Molecular E ﬀ ects of Glycerol on Lipid Monolayers at the Gas − Liquid Interface: Impact on Microbubble Physical and Mechanical Properties

: The production and stability of microbubbles (MBs) is enhanced by increasing the viscosity of both the formation and storage solution, respectively. Glycerol is a good candidate for biomedical applications of MBs, since it is biocompatible, although the exact molecular mechanisms of its action is not fully understood. Here, we investigate the in ﬂ uence glycerol has on lipid-shelled MB properties, using a range of techniques. Population lifetime and single bubble stability were studied using optical microscopy. Bubble sti ﬀ ness measured by AFM compression is compared with lipid monolayer behavior in a Langmuir − Blodgett trough. We deduce that increasing glycerol concentrations enhances stability of MB populations through a 3-fold mechanism. First, binding of glycerol to lipid headgroups in the interfacial monolayer up to 10% glycerol increases MB sti ﬀ ness but has limited impact on shell resistance to gas permeation and corresponding MB lifetime. Second, increased solution viscosity above 10% glycerol slows down the kinetics of gas transfer, markedly increasing MB stability. Third, above 10%, glycerol induces water structuring around the lipid monolayer, forming a glassy layer which also increases MB sti ﬀ ness and resistance to gas loss. At 30% glycerol, the glassy layer is ablated, lowering the MB sti ﬀ ness, but MB stability is further augmented. Although the molecular interactions of glycerol with the lipid monolayer modulate the MB lipid shell properties, MB lifetime continually increases from 0 to 30% glycerol, indicating that its viscosity is the dominant e ﬀ ect on MB solution stability. This three-fold action and biocompatibility makes glycerol ideal for therapeutic MB formation and storage and gives new insight into the action of glycerol on lipid monolayers at the gas − liquid interface.


■ INTRODUCTION
Microbubbles (MBs) for contrast enhanced ultrasound imaging are typically between 1 and 8 μm in diameter and consist of a biocompatible lipid, protein, or polymer shell encapsulating a gas core, which is usually a perfluorocarbon. These MBs oscillate in response to ultrasound (US) excitation, and their high acoustic impedance mismatch with the surrounding tissue means that they act as efficient contrast agents for ultrasound imaging. 1,2 Therapeutic MBs, in which the shell is functionalized to attach a drug payload and/or targeting agents, have been emerging as a potential route for targeted, triggered-release, drug delivery systems. 3,4 Multifunctional MB coating has been used for magnetic resonance imaging contrast enhancement and gene delivery at the same time by coupling super-paramagnetic iron oxide nanoparticles with albumin shells. 5,6 Typically, the MBs have a therapeutic agent conjugated to it, either in liposomes, 7,8 oil nanodroplets, 9 or genes attached to the MB shell, 10,11 or in some cases the gas core itself may be therapeutic. 12,13 For clinical and preclinical applications, control over MBs' size distribution, stability, and mechanical response to US are key parameters typically considered when optimizing MB production. These are controlled by the gas type, the MB shell, and the properties of the solution phase. The MB shell introduces a resistance to gas permeation and thus together with surface tension has an impact on MB size and lifetime. 14−19 Phospholipid monolayers are a commonly used coating that has good shell resistance as well as a subharmonic response to low amplitude ultrasound. 20,21 The gas core, and its solubility in the surrounding medium, 22−24 also impact the MB lifetime with the use of higher-molecular-weight, lesssoluble perfluorocarbons (PFCs), leading to enhanced MB stability. 17,22,25,26 Finally, the properties of the surrounding liquid medium are also important during storage or in vitro studies, where parameters such as temperature, pressure, viscosity, and concentration of the dissolved gas all play a role. 22,27 MBs have been widely investigated to increase their lifetime and stability in vitro and in vivo, either by, for example, altering the lipid shell composition 28−30 or by the encapsulated gas. 26,31 Recently in our group, we studied saturation of the surrounding medium with liquid PFC (C 6 F 10 ) to increase the stability of MBs, and we showed that PFC molecules were incorporated in the shell lipid monolayer, which resulted in a 25% reduction in the surface tension that reduced the Laplace driving pressure for dissolution. 22 Other groups have demonstrated that incorporation of viscosity agents within the production medium, such as 10% glycerol and 10% propylene glycol and 80% water, known as GPW, has an important role in producing smaller MBs and helps to control MB dispersity and stability. 27,32−34 In a more detailed study, it was shown that varying solution viscosity using glycerol from 5 to 75 wt % using microfluidic electrohydrodynamic focusing production at voltages up to 12 kV reduced MB size from 120 to 25 μm. 35 In another study using a flow focusing PDMS chamber, glycerol and propylene glycol were used as the viscous agents to form MBs in solution and implicated as one of the factors to tailor MB size distribution. They reported that smaller MBs were produced only in the case of using 10 vol % viscous agents 36 but not less. A similar effect was also noticed in the case of forming nanoemulsions with high concentrations of glycerol, up to 50%, leading to 38% reduction in droplet size and narrower size distribution. 37,38 Glycerol is a water miscible molecule that has been used for many years to stabilize the activity of enzymes and the native structure of proteins as well as enhancement of protein selfassembly. 39,40 It has also been broadly used as a cryoprotectant to preserve the functionality of the biological molecules. 41−43 Glycerol molecules contain only single bonds, making them flexible and able to adapt to optimize their hydrogen bonding with surrounding water. 44 Phospholipids in water with higher concentrations of glycerol (10−30%) have been shown to lead to the formation of so-called "glycerosomes," which have been shown to be more potent than aqueous liposomes for transdermal drug delivery. 45,46 Glycerol is known to be a kosmotropic agent that increases water−water H-bonding in the solvation shell and diminishes the solubilization of hydrophobic groups, thus it has an ordering effect on lipids on proteins stabilizing the folded state 39,47,48 There is ample evidence from the literature that sugars and sugar alcohols, such as glycerol, in solution affect the physical properties and the stability of lipid layers. 49,50 The key reason glycerol and other sugars are used as cryo-preservatives is their ability to form glassy water−glycerol structures around biointerfaces during the process of freezing cells and liposomes, 49−52 as well as forming water−glycerol bonding to stabilize proteins. 39,53 These effects have also been used for prolonged stability of oil nanoemulsions 38,54,55 A number of studies have investigated the influence of glycerol on different biological systems, and there have been debated conclusions regarding the exact mechanism and molecular behavior of glycerol at the membrane surface. Glycerol affects the viscosity of the solution, which is expected to influence the surface tension of the lipid membrane and alter its optimum curvature and solubility. 48,56,57 One of the reported mechanisms refers to glycerol's ability to permeate the bilayer membrane and suppress intracellular ice formation during freeze and thaw cycles. 41−43 An alternative hypothesis suggests that the similarity between water and glycerol leads to bonding with the lipid headgroup, which could suggest that glycerol partitions equally between the surface and the bulk. 58 A thermodynamic study on DMPC vesicles demonstrated that glycerol has two competing effects at the molecular level, both partitioning to the membrane but also preferential exclusion from the interface. 59 The second, possibly stronger, effect is that as glycerol concentration is increased the membrane hydration decreases. 59 In spite of its popular use in lipid systems there are surprisingly fewer reports discussing the precise molecular effects of glycerol on lipid monolayers. In 2011, Pocivavsek et al. reported that glycerol not only has a fluidizing effect on model Lung surfactant monolayers but also leads to a stiffening at higher packing densities, 62 which they attributed to the incorporation of glycerol molecules into the solvation shell, with the hypothesis that in the presence of glycerol the hydration shell retains two to three water molecules per lipid headgroup, increasing the stiffness of the monolayers. 61,62 The aim of this study is to understand how glycerol influences the formation, stability, and mechanical properties of phospholipid coated MBs. This is the first detailed investigation of the effects of glycerol on lipid-shelled microbubbles, to attempt to deconvolute purely viscous effects from molecular interactions. Glycerol concentrations were varied between 1 and 30 vol % in the aqueous media during MB formation in a microfluidic device and subsequent storage in vials. Taken together, the results suggest complex and competing effects of glycerol at the molecular level, but the stability of the MBs always increases with concentration, even though the properties of the lipid layer, its hydration shell, and the properties of the bulk solution are in competition. Microbubble Production and Characterization. MBs were prepared as previously described. 63,64 Briefly, the lipid shell was formulated with DPPC and 5 mol % DSPE-PEG 2000 . The lipid mixture was dried under a steady stream of nitrogen gas on the vial walls. The dried film was then suspended in a solution containing 4 mg/mL NaCl and varying glycerol concentrations of 1, 10, 20, and 30 vol % (Sigma-Aldrich, UK) to a final lipid concentration of 1 mg/mL. This solution was vortexed for 1 min, then placed in an ultrasonic Langmuir Article bath for 1 h. The lipid solutions were allowed to cool down in the fridge at 4°C for 5 min prior to use in the microfluidic-MB maker.
MBs were produced microfluidically according to our previously described protocols. 64 C 3 F 8 gas was used for the MB core, and the gas pressure was controlled using a Kukuke microprecision regulator (RS supplies, Leeds, UK). The liquid phase, containing the lipid mixture, was flow rate controlled using an Aladdin AL 2000 syringe pump (World Precision Instruments, Stevenage, UK).
For sizing and counting the produced MBs, a 10 μL sample collected from the middle of the homogeneous MB solution was diluted 10-fold. From this diluted sample, 30 μL was introduced in a 50 mm depth chamber on a glass slide. MBs were allowed to rise for ∼2 min before collecting images. An inverted microscope (Nikon, Japan) was used to image the MBs on a 60× magnification. A CCD camera (DS-Fil 5Mega pixel, Nikon, Japan) was used to capture 40 images for each sample, from which the concentration and size distribution were obtained using ImageJ freeware (http.//rsbweb.nih. gov/ij/) and analyzed using Origin Pro (Version 8.5 or later) to create the size distribution histograms.
Microbubble Stability. MB lifetime was measured as described previously. 14 Briefly, 500 μL of MB solution was introduced in 500 μL of cell medium [RPMI from Invitrogen, Life Technologies, UK with 10% (v/v) fetal calf serum (Sigma-Aldrich, UK)] and incubated at 37°C in a digital dry block heater (Model D1100, Labnet International, USA). The vial containing the sample was left open exposed to air, and 10 μL samples were collected every 15 min for sizing and counting. At each time point, the vial was closed and gently shaken, and a sample was collected. The MB samples were allowed 2 min to rise from the bulk to the top of the optical chamber, to ensure all bubbles in the sample were imaged and considered.
The shell resistance to gas permeation was monitored by following single MB dissolution as previously described. 65 In brief, MBs were injected in a 50 mm deep fluid cell with fluid inlet and outlet ports and a replaceable glass slide top. These inlets were to allow degassed bubble solution to be flowed through the fluid cell containing the MBs under a flow rate of 2 mL/min. Several rinsing passages were used to ensure complete rinsing and buffer exchange takes place in the chamber. The change in bubble radius was then imaged with light microscopy.
AFM Mechanical Testing. For AFM testing, a diluted MB sample interacted with a biotinylated surface, to allow attachment of MB at the surface. For this, a 20-nm-thick layer of gold with a 2.5 nm Cr adhesion layer was deposited on a glass coverslip by thermal evaporation (Edwards Auto 306). The gold coverslips were incubated overnight in a 90:10 mol % 6-mercaptohexanol (Sigma-Aldrich, Poole, UK)/biotinylated-thiol (Asemblon, WA, USA) 66 solution to form a biotinylated self-assembled monolayer (SAM). The biotinylated SAM surface was incubated with 0.1 μM Neutravidin for ∼1 h prior to adding the MBs. The sample was then placed into contact with the MB solution in an inverted configuration for 1 h to enable the buoyancy force to bind the MBs onto the surface. Repeat measurements with the AFM loading at the pole of the MB were performed as described previously 14,63 using tip-less cantilevers (Nanosensors, UK). The cantilevers were calibrated using the thermal tuning method 67 giving a spring constant ∼ 0.3 N/m. An MFP-3D AFM (Asylum Research, Santa Barbara, CA) was utilized to select a single MB within a small diameter range (3−4 μm) for mechanical testing using the integrated optics of the AFM. The cantilever was then positioned above the pole of the MB, and force curves were acquired in continuous mode. For all MB preparations, the loading velocity was fixed to 3 μm/s, and force curves with a range of maximum applied loads were acquired: 5, 10, 30, and 50 nN. For each MB under investigation, 50 force−distance curves were collected with the cantilever in continuous mode to ensure reproducibility. Force− distance curves were then converted to force-deformation curves with the AFM software accounting for the cantilever deflection and the calibrated deflection sensitivity.
Hysteresis between loading and unloading in the force−indentation curves collected has been observed for all samples and is related to the dissipation of energy during sample deformation. The viscoelastic behavior of the MBs is quantified by calculating the plasticity index (η) from the hysteresis (ratio between the areas under the unloading (A u ) and loading curves (A L )), η = 1 − (A u /A L ), which gives the relative plastic/elastic behavior of the material under force. For purely elastic samples, where η = 0, the loading and unloading curves overlap, and the two areas are equal. For fully plastic properties, η = 1, where the area under the unloading curve is zero, and the deformation of the MB remains constant when the load is removed. In our case, η typically lies between 0 and 1, representing a mixed viscoelastic behavior. 63,68 Langmuir Isotherms. A Langmuir trough (KSV Nima) was used as previously described 22 to measure the changes in monolayer compressability and elasticity. The trough was equipped with two movable PTFE barriers to compress the monolayer symmetrically. A Wilhelmy plate tensiometer (paper method) was used to measure the surface pressure of the monolayer. This experiment was performed for the same lipid monolayer composition used to produce MBs. A total of 20 μL of a 1 mg/mL solution of DPPC + 5% DSPE-PEG 2000 , in chloroform, was spread on the surface of the subphase. The subphase was mixed with MiliQ water at the same concentrations of glycerol as used for the MB productions.

■ RESULTS AND DISCUSSION
Viscosity agents are known to increase the lifetime of MBs and influence their size distributions. For the range of glycerol concentrations used here, the density and viscosity of the glycerol water mixtures have been calculated according to the formula and parameters published in 2008 by Cheng 69 (Table  1). The results show that increasing the glycerol concentration does not affect the solution density significantly but has a large effect on the solution viscosity.
Microbubble Production and Size Distribution. MBs were prepared with four different glycerol concentrations (1, 10, 20, and 30 vol %) in the surrounding medium. Figure 1a shows an example of a typical MB size distribution, for bubbles produced in the microfluidic "spray" regime. In this case, data are shown for the 1% solution, while histograms for the other solutions are presented in Figure S1 and show a slight reduction in the modal size and the fwhm (Figure 1b). The average MB concentration produced in all cases was in the range of (9 ± 2) × 10 8 MB/mL. MB stability for the different glycerol concentrations was determined by measuring the MB concentration at different time points during incubation under physiological conditions, in cell medium, at 37°C, in an open environment to allow gas exchange ( Figure 2). There was no significant effect on MB population stability for glycerol concentrations less than 10%, but considerable improvement in the lifetime was observed at 20% glycerol and higher. At 30% glycerol, the whole population essentially remained stable over the time course of the experiment. This compares to a loss of 90% in the Langmuir Article population concentration with the lowest glycerol concentration of 1% within ∼100 min. This effect is due to the glycerol directly influencing the organization in the lipid shell surrounding the MB and increasing resistance to permeation and thereby decreasing the rate of gas escape from the MB core to the surrounding medium. To address this, we also studied single bubble dissolution. MB dissolution was investigated by following the MB size as a function of time, for individual MBs, across the glycerol concentration range (Figure 3a). For these studies, slightly larger MBs were considered to allow observation for prolonged periods. In all cases, the MBs initially decreased nearly linearly with time. Example images from the 10% sample are presented in Figure 3b and show that MBs maintain their uniform circularity while the gas core escapes, indicating that lipids from the shell are shed during dissolution in a quasicontinuous manner. 15 By contrast, the 20 and 30% samples show discrete steps in the single MB dissolution behavior and the MBs adopt noncircular and faceted shapes (Figure 3c). We have previously observed this effect for other stiff shelled MBs (e.g., actin coated MBs and for diacetylene polymerized MB shells) and have attributed this step like behavior to the stiffening of the MB coating. 65 The effective MB shell resistance was determined by numerically fitting the experimental radius-time plots to the Epstein−Plesset equation by varying the shell resistance R shell for each case and assuming the following parameters: 24 the saturation of the surrounding medium (f) was 5/6, and values of 5.2 × 10 −4 and 7.45 × 10 −4 m 2 s −1 were used for the Ostwald coefficient (H) and diffusivity of C 3 F 8 in water (D w ) as previously determined. 15,24 We also assumed that the surface tension (σ) was constant throughout at a value appropriate for DPPC of 62 mN/m as determined by Borden et al. from measurements of the collapse pressure of a Langmuir monolayer. 15 The average values calculated for the shell resistance, R s , were plotted for the different glycerol concentrations in Figure  4 and increased monotonically and in a superlinear manner with increasing glycerol concentration. The calculated R s values ranged from (90 ± 1) × 10 2 s/cm for the 1% glycerol sample to (340 ± 2) × 10 2 s/cm for the 30% glycerol sample. These values are 2 orders of magnitude higher than the values obtained by Heath et al., 65 using an MB shell of DOPC/

Langmuir
Article DOTAP, and Borden et al., 15 using a range of PC lipids and PEG40 stearate in the shell. This increase could be related to the different shell composition used here, which contains 5% PEG 2000 -lipid, giving a thicker shell. In another study, Park et al. 70 tested different compositions of a diacetylene-PEG lipid MBs shell with and without polymerization in a 5% glycerol medium . The R s values predicted for the unpolymerized MBs were 1 × 10 4 s/cm, which is the same order of magnitude to the values found here.
Microbubble Mechanical Properties. The mechanical properties of the MBs at each glycerol concentration were measured by AFM using a tip-less cantilever. MB stiffness was determined from the tangent of force−indentation data at different maximum applied loads, to understand the MB compression behavior with increasing pressure. Figure 5a presents the characteristic compressibility behavior of MBs where MB stiffness increases with increasing the maximum applied load, for the different glycerol concentrations. All data sets show linearly increasing stiffness with applied load, and only for the 30% sample is the compressibility behavior (linear slope) different than all the other cases. Interestingly, for any given force, the stiffness was found to increase from 1% to 20% followed by a dramatic reduction in stiffness for the 30% samples (Figure 5b).
At low applied force (10 nN), AFM is assumed to probe predominantly the MB shell properties, which exhibit an increase up to 20% glycerol concentration. The pattern of behavior is the same across the range of applied forces but the variations of effective stiffness are larger at higher force. At these higher MB deformations, the effective stiffness of the whole bubble including the gas core is interrogated, rather than just the shell at low force (10 pN).
Extrapolating the stiffness−force data (Figure 5a) to zero applied force yields the effective stiffness of MBs, at zero deformation (K 0 ). Figure 5c presents the average K 0 at each glycerol concentration. The value of K 0 increased rapidly from 11 ± 0.6 mN/m at 1% glycerol to 25 ± 2 mN/m at 10% glycerol and then essentially remained constant up to 30% glycerol. The observed jump in effective stiffness between 1 and 10% glycerol implies an interaction of the additional glycerol molecules with the lipid monolayer, stiffening the shell. These AFM shell stiffening data imply shell stiffening that also confirms the results from the dissolution experiment, Figure 4.
The plasticity index (PI) values for all samples was calculated and provides an indication of the viscoelastic behavior of the MBs (Figure 5d). At all glycerol concentrations, MBs exhibit viscoelastic behavior, with the highest plasticity occurring between 10 and 20% glycerol. These trends in MB stiffness and PI support the idea that a glassy water− glycerol hydration shell is forming between 10 and 20% glycerol, which is lost at the higher 30% concentration due to the formation of glycerosomes in the bulk solution (see discussion below).
Monolayer Isotherms. Langmuir isotherms of the DPPC + 5% DSPE-PEG2000 lipid monolayers were collected on subphases of glycerol solutions, at concentrations ranging from 0 to 30 vol %. Figure 6 presents the isotherms for increasing glycerol concentration at room temperature (20°C). Region 1 indicates the initial area per lipid molecule, A o , at which the surface pressure rises above zero. The data ( Figure S2) show that A o increases with the glycerol concentration up to 20% followed by a decrease for the highest glycerol concentration of 30%. This effect agrees with previous observations and interestingly mirrors the stiffness/plasticity trends seen in the AFM data. 44,59,71,72 The inset in Figure 6 shows the liquid condensed phase (region 2), from which the area per lipid molecule is extrapolated from the isotherms. The area/molecule (at high surface pressure) and the slope of the tangent are used to estimate the isothermal compressibility, C, and the compres-  sion modulus (elasticity), K, of the lipid modified interface using 22 C = −(1/A(dπ/dA)) and K = 1/C. The results plotted in Figure 7 show that increasing glycerol concentration from 1 to 10% results in a slight increase in the mean molecular area from 0.52 ± 0.002 to 0.58 ± 0.007 nm 2 , which drops slightly as the glycerol concentration is increased to 30%. For the 10% concentration, if the glycerol molecules insert into the MB shell, this has been estimated to be equivalent to one glycerol molecule for every five lipid molecules. These changes in area per molecule in the liquid condensed phase are accompanied by a decrease in monolayer compression elasticity (k) from 242 to 147 mN/m with increasing glycerol concentration from 0 to 20%. This is followed by an increase to 188 mN/m at 30% glycerol.
In addition to the other techniques, we have carried out initial ultrasound measurements of MB populations prepared with 1, 20, and 30% glycerol concentrations. The attenuation of the different MB populations (concentration of 2 × 10 6 MB/mL) was measured with a set of three matched pairs of unfocused transducers (V323, V384, V310, Olympus Industrial, Essex, UK), similar to the approach used before. 8 These three transducers allowed for testing over a frequency range of 1.5−8.5 MHz in 0.1 MHz steps, with a peak acoustic pressure of 50 kPa. The acoustic attenuation as a function of frequency was used to estimate the shell parameters as shown in studies by Gorce et al. 73 and Hoff. 74 Preliminary results in Table S3 show that the shell stiffness peaks at 20% glycerol with the 30% values close to the 1% values. This trend is in agreement with both the AFM and LB trough data but will require more detailed investigation.
Glycerol has been used for a wide range of biological studies and bioinspired systems, mainly as a cryoprotectant, but the exact mechanism of glycerol on lipid membrane hydration and physical properties is still debated, mainly because it has more than one effect in any given system. It has been reported as having effects of both fluidizing and stiffening lipid membranes, while also dehydrating the lipid headgroup as well as forming structured glassy water−glycerol layers at interfaces. To further add to the complexity of behavior, it has been reported that above 10% glycerol, glycerosomes are formed with lipids.
On the basis of our results and current literature, we propose a working hypothesis for the molecular effects of glycerol on lipid-shelled MBs. For concentrations up to 10%, the glycerol molecules bind to the lipid headgroup, increasing the area per molecule. At low surface pressures (region 1 in Figure 6), the glycerol is known to fluidize the membrane, 62,72,75 but the lipid shell of the MBs will be in the liquid condensed phase (region 3 in Figure 6) where the binding of glycerol to the headgroups stiffens the membrane, which is what we observe in AFM compression measurements up to 20% glycerol. Glycerol binding to the lipids is more or less saturated at 10% glycerol (Figure 7). At the same time, glycerol−water glassy layers are forming around the lipid monolayer at the gas−liquid interface, increasing the effective stiffness of the MB coat. This effect continues up to ∼20% glycerol, beyond which glycerosomes start to form in solution. There is then a competition between use of glycerol to form these glassy layers and their disruption due to sequestering of glycerol into glycerosomes. By 30% glycerol, the glassy layer is all but abolished and glycerosomes dominate in solution. We now discuss in more detail how this picture aligns with the current literature and correlates with our measurements.
At low concentrations, up to 10%, glycerol replaces water and binds to the phospholipid headgroup causing an increase in the area per molecule and increases the alkyl chain disorder. Furthermore, the ability of a glycerol molecule to hydrogen bond to more than one lipid molecule (its cross-linking ability) increases the stiffness of the layer. It also attracts and binds more water molecules forming the solvation (hydration) shell. Increasing the glycerol concentration in the range between 10 and 20% forms the more ordered glassy water−glycerol layer. Both these effects contribute toward increased shell resistance, which in turn reduces the gas exchange and increases MB stability (Figures 2 and 3). The observed increase in shell stiffness up to 20% glycerol, from Langmuir isotherms on monolayers ( Figure 7) and from AFM mechanical testing on MBs (Figure 5b), is also consistent with the aforementioned cross-links and the glassy hydration layer. From molecular dynamics simulations of bilayers, 76 it was suggested that at 15% mol concentration of glycerol, the lipid surface is enriched with glycerol molecules binding at a ratio of 2:1 glycerol per phospholipid headgroup, with hydrogen bonding between the phosphate moiety−glycerol or between two glycerol molecules, such that the lipid headgroup is dehydrated but yielding a reordering effect on the acyl chain 76 . This would make the shell stiffer and also increase resistance to permeation At a high glycerol concentration (>20%), two possible mechanisms could take place either alternatively or simultaneously. Considering that glycerol binding to the lipids is more or less saturated at 10% glycerol concentration. Glycerol− water glassy layers are formed at the lipid/aqueous interface, increasing the effective stiffness of the MB coat. This effect continues until ∼20% glycerol, beyond which glycerol as a cosolvent reduces the polarity of the mixture and alters the charge interaction between the lipid head groups, which facilitates the exposure of the hydrophobic tails to water 48 and promotes the aggregation of the lipids to form micelles or glycerosomes 38,57 that start to form in solution. This leads to a competition for the glycerol to form the glassy layers versus their sequestration into glycerosomes. By 30% glycerol, the glycerosomes dominate in solution. Alternatively, glycerol molecules might sequester water molecules at the lipid interface and dehydrate the PEG chains, which will cause the PEG chains to aggregate and phase separate. This effect could explain the shell "cracking" behavior that was observed in the

Langmuir
Article individual MBs dissolution experiments ( Figure 3) and the unusual compressibility behavior in the AFM experiments for the 30% glycerol sample (Figure 5a), and in the associated Langmuir isotherms (Figure 6). It should also be noted that dehydration of lipid headgroups causes the area per lipid molecule to get smaller and increases ordering in the alkyl chains. The observed decrease in monolayer elasticity ( Figure  7) could be attributed to the chain ordering effect of glycerol that favors the gel phase packing. 47,59 In this phase, the repulsive interaction forces between the hydrophilic head groups increases, causing an increase in the free energy and the surface tension of the membrane. 57,77,78 From the presented Langmuir isotherms, because of glycerol (surface tension 64 mN/m 79 ) incorporation in the lipid layer, the surface tension of the monolayer (using 10% glycerol) is increased from 25 mN/m 24,80 to 31 mN/m. At 20 and 30% glycerol, the fraction of glycerol incorporated in the shell is reduced again to be approximately 1 glycerol molecule for every 10 lipid molecules. Our estimates of glycerol to lipid molecular ratio are based on a simple calculation of glycerol molecular size and an assumption of intercalation in between the head groups. Molecular modeling and other studies indicate that one should expect a 1:1 ratio of glycerol to lipid at 10% glycerol and that glycerol binds to the lipid headgroups but does not insert between them.
The plasticity index derived from the AFM loading and unloading curves, follows a similar trend to that of the MB stiffness (Figure 5d). The increased plastic behavior at 10 and 20% appears related to the formation of the glassy glycerol hydration layer, as the MBs at 30% return to a similar viscoelasticity to that at 1%. Again, this supports the idea of lipid dehydration at a high concentration of glycerol that can cause changes in phase transition and overall shell behavior.
It might be expected that changes in shell structure at high glycerol concentration should lead to a deterioration in MB lifetime; however, this was not observed and indicates there is another factor to be considered, which is the increased viscosity (from 1−2.24 Pa s) with increasing glycerol in solution from 1 to 30% (Table 1). Viscosity affects the mass transport kinetics at the membrane interface 57 and leads generally to reduced gas exchange and increased bubble lifetime (Figures 2 to 4). It also alters the optimum monolayer curvature 57 and leads to a reduction in MB size and lower polydispersity of the microspray on chip production ( Figure  1). A similar reduction of polydispersity index was reported by Saberi et al.; 38,55 when producing oil nanoemulsions with >20% glycerol concentrations, this effect was attributed to the tendency of glycerol to alter the interfacial tension, curvature, and characteristics of lipids 48,56,57 Whether or not glycerosomes are forming in solution, we see that the increased concentration of glycerol leads to a tighter size distribution of MBs; as a consequence, there is higher viscosity of the solution in the nozzle. There is a concomitant increase in the lifetime of the MBs, associated with a more stable bubble shell (lipid monolayer). Simple modeling with the Epstein−Plesset equation indicates that the effective shell resistance (R shell ) goes up with increasing glycerol. In reality, it is likely that this is predominantly due to the diffusion coefficient of the gas going down in the higher viscosity solution since lifetime increases from 0 to 30%. The molecular effects of glycerol could also contribute to increased shell resistance to gas permeation, at least up to 20% where glassy layers are formed.

■ CONCLUSION
Introducing glycerol into the solution during MB production and storage can greatly increase lifetime and reduce the polydispersity of the MB population. The multiple molecular effects of glycerol on the lipid monolayer at the gas−liquid interface can tune the physical and mechanical properties of lipid-shelled microbubbles. Despite modulation of the lipid monolayer properties through molecular interactions with glycerol, our results indicate that 30% glycerol is the optimal concentration for both MB formation and storage since it greatly enhances the lifetime and minimizes the polydispersity. One should bear in mind that once MBs are injected into the bloodstream, the in vivo solution environment will be very different; therefore higher concentrations of glycerol might be preferable for pure shelf life stability over optimization of shell properties. We expect that these findings may be relevant to MBs with different biomaterial shells, such as proteins, due to the molecular interactions of glycerol with water and biomolecules, as well as its dominant viscous behavior.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.langmuir.8b04130. Figure S1: Optical size distributions of MBs formed in 1−30% glycerol. Figure S2: Area per lipid molecule calculated from Langmuir isotherms at high and low surface pressure for the lipid monolayers above an aqueous subphase of 1−30% glycerol. Table S1: MB shell stiffness and viscosity derived from ultrasound attenuation from MB populations for 1−30% glycerol (PDF)