Applications of scanning probe microscopy in neuroscience research

Scanning probe microscopy techniques allow for label-free high-resolution imaging of cells, tissues, and biomolecules in physiologically relevant conditions. These techniques include atomic force microscopy (AFM), atomic force spectroscopy, and Kelvin probe force microscopy, which enable high resolution imaging, nanomanipulation and measurement of the mechanoelastic properties of neuronal cells, as well as scanning ion conductance microscopy, which combines electrophysiology and imaging in living cells. The combination of scanning probe techniques with optical spectroscopy, such as with AFM-IR and tip-enhanced Raman spectroscopy, allows for the measurement of topographical maps along with chemical identity, enabled by spectroscopy. In this work, we review applications of these techniques to neuroscience research, where they have been used to study the morphology and mechanoelastic properties of neuronal cells and brain tissues, and to study changes in these as a result of chemical or physical stimuli. Cellular membrane models are widely used to investigate the interaction of the neuronal cell membrane with proteins associated with various neurological disorders, where scanning probe microscopy and associated techniques provide significant improvement in the understanding of these processes on a cellular and molecular level.


Introduction
Integral to the functioning of the human brain are neuronal cells, responsible for the processing and transmission of messages in the brain and nervous system via electrical impulses and chemical signals.The reception and interpretation of these messages in the brain, spatially and temporally, allow for functional abilities, such as the perception and integration of sensory information, and higher-order processes, such as cognition and comprehension, among many others [1,2].
Neurodegenerative diseases, such as Alzheimer's disease, involve neuronal damage and death.Damaged neurons, unable to transmit and process these messages as in a healthy brain, are unable to communicate efficiently, resulting in compromised function of the central nervous system.There is currently no cure for Alzheimer's disease, and the molecular mechanisms involved in the progression of Alzheimer's disease are not well understood [3,4].The determination of the molecular mechanism of neurodegeneration and the development of novel therapeutic strategies relies on the early detection of structural and nanomechanical changes in the neurons preceding cell death.One of the most researched hallmarks of disease progression is the aggregation and accumulation of proteins (amyloid-β in Alzheimer's disease; α-synuclein in Parkinson's disease) in the brain [5,6].
Neuroscience involves the study of the structure and function of the brain, and as a result spans the study of patients and neuroimaging down to in vitro molecular studies.In this regard, high resolution microscopy and nanotechnology approaches provide a useful way to study the molecular mechanisms involved in neuronal function and to identify nanoscale structural and mechanical changes in individual neurons and their environment.Fluorescence microscopy is commonly used in cellular studies, but requires a fluorescent tag and is limited in resolution outside of super-resolution techniques [7].Atomic force microscopy (AFM) is a technique that is not limited by either of these factors, while maintaining a physiologically relevant environment.AFM, a technique in which a nanoscale probe is raster-scanned along a surface to form a topographical image, allows for the structural identification of cell topography and associated components with nanometer resolution [8][9][10].An associated technique, atomic force spectroscopy, involves measuring the force exerted on this probe as a function of the separation between it and the cell or sample; this allows for the nanoscale determination of mechanical characteristics such as elastic modulus and adhesion forces [11][12][13][14][15].These techniques combined could be applied to identify early changes in cells and their structures during neurodegeneration [16,17].
As incredibly complex as neurons and neuronal systems are, it is necessary to simplify them and their components experimentally.Most relevant is the in vitro modeling of the neuronal membrane by model lipid membranes, supported bilayers of synthetic lipid mixtures [18][19][20].This allows for the study of the interaction between this bilayer and disease-associated proteins, as well as the study of aggregation inhibitors [21][22][23][24].This review focuses on the applications of AFM, and related techniques, to the study of brain tissues, neurons, and lipid membranes modeling neurons and their interactions with proteins associated with neurodegenerative disease.

AFM & related techniques
Scanning probe microscopy began with the invention of the scanning tunneling microscope, using a biased tip to measure the tunneling current on a metal surface, in 1981 by Binnig, Rohrer and Gerber [25].Their invention of the atomic force microscope followed soon after, with the tip-sample interaction involving instead atomic forces instead of a tunneling current [26]; this method has since found widespread use in biological studies [8,27].

AFM imaging
AFM is a critical tool in biology, as it allows for the nanoscale characterization of cells and systems in a physiologically relevant (i.e.not vacuum) environment [9,28,29].Atomic force spectroscopy also allows for the measurement of nanomechanical properties of the sample, as well as the measurement of protein-membrane and protein-protein interactions [12,14,30].In AFM, a topographical image is generated by raster scanning a nanoscale probe over the sample surface.The forces interacting with the probe usually include mechanical contact, van der Waals forces, capillary forces and electrostatic forces [31]; probes may be chemically modified or coated, or have a bias applied, to measure different sample properties, such as magnetic properties and surface contact potential [32,33].These forces cause a bending of the probe as it passes over the sample.This deflection is tracked by a laser reflected off the cantilever and onto a position-sensitive photodiode.When the tip is in contact with the sample, the feedback system works to maintain constant forces between the tip and the sample, adjusting the height of the tip as the probe scans over the surface: this provides the sample topography [9].A schematic is shown in figure 1.
Contact mode AFM can result in high surface forces, detrimental to measurement of soft biological samples; the development of tapping mode AFM resulted in wider application to biological materials [8,34].In tapping mode, the cantilever is oscillated near its resonant frequency by piezoelectrics as it raster scans over the sample surface.The feedback system adjusts the z-piezo so that the amplitude of the cantilever oscillation remains constant, from which the sample topography can be extracted.In phase contrast imaging, the phase difference between the driven and measured oscillations of the cantilever is measured and mapped over the surface; these phase differences can be attributed to different material properties, giving an indication of qualitative differences in sample composition at the surface [35,36].

Atomic force spectroscopy
In contrast to the imaging modes of AFM which operate in the x-y plane of the sample, atomic force spectroscopy is an operation mode of the AFM that operates in the z-direction, measuring interaction forces on the probe as it approaches and retracts from contact.Nanoindentation experiments allow for the measurement of elastic modulus and adhesion of biological systems, values that cannot be quantified from imaging modes alone [37,38].
In force spectroscopy, the AFM tip is extended towards and retracted back from the sample surface, with the deflection graphed as a function of the extension of the piezoelectric: this forms the force-distance curve [38].The region of the force-distance curve where the tip is compressing the sample can be fit to a model in order to extract the elastic modulus; the choice of model is dependent upon the shape of the tip [13,15].If there is a sudden jump in the force, this is indicative of a rupture of the sample: the force necessary to achieve this is labeled the breakthrough force [11,19].
Single-molecule force spectroscopy In single-molecule force spectroscopy, where AFM is used to investigate interactions between individual biomolecules, including protein-protein interactions, a jump that is seen when a domain of the target protein is unfolded [37,39].On the retract curve, a negative peak may be observed as the tip adheres to the sample: this point defines the adhesion force [11].Functionalization of the probe is not limited to proteins, and alternate choices can allow for chemical recognition and mapping [38].
Nanomanipulation With fine control of the force of interaction between the tip and the sample, it is possible to manipulate and even perform lithography on surfaces and nanostructures.In cell biology, this has been applied to manipulate cells and associated structures [40,41], and nanolithography has been used to guide cell growth on surfaces [42].

Coupling scanning probe microscopy with other microscopic & spectroscopic techniques
AFM can be especially powerful when coupled with other techniques.In the following section, the methods combining AFM with optical microscopy and electrical measurements are discussed.

Scanning probe microscopy techniques
KPFM As an extension of AFM, Kelvin probe force microscopy (KPFM) measures changes in contact potential over a sample surface.Using a conductive probe with an AC voltage applied, the probe raster scans over the conductive sample surface.As the probe interacts with the electrostatic surface potential, oscillations are generated for the probe, generating feedback against the applied AC voltage to produce the contact potential difference (CPD) of the sample surface [32].This is performed concomitantly with AFM to produce topographical images along with CPD maps.Most commonly, this is done using a dual-pass technique, where the first pass measures the sample topography, the tip is raised by a set height, and the AC voltage is then applied, in order to measure the CPD of the surface [32,43].
SICM A limitation of conventional KPFM is that it is limited to the study of solid-gas interfaces, and so is limited in its applicability to studies in physiological conditions [44,45].Scanning ion-conductance microscopy (SICM) is a scanning probe microscopy technique that uses an electrode as the probe.It is well-suited to the measurement of biological samples, including living cells, as it requires an aqueous electrolyte-containing medium, and there is no requirement for sample conductivity.In SICM, the micro-or nanopipette probe holds an electrode, with another electrode in the bulk solution.A voltage is applied across these electrodes, creating an ionic current, and the conductance between the two electrodes is measured.When the probe is lowered to the surface, the ionic flux is restricted, reducing the conductance; at a set threshold, the position of the probe is recorded, and the tip is moved, repeating the measurement to record the topography of the sample [46,47].
FluidFM Fluidic force microscopy (FluidFM) is a scanning probe microscopy technique that combines AFM with microchannelled cantilevers, allowing for the dispensation of soluble molecules through an aperture in the AFM tip.The AFM force feedback controls the approach of the tip to the sample, allowing for extremely local stimulation and modification of surfaces in liquid environments [48][49][50].

Scanning probe combined with optical techniques
In applications on the microscale (e.g. in cells), AFM can be used in conjunction with conventional optical microscopy, such as fluorescence and confocal microscopy [51][52][53].For applications on the nanoscale, these techniques are limited by the diffraction limit of light [54].The following spectroscopic techniques have used the presence of the AFM tip as a probe for chemical identification of the sample in addition to the measurement of sample topography.
AFM-IR In order to measure the infrared spectrum of samples on the nanoscale, a technique called AFM-IR (AFM-infrared spectroscopy) is used.In AFM-IR, the sample is illuminated with a pulsed, tunable IR laser.The IR radiation is absorbed by the sample if it is in resonance with a molecular vibrational frequency.This causes local thermal expansion in the laser spot as the excited molecules return to their ground vibrational state.When probed by an AFM tip, this expansion corresponds to an oscillation of the cantilever proportional to the radiation absorbed by the sample.As the AFM tip scans along the surface, maps of IR absorbance and topographical data can be measured concomitantly with nanoscale detail [55,56].Further signal enhancement is achieved by using a quantum cascade laser (QCL) and tuning its repetition rate to match the resonant frequency of the AFM cantilever, increasing its oscillation amplitude.The IR spectrum of samples down to nanometer thickness can be acquired by using a gold-coated AFM tip, localizing the electric field at the tip apex, and a gold substrate [55,57].
S-SNOM In scattering-type scanning near-field optical microscopy (s-SNOM), a laser is focused on the apex of the probe as it interacts with the sample.As the tip moves over the sample, the laser light scattered from the tip changes as a function of the properties of the sample: s-SNOM involves tracking this scattered light [58].A QCL is focused on the surface of the sample through the use of a parabolic mirror [59].A mirror on top of the sample collects the light onto the detector.The backscattered signal of the laser on the tip while translating the reference mirror yields an interferogram.Subsequently, the Fourier transform of this interferogram can be calculated, resulting in the near-field spectrum of the sample.By tuning the QCL to a specific mode and scanning over the surface, a map of the intensity of this signal over the surface is obtained [58].
TERS Tip-enhanced Raman spectroscopy (TERS) involves the illumination of the tip-sample junction with a visible laser and measuring the enhanced Raman-scattered light.The tip, usually coated in gold, serves to both enhance the Raman signal and to locally perturb the evanescent waves containing nanoscale details of the sample, into propagating waves detectable in the far-field by a conventional detector.In this way, TERS results in an enhanced signal and spatial resolution beyond the diffraction limit of light [60,61].

Electrophysiology and scanning probe microscopy
In the study of ion channels and their electrical conductance in lipid bilayer, AFM has been combined with electrophysiology techniques.A lipid bilayer can be suspended over a nanopore in a microfabricated chip in a cell, with the AFM tip acting as one electrode and a platinum wire as the other.In this way, ion channels in the lipid bilayer can be both imaged simultaneously with the measurement of the electrical current flowing across the bilayer [62].

Applications of scanning probe microscopy in neuronal cells & tissues
Neurons are large enough to be imaged by a variety of microscopic techniques, such as fluorescence microscopy; with suitable fluorescent probes, super-resolution microscopy is possible, allowing for the visualization of cells and structures beyond the diffraction limit of light [7].For example, a two-photon fluorescent-lifetime-based probe has been used to monitor mitochondrial hydrogen peroxide and ATP in neurons in real-time [63].Imaging by AFM allows for different insights, such as the measurement of mechanoelastic properties, among the topics discussed here.

Surface properties and morphology of neurons
AFM-based techniques that generally rely on the electrical conductance of the sample, such as KPFM, are not widely applied to cells, as these techniques are not compatible with maintaining physiologically-relevant conditions.Recently, Zhao et al applied KPFM and electrostatic force microscopy to the study of dried PC12 cells (representative of the sympathetic nervous system) and hippocampal neurons (representative of the central nervous system), on the basis that other studies have found little difference in the CPD of dried biomaterials and the values obtained in liquid by other approaches [64,65].Their results show that the capacitance gradience of PC12 cells and hippocampal neurons are very similar, implying that the differences in nerve signaling and functions of the sympathetic and central nervous system are not related to the electric polarization properties.The CPD of neuronal spines is much more negative than that of cell bodies and processes, and the CPD of the cell bodies and processes of PC12 cells is smaller than that of the hippocampal neurons, indicating that surface potential is related to neural signal transduction functions, with neural spines playing a vital role in neural signal transmission [66].The AFM and CPD images of processes of PC12 cells and hippocampal neurons are shown in figure 2. A similar study on astrocytes, glial cells in the central nervous system, showed that the CPD of cell body was larger than that of the glial filaments, but their capacitance gradient was comparable [67].
Conventional SICM has been applied to many different studies, including that of neuronal cell migration [68], morphological changes of a neuron in the early stages of apoptosis [69], the measurement of cell stiffness due to amyloid-β [70], and to guide neuronal growth cones [71]; however, the temporal resolution of conventional SICM restricts time-lapse imaging of live cells.Takahashi et al have developed a scanning method in which the next imaging region is chosen by the prediction of the location of a cell, reducing the scanning speed of time-lapse SICM imaging by about half [72].They applied this to the visualization of nanoscale changes in hippocampal neurons.The application of this new algorithm allowed for the observation of the translocation of plasmalemmal precursor vesicles, which have no established fluorescent label, and the rearrangement of the neuronal cytoskeleton at the growth cone.

Mechanoelastic properties of neurons and tissues
In order to maintain the native state of the neuron, physiological conditions must be maintained: typically, the experiment is carried out in buffer with an adjustable temperature.Cells can also be fixed, but this can lead to mechanical artifacts, such as an increase in elastic modulus [73,74].Once the appropriate sample conditions are established, the AFM cantilever must be chosen.It is necessary to choose a cantilever with a similar spring constant to the sample in question, as cantilevers that are too stiff result in insensitive measurements and cantilevers that are too soft do not deform the sample sufficiently to measure sample stiffness.Probes used for the atomic force spectroscopy of cells range from microscale spheres and wedges to pyramids with apices on the scale of a few nanometers.Larger probes provide a larger sample contact, averaging out the mechanical properties measured.Cantilever and probe shape is also critical to consider in analysis of force spectroscopy data, as these are considered in choosing an appropriate model for the elastic modulus fit and extracting parameters such as indentation depth and adhesion force.Other cantilever measurements may also be relevant: for example, Ankundinov et al showed that, in the study of the elastic modulus of living sensory neurons, the elastic modulus decreased when the ratio of probe height to beam length of the cantilever increased [75].Another important consideration is the loading rate: as cells are highly complex systems with components with different mechanical responses, the properties of cells change nonlinearly with the loading rate [76][77][78].For this reason, mechanical properties of cells cannot be compared at differing loading rates.
In addition to probe properties, the local temperature and substrate hardness must be considered.In vivo, cells respond to the mechanical properties of their environment.Recently, it has been shown that AFM indentation of cortical neurons pushes them into soft substrates: previously, it had been shown that the stiffness of these cells was higher when they were placed on stiffer substrates.Further investigation has shown that including the indentation of soft substrates in the analysis resulted in cell stiffness measurements being largely independent of substrate stiffness [79].Temperature is also an important factors to consider.In AFM experiment and combined fluorescence experiments, the elastic modulus of neuronal somata was shown to decrease by approximately half when the temperature was adjusted from physiological to ambient temperature (i.e.37 • C-25 • C), while the volume of the soma was shown to increase by a factor of 1.3 [80,81].This was attributed to changes in the dynamics of the cytoskeleton components as a function of temperature [81].Therefore, in comparing nanomechanical responses of cells, it also important to consider substrate and temperature.
Studies made in whole tissues are further complicated by deformations, including extraction and slicing, resulting in a loss of native environment [82].In a comparison of tissue sectioning techniques, vibratomed sections of living tissues have been shown to allow for the nanomechanical measurement of cells in tissues, while cryotomed sections inhibited the measurement of viable cells [83].With an appropriate embedding matrix and modified AFM probes, measurement of the morphology and nanomechanics of live brain tissues has been demonstrated [84].
In a study of Alzheimer's disease, nanomechanical measurement of the stiffness of brain tissue in mice showed reduced stiffness [85].Nanomechanical indentation of human pituitary gland tissues has shown highly heterogeneous rigidities, with strong, subcellular gradients [86].Brain abscesses have been studied by atomic force spectroscopy, determining their elasticity and dissipated energy for the multilayer system to better estimate the compression and damage during their expansion [87].

Protein localization and protein-cell interactions
As different cell components have distinct nanomechanical properties, it is possible to map these components using force mapping, without relying on fluorescent labelling.Such mapping can be difficult due to nonspecific binding or multiple binding events.Recently, Lim et al mapped the location of LIMK1 proteins in cultured neuronal cells using an Anti-LIMK1 tethered AFM probe in force measurements.It was elucidated that LIMK1, a protein involved in the growth of neurons, are more abundant in dendritic spines [88].The force map showing the distribution of LIMK1 in neuronal somas is shown in figure 3.This high-resolution mapping can in the future be adapted using other antibody-protein target pairs, allowing for a better understanding of the spatial distributions of target molecules in a cell.In a similar manner, AFM tips tethered with a hybrid binding domain have been used to map microRNA in a cultured neuron using adhesion force mapping [89].
Interaction with amyloid-β, the accumulation of which is associated with Alzheimer's disease, has been shown to induce morphological changes in cell structure, inducing pore-like structures [90]; while these pore-like structures have not been imaged in living neurons, their presence has been supported by the observation of pore-forming oligomers in lipid membrane models [91,92].In addition to morphological changes, mechanical changes in the cell are critical to cell function.Changes in the lipid composition of the neuronal membrane associated with age have been shown to play a role in amyloid-β production and toxicity [93,94].In an AFM study of the effect of amyloid-β oligomers on living hippocampal neurons, it was demonstrated that changes in the neuronal elasticity were induced, with amyloid-β reducing the stiffness of living neurons, the magnitude of the reduction being correlated with the age of the neuronal preparation [16].

Nanomanipulation of neurons
When AFM is used concomitantly with other techniques, information on other features, such as chemical identity, can be obtained.In a study of rat cortical neurons, Gaub et al. developed an assay combining AFM with confocal microscopy, indenting the neurons by AFM and measuring their response by functional calcium imaging and cellular morphology by differential interference contrast microscopy.In comparing  nanoindentation to shear stress, they show that the neurons differentiate the magnitude and location of mechanical stimuli, and show either short-lived or sustained responses [95].
The study of the mechanical properties of neuronal cells can be extended to manipulating neurons and rewiring neuronal circuits.Injury to the central nervous system often leads to deficits in function, as it is not possible to regenerate axons over long distances.Using rat neurons guided by a beaded AFM probe, new, functional neurites have been created, resulting in a rewired neuronal network [96].It has been proposed that the creation of these new, functioning synapses can be extended to build an artificial neural network with biological neurons [97].Using a sharp cantilever, a force sufficient to sever the axon was applied, with the sample stage moving to make a clean cut.For reconnection, a different cantilever was used, with a bead coated in Poly-D-lysine (PDL) glued to the cantilever.By applying force on the axon with the beaded cantilever, a new neurite is formed, which is then reconnected; this is shown in figure 4.

Monitoring neuronal response to chemical stimuli
FluidFM allows for many applications in single-cell studies, including manipulating cells, attaching colloidal beads for use as indenters, and the delivery of biomolecules as chemical stimuli [50].FluidFM has been used to deliver the neurotransmitter glutamate to primary rate hippocampal neurons, and spiking activity was measured by concurrent microelectrode array and calcium recordings.This platform allows for the simultaneous modulation and recording of neuronal activity [49].FluidFM has also been shown to accommodate simultaneous SICM and AFM imaging on living neural cells.In this mode, the topography of sample is measured using the ionic current as feedback, with the force images of the sample generated simultaneously from the force feedback of cantilever [98].

Fluorescence microscopy and other SPM-based spectroscopic techniques for chemical identification
In the microscopy of cells, one of the most common bimodal techniques is the combination of AFM with optical microscopy, including fluorescence [99][100][101].These studies aim at understanding how nanomechanical properties, as measured by AFM, relate to the cell function and morphology [17,81,102].
While TERS has been applied to the study of amyloid-β at neuronal spines [103], optical techniques, such as TERS, AFM-IR and s-SNOM, are most applicable for the study of the proteins involved in neurodegenerative disease (see section 5.2), as these techniques are not currently well-suited to physiological conditions [56,104,105].In a proof-of-concept study, Freitas et al used s-SNOM to detect amyloid-β sheet structures on cell surfaces at the nanoscale [106].

Applications of scanning probe microscopy in cellular membrane models and brain lipid extracts
The cell membrane is a fundamental biological structure, studied extensively in membrane biophysics due to its role in neurodegenerative disorders such as Alzheimer's disease [22,107].They are often modeled using mixtures of synthetic lipids, reducing the complexity of the native membrane while mimicking native membranes to study their structure and function [10,23].In cells, lipids self-assemble into bilayers, with hydrophobic acyl tails surrounded by hydrophilic headgroups [108].Models made to mimic cell membranes can be of varying complexity, from a single lipid to a complex model of five or more lipids [22,109].These models can exhibit phase separation, forming micro-and nanoscale domains, which can be critical for membrane-protein interactions [110].Experimentally, these are formed via the vesicle fusion method, or by Langmuir Blodgett deposition on flat surfaces to form supported lipid bilayers (SLBs) [18,20].An alternative route is to form membranes using brain total lipid extracts (BTLE) [111].

Imaging and force spectroscopy of cellular membrane models
Complex lipid membranes exhibit structural nanodomains as a result of phase separation [112,113].Such nanostructures are expected to be ubiquitous in biomembranes, but currently super-resolution microscopy approaches are only able to observe large membrane domains [114].These domains are thought to play a role in cellular organization and signaling [115,116]; their influence on protein binding is thought to play a role in neurodegenerative disease, including Alzheimer's disease [117,118].Another aspect of model membranes that is important to consider is their symmetry, whether the domains align in the inner and outer leaflets; transmembrane asymmetry is a fundamental characteristic of biomembranes, but most model membranes exhibit symmetry [119,120].In the atomic force spectroscopy of neurons, the critical nanomechanical parameter is the elastic modulus, which is correlated to cell stiffness; conversely, in model lipid membranes, the breakthrough force, the force necessary to locally rupture the membrane, describes the resilience of the membrane [11].

Interaction with proteins involved in neurodegenerative disease
One of the most relevant and applications of AFM to the study of model lipid membranes is their interaction with proteins involved with neurodegenerative disease.

Amyloid-β
The hallmark of Alzheimer's disease is the accumulation of amyloid-β-containing plaques in the brain, and as such is the primary therapeutic target in experimental disease prevention and treatment strategies [121,122].On a ganglioside-containing model membrane, ganglioside cluster binding peptide was found to inhibit the formation of amyloid-β assemblies and removes fibrils deposited on the membrane [123].Recently, it has also been posited that while its misfolding and aggregation is associated with neurodegeneration, amyloid-β contributes to biological fitness, and effective therapies may require specificity towards toxic forms of amyloid [124].In a study using a complex neuronal model membrane it was shown that K162, a fluorene-based active drug candidate, inhibited permeation of amyloid-β into the lipid membrane, with only shallow defects formed.The aggregation of amyloid-β was modified to bypass toxic oligomers, allowing for the preservation of neurologically beneficial monomeric species; this is shown in figure 5 [109].This has also been shown for trodusquemine, which enhances the aggregation of amyloid-β, converting toxic oligomers to less toxic oligomeric species [125].
Different amyloid-β species have been found to have distinct effects on the model membranes.Oligomers of amyloid-β been shown to have a detergent-like behavior on PC/GM1/Cholesterol bilayers, causing holes within the bilayer, while fibers tended to embed into the lipid bilayer [126].On a BTLE bilayer, large amyloid-β oligomers were shown to accelerate the generation of fibrils, while smaller oligomers created pores and disintegrated into the membrane.Using AFM-based nanomechanical mapping, it was found that both forms decreased the elastic modulus of the membrane by ∼45%; both forms would have a neurotoxic effect, but their mechanism likely varies, given the differences in their interaction with the membrane [127].In a study of three mutants of amyloid-β peptides and their interactions with SM/PC/Cholesterol/GM1 membranes, high-speed AFM and nanoscale infrared techniques were used to characterize the amyloid peptides and their fibrillization products, with AFM-IRs providing information on their secondary structure.It was their conclusion that membrane disruption is associated with amyloid-β species exhibiting an anti-parallel β-sheet secondary structure, which interact strongly with the membrane and are only transient small oligomeric species in the early stages of aggregation [128].
Similarly, different membrane components have been found to have distinct interactions with amyloid-β.Accumulation of cholesterol in membranes is associated with the development of Alzheimer's disease, suggesting that it may be a factor in amyloid-β aggregation.With time-lapse AFM imaging of a PC/PS/Chol bilayer, physiological concentrations of cholesterol have been observed to enhance the aggregation of amyloid-β at nM monomer concentration, with both the quantity and size of oligomers increasing in the presence of cholesterol.The model also exhibited a dynamic aggregation process, with dissociation from the bilayer surface into the bulk solution a possibility [129].Cholesterol has also been shown to positively correlate with the average pitch size of twisting periodicity of amyloid-β fibrils [130].In a study of the effect of the model amyloid-β (1-42) peptide oG37C on different biomimetic membranes by high-speed AFM, the association of cholesterol with ganglioside GM1 was necessary to observe detergent-like effects on the lipid bilayer, with no membrane dissolution observed with either component alone.The authors, Ewald et al., suggest a two-step mechanism is involved, with the attachment and accumulation of oG37C to the GM1 domains of the membrane, followed by their insertion into the membrane via the nearby cholesterol [131].In a study of using microcavity SLBs, the amyloid-β interaction with DOPC bilayers was compared to that of amyloid and asymmetric DOPC/DOPS bilayers.Monomeric amyloid-β was found to adsorb weakly to DOPC membranes, but doping 10 mol% DOPS asymmetrically into the outer leaflet of the membrane resulted in oligomerization as observed by AFM, with membrane pore formation visible within half an hour [132].In order to assist the elucidation of peptide-membrane interactions, simpler model systems may also be required: the interaction of short peptides with a DPPC model membrane has been used to model the amyloid-membrane interaction.These peptides, called [RF] and [RF] 4 , adopted β-sheet structures after adsorption to the DPPC membrane, and bilayer pores and aggregation at different interfaces were observed, suggesting that the hydrophobic residues might facilitate membrane disruption [133].

α-synuclein
The self-assembly of the α-synuclein protein into various aggregates is associated with the development of Parkinson's disease.It has been postulated that the interaction of α-synuclein with lipid membranes is a critical factor in the initiation of the formation of aggregates, and directs the overall aggregation process.The efficiency of aggregation has been shown to be dependent upon the lipid composition of the membrane [134,135], and that assembled aggregates can dissociate from the surface, allowing for the possibility of surface aggregation as a mechanism in disease pathology [135].In an AFM-IR study of α-synuclein grown in the presence of phosphatidylcholine and phosphatidylserine, a gradual decrease in the amount of parallel β-sheet structures and an increase in the α-helix and unordered protein structures was observed over time, indicating that the presence of lipids in the oligomeric structures prevented an expansion into the parallel β-sheet structure upon their interaction with monomeric α-synuclein [136].The secondary structures of the oligomers grown in the presence of these two lipids differed from each other and from those grown in a lipid-free environment, likely due to the incorporation of lipids with different charges affecting protein aggregation [137].
In a study of the effect of cholesterol on the binding and aggregation of α-synuclein, it was found that the strongest binding occurred with membranes containing 10% cholesterol, corresponding to the highest vesicle clustering and lowest aggregation, indicating the lowest risk for neurodegenerative disease [138].The inclusion of ganglioside (GM1) has been shown to lead to a stronger interaction with α-synuclein in both its monomeric and oligomeric forms [139].In a complex model mimicking the composition of the inner and outer leaflets of neuronal membranes, the internal leaflet mixture was found to have a higher stability in its interaction with α-synuclein, while damage to the external leaflet was observed in a concentration-dependent manner [140].
In a lipid membrane exhibiting nanodomains (DOPC/SM/Cholesterol), protein monomers resulted in membrane thinning, targeting the disordered lipid domains.When combined with ferrous cations, the formation of protein aggregates was promoted, targeting the solid-ordered domains (figure 6) [24].This underlies a new mechanism of membrane-protein interaction in Parkinson's disease.

Tau fibrils
Tau proteins assemble into paired helical filaments, which aggregate into neurofibrillary tangles; these tangles are linked to neurodegeneration in Alzheimer's disease and related tauopathies.In the study of tau on supported brain lipid extracts, it was found to attach and self-assemble on the membrane dependent on the presence of cations: sodium triggers this process, but potassium inhibits it.The stability of these tau-BTLE assemblies was tested using the AFM tip to mechanically perturb them, and they were found to be stable at 1 nN of applied force, exhibited partial distortion at 2 nN, and dissection at 5 nN.They were found to remain stable in the presence of sodium and lithium, but disassembled in the presence of potassium and rubidium [141].The presence of these ions on either side of the neuronal membrane may be relevant: potassium in the intraneuronal fluid could inhibit the aggregation of tau, while sodium in the extracellular fluid could facilitate the interaction of tau with the membrane.
The tau construct K18 showed detergent-like membrane solubilization, with lipid composition modulating the behavior of the peptide.These deleterious effects were observed on the liquid disordered phases of zwitterionic membranes, with no prerequisite aggregation or fibrillization of K18 [21].

Huntingtin
Huntington disease is an inherited neurodegenerative disease caused by expansion of a glutamine repeat region of the huntingtin protein.This promotes the formation of a variety of oligomeric and fibrillar aggregates, which then accumulate to form protein plaques.Huntingtin is associated with cellular and subcellular lipid membranes, which may influence this aggregation.To determine the influence of these lipid types, Chaibva et al added exogenous sphingomyelin and ganglioside (GM1) to BTLE.The addition of either of these was found to decrease huntingtin insertion into the lipid monolayer; however, vesicles with an increased SM content were more susceptible to permeabilization [142].In a study of the influence of lipid headgroup identity in lipid vesicles, a complex relationship between head group and effect was found, with anionic lipids enhancing huntingtin fibrillization, but more membrane activity was observed in zwitterionic lipid systems [143].Lipids with different tail groups were also shown to affect fibrillization, but the degree of protein-lipid complexation did not correlate with huntingtin aggregation, indicating that different lipid systems uniquely alter the aggregation mechanisms [144].The oxidation of huntingtin fibrils resulted in unique morphological features and, with and without the presence of a BTLE membrane, this promoted oligomerization over fibril elongation.In the presence of a membrane, altered huntingtin-induced morphological changes as observed by AFM [145].

Applications of scanning probe microscopy in single-molecule studies
The in vitro microscopy and spectroscopy of proteins allows for the observation of their fundamental structure and behavior.

Single-molecule force microscopy of proteins and inhibitors
In single-molecule force spectroscopy, proteins adsorbed or bonded on a surface are manipulated with an AFM tip, and their behavior is analyzed by force-distance curve.Often, the probe is functionalized with a protein or molecule of interest, to test interaction with the protein immobilized on the substrate.In the force-distance curve, unfolding events are observed as jumps.The stretching region proceeding these jumps can be fit with an elasticity model, and distances between peaks can be used to assign domains to the unfolding events [146].Prion protein monomers and dimers have been studied by single-molecule force spectroscopy to investigate their potential initial oligomerization processes, as the misfolding is involved in prion disease; this has been widely studied using other techniques [147][148][149][150]. Two dimeric constructs were studied, with different orientations: N-C and C-C terminal orientations.The C-C dimer was found to unfold at a higher force compared to N-C, implying a more stable structure, and one that may be a building block in amyloid fibril formation [151].
In the analysis of force-distance curves, atomistic simulations can help identify interactions.Churchill et al have applied this approach to the study of dimers of α-synuclein, the simplest oligomer, with simulated single-molecule results compared to experiments.Simulated results included more transient intermediates than experimental observations, but were otherwise consistent, identifying β-rich dimer structures [152].Molecular dynamics can also be used to design and optimize structures to effect local binding.One strategy in the development of treatment of Alzheimer's disease is to prevent the formation of toxic oligomer; SG inhibitors are a class of pseudopeptides designed to inhibit amyloid aggregation [153,154].Using single-molecule force spectroscopy, these have been experimentally shown to inhibit amyloid-amyloid binding, and some have shown neuroprotective abilities in cell models [155].
Most commonly, probes are functionalized with the protein under study in order to determine aggregation pathways.Aptamers, sequences of DNA or RNA that bind a specific molecule, protein or peptide, have previously been used to target amyloid [156][157][158][159].It is also possible to perform single-molecule force spectroscopy with an aptamer-functionalized tip.Aptamer-functionalized force spectroscopy allowed for the observation of an increase in binding probability of amyloid oligomers after application of an electric field, while the electric field transformed amyloid-β monomers and fibrils to oligomeric structures [160].Extension of this work to incorporate surface plasmon resonance sensing has shown the detection of amyloid-β aggregates in cerebrospinal fluid [161].

Microscopy and spectroscopy of proteins
AFM and associated spectroscopic techniques have also been applied to the study of nucleation and aggregation pathways of neurodegeneration-associated proteins, as well as critical observations of their secondary structure.In the high-speed AFM study of amyloid nucleation and growth, Huang et al chose amylin as the model peptide, showing that protofibril aggregation was positively correlated with monomer concentration but not with fibril elongation.In maps of the elastic modulus, it was possible to distinguish growing and passivated ends of the fibril, with the growing end having a lower stiffness than the aggregate body and a stable structure and the passivated end experiencing rearrangements of β-structures [162].With AFM-IR, amyloid oligomers have been shown to exhibit structural heterogeneity (figure 7).This heterogeneity is retained in protofibrils and fibrils, with structurally different domains extant in the same fibril [163].Using TERS in liquid, Lipiec et al resolve the amide I and amide III bands of amyloid-β, allowing for detailed analysis during aggregation.During the conversion of protofibrils to fibrils, a rearrangement of β-sheets from antiparallel to parallel was observed [164].The mapping of amyloid fibril formation with TERS has allowed for the detection of an aggregation pathway, confirming that these pathways are related to dynamic conformational changes and the distribution of different secondary structures (turns and random coils vs. β-sheets) [165].
In AFM-IR, artifacts are possible, making interpretation of spectra for the determination of secondary structures difficult.Experimentally, important factors to consider include incident polarization, the tip material and illumination configuration: in the study of amyloid fibrils in particular, a bottom-up illumination, s-polarized incident beam and an Si-tip have been recommended [166].In the characterization of the suitability of different substrates for the AFM-IR of amyloid oligomers, silicon, zinc sulfide and calcium fluoride were found to be the best candidates, having low roughness and spectral background.Additionally, gold-coating the silicon substrates was shown to increase the enhancement factor of AFM-IR by about a factor of 7 [167].Deposition on zinc selenide has been shown to induce restructuring of amyloid fibrils, resulting in amorphous aggregates [168].
In the AFM study of the early aggregation of α-synuclein, monomers were shown to first aggregate to protofilaments, single-strand chain-like aggregates, with subnanometer diameter and micrometer length.In force spectroscopy experiments, the protofilaments were shown to have nanomechanical and thermodynamic properties that were distinct from those of mature amyloid fibrils, exhibiting a less rigid structure.Figure 8 shows the force responses of α-synuclein protofilaments [169].
Using total internal reflection TERS (TIR-TERS), Talaga et al showed a TERS enhancement nearly 30 times higher than in conventional bottom-illumination configuration, and 8 times higher when radially polarized.In the application of this method to the study of tau fibrils, aromatic amino acid residues can be distinguished from non-aromatic residues, and amide I and amide III bands can be detected.This allowed for the correlation between fibril composition and their structure: antiparallel β-sheets and fibril core β-sheets were found in amino acid-rich regions, where parallel β-sheets and random coil were less present [170].With AFM-IR, individual tau fibrils were investigated at different aggregation stages.Structurally distinct polymorphss were observed, with similar morphologies but different secondary structures; in particular, in early stages of aggregation, fibrils exhibiting a transient ordered parallel β-sheet structure were observed.In later stages, mature fibrils exhibited structures containing antiparallel β-sheets [171].

Conclusions
AFM and related scanning probe microscopy techniques are a powerful tool for visualizing the nanoscale structure of neurons and their components, lipid membrane and neurodegeneration-related proteins; its combination and extension, such as in KPFM, atomic force spectroscopy, AFM-IR, and TERS, allows for the simultaneous study of other properties, such as nanomechanics, electrostatic potential, and secondary structure.In both physiological conditions and in vitro, these techniques are capable of providing insight into physical nanostructures and biochemical interactions.The application of AFM and related techniques to neuroscience is continuously evolving, providing new insight into different physiological processes.

Figure 1 .
Figure 1.Schematic of atomic force microscopy on a cell.

Figure 2 .
Figure 2. The contact potential difference (CPD) of processes of PC12 cells and hippocampal neurons in air by KPFM.(a) The topography of PC12 cells, with (b) the magnification of the region denoted in (a); (c) the corresponding CPD image, the histogram of which is shown in (d); (e) the topography of hippocampal neurons, with (f) the magnification of the region denoted in (e); (g) the corresponding CPD image, the histogram of which is shown in (h).Reproduced from reference [66].CC BY 3.0.

Figure 3 .
Figure 3. High-resolution force mapping of LIMK1 on neuronal somas.(A) A fluorescence microscopic image: Hippocampal neuron at 14 d, fixed and immunostained.(B) AFM image of the region denoted in (A).(C) Force map of the arbitrary-chosen area denoted in (B), with LIMK1 proteins marked by circles.Map measures 150 × 150 nm, 50 × 50 pixels.(D) Measured densities (per 150 nm × 150 nm area) of LIMK1 for neuronal somas under different conditions.Three native somas and three depolarized somas were examined, and four regions of each soma were investigated.For control experiments, control 1 indicates the soma without the membrane removal, and control 2 indicates the soma blocked with free LIMK1 antibodies.Error bar represents the standard error of the mean.* * * P < 0.001.Scale bars are 20 µm.Reprinted with permission from reference [88], copyright (2022) American Chemical Society.

Figure 4 .
Figure 4. Axotomy & reconnection by AFM.(a) A single healthy axon is cut (dashed line) by a sharp cantilever.(b) The axon following the cut.(c) Using a bead glued to bottom of cantilever, a synapse is formed with the proximal end of the axotomized axon.(d) A new neurite is generated by the application of force on the axon by the cantilever and bead.(e) The neurite reconnects the axon's proximal and distal ends; the axons appear to be reconnected.Reproduced with permission from reference [97].CC BY 4.0.

Figure 5 .
Figure 5. AFM topography images of (a) the bilayer lipid membrane (50% DSPE, 15% DPPC, 25% Chol, 8% SM, and 2% GM1 by weight), (b) the membrane in the presence of amyloid-β oligomers and (c) the membrane in the presence of amyloid-β oligomers and K162, an active drug candidate.(d)-(f) show the cross-sectional profiles as extracted along the lines as indicated in the corresponding images (a)-(c).Reproduced with permission from reference [109].CC BY 4.0.

Figure 6 .
Figure 6.Interaction of iron-induced α-synuclein oligomers with raft-like supported lipid bilayer.(A) AFM topographic images of membrane-protein interaction monitored over time.Blue circles show the nucleation of oligomers and its time-evolution.The scale bars measure 1 µm.(B) Analysis of relative surface areas of lipid domains and oligomeric clusters vs time.Distribution of (C) heights and (D) areas of clusters.Reproduced from reference [24] with permission from the Royal Society of Chemistry.

Figure 7 .
Figure 7. Structural characterization of individual early-stage amyloid-β aggregates.(A) AFM topographic image of oligomers generated after 3 d of incubation in PBS buffer.(B) Representative IR spectra of the amide I region obtained from different individual oligomers.(C) AFM image of protofibril and (D) corresponding IR spectra recorded from different regions of the protofibril cluster.Reproduced with permission from reference [163], copyright (2022) American Chemical Society.

Figure 8 .
Figure 8. Force spectroscopy of α-synuclein protofilaments.(A) Model of pulling a single protofilament and force-distance curve, showing: (1) the AFM tip pressed strongly against a protofilament; (2) the strong pressure breaking the polymer and a constant force plateau due to its unzipping from the surface, (3) the tip-filament contact is lost during the pulling, and; (4) the tip return to its initial position.The force-distance curves of (B) one protofilament, showing a single force plateau; (C) two protofilaments, with a double plateau, and; (C) four protofilaments.Reproduced with permission, from reference [169].