Proc. Nati. Acad. Sci. USAVol. 89, pp. %32-9636, October 1992Biophysics

Scanning tunneling microscopy imaging of Torpedoacetylcholine receptor

(synaptic membrane/surface dimensions/central cavity)

A. BERTAZZON, B. M. CONTI-TRONCONI, AND M. A. RAFTERY*Department of Biochemistry, University of Minnesota, 1479 Gortner Avenue, St. Paul, MN 55108

Communicated by David S. Eisenberg, May 15, 1992 (received for review July 19, 1991)

ABSTRACT The synaptic surface of the acetylcholine re-ceptor in membranes from Torpedo calzfornica electric organhas been imaged by anig tunneling microscopy. The mol-ecule appears pentameric, with one major and four minorprotrusions rising above the surface, and these protrusionsencompass a large central cavity. The outer diameter of themolecule is 69 ± 10 A, while the diameter of the cavity,measured at the widest complete contour line delmiting theopening, is 26 ± 7 A. The images and dimensions obtained areconsistent with the structure determined from hybrid densitymaps obtained by x-ray diffraction and electron microscopy.Thus, scanning tunneling microscopy can be used to obtainoverall dimensions and low-resolution structural features ofthesurface of a membrane-embedded protein.

Given the difficulties encountered in preparing crystals frommembrane proteins suitable for x-ray or electron diffractionanalysis, it is worthwhile to explore other methods forobtaining structural information on these rather intractablesystems. The introduction of scanning tunneling microscopy(STM) for imaging of biological systems (1) can yield struc-tural data at reasonably good resolution. This method iscapable of atomic resolution for smooth and highly conduc-tive surfaces of metallic compounds but not, at least now, forbiological samples, which are considered poor conductors.Many such studies are currently under way (2). We wished totest the application ofSTM to biological membranes by usinga system for which some characteristic features of surfacestructure are known. The system should ideally be composedof a large protein embedded in a biological membrane that ishomogeneous with respect to protein composition. Suchcriteria are met by postsynaptic membrane preparations fromTorpedo species such as Torpedo californica because theycan be readily prepared (3, 4) and have been shown to containthe nicotinic acetylcholine receptor (AChR) with minimalcontamination by other proteins (5).

This AChR is a large protein complex (270 kDa) composedof five subunits (6, 7) with the stoichiometry a213y6 (6, 8). Athree-dimensional model of Torpedo AChR, based on densitymaps obtained from low-dose electron microscopy imagesand x-ray diffraction at 12.5-A resolution, has been described(9). The AChR protrudes 55 A from the outer surface of thelipid bilayer and the average diameter of this extracellularpart (50-60%o of the total mass) of the AChR molecule is65-80 A (10). The view from above the membrane ofa samplenegatively stained for electron microscopy shows a rosettewith a central cavity of diameter 25-30 A (11, 12). The mapfurther shows that five electron-dense regions make up therosette, possibly corresponding to the five constituent sub-units (9), with a characteristic protrusion of density risingabove the other regions (10). This preparation of a large

FIG. 1. STM image of AChR-rich membrane fragments fromelectric organ of T. californica. Two receptor molecules (arrows) arevisible. The original image was taken by scanning a field of230 x 230nm at a probe scan rate of 1400 nm/sec, sampling time of 18 jusec,tunneling current of 0.56 nA, bias voltage of 32 mV, and integral andproportional gains of 2.0 and 3.0, respectively. The enlargement hasbeen filtered using the x, y plane fit and the flatten functions. Colorscale is in nanometers. Backgrounds of all preparations revealed onlythe atomic structure of the support.

integral membrane protein whose dimensions and generalstructural features are known represents a suitable system forapplication of STM to study the surface structure of amembrane protein. We report here images obtained fromdifferent preparations of T. californica AChR in enrichedmembranes, an analysis of the outer surface dimensions, anda comparison ofthe results with those obtained from electronmicroscopic and x-ray analysis studies.

MATERIALS AND METHODSMembrane Purification. AChR-rich membranes were pre-

pared from T. californica electric organ (13) and treated at pH11 (14, 15) to remove extrinsic membrane proteins. Theprotein subunit composition was assessed by SDS/PAGE(16). Binding activity was determined by use, of 125I-labeleda-bungarotoxin and a DEAE disk assay (17). Octyl f3-D-glucopyranoside (1.0%, wt/vol) was used to solubilize theAChR, followed by centrifugation at 100,000 x g for 1 hr toeliminate undissolved debris. AChR preparations werestored at 40C and used within a few days.

Abbreviations: STM, scanning tunneling microscopy, AChR, ace-tylcholine receptor; HOPG, highly oriented pyrolytic graphite.*To whom reprint requests should be addressed.

%32

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Apr

il 25

, 202

1

Proc. Natl. Acad. Sci. USA 89 (1992) 9633

FIG. 2. Three-dimensional view of a single AChR molecule. (Upper) A filtered three-dimensional view of a single AChR molecule. (Lower)Top view of an unfiltered (Left) and a filtered (Right) image. Flattening and plane fitting reduce the sharpness of the image and result in a lossof detail, but they provide consistent images for determination of the average dimensions reported in Table 1. A pentagonal shape, with a typicalprotrusion and a central cavity, can be recognized. The lowest part of the surface is consistently localized on both sides of the major peak (bottomof the image). The original image was obtained at 50-mV bias voltage, 1.9-nA tunneling current, and scan rate of 3.2 cycles per second on afield 250 x 250 nm (1600 nm/sec). Color scale is in nanometers. In membrane preparations poor in AChR content no images were observedexcept for uninterpretable structures.

Instrumentation. Scans were performed with a NanoscopeII (Digital Instruments, Santa Barbara, CA) equipped with asize D head, 0.8-pim scanning field. Probes were cut from0.25-mm-diameter Pt/Ir (80:20) wire with sharp scissors orpurchased from Digital Instruments (Nanotips). Highly ori-ented pyrolytic graphite (HOPG) was purchased from UnionCarbide (Cleveland) (10 x 10 x 0.2 cm).Sample Deposition. AChR-enriched membranes were sus-

pended in 10 mM Tris HCl at pH 7.4 or pH 9.5. Membraneswere stored at pH 7.4 and the pH was adjusted to 9.5 not morethan 30 min before deposition onto HOPG. All samples werediluted to a protein concentration of0.1-0.3 mg/ml in 10 mMTris HCl. Membrane deposition was achieved by loading 5 Alof the solutions onto a clean, newly cleaved surface ofHOPGand spread over an area of about 0.2 cm2.Snning Parameters. All scans were performed in air at

room temperature. The quality of the tip was checked by itsability to properly image the graphite background. Typicalstarting bias voltages were 30-50 mV, with the positivedirection toward the tip (current flow from sample to tip).Large fields (1-5 ,m) were scanned at a rate of 2-5 cycles persecond using a bias voltage between 20 and 100 mV, withtunneling currents ranging from 0.4 to 2 nA, integral gain setfrom 2 to 6 and a proportional gain ranging between 3 and 9to maintain a constant ratio of 1.5 with the integral gain; thetwo-dimensional gain was 0.05. Data analysis and image

filtering were performed with software provided by DigitalInstruments. High-frequency noise was removed by using thelowpass function, image curvature was removed by using theplane fit function, which removes, in both x and y directions,a least-square fit first-, second-, or third-order polynomialfrom the data. Where the sample is imaged there is no visiblebackground. In areas with no sample, the HOPG can beclearly imaged at 10-nm field size; resolutions at 0.8-nm fieldshowing the typical graphite hexagon are obtained only inareas of the graphite surface where the sample was notdeposited.

Table 1. Average dimensions of the AChR (A)This work* Ref. 6 Ref. 7

Outer diameter 69 ± 10 (12) 65 72Inner diametert 26 ± 7 (15) 30 25Majorpeak* 9 ± 4 (13)*Dimensions were obtained from filtered images and are expressedin angstroms (mean ± SD). In parentheses is the number ofmolecules used. The images are stored on discs and can be furnishedto readers who are interested.tMeasured from contour maps (as in Fig. 3) and corresponding to thelargest diameter measured on the first complete contour line de-limiting the central cavity.*Height of the major peak measured from the first contour line.

Biophysics: Bertazzon et al.

Dow

nloa

ded

by g

uest

on

Apr

il 25

, 202

1

9634 Biophysics: Bertazzon et al.

:~~~~~~~ 1:

3 +

1

I A .-LI

FIG. 3. Subunit localization on unfiltered images of single AChRmolecules. The general pentameric geometry observed in filteredimages is generated by five separate peaks on the surface. (Upper)Enlargement from a 193 x 193-nm field, obtained with a bias voltageof 18 mV, a tunneling current of 1.9 nA, a scanning rate of 1.3 cyclesper second (650 nm/sec), and sampling of 18 lsec. The contour linesare spaced 0.3 nm in the vertical dimension. Color scale is innanometers. (Lower) Image obtained from a 230 x 230-nm field withinstrumental settings as above and a scanning rate of 2.7 nm/sec.Five areas can be distinguished.

Receptor Stability. The conditions used for STM sampledeposition were also employed for binding and structuralexperiments. Receptor suspensions (150 IlI) in 10 mMTris-HCl were allowed to dry on HOPG surfaces. Driedsamples were solubilized in 150 ,u of 10 mM Tris-HClcontaining 1% octyl ,B-D-glucopyranoside. Control experi-ments were performed with the same preparation simplysolubilized by adding octyl f-D-glucopyranoside to a finalconcentration of 1%, without deposition or drying.

Circular Dichroism (CD). Spectra were taken with a JascoJ 41 C spectropolarimeter (Jasco, Easton, MD) interfacedthrough an Adalab A/D converter (Adalab-PC InteractiveMicroware, State College, PA) to an IBM-XT compatiblecomputer. The instrument was calibrated with (+)-10-camphorsulfonic acid. Water-jacketed thermostatted quartzcuvettes with 0.1- and 1-mm path lengths were obtained fromHellma (Forest Hills, NY). Optical activity, expressed asmean residue elipticity (6) in degrees-cm2 dmol-1, was ana-lyzed by means of a constrained nonlinear least-square fit(18).

190 200 210 220 230 240 250 260

nanometer

FIG. 4. CD spectra of resuspended, solubilized AChR. Mem-brane-bound AChR was deposited onto graphite and allowed to dry.After 30 min (curve 2) and 3 hr (curve 3) the AChR was solubilizedin 10mM Tris-HCI containing 1% octyl 3-D-glucopyranoside. Curve1 was obtained from a membrane preparation in 10 mM Tris-HClsolubilized by addition of 1% octyl 3-D-glucopyranoside. Valuesobtained from analysis of the spectra are shown in Table 2.

RESULTS

When AChR-enriched membranes were deposited on theHOPG substrate at pH 7.4 they had a tendency to aggregatein layers 100-150 nm thick and were readily swept away bythe scanning tip, with formation ofnew aggregation patterns.However, when the sample was deposited at pH 9.5 suchaggregates were not observed and the membranes consis-tently formed layers 30-50 nm thick. An enlargement of a 15x 15-nm scanning field is shown in Fig. 1. The image was

taken at a scanning rate of 1400 nm/sec (3.2 cycles persecond; field, 230 x 230 nm), at a bias voltage of 32 mV, atunneling current of 0.56 nA, integral and proportional gains2.0 and 3.0, respectively, and the enlargement was filteredusing the plane fit (automatic x and y) and flatten functions.Two rosette-like structures, protruding 30-40 A from themembrane surface, were observed (center and upper right inFig. 1) with an irregular surface and typical asymmetricprotrusions.A three-dimensional view of a single receptor molecule is

shown in Fig. 2 Upper. These images can be filtered asdescribed above, and this results in a loss of detail but allowsa standardized determination of average dimensions (Fig. 2Lower). The dimensions obtained on several different sam-ples are presented in Table 1. The average outer diameter is69 ± 10 A, while the central cavity, taken on contour mapsoffiltered images at the largest delimiting line, is 26 ± 7 A, andthe major protrusion, calculated from the lowest point of thereceptor surface (typically on the side of the major peak), is

Table 2. Secondary structure analysisSample a-Helix p-Sheet (-Turn Random coil

This workControl membranes 38.6 32.1 1.2 28.130-min drying 33.2 33.4 0.8 32.63-hr drying 32.6 38.1 0.6 28.7

Literaturea* 34 29 0 37bt 35 24 10 31ct 22 41 8 29The analysis was carried out with the algorithm ofChang et al. (18).

*From Moore et al. (19)tFrom Wu et al. (20)tFrom Mielke and Wallace (21), referred to the constrained least-square fitting according to Chang et al. (18)

25000

15000

2-

c(1)

5000

-5000

-15000

-25000

Proc. Natl. Acad Sci. USA 89 (1992)

Dow

nloa

ded

by g

uest

on

Apr

il 25

, 202

1

Proc. Natl. Acad. Sci. USA 89 (1992) 9635

Table 3. Binding of 1251-labeled a-bungarotoxin tosolubilized AChR

Drying time, min Binding, % control n

15 82 ± 7 430 87 ± 8 460 79±11 4180 92 ± 4 4

One hundred fifty microliters of a receptor suspension, containingprotein at 0.3 mg/ml, was allowed to dry on a newly cleaved HOPGsurface. At defined time intervals the dried sample was solubilized inthe same volume of 10 mM Tris-HCI, pH 7.4/1% octyl 3-D-glucopyranoside. Binding assay was performed as described (17).Control binding was obtained with solubilized membranes and was2-4 nmol of 125I-a-bungarotoxin binding sites per milligram ofprotein.

9 ± 4 A. These data, presented in Table 1, compare nicelywith other published values for the outer surface dimensionsof the AChR molecule.

Fig. 3 Upper shows a contour map of the unfiltered imageof a single molecule, with the distance between lines being 3A (vertical dimension). The total height above the back-ground is 50 A, which is close to the dimensions reportedearlier for the extracellular moiety of the molecule (9-12)protruding from the plane of the membrane. Five peaks canbe observed, although two on the left side are not wellresolved. The major peak protrudes about 15 A from thelargest contour line delimiting the central pit, while the peakon the opposite side of the central opening (at top bar in Fig.3) protrudes -9 A. Of the two peaks poorly resolved on theleft side, the one on the lower left of the image protrudes 3 Aand the one on the upper left 9 A above this same plane. Onthe right side a peak protruding -6 A is observed. Thiscontour image closely matches those previously reportedfrom hybrid maps (9). The average width of the walls of thepseudosymmetric rosette surrounding the central pit is 25 A(determined from filtered images). The depth ofthe pit can befollowed for 9 A from the widest contour line. Steric limita-tions in imaging the cavity are due to the dimensions of thetip, and only the outer opening can be reliably measured. Thelargest circular contour line is taken as a reference for theouter opening and has a diameter, in this case, of 28-36 A (atthe smallest and largest widths, respectively).The useful imaging time is limited by the rate of dehydra-

tion, since the average time for stability of the sample withrespect to STM imaging was 60-90 min. After this length oftime membranes underwent a substantial change in aggrega-tion, resulting in smaller fragments moving freely under theprobe. Such observation of a limited imaging time suggests adeleterious effect ofdehydration on the sample. We thereforestudied preparations of AChR-enriched membranes, treatedas in the STM structural studies, with respect to their proteinsecondary structure by using CD spectroscopy and theirfunctional stability by using 125I-a-bungarotoxin binding overthe same time period. In Fig. 4 the CD spectra of membranesallowed to dry for two different times and then solubilized arepresented and compared with a sample of the same prepa-ration that was simply solubilized. Numerical values of thespectral data analyses are presented in Table 2. Only a slightdifference in the intensity of the two troughs at 209 and 218nm of the dried sample compared with the control prepara-tion was detected. Comparison of curve 1 (control) and curve2 (allowed to dry for 30 min) showed that there was a loss of6% in helical structure and an increase of -4% in therandom-coil signal. No significant differences were foundbetween the samples dried for 30 min (curve 2) and for 3 hr(curve 3). However, as noted above, we did observe changesin the STM images over the longer time period. The bindingof 125I-a-bungarotoxin to the AChR was not affected signif-

icantly by dehydration over the time periods used (180 min;Table 3). The slightly lower values of the dried samples (85± 11% of the control) may be due to small changes instructure of the protein or to incomplete solubilization of theAChR from the HOPG surface by the detergent.

DISCUSSIONSTM imaging of biological macromolecules has only recentlybeen applied to nucleic acids (22-25), globular proteins (26,27), and glycogen (28). A review of several such studies ofproteins is presented in ref. 2. In addition, similar studies ofmembranes containing bacteriorhodopsin have been de-scribed (29). A very recent study of Torpedo marmoratamembranes enriched in AChR showed arrays of receptormolecules rising above the membrane surface (30) withresolution comparable to negatively stained electron micro-graphs. DNA is a highly suitable substrate for STM imaging,presumably because of its characteristic shape and surfacecharge density. In the case of soluble proteins and glycogen,their overall shape and dimensions can be determined but fewdetails of the surface have so far been determined, eitherbecause of a lower density of surface charge or a lesscharacteristic shape compared with nucleic acids. In thestudy reported here the resolution obtained also does notreveal the structure at atomic resolution, but it has beenpossible to obtain overall dimensions for the outer (synaptic)surface of the AChR. The values obtained by using filteredimages agree well with those previously reported from othermethods. The characteristic shape and dimensions of theAChR, with a large central cavity, allowed facile identifica-tion of this membrane-embedded molecule. In this attempt todelineate the surface structure ofthis molecule by STM it wasimportant to eliminate the possibility of artifactual images(31) due to factors such as the substrate used for sampledeposition, the variability of the probes, or modification ofthe sample upon dehydration. Thus, comparison of theimages and their dimensions with those previously obtainedby other methods was critical (see ref. 10). Since no majorchanges in either secondary structure, as evidenced by CDspectroscopy, or in the extent of a-neurotoxin binding oc-curred over time up to 3 hr, it is likely that the loss of usefulinformation after 60-90 min of imaging could be due to theeffects of dehydration on the lipid bilayer. Related to this wefound that detergent-solubilized AChR was less stable forimaging purposes than the membrane-bound form. In addi-tion, the pH of the preparation affected the uniformity of itsspreading on the graphite surface, which was much betterunder slightly alkaline conditions than at neutral pH. Weobserved similar effects of pH on the spreading of AChRsamples on carbon-coated grids used for transmission elec-tron microscopy, in that larger membrane aggregates werealso observed at lower pH.The most revealing structural details ofthe AChR molecule

are obtained by using unfiltered images where major regionsof protein density are observed, two of which are somewhatoverlapping (Fig. 3). These main regions of density maycorrespond to the constituent subunits of the AChR. Due totheir disposition these features suggest a rather asymmetricstructure, despite the known extensive homology in primarystructure between the subunits (6, 7). We conclude that theSTM method can at present be profitably utilized to obtainboth surface dimensions and low-resolution structural infor-mation for membrane proteins when such membranes con-tain a high density of one specific protein such as the AChRor bacteriorhodopsin (29) or where two-dimensional orderingis observed in reconstituted systems where a highly purifiedmembrane protein has been reintroduced into lipid bilayers.It is important to note that not only do the outer and innerdimensions of the AChR images reported here conform to thevalues obtained by other methods but also that the pattern of

Biophysics: Bertazzon et al.

Dow

nloa

ded

by g

uest

on

Apr

il 25

, 202

1

9636 Biophysics: Bertazzon et al.

surface protrusions observed is also in close agreement (Fig.3 Lower). It will be interesting to conduct imaging studies ofsimilar channel proteins from different tissues as well as ofother proteins of the superfamily of ligand-gated channelproteins of this type now known to exist (reviewed in ref. 10).Furthermore, improvements in methodology should allowincreased resolution of such structures in the future.

This work was supported by National Institutes of Health Grant5R01-NS10294 (to M.A.R.), the ARO Contract DAMD 17-88-C-8120(to M.A.R. and B.M.C.-T.), and National Institute on Drug AbuseProgram Project Grant 5PO1-DA05698 (to M.A.R. and B.M.C.-T.).

1. Binnig, G. & Rohrer, H. (1984) in Trends in Physics, eds. Janta,J. & Pantoflicek, J. (Eur. Phys. Soc., Petit-Lancy, Switzer-land), pp. 38-46.

2. Horber, J. K. H., Ruppersberg, J. P., Schuler, F. M. &Schroter, K. H. (1990) in Abstracts STM '90, Proceedings ofthe Fifth International Conference on Scanning TunnelingMicroscopy/Spectroscopy, and Nano, I, First InternationalConference on Nanometer Scale Science and Technology,Baltimore (AIP, New York), abstr. 243.

3. Duguid, J. R. & Raftery, M. A. (1973) Biochemistry 12, 3593-3597.

4. Cohen, J. B., Weber, M., Huchet, M. & Changeux, J.-P. (1972)FEBS Lett. 112, 73-78.

5. Raftery, M. A., Vandlen, R., Reed, K. & Lee, T. (1975) ColdSpring Harbor Symp. Quant. Biol. 40, 193-202.

6. Raftery, M. A., Hunkapiller, M. W., Strader, C. D. & Hood,L. E. (1980) Science 208, 1454-1457.

7. Numa, S., Noda, H., Takahashi, T., Tanabe, T., Toyosato, M.,Furutani, Y. & Kikyotani, S. (1983) Cold Spring Harbor Symp.Quant. Biol. 48, 57-69.

8. Conti-Tronconi, B. M., Gotti, C., Hunkapiller, M. & Raftery,M. A. (1982) Science 218, 1227-1232.

9. Mitra, A. K., McCarthy, M. P. & Stroud, R. M. (1989) J. CellBiol. 109, 755-774.

10. Stroud, R. M., McCarthy, M. P. & Shuster, M. (1990) Bio-chemistry 29, 11009-11023.

11. Klymkowsky, M. W. & Stroud, R. M. (1979) J. Mol. Biol. 128,319-334.

12. Brisson, A. & Unwin, P. N. T. (1985) Nature (London) 315,474-477.

13. Elliot, J., Blanchard, S. G., Wu, W., Miller, J., Strader, C. D.,Hortig, P., Moore, H.-P., Racs, J. & Raftery, M. A. (1980)Biochem. J. 185, 667-677.

14. Elliot, J., Dunn, S. M. J., Blanchard, S. & Raftery, M. A.(1979) Proc. Natl. Acad. Sci. USA 76, 2576-2580.

15. Neubig, R. R., Krodel, E. K., Boyd, N. D. & Cohen, J. B.(1979) Proc. Natl. Acad. Sci. USA 76, 690-694.

16. Laemmli, U. K. (1970) Nature (London) 227, 680-685.17. Blanchard, S. G., Quast, U., Reed, K., Lee, T., Schimerlik,

M. I., Vandlen, R., Claudio, T., Strader, C. D., Moore, H.-P. H. & Raftery, M. A. (1979) Biochemistry 18, 1875-1883.

18. Chang, C. T., Wu, C.-S. C. & Yang, J. T. (1978) Anal. Bio-chem. 91, 13-31.

19. Moore, W. M., Holladay, L. A., Puett, D. & Brady, R. N.(1974) FEBS Lett. 45, 145-149.

20. Wu, C.-S. C., Sun, X. H. & Yang, J. T. (1990) J. Prot. Chem.9, 119-126.

21. Mielke, D. K. & Wallace, B. A. (1988) J. Biol. Chem. 263,3177-3182.

22. Arscott, P. G., Lee, G., Bloomfield, V. A. & Evans, D. F.(1989) Nature (London) 339, 484-486.

23. Beebe, T. P., Jr., Wilson, T. E., Ogletree, F., Kate, J. E.,Blahorn, R., Salmeron, M. B. & Siekhaus, W. J. (1989) Science243, 1226-1227.

24. Cricenti, A., Selci, S., Felicic, A. C., Generosi, R., Gori, E.,Djanzenko, W. & Chirotti, G. (1989) Science 245, 1226-1227.

25. Lindsay, S. M., Thundant, T., Nagahara, L., Knipping, U. &Rill, R. L. (1989) Science 244, 1062-1064.

26. Edstrom, R. D., Meinke, M. H., Yang, X., Yang, R. & Evans,D. F. (1989) Biochemistry 28, 4939-4942.

27. Miles, M. J., Carr, H. J., McMaster, T. C., I'Anson, K. J.,Belton, P. S., Morris, V. J., Field, J. M., Shewry, P. R. &Tatham, A. S. (1991) Proc. Natl. Acad. Sci. USA 88, 68-71.

28. Yang, X., Miller, M. A., Claus, D. F. & Edstrom, R. D. (1990)FASEB J. 4, 3140-3143.

29. Guckenberger, R., Hacker, B., Hartmann, T., Scheybani, T.,Wang, Z., Wiegrabe, W. & Baumeister, W. (1991) J. Vac. Sci.Technol. 9, 1227-1230.

30. Horber, J. K. H., Schuler, F. M., Witzemann, V., Schroter,K. H., Muller, H. & Ruppersberg, J. P. (1991) J. Vac. Sci.Technol. B 9, 1214-1218.

31. Clemmer, C. R. & Beebe, T. P., Jr. (1991) Science 251, 640-642.

Proc. Natl. Acad Sci. USA 89 (1992)

Dow

nloa

ded

by g

uest

on

Apr

il 25

, 202

1