Molecular and cellular level characterization of cytoskeletal mechanics using a quartz crystal microbalance

A quartz crystal microbalance (QCM) is an instrument that has the ability to measure nanogram‐level changes in mass on a quartz sensor and is traditionally used to probe surface interactions and assembly kinetics of synthetic systems. The addition of dissipation monitoring (QCM‐D) facilitates the study of viscoelastic systems, such as those relevant to molecular and cellular mechanics. Due to real‐time recording of frequency and dissipation changes and single protein‐level precision, the QCM‐D is effective in interrogating the viscoelastic properties of cell surfaces and in vitro cellular components. However, few studies focus on the application of this instrument to cytoskeletal systems, whose dynamic parts create interesting emergent mechanics as ensembles that drive essential tasks, such as division and motility. Here, we review the ability of the QCM‐D to characterize key kinetic and mechanical features of the cytoskeleton through in vitro reconstitution and cellular assays and outline how QCM‐D studies can yield insightful mechanical data alone and in tandem with other biophysical characterization techniques.

A variety of instrumentation has been used to observe and measure cytoskeletal mechanics workings through both molecular reconstitution assays and in cultured cells. Molecular assays, such as in vitro motility assays, allow researchers to replicate and record interactions between the cytoskeleton's individual components (Kron & Spudich, 1986;Umemoto & Sellers, 1990). Cytoskeletal filaments and proteins can be tagged with fluorescent proteins to visualize movement or density; however, labelling may create non-native complexes with unintended and physiologically irrelevant properties (Chen, Li, Penn, & Xi, 2011). Filaments and motor proteins can also be adsorbed to surfaces or microscopic beads to allow manipulation by devices like optical traps (OT). In an OT, a laser traps a micron-sized bead attached to a biomolecule and records changes in force and motion (Arbore, Perego, Sergides, & Capitanio, 2019;Mofrad, 2009). A magnetic micromanipulator can be used in a similar way, albeit with magnetic beads (Huang et al., 2005;Mofrad, 2009). Cytoskeleton mechanics have also been investigated as a collective unit in cells (Haga et al., 2000;Lamaze, Fujimoto, Yin, & Schmid, 1997;Li, Zheng, & Drubin, 1995). In these assays, cytomorphic agents, such as taxol or phalloidin that directly interact with cytoskeletal filaments, can be used to induce structural alterations (Fatisson, Azari, & Tufenkji, 2011). Microscopy and flow cytometry can be used to monitor such changes (Nebe, Bohn, Pommerenke, & Rychly, 1997). Similarly, micropipette aspiration can be applied to measure viscoelasticity and mechanical tension of the cellular membrane and affixed cytoskeleton (Discher, Boal, & Boey, 1998;Mofrad, 2009). These experimental approaches often require the cell to be fixed or lysed, which does not allow the cytoskeleton to behave as it would natively (Fatisson et al., 2011). As with molecular assays, cellular assays can be labeled via fluorescence, which carries with it the same issues. While time lapse video microscopy can be used to observe the changes over time without labelling or compromising the cell (Schneider, Sampathkumar, & Persson, 2019;Yang et al., 2012), the technique is labor intensive, time consuming, and its results are qualitative (Li et al., 1995). As such, the relatively noninvasive technique of atomic force microscopy (AFM) is especially popular (Radmacher, 1997;Yang et al., 2012). In AFM, a probe scans the top of the cell with a force-sensitive cantilever to create a force-displacement curve (Haase & Pelling, 2015).
Although it is high precision, AFM cannot study changes in the basal portion of cell, critical for the understanding processes like cell adhesion (Yang et al., 2012). Micropillar assays are also able to probe interactions between cells and substrates, as well as their functional effects due to differing substrate stiffness, yet requires the ability to fabricate specialized microdevices (Bavi, Richardson, Heu, Martinac, & Poole, 2019;Edwards & Schwarz, 2011;Tan, Choong, & Dass, 2010).
A powerful tool that has the potential to fill gaps in and complement current understanding of cytoskeletal mechanics is the quartz crystal microbalance (QCM), a piezoelectric sensor (Dixon, 2008).
The QCM is versatile, label free, noninvasive, and can be used in tandem with biophysical interrogation methods like AFM or OT (Braidotti et al., 2022;Yang et al., 2012). By measuring changes in the resonance frequency of a quartz crystal, molecular-level changes in mass can be detected as molecules adsorb to the surface (Dixon, 2008). A variant, the quartz crystal microbalance with dissipation monitoring (QCM-D), detects changes in the adhered system's viscoelastic properties (Dixon, 2008;Fatisson et al., 2011;Johannsmann, 2008). It has become increasingly popular for cell biology studies, with applications ranging from cell-cell adhesion, binding kinetics, signaling, and biomarker analysis, among others (Xi & Chen, 2013;Yang et al., 2012). However, the QCM remains underutilized in cytoskeletal studies (Yang et al., 2012). This technique can be used to study both the cytoskeleton's constituent parts and its macroscopic structure. For instance, the QCM can record the interactions of in vitro motility assays with artificial surfaces, like gold or silicon dioxide, and facilitate more complex and realistic design of assays (Lord et al., 2006;Ozeki et al., 2009). The QCM can also be used to study the cytoskeleton's interactions with other cell components, such as lipid membranes or specific proteins (Johjima et al., 2015;Tae, Park, Kim, Yorulmaz Avsar, & Cho, 2022). This instrument has the ability to noninvasively record and quantify the real-time reorganization of the cytoskeleton in situ within cells (Fatisson et al., 2011;Marx, Zhou, Montrone, McIntosh, & Braunhut, 2007). Further, potential pharmaceutical applications for QCM investigations include understanding how a cell responds mechanically to a certain agent (Bianco et al., 2018;Marx et al., 2007). Due to the versatility of information that can be gleaned from QCM studies and the potential for innovative collaborations with existing cytoskeletal experimental approaches, we will review how the QCM has been used to study cytoskeletal mechanics, specifically highlighting key assay preparations to make QCM usage more accessible to the cytoskeleton biophysics field.

| QUARTZ CRYSTAL MICROBALANCE
The quartz crystal microbalance ( Figure 1A) uses thin quartz disks, commonly coated in gold or silicon dioxide, with electrodes to record changes in frequency as samples are measured (O'Sullivan & Guilbault, 1999). Quartz crystals are piezoelectric in nature, and the excitation of the crystal via oscillating currents causes the crystal to deform (Dixon, Xi) (Dixon, 2008;Xi & Chen, 2013). Since resonance frequency is an inherent property of the crystal, it is used as the baseline for calculating the change in frequency that occurs during experiments (Johannsmann, 2008). The theory behind the piezoelectric nature of quartz crystals is reviewed extensively elsewhere (Guilbault, Jordan, & Scheide, 1988;O'Sullivan & Guilbault, 1999). In 1959, Günter Sauerbrey determined that the relationship between changes in frequency and mass on the crystal sensor were proportional (Dixon, Sauerbrey, Xi) (Dixon, 2008;Sauerbrey, 1959;Xi & Chen, 2013). This relationship is given by: where Δm is the change in mass, Δf is the frequency change with respect to the resonance frequency ( Figure 1B, green), n is the harmonic number, and C is mass sensitivity constant for the sensor which is approximately À17.7 Hz•ng/cm 2 (Dixon, 2008). Due to the high sensitivity of the sensors, mass changes can be quantified on the nanogram scale (Dixon, 2008).
As originally designed, the quartz crystal microbalance was limited in its capabilities as it could only characterize thin films without viscous qualities; thus prior to 1980, the QCM was mainly used to study thin films in a gaseous state (Dixon, 2008;Xi & Chen, 2013). In the mid-1980s, the QCM rose to popularity as it was shown to be effective in characterizing liquids; however, the viscosity of samples leads to overdamping the crystal oscillation (Dixon, 2008;Johannsmann, 2008). While the Sauerbrey equation is helpful for thin layers, the viscous qualities of liquid samples allow for dissipation of energy, therefore nullifying the Sauerbrey equation in such cases (Johannsmann, 2008;Xi & Chen, 2013). With the advancement of QCM technology, a quartz crystal microbalance with dissipation monitoring (QCM-D) was created. This allowed for analysis in cases where viscoelastic films were present. Dissipation ( Figure 1B, purple), which is dimensionless, is a comparison of the energy dissipated during a single oscillation, E Dissipated , and E Stored , which is the energy stored within the system during oscillation (Dixon, 2008;Xi & Chen, 2013). It can similarly be defined as a relationship between bandwidth, Γ, and resonance frequency (Johannsmann, 2008). The change in dissipation is given by the following equations: To assess the changes in dissipation, commonly referred to as ΔD, there are two methods used: analyzing impedance or monitoring crystal oscillation after excitation near the resonance frequency (Dixon, 2008;Johannsmann, 2008). Impedance analysis is based upon the electrical conductance curve; utilizing both the bandwidth and resonance frequency, a decay in voltage displacement can be quantified and converted into dissipation (Johannsmann, 2008). The other method, which is extensively reviewed in Dixon's, 2008 paper, quantifies dissipation by short circuiting the oscillating current (Dixon, 2008). When the current source is deactivated due to the short circuit, the crystal oscillation decays to a point of rest; the energy dissipated and stored can then be analyzed to obtain ΔD.
The 1996 development of the Q-Sense QCM-D, which simultaneously collects frequency and dissipation data, allowed researchers to expand the breadth of assays on the QCM-D to include biological samples (Dixon, 2008). With both frequency and dissipation data, parameters such as mass, film thickness, density, and viscosity of samples, as well as elastic modulus, can be extracted from the raw data for analysis. QCM-D systems can be purchased commercially (e.g., the QSense Analyzer from Biolin Scientific, openQCM, etc.) and offer customization options to fit specific research needs. The instrument price can exceed $100,000 USD, especially with added customizations. The basic instrument includes a flow chamber that can fit multiple sensors with microfluidic technology ( Figure 1A), a pump for the microfluidic setup, power supply, and computer for data acquisition and analysis.
The real-time monitoring aspect of the QCM-D is of particular interest to researchers as they can manipulate the assay conditions through the microfluidic setup and monitor the resultant effect on the system (Xi & Chen, 2013  The system has a small benchtop footprint, consisting of the QSense Analyzer with four chambers for sample analysis in series or parallel, the QSense Electronics Unit, a peristaltic pump, and a desktop computer (not shown). The tubing can be changed to be compatible with the sample if necessary. The amount of sample needed (right) will depend on the flow rate and number of sensors being utilized per experiment. (B) Upon applying voltage to a thin quartz disk with electrodes, the sensor is excited to resonance. Nanoscale mass changes can be monitored through changes in resonance frequency (green). Dissipation increases with sample "softness" or viscoelasticity and can be monitored in real-time (purple).

| QCM STUDIES OF RECONSTITUTED CYTOSKELETAL ASSEMBLIES
In vitro assays have been critical in determining how dynamic cytoskeletal components interact and function in isolation, small ensembles, and complex networks (Hanson, Viidyanathan, & Nicolau, 2005).
The real-time monitoring of the QCM and QCM-D is ideal for such assays as they are able to measure changes in binding state and viscoelastic properties as they occur. As early as 1994, researchers analyzed reconstituted in vitro assays of actomyosin depolymerization using piezoelectric quartz crystals (Kurosawa et al., 1994). While Kurosawa et al. did not use a QCM, the same principle of analyzing changes in the resonance frequency of a quartz crystal was present in their study (Kurosawa et al., 1994). As this was prior to the Using the QCM as a cell biosensor requires that cells be seeded directly onto the quartz crystal sensor. It is important to note that the sensors can be purchased with a variety of coatings with ranging levels of robustness to experimental conditions (such as gold and silicon dioxide to protein coatings like biotin). As such, the coatings must be taken into consideration before the following cleaning and preparation procedures take place. For gold and silicon dioxide coated sen- obtaining baseline values without the cells, the sensor is exposed to a cell suspension that is washed through in a serum (Tymchenko et al., 2012). The QCM pump can be stalled to allow cells to settle and attach to the sensor. A second set of baseline values is then collected with the adhered cells. These two baseline value sets may allow for better data analysis, and this technique may be adapted to suit a specific experiment's needs (Tymchenko et al., 2012). Next, cells are exposed to the agent or stimuli of interest through successive washes.
After collecting data, the number of cells attached to the sensor can be quantified through trypsinization and electronic counting (Marx et al., 2001). Due to variable results of cell assays, this experiment is typically repeated several times (Yang et al., 2012). To broaden the generality of the results, the experiment can also be performed with multiple cell types (Tymchenko et al., 2012). Control experiments without cells may be conducted to isolate the effects of agents on frequency and dissipation sans cells (Fatisson et al., 2011).
As the actin cytoskeleton can drive cell mechanical and morphological properties and has a prominent role in cell attachment, much of the cytoskeleton-based QCM literature has focused on agents that specifically target actin filaments. The effects of actin-perturbing agents are commonly captured through change in dissipation, ΔD. A well-studied agent is cytochalasin D (CytoD), which depolymerizes Factin and dramatically alters cell morphology (Bianco et al., 2018). This depolymerization affects the cell-substrate interface, which is detected by the QCM as a dissipation decrease (Bianco et al., 2018;Tymchenko et al., 2012). To interpret this result, Tymchenko et al.
claim that CytoD exposure leads to cell body retraction that causes the QCM-D response (Tymchenko et al., 2012). However, Braidotti et al. observed cell body retraction coupled with a decrease in cell body area and increase height using complementary imaging (Braidotti et al., 2022). The authors correlate the decrease in dissipation, or loss of "softness", with the change in cell-substrate adhesion with a transition to a more round-like shape (Braidotti et al., 2022). Both of these studies are exemplify the benefit of combining QCM-D measurements with microscopy to confirm interpretation of the QCM-D response.
Another actin modifying agent used in QCM-D study is bacterial lipopolysaccharide (LPS), an endotoxin involved in Gram-negative septic shock, that alters the actin cytoskeleton and induces cell rounding.
Triton-X 100, a nonionic surfactant, acts as a detergent to disrupt membranes. Fatisson et al. report that the QCM-D will detect the effect of cell treatment with both Triton-X 100 and LPS (Fatisson et al., 2011). Other actin-targeting cytotoxins that have been used in QCM studies include jasplakinolide and phalloidin (Cooper, 1987;Galli Marxer, Collaud Coen, Greber, Greber, & Schlapbach, 2003). Cell signaling pertaining to the cytoskeleton has also been manipulated in QCM studies. Bianco et al. used Y27632, an inhibitor of a Rhoassocated kinase that leads to actin filament depolymerization and contractility-loss, to induce a rapid increase in ΔD (Bianco et al., 2018;Maekawa et al., 1999). Chen et al. and Yang et al. both  Microtubule-targeting agents have also been studied using the QCM cell biosensor (Braidotti et al., 2022). Such agents include nocodazole and taxol (Galli Marxer et al., 2003;Marx et al., 2007). Agent exposure regimens can be altered to suit an experiment's needs. For example, Marx et al. utilized varied concentrations of nocodazole to study dose-dependent cytoskeleton alterations (Marx et al., 2001, Marx et al., 2007. Furthermore, the effects of some agents are reversible, like CytoD, and its effects can diminish upon washout (Tymchenko et al., 2012). To probe the effects of a novel agent, parallel experiments can be run in which Δf and ΔD values of the unknown agent can be recorded and compared to those of a well-understood control agent, like CytoD (Bianco et al., 2018). This makes the QCM a useful platform for cell-based drug screening applications (Fatisson et al., 2011).
Upon cell adhesion to the sensor, the QCM records a reversible and positive shift in frequency and dissipation ( Figure 3B) (Marx et al., 2001;Tymchenko et al., 2012). While early studies stressed that the total number of cells adhered to the sensor was proportional to frequency, Marx concluded that the Δf values are dependent on the total number of trypsinizable cells adhered to the QCM surface (Lord et al., 2006;Marx et al., 2001). Calibration curves of Δf versus the number of adhered cells yield a graph similar to that of a nonlinear binding isotherm (Marx et al., 2001). As such, proper cell counting can assist data analysis (Galli Marxer et al., 2003;Marx et al., 2001). However, others have suggested that an uneven distribution of cells may be responsible for the disproportionality. Furthermore, Lord et al.
noted that a scarcity of cells in the center of the sensor may negatively affect results (Lord et al., 2006). A further consideration is that QCM-D results can be affected by the thickness of the adhered cell layer (Tymchenko et al., 2012). The effects of extraneous substances used for cell attachment should also be assessed. For instance, when using extracellular matrix proteins, a Voight-based model can be used to calculate their mass and approximate a viscoelastic model (Lord et al., 2006). Results may also vary between different cell types due to different mechanical qualities. Accordingly, assay composition and conditions should be standardized and controlled for to enhance data comparison between different independent variables.
Despite the multiple contributing variables, QCM data shifts can be attributed to specific cytoskeletal restructuring and other changes.
As frequency alone is insufficient to characterize cytoskeletal changes, dissipation shifts are more heavily relied upon (Fatisson et al., 2011).
Frequency changes indicate mass changes, and ΔD is more relevant, revealing cytoskeletal changes within cells indirectly through viscoelastic changes (Bianco et al., 2018). However, current Voight-based models used to correlate QCM dissipation changes to viscoelasticity have constraints like the need for surface uniformity in film coverage and density, which makes quantitatively interpreting cytoskeletal data challenging; however, dissipation data to date has largely been interpreted qualitatively or in tandem with other characterization techniques, such as AFM (Biolin Scientific, 2017;Höök et al., 2001;Johannsmann, 1999). For instance, an actin-depolymerizing agent like CytoD will induce a characteristic dissipation decrease, signifying that the cell has become more rigid (Bianco et al., 2018;Chen et al., 2011).
Agents that act through more complex pathways may exhibit complicated data patterns (Bianco et al., 2018). The kinase inhibitor Y27631, for example, is reported to exhibit an initial dissipation decrease followed by an increase (Bianco et al., 2018). Additionally, the intensity of these values can reveal relative severity of different agents upon comparison (Bianco et al., 2018). Linear regression analysis of initial ΔD and Δf slopes as a function of agent concentration can also be conducted to confirm cytoskeletal alterations and to decipher the effects of competing cytomorphic agents (Fatisson et al., 2011;Marx et al., 2001).
While data from the cellular QCM biosensor is frequently used qualitatively to characterize cytoskeletal restructuring, whether the data can be quantified remains disputed. One major contributor is the absence of suitable models of cell layers on QCM (Braidotti et al., 2022). Quantitative modeling of these adhered cells-which do not act as an elastic mass-is a difficult task as they are complex systems that are heterogenous in both dimensions (Marx et al., 2001;Tymchenko et al., 2012). At the time of publication, there are no complex models that tackle the heterogeneity of these cell layers (Tymchenko et al., 2012). Although they did not attempt to make it quantitative, Tymchenko et al. proposed a two-layer model for interpreting QCM-D cellular data (Tymchenko et al., 2012). While the investigators cautioned against quantifying QCM data, they measured the effect of CytoD treatment by comparing its maximal dissipation shift to the dissipation shift observed in the cells prior to CytoD exposure (Tymchenko et al., 2012). Most notably, investigators have relied on the semi-quantitative ΔD/Δf plots (or simply Df plots) developed by Fredriksson et al. (Fredriksson, Kihlman, Rodahl, & Kasemo, 1998;Xi & Chen, 2013). These plots provide a unique profile and can reveal internal changes that a scanning electron microscope or another other form of microscopy cannot (Fatisson et al., 2011;Xi & Chen, 2013).
Df plots can deduce response types, including whether mass is being removed, and to differentiate between cytotoxic and noncytotoxic responses (Fatisson et al., 2011). Df slopes from different cytomorphic agents and controls can be compared to one another to isolate actions. Further, the biochemical and mechanical pathway by which an agent acts may result in major differences in the observed Df plot (Fatisson et al., 2011).
Due to some of the above nuances regarding QCM data interpre-  (Braidotti et al., 2022).

| CONCLUSIONS AND OUTLOOK
Examining the cytoskeleton and its components using a quartz crystal microbalance with dissipation monitoring can aid in elucidating the viscoelastic and kinetic properties of such assays. Changes in resonance frequency can be interpreted as the addition or reduction of mass on the sensor surface, as well as how binding conformation changes in real time (Ozeki et al., 2009). Dissipation data further allows researchers to analyze how the structural components of the cytoskeleton mechanically couple within an architecture to produce a desired output (Bianco et al., 2018;Chen et al., 2011;Fatisson et al., 2011). Here, we review approaches to constructing both protein and cellular assays using a QCM-D where the emergent properties of cytoskeletal ensembles that result from dynamic, viscoelastic filament and protein subunits can be obtained in real-time as the assays are assembled on the sensor surface.
Previous studies involving in vitro protein assays reveal limited use of the QCM-D capabilities as it is often used as a simple mass balance Suda et al., 2004). This is partly due to the difficulty in interpreting and lack of appropriate models for protein assays, especially if they do not cover the entire sensor surface, making conclusions thus far more qualitative than quantitative. However, motility and kinetic assays allow for real-time monitoring of cytoskeletal networks; using the QCM-D to analyze changes in structural mechanics as conformational changes occur could foster deeper understanding of cytoskeletal function and communication. Adding cross linkers to multi-filament assays has been shown to alter the viscoelasticity and architecture of the cytoskeleton, which would be suitable for study by QCM-D (G. Lee et al., 2021;Ricketts et al., 2019).
The QCM-D can also be used in pharmaceutical studies as a cell biosensor to test the impact of drugs or other reagents on structural rearrangements and rigidness (Fatisson et al., 2011;Marx et al., 2001).
Another important application that could be intuitively adapted for As the QCM is still an emerging technique in cytoskeletal studies, there are many opportunities for refinement in assay preparation and data analysis, as well as applications to novel research. For instance, the QCM is still hampered by the lack of an encompassing mathematical model that can link dissipation changes to quantitative viscoelastic properties of the sample. Variations on the Voight model have been used for this purpose, but one of the constraints in its use is that the sensor needs to be evenly coated with the sample material with uniform thickness and density (Biolin Scientific, 2017;Höök et al., 2001;Johannsmann, 1999). This could perhaps be useful for high-concentration motor gliding assays, active matter studies of cytoskeletal meshes, or cells that reach confluence on the sensor surface. However, in the case of reconstituted or cellular assays that attempt to probe certain protein or motor interactions within a complex environment, not only will there not be an evenly coated surface, but the events being monitored are also not uniform across the sensor. Thus, great care will be required when designing cytoskeletal QCM assays to ensure proper controls and defined intervals are used for addition of mechanosensitive agents to qualitatively interpret frequency and dissipation data until a more accurate model is developed. In addition, determining what level of sensor coverage and uniformity is close enough to still use the existing Voight model would be useful.
Further, coupling QCM-D measurements with more established and complementary techniques such as fluorescence microscopy, AFM imaging, or optical trapping force measurements will aid in the interpretation of frequency and dissipation shifts.
As biophysicists work to explore the grand challenges of the field, such as building reductionist yet functional sub-cellular machinery like a working sarcomere or mitotic spindle, all the way to building a syn-