Current Biology 19, 442���447, March 10, 2009 ��2009 Elsevier Ltd All rights reserved DOI 10.1016/j.cub.2009.01.058 Report The Processivity of Kinesin-2 Motors Suggests Diminished Front-Head Gating Gayatri Muthukrishnan,1 Yangrong Zhang,1,2 Shankar Shastry,1 and William O. Hancock1,* 1Department of Bioengineering The Pennsylvania State University 205 Hallowell Building University Park, PA 16802 USA Summary Kinesin-2 motors, which are involved in intraflagellar trans- port and cargo transport along cytoplasmic microtubules, differ frommotors inthe canonical kinesin-1 family byhaving a heterodimeric rather than homodimeric structure and possessing a three amino acid insertion in their neck linker domain. To determine how these structural features alter the chemomechanical coupling in kinesin-2, we used single-moleculebeadexperimentstomeasuretheprocessiv- ity and velocity of mouse kinesin-2 heterodimer (KIF3A/B) and the engineered homodimers KIF3A/A and KIF3B/B and compared their behavior to Drosophila kinesin-1 heavy chain (KHC). Single-motor run lengths of kinesin-2 were 4-fold shorter than those of kinesin-1. Extending the kinesin-1 neck linker by three amino acids led to a similar reduction in processivity. Furthermore, kinesin-2 processivity varied inversely with ATP concentration. Stochastic simulations of the kinesin-1 and kinesin-2 hydrolysis cycles suggest that ������front-head gating,������ in which rearward tension prevents ATP binding to the front head when both heads are bound to the microtubule, is diminished in kinesin-2. Because the mechanical tension that underlies front-head gating must be transmitted through the neck linker domains, we propose that the diminished coordination in kinesin-2 is a result of its longer and, hence, more compliant neck linker element. Results and Discussion Kinesin processivity relies on maintaining the hydrolysis cycles of the two heads out of phase such that one head remains bound to the microtubule at all times. Figure 1A shows our working model for the kinesin-1 hydrolysis cycle that accounts for a large body of kinesin mechanical and biochem- ical experiments [1]. Features of the hand-over-hand model that ensure processivity can be described by two nonexclu- sive mechanisms: front-head gating and rear-head gating (for consistency with the literature, we use the term ������gating������ but emphasize that this refers to gating of a given head and not gating by a given head) [2���5]. Both of these mechanisms involve mechanical tension between the two heads that is transmitted through the flexible neck linker of each head and their shared neck coiled-coil domain. In front-head gating, rearward strain on the leading head in state 1 prevents ATP from binding until the trailing head detaches (state 2). This mechanism prevents premature ATP hydrolysis and subse- quent detachment of the leading head and ensures that the trailing head is primed to advance to the next binding site when ATP binds to the leading head. In the rear-head gating model, the bound head in state 4 dissociates slowly in compar- ison to the overall cycle time, and binding of the second head (state 1) produces forward strain that leads to rapid detach- ment (state 2). Becausebothfront-headgatingandrear-headgatinginvolve mechanical tension between the head domains, modifications that increase the mechanical compliance of the flexible neck linker are predicted to reduce motor processivity. Based on sequence alignments and comparisons of crystal structures, the neck linker domain in kinesin-2 motors is three amino acids longer than the neck linker domain of kinesin-1 (17 versus 14 amino acids) (Figure 1B). If this extension increases the compliance of the neck linker, then either front-head gating, rear-head gating, or both mechanisms may be diminished in kinesin-2. This prediction was tested by measuring the proces- sivity of kinesin-1 and kinesin-2 motors and interpreting the results with a stochastic model of the hydrolysis cycle. Kinesin-2 Is Less Processive Than Kinesin-1 The velocities and run lengths of individual KIF3 and KHC motors attached to polystyrene beads were analyzed first at saturating ATP levels. Motor dilution profiles were generated to ensure that experiments were carried out in the single- molecule regime (see Supplemental Data available online). The mean velocity of the KIF3A/B heterodimer was 436 6 129 nm/s (mean 6 SD, n = 90), and the velocity of KHC motors was 703 6 136 nm/s (n = 58). The KIF3 homodimers KIF3A/A (455 6 115 nm/s, n = 101) and KIF3B/B (458 6 106 nm/s, n = 102) had similar velocities to the wild-type heterodimer. Whereas the KIF3B/B velocity was consistent with previous measurements from gliding assays, the KIF3A/B and KIF3A/A velocities were significantly higher than previously reported [6]. The velocity differences are due to mutations that were discovered in the motors used in the previous work (see Supplemental Data) all of the velocities presented here are from the corrected sequences. In contrast to velocity differences, which were within a factor of two, the run length of wild-type kinesin-2 was w4-fold shorter than that of kinesin-1. The mean run length for KIF3A/B was 449 6 30 nm (mean 6 SEM, n = 88) compared to 1747 6 199 nm (n = 57) for KHC (Figure 2). The reduced processivity of kinesin-2 is consistent with the hypothesis that the longer neck linker domain in kinesin-2 reduces the degree of mechan- ical communication between the two heads and ������uncouples������ the hydrolysis cycles of the two heads. To test this hypothesis more directly, we extended the 14 amino acid neck linker of kinesin-1 by inserting the last three amino acids of the kine- sin-2 neck linker (DAL) into kinesin-1 at the neck linker/neck coil junction. The run length of this kinesin-1+DAL construct (355 6 14 nm) was 5-fold shorter than that of wild-type (Figure 2), consistent with disruption of the mechanochemical coupling betweenthe twoheads.Despite the significant reduc- tion in processivity, kinesin-1+DAL moved at 552 6 103 nm/s, *Correspondence: wohbio@engr.psu.edu 2Present address: Department of Biomedical Engineering, Northwestern University, Evanston, IL 60208, USA

only 20% slower than wild-type kinesin-1. The behavior of kinesin-1+DAL is qualitatively consistent with work from Hackney, who found that the biochemical processivity of Drosophila kinesin-1 was reduced when the neck linker was ar- tificially extended [7]. However, it contrasts with recent results from Yildez et al., who extended the neck linker of Cys-lite human kinesin-1 and found that the processivity fell by less than a factor of two for inserts as large as 29 amino acids [8]. Possible explanations for why the Yildiz results differ from ours include: (1) their use of a cys-lite modified kinesin, which has been shown to have altered motor kinetics [4], (2) the use of axonemes rather than taxol-stabilized microtubules, (3) the inclusion of two positively charged lysines in the inserts, and (4) the low ionic strength buffer (12 mM PIPES versus 80 mM PIPES used here), which will enhance electrostatic tethering of the motors to the axonemes. The measured run lengths of KIF3A/A and KIF3B/B were in the range of KIF3A/B, but the KIF3A/A run length (410 6 35 nm, n = 85) was moderately shorter than KIF3A/B, whereas the KIF3B/B run length (704 6 81, n = 83) was somewhat longer Figure 1. Kinesin Model and Structural Comparison (A) Working model for the kinesin-1 chemomechanical pathway. T, ATP D, ADP DP, ADP.Pi f, no nucleotide. Motors in solution have high affinity for ADP and, upon binding to the microtubule (state 2), release one ADP molecule [20]. The motor waits in this state until ATP binds to the front head, which results in the docking of the neck linker and a displacement of the tethered head toward the next binding site (state 3) [21]. While the tethered head diffusively searches for the next binding site, the ATP on the bound head is hydrolyzed (state 4). Following hydrolysis (state 4), there are two possibilities. Most of the time (w99% for kinesin-1), the tethered head will bind to the next biding site and release its ADP (state 1), and then the rear head will detach (state 2), completing an 8 nm step. Alternatively, the bound ADP.Pi head in state 4 will unbind from the microtubule, terminating the processive run. Whereas detachment from state 4 is the predominant termination step, at limiting ATP concentrations, the motor can occasionally detach from state 2. (B) Neck linker sequences for kinesin-1 and kinesin-2 motors. The sequences of human conventional kinesin heavy chain (HsKHC), Drosophila melanogaster kinesin (DmKHC), and mouse KIF3A and KIF3B were aligned based on the known crystal structures of KHC (PDB: 3KIN, 2KIN, and 1MKJ) and KIF3 (3B6U). The kinesin-1 neck linker is defined as the 14 amino acids that span between the end of a6 and the first hydrophobic residue of the a7 coiled coil (A323 to T336 in human numbering). The start of the a7 coiled coil was taken from the rat kinesin-1 dimer structure (3KIN) and the human KIF3B (3B6U) and Giardia KIF3A (GiKIN2a) monomer structures [22]. Hydrophobic a and d residues in the heptad repeat are underlined. (C) Crystal structure of rat monomeric conventional kinesin from [23], showing the start and end of the 14 residue neck linker in kinesin-1. (D) Comparison of the length and conformation of the KHC and KIF3B neck linkers. The figure was created by aligning the a6 helix of the 2KIN and 3B6U structures. Figure 2. Run Length Distributions for Single Kinesin-1, Kinesin-2, and Kinesin-1+DAL Motors Attached to Beads (A���C) Run lengths were estimated by fitting the data to an exponential in which the first bin (0���0.5 mm for kinesin-1 and 0���0.25 mm for kinesin-2 and kinesin- 1+DAL) was ignored due to uncertainties in detecting events below 250 nm. Error bars represent SE of fits. Kinesin-2 Processivity 443