Preview

PDF, English - main document Download (9MB) | Terms of use

Abstract

The metabolism of all living organisms, and specifically also of their smallest constituents, the cell, is based on chemical reactions. A key factor determining the speed of these processes is transport of reactants, energy, and information within the and between the cells of an organism. It has been shown that the relevant transport processes also depend on the spatial organization of the cells. Such transport processes are typically investigated using fluorescence correlation spectroscopy (FCS) in combination with fluorescent labeling of the molecules of interest. In FCS, one observes the fluctuating fluorescence signal from a femtoliter-sized sub-volume within the sample (e.g. a cell). The variations in the intensity arise from the particles moving in and out of this sub-volume. By means of an autocorrelation analysis of the intensity signal, conclusion can be drawn regarding the concentration and the mobility parameters, such as the diffusion coefficient. Typically, one uses the laser focus of a confocal microscope for FCS measurements. But with this microscopy technique, FCS is limited to a single spot a every time. In order to conduct parallel multi-spot measurements, i.e. to create diffusion maps, FCS can be combined with the lightsheet based selective plane illumination microscopy (SPIM). This recent widefield microscopy technique allows observing a small plane of a sample (1-3um thick), which can be positioned arbitrarily. Usually, FCS on a SPIM is done using fast electron-multiplying charge-coupled device (EMCCD) cameras, which offer a limited temporal resolution (500us). Such a temporal resolution only allows measuring the motion of intermediately sized particles within a cell reliably. The limited temporal resolution renders the detection of even smaller molecules impossible. In this thesis, arrays of single photon avalanche diodes (SPADs) were used as detectors. Although SPAD-based image sensors still lack in sensitivity, they provide a significantly better temporal resolution (1-10us for full frames) that is not achievable with sensitive cameras and seem to be the ideal sensors for SPIM-FCS.

In the course of this work, two recent SPAD arrays (developed in the groups of Prof. Edoardo Charbon, TU Delft, the Netherlands, and EPFL, Switzerland) were extensively characterized with regards to their suitability for SPIM-FCS. The evaluated SPAD arrays comprise 32x32 and 512x128 pixels and allow for frame rates of up to 300000 or 150000 frames per second, respectively. With these specifications, the latter array is one of the largest and fastest sensors that is currently available. During full-frame readout, it delivers a data rate of up to 1.2 GiB/s. For both arrays, suitable readout-hardware-based on field programmable gate arrays (FPGAs) was designed. To cope with the high data rate and to allow real-time correlation analysis, correlation algorithms were implemented and characterized on the three major high performance computing platforms, namely FPGAs, CPUs, and graphics processing units (GPUs). Of all three platforms, the GPU performed best in terms of correlation analysis, and a speed of 2.6 over real time was achieved for the larger SPAD array.

Beside the lack in sensitivity, which could be accounted for by microlenses, a major drawback of the evaluated SPAD arrays was their afterpulsing. It appeared that the temporal structure superimposed the signal of the diffusion. Thus, extracting diffusion properties from the autocorrelation analysis only proved impossible. By additionally performing a spatial cross-correlation analysis such influences could be significantly minimized. Furthermore, this approach allowed for the determination of absolute diffusion coefficients without prior calibration. With that, spatially resolved measurements of fluorescent proteins in living cells could be conducted successfully.