Technical Note pubs.acs.org/ac
Method for Estimating the Tip Geometry of Scanning Ion Conductance Microscope Pipets Matthew Caldwell,*,†,‡ Samantha J. L. Del Linz,†,‡ Trevor G. Smart,‡ and Guy W. J. Moss†,‡ †
Centre for Mathematics & Physics in the Life Sciences & Experimental Biology, University College London, London, U.K. Department of Neuroscience, Physiology and Pharmacology, University College London, London, U.K.
‡
ABSTRACT: Scanning ion conductance microscopy (SICM) offers the ability to perform contact-free, high-resolution imaging of biological cells and tissues at physiological conditions. However, imaging resolution is highly dependent on the geometry of the SICM probe, which is generally not known. Small, high-resolution probes are too fine to image optically and, to date, geometry estimation has usually required electron microscopy (EM). This is time-consuming and prone to failure and cannot provide information about the crucial internal geometry of the probe. Here we demonstrate a new method for determining SICM tip geometry that overcomes the limitations of EM imaging. The method involves fitting an analytical model to current changes during quasi-controlled breakage of the pipet tip. The data can be routinely obtained using the SICM apparatus itself and our method thus opens the way for substantially better quantification in SICM imaging and measurement.
■
S
EXPERIMENTAL SECTION Apparatus and Materials. Scanning ion conductance microscopy was performed using an Ionscope ICnano SSH01 system, equipped with Physik Instrumente Hera P-621.1CL and LISA P-753 piezos for sample and pipet positioning, respectively. The SICM head was mounted on the stage of a Nikon Eclipse TE-2000U optical microscope, in turn mounted on a Halcyonics Active Workstation 900-369 M6/25 vibration isolation table. SICM pipets were pulled from borosilicate glass with an outer diameter of 1.0 mm and an inner diameter of 0.5 mm (Sutter Instrument) using a P-2000 laser puller (Sutter Instrument). The electrode voltage was applied and ion current monitored using an Axon Instruments Axopatch 200B patch clamp amplifier and CV203BU head stage. Ion currents, driving voltages, and positional signals from the piezo capacitative sensors were digitized via an Axon Instruments Digidata 1440 interface and recorded at a 10 kHz sample rate using Clampex 10 software. The SICM was controlled via the IonScope ScanIC application, operating in hopping mode, with a hop height of 4 μm, a set point of 0.3%, and a fall rate of 50 nm/ms. Recordings were made in HEPES-buffered physiological saline containing NaCl 140 mM, KCl 4.7 mM, MgCl2 1.2 mM, CaCl2 2.5 mM, HEPES 5 mM, D-glucose 11 mM; pH adjusted to 7.4 with NaOH. All reagents were purchased from VWR. Solution resistivity was measured as 64 Ω cm at 23 °C, using a Radiometer Copenhagen CDM 83 conductivity meter.
canning ion conductance microscopy (SICM) measures surface topography by using attenuation of the ion current passing through a fine glass pipet tip as a proximity probe.1 Images are obtained in saline solution at physiologically relevant temperatures, so the technique is especially applicable to living samples.2 Originally designed around a constantdistance feedback protocol, the technique has been adapted by several groups to use variations on a “pulse” or “hopping” mode in which the probe is withdrawn after each surface measurement.3−6 This minimizes the risk of collisions during lateral movement and allows more convoluted samples to be successfully scanned. A key determinant of detection sensitivity and image resolution is the size and shape of the pipet tip.7−9 SICM tips are too small to examine optically and must be imaged by other means. Traditionally this has required scanning electron microscopy (SEM), a difficult and time-consuming process that is impractical for routine application. An alternative method using atomic force microscopy (AFM) has recently been reported,10,11 which should be more practical where the instrumentation is available. Most commonly, however, a rather crude estimate of tip size is made from the pipet resistance, together with some “rules of thumb” concerning the behavior of capillary glass during the pipet pulling process.12 Pipet resistance is a function of several factors, and a single measured value does not uniquely identify the tip geometry. For simple, conical geometries, the important factors are the pipet tip inner radius, ri, and the half cone angle θ. Here we present a method for estimating both factors from multiple resistance measurements during quasi-controlled breakage of the tip. Such measurements can readily be made by exploiting the repeated surface approaches of the hopping protocol. © 2012 American Chemical Society
Received: June 25, 2012 Accepted: August 14, 2012 Published: October 19, 2012 8980
dx.doi.org/10.1021/ac301851n | Anal. Chem. 2012, 84, 8980−8984
Analytical Chemistry
Technical Note
SEM images were obtained using a Jeol JSM-6480LV scanning electron microscope. Pipet tips were mounted on Agar Scientific 45° angled pin stubs. Agar Scientific quick drying colloidal silver was painted onto the pipet body to improve grounding. The tips were coated with gold using a Polaron E500 sputter coater. Distances and angles were measured from the resulting TIFF images using PixelStick software (Plum Amazing). Positional and ion current recordings were converted to text format using Clampfit 10 (Axon Instruments) and analyzed in R.13 Model parameters were fit by numerical minimization of the residual sum of squares. The custom R code used for analysis and fitting has been posted to the R online repository at http://cran.r-project.org under the package name breakage. Resistance Measurements with Quasi-Controlled Tip Breakage. In hopping mode SICM, the probe is repeatedly moved towards the sample while monitoring the ion current for a drop, indicating proximity to the surface.5 If the current drop is not detected in time, a collision occurs. During normal operation such collisions are highly undesirable, as they can damage delicate samples and also the pipet tip itself. However, collision with the coverslip can be used to perform quasicontrolled breakage of the tip. In principle, any intervention that obscures the current drop can be used to induce collisions. For simplicity, we applied a brief voltage pulse to saturate the current signal. The facility to produce such a “zap” voltage is built into many patch clamp amplifiers. This voltage is much greater than the pipet bias potential used for normal scanning. On the Axopatch 200B, for example, the zap is at 1.3 V, compared to a typical scanning potential of 100−200 mV. For the duration of the zap, the resulting current is beyond the measurement range of the amplifier and any variations due to surface proximity are rendered undetectable. The pipet was first positioned over an empty region of coverslip, in order to avoid damage to the sample or contamination of the tip during breakage. The pipet was then brought to the surface, hopping under control. Both the ion current and the pipet Z position, as reported by the piezo capacitative sensors, were recorded continuously while the zap was applied. Zapping is invoked manually via a button on the amplifier, so its onset was not synchronized with the hopping cycle. The probability of collision on any individual zap was thus dependent on the pulse duration relative to the hop period. For our experiments, with a hop height of 4000 nm and a fall rate of 50 nm/ms, the period was ∼80 ms, and a zap duration of 50 ms was found to give good results. When a collision resulted in tip breakage, this appeared in the recordings as an increase in resting current and an apparent change in surface position. A representative breakage sequence from one set of current and position traces is shown in Figure 1a. If required after each break, the sample was repositioned so that the probe was again over an empty region, clear of any fragments of the broken tip and the pipet voltage was reduced to ensure the ion current remained in a measurable range. The process was then repeated. To reconstruct the resistance data, traces for each break sequence were divided into hops. The pipet tip position was measured as the bottom of each hop. The corresponding ion current was taken as the median value measured in between consecutive bottom points; this was converted to resistance by
Figure 1. Tip breakages result in changes in pipet resistance and measured surface position. (a) Short (5 s) segment from simultaneous recordings of ion current (top) and position (bottom), while hopping. Onset of the zap voltage is indicated by an arrow. The SICM is unable to detect the surface during the zap and the pipet tip collides, causing a small break. (b) Resistances corresponding to different surface detection positions. Apparent changes in surface position result from change in pipet length after breakage. Data shown were derived from the same recordings excerpted in panel a.
division into the pipet potential. Because the coverslip was not perfectly horizontal, lateral repositioning could cause a change in the pipet baseline, so the pipet position after each shift was offset to match that before the move. Median estimates of all levels for a single pipet were then plotted together, as shown in Figure 1b. The overall highest position was taken as that of the initial unbroken pipet and all other positions subtracted from that baseline to obtain an estimate of the breakage distance. In normal operation at low set points, the SICM pipet detects the surface at a distance of order ri.8,9 We expect the pipet to have a conical shape and thus for ri to increase with breakage. Therefore, the detection distance would also be expected to increase, leading to an underestimation of the true breakage distance z by an amount ∼ z tan θ (see eq 3, below). However, this deviation should be negligible provided θ is small, which it typically is for SICM pipets.12,14 8981
dx.doi.org/10.1021/ac301851n | Anal. Chem. 2012, 84, 8980−8984
Analytical Chemistry
Technical Note
Figure 2. Estimation of the tip geometry by fitting breakage data. (a) Predicted variation of resistance with breakage for pipets of different values of the half cone angle θ. (b) Representative example of the model curve fitted to experimental data. (c) Summary results from n = 14 pipets suggest the tip geometry is reasonably consistent, with ri = 38 ± 6 nm and θ = 1.6 ± 0.2° (mean ± SD). (d) Contour map of goodness of fit, quantified by coefficient of determination R2, as a function of model parameters. Dotted contour indicates the region for which R2 > 0.99. Plot shown is for the data and fit in panel b.
■
RESULTS AND DISCUSSION Fitting a Resistance Model to Estimate Tip Geometry. Total resistance was modeled analytically as the sum of the internal pipet resistance, Rpip, and the external access resistance, Racc. The pipet tip was considered as a single truncated cone of filler solution with tip radius ri, base radius rb, and half cone angle θ. For filler with resistivity ρ, the resistance was then15 R pip =
ρ ⎛1 1⎞ ⎜ − ⎟ π tan θ ⎝ ri rb ⎠
broken pipets to compare the geometry away from the tip. The position of the break was not controlled but could be estimated by comparison with the unbroken geometry. Representative images are shown in Figure 3a−d. The pipet glass is an electrical insulator, so tips must be coated with a conductor such as gold before SEM imaging. This obscures the true surface position, introducing an error into the distance measurements. The effect is negligible at larger scales but is problematic for the very smallest part of the tip, where the geometry is critical. The sputter coater’s nominal coating thickness was 5−20 nm. Lacking more precise information, estimates were calculated using both ends of this range. We assumed the coating was uniform, so the same adjustment was made both to increase the measured pore radius and decrease the outer radius. Results calculated for the inner tip radius are plotted in Figure 3e, alongside the estimates from breakage data. SEM only shows the exterior of the tip and does not reveal the internal geometry. We obtained an approximate estimate of the internal angle by assuming that the proportional wall thickness is maintained in the neighborhood of the opening; the validity of this assumption is discussed below. Measurements of the exterior angle were assumed not to be significantly affected by the gold layer. Given a measured external angle θo and radius ro, the interior cone angle was calculated as
(1)
The external access resistance to the open pore was calculated using the formula of Hall:16 ρ R acc = 4ri (2) For conical geometry, breakage of the tip at an axial distance z would leave θ and rb unchanged and modify ri as ri(z) = ri + z tan θ
(3)
Predicted curves for the variation of resistance with breakage are shown in Figure 2a for several values of θ. It can be seen that resistance is highly concentrated at the tip and thus changes rapidly for small breakage distances. Provided that rb is relatively large (i.e., the pipet is much longer than the tip radius), the resistance contribution from the bulk of the pipet is negligible. For fitting purposes, the pipet length was assumed to be 1 mm, which is likely shorter than it ever would be in practice. Increasing the length had a negligible effect on the fitted values (data not shown). Model fitting to estimate ri and θ was performed for the reconstructed breakage data from a number of pipets pulled using the same program settings. A representative example is shown in Figure 2b, while summary results are shown in Figure 2c. The fit between model and experimental data was typically very close. Sensitivity analysis was performed by varying the model parameters and observing the effect on residuals. An acceptable fit (R2 > 0.99) was found only in a small region of the parameter space, suggesting that the optimization was reliable (Figure 2d). Comparing the Estimated Geometry with Electron Microscopy. To gauge the accuracy of the geometry estimates, we produced SEM images of similar pipet tips for comparison. (SEM and tip breakage are destructive, so it is not possible to apply both techniques to the same pipet.) In addition to the unbroken tips, a smaller number of SEM images were taken of
⎛ r tan θo ⎞ θ = arctan⎜ i ⎟ ⎝ ro ⎠
(4)
These results are plotted in Figure 3f, again alongside the breakage results. The parameters estimated from the SEM images agree well with those produced by model fitting provided the gold thickness estimate is not at the extremes of its range. Using a Mann−Whitney U test to compare the groups, the estimates for ri are not significantly different for gold thickness in the range 5−17 nm, while those for θ are indistinguishable for gold thickness in the range 7−19 nm. In estimating the internal cone angle from SEM images, it was necessary to assume a consistent relationship between the internal and external walls. If the true internal geometry deviated substantially from that observed externally, then any conclusions drawn could be wrong. Such assumptions are almost inevitable when attempting to estimate tip sizes from SEM images. A common heuristic is that the ratio of the outer diameter (o.d.) to the inner diameter (i.d.) is everywhere the same as for the original capillary glass.17,18 This rule does not appear to hold for our pipets (Figure 3g). 8982
dx.doi.org/10.1021/ac301851n | Anal. Chem. 2012, 84, 8980−8984
Analytical Chemistry
Technical Note
behavior of a SICM probe. We see no evidence of cone angle changes close to the tip in the breakage data. Thus, the transition appears to be gradual and/or remote, and the assumptions used to estimate the internal cone angle from the external SEM data should be sound.
■
CONCLUSIONS One of the great attributes of SICM lies in its ability to obtain images of cells and tissues at high resolution under physiological conditions. A second important feature is that high-quality probes can be manufactured by a very simple method, the heating and pulling of glass capillaries. However, these advantages can be substantially curtailed by the difficulty in obtaining information about probe geometry, which is central to estimating image resolution. Geometry measurement has traditionally used electron microscopy, which necessitates careful mounting and gold coating of the probe and is timeconsuming and prone to failure. Even when successful, the images obtained by this method show only the outside surfaces of the probe, which is helpful for determining the aperture at the probe tip but does not clearly define the inner geometry. Very recently, tip calibration via an AFM cantilever has been reported. By deconvolving the AFM probe geometry, this can provide an accurate measure of the pipet tip inner radius, and it may be integrated into the SICM system so as to make such measurement more routine.11 Further, AFM imaging is nondestructive and can also provide some measure of the orthogonality of the pipet to the sample plane. However, such hybrid instrumentation is not yet in common use, and the AFM tip has very limited access to the interior geometry. Because of the difficulties of measurement, many SICM studies give little information about the tip and instead estimate resolution based on the smallest objects apparent in the images. We have used a simple “zap” method to induce quasicontrolled collisions with the coverslip and record the variation of pipet resistance with breakage distance. This variation can be closely fitted to an analytical model of the geometry to estimate the original inner radius and cone angle, important determinants of the imaging capabilities. Comparison with SEM suggests the values obtained with our method are correct, as they agree when reasonable estimates for the thickness of the gold coating are made. Since SEM images are subject to uncertainty because of the need for estimating coating thickness, we suggest that our method is likely to be more reliable than SEM. The method is easy to implement and requires no additional apparatus or sample preparation. The model is simple to understand and validate, and source code for the analysis is provided. The breakage process is only quasi-controlled and may fail, for example, if the tip flexes too much or breaks too far. We have found the success rate with fine-tipped borosilicate pipets to be high, but it is possible that pipets with very different geometries or material properties might not produce sufficient data or may fit poorly due to violation of the model assumptions. However, our pipets are typical of those used for a wide range of SICM applications. The investment of time and resources to record the data is minimal, on the order of a few minutes per pipet. We suggest that the method could become a routine step for many users of SICM.
Figure 3. Estimating geometry with SEM: (a) SEM side view of an unbroken pipet tip (scale bar 500 nm); (b) SEM end view of an unbroken pipet tip (scale bar 100 nm); (c) SEM side view of a broken pipet tip (scale bar 10 μm); (d) SEM end view of a broken pipet tip (scale bar 1 μm). (e) Estimates of inner radius ri of unbroken tips, obtained from SEM images assuming a gold thickness of 5 and 20 nm, along with estimates from breakage data. (f) Estimates of internal half cone angle θ of unbroken tips, obtained from SEM images assuming a gold thickness of 5 and 20 nm, along with estimates from breakage resistance fitting. (g) Comparison of o.d./i.d. ratios of broken and unbroken tips, estimated from SEM images assuming gold thickness of 5 nm (top) and 20 nm (bottom). Dotted lines indicate the o.d./i.d. ratio of the original capillary glass.
For the unbroken tips, the unadjusted data suggests a much greater o.d./i.d. ratio. Of course our estimate of this ratio is dependent on the figure used for the gold thickness. Although this is uncertain, an unchanged o.d./i.d. ratio would require a gold coating of >20 nm, which is quite unlikely. Further, for the broken tips, the measured ratio is already less than that of the glass, and compensation for the gold layer can only decrease it. The estimated ratios in the two cases are probably distinct (Mann−Whitney U test, p = 0.035), suggesting that the idea of a preserved ratio is wrong. However, the differences are not large (1.6 ± 0.1 for the broken tips, 2.2 ± 0.3 for the unbroken, assuming 20 nm gold thickness). Further, comparison between the external profiles of the broken and unbroken pipets suggests that the breaks are >100 μm from the tip, and geometry variations at such a distance would not be expected to have any effect on the
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. 8983
dx.doi.org/10.1021/ac301851n | Anal. Chem. 2012, 84, 8980−8984
Analytical Chemistry
Technical Note
Notes
The authors declare the following competing financial interest(s): G.M. has a CASE studentship supported by Ionscope Ltd, UK, a small spin−out company manufacturing scanning ion conductance microscopes. M.C, S.D.L. and T.G.S. declare no competing interests.
■
ACKNOWLEDGMENTS This work was supported by funding from the Biotechnology and Biological Sciences Research Council and the Engineering and Physical Sciences Research Council. The authors would like to thank Pavel Novak and Yuri Korchev of Imperial College London for their invaluable contribution to the development of this manuscript.
■
REFERENCES
(1) Hansma, P. K.; Drake, B.; Marti, O.; Gould, S. A.; Prater, C. B. Science 1989, 243, 641−643. (2) Korchev, Y. E.; Bashford, C. L.; Milovanovic, M.; Vodyanoy, I.; Lab, M. J. Biophys. J. 1997, 73, 653−658. (3) Mann, S. A.; Hoffmann, G.; Hengstenberg, A.; Schuhmann, W.; Dietzel, I. D. J. Neurosci. Methods 2002, 116, 113−117. (4) Happel, P.; Hoffmann, G.; Mann, S. A.; Dietzel, I. D. J. Microsc. (Oxford, U. K.) 2003, 212, 144−151. (5) Novak, P.; Li, C.; Shevchuk, A. I.; Stepanyan, R.; Caldwell, M.; Hughes, S.; Smart, T. G.; Gorelik, J.; Ostanin, V. P.; Lab, M. J.; Moss, G. W. J.; Frolenkov, G. I.; Klenerman, D.; Korchev, Y. E. Nat. Methods 2009, 6, 279−281. (6) Rheinlaender, J.; Geisse, N. A.; Proksch, R.; Schäffer, T. E. Langmuir 2011, 27, 697−704. (7) Rheinlaender, J.; Schäffer, T. E. J. Appl. Phys. 2009, 105, 094905. (8) Del Linz, S. J. L. Scanning ion conductance microscopy: modelling and approaches to studying peptide secretion. Ph.D. Thesis, University College London, U.K., 2011. (9) Del Linz, S. J. L.; Willman, E. J.; Caldwell, M.; Klenerman, D.; Fernández, F. A.; Moss, G. W. J. submitted to Biophys. J. (10) Pellegrino, M.; Orsini, P.; Pellegrini, M.; Baschieri, P.; Dinelli, F.; Petracchi, D.; Tognoni, E.; Ascoli, C. Micro Nano Lett. 2012, 7, 317−320. (11) Pellegrino, M.; Pellegrini, M.; Orsini, P.; Tognoni, E.; Ascoli, C.; Baschieri, P.; Dinelli, F. Pfluegers Arch. 2012, DOI: 10.1007/s00424012-1127-6. (12) Ying, L.; Bruckbauer, A.; Rothery, A. M.; Korchev, Y. E.; Klenerman, D. Anal. Chem. 2002, 74, 1380−1385. (13) R Development Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2010. (14) Ying, L.; White, S. S.; Bruckbauer, A.; Meadows, L.; Korchev, Y. E.; Klenerman, D. Biophys. J. 2004, 86, 1018−1027. (15) Snell, F. M. In Glass Microelectrodes; Lavallée, M., Schanne, O. F., Hébert, N. C., Eds.; John Wiley and Sons: New York, 1969. (16) Hall, J. E. J. Gen. Physiol. 1975, 66, 531−532. (17) Corey, D. P.; Stevens, C. F. In Single Channel Recording, 1st ed.; Sakmann, B., Neher, E., Eds.; Plenum Press: New York, 1983; Chapter 3, pp 53−68. (18) Brown, K. T.; Flaming, D. G. Advanced Micropipette Techniques for Cell Physiology; John Wiley and Sons: Chichester, U.K., 1986.
8984
dx.doi.org/10.1021/ac301851n | Anal. Chem. 2012, 84, 8980−8984
