AFM/TIRF force clamp measurements of neurosecretory vesicle tethers reveal characteristic unfolding steps

Although several proteins have been implicated in secretory vesicle tethering, the identity and mechanical properties of the components forming the physical vesicle-plasma membrane link remain unknown. Here we present the first experimental measurements of nanomechanical properties of secretory vesicle-plasma membrane tethers using combined AFM force clamp and TIRF microscopy on membrane sheets from PC12 cells expressing the vesicle marker ANF-eGFP. Application of pulling forces generated tether extensions composed of multiple steps with variable length. The frequency of short (<10 nm) tether extension events was markedly higher when a fluorescent vesicle was present at the cantilever tip and increased in the presence of GTPγS, indicating that these events reflect specifically the properties of vesicle-plasma membrane tethers. The magnitude of the short tether extension events is consistent with extension lengths expected from progressive unfolding of individual helices of the exocyst complex, supporting its direct role in forming the physical vesicle-plasma membrane link.

acting on the cantilever.
To determine the error in D, 20 consecutive measurements of D were performed with the same cantilever, and the RMS deviation was determined to be 1.88 nm/V. For the force curve experiments, 5 D values were averaged to determine the value used in the conversion factor, so the final RMS deviation was: . (1) The Agilent Thermal k software reported the k value with good reproducibility to the thousandths place in N/m. Therefore, an error of 0.0005 nN/nm was used, or: . ( Thus, the total error in C was: . ( To examine the dependence of V defl on z, a subset of 60 of the 390 collected force curves was selected using the random number generator in the program R. Of the 60 force curves chosen, 4 were discarded because they were unusable due to problems that occurred during data collection. For the region of a force curve during which the servo was approaching the surface, a linear fit to the V defl vs time trace and the z vs time trace were performed. The ratio of the slopes of the fits gave a value for the V defl drift in V/μm. A histogram of all the V defl drift values was generated, and a Gaussian fit to the histogram (Fig. S3) had a drift value of -0.05±0.043 V/μm (mean ± sd). To account for the drift effect, each V defl measurement was adjusted by: .
Also, the V defl measurements had an uncertainty of: .
To confirm that the drift effect was due to servo motion, a histogram of V defl slopes was plotted for the region before the servo was moved (not shown). In this case, the mean drift value was 0.00042 V/μm, suggesting no significant V defl drift.
Baseline V defl at the sample surface was estimated at both the beginning and end of each force curve. During cantilever approach, surface contact was confirmed by an increase of V defl by 0.05 V above the initial value, then a second increase of 0.05 V that occurred with <50 nm further z displacement. The second increase was a test to confirm contact, so the actual point of contact should almost always occur during the first 0.05V increase. Therefore, the V defl at initial contact with the surface was taken to be: , where V contact was the value of V defl at which the script reported confirmed contact. Error was then: .
A second estimate of baseline V defl was taken after the servo was set to z contact + 500 nm at the end of each force curve. In this case, the error came from the V defl drift with z, so the baseline V defl at the end was: .
where V end was the V defl value measured at the end of the force curve. The error was: .
One more effect had to be accounted for before V defl could be converted to a pull force.
V defl was observed to drift negative with time. To account for this, the V defl and time values at surface contact (Eq. 11 and Eq. 13) at the beginning and end of the run were taken as endpoints for a line, and the slope of that line was calculated. For each pull segment, this slope value was multiplied by the time at the segment center and added to the V defl value of that segment. After accounting for the above corrections, the final conversion from V defl to force (F) was: where F was the pull force value for the segment, V defl was the deflection value for the segment, t contact was the time at which V contact was measured, t end was the time at which V end was measured, and t segment was the time at the center of the segment. The minus sign was used to give pull forces positive values. When the above errors were included, the variance of each force measurement was: .

Alignment of AFM laser and cantilever calibration
The degree of cantilever bending was measured by a laser that was reflected off of the cantilever and onto a quad photodiode (QPD), producing the cantilever deflection signal, V defl . At the beginning of each day of experiments, the AFM laser was aligned on the cantilever such that it was reflected to the approximate center of the QPD. This was done in buffer on the AFM/TIRF assembly using the 405 filter set and Hitachi camera to show the position of the AFM laser on the cantilever. To convert V defl into a pull force, two calibrations were required: deflection sensitivity (D) and cantilever spring constant (k). D is the amount by which the cantilever must be deflected, as measured by cantilever tip displacement from rest, to result in a given change in V defl . To measure this, the cantilever was pressed onto the surface of a glass coverslip as used for experiments with buffer but without PDL or cells on it. Thus, the tip would remain fixed while z was adjusted, causing the cantilever to Fig. S4 shows this procedure and the resulting V defl . A plot of V defl vs z shows a straight line for the region in which the tip is pressed onto the surface, and the slope of that line is -1/D.
The slope was measured with a procedure built into the PicoView software.
To determine k, the Thermal k method 1,2 built into the PicoView software was used. The Thermal k calibration was performed at a height of 50 μm above the same coverslip with buffer that was used for the deflection sensitivity measurement. To minimize mechanical noise in the power spectrum, the thermal k calibration was performed with the iXon camera water cooling pump turned off.

Alignment of membrane sheets, TIRF objective, and AFM cantilever
Immediately after cell lysis, the coverslip with the membrane sheets was mounted on the AFM, which was slid into place over the TIRF objective using the Quick Slide stage. The halogen lamp was turned on, and the Hitachi camera was used along with a 4x objective and the 488 filter set to observe the cantilever and the lysed region of the sample, where the membrane sheets were located. The lysed region was recognizable by a lack of cells. Micrometers on the AFM stage were then adjusted by hand to position the sample so that the lysed region was located below the cantilever tip, and micrometers on the Quick Slide stage were used to center the AFM tip in the TIRF microscope field of view.
The AFM motor was then used approach the sample surface using the PicoView software.
Contact with the sample surface was indicated by V defl exceeding a preset threshold. The AFM tip was centered again in the field of view, using the micrometers on the Quick Slide stage. The 4x objective was switched out for a 40x objective, and the AFM tip was centered again. The 40x objective was then switched out for the TIRF objective. Due to the short working distance of the TIRF objective, it was no longer possible to view the AFM tip directly. However, by using the 405 filter set, the light from the AFM laser could be seen. The TIRF objective was raised toward the sample until the shadow of the tip was visible within the laser light. When the shadow of the tip was nearly in focus, the objective was close to the sample. The position of the tip was again centered in the field of view, and then the illumination source was changed to the 488 nm laser with the 488 filter set. The eGFP-labeled vesicles were visible in the TIRF illumination. The shadow of the AFM tip could also be seen by allowing a small amount of transmitted light from the microscope's halogen lamp.
The cantilever tip was moved to a new spot away from the original landing spot in case the sample was damaged by the coarse motor approach. To do this, the AFM servo was used to lift the cantilever 1 µm off the surface, and the AFM micrometers were used to move the sample without moving the cantilever. Once a membrane sheet was located, the tip was positioned over it.

Correlation of AFM and TIRF data
For analysis, the V defl , z, and camera FIRE signal traces recorded by PicoView and the Python script were imported into Igor Pro, along with the corresponding time traces. A custom Igor function checked every data point in the FIRE signal. If a point had a value < 3, the camera was considered to not be exposing. Otherwise, the camera was exposing. Using the fact that the last camera frame was known to be frame 1000, a new wave, called fnum, was generated that had value equal to the current frame number x 10 -3 at all points for which the camera was exposing and a lesser value otherwise (Fig. S5). This approach directly matched each camera frame exactly to the corresponding deflection and z sensor data. Using the known cycle time of the camera (0.05091 s), the frame number of the first full frame after the start of the force segments, and the start time of that frame, the gap between the end of the script data recording and the beginning of the force clamp segments was determined.
During the approach phase, the time resolution of the recording of the FIRE signal was not sufficient to resolve individual camera frames. Therefore, the intensity data was evenly spaced within the timeframe the camera was recording.

Semiautomatic detection and analysis of tether extension events
The detection of the force transients was thus based on the software developed by Mosharov and Sulzer for amperometric spike analysis [3] using the cantilever voltage signal V corresponding to negative force, as illustrated in Fig. S6 using a rather large tether extension event for clarity.
First, the program calculates the time derivative of the force trace ′ = !" !" followed by appropriate smoothing, which facilitates the detection of the rapid rise of the force transient [4].
The time derivatives of all traces analyzed here were smoothed by the Box-Car smoothing method using a box size of 80 data points. This filter was chosen based on the assertion that the primary sources of background fluctuations in the data could be characterized as white noise. The time points t start , and t end of the force transient (panel a of Fig. S6) are usually taken as the first time points before and after t(Vmax) at which the cantilever deflection voltage (and applied force) returns to the baseline level V(F clamp ). If V(F clamp ) is unstable, then this method is not reliable and the time derivative V' is used instead of V to determine t start , and t end . In this case t start is set to the time of the first zero of V' to the left of t(V'max), the time where V' has its maximum, and t end is set to the time after t(V max ) when V returns to the level measured at t start .
Determinaton of t start and t end is complicated by the broad variability of possible force transient shapes and arrangements. Therefore, a supplemental algorithm was designed to increase the accuracy of t start and t end detection accounting for such complications during automated analysis.
If activated, the algorithm will first divide the V trace into segments of size Δt seg . The default initial guess for Δt seg is ∆tmax (see above). Starting at t(V max ), the program iteratively searches in positive and negative direction until two successive segments are found to have average forces within one SD V of each other, where SD V is the rms noise of V. We then set t start and t end as the first time-points at this steady-state force on the corresponding side of t(V max ),. If subsequent force transients are found to overlap, they are discarded, analyzed by separation, or considered as a single complex tether extension event.

Determination of tether extension magnitudes
Tether extension steps coincide with force transients. Therefore, t start and t end of a tether extension are assigned the time-values obtained from the associated force transient (panel b of Fig. S6). An estimate of the magnitude of a tether extension event can be obtained from the di!erence between the z-servo position at t end and the z-servo position at t start . While this may be suitable for characterizations of large steps, estimates of the magnitudes of smaller steps become grossly inaccurate due to the noise in the V defl trace.
We therefore developed a more robust approach based on a method originally developed to determine membrane capacitance step sizes associated with vesicle fusion [5] that uses linear extrapolation to find a stable value for the z-position before t start and after t end . The method was implemented to fit a line to user-determined time intervals (default = 200 ms) both before t start and