This protocol provides instructions for implementing multiphoton lithography to fabricate three-dimensional arrays of fluorescent fiducial markers embedded in poly(ethylene glycol)-based hydrogels for use as reference-free, traction force microscopy platforms. Using these instructions, measurement of 3D material strain and calculation of cellular tractions is simplified to promote high-throughput traction force measurements.
Quantifying cell-induced material deformation provides useful information concerning how cells sense and respond to the physical properties of their microenvironment. While many approaches exist for measuring cell-induced material strain, here we provide a methodology for monitoring strain with sub-micron resolution in a reference-free manner. Using a two-photon activated photolithographic patterning process, we demonstrate how to generate mechanically and bio-actively tunable synthetic substrates containing embedded arrays of fluorescent fiducial markers to easily measure three-dimensional (3D) material deformation profiles in response to surface tractions. Using these substrates, cell tension profiles can be mapped using a single 3D image stack of a cell of interest. Our goal with this methodology is to make traction force microscopy a more accessible and easier to implement tool for researchers studying cellular mechanotransduction processes, especially newcomers to the field.
Traction force microscopy (TFM) is the process of approximating cellular tractions using interpolated displacement fields of fiducial markers generated by an adherent and contractile cell. Using TFM, the influence of mechanical cues in the extracellular environment on important cellular processes such as proliferation, differentiation, and migration can be investigated1,2,3,4,5,6,7,8,9,10,11,12. Unfortunately, many existing approaches can be difficult to implement or require familiarity with highly specialized analytical and computational tools making TFM difficult for inexperienced researchers to use. We describe a methodology to generate a TFM platform that eliminates some of the difficulty in analysis while also providing high-throughput data acquisition.
Of the existing TFM approaches, the most commonly used for quantifying material strain involves incorporation of small fluorescent markers (typically nano- or micrometer-sized fluorescent beads) into a deformable hydrogel, such as polyacrylamide (PAA) or poly(ethylene glycol) diacrylate (PEGDA)13,14,15. These bead-based approaches provide the ability to densely cluster fiducial markers around a cell of interest to maximize displacement sampling. Unfortunately, the distribution of the beads throughout the hydrogel cannot be directly controlled so the spatial organization is random. This random placement leads to problems such as beads which are too close to each other to accurately resolve, or so spread that patches of the substrate yield low quality data. The inability to predict where fiducial markers lie in the absence of cells also creates a constraint that, for every collected set of cell traction data, an additional reference image of the underlying markers in a relaxed state must also be captured. The reference image is required so that displacement in the stressed image can be approximated as the difference between the stressed and unstressed images. To achieve a relaxed state, the cells being measured are either chemically relaxed or completely removed. This process often prevents acquisition of further experimental measurements, inhibits long-term cell studies, and limits throughput. A reference image also requires image registration techniques to accommodate for drift which may have occurred during experimentation, often leading to cumbersome manual matching of stress state images to reference images.
Other TFM methods deemed reference-free, implement some form of control over the distribution of fiducial markers, either by high resolution lithography, microcontact printing, or micromolding16,17,18,19,20. Reference-free TFM is achieved through the assumption that the relaxed state for each fiducial marker can be predicted based on how marker positions were prescribed during the fabrication process. These methods allow for complete capture of a cell’s tension state within a single image capture in which fiducial marker displacements are measured in comparison to an implied reference than can be inferred from the fiducial marker geometry. While consistency in marker placement is typically achieved using these platforms, they generally suffer from their own shortcomings relative to the widely used bead-based approaches including: 1) decreased traction resolution; 2) decreased accuracy of out-of-plane displacements (in some cases a complete inability to measure); and 3) decreased customizability of platform substrates and materials (e.g., ligand presentation, mechanical properties).
To address these shortcomings, we designed a new reference-free TFM platform. The platform utilizes multiphoton activated chemistry to crosslink a small volume of a fluorophore into specific 3D locations within the hydrogel that serve as fiducial markers to measure material strain. In this way, we have designed a platform that operates similarly to bead-based approaches but with the significant benefit that fiducial markers are organized into gridded arrays allowing for reference-free material strain tracking. This reference-free property affords many advantages. First and foremost, it allows for non-intrusive monitoring of cellular traction states (i.e., circumvents the need to relax or remove cells to acquire reference positions of displaced fiducial markers). This was our primary goal in designing this system, as we intended to incorporate other downstream analytical methods in tandem with TFM, which can be difficult with destructive end-point TFM approaches. Second, using an implied reference based on gridded arrays allows for near-complete automation of displacement analysis. The regularity of the arrays creates a predictable workflow where the occurrence of exceptional cases (i.e., sample cell data containing unanticipated artifacts such as suboptimal marker spacing or registration mismatches) can be maintained at a minimum. Third, forgoing the need to acquire a reference image provides the freedom to monitor many cells on a single sample over extended periods of time. This contrasts with traditional bead-based approaches, where, depending on the fidelity of the microscope’s automated stage movements, errors in positioning can accumulate and increase the difficulty of properly registering reference images to cell tension images. Overall, this platform facilitates higher throughput in collecting cellular tension data.
With this protocol, we hope to familiarize the readers with the two-photon, laser scanning lithography technique that we implemented to generate this reference-free TFM platform to measure in-plane and out-of-plane traction components generated by cells seeded on the surface. Not covered in this protocol is the synthesis of some of the monomeric components. In general, these reactions include nearly identical “one-pot” synthesis reaction schemes described previously21, and alternatives to these products can also be purchased. We also aim to familiarize readers with the software-based tools we generated to promote the use of commercially-available laser-scanning microscopes as 3D printing tools and to facilitate analysis of fiducial marker displacements.
1. Photopolymerizing a PEGDA base hydrogel
2. Creating patterning instructions
3. Fabricating fiducial marker arrays
4. Visualizing fiducial marker arrays
5. Performing TFM using photopatterned hydrogels
6. Analyzing the Images
Throughout the protocol, there are a number of checkpoints providing feedback to assess the quality of the patterning procedure. To provide some insight concerning how to assess progress at each of these checkpoints, we provide representative results of an actual experiment. The results highlight the application of this protocol performed on a photopatterned hydrogel prepared for TFM analysis of human umbilical vein endothelial cells (HUVECs). The results include raw image data as well as processed data outputs at each critical step.
The first checkpoint occurs at step 4, once the fiducial marker array has been photopatterned in the hydrogel. When collecting an image stack, the resulting images should display a regular array of patterned features (Figure 1A-C) that oscillate in intensity as a function of z-position within the image stack. If the patterned surface was not perfectly level, different regions of each image slice may seem out of phase with each other with respect to these oscillations; this is expected and should not affect analysis except in extreme cases.
The remaining checkpoints utilize outputs from each of the scripts run during analysis of a given COI. The tracking.m script provides several diagnostic images including a z-projection of fluorescent fiducial markers (Figure 2A) to assess pre-processing quality, a plot of detected centroids color coded as a function of z-position (Figure 2B) to assess object detection quality, and a plot of tracks representing detected marker centroids which have been linked into columns in the z-direction (Figure 2C) to assess object tracking quality.
For accurate 3D centroid detection necessary to calculate reference coordinates and displacements, fluorescent fiducial markers should resemble ellipsoidal shapes with intensity decreasing radially from the center. The disp3D.m script provides a plot of intensity as a function of z-position for each column of detected features in a given image stack (Figure 3A,B) to assess the quality of the fiducial marker intensity profiles. Both disp3D.m and dispShear.m together provide histograms of measured displacement noise in each of the Cartesian coordinate dimensions (Figure 4A-C) as well as interpolated heat maps of displacement (Figure 4E,F). Finally, the interpFinal3D_2.m provides heat maps of surface tractions calculated using outsourced code (Figure 5)31.
Figure 1: Typical patterning results.
(A) Fluorescent images of fiducial markers displaying intensity fluctuations through the Z-dimension from the hydrogel surface to 12.4 μm within the hydrogel. (B) A processed volumetric rendering reveals the ellipsoidal shape of the fiducial markers. (C) Sectioned profile views of the volumetric rendering display raw marker data. A,C: Scale bar =5 μm. Please click here to view a larger version of this figure.
Figure 2: Diagnostic outputs from the particle tracking algorithm.
The object detection and tracking software outputs several diagnostic images including (A) a Z-weighted projection of the fluorescent fiducial markers and (B) a scatter plot of marker positions with color coded Z-position. (C) An additional output displays a z-weighted projection, as in (A), with 2D detections from each frame linked into columns. (D) A closer look at (B) more clearly shows individual 2D detections color coded by z-position. A: Scale bar =20 μm. C: Scale bar =5 μm. Please click here to view a larger version of this figure.
Figure 3: Diagnostic outputs from the 3D displacement algorithm.
(A, B) Line plots of marker intensity as a function of frame position display oscillations between markers grouped in the vertical direction. These line plots allow for qualitative assessment of grid alignment with the imaging plane, as well as overall quality of the patterned fiducial markers. (A) Overlapping oscillations in marker intensity confirm acceptable alignment of the imaging plane with the patterned fiducial marker arrays. Including empty frames at the top of a Z-stack allows for automated calculation of a noise threshold in the processing software. (B) Poor overlap between oscillations is often indicative of poor alignment of the patterned grids with the imaging plane but may also suggest that the grid themselves are misaligned and may yield poor results. Please click here to view a larger version of this figure.
Figure 4: Diagnostic outputs from the 2D and 3D displacement algorithms.
(A–C) Histograms of measured displacements in regions considered ‘non-deformed’ (i.e., displacement noise) provide a quantitative assessment of the accuracy of calculated reference lines for approximating marker reference positions. Displacement fields provide a visual representation of both (D) in-plane (shear) and (E) out-of-plane (normal) measured displacements. D: Scale bar =20 µm. Please click here to view a larger version of this figure.
Figure 5: Diagnostic Outputs from the Traction Conversion Algorithm.
(A-C) Traction stresses can be calculated based on interpolated displacement fields to determine the (B) total traction, (C) shear traction, and (D) normal traction. A: Scale bar =20 µm. Please click here to view a larger version of this figure.
The goal of this protocol is to provide a workflow that alleviates much of the difficulty associated with the generation and analysis of TFM data. Once prepared, the photopatterned hydrogels are simple to use, requiring only knowledge of standard tissue culture practices and fluorescence microscopy. The reference-free aspect enables carefree navigation on cell-laden hydrogels and eliminates cumbersome image processing steps such as image registration between reference and deformed images. The resulting analysis is nearly completely automated and data from individual COIs can be analyzed from start to finish in less than 10 min.
The primary challenges in this protocol stem from the fabrication of the photopatterned hydrogel. Because most of the reagents used in the protocol are synthesized in-house, we expect that users with little experience using synthetic hydrogel materials may be somewhat deterred from attempting this protocol. That said, all of the PEGylated components used in this protocol are synthesized from nearly identical protocols using one-pot reactions and many of the components can be purchased from commercial vendors.
This protocol also requires some mastery over the use of laser-scanning microscopes. The laser-scanning parameters need to be optimized for every different microscope, as laser intensities, objective lenses, and other microscope components will vary, even between microscopes of the same model. A simple approach to determining the best microscope settings for photo-patterning is to perform a panel of scans with varied settings. Any setting that affects the total laser exposure received by the sample will affect the size and quality of the patterned markers. This includes: scan speed, pixel size, averaging number, laser power/intensity/fluence, magnification, and numerical aperture32,33. It is often easiest to adjust laser power and scan speed and so it is recommended to start with those two parameters. It should also be noted that the software used to generate regions files was designed specifically for the brand of microscopes we use and may need to be augmented to accommodate other microscope makes and models. With the settings provided in the protocol, the user should expect to generate ellipsoidal markers with 3D gaussian-like intensity profiles of full-width half-max dimensions of 0.84 ± 0.11 μm in XY and 3.73 ± 0.30 μm in Z. The object detection algorithm uses a center-of-mass of intensity to identify marker centroids and will perform best when markers present a definite gradient of intensity.
Properly rinsing the hydrogel and protecting it from light and from contaminants is critically important for cell culture. As mentioned in the protocol, some of the patterning components may crash out of solution and deposit on the surface of the hydrogel. If these components are not completely rinsed from the surface, cell adhesion may be unpredictably affected. If exposed to UV spectrum light, even in small doses, while the hydrogel is soaking in the patterning solution, the patterning solution will begin to polymerize and will yield poor results. Finally, airborne particulates or unfiltered particulates will complicate analysis, especially if they are fluorescent. Special care should be taken to prevent these contaminants from entering the sample at all steps in the protocol.
The analysis code predicts reference locations of deformed fiducial markers by predicting the original array from non-deformed markers in the same row or column as the deformed marker. While this greatly simplifies analysis, it also imposes some restrictions on what can or cannot be analyzed. At a minimum, each deformed marker being analyzed should have 2 or more members of both its column and row of markers which are not deformed so that an accurate prediction can be made. A row is defined by the group of dots printed in a single line scan on the microscope, and a column is defined by markers patterned on a single vertical axis using a z-stack function. This limits analysis to single cells or small cell clusters based on the length and depth of each patterned area. It is important to note that this is not remedied by placing two patterned regions perfectly next to each other. This is because although the XY components may line up, it is not feasible to perfectly line up the Z positional components of the arrays, unless the sample surface is perfectly level with both the focal plane and the stage. The best way to avoid this limitation is to employ additional rounds of patterning that specifically dictate adhesion ligand placement. If aligned properly, cell adhesion can be restricted to usable areas on the patterns.
The analysis code has been optimized for square arrays where markers are spaced approximately 3.5 μm apart in Z and 2.1 μm in X and Y. The final spacing will vary based on patterning fidelity and swelling, but this should not greatly impact performance if in-row spacing is consistent (inter-row spacing variability will not affect tracking). While the code may run to completion on differently prescribed arrays, its current state may yield poor results if the arrays do not match the parameters listed here. Our current goals for the analysis code are to include support for variable spacing as well as triangular arrays, which may improve traction reconstruction accuracy and reduce regional bias as compared to square arrays.
To conclude, we provide an approach to TFM that allows for facile collection and analysis of TFM data without perturbing cellular function. Our goal with this approach is to provide a more accessible and non-intrusive TFM approach, without compromising the resolution and quality of TFM data. We expect that this methodology will promote the convergence of biochemical knowledge of cellular phenotypes with the observed physical properties of cells.
The authors have nothing to disclose.
O. A. Banda was supported by funding from a NSF IGERT SBE2 fellowship (1144726), startup funds provided by the University of Delaware, and the National Institutes of Health/National Cancer Institute IMAT Program (R21CA214299). JHS is supported by funding from the National Institutes of Health/National Cancer Institute IMAT Program (R21CA214299) and the National Science Foundation CAREER Award Program (1751797). Microscopy access was supported by grants from the NIH-NIGMS (P20 GM103446), the NSF (IIA-1301765) and the State of Delaware. The structured illumination microscope was acquired with funds from the State of Delaware Federal Research and Development Grant Program (16A00471). The LSM880 confocal microscope used for two-photon laser scanning lithography was acquired with a shared instrumentation grant (S10 OD016361).
Acrodisc Syringe Filter, 0.2 μm Supor Membrane, Low Protein Binding | Pall | PN 4602 | Allows for filtering of macromer solutions prior to base gel synthesis and subsequent lithography steps. |
Acrylate-Silane Functionalized #1.5 Coverslips | in-house | in-house | Acrylates allow binding of base hydrogel to the glass surface to immobilize the hydrogels. See reference: 21-24 |
Axio-Observer Z1 w/Apotome | Zeiss | Widefield microscope with structured illumination module used to capture images for TFM. | |
Chameleon Vision ii | Coherent Inc. | Equipped on laser-scanning microscope used for multiphoton Lithography. | |
Double Coated Tape, 9500PC, 6.0 mil | 3M | Binds acrylate-silane functionalized coverslips to Petri dishes. | |
Flexmark90 PFW Liner | FLEXcon | FLX000620 | Allows lining of double coated tape enabling feeding of tape into plotter. |
LSM-880 | Zeiss | Laser-Scanning microscope used for Multiphoton Lithography. | |
MATLAB | Mathworks | R2018a | Runs custom scripts to generate lithography instructions for microscope and for analysis of TFM data. |
Model SC Plotter | USCutter | SC631E | Cuts double coated tape into rings to bind coverslips to petri dishes. |
Objective C-Apochromat 40x/1.20 W Corr M27 | Zeiss | Equipped on both widefield microscope and laser-scanning microscope to be used for both lithography and TFM. | |
PEG-AF633 | in-house | in-house | Fluorophore-labeled acrylate PEG variant for creating fiducial markers. See reference: 21 |
PEG-DA | in-house | in-house | Base material for hydrogels. See reference: 21 |
PEG-RGDS | in-house | in-house | RGDS peptide-labeled mono-acrylate PEG variant for promoting cell-adhesion. See reference: 21 |
Petri Dishes | CELLTREAT | 229638 | 8mm holes are cut into the center of each dish using a coring bit to fit base hydrogels. |
Sylgard 184 Silicone Elastomer Kit | Dow Corning | 3097358-1004 | For creating spacers to control base hydrogel thickness (aka PDMS). |
Syringe, Leur-Lok, 1 mL | BD | 309628 | Allows for filtering of macromer solutions prior to base gel synthesis and subsequent lithography steps. |
UV Lamp | UVP | Blak-Ray® B-100AP | Polymerizes base hydrogel. |
1-vinyl-2-pyrrolidinone (NVP) | Sigma-Aldrich | V3409-5G | Radical accelerant and co-monomer. Improves pegylated fluorophore incorporation during lithography. |