1. Introduction

Near-field scanning optical microscopy (NSOM) has made significant impact in many areas of optical physics [1]. One approach to such imaging is to use a cantilevered glass probes for a variety of optical measurements. However, the ability of these cantilevers to image living biological cells has been severely limited due to the mechanical stiffness of glass with force constants ranging from 1 to 10N/m. Nonetheless, optical and other functional imaging applications of live cells could be most significant.

To give a perspective of this and the goals of the present paper, it is important to cite the recent work of Ossola et al [2]. Ossola et al [2] have tried to hybridize patch clamping with AFM force control and topographic imaging. Limiting their advance was the inability to use glass cantilevers. As these workers noted [2], even though “the field of force-controlled glass pipettes was pioneered by Shalom et al.” [3] functional biological applications with such high force constant cantilevers has been impossible for the last 25 years.

Besides optical techniques such as NSOM which give correlated optical and structural imaging, force controlled scanning electrochemical microscopy (SECM) [4], scanning thermal microscopy (SThM), patch clamping correlated with nano topography [2] etc are all potential beneficiaries of the advances in this paper.

2. A review of retarding issues of AFM live cell imaging

It should be noted, that even topographic imaging with atomic force microscopy (AFM) has been retarded by other issues. For example, a force sensitivity of several hundred pN is needed for not penetrating the ~4nm fluid phospholipid bilayer thickness of a cell membrane and a few tens of pN are needed to image microvilli emanating from a cell membrane. The latter was not achievable until recently. Pioneering studies by Schiller et al [5] imaged microvilli on live Madin–Darby canine kidney (MDCK) cells with specialized silicon cantilevers with a 0.0611N/m force constant and a new geometry. They also employed a new algorithm to address issues, which limit laser beam deflection (LBD) force feedback with soft cantilevers. Specifically, with such cantilevers and feedback, discontinuities are seen as a probe tip approaches and retracts from a surface. To address such issues new algorithms have been constantly evolving, taking force sensitivity from the original 50nN regime of force volume techniques to the more recent methods of full digitization of the force approach curve and subsequent algorithmic deconvolution to reach beyond 100pN [6]. These advances, as noted herein, have been used by Schiller et al [5] to image the brush of microvilli emanating from the apical surface of MDCK cells [5].

There are also geometrical issues with silicon cantilevers. Schiller et al [5] were able to address three such issues in their new design of their cantilevers. First, the height and shape of a confluent monolayer of in vitro cultured epithelial cells, such as MDCK cells, which are known to be well-differentiated cells, is not uniform [5]. This poses a serious challenge for standard AFM tip lengths that can be on the order of 5 μm. To address this issue Schillers et al [5] developed their low force constant cantilevers (0.0611N/m) with a relatively large tip length of 17µm.

In addition, silicon probes are geometrically not angled and protruding outward imposing high pressure on the cell membrane. Schiller et al’s [5] new design had a cantilever/tip angle of 15° and a tip diameter (130nm) to reduce cell pressure and prevent cell penetration even with soft cantilevers. This geometry has greater potential for minimizing the significant issue with squeeze layers. This is an effect in which a generally straight probe tip squeezes out water layers and any protruding structures such as microvilli emanating from the cell membrane. An angular geometry somewhat alleviates this effect and helps reduce the shadowing effect often seen when silicon AFM cantilevers image cells. Nonetheless, the lack of high aspect ratios in these tips compromises deep penetration of the tip into the invaginations in a cell membrane and between cells.

Even with these advances damping of tapping mode AFM oscillations of a silicon tip in biological media reduces the Q (quality factor), i.e. sensitivity of the probe, to single digit values. Also, LBD only allows an estimate of the point of contact of the tip with the surface. This is a critically important parameter for quantifying cellular elasticity.

The present paper, where the focus is the use of cylindrical glass cantilevers, has the potential to overcome the remaining geometric and feedback limitations noted above for silicon cantilevers. Also, such probes permit a whole group of functional bioimaging applications that have been difficult in the past. In addition, glass probes with their cylindrical rather than flat cantilevers are resistant to damping. Furthermore, these probes have slender probe tips of over 100µm length with high aspect ratios and long penetration depths and allow adjustable opening angles.

3. The road to a solution

When such glass cantilevers are complexed to normal force tuning fork feedback many of these issues could be addressed.

In fact tuning forks were developed many decades ago due to the seminal work of Karai & Grober, who introduced such feedback for NSOM [7]. Due to the sharpness of resonance frequencies inherent in tuning forks in air and vacuum ultrasmall frequency changes can be detected and this permits ultra-sensitivity of the force interactions to the single picoNewton regime [8]. Even the force of single photon, 1.6pN, has been measured with an NSOM probe attached to a tuning fork based AFM/NSOM system [8], and this agrees well with theoretical calculations of the fundamental force sensitivity of tuning forks [9].

In fact, all probes used for either NSOM or solely topographic imaging in this study or in the previous studies mentioned above [7–9] are produced from glass. They have force constants in the range of ~5-10N/m. Tuning forks in the past have been viewed generally as a simple, convenient way, to obtain feedback for NSOM. This is certainly an important aspect of tuning forks. But, in this paper, an additional advantage is emphasized. Namely mounting the probes to get maximum Q (quality factor) or force sensitivity with such high force constant glass probes.

The approach that was used to get high Q factors is a result of on-line monitoring of the Q factor during the mounting and limiting the adhesive used to connect the probe to the tuning fork. In addition, for any amount of adhesive used on the tang of the tuning fork to which the probe is attached, additional adhesive is added to the opposing tang in order to balance the weight between the two tangs.

As a result of these efforts, Q factors (quality factors) in the 2000-5000 in air have been obtained for these probes. Such Q factors produce very sharp resonances. As a result, feedback based on frequency or Q can be chosen. For such feedback, frequency shifts of 0.25-0.5Hz from the resonance frequency of around 32KHz are typically chosen as the set point or error signal. The feedback returns the system to these chosen values after an alteration in force is detected due to scanning the topography and the resulting forces that trigger such feedback corrections are on the order <100pN. This is based on estimates published in reference [10] that Q values of 300-600 lead to <500pN forces.

However, in spite of these high Q factor probes, a general issue with the use of tuning forks in live cell imaging has been their inability to operate in aqueous solutions. In fact, an early approach that was suggested for NSOM glass probes was to insert only the tip of the probe in the liquid [11]. This approach is readily possible with glass probes due to their long tips with high aspect ratios. Nonetheless, in all the years since the publication of this approach in 2003 [11], there have been very few examples of imaging of living cells with NSOM.

To understand what has prevented progress in the approach of Höppener et al [11], careful analysis of the unsuccessful experiments in our laboratory allowed for the discovery that the root cause of the difficulties was uncontrolled variability in the Q of the tuning fork during imaging. This resulted in non-repeatable experimental conditions and unstable imaging. This was the case for the possible geometries of mounting of the NSOM probe to the tuning fork either in a straight (shear force) geometry as used by Höppener et al [11] or a normal force geometry also employed in some measurements reported in this paper based on the extensive and elegant methods developed by Giessibl et al. [12].

To better understand the problem, the Q of the tuning fork as function of the depth of insertion of the probe tip into the liquid was investigated. This is plotted in the graph shown in Fig. 1(a).

 

Fig. 1 (a) Functional dependence of quality factor, Q, with probe tip penetration into the liquid at room temperature. (b) Height reduction as a function of time, at room temperature of the liquid layer contained in the liquid cell due to liquid evaporation in an open cell without the suggested solution in (c); (c) A diagrammatic illustration of the liquid cell with a parting of the body of water or a “Moses Like Effect” generated by the wetting angle of the Perspex coverslip with an aperture surrounding the probe. This resulted in maintaining a small, stable probe tip penetration with a large Q factor and protected a reservoir of liquid on either side, with reduced evaporation. Such a development allowed a stable thin layer of liquid above the cells at the point where the probe tip images the sample. Thus, the region of the cell membrane stained with a membrane voltage sensitive dye Di-4-AN(F)EPPTEA remains hydrated for NSOM absorption/fluorescence and topographic imaging. The configuration of the liquid cell works for both straight and cantilevered glass probes.

Download Full Size | PDF

As can be seen in Fig. 1(a) there is a near linear reduction in the Q of the probe as the tip penetrates into the liquid. The data shown indicates that the Q of the probe dropped from 3,300 to under 2,900 by penetrating 100 μm into the liquid. This is a decrease of >10% for every 100 μm of penetration. The effect of Q factor reduction with penetration depth occurred for probes with numerous initial values of Q.

A very important corollary of this effect is seen in Fig. 1(b) where the evaporation in uncovered standard liquid cells used with tuning forks with a volume similar to the one used results in a nearly linear alteration as a function of time in terms of volume. This results in varying penetration of the tip of the probe in liquid. Critically associated with this variable depth of penetration of the tip was a continually varying Q factor and extreme feedback instability even at room temperature where the measurement of Fig. 1(b). This would certainly have been considerably larger at a temperature of 37°C which is needed for general cell viability. The resulting conclusion was that the layer of liquid above the cells at the point of tip/membrane contact had to be very thin while maintaining an appropriate liquid reservoir to keep this thin layer unchanged in spite of evaporation.

On face value this appears drastic. However, learning from standard AFM imaging it was realized that AFM liquid cells employ a very small volume of liquid and evaporation is reduced significantly by a cover slip required above the probe. Historically, such a requirement evolved for conventional laser feedback due to the need for laser beam reflection off the probe cantilever which had to remain stable. This was impossible without a transparent cover above the probe. The solution prevented a continual changing of the surface of water which caused noise in the reflection. Thus, a sandwich geometry of glass, liquid and sample arose in LBD feedback in liquids. Even though this is not essential for tuning fork feedback, such a geometry complexed with the ideas of Höppener et al [11] has led to a resolution of the problems that were faced in this research with tuning fork glass cantilever combinations. The solution is shown in Fig. 1(c).

Since glass probes, unlike silicon cantilevers, can be prepared with very long tips, as long as 2mm, one can readily develop a hybrid solution that takes the best of conventional AFM liquid cells with the specific requirements of tuning fork based glass probes. As seen in Fig. 1(c) the solution was a liquid cell with a thin liquid layer, ≤800µm covered with a hydrophobic Plexiglass coverslip having a thickness of 300µm. The aperture in the coverslip created a separation in the body of the liquid due to the wetting angle of the hydrophobic coverslip around the aperture. This has allowed for uninterrupted stable imaging for up to 2 hours in physiological media.

To put this development in context of previous approaches, both the frequency of the tuning fork [13,14] and the Q factor [14,15] as a function of the depth of penetration in liquid of the probe has been reported previously but none of these studies emphasized the variability in imaging parameters that could result from evaporation. Nonetheless, it is important to note that long before these studies and a year after the work of Höppener et al [11], van Hulst and Garcia-Parajo introduced a 'diving bell' method for NSOM liquid imaging [16]. Even though the early papers did not emphasize live cell imaging, a later investigation measured fluorescence correlation spectra at a point on living cells using the sub-wavelength illumination of an NSOM probe [10]. In this study shear force feedback was employed and Q factors were reported between 300 and 600. As noted by the authors, this allowed for recording stable fluorescence for about 30sec. Thus, imaging was not reported.

As indicated by references 13–16 published 7-10 years after the innovative introduction of the 'diving bell' idea, problems in liquid cell imaging with live cells continued. It is assumed from the present study but not reported in any of the above publications that the Q factor variability along with the initial Q factor of the tuning fork after mounting the probe has led to the lack of live cell imaging.

Very recently, Park et al [17] have used the diving bell idea but emphasized a new way of mounting the probe in a shear force geometry which was based on an active control of the mechanical resonance by adjusting the position of two nodal wedges along the shaft of the probe’s attachment to the tuning fork [17]. With such an approach they were able to achieve a Q in liquid of 2800. However, there is no report in this paper of live cell imaging.

4. NSOM topographic and optical imaging of living cell membranes

We have achieved live cell imaging with the new approaches developed in this paper. In addition to the above, an important parameter has been the mounting of the probes. In the mounting, the amount of glue touching the tuning fork has been limited and this along with the specialized liquid cell allows for Q factors in liquid that are more than sufficient for live cell imaging. Both these developments have allowed for the imaging of live cells with a stable Q of 2000 in liquid for up to 2 hours at room temperature where MDCK cells are stable.

The result of these advances is topographically correlated NSOM imaging of live cells with NSOM probes mounted in the shear force tuning fork configuration. Shown in Fig. 2 are NSOM images, which can now be correlated with AFM topographic imaging of MDCK cells obtained using a straight NSOM probe mounted in the shear force configuration. The probe had a 1.5mm tip length with a high enough Q factor of ~2000 in liquid. In air the same probe had a Q factor of ~2,800. In both in the shear force and normal force geometry when it is employed the amplitude of the oscillation of the tuning fork is considerably less than 5nm and can be as little as 1nm.

 

Fig. 2 Comparison of four imaging modes of MDCK cells clearly showing different contrast & complementarity of the information. Topographic imaging and simultaneously obtained NSOM absorption (a, b) and fluorescence imaging (d, e). All NSOM (b, c and e) & scanning confocal fluorescence image (f) were obtained on cells stained with a membrane bound voltage sensitive dye, Di-4-AN(F)EPPTEA. The NSOM absorption image (b) shows defined contrast of dark dots surrounded by bright regions. These dark dots due to absorption are interpreted as membrane emanating microvilli filled with dye (see text). Similarly, in the NSOM absorption image (b) a large dark circular region is correlated with a large protrusion in the topography (a). This is indicated with a blue arrow. It is interpreted as a cilium which is known to be seen in such cells and would have a larger membrane surface area and therefore more absorption. The NSOM fluorescence (e) obtained in a second scan of the same location shows opposite contrast. The dark cilium in absorption is now a bright fluorescent region. A blue arrow has been placed between (d) and (e) indicating this. To further guide the reader a gold arrow correlates a depression in the topography with a dark region in the fluorescence. Contrast reversal is also seen for the dark dot in (b) interpreted as a microvillus which is correlated with a bright dot in (e). A green arrow has been placed as a guide. The laser scanning confocal fluorescence imaging, which is not surface centric as is the case in NSOM, shows different contrast complementing the NSOM results. A region of the absorption image at higher resolution is shown in (c). The line scan shows an edge sharpness of 150nm. Other imaging parameters include: a straight NSOM probe mounted in shear force geometry, Q-factor in liquid 2000, scan area 40µm X 40µm, 12msec/pixel with 488nm laser excitation having an input power of 12mW.

Download Full Size | PDF

For both the NSOM and laser scanning confocal imaging in Fig. 2, the cell membrane was labeled with a membrane anchored, membrane potential sensitive dye Di-4-AN(F)EPPTEA [18]. This dye is amphiphilic. in other words, both hydrophobic and hydrophilic chemical structures are part of the molecular structure of the dye. This dye contains two long hydrophobic aliphatic carbon chains that act to anchor and embed the dye in the upper leaflet of the bilayer membrane. The dye chemical structure, that is attached to this aliphatic section of the molecule, is hydrophilic. And, even though this hydrophilic region should emanate from the cell membrane into the space outside the cell it is actually partially sunk into the upper membrane leaflet due to chemical factors [18].

The NSOM imaging shown in Fig. 2 was recorded in transmission mode and correlated with the simultaneously obtained image of the AFM topography. Depending on the filters employed either a transmission absorption or fluorescence image was recorded. NSOM is a membrane centric optical imaging method with a depth of field that is only about 50nm [19]. Therefore, these images highlight the dye absorption in the plasma membrane. With such absorption contrast, that is hard to achieve in confocal imaging, dark areas represent higher absorption or higher dye concentration while bright regions represent lower absorption and smaller concentration of dye in the path length of the light.

A good example in this image is the region highlighted by a blue arrow. The AFM of this region shows a structure, which is approximately 1-2µm high. MDCK cells exhibit primary cilia which would be protrusions in the µm range depending on the age of the culture [20]. These cilia are plasma membrane protrusions with a larger membrane area that would be expected to concentrate a dye embedded in the cell membrane. However, one cannot exclude such dye filled structures being other protruding membrane structures such as microvilli. Against this deduction is the height seen in the correlated AFM which is not consistent with microvilli (see below).

A comparison of the NSOM [Fig. 2(b)] and AFM topography image in the region of the protrusion [Fig. 2(a)] shows a large absorption cross section well correlated with the topographic feature. Such correlation of optical imaging and topography gives confidence in the assumption that this feature is a protrusion from the cell membrane. It is very difficult to observe such cilia by confocal imaging with simple plasma membrane dye staining probably due to out-of-focus light. Thus, it is assumed that the contrast seen in the confocal image in Fig. 2(f) is probably due to internal structures not seen in NSOM.

Also supporting this interpretation is the NSOM fluorescence [Fig. 2(e)] with its correlated topography [Fig. 2(d)] that arises from essentially the same field of view seen in Figs. 2(a) and 2(b). In fluorescence, a protrusion that appears dark in the absorption image of the membrane [Fig. 2(b)] would be bright in the fluorescence image [Fig. 2(e)]. Thus, the opposite contrast seen in the NSOM fluorescence is completely consistent with our explanation of a large amount of dye in a protruding structure due to a local membrane surface area increase. Interestingly, below the bright spot of fluorescence there is a dark region shown by a gold arrow in Fig. 2. This correlates well with a depression in the AFM topography where, if dye is in fact in this region, it would be difficult for the NSOM probe to track this invagination into the cell and excite the fluorescence in the near-field.

In addition to such cilia, MDCK cells are known to have microvilli which are much shorter membrane protrusions also emanating from and distributed over the cell surface [21]. As noted in the introduction, the limitations of the force sensitivity of laser beam feedback had prevented effective imaging such microvilli until recently [5]. An inspection of the membrane centric absorption contrast NSOM image shows dark dots surrounded by regions of brighter contrast. It is reasonable to assume that these regions could be due to increased membrane area at these points where there would a larger concentration of embedded dye due to the microvilli protrusions. The distribution of such dots is also consistent with microvilli distribution.

The XY dimension of these dots as measured in the NSOM, namely <1µm, is larger than such microvilli and probably represents the imaging of the NSOM aperture which was 250nm. Support of this interpretation of microvilli comes from the NSOM fluorescence that shows a bright rather than a dark spot in the image and this is highlighted by a green arrow in Fig. 2.

In terms of an edge sharpness of 150nm obtained with a 250nm aperture this is a result of moving the sample in steps of less than 250nm. Thus, even when a feature is only partially in the aperture of the probe a signal alteration could be detected if the signal to noise is high enough. Thus, with APDs, that are low background noise detectors, such sub-aperture movements can be detected with high statistical confidence. Therefore, the imaging resolution of an NSOM probe, whose illumination is a top hat rather than a Gaussian profile, can be higher than the aperture diameter.

As would be expected the contrast in the laser scanning confocal fluorescence image is quite different with the same dye staining. This is due to the fact that membrane centric imaging of NSOM is not possible in confocal which is more sensitive to the internal structures of the cells where the invaginated dye stained membrane has stained the cytoplasm. This combination of images highlights the complementarity of NSOM and confocal imaging.

The stability of the MDCK cell culture during the imaging protocols reported in this paper is an essential element of the present work. As is noted above, such cells live generally for upto two hours at room temperature. But, we have seen cultures in a MEM-Hepes medium behaving normally at ambient atmosphere for even several hours.

This is exemplified by the ability of these cells to be maintained in this medium to form a tight cell monolayer, as demonstrated by our group [22].

The nature of such living MDCK cells and their stability is initially assessed by on-line optical microscope inspection of their morphological structure before and after NSOM/AFM or AFM imaging. Cell death is generally indicated in this cell line by a breaking of the characteristic mosaic structure of the confluence of cells. In addition, as a result of death, cells start floating in the liquid. None of this was observed for any of the cell cultures for which imaging is reported in this paper.

However, in addition to this general indication of cellular integrity which is qualitative, we have used in this study a dye, sensitive to membrane potential, to stain the cell membrane. Using this dye, we also quantitatively show in Section 6 of this paper that cells, subjected to identical scanning probe imaging protocols, show membrane potential changes characteristic of live cells. Thus, both in terms of the mosaic morphological integrity of the cells in the optical microscope and, in terms, of the membrane potential changes that have been imaged by a glass scanning NSOM probe in liquid the cells stay alive during the imaging process. This is uniquely allowed in this paper by the protocols developed and discussed below.

Such a metric of assessing living cells in scanned microscopic protocols has not been reported previously for either NSOM or AFM of cells. This includes reference 5 which recently reported live cell imaging of MDCK cells with specialized silicon cantilevers and specialized digitized algorithms for force sensing. As indicated in reference 5, the structural imaging of membrane protruding cilia is also an indication of MDCK cell integrity and this is shown in the next section of our paper.

5. Fine topographic imaging of microvilli

The NSOM imaging in Fig. 2, which had Q factors of upto 2000 should have had sufficient force sensitivity to image topographically the microvilli. However, an NSOM probe has, as a result of its metal coating, outer diameters too large for this imaging task. To further verify this and assuming that such tuning forks should have the sensitivity to topographically image thin structures such as microvilli directly and without the need for digitization or special algorithms, a large force constant, 30nm outer diameter AFM glass probe attached to tuning forks was employed using the liquid cell developments in this paper. The resulting image with a cantilevered probe mounted in normal force geometry is shown in Fig. 3. The image clearly shows that tuning fork feedback is highly sensitive and capable with such glass probes with their long slender tips to image microvilli in live cells in solution directly and without any complex digitization or algorithms. The data shown in Fig. 3 is the raw data. The dimensions in terms of length and width of the cilia we have imaged are very similar to those recently detected in reference 5 with ultra low force constant silicon cantilevers and new digitization algorithms to achieve ~100pN forces.

 

Fig. 3 (a) Imaging live MDCK cell microvilli with a cantilevered glass probe attached to a tuning fork and mounted for normal force imaging. Other imaging parameters were a Q factor in liquid 5000, scan area 30µm X 30µm, 12msec/pixel. A reduction of a few hundred in the Q factor from air is seen in liquid depending on the depth of penetration of the probe. (b) A line scan showing the resolution of the image. (c) and (d) Width and length of microvilli.

Download Full Size | PDF

6. Imaging the membrane potential

All indications of the behavior of the cells during such imaging indicated that the cells had remained alive. This was deduced by qualitative observations, namely, that the cells stayed well adhered and maintained normal morphology. However, NSOM fluorescence of cell membranes has the unique ability to report the voltage of the cell, with the membrane anchored potential sensitive dye staining of the membrane that was used. This can quantitatively verify our assumption that the cells are indeed alive during such imaging.

If indeed the cells are alive the intensity of the fluorescence of the dye should change if a reagent is added to depolarize the cell membrane. In Fig. 4, MDCK cells stained with Di-4-AN(F)EPPTEA, were imaged with laser excitation at 514nm which is the red tail of the absorption. As a result of such red tail excitation it is expected that a large alteration would be detected if the extinction changed as a function of the change in membrane voltage around this dye. This effect is due to a fast Stark effect response to membrane potential for such dyes [18]. In fact, depolarization of the cell under these conditions should result in a general decrease in the fluorescence intensity due to an electrochromic shift in the absorption of this dye. This is clearly seen in a point measurement with the NSOM probe where an approximately 20% change was observed (see Fig. 4(a)).

 

Fig. 4 (a) Monitoring and (b, c) Imaging cell depolarization with addition of 5mM KCl using NSOM fluorescence imaging of Di-4-AN(F)EPPTEA stained membrane. Imaging parameters included a straight NSOM probe mounted in shear force geometry, Q-factor in liquid 300 (in air: 2100), scan area 50µm X 19µm, 12msec/pixel with 514nm laser excitation having an input power of 12mW.

Download Full Size | PDF

This point measurement was recorded with the same NSOM probe used to form the image in Fig. 4(b).

Before and after addition of KCl the fluorescence intensity is very stable. This is the case even though with the NSOM tip there should certainly be bleaching at the point of illumination. This is the case even though the intensities through the tip are low. However, as is known, the membrane is fluid and the diffusion time of the dye is sufficiently fast and the bleaching is sufficiently low so that there is effective replenishment of any bleached molecules.

The robust change in the fluorescence due to 5mM KCl addition could also be indicative of the nature of the NSOM measurement. The depth of focus of NSOM, as noted, is very low and thus this measurement which is membrane centric has little out-of-focus contributing noise. This is also seen in the images obtained before and after the KCl addition. If one looks at the brightest pixel in the image before and after KCl there is a considerable reduction seen. However, such a measurement is not as telling as the point measurement since the brightest pixel in the whole image could result from some region of aggregated dye and does not have the average view over time that the point measurement allows.

One caveat on the fluorescence imaging measurements is that the images shown in Fig. 2 were obtained with 488nm excitation which is the peak of this membrane embedded dye’s absorption and thus a relatively small NSOM aperture could be used. However, this wavelength is not effective for the measurement of alteration of fluorescence intensity with depolarization for this voltage sensitive dye since it is not at the red edge of the absorption of the dye [18]. For such measurements it is required to excite in the red tail of the dye’s absorption. However, at this wavelength of 514nm the extinction is low. At least at this stage in our development this required a large probe tip of ~1µm for sufficient intensity to measure the membrane potential change. Thus, the detail in the fluorescence image in Fig. 4 is not what could be seen with 488nm excitation at the peak of the dye absorption. Nonetheless, it should be noted that even for such a large tip the diffraction is very large and the depth of penetration is very low. This is supported by earlier studies done in our laboratory where large probe tips were used to effectively record fluorescence correlation spectra of rhodamine in solution indicating that, in spite of the large XY extent, the Z extent due to diffraction was very small allowing for even 100nM FCS data to be recorded [23].

Another aspect of the imaging is that the NSOM probe used in Fig. 4 had a very low Q factor of 300 because of its size. Manzo et al [10] used probes with similar Q factors and were only able to image at a point for 30secs in a living cell. It is possible that the difference in our ability to use such probes could be caused by the structure of the liquid cell developed in this paper which minimizes the effect of evaporation and the alteration in the Q of the probe due to alterations in the penetration depth.

In summary, this paper has achieved correlated topographic and optical imaging with an AFM controlled NSOM probe on live MDCK cells. This has also permitted with such probes the sensing and the imaging of membrane potential with good signal to noise ratio, highlighting the membrane centric nature of NSOM imaging. Such sensing proves beyond doubt that the cells being imaged with such NSOM probes are alive and remain so during the imaging. It is also demonstrated that with an appropriate diameter glass probe it is possible to image microvilli emanating from the cell surface of these living cells without any need for digitization or complex algorithms that are associated with LBD methods of live cell imaging. In comparison with the recent reported advance of imaging, such microvilli in this same cell line by LBD techniques [5], the tuning fork high force constant glass probe images are certainly competitive. The solutions developed as part of this paper for maintaining high and stable Q factors during liquid cell imaging with glass cantilever complexed tuning forks should contribute to other AFM ultrafine imaging tasks. The results of this paper clearly give great hope for the use of NSOM and other glass based imaging techniques in cell biology, which many workers have attempted in the past with variable results. The results allow for new horizons of functional biological imaging with scanned probe microscopy.

7. Materials and methods

Cell culture

For our research we used MDCK II cells. These are the same cells that have been the subject of the recent investigations of Schiller et al [5] to demonstrate the significant progress they made in imaging surface structures of live cells, such as microvilli. These cells are excellent both in terms of comparing our results with large force constant probes with those of Schiller et al using ultrasoft cantilevers. Furthermore, these cells allow one to address many important biological questions and the cell line is a model system for various diseases that extend from kidney disease [24] to cancer [25]. Finally, they can be differentiated to highly polarized epithelial cells under specific cell culture conditions.

MDCK cells were cultured routinely as described previously [26]. For AFM experiments, cells were seeded in a density of 100,000 cells on 16mm diameter glass bottom Petri dishes and cultured for 5 days in growth medium, until forming a continuum cell monolayer. Before measurement, the growth medium was exchanged for MEM Hank's salts medium lacking phenol-red, supplemented with 15 mM Hepes, pH 7.5 (MEM-Hepes medium).

For the NSOM and laser scanning confocal imaging the cell membrane was stained with a membrane bound, voltage sensitive dye, Di-4-AN(F)EPPTEA. This was accomplished by incubating the dye with the cells for one hour at 37°C.

AFM and NSOM imaging

AFM and NSOM imaging was accomplished with an MV2000 (Nanonics Imaging Ltd., Jerusalem, Israel). The MV2000 was placed on a custom dual microscope based on an Olympus BXM free space microscope. The probes used for both NSOM and AFM that were suitable for this SPM/NSOM system employed. The probes were obtained from the same manufacturer as the SPM/NSOM system which was Nanonics Imaging Ltd, Jerusalem, Israel. A description of the mounting of the glass probes on tuning forks for achieving high Q factors is in the main text of the article. For the NSOM imaging an argon ion laser at 488nm and 514nm was employed. The photodetector used was an avalanche photodiode (Perkin-Elmer SPCM-AQR-14) and a photon counting module (Nanonics APD Photon Counter). All imaging was done in transmission mode. For the fluorescence imaging we used a 542nm long pass filters. For the absorption image, we do not used filters.

Laser scanning confocal

Images were taken with a confocal microscope (Olympus FV-1000) using a 60x oil immersion objective with 1.42 NA (numerical aperture). For excitation, a 488nm scanning laser was used and the fluorescence signal was collected at the range of 500 to 680nm for Di-4-AN(F)EPPTEA. Z-scans were performed for each worm and in every measurement reported. The average fluorescence intensities from the Z projections were produced from every scan using ImageJ software.