FP fusions to KHC and KLC
To analyze the structure of Kinesin-1 in living cells by FRET, donor (monomeric ECFP [mECFP]) and acceptor (monomeric Citrine [mCit]) FPs were fused to the N and/or C termini of both KHC and KLC (). COS cells were chosen for their flat morphology and because their low levels of endogenous Kinesin-1 are unlikely to interfere with formation of donor–acceptor FP complexes (Fig. S1 A, available at http://www.jcb.org/cgi/content/full/jcb.200605097/DC1
; unpublished data). Only cells expressing low-to-medium levels of FP proteins were chosen for data analysis to avoid artifacts caused by protein aggregation and ATP-independent microtubule interactions (Fig. S1 B). FRET stoichiometry (which is discussed in the following two paragraphs) and coimmunoprecipitation (Fig. S1 A) experiments verified that FP fusions to KHC and KLC did not alter their interactions.
Figure 1. Localization and activity of FP-tagged Kinesin-1 in COS cells. (A, left) Schematic diagram of KHC (red) and KLC (orange) domain structure and positions of the epitope and FP tags. (right) Current model of Kinesin-1 structural organization. Red ovals, (more ...)
FP-KHC and -KLC expressed in COS cells demonstrated similar localization patterns to those described previously for other tagged Kinesin-1 motors (Fig. S1 B). Because steady-state fluorescence patterns do not indicate the activity of kinesin motors, we developed an assay to delineate between active and inactive motors in vivo. To do this, we took advantage of the ability of the nonhydrolyzable ATP analogue AMPPNP to block the release of active kinesin motors from microtubules (Kawaguchi and Ishiwata, 2001
). Live cells were transiently permeabilized with low levels of the bacterial toxin streptolysin O (SLO), and active FP-Kinesin-1 motors were trapped on microtubules by the addition of AMPPNP. FP-Kinesin-1 (e.g., mCit-KHC + HA-KLC; , column 1) did not become trapped on microtubules, but remained diffuse and cytosolic upon addition of AMPPNP, indicating that the Kinesin-1 holoenzyme is in an inactive state in vivo. In contrast, in cells expressing FP-KHC alone (e.g., mCit-KHC; , column 2), the motor rapidly accumulated on microtubules after exposure to AMPPNP, indicating that the KHC subunit exists in an active state in vivo.
Several lines of evidence verify that FP-KHC alone is capable of ATP-dependent microtubule motility, and thus represents the Kinesin-1 active state. First, single molecule motility assays demonstrate that FP-KHC molecules are capable of microtubule-based motility in vitro (Fig. S1 C). Second, removal of the cryptic ATP-independent microtubule-binding site in the KHC tail (KHC[1–891]) resulted in a KHC molecule that retained ATP-dependent microtubule binding (, mCit-KHC[1–891]). Third, this microtubule localization was caused by direct interaction between the KHC motor domain and the microtubules because FRET between mCit-KHC(1–891) and mECFP-tubulin increased after addition of AMPPNP (Fig. S4). Fourth, mutation of the microtubule-binding site in the KHC motor domain (Δloop12 mutation; Woehlke et al., 1997
) abolished the ability of FP-KHC to be locked in a microtubule- bound state after addition of AMPPNP (, mCit-KHC[1–891]/Δloop12). Collectively, these results indicate that KHC homodimers are active for microtubule binding and motility, whereas the complete Kinesin-1 holoenzyme (KHC + KLC) remains inactive and predominantly in the cytosol. In addition, these results validate the use of fluorophore-tagged subunits to study Kinesin-1 structure and function in vivo.
FRET stoichiometry reveals conformational changes in Kinesin-1 in live cells
For FRET stoichiometry of Kinesin-1, various combinations of KHC and KLC FRET pairs were cotransfected into COS cells, and 24 h later the data were collected on a wide-field fluorescence microscope calibrated for FRET stoichiometry. FRET stoichiometry uses three fluorescence images from a calibrated microscope to calculate three parameters that describe each pixel (Fig. S2, available at http://www.jcb.org/cgi/content/full/jcb.200605097/DC1
; Hoppe et al., 2002
; Beemiller et al., 2006
): (a) RM
, the mole ratio of acceptor- to donor-labeled proteins, (b) EA
, the apparent acceptor FRET efficiency (FRET efficiency × fraction of acceptor molecules in complex), and (c) ED
, the apparent donor FRET efficiency (FRET efficiency × fraction of donor molecules in complex). EA
range between 0 and 100%, where 100% indicates all acceptor and donor molecules in the FRET complex and with complete energy transfer. Because protein expression levels influence the fraction of donor or acceptor molecules in FRET complex for nonlinked molecules, we analyzed cells with RM
close to 1.0 and we calculated an average FRET efficiency, EAVE
)/2, which is less sensitive to expression ratio (Beemiller et al., 2006
). For the control calibration molecule, mECFP-16aa-mCit, EAVE
≈ 37% ().
Figure 2. FRET monitors conformational changes in Kinesin-1 in live cells. (A) Schematic diagram of the linked mCit-16aa-mECFP calibration molecule, as well as FP-KHC and -KLC constructs. Yellow rectangles, mCit; cyan rectangles, mECFP. (B and D) FRET stoichiometry (more ...)
Because the subunits of Kinesin-1 interact with very high affinity and are of known stoichiometry, changes in EAVE
should reflect structural changes in the Kinesin-1 molecule. Modeling the spatial arrangements between the FP and the KHC motor domain based on crystal structures supports this assumption, as the short linker sequences (4 or 5 aa) limit the flexibility of the FP (Fig. S1 D). To verify that FRET stoichiometry can detect conformational changes in Kinesin-1 in live cells, we obtained FRET efficiencies under ion concentrations known to induce Kinesin-1 conformational changes in vitro (Hackney et al., 1992
). To monitor Kinesin-1 motor-to-tail FRET, FRET pairs were placed on the N and C termini of the same KHC polypeptide (mCit-KHC-mECFP; ). Coexpression with Myc-KHC was required to prevent aggregation of the four FPs in the KHC homodimer (Fig. S3, available at http://www.jcb.org/cgi/content/full/jcb.200605097/DC1
). COS cells expressing mCit-KHC-mECFP + Myc-KHC + HA-KLC () were transiently permeabilized with SLO under physiological salt conditions (I ≈ 0.15). After 5 min, the cells were exchanged into high ionic strength buffer (I ≈ 0.8). High motor-to-tail FRET efficiencies were observed before permeabilization (EAVE
= 11.5 ± 1.9%; , EAVE
), indicating a close association of the KHC motor and tail regions. FRET remained high during permeabilization at physiological ionic strength (, B [EAVE
] and C); however, high ionic strength buffer resulted in a rapid and significant decrease in FRET efficiency (EAVE
= 4.0 ± 1.1%; , B [EAVE
] and C). COS cells expressing the mCit-16aa-mECFP calibration molecule exposed to the same conditions showed no significant change in FRET efficiency (EAVE
= 37.8 ± 2.1% at physiological ionic strength and 37.4 ± 1.8% at high ionic strength; , C and D [EAVE
]). Note that RM
remained constant in both cases, indicating negligible differences in acceptor and donor photobleaching. These results indicate that Kinesin-1 is folded (high motor-to-tail FRET) at physiological ionic strength, but is more extended (low motor-to-tail FRET) under high ionic strength conditions. Thus, FRET stoichiometry can detect conformational changes in Kinesin-1 in living cells.
Structural relationships within inactive Kinesin-1 molecules
To probe the overall structure of inactive Kinesin-1 in vivo, FP-labeled KHC and KLC were coexpressed in COS cells. We first measured FRET efficiencies for FRET pairs located on the KHC subunit (, ). For the KHC motor-to-tail relationship, higher FRET efficiencies were obtained for FRET pairs on the same KHC polypeptide (EAVE = 12.4 ± 1.0% for mCit-KHC-mECFP + Myc-KHC + HA-KLC; , ) than for FRET pairs on separate KHC polypeptides (EAVE = 4.8 ± 0.5% for mCit-KHC + KHC-mECFP + HA-KLC; , ). Although these data cannot distinguish the relationship between each motor and its tail domain because of differences in fraction of FP protein in complex and orientation of the FPs, these FRET efficiencies demonstrate that inactive Kinesin-1 molecules are in a folded conformation in vivo. For KHC motor-to-motor measurements, the FRET efficiency was low (EAVE = 2.4 ± 0.5%; , ), suggesting that the KHC N-terminal motor domains are separated in the inactive molecule. In contrast, for KHC tail-to-tail measurements, the FRET efficiency was higher (EAVE = 8.6 ± 0.9%; , ) indicating that the KHC C-terminal tail domains are relatively close together in vivo.
Figure 3. Structural organization of inactive and active Kinesin-1 in live COS cells. (A) FRET stoichiometry of inactive Kinesin-1 (KHC + KLC). (top left) Representative fluorescence image. (top right) The white boxed area of the image was enlarged, binned, (more ...)
Figure 4. Conformational changes upon Kinesin-1 activation. Inactive Kinesin-1 (left) is in a folded conformation such that the KHC motor and tail domains are in close proximity (green arrow), but the KHC motor domains are pushed apart from each other (blue arrow). (more ...)
We next measured FRET efficiencies within inactive Kinesin-1 molecules for FRET pairs located on the KLC subunit (, ). Little to no FRET was detected between the N and C termini of KLC (EAVE = 0.2 ± 0.1%; , ), indicating that the KLC subunit is in an extended conformation. Low FRET efficiencies obtained for the C termini of KLC indicate that these regions are separated (EAVE = 2.3 ± 0.4%; , ), whereas the higher FRET efficiencies obtained for the N termini of KLC indicate that these regions are in close proximity (EAVE = 11.2 ± 1.7%; , ), presumably because of dimerization via the heptad repeats.
Figure 5. The KHC motor/neck domains are separated in the inactive molecule by the presence of KLC. (A) Schematic diagram of mECFP-tagged KHC(Cys344) and KLC in the Kinesin-1 holoenzyme. (B) Lysates of COS cells expressing KHC(Cys344) alone (left) or with KLC (right) (more ...)
Figure 6. The KHC tail domain contributes to autoinhibition of Kinesin-1, but not conformational changes. (A) Live-cell microtubule-binding assay. mCit fluorescence images of COS cells expressing Myc-KHC + KLC-mCit (right) or Myc-KHC(1–891) + (more ...)
Figure 7. The KLC subunit contributes to both autoinhibition and conformational changes. (A, 1–6) The FRET pair being analyzed is indicated vertically to the left of the panels, the transfected constructs are shown schematically in the middle left, and (more ...)
Finally, we measured FRET efficiencies within inactive Kinesin-1 molecules for FRET pairs located on both the KHC and KLC subunits (, ). Moderate FRET efficiencies between the C terminus of KLC and either the N terminus of KHC (EAVE
= 5.8 ± 0.5%; , ) or the C terminus of KHC (EAVE
= 6.4 ± 0.4%; , 9) suggest that the KLC C terminus is in close proximity to both the KHC motor and tail domains. In contrast, negligible FRET efficiencies were observed between the N terminus of KLC and either the N terminus of KHC (EAVE
= 0.4 ± 0.1%; , 10) or the C terminus of KHC (EAVE
= 0.6 ± 0.3%; , 11). This suggests that the N terminus of the KLC subunit is close to the region in the KHC stalk that allows folding. These data also indicate that the KLC subunits lie in a direction parallel to the KHC subunits (N′ to N′ and C′ to C′; ). Collectively, these results support the overall structure of Kinesin-1 gleaned from various in vitro experiments (Vale, 2003
) and demonstrate that inactive Kinesin-1 molecules are in a folded conformation in intact cells.
Figure 8. Model for activation of Kinesin-1. Full activation of Kinesin-1 requires that the inhibitory effects of both the KHC tail and the KLC subunit must be relieved. This likely requires both cargo (green stars) binding to the KLC TPRs (shown) and cargo or (more ...)
Structural relationships within active Kinesin-1 molecules
To probe the structure of active Kinesin-1 in vivo, we measured FRET efficiencies from combinations of FP-KHCs expressed in COS cells (). FRET efficiencies from KHC molecules accumulated at the cell periphery in highly expressed cells (, ) were very high (EAVE > 20%), regardless of FP position, and correlated with fluorescence intensities (Fig. S2 C), indicating that intermolecular FRET occurs between crowded KHC molecules accumulated at the plus ends of the microtubules. FRET efficiencies for FP-KHC molecules localized in the rest of the cell (, ) remained constant despite variations in fluorescence intensity (Fig. S2 C), suggesting that these FRET measurements represent only intramolecular FRET. Thus, we only collect data from these regions or from cells with low-to-medium expressions to avoid artifacts caused by KHC accumulation.
For KHC motor-to-tail measurements, moderate FRET efficiencies were obtained for FRET pairs on the same KHC polypeptide (EAVE = 7.9 ± 1.5%; , ) and on separate KHC polypeptides (EAVE = 4.7 ± 1.0%; , ), indicating that the motor and tail domains of KHC remain in relatively close proximity upon activation. Moderate FRET efficiencies were also obtained for KHC motor-to-motor FRET pairs (EAVE = 6.1 ± 1.2%; , ) indicating that the two motor domains are in close proximity, as expected for active Kinesin-1. FRET efficiencies obtained for KHC tail-to-tail FRET pairs (EAVE = 8.3 ± 2.8%; , ) indicate that the KHC tails are also in close proximity.
To compare the structure of KHC motor domains engaged with microtubules with those in cytosol, FRET efficiencies were measured for KHC molecules forced on or off the microtubules. FP-KHC(1–891) was forced to remain on the microtubule by addition of AMPPNP (, 9) or was prevented from binding to microtubules by mutation of the microtubule-binding site in the motor domain (Δloop12 mutation; , 10; Woehlke et al., 1997
). Similar motor-to-motor FRET efficiencies were obtained for microtubule-bound and unbound motors (EAVE
= 6.7 ± 1.6% and 6.4 ± 0.6%, respectively). These results indicate that KHC motor domains in active molecules likely stay in close proximity regardless of whether they are on or off the microtubules.
Two conformational changes in Kinesin-1
To identify conformational changes within Kinesin-1 upon activation, we compared the FRET efficiencies of inactive (KHC + KLC; ) and active (KHC alone; ) molecules. KHC motor-to-tail FRET pairs on the same KHC polypeptide had higher FRET efficiency in the presence (EAVE = 12.4 ± 0.1%) than in the absence (EAVE = 7.9 ± 1.5%) of KLC. This difference is statistically significant (P < 0.001; ) and indicates a smaller distance between the KHC motor and tail domains in the inactive state. This global conformational change (, green arrows) likely displaces the KHC tail from the KHC motor domains for Kinesin-1 activation.
t test comparing FRET efficiencies of inactive and active Kinesin-1
For KHC motor-to-motor FRET pairs, a lower FRET efficiency was observed in the presence (EAVE
= 2.4 ± 0.5%) than the absence (EAVE
= 6.1 ± 1.2%) of KLC. This difference is statistically significant (P < 0.001; ) and indicates a larger distance between the two KHC motor domains in the inactive state. That the two KHC motor domains are pushed apart in the inactive holoenzyme was surprising because crystallography and 3D cryoelectron microscopy suggested that the motor domains of truncated KHC molecules are closer together when free in solution than when engaged with a microtubule (Marx et al., 2006
). A local conformational change (, blue arrows) upon activation is, thus, likely required to position the motor domains for processive motility.
To confirm that the two KHC motor domains are pushed apart in the inactive state, we tested biochemically whether the KHC neck coiled-coil segments are closer together in the active state (absence of KLC) than in the inactive state (presence of KLC). A Cys residue was introduced into the neck coiled coil of a Cys-lite version of KHC (KHC[Cys344]; ) at a position accessible to cross-linker, but demonstrated to have no effect on the motile properties of the truncated KHC (Tomishige and Vale, 2000
). When COS cell lysates expressing KHC(Cys344) in the absence of KLC (i.e., active Kinesin-1) were treated with the cross-linker 3-carboxy-4-nitrophenyl disulfide 6,6′-dinitro-3,3′-dithiodibenzoic acid Bis(3-carboxy-4-nitrophenyl) disulfide (DTNB), nearly all of the KHC(Cys344) was rapidly cross-linked as indicated by a shift to a slower mobility form (, lanes 2–5). In contrast, in the presence of KLC (i.e., inactive Kinesin-1), little to no cross-linking of KHC(Cys344) was observed (, lanes 7–9). Incubation in the presence of DTNB for long periods of time resulted in cross-linking of KHC(Cys344) + KLC (, lane 10), presumably caused by “breathing” of the Kinesin-1 holoenzyme. These results confirm that the KHC neck coiled coil is more separated in the inactive state than in the active state.
Contribution of the KHC tail domain to Kinesin-1 autoinhibition
The KHC globular tail domain has been implicated in contributing to both the folded conformation and autoinhibition of KHC in vitro (Coy et al., 1999
; Friedman and Vale, 1999
; Hackney and Stock, 2000
). In particular, a conserved stretch of residues in the KHC tail domain (the IAK region) is critical for autoinhibition of motor activity in vitro (Hackney and Stock, 2000
). To determine whether the KHC tail and/or the IAK region play a role in autoinhibition or conformational changes in the Kinesin-1 holoenzyme in vivo, we expressed truncated (KHC[1–891]) and mutated (KHC[ΔIAK]) versions of KHC in COS cells. KHC(1–891) + KLC did not localize to microtubules or accumulate at the cell periphery at steady state, but became locked on microtubules upon exposure of permeabilized cells to AMPPNP (, A [left] and B). The microtubule-bound state of KHC(1–891) + KLC reflects a direct interaction between the motor domain of KHC(1–891) and the microtubule because FRET efficiency between the mCit-KHC(1–891) motor domain and mECFP-tubulin significantly increased upon AMPPNP addition (Fig. S4, available at http://www.jcb.org/cgi/content/full/jcb.200605097/DC1
). Like the tail-truncated molecules, Myc-KHC(ΔIAK) + mCit-KLC molecules were capable of microtubule binding (locked on microtubules with AMPPNP; Fig. S5, D and E), but not processive motility (did not accumulate at ends of microtubules; ). These results indicate that removal of the KHC tail, or mutation of the IAK region, results in a Kinesin-1 holoenzyme that is active for microtubule binding, in contrast to KHC + KLC (, A [right] and B). Thus, the IAK segment of the KHC tail plays an important role in autoinhibition in vivo, specifically in preventing the microtubule association of Kinesin-1.
To test whether activation of Kinesin-1 by mutation of the IAK segment results in a global conformational change in Kinesin-1, we measured motor-to-tail FRET for FRET pairs on the same KHC(ΔIAK) polypeptide (). High FRET efficiencies were obtained for KHC(ΔIAK) + KLC molecules in the absence of AMPPNP (EAVE = 14.4 ± 1.8%, ), which is similar to wild-type Kinesin-1 (KHC + KLC) molecules with the same labeling (EAVE = 12.4 ± 1.0%; , ), and no change (P > 0.5) was detected upon addition of AMPPNP (EAVE = 14.5 ± 2.2%; ). Thus, despite a statistically significant difference (0.001< P <0.01) in the Relocation Index of KHC(ΔIAK) + KLC (, red), no difference in the motor-to-tail spatial relationship (, black) was detected (0.1< P <0.5), even after 30 min of exposure to AMPPNP.
We next looked for a local conformational change in active KHC(ΔIAK) + KLC molecules by measuring motor-to-motor FRET (). Low FRET efficiencies were obtained in the absence of AMPPNP (EAVE
= 2.5 ± 1.5%; ), similar to the values obtained for wild-type Kinesin-1 (EAVE
= 2.4 ± 0.5%; , ), and no change (P > 0.5) was detected after 10 min of AMPPNP exposure (EAVE
= 3.1 ± 0.1%; ). Interestingly, if left in the presence of AMPPNP for 30 min, KHC(ΔIAK) + KLC molecules showed a statistically significant (P < 0.001) increase in motor-to-motor FRET (EAVE
= 5.7 ± 0.7%; , black). This may reflect the ability of Kinesin-1 motors to exist in single- and double-headed binding states in the presence of AMPPNP, with the double-headed state predominating at low load in vitro (Kawaguchi and Ishiwata, 2001
) and after prolonged incubation in vivo.
Collectively, these results indicate that the IAK inhibitory region plays an important role in Kinesin-1 autoinhibition in vivo by preventing the microtubule association of inactive Kinesin-1 molecules. The IAK inhibitory region does not, however, contribute to the autoinhibited conformation in vivo because mutant Kinesin-1 molecules remained tightly folded with the motor domains pushed apart. Thus, other parts of the Kinesin-1 molecule must be required for generating the folded conformation and for keeping the motor domains pushed apart in the absence of cargo.
Contribution of the KLC subunit to Kinesin-1 autoinhibition
The KLC subunit contributes to both the folded conformation of Kinesin-1 and to the separation of the KHC motor domains (). To determine the regions of KLC that contribute to autoinhibition in vivo, we used a truncated version of KLC (KLC[1–176]) that lacks the tetratricopeptide repeat (TPR) motifs required for cargo binding, but retains the heptad repeats required for association with KHC (Verhey et al., 1998
). FP-KHC + FP-KLC(1–176) localized to the cytosol at steady-state and after exposure of cells to AMPPNP, similar to wild-type Kinesin-1 (Fig. S5, D and E), indicating that the heptad repeat region of KLC is sufficient for autoinhibition. This is likely caused by the ability of the KLC heptad repeats to maintain the folded conformation (, top), as no statistically significant difference was seen (0.1 < P < 0.5) in the motor-to-tail FRET efficiency of mCit-KHC-mECFP + KLC(1–176) (EAVE
= 12.1 ± 0.7%; , ) when compared with that of wild-type Kinesin-1 (mCit-KHC-mECFP + KLC, EAVE
= 12.4 ± 1.0%; , ).
We next looked for a local conformational change in KHC + KLC(1–176) molecules by measuring KHC motor-to-motor FRET (, ). Significantly higher FRET efficiencies (P < 0.001) were obtained for FP-KHC + KLC(1–176) (EAVE = 5.3 ± 0.5%; , ) than for wild-type Kinesin-1 (EAVE = 2.4 ± 0.5%; , ), indicating that the KHC motor domains are closer together when the KLC TPR motifs are removed (, bottom). To test whether the KHC tail domains also play a role in separating the motor domains, we compared the motor-to-motor relationships of molecules containing truncations of KLC, KHC, or both. KHC motors that are incapable of binding to microtubules (KHC[1–891]/Δloop 12 mutant) were used to eliminate potential effects that microtubule binding may have on motor-to-motor distances. Truncation of the KHC tail domain (KHC[1–891]/Δloop12 + KLC) caused no significant change (P > 0.5) in motor-to-motor FRET when compared with wild-type Kinesin-1 molecules (KHC + KLC; EAVE = 2.3 ± 0.8%; , , vs. EAVE = 2.4 ± 0.5%; , , respectively) and a small increase (0.01 < P < 0.02) in motor-to-motor FRET when compared with KLC-truncated Kinesin-1 molecules (EAVE = 6.2 ± 1.9%; , vs. EAVE = 5.3 ± 0.5%; , ). In contrast, truncation of KLC caused a significant change (P < 0.001) in motor-to-motor FRET when compared with either wild-type Kinesin-1 molecules (EAVE = 5.3 ± 0.5%; , vs. EAVE = 2.4 ± 0.5%; , , respectively) or to KHC-truncated Kinesin-1 molecules (EAVE = 6.2 ± 1.9%; , vs. EAVE = 2.3 ± 0.8%; , , respectively). These results indicate that the major contribution for separation of the KHC motor domains in the inactive conformation is provided by the KLC TPR motifs.