Our observations and simulations support a model in which OM growth occurs in discrete insertion events that are uniformly distributed over the cylindrical cell surface, forming an inhomogeneous mixture of newer and older OM on a length scale set by the size of the insertion events. The appearance of fluorescent puncta can be attributed to a large average area per insertion event (), and simulations based on our fluid dynamics model indicate that the observed growth patterns do not require spatial organization of insertion events, mechanical stresses, or biochemical interactions between molecules. Similarly, the pattern of newly inserted porins in S. typhimurium
observed using electron microscopy consists of patches with dimensions of 50–100 nm that increase in number over time but do not increase in size 
. We observed this patchy insertion directly in our induction time-course experiments, in which longer induction times produced more puncta that were more densely packed on the cylindrical cell surface (), and indirectly in our growth experiments, in which new unlabeled material dispersed the older, labeled OM into punctate spots (). Additionally, other direct and indirect measurements indicate that active translocation events occur at discrete sites on the cylindrical cell surface 
. While the cylindrical section of the cell undergoes patchy growth, at the division septum, OM material that was once part of the actively growing cylindrical portion of the cell becomes trapped in the inert polar region. Our experiments showed that at short induction times, where new insertion events resulted in new fluorescent puncta, LamB puncta appeared more often in the cylindrical portion of the cell than at the poles (), indicating that the rate of insertion was indeed lower at the poles. This observation is consistent with our bulk label-and-chase experiments that showed faster dilution of the fluorescence distribution in the cylindrical region, and, after a few generations, the emergence of relatively bright “old" poles that resulted from the lack of growth in the polar region after the initial labeling ( and ). While our data cannot be used to rule out the possibility that groups of molecules much smaller than the diffraction limit are occasionally being inserted, when combined with the mechanistic insights from our model, our data suggests that the formation, evolution, and movement of fluorescent OM puncta is well explained by fairly large (~10−2
) discrete insertion events that are uniformly distributed over the cylindrical cell surface, with an exponentially distributed duration that has a single, fixed time constant.
Within the production and transport pathway of proteins bound for the OM, several possible molecular mechanisms could produce the punctate growth pattern observed here. First, nearly all OM proteins (with the known exception of PulD 
) are transported from the periplasm, through the rigid cell wall, and inserted into the OM by BAM complexes, and hence one possible interpretation of our data is that the appearance of fluorescent LamB puncta reflects the location of active BAM complexes and OM protein insertion in general. A second, independent possibility is that the uniformly distributed Sec apparatuses 
, responsible for translocating proteins from the cytoplasm to the periplasm where BAM complexes then effect OM insertion, are only active at specific locations, and hence determine the pattern of protein insertion in the OM. A third interpretation arises from the recent discovery that chromosomally expressed mRNAs remain localized to their transcription sites and form diffraction-limited spots, with mRNA diffusion coefficients orders of magnitude lower than expected for free diffusion 
. While this observation has not been replicated for plasmid-expressed mRNAs, several multicopy plasmids have been shown to localize to a handful of foci in E. coli
. The mean duration of an OM insertion event (2 to 30 sec) is considerably shorter than the half-life of E. coli
mRNA (~400 sec) 
, thus bursts of cell growth are likely not due to transient mRNA production. However, we speculate that the combination of plasmid localization and mRNA localization to a few foci per cell could give rise to localization of secreted LamB. This list is by no means exhaustive; as data on the lifetimes of candidate molecular complexes (e.g. BAM and Sec) become available, the regions of parameter space we have identified through screening of computational results () could provide a useful means of distinguishing among possible mechanisms producing the patchy patterns of OM protein insertion.
Our observations suggest that bacterial cell surface organization is strongly influenced by growth patterning. Calculations based on ferritin labeling of newly translocated LPS in Salmonella
predict that an LPS molecule would diffuse less than 300 nm 
, or approximately one twelfth of the cylindrical circumference, on the time scale of bacterial cell division. These ferritin-labeled LPS molecules initially appear in clusters assumed to be membrane translocation sites. Electron micrographs taken at regular intervals after labeling revealed that these clusters gradually disperse from each other across the surface of Salmonella
, beyond the 300 nm diffusion limit. In comparison to the doubling of OM area that occurs on this time scale, it is reasonable to expect that the motion of OM proteins is dominated by patterns of OM incorporation, not diffusion.
The spatial pattern of protein expression dictates not only the distribution of proteins within a single cell, but also the distribution of OM proteins in a population. The unequal partitioning of proteins into daughter cells can produce heterogeneity among isogenic bacterial populations, and in a growing bacterial population, the fraction of cells harboring ‘old poles’ shrinks exponentially. If an OM protein is expressed and subsequently repressed early in the growth curve, polar retention will cause it to be concentrated at these few old poles. In this manner, OM protein inheritance could generate heterogeneity based on generational age that has important phenotypic consequences for differential permeability to essential nutrients, drug concentration, or phage infection. The agreement between our simple model and the observed motion of LamB puncta suggests that heterogeneity may be a general outcome in any bacterium in which the OM grows in discrete bursts.