One of our past and current foci is the development of novel optical techniques for high-speed imaging. Together with our collaborators, we apply our methods to study neuronal activity and cellular dynamics in a range of model organisms. Amongst others, we have put forward a two-photon microscopy technique based on light-sculpting (Fig. 1) that has enabled the first whole-brain calcium imaging in C. elegans (Fig. 2). (Schrodel et al., 2013)

Fig. 1: Principle and realization of temporal focusing (TeFo).Schematic illustration of temporal focusing setup, illustrating spatial dispersion and pulse width broadening outside the focal (sample) region.

Fig. 2: Whole brain imaging in C.Elegans.Left: Micrograph of the C. Elegans nematode, brain region is highlighted by red box. Right: Single axial plane acquired with the WF-TeFo method. Dashed lines indicate yz and xz cross-sections shown.

Recently, we extended our light-sculpting imaging methods to the scattering tissue domain and demonstrated fast volumetric calcium imaging across the majority of a cortical column in the mouse (Fig. 3). (Prevedel et al., 2016)

Fig. 3: Fast volumetric Ca²⁺-imaging across a cortical column in the in-vivo mouse.(a) Cartoon depicting the extend of a cortical column as well as its cortical layers. Red line indicates maximum depth during imaging. (b) Left: 3D rendering of imaged volume (500x500x500µm), individual neurons are clearly resolved. Right: Example Ca²⁺ signal (∆F/F0) of GCaMP6 fluorescence extracted from the data. Time traces of 16 neurons out of a total of ~4000 are shown. Volume acquisition rate is 3Hz.

In other work, we have established light-field deconvolution microscopy, an elegant approach to perform volumetric imaging that achieves unprecedented acquisition speeds while requiring no mechanical scanning. This can be applied to image neuronal activity across entire, small organisms or small animal brains such as zebrafish larvae (Fig. 4). (Prevedel et al., 2014)

Fig. 4: Light-field deconvolution microscopy.Left: Experimental realization depicting the microlens array appended to the camera port of a wide-field microscope. Right: Whole-brain Ca²⁺ imaging of larval zebrafish in-vivo.

To further push the capabilities of light-field imaging, we have combined selective volume illumination with orthogonal light-field detection to obtain higher and more isotropic resolution, while significantly reducing reconstruction artefacts. With this we were able to image the beating heart of Medaka fish as well as its blood flow dynamics at up to 200Hz volume rate (Fig. 5). (Wagner et al., 2019)

Fig. 5: Isotropic light-field microscopy.Left: Schematic showing the mutually orthogonal illumination and dual-view detection geometry. Right: Maximum-intensity projections of Medaka heart cells, comparing resolution in standard LFM (top) and dual-view LFM (bottom).

Project funded with an ERC Consolidator grant.

Brillouin microscopy is a novel microscopy technique that measures mechanical properties such as elasticity and viscosity of biological samples in a 3D, nondestructive, and fully optical fashion.

Fig. 1: Principle of Brillouin microscopy.Laser light is scattered by thermally generated acoustic waves and undergoes a frequency shift (left). The high shift corresponds to a large longitudinal modulus of the probed volume, which can be interpreted as being ‘stiffer’ (right).

Mechanical properties have been shown to play an important role in several biological processes, including control of malignancy in tumors, stem cells differentiation, as well as in morphogenesis of cells and tissues. Standard techniques in the field of mechanobiology, however, typically rely on external perturbations making them invasive, while other approaches often suffer from poor resolution. Instead Brillouin microscopy exploits a light-matter interaction, called Brillouin scattering, to probe mechanical properties in the GHz regime. Due to its all-optical nature, it can achieve, high, diffraction-limited resolution in 3D. (Bevilacqua et al., 2019)

Fig. 2: Brillouin imaging of zebrafish tail.Brillouin microscopy is capable of revealing the thin (400 nm) ECM layer and measuring its mechanical properties and thickness in-vivo.

The physical principle is outlined in Figure 1. Laser light is focused on the sample and interacts with sound waves, intrinsically present in any material because of thermal agitation. During the scattering process, the light experiences a positive or negative frequency shift. The amount of the shift is proportional to the speed of sound inside the material and is thus informative of visco-elastic properties. In particular, it is proportional to the longitudinal modulus, from which elastic and viscous parameters can be obtained. (Prevedel et al., 2019)

Our lab has developed a confocal Brillouin microscope capable of fluorescence co-detection for long-term imaging of biological processes. We are applying our Brillouin microscope, in collaboration with other groups, to interesting processes in development and cell biology. Recently we optimized our Brillouin microscope to image the properties of sub-micron thick layers of extracellular layers in live zebrafish (Fig.2). (Bevilacqua et al., 2019)

Fig. 1: in-vivo imaging of dendrites and synapses.Three-photon microscopy of superficial cortical layers in a Thy1-GFP expressing mouse.

Fig. 2: Attenuation in brain tissue.Two optical windows in the near-infrared at 1300nm and 1700nm are optimal for deep tissue imaging. Two photon-microscopy at these windows enables excitation off far-red shifted fluorophores while three-photon excitation enable excitation of common fluorophores such as GFP or RFP. Modified from Horton et al. 2013.

In order to study dynamic biological processes in-vivo in mammalian organisms such as the mouse, techniques are required which enable non-invasive imaging at large tissue depth with sub-cellular resolution. Multi-photon microscopy is currently the technique of choice for in-vivo imaging of opaque and highly scattering tissue samples. However, scattering and optical aberrations lead to degradation of the point spread function (PSF) and hence, reduced image contrast, resolution and excitation power at depth. To enable imaging of cellular- and subcellular structures at high spatial resolution deep inside mammalian tissue in-vivo, we combine two powerful optical techniques: multiphoton microscopy and adaptive optics.

Fig. 2: Attenuation in brain tissue.Two optical windows in the near-infrared at 1300nm and 1700nm are optimal for deep tissue imaging. Two photon-microscopy at these windows enables excitation off far-red shifted fluorophores while three-photon excitation enable excitation of common fluorophores such as GFP or RFP. Modified from Horton et al. 2013.

Fig. 3: Optical aberrations in multi-photon microscopy.(a) Aberrations stemming from the system and sample lead to wavefront distortions and degradation of the PSF, which in turn lead to a loss of resolution, contrast and signal intensity. However, wavefront aberrations can be corrected by determining an ideal input wavefront which recovers diffraction limited resolution inside the tissue (b).

For deep tissue imaging we are exploiting the advantageous properties of three- photon excitation. The longer wavelength employed for excitation and the later on-set of out-of-focus fluorescence increases the signal-to-noise ratio at depth (Fig.2). This can be combined with red-shifted fluorophores to further reduce scattering. (Qi et al., 2018)

Adaptive optics is a technique to improve imaging performance by correcting wavefront aberrations introduced by the biological sample and the optical system. We are working on direct and indirect wavefront sensing approaches to enable diffraction limited resolution deep inside the tissue (Fig. 3).

We apply our adaptive multi-photon technique to open questions at the forefront of biology and neuroscience.

Fig. 3: Optical aberrations in multi-photon microscopy.(a) Aberrations stemming from the system and sample lead to wavefront distortions and degradation of the PSF, which in turn lead to a loss of resolution, contrast and signal intensity. However, wavefront aberrations can be corrected by determining an ideal input wavefront which recovers diffraction limited resolution inside the tissue (b).

The strong scattering properties of many biological tissues prevent deep imaging using visible light. Although approaches exist that partly circumvent this problem, such as multi-photon microscopy or Optical Coherence Tomography, they still hit a hard depth limit at about 1-2 mm in practice. To image beyond, we need to exploit other imaging modalities not based on classical light microscopy. One of the most promising techniques is photoacoustic imaging which allows to achieve imaging depth and resolution similar to that of ultrasound tomography while retaining molecular specificity.

Fig. 1: Photoacoustic tomography.(A) Physical principles of the photoacoustic effect. (B) Overview of detection and image reconstruction. Curved black line visualises the curvature of the wave being encoded in the time delay of incoming pulses.

Fig. 2: Photoacoustic tomography of 3dpf zebrafish embryo.Right: Bright-field image showing the zebrafish anatomy as well as light-absorbing melanocytes which surround the notochord. Left: Photoacoustic tomography image highlighting melanocytes and eye pigmentation.

In simple terms, the photoacoustic imaging is based on generating sound with the use of light (Fig. 1A). The use of high-energy, short laser pulses tuned to the absorption spectrum of the molecules of interest causes rapid transient heating of the tissue at the sites of light absorption. Rapid heating in turn causes local expansion of the tissue and generates propagating acoustic waves which can be used for ultrasound-based imaging. We are developing approaches based on photoacoustic tomography which allows to fully harvest the deep tissue imaging capabilities (<10mm) of the technique. In this approach, instead of acquiring images pixel by pixel, we excite the entire tissue at once and record the overall generated acoustic field over time from multiple directions. We then use a computational approach based on simulating the acoustic field propagation to reconstruct the absorbing molecule distribution from the acoustic field they generate (Fig. 1B).

Fig. 2: Photoacoustic tomography of 3dpf zebrafish embryo.Right: Bright-field image showing the zebrafish anatomy as well as light-absorbing melanocytes which surround the notochord. Left: Photoacoustic tomography image highlighting melanocytes and eye pigmentation.

Fig. 1: Working principle of the system.Excitation light (blue) enters the media and was collected by the radially spaced detector array (D1-D6). Photon propagation follows the “banana shape” phenomena. The further the detector located from the source, the deeper the light penetrate into the medium. Subsequently, fluorescence inclusion will be excited by those photons and the emitted light will be collected by detectors. Relative proximity to the fluorescence inclusion will affect the acquired intensity value.

Intravital Fluorescence Tomography (IFT) is a non-invasive, thick tissue imaging technique enabling longitudinal studies. IFT relies on collecting multiply scattered fluorescent light from the surface of a turbid media with a detector array which is radially spaced away from the excitation light source (Fig.1). The different separation between the source and detectors yields depth information of the fluorescence emission inside the tissue. The collected fluorescent light from different detectors constitutes a tomographic data set for 3D volumetric reconstruction of the fluorescent target.

Tumor volume assessment is essential for staging of tumor progression and evaluation of drug treatment efficacy. However, non-invasive assessment of the volume, in-vivo, is still a challenge for traditional imaging modalities, owing to the large interrogation area and deep-seated molecular (fluorescence) signal through a highly scattering skin tissue. Intravital Fluorescence Tomography (IFT) offers a solution to capture the in-vivo tumor volume in 3D with sufficient spatiotemporal resolution with the help of Non-invasive Intravital Imaging Window (NIIW) for tissue stabilization (Fig.2).

Fig. 2: 3D image shows a reconstruction of mCherry expressing tumor cells.Through the fluorescence intensity data acquired by an array of cells, we reconstructed the relative cell density within a 10x10x3 mm³ volume. Colorbar represents the relative cell count within a tumor mass.

With the help of an Doxycycline inducible mammary gland tumor model, one can replicate the tumor lifecycle in mice: the tumor onset, growth, regression, and relapse. IFT provides the platform for long-term tumor monitoring without interfering with the tumor microenvironment and reveals the tumor volume progression over its entire lifecycle. This technique is useful for studying tumor biology, understanding tumor development, and investigating the drug response of a selected tumor type in-vivo in the native tissue. Furthermore, IFT alleviates the burden put on animals, usually during the imaging sessions and reduces the overall required animal number by enabling longitudinal studies. 

Overall, IFT is a powerful 3D intravital imaging tool of biological imaging, which we apply for shedding light on tumor development questions.

The term optical coherence imaging refers to the optical coherence tomography (OCT) and its another modality optical coherence microscopy (OCM) which has a higher transverse spatial resolution. Similar to ultrasound in principle, OCT/OCM obtains cross sectional images of microstructure in biological systems with optical resolution by measuring the echo time delay of optical backscattering. In general, OCT/OCM allows 3D imaging with fast speed, high resolution and deep penetration. These features make it a powerful intravital imaging tool with applications spanning many multiple clinical specialties as well as fundamental scientific and biological research.

Fig. 1: Working principle of OCTBased on the Wiener-Khintchine Theorem, the depth scan can be calculated by an inverse Fourier-transform from the acquired spectra since the spectral modulation frequency in the space of wavenumber is proportional to the optical pathlength difference.

Fig. 2: Volumetric display of mouse skin obtained by OCTA mammary gland (breast) tumor was implanted orthotopically into the mammary fat pad (the black hole at the corner) and kept developing. The total volume of the image is 5x5x0.5(depth) mm.

Fig. 3: in-vivo OCT scan of mouse earThis 3D image was acquired by scanning the laser beam on the dorsal side of a mouse ear. The cross sections (B-scan) show the ear structure in a width of 2.5 mm. The thicknesses of mouse ear and its cartilage in the image are around 350 μm and 70 μm respectively.

In our lab, we are developing a high-resolution OCT system for intravital applications such as tumor anatomy imaging (Fig 2).

Specific tasks need specific tools. Our lab develops bespoke methods to unlock long standing questions in neurobiology. Currently our research is aimed at:

Fig. 1: Linking neuronal population activity to facial behavior.Top: combination of face imaging and fast volumetric sTeFo 2P Ca²⁺ imaging. Bottom: fully automated videographic beahvioral analysis and quantification.

1) What is the neural basis of behavior? How does our brain control and generate such diversity of actions? Tackling this question requires tools that allow the interrogation of neural function at the systems level and, simultaneously, a comprehensive quantitative readout of complex behavior. To address this challenge, we have combined our high-speed bio-imaging system with unbiased videographic tracking of behavior (Fig. 1) which allows simultaneous sampling of behavioral sequences and neuronal population activity in head fixed but awake and behaving mice. We developed an unbiased, unsupervised image analysis algorithm to quantify on-going facial motor behavior in head fixed mice (GitHub), which enabled us to find a functional link between neuronal population activity at the motor cortex and face movements. (Boffi et al., 2021)

2) One of the key features of our brains is its ability to categorize sensory inputs. For example, we easily attribute different meaning to the sounds of the words ‘pea’ and ‘tea’, or to sounds coming from the left or the right side of the head versus the front when we are walking across a street. However, if we are passively listening to a lecture about food and drinks, we might find periods where we can’t figure out if the speaker is talking about peas or tea. Equally so, the lecturer might have a hard time locating the sound of a student phrasing a question from the left or right side of the auditorium. To interrogate the mechanisms by which neuronal populations reliably categorize auditory inputs we modified our fast volumetric sTeFo 2P Ca²⁺ imaging platform to include a fully automated ultrasonic speaker system which delivers auditory stimuli from flexible directions to ultrasonic hearing animal models like mice.