Volume 124, No. 3, September 2004. Pages 259–272.

On further examination of the data obtained from the three-dimensional reconstructions we realized that in the two places where we took the data as being derived from a single channel there was more than one channel present. Although the conclusions reached from the data would not be altered, we want to inform the readers of this error.

For the data shown in Fig. 2 on page 265, although only one fluorescence transient was detected, there was a second caffeine-activated channel that opened before and closed after the fluorescence transient. Any fluorescence increase due to Ca influx through that channel was out of the image field and, therefore, would not contribute to the fluorescence measurements obtained for the figure. However, the brief closures in the current trace could have come from either channel. When constructing the figure, we had assumed all of the brief closures were from the channel of interest. To estimate how much of an error this assumption would produce, we recalculated the slope of ΔFtotal (or signal mass) versus charge (ΔQ) assuming the opposite situation (all of the brief closures being from the other channel). The average difference in the slopes for all six planes remained nearly the same (−1.8 ± 0.7% versus −1.3 ± 0.7% for original versus recalculated data, respectively). The latter value is the same as that indicated in the original text by plotting ΔFtotal versus time instead of charge. Therefore, the brief closures did not contribute significantly to the differences among the slopes of ΔFtotal versus ΔQ as viewed from each optical plane.

The stretch-activated channel current (referred to at the top of the second column on page 266), recorded in a cell-attached patch, could have been derived from more than one channel. However, unlike for the caffeine-activated channels, the opening of more than one channel in the cell-attached patch would not have altered the slope of ΔFtotal versus ΔQ because both measurements were obtained from the same patch with the channels essentially in the same plane. Also, there was essentially the same average difference in the slopes of ΔFtotal versus ΔQ or versus time (−4.5 ± 4.9% and −3.3 ± 5%, respectively). Both values were also in the original text.

We regret we did not catch these before the publication of the paper and hence did not include the above in the published analysis. However, none of what we have described above would have changed the conclusions based on this data.