Dr. Khaled Khairy
EMBL Heidelberg

Meyerhofstraße 1, 69117 Heidelberg, Germany

khairy@embl.de

Research interests / Publications / Scientific software

Research interests

Cell Biophysics

- Mechanics of the cell membrane and the cytoskeleton

What is the basis for cell and tissue morphology? The answer lies in the mechanics of cell membranes and the associated cytoskeletal and other structures. The human red blood cell (RBC) represents an ideal candidate for the study of membrane mechanics in a cell biological context because it is readily available and contains no organelles or large cytoskeletal structures. Instead, the membrane is associated with a "soft" cytoskeletal network of the protein spectrin. Using state-of-the-art experimental and computational techniques, we showed that the bending energy-based theoretical model for the lipid membrane bilayer is able to explain observed three dimensional RBC geometries that have been segmented from three dimensional confocal microscopy images. The models explain emergence of the typical canonical shapes of RBCs (cup-shaped, biconcave disc, and spiculated).

Confocal microscopy recordings (left) and theoretically predicted (right) canonical RBC shapes. Color code for echinocyte: local mean curvature. Bar 5 micron.

Moreover, we showed that disruption of the spectrin network leads to the formation of the elliptocyte, an elliptical biconcave structure (see image below). This is indeed seen in the human blood disease hereditary elliptocytosis, where the spectrin dimers are not able to form the tetramers required for a healthy cytoskeletal network. Our theoretical prediction is that with the lack of the shear resistance provided by such a network, the default shape of the human RBC is not the discocyte, but the elliptocyte.

RBCs after treatment with low concentrations of urea (which disrupts spectrin) turn into elliptocytes

Theoretically predicted shape (elliptocyte), using the bending and bilayer couple models, matches the medical condition induced above.

For more details see: Khairy K., J. Foo, J. Howard, "Shapes of Red Blood Cells: Comparison of 3D Confocal Images with the Bilayer-couple Model",Cellular and Molecular Bioengineering, 1:173-181 (2008). Also: K. Khairy and Howard J."Minimum-energy vesicle and cell shapes calculated using spherical harmonics parameterization" Soft Matter 7(5):2138-2143 (2011).

- Cell motility

Collagen fibers within the extracellular matrix lend tensile strength to tissues and form a functional scaffold for cells. Cells can move directionally along the axis of fibrous structures, in a process important in wound healing and cell migration. The precise nature of the structural cues within the collagen fibrils that can direct cell movement are not known. We found that structural features of collagen on the scale of a few nano meters are sufficient to induce directional motility of mouse dermal fibroblast cells. We analyze cell morphodynamics on two-dimensional collagen surfaces that have been manipulated to produce predefined patterns and fiber spacings.

Directed motion of a fibroblast cell on a collagen surface that has been aligned in the arrow direction. Bar:40 micron

For related articles see:

Poole K., K. Khairy, J.Friedrichs, D. Cisneros, J. Howard, D. Mueller, “Molecular-scale Topographic Cues Induce the Orientation and Directional Movement of Fibroblasts onTwo-dimensional Collagen Surfaces”,Journal of Molecular Biology, 349:380-386, (2005). (Note: the first three authors contributed equally to this work).

and

Jiang, F., K. Khairy, K. Poole, J. Howard, and D. Mueller, “Creating nanoscopic collagen matrices using atomic force microscopy”, Microscopy Research and Technique, 64:435-440, (2004).

- Biological Morphometrics

“Morphology is the fundamental observation in biology.” {Quoted from: MacLeod N., “Morphology, Shape and Phylogeny”, Taylor and Francis, London 2002). It is a macroscopic representation of the organization and physiological state of organelles, cells, tissues and organisms. However, a unified mathematical description of morphology, which would lead to a more quantitative biology, remains elusive. To remedy this situation, we are investigating new methods for biological morphometrics. Biological surfaces are often topologically equivalent to the sphere and thus the new approach uses an economic description (with no symmetry restrictions) which is a parametric Fourier series expansion, the spherical harmonics parameterization. These are powerful functions that can describe highly complicated geometries with high fidelity using only very few parameters.

Triangulated surface and its parametric spherical harmonic model

The spherical harmonics parameterization is able to describe complicated shapes with high fidelity. Grey: fitted shapes. Green: original mesh.

For related articles see:

- Mayer J., K. Khairy, J. Howard, "Drawing an elephant with 4 parameters", American Journal of Physics (accepted).

- Khairy K., J. Howard, “Spherical Harmonics-Based Parametric Deconvolution of 3D Surface Images using Bending Energy Minimization” Medical Image Analysis, 12(2): 217-227, (2008).

Image processing

Biological structures are increasingly being imaged in three dimensions using fluorescence and other microscopy techniques. The resulting datasets are large and often difficult to analyze quantitatively. The three main categories of problems that we work on are a) segmentation-less image analysis for reducing data size, b) image segmentation, with the main goal of determining the number, geometries and positions of objects, and c) image fusion, which is necessary --for example-- when samples have been imaged from different angles in Light-Sheet based Fluorescence Microscopy (LSFM).

- Image Segmentation using diffusion gradient vector fields

The main idea is that the second derivative of the intensity image is simulated as a force that gives rise to a “flow” vector field, which is smoothened by diffusion. Voxels are simulated to move along the flow directions, thus defining the (sinks) centers of biologically interesting objects. After finding these object centers, local adaptive thresholding is performed, followed by fitting to surface harmonics in order to obtain the desired object geometries and their renderings.

- Image Segmentation using biophysical priors and Markov-Chains

In this method, a Bayesian inference framework is established, which assigns a (posterior) probability to a current model describing the structures inside the image.. To maximize this probability, a Markov-Chain Monte-Carlo approach is used. The key is to take advantage of a) mechanical and biological image priors (such as the smoothness of a surface, the bending energy of a biological membrane, or the expected number of objects), and b) the use of an economic description of the unknown geometries to reduce the number of fitting variables to a manageable size. To accomplish this we take advantage of the spherical harmonics parameterization. This type of three dimensional image segmentation yields parametric 2-manifold surfaces.

For more details see:Khairy K., E. Reynaud, E. Stelzer, "Detection of Deformable Objects in 3D Images using Markov Chain Monte Carlo and Spherical Harmonics", (peer-reviewed) Conference Proceedings MICCAI, NY 2008.

- Image Fusion in object parameter space

Image fusion becomes necessary when an object has been imaged from different angles, yielding different views that are complementary. These views must subsequently be fused to obtain one enhanced image that contains all relevant information. Classically, image fusion involves image registration followed by the selection of voxel intensities based on some quality criteria. In our view this is not necessary when the final goal is not just the fused image, but a segmentation. For this case, we developed a method that depends on segmenting each view, interpreting the objects in terms of spherical harmonics parameterizations and “fusing” the different instances of the object using its parametric representation.