A simple screen to identify promoters conferring high levels of phenotypic noise

Abstract

Genetically identical populations of unicellular organisms often show marked variation in some phenotypic traits. To investigate the molecular causes and possible biological functions of this phenotypic noise, it would be useful to have a method to identify genes whose expression varies stochastically on a certain time scale. Here, we developed such a method and used it for identifying genes with high levels of phenotypic noise in Salmonella enterica ssp. I serovar Typhimurium (S. Typhimurium). We created a genomic plasmid library fused to a green fluorescent protein (GFP) reporter and subjected replicate populations harboring this library to fluctuating selection for GFP expression using fluorescent-activated cell sorting (FACS). After seven rounds of fluctuating selection, the populations were strongly enriched for promoters that showed a high amount of noise in gene expression. Our results indicate that the activity of some promoters of S. Typhimurium varies on such a short time scale that these promoters can absorb rapid fluctuations in the direction of selection, as imposed during our experiment. The genomic fragments that conferred the highest levels of phenotypic variation were promoters controlling the synthesis of flagella, which are associated with virulence and host-pathogen interactions. This confirms earlier reports that phenotypic noise may play a role in pathogenesis and indicates that these promoters have among the highest levels of noise in the S. Typhimurium genome. This approach can be applied to many other bacterial and eukaryotic systems as a simple method for identifying genes with noisy expression.

Figure 1. Noise in GFP expression in clones from selected and control populations.

Clones from selected populations (red) show a higher level of noise than do clones from control populations (blue) (univariate GLM, p = 0.016). Open circles indicate clones that contain the promoter sequence for the fliC gene driving GFP expression, which were significantly enriched in the selected populations. Each data point represents the coefficient of variation of the GFP expression of several thousand individual cells.

Figure 2. Histograms of GFP expression from clones exhibiting the highest level of noise in each population.

Clones from each of the ten populations were ranked according to the amount of noise in GFP expression produced. A histogram of GFP expression was plotted for a single clone from each population with the highest level of noise. Clones from selected populations (red, orange, and yellow lines) show a much higher level of noise than the control those from control populations (blue lines). Clones containing the fliC promoter are orange and a clone containing the flgK promoter is yellow.

Figure 3. Phenotypic noise in a microcolony in gfp expression from the fliC promoter.

A. An image of a microcolony containing the plasmid-borne fliC promoter driving expression of GFP. The colony was started from a single cell and grown for about 6 generations. B. A lineage tree of this microcolony with GFP expression plotted in green (light colored boxes represent high levels of GFP, and dark boxes represent low levels), illustrating the temporal pattern of switching of the fliC promoter. The image and the lineage tree are based on Movie S1.