Abstract
Optical tweezers are an invaluable tool for noncontact trapping and micromanipulation, but their ability to facilitate highthroughput volumetric microrheology of biological samples for mechanobiology research is limited by the precise alignment associated with the excitation and detection of individual bead oscillations. In contrast, radiation pressure from a lownumerical aperture optical beam can apply transversely localized force over an extended depth range. Here we present photonic force optical coherence elastography (PFOCE), leveraging phasesensitive interferometric detection to track subnanometer oscillations of beads, embedded in viscoelastic hydrogels, induced by modulated radiation pressure. Since the displacements caused by ultralow radiationpressure force are typically obscured by absorptionmediated thermal effects, mechanical responses of the beads were isolated after independent measurement and decoupling of the photothermal response of the hydrogels. Volumetric imaging of bead mechanical responses in hydrogels with different agarose concentrations by PFOCE was consistent with bulk mechanical characterization of the hydrogels by shear rheometry.
Introduction
Optical manipulation has had a revolutionary impact in the biological and nanoscale sciences^{1,2,3}. Developed by Ashkin et al.^{4}, optical tweezers (OTs) have enabled the manipulation of biological systems at the moleculartocellular scale. This has led to many seminal studies, including measurement of the elastic properties of bacterial flagella^{5}, direct observation of the movement and forces generated by molecular motors^{6,7}, the study of mechanotransduction pathways in living cells^{8}, and measurement of the mechanical properties and biophysical interactions of DNA^{9,10}. OTs utilize the gradient force of a highnumerical aperture (NA) laser beam to achieve trapping and manipulation of micrometersized particles near the beam focus. Typical forces that can be achieved range from femtonewtons to hundreds of piconewtons, which covers the range of forces associated with the biophysical processes of life at the moleculartocellular scale.
More than a decade before the invention of OTs, Ashkin^{11} demonstrated that lowNA laser beams could exert sufficient radiationpressure forces to accelerate micrometerscale dielectric particles along the beam path, or could be used to form a dualbeam trap based on counterpropagating beams. Guck et al.^{13} adopted Ashkin’s original dualbeam trapping configuration to develop the optical stretcher for the study of cell mechanics^{12}, and used it to study the mechanical properties of cancer cells and human blood cells^{14}. Radiation pressure has been utilized on other microfluidic platforms for nanoparticle sorting and chromatography^{15,16}. Compared to OTs, however, optical manipulation based on radiation pressure has led to fewer applications in the life sciences.
The rapidly growing field of mechanobiology^{17,18} has created new opportunities for application of optical manipulation across the micrometertomillimeter scale. Over the last decade, mechanobiology research has uncovered the integral role that extracellular matrix (ECM) mechanics and biophysical cell–ECM interactions play in biological processes, including the onset and progression of cancer^{19,20,21,22}, stem cell differentiation^{23,24,25}, morphogenesis^{26}, and wound healing^{25}. Biophysical interactions play an integral role across all spatial scales, from molecular processes at the nanoscale^{6,7,8,9,10}, to collective (emergent) behavior at the micro to mesoscale^{27,28,29}. At the micro to mesoscale, cellular behavior is known to be different in threedimensional (3D) versus twodimensional (2D) environments^{30,31}. Consequently, there is an important trend toward the adoption of 3D cell culture systems^{22,30,31} that is driving the need for new imaging approaches that can support volumetric imaging of microscale spatial variations in ECM mechanics across time. Atomic force microscopy has been the preferred method in mechanobiology, but it can only interrogate the sample surface. Laser tweezerbased active microrheology^{32,33} (AMR) is a promising approach that utilizes OTs to induce and detect nanometerscale displacements from microbeads randomly distributed within 3D cell culture. However, its volumetric throughput is limited by the need for precise 3D alignment of highNA trapping and position detection beams to each probing bead, prior to actuation with a transversely oscillating optical trap. Although simultaneous trapping and detection of multiple beads can be achieved with timeshared OTs^{34} or holographic OTs^{35,36}, multiplexed manipulation of beads randomly distributed over a depth range of a hundred micrometers or greater with OTs has not been demonstrated. Other emerging approaches include Brillouin microscopy^{37,38} and optical coherence elastography^{39,40,41} (OCE). However, none of the above methods have demonstrated the ability to support mechanical microscopy with both 3D cellularlevel resolution and sufficient throughput to facilitate timelapse imaging studies over millimeterscale volumes in mechanobiology.
In order to address the unmet need for volumetric, timelapse mechanical microscopy of engineered systems in mechanobiology research, we revisited Ashkin’s original idea of using radiation pressure from a lowNA beam^{11}—now as a potential mechanism to apply localized mechanical excitation to microbeads embedded in aqueous biological media. The use of lowNA radiation pressure to apply transversely localized axial force over an extended depth range can be advantageous for volumetric data acquisition, but comes with the challenges that a lowNA beam exerts significantly lower force than highNA OTs, and that historically, studies on isolating the effects of photonic radiation pressure have been hindered by accompanying photothermal responses^{11,42,43}. OCT has been used previously to monitor the radiationpressureinduced trajectories of beads in liquid media^{44}; but this method has yet to be applied to solid viscoelastic materials.
In this paper, we present photonic force OCE (PFOCE) as a technique for 3D mechanical microscopy, leveraging the interferometric displacement sensitivity of OCT to detect picometertonanometer bead oscillations induced by modulated radiation pressure from a lowNA beam. We address the challenge of isolating radiationpressure effects within absorbing aqueous media via a linear model to decouple mechanical and photothermal responses, combined with a differential scattering approach. PFOCE has the potential to provide a new platform for largescale volumetric mechanical microscopy—probing micrometerscale spatial variations in the mechanical responses of the medium, with a spatial sampling that is statistically controlled by the distribution of the beads inside the medium. Such a capability may readily find applications in cell mechanics and mechanobiology research, for instance, by enabling the mapping of spatiotemporal variations in ECM mechanics for 3D traction force microscopy (TFM)^{27,29,31,45}.
Results
Principles of PFOCE
Based on mechanical excitation via harmonically modulated radiationpressure force from a lowNA beam (hereafter referred to as the PF forcing beam), PFOCE measures the resulting oscillations of beads embedded in viscoelastic media induced by the PF forcing beam (the mechanical response) by compensating for the accompanying absorptionmediated photothermal effects of the aqueous medium (the photothermal response).
We begin with the theoretical basis for using harmonically modulated radiation pressure to induce oscillations of beads embedded in a viscoelastic medium. Generalizing Ashkin’s simplified expression^{11} to include the contribution of both photon scattering and absorption to the net change in linear momentum, the axial radiationpressure force, F_{rad}, exerted on a neutral particle by a weakly focused beam with optical power P, is given by:
where c denotes the speed of light in vacuum, n_{med} denotes refractive index of the medium, and \({\hat{\mathbf z}}\) denotes a unit vector pointing in the propagation direction of the forcing beam. The proportionality constants q_{a} and q_{s} define the fractions of incident photon momentum that are imparted to the bead in the direction \({\hat{\mathbf z}}\) as a result of absorption and scattering, respectively. For a nonabsorbing particle, such as the latex (polystyrene) beads used in Ashkin’s experiments, the contribution of q_{a} to the radiationpressure force is neglected. The proportionality constant q_{s} accounts for the effects of shape, size, and scattering cross section of the bead in relation to the characteristics of the forcing beam (for example, wavelength and beam waist radius or NA), the refractive indices of the bead and medium, and the position of the bead in 3D space relative to the forcing beam.
Together with the restoring force from the viscoelastic medium, harmonically modulated F_{rad} induces oscillatory motion of the bead. Oestreicher provided a theoretical model for the impedance of an oscillating sphere in a linear viscoelastic medium^{46}. We inverted equation (18) in Oestreicher’s paper^{46} to obtain an expression for the oscillation amplitude of the bead as a function of bead radius, a, complex shear modulus, \(G^\ast\)(ω) = G′(ω) + iG″(ω), and mass density, ρ, of the medium, given by
where
In Eq. (2), u_{0} describes the oscillation amplitude of the bead resulting from harmonically modulated axial radiationpressure force with peak magnitude \(F_{{\mathrm{rad}}} = \left\ {{\mathbf{F}}_{{\mathrm{rad}}}} \right\\) and modulation (angular) frequency ω. We observe that u_{0} is directly proportional to the magnitude of axial radiationpressure force but inversely proportional to higherorder powers of the bead radius.
For nonabsorbing beads embedded in an aqueous medium, the scatteringmediated radiation pressure exerted on the bead from the PF forcing beam is accompanied by the absorptionmediated photothermal response of the medium. Absorptionmediated responses form the basis of multiple functional imaging modalities. Photoacoustic tomography uses short laser pulses to generate ultrasonic pressure waves from absorptioninduced thermoelastic expansion^{47}. Highpower laser pulses have also been used to generate propagating surface acoustic waves caused by thermal expansion for elastography applications^{48}. Absorptioninduced optical path length (OPL) change, governed by the thermooptic effect and thermal expansion, allows photothermal OCT (PTOCT) to detect the presence of chromophores in biological samples^{49,50}.
In order to understand and account for the effects of absorption on OPL, consider the case of a nonabsorbing dielectric bead embedded at depth L in a homogeneous absorbing medium with uniform (spatially invariant) refractive index n_{med}. The OPL to the bead measured by OCT is encoded in the phase of the complex OCT signal, given by Φ = (4π/λ) ⋅ n_{med}L. The OPL to the bead corresponds to the product OPL = n_{med}L. Both n_{med} and L can vary with a change in temperature via two different phenomena—n_{med} via the thermooptic effect and L via thermal expansion. The OPL change with respect to the change in temperature can be expressed via the product rule of differentiation as,
The derivatives dn_{med}/dT and dL/dT are the thermooptic coefficient and the thermal expansion coefficient of the medium, respectively. Equation (4) describes how the measured OPL change due to the change in temperature of the medium originates from both the change in refractive index and thermal expansion, and does not directly correspond to physical displacement of the bead.
Assuming that scattering and absorption are independent events within the context of PFOCE, we model the measured bead OPL oscillation (the total response), ΔOPL_{tot}, due to the modulated PF forcing beam as a linear combination of the complex mechanical response of the bead, ΔOPL_{mech}, and the complex photothermal response of the medium, ΔOPL_{PT}, given by
where ΔOPL_{mech} and ΔOPL_{PT} are given by
and
The vector r = (x, y, z) denotes the spatial coordinates of each pixel in the OCT image and φ_{drive} denotes the phase of the PF forcing beam drive waveform at time t = 0. A_{mech} and φ_{mech} denote the amplitude and phase of the complex mechanical response, respectively. Likewise, A_{PT} and φ_{PT} denote the amplitude and phase of the complex photothermal response, respectively. The goal of PFOCE is to isolate the complex mechanical response of the bead, which is dependent on the mechanical properties of the surrounding viscoelastic medium, from the measured total response by subtracting the accompanying photothermal response of the medium (Fig. 1a).
Theoretical simulations
In order to understand the effects of each design parameter (for example, NA of the PF forcing beam and bead size) on the measured response, and to obtain an estimate of the expected magnitude of the mechanical and photothermal responses, we simulated the contributions of both radiationpressure force and photothermal response to the total OPL oscillation of a polystyrene bead. Unless stated otherwise, all numerical results presented in this section were obtained from theoretical simulation assuming a Gaussian PF forcing beam with wavelength λ = 976 nm and waist radius w_{0} = 3.19 μm and a spherical bead with refractive index n_{bead} = 1.58. Refractive index of the medium was assumed to be n_{med} = 1.34 for biological hydrogels.
An accurate estimate of the scattered (and absorbed) photon energy is critical to predict the magnitude of radiationpressure force on a bead. Several approaches have been used to estimate q_{s} and q_{a} in Eq. (1). Among them, Generalized LorenzMie Theory (GLMT)^{51} is applicable for estimating F_{rad} from a focused Gaussian beam on a bead of arbitrary shape and size. Our MATLAB implementation of GLMT showed that the normalized axial radiationpressure force (that is, F_{rad} per unit power), \(\bar F_{{\mathrm{rad}}}\), from the PF forcing beam on a bead was on the order of 0.2–0.3 pN mW^{−1} at the focal plane when the focal spot size of the beam was comparable to the bead diameter (Fig. 1b). The force was lower as the focal spot size deviated from the bead diameter. Moreover, \(\bar F_{{\mathrm{rad}}}\) increased monotonically with both beam waist and bead diameter when the two dimensions were comparable, whereas \(\bar u_0\) decreased beyond a certain point (Fig. 1c) because the oscillation amplitude is both directly proportional to \(\bar F_{{\mathrm{rad}}}\) and inversely proportional to a (Eqs. (2) and (3)).
In order to simulate the photothermal response, we solved the heat transfer equation to estimate the change in temperature of the medium due to absorption and then modified a theoretical model given for PTOCT by Lapierre et al.^{50} to estimate the resulting cumulative OPL change for a general case with spatially varying n_{med} and T. Detailed descriptions of the theoretical model and parameters used for the simulation can be found in the Supplementary Methods and Supplementary Table 1. In an aqueous medium, the normalized change in temperature per unit power, \(\overline {\Delta T}\), due to absorption by water molecules was simulated to be on the order of 10^{−2} K mW^{−1} after 50 ms of continuous exposure to the PF forcing beam (Fig. 1d). The resulting normalized cumulative OPL change, \(\overline {\Delta {\mathrm{OPL}}}\), due to photothermal effects was on the order of 0.5 nm mW^{−1} (Fig. 1e), approximately an order of magnitude larger than \(\bar u_0\) measured for the same beam parameters (λ = 976 nm, w_{0} = 3.19 μm, and n_{med} = 1.34; Fig. 1c, e).
These simulations assume that the OPL change induced by the photothermal effects of a PF forcing beam on the medium is unaffected by the presence or size of the beads. On the other hand, both the radiationpressure force and the resulting bead oscillation amplitude for a given medium and PF forcing beam can vary by an order of magnitude depending on the size of the bead alone. This provides a guide for designing experimental conditions that affect the mechanical response without disturbing the photothermal response, forming the basis for the isolation of bead mechanical response from the measured total response (see Isolation of bead mechanical response).
Experimental setup and data acquisition
To enable simultaneous mechanical excitation by radiation pressure from the PF forcing beam and detection of resulting OPL oscillations by phasesensitive OCT, we combined the PF forcing beam (λ = 976 nm) with the sample arm beam of a spectraldomain (SD)OCT system via a freespace beam control module (BCM) and a dichroic filter (Fig. 2a). In this configuration, both the OCT and the PF forcing beams were collinearly scanned in a raster pattern by the same galvanometer and focused by the objective lens to the same position in 3D space. The waist radius of the PF forcing beam was measured to be 3.19 µm (NA = 0.1) at the focal plane (Supplementary Fig. 1 and Supplementary Methods). Although the theoretical simulation suggests that a PF forcing beam with smaller waist radius (when paired with comparable bead size) is optimal for maximizing the bead oscillation amplitude (Fig. 1b), we chose not to increase the NA of the PF forcing beam beyond 0.1 to ensure that radiationpressure force would be applied over an extended depth range for highthroughput volumetric measurement.
In this paper, we used agarose hydrogels of different concentrations confined in a glass chamber as examples of viscoelastic substrates used in cell imaging applications (Fig. 2a). Polystyrene beads with mean diameter of 3 µm were added to the hydrogels to serve as scattering particles for oscillation by radiation pressure. With a typical shear modulus in the range of 0.1–1 kPa^{52}, we expected \(\bar u_0\) to be approximately 0.03 nm mW^{−1}. Agarose hydrogels of uniform concentrations were used to validate PFOCE with shear rheometry, and a sidebyside sample of two agarose concentrations on either side of the sample chamber was used to demonstrate the volumetric capabilities of PFOCE. To enable 3D volumetric measurements of OPL oscillations of the 3 µm beads, we adopted a 3D BMmode acquisition scheme, wherein the beams were repeatedly raster scanned to acquire multiple Bscans at each slowaxis location (Fig. 2b, c). This scanning configuration allows continuous beamscanning acquisition along the fast axis (Bscan) while also supporting OPL tracking at each spatial location over time (Mscan) (refer to Methods for further details on the acquisition scheme).
The BMmode acquisition scheme has two key implications for the implementation of PFOCE in biological systems typically used for livecellimaging studies. First, the fastaxis beamscanning effectively resulted in a pulsetrain mechanical excitation on each of the 3µm beads instead of a continuous sinusoidal waveform provided by the function generator (Supplementary Fig. 2). As a Bscan was acquired, the PF forcing beam would dwell on each of the 3μm beads for approximately 67 μs (corresponding to four Ascans), after which the bead received no force until the PF forcing beam scanned over the bead again in the next frame. In other words, the actual excitation on each bead is a frequency comb with a 20Hz fundamental frequency and additional higherorder harmonics. This implies that the frequencydependent response of the medium must be accounted for when quantitatively reconstructing absolute mechanical properties of the medium from the bead mechanical responses; this subject will be addressed in a future manuscript. Under this type of excitation, the timeaveraged optical power imparted on each 3μm bead by the PF forcing beam was only 0.3 mW for a peak power of 112 mW, as opposed to 56 mW that would have been expected from a continuous sinusoidal excitation with the same peak power. Although this outcome is expected to result in a lower bead oscillation amplitude compared to the continuous excitation case, the two orders of magnitude reduction in the timeaveraged optical power imparted on the sample is beneficial for biological studies where cell viability is a concern.
Second, there is a tradeoff between acquisition speed and OPL oscillation measurement sensitivity. In the shotnoise limit, the sensitivity of OPL oscillation amplitude measurement by OCT is approximately inversely proportional to the squareroot of the number of BMmode frames acquired per slowaxis position^{53,54}. Prioritizing sensitivity over speed, we acquired up to 6144 frames per slowaxis position at a rate of 200 frames per second and achieved an OPL oscillation amplitude noise floor of 105 pm, approximately 50 pm above the theoretical shotnoise limit (Supplementary Methods), for OCT signals with signaltonoise ratio (SNR) of 25 dB (Supplementary Fig. 3). This acquisition scheme requires at least 5 min to acquire OPL oscillation data over a transverse field of view (FOV) of 200 μm × 10 μm at a spatial sampling of 1 µm per pixel; this spatial sampling density ensures each 3µm bead was sampled multiple times along the fast and slow axes. For typical volumes used in biological studies such as the sidebyside sample presented in this paper with a FOV of 200 μm × 125 μm, acquiring 3072 frames at 490 frames per second offers a 5× reduction in the acquisition time to 13 min, at the cost of increasing the OPL oscillation amplitude noise floor to 180 pm. Alternatively, the acquisition time can be shortened without sacrificing the OPL oscillation amplitude sensitivity by increasing the frame rate while maintaining spatial sampling over a large FOV; this can be achieved with a resonant scanner or other highspeed beamscanning options^{55}.
Isolation of bead mechanical response
Based on the linear model in Eqs. (5)–(7), the mechanical response of the 3µm beads can be isolated from the measured total response by subtracting the photothermal response of the medium. This approach requires availability of a reliable estimate of the photothermal response. In principle, the complex photothermal response in a uniformly absorbing medium may be theoretically obtained from the model of the absorptionmediated OPL change in PTOCT^{50}. However, we were not able to ascertain the accuracy of the theoretical simulation (Supplementary Methods) under our experimental conditions due to the lack of available material properties for the agarose hydrogels used in our experiments as well as the added contribution of the confined glass chamber in the experimental setup to the heat transfer process and thermal expansion model (Supplementary Discussions). Alternatively, under the premise that weak scattering signals from the medium would produce adequate OCT signals for OPL measurements but would not be sufficient to produce detectable mechanical response induced by scatteringmediated radiationpressure force, a differential scattering approach could be employed wherein the complex photothermal response may be measured experimentally from weak scatterers in the sample. These weak scatterers, thus, act as reporters of the photothermal response, without producing measurable displacements resulting from photon momentum transfer.
In implementing this differential scattering approach, we leveraged the sizedependence of backscattering intensity^{3} and added 0.1µm polystyrene beads to the sample to provide weak background scattering signals for measuring the photothermal response. In the Supplementary Methods, we estimate that \(\bar u_0\) for the 0.1µm beads is expected to be four orders of magnitude smaller than \(\bar u_0\) for the 3μm beads, based on their differences in OCT scattering intensity and theoretical predictions.
The process to isolate the mechanical responses of the 3µm beads is summarized in a flow chart (Fig. 3). In order to obtain a reliable estimate of the photothermal response, it was important to account for the consequences of the low SNR of OCT signals^{53,54} from the weakly scattering 0.1μm beads. The low OCT SNR (<12 dB) of the 0.1µm beads resulted in relatively large OCT phase noise on the measured OPL oscillations (that is, the raw photothermal response). To reduce the contribution of individual OCT phase measurement errors, we calculated the depthdependent photothermal response (with amplitude A_{PT} (z) and phase φ_{PT} (z)) by performing curvefitting of the amplitude and phase of the raw photothermal response as a function of depth (Fig. 3), using the theoretical simulation of the absorptionmediated OPL change^{50} for the amplitude and a cubic polynomial function for the phase. For small FOVs, we assumed that water was the only absorber in the sample and that the water concentration was transversely uniform across the sample, resulting in a photothermal response that is dependent on depth alone. This combined experimental and theoretical approach yielded estimates of the depthdependent photothermal response from the fitted curves with uncertainties of approximately ±0.7 nm for A_{PT} (z) and ±0.5 rad for φ_{PT} (z) (Supplementary Figs 4 and 5). For larger volumes, such as the sidebyside sample presented here, this uncertainty can be maintained or reduced by performing depthdependent photothermal curvefitting on smaller localized subvolumes. The mechanical responses of the 3µm beads, ΔOPL_{mech}, were subsequently isolated from the measured total responses, ΔOPL_{tot}, by subtracting the depthdependent photothermal responses (values obtained from the bestfit curves at the depths of the 3µm beads) from the total responses.
This differential scattering approach relies on four fundamental assumptions. First, we assumed that absorption events and scattering events that occurred in the sample were independent and separable. This assumption must hold for the linear model in Eqs. (5)–(7) to be valid. Second, we assumed that the mechanical response of the 0.1μm beads was negligible. This assumption was supported by both theoretical predictions and experimental observation of F_{rad}^{44}, and experimental observation of OCT scattering intensity (Supplementary Methods). Third, we assumed that the photothermal response in our singleconcentration samples was transversely uniform within the volume considered for photothermal fitting. Under this assumption, the uncertainty from the curve fits may impose a depthdependent systematic error that affects the accuracy of the isolated amplitude and phase of the mechanical responses relative to their true values, but does not degrade the precision for distinguishing microscale variations in the mechanical responses within a sample. However, if the photothermal response was to have a transverse variation that was unaccounted for, these uncertainties could also impose a random error that adversely affects the ability of PFOCE to distinguish relative differences in the mechanical responses within a sample. Lastly, we assumed the photothermal response measured on the 3μm beads was equivalent to the photothermal response measured on the 0.1μm beads at the same depth. In other words, we assumed that any perturbations of the photothermal response specifically due to the presence of the 3μm beads were negligible.
Comparison to shear rheometry
The mechanical responses of the 3µm beads were measured in four agarose hydrogel samples with different mechanical properties and compared to the bulk characterization of the hydrogels by shear rheometry (Supplementary Methods). The complex total responses measured on the 3µm beads, the fitted photothermal responses at the corresponding depths, and the isolated mechanical responses are displayed on the complex plane (Fig. 4a). We qualitatively observed that ΔOPL_{tot} and ΔOPL_{PT} had comparable magnitudes, and were distributed over the same quadrant of the complex plane for all agarose concentrations. In contrast, ΔOPL_{mech} was approximately four times smaller in magnitude and was in a different quadrant. These observations suggest that the measured total responses were dominated by contributions from photothermal effects, which can be explained by absorption of water molecules in the hydrogel samples at the PF forcing beam wavelength^{56}. The fact that ΔOPL_{PT} and ΔOPL_{mech} are in different quadrants of the complex plane implies that the photothermal responses and the mechanical responses do not occur inphase. Additionally, ΔOPL_{mech} followed a general trend of decreasing magnitude with increasing agarose concentration (Fig. 4a), consistent with our expectation that a higher agarose concentration would produce a stiffer hydrogel. This trend is reflected in Fig. 4b, where the median amplitude of the mechanical response (calculated from a set of spatial pixels corresponding to each 3µm bead as described in Methods) is overlaid on top of the OCT image of 0.2 and 0.4% agarose samples. Similarly, φ_{mech} of 0.4% agarose was closer to π than 0.2% agarose (Fig. 4c). In addition to mechanical contrast between samples, these maps also revealed the variability in A_{mech} and φ_{mech} of different beads within each sample, which may reflect microscale heterogeneity in the structural and mechanical properties of lowconcentration agarose hydrogels^{57,58,59}. For instance, beads with higher A_{mech} within a sample could be those inside larger pores, diffusing in the fluid phase of the biphasic porous hydrogel, whereas those with lower A_{mech} could be trapped in the solid agarose polymer matrix^{57,58,59}.
For quantitative comparisons to standard shear rheometry, A_{mech}, A_{PT}, φ_{mech}, and φ_{PT} measured by PFOCE are displayed as box plots next to the magnitude of complex shear modulus, \(G^\ast \), and phase delay, \(\varphi _{{\mathrm{rhe}}} = \tan ^{  1}(G{\prime\prime} {\mathrm{/}}G{\prime} ),\) measured by shear rheometry (Fig. 5). The total response is omitted here but can be found in Supplementary Fig. 6. Complete results from shear rheometry can be found in Supplementary Fig. 7. The mechanical response of the three imaging locations for each concentration of hydrogel can be found in Supplementary Fig. 8. We found a statistically significant (p_{C} < 0.05, see Methods for details) monotonically decreasing trend in A_{mech} versus agarose concentration (Fig. 5a). This behavior agrees with progressively increasing \(G^\ast \) of the hydrogels as the agarose concentration increased (Fig. 5c). In contrast, no significant trend (p_{C} = 0.15) or difference across agarose concentrations was observed for A_{PT} (Fig. 5b). We also found a similar increasing trend (p_{C} < 0.05) in both φ_{mech} and φ_{rhe} toward π as the agarose concentration increased (Fig. 5d, f). The phase delay approaching π is consistent with the response of a predominantly elastic material excited above its damped natural frequency. In contrast, φ_{PT} followed an opposite trend (p_{C} < 0.05) and was closer to 0 for all hydrogels (Fig. 5e). We note that φ_{mech} was up to π/4 rad smaller than φ_{rhe} for all concentrations. The discrepancies may reflect the differences between bulk responses of the hydrogels measured by shear rheometry and microscale mechanical responses measured by PFOCE. Nevertheless, both rheometry and PFOCE suggest that the hydrogels become more elastic as the agarose concentration increases, which is consistent with previous studies that reported decreases in porosity and influence of viscous drag (due to fluid flow through pores) at higher agarose concentrations^{58,59}. Our results demonstrate that the isolated complex mechanical responses of the 3µm beads from PFOCE can be used to distinguish different viscoelastic properties of agarose hydrogels, whereas no significant trend that directly correlates to the mechanical properties was observed in the photothermal responses.
Volumetric mechanical microscopy of agarose hydrogels
To demonstrate the volumetric imaging capabilities of PFOCE, the mechanical response of the beads embedded in the sidebyside sample (0.2% agarose hydrogel on one side and 1% agarose hydrogel on the other) was measured. The sample was aligned such that the boundary was roughly located at the center of the fastaxis (xaxis) scanning field, with 0.2% agarose hydrogel on the left half and the 1% agarose hydrogel on the right. Figure 6a shows the 3D distribution of A_{mech} (x_{b}, y_{b}, z_{b}), which reveals a clear contrast between the two halves of the sample. This is particularly pronounced for the 3μm beads between 0 μm < z < 75 μm, where the beads on the left (of the xaxis) mostly have larger A_{mech} (x_{b}, y_{b}, z_{b}) (shaded yellow to red) compared to the those on the right (shaded blue; Fig. 6b). Notably, a sharp boundary is also apparent in the en face projection of the photothermal responses at these depths due to the drastic increase in the number of the lowintensity photothermal reporters on the right side (Fig. 6c). This can be attributed to the intrinsic scattering (in addition to the exogenous 0.1µm beads added) in the 1% agarose hydrogel, which is present to a significantly higher extent than in the lowerconcentration 0.2% counterpart. Thus, the contrast in the bead mechanical responses in Fig. 6b is consistent with the sharp boundary between the two agarose concentrations revealed by the intrinsic background scattering in Fig. 6c. In contrast, the concentration of lowintensity photothermal reporters appears more uniform on each of the sides for depths −75 μm < z < 0 μm (Fig. 6c). The spatial variations of A_{mech} (x_{b}, y_{b}, z_{b}) at these depths similarly show a more gradual decrease from the softer 0.2% agarose hydrogel to the stiffer 1% agarose hydrogel (Fig. 6b). We speculate that this could indicate a larger extent of interaction (water exchange^{57,58,59}) between the two concentrations at these depths. Overall, we observe that A_{mech} (x_{b}, y_{b}, z_{b}) of the beads embedded in the 0.2% hydrogel is approximately 2 nm larger than those in the 1% hydrogel on average, confirming that the 0.2% hydrogel is softer (Fig. 6a, e). Additionally, the photothermal response of the 0.1µm beads appears transversely homogenous and increases with depth in accordance with the theoretical trends (Fig. 6d, f).
In addition to the expected contrast between the two sides, the 3D distribution of 3μm beads in Fig. 6a also highlights the local variations in A_{mech} (x_{b}, y_{b}, z_{b}) similar to those in the singleconcentration datasets (Fig. 4b, c), which could arise due to fluid transfer between the two sides, or the heterogeneous nature of hydrogels. Comprehensive discussions of sources of variability in PFOCE measurements can be found in Supplementary Discussions. We note that A_{mech} (x_{b}, y_{b}, z_{b}) in the 0.2% agarose hydrogel in this dataset is higher than in the singleconcentration dataset (Figs. 4b and 5a). This could be attributed to improvements made to the beam profile of the PF forcing beam as the BCM was readjusted at the beginning of the sidebyside experiment. Finally, with the current acquisition settings, the 3µm beads (N = 220 beads in total) analyzed in this 3D volume were acquired in 13 min resulting in an average acquisition time of 3.55 s per bead, which represents a greater than 2× speed improvement compared to stateoftheart OTAMR techniques^{33}.
Discussion
The use of radiation pressure to induce bead oscillations has two key implications for the basis of PFOCE. First, the ultralow radiationpressure force from a lowNA beam results in picometertonanometer bead oscillation amplitudes, which pushes the limit of interferometric axial displacement sensitivity of phasesensitive OCT. Our results demonstrate an oscillation amplitude measurement sensitivity of ≤105 pm (for OCT SNR ≥25 dB), which, in practice, corresponds to the smallest detectable oscillation amplitude of 150 pm for a detection threshold of 3 dB. The current sensitivity was achieved after optimization of hardware synchronization and acquisition scheme to minimize galvanometer motion instability.
Second, the use of radiation pressure from a lowNA beam in aqueous media is accompanied by absorptionmediated thermal effects, which can produce responses that may be an order of magnitude larger than the mechanical response induced by radiationpressure force (Fig. 1a–d). Employing a differential scattering approach with 3 and 0.1µm beads, our experiments in agarose hydrogels demonstrate that PFOCE is able to isolate bead mechanical responses induced by modulated radiation pressure by compensating for the photothermal responses of the surrounding hydrogels. The measured total response from the 3µm beads did not directly correspond to either the mechanical response or the photothermal response alone. These results confirm the importance of accounting for both scatteringmediated radiation pressure and absorptionmediated photothermal responses when using 976nm photonic excitation from a lowNA beam in aqueous media.
One potential area of application of PFOCE is the study of dynamic cell–ECM biophysical interactions in 3D environments. Particularly, the embedded scattering beads used in PFOCE may also readily serve as the fiducial beads in TFM^{45}. Recent innovations in TFM strive to extend traditional quasistatic and 2D dynamic studies to 3D volumetric measurements of dynamic cell–ECM interactions, including collective (emergent) cellular behaviors^{22,27,28,45}. Previous work on mechanical characterization of hydrogels typically used as ECM in mechanobiology studies reveals drastic variations in the pore shapes and sizes of fibrin hydrogels^{60}, and as a result, their local mechanical properties can vary by over an order of magnitude^{32,33}. An important capability that would enable such studies is the characterization of spatiotemporal variations in the mechanical properties of hydrogel substrates at subcellular length scales over a millimeterscale volume at timescales of minutes to hours. OTAMR is currently still the leading technique for characterization of ECM mechanical properties during livecellimaging studies^{33}. However, a practical challenge that persists in current OTAMR studies is the precise alignment of the trapping beam to the center of the marker beads to within 0.1 μm, which must be achieved on each individual bead being probed at a given time. This results in a net measurement time of 8 s per bead achieved by recent stateoftheart OTAMR studies^{33}. Furthermore, the detection of transverse bead displacements (typically on the order of 10^{1}–10^{2} nm) in OTAMR is implemented via a transillumination geometry, which limits the sample thickness (and turbidity) that can be imaged. In this respect, PFOCE has the benefit of epiillumination, continuousscanning acquisition that probes multiple beads at various depths in each Bscan, owing to the use of a lowNA beam to exert transversely localized axial radiationpressure force over an extended depth range.
Moving toward future implementation of PFOCE in 3D livecellimaging studies, the acquisition scheme presented here can be expanded to accommodate largerscale volumetric acquisition over timescales of minutes to hours. For instance, consider a 3D multicellular imaging study over a volumetric FOV of 1 mm × 1 mm × 200 μm (depth range about the PF beam focal plane with viable PFOCE mechanical response data) with statistical spatial sampling determined by a 15µm average edgetoedge bead separation (corresponding to approximately 48 500 beads within the volume). A feasible PFOCE acquisition scheme may involve acquiring two volumes serially, each consisting of 3000 BMmode frames per slowaxis position (achieving 190 pm sensitivity instead of the current 105 pm) and a frame rate of 200 Hz over a fastaxis scan range of 500 µm (the largest scan range achievable by our current galvanometer at this frame rate). At the transverse spatial sampling of 1 μm per pixel, the total PFOCE acquisition time for the entire volume would be 8.3 h, whereas an OTAMR measurement (at 8 s per bead, assuming that volumetric measurements can be conducted up to 200 µm depth) would take up to 4.5 days. If we were to use a resonant scanner^{55} and acquire the same PFOCE dataset at the frame rate of 500 Hz over a fastaxis scan range of 1 mm (for an exposure time of 2 µs per Ascan while maintaining the spatial sampling required to ensure sufficient SNR of OCT images), the total acquisition time would reduce to only 100 min. In this case, PFOCE offers over 60× improvement in volumetric throughput over the current stateoftheart OTbased microrheology techniques. Future studies will also address quantitative reconstruction of the local complex shear modulus from the measured PFOCE bead mechanical responses by utilizing OCTbased depthresolved measurement of radiationpressure force^{44}. Additionally, such studies may warrant bead sizes smaller than the 3µm ones used here; the NA of the PF forcing beam could be appropriately increased to optimize forces and the corresponding displacements (Fig. 1b, c). Based on our differential scattering approach, there is also a future possibility of completely endogenous PFOCE (that is, without embedded beads) that relies on native scattering of different constituents in the biological sample. With the potential to reconstruct microscale viscoelastic properties of hydrogels from variations in complex mechanical responses of embedded beads over millimeterscale volumes, PFOCE may unlock new research directions in cell mechanics and mechanobiology. A potential example of this is the study of how micromechanical properties and biophysical cell–ECM interactions impact collective behavior^{27,28,29}, such as the emergent 3D migration patterns of invasive cancer cells.
Methods
Sample preparation
First, a sample chamber shown in Fig. 2a was fabricated from microscope glass coverslips (Electron Microscopy Sciences, 22 × 22 mm, #0) for each agarose hydrogel sample. The coverslips were bonded together by an RTV silicone adhesive (Permatex, 80050) as shown in Fig. 2a. The chamber was left to cure for 24 h before use.
The five agarose hydrogel concentrations were made by mixing solid agarose polymer (Fisher Scientific, BP1423) with roomtemperature distilled water at the concentrations of 0.2, 0.3, 0.4, 0.5, and 1% (w/w). The mixture was repeatedly heated in a microwave oven for 5 s then stirred for 10 s until all visible agarose solid had dissolved and the mixture became clear. Throughout the process, the total weight of the mixture was constantly checked for any loss of water to evaporation; distilled water was added accordingly. Threemicrometer polystyrene microsphere suspension (SigmaAldrich, LB30) was added to the dissolved mixture at the concentration of 6 μL mL^{−1} to achieve 15μm mean edgetoedge particle spacing. Then, 0.1μm polystyrene microsphere suspension (SigmaAldrich, LB1) was added at a concentration of 0.12 μL mL^{−1} to achieve 2μm mean edgetoedge particle spacing. The mixture was stirred by hand to ensure all particles were dispersed before injecting into the premade glass chamber. The individual samples were injected such that the sample chamber was filled completely with no air gaps/bubbles. For the sidebyside samples, the sample chamber was partially injected with 1% agarose hydrogel such that it filled onehalf of the chamber without any gaps and left to set. Later, the remaining volume in the chamber was filled with 0.2% agarose hydrogel. Finally, the glass chamber was sealed on both opening ends by a liquid sealing glue (Bob Smith Industries, InstaCure+). We found that it was crucial to carefully seal the open ends of the chamber to produce a confined container. When the open ends of the glass chamber were not sealed, the hydrogels underwent drastic structural and compositional change due to evaporation of water and the motion of the beads was apparent under the OCT system during data acquisition.
Separate hydrogel samples were made in aluminum trays (The Lab Depot, TLDD43–100) from the same agarose–distilled water mixture for rheometer testing. The samples for the rheometer testing were cut into a disk with diameter of 40 mm and thickness of 2 mm.
Optical setup
The optical setup (Fig. 1a) for measuring complex OPL oscillation consisted of an SDOCT system with a broadband superluminescent diode (Thorlabs, LS2000B) with center wavelength of 1300 nm. The OCT beam focused with an NA of 0.14. The transverse and axial resolutions were 4.5 and 3.7 μm in air, respectively (we note that unlike confocal microscopy, OCT does not need high NA to achieve cellular resolution^{61}). Combined in free space with the sample arm of the OCT system is a laser diode (RPMC Lasers, R0976SB0500P) at the wavelength of 976 nm (in air) acting as the PF forcing beam. We selected the PF forcing beam wavelength of 976 nm for its low to negligible cytotoxicity reported in previous OTs studies^{62}. A BCM was used to optimize the PF forcing beam and ensure that it focused to the same position in 3D space as the OCT beam after going through the same OCT sample arm objective lenses (refer to Supplementary Methods). The waist radius of the PF forcing beam was measured to be 3.19 μm at the focal plane, corresponding to an NA of 0.1 (Supplementary Fig. 1).
Data acquisition
To maximize the force exerted by the PF forcing beam during acquisition, the alignment between the OCT beam and the PF forcing beam was checked at the beginning of each experiment (Supplementary Methods). A 3D BMmode acquisition scheme (Fig. 2b) was adopted. For each concentration in the individual samples of 0.2–0.5% agarose hydrogels, each 3D volume consisted of 10 BMmode datasets, acquired at 10 slowaxis positions along the yaxis. Each BMmode dataset consisted of 6144 frames (200 frames per second) with 256 Ascans per frame (Fig. 2c). This acquisition scheme provides a transverse FOV of 200 μm × 10 μm at the spatial sampling density of 0.8 μm per pixel along the fast xaxis and 1 μm per pixel along the slow yaxis. Each spatial voxel contains 6144 measurements of the sample response over time. During the acquisition of each BMmode dataset, the PF forcing beam power was modulated by a function generator (Tektronix, AFG3051C), which sent a continuous 20Hz sinusoidal modulation waveform to the laser diode controller. Since the modulation was provided externally by a function generator, asynchronous to the OCT acquisition control, we measured the function generator output at the mth Ascan in each frame to reconstruct the full PF drive waveform. The function generator voltage at the mth Ascan reflected the real part of the complex drive waveform, \(\tilde V\left( t \right)\), at that Ascan. From this measurement, we calculated the phase of the drive waveform at the first acquired Ascan from φ_{drive} = φ_{ m } − ωt_{ m }, where φ_{ m } and t_{ m } denote the phase and time at the mth Ascan, respectively. Then, we reconstructed the complex drive waveform as \(\tilde V(t) = V_0\exp (i(\omega t + \varphi _{{\mathrm{drive}}}))\), where V_{0} denotes the modulation amplitude. At the frame rate of 200 Hz, the OPL oscillation due to the 20Hz modulation was sampled at 10 distinct phases per modulation cycle. Three such volumes were acquired at different imaging locations in each sample. All synchronization and instrument controls were accomplished via a custom LabVIEW software.
The sidebyside sample was acquired with 208 Ascans spanning a fastaxis range of 208 μm at 490 frames per second and 3072 frames per slowaxis location. Acquiring 125 such slices at 1μm spacing yielded a 200 μm × 125 μm transverse FOV. Furthermore, the 20Hz waveform for modulating the PF forcing beam was generated internally via the LabVIEW software (rather than the function generator) in order to ensure precise synchronization with OCT Ascans.
Data processing
For the individual (singleconcentration) samples of agarose hydrogels, all the data processing was implemented in MATLAB. The spatialdomain OCT image was reconstructed with standard procedures (background subtraction, spectrum resampling, dispersion correction, and inverse Fourier transformation). In order to efficiently process large 3D BMmode datasets, only the depths containing the sample (601 pixels in depth out of 2048 acquired) were reconstructed; this was implemented with a highspeed SDOCT processing method for depthselective reconstruction^{45,63}. The reconstructed spatialdomain complex OCT image was first segmented into a photothermal data region (corresponding to the 0.1µm beads and bottom glass surface of the sample chamber) and a total response data region (corresponding to the 3µm beads) via thresholds based on magnitude of the reconstructed OCT image and OCT SNR (Fig. 3). The thresholds used to generate the results here are defined in Supplementary Table 2. Unless stated otherwise, all remaining processing steps outlined in this section were performed independently on each spatial pixel in the 3D OCT image that passed the thresholds, which encoded OPL oscillation resulting from 20Hz modulation of the PF forcing beam power.
The OPL oscillations due to the modulated PF forcing beam were estimated with a previously described method to reconstruct complex sample displacement in phasesensitive OCE^{64}. Briefly, the complex phase differences were calculated between every adjacent BMmode frame at each spatial pixel. The complex phase differences, expressed as e^{iΔΦ(r,t)}, were first registered to that of the top glass surface of the sample chamber to remove systematic noise and phase drifts across BMmode frames, then filtered by a median filter (3 × 3 kernel, applied separately to the real and imaginary parts of e^{iΔΦ(r,t)}). The realvalued phase differences, ΔΦ(r,t), were obtained from the phase angle of e^{iΔΦ(r,t)}, then, filtered by a Butterworth bandpass filter (±1 Hz passband centered around 20 Hz). The complex OPL oscillation at each spatial pixel was obtained after cumulative summation (integration in time) and Hilbert transformation of ΔΦ(r,t) along the time axis. This complex OPL oscillation corresponded to the raw photothermal response, ΔOPL_{PT} (x, y, z, t) and total response, ΔOPL_{tot} (x_{b}, y_{b}, z_{b}, t), for spatial pixels in the photothermal response data region and total response data region, respectively. The vector (x_{b}, y_{b}, z_{b}) refers the set of pixel coordinates of each spatial pixel comprising the segmented 3µm bead regions. Note that we omit the argument ω included in Eqs. (5)–(7) because ω = 2π (20 Hz) is implied for all PFOCE measurements.
The amplitude and phaseshift w.r.t. φ_{drive} of the raw ΔOPL_{PT} (x, y, z, t) in each 3D dataset were curvefit as a function of depth by the theoretical curve obtained from simulation and by a cubic polynomial function, respectively. A cubic polynomial function was used to fit the phase data because the theoretical simulation did not to reproduce the nonzero and depthdependent phase delay observed experimentally (Supplementary Figs 4 and 5). The curvefitting was done by minimizing the weighted sumsquare error (SSE). To accommodate for lowSNR data with large phase noise, the SSE calculation was weighted by the OCT SNR in each spatial pixel such that measurements with higher SNR were weighted more heavily. The weights, W, were given by
For each 3D dataset, the bestfit curves for the amplitude and phase data yielded A_{PT} (z) and φ_{PT} (z), respectively. Then, the depthdependent complex photothermal response, ΔOPL_{PT} (z, t), was obtained from \({\mathrm{\Delta OPL}}_{{\mathrm{PT}}}(z,t) = A_{{\mathrm{PT}}}(z)e^{i(\omega t + \varphi _{{\mathrm{drive}}} + \varphi _{{\mathrm{PT}}}(z))}\).
The complex mechanical response at each spatial pixel that comprised the segmented 3µm bead regions was obtained from ΔOPL_{mech} (x_{b}, y_{b}, z_{b}, t) = ΔOPL_{tot} (x_{b}, y_{b}, z_{b}, t) − ΔOPL_{PT} (z_{b}, t). The isolated ΔOPL_{mech} (x_{b}, y_{b}, z_{b}, t) was further filtered, via multiplication by a brickwall filter (±0.2 Hz passband) in the frequency domain, before its amplitude A_{mech} (x_{b}, y_{b}, z_{b}) and phase φ_{mech} (x_{b}, y_{b}, z_{b}) were extracted. In order to obtain A_{mech}, φ_{mech}, A_{PT}, φ_{PT}, A_{tot}, and φ_{tot} for each of the 3µm beads, the spatial pixels that belonged to a given 3µm bead were identified and grouped together by their pixel coordinates (x_{b}, y_{b}, z_{b}). Then, the median values of A_{mech}, φ_{mech}, A_{PT}, φ_{PT}, A_{tot}, and φ_{tot} were calculated for each group, yielding the responses of each 3µm bead. These median values were used to generate the maps of A_{mech} and φ_{mech} (Fig. 4b, c).
For the sidebyside samples, all data was processed using the same procedure as described above, except for the curvingfitting step to estimate ΔOPL_{PT} (z, t), which was performed over eight subvolumes, centered at different positions in the xy plane. Six peripheral subvolumes had transverse dimensions of 100 µm × 25 µm, and two central subvolumes had transverse dimensions of 100 µm × 50 µm. All sidebyside data processing, except for the photothermal curvefitting and the final step to estimate the median values of the responses from 3µm beads, was performed on a NVIDIA Titan Xp GPU with custom C++ software using the CUDA v8.0 Toolkit. This was a key step for practical implementation of PFOCE for volumetric datasets that resulted in a processing time of 90 min, a 40× improvement over its equivalent in MATLAB.
Statistical analysis
All statistical analysis was implemented in MATLAB. Two statistical tests were performed. First, a Wilcoxontype nonparametric test for ordered groups, proposed by Cuzick^{65}, was implemented to test the null hypothesis that that there was no statistically significant trend across the four agarose concentrations (that is, the responses from the four samples were not ordered) against the alternative hypothesis that there was a statistically significant trend. The normalized test statistics, z_{C}, and the associated twosided pvalue, p_{C}, are reported. The data were ordered such that a z_{C} > 0 indicates an increasing trend while a z_{C} < 0 indicates a decreasing trend. We considered a trend to be statistically significant if p_{C} < 0.05. Second, a multiple comparison based on Wilcoxontype nonparametric Kruskal–Wallis test of variance was implemented to determine if there were statistically significant differences between measurements from any two agarose concentrations. The reported pvalues reflect the significance of chisquared (χ^{2}) statistics on the groupadjusted (Bonferroni correction for multiple comparisons among groups) twosided pairwise comparison between two agarose concentrations. In both tests, rankbased nonparametric methods were chosen to accommodate for deviation from a normal distribution (Anderson–Darling test for normality) and unequal variances (Barlett’s test for equal variances) among measurements in different agarose concentrations.
Although the variations in the local mechanical responses within a sample can be attributed to the heterogeneity of agarose hydrogels (see Supplementary Discussions for possible sources of variability within a sample), in order to compare these responses to the bulk mechanical properties measured via rheometry and to observe the overall trend of the mechanical response across different concentrations, the reconstructed data (that is, A_{tot} (x_{b}, y_{b}, z_{b}), φ_{tot} (x_{b}, y_{b}, z_{b}), A_{mech} (x_{b}, y_{b}, z_{b}), φ_{mech} (x_{b}, y_{b}, z_{b}), A_{PT} (z_{b}), and φ_{PT} (z_{b})) were subjected to further thresholding to exclude outliers from the responses of each agarose concentration prior to performing the statistical tests. Any spatial pixels that contained A_{mech} (x_{b}, y_{b}, z_{b}) values above the 85th percentile or below the 15th percentile of all A_{mech} (x_{b}, y_{b}, z_{b}) values in each 3D dataset (that is, the percentiles were calculated separately for each imaged location in each sample), were excluded from the statistical tests. The acceptance or rejection of the null hypothesis by the Cuzick’s test for trend was not affected by this exclusion (Supplementary Fig. 9). The responses of each of the 3µm beads were then obtained from the median values among all remaining spatial pixels in each 3D dataset that constituted each bead.
All statistical analysis, including all box plots (Fig. 5 and Supplementary Figs 6 and 7), were performed using the median values for each of the 3µm beads after exclusion of outlier pixels. The total number of beads included in the statistical analysis was N = 54, 74, 58, and 52 beads for 0.2%, 0.3%, 0.4%, and 0.5% agarose hydrogel samples, respectively.
Data availability
All relevant data are available from the authors.
References
 1.
Grier, D. A revolution in optical manipulation. Nature 424, 21–27 (2003).
 2.
Stevenson, D. J., GunnMoore, F. & Dholakia, K. Light forces the pace: optical manipulation for biophotonics. J. Biomed. Opt. 15, 041503 (2010).
 3.
Bowman, R. W. & Padgett, M. J. Optical trapping and binding. Rep. Prog. Phys. 76, 026401 (2013).
 4.
Ashkin, A., Dziedzic, J. M., Bjorkholm, J. E. & Chu, S. Observation of a singlebeam gradient force optical trap for dielectric particles. Opt. Lett. 11, 288–290 (1986).
 5.
Block, S. M., Blair, D. F. & Berg, H. C. Compliance of bacterial flagella measured with optical tweezers. Nature 338, 514–518 (1989).
 6.
Svoboda, K., Schmidt, C. F., Schnapp, B. J. & Block, S. M. Direct observation of kinesin stepping by optical trapping interferometry. Nature 365, 712–727 (1993).
 7.
Molloy, J. E., Burns, J. E., KendrickJones, J., Tregear, R. T. & White, D. C. S. Movement and force produced by a single myosin head. Nature 378, 209–212 (1995).
 8.
Wang, Y. et al. Visualizing the mechanical activation of Src. Nature 434, 1040–1045 (2005).
 9.
Bustamante, C., Bryant, Z. & Smith, S. B. Ten years of tension: singlemolecule DNA mechanics. Nature 421, 423–427 (2003).
 10.
Forties, R. A. & Wang, M. D. Discovering the power of single molecules. Cell 157, 4–7 (2014).
 11.
Ashkin, A. Acceleration and trapping of particles by radiation pressure. Phys. Rev. Lett. 24, 156–159 (1970).
 12.
Guck, J. et al. The Optical Stretcher: a novel laser tool to micromanipulate cells. Biophys. J. 81, 767–784 (2001).
 13.
Guck, J. et al. Optical deformability as an inherent cell marker for testing malignant transformation and metastatic competence. Biophys. J. 88, 3689–3698 (2005).
 14.
Ekpenyong, A. E. et al. Viscoelastic properties of differentiating blood cells are fate and functiondependent. PLoS ONE 7, e45237 (2012).
 15.
Dholakia, K., MacDonald, M. P., Zemanek, P. & Cizmar, T. Cellular and colloidal separation using optical forces. Methods Cell Biol. 82, 467–495 (2007).
 16.
Jonas, A. & Zemanek, P. Light at work: the use of optical forces for particle manipulation, sorting, and analysis. Electrophoresis 29, 4813–4851 (2008).
 17.
Eyckmans, J., Boudou, T., Yu, X. & Chen, C. S. A Hitchhiker’s guide to mechanobiology. Dev. Cell 21, 35–47 (2011).
 18.
Discher, D. E., Janmey, P. & Wang, Y. L. Tissue cells feel and respond to the stiffness of their substrate. Science 310, 1139–1143 (2005).
 19.
Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 8, 241–254 (2005).
 20.
Levental, K. R. et al. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell 139, 891–906 (2009).
 21.
Butcher, D. T., Alliston, T. & Weaver, V. M. A tense situation: forcing tumor progression. Nat. Rev. Cancer 9, 108–122 (2009).
 22.
Wirtz, D., Konstantopoulos, K. & Searson, P. C. The physics of cancer: the role of physical interactions and mechanical forces in metastasis. Nat. Rev. Cancer 11, 512–522 (2011).
 23.
Guilak, F. et al. Control of stem cell fate by physical interactions with the extracellular matrix. Cell Stem Cell 5, 17–26 (2009).
 24.
Gilbert, P. M. et al. Substrate elasticity regulates skeletal muscle stem cell selfrenewal in culture. Science 329, 1078–1081 (2010).
 25.
Wong, V. W., Akaishi, S., Longaker, M. T. & Gurtner, G. C. Pushing back: wound mechanotransduction in repair and regeneration. J. Invest. Dermatol. 131, 2186–2196 (2011).
 26.
Mammoto, T., Mammoto, A. & Ingber, D. E. Mechanobiology and developmental control. Annu. Rev. Cell Dev. Biol. 29, 27–61 (2013).
 27.
Trepat, X. et al. Physical forces during collective cell migration. Nat. Phys. 5, 426–430 (2009).
 28.
Tambe, D. et al. Collective cell guidance by cooperative intercellular forces. Nat. Mater. 10, 469–475 (2011).
 29.
RocaCusachs, P., Conte, V. & Trepat, X. Quantifying forces in cell biology. Nat. Cell Biol. 19, 742–751 (2017).
 30.
Pampaloni, F., Reynaud, E. G. & Stelzer, E. H. K. The third dimension bridges the gap between cell culture and live tissue. Nat. Rev. Mol. Cell Biol. 8, 839–845 (2007).
 31.
Legant, W. R. et al. Measurement of mechanical tractions exerted by cells in threedimensional matrices. Nat. Methods 7, 969–971 (2010).
 32.
Kotlarchyk, M. A. et al. Concentration independent modulation of local micromechanics in a fibrin gel. PLoS ONE 6, e20201 (2011).
 33.
Keating, M., Kurup, A., AlvarezElizondo, M., Levine, A. J. & Botvinick, E. Spatial distribution of pericellular stiffness in natural extracellular matrices are dependent on cellmediated proteolysis and contractility. Acta Biomater. 57, 304–312 (2017).
 34.
Mejean, C. O., Schaefer, A. W., Millman, E. A., Forscher, P. & Dufresne, E. R. Multiplexed force measurements on live cells with holographic optical tweezers. Opt. Express 17, 6209–6217 (2009).
 35.
Schwingel, M. & Bastmeyer, M. Force mapping during the formation and maturation of cell adhesion sites with multiple optical tweezers. PLoS ONE 8, e54850 (2013).
 36.
Curtis, J. E., Koss, B. A. & Grier, D. G. Dynamic holographic optical tweezers. Opt. Commun. 207, 169–175 (2002).
 37.
Scarcelli, G. & Yun, S. H. Confocal Brillouin microscopy for threedimensional mechanical imaging. Nat. Photon. 2, 39–43 (2008).
 38.
Scarcelli, G. et al. Noncontact threedimensional mapping of intracellular hydromechanical properties by Brillouin microscopy. Nat. Methods 12, 1132–1134 (2015).
 39.
Mulligan, J. A., Untracht, G. R., Chandrasekaran, S. N., Brown, C. N. & Adie, S. G. Emerging approaches for highresolution imaging of tissue biomechanics with optical coherence elastography. IEEE J. Sel. Top. Quantum Electron. 22, 6800520 (2016).
 40.
Kennedy, B. F., Wijesinghe, P. & Sampson, D. D. The emergence of optical elastography in biomedicine. Nat. Photon. 11, 215–221 (2017).
 41.
Larin, K. V. & Sampson, D. D. Optical coherence elastography—OCT at work in tissue biomechanics. Biomed. Opt. Express 8, 1172–1202 (2017).
 42.
Nichols, E. F. & Hull, G. F. A preliminary communication of the pressure of heat and light. Phys. Rev. 13, 307–320 (1901).
 43.
Ma, D., Garrett, J. L. & Munday, J. N. Quantitative measurement of radiation pressure on a microcantilever in ambient environment. Appl. Phys. Lett. 106, 091107 (2015).
 44.
Leartprapun, N., Iyer, R. R. & Adie, S. G. Depthresolved measurement of optical radiationpressure forces with optical coherence tomography. Opt. Express 26, 2410–2426 (2018).
 45.
Mulligan, J. A., Bordeleau, F., ReinhartKing, C. A. & Adie, S. G. Measurement of dynamic cellinduced 3D displacement fields in vitro for traction force optical coherence microscopy. Biomed. Opt. Express 8, 1152–1171 (2017).
 46.
Oestreicher, H. L. Field and impedance of an oscillating sphere in a viscoelastic medium with an application to biophysics. J. Acous. Soc. Am. 23, 707–714 (1951).
 47.
Wang, L. V. & Hu, S. Photoacoustic tomography: in vivo imaging from organelles to organs. Science 335, 1458–1462 (2012).
 48.
Li, C. et al. Laser induced surface acoustic wave combined with phase sensitive optical coherence tomography for superficial tissue characterization: a solution for practical application. Biomed. Opt. Express 5, 1403–1418 (2014).
 49.
Skala, M. C., Crow, M. J., Wax, A. & Izatt, J. A. Photothermal optical coherence tomography of epidermal growth factor receptor in live cells using immunotargeted gold nanospheres. Nano Lett. 8, 3461–3467 (2008).
 50.
LapierreLandry, M., TuckerSchwartz, J. M. & Skala, M. C. Depthresolved analytical model and correction algorithm for photothermal optical coherence tomography. Biomed. Opt. Express 7, 2607–2622 (2016).
 51.
Ren, K. F., Grehan, G. & Gouesbet, G. Prediction of reverse radiation pressure by generalized LorenzMie theory. Appl. Opt. 35, 2702–2710 (1996).
 52.
Oyen, M. L. Mechanical characterization of hydrogel materials. Int. Mater. Rev. 59, 44–59 (2014).
 53.
Park, B. H. et al. Realtime fiberbased multifunctional spectraldomain optical coherence tomography at 1.3 um. Opt. Express 13, 3931–3944 (2005).
 54.
Chang, E. W., Kobler, J. B. & Yun, S. H. Subnanometer optical coherence tomographic vibrography. Opt. Lett. 37, 3678–3680 (2012).
 55.
Klein, T. & Huber, R. Highspeed OCT light sources and system. Biomed. Opt. Express 8, 828–859 (2017).
 56.
Hale, G. M. & Querry, M. R. Optical constants of water in the 200nm to 200um wavelength region. Appl. Opt. 12, 555–563 (1973).
 57.
Pernodet, N., Maaloum, M. & Tinland, B. Pore size of agarose gels by atomic force microscopy. Electrophoresis 18, 55–58 (1997).
 58.
Narayanan, J., Xiong, J. Y. & Liu, X. Y. Determination of agarose gel pore size: absorbance measurements vis a vis other techniques. J. Phys. Confer. Ser. 28, 83–86 (2006).
 59.
Boral, S., Saxena, A. & Bohidar, H. B. Syneresis in agar hydrogels. Int. J. Biol. Macromolec. 46, 232–236 (2010).
 60.
Kotlarchyk, M. A., Botvinick, E. L. & Putnam, A. J. Characterization of hydrogel microstructure using laser tweezers particle tracking and confocal reflection imaging. J. Phys. Condens. Matter 22, 194121 (2010).
 61.
Swanson, E. A. et al. In vivo retinal imaging by optical coherence tomography. Opt. Lett. 18, 1864–1866 (1993).
 62.
Liang, H. et al. Wavelength dependence of cell cloning efficiency after optical trapping. Biophys. J. 70, 1529–1533 (1996).
 63.
Chelliyil, R. G., Ralston, T. S., Marks, D. L. & Boppart, S. A. Highspeed processing architecture for spectraldomain optical coherence microscopy. J. Biomed. Opt. 13, 044013 (2008).
 64.
Adie, S. G. et al. Spectroscopic optical coherence elastography. Opt. Express 18, 25519–25534 (2010).
 65.
Cuzick, J. A Wilcoxontype test for trend. Stat. Med. 4, 87–90 (1985).
Acknowledgements
The authors gratefully acknowledge useful discussion with Dr. Nozomi Nishimura regarding statistical analysis. This work was funded in part by National Institutes of Health (NIBIBR21EB022927, Adie) and Cornell Discovery and Innovation Research Seed Award (Adie). This work made use of the Cornell Center for Materials Research Shared Facilities, which are supported through the NSF MRSEC program (DMR1120296). Additional information can be found at http://adie.research.engineering.cornell.edu.
Author information
Affiliations
Contributions
N.L. and R.R.I. set up and conducted PFOCE experiments and theoretical simulations; N.L. conducted rheometry measurements; N.L. and R.R.I. processed and analyzed all data; J.A.M. developed the MATLAB code for depthselective OCT reconstruction; G.R.U. and J.A.M. helped conceive theoretical simulations; N.L. and R.R.I. implemented the MATLAB based GLMT simulations; R.R.I. implemented the GPUbased processing software for data analysis; G.R.U. designed and implemented the LabVIEW acquisition software; S.G.A. conceived the principle of PFOCE and the experiments presented here; N.L. wrote this manuscript; S.G.A. supervised the writing of this manuscript; all coauthors assisted in the revision of this manuscript.
Corresponding author
Ethics declarations
Competing Interests
S.G.A., N.L., and G.R.U. are listed as inventors on a patent (U.S. Patent Application No. 15/162,608) assigned to Cornell University related to the methods presented in this manuscript. The remaining authors declare no competing interests.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Leartprapun, N., Iyer, R.R., Untracht, G.R. et al. Photonic force optical coherence elastography for threedimensional mechanical microscopy. Nat Commun 9, 2079 (2018). https://doi.org/10.1038/s41467018043578
Received:
Accepted:
Published:
Further reading

Ultrahighsensitive optical coherence elastography
Light: Science & Applications (2020)

Elastography mapped by untangling compressional and shear deformation
Extreme Mechanics Letters (2020)

Stiffness Sensing by Cells
Physiological Reviews (2020)

Interstitial magnetic thermotherapy dosimetry based on shear wave magnetomotive optical coherence elastography
Biomedical Optics Express (2019)

Optical pulling at macroscopic distances
Science Advances (2019)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.