter pL (pH or pD) profile of LRRKtide or LRRKtide S phosphorylation was carried out for the t-I2020T mutant in a triple buffer consisting of 50 mM MES, 100 mM Tris, and 50 mM acetic acid along with other components as described under "Experimental Procedures." The combination of the three buffers has the advantage of providing a constant ionic strength for the entire pL range studied. The pH stability of the mutant was determined by incubating the enzyme at the desired pH and assaying aliquots at pH 7.5. The enzyme was determined to be stable from pH 6 to 9.3. Small activity loss occurs at pH Ͻ 6.0. However, the small activity loss does not affect the measurement of initial velocities, because enzyme was added from a stock solution at pH 7.5 to the assay mixture and assayed for 30 min in the presence of substrate, which is likely to provide some protection against denaturation. The t-I2020T-catalyzed phosphorylation of LRRKtide was monitored in a pL range of 4.9 -8.5 for k cat (initial velocities were measured at saturating ATP and LRRKtide concentrations of 60 M and 2 mM, respectively) and for k cat /K m (initial velocities were measured at ATP and LRRKtide concentrations less than the K m values, 1 and 50 M, respectively). The pL profile of k cat for the t-I2020T mutant revealed an acidic limb ( Fig. 2A) and was fit to Equation 4, yielding one apparent pK a of 6.8 and 7.3 from pH and pD profiles, respectively.
y ϭ C 1 ϩ 10 ϪpH /10 ϪpKa (Eq. 4) k cat /K m exhibited a similar pL dependence as k cat , revealing one apparent pK a of 7.2 and 7.5 from pH and pD profiles, respectively (Fig. 2B). The pL-independent steady-state parameters revealed SKIE values of 1.01 and 1.10 for k cat and k cat /K m , respectively, suggesting that like t-WT or the t-G2019S mutant, the rate-limiting step for t-I2020T-catalyzed LRRKtide phosphorylation is product release.
The same pL-dependent studies were conducted for LRRKtide S as well. The pL profiles revealed similar shape with one apparent pK a of 7.7 and 8.4 from pH and pD profiles, respectively, for k cat ; and one apparent pK a value of 8.1 and 8.7 from pH and pD profiles, respectively, for k cat /K m (Fig. 2, C and  D). The pL-independent steady-state parameters revealed SKIE values of 0.50 and 0.61 for k cat and k cat /K m , respectively, suggesting that the rate-limiting phosphoryl transfer step is the characteristics of the t-I2020T-catalyzed LRRKtide S phosphorylation. To directly monitor the effects of the t-I2020T mutant on phosphoryl transfer of LRRKtide phosphorylation, we conducted pre-steady-state kinetics as discussed below.
Pre-Steady-state Kinetics-LRRKtide phosphorylation was carried out at saturating concentrations of LRRKtide and ATP in the presence of 12 nM enzyme. The reaction was stopped by the rapid quench flow instrument. As shown in Fig. 3, A and B, burst kinetics or biphasic kinetics was observed for both t-WT and the mutant. Fitting data to a single exponential followed by a steady-state rate equation: y ϭ A[1 Ϫ exp(Ϫk 1 t)] ϩ k 2 t, yielding values for A that represents the concentration of active enzymes, a pseudo first-order rate constant k 1 for the burst phase of the reaction, and a steady-state rate k 2 . Averaged values of these parameters determined from two experiments are summarized in Table 2. Comparison of the parameters for t-WT and the t-I2020T mutant reveals that (i) the concentration of the active form of the mutant is significantly lower than that of t-WT; (ii) the t-I2020T mutant increases the rate of the phosphoryl transfer step of LRRKtide phosphorylation by greater than 3-fold compared with t-WT; (iii) first glance at the steady-state rate k 2 , the rate of product release, we found that the mutant is associated with a slower product release rate than t-WT. However, when we take into account the concentrations of the active forms of both enzymes, a product release rate constant (k 2 /A) similar to t-WT was revealed for the t-I2020T mutant ( Table 2).
Inhibition Studies of Two Known DFG-out Inhibitors-In an effort to develop LRRK2 inhibitors, we have identified that ponatinib and imatinib, two DFG-out inhibitors of Abl kinase (39,40), inhibit t-WT LRRK2 with IC 50 values of 160 nM and 8.3 M, respectively; and inhibit the t-I2020T mutant with IC 50 values of 800 nM and 90 M, respectively (Table 3). Next, the mechanism of inhibition of these two type II inhibitors was studied and data were analyzed using the method of replots as described in detail previously (41). Briefly, initial velocities of LRRK2-catalyzed LRRKtide phosphorylation were measured as a function of inhibitor concentration [I] at varied ATP concentrations and at a fixed LRRKtide concentration. The shape of replots of (k cat ) ATP versus [I] and (k cat /K m ) ATP versus [I] were used to determine the inhibition mechanism. For ponatinib, as expected based on its mechanism of inhibition with its targeted kinase Abl (40), it shows ATP non-competitive inhibition toward t-WT; both (k cat ) ATP and (k cat /K m ) ATP are dependent on [ponatinib] (Fig. 4, B and C). However, surprisingly ponatinib inhibits the t-I2020T mutant via an ATP competitive mechanism, (k cat ) ATP is independent of [ponatinib] (Fig. 4E). To better judge the two inhibition mechanisms, noncompetitive versus competitive inhibition, data were also globally fit into equations reflecting the two mechanisms, respectively. The goodness of fit was judged by an F test, in which a probability value lower than 0.05 indicates that one of the two equations fits the data better. F test suggests that data for t-WT fit better to the noncompetitive inhibition (probability value Ͻ Ͻ 0.001), and data for the mutation t-I2020T fit better to the competitive inhibition (probability value Ͻ Ͻ 0.001), consistent with the method of replots. The data for imatinib were analyzed the same way using both methods of replots and global fitting (data not shown). The same inhibition patterns were also observed for imatinib, it inhibits the t-I2020T mutant in an ATP competitive manner, whereas it inhibits t-WT in an ATP noncompetitive mechanism. As a comparison, an ATP competitive inhibitor GSK3-XIII shows competitive inhibition against both t-WT and the t-I2020T mutant (data not shown). Its IC 50 values are summarized in Table 3.
Effects of the I2020T Mutant on the Structure of LRRK2 and ATP Binding-A structural model for LRRK2 in the DYG-out form (inactive conformation) has been previously published (22). In the DYG-out form of LRRK2 (DFG-out in many other kinases), the tyrosine side chain (Phe in other kinases) moves to a position that would otherwise be occupied by ATP. The kinase in this form therefore no longer binds ATP. However, in this conformation, the kinase structure still remains competent to bind hinge-binding elements of type II inhibitors such as ponatinib. Our model is consistent with x-ray structures of ponatinib bound to other kinases such as cAbl in their DFG-out state (40). In our LRRK2 model, residue Ile-2020 lies in the activation loop adjacent to the DYG motif as shown in Fig. 5A. The isoleucine residue at position 2020 was mutated to threonine using PyMol and the resultant I2020T mutant could be accommodated within the LRRK2 structure as evidenced by its stability after MD thermalization. Analysis of the energy minimized structure reveals that the side chain of Thr-2020 makes a hydrogen bond with the backbone carbonyl of Asp-2017 (Fig.  5B). This in turn results in the side chain of Asp-2017 making a salt-bridge network with Lys-1906 and Glu-1920 (Fig. 5B). To fully understand the thermodynamic consequence of this

TABLE 2 Pre-steady-state kinetics of t-WT and the t-I2020T mutant
A represents the concentration of active enzymes; k 1 is a pseudo first-order rate constant for the burst phase of the reaction; k 2 is a steady-state rate; and k 2 /A is a steady-state rate constant.

t-WT t-I2020T
A     published previously by our group (22). The change in the center of mass for the C␣ atoms of residues Tyr-2018, Ala-2021, Cys-2024, Arg-2026, and Met-2027 was an effective descriptor for the motion of the activation loop and represented the second collective variable (z) in our calculation. These final set of atoms used to define the collective variables for these calculations were reached after several pilot calculations to ensure that the MD path leads to both activation-loop movement and flip of the DYG motif from DYG-in to DYG-out occurred in a correlated fashion. The results of the metadynamic simulation are plotted as a FES, which also represents the barriers involved in the motion of the kinase as it switches between the DYG-in and DYG-out states. The FES for WT LRRK2 shows a smooth transition between DYG-in and DYG-out, whereas the FES for I2020T shows two clear energy minima representing the two states (Fig. 6). There is an average 4.7 kcal/mol energy barrier for the mutant that separates the two states from each other.
To measure the affinity of ATP for the binding site of LRRK2, we performed 50-ns metadynamic simulations with the ATPbound form of WT LRRK2 and the I2020T mutant. In this form of MD, a root mean square deviation type collective variable was used in conjunction with an angular collective variable between ATP and two fixed water oxygen atoms in the bulk solvent. Through several trial simulations, we found that this set of collective variables is optimal because it encourages ATP to move away from the binding site. The type and number of barriers encountered during this process estimates the energy required to dissociate from the binding site, which is a measurement of K m for ATP. The FES for dissociation of ATP from WT LRRK2 and I2020T are shown in Fig. 7. The FES of WT LRRK2 shows that the ATP dissociation path has an average barrier of 3.8 kcal/mol, which results in half-life of the ATP binding site at the active site for 3 s. The mutant I2020T shows an average barrier of 5.5 kcal/mol, which results in a half-life of ATP at the binding site for 75 s. A key distinction between the ATP dissociation mechanism of WT and I2020T is the number of intermediate steps involved. WT LRRK2 shows four distinct states including bound/unbound states and two intermediates (Fig. 7A). The mutant on the other hand shows more of a twostate dissociation with a deep distinctive low energy minima corresponding to the bound conformation and a second one corresponding to the unbound form (Fig. 7B).

DISCUSSION
Unlike the most common PD-linked LRRK2 mutation G2019S, which has been consistently reported to increase kinase activity (7,8,11,17,43), the effect of the I2020T mutant on kinase activity is controversial. It was reported to have a lower kinase activity in assays measuring LRRKtide phosphorylation (14, 16) but has either higher or unchanged activity in assays monitoring autophosphorylation depending on the autophosphorylation sites (12,15,44). To better understand the substrate-dependent effects of the I2020T mutant, we chose two peptide substrates, LRRKtide and LRRKtide S . They differ only by the phosphorylatable sites, threonine in LRRKtide versus serine in LRRKtide S . However, as a result, the rate-limiting steps of their phosphorylation are quite different as reported previously for t-WT LRRK2 and the t-G2019S mutant (17). In this study, we found that these features remained for the t-I2020T mutant, LRRKtide phosphorylation is rate-limited by a slow product release step as evidenced by a SKIE value close to 1 for k cat , whereas a slow phosphoryl transfer step is associated with LRRKtide S phosphorylation as evidenced by an inverse SKIE value of 0.5 for k cat . Using these two peptide substrates, we could directly evaluate the effects of the t-I2020T mutant on the two steps, phosphoryl transfer and product release steps even in steady-state kinetic studies.
Initial velocity studies suggest that the t-I2020T mutant has a lower kinase activity in catalyzing LRRKtide phosphorylation than t-WT LRRK2. However, for LRRKtide S phosphorylation, we surprisingly observed a higher kinase activity for the t-I2020T mutant, 2 times t-WT kinase activity. These steadystate kinetic studies suggest that the t-I2020T mutant increases phosphoryl transfer but it seems to slow down product release. Pre-steady-state kinetic studies for LRRKtide phosphorylation also revealed a faster phosphoryl transfer rate for the t-I2020T mutant, greater than 2 times that for t-WT. Although we were well aware that the I2020T mutant degrades faster than WT in cells (45), we were surprised to see that the concentration of the active form of the mutant is significantly lower than that of t-WT. Taken into account the real concentrations of the active forms of both enzymes, we found that (i) for LRRKtide phosphorylation, the t-I2020T mutant releases products with a rate similar to t-WT and (ii) for LRRKtide S phosphorylation, the mutant increases kinase activity or the rate of phosphoryl transfer by 7-fold compared with t-WT. Combined results from our steady-state and pre-steady-state kinetic studies suggest that the t-I2020T mutant increases kinase activity by accelerating the phosphoryl transfer rate. However, this increase was masked in some reactions in which the rate-limiting step is product release, like in LRRKtide phosphorylation.
The LRRK2 research community has put in extensive efforts in identifying LRRK2 substrates since LRRK2 was linked to PD in 2004. As a result, a number of proteins have been reported to be phosphorylated by LRRK2 in vitro although none of them have been demonstrated as LRRK2 natural substrates. The first such potential LRRK2 substrate is moesin, a member of the ERM (ezrin/radixin/moesin) family that anchors the actin cytoskeleton to the plasma membrane. It was reported in 2007 by Jaleel et al. (14) that LRRK2 phosphorylated moesin at Thr 558 . LRRK2 also phosphorylates ezrin and radixin at the residues equivalent to Thr 558 .
Although ezrin/radixin/moesin as real substrates of LRRK2 is debatable in the LRRK2 field, a peptide encompassing Thr 558 of moesin was derived and named LRRKtide. LRRKtide has since been widely used in the LRRK2 field for activity and inhibition studies. Inhibitors evaluated using LRRKtide in vitro shows consistent potency in cells. Our previous publication (22) on the G2019S mutant based on LRRKtide phosphorylation and modeling studies made the same conclusion as a crystal structural study on Roco4, the D. discoideum version of LRRK2 (23). All these observations make our assessment of the I2020T mutant based on LRRKtide and its analog reliable.
Currently the physiological substrate of LRRK2 is unknown. Recent studies reported that LRRK2 phosphorylates Tau (46,47) and MKK6 (48). However, due to the slow kinetics of the phosphorylation of these proteins catalyzed by LRRK2, we were unable to conduct the pre-steady-state studies using MKK6 or Tau proteins and were unable to determine whether the I2020T mutant increases the phosphoryl transfer rate of protein substrate phosphorylation. In the future, with the hope that LRRK2 natural substrate would be identified, we will re-evaluate this mutant.
It has been reported that the I2020T mutant exhibits resistance toward ATP competitive inhibitors due to its small K ATP (49). We were able to replicate this finding using an ATP competitive inhibitor GSK3-XIII (Table 3). We rationalize that DFG-out or type II inhibitors could overcome this issue, because they are known to bind noncompetitively with ATP by interacting with an allosteric site that exists when the kinase switches to an inactive DFG-out form (39, 40, 50 -52). However, two known DFG-out inhibitors, ponatinib and imatinib, inhibit the t-I2020T mutant with a potency 5-10-fold weaker than they inhibit t-WT (Table 3). Inhibition mechanism studies revealed that ponatinib and imatinib exhibit a surprising difference in their inhibition mechanisms between t-WT and t-I2020T mutant. Both ponatinib and imatinib inhibit t-WT in an expected ATP non-competitive mechanism, whereas they inhibit the t-I2020T mutant in an ATP competitive mechanism.
To explain the observed inhibition mechanism and kinetics, we compared homology models of LRRK2 in the DYG-in FIGURE 7. A, the FES of the undocking process of ATP in WT LRRK2 as a function of the angle and root mean square deviations of CVs. The FES shows an internal, partially bound and an external energy minimas. A detailed description of the various intermediate steps is provided in text. Briefly, the dissociation path of ATP involves two metastable states (states 2 and 3) with a relatively smooth transition between bound and unbound states. B, the FES of the undocking process of ATP in I2020T as a function of the angle and root mean square deviations of CVs. Unlike WT LRRK2, the dissociation appears to follow a two-step-like mechanism with a highly stable bound state that persists much longer compared with WT and then dissociate very quickly to an unbound external energy minima.
(active) and DYG-out (inactive) forms for both WT and the I2020T mutant. Almost in all kinases, there are two synergistic motions that occur as a result of kinase activation. At the N-terminal of the flexible activation loop there exists a conserved three-residue motif, DYG for LRRK2 (DFG for most kinases). The Gly residue acts as a bipositional switch, which allows the Asp and Tyr/Phe to flip and exchange positions during the activation process. In addition, the activated configuration of DYG flip facilitates an inward movement of C-helix making a saltbridge with a conserved Lys residue in the N-terminal ␤-sheet domain (53,54). In the case of LRRK2, this salt bridge is between Lys-1906 and Glu-1920. The motions of two lobes thus become coordinated for precisely positioning ATP bound to the N-lobe with respect to the catalytic site on the C lobe. Homology modeling of LRRK2 suggests that the mutation I2020T is in fact right next to this critical regulatory conserved DYG motif. The threonine side chain of the mutant makes a hydrogen bond with the backbone carbonyl of Asp-2017. This would in fact rigidify the DYG-motif and have a stabilizing effect on the DYG-in conformation. The FES associated with FIGURE 8. A, a simulation interaction diagram showing prolonged interactions between ATP, amino acid residues of WT LRRK2, and water molecules over the course of simulation. The simulation interaction diagram for WT LRRK2 shows several interactions between ATP phosphate groups and water molecules. This indicates that the ATP molecule moved fairly rapidly from the binding site to other non-native sites. The only strong persistent interaction (47%) is observed between the ␥-phosphate of ATP and Arg-110 (structurally equivalent to R2026 in native LRRK2 sequence). B, a simulation interaction diagram showing prolonged interactions between ATP, amino acid residues of I2020T mutant, and water molecules over the course of simulation. Unlike WT LRRK2, the simulation interaction diagram for I2020T shows direct interaction between the ATP molecule and protein side chains that make up the hinge region and the ATP binding cavity. This indicates that the molecule has higher residence time in the ATP binding site compared with WT. In addition to the strong interaction with Arg-110 (structurally equivalent to Arg-2026 in native LRRK2 sequence), the ATP molecule also makes strong and persistent (51% total) water-mediated hydrogen bonds with Asp-170 (structurally equivalent to Asp-2017) of the DYG motif.
the activation loop transition indicates that the I2020T mutant has a barrier of 4.7 kcal/mol that separates the DYG-in active state from the DYG-out inactive state. On the other hand, WT LRRK2 has no such barrier. In other words, the I2020T mutant stabilizes the DYG-in active form and may not easily switch to the DYG-out inactive form. As a result, the I2020T mutant leads to increased kinase activity (increased phosphoryl transfer rate). In addition, the I2020T mutant is able to bind traditional type II DFG-out inhibitors but in an unusual type I-like ATP competitive inhibitor binding mode. A detailed study of the binding mode for these inhibitors has been previously published from our research group (42).
To measure the effect of the energy barrier of the activation loop transition associated with the mutant on ATP binding, we carried out metadynamic simulation, in which the bound ATP molecule is encouraged to move away from the active site during the course of the simulation. As the ATP molecule dissociates, the energy barriers of dissociation are measured and the highest energy barrier can be converted to off-rates or residence time using Eyring's equation. Analysis of the FES from metadynamic simulations for WT and the I2020T mutant indicates a similar but not identical path for ATP dissociation but with slightly different barriers. The majority of barriers in ATP dissociation from WT LRRK2 involve a series of hydrogen bonds between ATP and the hinge region, glycine-rich loop, and Arg-2026. In the case of the I2020T mutant, the side chain of Thr-2020 makes a hydrogen bond with the backbone and side chain of Asp-2017. This rigidifies the activation loop region leading to a less mobile DYG region (part of ATP binding pocket), whose motion is necessary for dissociation of ATP. We used a simulation interaction diagram, which measures the contribution of each amino acid and water molecule toward the motion of the ligand to analyze the MD trajectories for WT LRRK2 and I2020T mutant (Fig. 8). The analysis shows that the rigidified DYG region of I2020T makes strong and persistent hydrogen bonds with the phosphate groups of ATP thereby contributing toward prolonged residence time of ATP. These interactions are absent from the WT LRRK2 simulation trajectory. Metadynamic simulations estimated a half-life of 3 s at 37°C for ATP bound to WT LRRK2, whereas a half-life of 75 s for ATP bound to the I2020T mutant. This prolonged ATP-I2020T complex explains the kinetic observation that the I2020T mutant has an unusually high binding affinity toward ATP.

CONCLUSION
Although the x-ray structure for LRKK2 is yet to be determined (a closely related Roco4 kinase domain is now available (23)), enzyme mechanism studies and molecular modeling have provided significant insights into the nature of LRRK2 and the PD-linked mutation I2020T. Given the unique insights gained in this work with traditional type II kinase inhibitors binding to the I2020T mutant in an ATP-competitive way, it is conceivable that new classes of inhibitors will be needed to gain the desired affinity and selectivity to target the I2020T mutant in the pursuit to treat Parkinson disease. The presence of two exposed cysteine residues (Cys-2024 and Cys-2025) close to the ATP binding site opens up the possibility of developing cova-lent or irreversible inhibitors for LRRK2. Such inhibitors could provide a sustained duration of inhibition for this resistanceconferring mutant. In addition, with further studies, possibly including x-ray crystallography or long time scale molecular dynamics simulations, it may eventually be possible to understand the nature of the allosteric pocket in the I2020T mutant enzyme and develop highly selective inhibitors that will also overcome the resistant issue.