My work on cell adhesion started with revealing the molecular diversity of cell-matrix adhesions sites (Zamir et al., 1999). Along this line, I identified a new type of adhesion sites, fibrillar adhesions and the mechanism of its formation (Zamir et al., 2000) (Fig. 1).
Conceptually, a fundamental implication of these results is that they indicated the tight coupling between the dynamics, molecular composition and functions of cell-matrix adhesion sites (Zamir and Geiger, 2001a and Zamir, 2016).
In order to visualize and resolve comprehansively the molcular diversity of adhesion sites I developed compositional imaging (Fig. 1) and thus revealed compositional heterogeneity within focal adhesions (Zamir et al., 2008).
We study how the integrin adheosme can give rise to diverse adheison sites, locally, rapidly and correctly -- solely be self-organization prcosses.
To address this question we investigated the state of the integrin adhesome network in the cytosol and its spatiotemporal relation with focal adhesions.
Using systematic fluorescence cross-correlation spectroscopy (FCCS), fluorescence recovery after photobleaching (FRAP) and fluorescence lifetime imaging microscopy (FLIM) measurements we found that the integrin adhesome is extensively pre-assembled in the cytosol, thereby forming multi-protein building blocks for adhesion sites (Hoffmann et al.,2014, eLife 3:e02257).
These building blocks are combinatorially diversified, confined in their size and correlate with the structural and functional organization of proteins across focal adhesions (Fig. 1).
The building blocks enter and exit focal adhesions without being altered, thus preserving their specifications and the assembly logic of the system (Fig. 1).
How can diverse adhesion sites assemble correctly despite the alternative ways their components can interact with each other and stochastic biochemical noise ? To address this question, we (in collaboration with Hernan Grecco) established high-throughput cyclic immunofluoescence (cycIF) imaging of intracellular proteins which, by serial cycles of immunolabeling, imaging and bleaching, enables to image unprecedented large number of proteins and phosphorylation states in thousands of individual adhesion sites (Fig. 2) (Harizanova et al., 2016, Plos One).
By implementing these CycIF and live cell imaging in a high-throughput manner, thousands of individual focal adhesions were imaged obtaining an unprecedented multiplexed quantification of the molecular content of these structures. Based on changes in the variances of the densities of these components and in the correlations between them, we inferred changes in noise levels in focal adhesions (Fig. 3).
To reveal high-order statistical relations between the densities of components in focal adhesions, we examined how well artificial neural networks can predict the densities of a target component based on the densities of the other components in these structures (Fig. 4).
Our results revealed that the noise in the molecular content of focal adhesions is suppressed as they grow. This noise suppression is coupled with the size and internal density of focal adhesions, but not with their age or shape. The levels of two components, zyxin and paxillin, are particularly tuned, creating a hub of order within the focal adhesion structure. Our results indicate a mechanism for the emergence of quality control as a system property of assembling focal adhesions (Fig. 5)(Harizanova et al., 2016, Plos One).
The integrin adhesome is interconnected by a dense network of interactions in which most proteins have multiple potential binding partners. While this enables the integrin adhesome to construct different adhesive devises with different properties, it also creates a fundamental challenge for confining the set of possible assembly and regulations models. Particularly, a network description of such biochemical systems (with nodes and edges corresponding to proteins and interactions, respectively) does not incorporate information about the dependencies of each interaction and therefore is not sufficient for inferring assembly and regulation mechanisms. Boolean dependencies between protein interactions can arise for example if two proteins compete on the same binding domain along a third protein or if one interaction induces a conformational change on a protein and thereby affecting allosterically its other interactions. To address this, we (in collaboration with Johannes Koester and ven Rahmann) developed approaches for system-level integration of information about interaction dependencies toward predictive protein hyper-networks. Specifically, we applied these approaches for the construction of the integrin adhesome hypernetwork and inference of potential protein complexes, paths of assembly and regulatory switches (Fig. 6).
Cell biology research is fundamentally limited by the number of intracellular components, particularly proteins, that can be co-measured in the same cell. Therefore, cell-to-cell heterogeneity in unmeasured proteins can lead to completely different observed relations between the same measured proteins. Attempts to infer such relations in a heterogeneous cell population can yield uninformative average relations if only one underlying biochemical network is assumed. To address this, we (in collaboration with Katja Ickstadt and Hernan Grecco) developed a method that recursively couples an iterative unmixing process with a Bayesian analysis of each unmixed subpopulation.
The study of cell-matrix adhesion sites is fundamentally challenged by: (1) their molecular and functional diversity, (2) their large number of components (3) the complex interaction network between their components.
Therefore, we develop methods that address these challenges and allow to study large protein networks at high spatiotemporal resolution in intact cells.
Perturbations: genetic, pharmacological and chemical-biology perturbations.
Optical methods: Multi-color live cell imaging, fluorescence correlation spectroscopy (FCS), fluorescence lifetime imaging (FLIM), fluorescence recovery after photobleaching (FRAP), fluorescence lifetime correlation spectroscopy (FLCS), cyclic immunofluorescence (cycIF) and other approaches.
Computational methods: Compositional imaging, multi-dimensional data analyses, reverse-engineering, protein hypernetworks analysis.