Molecular unbinding forces between protein-specific ligands and surface-bound proteins have been recently measured by using AFM (see Table 1). The force required to separate a ligand from its specific binding site is different from the force needed to remove a non-specifically bound ligand (Pierce et al., 1994b). Since the AFM can measure the differences of the specific adhesion forces in a spatially resolved manner, the ‘‘specific adhesion force contrast’’ can be used to identify bound molecules and determine their distribution on the surface of the specimen. When the specificity of the protein interaction results in a force contrast, we can separate the topographical and chemical surface information. In biological systems, where molecular recogni-tion events determine the specificity of interacrecogni-tions, the force contrast can be used to identify and map the distribution of those surface-immobilized proteins which are capable of binding AFM tip-bound ligands.
The unbinding force (i.e. rupture force) of protein–ligand interactions can be measured by employing the force–distance curve of the AFM. In a typical ligand rupture experiment the protein is bound to a substrate surface and the ligand bound to the AFM probe. The ligand-covered probe is then brought into contact with the substrate and binding takes place. The probe is then withdrawn from the surface, pulling the ligand out of its binding pocket. Then, the deflection of the lever, which is similar to a force exerted on the probe by the protein/protein-specific ligand interaction, is recorded. The lever deflection is measured as it approaches and as it retracts from the substrate surface. The difference between these two traces occurs through adhesion between the probe and its surface and can be attributed to the protein/ligand interaction. The maximum adhesive force measured is thus known as the ligand unbinding force. In this article, we review AFM measure-ments of the unbinding force between surface-bound proteins and an AFM tip chemically modified with a protein-specific ligand.
2.2. Theory
When an AFM is used to measure the unbinding force between a protein and a protein-specific ligand, it is important to convert the plot of a deflection signal from a position sensitive photo-diode (PSPD) to a force unit. The cantilever deflection signal, measured in units of nanoamperes (nA) or voltage (V), needs to be first converted to a deflection distance (nm), d, using the gradient (sensitivity, nm/A or nm/V) of the linear portion of the retract trace in the contact region of the force curve (Allen et al., 1996). This deflection distance, d, is then converted into a force value (nanoNewton, nN) acting on the AFM probe using the cantilever spring constant, k, and Hooke’s law (Allen et al., 1997):
F¼ k d (1)
The spring constant of a cantilever is determined from the individual frequency resonances with its shape factor which is expressed as where L is the length of cantilever, w the width of the cantilever, r the density of the cantilever material, E the elastic modulus (Young’s modulus) of the cantilever material, and f is the mea-sured resonant frequency (Cleveland et al., 1993). In this force curve, the unbinding force (i.e. adhesion force) is characterized as the maximum force needed to begin separation of the two paired partners after contact. Therefore, protein–ligand (or protein–
protein) interaction between a tip coated with one half of a pair of the interacting species and a surface coated with the other half, can be identified by the increase in the magnitude of the force.
In most experimental set-ups, the average unbinding force as a function of the pulling speed is usually measured. With a series of generally accepted assumptions: (1) the rupture time is longer than diffusional relaxation time, (2) the acting force on the complex increases with a constant loading rate and (3) the binding process is on the quasi-equilibrium state (Evans and Ritchie, 1997, 1999), it can be shown that the most probable unbinding force is stated as
F¼kBT where r is the loading rate, defined as the time derivative of the force applied to the bond. The loading rate r is assumed to be equal to the product of the force constant of the AFM cantilever and its retraction speed. It is equal to the true loading rate r only if the bond is fixed in position. This assumption has been used to analyze experimental AFM results for intermolecular protein–
ligand interactions (Paci et al., 2001).
2.3. Measurements
During a force measurement cycle (Fig. 1A), the piezo-scanner moves toward the AFM tip by piezo expansion and velocity is kept constant until it is brought into contact with the tip (point B). As the forward motion continues, the cantilever is pressed into the sample surface until a point of maximum load is reached (point C). The direction of motion is then reversed and the piezo-scanner is withdrawn from the AFM tip. During the retraction process for the force measurement, i.e., when the piezo-scanner withdraws from the tip surface, the tip adheres to the surface due to the interaction between the tip and the sample. In addition, a change in the slope of the curve (from points C–D to D–E) will occur during the retraction process which can be attributed to a decrease in its effective spring constant. The strength of the unbinding force is then calculated from the difference between the maximum cantilever deflection (point E) during the retraction process of the curve and the point of zero cantilever deflection (point A).
However, we need to keep in mind that the interpretation of the unbinding force (i.e. adhesion force) may be complicated by
Table 1
Protein–ligand unbinding force
Molecular partners Substrate AFM tip Pulling
velocity (mm/s)
Average force (pN)
Reference
Avidin/biotin system
Avidin/biotin Biotin Avidin Not given 160 20 Florin et al. (1994)
Avidin/iminobiotin Avidin Iminobiotin Not given 85 10 Moy et al. (1994)
Streptavidin/biotin Streptavidin Biotin Not given 257 25
Avidin/desthiobiotin Avidin Desthiobiotin Not given 94 10 Moy et al. (1994)
Streptavidin/iminobiotin Streptavidin Iminobiotin Not given 135 15
Biotin/streptavidin Streptavidin Biotinylated BSA Not given 340 120 Lee et al. (1994a)
Strepavidin/biotin Strepavidin Biotin Not given 300 Allen et al. (1996)
Biotin/streptavidin Streptavidin Biotin Not given 200 Wong et al. (1998)
Avidin/biotin Avidin Biotin 5 173 19 Lo et al. (1999)
Strepavidin/biotin Strepavidin Biotin 5 326 19
Streptavidin/biotin Biotin Streptavidin 0.086–1 126 2.3–207 5.8 Yuan et al. (2000)
Strepavidin/biotin Strepavdin Biotin 1–200 167 20–442 17 Lo et al. (2001)
Strepavidin/biotin Strepavdin Biotin Not given 454 Stevens et al. (2002)
Antigen/antibogy (Ag/Ab) interaction
Fluorescein/anti-fuorescyl Ab Anti-fuorescyl Ab Fluorescein 0.182 200 Stuart and Hlady
(1995) Human serum albumin
(HSA)/anti-HSA
HSA Anti-HSA 0.2 244 22 Hinterdorfer et al.
(1996)
Ferritin/anti-ferritin Anti-ferritin (Ab) Ferritin Not given 49 10 Allen et al. (1997)
Intercellular adhesion molecule-1
(ICAM-1)/anti-ICAM-1 Ab
ICAM-1 Anti-ICAM-1 Ab 4.68 100 50 Willemsen et al.
(1998) Ryanodine receptor 1
(RYR1)/anti-RYR1
Ryanodine receptor 1 (RYR1)
Anti-RYR1 0.066–0.3 42–73 Kada et al. (2001)
Human chorionic gonadotrophin (hCG)/anti-hCG
hCG Anti-hCG Not given 507 Stevens et al. (2002)
Mercaptopropanoic acid derivative of atrazine (MPAD)/anti-MPAD
Anti-MPAD Not given 511.7 244.1 Kaur et al. (2004)
Sendai-purple membrane (Sendai-PM)/anti-Sendai antibody
Sendai-PM Anti-Sendai 0.2–5 70–170 Kienberger et al.
(2005)
Other protein/protein interaction
Cell adhesion proteoglycans Proteoglycans Proteoglycans Not given 125 Dammer et al. (1995)
Muscle proteins actin/myosin Ultraavidin-coated fluorescent acrylamide nanobead with biotinylated myosin
Biotin 0.0334 14.8 4 and
24.7 1.4
Nakajima et al. (1997)
Recombinant P-selectin/P-selectin glycoprotein ligand-1 (PSGL-1)
P-selectin PSGL-1 2.8 165 Fritz et al. (1998)
Ab single-chain Fv (scFv) fragment/fluorescein
Ab single-chain Fv fragment (scFv) with C-terminal Cys residue
Fluorescein 1 50 4 Ros et al. (1998)
Citrate synthase/E. coli chaperonin GroEL
GroEL Citrate synthase Not given 420 100 Vickier et al. (1998)
b-Lactamase/E. coli chaperonin GroEL
GroEL b-Lactamase Not given 240 70
Insulin/insulin Insulin Insulin Not given 1300 Yip et al. (1998)
Myelin basic protein/lipid bilayers
Lipid bilayers Myelin 1 140 60 Mueller et al. (1999)
Osteopontin/avb3integrin avb3Integrin Osteopontin 1–50 50 Lehenkari and Horton
(1999) Vascular endothelial
(VE)–cadherins–Fc
VE–cadherins–Fc VE–cadherins–Fc 0.2–4 15150 Baumgartner et al.
(2000)
contributions from non-specific interactions.Fig. 1B shows a schematic diagram of a typical non-specific interaction. The difference between the curves ofFig. 1A (specific interaction) andFig. 1B (non-specific interaction) is thatFig. 1A shows a change in the slope during the retraction process which is a result of a decreased effective spring constant. This change indicates that at the beginning of the retraction process, the cantilever is relaxed, while during further retraction, the cantilever as well as the spacer, ligand and protein, become stretched. In contrast, the curve of Fig. 1B retains the same slope during the process of retraction. This is a non-specific interaction between a ligand at the tip and a protein on the substrate without the involvement of spacers, so therefore, we can only assume the cantilever is bent (Willemsen et al., 1998).
Non-specific interactions can arise from an improper spatial orientation of the protein and/or ligand that prohibits specific binding during the approach cycle and/or while in contact. Thus the challenge is to identify those interactions that are specific, as opposed to those that are non-specific, in nature. This separation can be accomplished by conducting control experiments where, for example, the binding site on one of
the partners in the protein–ligand or protein–protein pair is blocked. Because of the often random spatial orientation of the binding partners at the tip and substrate surface, it is usually necessary to collect several hundred individual force curves to determine the distribution of the binding force.
Another essential requirement for the quantitative mea-surement of interaction forces is to develope an accurate method for the calibration of the spring constant, k, of the AFM cantilever. Since widely used commercially available silicon nitride cantilevers vary significantly in their spring constant (Cleveland et al., 1993), several methods for calibration of AFM cantilevers have been proposed and used in various laboratories (Cleveland et al., 1993; Hutter and Bechhoefer, 1994; Senden and Ducker, 1994). However, it has been shown that measured values for the spring constants can differ significantly from the original manufacturer’s specifications (Cleveland et al., 1993).
In light of the fact that the AFM is now used more frequently as a technique to measure interaction forces, it has become more critical that exact values are measured and determined.
However, an important feature of these interaction assays is that
Table 1 (Continued )
Molecular partners Substrate AFM tip Pulling
velocity
ConA Not given 86 2.6 Chen and Moy (2000)
Lactose/bovine heart (BHL) BHL Lactose 1.2 34 6 Dettmann et al. (2000)
Lactose/lactose-binding immunoglobulin G (IgG)
Lactose IgG 1.2 36 4
Lactose/Viscum album (VAA) VAA Lactose 1.2 47 7
Lactose/Ricinus communis (RCA) Lactose RCA 1.2 58 9
Asialofetuin (ASF)/BHL BHL ASF 1.2 37 3
ASF/VAA VAA ASF 1.2 43 5
ASF/IgG ASF IgG 1.2 45 6
ASF/RCA ASF RCA 1.2 65 9
Nitrilotriacetate (NAT)/
histidine 6 (His6)
NAT His6 0.09–0.27 150–194 Kienberger et al.
(2000) Human ocular mucins Human ocular mucins Human ocular
mucins
0.0205–0.3488 100–1200 Berry et al. (2001)
Ligand (RGD)/human platelet aIIbb3receptor
Human platelet aIIbb3
receptor
Ligand (RGD) 0.2 93 Lee and Marchant
(2001)
Phospholipid bilayers/recoverin Phospholipid bilayer Recoverin 5 48 5 Desmeules et al.
(2002) Function-associated antigen-1
(LFA-1)/intercellular
adhesion molecule-1 (ICAM-1)
iICAM-1 iLFA-1 0.1–15 Not given Zhang et al. (2002)
Function-associated antigen-1 (LFA-1)/intercellular
adhesion molecule-1 (ICAM-1)
hICAM-1 iLFA-1 0.1–15 Not given
Ganglioside GM1/cholera toxin B-oligomer (ctB)
Ganglioside GM1 Cholera toxin B-pentamer (ctB)
0.04–4 54 46–62 30 Cai et al. (2003)
P-selectin/P-selectin glycoprotein ligand-1 (PSGL-1)
P-selectin PSGL-1 0.25 20 Marshall et al. (2003)
Receptor-associated protein
Cell adhesion proteoglycan Proteoglycans Proteoglycans Not given Not given Popescu et al. (2003)
Epimerase AlgE4/mannuronan Epimerase AlgE4 Mannuronan 0.2–4 74–144 Sletmoen et al. (2004)
Human a5b1integrins/GRGDSP GRGDSP a5b1Integrins 1–50 32 2 Kokkoli et al. (2004)
the measured forces are not contingent only on the nature of the pair, but depend also on the loading rate at which force is applied to the complex. The unbinding force usually scales linearly with the logarithm of the loading rate. For a single a single barrier, this would give rise to a simple, linear force spectrum versus ln(r) (Fig. 2A). In cases involving more than one barrier, and assuming that all barriers lie along a single, one-dimensional escape path, the spectrum is predicted to follow a continuous sequence of linear regimes (Fig. 2B) (Merkel et al., 1999; Nevo et al., 2003).
2.4. Examples
2.4.1. Ran–importin b1 interaction
Nevo et al. (2003)applied AFM to study the interaction of Ran with the nuclear import receptor importin b1 (impb) at the single-molecule level. It was shown that the probability density of rupture events under a ramp of force is described by a random (Markov) process, which predicts the likehood of bond survival over time. The most probable force for unbinding, taken as the maximum of the force distribution, is related to the loading rate through the Eq.(3). They first studied the interaction between impb and Ran loaded with GDP. Next, they investigated the interaction between impb and Ran loaded with the nonhydrolyz-able GTP analog GppNHp. In contrast to GDP, which can be readily loaded into Ran, loading GppNHp into Ran is 80%
efficient, even in the present of alkaline phosphatase. Unbinding of Ran–GppNHp from impb gave rise to a distinct set of force distributions, which were shifted to higher forces compared with those measured for Ran–GDP–impb. These distributions were proven to have a unique bimodal appearance (Fig. 3A).
Moreover, a semilog plot of the most probable unbinding forces measured for each of the two populations was also shown (Fig. 3B). Those results indicate that interaction between Ran–
GppNHp and impb could lead to two distinct bound states, each associated with a dissociation path of its own.
Fig. 2. (A) Dynamic force spectra. The most probable unbinding force vs. the logarithm of the loading rate is a straight line in each regime of the spectrum.
(B) A piece-wise linear dynamic force spectrum for a cascade of two sharp energy barriers, which is predicted to follow a continuous sequence of linear regimes.
Fig. 1. (A) A schematic diagram of a typical specific force measurement curve, using an AFM to measure the force required to separate individual intermo-lecular protein–ligand (protein) interaction. Approach (dash line): as the piezo-scanner moves towards the AFM tip at a constant velocity from A to B, and then where the AFM tip comes into contact with the surface at point B. As the scanner continues approaching the tip, the cantilever bends upward until it reaches point C. The retract (solid line): occurs when the tip reaches point C, at which the piezo-scanner moves away from the AFM tip and the cantilever then begins to retract. During the retraction process, the tip adheres to the surface due to the interaction between the tip and sample and results in a change in the slope of the curve (points C–D to D–E). As the scanner continues to retract, the cantilever is bent downwards until it reaches point E. The tip-sample starts to break from point E. Finally, the cantilever then returns to its original equilibrium state at point A. (B) A schematic diagram of a typical nonspecific force measurement curve. (A) A change in the slope of the curve (points C–D and D–E) during the retraction process. The curve of (B) retains the same slope (points C–E) during the process of retraction.
2.4.2. Mercaptopropanoic acid derivative (MPAD)–anti-MPAD interaction
Kaur et al. (2004)used an AFM to directly evaluate specific interaction forces between pesticides and antibodies on a bio-sensor surface. An oriented immobilization of antibodies against two herbicide molecule mercaptopropanoic acid derivatives of atrazine (MPAD) and 2,4-dichlorophenoxyacetic acid (2,4-D) on a gold substrate, was carried out to simulate an immuno-biosensor surface. The adhesive force between the immobilized antibodies and their respective antigens was measured by force spectroscopy using a hapten-carrier protein functionalized AFM cantilever. They demonstrated typical force–distance curves for the interaction between the BSA–
MPAD and BSA–2,4-D functionalized cantilevers with a corresponding antibody as well as only protein A-coated surfaces. The interaction between the anti-atrazine antibody
and BSA–atrazine-coated cantilevers showed a wide range distribution value of the pull-off force, varying from 200 to 940 pN with the majority of the data falling between 200 and 500 pN. The mean value was calculated to be 511.62 244.1 pN (S.D.). This mean value was significantly lower (P < 0.001) when the BSA–atrazine-coated cantilever was allowed to interact with only the protein A-coated surface (31.56 19.8 pN). Furthermore, in the case of the BSA–2,4-D-coated cantilevers, the pull-off force with the anti-2,4-D-BSA–2,4-D-coated surface varied from between 80 and 700 pN. The mean value of 228.62 163.1 pN was significantly higher (P < 0.001) than the corresponding value of 25.66 33.5 pN for the BSA–2,4-D interaction with protein A.
2.4.3. Receptor-associated protein (RAP)-receptor on fibroblast cells
Osada et al. (2003)used an AFM to examine the distribution of the receptor-associated protein (RAP) binding protein and the adhesion force between a RAP and its binding protein on living fibroblast cells. The distribution of the RAP binding protein was obtained at 256 (16 16) locations in 2 mm 2 mm sections over the surface of the living cells. They also measured the adhesion forces between the RAP and the binding protein with an AFM tip functionalized with RAP. It was found that the most frequently observed force value was 120 pN. In the presence of RAP in the scanning solution, the force curves showed a large decrease of adhesion force. The results indicated that the measured adhesive forces represent specific binding between the RAP and the binding protein.
2.4.4. Epimerase AlgE4–mannuronan interaction
Sletmoen et al. (2004)determined the interaction between AlgE4 epimerase and mannuronan by dynamic force spectro-scopy. Alginate biosynthesis involves C-5-mannuronan epi-merases catalyzing the conversion of b-D-mannuronic acid to a-L-guluronic acid at a polymer level. It is known that mannuronan epimerases are modular enzymes where their various modules can yield specific sequential patterns of the converted residues in their polymer products.Sletmoen et al.
(2004) studied the specific unbinding force between the molecular pairs of mannuronan and AlgE4 as well as its two modules, A and R, respectively, as a function of the force loading rate. They found that the mean protein-mannuronan unbinding force was in the range 73–144 pN depending on the protein, and possessing an loading rate of 0.6 nN/s which increased with an increased loading rate. Moreover, the position of the activation barrier was determined to be 0.23 0.04 nm for the AlgE4 and 0.10 0.02 nm for its A-module. The lack of interaction observed between the R-module and mannuronan suggests that the A-module contains the binding site for the polymer substrate.
2.4.5. Glucagon–anti-glucagon interaction
Lin et al. (in press) studied the dynamic responses of glucagon/anti-glucagon pairs with multiple pull-off steps to pulling velocity and pH by AFM. Force–distance curves of a specific glucagons–anti-glucagon interaction system with
Fig. 3. Molecular unbinding of different Ran–impb complexes. (A) Unbinding force distributions. To reveal dynamic aspects of bond rupture, measurements were carried out at various probe velocities, leading to different loading rates.
(B) Force spectra. Shown in the main figure are plots for the two populations observed for Ran–GppNHp–impb (solid lines) and the single population observed for RanQ69L–GTP–impb (dotted line). Inset, curves obtained for Ran–GDP–impb (solid line) and RanQ69L–impb (dotted line); dashed line describes noise level. Reprinted with permission fromNevo et al. (2003).
Copyright 2003 Nature Publishing Group.
mono-, di-, and multi-unbinding events were recorded, which were attributed to a single, sequential or multiple breaking of interacting bond(s) between glucagon and anti-glucagon. They reported the dynamic response of glucagons–anti-glucagon pairs to various pulling velocities (16.7–166.7 nm/s). It was found that the mean value of the unbinding force was shifted toward higher values with increasing pulling velocity at pH 7.
Moreover, the dynamic response of glucagons–anti-glucagon pairs to pH (4–10) with different pulling velocities was reported. Within the acid range, the bond strength between the glucagon/anti-glucagon complex showed a rapid increase from pH 4 to 7 and reached a maximum (256.4 48.9 pN at 166.7 nm/s) at neutrality, followed by a sharp decrease with increasing pH (7–10) (Fig. 4). This study demonstrated that the pH dependence of multiple antigen–antibody bond-rupture forces could be measured by a force-based AFM biosensor.
3. Determination of dissociation rate