Background
Myosin-1c (Myo1c), the myosin previously called
myosin-Iβ, myr 2, or MI-110K [ 1 ] , is an unconventional
myosin isozyme implicated in nuclear transcription [ 2 ] ,
lamellopodia dynamics of motile cells [ 3 4 ] ,
brush-border dynamics of proximal-tubule cells of the
kidney [ 5 6 ] , and adaptation of mechanoelectrical
transduction in hair cells, the sensory cells of the inner
ear [ 7 ] . Myo1c belongs to the myosin-I class, which
contains eight members in humans [ 8 ] and mice [ 9 ] ; the
bullfrog genome possesses at least two members [ 10 ] .
Members of the myosin-I class have a single globular motor
domain, followed by a neck region and a relatively short
(30-40 kD) tail domain (Fig. 1A). This latter domain is
highly basic and binds to acidic phospholipids [ 11 ] .
Like all biochemically characterized unconventional
myosins, Myo1c binds calmodulin light chains in its neck
region [ 12 ] ; this region also interacts with
non-calmodulin receptors in hair cells [ 13 ] .
Unconventional myosins contain from one to several IQ
domains, which are calmodulin-binding motifs that adhere to
the general consensus sequence IQX
3 RGX
3 R [ 14 ] . Calmodulin, which can bind
up to four Ca 2+ions, generally binds IQ domains in the Ca
2+-free conformation; interaction of Ca 2+-bound calmodulin
to other proteins occurs through alternative binding motifs
[ 14 ] .
Myo1c contains three readily recognized IQ motifs of 23
amino acids each (Fig. 1C; refs. [ 10 15 16 17 18 ] ).
Purified Myo1c apparently includes 2-3 calmodulins per
Myo1c heavy chain [ 19 20 21 ] ; calmodulin supplementation
can increase the stoichiometry to as many as 4 calmodulins
per Myo1c [ 21 ] . Unfortunately, the lack of appropriate
quantitation standards for the Myo1c heavy chain in those
experiments limits the reliability of these values.
How Ca 2+and calmodulin regulate Myo1c or indeed any
myosin-I is unclear. Although Ca 2+increases ATPase
activity of most myosin-I isozymes,
in vitro motility is usually blocked
under identical conditions [ 11 ] . Ca 2+dissociates one or
more calmodulins from the myosin-calmodulin complex, which
apparently elevates ATPase activity and inhibits motility [
11 ] . In conventional myosin, light chains related to
calmodulin appear to be essential for stabilization of the
myosin lever arm [ 22 ] , a domain that is vital for
efficient conversion of chemical energy into mechanical
work [ 23 ] . Calmodulin probably plays a similar lever-arm
stabilizing role for Myo1c; Ca 2+-induced calmodulin
release would reverse the stabilization and inhibit
motility.
To better understand the regulation of Myo1c activity by
calmodulin, we sought to more accurately determine how
calmodulin binds to Myo1c by measuring the Ca 2+-dependence
of calmodulin binding to individual Myo1c IQ peptides. In
addition, to examine the consequences of calmodulin binding
to adjacent IQ domains, we measured hydrodynamic properties
of recombinant Myo1c-calmodulin complexes, under differing
conditions of Ca 2+, calmodulin, and temperature. These
measurements allowed us to determine the molecular mass and
hence stoichiometry of the Myo1c complex. Our results
indicate that IQ1, IQ2, and IQ3 have calmodulin bound when
the concentration of Ca 2+is low, and that increased Ca
2+induces release of calmodulin from IQ1 and IQ2.
Results
Sequence analysis of IQ domains
Examination of the primary sequence of the bullfrog
Myo1c neck region reveals an exact repeat of five amino
acids located both in the IQ3 region (YRNQP; residues
761-765) and at residues 786-790. Alignment of the
residues surrounding the repeat revealed reasonable
homology with the three known IQ domains, with particular
similarity to IQ3, suggesting that this region may be a
fourth IQ domain (Fig. 1C). Although the pair of amino
acids (LM; residues 782 and 783) that align with the RG
of the IQ consensus motif are not conserved, the first
pair of amino acids (IR; residues 777 and 778) that align
with the consensus IQ adhere to the consensus better than
those of IQ3. Because of the sequence similarity to IQ3
and because this peptide binds calmodulin (albeit weakly;
see below), we refer to this domain as IQ4.
IQ - Alexa-calmodulin interaction on plastic
plates
To investigate the calmodulin-binding properties of
each Myo1c IQ domain, we measured interaction of a
fluorescently labeled calmodulin (Alexa-calmodulin) with
individual Myo1c IQ peptides that had been conjugated to
wells of a plastic plate. We used an IQ peptide from
neuromodulin [ 24 ] as a positive control; calmodulin
binds to this site, with its interaction reduced by high
ionic strength [ 24 25 ] . As a negative control, we used
a peptide (PVP), corresponding to the 25 amino acids of
Myo1c immediately following IQ4. Indeed, Alexa-calmodulin
bound to wells derivatized with the neuromodulin IQ
peptide and did not bind to the PVP peptide (Fig.
2A).
Substantial amounts of Alexa-calmodulin bound to wells
derivatized with IQ1, IQ2, and IQ3; by contrast,
relatively little bound to IQ4-coated wells under these
conditions (Fig. 2A). As has been noted for the
neuromodulin IQ domain [ 25 ] , increasing the KCl
concentration reduced binding to each IQ peptide.
Although the data shown in Fig. 2Awere obtained at room
temperature, we saw a similar rank order of binding -
albeit with lower total Alexa-calmodulin bound - at 4°C
(data not shown).
To confirm the approximate binding strength reported
by this assay, we used free IQ peptides to prevent
Alexa-calmodulin binding to an IQ3-derivatized plate.
Because the IQ peptides strongly quenched
Alexa-calmodulin fluorescence when bound, we corrected
fluorescence measurements using an identical assay in an
underivatized plate. Although this quenching correction
introduced substantial scatter into the data, we found
that the apparent affinities for binding of peptides to
Alexa-calmodulin followed the order IQ3 > IQ1 ≈ IQ2
> IQ4 (Fig. 2B).
IQ - calmodulin interaction by quenching of
Alexa-calmodulin fluorescence
As noted above, Alexa-calmodulin fluorescence was
quenched upon binding to an IQ peptide (Fig. 3A). Because
an excess of unlabeled calmodulin was able to reverse
70-95% of the quench (Fig. 3A), we inferred that most
Alexa-calmodulin bound to the same site as unlabeled
calmodulin. We used this fluorescence-intensity quench
empirically to measure the affinity of each IQ peptide
for Alexa-calmodulin (Fig. 3B). In some experiments, Ca
2+was held at <30 nM by chelation with 100 μM EGTA; in
other experiments, we added 25 μM exogenous CaCl
2 in the absence of EGTA. These two
concentrations mimic the low- and high-Ca 2+conditions
that Myo1c may encounter in hair cells when the
transduction channel is closed or open [ 26 ] . In the
presence of 100 μM EGTA, K
d values followed the order IQ3 <
IQ1 ≈ IQ2 << IQ4. Although the data were fit
somewhat better with a modified Hill equation that with a
standard bimolecular-binding isotherm (Fig. 3B), the
physiological significance of Hill coefficients >1 is
uncertain, particularly given the 1:1 peptide:calmodulin
stoichiometry (see below). Ca 2+had only modest effects
on the affinity of the Myo1c IQ peptides for
Alexa-calmodulin (Table 1).
Despite only minor effects on binding affinity, Ca
2+did influence the calmodulin-peptide complex, as
signaled by changes in Alexa-calmodulin fluorescence.
Changes in fluorescence intensity during manipulation of
a single parameter, like Ca 2+concentration, should
report conformational changes in Alexa-calmodulin. For
example, the fluorescence intensity of free
Alexa-calmodulin in solution was ~15% lower in 25 μM CaCl
2 than in 100 μM EGTA (left-hand
limits in Fig. 4). Because the dye moiety itself is not
Ca 2+sensitive [ 27 ] , the Ca 2+-dependent fluorescent
change reflects changes in the dye's surrounding
environment, probably signaling the compact-to-open
structural change seen when Ca 2+binds to calmodulin [ 28
] . In contrast to the reduction of free Alexa-calmodulin
fluorescence by Ca 2+, fluorescence of Alexa-calmodulin
when saturated by IQ peptides was 1.5- to 2-fold
greater in 25 μM CaCl
2 than in 100 μM EGTA (Fig. 4; Table
1). Thus, when Alexa-calmodulin was bound to IQ peptides,
Ca 2+induced a conformational change that was
substantially different from that seen in the
peptide-free state.
IQ - calmodulin interaction under
stoichiometric-titration conditions
To determine the affinities of the Myo1c IQ peptides
for unlabeled calmodulin, we used Alexa-calmodulin as a
reporter (Alexa-calmodulin : unlabeled calmodulin ratio
of 1:100) in our binding studies. This approach assumes
that Alexa-calmodulin is functionally equivalent to
unlabeled calmodulin.
We determined affinities by fitting the IQ-peptide
concentration vs. fluorescence quench data with an
appropriate model. If the IQ peptides bound only
Alexa-calmodulin and not unlabeled calmodulin, the K
d and F
IQ /F values of Table 1would have
described the fit to the concentration-quench plots. The
line derived from these values did not fit the data (Fig.
5), indicating that, as expected, unlabeled calmodulin
binds to the Myo1c IQ peptides.
These experimental conditions resembled a
stoichiometric titration, where the total concentration
of calmodulin was higher than the K
d values for IQ1, IQ2, and IQ3. Under
true stoichiometric-titration conditions (fixed
concentration of receptor at 100-fold or more than the K
d , varying the ligand concentration
up to and beyond the receptor concentration), almost all
of the added IQ peptide would bind tightly to calmodulin
and linearly decrease the fluorescence; at the point
where the IQ-peptide concentration exceeds the calmodulin
concentration multiplied by the peptide:calmodulin
stoichiometry (
m ), a plateau in the fluorescence
intensity would be reached. Because the relatively weak
affinities observed here make such true stoichiometric
titration impractical, we used an intermediate
concentration of calmodulin (50 μM, ~10-fold larger than
K
d ) and used equation (7) to describe
the equilibrium precisely. This approach allowed us to
determine both
m and K
d in the same experiment.
In the presence of EGTA, the IQ1, IQ2, and IQ3 binding
data were much better fit by
m = 1 than they were to
m = 2, indicating that the binding
stoichiometry of peptide to calmodulin was 1:1 (Fig. 5).
The K
d values determined with equation (7)
were very similar to those determined for binding to
Alexa-calmodulin alone (Tables 1and 2), confirming that
under these conditions, the IQ peptides bind to unlabeled
calmodulin and Alexa-calmodulin similarly.
In the presence of Ca 2+, IQ3 also bound to calmodulin
with a stoichiometry of 1:1 (Fig. 5). The fits to
m = 1 and 2 were equally good for
IQ1 and IQ2 in the presence of Ca 2+, signifying the
inability for this analysis to determine precise binding
stoichiometry of IQ1 and IQ2 under these conditions. In
addition, these data indicate that the apparent
affinities of IQ1 and IQ2 for unlabeled calmodulin were
substantially weakened by Ca 2+(Table 2), unlike results
with Alexa-calmodulin alone (Table 1). Because the
assumption that affinities of Alexa-calmodulin and
unlabeled calmodulin for IQ peptides are identical was
violated for IQ1 and IQ2, the actual affinities of these
IQ peptides for unlabeled calmodulin may be even weaker
than those reported in Table 2. By contrast, Ca 2+had
only a very modest effect on IQ3 affinity for unlabeled
calmodulin.
Binding of IQ4 to unlabeled calmodulin was distinct
from that of the other IQ peptides. The data with IQ4
were best fit with a Hill equation (equation 4), with a
Hill coefficient of greater than 2 (Fig. 5, thick solid
lines), suggesting that binding of two peptides per
calmodulin may be required for the fluorescence change.
The apparent affinities (~100 μM) were similar to the
concentration of calmodulin (50 μM), however, indicating
that the apparent affinities did not accurately reflect K
d values. These results with a mixture
of unlabeled and Alexa-calmodulin were different from
those with Alexa-calmodulin alone, where IQ4 Hill
coefficients were close to 1 (data not shown).
Nevertheless, these data show that unlabeled calmodulin
can bind to IQ4, albeit with weak affinity and uncertain
stoichiometry.
Hydrodynamic analysis of full-length Myo1c
To determine the stoichiometry and Ca 2+-dependent
regulation of calmodulin binding to Myo1c with all four
IQ motifs, we co-expressed calmodulin and full-length
bullfrog Myo1c in insect cells using baculoviruses and
subjected the purified Myo1c-calmodulin complexes (Fig.
1B) to hydrodynamic analysis (Table 3). We carried out
velocity sedimentation of Myo1c-calmodulin complexes on
5-20% sucrose gradients to determine sedimentation
coefficients. We measured the Stokes radius of
Myo1c-calmodulin complexes using gel filtration on
Superdex 200 under temperature and buffer conditions
identical to those of the velocity-sedimentation
experiments (Table 3). Although most experiments used 400
mM KCl (which prevented adsorption to the gel-filtration
matrix), we obtained identical sedimentation coefficients
in the presence of 150 or 250 mM KCl (not shown).
Velocity-sedimentation and gel-filtration experiments
were carried out at 4°C, the temperature used for Myo1c
purification, as well as at 25°C, a physiologically
relevant temperature for a bullfrog.
To calculate the molecular mass of Myo1c-calmodulin
complexes, we applied the modified Svedberg equation,
which relates mass to the diffusion constant (calculated
here from Stokes radius) and the sedimentation
coefficient [ 29 ] . The partial specific volume of each
protein complex was determined using the amino-acid
composition of the constituent proteins (Table 3; ref. [
30 ] ). Although the uncertainty in calmodulin
stoichiometry leads to ambiguity in this calculation, the
calculated partial specific volumes were so close (
e.g. , 0.734 for one and 0.731 for
three calmodulins per Myo1c complex) that the precise
value did not significantly affect the final
molecular-mass value.
Full-length Myo1c bound ~3 calmodulins per Myo1c at
4°C in the presence of EGTA or CaCl
2 (Table 3). One of the bound
calmodulins was only weakly associated, as elevation of
the temperature to 25°C induced the release of 1 mole of
calmodulin in the presence of EGTA. When Ca 2+was
elevated to 25 μM at 25°C, however, we could not detect
substantial full-length Myo1c in solution after
sucrose-gradient centrifugation or gel filtration,
suggesting that the protein had aggregated.
Hydrodynamic analysis of T701-Myo1c
Because the size of full-length Myo1c (125 kD,
including purification and detection tags) is much larger
than calmodulin (16.7 kD), we improved our ability to
determine stoichiometry from molecular mass by examining
a smaller (45 kD) neck-tail recombinant fragment of
Myo1c. This construct, T701-Myo1c, contained amino acids
701-1028 of bullfrog Myo1c, including all four IQ
domains, the entire C-terminal tail, and N-terminal
purification and epitope tags (Fig. 1A,1B).
T701-Myo1c bound 2.5 moles of calmodulin per mole of
heavy chain at 4°C in the presence of 100 μM EGTA (Fig.
6; Table 4). As with full-length Myo1c, elevation of the
analysis temperature to 25°C induced the release of ~0.7
mole of calmodulin. In contrast to the results seen with
full-length Myo1c, elevation of the CaCl
2 concentration to 25 μM at 4°C also
induced the release of ~0.7 mole of calmodulin. The
amount of T701-calmodulin complex recovered on sucrose
gradients or by gel filtration decreased substantially
when the CaCl
2 concentration was elevated to 25 μM
at 25°C, signaling the formation of aggregates, as seen
with the full-length complex. Furthermore, the calculated
calmodulin stoichiometry of the observed T701-calmodulin
complex under these conditions was only ~0.3 mole of
calmodulin per mole of Myo1c, reinforcing the suggestion
that Ca 2+induced the dissociation of most calmodulins at
25°C and that this loss of light chains resulted in
aggregation.
We could not prevent the release of calmodulin at 25°C
by saturating T701-Myo1c with excess calmodulin
immediately prior to centrifugation (preloading). In
EGTA, the sedimentation coefficient of
calmodulin-preloaded T701-Myo1c measured at 25°C (3.85 ±
0.07 S; n = 2) was nearly identical to that measured
without preloading (3.83 S; Table 4). Likewise, the
sedimentation coefficient of calmodulin-preloaded
T701-Myo1c measured at 25°C and in 25 μM CaCl
2 (2.80 ± 0.42 S; n = 2) was similar
to that measured without preloading (3.13 S; Table
4).
By contrast, we could prevent the
temperature-dependent loss of calmodulin by carrying out
sedimentation in the continuous presence of 5 μM
calmodulin (Fig. 6C,6D; Table 5). Although gel-filtration
analysis was impractical with this high calmodulin
concentration, we assumed that the Stokes radius of
T701-Myo1c in the presence of calmodulin was identical to
the value obtained in the absence. Sedimentation at 25°C
in EGTA gradients supplemented with 5 μM calmodulin
resulted in the retention of ~3 calmodulins per
T701-Myo1c. In 25 μM CaCl
2 , supplementation with 5 μM
calmodulin resulted in ~1 calmodulin bound per
T701-Myo1c. In addition, protein loss due to aggregation
was minimal under these conditions.
Myo1c-calmodulin stoichiometry by gel
scanning
To measure Myo1c-calmodulin stoichiometry by an
independent method, we separated calmodulin standards and
T701-Myo1c by SDS-PAGE (Fig. 7A). Using densitometry, we
quantified the staining intensity of the calmodulin
standards to generate a standard curve (Fig. 7B) and
determined the amount of calmodulin present in each
T701-Myo1c sample. Applying the analysis described in
Experimental Procedures and equation (13), we found an
average of 2.6 ± 0.2 calmodulins per T701-Myo1c (mean ±
standard error) in six experiments, three separate
preparations analyzed in duplicate. This value was very
close to the value of 2.5 ± 0.1 bound calmodulins
determined independently by hydrodynamic analysis (Table
4).
Discussion
Calmodulin interaction with individual Myo1c IQ
domains
To examine how calmodulin binds to the Myo1c IQ sites,
we developed two binding assays using a commercially
available fluorescent calmodulin and individual IQ
peptides. In one assay, we covalently attached peptides
to plastic plates, then measured the amount of
fluorescent calmodulin that remained unbound after
incubation with the peptide-derivatized plate. This assay
was simple and fast, and allowed us to measure binding
under a wide variety of conditions. In our second assay,
we exploited the empirical observation that the
Alexa-calmodulin fluorescence intensity is quenched by
binding of IQ peptides. As in other assays with
fluorescently labeled calmodulins (e.g., ref. [ 31 ] ),
binding of the peptides to Alexa-calmodulin did not
perfectly mimic binding to unlabeled calmodulin. For
example, Alexa-calmodulin bound IQ peptides more strongly
in the presence of Ca 2+than did unlabeled calmodulin.
Moreover, excess unlabeled calmodulin could not fully
reverse the quenching of Alexa-calmodulin fluorescence
induced by IQ peptides, suggesting that IQ peptides could
bind to Alexa-calmodulin at two sites, including one
where unlabeled calmodulin could not bind. Indeed,
binding of IQ peptides to both sites on a single
fluorescent calmodulin could account for Hill
coefficients of >1 seen in some experiments (e.g.,
Figs. 3Band 4). Nevertheless, these discrepancies should
not prevent use of Alexa-calmodulin for measuring
interaction with calmodulin's targets, particularly if
the interaction with unlabeled calmodulin is compared to
the interaction with Alexa-calmodulin.
Calmodulin bound to peptides corresponding to each of
the four Myo1c IQ domains, although with differing
affinity and Ca 2+sensitivity. Affinities for calmodulin
binding to IQ1 and IQ2 were relatively modest (K
d values of ~5 μM). As with other IQ
domains [ 14 ] , Ca 2+weakened the affinity of calmodulin
for IQ1 and IQ2 by more than 10-fold.
By contrast, calmodulin binding to IQ3 was slightly
stronger and was affected much less by Ca 2+. Because
calmodulin binds strongly to classic IQ domains only in
the absence of Ca 2+ [ 14 32 ] , its strong binding to
IQ3 in the presence of Ca 2+suggests the participation of
an additional Ca 2+-requiring binding motif. Two common
calmodulin-binding motifs, called 1-8-14 and 1-5-10 for
the pattern of hydrophobic amino-acid residues, require
Ca 2+for calmodulin binding [ 14 ] . IQ3 has two nearly
perfect 1-5-10 domains that are at +2 net charge instead
of the minimum +3 in the consensus [ 14 ] . In addition,
IQ3 has a 1-8-14 motif with a proline residue at position
14 instead of phenylalanine, isoleucine, leucine, valine,
or tryptophan. Because most proteins that bind calmodulin
through the 1-8-14 and 1-5-10 motifs do so strongly, the
relatively modest affinity of IQ3 for calmodulin in the
presence of Ca 2+suggests that calmodulin binds through
one of these imperfect motifs located within this IQ
domain. To interact with an alternate set of residues, Ca
2+-calmodulin must adopt a new conformation. A similar Ca
2+-dependent rearrangement was predicted for the complex
of calmodulin and the first IQ domain of myosin-1a
(brush-border myosin I) [ 32 ] .
In support of this view, we observed evidence for Ca
2+-dependent conformational changes in calmodulin while
bound to IQ peptides. When Alexa-calmodulin was bound to
Myo1c IQ peptides, its fluorescence was higher in the
presence of Ca 2+than in its absence, suggesting that
that Ca 2+-bound Alexa-calmodulin binds to the IQ
peptides in a different conformation than does Ca 2+-free
Alexa-calmodulin. For example, in the absence of Ca 2+,
Alexa-calmodulin may bind to IQ peptides in a more
compact conformation, quenching fluorescence by burying
dye moieties in a less polar environment. Although the Ca
2+-induced conformational change could be a property of
Alexa-calmodulin rather than calmodulin itself, the Ca
2+-dependent changes in affinity of calmodulin for IQ1
and IQ2 (Table 2) and calmodulin's likely shift to a new
binding site on IQ3 suggests that the conformational
change is probably also a property of authentic
calmodulin.
Calmodulin also bound to a newly identified domain,
IQ4. Because the affinity of calmodulin for IQ4 is very
weak, calmodulin should only occupy IQ4 in subcellular
locations with a low Ca 2+concentration and a high level
of free calmodulin. For example, a small population of
Myo1c molecules with calmodulin bound to IQ4 should be
present in the stereocilia of inner-ear hair cells, which
contain ~35 μM free calmodulin [ 33 ] . Although most
tissues contain less free calmodulin [ 34 ] ,
concentrations in other individual organelles can reach
the millimolar range [ 35 ] . On the other hand, the weak
affinity of this IQ domain for calmodulin suggests that
IQ4 may play another role, such as interacting with
another protein.
Calmodulin interaction with Myo1c
The binding affinities of calmodulin for the
individual IQ peptides do not reflect exactly the
affinities of calmodulin for the IQ domains within Myo1c.
For example, despite micromolar K
d values for calmodulin-IQ peptide
interactions, calmodulin remains bound to Myo1c during
long gel-filtration or centrifugation experiments, even
at nanomolar Myo1c concentrations (Fig. 6). This result
suggests that calmodulin binds to some of Myo1c's four
tandem IQ domains substantially more strongly than to the
individual peptides. For example, other regions of Myo1c
could constrain the IQ domains in conformations that are
substantially more (or less) favorable for calmodulin
binding than the population of conformations adopted by a
soluble IQ peptide. Moreover, calmodulin binding to Myo1c
could be influenced by interactions with adjacent
calmodulin molecules or to the Myo1c head or tail
domains.
To examine calmodulin binding to IQ domains in Myo1c,
we determined the molecular mass (and hence
calmodulin:Myo1c stoichiometry) and shape of Myo1c under
the appropriate conditions of temperature and Ca 2+.
Although analytical ultracentrifugation is more commonly
used to measure molecular size of protein-protein
complexes [ 36 ] , we instead used classic hydrodynamic
methods of velocity sedimentation on sucrose gradients to
obtain sedimentation coefficients and gel filtration to
obtain Stokes' radius. One advantage of this approach was
that by detecting Myo1c using a sensitive ELISA method,
we were able to use very low concentrations of Myo1c.
Furthermore, we were able to carry out sedimentation in
the presence of a high concentration of calmodulin, a
manipulation that prevents Myo1c detection in a standard
analytical ultracentrifugation experiment. A disadvantage
of this approach was the need for high concentrations of
sucrose, which in rare conditions can substantially
affect the hydrodynamic properties of a protein [ 37 ] ;
nevertheless, changes in Myo1c size were observed both in
velocity sedimentation (in the presence of sucrose) and
in gel filtration (in its absence). Another disadvantage
of our classic approach to molecular-mass determination
was that the gel filtration and velocity sedimentations
were done on different time scales (~1 hour vs. 15-18
hours). If calmodulin slowly dissociated during the
analysis (which in both assays diluted Myo1c well below 1
μM), the degree of dissociation would be larger in the
velocity sedimentation experiments than in the gel
filtration experiments. Nevertheless, our approach was
validated by the demonstration that the number of
calmodulins per T701-Myo1c was identical in hydrodynamic
and gel-scanning experiments, at least in EGTA at
4°C.
Because T701-Myo1c mimicked properties of the
full-length protein (except under low-temperature,
high-Ca 2+conditions), we exploited the neck-tail
construct for a more detailed analysis of calmodulin
binding. As expected from the large Ca 2+-dependent
weakening of calmodulin affinity for IQ1 and IQ2 (Fig. 5;
Table 2), Ca 2+decreased the number of calmodulins bound
to T701-Myo1c at high ionic strength. When Ca 2+was low
at 25°C, each T701-Myo1c had about two bound calmodulins,
with a third bound if the calmodulin concentration
reached 5 μM. At this calmodulin concentration, IQ
domains 1, 2, and 3 are likely occupied by calmodulin.
When Ca 2+is high at 25°C, all but one calmodulin
dissociated from T701-Myo1c in the presence of 5 μM free
calmodulin. The strong affinity of IQ3 for Ca
2+-calmodulin suggests that the remaining calmodulin was
bound to this IQ domain.
How many calmodulins are bound to Myo1c in the cell at
increased Ca 2+concentrations? The elevated ionic
strength used for the hydrodynamic analysis probably
weakened the affinity of the calmodulin for IQ3 (Fig.
2A), requiring 5 μM free calmodulin to maintain occupancy
of that site. We therefore infer that at a
physiologically significant temperature and at a cellular
ionic strength, Ca 2+triggers release of calmodulins from
IQ1 and IQ2 from T701-Myo1c, leaving only IQ3 occupied.
Although these results contrast with those reported for
mammalian Myo1c, where only one of three calmodulins is
released by Ca 2+ [ 19 38 ] , our T701 construct lacks
Myo1c's motor domain. It is entirely plausible that even
in the presence of Ca 2+, calmodulin remains bound to
IQ1, albeit in a different conformation and dependent on
interactions with the myosin head. Our results therefore
suggest that Ca 2+either induces the release of
calmodulin from IQ1 or causes it to change its
interaction with Myo1c substantially.
Of the three calmodulins bound to Myo1c, one of these
binds relatively weakly at 25°C, even in EGTA. To which
IQ domain does this weakly bound light chain bind?
Although calmodulin binds to IQs 1-3 with approximately
the same strength in the presence of EGTA, we suggest
that the readily released calmodulin is likely to be that
bound to IQ2. To bind three calmodulins, IQs 1-3, each of
which are only 23 amino acids long, must be arranged
without kinks [ 32 ] ; this arrangement may produce
unfavorable strain on each of the calmodulin molecules.
Release of calmodulin from IQ2 would relieve all of that
strain; release from IQs 1 or 3 would not. Strain relief
also may accelerate calmodulin release in the presence of
Ca 2+; because Ca 2+apparently rearranges the
three-dimensional interaction of calmodulin with IQ3,
binding of an adjacent calmodulin - on IQ2 - might be
destabilized even more [ 32 ] .
Despite the loss of calmodulin from T701-Myo1c induced
by elevation of the temperature from 4°C to 25°C, the
frictional ratio (a measure of the protein's asymmetry)
increased (Table 4). The neck-tail region of Myo1c thus
appears to adopt a compact structure at 4°C, becoming
more extended at 25°C. Less calmodulin may be released at
lower temperatures because the Myo1c tail may bind to and
stabilize calmodulin's interaction with the Myo1c
neck.
Implications for Myo1c activity
The Ca 2+-dependent change in interaction of
calmodulin with IQ1, the IQ domain closest to the motor
domain, has important implications for Myo1c
mechanochemical function. Although Ca 2+increases Myo1c
ATPase activity, the ion completely halts
in vitro motility [ 19 ] . Ca
2+-dependent changes in conformation may prevent
amplification of a small converter-domain movement into a
large motor step. In the presence of an external force,
as is seen by Myo1c during an excitatory mechanical
stimulus in a hair cell [ 39 ] , Ca 2+(which enters the
cell through open transduction channels), should permit
Myo1c to go through its ATPase cycle, binding and
unbinding from actin, but the altered interaction of
calmodulin and IQ1 may prevent force production by the
motor. We predict that Ca 2+will decrease the stiffness
of a Myo1c-actin interaction, preventing coupling of the
energy released by ATP hydrolysis to the swing of the
neck [ 40 ] . This behavior will assist Myo1c in its role
of adaptation in hair cells, where the motor reduces
force applied to the hair cell's transduction
channel.
A limitation of our experiments is the restriction of
Myo1c binding to a single type of light chain,
calmodulin. Other light chains can interact with IQ
domains, including essential light chain isoforms [ 41 ]
and calmodulin-like protein [ 42 ] . Although purified
bovine adrenal Myo1c does not appear to have alternative
associated light chains [ 12 ] , we can not rule out the
possibility that other light chains bind in a cellular
context. Nevertheless, purified recombinant full-length
Myo1c associated with calmodulin light chains exhibited
actin-activated ATPase activity and motility
in vitro [ 43 ] , indicating that
calmodulin can function as a Myo1c light chain.
That Myo1c does not bind calmodulin tightly is, at
first glance, surprising. Weak calmodulin binding may,
however, permit access of IQ domains to intracellular
Myo1c receptors. Accordingly, we have found that a Myo1c
fragment containing only IQs 1-3, partially complexed
with calmodulin, binds avidly to hair-cell receptors;
excess calmodulin blocks this interaction, probably by
binding to an unoccupied IQ site on the Myo1c fragment [
13 ] . IQ2 is highly conserved between species, leading
us to propose that hair-cell receptors interact through
this region [ 13 ] . Because Myo1c-interacting proteins
in hair cells and elsewhere may interact through IQ
domains, regulation of calmodulin binding to Myo1c - for
example, by Ca 2+- likely affects coupling of the motor
protein to its cargo.
Conclusions
Under low Ca 2+conditions and normal ionic strength,
calmodulin binds moderately tightly to three Myo1c IQ
domains, IQ1, IQ2, and IQ3. IQ4 will only be occupied when
the calmodulin concentration is very high. When linearly
arranged in the Myo1c molecule, at least one calmodulin
(most likely that bound to IQ2) is bound less tightly,
probably due to steric constraints. Upon binding Ca 2+,
calmodulin bound to IQ2 dissociates; that bound to IQ1
either dissociates or changes its conformation sufficiently
that chemomechanical coupling cannot ensue.
Methods
Peptide - calmodulin interaction on plates
Bullfrog Myo1c IQ peptides were synthesized (Genemed
Synthesis, South San Francisco, CA) with N-terminal
cysteine residues: IQ1 (residues 698-720),
CRKHSIATFLQARWRGYHQRQKFL; IQ2 (721-743),
CHMKHSAVEIQSWWRGTIGRRKAA; IQ3 (744-766),
CKRKWAVDVVRRFIKGFIYRNQPR; and IQ4 (767-791; native
cysteine at residue 767), CTENEYFLDYIRYSFLMTLYRNQPK.
Peptide concentrations were measured by determining
optical density at 280 nm, using calculated molar
extinction coefficients of 7090 (IQ1), 11500 (IQ2), 7090
(IQ3), and 5240 M -1cm -1(IQ4). We also synthesized a
negative-control peptide ("PVP") corresponding to amino
acids 792-816 of frog Myo1c (SVLDKSWPVPPPSLREASELLREMC;
native C816) and a positive control IQ-peptide ("NM")
corresponding to amino acids 29-52 of bovine neuromodulin
with an added C-terminal cysteine
(KAHKAATKIQASFRGHITRKKLKC) [ 24 ] .
For measuring interaction of calmodulin with peptides
conjugated to plastic plates, we incubated 10 μM peptide
in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM
KCl, 4.3 mM Na
2 HPO
4 , 1.4 mM KH
2 PO
4 , pH 7.4) overnight at room
temperature in a maleimide-derivatized 96-well plate
(Pierce, Rockford, IL). Peptide was present in large
excess over free binding sites (25-50 pmol) on the
plates. To remove unconjugated peptides, plates were
washed with PBS; unreacted sites were saturated by
incubating with 10 μg/ml cysteine for 1 hour. We then
incubated the peptide-conjugated plates with 50 nM Alexa
Fluor 488 calmodulin (Alexa-calmodulin; Molecular Probes,
Eugene, OR) in 100 μl of a solution that contained 150 or
400 mM KCl, 1 mM MgCl
2 , 100 μM ethylene
glycol-bis(β-aminoethylether)-N,N,N',N'-tetraacetic acid
(EGTA) or 25 μM CaCl
2 , and 15 mM
2-[4-(2-hydroxyethyl)-1-piperazinyl] ethanesulfonic acid
(HEPES) at pH 7.5. According to the manufacturer,
Alexa-calmodulin had two dye moieties per calmodulin
molecule; the modified residues were likely Lys-75 and
Lys-94, the most reactive of calmodulin's lysine residues
[ 44 ] . After incubation for 2 hours at room
temperature, we transferred 50 μl of the solution to
another 96-well plate and measured fluorescence
(excitation 485 nm; emission, 520 nm) using a BMG
Labtechnologies Fluorostar 403 microplate fluorometer
(Durham, NC). Under the assay conditions, the
inner-filter effect (absorption of excitation or emission
photons by the sample) was negligible. From this
measurement, we calculated the amount of calmodulin bound
to the conjugated peptides. In some experiments, we also
included 0.1-100 μM unconjugated IQ peptide; in that
case, we carried out duplicate control reactions in
underivatized 96-well plates to correct for fluorescence
quenching exerted by IQ peptides.
Peptide - calmodulin interaction by fluorescence
quench
We used empirically observed changes in the
fluorescence intensity of Alexa-calmodulin, large in
magnitude, to measure binding of IQ peptides to
calmodulin. Peptides and 50-500 nM Alexa-calmodulin were
mixed in 96- or 384-well microtiter plates with 150 mM
KCl, 1 mM MgCl
2 , 100 μM EGTA or 25 μM CaCl
2 , 0.5 mg/ml bovine serum albumin,
and 15 mM HEPES at pH 7.5; in some experiments we added
50-75 μM bovine-brain calmodulin. Total volume varied
from 10 μl (384-well plates) to 100 μl (96-well plates).
After 1-2 hours at room temperature, fluorescence was
read directly.
When IQ peptides bound to Alexa-calmodulin, the
fluorescence intensity was reduced as the quantum yield
decreased (fluorescence quenching). We assumed that two
fluorescent species were present, Alexa-calmodulin and IQ
peptide-bound Alexa-calmodulin, and that the fluorescence
intensity (
I ) was a linear combination of the
fluorescence of the two species:
I =
f
CaM
I
CaM +
f
CaM-IQ
I
CaM-IQ (1)
where
f
CaM and
f
CaM-IQ are the mole fractions of the
two components and
I
CaM and
I
CaM-IQ are their fluorescence
intensities. Because the quantum yield of
Alexa-calmodulin is reduced when IQ peptides bind,
I
CaM-IQ <
I
CaM . The fraction of peptide bound
is:
To calculate K
d , we fit the data with a
bimolecular-binding isotherm:
where [IQ] was the free IQ-peptide concentration
added. Because we used concentrations of Alexa-calmodulin
in our experiments that were much less than the K
d , we approximated [IQ] using the
total IQ peptide concentration.
In other cases, however, the binding data were fit
better with a modified Hill equation:
where
h is the Hill coefficient. A value
for
h greater than one suggests the
fluorescence change arose from a more complex equilibrium
than just one peptide binding per calmodulin.
To carry out stoichiometric-titration experiments
(calmodulin concentration greater than the K
d ), we used a low concentration of
Alexa-calmodulin as a reporter and added an excess of
unlabeled calmodulin. For simplicity in analysis, we
assumed that Alexa-calmodulin behaved identically to
calmodulin, and thus this calmodulin mixture was
equivalent to a decrease in specific activity
(fluorescence quench) of calmodulin. We then solved the
bimolecular-binding isotherm to enable us to plot the
total ligand concentration (
T ) added versus fluorescence
intensity (
I ). The concentration of peptide
bound (
B ) was:
B =
m [CaM]
f
CaM-IQ (5)
where
m is the number of binding sites
per calmodulin and [CaM] is the fixed concentration of
calmodulin. The free concentration of IQ peptide (
F ) was
T -
B . We substituted the expression
for
B in equations (2) and (5)
into:
Note that
n [CaM] is the maximum amount of IQ
peptide that can bind (B
max ). We then solved equation (6) for
fluorescence intensity using Mathematica 4.0 (Wolfram
Research, Champaign, IL):
For
m = 1, the only free parameters
were K
d and
I
CaM-IQ . We were forced to include
I
CaM-IQ as one of the fit parameters;
the limited solubility of IQ peptides in the assay
solution prevented us from using very high peptide
concentrations that would independently establish its
value by producing a plateau in the
T vs.
I plot. We then used the value of
I
CaM-IQ determined from the
m = 1 fit and refit the data for
m = 2, using K
d as the only free parameter. To judge
the stoichiometry, we compared by eye the effectiveness
of the fit under the two conditions.
Baculovirus constructs
Using methods described previously for rat Myo1c [ 43
] , we cloned full-length bullfrog Myo1c into the
baculovirus transfer vector pBlueBacHis2B (Invitrogen,
Carlsbad, CA), introducing an N-terminal hexahistidine
tag for purification and a DLYDDDDK epitope tag for
antibody detection. Baculoviruses were generated,
purified, and characterized using standard techniques [
43 45 ] .
Protein expression and purification
Bullfrog Myo1c or its neck-tail fragment (Fig. 1A,1B)
were co-expressed with
Xenopus calmodulin in Sf9 cells
using methods described previously [ 43 ] .
Xenopus calmodulin is identical to
all other sequenced vertebrate calmodulins, including
bovine calmodulin [ 46 ] ; we presume that bullfrog
calmodulin is also identical. Recombinant proteins were
partially purified by centrifugation of an Sf9-cell
extract and Ni 2+-nitrilotriacetic acid chromatography [
43 ] ; further purification was achieved using gel
filtration at 4°C on a 25-ml Superdex 200 HR 10/30 column
run at 0.5 ml/min in 400 mM KCl, 1 mM MgCl
2 , 100 μM EGTA, 15 mM HEPES pH 7.5
with an AKTA-FPLC system (Amersham Pharmacia Biotech,
Piscataway, NJ). The concentration of each purified
recombinant protein was calculated by measuring
absorption at 280 nm and using extinction coefficients
calculated from the appropriate aminoacid sequence using
the ExPASy ProtParam tool
http://www.expasy.ch/tools/protparam.html, assuming 2.5
calmodulins per full-length Myo1c (53,619 M -1cm -1) or
T701 fragment (65,565 M -1cm -1). We typically obtained
100-300 μg of recombinant protein from ~10 9Sf9 cells.
Full-length Myo1c had NH
4 Cl-activated ATPase activity [ 12 ]
of 1.8 ± 0.7 s -1, with a K
m for ATP of 0.3 ± 0.1 mM. Actin
activated basal Mg 2+-ATPase activity ~15-fold.
Calmodulin was purified from bovine brain (Pel-Freez,
Rogers, AR) by isoelectric precipitation and
phenyl-agarose (Sigma, St. Louis, MO) chromatography [ 47
] ; its concentration was measured assuming a molar
extinction coefficient of 3030 M -1cm -1at 276 nm [ 48 ]
.
Gel filtration
Stokes radii of Myo1c and T701-Myo1c were measured
using gel filtration on a 25-ml Superdex 200 HR 10/30
column at either 4°C or room temperature (23-25°C).
Columns were run at 0.5 ml/min in 400 mM KCl, 1 mM MgCl
2 , 15 mM HEPES pH 7.5, and either 100
μM EGTA or 25 μM CaCl
2 ; 5-20 μg of recombinant protein was
applied to the column. Columns were calibrated using
20-200 μg each of globular proteins of known Stokes radii
(thyroglobulin, 8.50 nm; ferritin, 6.10 nm; catalase,
5.22 nm; aldolase, 4.81 nm; bovine serum albumin, 3.55
nm; ovalbumin, 3.05 nm; chymotrypsinogen, 2.09 nm; and
RNase A, 1.64 nm; all obtained from Amersham Pharmacia
Biotech). Proteins were detected by absorption at 280
nm.
Velocity sedimentation on sucrose gradients
Sedimentation coefficients of full-length and
T701-Myo1c were measured using linear 5-20% sucrose
gradients in 11.5 ml of 400 mM KCl, 1 mM MgCl
2 , 15 mM HEPES pH 7.5, 0.2 mM
phenylmethylsulfonyl fluoride, 10 μM leupeptin, 10 μM
pepstatin, and either 100 μM EGTA or 25 μM CaCl
2 . Gradients were calibrated with
2-20 μg internal standards of known sedimentation
coefficients (catalase, 11.3 S; bovine serum albumin,
4.31 S; lysozyme, 1.91 S; all obtained from
Sigma-Aldrich). After centrifugation at 33,000-40,000 rpm
in an SW 41 rotor for 15-18 hours at 4°C or 25°C,
gradients were fractionated from the bottom into ~30
fractions. Calibration proteins were located using a
Bradford protein assay [ 49 ] ; Myo1c-containing
fractions were located by ELISA [ 43 ] using an antibody
against the Myo1c tail (mT2/M2; ref. [ 50 ] ) or against
the DLYDDDDK epitope tag (anti-Xpress; Invitrogen). To
determine the location of protein peaks, plots of
fraction number versus the levels of Myo1c or calibration
proteins were fit with either one, two, or three Gaussian
curves.
For calmodulin preloading of T701-Myo1c, 10 μM
purified calmodulin was incubated with 1 μM Myo1c in a
solution containing either 100 μM EGTA or 25 μM CaCl
2 for 60 min at room temperature prior
to centrifugation.
Determination of molecular mass
We used the modified Svedberg equation for
molecular-mass determination:
where
M = molecular mass, η = viscosity
of the medium,
N = Avogadro's number,
a = Stokes radius,
s = sedimentation coefficient, =
partial specific volume, and ρ = density of the medium.
Partial specific volume was calculated from the
composition of Myo1c or T701-Myo1c, along with the
appropriate number of calmodulins, by summing the partial
specific volumes of each amino acid [ 30 ] . We used η =
1.002 × 10 -2g cm -1s -1and ρ = 0.998 g cm -3.
Errors in molecular mass were propagated from standard
deviations for Stokes radius and sedimentation
coefficient measurements. To calculate error in
calmodulin stoichiometry, we used the conservative
assumption that all error in the molecular-mass
measurement was due to variability in the number of
calmodulins.
The frictional ratio was determined from:
where
f is the frictional coefficient of
the Myo1c-calmodulin complex and
f
0 is the frictional coefficient of a
sphere of equal volume. Accordingly, the frictional ratio
of a globular protein will be 1; that of an elongated
protein will be >1.
Stoichiometry determination by gel scanning
T701-Myo1c and bovine-brain calmodulin were separated
by sodium dodecyl sulfate gel electrophoresis (SDS-PAGE)
and stained with Coomassie blue R250. Gels were scanned
with a flatbed scanner; calmodulin was quantified using
analysis of the resulting images with NIH Image version
1.62. The concentration of the T701-Myo1c heavy chain was
determined by measuring absorbance at 280 nm, although
the analysis was complicated by the uncertain calmodulin
stoichiometry (
p ). To circumvent this problem, we
solved several simultaneous equations for
p . The molar extinction
coefficient of the T701-Myo1c/calmodulin complex (ε
T701-CaM ) is given by:
ε
T701-CaM = ε
T701 +
p ·ε
CaM (10)
where ε
T701 is the extinction coefficient of
the T701-Myo1c heavy chain alone (57,990 M -1cm -1), ε
CaM is the extinction coefficient of
calmodulin (2560 M -1cm -1), and
p is the calmodulin:T701
stoichiometry. The concentration of T701-Myo1c heavy
chain is given by:
where A280 is the absorbance of the complex at 280 nm
for a 1 cm pathlength. Finally,
p ·[T701
HC ] = [CaM] (12)
where [CaM] is the calmodulin concentration determined
by gel scanning. Solving for
p :
Other methods
We measured the free Ca 2+concentrations in our
solutions using spectrofluorometry with Calcium Green-2
(Molecular Probes). SDS-PAGE was carried out with 18%
acrylamide Criterion gels (Bio-Rad Laboratories;
Hercules, CA); gels were stained with Coomassie blue
R250.
Authors' contributions
Author 1 (PGG) conceived of the experimental approach,
carried out many of the experiments, supervised the
technician who performed the remainder, developed the
methods for analysis, analyzed and interpreted the data,
and wrote the manuscript. Author 2 (JLC) contributed to the
development of the experimental approach, helped analyze
and interpret the data, and edited the manuscript.
