Changes in microbubble dynamics upon adhesion to a solid surface

Microbubbles are currently used in biomedical ultrasonics for contrast-enhanced ultrasound imaging, super-resolution imaging, molecular imaging, and image-guided therapy. Biomedical microbubbles comprise a 1–10 μm diameter gas core suspended in an aqueous medium and commonly coated with a phospholipid monolayer shell.¹ The resonance frequency of a lipid-coated microbubble is typically between 1 and 10 MHz, depending on the diameter and shell stiffness.² Microbubble resonance fortuitously is in the range of most clinical ultrasound imaging and therapy devices. Optimal sensing and actuation of the microbubble requires matching of the transducer bandwidth and the ultrasound pulse center frequency to the microbubble resonance frequency.³ 

Prior research has primarily focused on characterizing microbubble resonance in an isotropic medium.^4,5 However, the interaction between an acoustically driven microbubble and a surface is important for many applications. There have been some studies comparing the acoustic response and stability of individual microbubbles near a boundary.^6–10 In each of these studies, microbubbles were driven with sufficiently large acoustic pressure to induce a nonlinear response and microbubble instability. The response of adherent microbubbles under stable, linear oscillations, although fundamental to understanding the physical process, has received comparatively little attention. In this letter, we investigate the dynamics of microbubbles adherent to a rigid substrate under stable, linear oscillation using a photoacoustic technique. A modulated laser source was used to generate ultrasound in the liquid adjacent to a microbubble to drive oscillations. The resulting radial displacement, on the order of tens of picometers, was detected using forward light scattering. We show that adhesion to a rigid substrate significantly increases the resonance frequency over that of non-adherent microbubbles resting against the wall. The response is consistent with an increase in stiffness, rather than an increase in mass, and can be captured quantitatively through an increase in the apparent shell elasticity. Bound microbubbles also show a decrease in both the damping ratio (friction) and the maximum radial displacement.

In this study, phospholipid-coated microbubbles were generated by shaking.¹¹ The microbubble shells comprised 90 mol. % of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) and 10 mol. % of 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (DSPE-PEG2000). Biotinylated microbubbles were prepared by substituting 10 mol. % of their PEGylated lipid (1 mol% overall) with 1,2-distearoyl-sn-glycerol-3-phosphoethanolamine-N-[biotinyl-(polyethylene glycol)-2000] (DSPE-PEG2000-B). These lipids were suspended in phosphate buffered saline (PBS) at a lipid concentration of 2 mg/ml and were mixed and heated to 45 °C by probe sonication. A 3-ml glass serum vial was filled with lipid suspension and sealed with a gas headspace containing perfluorobutane (PFB). Microbubbles were generated by vigorously shaking the serum vial for 40 s using a dental amalgamator, and the vial was quickly quenched to room temperature in an ice bath. A centrifugation technique¹² was used to wash and separate the microbubbles by removing undesired bubble sizes and excess lipid in the solution. Microbubbles were immediately diluted with PBS (approximately 5 × 10⁵ bubbles per ml and 4–10 μm in diameter) and pipetted between a microscope slide and a modified glass coverslip with a thin gasket separating them.

We explored adherent and non-adherent microbubbles, where adhesion resulted from either specific (biotin/avidin) or nonspecific (lipid/glass) interactions. For specific binding, biotinylated microbubbles were brought into contact with a streptavidin-coated glass coverslip, which had been incubated with 33 μg/ml streptavidin for one hour, rinsed with PBS, and then air-dried immediately prior to the experiment. This condition was compared to a non-adherent microbubble condition, where non-biotinylated microbubbles were brought into contact with the streptavidin-coated glass coverslip surface. For nonspecific binding, non-biotinylated microbubbles were brought into contact with a clean glass coverslip. This condition was compared to a non-adherent condition, where non-biotinylated microbubbles were brought into contact with a glass coverslip surface where a supported lipid bilayer was deposited to block adherence. The glass coverslip had been incubated with 2 mg/ml DPPC for one hour, rinsed with PBS and then air-dried immediately prior to the experiment. Microbubbles were allowed to equilibrate with the glass coverslips for 30 min prior to testing.

Total internal reflection fluorescence (TIRF) and epifluorescence microscopy were used to characterize the adhesion between the microbubbles and the glass coverslips. Here, microbubbles were fluorescently labeled with DiO (ex/em = 484/501 nm) during preparation. Figure 1(a) shows a schematic of the prism-TIRF microscopy setup, which incorporated a Nikon inverted microscope with a 60X objective. A 491-nm excitation laser source was directed through an index matched prism to illuminate the sample surface at an incident angle greater than the critical angle for total internal reflection at the glass-water interface. This generated an evanescent field extending from the glass-water interface (∼100 nm deep) where only the fluorophores in the microbubble shell near the surface (i.e., in the region associated with adhesion contact) were excited. The emitted light from the microbubble contact region was then directed with a dichroic mirror (505 nm) and captured by an electron-multiplying CCD camera. For epifluorescence microscopy, the microbubbles were illuminated through the microscope objective using a 488/10 nm bandpass excitation filter, and fluorescent images were captured with the focal plane at the microbubble mid-plane (i.e., the widest projection of the bubble) to quantify the resting radius.

Figures 1(b) and 1(c) show representative epifluorescence and TIRF microscopy images of typical microbubbles for each of the different adhesion conditions. The bubble diameter was quantified using the averaged outer full-width-at-half-maximum taken from the vertical and horizontal pixel intensity line profiles across the center of the epifluorescence image. Similarly, the diameter of the contact region was quantified using the averaged full-width-at-80%-maximum from the line intensity profiles for TIRF images. Figures 1(b) and 1(c) show measurements of adhesion contact radius vs the bubble resting radius for the different conditions. It is interesting to note the irregular and less consistent contact regions formed for the biotin/avidin condition, possibly owing to heterogeneity in the physisorption of streptavidin to the glass, or streptavidin coating some of the microbubbles.

Kooiman et al.¹³ and Lajoinie et al.¹⁴ reported microbubble adhesion deformation into a spherical cap geometry and quantified this deformation by calculating a contact ratio. Here, the contact ratio is defined as $Rcontact/R0$⁠, where $Rcontact$ is the radius of the adhesion contact region and $R0$ is the resting radius of the microbubble. Biotinylated microbubbles resting against a streptavidin surface had a median contact ratio of 0.28, compared to non-biotinylated microbubbles that had a median contact ratio of 0.13. This result confirmed our assumptions that the biotinylated microbubbles were adhered via specific interactions and that non-biotinylated microbubbles were not adhered to the streptavidin surface. Similarly, microbubbles resting against a clean glass surface had a median contact ratio of 0.56, compared to those resting against a pre-coated lipid-on-glass surface that had a median contact ratio of 0.15. Again, this result confirmed nonspecific adhesion and non-adhesion for the two conditions, respectively. Brownian motion of microbubbles in the non-adherent conditions was also observed and supports our findings. A Shapiro–Wilk test for normality showed that the contact ratios for each case were not all normally distributed, so nonparametric statistical testing was conducted using OriginPro software and was used for all the data throughout this study. Using the Mann–Whitney U-test, the contact ratios were significantly greater for the adherent microbubbles than for the non-adherent microbubbles for both specific and nonspecific interactions (p < 0.01).

The resonant behavior of single microbubbles bound to a solid surface were quantified using a photoacoustic technique.¹¹ Briefly, a microbubble was driven into small-amplitude radial oscillations by ultrasound waves generated by an amplitude-modulated continuous wave laser source. This 1550-nm infrared generation laser source illuminated water in the sample (3 μm 1/e diameter and 5.5 mW) approximately 100 μm away from the bubble. Absorption of the laser energy by water caused localized thermal expansion and ultrasound generation through the photoacoustic effect. We previously estimated that the laser source also produces a local temperature rise of approximately 2.5 °C at the microbubble position, and all measurements were conducted at room temperature. A second continuous wave laser source with a wavelength of 532 nm (9.5 μm 1/e diameter and 1.05 mW) was used to illuminate the microbubble and track its oscillations by means of forward scattered light. A photodetector and lock-in amplifier were used to track the light scattering from the microbubble at a single driving frequency. The frequency of the amplitude modulation for the generation laser was swept across a range of at least 2 MHz in 50 kHz steps, producing a resonance curve for a single microbubble. A microscope image of the microbubble was captured prior to measurement using a CCD camera to determine the bubble resting radius.¹¹ 

Figure 2(a) shows resonance curves for a biotinylated microbubble and a non-biotinylated microbubble resting against streptavidin-coated glass, both with a resting radius of 2.7 μm. The adherent microbubble had a resonance frequency of 3.4 MHz, which was much higher than the non-adherent microbubble resonance frequency of 1.5 MHz. Figures 2(b) and 2(c) show plots of eigenfrequency vs bubble radius for the different conditions. As with the contact ratio measurements, more variability was seen in the frequency response for the biotin/avidin system than for the lipid/glass system.

The lipid shell elasticity, $χ$⁠, was calculated for each measured microbubble response using the linearized form of the modified Rayleigh–Plesset equation,^3,15–17 which describes the small-amplitude radial oscillations of a microbubble. The shell elasticity is given as

where $f$ is the bubble resonance frequency, $ρL$ is the density of surrounding liquid, $R0$ is the resting radius of the microbubble, $ζ$ is the damping ratio, $κ$ is the polytropic exponent for the gas core, $P0$ is the ambient pressure, and $σ$ is the surface tension. The surface tension was assumed to be zero to meet the criterion for bubble stability against dissolution in a saturated medium.¹⁸ The damping ratio is given by $ζ=Δf/2f$⁠, where $Δf$ is the full-width-at-half-maximum at the peak of the resonance curve. Equation (1) includes a correction for the expected (2/3)^0.5 decrease in resonance frequency from the acoustic effect of a rigid boundary.¹⁹ 

The biotinylated microbubbles adherent to the streptavidin surface had a greater median elasticity of $χ$ = 2.71 N/m, compared to the non-adherent microbubbles on the streptavidin surface where $χ$ = 0.57 N/m (p < 0.01). Similarly, microbubbles adherent to the clean glass surface had a greater median elasticity value of $χ$ = 2.84 N/m, compared to non-adherent microbubbles on the lipid surface where $χ$ = 0.57 N/m (p < 0.01). The apparent elasticity of adherent microbubbles agrees with our prior photoacoustic measurements,^11,19 while that for non-adherent microbubbles agrees well with a prior high-speed videomicroscopy study for similar shell compositions.²⁰ These median elasticity values were used with Eq. (1) to plot the solid curves seen in Figs. 2(b) and 2(c). The dashed curves represent the frequency response of uncoated bubbles, where $χ$ = 0 N/m. Interestingly, the frequency response of the adherent microbubbles was well-described by the modified Rayleigh–Plesset model even though the model assumes spherical oscillations in an unbounded medium.

Figures 3(a) and 3(b) show the damping ratio vs bubble radius for the same bubbles measured in Figs. 2(b) and 2(c), respectively. The solid lines are the median values. The adherent biotinylated microbubbles had a lower median damping ratio (ζ = 0.14) than the non-adherent microbubbles (ζ = 0.23) in contact with the streptavidin surface (p < 0.01). Similarly, the nonspecific adherent microbubbles had a lower median damping ratio (ζ = 0.10) compared to their non-adherent counterparts (ζ = 0.20) (p < 0.01).

The scatter plots in Figs. 3(c) and 3(d) show the maximum radial displacements vs bubble radius. The maximum radial displacements were calculated from the maximum signal amplitude of the resonance curve using the calibration procedure described previously.¹¹ In the conversion from the measured response to the radial displacement, we assume spherically symmetric oscillations for both adherent and non-adherent bubble populations. The displacement was also found from the measured damping ratio and resonance frequency using the harmonic oscillator model. The maximum radial displacement, $ΔRmax$⁠, is given by

where P_a is the acoustic driving pressure. The parameters $ζ$ and f were set to the corresponding median values determined experimentally. P_a was estimated to be 2.8 Pa for all cases from a best fit to the experimental data [solid curves in Figs. 3(c) and 3(d)]. Adherent microbubbles had lower maximum radial displacements than non-adherent microbubbles for both the specific and nonspecific interactions (p < 0.01).

To summarize, we report on the behavior of individual adherent microbubbles undergoing small-amplitude oscillations. While it is difficult to compare our findings directly with past literature due to differences in experimental parameters, we can compare our results to a similar study by Rooij et al.,¹⁰ who investigated the dynamics of individual biotinylated microbubbles with lipid coatings composed primarily of DPPC lipid that were bound to a streptavidin surface. They also reported higher resonance frequencies for adherent DPPC microbubbles compared to non-adherent microbubbles driven by 50 kPa ultrasound although the increase was not statistically significant. However, Rooij et al. also reported an increase in maximum radial displacements for adherent microbubbles compared to non-adherent microbubbles, while we observed a decrease in maximum radial displacements for adherent microbubbles. This disagreement could involve how Rooij et al. monitored the microbubble oscillations optically in one plane-parallel to the contact surface using a high-speed camera while our measurements are sensitive to any dynamic changes in the optical scattering cross sections of the microbubbles. We note that our reported maximum radial displacements could also be influenced by the contact ratio and non-spherical oscillations of microbubbles in the adherent bubble populations. The disagreement with maximum radial displacements could also be due to the differences in amplitudes of oscillation, where our study involved very small amplitudes of oscillation (tens of picometers) and Rooij et al.'s microbubbles were driven into larger amplitudes of oscillation (hundreds of nanometers).

Additionally, Rooij et al. used a polymer substrate (OptiCell) instead of glass. For comparison, we repeated our nonspecific binding experiment using the OptiCell (Fig. 4). TIRF microscopy confirmed adhesion and non-adhesion for the uncoated OptiCell vs lipid-coated OptiCell, respectively. Consistent with our results for glass, microbubbles adherent to the rigid polymer substrate showed an increase in resonance frequency, decrease in damping ratio, and decrease in maximum radial displacement. Changes in shell elasticity, damping ratio, and maximum radial displacement were all statistically significant (p < 0.01). Thus, our results are consistent for both glass and rigid polymer surfaces.

The surfaces used here (glass and OptiCell) are sufficiently rigid compared to the microbubble that their acoustic response is negligible. This was confirmed in our experiments without a microbubble, where no resonance peak was observed. Additionally, the non-adherent microbubble measurements for both substrates (coated-glass or coated-OptiCell) had very similar resonance frequencies and responses, leading us to conclude that the forward light scattering technique was independent of the substrate mechanics.

The mechanism for the observed stiffening is currently unknown. One possible explanation is that the contact line of the adhesion zone imparts an additional tension in the lipid shell that manifests as an increase in the apparent elasticity. For liposomes, adhesion is known to induce a significant force normal to the contact line of the adhesion zone.²¹ This adhesive tension was shown to deform a soft substrate (hence it was called a “traction force”), pulling up a rim of soft substrate material along the contact line.²¹ One can imagine a similar situation for microbubbles, where an additional tension in the shell induced by adhesion increases the overall system stiffness and apparent microbubble shell elasticity. Future work combining modeling and experimentation may further elucidate the mechanism.

In conclusion, we verified microbubble adhesion and non-adhesion under specific or nonspecific surface interactions by use of TIRF microscopy. We then measured stable, linear oscillations using a photoacoustic technique to show that adherent microbubbles exhibit an increase in resonance frequency, decrease in damping ratio, and decrease in maximum radial displacement compared to non-adherent microbubbles. Comparison to the linearized Rayleigh–Plesset model with shell viscoelastic terms indicates that adhesion stiffens the system, instead of increasing its inertia or friction. While the apparent shell stiffness is observed to increase, the stiffening effect occurs over the system as a whole. Our working hypothesis is that the adherent microbubble is effectively pinned at the microbubble/substrate contact perimeter, causing a restoring force to be applied across the rest of the microbubble that is in contact with the aqueous medium. In our prior photoacoustic measurements,^11,19 it is likely that the microbubbles were adherent as they gave similar shell elasticities. Future studies seeking to characterize the mechanics of free lipid-coated microbubbles should avoid adhesion by pre-incubating the substrate with phospholipid. The observed microbubble stiffening may lead to deeper understanding of the dynamics of microbubbles adherent to different surfaces and shows that surface forces dominate the acoustic image-bubble effect for small-amplitude oscillations.

This work was supported by the National Institutes of Health (No. R01CA195051). J.L. acknowledges support from the CU Boulder Mechanical Engineering Research Innovation Fellowship. D.F.K. acknowledges support from the U.S. Defense Threat Reduction Agency (Award No. HDTRA1-16-1-0045). The authors are grateful to Gazendra Shakya and Hendrik Vos for insightful discussions about the experimental results.