N-terminal splicing extensions of the human MYO1C gene fine-tune the kinetics of the three full-length myosin IC isoforms

The MYO1C gene produces three alternatively spliced isoforms, differing only in their N-terminal regions (NTRs). These isoforms, which exhibit both specific and overlapping nuclear and cytoplasmic functions, have different expression levels and nuclear–cytoplasmic partitioning. To investigate the effect of NTR extensions on the enzymatic behavior of individual isoforms, we overexpressed and purified the three full-length human isoforms from suspension-adapted HEK cells. MYO1CC favored the actomyosin closed state (AMC), MYO1C16 populated the actomyosin open state (AMO) and AMC equally, and MYO1C35 favored the AMO state. Moreover, the full-length constructs isomerized before ADP release, which has not been observed previously in truncated MYO1CC constructs. Furthermore, global numerical simulation analysis predicted that MYO1C35 populated the actomyosin·ADP closed state (AMDC) 5-fold more than the actomyosin·ADP open state (AMDO) and to a greater degree than MYO1CC and MYO1C16 (4- and 2-fold, respectively). On the basis of a homology model of the 35-amino acid NTR of MYO1C35 (NTR35) docked to the X-ray structure of MYO1CC, we predicted that MYO1C35 NTR residue Arg-21 would engage in a specific interaction with post-relay helix residue Glu-469, which affects the mechanics of the myosin power stroke. In addition, we found that adding the NTR35 peptide to MYO1CC yielded a protein that transiently mimics MYO1C35 kinetic behavior. By contrast, NTR35, which harbors the R21G mutation, was unable to confer MYO1C35-like kinetic behavior. Thus, the NTRs affect the specific nucleotide-binding properties of MYO1C isoforms, adding to their kinetic diversity. We propose that this level of fine-tuning within MYO1C broadens its adaptability within cells.

Myosins are molecular motors that utilize ATP binding, hydrolysis, and product release to perform mechanical work along actin filaments. All myosins share a highly conserved motor domain, a lever arm, and a tail domain that exhibits substantial diversity. The mechanochemical transduction pathway of the ATPase cycle and the major structural biochemical intermediates are conserved across the myosin family (1). Importantly, to perform their myriad biological functions, myosins have evolved unique kinetic adaptations by modulating the rate and equilibrium constants of the ATPase cycle (2)(3)(4).
Myosin IC (MYO1C), 2 a class I myosin, is produced as three splice isoforms that differ only at the N-terminal region (NTR) (Fig. 1) (5). Although they share identical motor domains, three calmodulin-binding IQ motifs containing a nuclear localization signal (6) and a membrane-binding tail domain (Fig. 1A), each isoform has acquired differences in function and nuclearcytoplasmic partitioning. MYO1C C , the first isoform to be identified, localizes mostly to the cytoplasm and interacts with plasma membrane phosphoinositides via its PH domain; functionally, it participates in the generation of membrane tension, cell migration, vesicle trafficking, signal transduction, and hearing (7)(8)(9)(10)(11)(12)(13)(14). MYO1C C nuclear import is regulated by calcium (15), but its nuclear functions remain unknown. MYO1C 16 has a 16-amino acid (aa) extension at the NTR, which includes a nucleolar localization signal (16). MYO1C 16 , previously referred to as nuclear myosin I (or NMI), localizes mostly to the nucleus (17), and its nuclear localization is moderately affected by calcium (15). In the nucleus, MYO1C 16 is involved in transcription by all three RNA polymerases, mRNA maturation, chromatin remodeling, and chromosome movement (18 -23). MYO1C 16 also localizes to the cytoplasm, where, like MYO1C C , it contributes to the regulation of cell membrane tension (24). MYO1C 16 knock-out mice do not exhibit any obvious phenotype (25). MYO1C 35 has an additional 35-aa extension at the NTR (partially overlapping with MYO1C 16 (Fig. 1B)) and also localizes mostly to the nucleus, where it interacts with RNA polymerase II but not RNA polymerase I (5). In mouse, MYO1C 35 is expressed at low levels in a tissue-specific pat-tern, whereas MYO1C 16 and MYO1C C are ubiquitously expressed at comparable levels in most tissues (26).
Although the biological functions and localization of all three MYO1C isoforms have been thoroughly investigated, only MYO1C C has been characterized enzymatically. MYO1C C is a low-duty-ratio myosin (i.e. it spends most of the ATPase cycle in the weak actin-binding states), despite the fact that its actin attachment lifetime is relatively long (27,28). It exhibits a weak coupling between ADP and actin binding, and ADP release is coupled to an additional lever arm movement, which facilitates additional work subsequent to the power stroke (29,30). The force-sensitive transition in the ATPase cycle is the isomerization that follows ATP binding, and the rate-limiting step has been proposed to be a transition that precedes entry to the strong-binding state (28). Because of these properties, MYO1C C may be viewed as a tension sensor or slow transporter (3,28,29). The kinetics of the additional two isoforms have not been investigated previously, and the impact of alternative splicing on the enzymology of MYO1C isoforms is not fully understood.
A recently solved crystal structure of the closely related protein Myo1b shows that its NTR interacts with calmodulin bound to the first IQ motif (31). Deletion of nine aa of the NTR of MYO1C C , or replacement of this sequence with residues from the N terminus of Myo1b, dramatically changes the kinetics and tension-sensing properties of MYO1C C (32). This region is shared by all three isoforms of MYO1C, but the functions of the NTR extensions remain to be elucidated.
We investigated how the kinetic properties of the MYO1C isoforms are affected by alternative splicing of the NTR. To this end, we expressed and purified the three full-length human MYO1C isoforms in a human cell line and characterized their steady-state ATPase and nucleotide-binding activities by both kinetic and equilibrium measurements. Our detailed kinetic analysis revealed that the diverse NTRs affect isomerization of the nucleotide pocket. Specifically, the NTR stabilizes both the AM O (open state) and the AMD C (ADP closed state preceding isomerization). The open-closed transitions are dependent on the length of the NTR. Next, we computationally analyzed the interaction between the 35-aa NTR with the myosin heavy chain by molecular dynamic modeling and docking to the recently solved crystal structure of MYO1C C -1IQ (33). Finally, we found that the addition of the NTR 35 peptide in trans to the MYO1C C isoform mimicked the ATP-induced acto⅐MYO1C⅐ ADP dissociation kinetics of the MYO1C 35 isoform and stabilized the AMD C state.

N-terminal sequence alignments and purification of fulllength human MYO1C isoforms
MYO1C 35 and MYO1C 16 share 10 aa from exon 1 and an additional 6 or 25 aa from exons Ϫ1 or Ϫ2, respectively (Fig. 1,  A and B). The remaining identical 1028 aa constitute the motor domain, which binds nucleotides and actin; the lever arm, which consists of three IQ motifs for calmodulin light-chain binding; and the PH domain in the tail, which binds phosphoinositides. To better match the cellular environment in which MYO1C functions, we established an expression and purification system for human myosins using a suspension adapted human HEK293SF-3F6 cell line. This system may be of use in future studies aimed at expanding the options for expression of other myosins with more complex architectures. For our experiments, we purified the three human MYO1C isoforms as full-length constructs with calmodulin as their light chain (Fig. 1C) and determined whether they display full calmodulin light-chain binding motif occupancy by preforming actin co-sedimentation followed by SDS-PAGE and densitometry (supplemental Fig. S1) with calmodulin standards to determine light-chain concentration. We found that all MYO1C isoforms bound three molecules of calmodulin (2.9 Ϯ 0.10, 2.75 Ϯ 0.13, and 3.1 Ϯ 0.07 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively).

Actin-activated steady-state ATPase activity of full-length human MYO1C isoforms
Under our reaction conditions, all three isoforms exhibited actin-activated steady-state ATPase activity with hyperbolic dependence on actin concentration, which allowed us to determine the steady-state parameters according to the Michaelis-Menten model (supplemental Fig. S2 and Table S2). Our results revealed that the steady-state kinetic parameters did not indicate whether the different NTRs of these isoforms impact their enzymology. Hence, we reasoned that a detailed kinetic dissection of the ATPase cycle of each isoform might reveal modulation of the rate constants of their ATPase cycle (see Scheme 1).

N-terminal splicing fine-tunes MYO1C kinetics ATP induced an acto⅐MYO1C population of weakly bound states
The increase in pyrene-labeled actin fluorescence and the decrease in light scattering upon ATP binding to acto⅐MYO1C isoforms were monitored as a function of [ATP]. The signal arises from the induced weak-binding states (pyrene-actin, Fig.  2A) or dissociation from actomyosin (light scattering, Fig. 2A). For all three isoforms, both signals were best fitted to doubleexponential equations, suggesting that the isoforms have similar ATP-binding mechanisms ( Fig. 2A). This is explained by actomyosin existing in two states: the fast phase reflects ATP binding to the AM O state, and the slow phase represents the conversion from AM C to AM O (29,30). The equilibrium for this transition is termed KЈ ␣ (where KЈ ␣ ϭ kЈ Ϫ␣ /kЈ ϩ␣ ) as shown in Scheme 2. The observed rate constants of the fast phase (k obs,fast ) were hyperbolically dependent on [ATP] (Fig. 3B), yielding KЈ 1T and kЈ ϩ2T (Scheme 2 and Equation 1).
The observed slow phase in ATP-induced dissociation from actomyosin mostly represents the decrease in the population of the AM C state over time. The [ATP]-dependence of the observed slow-phase rates were fitted to a rectangular hyperbola, yielding k' ϩ␣ ( Fig. 2B)

N-terminal splicing fine-tunes MYO1C kinetics
observed changes in KЈ ␣ in both assays were consistent with an overall shift toward population of the AM O state as the NTR grew longer.

Global numerical curve fitting of ATP-induced population of weakly bound acto⅐MYO1C states
The explicit solutions for the mechanism shown in Scheme 2 cannot extract the four parameters kЈ ϩ1T , kЈ Ϫ1T , kЈ Ϫ2T , and kЈ diss . Therefore, we performed numerical integration (simulation) using KinTek Explorer (35,36) by globally fitting our time-dependent reaction curves. The kinetic parameters extracted by performing simulation on the complete data sets for each isoform are presented in Fig. 3 and Table 2. The signal represents the sum of the AM⅐ATP and A states (weak-binding states). As shown in Table 4, the results of the simulation were in good agreement with the experimentally determined parameters KЈ 1T and kЈ ϩ2T , and the data provided numeric solutions for kЈ ϩ1T , kЈ Ϫ1T , kЈ Ϫ2T , and KЈ 2T . kЈ ϩ1T decreased, whereas kЈ Ϫ1T did not differ between the isoforms (kЈ ϩ1T ϭ 6.6, 5.1, and 4.3 s Ϫ1 and kЈ Ϫ1T ϭ 548, 568, and 531 s Ϫ1 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively). kЈ Ϫ2T of MYO1C 35 was 2-and 8-fold lower than the values for MYO1C 16 and MYO1C C , respectively ( Table 2). KЈ 2T , the equilibrium constant for isomerization after ATP binding, was 2-and 6-fold higher in MYO1C 35 than in MYO1C 16 and MYO1C C , respectively ( Table 2). Both the forward and reverse rate constants for the nucleotide-binding isomerization from the closed-to-open state, kЈ ϩ␣ and kЈ -␣ , were smaller than the experimentally determined constants ( Table  2). The equilibrium rate constant for the closed-to-open isomerization, KЈ ␣ , was similar between the experiment and simulation for MYO1C C . However, for MYO1C 16 and MYO1C 35 , the predicted values were 1.5-and 2-fold larger, respectively, than the experimentally measured values (KЈ ␣ ϭ 3.0, 1.5, and 0.8 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively). Overall, the results were statistically significant in terms of the goodness of the fits per set of each isoform (Fig. 3, A-C, simulated data sets of the ATP-induced population of weakly bound acto⅐MYO1C states fitted to raw data from MYO1C 35 , MYO1C 16 , and MYO1C C . Each graph shows in solid lines the time courses of data collected at 0.03125 (blue), 0.0625 (brown), 0.125 (yellow), 0.25 (purple), 0.5 (green), 1 (light blue), 2 (dark red), 4 (blue), and 8 (orange) mM ATP as presented in Fig. 2A for pyrene-labeled acto⅐MYO1C ATP-induced dissociation. The solid lines through the data sets are the fitted curves resulting from performing global numerical analysis on the entire set of data for each isoform. The fitting was to the sum of the AM⅐ATP and A states according to Scheme 2. D-F, time-dependent distribution of biochemical intermediates of the reaction according to the simulation mechanism shown in Scheme 2 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively. Blue, AM C ; red, AM O ; yellow, AM(ATP); purple, AM⅐ATP; green, A state. The light blue line represents the sum of the AM⅐ATP and A states, which reflect the pyrene signal of weakly bound or dissociated states.

N-terminal splicing fine-tunes MYO1C kinetics
A-C). The confidence of the fitting results is expressed as the ratio of 2 /minimum 2 for the forward versus reverse constants, which assures that the fits reached a global minimum. The signal arising from the summation of the weak-binding states in the simulation allowed determination of the rate of actomyosin dissociation, kЈ diss . Interestingly, we found that the value for MYO1C C (kЈ diss C ϭ 1.0 s Ϫ1 ) was 5-fold smaller than the values for MYO1C 35 and MYO1C 16 (kЈ diss ϭ 5.1 and 5.7 s Ϫ1 , respectively). This analysis yielded additional rate constants that describe ATP binding to acto⅐MYO1C in greater detail and, more importantly, identified specific steps that differed among MYO1C isoforms.

ADP-binding kinetics to acto⅐MYO1C isoforms
The rate of ADP release can be determined by measuring the kinetics of ATP-induced dissociation of actomyosin⅐ADP as a function of [ADP] (37,38). Fig. 4A shows representative time courses of acto⅐MYO1C⅐ADP premixed with 3.75 M ADP (final concentration) upon rapid mixing with 1 mM ATP. The transient time courses were best fitted to a sum of two exponentials for all three isoforms (supplemental Fig. S3). The fast phase is thought to reflect ATP binding to free actomyosin, whereas the slow phase reflects the fraction of ADP dissociation from the acto⅐MYO1C⅐ADP complex (37,39). Although acto⅐MYO1C C and acto⅐MYO1C 16 sustained the fast and the slow phases throughout the entire range of [ADP], acto⅐ MYO1C 35 ⅐ADP prebound to ADP at a concentration of 7.5 M or higher lost the fast-phase component due to loss of the fast amplitude ( Fig. 4, B and D). Similar behavior has been reported for MYO1C C -1IQ and MYO1C C -3IQ (28,30).
For all three isoforms, the dependence of both fast and slow k obs on [ADP] exhibited hyperbolic behavior, suggesting that ADP dissociation occurs via at least two transitions preceding the complete dissociation from actomyosin (Fig. 4, B and C). This is described as actomyosin⅐ADP isomerization from AMD C (ADP closed binding state) to AMD O (ADP open binding state), which has been observed in several other myosins (34,40). However, for MYO1C C , AMD isomerization has not been kinetically identified in previous studies. A reaction mechanism that accounts for these events is presented in Scheme 3.
According to Scheme 3, in the absence of ADP, the signal arises from ATP-induced dissociation of the AM C and AM O states. Preincubation with higher [ADP] increases population of the AMD O and AMD C states, and at saturating ADP, AMD O and AMD C are the predominant states. Thus, the fast and slow k obs values measured upon rapid mixing with 1 mM ATP reflect the summation of all of these states. k obs,fast and k obs,slow are described by Equation 3 (Fig. 4, B and C).
The  16 , and MYO1C C , respectively (Table 3)) are very similar to the rate of isomerization after ATP binding, kЈ ϩ2T (Table 2 and Scheme 3 for 1 mM ATP). At saturating ADP, AMD C and AMD O are the predominant states, and k obs,fast reports the decay of the AMD O state. ADP release was ϳ4-fold faster for MYO1C 16 than for MYO1C C (kЈ min,fast ϭ 8.4 Ϯ 3.90 and 2.2 Ϯ 2.69 s Ϫ1 for MYO1C 16 and MYO1C C , respectively (Table 3)). The fast phase of MYO1C 35 was not observed above 7.5 M ADP due to a loss of the fast amplitude (Fig. 4D). This suggests that, for MYO1C 35 , AMD C is the predominant state and ADP release occurs sequentially. For MYO1C 16 and     16 and MYO1C C , respectively). The y intercept of k obs,slow (I slow ) represents the rate of AM C -to-AM O isomerization (Fig. 5C). Indeed, I slow values (I slow ϭ 0.9 Ϯ 0.02, 1.5 Ϯ 0.02, and 1.9 Ϯ 0.07 s Ϫ1 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively (Table 3)) were very similar to kЈ ϩ␣ (Table 3). At saturating ADP, the slow k obs represents the rate of AMD C decay. kЈ min,slow differed among the isoforms (kЈ min,slow ϭ 0.4 Ϯ 0.03, 0.5 Ϯ 0.02, and 0.7 Ϯ 0.10 s Ϫ1 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively (Table 3)). For MYO1C 35 16 , and MYO1C C , respectively). For all three isoforms, the slow-phase amplitudes remained constant as a function of [ADP] (Fig. 4D) in contrast to k obs,slow , which exhibited hyperbolic dependence on [ADP]. At saturating ADP, both AMD O and AMD C should be fully occupied. Thus, the KЈ 0.5 value for the fraction of the slow amplitude as a function of  [ADP] indicates the overall affinity for ADP. The fraction of the slow phase (Fig. 3E) was fitted according to Equation 4.

Global numerical curve fitting of ATP-induced acto⅐MYO1C⅐ADP dissociation
The minimum mechanism for dissociation of prebound ADP from actomyosin⅐ADP upon ATP binding involves at least five biochemical transitions (Scheme 3). Previously, we analyzed our data only in the defined AM or AMD states in the absence of ADP or under saturating ADP conditions. However, to analyze our data throughout the entire range of [ADP], considering all intermediates, we globally fitted the entire data set according to Scheme 3 (Fig. 5). To constrain the simulation, we used the parameters determined in the ATP-induced dissociation experiments (Table 4). Only four rate constants describing ADP binding and dissociation were determined by the model [kЈ ϩ1D (s Ϫ1 ), kЈ Ϫ1D (M Ϫ1 ⅐s Ϫ1 ), kЈ ϩ2D (s Ϫ1 ), kЈ -2D (s Ϫ1 )]. The fitting iterations to the data sets were allowed to run until they converged to the best possible fitting parameters to reach a global minimum ( 2 /degree of freedom Ͻ 1.1, S.D. () Ͻ 0.85). The good of the fitting results is expressed as the ratio of 2 /minimum 2 for the forward versus reverse constants, which assures that the fits reached a global minimum (supplemental Fig. S5). The rate constant for ADP release from the AMD O state, kЈ ϩ1D , was 1.4-fold larger for MYO1C 35 than MYO1C 16/C ( Table 4). The rate for ADP binding for MYO1C 35 , however, was 3.7-and 2.7-fold smaller than those of MYO1C 16 and MYO1C C ( Table 4). As a result, the ADP-binding affinities of MYO1C 16 and MYO1C C were 5.2-and 3.2-fold stronger than that of MYO1C 35 (Table 4). The rate constant for isomerization of AMD C to AMD O , kЈ ϩ2D , increased gradually as the NTR became shorter ( Table 4). The rates for the closed-to-open nucleotide pocket isomerization were very similar to those for the isomerization of AM C to AM O (Table 3 and Scheme 2). By contrast, the reverse rate for the open-to-closed isomerization decreased as the NTR became shorter (Table 4). Consequently, two of the three isoforms populate the AMD C state to a greater extent than the AMD O state but with different ratios.  Table 4). The overall ADP affinity (KЈ AD ϭ 0.5, 0.2, and 0.7 M for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively) agreed with the measured KЈ AD for MYO1C 35/16 , but was 5-fold smaller than the measured KЈ AD for the measured MYO1C C .
The simulated progression curves allowed us to follow the decay upon addition of ATP of AMD C and AMD O equilibrated with 30 M ADP. AMD C decayed exponentially, with k obs and amplitude similar to the measured k obs,slow for all three isoforms (Figs. 4 and 5 and Tables 3 and 4). However, AMD O exhibited double-exponential decay with a fast phase that was not completely consistent with the measured fast phase, even though the fitted curve yielded a fast phase identical to the measured data (Figs. 4 and 5 and Tables 3 and 4). This suggests that the measured slow phase reflects AMD C decay, whereas the measured fast phase reflects a mixture of several states in the system.

Structural homology modeling and molecular dynamic docking of the NTR 35 domain with MYO1C C
To gain further insights into how the NTRs impact the structural properties of MYO1C isoforms, we applied structurebased molecular dynamics and docking routines. First, we used PSIPRED version 3.3 to perform a secondary structure prediction of the NTR of MYO1C 35 , which indicated that this region consists of four ␤-strands interrupted by three coils (Fig. 6A). A FASTA search against structured proteins revealed that the NTR of MYO1C 35 shares 18.4% sequence identity (44.7% sequence similarity) with desulforedoxin in (PDB ID: 1DHG) (Fig. 6B). Remarkably, desulforedoxin adopts a compact ␤-barrel fold comprising four ␤-strands, which is similar to the PSIPRED prediction. Therefore, the homology model of MYO1C 35 -NTR was based on desulforedoxin (PDB ID: 1DHG). In that model, a HPH motif (similar to the WPH motif of Myo1B (31)) is exposed at the tip of the ␤-barrel. Next, we performed docking experiments of the predicted folded motif of MYO1C 35 -NTR. For this procedure, we used Maestro in the Schrödinger Software Suite to generate different states of the predicted homology model by molecular dynamics simulations (41,42). These different conformations of the NTR model were used as ligands for protein-protein docking studies using the crystal structure of MYO1C C (0, 6.4, 8.6, 15.5, and 25 ns after energy minimization) (PDB ID: 4BYF). For this purpose, we used the ClusPro online docking tool for protein-protein docking studies (43). Different constraints were added for the docking procedures. For example, the HPH motif (H18-P19-H20) of the NTR model was involved in the interaction, and the C-terminal Phe-57 of the NTR model exhibited a repulsive interaction. Most of the resulting docking models were out of range of the experimentally solved N terminus. The best hit within a reachable range is shown in Fig. 6D. In this model, the MYO1C 35 NTR interacts with the experimentally solved crystal structure. Interestingly, according to this model, the NTR of the MYO1C 35 interacts with amino acid region 619 -636, which was identified by Schwab et al. (16) as corresponding to one of the two nucleolar localization signals of MYO1C 16 . The second

N-terminal splicing fine-tunes MYO1C kinetics
nucleolar localization signal of MYO1C 16 is located on the NTR itself, suggesting that the two are connected. Besides numerous electrostatic and hydrophobic contacts between the N termi-nus of the crystal structure and the nucleolar localization signal, one specific interaction stands out: residue Arg-21 engages in polar contact with Glu-469 of the loop directly after the relay helix, which could have a mechanical impact on the myosin power stroke. Consistent with this model, our kinetic studies revealed that MYO1C 35 has a Ͼ2-fold faster AMD O -to-AMD C isomerization and a 3-fold slower ADP-binding rate constant than MYO1C 16 and MYO1C C .

Determination of NTR 35 , NTR 35 -R21G, and NTR 16 secondary structure by circular dichroism
To test our structural homology modeling, we synthesized three peptides corresponding to the 35-aa NTR (NTR 35 ), NTR 35 -R21G (a mutant peptide based on the results of molecular dynamic modeling), and 16-aa NTR (NTR 16 ). We then determined their secondary structures by performing circular dichroism (CD) measurements and deconvolution of their  Table 4.

N-terminal splicing fine-tunes MYO1C kinetics
spectra ( Fig. 7 and supplemental Tables S4 and S5). NTR 35 exhibits a strong negative peak at 208 nm. Deconvolution of its spectrum yielded a prediction of 38% anti-parallel ␤-sheet, 47% unstructured, and 15% turns at 20°C (Fig. 7 and supplemental Table S5). This is highly similar to the predicted model based on the structural homology presented in Fig. 6C. Point mutation of R21G within this peptide to generate NTR 35 -R21G shifted the negative peak to 209 nm (Fig.  7). We also determined the secondary structure of the shorter NTR 16 , which shares 10 aa with NTR 35 (Fig. 7). According to the predicted secondary model, ϳ43% of the identical amino acids fall within the predicted anti-parallel ␤-sheet (Fig. 6C).

NTR 35 peptide added in trans to MYO1C C induces MYO1C 35 -like kinetic behavior
The CD measurements of the various NTRs showed that these peptides form independently folded domains. Hence, they may invoke similar effects whether they are present in trans or covalently attached to the polypeptide chain. We studied the effect of the NTR 35 and NTR 35 -R21G peptides on ATPinduced dissociation of acto⅐MYO1C C ⅐ADP and compared this effect among the three isoforms (Fig. 8). These experiments were performed in the presence of 20 M folded peptide, high enough to saturate binding to MYO1C C (both 50 and 100 M folded peptides yielded similar results). Interestingly, we observed the same kinetic behavior in k obs,fast and k obs,slow of ATP-induced dissociation of acto⅐MYO1C 35 ⅐ADP as with acto⅐MYO1C C ⅐ADP preincubated with NTR 35 . This remarkable finding demonstrates that NTR 35 impacts the nucleotidedependent transition in trans in the same way as when it is present on a continuous polypeptide chain. NTR 35 -R21G did not affect MYO1C C kinetic behavior to the same extent, confirming the predicted interaction of Arg-21 with the rest of the myosin heavy chain. Finally, the addition of the NTR peptides in trans influenced the fraction of the slow amplitude (Fig. 8B). This shifted the population of AMD C and AMD O toward the closed states, as predicted by our model.

Discussion
We performed comparative studies of MYO1C splice isoforms in the context of their full-length proteins with the goal of achieving accurate allosteric awareness as proposed by Preller and Manstein (44). The steady-state parameters did not show dramatic changes in overall ATPase behavior among the isoforms. However, a detailed kinetic analysis revealed intrinsic divergence among the isoforms that either balanced out or had a low impact on k cat and K ATPase .
We compared our results with those obtained to date with truncated MYO1C C constructs. We determined the equilibrium constant of MYO1C nucleotide-binding pocket isomerization using both pyrene-actin and light-scattering approaches. Previous studies on the MYO1C C -1IQ/3IQ motor domain with variable lever arm constructs showed that the MYO1C C isoform populates mostly the AM C state   (27,28,29,30). Our results supported these findings but also demonstrated that the equilibrium of full-length MYO1C C is shifted from AM C toward AM O by 7-fold relative to the 1IQ construct and 2-fold relative to the 3IQ construct.

ADP-binding kinetics to acto⅐MYO1C isoforms reveal conservation of ADP isomerization states
ADP-binding kinetics contribute to the dwelling time of myosin in the strong-binding state and hence its effect on the duty ratio (40,45). Both EM and mechanical measurements show that the ADP release mechanism of MYO1C C is biphasic and depends upon additional movement of the lever arm and load (29,30). However, kinetic measurements of ATP binding to prebound actomyosin⅐ADP using the 1IQ and 3IQ MYO1C C constructs could not detect intrinsic isomerization prior to ADP release due to the loss of the fast-phase k obs (27)(28)(29)(30). Here, we show for the first time that full-length MYO1C undergo closed-to-open isomerization before ADP is released. This isomerization was also detected using truncated constructs of the closely related Myo1b (32). Therefore, we suggest that the addition of the tail domain shifts the equilibrium toward the AMD O state.

NTR effect on actomyosin nucleotide pocket isomerization
The NTR extensions altered the closed-to-open isomerization of the nucleotide pocket by stabilizing the AM O and AMD C states. As a result, the differences in behavior imply that each isoform has different kinetics depending on the ATP/ADP ratio. Related to this finding, the simulated models (Figs. 3 and  5) demonstrated that the larger the population of the AM O and AMD C states, the larger the population of the AM(ATP) state and the faster it forms, at a given [ATP]. The simulation also revealed differences in kЈ Ϫ2T and kЈ diss . MYO1C C populated the AM⅐ATP (weak-binding state) longer than MYO1C 35 , suggesting that the NTR extensions destabilize the weak-binding state and could be responsible for differences in the tension-sensing features between the two isoforms (32).

Communication between nucleotide-and actin-binding sites
The communication between the nucleotide-and actinbinding sites can be described by a closed thermodynamic square, in which ADP (D) and actin (A) binding to myosin (M) are linked by four equilibrium constants as shown in Scheme 4.
Myosins that generate rapid sliding velocities (e.g. muscle myosins) have large thermodynamic coupling constants (Ͼ10) and hence strong coupling between actin and ADP binding. On the other hand, myosins that function as gated/processive or tension sensors (e.g. myosins V, VI, and VII) (3,40,45,46) have small thermodynamic coupling constants (Ͻ5). Table 3 and supplemental Fig. S4 and Table S3 show KЈ A , KЈ DA , and KЈ AD for each of the isoforms (see supplemental information text, Fig.  S4, Scheme S1, and Equation S1 for a description of the equilibrium-binding experiments). The K D values of each isoform according to Scheme 4 are KЈ DA /KЈ A ϭ 1.40, 0.99, and 1.39 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively, and the affinities for ADP in the absence of actin are K D ϭ 0.3, 0.18, and 0.28 M, respectively. The results support weak thermodynamic coupling between ADP and actin binding, consistent with a role for MYO1C isoforms as tension sensors or slow transporters in ensembles.

Consideration of the structural impact of the NTR in light of current models
Greenberg et al. (32) studied how MYO1C C -3IQ NTR impacts load dependence and kinetics, either by deleting the first nine residues of MYO1C C or replacing them with the first 13 residues of Myo1b. This region is shared by all three MYO1C spliced isoforms. Greenberg et al. (32) performed comprehensive biochemical and mechanical (under load) studies to investigate how such structural changes affect the motor properties of MYO1C C ; specifically, the results of the unloaded kinetics revealed that isomerization after nucleotide binding and AM Cto-AM O isomerization are strongly affected by the identity of the NTR, which can alter kЈ ϩ2 , kЈ ϩ␣ , and kЈ Ϫ␣ quite significantly. They found that the addition or deletion of these structural

N-terminal splicing fine-tunes MYO1C kinetics
elements affects the active site isomerization by increasing its flexibility. Finally they proposed that the NTR plays an important role in stabilizing the post-power-stroke conformation (32). Our results show that the extended NTRs affect nucleotide pocket isomerization by decreasing both kЈ ϩ␣ and kЈ Ϫ␣ without affecting kЈ ϩ2 . Moreover, unlike MYO1C C , which tended to populate the AM C state in the ATP-induced dissociation experiment, MYO1C 35 populated mostly the AM O state, whereas MYO1C 16 populated both states equally (Fig. 2). Together, our results indicate that the lengthening of the NTR increases the rigidity of the nucleotide pocket and stabilizes the AM O state. In the prebound ADP measurements, all three isoforms tended to populate the AMD C state but to varying extents (MYO1C 35 Ͼ MYO1C 16 Ͼ MYO1C C ) (Fig. 4). This suggests that the NTR extensions stabilize the post-power-stroke state, similar to what was observed previously in NTR mutants as well as Myo1b (32). Our results indicate some degree of correlation between ATP-and ADP-binding kinetics in all three isoforms. It may be that these transitions are linked in terms of structural reorganization, i.e. the open-to-closed isomerization of the nucleotide-binding pocket. In addition, consistent with the findings of Greenberg et al. (32), our results support the idea that the nucleotide-binding pocket is affected by the NTR region. We propose that the extended NTRs of the isoforms form a structural domain (Fig. 7) that affects pocket rigidity and stabilizes the AM O and AMD C states.

Higher level of fine regulation by MYO1C NTRs
Our results show changes in the kinetic parameters that may yield additional specific kinetic adaptations for each of the three isoforms. Several studies have suggested that in addition to their distinct functions, some overlap could occur in the event that one isoform is lost. Knock-out mice lacking the NM1 (MYO1C 16 ) start codon (without affecting MYO1C C or MYO1C 35 ) exhibit interchangeability and redundancy of myosin isoforms in the cell nucleus, suggesting that both isoforms can substitute for each other in nuclear processes (25). Partial rescue and functional overlap between closely related MYO1C isoforms are likely to minimize the observed cellular and whole-animal knockdown phenotypes (47). MYO1C 16 , although displaying specific nuclear functions, localizes to the plasma membrane. Furthermore, knock-out of MYO1C 16 has strong effects on the elasticity of the plasma membrane around the actin cytoskeleton, as determined by atomic force microscopy (24). Overall, MYO1C isoforms possess overall nearly identical structural domains and most likely are subject to similar post-translational modifications and binding to similar partners. Thus, different mixtures or ensembles of MYO1C isoforms could serve to fine-tune a specific biological function. Finally, to distinguish between the ensemble effects and redundancy of isoforms, all three knockouts should be compared individually.

Reagents
All chemicals and reagents were of the highest purity commercially available. ATP was purchased from Roche Applied Science, and ADP was purchased from Bio Basic (Markham, Ontario, Canada). Nucleotide concentrations were determined by measuring absorbance at 259 nm using ⑀ 259 ϭ 15,400 M Ϫ1 cm Ϫ1 . In all experiments, 1 molar equivalent of MgCl 2 was added to nucleotide solutions immediately before use. N-(1-Pyrene)iodoacetamide (Molecular Probes, Eugene, OR), MOPS, EGTA, apyrase (potato grade VII), and phalloidin were purchased from Sigma-Aldrich. MgCl 2 ⅐6H 2 O came from Bio Basic and KCl from Merck (Darmstadt, Germany).

Cell culture
All media reagents were purchased from Sigma. Fetal calf serum, L-glutamine, HEPES-KOH, pH 7.4, penicillin, streptomycin, and amphotericin B were purchased from Biological Industries (Beit Haemek, Israel).

Cloning of full-length human MYO1C isoforms
Full-length human MYO1C isoforms (residues 1-1063, 1044m, and 1028 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively) were cloned into the HaloTag-pFC14K Flexi vector (Promega). Human cDNA (HsCD00365758 clone ID), purchased from the ORFeome Collaboration, was used as the template for cloning human isoforms by primer extension PCR using the primers listed in supplemental Table S1. All constructs were fully sequenced and compared with the published sequences of human MYO1C isoform (NCBI RefSeq NM_ 001080779.1, NM_001080950.1, and NM_033375.4 for MYO1C 35 , MYO1C 16 , and MYO1C C , respectively). Human calmodulin was cloned into pF4A.

N-terminal splicing fine-tunes MYO1C kinetics Expression and purification of other proteins
Actin was purified from rabbit or chicken skeletal muscle (labeled with pyrene if needed) and gel-filtered over Sephacryl S-300 HR (49). Ca 2ϩ -actin monomers were converted to Mg 2ϩactin monomers by the addition of 0.2 mM EGTA and 40 M MgCl 2 (excess over [actin]) immediately prior to polymerization by dialysis against KMg50 buffer (20 mM MOPS, 50 mM KAc, 2 mM MgCl 2 , 0.2 mM EGTA, and 1 mM DTT, pH 7, at 25°C). The final dialysis was performed against KMg25 buffer (20 mM MOPS, 25 mM KAc, 2 mM MgCl 2 , 0.2 mM EGTA, and 1 mM DTT, pH 7 at 25°C). Phalloidin (1:1 molar ratio) was used to stabilize actin filaments. Actin was purified from rabbit skeletal muscle, labeled with pyrene, and gel-filtered over Sephacryl S-300 HR 28 (49) Ca 2ϩ -actin monomers were converted to Mg 2ϩ -actin monomers with 0.2 mM EGTA and 50 M MgCl 2 (excess over [actin]) immediately prior to polymerization by dialysis against KMg50 and followed by KMg25 buffer. Phalloidin (1.1 molar equivalent) was used to stabilize the actin filaments. Calmodulin was expressed in bacteria and purified as described (50).

Determination of light-chain calmodulin occupancy to the myosin isoforms
Light-chain calmodulin occupancy of myosin isoforms was performed by actin co-sedimentation of the MYO1C constructs as described elsewhere (51) with minor changes. Briefly, 60 nM MYO1C isoforms was incubated with 1 M actin in KMg25 buffer for 30 min followed by ultracentrifugation in a TLA55 rotor (Beckman) at 186,000 ϫ g for 25 min at 4°C. The pellets were then washed gently with the KMg25 buffer and resuspended in protein sample buffer in the presence of 3 mM EGTA, heated for 5 min at 95°C, resolved by gradient SDS-PAGE (10 -20%), and visualized by staining with InstantBlue TM (Expedeon, San Diego, CA). Calmodulin band intensities were quantified with ImageLab software using known quantities of calmodulin resolved on the same gel as standards.

Steady-state ATPase activity
The actin-activated steady-state ATPase activity of MYO1C was measured at 20 Ϯ 0.1°C in KMg25 buffer supplemented with 2 mM MgATP by monitoring changes in absorption at 340 nm for 10 min at 1-s intervals on a UV-spectrometer (37). The concentration of all myosins was 100 nM. Myosin concentrations were determined as described above and verified by gel densitometry.

Equilibrium fluorescence binding of pyrene-actin to myosin and actomyosin
Fluorescence measurements were performed on a PC1 spectrofluorimeter (ISS Inc., Urbana-Champaign, IL) set up in an L-format configuration using an emission channel monochromator. Samples were equilibrated (60 min at room temperature), measured with ex ϭ 365 nm, and monitored with emission monochromators scanning from 390 to 430 nm with a peak at em ϭ 409 nm.

Stopped-flow measurements
All experiments were performed in KMg25 buffer on a Hi-Tech Scientific SF-61DX2 stopped-flow apparatus (TgK Scien- where F(t) is the signal at time t, F ∞ is the final signal value, A i is the amplitude, k i is the observed rate constant characterizing the i-th relaxation process, and n is the total number of observed relaxations. The value of n was either 1 (single exponential) or 2 (double exponential). Fitting was limited to data beyond 1 ms to account for the instrument dead time and to exclude data acquired during the continuous-flow phase of mixing, as recommended by the manufacturer.
Uncertainties are reported as standard errors in the fits, unless stated otherwise, and were propagated using the general formula shown in Equation  where the experimental measurements x 1 , x 2 …x n have uncertainties dx 1 , dx 2 …dx n and a is a function of x 1 , x 2 …x n . The Levenberg-Marquardt algorithm was used to solve the minimization of nonlinear least squares curve fitting.

Nucleotide-binding kinetics
The time courses of nucleotide binding were acquired under pseudo first-order conditions with [nucleotide] Ͼ Ͼ [myosin or actomyosin]. Actomyosin samples were prepared by mixing equimolar amounts of MYO1C and actin filaments or, where specified, with [actin] Ͼ Ͼ [myosin]. Actomyosin samples were treated with apyrase (0.01 unit/ml), used to deplete ATP and ADP from actomyosin when relevant, and equilibrated on ice for 10 min before measurements were made. After mixing, the final concentration of apyrase was 0.005 unit/ml (37).

Circular dichroism measurements
Far-UV CD was observed with an Applied Photophysics PiStar CD spectrometer (Surrey, UK) equilibrated with nitro-N-terminal splicing fine-tunes MYO1C kinetics gen gas, with the temperature regulated by a thermostat at 20 Ϯ 0.1°C. Changes in ellipticity were followed from 280 to 190 nm in 20 mM MOPS, 25 mM KCl in a 1-mm path-length fused quartz cell with a step size of 0.5 nm and bandwidth of 10 nm. Typically, three scans were averaged prior to analysis. Peptide concentrations were 0.2 mg/ml, corresponding to 54, 56.1, and 115 M NTR 35 , NTR 35 -R21G, and NTR 16 , respectively (supplemental Table S4). The mean residue ellipticity ([], in mdeg⅐cm 2 ⅐dmol Ϫ1 ⅐ residue Ϫ1 ) was derived from the raw data (, in millidegrees (mdeg)) using the following formula: [] ϭ ϫ 100/(l ϫ n ϫ c), where l is the path length of the cuvette, c is the molar concentration of the peptides, and n is the number of residues in the peptides (i.e. 35 or 16 residues). The CD spectrum was deconvoluted using the BestSel server (http://bestsel.elte.hu/) 3 (53). The normalized root-mean-square deviation represents the goodness of the deconvolution ( NRMSD, supplemental Table S5).

Kinetic simulations and modeling
We used KinTek Explorer (35,36) to simulate the complex reaction shown in Schemes 3 and 4, utilizing numerical integration and global fitting of a family of data sets to a single model to extract mechanistic information directly from kinetic data. First, the transients were fitted to an analytic function to derive their standard deviation (), which was then incorporated into the statistical analysis of the goodness of the fits. The fitting parameters were then loaded into the Ode45 function in Matlab to produce the model and fitting.

Molecular dynamics simulation
For molecular dynamics simulation and energy minimization processes, we used Desmond in the Maestro 11 work suite (41,42). The OPLS_2005 force field was used in a minimized 10 Å orthorhombic water box with 0.05 M sodium chloride ions. The simulation was performed with the TIP3P water model, including a 0.03-ns quick relaxation step. The simulation time was 25 ns at 300 K and 1 atmosphere.

Protein docking
Protein-protein docking was performed using the ClusPro online docking tool (43). The MYO1C C crystal structure (PDB ID: 4BYF) was used as the receptor molecule. The ligand molecule was a homology model of an NTR in minimized conformation with four additional different plateau states of the molecular dynamics simulation (6.4, 8.6, 15.5, and 25 ns). Different parameters of attraction and repulsion were used, including determined attraction for His-18 -Pro-19 -His-20 of the NTR and repulsion for Val-31.