Imaging of Extracellular Vesicles by Atomic Force Microscopy

Summary

A step-by-step procedure is described for label-free immobilization of exosomes and extracellular vesicles from liquid samples and their imaging by atomic force microscopy (AFM). The AFM images are used to estimate the size of the vesicles in the solution and characterize other biophysical properties. 

Abstract

Exosomes and other extracellular vesicles (EVs) are molecular complexes consisting of a lipid membrane vesicle, its surface decoration by membrane proteins and other molecules, and diverse luminal content inherited from a parent cell, which includes RNAs, proteins, and DNAs. The characterization of the hydrodynamic sizes of EVs, which depends on the size of the vesicle and its coronal layer formed by surface decorations, has become routine. For exosomes, the smallest of EVs, the relative difference between the hydrodynamic and vesicles sizes is significant. The characterization of vesicles sizes by the cryogenic transmission electron microscopy (cryo-TEM) imaging, a gold standard technique, remains a challenge due to the cost of the instrument, the expertise required to perform the sample preparation, imaging and data analysis, and a small number of particles often observed in images. A widely available and accessible alternative is the atomic force microscopy (AFM), which can produce versatile data on three-dimensional geometry, size, and other biophysical properties of extracellular vesicles. The developed protocol guides the users in utilizing this analytical tool and outlines the workflow for the analysis of EVs by the AFM, which includes the sample preparation for imaging EVs in hydrated or desiccated form, the electrostatic immobilization of vesicles on a substrate, data acquisition, its analysis, and interpretation. The representative results demonstrate that the fixation of EVs on the modified mica surface is predictable, customizable, and allows the user to obtain sizing results for a large number of vesicles. The vesicle sizing based on the AFM data was found to be consistent with the cryo-TEM imaging. 

Introduction

Extracellular vesicles (EVs) are present in all body fluids, including blood, urine, saliva, milk, and the amniotic fluid. Exosomes form a district class of EVs differentiated from other EVs by endosomal biogenesis, the markers of the endosomal pathway, and the smallest size among all EVs. The size of exosomes is often reported with substantial variability between studies. The sizing results were found to be method dependent, reflecting the difference in physical principles employed by different analytical techniques to estimate EV sizes^1,2. For example, the nanoparticle tracking analysis (NTA) ― the most widely used size characterization technique ― estimates the size of EVs as their hydrodynamic diameters, which characterize the resistance to the Brownian mobility of EVs in the solution. A larger hydrodynamic diameter of a vesicle implies its lower mobility in liquid. The coronal layer around vesicles, consisting of surface proteins and other molecules anchored or adsorbed to the membrane surface, substantially impedes the mobility and increases the hydrodynamic size of EVs. In relative terms, this increase is particularly large for the exosomes³, as illustrated in Figure 1. 

The cryogenic transmission electron microscopy (cryo-TEM) imaging is a definitive technique in characterizing vesicle sizes and morphology in their hydrated states. However, the high cost of the instrumentation and the specialized expertise needed to use it correctly motivate the exploration of alternative techniques that can image hydrated EVs. A relatively small number of EVs observed or characterized in the acquired cryo-TEM images is another notable disadvantage of this technique.

Atomic force microscopy (AFM) visualizes the three-dimensional topography of hydrated or desiccated EVs^4,5,6 by scanning a probe across the substrate to raster the image of the particles on the surface. The essential steps of the protocol to characterize EVs by AFM are outlined in this study. Before imaging the vesicles in liquid, they must be immobilized on a substrate by either tethering to a functionalized surface, trapping in a filter, or by electrostatic attraction⁷. The electrostatic fixation on a positively charged substrate is a particularly convenient option for immobilization of exosomes known to have a negative zeta potential. However, the same electrostatic forces that immobilize the extracellular vesicles on the surface also distort their shape, which makes post-imaging data analysis essential. We elaborate this point by describing the algorithm that estimates the size of the globular vesicles in the solution based on the AFM data on the distorted shape of the exosomes immobilized on the surface.

In the developed protocol, the procedure for the robust electrostatic immobilization of vesicles is presented and followed by the steps needed to perform atomic force imaging in the hydrated or desiccated states. The factors that influence the surface concentration of the immobilized vesicles are identified. The guidance is given on how to perform the electrostatic immobilization for samples with different concentrations of EVs in the solution. The selection of experimental conditions permitting the estimation of empirical probability distributions of different biophysical properties based on a sufficiently large number of immobilized vesicles is discussed. Examples of post-imaging analysis of the AFM data are given. Specifically, an algorithm is described for determining the size of vesicles in the solution based on the AFM characterization of immobilized EVs. The representative results show the consistency of the vesicle sizing by AFM with the results of cryo-TEM imaging. 

Protocol

1. Isolation of EVs from a biofluid

1. Isolate EVs by one of the established methods, such as the differential ultracentrifugation⁸, precipitation, or size-exclusion chromatography⁹.
2. Confirm the presence of expected surface and luminal biomarkers and the absence of biomarkers indicating cross-contamination of the preparation. Confirm the lipid bilayer morphology of the isolated particles by electron microscopy.
   NOTE: When isolating the exosomes, the hydrodynamic size distribution measured by nanoparticle tracking analysis (NTA) or dynamic light scattering should be in the expected range. The details of EV and exosome isolation are beyond the scope of this protocol. The selected method will depend on experimental questions and the goal of the study¹⁰. The following steps provide a concrete illustration of the procedure to enrich the exosomes by precipitation from the growth medium of MCF-7 breast cancer cells using a commercially available precipitation kit (Table of Materials).
3. Before cell culture expansion, store MCF-7 breast cancer cells in liquid nitrogen. Thaw cells to subculture.
4. Following aseptic practices, perform cell plating on 150 mm plates. Use the growth medium composed of the Eagle’s minimum essential medium, 0.01 mg/mL human recombinant insulin, and 10% exosome-free fetal bovine serum.
5. Aerate the culture by 95% air and 5% CO₂ and incubate at 37 °C.
6. After the cells are settled (approximately 24 h after plating), change the media. Split the plate at 1:10 ratio and culture ten plates, each containing 20 mL of media.
7. Harvest and pool media from 9 of these plates (180 mL) at ~70−80% confluence when cells are still in the growth phase.
8. Divide the media into 60 mL and 120 mL, further split into 30 mL/tube, and centrifuge at 3,000 x g for 15 min.
9. Transfer the supernatant from each tube to a new sterile 50 mL tube and perform the exosome isolation.
10. Isolate exosomes by precipitation according to published protocols (see, for example, reference¹¹) or follow the manufacturer’s instructions if a commercial isolation kit (Table of Materials) is used. As a first step in the latter case, centrifuge cell medium at 3,000 x g for 15 min. Withdraw supernatant and discard cells and cell debris.
11. Add the precipitation solution (1:5 volume ratio) to the supernatant, mix, and refrigerate overnight.
12. Centrifuge at 1,500 x g for 30 min at room temperature. Discard the supernatant after centrifugation.
13. Spin the remaining exosome pellet for another 5 min at 1,500 x g. Without disturbing the pellet, remove the remaining precipitation solution by aspiration.
14. Resuspend the pellet in 100−500 µL of 1x phosphate-buffered saline (PBS) buffer and divide into multiple aliquots as needed for the downstream analysis.
15. Immediately proceed to the surface immobilization of the isolated exosomes for AFM imaging. If necessary, freeze the aliquots at -80 °C for later use while taking precautions to avoid damage to the sample during the freeze-thaw cycle.

2. Surface fixation of extracellular vesicles

1. Use strong double-sided tape, epoxy, or an alternative adhesive to firmly attach a mica disk to an AFM/scanning tunneling microscope (STM) magnetic stainless-steel specimen disk.
2. Cleave mica disc by using a sharp razor or utility knife, or by attaching an adhesive tape to the top surface and then pealing it off to remove a layer of material.
   NOTE: Either method should reveal a virgin surface by removing a thin layer of mica previously exposed to the environment. After the procedure, the attachment of mica to the AFM/STM metal specimen disk must remain firm.
3. At room temperature, treat the top surface of mica for 10 s with 100 µL of 10 mM NiCl₂ solution, which modifies the surface charge from negative to positive.
4. Blot NiCl₂ solution with a lint-free wipe or blotting paper. Wash the mica surface 3x with deionized (DI) water and dry it with a stream of dry nitrogen.
   NOTE: It is a good practice to scan the modified surface with AFM to confirm it is free from contaminants.
5. Place the AFM specimen disk with the attached surface-modified mica in a petri dish.
6. Dilute the exosome sample from step 1.14 with 1x PBS to obtain a concentration between 4.0 x 10⁹ and 4.0 x 10¹⁰ particles per mL of solution. Validate the diluted particle concentration using NTA.
7. Form a sessile drop on the surface of mica by emptying 100 μL of the diluted exosome solution from a pipette.
8. Place lid on the petri dish and seal it with a paraffin film to reduce sample evaporation. Incubate the sample for 12−18 h at 4 °C.
   NOTE: The surface density of the immobilized exosomes will increase with the incubation time and the concentration of EVs in the liquid. Longer incubation time may be necessary if exosomes are present in the sample at lower concentrations.
9. After incubation, aspirate 80−90% of the sample without disturbing the surface. At this point, the exosomes will be electrostatically immobilized on the mica substrate.
10. Before imaging hydrated EVs, rinse the surface with 1x PBS. Repeat 3x. Take care to keep the sample hydrated throughout the rinsing process.
11. After washing the mica surface with 1x PBS, remove 80%−90% of liquid, and pipette ~40 μL of fresh 1x PBS to cover the sample.
12. When imaging the desiccated EVs, rinse the substrate with DI water. Repeat 3x.
    NOTE: Rinsing with DI water will prevent the formation of salt crystals and the deposition of solutes on the surface as the substrate dries.
13. Before imaging desiccated EVs, aspirate as much liquid as possible without touching the surface and dry the rest with a stream of dry nitrogen.

3. AFM imaging

1. To image the desiccated EVs, select a cantilever designed for scanning in the air in tapping and non-contact imaging modes and mount it onto the probe holder.
   NOTE: The characteristics of an example cantilever listed in Table of Materials (123 μm length, 40 μm width, 7 nm tip radius, and 37 N/m spring constant) may be used as a guide when selecting a probe compatible with the available AFM instrumentation.
   1. Place the preparation from step 2.13 on the AFM stage. The magnetic stainless-steel specimen disk will immobilize the sample on the stage. Allow time for the preparation and the stage to equilibrate thermally.
   2. Use the tapping mode to scan a sufficiently large area of the mica’s surface. For example, choose an area of 5 x 5 µm, rastered in 512 lines at a scan rate of ~1 Hz. Acquire both the height and phase images as they provide complementary information on the topography and the surface properties of the sample.
      NOTE: The scan time will increase with the imaged area and the number of lines selected to form the image but decrease with the scan rate defined as the number of lines scanned per second. Fast scan rates may impact the image quality. Therefore, the speed of rastering should judicially balance the tradeoff between the acquisition time and the image quality.
2. To image hydrated vesicles, select a cantilever appropriate for scanning soft, hydrated samples and mount the cantilever onto a probe holder designed for scanning in liquids.
   NOTE: When selecting a probe compatible with the available AFM instrumentation, the specifications of the probe listed in Table of Materials (triangular cantilever with 175 µm nominal length, 22 µm width, 20 nm tip radius, 0.07 N/m spring constant, and optimized for imaging with the drive frequency in the range between 4 to 8 kHz) may be used as a guide.
   1. Wet the tip of the cantilever with 1x PBS to reduce the likelihood of introducing air bubbles into the liquid during scanning.
   2. Place the preparation from step 2.11 on the AFM stage. The magnetic stainless-steel specimen disk will immobilize the attached mica containing immobilized EVs on its surface.
   3. Allow time for the preparation and the AFM stage to equilibrate thermally.
   4. Image the hydrated mica surface in the tapping mode. Acquire both the height and phase images.
      NOTE: The imaging quality is influenced by the instrumentation, selected probe, and scan parameters. When optimizing the scanning conditions, the following choices may be used as a starting point: 5 x 5 µm area scanned in 512 lines with ~0.8-1.0 Hz scan rate and drive frequency between 4 to 8 kHz.

4. Image analysis

NOTE: The following data processing and analysis steps are applied to the acquired height images. A similar procedure may be adapted to analyze the phase data. The description below is specific to Gwyddion¹², a free and open source software available under GNU General Public License. Similar capabilities are available in alternative software tools.

1. Go to Data Process, SPM modes, Tip and choose Model Tip (Figure 2). Select the geometry and the dimensions of the tip used to scan the sample and click OK.
2. Correct the tip erosion artifacts by performing the surface reconstruction. Open the image. From the menu, select Data Process, SPM modes, Tip, then choose Surface Reconstruction and click OK (Figure 3).
3. Align the imaging plane to match the laboratory XY plane by removing the tilt in the substrate from the scan data. To accomplish this task, select Data Process, Level and choose Plane Level (Figure 4).
4. Align rows of the image by selecting Data Process, Correct Data and then choose Align Rows. Several alignment options are available (Figure 5). For example, Median is an algorithm that finds an average height of each scan line and subtracts it from the data.
5. Go to Data Process, Correct Data and choose Remove Scars (Figure 6), which removes common scanning errors known as scars.
6. Align the mica surface at the zero height, Z = 0, by selecting Flatten Base in Level drop-down menu accessible from Data Process (Figure 7).
7. Identify EVs on the scanned surface by using Mark by Threshold in Grains drop-down menu (Figure 8A). This algorithm identifies surface-immobilized exosomes as particles protruding from the zero-surface substrate by the height above the user-selected threshold. Select a threshold in the range between 1 and 3 nm, which will eliminate most of the background interference. Smaller thresholds are used with cleaner background.
   NOTE: The threshold in Figure 8A is 1.767 nm. The outcome of the MCF-7 exosome identification with this thresholding is shown in Figure 8B. Gwyddion offers several alternatives to thresholding as the algorithm to automatically identify vesicles in the image, including automated thresholding (Otsu's method), edge detection, and the watershed algorithm.
   NOTE: Particle agglomerates, if present in the AFM image, may be masked and excluded from the analysis.
8. Perform geometric and dimensional characterization of the identified EVs using the available Distributions algorithms accessible from Grains menu.
   NOTE: Gwyddion provides tools to assess the distribution of scalar-valued, areal, volumetric, and other properties of immobilized EVs in a hydrated or dessicated state. An example of a scalar-values property is shown in Figure 9, which gives the distribution of maximum heights within the footprint of each identified exosome.
9. Export the AFM data from Gwyddion for specialized analysis by other computational tools and custom computer programs.

Representative Results

Surface fixation of EVs is a critical step in the imaging sequence. Electrostatic surface immobilization of exosomes, known to have a negative zeta potential, will robustly occur after the mica’s substrate is modified to have a positive surface charge. Without the treatment with NiCl₂ to impart positive surface changes, the immobilization of EVs on the substrate was found to be ineffective. The height image in Figure 10A, acquired in the air after the MCF-7 exosome sample containing 2.59 x 10¹⁰ vesicles per mL of PBS was incubated for 12 h on unmodified surface of freshly cleaved mica, shows very few vesicles remaining on the surface after it was cleaned with DI water. The vesicles visible in Figure 10A are, most likely, the result of incomplete aspiration of DI water, which resuspended vesicles not fixed to the surface and then deposed them on the substrate as it evaporated.

After modifying the surface charge with nickel chloride, it is advisable to confirm that the surface remains free of contaminants after the treatment. The height image in Figure 10B (obtained in the air) gives an example of a clean surface after it was treated with NiCl₂ and then washed three times with DI water. The roughness of cation-derivatized surface was below 0.3 nm, which is consistent with the previous report¹³. 

The dramatic positive impact of the surface charge modification on the efficiency of the fixation of MCF-7 exosomes is illustrated by Figure 10C,D. These two panels show the height scans acquired in the air after the sample, previously imaged in Figure 10A, was incubated for 24 h and 12 h, respectively, on the surface treated with nickel chloride.

The time a given sample is incubated on the treated surface determines the surface concentration (vesicles per area) of the immobilized EVs. The height image in Figure 10C illustrates the case of excessively dense surface coverage by the immobilize vesicles obtained after the described MCF-7 exosome sample was incubated for 24 h. A number of algorithms rely on having sufficient unoccupied substrate between the grains to perform image correction and data analysis. For example, leveling and shifting the substrate to the zero plane, line correction, and estimation of the grains’ volume need the intervening flat surface to perform accurate calculations. When the concentration of the immobilized vesicles is as high as in Figure 10C, these algorithms will not function reliably. An example of an adequate surface concentration of vesicles immobilized from the same MCF-7 sample is shown in the height image in Figure 10D, which was obtained after shorter (12 h) incubation.
 
The post-processing of the acquired raw AFM data is needed to correct for common scanning errors. The following description is specific to Gwyddion. Similar functionality is available in other AFM/SPM data analysis tools.

Within Gwyddion, Plane Level function is used to correct for a tilt in the substrate. Such background correction is accomplished by first finding the plane of the substrate using all data points in the image and then subtracting it from the raw data. The correction along the scan lines is accomplished by Align Rows function. For example, one of the implemented algorithms performs the alignment by computing the median height of each scan line and then subtracting the result from the corresponding row of image data. The contribution of the local faults in the feedback loop can be removed by applying Remove Scars function, which fills the gaps in the aligned data and eliminates the scars by comparing the data in the adjacent scan lines. The shift of the substrate to the elevation Z = 0 can be accomplished by a combination of a facet and polynomial leveling of the surface after masking grains and other features. Gwyddion’s Flatten Base tool performs this task autonomously or with a user-specified mask. After the described background and line corrections, the electrostatically fixated vesicles can be identified on the substrate by executing Mark Grains function. 

Figure 11A and Figure 11B show height and phase images of hydrated MCF-7 exosomes immobilized on a mica surface and acquired in PBS using the tapping mode. A total of 561 hydrated vesicles were identified in the scanned area using Threshold algorithm of Mark Grains function with the threshold value set to ~20%. The phase lag of the probe’s response at the drive frequency is sensitive to localized stiffness variations in soft samples. The consistency between the height and phase images, seen in Figure 11A,B, is, therefore, an important confirmation that the imaged grains are, indeed, soft vesicles immobilized on the substrate.
 
Figure 11C shows the cross-section of the height image through exosomes located on the white line in Figure 11A. While the exosomes in a biofluid have a globular geometry^1,14,15,16, their shape on the substrate is severely distorted by the electrostatic attraction to the positively charged surface. The oblate pancake-like geometry of electrostatically immobilized vesicles is further illustrated in Figure 11D by the close-up height image (and its cross section) of an exosome boxed in Figure 11A. The corresponding phase image is shown in Figure 11E. The empirical probability density function (pdf) of peak heights above the surface for all 561 hydrated vesicles identified in the AFM scan is shown in Figure 12A. The mean value for this distribution is 7.9 nm, which is approximately equal to twice the thickness of a phospholipid bilayer¹⁷ in the absence of deforming forces.
 
The area on the substrate occupied by an immobilized exosome was approximated as a circle with the diameter equal to the mean distance from the vesicle’s “center of mass” to its boundary on the mica’s surface. The distribution of these projection diameters is shown in Figure 12A and has the mean equal to 69.6 nm. The obtained height and the diameter distributions further quantify the significant impact of the electrostatic surface immobilization on the distorted shape of immobilized exosomes.
 
The robustness and repeatability of the protocol procedures were confirmed by reanalyzing the same MCF-7 sample three times, from sample preparation to imaging, with each repeat producing results statistically similar to those shown in Figure 12.

The deformation of immobilized vesicles caused by electrostatic forces may be compensated or interpreted to provide an insight into the properties of the imaged EVs. For example, the AFM data may be used to estimate the globular size of the vesicles in the solution. As a starting point, we can calculate the volume encapsulated by the membrane envelopes of immobilized vesicles. The volume is found by integrating the difference between the surface level of the identified vesicles and the substrate elevation underneath them. The substrate level under the vesicles is not directly accessible but can be estimated by the Laplace or alternative interpolation of data points for unoccupied substrate surrounding the vesicles. Within Gwyddion, such volume calculation is performed using Distribution of Various Grain Characteristics function. The result exported from Gwyddion can then be mapped into the diameters of volume-equivalent spheres. 

The application of the described algorithm to the AFM data for 561 analyzed hydrated MCF-7 vesicles produced the distribution of the diameters of volume-equivalent spheres shown in Figure 12B. This distribution estimates the size of membrane vesicles in their innate globular form in a biofluid before their electrostatic fixation on the mica surface. The vesicle sizing obtained from the analysis of the AFM data was compared with the results of cryo-TEM imaging of the same sample and was found to be in close agreement³ (Figure 12B). The comparison of the hydrodynamic diameters measured by the NTA with the obtained vesicle sizes (Figure 1) indicates that the mobility of exosomes is much smaller than would be expected from the size of their vesicles determined from the AFM and cryo-TEM measurements. The difference between the hydrodynamic and vesicle sizes characterizes the thickness of the coronal layer surrounding exosomal vesicles. 

Figure 1: Comparison of hydrodynamic and geometric diameters of EVs. The geometric size of the exosomal vesicle is substantially smaller than its hydrodynamic size determined from its diffusion in a liquid. The difference is the coronal layer formed by membrane-conjugated and adsorbed molecules that impede the mobility of EVs. This figure is modified from reference³ and reprinted with permission. 

Figure 2: Properties of the AFM probe. The geometry and the dimensions of the AFM probe can be specified using Model Tip function. Please click here to view a larger version of this figure.

Figure 3: Correction of imaging artifact caused by tip-sample convolution. By performing Surface Reconstruction, the acquired AFM data can be corrected for tip artifacts.  Please click here to view a larger version of this figure.

Figure 4: Correction for a tilt in the substrate. Plane Level determines the plane of the substrate and subtracts it from the AFM data. Please click here to view a larger version of this figure.

Figure 5: Correction of misalignments in scan rows. A conventional algorithm to align the scan data is to find an average height along each scan line and subtracts the result from the corresponding row of data points in the image. Please click here to view a larger version of this figure.

Figure 6: Correction for the gaps in the aligned data. Common scanning errors, known as scars, can be removed from the AFM data by applying Remove Scars function.  Please click here to view a larger version of this figure.

Figure 7: Alignment of a substrate at zero elevation. Flatten Base option in Level menu allows the user to place the substrate surface at the base level corresponding to the zero height.  Please click here to view a larger version of this figure.

Figure 8: Identification of immobilized vesicles on the scanned surface. (A) The surface-immobilized exosomes are identified as grains protruding above the substrate by a user-selected height threshold specified in Mark by Threshold. (B) The outcome of the identification. Please click here to view a larger version of this figure.

Figure 9: Analysis of the AFM data. The distribution of maximum heights above the substrate within the area occupied by the identified exosomes is shown as compiled by Grain Distributions tool.  Please click here to view a larger version of this figure.

Figure 10: Impact of surface modification and EV concentration of the surface density of immobilized vesicles. (A) The AFM height image of freshly cleaved mica substrate after 12 h incubation with MCF-7 exosome sample followed by cleaning with DI water and drying. The immobilization of EVs from the liquid to the substrate is inefficient without imparting a positive charge to mica’s surface. Few particles seen in the scan are likely the result of incomplete removal of the MCF-7 sample before the substrate was dried. (B) The height scan of mica’s surface in the air after the treatment with nickel chloride shows the substrate free of contaminations. Panels (C) and (D) show AFM height scans obtained after the modification of the surface charge and the incubation with the same MCF-7 sample as in panel (A) for 24 h and 12 h, respectively. The surface concentration of immobilized vesicles is excessively dense after 24 h incubation. The 12 h incubation leads to fewer exosomes immobilized on the surface and the scan data that are easier to analyze accurately.  Please click here to view a larger version of this figure.

Figure 11: AFM images of hydrated MCF-7 exosomes electrostatically immobilized on the modified mica surface. (A) The height image. (B) The corresponding AFM phase image confirms that the grains in the height image are soft nanoparticles, as should be expected for membrane vesicles. (C) The height data for the three vesicles crossed by the line shown in panel (A) illustrate a flattened shape caused by the electrostatic attraction of exosomes to the positively charged surface of the modified mica. (D) The shape distortion is apparent in an enlarged view the immobilized vesicle boxed in panel (A) and its cross section. The phase image of the same vesicle is shown in (E). This figure is modified from reference³ and reprinted with permission. Please click here to view a larger version of this figure.

Figure 12: Dimensional characterization of hydrated vesicles immobilized on the surface and the estimation of their globular size in the solution. (A) The distribution of peak heights above the surface (red curve) has the mean equal to 7.9 nm. The area occupied by immobilized exosomes has 69.6 nm average diameter (blue curve). (B) AFM height image for one of the immobilized exosomes illustrates its highly oblate shape caused by electrostatic forces. The globular size of exosomal vesicles in the solution can be estimated by matching volumes enclosed by surface-immobilized and spherical membrane envelopes. (C) The size distribution of globular vesicles in the solution (red curve) was determined from the AFM data of 561 immobilized vesicles. The vesicle sizes in cryo-TEM images (blue curve) are consistent with the AFM results. This figure is modified from reference³ and reprinted with permission. Please click here to view a larger version of this figure.

Figure 13: Surface concentration and size segregation artifacts during passive deposition of EVs from evaporating liquid. (A) The scanning electron microscopy (SEM) image shows that the surface concentration of exosomes passively deposited from a drying liquid is spatially variable when surface immobilization from a suspending biofluid is not performed. (B) Passive deposition of EVs from a drying sample causes vesicles size segregation. The substantial size variability is quantified by the probability density functions (pdf) for the vesicles in different regions in the image (A) defined by white diagonal lines. This figure is modified from reference¹ and reprinted with permission. Please click here to view a larger version of this figure.

Figure 14: Cup-shaped geometry of desiccated vesicles passively deposited on the surface during the liquid evaporation. The surface desiccation of vesicles which were not immobilized by electrostatic forces is known to result in a cup-shaped appearance often observed in SEM images of EVs. This figure is modified from reference¹ and reprinted with permission. Please click here to view a larger version of this figure.

Discussion

The immobilization of EVs from a biological fluid, surface scanning, and image analysis are the essential steps of the developed protocol for the AFM characterization of EVs in liquid. The number of vesicles amenable to AFM imaging scales with the imaged surface area and the surface concentration of the vesicles immobilized on the substrate. Given a negative zeta potential of EVs and exosomes¹⁸, we advocate electrostatic fixation of EVs from liquid samples to the AFM substrate. The immobilization is effective when the surface is positively charged. Prior to EV immobilization, the positive surface charge may need to be imparted to the substrate, as in the case of mica ― a layered silicate mineral with general formula KAl₂(AlSi₃O₁₀)(OH)₂. Freshly cleaved mica’s surface is close to perfectly flat, which is ideal for imaging nanoparticles by the AFM, but its surface charge is negative and, thus, must be modified. The protocol describes the procedure to impart a positive surface change to the AFM substrate. The representative results show marked improvement in EV fixation from a biofluid to the modified mica substrate.
 
When imaging hydrated vesicles, it is important to minimize sample evaporation which causes the surface deposition artifacts and convective flows and increases the liquid concentration of the vesicles with time, leading to higher surface concentration of immobilized EVs than expected, especially during prolonged incubations. Probe holders explicitly designed for liquid samples eliminate or slow evaporation and should be used to image hydrated EVs. Nonspecific bindings to the scanning probe are reduced in the presence of ionic species. Therefore, when imaging hydrated EVs, it is preferable to cover the substrate with a buffered medium, such as PBS, instead of DI water.

Importance of surface immobilization 
The consistent and predictable immobilization of EVs on the modified substrate removes the primary source of variability in the AFM results. All downstream steps, from scanning to data analysis, are more easily controlled by the selection of instrumentation, probes, scanning parameters, and the data analysis sequence and algorithms. The user should be aware of upstream variability in biological samples and EV isolation protocols, which are important issues beyond the scope of this work.

We recommend performing surface immobilization of EV on the modified mica’s surface from liquid samples even when the goal is to characterize desiccated vesicles in the air — the need is less evident since vesicles will unavoidably deposit on any substrate as the liquid evaporates. In fact, the AFM results for desiccated EV obtained without modifying mica’s surface charges, which is a prerequisite step for electrostatic immobilization of EVs from a liquid, have been reported in the past^19,20,21. When EVs are not fixed to the surface from the liquid sample, however, their passive deposition by evaporation will produce artifacts collectively known as a coffee ring effect²². Two such artifacts, occurring as a drying liquid recedes, are illustrated in the SEM image (Figure 13A) of serum exosomes deposited by evaporation on a negatively charged glass surface. Significant variations in the surface concentration of precipitated vesicles are immediately apparent. The second artifact, quantified in Figure 13B, is the considerable variability in vesicle sizes in different areas within the perimeter of the dried sample. Given these artifacts, the AFM characterization of passively deposited vesicles from a drying liquid may produce biased or inconsistent results unless the entire surface area initially occupied by a now-dried liquid sample is scanned. 

Two additional issues should be considered when imaging the desiccated samples obtained without firm immobilization of vesicles on the substrate. Recall that our protocol instructs users to thoroughly wash the surface with DI water after the vesicles are immobilized from a liquid sample. This step intends to prevent ionic and other non-vesicular solutes from forming surface deposits during the evaporation of complex biofluids with considerable osmolarity. If EVs are not fixed, thorough washing will detach a large number of vesicles from the surface, potentially biasing the results and leaving too few particles for analysis. Another common difficulty, reduced by immobilizing EVs on the modified mica surface before the AFM imaging, is the adhesion of particles to the probe²³ and the misleading artifacts caused by this phenomenon.
 
Control of surface density of immobilized EVs
The two easily controllable factors identified in the protocol allow the user to customize the surface concentration of the immobilized EVs on the modified mica substrate: the concentration of the vesicles in the liquid sample and the time the sample is incubated on the substrate. A high density of immobilized vesicles, achieved with longer incubation times and higher concentration of EVs in the liquid, increases the number of vesicles analyzed during the scanning and the statistical power of conclusions arrived by the analysis of the AFM data. At the same time, an excessively dense surface concentration, as in the case shown in Figure 10C where particles tightly cover the entire surface with no intervening areas of the substrate, complicates the image analysis and the interpretation of results and may lead to scanning artifacts caused by the interaction between closely spaced particles.

Influence of electrostatic screening and hydrodynamic mobility of EVs
The transparent control over the surface concentration of immobilized EVs as a function of factors influencing it allows a user to customize experimental conditions to meet the specific needs of a study. When performing customization, it is important to recognize that the electrostatic surface immobilization is a transport-limited process influenced by the ionic strength of the biofluid. 

The concentration of ionic and positively charged species inversely impacts the Debye length over which the surface and vesicle charges are screened. Beyond this length, the electrostatic forces are negligible. The boundary layer of the substrate’s electrostatic attraction will be much smaller in ionically-rich PBS than in DI water. This difference implies that, after a short incubation corresponding to the time needed to deplete the layer of liquid where the electrostatic attractions are felt, a surface density of EVs immobilized from the suspension in DI water will be higher than from PBS suspension, assuming the concentration of EVs is the same in both liquids. Put differently, more vesicles must be immobilized to deplete a thicker attraction layer in DI water than in PBS under otherwise identical conditions. 

After the vesicles are depleted from the boundary layer, the immobilization becomes an entirely transport-limited process. In this regime, the rate of the deposition will not depend on the suspending medium (e.g., DI water or PBS) as long as the viscosity is the same and the transport is entirely diffusive. However, the transport of vesicles into the attraction boundary layer may not be entirely diffusive. For instance, if the sample in a sessile drop on the AFM substrate partially evaporates during the incubation, the fluid inside the drop will be subjected to the evaporation-driven flow, and the transport of the vesicles towards the substrate will have, both, diffusive and convective contributions. When the evaporation is not adequately controlled, the contribution of a convective transport will be considerable, and the rate of immobilization will be higher than expected. The impact of the convective transport will change with the thickness of the attraction layer, which itself depends on the ionic content of the liquid. Furthermore, the evaporation will enhance the vesicle immobilization on the substrate by concentrating EVs in the solution. At higher EV concentrations, the concentration gradient between the attraction layer and the adjacent liquid will increase, creating a larger thermodynamic driving force to the migration of vesicles towards the substrate.

Immobilized vesicles may represent a liquid sample with a bias. For the case when the rate of immobilization is limited by diffusion, vesicles with smaller hydrodynamic sizes, determined by the combination of the vesicle size and the thickness of the coronal layer surrounding it (Figure 1), are more likely to enter the attraction layer because of their higher mobility. Consequently, after the initial depletion period, the hydrodynamically small EVs will be overrepresented on the substrate compared to their contribution to the EV population in the liquid sample. Note that a smaller hydrodynamic size does not automatically point at EVs with smaller vesicle sizes because of the heterogeneity in the thickness of the coronal layer³. The biased representation is avoided with long incubations that deplete the entire population of EVs in the liquid by its immobilization on the substrate. When a user aims at immobilizing all EVs from the biofluid, to avoid excessively dense coverage of the surface with the immobilized vesicles, it may be necessary to reduce the EV concentration in the liquid below the range suggested in the protocol. 

Deformation of EVs on the substrate
Extracellular vesicles in their native hydrated state and after desiccation can be characterized by the AFM, as described in the protocol. The electrostatic forces²⁴ that immobilize EVs on the mica surface also distort their shape from globular geometry in which they exist in the solution. The impact of the desiccation on the size and morphology of immobilized EVs may be analyzed by rescanning the same surface area before and after the sample is allowed to dry. 

It is instructive to examine the impact of the sample preparation on the shape of the desiccated EVs. The electrostatically immobilized EVs maintain the highly oblate geometry after drying but are further flattened by the desiccation. The height above the surface of the desiccated vesicles becomes smaller than in Figure 12A, while their footprint area increases (data not shown). On the other hand, when vesicles are deposited passively during the liquid evaporation and without prior immobilization on the surface, they tend to attain a cup-shape geometry upon desiccation, as has long been observed in the SEM images and, more recently, in AFM scans. This cupped shape is now recognized as a sample preparation artifact²⁵ caused by non-uniformity in capillary forces during surface desiccation, as mechanistically explained in Figure 14¹.
 
Image analysis and interpretation of AFM data
The responses to electrostatic and capillary forces acting to distort the shape of EVs provide valuable information on structural and compositional properties of EVs. For example, a multidimensional set of biophysical characteristics, such as the deformed size and shape extracted from the AFM data, were recently used to demonstrate the feasibility to differentiate between exosomes secreted by different host cells⁵. The distortions can also be taken into account and compensated. For example, we showed how to use the AFM data to characterize the globular size of vesicles in the solution by estimating the diameters of spheres that encapsulate the same volume as immobilized exososmes³. 

Materials

Name	Company	Catalog Number	Comments
AFM/STM Controller	Bruker	Multimode Nanoscope V	This AFM controller supports imaging of biological samples in liquid and air.
AFM/STM metal specimen discs (10 mm)	TedPella	16207	Metal specimen disc on which a mica disk is attached by a double-sided tape or other means.
AFM/STM Mica discs (10 mm)	TedPella	50	Highest quality grade V1 mica, 0.21mm (0.0085”) thick. Interleaved, in packages of 10. Can be mounted on AFM/STM discs. Available in four diameters
AFM probe for imaging in the air	Bruker	TESP-V2	High quality etched silicon probes for tapping mode and non-contact mode for scanning in the air.
AFM probe for soft sample imaging in liquid	Bruker	MLCT	Soft silicon nitride cantilevers with silicon nitride tips, which are well-suited for liquid operation.  The range in force constants enables users to image extremely soft samples in contact mode as well as high load vs distance spectroscopy.
Double sided tape	Spectrum	360-77705	Used to fix the mica disk on the metal specimen disc.
ExoQuick-TC	System Biosciences	EXOTC50A-1	ExoQuick-TC is a proprietary polymer-based kit designed for exosome isolation from tissue culture media.
Glass probe AFM holder for imaging in liquid	Bruker	MTFML-V2	This glass probe holder is designed for scanning in fluid with the MultiMode AFM.  The holder can be used in peak force tapping mode, contact mode, tapping mode, and force modulation.  The probe is acoustically driven by a separate piezo oscillator for larger amplitude modulation.  The holder is supplied with two ports, required fittings, and accessories kit for adding and removing fluids.
Gwyddion	Czech Metrology Institute.	Version 2.52	Open Source software for visualization and analysis of data fields obtained by scanning probe microscopy techniques.
Lint-free blotting paper	GE Healthcare Whatman	Grade GB003 Blotting Paper	Use this blotting paper to remove NiCl2 after the modification of the mica's substrate.
Lint-free cleanroom wipes	Texwipe	AlphaWipe TX1004	Use these polyester wipes for surface cleaning.
Nickel(II) chloride (NiCl2)	Sigma-Aldrich	339350	Powder used to make 10 mM NiCl2 in DI water
Phosphate Buffered Saline (1x)	Gibco	10010023	PBS, pH 7.4