Langmuir 2006, 22, 11311-11321
11311
Microscale Correlation between Surface Chemistry, Texture, and the Adhesive Strength of Staphylococcus epidermidis Ray J. Emerson IV,† Torbjorn S. Bergstrom,‡ Yatao Liu,† Ernesto R. Soto,§ Christopher A. Brown,‡ W. Grant McGimpsey,§,| and Terri A. Camesano*,† Department of Chemical Engineering, Department of Mechanical Engineering, Department of Chemistry and Biochemistry, and Bioengineering Institute, Worcester Polytechnic Institute, 100 Institute Road, Worcester, Massachusetts 01609 ReceiVed July 9, 2006. In Final Form: September 21, 2006 Staphylococcus epidermidis is among the most commonly isolated microbes from medical implant infections, particularly in the colonization of blood-contacting devices. We explored the relationships between surface wettability and root-mean-square roughness (Rq) on microbial adhesive strength to a substrate. Molecular-level interactions between S. epidermidis and a variety of chemically and texturally distinct model substrata were characterized using a cellular probe and atomic force microscopy (AFM). Substrata included gold, aliphatic and aromatic self-assembled monolayers, and polymeric and proteinaceous materials. Substrate hydrophobicity, described in terms of the water contact angle, was an insufficient parameter to explain the adhesive force of the bacterium for any of the surfaces. Correlations between adhesion forces and Rq showed weak relationships for most surfaces. We used an alternate methodology to characterize the texture of the surface that is based on a fractal tiling algorithm applied to images of each surface. The relative area as a function of the scale of observation was calculated. The discrete bonding model (DBM) was applied, which describes the area available for bonding interactions over the full range of observational scales contained in the measured substrate texture. Weak negative correlations were obtained between the adhesion forces and the area available for interaction, suggesting that increased roughness decreases bacterial adhesion when nano- to micrometer scales are considered. We suggest that modification of the DBM is needed in order to include discontinuous bonding. The adhesive strength is still related to the area available for bonding on a particular scale, but on some very fine scales, the bacteria may not be able to conform to the valleys or pits of the substrate. Therefore, the bonding between the bacterium and substrate becomes discontinuous, occurring only on the tops of ridges or asperities.
Introduction Microbial infections of medical implants occur in more than 2 million surgical cases each year in the United States alone, increasing patient morbidity, mortality, cost, and recovery time.1-3 Surgical excision of the infected device is often the only treatment.4,5 Clinically, it is of interest to determine the factors affecting microbial adhesion, the precursor to infection, and to formulate adhesion-resistant materials possibly with specially designed surface textures, or roughness, that are effective for either short- or long-term use. The roughness of the substrate has been identified as an important factor in determining whether bacteria will adhere. In general, experiments have demonstrated that increased roughness leads to increased surface area available for contact and thus higher bacterial adhesion.6 Rougher surfaces may also protect bacteria from detachment induced by shear.7 There is some disagreement in the literature as to whether there may be a threshold below which less roughness does not prevent bacterial * Corresponding author. E-mail:
[email protected]. Phone: 508-831-5380. † Department of Chemical Engineering. ‡ Department of Mechanical Engineering. § Department of Chemistry and Biochemistry. | Bioengineering Institute. (1) Arciola, C. R.; Campoccia, D.; Gamberini, S.; Donati, M. E.; Montanaro, L. Biomaterials 2004, 25, 4825-4829. (2) Correa, L.; Pittet, D. J. Hosp. Inf. 2000, 46, 89-95. (3) Khardori, N.; Yassien, M. J. Ind. Microbiol. 1995, 15, 141-147. (4) Gristina, A. G. Science 1987, 237, 1588-95. (5) Heard, S. O. Ann. Acad. Med. Singapore 2001, 30, 419-429. (6) Taylor, R. L.; Verran, J.; Lees, G. C.; Ward, A. J. P. J. Mater. Sci.: Mater. Med. 1998, 9, 17-22. (7) Quirynen, M.; Bollen, C. M. L. J. Clin. Periodont. 1995, 22, 1-14.
attachment or an upper threshold where increased roughness does not promote more attachment.8,9 One limitation is that the root-mean-square roughness parameter (Rq) does not always correlate with bacterial attachment, even in cases where a qualitative correlation is evident. For example, the attachment of streptococci to stainless steel was independent of Rq, but the authors could observe trapping of bacteria in surface defects that were on the order of the size of a bacterium.10 Because the quantitative interpretation of the effect of roughness has been limited, we are not yet able to design a material with a texture that is known to be anti-adhesive for bacteria. In addition, most studies have attempted to correlate roughness with bacterial retention or adsorption onto a surface. For example, roughness parameters were correlated with the number of bacteria that could attach to commercially available polymers used for contact lenses.11 Although these types of experiments can be useful for a particular application and from a practical perspective, fundamental studies that will lead to mechanistic understanding, in which nanoscale adhesion forces between the bacterium and substrate are related to surface roughness, have not been performed. In addition to surface texture, cell surface hydrophobicity and charge have received significant attention as determining factors in establishing the propensity of a bacterium to adhere to a (8) Barnes, L.-M.; Lo, M. F.; Adams, M. R.; Chamberlain, A. H. L. Appl. EnViron. Microbiol. 1999, 65, 4543-4548. (9) Tide, C.; Harkin, S. R.; Geesey, G. G.; Bremer, P. J.; Scholz, W. J. Food Eng. 1999, 42, 85-96. (10) Flint, S. H.; Brooks, J. D.; Bremer, P. J. J. Food Eng. 2000, 43, 235-242. (11) Alava, J.; Garagorri, N.; Briz, N.; Mendicute, J. J. Mater. Sci.: Mater. Med. 2005, 16, 313-317.
10.1021/la061984u CCC: $33.50 © 2006 American Chemical Society Published on Web 11/11/2006
11312 Langmuir, Vol. 22, No. 26, 2006
surface.12-14 Hydrophobicity is determined by water contact angles measured on dried microbial lawns,14 and the surface charge is reported in terms of the zeta potential of a microbial suspension in buffered media of varying ionic strength.12 In conjunction with the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory of colloid stability,15 these parameters may be used to calculate the energy profiles arising from London-van der Waals and electrostatic interactions. Extensions to the classical theory have been developed to account for exopolymeric steric interactions,16 acid-base interactions,17 and various other specific and nonspecific interactions.18 Although hydrophobicity is generally accepted as correlating with bacterial adhesion,19 the ability to design materials or predict bacterial adhesion to a biomaterial is not possible solely on the basis of considerations of cellular and/or substrate hydrophobicity. Researchers have since turned to more sophisticated tools for probing bacterial surfaces. Significant progress has been made in the ability to quantify the adhesion forces between living (unfixed) bacterial cells and virtually any kind of substrate on the basis of atomic force microscopy (AFM) measurements,20,21 where it is even possible to map the individual recognition forces between a bacterial adhesin and its glycoconjugate receptor.22 Forces can be quantified over a wide range with the AFM, including those for bacterial-substrate pairs where the interactions are in the piconewton,23,24 nanonewton,25,26 and micronewton ranges,23 thus providing a more quantitative and direct characterization of bacterial adhesion for numerous systems. There remains an important challenge as researchers try to relate bulk properties, nanoscopic properties, and adhesion behavior for complex bacterial systems. Some studies have attempted to correlate chemical and physical properties of substrates with bacterial attachment by applying multiple linear regressions.27 Such statistical methods were also used to explore potential correlations between macroscale measurements and microscale behavior as measured by the AFM.28 Both of these studies showed some interesting relationships, but more importantly, they further highlighted the complexity of the microbial (12) Wilson, W. W.; Wade, M. M.; Holman, S. C.; Champlin, F. R. J. Microbiol. Methods 2001, 43, 153-64. (13) Bos, R.; van der Mei, H. C.; Busscher, H. J. FEMS Microbiol. ReV. 1999, 23, 179-230. (14) Busscher, H. J.; Bialkowska-Hobrzanska, H.; Reid, G.; van der KuijlBooij, M.; van der Mei, H. C. Colloids Surf., B 1994, 2, 73-82. (15) Hunter, R. J. Foundations of Colloids Science; Clarendon Press: Oxford, U.K., 1986; Vols. 1 and 2. (16) Butt, H.-J.; Kappl, M.; Mueller, H.; Paiteri, R.; Meyer, W.; Ruhe, J. Langmuir 1999, 15, 2559-2565. (17) van Oss, C. J. Interfacial Forces in Aqueous Media; Marcel Dekker: New York, 1994. (18) Grasso, D.; Subramaniam, K.; Butkus, M.; Strevett, K.; Bergendahl, J. ReV. EnViron. Sci. Biotechnol. 2002, 1, 17-38. (19) Hogt, A. H.; Dankert, J.; de Vries, J. A.; Feijen, J. J. Gen. Microbiol. 1983, 129, 2959-68. (20) Cail, T. L.; Hochella, M. F. Geochim. Cosmochim. Acta 2005, 69, 29592969. (21) Lower, S. K.; Hochella, M. F., Jr. Beveridge, T. J. Science 2001, 292, 1360-1363. (22) Dupres, V.; Menozzi, F. D.; Locht, C.; Clare, B. H.; Abbott, N. L.; Cuenot, S.; Bompard, C.; Raze, D.; Dufreˆne, Y. F. Nat. Methods 2005, 2, 515-520. (23) Tsang, P. H.; Li, G. L.; Brun, Y. V.; Ben Freund, L.; Tang, J. X. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 5764-5768. (24) Touhami, A.; Jericho, M. H.; Boyd, J. M.; Beveridge, T. J. J. Bacteriol. 2006, 188, 370-377. (25) Vadillo-Rodriguez, V.; Logan, B. E. EnViron. Sci. Technol. 2006, 40, 2983-2988. (26) Liu, Y.; Black, M. A.; Caron, L.; Camesano, T. A. Biotechnol. Bioeng. 2006, 93, 297-305. (27) Bruinsma, G. M.; Rustema-Abbing, M.; de Vries, J.; Busscher, H. J.; van der Linden, M. L.; Hooymans, J. M. M.; van der Mei, H. C. Biomaterials 2003, 24, 1663-1670. (28) Vadillo-Rodriguez, V.; Busscher, H. J.; Norde, W.; De Vries, J. A.; van der Mei, H. C. Langmuir 2003, 19, 2372-2377.
Emerson et al.
adhesion process because clear relationships could not be obtained for all considered parameters. The main objective of this research was to determine quantitatively the main factors influencing the adhesion of Staphylococcus epidermidis to a variety of substrates. Where macroscale correlations have proved elusive, we explored their use in conjunction with AFM techniques, specifically in comparing results obtained from macroscale contact angle experiments and microscale texture analysis to characterize the surface texture. The adhesion forces were measured between immobilized cells of S. epidermidis used as a force probe to determine the interactions with each member of a series of chemically and texturally distinct films, including self-assembled monolayers (SAMs) with terminal aliphatic and aromatic groups, polyamino acids, and proteins. Adhesion forces were correlated to both the relative hydrophobicity (contact angle) of each surface and to the root-mean-square roughness (Rq) of each surface. Recognizing that Rq may be insufficient to characterize the aspects of surface texture responsible for controlling bacterial adhesion, we also investigated the application of a discrete bonding model.29,30 In this model the adhesive strength of an interaction is related to the area available for interaction on a particular scale. The model is based on a fractal representation to characterize the substrate texture measurement, which shows how the apparent, or calculated, areas change with the scale of observation, or calculation.31 This technique offers many benefits because it partially overcomes some of the resolution dependence of the AFM measurements and its use allows for the identification of the areal scales on which individual interactions occur. Materials and Methods Preparation of Self-Assembled Monolayer Surfaces and Randomly Adsorbed Surfaces. A variety of substrate surfaces were used in these experiments. By starting with alkanethiol-based selfassembled monolayers (SAMs), we could vary the terminal functional group so that the effect of a particular type of chemical group could be examined. In addition to these SAMs, we used some physisorbed chemicals to represent commonly used coatings. These latter surfaces could serve as benchmarks for the performance of the SAMs because no prior research was done to study the role of these SAMs on bacterial adhesion forces. In total, six SAMs and four surfaces coated through physisorption were constructed with varying terminal groups so that the chemical nature of each group could be explored (Table 1, Figure 1.) Dodecanethiol (DDT), hexadecanethiol (HDT), and ω-mercaptoundecanoic acid were purchased from Aldrich (Milwaukee, WI) and used as received. DDT was used in a 1 mM solution in ethanol. HDT was prepared in 10 and 20 mol/L solutions in ethanol (termed HDT10 and HDT20, respectively). IPA. 5-(10-Mercaptodecyloxy)isophthalic acid (IPA) was synthesized following the procedure described in ref 32. The compound was obtained in a three-step synthesis starting from a coupling reaction of 1,10-dibromodecane and diethyl 5-hydroxyisophthalate. The bromine group in the resulting diethyl 5-(10-thioacetyl-decyloxy)isophthalate was substituted with a thioacetate group. The resulting compound was treated in a potassium hydroxide/ethanol solution. IAG. Films of Ag(I) + IPA (called IAG) were prepared by transferring the monolayer of IPA to a 5 mM solution of silver(I) nitrate (AgNO3) in acetonitrile. Ag(I) complexation of the surface was obtained after 3 h. (29) Siegmann, S.; C. A. Brown. Surface Texture Correlations with Tensile AdhesiVe Strength of Thermally Sprayed Coatings Using Area-Scale Fractal Analysis; United Thermal Spray Conference, 1999. (30) Brown, C. A.; S. Siegmann Int. J. Mach. Tool Manuf. 2001, 41, 19271933. (31) Surface Texture: Surface Roughness, WaViness and Lay; American Society of Mechanical Engineers: New York, 2002. (32) Soto, E.; MacDonald, J. C.; Cooper, C. G.; McGimpsey, W. G. J. Am. Chem. Soc. 2003, 125, 2838-9.
Chemistry Vs Texture Vs S. epidermidis Adhesion
Langmuir, Vol. 22, No. 26, 2006 11313
Table 1. Summary of Chemicals Used to Prepare Coated Substrates concentrationa
molecular weight
sample
abbreviation
bovine serum albumin dodecanethiol hexadecanethiol (10 mM) hexadecanethiol (20 mM) 5-(10-mercaptodecyloxy)isophthalic acid Ag(I)-IPA 4-(10-mercaptodecyloxy)pyridine poly-L-lysine poly-L-lysine poly-L-lysine
BSA DDT HDT10 HDT20 IPA
10 mg/mLb 1 mM 10 mM 20 mM 1 mM
50-80 kDa 201.4 Da 257.5 Da 257.5 Da 353.5 Da
IAG PDT
1 mM 1 mM
459.3 Da 266.4 Da
PLL1 PLL10 PLL100
0.01% w/vc 0.001% w/vc 0.0001% w/vc
0-300 kDa 50-300 kDa 50-300 kDa
c
a In ethanol, unless specified as another solution. b In ultrapure water. In 100 mM MES buffer.
PDT. 4-(10-Mercaptodecyloxy)pyridine (PDT) was synthesized from 1,10-dibromodecane and 4-hydroxypyridine. The resulting 4-(10-bromo)-decyloxy-pyridine was converted to the final compound in one step according to the trimethylsilythioxy-dehalogenation reaction reported by Hu and Fox.33 PLL. Poly-L-lysine (PLL, MW ) 50-300 kDa, Sigma-Aldrich) solutions were prepared at 0.01% w/v (PLL1) and also at 10× and 100× dilutions (PLL10 and PLL100) in 100 mM MES buffer [2-(Nmorpholino)ethanesulfonic acid]. Solutions were mixed for 5 to 6 h at ambient temperature. BSA. Bovine serum albumin (MW ) 50-80 kDa, Sigma) was prepared at a concentration of 10 mg/mL in ultrapure water (Milli-Q water, Millipore Corp., Billerica, MA). The solution was mixed at ambient temperature for 5 to 6 h. The SAMs and randomly deposited coatings were adsorbed onto commercially prepared gold substrates (Evaporated Metal Films, Ithaca, NY). The slides are glass, coated with 5 nm of chromium or titanium followed by 100 nm of gold, prepared using an evaporation procedure. Gold slides were cleaned in acid piranha solution (70% sulfuric acid/30% hydrogen peroxide) at 90 °C for 10-15 min, rinsed with water and ethanol, dried with a stream of nitrogen, and used immediately. For the SAMs, monolayers were prepared by immersing the clean gold slides in a 1 mM ethanol solution of the desired compound (unless another concentration/solvent is specified in Table 1) for at least 24 h. The films were rinsed with ethanol and dried with nitrogen. For the randomly adsorbed films (PLL series and BSA), 10 mL of the desired solution was added to the cleaned gold surface. Slides were agitated at 0.81g for 8 h. Substrate Characterization. Water contact angles on each substrate were measured with a goniometer (Rame´-Hart model 10000; Netcong, NJ). Drops of ultrapure water (2 µL) were deposited with a micropipet, and the sessile drop contact angle was measured in triplicate. Film thickness measurements were obtained from ellipsometry (Rudolf 439L633P; Rudolph Instruments, Fairfield NJ). The measurements were taken at a 70° angle of incidence using a HeNe laser (principal wavelength ) 632.8 nm). Three samples per coating were examined. On a given sample, measurements were taken at three different locations separated by >5 mm. A bare gold substrate was used to determine the refractive index (n ) 0.15) and the extinction index (k ) 3.3) of gold, which were similar to values that others have reported.34 For the SAMs, values of n ) 1.457 and k ) 0 were assumed on the basis of previous work.35 We determined that all of the SAMs and physisorbed substrates were neutral on the basis of measuring the electrical charge across the surface and comparing with a reference electrode. Although the ionization of charged groups is possible, the buffer used in all AFM (33) Hu, J.; Fox, M. A. J. Org. Chem. 1999, 64, 4959-4961. (34) Ordal, M. A.; Long, L. L.; Bell, R. J.; Bell, S. E.; Bell, R. R.; Alexander, R. W., Jr.; Ward, C. A. Appl. Opt. 1983, 22, 1099-1119. (35) Folkers, J. P.; Laibinis, P. E.; Whitesides, G. M. Langmuir 1992, 8, 13301341.
experiments (100 mM MES) likely provided some screening of electrostatic charges. Therefore, we did not consider the effect of charge variability of the substrates or determine the role of charge in influencing the bacteria adhesion behavior. Microbial Growth and Storage. Clinically isolated samples of S. epidermidis were kindly provided by Dr. Stephen Heard (Department of Anesthesiology, University of Massachusetts Medical School, Worcester, MA). Cultures were maintained on tryptic soy agar (40 g/L; Sigma) plates and were repitched every 14 days. Cells were cultured in tryptic soy broth (30 g/L; Sigma) at 37 °C until reaching the mid-exponential growth phase (optical density at 600 nm ) 0.9 ( 0.05). For cell probe experiments, 10 mL of cell suspension was transferred to a centrifuge tube and formed into pellets at a relative centrifugal force of 1360g for 10 min (Centrific Centrifuge, Fischer) at 25°C. The supernatant was then eluted, and the pellet was resuspended in 4 mL of 100 mM MES buffer. Cell Probe Functionalization. Atomic force microscope probes (CSC38 B; Mikromasch, Portland, OR) were functionalized with S. epidermidis according to the procedure given in ref 36. Briefly, a custom triaxial micromanipulator was used to immerse the tip of the cantilever in a solution of 20 mM HDT in ethanol. The cantilever was removed from the solution and allowed to dry for 5 min and was then immersed in a suspension of S. epidermidis, concentrated to an approximate cell density of 1 × 1011 cells/mL. After 5 min, the cantilever was removed from the suspension and allowed to dry for an additional 5 min to remove excess moisture. Cell attachment and orientation were verified by scanning electron microscopy (SEM) as well as by control force curve measurements on gold. For the SEM investigations, the bacteria-coated tip was placed in a closed AFM tip box and allowed to dry under ambient conditions. The silicon chip attached to the AFM tip was affixed to an SEM specimen stub using double-sided carbon tape (Electron Microscopy Science, Washington, PA). Bacterial cells did not need to be coated prior to SEM imaging. The SEM (JEOL JSM-840) was operated at 15 kV with a magnification of 10 000×. An example SEM image of a few cells of S. epidermidis attached to the AFM tip is shown in Figure 2, which represents a typical probe used for our experiments. Substrate Texture Measurements Using the AFM. The surface texture of each substrate was measured using an atomic force microscope (Dimension 3100 AFM with Nanoscope IIIa controller, Veeco Metrology, Woodbury, NY). The AFM was calibrated according to the manufacturer’s specifications37 as well as using the methods described in refs 38-40 to characterize the probe spring constant and piezoactuator travel distance. Cantilever spring constant values were 0.04 ( 0.01 N/m. Calibration of the piezoactuator correction factor yielded a mean value of 0.997, indicating that the actual piezoactuator travel distance is less than that reported by the AFM software. When we designed the SAMs and other substrates, it was not the intent to vary the surface texture over a wide range. We wanted to focus on surfaces with well-defined chemistries that were all (on some scale) “smooth”. However, even subtle differences in their surface textures could be important in influencing bacterial adhesion and must be characterized. To this end, measurements of substrate textures were made for each surface, using contact mode AFM, with a bare silicon probe (CSC38 B, Mikromasch, Portland, OR). The measured regions were (1 × 1) µm2 and were composed of 256 × 256 pixels2, resulting in sampling intervals of about 2 nm. Five measurements were made on each sample using a scan rate of 1 Hz. Root-mean-square roughness (Rq) calculations were performed for (36) Emerson, R. J.; Camesano, T. A. Appl. EnViron. Microbiol. 2004, 70, 6012-6022. (37) Dimension 3100 Manual, version 4.43B. Digital Instruments, Veeco Metrology Group, 1997. (38) Burnham, N. A.; Chen, X.; Hodges, C. S.; Matei, G. A.; Thoreson, E. K.; Roberts, C. J.; Davies, M. C.; Tendler, S. J. B. Nanotechnology 2003, 14, 1-6. (39) Emerson, R. J.; Camesano, T. A. Ultramicroscopy 2006, 106, 413-22. (40) Thoreson, E. K.; Burnham, N. A. ReV. Sci. Instrum. 2004, 75, 13591362.
11314 Langmuir, Vol. 22, No. 26, 2006
Emerson et al.
Figure 1. Schematic representation of the substrates examined. All species were bound to gold by thiol linkages, with the exception of PLL and BSA, and confluence of the layers was determined by contact angle and ellipsometric measurements (Table 2). Molecular modeling suggests that the Ag+ group of the IAG SAM is most likely associated with one of the carboxyl groups (data not shown). The 3-D structure of BSA has not been completely established, but it is known that BSA contains 3 homologous domains (1 domain is depicted) and each domain contains 9 loops connected by 17 disulfide bonds. Each domain can be further divided into 10 helical segments.54 BSA may contain both R-helices and other types of tertiary structures.
Figure 2. SEM micrograph showing S. epidermidis adsorbed to the AFM tip. each measurement using the AFM software (Nanoscope v. 4.43r8, Digital Instruments).37,41 Bacterial Adhesion Force Measurements. Using the cellular probes, force cycles were captured at a scan rate of 1 Hz, a nominal ramp size of 1000 nm, and a resolution of 512 samples per curve. Five cycles were captured for each of 10 areas per substrate and were analyzed according to ref 36. Force curves were recorded with bare probes and HDT-coated probes as controls. All force measurements were made in 100 mM MES buffer. In the retraction portion of every force cycle (i.e., after the probe and sample have come into contact and the tip is being moved away from the substrate), the separation distance and magnitude of each pull-off event was recorded and processed separately to generate histogram data of percent (41) Scanning Probe Microscopy Training Notebook; Digital Instruments, Veeco Metrology Group, 1998; pp 39-40.
normalized frequency against the pull-off force. Each pull-off event represents an adhesive interaction between the probe and the substrate. The histogram data sets were analyzed in SigmaStat v. 2.03 (Systat, Richmond, CA) using the Kruskal-Wallis one-way analysis of variance (ANOVA) on ranks to determine the statistical significance. In general, each substrate-bacterial probe combination was performed on a different day and with a new bacterial probe, prepared just before the given day’s experiments. A given probe would therefore be used only for a few hours, as long as it would take to examine a single substrate. Because the AFM experiments were always performed in buffer (MES), with no nutrients added, we do not think that the bacteria were growing during this time. The doubling time of S. epidermidis is ∼1 h in rich growth media. In our experiments, the bacteria were kept in the buffer for a period of up to ∼3 h, depending on how long it took to capture the force measurements for a particular substrate. No growth is likely to have occurred during this time and under these conditions. The general procedure for making force measurements was to prepare the cellular probe, make several force measurements on cleaned gold, change to the substrate desired, and then return to measurements on gold. Making these “before” and “after” measurements on the gold would allow us to identify any cases where the tip picked up contamination or lost the bacterial cells. Area-Scale Fractal Analysis and the Discrete Bonding Model. For each AFM measurement, the relative areas were determined over a range of scales using specialized software (SFrax; www.surfract.com), following the method developed by Brown et al.29,42 Briefly, the software applies triangular patches to virtually tile the surface (Figure 3). In each tiling exercise, all of the triangles have the same area but not necessarily the same shape, which is adjusted so that they fit together. The area of the triangle represents the scale of calculation or observation. The tiling exercises are repeated many times so that different scales can be represented, with the area of the triangle varying from one iteration to the next. Each measured region was fit with triangles that ranged in size from 7.69 nm2 to 0.5 µm2. This range was selected by considering the sampling interval and the size of the measured region. The tiled area on each (42) Brown, C. A. Scale-Area Analysis and Roughness: A Method for Understanding the Topographic Component of AdhesiVe Strength; Seventeenth Annual Meeting and Symposium on Particle Adhesion, The Adhesion Society, February 1994, Orlando, FL.
Chemistry Vs Texture Vs S. epidermidis Adhesion
Langmuir, Vol. 22, No. 26, 2006 11315
Figure 3. Schematic of triangle tiling that is applied to each AFM image of a substrate. The numbers of triangles used to tile the image were (A) 32, (B) 5202, and (C) 32 413. The relative areas are 1.000011, 1.000858, and 1.004543, respectively. scale is normalized to the projected area of the surface covered by the tiling exercise on that scale, and this ratio is called the relative area. The relative area (Rea) is determined by dividing each virtually measured area by the projected area and is quantified as t Rea ) N A
(1)
where N is the number of virtual tiles, t is the area of the tile and represents the scale of the virtual measurement, and A is the projected area for that tiling exercise.43 The relative area on a particular scale is related to the inverse cosine of slopes on the surface for that scale such that greater relative areas correspond to steeper average slopes on the surface.44 Once the relative areas are calculated for different scales of observation, the discrete bonding model (DBM) can be applied. In this methodology, it is assumed that the adhesive strength on a rough surface can be predicted on the basis of summing the contributions of individual bonds.30 The adhesive strength is defined, in this case, as the mean of the force maximum measured for 50 AFM retraction curves for each substrate. Plots of relative areas for several surfaces on a certain scale against adhesion forces on those surfaces are then generated for each observational scale. A linear regression analysis is applied to the relative area versus adhesive strength plot on each scale. The regression coefficient, R2, is recorded and plotted against the scale of observation for all measurements, yielding what is termed a Siegmann plot.30 High values of R2 indicate scales of observation, or calculation, on which the surface texture has significant influence over the adhesive strength. The presence of a local maximum identifies the fundamental adhesive scale, which describes a geometric characteristic of the adhesion interaction. This defines the scale on which adhesion is preferred or the scale on which the bacterium “sees” the substrate.
Results Bacterial Probe Validation. We first verified that our technique for bacterial immobilization on the AFM tips was reproducible and that the forces we measured were due to bacterial interactions rather than interference due to the use of HDT in the attachment. Example approach and retraction curves are shown for the probing of a gold substrate with a bare silicon tip, a silicon tip coated with HDT, and a silicon-HDT tip with attached bacteria (Figure 4A,B). These three conditions are easily distinguishable on the basis of the representative force data shown. (43) Brown, C. A.; Charles, P. D.; Johnsen, W. A.; Chesters, S. Wear 1993, 161, 61-67. (44) Brown, C. A.; Johnsen, W. A.; Butland, R. M. Ann. CIRP 1996, 45, 515-518.
Figure 4. Representative (A) approach and (B) retraction curves to illustrate how adding bacteria to the tip changed the overall interactions and to isolate the effect of HDT, which was used to help the bacteria attach to the AFM tip. All force measurements were made on different areas of a clean gold surface. The tip was bare (silicon), silicon coated with HDT only, or silicon coated with HDT and S. epidermidis.
Therefore, we are confident in our ability to probe the desired substrate with an S. epidermidis probe. In addition, we note that a new bacterial probe (freshly prepared) was used to examine each substrate (SAM or other adsorbed material). The bacterial probe was used for a period of up to ∼3 h and then discarded. In each case, force measurements with the bacterial probe on cleaned gold were made before and after the measurements on the desired substrates, ensuring that the tip did not become contaminated or lose its bacterial cells during the course of an experiment. Furthermore, we confirmed that different bacterial probes could give comparable results. Representative data for two S. epidermidis probes interacting with the BSA-coated surface, made several months apart, are shown in the Supporting Information (Figure S1). Adhesion Force Measurements. Adhesion forces were measured for each substrate using a control probe and an S. epidermidis probe (Figure 5). The adhesive forces ranged from ∼8.5 nN for Au to e50 pN for IPA. Given the wide range of adhesion force magnitudes, the data is shown on two different scales to highlight the retraction profiles for all substrates. The most promising coatings, from the perspective of having low adhesive force values for the bacterial probe, are IPA and IAG, with force magnitudes e50 pN. These values were at or below the instrument noise level for our AFM. The accumulation of all adhesion events is presented fully in histogram form in Figure 6. Comparisons of the top and bottom
11316 Langmuir, Vol. 22, No. 26, 2006
Emerson et al.
Figure 5. Example retraction curves that demonstrate adhesive interactions between the bacterial probe and each substrate. The same data were plotted on two different scales so that small forces can be illustrated more clearly. The adhesion forces between the S. epidermidis probe and either IPA or IAG were e50 pN, which means that these surfaces interact more weakly than we could resolve. All force measurements were made in 100 mM MES.
halves of the plots are used to highlight the differences in the adhesion forces between S. epidermidis-substrate pairs and bare probe-substrate pairs. For example, when considering the DDT substrate being probed with S. epidermidis, a distribution of adhesion forces was obtained ranging from 0.4 nN to >3.4 nN. Yet when DDT was probed with the control (silicon) probe, all of the adhesion force values (100% of events) were in the range of 0.4-0.9 nN. When the top and bottom halves of the figure are dissimilar, this indicates more of a difference in how the substrate interacted with either the bare probe or the bacteriacoated probe. We note that the range of forces is so large that in order to show all of the data on a single histogram some of the adhesion force data has to be compressed, and we fail to show a full distribution from the histogram. Therefore, we have used a second histogram representation (Figure 7) in which we highlight the data between 0.2 and 1.0 nN. Any forces larger than this value were lumped together in the “more” category. When viewing this force range in more detail, the differences between PLL interacting with the control probe and the cellular probe become more apparent. For the substrates with very low forces, such as IPA and IAG, and PDT, even in this narrower force range, it is still difficult to appreciate the differences in the adhesion behavior of the substrate for the cellular and control probes. Because of the difficulties in illustrating the differences for this many
combinations of substrate-probe pairs, we need to use both pictorial representations (i.e., these histograms) and quantitative statistical analysis to formulate conclusions describing the behavior of the coatings. Statistical analysis of the retraction data using the KruskalWallis one-way ANOVA on ranks showed statistically significant differences (P < 0.001) between all control/cell probe experiments, although DDT displayed the most noticeable change in adhesion when comparing the bare probe to the bacterial probe. In general, all of the adhesion forces were low in comparison with those in other systems studied in our laboratory. For the interactions between gram-negative bacteria, including Escherichia coli JM109 or Pseudomonas aeruginosa ATCC 10145, with uncoated AFM tips (silicon nitride or silicon), adhesion forces were observed that were several orders of magnitude higher.36,45 It is interesting that HDT yielded very low forces with the S. epidermidis probe, yet this same chemical could be used to help the bacteria adsorb to the AFM tip. The reasons for the difference include the different time scales and concentrations used in the two systems. For an AFM force measurement, a few S. epidermidis cells are in contact with the HDT for 5 min. This difference highlights the fact that adhesion measurements can depend on both time and concentration. Although the effects of these two variables are beyond the scope of this study, they may be worth investigating in the future. SAM Characterization. All surfaces are hydrophobic to varying degrees (Table 2). Specifically, Au, IPA, and the PLL series, having θw ) 34-59°, are moderately hydrophobic. IAG and PDT are hydrophobic (θw ) 74-90°). BSA, DDT, and HDT10/20 are highly hydrophobic, having θw values >100° For the SAMs, ellipsometry results showed film thicknesses on the order of 1 nm with low standard deviations among repeated
measurements, indicating that confluent monolayers were formed. Because ellipsometry requires optical homogeneity, isotropism, and negligible interactions between molecules, we could not use this technique to measure the film thicknesses for PLL and BSA. Representative topographical measurements, or texture measurements, of all surfaces were made (Figure 8), and Rq values were calculated (Table 2). The maximum vertical scale on each measurement is 10 nm, with the exception of PLL1 (15 nm) and BSA (30 nm). The roughness values spanned a range of Rq ∼0.09 nm for IPA to ∼1.58 nm for Au. The roughness was the most difficult to characterize and was the most variable for the non-
11318 Langmuir, Vol. 22, No. 26, 2006
Emerson et al.
forces from the AFM experiments (Figure 10). Although there are a small number of chemically related molecules, we did check if it was possible to correlate Rq with the adhesion force for individual classes of molecules (i.e., aliphatic only, aromatic only). However, no grouping of molecules could be found that would lead to a satisfactory correlation between Rq and the adhesive force. Theoretical Tensile Strength Modeling. An example plot demonstrating how the relative area changes as a function of the scale of observation is shown for all substrates (Figure 11). When the triangle size is large, the slope of the line becomes zero, and the surface can be considered to be smooth. On finer scales (i.e., with smaller triangles), the relative areas are much greater than unity, and the surfaces can be considered to be rough with fractal geometries. The areal scale at which a “smooth-to-rough crossover” occurred, indicated by the changing slope of the plotted data, varied depending on the substrate from a low value of 0.003 µm2 for PLL10 to a high value of 0.05 µm2 for HDT10. For several scales, plots of mean pull-off force versus relative area were generated. (Figure 12 shows a few representative cases.) These plots showed weak negative correlations between the adhesive strength and relative area. On the finest scales, the strength of the correlations increased regularly with decreasing scale. This trend can be seen in Figure 13, where the correlation coefficients are plotted against the scale of observation. The maximum R2 of 0.35 occurred on the coarsest scale investigated, 0.5 µm2. R2 decreased to zero at about 100 nm2 and increased steadily to about 0.125 at about 8 nm2, the finest scale of observation tested in this study. Figure 7. Alternate form of a histogram that focuses on the adhesion behavior up to 1.0 nN for the (A) cellular probe and (B) control probe with each of the substrates for force measurements made in 100 mM MES. Table 2. Substrate Characterization sample Au BSA DDT HDT10 HDT20 IPA IAG PDT PLL1 PLL10 PLL100
water contact film thickness (nm)b angle (deg)a 36.4 ( 0.1 109 ( 3 105 ( 2 114 ( 3 108 ( 2 52 ( 3 74 ( 3 90 ( 3 34 ( 3 47 ( 3 59 ( 3
e e
1.0 ( 0.2 1.3 ( 0.2 1.3 ( 0.2 1.2 ( 0.3 1.5 ( 0.3 1.3 ( 0.3 e e e
Rq (nm2)c
mean pull-off force (nN)d
1.58 ( 0.879 1.25 ( 0.263 1.07 ( 0.660 0.31 ( 0.060 0.10 ( 0.020 0.09 ( 0.001 0.79 ( 0.170 0.47 ( 0.100 0.17 ( 0.070 1.02 ( 0.090 1.12 ( 0.390
11.6 ( 1.68 7.9 ( 1.40 4.7 ( 1.11 0.2 ( 0.70 0.1 ( 0.10 0.2 ( 0.09 0.2 ( 1.24 3.2 ( 0.85 0.3 ( 2.17 0.3 ( 2.34 0.2 ( 0.79
a Water contact angle; N ) 9. b Ellipsometric film thickness; N ) 9. Root-mean-square roughness, measured region ) 1 µm2; N ) 5. e Mean pull-off force; N ) 50. f Film thicknesses were not measured for Au, BSA, or PLL.
c
SAMs (i.e., BSA and the PLL series) and for uncoated gold. For PLL and BSA, we observed nonhomogeneous adsorption onto the slide. Correlation of Hydrophobicity with Adhesion Forces. We attempted to correlate the water contact angle with the mean pull-off force for each substrate (Figure 9), but no correlations were possible. This was true regardless of whether the data was considered all together or in subgroups based on chemical properties. The R2 was essentially zero for the relationship between the water contact angle and adhesion force between the S. epidermidis probe and the substrates examined. Correlations of Root-Mean-Square Roughness with Adhesion Forces. It was not possible to correlate Rq with the pull-off
Discussion The objective of this study was to determine whether surface chemistry and surface texture would be correlated with the adhesion of S. epidermidis to various substrates, which can represent models for biomaterials. Specifically, we related adhesion force measurements between S. epidermidis and a variety of substrates to parameters that could characterize surface texture of a physicochemical nature, such as the root-mean-square roughness parameter, the water contact angle, and the areal scale of interaction. This latter parameter was obtained by applying a fractal-based tiling algorithm to each substrate to help determine the critical areal scale that would control the adhesion behavior for the bacterium to each substrate. Furthermore, we explored the use of a discrete bonding model (DBM) and proposed an alternate model that could be used to relate surface texture to bacterial adhesion. Surface Wettability Correlation with Adhesion Forces. The correlation between the water contact angle and adhesive strength was weak, with an R2 of essentially zero. This means that surface hydrophobicity alone cannot be used to predict the adhesive strength of the bacteria for these substrates. Surface Roughness Correlation (Rq) with Adhesion Forces. Correlations between Rq and mean adhesive strength were poor, with significant weight placed on one or two data points in the regression. This may be explained by the fact that, although the surfaces were chemically unique, producing a range of adhesive strengths spanning ∼12 nN, the calculated root-mean-square surface roughness values varied by only a few nanometers among all samples. Weak relationships were observed between Rq and adhesive strength, except for in the case of the aliphatic molecules sampled at 1 µm2 area (i.e., considering the circles only in Figure 8A). However, the small number of data points used for that correlation (n ) 3) prevents us from concluding that Rq is directly related to the adhesive strength on this scale.
Chemistry Vs Texture Vs S. epidermidis Adhesion
Langmuir, Vol. 22, No. 26, 2006 11319
Figure 8. AFM topographies of all substrates: (A) bare gold, (B) BSA, (C) DDT, (D) HDT10, (E) HDT20, (F) IAG, (G) IPA, (H) PDT, (I) PLL10, and (J) PLL1. Although Rq was calculated on the basis of (1 × 1) µm2 measurements, representative measurements from (5 × 5) µm2 areas are shown. All measurements were at 256 pixel resolution.
Figure 9. Plots of Rq versus mean pull-off force for aliphatic, aromatic, and non-SAMs species. Aliphatic compounds are DDT and the two HDT solutions of different concentrations. Aromatic compounds are IPA, IAG, and PDT. The non-SAMs category includes gold and the randomly deposited molecules. Rq was calculated for (1 × 1) µm2 measurements.
Figure 10. Plots of water contact angle versus mean pull-off force for aliphatic, aromatic, and non-SAMs species. Compounds are as defined in the Figure 9 caption.
Rq is dependent upon the root-mean-square value of all of the heights measured. However, no corrections exist within this analytical expression to characterize the sampling range, sampling interval, number of heights measured, or actual area sampled by the probe to determine each height (i.e., resolution of the probe), meaning that each individual height measurement has equal
Figure 11. Relative area versus scale of observation based on (1 × 1) µm2 measurements on all substrates.
weight regardless of the sampling strategy or probe resolution. Rq is not able to characterize surfaces or even the same surface measured over different regions because the smaller regions do not include the larger-amplitude wavelengths and Rq is sensitive to the largest amplitudes present in the measurement. It may be possible to explore the use of other parameters to characterize the texture of the surface. For example, in the context of protein adhesion or cellular deposition on surfaces, 2-D Fourier transform (or power spectral density) analysis has been useful for helping to characterize surface topography46,47 and relate to biomolecule deposition. However, for a very different type of system (ground polyethylene), Jordan and Brown recently showed that area-scale analysis was more sensitive and could better differentiate the surfaces on very fine scales than Fourier transform or other conventional analysis techniques.48 Although our objective was not to compare different methodologies for surface texture characterization, this may be a relevant topic for investigation in future studies. Application of the Discrete Bonding Model. On the basis of the a priori assumption that the texture of the surface will, to some extent, influence its adhesive characteristics, the discrete (46) Buchko, C. J.; Kozloff, K. M.; Martin, D. C. Biomaterials 2001, 22, 1289-1300. (47) Denis, F. A.; Hanarp, P.; Sutherland, D.; Gold, J.; Mustin, C.; Rouxhet, P. G.; Dufreˆne, Y. F. Langmuir 2002, 18, 819-828. (48) Jordan, S. E.; Brown, C. A. Wear 2006, 261, 398-409.
11320 Langmuir, Vol. 22, No. 26, 2006
Figure 12. Force versus relative area for several representative scales of observation. The correlations are all negative because the adhesive force decreased with increasing relative area. Linear regressions were performed for the three indicated scales, and regression coefficients are shown (R2). Although we chose to show representative data for three scales of observation, this regression was performed using many other triangle sizes, and the cumulative R2 values over all of the scales of observation tested are shown in Figure 13.
Figure 13. Siegmann plot showing regression coefficients versus observational scale for (1 × 1) µm2 measurements. Although the R2 values are never high, the pattern of the plot suggests that adhesion may be controlled by competing mechanisms or different mechanisms, depending on the scale, because two local maxima in R2 were seen.
bonding model (DBM) was employed. Area-scale analysis provides a fractal-based representation of the original surface that shows how the surface area available for bonding interactions depends on the scale on which the interaction is taking place. Differences in relative area from scale to scale tend to be largely independent of the original measured region. Whereas Rq depends only on the height measurements and not their relative positions or their spacing, the area-scale fractal representation depends on both the spacing of the measured heights and the order of the individual height measurements and therefore contains more information about the texture and could be said to be a more accurate representation of the texture. Geometric features of rough surfaces are not unique but vary as a function of the scale of observation, or calculation. The mathematician Lewis Fry Richardson first pointed out this phenomenon in the early 1900s as he speculated on the length of the coastline of Britain and other natural geographic barriers, noting that their length would change as the measurement scale
Emerson et al.
changed.49 Area changes with the observation scale in the same way as does length. On sufficiently fine scales, even smooth and highly polished surfaces appear to have some chaotic components to their texture, which is commonly called roughness. Below the scale on which a surface begins to appear rough, observations on finer scales, or resolutions, tend to reveal features that are not evident on larger scales. These features increase the observed surface area. The observed surface area therefore tends to increase as the scale of observation diminishes. The scale-based comparison determines the patterns in the correlations between the geometric parameters, such as the relative area and aspect ratios of the pits and valleys, and the adhesive force as a function of scale. On each scale available in the measured texture, for the several surfaces being studied, the adhesive forces for each surface were regressed against the geometric parameters for those surfaces, and the correlation coefficient (R2) was calculated. The R2 values were then plotted versus the scale of observation. A pattern of increasing R2 values to some relative maximum is an indication that the scale corresponding to the maximum is important to the adhesion mechanism. This type of analysis has never been applied to the description of bacterial adhesion to a textured surface or to any similar biological system. However, this methodology has proved successful for several diverse systems, such as the characterization of skin surfaces,50 chocolate cross-sections,51 and anthropological tooth samples.52 In the case of grit-blasted surfaces, the discrete bonding model (DBM), using the results of the area-scale fractal analysis, was able to correlate substrate roughness with adhesive strength with R2 values of nearly 0.9 for thermal spray coatings.20,29 However, this high correlation was achievable only when a single type of alloy was considered, indicating that the effects of different chemistries of the surfaces will always play some role, in addition to surface texture. The DBM supposes that the adhesive force is dependent on the number of individual adhesive bonds that form and that the number of bonds depends on the surface area available for forming the bonds. Because the surface area depends on the scale of observation, or calculation, then clearly the scale is important. In the present study, all measurements show a negative relation between the adhesive force and relative area, meaning increased relative area corresponds to decreased forces. The negative relationship between relative area and adhesive forces suggests that a modification of the Discrete Bonding Model may be needed. Increased roughness of a certain kind of character could lead to discontinuous contact, thus reducing the bonding opportunities as the roughness increases. When the slopes on the surface are sufficiently steep and the valleys and pits are sufficiently deep and narrow, the bacteria may not conform to the surface, and the contact will be discontinuous. This discontinuous contact should be scale-dependent, at least over some scale ranges. Discontinuous contact has never been studied for its role in biological adhesion, but it has been found to be important in controlling capillary adhesion leading to friction on materials such as ski bases and hard drives.48,53 A possible (49) Mandelbrot, B. B. Fractals, Form, Chance, and Dimension; Freeman: San Francisco, 1977. (50) Articus, K.; Brown, C. A.; Wilhelm, K. P. Skin Res. Technol. 2001, 7, 164-167. (51) Pedreschi, F.; Aguilera, J. M.; Brown, C. A. Int. J. Food Prop. 2002, 5, 523-535. (52) Scott, R. S.; Ungar, P. S.; Bergstrom, T. S.; Brown, C. A.; Grine, F. E.; Teaford, M. F.; Walker, A. Nature 2005, 436, 693-695. (53) Kennedy, F. E.; Brown, C. A.; Kolodny, J.; Sheldon, B. M. ASME J. Tribol. 1999, 121, 968-974.
Chemistry Vs Texture Vs S. epidermidis Adhesion
explanation for the increasingly strong correlations between increasing adhesive strength and decreasing relative area below scales of 100 nm2 is that the bacteria cannot conform to the valleys and pits on the surface on these scales. Therefore, the actual contact area may even be less than the projected area of the rough surface. Whereas on larger scales the texture features may provide more area for bonding and increasing the bond strength with roughness, on fine scales certain kinds of texture features limit the contact area and decrease the adhesive strength. For certain bacteria, particularly gram-negative strains, it is expected that extracellular polymeric substances (EPS), especially lipopolysaccharides (LPS), polysaccharides, and proteinaceous molecules such as flagella and fimbriae, could penetrate into these smaller areas and aid in the adhesion. It appears that, under the conditions of this study, the S. epidermidis cells were either not producing EPS or their surfaces may have been dominated by protein molecules that were not of the right conformation or did not have a strong affinity for the substrate features. The pattern of correlation coefficients with respect to the scale of observation in Figure 13 is intriguing, even though the R2 values are low. There are two regions of the plot where the R2 values improve, suggesting that there may be two distinct bonding mechanisms controlling bacterial adhesion. On the largest scale of observation, 0.5 µm2, approaching the size of a bacterium, the R2 values were the highest. Perhaps on these scales and larger, increasing area allows for increased bonding, possibly laterally on the sides as well as at the end of a bacterium. This would be consistent with the original discrete bonding model, which proposes that increased area leads to increased bonding. However, this assumes uniform and complete contact over the available area. Another texture-related mechanism influencing adhesion could be limiting contact with increasing roughness because a bacterium is not compliant enough to maintain contact over sharp features of the surface. Hence, we are proposing a modification to the discrete bonding model that recognizes the diminishing adhesive strength with limited contact. It could be the combination of the two mechanismssone increasing and the other decreasing adhesive force with increased roughnesssthat leads to the low correlation between surface roughness and adhesion. On sufficiently fine scales, the adhesive force limitations of discontinuous contact start to dominate. Below 100 nm2, well away from those larger scales where the contact area could increase significantly through lateral bonding, one mechanism is clearly dominant, and the strength of the negative correlation between adhesive strength and roughness begins to increase regularly. Other factors may have contributed to the weak correlations between relative areas and adhesive strength, including region(54) Friedli, G.-L. Interaction of Deamidated Soluble Wheat Protein (SWP) with Other Food Proteins and Metals; University of Surrey: 1996.
Langmuir, Vol. 22, No. 26, 2006 11321
to-region variations in surface texture and the dependence of bond formation on chemical properties of the molecules. To more completely test the applicability of the DBM with the modification for discontinuous contact for bacteria, the variations in surface texture at the level of the region over which the force measurements are being made need to be established so that the number and resolution of the texture measurements can be adjusted appropriately. Variations in the chemistry of the surfaces also need to be considered because composition will influence the nature and strength of the bonding.
Conclusions A series of SAMs and related compounds were prepared on gold substrates so that chemical and textural properties could be related to bacterial adhesion forces. The surface hydrophobicity and the root-mean-square roughness values were insufficient parameters to correlate with the adhesion force of S. epidermidis to each surface. The surface texture was characterized through a fractal tiling algorithm so that the critical areal scale important for the bacterial adhesive interactions could be determined. Contrary to expectations, our calculations showed a negative correlation between bacterial adhesion and surface roughness. This may be because most previous studies have focused on larger-scale surface roughness and have measured bacterial retention rather than adhesive forces. Whereas the original discrete bonding model, which is based solely on increasing available area and therefore opportunities for bonding on finer scales, does not appear to be appropriate for the present system, we suggest that modifications could be made to this model to account for discontinuous contact. The model should describe the case where the valleys and pits in the substrate are deep and distinct enough so that the bacteria do not conform fully to the surface and the contact area available for bonding is diminished. Acknowledgment. This publication was made possible in part by a grant from the National Science Foundation (BES0238627). We also acknowledge the donors of the Petroleum Research Fund of the American Chemical Society for partial support of this work (PRF 38988-G2). We thank Dr. Stephen Heard, Mr. Eftim Milkani, and Mr. Pete Driscoll for helpful discussions throughout the course of these studies, Ms. Xiaoshu Dai and Professor Satya Shivkumar for their assistance with the SEM measurements, and Dr. Christopher Lambert for helpful comments on a previous version of this manuscript. Supporting Information Available: Representative examples of reproducibility in force measurements made for cellular probes prepared over a period of several months. This material is available free of charge via the Internet at http://pubs.acs.org. LA061984U
