A multi-modal microscope for integrated mapping of cellular forces and Brillouin scattering with high resolution

Mechanical forces and stiffness play key roles in the health and development of cells and tissue, but despite the physical connection between these quantities, they cannot be monitored in parallel in most cases. Here, we introduce a fully integrated microscope that combines a method for high-resolution cell force imaging (elastic resonator interference stress microscopy, ERISM) with non-contact mapping of the elastic properties of cells (via Brillouin microscopy). In order to integrate both techniques, we had to account for the strong back reflection on the surface of the microcavity used for ERISM measurements as well as the local destruction of the cavity under illumination for Brillouin microscopy measurements. Therefore, we developed an elastic optical microcavity with minimal absorption that can perform ERISM measurements without sustaining laser damage during Brillouin microscopy. Furthermore, an unequal-arm Michelson interferometer was designed to suppress the back reflection of the laser on the ERISM microcavity surface using division by amplitude interference to reduce the reflected light and enhance the Brillouin signal. We show the utility of our integrated microscope by simultaneously mapping cellular forces and Brillouin shifts in cultures of fibroblast cells.


Introduction
Recent developments in mechanobiology have repeatedly shown the importance of physical properties in cell and tissue development and in disease mechanisms [1,2].Cells can adapt their response and signalling pathways to their environment based on mechanosensing [3][4][5][6] which on a larger scale impacts tissue organization as well as the onset and progression of diseases [7][8][9].Measuring and understanding the interplay between different physiological properties remains a complex task, in part due to the lack of methods to monitor the different properties of interest in parallel.Recent developments of new materials [10][11][12][13] and their analysis [14,15] have improved the quality and increased the number of possible applications of optical techniques in the last few years.
A variety of innovative techniques have emerged to quantify the key mechanical properties of cells and tissue, i.e. the forces exerted by cells on their environment and the viscoelastic properties of cells and tissue.To measure force, traction force microscopy [16][17][18], elastic micropillars [19,20] and FRET based tension sensors [21][22][23] are the most ubiquitous and widely accepted methods.With elastic resonator interference stress microscopy (ERISM), we introduced a new, intrinsically non-invasive technique to detect cell forces based on optical interference [24][25][26][27][28].When illuminated with collimated, monochromatic light, optical interference occurs within an elastic microcavity, with the position of the interference depending on the local thickness of the cavity and the illumination wavelength.Cells can then be cultured on top of the cavity, with any exerted forces resulting in a thickness change of the cavity and thus a change in the interference pattern.
By scanning through multiple wavelengths and fitting the resulting data with an optical model, an accurate deformation map can be created, which provides a measure of the mechanical force exerted on the cavity surface.Key advantages of ERISM are that it does not require a zero-force reference image, shows low phototoxicity, high spatial resolution and exceptional measurement precision.Due to these qualities, ERISM enables long-term imaging of living cells for up to several weeks.
To determine information about stiffness and viscosity, atomic force microscopy (AFM) [29,30], mechanical rheology [31], micropipette aspiration [32,33], optical stretchers [34] and microfluidic methods [35] are frequently used [36].However, most of these techniques function by exerting forces on the sample and thus require physical contact, hence they are invasive and can affect the behavior of cells.By contrast, Brillouin microscopy is an all optical technique that enables non-invasive, label-free probing of viscoelastic properties at high resolution [37,38].It is based on inelastic scattering of light by thermally induced acoustic phonons with high frequency (typically in the GHz range).The position of the peak in the resulting Brillouin spectrum is related to the real part of the longitudinal elastic modulus of the investigated material.Using a scanning confocal setup, it is thus possible to obtain microscopic 3D maps of the local elastic properties of a sample.Therefore, Brillouin microscopy is utilized in various applications, e.g. to study cells [7,39], tissue [40][41][42] and embryo development [43,44].To gain a better understanding of the complex functioning of these systems, Brillouin microscopy has recently been combined with other methods like optical coherence tomography [45,46], optical diffraction tomography [47], fluorescence microscopy [48] and optical tweezer based microrheology [7].However, to the best of our knowledge, there has only been one report, from 2019, on combining Brillouin microscopy with cell force microscopy [49].This work investigated fixed cells on a micropillar array, and recorded stiffness and force on separate setups.
Here, we present a fully integrated setup that combines ERISM and Brillouin microscopy providing high resolution maps of both cellular force and intracellular stiffness.Several modifications of the two modalities were made to ensure their mutual compatibility.In particular, a non-absorbing ERISM microcavity featuring ZrO 2 based dielectric reflectors was developed to prevent thermal laser damage to the cavity during Brillouin imaging.Additionally, an unequal-arm Michelson interferometer was utilized to minimize bleed-through of light reflected from the ERISM microcavity into the Brillouin spectrometer by utilizing self-interference.Both innovations have implications for ERISM and Brillouin microscopy beyond the multi-modal instrument presented here.To demonstrate the unique capability of our combined microscope, we show integrated measurements of cell forces and cell stiffness using cultures of live fibroblasts.

Background
The measurement principle of ERISM is described in detail by Kronenberg et al [24] Briefly, the ERISM sensor is formed by an elastic microcavity consisting of a layer of a soft elastomer sandwiched between two ultrathin mirrors.Cells cultured on top of the microcavity surface exert forces and deform the upper mirror and thus alter the cavity thickness.Optical interference in the cavity is read out under epi-widefield illumination with monochromatic light.If the optical path difference of light in the cavity nd (n, effective refractive index of the cavity interior; d, cavity thickness) corresponds to an integer number of wavelengths mλ, constructive interference occurs, i.e. light is transmitted.Accordingly, if the path difference corresponds to a half integer number of wavelengths, destructive interference occurs and the light is reflected This results in an interference pattern on the sensor that depends on the local profile in cavity thickness.By scanning the wavelength of illumination, a spectrum of reflectance at each point is obtained.The resonance wavelengths are extracted from this spectrum and compared to an optical model of the cavity.This model provided the reflectance spectrum for the microcavity as function of the elastomer layer thickness calculated using a transfer matrix method simulation and such the comparison of experimental and modelled data allowed to determine the expected reflectance spectrum of the microcavity as a function of the elastomer layer thickness.From this, the local cavity displacement at each point in the image is determined, which can then be related to a map of mechanical stress exerted by the cells.
Spontaneous Brillouin microscopy measures the inelastic scattering of light from thermally induced density fluctuations in the sample (see figure 1(c)).The frequency shift ν B of the scattered light depends on the local mass density ρ, the refractive index n and the longitudinal elastic modulus M ′ of the material In a backscattering configuration (as used in our work), the scattering angle θ equals 180 • and the magnitude of the frequency shift is maximal.Under the assumption that the ratio of n √ ρ does not vary significantly across the biological sample, the Brillouin shift is then directly related to the longitudinal modulus and is therefore a measure for the local stiffness of the sample [38,39,50].The Brillouin frequency shifts of cells for green illumination are typically in the range of ∼7.5-8 GHz, which is close to but clearly distinguishable from the Brillouin frequency shift of water (7.46 GHz).Compared to elastic moduli in the Hz to kHz range that are usually probed in biological samples, the Brillouin shift provides information about elastic properties in the frequency region of a few GHz and hence probes mechanical properties on sub-nanosecond timescales.Higher frequency shifts correlate with higher sample stiffness/longitudinal elastic modulus, because acoustic phonons travel faster in media with a higher stiffness, which in turn leads to larger Brillouin frequency shifts of scattered laser light from the faster phonons.

ERISM microscope and data analysis
A diagram of the ERISM setup used in this work is given in figure 1(a).Illumination consisted of a white light halogen source combined with a scanning monochromator (CM110, spectral products).Light exiting from the monochromator slit was focused onto a pinhole for spatial cleaning.After recollimation, the light was coupled into an inverted microscope (Eclipse Ti, Nikon) and passed through a 550 nm long pass dichroic mirror that combines its path with that of the laser used for Brillouin microscopy (see below).The illumination light was focused onto the back aperture of the objective using a 150 mm focal length lens, such that collimated, widefield illumination of the microcavity was achieved.A 10× objective (Plan Fluor 10× NA 0.3, Nikon) was used to maximize fringe contrast during the measurement.A 50:50 beamsplitter placed in the beam path under the objective first reflected the illumination light towards the microcavity and then transmitted part of the light reflected by the microcavity towards an sCMOS camera (Zyla 4.2, Andor).Camera and monochromator were synchronized via custom LabView software.For a typical measurement, the monochromator was scanned from 560 nm to 765 nm in 1 nm steps (i.e.206 different wavelengths) and a reflection image of the ERISM microcavity was acquired by the camera for each wavelength increment.The range of illumination was chosen to avoid reflection from the dichroic mirror used for the Brillouin illumination (see section 2.3) and to obtain a precise measurement of the cavity thickness by ensuring that multiple resonant modes are covered within each ERISM scan.Details on the considerations that go into selecting a suitable range of illumination wavelengths for ERISM are given by Liehm et al [25].
Interference images corresponding to each wavelength were stored as a sequence of TIF files, which were then analysed as a single stack to extract the spectra at each pixel and determine the cavity thickness using the resonant mode positions.The image analysis was performed using custom-written Python-code, which compared detected resonance positions in the experimental spectra to a lookup table of resonances and associated cavity thickness produced using a transfer matrix method simulation of the same microcavity structure.The transfer matrix method calculates the transmittance and reflectance of the microcavity as a function of cavity thickness and wavelength of illumination.The key physical parameters that enter the transfer matrix calculation are the refractive index (n) and extinction coefficient (k) as a function of wavelength for each layer/material forming the cavity.These values are determined using ellipsometry measurements on thin films of the respective materials.The final displacement maps were then calculated by subtracting a background from the thickness maps.The accuracy of the ERISM setup to measure deformations of the microcavity with nanometre precision was validated in previous publications using AFM [24].As the cavity displacement is proportional to the forces and stresses exerted on its surface, displacement is frequently used as a proxy for cellular forces.

Brillouin confocal microscope and data analysis
Confocal Brillouin measurements (see figures 1(a) and (b)) were performed with a 532 nm single-mode, mode-locked laser with vertical polarization and a beam waist of 1.54 mm (Torus 532, Laser Quantum).The choice of laser wavelength was guided by the availability of lasers with high spectral stability and the higher Brillouin scattering signal at this wavelength relative to the red and near infrared part of the spectrum.After passing through a Faraday Isolator (IO-5-532-HP, Thorlabs) to prevent back reflection of light into the laser cavity, the power of the beam was adjusted using a rotating half-waveplate and a polarizing beamsplitter.A further 90:10 nonpolarizing beamsplitter reflected 10% of the laser light onto the optical path of the inverted microscope and passed Brillouin scattered light to the spectrometer (see below).
The laser beam was spatially cleaned and expanded to a diameter of 7.51 mm with a telescope consisting of an objective (DIN 10× NA 0.25, Edmund Optics), an f = 75 mm achromatic doublet lens, and a 20 µm diameter pinhole.The expanded beam was combined with the light for the ERISM scans using a periscope and the 550 nm long pass dichroic mirror mentioned above.The 150 mm lens used for the ERISM illumination was removed from the optical path for Brillouin measurements, so that the laser beam was reflected up towards the objective lens using a further 550 nm long pass dichroic mirror placed on top of the 50:50 beamsplitter used for ERISM illumination.Brillouin measurements were performed with a 40× objective (S Plan Fluor ELWD NA 0.6, Nikon) to tightly focus the beam.
Backscattered Brillouin light was collected with the same objective and sent back along the laser path.The fraction of light transmitted through the 90:10 beamsplitter was passed to a Michelson interferometer set up such that the Brillouin signal interfered constructively while Rayleigh scattered and reflected laser light interfered destructively.For details on interferometer alignment, see section 2.4.Finally, the signal was coupled into a single-mode fibre (P3 460B FC 5, Thorlabs) using an aspheric collimating lens.
On the far-side of the fibre, a further aspheric collimating lens (F110APC 532, Thorlabs) was used to form a collimated beam with a waist diameter of 1.14 mm.This relatively narrow beam width maximized the signal to noise ratio by centring the signal intensity over two central Brillouin modes, rather than spreading it over a larger number of angular orders.To separate the different spectral components of the signal, it was passed through a cross-axis virtually-imaged phased array (VIPA) configuration equipped with slits, apodization filters and Lyot stop, similar to the setup described by Zhang and Scarcelli [51] (see supplementary note 4).The final spectrometer setup featured an overall extinction of ∼130 dB.The spatially separated spectral components of the signal were recorded by an EM-CCD camera (iXon Ultra, Andor).
To calibrate the Brillouin shift to pixel numbers on the EM-CCD camera, a sample of water was measured right before each experiment.Additionally, a measurement of the free spectral range was performed with open spectrometer slits at reduced exposure time and camera gain.To record Brillouin maps, the focused 532 nm laser beam was raster scanned relative to the sample by moving the sample in 2.5 µm steps with a motorized microscope stage (H117, Prior Scientific Instruments) and acquiring a Brillouin spectrum at each position.Raster scanning and triggering of the EM-CCD camera were controlled and automated with the ImageJ plugin µManager.
A diagonal line profile with a width of 4 pixels was drawn through the Stokes and anti-Stokes Brillouin peaks in the first EM-CCD image.A Gaussian was fitted to each peak in the resulting profile, and the separation between the centres of the Gaussians was recorded.For each subsequent EM-CCD image, an automated script extracted a line profile from the image along the same line and repeated the fitting procedure.This yielded the separation of the Stokes and anti-Stokes Brillouin peak for each EM-CCD image and hence for each position across the sample.Using the calibrated free spectral range and Brillouin shift of water, the pixel-separations of the Stokes and anti-Stokes peak were then converted to absolute Brillouin shifts.For the final maps of Brillouin shift, a two times interpolation was applied to the raw data to expand their size and thus allow a better side-by-side comparison to the corresponding ERISM maps.

Control of Michelson interferometer
Control of the path difference in the Michelson interferometer was achieved using four piezo chips (PA4GEW, Thorlabs), with two chips glued behind each interferometer mirror.Voltage was provided to each pair of chips via an analogue driver board (USB-6001, National Instruments) that was controlled via simple, semi-manual software (LabView, National Instruments).Varying the voltage in mV steps provided nm-scale control over mirror displacement.Mirrors were adjusted to maximize rejection of Rayleigh scattered and reflected laser light.Generally, the interferometer remained stable for well over ten minutes, indicating that the interferometer path length difference did not drift significantly during the acquisition of a Brillouin map.The interferometer was carefully adjusted prior to each measurement to ensure maximal signal intensity.For more information on the theory, setup and alignment of the interferometer, see supplementary notes 1 and 2.

Integrated measurements on NIH-3t3 cells
Integrated ERISM and Brillouin measurements were performed in sequence.For the 206-wavelength ERISM scan, an exposure time for each wavelength of 100 ms was used, resulting in a total measurement time of approximately 20 s.ERISM scans were performed with the 10× objective and with the f = 150 mm lens in place (see 2.2).
Brillouin measurements were performed at a laser power of 4-6 mW, an EM-CCD camera gain of 300 and an exposure time of 200-300 ms per point to maximise the Brillouin signal without significantly increasing the total measurement time or inducing photodamage to the cells.The slightly longer exposure time relative to what is reported for similar Brillouin spectrometers in the literature [39,[51][52][53] was chosen to account for the reflection from the ERISM microcavity mirrors, which reduce the amount of laser light reaching the cells on the top surface.Brillouin measurements were performed with the 40× objective and the f = 150 mm lens removed from the optical path (see 2.3).
In addition to recording ERISM and Brillouin data, normal microscopy images were captured for both the 10× and the 40× objectives using the sCMOS camera also used for ERISM measurements to identify the cell outlines and asses the health of the cells, in particular regarding any effect of the Brillouin measurements.

ERISM microcavity fabrication
ERISM microcavities were fabricated on No #5 thickness glass substrates with dimensions of 24 × 24 mm 2 .Following extensive cleaning, a 100 nm thick ZrO 2 bottom mirror was prepared by atomic layer deposition (ALD; Savannah S200, Ultratech).The ALD precursors used were TDMAZr (Pegasus Chemicals) and H 2 O; the reaction was performed at 80 • C. The pulse sequence used was 0.3 s of TDMAZr, followed by a 7 s purge with N 2 at 20 sccm, 0.03 s pulse of H 2 O, and a final purge of 7 s with N 2 .This sequence was repeated for a total of 555 cycles to create a ZrO 2 layer with an approximate thickness 100 nm.
Next, the two precursors of an ultra-soft silicone-based elastomer (Gel8100, Nusil) were mixed in equal parts.The resulting mixture was spin-coated onto the ZrO 2 mirror at 3000 rpm for 60 s to create an approximately 8 µm thick film.The film was crosslinked on a hotplate set to 125 • C for one hour.The hydrophobic surface of the elastomer was oxidized using an oxygen plasma in a sputter chamber (NexDep, Angstrom Engineering) to create a hydrophilic surface that promotes adhesion of the final ZrO 2 mirror and prevents the formation of nano-islands.The process for elastomer deposition and oxygen plasma treatment followed the procedure developed by Kronenberg et al [24], with the exception that the oxidation time was increased to 60 s.
Finally, a top ZrO 2 mirror was deposited at an ALD reactor temperature of 50 • C, which was necessary to prevent wrinkling of the microcavity due to the thermal expansion coefficient mismatch between the elastomer and the ZrO 2 mirror (see supplementary note 5).[54] As the TDMAZr precursor becomes significantly more adherent at this low temperature, [55] an extended 60 s N 2 purge was used to remove excess precursor between each TDMAZr and H 2 O pulse to create the ZrO 2 top mirror in a total of 60 cycles.
To allow culturing of four different cell cultures on a single ERISM microcavity, a four-well silicone structure (prepared from removable 12 well chambers, ibidi) was placed on top of the microcavity.

Cell culture
To create a suitable surface for cells to adhere to, the surface of the top mirror of the ERISM microcavity was coated with fibrous collagen.A 100 µg ml −1 collagen solution was created by mixing a 10 mg ml −1 collagen 1 stock solution (354 249, Corning) in cell medium composed of Phenol red free DMEM, supplemented with 10 vol% Fetal Bovine Serum, 1 vol% penicillin/streptomycin solution, and 1 vol% 100× concentration Glutamax.Neutral instead of acidic buffer solution was used to avoid the formation of monomeric collagen and instead obtain fibrous collagen which led to better cell adhesion and correspondingly stronger displacement of the microcavity.400 µl of the collagen solution were added to each of the four silicone wells and incubated at 37 • C for four hours before washing the chambers with cell medium.
NIH-3t3 cells were trypsinised from a cell culture flask, centrifuged and suspended in fresh medium (same medium as described above).Approximately 1000 NIH-3t3 cells were seeded in each chamber of the microcavity and cells were incubated at 37 • C for two days prior to performing the combined ERISM and Brillouin measurements to ensure strong adhesion of the cells while simultaneously avoiding confluency to enable individual cell measurements.

Integrated Brillouin and ERISM microscope
To make combined measurements of cellular force and stiffness, both ERISM and Brillouin microscopy were integrated on an inverted microscope (figure 1(a)).While ERISM works most efficiently as a wide-field imaging modality, Brillouin maps are formed by confocal raster scanning of the sample.Combining an ERISM measurement with a scanning Brillouin measurement of the same cell thus requires focusing the Brillouin laser just above the microcavity top mirror (figure 1(b)).
The measurement setup for ERISM consisted of a white light source, scanning monochromator and fast scientific-CMOS camera and was largely identical to the setup developed by Kronenberg et al [24]; however, the microcavity had to be modified (see below).The Brillouin confocal setup was of a similar design to that described by Zhang and Scarcelli [51], comprising a single-mode stabilized 532 nm Brillouin laser and a spectrometer containing two VIPA etalons placed in a cross-axis configuration [56].In addition, an unequal-arm Michelson interferometer was specifically developed for the combined ERISM/Brillouin instrument to further improve the extinction of non-Brillouin light, as described below.Dichroic mirrors were used to combine the ERISM and Brillouin illumination paths and then separate them again after reflection/scattering at the sample.
In addition, several modifications were made to both the confocal and spectrometer part of the Brillouin setup in order to improve light coupling efficiency and reduce alignment time.The alignment of the confocal and spectrometer part of the Brillouin setup could be carried out within a few hours, including the adjustment of the Michelson interferometer (see supplementary note 2) and the alignment of the spectrometer [51].Most importantly, a telescope setup featuring a 20 µm diameter pinhole spatial filter was used to clean and expand the laser and thus overfill the back aperture of the imaging objective.Its positioning after the 90:10 beamsplitter simplified coupling of the backscattered Brillouin light to the single-mode fibre, as it reduced the diameter of the backscattered beam to better match the aperture of available fibre coupling lenses.We found that this placement of the pinhole did not cause any significant loss in the backscattered Brillouin signal as the latter follows the same optical path as the incident laser beam.

Non-absorbing elastic fabry perot etalon microcavities
The principal modification to the original ERISM design was made to the elastic microcavity.In the original implementation, the top and bottom mirrors consisted of 10-15 nm thick gold layers (figure 2(d)) [24].However, as a metal, gold is highly absorbing and in addition, for such thin layers, nano-islands form in the gold film which leads to plasmonic resonances and thus additional absorption around the wavelength of the Brillouin laser [57].This absorption was found to cause rapid heating of the sample during a Brillouin measurement, to an extent that the elastomer in the microcavity deformed [58], which damaged the ERISM sensor and thus interfered with cell force measurements (figures 2(a)-(c)).Therefore, it was necessary to replace the gold mirrors with a non-absorbing material.For this purpose, the high refractive index metal oxide ZrO 2 was chosen (n = 2.2; the imaginary component of the refractive index of ZrO 2 , and hence the material absorption, is smaller than the minimum value that can be reliably resolved by spectroscopic ellipsometry).The reflection in this case was provided by the refractive index contrast at the interface with the elastomer (n = 1.4).Details of the fabrication of the ZrO 2 based ERISM microcavities are given in section 2.6.The major challenge in the development of a process for fabrication of the ZrO 2 based microcavities was the elevated temperature required for the ALD procedure.Due to a mismatch in the thermal expansion of ZrO 2 and the underlying elastomer, the microcavity would wrinkle when cooling it down to room temperature following deposition of the ZrO 2 at 80 • C, which is the standard temperature for ALD based ZrO 2 deposition.To prevent this, a stronger plasma oxidation of the elastomer was used in order to create a mechanical buffer layer before deposition of the metal oxide.Additionally, the temperature for ALD of the ZrO 2 top mirror was reduced to 50 • C which required longer purge times than for the standard 80 • C ALD process.The stiffness of the final microcavity was characterized by AFM.An apparent stiffness in the range of ∼10-25 kPa was obtained (see supplementary note 3), which is within the physiological stiffness range of several types of biological tissue [59].

Division by amplitude interference for Brillouin measurements on a ZrO 2 microcavity
In addition to the issue of laser absorption by the ERISM microcavity, reflection of the Brillouin laser at the reflective interfaces of the microcavity, in particular the reflection from the top mirror surface, represents a further challenge.We found, that when performing a Brillouin measurement, a two-stage VIPA-based spectrometer, even when equipped with slits, Lyot stop and apodization filters, did not provide sufficient extinction to filter this reflection from the Brillouin signal.To circumvent this issue, we introduced an unequal-arm Michelson interferometer as an additional tuneable narrowband self-interference filter.
The Brillouin signal can be many orders of magnitude weaker in intensity than the light from the Brillouin laser that is elastically reflected or scattered by the sample and the microcavity interface, and it differs from this reflected and Rayleigh scattered light by only a few picometers in wavelength.Therefore, discerning the signal from the strong background is a general challenge in Brillouin microscopy.Many innovative solutions have been proposed to improve the extinction of a Brillouin spectrometer, including an iodine absorption notch filter [60], Rubidium gas filter [61], spatial light modulation for enhanced spectrometer apodization [62], background deflection aperture masks in the spectrometer [52], and interference-based approaches involving both etalons [63] and interferometer-based setups [64,65].Antonacci et al used a reference laser beam to destructively interfere the reflection from a specific interface in a Michelson-interferometer-style setup [64], while a division by wavefront-interference method using a prism to split the beam was employed by Lepert et al to filter both reflection and Rayleigh scatter at all planes [65].
For our system, we made use of a division by amplitude Michelson interferometer design to filter both Rayleigh and reflected light from the signal at all planes.Such an approach has previously been suggested as a filtering method for Brillouin spectroscopy studies [66], however, has not yet been applied to Brillouin microscopy.Filtering light at all planes is particularly relevant considering the ZrO 2 top mirror is deformed by hundreds of nanometres by cellular force and hence is not located in a fixed and defined plane.Furthermore, there is a second bottom mirror placed approximately 8 µm below this; having a strong background signal originate from at least these two planes makes a reference-beam based approach unfeasible.The advantage of using division by amplitude compared to division by wavefront, is that the system is immune to spatial incoherence in the scattered light, and so an additional single-mode fibre before the interferometer is not required to spatially filter the beam, i.e. the interferometer can be placed before the fibre coupling to the spectrometer.
The design of the interferometer consisted of two mirrors custom mounted on piezo chips, the expansion of which could be controlled in nanometre steps by applying an adjustable potential difference across the chips (section 2.4, figure 3(a)).The alignment and coupling of each interferometer mirror into the single-mode fibre is straightforward and quick to perform by coupling a fibre reference beam from the opposite end of the fibre to highlight the fibre mode (supplementary note 2).The macroscopic path difference between the two interferometer arms was adjusted to create a free spectral range for the interferometer that was approximately twice the Brillouin shift of water (7.46 GHz), so that constructive interference of the Brillouin signal overlapped with destructive interference of the laser.Both, the Stokes and anti-Stokes components of the Brillouin signal are separated from the laser signal by the same distance in frequency space, so for the correct path length difference they both interfere constructively when the laser signal undergoes destructive interference (figure 3(b), supplementary note 1).
With a fine adjustment of the path length difference in the interferometer, an extinction of up to ∼30 dB of Rayleigh and reflected light from the Brillouin signal was achieved before coupling the signal into the single-mode fibre (coupling efficiency ∼50%-60%).Figures 3(c) and (d) demonstrate the effect of applying the interferometer when using a ZrO 2 based ERISM microcavity as the substrate for cells.To ensure the maximum possible extinction of the reflected light and best possible throughput for Brillouin signal, the interferometer was adjusted prior to every measurement (section 2.4, supplementary note 2).Combining the interferometer with the complete two-stage cross-axis spectrometer resulted in an instrument extinction rate of approximately 130 dB which proved sufficient to filter the reflected and Rayleigh scattered light to a level where the Brillouin signal could be detected.

Integrated measurement of NIH-3t3 fibroblast cells
With the interferometer in place, integrated ERISM and Brillouin measurements of cells cultured on a ZrO 2 based microcavity were now possible.Figure 4 shows proof of principle measurements of NIH-3t3 fibroblasts, a cell line that is well characterized in terms of the forces they exert on their surroundings [67,68].The ERISM measurements are free of artefacts and resolve deformations of the microcavity down to the nm-scale, similar to ERISM measurements recorded on conventional ERISM microcavities.In particular, the ERISM maps clearly show points of contraction and regions where pulling and pushing forces are exerted (figure 4(b)).The Brillouin images obtained for the same cells and on the same substrate and with the same microscope are of a high contrast and free of reflection artefacts.The Brillouin images show a region of increased stiffness within the cell interior (figure 4(c)).This region roughly aligns with the expected location of the cell nucleus, which is anticipated to be the stiffest part of the cell [69].To compare ERISM and Brillouin data from multiple cells, we first computed the mean ERISM displacement and the mean Brillouin shift for each cell.For ERISM, we take the average downward displacement underneath the cells (including all pixels with a negative indentation and an absolute value >20 nm, which ensures non-uniformity in cavity surface and fluctuations at the cell periphery are reliably avoided).In a similar manner, for the Brillouin shift, the image is thresholded at a shift of between 7.50 and 7.56 GHz for each cell to remove regions with culture medium measured rather than the cell, and the mean shift is then calculated across the cell area.Comparing the resulting mean ERISM and Brillouin data, cells expressing higher forces have a lower internal stiffness (figure 4(d)).A further example of an integrated measurement, this time for cell division, is shown in supplementary note 6.As far as we are aware, these images represent the first demonstration of combined force and stiffness measurements on live cells that were performed on a single integrated microscope.

Conclusion
In this study we demonstrated the successful combination of ERISM and Brillouin microscopy for integrated measurements of cell force and cell stiffness.Our results demonstrate that ERISM and Brillouin microscopy are overall well suited for cross-integration, in particular due to the robustness of ERISM (e.g.no requirement for zero-force reference images) and the unique non-contact nature of Brillouin-based stiffness measurements.However, the integration still required adapting both techniques; this involved the creation of a new form of ERISM microcavity with ZrO 2 metal oxide mirrors that do not absorb the Brillouin laser light and the use of division by amplitude-interference in a Michelson interferometer to efficiently filter Rayleigh and reflected laser light while allowing the Brillouin signal to pass.With these improvements, we obtained high resolution Brillouin microscopy images of the stiffness profile of NIH-3t3 fibroblasts while also imaging the forces they exert via ERISM.The anti-correlation we found between the mean Brillouin shift and the mean substrate indentation generated by individual cells is in contrast to the weak positive correlation reported by Coppola et al [49], in a study that used micropillar arrays to estimate forces exerted by fixed cells.Given that the nucleus is the stiffest region of the cell, a change in its mechanical properties will strongly impact the mean stiffness of the cell.With this in mind, a possible explanation for the observed anticorrelation might be nuclear softening in response to cell forces.Prior research has shown that on stiffer substrates, where cells would exert stronger force, the nucleus is stretched [70] due to the mechanical anchoring between the actomyosin cytoskeleton and the nuclear envelope provided by the LINC complex protein [71].Additionally, stretching the nucleus using external force is known to lead to a softening effect as a protective measure against genome damage [42,72].Combined, the above points may explain the observed decrease in Brillouin shift with increasing cell force as nuclear stretching and concomitant softening due force exposure.However, we stress that further studies are required to further explore and validate or dispel this conjecture.
In the future, our combined instrument can prove useful to perform measurements of changes in cell forces and the stiffness of cellular components under a range of different stress conditions, e.g. in neurons [73,74].Another field where our integrated measurement system adds relevant capability is in the study of cancer cells, which are known to perform intricate mechanosensing and to adapt their behaviour, mechanical properties, and architecture accordingly [7,75,76].Furthermore, it is known that tumour formation itself is strongly dependent on the tumour microenvironment [77].There have already been several efforts to utilize Brillouin microscopy [7,78,79] as well as cell force measurements [80,81] to study the mechanobiology of cancer on the cellular level.Finally, our integrated ERISM/Brillouin microscope may prove helpful for screening the behaviour of cells under various conditions, e.g. when introducing drugs, different media or changing the mechanical stiffness of their substrate.
In addition to their direct application for integrated ERISM/Brillouin microscopy, the innovations described here create new possibilities for ERISM and Brillouin microscopy.For instance, an ERISM microcavity that does not absorb light will also be useful for combining ERISM with other imaging and manipulation modalities that require a high light intensity, such as more efficient light collection in epi-fluorescence microscopy, compatibility with two-photon [82] and light sheet [83] microscopy, and optical tweezers [84].Further specific applications of the new ERISM microcavities might involve combining ERISM force measurements with intracellular laser based cell tracking and sensing [85][86][87] and integrating ERISM with optogenetics, i.e. light driven activation of genetically encoded ion channels in cells, usually neurons [88] and cardiac cells [89].ERISM measurements have already been used in conjunction with cardiomyocytes [28] and neuronal growth cones [25].A combination of ERISM with optogenetics would allow to probe the change of cellular forces in response to controlled changes in cellular behaviour that are evoked by targeted exposure to light.
Combining the conventional two-stage cross-axis Brillouin spectrometer with the Michelson interferometer resulted in an instrument with exquisite extinction that is uniquely suited to perform measurements in environments with significant amounts of Rayleigh scattering and reflection, potentially well beyond the current application of imaging cells on a partially reflective ERISM microcavity.This might apply, for example, to recent efforts to use Brillouin microscopy to characterize certain membrane-less cell-organelles, such as the stress granules forming in mutated HeLa cells [47,52].A further area where additional extinction of Rayleigh scattered light would be useful is for measurements on tissue slices or whole animals where propagation of the laser trough the sample results in significant Rayleigh scattering.

Figure 1 .
Figure 1.Integrated ERISM and Brillouin microscope.(a) Schematic of the integrated microscope.The ERISM light path is shown in red with the illumination and imaging units marked by dashed boxes.The Brillouin illumination and detection paths are shown in green.The two illumination paths are combined using a 550 nm longpass dichroic mirror.To focus the Brillouin laser on the sample, the lens used for widefield ERISM illumination is removed prior to Brillouin imaging.The ERISM and Brillouin signal are collected by the same objective and directed to their respective detection paths by a 50:50 beamsplitter (ERISM) or another 550 nm longpass dichroic mirror and a 90:10 beamsplitter (Brillouin), respectively.(b) Close up of the illumination on the ERISM microcavity.Under monochromatic wide-field illumination of the sample, interference forms inside the cavity, with the position of fringes depending on the cell induced deformations of the cavity.To record Brillouin maps from the cells of interest, the Brillouin laser is focused to a point slightly above the top surface of the ERISM microcavity.(c) Principle of Brillouin scattering.The incident laser light is inelastically scattered by thermally induced phonons within the sample.Depending on the direction of the phonons, the backscattered light has a higher (anti-Stokes) or lower (Stokes) frequency.The frequency shift is symmetric and depends on the high frequency elastic longitudinal modulus of the material.

Figure 2 .
Figure 2. Optimization of the ERSIM microcavity for compatibility with Brillouin microscopy.(a) Brightfield microscopy image of the top surface of a conventional ERISM microcavity with gold mirrors after exposure to the focused beam of the 532 nm Brillouin laser at different laser powers.The exposure was either pulsed (100 ms, using a shutter to block the beam in between exposures; for 2 mW, 5 mW) or continuous (for 10 mW), with the later condition resulting in the most dramatic damage.(b) Displacement map of the 2 mW region marked by the black square in (a) obtained by performing a standard ERISM measurement on the microcavity after exposure to the Brillouin laser.(c) Profile plot along the black line in (b), indicating a permanent deformation of the microcavity by about 80 nm caused by the Brillouin laser.(d) Schematic of the conventional ERISM microcavity and the new microcavity design.For the latter, the bottom and top mirrors are created via atomic layer deposition of the high refractive index transparent metal oxide ZrO2.(e) Phase contrast microscopy image of a NIH-3t3 fibroblast cell cultured on a ZrO2 based ERISM microcavity, with reflection from the 532 nm Brillouin laser focused on the cavity surface partially visible at the top left of the image.(f) A displacement map of the cell in (e) obtained by an ERISM measurement, with the cell outline overlaid as a black dotted line.Scale bars, 50 µm (a) 5 µm (b), 20 µm (e), (f).

Figure 3 .
Figure 3. Michelson interferometer for filtering of Rayleigh and reflected light from Brillouin signal.(a) Schematic of the Michelson interferometer used to filter the Rayleigh/reflected laser light from the Brillouin signal before coupling the signal into the single-mode fibre leading to the spectrometer.The optical path length difference between the two arms is precisely adjusted with piezo mirrors, such that the Brillouin signal interferes constructively while the laser signal undergoes destructive interference.(b) Illustration of the normalized intensity of Stokes and anti-Stokes Brillouin signal and Rayleigh laser light as they exit the Michelson interferometer, shown as a function of the path length difference between the two mirrors.(c) 2D raw Brillouin spectra recorded by the EM-CCD camera of the Brillouin spectrometer with and without (w/o) the Michelson interferometer in place and with the Brillouin laser focused to a point immediately above the top surface of a ZrO2 based ERISM microcavity.(d) Profile plots along the red diagonal lines in (c), showing the effect of image saturation caused by reflection from the ERISM microcavity (red), and the subsequent removal of this reflection through the Michelson interferometer (grey).

Figure 4 .
Figure 4. Combined ERISM and Brillouin measurements.(a) Brightfield microscopy images of a NIH-3t3 fibroblast cells on ZrO2 based ERISM microcavities, with the 532 nm laser used for Brillouin microscopy visible next to the cell in each image.(b) ERISM displacement maps for the cells in (a), showing several points of force exertion at the cell periphery and the characteristic indentation from the counterbalancing force under the cell.(c) Maps of Brillouin shift obtained by confocal scanning for the cells in (a), showing the largest Brillouin shifts towards the centre of each cell, consistent with earlier reports in the literature [42, 47].(d) Mean indentation taken from the ERISM displacement maps (threshold −20 nm) versus the corresponding mean Brillouin shift (threshold 7.50-7.56GHz) for n = 11 cells.A variable Brillouin shift threshold was used to exclude outliers outside the region of the cell.Red line represents a linear fit (R 2 = 0.48).Scale bars, 20 µm.