The correct assembly of macromolecular protein complexes such as ribosomes or multisubunit membrane channels is essential for their function. Alterations in the process can lead to disease either by loss of function or by acquisition of toxic function. However, the precise mechanisms by which multimeric protein complexes assemble are generally poorly understood owing to the difficulty in reconstituting the process in-vitro. Thus, systems that undergo dynamic assembly and disassembly in solution, such as cytoskeletal proteins, have been particularly attractive. By contrast, the assembly of multi-subunit membrane protein complexes has been notably difficult to study. An interesting, somewhat in between, situation is provided by pore-forming proteins, the best-characterized subclass of which are bacterial pore-forming toxins (PFTs). These proteins can generally be produced recombinantly and in a highly stable soluble monomeric form, but assemble into homo-oligomeric ring-like structures upon addition to cells expressing the appropriate cell surface receptor. These rings concomitantly, or subsequently depending of on the PFT, undergo a conformational change that leads to exposure of hydrophobic surfaces and spontaneous membrane insertion. The built pores then perforate the membrane of the host cell. PFTs do not only occur in bacteria and have a very broad taxonomic distribution, occurring in organisms such as parasites, plants and mammals. However, for pathogenic organisms the PFTs are major virulence factors contributing to the disease caused by the organisms producing them. For this reason bacterial PFTs have been studied extensively during the last century and a lot of insight into their mode of action has been gained. However, despite the increasing knowledge on PFTs, the mechanisms and the kinetics of self-assembly of these complexes remain largely enigmatic. In particular it is unknown whether oligomerization occurs through the sequential addition of monomers or through interaction of multimeric intermediates, whether oligomerization is the rate-limiting step during the pore-formation process and whether the same rules apply to all toxins. This is mostly due to the fact that intermediates have not been visualized by either biochemical or structural methods. Also, with few exceptions, functional assays on the activity of PFTs have always been performed at the level of a population of cells. The aim of this thesis was to measure pore-formation at the single cell level to extract mechanistic information from the analysis of the stochasticity of the pore-formation process. We chose to study three PFTs produced by human pathogens: aerolysin, PFO and Cytolysin A (ClyA). Aerolysin is produced by Aeromonas hydrophila and forms pores of 1-2 nm in diameter, perfringolysin O (PFO) occurs in Clostridium perfringensis and is a cholesterol-dependent cytolysin (CDC) that builds channels of 25-30 nm. Both of these PFTs build structurally similar pores made of transmembrane β-barrels even though the number of strands in the barrel varies by more than 1 order of magnitude. On the contrary, ClyA is a PFT produced by both E. coli and Salmonella enterica strains and forms rings of 12-13 protomers with the membrane spanning domain composed of a-helices. Hence, the pores formed by these three PFTs vary greatly from one another and are representative for the diversity of PFTs. We measured cell permeabilisation at the single cell level using two independent live-cell imaging methods, in erythrocytes and in nucleated cells. Both assays revealed that the assembly mechanism of PFTs is stochastic in its nature and that lowering the initial monomer concentration increases the variability and average time to successfully build first functional pores. The measured stochasticity reflects the chemical kinetics behind the assembly reaction and therefore carries important information about the underlying mechanism. In fact, analysis of the stochasticity of the processes led to the robust determination for each toxin of the minimal number of independent rate-limiting steps required for pore formation. Furthermore, scaling relationships in our data upon changes in initial toxin concentration constrain the possible underlying mechanisms for assembly and reveal that the steps leading to the aerolysin heptamer are composed of at least seven independent reactions whose rates scale with toxin concentration. Taken together, our findings suggest that that what is limiting for aerolysin is the conversion for each monomer to an assembly competent state, while for PFO a single limiting step is observed, likely corresponding to membrane insertion. A second aim of this thesis was to clarify the cellular consequences that occur upon pore formation by different PFTs and to find out how PFT-affected cells sense this event. This was done in collaboration with other members of the lab. Our results suggest that the cells monitor the breaches in the plasma membrane by assessing the correct homeostasis of the potassium ion inside the cell through a yet unknown mechanism. Interestingly, the cells seem to "monitor" the integrity of the membrane through the most upstream effect possible after pore formation, namely the flux of ions. Incidentally, the assays we developed in the first part of this thesis to "monitor" initial pore formation events are based on the same principle, namely the immediate loss of gradients that occur upon pore formation.