Visible spectrum extended-focus optical coherence microscopy for label-free sub-cellular tomography
Paul J. Marchand, Arno Bouwens, Daniel Szlag, David Nguyen, Adrien Descloux, Miguel Sison, Séverine Coquoz, Jérôme Extermann, Theo Lasser
VVisible spectrum extended-focus optical coher-ence microscopy for label-free sub-cellular tomog-raphy
Paul J. Marchand, , ∗ Arno Bouwens, Daniel Szlag, David Nguyen, AdrienDescloux, Miguel Sison, S´everine Coquoz, J´erˆome Extermann, and TheoLasser Laboratoire d’Optique Biom´edicale, Ecole Polytechnique F´ed´erale de Lau-sanne, CH-1015 Lausanne, Switzerland ∗ paul.marchand@epfl.ch Abstract
We present a novel extended-focus optical coherence microscope (OCM)attaining 0.7 µ m axial and 0.4 µ m lateral resolution maintained over adepth of 40 µ m, while preserving the advantages of Fourier domain OCM.Our method uses an ultra-broad spectrum from a super-continuum lasersource. As the spectrum spans from near-infrared to visible wavelengths(240 nm in bandwidth), we call the method visOCM. The combinationof such a broad spectrum with a high-NA objective creates an almostisotropic 3D submicron resolution. We analyze the imaging performanceof visOCM on microbead samples and demonstrate its image quality oncell cultures and ex-vivo mouse brain tissue. Over the past decades, optical microscopy has allowed investigating biologicalsystems at high spatial and temporal resolution. Confocal fluorescence mi-croscopy [1] and light-sheet microscopy[2], through their capabilities in three-dimensional imaging, have become the mainstay for cellular and subcellularimaging. Nevertheless, while fluorescence provides molecular specificity, the in-fluence of these agents on cellular processes is ambiguous as they might interferewith the functioning of the cell. These effects combined with photobleachingultimately hinder the possibility to perform long-term imaging.In such studies, label-free microscopy offers an interesting alternative as itcan provide wide-field images at high acquisition rates without using exoge-nous agents. Moreover, the absence of labels facilitates the sample preparation.Recent advances in phase microscopy [3] and ptychography [4] have allowed per-forming three-dimensional imaging of cellular cultures or embryos but remainlimited to thin single layer structures.Optical coherence tomography (OCT) is an interferometric imaging tech-nique sensitive to refractive index contrast in the sample [5]. In OCT, the axialresolution is defined by the width of the illumination spectrum and an entiredepth profile can be obtained from a single recording of the output spectrum.1 a r X i v : . [ phy s i c s . b i o - ph ] A p r s such, only a two-dimensional scan is required to obtain a three-dimensionalimage.Optical coherence microscopy (OCM), the microscopy analogue to OCT,uses high-NA objectives to obtain a higher lateral resolution. In standard OCMsystems, however, the axial field of view is dictated by the Rayleigh range andthus decreases quadratically ( ∝ / NA ) with the improvement in lateral res-olution ( ∝ / NA). This compromise can be circumvented by engineering anextended-focus illumination through the use of so-called diffraction-less beamssuch as Bessel beams [6].In order to maintain a good collection efficiency of the scattered light signal, aGaussian detection mode is used. Therefore, separate illumination and detectionmodes are required: a Bessel illumination mode and a Gaussian detection mode.This split between modes can further be exploited to filter specular reflectionsand obtain a dark-field OCM system [7]. The dark-field property is particularlyimportant when investigating weakly scattering structures, such as cell samples,as it suppresses light reflected from the sample support which would otherwisestrongly reduce the usable dynamic range of the detector. As such, all availabledynamic range can be devoted to the desired, but weak, scattered light signal.In this paper, we present visible spectrum optical coherence microscopy (vi-sOCM). The system builds upon our previous dark-field OCM design, and im-proves its imaging capabilities for sub-cellular structures by using a large band-width illumination spectrum spanning visible to near-infrared wavelengths anda high-NA objective. The resulting system possesses an almost isotropic sub-micron resolution (0.4 µ m laterally and 0.7 µ m axially) maintained over a largedepth of field (¿40 µ m). Hence visOCM extends the capabilities of our previ-ous non-imaging visible light optical coherence correlation spectroscopy (OCCS)system and is optimized three-dimensional cellular tomography [8]. We presenta strategy for dispersion compensation and demonstrate the system’s 3D resolu-tion on microbead samples. We demonstrate visOCM’s image quality and con-trast on living cell cultures as well as fixed brain slices of healthy and alzheimericmice.Besides imaging the structure of a sample at a given time-point, there isalso much interest in monitoring intracellular dynamics to understanding cellfunction. As such, several optical microscopy techniques have been developedto analyse cell trafficking and intracellular motility[9, 10, 11]. Recently, OCTmethods developed to obtain qualitative and quantitative information on vas-cular function have been used to reveal sub-cellular compartments and quantifytheir activity [12, 13]. Being a Fourier-domain method, visOCM is capable ofrapidly acquiring tomograms, and therefore these dynamic signal imaging meth-ods can be applied to visOCM as well. We demonstrate dynamic signal imagingwith visOCM on living cells. 2 Materials and Methods
As illustrated in Figure 1, the optical setup is based on a Mach-Zehnder inter-ferometer, allowing for a separation of the illumination and detection modes,necessary to obtain the desired Bessel-Gauss configuration. The output of asupercontinuum laser (Koheras SuperK Extreme, NKT Photonics) is first fil-tered through three broadband dielectric mirrors (BB1-EO2, Thorlabs) to rejectthe source’s strong infrared emission. Then, the light is split by a polarizingbeam-splitter (PBS251, Thorlabs), and injected in the interferometer using asingle-mode fiber (P3-460B-FC-2, Thorlabs). As shown in Figure 3(a), thespectrum in the interferometer is centred at 647 nm and has a 246 nm band-width. After collimation, the light passes through a polarizer and then a firstbeamsplitter (BS1) splits the light into the reference and illumination paths,where an axicon lens (Asphericon, apex angle 176 ◦ ) generates a Bessel beam.The beam is then Fourier filtered in a telescope to remove stray light originat-ing from the axicon’s tip, after which it is steered to the scan-unit, and focusedon the sample by a 40x high NA objective (Olympus, effective NA = 0.76).The back-scattered light is collected by the objective, de-scanned by the scan-unit and directed to the detection arm through a second beam splitter (BS2)where it is coupled to a custom-made spectrometer with a single-mode fiber(P3-460B-FC-2, Thorlabs). The spectrometer is comprised of a transmissiongrating (600 lines/mm, Wasatch Photonics) and a fast line-scan camera (BaslerspL2048-140km). Dispersion matching between both interferometer arms wasperformed by adding a combination of prism pairs in the reference arm. Thedetailed dispersion matching strategy is presented in 2.2.To obtain a dark-field configuration, the specular reflection was suppressedby spatially filtering the reflected Bessel ring through a mask in the detectionarm (mask 2) and by properly filtering the stray light from the tip of the axiconlens (mask 1). The use of a broad illumination spectrum in visOCM, renders dispersion match-ing more challenging, particularly in the visible spectrum where the relationshipbetween the refractive index and the wavelength becomes increasingly non-linearat shorter wavelengths. Here we opted for a physical dispersion compensationscheme where we first estimated the dispersion caused by the illumination andsample arms prior to constructing the optical system. This allowed us to calcu-late the best composition of prisms to place in the reference arm (Figure 2(a)).In a second step, during the construction of the setup, we developed a real-timeinterface in LabView to display the amount of residual dispersion and used amotorized stage to finely adjust the thickness of the different glasses (Figure2(b-c)). 3 canUnitSupercontinuumLaser SourceSpectrometerFine Dispersion Control Unit delay linemask 1 mask 2PBS1 BS2PBSBBDMBBDM BBDM
Figure 1: Schematic of the extended-focus OCM using a broad spectrum in thevisible wavelength range and a high NA objective for high axial and lateral res-olution. By combining a Bessel illumination, generated by an axicon lens, and aGaussian detection, a dark-field extended-focus system can be obtained. P: Po-larizer, BBDM: Broadband dielectric mirror, BS: Beamsplitter, PBS: Polarizingbeamsplitter. 4 .2.1 Estimation of the dispersion in the optical system
Prior to constructing the optical system, we first estimated the thickness ofeach glass type present in the different arms of the interferometer. The glasstype and respective thickness of each optical element (lenses and beamsplitters)were obtained from the data-sheets provided with each element by Thorlabs.The objective was not modelled in this initial assessment as its composition isunavailable. As depicted in Figure 2(a) and described in Equation 1, we thenproceeded to estimate the k OP D ( k ) = (cid:2) n sample,SF ( k ) , . . . , n sample,UV F S ( k ) (cid:3) d sample,SF ... d sample,UV F S . . . + (cid:2) n ill,SF ( k ) , . . . , n ill,UV F S ( k ) (cid:3) d ill,SF ... d ill,UV F S . . . − (cid:2) n ref,SF ( k ) , . . . , n ref,UV F S ( k ) (cid:3) d ref,SF ... d ref,UV F S (1)The obtained OPD was then fit to a set of dispersion curves through amultivariate linear regression to find the thickness of the set of glasses to bestbalance the dispersion. As described in Equation 2, the multivariate linearregression tries to model the OPD as the sum of the refractive indexes (infunction of k ) of each glass multiplied by their respective thicknesses and anerror term (cid:15) ( k ). By performing this analysis, we can therefore retrieve thethicknesses of each glass required to compensate the OPD of the optical system.An initial analysis of several sets of glasses revealed that combining BK7 andSF10 allowed fully balancing the dispersion of the system. OP D ( k ) = (cid:2) n BK ( k ) , n SF ( k ) (cid:3) (cid:20) d BK d SF (cid:21) + (cid:15) ( k ) (2) During the construction and alignment of the system, the dispersion was finelybalanced by calculating the residual dispersion in the system and by adjustingthe thickness of each individual glass through a pair of motorized mechanicalstages. The residual dispersion was measured by placing a sample consistingof a cover glass and a mirror. Then the interferogram’s phase was calculated5y Hilbert transformation [14], and a linear fit was subtracted (Figure 2(c)).If the dispersion is perfectly balanced between the arms of the interferometerthen the residual dispersion curve should be null throughout the spectrum. Themotorized mechanical stages, as depicted in Figure 2(b), consists of a steppermotor operating a cogwheel mechanism. The mechanism allows varying theangle between two glass windows, in order to fine-tune the amount of that glasstype in the reference arm. Moreover, as the system is symmetrical, changesin the beam’s transverse position caused by the passage of the light throughthe first window are compensated by the second window. As such, varying theamount of glass causes minimal changes in the alignment of the optical system.The combination of these two effects (fine-tuning and minimal misalignment)facilitates the dispersion compensation procedure. By observing the residualdispersion and iteratively changing the thickness of each glass we could balancethe dispersion of the two arms of the microscope. We used SF6 windows, insteadof SF10, for dispersion fine tuning.
All images were acquired at a 20 kHz A-scan rate (with an integration time of43 µ s) and with a power varying from 1.5 mW to 3 mW in the back-focal planeof the objective. With the exception of the dynamic signal imaging protocol(presented in 3.4), the size of each image was 512 × × x , y and k respectively). Large field-of-views were obtained by stitching several tomograms(each having a lateral field-of-view of either 60 µ m × µ m, or 120 µ m × µ m) with a 30% overlap between each tile of the mosaic in both directions.Any tilt (angle with respect to the optical axis), was corrected on both axes( x-z and y-z ) prior to stitching. The tomogram processing, tilt-correction andstitching were performed through a custom-coded MATLAB graphical interface.The tomograms presented in Figures 4–7 are displayed with the intensity inlogarithmic scale for visualisation purposes. The tomograms were convolvedwith a 3D Gaussian kernel ( σ x,y = 0.187 µ m, σ z = 0.22 µ m) and were thenresized to obtain an isotropic sampling using ImageJ. All experiments were carried out in accordance to the Swiss legislation on animalexperimentation (LPA and OPAn). The protocols (VD 3056 and VD3058) wereapproved by the cantonal veterinary authority of the canton de Vaud, Switzer-land (SCAV, D´epartement de la s´ecurit´e et de l’environnement, Service de laconsommation et des affaires v´et´erinaires) based on the recommendations issuedby the regional ethical committee (i.e. the State Committee for animal experi-ments of canton de Vaud) and are in-line with the 3Rs and follow the ARRIVEguidelines. Brain slices were obtained by perfusing transcardially B6SJL/f1mice with PBS followed by 10% Formalin (HT501128, Sigma-Aldrich). The6 tepperMotorwindowscogwheel screw (c)
Residual dispersion calculation pipeline (a)
Dispersion estimation
ModelOPD MultivariateRegression d
BK7 d SF10
SF5k n k n UVFS k O P D d d substracting linear component λ I n t e n s it y λ -k mapping k Hilbert transform k P h a s e k R e s i du a l D i s p e r s i on (b) Fine compensation control unit d bs d ax ... k n SF10 k n BK7 lightbeam
Figure 2: Dispersion compensation strategies: (a) Prior to aligning the opticalsetup, a first estimation of the dispersion mismatch in the system is performedby modelling the optical path length of the different arms of the interferome-ter. The thicknesses and dispersion curves of the different glasses are used tomodel the OPD between the arms of the system. The OPD is then fit with amultivariate regression using the dispersion curves of SF10 and BK7 as set ofregressors to obtain the respective thicknesses to fully balance the dispersion inthe interferometer. (b) In order to match the dispersion, a compensation unitwas devised allowing a fine control of the thickness of the glass. A mechanismcomprising a stepper motor and cogwheels allows changing the angle of a pairof glass windows to vary the length of glass traversed by light. (c) During thealignment, the dispersion was matched by observing and minimizing the residualdispersion present in the system, obtained by a simple set of processing steps:The interferogram first undergoes a λ -k mapping step from which the phase isthen extracted by a Hilbert transform. The linear component of this phase isthen extracted and removed to reveal the residual dispersion.7ice were injected subcutaneously with Temgesic prior to the perfusion withheparinized PBS. The extracted brains were then left in 4% PFA overnight,and then placed in a solution of 30% glucose. Finally, the brains were cut intoslices using a microtome at a thickness of ∼ µ m and placed on a glass cover-slide. Brain slices from 5xFAD mice, a mice model of amyloid pathology, wereobtained using the same protocol. The amyloid plaques were stained using asolution of Methoxy-X04 in DMSO, which was administered through two I.P.injections 24h and 2h before the perfusion, as described by J¨arhling et al. [15]. In addition to mice brain slice imaging, we imaged live murine macrophages(cell line RAW 264.7) with the visOCM platform. The RAW 264.7 cells werecultured in an incubator at 37 ◦ C and 5% CO using DMEM high glucose withpyruvate (4.5 g l-1 glucose, Roti ® -CELL DMEM, Roth) supplemented with10% fetal bovine serum and 1 × penicillin-streptomycin (both gibco ® , ThermoFisher Scientific). Prior to imaging (1-2 days), the cells were seeded in Fluo-roDish Sterile Culture Dishes (35 mm, World Precision Instruments). The system’s lateral resolution was characterized by imaging a solution of nanopar-ticles of 30 nm in diameter embedded in a slab of PDMS. The small size of theseparticles allowed interrogating the point-spread function (PSF) of the opticalsystem. The depth-dependence of the lateral resolution was assessed by isolatingand averaging multiple ( ∼
10) measurements of the PSF at 7 different depths.The lateral profile of the measured PSF at each depth was then extracted, andthe position of the first zero served as a measure of the lateral resolution. Asshown in Figure 3(c) the width of the central lobe is maintained at 400 nm over40 µ m.The axial resolution of the system was measured by imaging a mirror placedon a glass coverslip in the sample arm. The reference power was adjusted toobtain maximum visibility of the interference pattern. The axial PSF, measuredas mentioned previously, is plotted in Figure 3(b) and has a FWHM of 0.92 µ min air, corresponding to a width of 0.69 µ m in water. The imaging performance of visOCM was first demonstrated by imaging cor-tical structures in fixed mice brain slices ( ∼ µ m thick) of both healthy andalzheimeric mice.The lateral and axial resolution and contrast offered by the visible spectrumallows resolving several different entities present in the brain such as myelinatedfibers, vascular structures, cell bodies and amyloid plaques.8 b) Axial PSF -15 -10 -5 10 1500.51 50 0.92 μ m / 0.69 μ m depth [ μ m] I n t e n s it y [ a . u . ] S i gn a l [ a . u . ] (c) Lateral PSF r [ μ m] -15 μ m -2 -1 0 1 20.73 μ m -10 μ m -2 -1 0 1 20.61 μ m Focus μ m -2 -1 0 1 2 μ m μ m μ m μ m μ m μ m μ m μ m0.40 μ m r [ μ m] r [ μ m] r [ μ m] r [ μ m] r [ μ m] r [ μ m] (a) Spectrum
500 550 600 650 700 750 800 850wavelength [nm]00.250.50.751 246 nm I n t e n s it y [ a . u . ] Figure 3: Characterization of the visOCM system: (a) Illumination spectrumspanning from the visible to the near-infrared range, centred at 647 nm and 246nm wide. (b)The ultra-broad spectrum leads to a submicrometric optical sec-tioning capability. (c) Plots and heatmaps (in linear scale) displaying the lateralPSF along the depth of focus of the objective, illustrating that the diameter ofthe central lobe is maintained at ∼
400 nm over 40 µ m in depth. Scalebar: 500nm 9 a) Transmission imaging of entire brain slice hemisphere (b) visOCM mosaics of selected region z zxy zz
PTLp cortex xy z = 28 μ m z = 33 μ m z = 40 μ m z = 45.7 μ mz = 22.4 μ m xy (c) close up of region with large cells (d) close up of region of dark cells xy rzxy rz (e) close up of dark capillaries (f) close up of bright capillaries (g) en-face images of (e) at different depth positions (b)(c)(e,g) (d,f) xy Figure 4: ex-vivo visOCM imaging of the PTLp cortex in a B6SJL/f1 mousebrain slice. (a) A transmission image of the entire mouse hemisphere was firstacquired to locate the desired area (green rectangle), which was then imagedwith visOCM (b). The mosaic of part of the PTLp cortex, acquired with vi-sOCM, reveals a variety of cortical structures, such as fibers, cell bodies andvascular entities (en-face view). In the mosaic, mainly two types of cells can bevisualized, large cells as shown in (c) and smaller darker cells as pointed by ar-rowheads in (d). The orthogonal views in (c–d) highlight the three-dimensionalrepartition of these cell types within the depth of the slice. Capillary vessels canbe discriminated from the tissue as either dark or bright structures as shownin (e) and (f) respectively. These different contrasts are further revealed in theorthogonal slices accompanying the close-ups, where one can trace the path ofthe hollow dark lumen or the bright vessel border, pointed by the arrowheads.En-face images at different depths show that visOCM can perform imaging over¿20 µ m (g). Scalebars: 150 µ m (b), 50 µ m for the en-face and 20 µ m for theorthogonal views (c–g). 10 a) Transmission imaging of entire brain slice hemisphere (b) visOCM mosaic of the selected region
RSPd cortex xy (c) Close-up of large caliber vessel (d)
Close-up of corpus callosum, cingulate bundle and cortex (e)
Close-ups of cells and subcellular structures (d)(e) (c)(b) zxy xy zz (f)(g) (h) (f)(g) (h) zxyz xyz zz zxyxy
Figure 5: ex-vivo visOCM imaging of the RSPd cortex in a B6SJL/f1 mousebrain slice. (a) A transmission image of the entire half hemisphere was per-formed to locate part of the RSPd cortex (blue rectangle). A mosaic of thearea of interest was then obtained with visOCM (b), where one can appreciatethe presence of fibers, vessels and cells. A large penetrating vessel (c) can beobserved through the difference in contrast between its hollow lumen and theback-scattering of the surrounding tissue. Examples of bifurcations and poten-tial clogging of the vessel are pointed by arrowheads in the orthogonal view.Fibers appear as thin oriented bright structures and are present in the cortexand in the corpus callosum (d). Finally, sub-cellular features can also be ob-served as darker spots within the cell bodies, as shown in (e–h) and pointed byarrowheads in (e). Scalebars: 150 µ m in (b), 50 µ m in the en-face view of (c–d),20 µ m in the orthogonal views of (c–d) and in the en-face view of (e), 10 µ m inthe en-face and orthogonal views of (f–g).11 .2.1 Myelin fibers Similarly to previous studies performed with OCM at longer wavelengths [16,17, 18, 19], myelin and neural fibers appear as bright linear structures throughtheir increased back-scattering. In Figures 4(a) and (c), fibers emerge from thecorpus callosum and the cingulum bundle and spread throughout the corticalcolumn. The corpus callosum, shown in Figure 5(d), contains a high density offibers and can be distinguished as a bright region with orientated stripes. Thecortex is characterized by a lower density of fibers with a higher variability intheir orientation.
Vascular compartments, from large calibres (penetrating arterioles and ascend-ing venules) to the smallest capillaries can be discriminated from the backgroundtissue as either hollow tubes (from the empty lumen) or as thin bright struc-tures. Figure 5(c) shows the en-face and orthogonal slice of a large penetratingvessel where one can distinguish the hollow lumen from the tissue and visualisebifurcations along the propagation of the vessel. Additionally, the edges of thevessel exhibit an increased signal, which could either be caused by a change inthe scattering properties of the vessel’s membrane or its surrounding tissue (forexample vascular smooth muscle). Smaller vessels, such as capillaries, can alsobe observed in the tomograms and appear either as dark or bright structurescompared to the surrounding tissue. Figures 4(e) and (f) show both dark andbright capillary structures and show that it is possible to trace their trajec-tory regardless of their contrast. Similarly to the large vessel in 5(c), the darkcontrast is indicative of a lack of scatterers within the lumen of the capillary.The bright contrast, on the other hand, could originate from scatterers fillingthe vessel’s lumen (i.e. clogging during the perfusion procedure) or from thedifferent scattering properties of the vessel’s boundary.
In addition to neuronal fibers and capillaries, visOCM imaging allows visualizingdifferent cell body types through their different contrast with respect to theextracellular space. Figures 4(b) and 5(b) present mosaics covering parts of theposterior parietal association areas (PTLp1) and the retrosplineal area (RSPd)respectively, where one can observe two main cell body types with differentcontrasts and shapes. As highlighted in Figure 4(d), some of the cells appear asdark spherical shapes, due to a decreased back-scattering. The second type ofcells visible in Figure 4(c) have similar contrast than the extracellular space andare larger. In addition to their different shapes and back-scattering properties,the two cell types also appear to be present in different regions of the corticalcolumn: the darker and smaller cells are denser in the upper layers whereasthe larger cells seem more prominent in the deeper layers, closest to the corpuscallosum. Cells in the cornu ammonis area 1 (CA1), as shown in Figure 6(b), arealso characterized by a darker contrast, similar to the cells in the upmost layers12f PTLp1. Additionally, one can notice in the RSPd brain slice the presenceof a small darker substructure within the body of certain cells. Figure 5(e–h)displays a selection of these cells and their dark subcellular structure. Althougha more complete study is necessary to elucidate the nature of this feature, ourexperience in live-cell imaging (results shown in Figure 7) has shown that asimilar contrast is present in what seems to be the nucleus. The orthogonalviews and tile (g) of Figure 4 show that the signal acquired with visOCM extendsthroughout the depth of the tissue slice, although a loss in intensity and blurringcan be observed in the deeper layers.
Previous work from our group has shown that amyloid plaques can be distin-guished from cerebral tissue using xfOCM operating at 800 nm through theirincreased scattering [16]. In continuation of this work, we imaged brain slices ofan alzheimeric mouse model with our novel visOCM system, with the expecta-tion that the increased spatial resolution and different illumination wavelengthwould shed light on the details of these aggregates. We therefore imaged a partof the PTLp cortex and subcortical structures (CA1) of a 5xFAD mouse. Asshown in Figure 6(b) and similarly to the results in Figures 4(b) and 5(b), thecortex is characterized by a high density of fibers and of cell bodies. The subcor-tical region, below the corpus callosum, has a slightly lower intensity comparedto the cortex and presents a line of cells (CA1). Similarly to the results obtainedat 800 nm [16], the amyloid plaques manifest themselves as high intensity regionswith a darker core. The increased spatial resolution of the system reveals withgreat detail the irregular shape of these aggregates, as shown in 6(c–d). Theplaques are, as expected, present in cortical and also subcortical regions [20],where the slightly decreased intensity of the cerebral tissue provides a highercontrast between the plaques and the background. The location of the plaqueswas colocalized with fluorescence imaging of Methoxy-X04 using a commercialwidefield microscope (Axiovert 200M, Zeiss), a 20x / 0.5 NA objective and theDAPI filter set (Excitation filter: 365 nm, dichroic mirror: 395 nm, Emissionfilter: 445/50 nm). As shown in tiles (c–d) of Figure 6, the locations of the ag-gregates in visOCM are in agreement with the location of the labelled structuresin the fluorescence image.
In addition to the imaging of tissue structures, we performed imaging of livemacrophages in a cell culture. A mosaic of tomograms of these macrophages isshown in Figure 7(a). The capabilities of the extended-focus can be appreciatedin the orthogonal views of the tomogram, where the signal extends sufficiently indepth to reveal the three-dimensional organisation of the culture, with certaincells lying on top of other cells. The strongly scattering cytoplasm of the cellappears as a bright structure surrounding a darker subvolume, which most likelycorresponds to the cell nuclei. The increased lateral resolution of the system13 a) Transmission imaging of entire brain slice hemisphere xy (b) visOCM mosaics of selected region xy zz * *** ***** *** xy (c) close up of a cortical region with plaques and corresponding fluorescence image Methoxy X04 z zxy * **** ***** ** xy (d) close up of a subcortical region with plaques and corresponding fluorescence image Methoxy X04 (b)(c) (d)
PTLp cortex
Figure 6: ex-vivo imaging of cortical and subcortical structures in a 5xFADmouse brain slice. (a) A transmission image shows the location of the areawhere a visOCM image mosaic was obtained (b). The visOCM mosaic revealsfibers, cells and amyloid plaques in both cortical and subcortical structures.Close-ups of areas of interest containing amyloid plaques, in both cortical andsubcortical regions, are displayed with their en-face views and correspondingfluorescence image (c–d). In the visOCM image, plaques can be seen as irregularhigh intensity regions. Scalebars: 150 µ m in (b), 50 µ m in the en-face and 20 µ m m in the orthogonal views of (c–d).14llows resolving the filopodia on certain cells. In a second step, we analysed the dynamic properties of the scattering signalfrom living cells. In contrast to previous dynamic signal imaging techniquesusing the autocorrelation function or the standard deviation of the OCT sig-nal [12, 10, 13], we extracted the dynamic component of the back-scatteringthrough a point-wise subtraction of the complex OCM signal (Figure 7(b)), asdeveloped by Srinivasan et al. for OCT angiography [21]. A time-series of scat-tering fluctuations per voxel was obtained by sampling each transverse position(B-scan) 32 times with a timestep of Δ t = 27 ms. Each timepoint was thentemporally high-pass filtered (through a point-wise complex subtraction) andthen averaged. A temporally averaged image was also obtained by averagingthe repeated acquisitions. The results of this operation are shown in Figure 7(c)and Figure 7(d) showing the averaged and dynamic signals respectively. Inter-estingly, in addition to the lower intensity of the cell nucleus already present inthe static imaging, the dynamic imaging further reveals details within the cellbody. As shown in Figure 7(d), darker regions are present within the nuclei ofthe cells and brighter spots can be observed within the cytoplasm. Finally, theinterface between the nuclei and the cytoplasm appears as a fine dark structurein certain cells. These differences are further revealed in Figure 7(e) and (f)showing close-ups of a selected cell in the temporally averaged and dynamicsignal image respectively. In this work, we presented a novel OCM system, called visOCM, combiningan extended-focus [6], a dark-field detection [7], a high-NA objective and anultra-broad illumination spanning from the visible to the near-infrared wave-length range. As demonstrated here, the combination of these features providesan almost isotropic submicron resolution, maintained over ¿ 40 µ m in depth.The capabilities of visOCM were demonstrated by imaging brain tissue slicesof healthy and alzheimeric mice and macrophages cell cultures. The imaging ofbrain tissue with visOCM reveals several cortical structures as vessels, capillar-ies and cells. Interestingly, these different structures exhibited a wide range ofcontrasts, even within the same structure type. Capillaries could be observed asboth dark or bright hollow structures. Although one cannot discard changes intissue caused by the sample preparation (i.e. clotting and vessel collapsing dueto the perfusion), the bright contrast could emanate from the presence of tissuebordering the vessels (such as smooth muscle cells or pericytes). Conversely, thedark contrast arises from a lack of scatterers from the hollow lumen of the ves-sel. Previous studies involving OCT and OCM imaging of brain tissue showedthat certain cells could be identified through their different contrast within thecerebral tissue [12, 17, 19, 22]. Srinivasan et al. and Tamborski et al. observed15 y zz zxyz xy μ ... Dynamic Signal B-Scan(d)
B-Scan 1B-Scan 2B-Scan 3B-Scan 4B-Scan 31B-Scan 32 fast-axis s l o w - a x i s l.p. 1l.p. 2l.p. 3l.p. 4l.p. 5l.p. 512 ... ...
32 repeated B-scans at each lateral position (l.p.) (a) visOCM imaging of live macrophages (b) dynamic signal imaging protocol(d) dynamic signal imaging of live macrophages(c) temporal averaged image of live macrophages zxyz μ ... Averaged B-Scan (c)
B-Scan 1B-Scan 2B-Scan 3B-Scan 4B-Scan 31B-Scan 32 ... xy zzxy zz (f) close-up of cell in dynamic signal image(e) close-up of cell in temporal averaged image
Figure 7: visOCM imaging of murine macrophages. (a) A mosaic ofmacrophages obtained with visOCM with its orthogonal views reveals the three-dimensional organisation of cells in a culture. We further explored the capabil-ities of visOCM by applying a protocol similar to OCT angiography (b). Theprotocol entails imaging each lateral position along the slow axis 32 times (32 re-peated B-scans per location). These 32 B-scans are either averaged or undergoa point-wise complex subtraction to obtain an averaged image (c) or a viewof the dynamic components of the tomogram (d) respectively. The averagedimage is identical in contrast to (a), whereas the dynamic signal image furtherreveals compartments within the cell, as either darker or brighter subregions (d)Close-ups of a selected cell in (c) and (d) are shown in (e) and (f) respectively,highlighting the differences in contrast between the averaged and dynamic im-ages. Scalebars: 50 µ m in the en-face view in (a), 20 µ m in the orthogonal viewsin (a) and in the en-face views of (c–d), 10 µ m in the orthogonal views in (c–d)and in the en-face views of (e–f), 5 µ m in the orthogonal views in (e–f).16hat certain neuron types could be discriminated from the tissue by their darkcontrast using an OCM operating at 1300 nm and 800 nm respectively [17, 22].By exploiting a higher lateral resolution and by shifting the illumination spec-trum to the visible wavelength range, we show that this intrinsic contrast seemsto vary between cell types and thus could be used in the future to identify differ-ent types of neurons and inter-neurons. Furthermore, the increased resolutionof the visOCM system reveals subcellular features, potentially the cell nuclei.Future work will focus on elucidating the causes of these different contrastsand attempt to discriminate and potentially classify the cells according to theirintrinsic scattering properties.In a second step, we built upon a previous study performed at 800 nm byimaging brain slices of alzheimeric mice models with our novel system [16].Similarly to our aforementioned results, the plaques appear as bright irregularstructures in the tomograms obtained with visOCM. Although the exact natureof this particular contrast remains unknown, the presence of metals (such asFe) in amyloid plaque cores, as described by Plascencia-Villa et al. [23], mightprovide a first hint into the cause of this phenomenon. In fact, a previousstudy from our group showed that the contrast of the Langerhans islets inOCM images originated from the presence of Zn crystals [24]. Alternatively,this contrast could also arise from polarization effects as the plaques have beenshown to have birefringent properties [25].Finally, we demonstrated the performance of visOCM by imaging live-cellsin cultures. The extended-focus and high acquisition rates available with theplatform allows fast imaging of the three-dimensional structure of cell cultures.The shift in illumination wavelength and increase in NA provides the resolutionnecessary to identify thin structures such as filopodias and sufficient contrastto visualize seemingly sub-cellular structures. Recent studies have highlightedthe importance of studying the dynamic properties of cellular back-scattering tounderstand cellular function [26]. In this context, we explored the possibility toperform dynamic imaging, as already performed in full field OCM [13], OCM [9]and phase imaging [11], with our novel imaging platform to further discriminatebetween subcellular compartments. We applied a protocol devised for OCT an-giography [21] and show that a dynamic contrast can be obtained in a cultureof macrophages. Although more work is needed to identify the different regionsand the nature of the signals, one can appreciate the increased contrast pro-vided by the protocol. Overall, the combination of the extended-focus, the highisotropic resolution and high acquisition rates make visOCM an ideal platformto monitor fast processes occurring at the subcellular level in cell cultures.In addition to our demonstration of the capabilities of our novel visOCM sys-tem, we have introduced a strategy for physical dispersion compensation. Weused a multivariate linear regression to model the OPD of the system prior toalignment and present a processing pipeline and a hardware unit to finely mini-mize the dispersion during the construction of the microscope. In this work, weopted for a hardware dispersion compensation strategy, however numerical dis-persion compensation techniques could also have been used and will be exploredin future work. 17ltimately, visOCM offers label-free 3D imaging of tissue and cell struc-tures but remains limited through its lack in specificity. Our results show thatit is possible to discriminate between structures in tissue and cells throughtheir individual back-scattering properties and potentially through their dy-namic signatures (as revealed by our dynamic imaging). Nevertheless, thesecontrast mechanisms are limited and do not always provide the desired molecu-lar specificity. In such regards, the visOCM platform could be augmented with acollinear fluorescence channel. This configuration would allow molecular-specificfluorescent imaging to be complemented with visOCM imaging, providing in-formation on the overall structural context of the organism under investigation.In addition to fluorescence, the visible spectrum could be used to explore thespectroscopic signatures of specific molecules. Spectroscopic OCT has showngreat promise in its ability to provide additional contrast mechanisms [27, 28].Having an illumination in the visible wavelength range could be used to visual-ize and discriminate between different endogeneous and/or exogenous contrastagents within the imaged tissue or cells. Funding
This study was partially supported by the Swiss National Science Foundation(205321L 135353 and 205320L 150191), the Commission for Technology andInnovation (13964.1 PFLS-LS and 17537.2 PFLS-LS) and by the EU FrameworkProgramme for Research and Innovation (686271).
Acknowledgments
We thank Kristin Grussmayer for preparing the cells and B. Deplancke (LSBG,EPFL) for kindly giving us the murine macrophages cell line RAW 264.7.
References [1] J. Pawley,
Handbook of Biological Confocal Microscopy (Springer, 2006).[2] J. Huisken, J. Swoger, F. Del Bene, J. Wittbrodt and E. H. K. Stelzer, “Op-tical Sectioning Deep Inside Live Embryos by Selective Plane IlluminationMicroscopy,” Science (5686), 1007–1009 (2004).[3] T. Kim, R. Zhou, M. Mir, S. D. Babacan, P. S. Carney, L. L. Goddard. andG. Popescu, “White-light diffraction tomography of unlabelled live cells,”Nat. Photon. , 256–263 (2014).[4] L. Tian and L. Waller, “3D intensity and phase imaging from light fieldmeasurements in an LED array microscope,” Optica (2), 104–111 (2015).185] A. F. Fercher, W. Drexler, C. K. Hitzenberger, and T. Lasser, “Opticalcoherence tomography - development, principles, applications,” Rep. Prog.Phys. , 239–303 (2003).[6] R. A. Leitgeb, M. Villiger, A. H. Bachmann, L. Steinmann, and T. Lasser,“Extended focus depth for Fourier domain optical coherence microscopy,”Opt. Lett. (16), 2450–2452 (2006).[7] M. Villiger, C. Pache, and T. Lasser, “Dark-field optical coherence mi-croscopy,” Opt. Lett. (20), 3489–3491 (2010).[8] S. Broillet, D. Szlag, A. Bouwens, L. Maurizi, H. Hofmann, T. Lasser, andM. Leutenegger, “Visible light optical coherence correlation spectroscopy,”Opt. Express (18), 21944–21957 (2014).[9] M. Sison, S. Chakrabortty, J. Extermann, A. Nahas, P. J. Marchand, A.Lopez, T. Weil, and T. Lasser, “3D Time-lapse Imaging and Quantificationof Mitochondrial Dynamics,” Sci. Rep. , 43275 (2017).[10] A. Oldenburg, X. Yu, T. Gilliss, O. Alabi, R. M. Taylor II, and M. A.Troester, “Inverse-power-law behavior of cellular motility reveals stromal-epithelial cell interactions in 3D co-culture by OCT fluctuation spec-troscopy,” Optica , 877–885 (2015).[11] L. Ma, G. Rajshekhar, R. Wang, B. Bhaduri, S. Sridharan, M. Mir, A.Chakraborty, R. Iyer, S. Prasanth, L. Millet, M. U. Gillette and G. Popescu,“Phase correlation imaging of unlabeled cell dynamics,” Sci. Rep. , 32702(2016).[12] J. Lee, H. Radhakrishnan, W. Wu, A. Daneshmand, M. Climov, C. Ay-ata and D. A. Boas, “Quantitative imaging of cerebral blood flow velocityand intracellular motility using dynamic light scattering-optical coherencetomography,” J. Cereb. Blood Flow Metab. (6), 819–825 (2013).[13] C. Apelian, F. Harms, O. Thouvenin, and A. C. Boccara, “Dynamic fullfield optical coherence tomography: subcellular metabolic contrast revealedin tissues by interferometric signals temporal analysis,” Biomed. Opt. Ex-press (4), 1511–1524 (2016).[14] M. Wojtkowski, V. J. Srinivasan, T. H. Ko, J. G. Fujimoto, Andrzej Kowal-czyk and J. S. Duker, “Ultrahigh-resolution, high-speed, Fourier domainoptical coherence tomography and methods for dispersion compensation,”Opt. Express (11), 707–709 (2004).[15] N. J¨ahrling, K. Becker, B. M. Wegenast-Braun, S. A. Grathwohl, M. Jucker,H.-U. Dodt, “Cerebral β -Amyloidosis in Mice Investigated by Ultrami-croscopy,” PLOS ONE (5), 1–13 (2015).1916] T. Bolmont, A. Bouwens, C. Pache, M. Dimitrov, C. Berclaz, M. Villiger, B.M. Wegenast-Braun, T. Lasser, and P. C. Fraering, “Label-free imaging ofcerebral β -amyloidosis with extended-focus optical coherence microscopy.”J. Neurosci. , 14548–14556 (2012).[17] V. J. Srinivasan, H. Radhakrishnan, J. Y. Jiang, S. Barry, and A. E. Ca-ble, “Optical coherence microscopy for deep tissue imaging of the cerebralcortex with intrinsic contrast.” Opt. Express (3), 2220–2239 (2012).[18] C. Leahy, H. Radhakrishnan, and V. J. Srinivasan, “Volumetric imagingand quantification of cytoarchitecture and myeloarchitecture with intrinsicscattering contrast.” Biomed. Opt. Express (10), 1978–1990 (2013).[19] C. Magnain, J. C. Augustinack, E. Konukoglu, M. P. Frosch, S. Sakadzic,A. Varjabedian, N. Garcia, V. J. Wedeen, D. A. Boas, and B. Fischl, “Opti-cal coherence tomography visualizes neurons in human entorhinal cortex.”Neurophotonics (1), 015004–015008 (2015).[20] S. Jawhar, A. Trawicka, C. Jenneckens, T. A. Bayer, and O. Wirths, “Mo-tor deficits, neuron loss, and reduced anxiety coinciding with axonal de-generation and intraneuronal A β aggregation in the 5XFAD mouse modelof Alzheimer’s disease.” Neurobiol. Aging (1), 196.e29–196.e40 (2012).[21] V. J. Srinivasan, J. Y. Jiang, M. A. Yaseen, H. Radhakrishnan, W. Wu,S. Barry, A. E. Cable, and D. A. Boas, “Rapid volumetric angiography ofcortical microvasculature with optical coherence tomography.” Opt. Lett. (1), 43–45 (2010).[22] S. Tamborski, H. C. Lyu, H. Dolezyczek, M. Malinowska, G. Wilczynski,D. Szlag, T. Lasser, M. Wojtkowski, and M. Szkulmowski, “Extended-focusoptical coherence microscopy for high-resolution imaging of the murinebrain.” Biomed. Opt. Express (11), 4400–4414 (2016).[23] G. Plascencia-Villa, A. Ponce, J. F. Collingwood, M. J. Arellano-Jim´eanez,X. Zhu, J. T. Rogers, I. Betancourt, M. Jos´e-Yacam´an and G. Perry “High-resolution analytical imaging and electron holography of magnetite parti-cles in amyloid cores of Alzheimer’s disease,” Sci. Rep. , 24873, (2016).[24] C. Berclaz, A. Schmidt-Christensen, D. Szlag, J. Extermann, L. Hansen,A. Bouwens, M. Villiger, J. Goulley, F. Schuit, A. Grapin-Botton, T.Lasser, and D. Holmberg “Longitudinal three-dimensional visualisation ofautoimmune diabetes by functional optical coherence imaging,” Diabetolo-gia (3), 550–559 (2016).[25] B. Baumann, A. Woehrer, G. Ricken, M. Augustin, C. Mitter, M. Pircher,G. G. Kovacs, and C. K. Hitzenberger “Visualization of neuritic plaques inAlzheimer’s disease by polarization-sensitive optical coherence microscopy,”Sci. Rep. , 43477 (2017). 2026] C.-E. Leroux, F. Bertillot, O. Thouvenin, and A. C. Boccara “Intracellulardynamics measurements with full field optical coherence tomography sug-gest hindering effect of actomyosin contractility on organelle transport,”Biomed. Opt. Express (11), 4501–4064 (2016).[27] U. Morgner, W. Drexler, F. X. K¨artner, X. D. Li, C. Pitris, E. P. Ippen,and J. G. Fujimoto “Spectroscopic optical coherence tomography,” Opt.Lett. (2), 111–113 (2000).[28] F. E. Robles, C. Wilson, G. Grant, and A. Wax “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photon.5