Mapping mechanical properties of biological materials via an add-on Brillouin module to confocal microscopes

Abstract

Several techniques have been developed in the past few decades to assess mechanical properties of biological samples, which has fueled rapid growth in the fields of biophysics, bioengineering and mechanobiology. In this context, Brillouin optical spectroscopy has long been known as an intriguing modality for non-contact material characterization. However, limited by speed and sample damage, it had not translated into a viable imaging modality for biomedically-relevant materials. Recently, based on a novel spectroscopy strategy that substantially improved the speed of Brillouin measurement, confocal Brillouin microscopy has emerged as a unique complementary tool to traditional methods as it allows non-contact, non-perturbative, and label-free measurements of material mechanical properties. The feasibility and potential of this innovative technique on both cell and tissue level has been extensively demonstrated in the past decade. As Brillouin technology is rapidly recognized, a standard approach for building and operating Brillouin microscopes is required to facilitate the widespread adoption of this technology. In this protocol, we aim to establish a robust approach for instrumentation, data acquisition and analysis. By carefully following this protocol, we expect the Brillouin instrument can be built in 5–9 days by a person with basic optics knowledge and alignment experience; the data acquisition as well as post-processing can be accomplished within 2–8 h.

EDITORIAL SUMMARY

A protocol for implementing Brillouin microscopy to study biological materials. The procedure contains instructions on how to integrate an add-on Brillouin module onto existing confocal microscope, calibration, data collection and processing.

TWEET

A new Protocol for confocal #Brillouin microscopy of biological samples.

COVER TEASER

Brillouin microscopy of biological materials

Introduction

Biomechanical interactions are recognized as a central player in the regulation of cellular functions (e.g. proliferation, migration, and gene expression)^{1, 2} as well as system-level behaviors (tissue morphogenesis, cancer metastasis, and angiogenesis)^3–5. To characterize the mechanical interplay within biological samples, such as cells, tissue, and biomaterials, we need tools to identify mechanically-related pathways⁶, measure deformations and forces⁷, and quantify material mechanical properties^{8, 9}. Assessing material mechanical properties has been greatly advanced by the development of many important techniques, which can be loosely classified into three categories. (a) Contact-based techniques, where a stress is applied on the sample and the corresponding strain is quantified to extract modulus. These include indentation methods such as atomic force microscopy (AFM)^10–14 and other cantilever-based tools^{15, 16}, micropipette aspiration^{17, 18}, parallel-plate rheometer¹⁹ and stretching substrate²⁰. (b) Bead-based techniques, where external beads are attached to or injected in the sample and their behavior is monitored to extract material properties. These include optical tweezer^{21, 22}, magnetic twisting cytometry (MTC)^{23, 24}, magnetic tweezer²⁵, passive microrheology²⁶ and microdroplet-based sensors²⁷. (c) Elastography techniques, where the application of an external force is coupled with imaging techniques to measure displacement and back-calculate sample strain. These include optical stretching²⁸ and optical coherence elastography (OCE)^{29, 30}. Interestingly, many mechanical testing concepts have been recently combined with or adapted to microfluidics technology^31–34. Each of these techniques has peculiar strengths and limitations, which have led to a specialization of their use for particular applications; for example, contact-based techniques are considered the gold-standard when direct assessment of the Young’s modulus of the cell and tissue samples is needed; bead-based techniques have been extensively used to extract the localized mechanical properties of the sample, mostly at the cellular and subcellular scales; elastography techniques are powerful to characterize and map the mechanical properties of biological tissue; and microfluidics-based techniques can assess the mechanical properties of cells in medium with high throughput.

In this context, an entirely different mode of probing material mechanical properties, via all-optical Brillouin light scattering, has been known and used for decades³⁵. Since its invention, Brillouin spectroscopy has been workhorse for non-contact material characterization. However, its application in biological field has mostly remained limited to single point measurements and ex vivo because of speed and sample damage limitations. In the past ten years, due to the development of a faster spectroscopy strategy, confocal Brillouin microscopy has emerged as an imaging modality suitable for live biological samples³⁶, which allows non-contact, non-perturbative and label-free measurements^{30, 37, 38}. Brillouin microscopy is an all-optical imaging modality, in which a laser beam is used to probe the mechanical properties of the sample on the micron-level scale, and a 2D/3D mechanical image with diffraction-limited resolution can be acquired by scanning the sample in a confocal configuration³⁹. We, along with colleagues from different labs over the world, have demonstrated non-disturbing characterization of numerous biological materials with Brillouin microscope on both cell and tissue level^39–51. Although the technique is rapidly being recognized, a standardized approach for Brillouin microscopy has not been established, which is a significant obstacle towards the widespread adoption of the technology. In this protocol, we aim to fill this gap and provide a robust approach for instrumentation, data acquisition and analysis. By following this protocol, we expect a reliable Brillouin microscope can be built by an engineer, postdoc or graduate student with non-expert optical knowledge and high-quality data can be obtained by a non-specialist user.

Development of the protocol

Brillouin microscopy is based on the physical process of spontaneous Brillouin scattering, where the interaction between the incident light and inherent acoustic waves inside a sample (generated by spontaneous thermal fluctuations) introduces a frequency shift (i.e. Brillouin shift) to the outgoing scattered light (Fig. 1a)³⁵. A typical Brillouin spectrum features a central laser peak with a symmetrical Brillouin doublet (Stokes and Anti-Stokes components) (Fig. 1b). Because the propagation of the acoustic wave is governed by the mechanical properties of the material, the Brillouin shift v_B can probe the material’s longitudinal modulus M′ via the relationship:

$$v_B=\frac{2n}{\lambda }\sqrt{\frac{M\text{'}}{\rho }}\text{sin}\left(\frac{\theta }{2}\right),$$ (1)

where n and ρ is the refractive index and mass density of the material, λ is the light wavelength, and θ is the collection angle of the scattered light. Brillouin microscopy usually collects backward scattered light thus θ = 180°. The Brillouin shift of common materials is in the order of GHz, which corresponds to less than 0.01 nm in wavelength. To resolve such a small frequency difference, common grating spectrometers used in fluorescence or Raman spectroscopy cannot be used. Historically, Brillouin spectroscopy has been enabled by the development of multi-pass scanning Fabry-Perot (FP) etalon interferometry by JR Sandercock⁵². For decades, FP-based Brillouin spectroscopy has been used for material characterization³⁵ and remote sensing⁵³. In biological fields, FP-based Brillouin spectroscopy measurements were reported for collagen fibers^54–56, cornea and crystalline lens of the eye^{57, 58}, biofilms⁵⁹ and bone⁶⁰. However, these early demonstrations in the biological domain were single point measurements and mostly ex vivo, because scanning FP spectrometers acquire the spectral components in a sequential manner and at low-throughput, which typically results in long acquisition times (minutes) per each Brillouin spectrum. Non-scanning FP spectrometers based on angular dispersion were subsequently demonstrated, which shortened the acquisition time⁶¹ and enabled the first 2D Brillouin imaging of liquid-polymer sample⁶². However, angular-dispersive FP spectrometers remained fundamentally limited in throughput; thus, in order to increase speed, they sacrificed the ability to reject noise by limiting to single-pass operation. Thus, while Brillouin spectroscopy had been widely utilized for noncontact material characterizations, biological applications remained underexplored due to speed limitations and/or high-power laser needed.

Figure 1. Principle of spontaneous Brillouin scattering.

(a) Brillouin scattering is caused by the interaction of incident laser and acoustic phonons of the sample. (b) A representative Brillouin spectrum. The subscripts S and AS indicate Stokes and Anti-Stokes components, respectively. V represent the frequency.

The instrument described in this protocol is the result of the past fifteen years of optical development dedicated to address the speed and sample damage limitations for biological applications. The instrument is based on a different type of angular-dispersive etalon, virtually imaged phased array (VIPA), where the light is coupled into the etalon through a window and propagates forward after multiple reflections between the two surfaces of the etalon⁶³. Since the entry surface of the VIPA has ~100% reflection, and the output surface has a partial reflection (>95%), light coupled into the etalon will be all transmitted without forming a reflected interference pattern. In 2008, we demonstrated the Brillouin microscope using a VIPA-based Brillouin spectrometer which allowed much higher throughput than traditional FP etalon³⁶. In subsequent years, a multi-stage VIPA-based Brillouin spectrometer was developed to improve the noise rejection ability at high throughput^{64, 65}. Meanwhile, other techniques such as destructive interference⁶⁶, absorptive gas chamber⁶⁷, spectral apodization³⁹, multi-stage FP filter⁶⁸ and spectral coronagraphy⁶⁹ were developed to further reject non-Brillouin light components. In addition, Brillouin signal strength has been improved by using optimal excitation wavelength (660 nm) such that the acquisition time can be shortened without losing signal-to-noise ratio (SNR)⁷⁰. Currently, the state-of-art Brillouin microscope described here has acquisition time of 40 ms and can operate under shot-noise limited condition with noise rejection of 75 – 80 dB, allowing it to measure non-transparent tissues^{46, 71, 72} and monitor biological processes in situ⁵¹.

Overview of the Procedure

The procedure described in this protocol consists of 6 stages. First, we describe how to set up the microscope and add-on optics (Steps 1–10), where it is critical to install the setup in a standard lab room with least temperature fluctuation and to choose the proper port of the microscope for beam coupling. Next, we describe how to build and optimize the Brillouin spectrometer (Steps 11–36), which are the most time-consuming stages of this protocol and require substantial attention. Then, we describe how to prepare the samples for Brillouin measurement by taking NIH 3T3 cells as an example (Step 37). Next, we describe how to image the sample using the Brillouin microscope (Steps 38–44), in which the periodical calibration is critical in case the frequency drift of the laser is a concern. Lastly, we describe how to post-process the data and present the results with 2D Brillouin image (Steps 45–48).

Applications of the method

Ocular biomechanics.

Mechanics of the cornea is crucial for the variability in refractive procedures and one of the main factors in eye disease like keratoconus⁷³. Existing methods for quantifying corneal mechanics in vivo have low sensitivity and spatial resolution, and the results rely on multiple assumptions and modeling⁷⁴. Brillouin microscopy provides a promising approach to address this clinical need, as it can provide the mechanical map of the cornea in vivo with high resolution and sensitivity^{41, 43, 75, 76}. Clinical data have shown Brillouin technique has the potential for diagnosis and treatment monitoring of keratoconus patients⁷⁷. In addition, Brillouin microscopy has demonstrated feasibility of measuring the biomechanics of crystalline lens^{40, 78, 79} and retina⁸⁰.

Developmental biomechanics.

The morphological evolution during embryo development involves cell alignment and folding as well as tissue reshaping and patterning, accompanied by dramatic mechanical changes of embryonic tissue^{4, 81}. As a non-contact technique, Brillouin microscopy has shown enough sensitivity and spatial resolution in quantifying tissue mechanics of both mouse and zebrafish embryos^{72, 82}. Specifically, this technique has demonstrated the possible applications in biologically relevant processes, such as the neural tube closure⁵¹, the development and injury of spinal cord⁴⁵, and the epithelial cell differentiation across the corneal surface⁴⁸.

Cellular biomechanics.

Mechanical cues from microenvironment, together with gene expression and signaling pathway, regulate cell functions and activities through the mechanism of mechanotransduction^{83, 84}. In response, cells alter their mechanical properties, suggesting that mechanical signatures are promising indicators of cell behaviors. In 2015, we reported 3D mechanical mapping of live cells and revealed the mechanical changes due to cytoskeletal modulation and cell-volume regulation with Brillouin microscopy³⁹. Brillouin microscopy was then applied to quantify subcellular mechanical properties of both live animal and plant cells^{44, 47, 85–87}. Brillouin microscopy is also suitable to assess the mechanical properties of the nucleus, which is embedded in the surrounding cytoplasm and not directly accessible with traditional methods⁸⁸. Using Brillouin technique, we have revealed that the nuclear mechanics is regulated by both the cytoskeletal network and internal nanostructures⁵¹, which can help understand how the intracellular mechanisms alters nuclear mechanics and thus affect cell behaviors in physiological and pathological conditions.

Cancer metastasis and 3D tumorigenesis.

Substantial evidence has shown cancer cells have altered mechanical properties comparing with healthy cells^{33, 89}. Specifically, metastatic cancer cells usually are softer (more deformable) than non-cancer ones^90–94. In this context, we have revealed that nuclear mechanics is a crucial regulator as tumor cells migrated through confined region, which represents a key step in the metastatic cascade⁹⁵. In addition, Brillouin microscopy has been validated in monitoring the biomechanics of 3D nodules and tumoral tissues^{49, 50}, paving the way for studying tumor progression and drug therapy with 3D in vitro model.

Biomaterial characterization.

The non-contact and non-invasive features of the technique may open up new possibility for material characterization that are challenging for conventional methods. For example, Brillouin spectroscopy technique allowed the first ever quantification of the entire stiffness tensors of spider silk, whose diameter is small, and the linear elastic properties could be altered upon mechanical deformation⁹⁶. In addition, Brillouin technique has been used to characterize the mechanical properties of other biomaterials, such as collagen⁹⁷, fibrous proteins of the extracellular matrix⁹⁸, protein crystals⁹⁹, and hydrogels^100–102.

Medical applications.

In addition, Brillouin microscopy has been suggested as a useful tool in a variety of medical applications, such as Barrett’s oesophagus⁹⁸, atherosclerosis¹⁰³, Bacterial meningitis¹⁰⁴, Alzheimer’s disease^{105, 106}, obesity¹⁰⁷, amyotrophic lateral sclerosis⁴⁷, melanoma¹⁰⁸ and dental diagnostics and restoration¹⁰⁹.

Comparison with other methods

FP etalon-based Brillouin spectroscopy.

The VIPA-based Brillouin microscopy described in this protocol is based on the same physical process of spontaneous Brillouin scattering as FP etalon-based spectroscopy. The main difference between the two instruments lies in the spectral dispersion elements. Although both the FP and VIPA use a parallel etalon to disperse light, the FP etalon has intrinsically lower throughput because it forms an interference pattern in both transmission and reflection. For practical applications, the reflected pattern represents a significant loss of signal which increases as the finesse (and thus spectral resolution) of the spectrometer increases. Of note, clever attempts to remediate this issue by re-circulating the reflected light have been recently demonstrated¹¹⁰. The VIPA etalon overcomes the throughput issue with peculiar surface coating: the entry surface of the VIPA has ~100% reflection, other than a small anti-reflection (AR) window for coupling the input beam, while the output surface has a partial reflection (>95%). As a result, light coupled into the etalon will be all transmitted without forming a reflected interference pattern resulting into an increase in throughput on the order of the interferometer finesse. In addition, the non-scanning configuration allows VIPA to acquire the entire spectrum with one shot, while FP etalon has to acquire each spectral component in sequence by physically scanning the cavity of the etalon. Considering these factors, a typical VIPA-based spectrometer is much faster than a FP etalon-based spectrometer. On the other hand, FP-based spectrometers can be tuned to allow a finer spectral analysis with better linewidth. In addition, the multi-pass multi-stage FP-etalon spectrometer can provide higher spectral extinction (>150 dB) than the highest demonstrated with VIPAs (85 dB)^{64, 68, 111}. This extra extinction is valuable for the characterization of highly reflective materials such as metal; however, for biological applications, current VIPA-based spectrometers have sufficient extinction to operate in shot-noise limited regime even in opaque tissues or very close to the reflective interfaces. In short, as a pioneer Brillouin technology, F-P etalon-based Brillouin spectroscopy is still actively involved in biological applications nowadays^{96, 98, 112, 113}. However, because the spectral acquisition of the scanning F-P etalon is slow, this technique is mostly limited to single-or few-point analysis.

Stimulated Brillouin scattering (SBS) microscopy.

This technique is based on the stimulated Brillouin scattering, in which the acoustic phonons are excited by the presence of two counterpropagating continuous-wave (CW) lasers¹¹⁴ (Fig. 2a). Since the stimulated process has higher scattering efficiency, the generated Brillouin signal can be much stronger than in the case of spontaneous Brillouin scattering^115–117. As such, the acquisition time of the SBS microscopy can potentially be shortened comparing with spontaneous confocal Brillouin microscopy. Additional advantages include higher spectral resolution and less background elastic scattering noise, which enables measurements of linewidth and Brillouin strength, which in turn can provide viscosity and density information. Very recently, SBS microscopy demonstrated its first biological application by mapping live C elegans samples¹¹⁸. However, current SBS microscopy demonstrations use CW lasers, thus they do not fully exploit the nonlinear effect advantage and need high laser power to reach the threshold of stimulated process. Consequently, potential phototoxicity is a concern for a biological sample exposed under such high laser power. While C. elegans has been shown to be least affected through SBS mapping, the photodamage for other biological samples should be carefully evaluated before implementing this technique as the phototoxic effect strongly depends on samples. In addition, measurements have to be implemented in a transmission geometry and thus access to the sample from both sides is needed.

Figure 2. Schematics of SBS and ISBS.

The dashed line represents the output laser beam containing Brillouin signal of the material. The line set inside the sample indicates the excited acoustic phonons. PD: photo-detector.

Impulsive stimulated Brillouin scattering (ISBS) microscopy.

This technique is also based on the stimulated scattering process, but an ultrashort pulse laser is used to excite the acoustic phonons¹¹⁹ (Fig. 2b). In the implementation of ISBS microscopy, the acoustic phonons are generated within the interference pattern of a pulsed pump laser and detected by a probe laser under Bragg diffraction condition^{120, 121}. Currently, the spatial resolution of ISBS microscopy does not reach single-cell level (lateral: 10 μm; axial: 230 μm) due to the optical configuration, and the setup has to be implemented in a transmission geometry. The validation of the technique has been demonstrated by measuring liquid samples and hydrogels, but the biological application has not been reported so far.

Limitations

• The mechanical property probed by Brillouin technique via the Brillouin shift is the microscopic high-frequency longitudinal modulus within the sample. This quantity is not the same as quasi-static elastic modulus (e.g. Young’s modulus, shear modulus) measured by conventional stress-strain methods on macroscopic samples. Currently, there is a lack of theoretically-established relationship between two moduli. However, strong correlations have been observed in many physiological and pathological processes^{39, 40, 50, 51, 75, 122}, which are expected because most material changes alter both moduli in the same direction. For highly hydrated gels (~95% water content) and large hydration variations, Brillouin shift is dominated by the change of water content and thus stops being a reliable estimator of conventional elastic modulus^{123, 124}; instead, in the regime of cells and tissues (~70% water content), Brillouin technology can be used to estimate traditional mechanical properties after building a correlation working curve between the two moduli³⁹.

• Currently, the pixel dwell time of the Brillouin microscope is 40 ms – 100 ms, which is slow comparing to other imaging modality such as fluorescent confocal microscopy. This is because the Brillouin signal generated from the spontaneous scattering process is intrinsically weak. Increasing the input laser power could reduce the dwell time thus improve the speed. However, high laser power could introduce possible photodamage via absorption of the incident light by the sample thus the ultimate improvement in this respect is limited⁷⁰. As a result, for large samples, the acquisition time could become a crucial limiting factor and should be considered carefully.

• The sensitivity of the instrument is limited by the spectral precision of the spectrometer. As demonstrated in the Procedures, the spectral precision of the described instrument is about 8 MHz in the measurement settings analyzed here (Fig.8). Considering that biological samples typically have a Brillouin shift of ~ 6 GHz, the relative precision of Brillouin frequency is about 0.13%. According to the relationship between longitudinal modulus and Brillouin shift (Equation 1), the relative precision of longitudinal modulus is 0.26%. This can be translated to a precision in the estimation of Young’s modulus on the order of few percent according to the quantified correlation of two moduli in cells and tissues^{39, 40, 51}.

• The calibration method used in this protocol assumes the spectral dispersion is constant across the spectral pattern, which, however, is not exactly true in practical scenarios. As the dispersion is related to the angle of the light within the VIPA cavity, higher-order dispersion would have less displacement on the EMCCD camera¹²⁵. Consequently, Brillouin components from adjacent dispersion orders have slightly different spectral dispersion, which results in a nonlinearity of VIPA pattern and might introduce artefact to the calibration result. If the Brillouin shift of a sample is far from those of the calibration materials, the inaccuracy of the derived Brillouin shift may be observable. In this case, more standard materials whose known Brillouin shifts covers broader region of the spectral pattern can be used in calibration¹²⁶, and the calibration parameters can be obtained by fitting multiple Brillouin peaks with a polynomial function.

• The spatial resolution primarily depends on the voxel size of the focused laser beam, which is determined by the diffraction limit of the optical system, and thus by the NA of the objective lens (Objective 1) chosen in the experiment. For objective lenses with high numerical aperture, the spatial resolution will also be affected by the mean free path of the acoustic phonons¹²⁷. Therefore, mechanical properties of the sample smaller than the spatial resolution will not be resolved by Brillouin microscope. Instead, the output result will be the averaged Brillouin shift of all the materials within that voxel. In practice this limits the ultimate mechanical resolution of Brillouin microscopy on the order of 1 micron.

• The penetration depth of the Brillouin microscope depends on the transparency of the sample; at the demonstrated spectral extinction, the Brillouin confocal microscope has similar penetration depth as reflectance confocal microscopy. In practice this corresponds to ~200 μm of mouse embryo at 532 nm⁴⁶, > 500 μm of zebrafish embryo at 780 nm⁴⁵, 360 μm of retina at 532 nm⁸⁰, ~90 μm of tumor nodules at both 532 nm and 660 nm⁵⁰, and 100 μm of chicken muscle tissue at 532 nm⁶⁸.

• To avoid any crosstalk, the optical path for the Brillouin channel and the fluorescence channel are physically separated. As such, Brillouin and fluorescent images are acquired in sequence rather than simultaneously. Although this will slightly increase the overall acquisition time, the accuracy of the image co-registration in the data postprocessing are not affected.

• The add-on optics in this protocol are designed for a specific commercial microscope (IX81, Olympus); depending on the configuration of different microscopes, tweaks to the optical setup may be required to ensure the protocol can work as expected.

Figure 8. Characterization of the Brillouin spectrometer.

(a) Time trace of 300 times measurement of water. (b) Histogram of the Brillouin shift precision. Inset is the zoomed-in plot. The linewidth of the histogram is 8 MHz. (3) Shot-noise limited curve of water sample with 10 mW laser power and various acquisition times. Dots are measured data, and the red curve is obtained from the linear fitting of the data under log-log scale. The slope is 0.4934.

Expertise needed to implement the protocol

A skilled engineer or applied physicist, graduate student or postdoc with optics background and experience of optical alignment should be able to build the instrument within 5–9 days by carefully following this protocol. An undergraduate student with computer engineering background and experience of LabVIEW-based programming should be able to develop the software interface within a month. A non-specialist user is expected to optimize and run the instrument to obtain high-quality data after necessary training (2–4 weeks). All these estimates are based on our experience training instrument builders and users.

Experimental design

Instrumentation.

The instrument includes a commercial inverted confocal microscope (e.g. IX81 + Disk Spinning Unit (DSU), Olympus) and a homemade add-on Brillouin module (Fig. 3). To briefly describe the operation of the commercial microscope, Lamp 1 and 2 are light sources for brightfield and fluorescent imaging, respectively. For both brightfield and fluorescent/confocal imaging, the Filter turret 1 is set to a 45° mirror to direct light to/from the DSU which consists of a Filter turret 2 and the spinning disk. A CMOS camera (Neo, Andor) is installed on the DSU to acquire both brightfield and fluorescent/confocal images. For sample translation, the microscope is equipped with a 2D translational stage (Prior Scientific) for holding and scanning the sample in x-y plane. Scanning along z-axis is achieved by moving the objectives (Objective 1) with the microscope’s own stage.

Figure 3. Schematic of the confocal Brillouin microscope.

(a) Integration of a confocal microscope and the add-on Brillouin module. Filter turret 1/2: commercial microscope turret for housing the fluorescence filter cubes; DSU: disk spinning unit; ND filter: neutral density filter; M1, M2, M3, M5: reflection mirrors; M4: dichroic mirror. HWP: half-wave plate; PBS: polarized beam splitter; QWP: quarter-wave plate. (b) Optical design of the Brillouin spectrometer. M6-M7: reflection mirrors; C1-C2: cylindrical lens; SL1-SL4: spherical lens; SF: spatial filter; LP: lens pair. EMCCD: electron multiplying charge coupled device. Filter 1/2: linear variable ND filter.

The add-on Brillouin module includes a CW laser source, an optical setup outside the microscope body, a Brillouin excitation/emission “filter cube” to guide light in and out of the microscope body, and a Brillouin spectrometer. The Brillouin filter cube is a home-made customized mount to fit a slot of the commercial microscope’s turret which usually houses the fluorescence filter cubes. To enable Brillouin measurement “channel”, the Brillouin filter cube is easily placed into the beam path by rotating the Filter turret 1.

The laser source should have a single frequency output with a narrow bandwidth (< 1 MHz), an excellent spectral purity (> 60 dB) and a good frequency stability (< 0.2 GHz in 0.5 h). In this protocol, a 660-nm laser (Torus, Laser Quantum) was used as the light source. After passing through an optical isolator, neutral density (ND) filter, beam expander (which is designed such that the output beam can overfill the back focal aperture (BFA) of the Objective 1) and a half-wave plate, the laser beam goes through a polarized beam splitter (PBS). Depending on the polarization of the incoming light, controlled by Polarizer 1, the PBS guides the beam to either the mirror M2, for Brillouin measurements of samples, or M5, for Brillouin measurements of calibration materials. For Brillouin measurements of samples, the laser beam is coupled into the microscope body through its right-side port by mirrors M2 and M3. Then the beam is focused into the sample after passing through mirror M4, the Brillouin filter cube (installed on a free slot of the Filter turret 1), and the objective lens (Objective 1). The backward scattered Brillouin light is collected by the same objective. Since the input laser and Brillouin signal have nearly identical wavelengths, no traditional filter or dichromatic mirror can be used to separate the two. For this reason, the Brillouin filter cube contains a quarter-wave plate (QWP); this allows to confer orthogonal linear polarizations to input laser and backward scattered Brillouin signal and separate them at a PBS. The Brillouin scattered light is transmitted through the PBS to a second objective lens (Objective 2), which couples light to the Brillouin spectrometer through a single mode fiber. For calibration purpose, the initial polarization of the laser beam can be rotated by 90° so that it is transmitted through the PBS to the calibration arm instead of the microscope body. In the calibration arm, the beam first passes through a QWP2 and is focused into the reference material under examination by a positive lens L3. The scattered Brillouin light traces its path back, is reflected at the PBS and is coupled into the fiber by Objective 2 to be analyzed by the Brillouin spectrometer.

The choice of Objective 1 depends on the smallest length scale that one wants to observe, and its numerical aperture (NA) primarily determines the spatial resolution of the Brillouin microscope. Since the Brillouin scattering is a result of the interaction between light and acoustic phonon, the actual resolution of Brillouin imaging will also be affected by the propagation of the acoustic wave, which can be quantified by the wavelength and the mean free path of the phonon within the sample¹²⁷. Therefore, the length scales of both the optical beam and the acoustic wave should be taken into consideration when estimating the mechanical resolution. To ensure the confocality and maximize the collection of the Brillouin signal at the fiber port, the Objective 2 is chosen such that the output beam matches the NA of the single mode fiber. Since the diameter of the collected Brillouin beam is determined by the BFA of Objective 1, which is usually smaller than the BFA of Objective 2, the effective NA of Objective 2 is scaled by BFA₂⁄BFA₁. Taking the choices of this protocol as an example, as the Objective 1 (40x/0.6 NA) has BFA₁ = 5.4 mm and the Objective 2 (10x/0.25 NA) has BFA₂ = 9 mm, the effective NA of Objective 2 becomes 0.15, which is close to the NA (0.1–0.14) of the fiber. Alternatively, the beam expander consisting L1 and L2 can be designed such that the size of the outcoming laser beam matches the BFA of Objective 2 to optimize the fiber coupling, and an additional beam expander can be placed between the PBS and the microscope body to match the BFA of Objective 1. In case the microscope port with a tube lens is used for the coupling of the add-on Brillouin module, the tube lens itself can be part of the additional beam expander then only one more lens is needed to adjust the beam size at the BFA of Objective 1.

The Brillouin spectrometer (Fig. 3b) features two-stages with orthogonally oriented VIPAs (free spectral range FSR = 15 GHz, LightMachinery). The outcoming light from the fiber is coupled into the VIPA 1 by the cylindrical lens C1 in the vertical direction, and the spectral pattern at the output of the VIPA 1 is projected onto the Mask 1 by the cylindrical lens C2. The spherical lens SL1 then couples the pattern on Mask 1 into the VIPA 2 in the horizontal direction, and the spherical lens SL2 projects the spectral pattern at the output of VIPA 2 on the Mask 2. After being relayed by a 4-f system consisting of the spherical lens SL3 & SL4 and spatial filter SF⁶⁹, the spectral pattern is imaged onto the electron multiplying CCD (EMCCD) camera (iXon, Andor) by a lens-pair LP. To improve the extinction of the spectrometer, linear variable ND filters (Filter 1 & 2) are placed right after the VIPAs as apodization filters to convert the intensity profile of the VIPA pattern from an exponential shape to a Gaussian shape³⁹.

Alternatives of key equipment.

1. Microscope body. This protocol can be adapted to other inverted microscopes from different manufacturers. For the microscope body with an infinity port (e.g. Leica DMi8), which can guide the laser beam to the Objective 1 without a tube lens, the add-on Brillouin module can be coupled into the microscope through the infinity port without any change. For other microscopes (e.g. Zeiss Axiovert; Nikon Eclipse Ts2R), the imaging port (side port) instead can be used for coupling. In this case, since the laser beam has to pass through a tube lens before reaching Objective 1, an additional lens can be inserted between M3 and M4 for relaying purposes. This lens has the same focal length as the tube lens and is placed at twice the focal length of the tube lens, so that it and the tube lens together build a 1:1 imaging system to make sure the laser beam after the tube lens is collimated and has the same diameter as it is before M3.

2. Laser source. As mentioned earlier, to obtain high-quality Brillouin data, the spectral linewidth, purity, and the frequency stability are crucial parameters for choosing a laser source. Beyond that, the trade-off between the scattering efficiency and possible phototoxicity should be considered when choosing the wavelength⁷⁰. Shorter wavelengths (e.g. 532 nm and 561 nm) will provide higher Brillouin signal than longer wavelengths (e.g. 660 nm, 671 nm, and 780 nm) because the scattering efficiency is proportional to λ⁻⁴, but the light absorption by melanin, hemoglobin and lipid in cell and tissue in short wavelength may introduce significant photodamage when the laser power is high¹²⁸. On the other hand, water absorption will be dominant in infrared region. Another advantage of using short wavelength is to achieve a better spatial resolution. The wavelength used in this protocol is not the only choice but a compromise considering the signal strength and the photodamage. As a matter of fact, the feasibility of other wavelengths has also been demonstrated for different applications40,^{47, 48, 86, 96, 129, 130}.

3. Brillouin spectrum camera. When choosing a camera, several parameters should be considered, including dark current, read noise, quantum efficiency, and pixel readout rate. As the signal from spontaneous Brillouin scattering is relatively weak, a camera with low noise level and high quantum efficiency is preferred. In this protocol, an EMCCD camera is recommended for the best performance. We notice that sCMOS camera has also been used in Brillouin spectrometers^{86, 111}. Comparing with the EMCCD (e.g., iXon Ultra 897, Andor), the sCMOS camera (e.g., Neo 5.5, Andor) has only 60% quantum efficiency but >20 times dark current. On the other hand, the sCMOS has much higher pixel readout rate than the EMCCD. Taken together, a sCMOS camera could be under consideration when the Brillouin signal is fairly strong and fast acquisition is desired.

4. FSR of the VIPA. The unambiguously range of the Brillouin shift that can be measured by the spectrometer is about half of the VIPA’s FSR. The choice of FSR in this protocol is based on the fact that most cells and some tissue samples (such as cornea and early-stage embryo) have the Brillouin shift in the range of 6–7 GHz at 660 nm. For very stiff biological samples, such as crystalline lens⁴⁰, fibers⁹⁶, cartilage⁹⁸, and bone^{131, 132}, a VIPA with larger FSR is needed.

Brillouin spectrum and calibration.

After dispersion by a VIPA etalon, the input light is separated in space based on its frequency (Fig. 4a). The spectral pattern imaged onto the EMCCD camera could include multiple dispersion orders. Each order features a triplet (laser peak, Stokes, and anti-Stokes Brillouin peaks) and the successive orders are spaced by a frequency spectral range (FSR) (Fig. 4b). In most cases of experiments, the intensity of the laser peak is much higher than the Brillouin signal because of the light reflection at interface as well as the strong scattering in non-transparent samples. As such, the camera can be easily saturated by the overwhelming laser peaks thus the weak Brillouin signal will be distorted or even buried. To avoid this, instead of using a single triplet to determine the Brillouin shift v_B, we use the anti-Stokes Brillouin peak from one dispersion order and the Stokes Brillouin peak from the successive order, as shown in the highlighted region of Fig. 4b, and all the remaining peaks are blocked by the Masks 1 and 2 within the spectrometer. With the measured peak separation ΔD and the calibrated FSR, the Brillouin shift v_B can be readily determined by

$$v_B=\frac{FSR-\Delta D}{2}.$$ (2)

Figure 4. Brillouin spectrum acquisition and calibration.

(a) Principle of VIPA-based spectrometer. (b) Brillouin spectrum acquired by a double-stage VIPA spectrometer. FSR: free spectral range. The dashed square indicated the region acquired by the camera. (c) Acquired Brillouin spectra of standard materials and the intensity profiles after curve fitting. d_m and d_w is the peak separation of the Brillouin spectrum for methanol and water, respectively. Dots are raw data and red lines are fitted curves.

Since the peak separation recorded on the camera is counted by pixels, the pixel-to-frequency conversion ratio (PR) has to be calibrated to calculate ΔD.

In the calibration process, the Brillouin spectra of two standard materials with known Brillouin shift (e.g. water and methanol) are acquired through the calibration arm, and the results are then used to determine the two unknown parameters (FSR and PR) through the relationship:

$$FSR-PR\cdot d_w=2v_w$$ (3a)

$$FSR-PR\cdot d_m=2v_m$$ (3b)

where d_w and d_m are the peak separations of the Brillouin signals from water and methanol samples, respectively, which are determined from the acquired spectra (Fig. 4c). The above process assumes the laser frequency is constant. In practice, the distinct drift of laser frequency could introduce artifact when applying a single calibration result to the experimental data collected over a long time. In this case, more frequent calibrations during the experiment are needed to obtain reliable data. To achieve this, the Polarizer 1 is installed on a motorized rotation mount, and a motorized flip mirror is placed after the L3 and set 45° to the optical path for switching between standard materials. The periodical calibration (e.g. per 0.5 h) can be implemented by automatically controlling the motorized mounts with a computer program.

High-NA objective lens will introduce a small shift of the Brillouin peak (e.g., 2% at NA= 0.8) and a distinct broadening of the spectral full-width-at-half-maximum (FWHM)¹³³. In this protocol, the collection NA of the reference arm (NA= 0.14) is smaller than that of the measurement arm (NA= 0.6), but the calibration error caused by the peak shift (<1%) can be neglected. However, if higher-NA objective lens is used in the measurement, the potential artifact might need to be considered and the direct comparison of the Brillouin data collected from low- and high-NA based on the same calibration arm should be careful. In case this is a concern, we recommend matching the NA of both arms to remove this artifact.

Sample preparation.

The sample is usually placed on a glass-bottomed petri dish with small amount of culture medium. For easier measurements, it is suggested, if possible, to coat the bottom of the dish with a thin hydrogel layer before seeding the sample to avoid strong reflection at the interface of glass and medium. The thickness of the layer can be adjusted based on the working distance of the objectives (Objective 1) and the thickness of the sample. For long-term live imaging, the dish can be housed by a mini-incubator designed for microscope to control the temperature, humidity, and CO₂ level.

Data acquisition.

The data acquisition is implemented by a homemade LabVIEW script (Supplementary Figure 1), and the procedure is shown in the flow chart of Figure 5. In the initialization of the spectrometer, the tilt angle of the VIPAs needs to be carefully adjusted to both maximize the signal and balance the intensity of two peaks. In spectrometer calibration, multiple samplings (e.g. n = 300) are taken for each standard material. Next, the optical path is switched to bright-field/fluorescent channel by rotating the Filter turret 1(Fig. 3a). Next, when the sample of interest is located in the field of view, the scanning strategy is chosen in terms of the scanning plane (e.g. x-y, x-z, or y-z), range, step size, and the exposure time of the EMCCD camera. After the optical path is switched back to Brillouin channel, the data acquisition is launched. During this step, Brillouin spectrum at each position of the sample is captured and saved as a figure file for data postprocessing. To monitor the status of the measurement in real time, the corresponding Brillouin shift is obtained by fitting the spectrum with a Lorentzian function, and the value is displayed as a pixel of a color image in the LabVIEW program. When a session of measurement is completed, it is advisable to calibrate the spectrometer again, because the laser frequency fluctuation as well as the mechanical drift of the VIPAs might introduce artifact to the result. Depending on the frequency stability of the laser and the measurement time of each sample, spectrometer calibration should be implemented whenever the drift is suspected.

Figure 5.

Flowchart of the data acquisition process for Brillouin imaging.

Data post-processing.

We use MATLAB to process the raw data (example codes as well as representative raw data are provided in Supplementary Data 1). The key procedures are summarized in Fig. 6. First, FSR and PR are calculated based on Equation (3) with each sampling point of the calibration data, and the average is used as the ultimate value of the parameters for retrieving Brillouin shift of the sample. For the same-day experiments, it is important to check the values of FSR and PR from all calibration datasets. In case any drift is observed, the values obtained from the nearest calibration dataset should be used for retrieving Brillouin shift of the related measurement. With the 1D vector of retrieved Brillouin shift, the 2D Brillouin image can be reconstructed with the information of scanning range and step size. Further analysis can be performed by co-registering Brillouin image with the corresponding fluorescent image, thus the Brillouin shift of either the whole sample or any subregion of interest can be extracted. Furthermore, with the known refractive index and mass density of the sample, the longitudinal modulus derived by Brillouin shift can be obtained based on Equation (1).

Figure 6.

Flowchart of the data post-processing.

Materials

Biological materials

• Cells of interest. Any adherent cells should work with the method described in the protocol, such as NIH 3T3 fibroblast cells (ATCC Cat# CRL-6442, RRID:CVCL_0594). Caution: The cell lines used for your research should be regularly checked to ensure that they are authentic, not cross-contaminated and not infected with mycoplasma.

Reagents

• Methanol (Fisher Scientific, cat. no. A453-500). Caution: Methanol is highly flammable and toxic. When handling methanol, avoid direct exposure by wearing gloves and cover the lid of the container properly.

• Deionized water (Fisher Scientific, cat. no. LC267402)

• Cell culture media. For NIH 3T3 cells, we use DMEM (Thermo fisher, cat. no. 11995065) with 10% fetal bovine serum (Thermo fisher, cat. no. 30-2020) and 1% penicillin-streptomycin (Thermo fisher, cat. no. 15070-063).

• (Optional) Hydrogel. For example, polyacrylamide (MilliporeSigma, cat. no. GF61308536-1EA).

• (Optional) Fluorescent dyes. For example: Hoechst 33342 (Thermo Fisher, cat. no. 62249).

Equipment

• Optical table (e.g. 4 × 6 ft, Newport, cat. no. INT4-46-8-A)

• Microscope body (Olympus, IX81+DSU)

• Motorized 2D stage for optical microscope (Prior Scientific, cat. no. H117E2)

• Objective lens (Objective 1: Olympus, cat. no. LUCPLFLN40X; Objective 2: Thorlabs, cat. no. RMS10X)

• EMCCD camera (Andor, iXon 897)

• CMOS camera (Andor, Neo 5.5 sCMOS)

• VIPA (FSR = 15 GHz, LightMachinery, cat. no. OP-6721-6743-3)

• Holder for VIPAs (Thorlabs, cat. no. KM100C, x2)

• Mount adapter for EMCCD camera and lens pair (Thorlabs, cat. no. SM1A39)

• Laser source (Laser Quantum, torus 660)

• Isolator (Thorlabs, cat. no. IO-5-670-HP)

• ND filter (Thorlabs, cat. no. NDC-25C-4)

• Beam expander (f₁ = 16 mm, Thorlabs, cat. no. AC080-016-B-ML; f₂ = 100 mm, Thorlabs, cat. no. AC254-100-B-ML)

• Mirrors (M4: Chroma, cat. no. T660dcrb; all the rest: Thorlabs, cat. no. BB1-E02-10)

• Polarizer (Newport, cat. no. 10LP-VIS-B)

• Half-wave plate (Thorlabs, cat. no. WPH10M-670)

• Quarter-wave plate (Thorlabs, cat. no. WPQ10M-670, x2)

• Polarized beam splitter (Thorlabs, cat. no. CCM1-PBS252)

• Single mode fiber (Thorlabs, cat. no. P1-460Y-FC-2)

• 3D translational stage (Thorlabs, cat. no. MBT616D)

• Lens for calibration arm (f₁ = 45 mm, Thorlabs, cat no. AC254-045-B-ML)

• Optical components for Brillouin spectrometer (all from Thorlabs. Fiber coupler, f = 11 mm, cat. no. PAF2P-11B; cylindrical lens (x2), f = 200 mm, cat. no. LJ1653L1-B; spherical lens (x2), f = 200 mm, cat. no. AC508-200-B; spherical lens (x2), f = 40 mm, cat. no. AC254-040-B-ML; lens pair, MAP103030-B; Variable ND filter (x2), cat. no. NDC-25C-4.)

• Slits for masks (Thorlabs, cat. no. VA100)

• Iris and 2D translation mount for spatial filter (Thorlabs, cat. no. SM1D12, cat. no. CXY1)

• 1D translational stages (Newport, cat. no. 423, x9)

• Vertical stages (Edmund Optics, cat. no. #66-499, x3)

• Components for building the optical enclosure (all from Thorlabs. Rails, cat. no. XE25L09, XE25L12; Black hardboard, cat. no. TB4; Blackout fabric, cat. no. BK5; Black masking tape, cat. no. T137-2.0)

• Cuvette for holding standard liquid materials (Thorlabs, CV10Q3500F)

• Petri dish (35 mm, glass bottom, Fisher Scientific, cat. no. NC9732969)

• MATLAB software (ver. R2020a, MathWorks, https://www.mathworks.com)

• MATLAB example codes (Supplementary Data 1)

• LabVIEW software (ver. 2016, National Instruments, https://www.ni.com/enus/shop/labview.html)

Alignment tools

• Power meter (Thorlabs, cat. no. PM130D)

• Shearing interferometer (Thorlabs, cat. no. SI050, SI100P)

• Home-built alignment assembly (Step 6), including irises (Thorlabs, cat. no. SM1D12C, x2), cage plates (Thorlabs, cat. no. CP33, x3), cage assembly rods (Thorlabs, cat. no. ER18, x4), and threading adapters (Thorlabs, cat. no. SM1A4, SM1T2)

Customized components

• Mount for Brillouin filter cube (Supplementary Data 2)

• Mounting base for laser head, including three stacked plates (Supplementary Data 3)

• Mounting stand for EMCCD camera, including base component, front component, circle component, and spacers for height adjustment (Supplementary Data 4)

Procedure

Setting up the microscope and add-on optics, Timing: 2–4 d

• 1Install and fix the microscope body on the optical table, making it about 3 ft from the right short end and 2 in from the bottom long side, and facing to the top long side of the table (Supplementary Figure 2).

  Critical step: Since the frequency stability of the laser and mechanical mounts inside the spectrometer are sensitive to the environmental temperature, the setup should be installed in a standard lab room with least temperature fluctuation. Thermally stabilized mounts can be used when temperature fluctuation is a concern.

• 2Place the laser head near the right short end of the table and install it on the customized aluminum base. The height of the base is determined such that the laser beam ultimately has the same height as the center of the right-side port of the microscope. Alternatively, a periscope optical arrangement would be needed to facilitate coupling the laser beam into the microscope body. Adjust the orientation of the laser head such that the output laser beam roughly aligns with the table holes. Install the isolator in front of the laser head. Install a variable ND filter after the isolator and adjust the laser power to the minimum visibility for alignment.
• 3Install the beam expander, including lens L1 (f₁ = 16 mm) and L2 (f₂ = 100 mm), right after the ND filter. Place L2 on a 1D translational stage. Finely adjust its position to make the output beam collimated, which can be checked by using a standard shear plate interferometer (Thorlabs). The magnification of the beam expander (f₂/f₁) is chosen by considering both the size of the laser beam and the BFA of Objective 1. Here, the size of the laser beam is about 1.8 mm, and it becomes 11.25 mm after expansion so that the beam can overfill the BFA of several common objectives. In case a different laser source and/or Objective 1 are used, the combination of L1 and L2 needs to be adjusted accordingly.
• 4Install the mirror M1 about 450 mm away from the laser. Set the mirror about 45° against the incoming laser beam so that the reflected beam aligns with the table holes. Adjust the tilt of the mirror to make the reflected beam leveled against the surface of the table. Install the half-wave plate (HWP) and linear polarizer (Polarizer 1) after M1. The linear polarizer is used to switch laser beam between the measurement arm and the calibration arm and ensure the polarization purity. The laser power after Polarizer 1 can be tuned by adjusting the orientation of the HWP.
• 5Install the PBS about 350 mm away from M1. Adjust the orientation of the PBS to make the outgoing vertically polarized beam nearly align with the table holes. This can also be checked by looking at the weak surface-reflected beam spot on the position of the Polarizer 1 with a small target card.
• 6Install the mirrors M2 and M3 to guide the beam into the microscope body and make the beam propagate upright after the mirror M4. To achieve good alignment, build an alignment assembly using cage mounts which includes two irises spaced by more than 400 mm. Make sure the aligning assembly ends with a mounting adapter to screw the assembly onto a free slot of the objective turret. To align, first uninstall the microscope translational stage, rotate the objective turret to a free slot and mount the aligning assembly. Adjust the beam direction to make the beam pass through the center of two irises simultaneously; this can be done by tuning the adjustment screws of the mirror mounts (two screws per mirror, which tune the horizontal and vertical tilt degrees of freedom). Once alignment is achieved, remove the aligning assembly, mount back the translational stage onto the microscope body.

  Critical Step: After M4, the beam should be upright without tilt. If it is tilted, the alignment of fiber coupling in Step 9 will not work for Brillouin beam path, which will significantly affect the collection efficiency of Brillouin signal. Before implementing the alignment assembly, it is important to make sure two irises are coaxial on the cage mount. If not, either label the correct position on the iris far from the threading adapter or use only one iris for alignment by sliding it along the cage rods.

  ? TROUBLESHOOTING

• 7Mount a mirror (not shown in the figures) above the translational stage with the reflective surface facing the laser beam to reflect the laser beam back to its original path. Adjust the tip/tilt of the mirror such that the reflected beam spot overlaps with the original one on M2. An iris can be used for this step within the M1-M2 beam path for ease of visual alignment. In this case the iris is centered on the incident beam, and the reflected beam spot should be adjusted to retrace back to the center of the iris.
• 8Install the Brillouin filter cube in one position of the Filter turret 1 of the commercial microscope. Set the fast axis of the quarter-wave plate (QWP) inside to be 45° to the polarization orientation of the beam. To achieve this, place a power meter between the PBS and Objective 2 and then rotate orientation of the QWP until the maximum power is detected.
• 9Place a target card 20 cm away from the PBS and mark the center of the beam. Install Objective 2. Adjust the height and tilt of Objective 2 to make sure the center of the output beam overlaps with the marker then move the card away. Install one end of the single mode fiber on a 3D translational stage. In case the height of the stage does not match the laser beam, install the stage on a stable post (e.g. 1.5” stainless steel) or a customized base. Finely adjust the position of the fiber port to achieve the maximum coupling of the laser beam, which can be monitored by using a power meter at the output end of the fiber.

  Critical Step: Make sure the axes of the Objective 2 (by adjusting the height and tilt) and the fiber port are aligned with the optical axis of the beam. Otherwise, the coupling efficiency could be greatly affected.

  ? TROUBLESHOOTING

• 10Align the calibration arm (Fig. 3b). Adjust the orientation of the polarizer 1 to allow a portion of beam to pass through the PBS. Install the mirror (M5) about 120 mm away from the PBS. Set the mirror about 45° against the incoming laser beam so that the reflected beam aligns with the table holes. Install the QWP 2 and the lens (f = 45 mm). Set the fast axis of the QWP 2 to be 45° to the polarization orientation of the incoming beam, which can be achieved by following the same procedure for installing Brillouin filter cube (Step 8).

Building the Brillouin spectrometer, Timing: 3–5 d

• 11Identify at least 1800 mm by 300 mm free space on the optical table (Supplementary Figure 2). Mount the EMCCD camera on the customized mounting stand and install the stand at one end of the free space. Attach the lens pair (LP) to the camera window.

  Caution: the EMCCD camera is heavy. Make sure the mounting stand is properly fixed before installing the camera.

• 12Temporarily place the fiber coupler close to the EMCCD camera. Adjust the height of the beam to be in the center of the EMCCD camera. To avoid the damage of the camera under direct laser illumination, the power of the laser beam should be very low (<0.01 mW), and camera should be set as no gain and low acquisition time (e.g. <1 ms). The parameters can be further adjusted for best visual inspection on the camera.

  Caution: To avoid the damage of the camera, the laser power should be low enough and the gain of the camera should be disabled before the illumination of the laser beam.

• 13Install the fiber coupler at the opposite end of the camera along the table’s free space (Fig. 3b). Install and adjust the mirrors M6 and M7 such that the laser beam after M7 is aligned, i.e. straight to table holes, parallel to the table height and illuminating the center of the EMCCD camera.
• 14Assemble the 4-f unit (SL3, SL4: f = 40 mm) using cage system mounting and fully open the aperture of the iris SF, which is mounted on a 2D translational stage. Place a target card in front of the EMCCD camera and mark the position of the laser spot. Place the 4-f unit far away from the target card. Adjust the distance between the lens SL3 and SL4 such that the size and position of the beam spot on the target card does not change. This can be double checked by using a commercial shear plate interferometer. Place a power meter after the 4-f unit. Close the aperture of the iris to the minimum. Finely adjust the position of the iris SF along optical axis and the 2D translational stage until the output power is maximum. This ensures the iris is on the back focal plane of the lens SL3. When this is done, fully open the iris.
• 15Reinstall the 4-f unit right in front of the EMCCD camera. Make sure the position of the beam spot does not change after installation.

  Critical step: It is important to ensure the 4-f unit is aligned with the optical axis. To facilitate the alignment, two irises can be used for determining the beam path by temporarily attaching them to the unit before SL3 and after SL4, respectively. The beam position can be checked with the EMCCD camera.

• 16Place the slit (Mask 2) horizontally on a 1D translational stage. Install the stage in front of the 4-f unit so that the mask is about 40 mm away from the first surface of SL3. Adjust the position of the slit along the beam path (z- axis) such that its blades are sharply imaged onto the EMCCD camera; adjust the lateral position of the slit (x-y plane) so that it is centered on the light beam.
• 17Install a 1D translational stage about 200 mm away from the Mask 2 and make its moving axis is along the optical axis (z- axis). Install the lens SL2 (f = 200 mm) on the stage. Adjust the stage such that the back focal plane of the lens SL2 is on the Mask 2. Install another 1D translation stage about 400 mm away from SL2 and install the lens SL1 (f = 200 mm) on the stage. Adjust the position of SL1 to make the output beam of SL2 collimated, which can be checked with a shearing interferometer.
• 18Carefully mount the VIPA 2 onto the holder with the entrance AR coated window oriented vertically (Fig. 3b). Install a 2D translation stage (combination of two 1D stages) about 200 mm away from the SL2. Install the VIPA 2 holder on the 2D stage. Set the height of the VIPA 2 such that the beam spot is located close to the top of the entrance window.
• 19Carefully slide the VIPA 2 into the beam path such that the focused beam spot is coupled into the entrance window. Adjust the horizontal translation stage and the horizontal-tilt degree of freedom of the VIPA 2 to optimize the coupling, which can be monitored by using a target card after the VIPA 2. Finely adjust the location of the VIPA 2 along the optical axis to make the last surface of the VIPA 2 overlap with the back focal plane of the lens SL1. To achieve this, place a power meter at the output of the VIPA 2, and repeat adjustments of the VIPA 2 until the maximum power is detected.

  Critical Step: The throughput of the VIPA 2 should be optimized in this step, at least 50% should be achieved. Since the VIPA pattern is large, in case the sensor is not big enough, a positive lens with short focal length can be used to help light collection for the power meter.

  ? TROUBLESHOOTING

• 20Finely adjust the position of the lens SL2 along optical axis until the spectral lines appear sharp on the EMCCD camera (Fig. 7a). Record the spectrum and calculate the experimental finesse of the spectrometer, which is defined as the ratio of the distance of adjacent diffraction order lines to the FWHM of one line. Aim for finesse > 30.

  Critical step: To achieve the expected finesse, the vertical-tilt degree of freedom of the VIPA 2 may need slight adjustment such that the output pattern of VIPA 2 is leveled horizontally, which can be checked using a target card right after VIPA 2 (Supplementary Figure 3). Exercise on the adjustment of both horizontal- and vertical-tilt degrees of freedom is suggested to familiarize their functions.

  ? TROUBLESHOOTING

• 21Slide the VIPA 2 fully out of the beam path.
• 22Install a 1D translational stage about 200 mm away from the lens SL1 and align its moving axis with the optical axis. Assemble a vertical stage on the translation stage. Install the slit (Mask 1) vertically on the vertical stage. Adjust the position of the slit such that its blades are sharply imaged onto the EMCCD camera.
• 23Install a 1D translation stage about 200 mm away from the Mask 1 by the same way as last step. Install the cylindrical lens (C2) on the stage such that its height axis is vertically aligned. Adjust the stage such that the back focal plane of C2 is on the Mask 1.
• 24Install a 1D translation stage about 400 mm away from the C2 by the same way as last step. Install the cylindrical lens (C1) on the stage by the same way as last step. Finely adjust the position of C1 such that the output beam of C2 is collimated, which can be checked by the commercial shear plate interferometer.
• 25Carefully mount the VIPA 1 onto the holder with the entrance window oriented horizontally and located on the top of the holder (Fig. 3b). Install a 1D translation stage about 200 mm away from the C2 by the same way as last step. Assemble a vertical stage on the translation stage. Install the VIPA 1 on the vertical stage such that the entrance window is lower than the beam.
• 26Carefully raise the VIPA 1 into the beam path such that the focused beam spot is coupled into the entrance window at the middle. Adjust the vertical-tilt degree of freedom of the VIPA 1 and the vertical stage to optimize the coupling, which can be monitored by using a target card after the VIPA 1. Finely adjust the location of the VIPA 1 along the optical axis to make the last surface of the VIPA 1 at the back focal plane of the C1. To achieve this, place a power meter at the output of the VIPA 1, and repeat adjustments of VIPA 1 until the maximum power is detected.

  Critical Step: The throughput of the VIPA 1 should be optimized in this step, at least 50% should be achieved.

• 27Finely adjust the position of the CL2 along optical axis until the spectral lines appear sharp on the EMCCD camera (Fig. 7b). Record the spectrum and calculate the finesse (as defined in Step 20). Aim for finesse > 30.

  Critical step: To achieve the expected finesse, the horizontal-tilt degree of freedom of the VIPA 1 may need slight adjustment such that the output pattern of VIPA 1 is straight vertically, which can be checked using a target card right after VIPA 1. Exercise on the adjustment of both horizontal- and vertical-tilt degrees of freedom is suggested to familiarize their functions.

  ? TROUBLESHOOTING

• 28Carefully slide in VIPA 2 until observe horizontally and vertically spaced dots, which are the spectra of the single-mode laser. Tune both VIPAs to the third or fourth order by finely adjusting the tilt angle (the vertical-tilt degree of freedom for VIPA 1 and the horizontal-tilt degree of freedom for VIPA 2) such that the central four brightest spots are equally spaced (Fig. 7c).

  Caution: Before implementing adjustments, make sure the correct degree of freedom of the tilting is located. Adjusting the wrong degree of freedom will distort the VIPA pattern.

  ? TROUBLESHOOTING

• 29Slightly detach the EMCCD camera from the mounting stand by loosening the screws. Carefully rotate the camera by 45° anti-clockwise such that the two dots previously in diagonal (Fig. 7d). Reattach the camera to the mounting stand by fastening screws. This step allows the Brillouin spectrum to be dispersed along a horizontal axis on the camera, which can be helpful for image processing and binning.

  Caution: The camera is heavy. It is strongly suggested to implement this step with a teamwork of two person.

• 30Use the construction rails and black hardboard to build an optical enclosure for the entire spectrometer. A recommended size is 70 in (length) × 11 in (width) × 13 in (height). Keep one side of the enclosure open to ensure the easy accessibility of the components (especially the VIPAs and masks) in the spectrometer. Cover all the gaps between the enclosure and the table surface with black tape. Cover the enclosure with a blackout fabric (9 ft × 5 ft).

  Critical step: The EMCCD is a main heat source and itself needs air cooling. To get rid of the effect of thermal fluctuation on the stability of the spectrometer, the EMCCD should be placed outside of the enclosure. In case the temperature fluctuation is a concern, the temperature inside the enclosure can be monitored by a thermometer.

Figure 7. VIPA pattern of the Brillouin spectrometer.

(a) Dispersion pattern of the VIPA 2 along horizontal axis. (b) Dispersion pattern of the VIPA 1 along vertical axis. (3) Pattern of the combination of two VIPAs. (d) VIPA pattern after rotating the camera by 45°. The arrows indicate the dispersion axes.

Optimizing the Brillouin microscope, Timing 8–10 h

• 31Close the two masks such that the laser spots just disappear. Remove the mirror installed on top of the microscope. Restore the objective turret to Objective 1 (40x/0.6 NA, Olympus). Place a petri dish containing methanol into the sample holder.
• 32Adjust the orientation of the polarizer 1 to ensure that the laser power at the output of the Objective 1 is at least 10 mW. Enable the gain of the EMCCD camera and set the value of gain to the maximum (i.e. 300). Set the acquisition time to 0.1 s. In case the laser spots are saturating the camera, further adjust the masks to block them.
• 33Adjust the position of the Objective 1 such that the laser beam is focused into the sample. Finely adjust the collar of the Objective 1 to optimize the intensity of the Brillouin signal. Carefully adjust the 3D stage that carries the fiber port to optimize the fiber coupling. Finely adjust the location and tilt angle of two VIPAs to further optimize the intensity of the Brillouin signal and ensure the two Brillouin peaks have balanced intensity.

  Critical Step: Since the pre-alignment of the fiber coupling might be imperfect, high laser power is preferred to generate strong Brillouin signal that can be observed by the camera. The alignment of the fiber coupling without indication of the Brillouin signal should be mostly avoided.

  ? TROUBLESHOOTING

• 34Adjust the spatial filter SF to optimize background suppression. Place a glass-bottom dish containing water into the sample holder. Move the focal plane of the Objective 1 close to the interface between water and dish so that both the Brillouin signal of the water and the back-reflection noise appear. Slowly close the aperture of SF to suppress the background noise until observing the drop of the Brillouin signal. Slightly adjust the 2D translational mount to optimize the noise suppression while maintaining/resuming the Brillouin signal.
• 35Install the apodization filters. Adjust the focal plane of the Objective 1 above the bottom of the dish and record the spectrum (B_spec 1). Then move the focal plane close to the interface between the sample and the bottom of the dish such that some background noise appeared and record the spectrum (N_bgd 1). Install a vertical stage right after VIPA 1. Install Filter 1 on the stage such that its apodization edge is parallel with the entrance window of VIPA 1 (Fig. 3b). Install a horizontal stage right after VIPA 2. Install Filter 2 on the stage such that its apodization edge is parallel with the entrance window of VIPA 2 (Fig. 3b).

  Critical Step: Carefully adjust the position of the Filters 1 and 2 to such that the background noise is largely removed while the loss of Brillouin signal is small, then record the spectrum (N_bgd 2). Resume the Objective 1 to its initial position and record the spectrum again (B_spec 2). The suppression of background noise and loss of Brillouin signal are then quantified by using the recorded spectra (N_{bgd 1}, N_bgd 2) and (B_spec 1, B_spec 2), respectively. With both filters properly positioned, the background noise can be reduced by 10–15 dB while the loss of Brillouin signal is <10%.

• 36Measure the spectral precision and SNR of the spectrometer; confirm it is shot-noise limited curve to verify the optimization of the spectrometer. To quantify spectral precision, repeatedly take the measurements of a standard sample (methanol or water) for 300 times. With 10 mW input power and 100 ms acquisition time, the spectral precision should be within 10 MHz (Fig. 8a & b). In case needed, the long-term stability can be estimated by Allen variance¹³⁴. To verify shot-noise limited operation, quantify the SNR of the Brillouin signal at various power levels and/or acquisition times. Plot the results of SNR versus input energy under log-log scale. Under shot-noise limited operation, the slope should be 0.5 (Fig. 8c). For SNR quantification, 300 frames are recorded at each power level, and the averaged peak intensity as well as its standard deviation of the Brillouin signal are calculated after deducting the background. The SNR is then determined by the ratio of the average to the standard deviation.

  Critical Step: The spectral precision and the shot-noise limited curve are key parameters to quantify the performance of the spectrometer. If the anticipated value is not achieved, Step 11–34 should be carefully revisited.

  ? TROUBLESHOOTING

Preparing samples for Brillouin imaging, Timing 1–3 d

• 37Coat a glass-bottomed petri dish with a thin hydrogel layer whenever possible¹³⁵. Seed live cells into the dish (30,000 cells/dish for NIH 3T3 cells) in 2 ml cell culture media and incubate them for > 12 hours to ensure cells are well attached on the substrate. If needed, stain the cells with fluorescent dyes according to the manufacturer’s instructions.

  Critical Step Preparation of cellular samples may need to be optimized/adjusted depending on the cell line and the specific biological question to be addressed.

Imaging the sample using the Brillouin microscope, Timing 1–8 h

• 38Initialize Brillouin spectrometer. Place a dish containing standard material (e.g. water) on the sample stage. Adjust the orders and coupling of the VIPAs such that the intensities of two Brillouin peaks are optimized and roughly equaled. Adjust the position of two masks such that the laser peaks are properly blocked. Adjust the focal plane of the objective 1 close to the interface between water and glass such that some amount of background noise appears in the spectrum. Carefully adjust the position of two filters to suppress the noise.

  Critical Step: Due to the drifts of laser frequency and the mechanical components, the Brillouin spectrometer should be optimized from day to day. This routine maintenance ensures high quality data can be obtained in experiment.

  ? TROUBLESHOOTING

• 39Calibrate Brillouin spectrometer. Using the calibration arm of the spectrometer (Fig. 3b) to record the Brillouin signals of two standard materials (e.g. water and methanol). Repeatedly record the signal of each material for at least 300 times.

  Critical Step: The calibration ensures the experimental data are not affected by the instrumental drift, thus makes the results obtained from different days comparable. Hence, the calibration process should be implemented whenever the instrumental drift is a concern. For example, when the Brillouin peak has shifted as large as 1 pixel over time or the intensities of two peaks are distinctly unbalanced (ratio < 50%), a recalibration should be implemented after resuming Brillouin signal by adjusting two VIPAs.

• 40Place the dish containing sample from Step 37 into the stage. Switch to bright-field/fluorescent channel by rotating the turret (Filter turret 1) that carries the Brillouin filter cube. Adjust the 2D stage and the focus of the Objective 1 to locate the sample of interest. Save the bright-field/fluorescent image by the CMOS camera. Switch back to Brillouin channel. Adjust the laser power to make sure the Brillouin signal of the sample is strong enough (e.g. peak intensity >10,000 count under EM gain).
• 41Set up the scanning strategy. Use raster scanning mode for 2D mapping. Based on the location and size of the sample, determine the scanning region, step size, scanning plane, and the acquisition time. Save the scanning information for future use in data post-processing.
• 42Start data acquisition of the sample by raster scanning. A straightforward home-built LabVIEW program can automatically save the Brillouin spectrum of each point.
• 43Repeat step 40 – 42 until all the samples are measured.

  ? TROUBLESHOOTING

• 44Re-calibrate the Brillouin spectrometer with standard materials according to Step 39 before finishing the experiments.

Data post-processing, 1–8 h

• 45Calculate the calibration parameters FSR and PR. Load the data file of the calibration from Step 39 and 44. Fit the spectrum of each sampling point with a Lorentzian function to extract the peak distance using MATLAB program. Use Equation (3) to calculate the corresponding FSR and PR with known Brillouin shifts of standard materials (for 660 nm laser source, ν_w = 6.01 GHz, ν_m = 4.49 GHz). The average of all the samplings can be used as the value of FSR and PR for retrieving Brillouin shift of the sample. Example MATLAB codes are provided in Supplementary Data 1.
• 46Retrieve Brillouin shift of the sample. Load the data file of the measurement. First interpolate then fit the spectrum of each pixel with a Lorentzian function to extract the peak distance using MATLAB program. Use the calibrated values of FSR and PR to retrieve the Brillouin shift at each pixel. Save the results into a 1D vector. If there is more than one sample, repeat this step until all the samples are processed. Example MATLAB codes are provided in Supplementary Data 1.
• 47Reconstruct 2D Brillouin image of the sample. In case the data is collected by raster scanning, use the ‘reshape’ function of the MATLAB to convert the 1D vector of Brillouin shifts from last step into 2D Brillouin image.
• 48Analyze mechanical properties of the sample. Extract the Brillouin image of the entire sample or its subregion by co-registering the 2D Brillouin image with the corresponding bright-field/fluorescent image. Use the averaged Brillouin shift of selected region to represent the mechanical property.

Troubleshooting

Troubleshooting guidance is provided in Table 1.

Table 1.

Troubleshooting table.

Step	Problem	Possible reason	Solution
6	Beam after M4 is straight but not in the center of the iris	Position of the M3 is not correct.	Carefully shift M3 along the optical axis without changing the reflection angle. Observe the position of the beam spot. Repeat this process until the beam passes through centers of both irises
9	No light detected at the output port.	Coupling of the fiber is poor	Check the beam quality in front of the back focal plane of the Objective 2 to ensure there is no beam clipping or distortion. Increase the laser power until observe laser light at the output port of the fiber. Then adjust the 3D stage to optimize the coupling. In case no signal was observed at maximum laser power, start over this step to make sure the beam position was not altered after installing Objective 2.
19	No light at the output of the VIPA	The tilt angle of the VIPA is too small or the beam is not on the entrance window	Increase the tilt angle of the VIPA to make sure the reflection angle of the incoming focused beam is large enough to be coupled into VIPA’s cavity. Meanwhile, carefully adjust the horizontal translation stage to make sure the beam is on the entrance window.
20	Spectral line is not leveled or sharp	Vertical-tilt degree of freedom of the VIPA is misaligned	Place a target card at the output of the VIPA to check if the pattern is tilted. If so, adjust the vertical-tilt degree of freedom of the VIPA to make them aligned. The good alignment will ensure a sharp pattern at the focal plane of the lens SL2.
27	Spectral line is not straight or sharp	Horizontal-tilt degree of freedom of the VIPA is misaligned	Use the same solution as #20.
28	Spectral dots are not circular	SL2 is slightly misaligned	Slide the VIPA 1 out of the path, readjust the position of SL2 such that the pattern of the VIPA 2 is sharp. Then slide in the VIPA 1 to get spectral dots. If it is still not circular, slightly adjust the position of C2.
33	No Brillouin signal is observed	Position of the fiber port is misaligned	Increase the laser power and the acquisition time of the EMCCD camera until Brillouin signal can be observed. Then adjust the 3D stage to optimize the signal.
	Background noise is too high	Back-reflection from optical component	Move away the sample dish. If noise disappears, clean the bottom of the dish, or change to a new dish. If not, check if there is any back reflection of optical components on either the measurement or the collection arm. This can be implemented by using a dark paper card to block the beam path of each component sequentially. Once locate the component, slightly tilt it to remove the back reflection.
36	Result is not shot-noise limited	Stray light from ambient or sample	Adjust the focus plane of the Objective 1 far away from the interface to avoid any reflection light. Carefully check the optical enclosure to make sure there is no stray light from ambient environment.
38	Brillouin peaks show quickly shifting	Frequency of the laser is drifting abnormally.	Ensure the room temperature is (20–25 °C) and stable. Avoid placing laser close to the vent of the room. Ensure the laser is fully warmed up before experiment (> 30 min at least). Check the temperature of the laser head to make sure it is properly cooled.
43	Brillouin peaks show slowly shifting over time	Frequency of the laser is drifting normally.	More frequent calibrations should be implemented. Follow the guideline regarding periodical calibration in section ‘Brillouin spectrum and calibration’.

Timing

• Steps 1–10, setting up the microscope and add-on optics: 2–4 d

• Steps 11–30, building the Brillouin spectrometer: 3–5 d

• Steps 31–36, optimizing the Brillouin microscope: 8–10 h

• Steps 37, preparing samples for Brillouin imaging: 1–3 d

• Steps 38–44, imaging the sample using the Brillouin microscope: 1–8 h

• Steps 45–48, data post-processing: 1–8 h

Anticipated results

Figure 9 shows the typical Brillouin images of live 3T3 cells in both control and treated conditions. The co-registration of the Brillouin image and the bright-field/fluorescent image can be implemented in either ImageJ or MATLAB. As the Brillouin image usually has lower resolution than the bright-field/fluorescent image, the latter needs to be resized to match the former before co-registration. The subcellular regions can then be outlined by overlaying two images. With the co-registered bright-field/fluorescent image (left panels in Fig. 9a and 9b), Brillouin shifts of subcellular regions (i.e., cytoplasm and nucleus) can be extracted and quantified (Fig. 9c). Generally, the nucleus has higher Brillouin shift than the cytoplasm. After treatment with cytochalasin D, which is used to depolymerize the actin filament, the Brillouin shift of both the cytoplasm and nucleus decreased.

Figure 9. Brillouin image of live 3T3 cells.

(a) Cell in control condition. (b) Cell with the treatment of actin depolymerization drug (Cytochalasin D). In each subfigure, the left and the right panel represents the bright-field/fluorescent image and Brillouin image, respectively. The blue shadow and the dashed curves indicate the nucleus. The scale bar is 5 μm. (c) Brillouin shifts of subcellular regions (i.e., nucleus and cytoplasm) of cells from both control and treated cells. Dots represent pixels, and values represent the mean ± standard deviation.

Supplementary Material

Supplementary material

Supplementary Data 1: MATLAB example codes and raw data.

Supplementary Data 2: technical drawing of the mount for Brillouin filter cube.

Supplementary Data 3: technical drawing of mounting base for laser head.

Supplementary Data 4: technical drawing of mounting stand for EMCCD camera.