In microbiology it is common to assess bacteria by their ability to metabolize certain compounds. Bacteria are grown on particular indicator media, and if the bacterial cells are capable of metabolizing a compound in the media, the media will change to a different color. Because bacterial morphologies tend to be very simple and indistinguishable between species, such tests are important to allow bacteria to be characterized and identified by other means.

In some cases, however, the bacteria themselves may change color. Although relatively few, there are some species which may produce pigments in response to particular environmental stimuli. For example, some photoautotrophic purple bacteria produce photoactive pigments only in anaerobic or microaerophilic conditions, but remain unpigmented under oxygen-replete conditions. This color change is clearly linked to changes in the environmental conditions, as the cells alter their metabolic strategy depending on the availability of oxygen.

An even more striking example is found in the bacterium Serratia marcescens. As you have observed, this bacterium produces a bright red pigment (which historically has even been mistaken for blood) when grown at temperatures below 35-37C, but does not produce this pigment at higher temperatures, resulting in a pale off-white color. Moreover, it has further been observed that the intensity of this pigment per cell is greater when the bacteria are grown at higher densities, and growth in nutrient-poor media reduces the intensity, regardless of population density. Whereas in the case of photoautotrophic bacteria the production of a pigment is easily explained by its role in photosynthesis, the production of an apparently superfluous pigment in Serratia is less easily attributed to any metabolic purpose.

It is currently thought that the pigment produced by Serratia, called prodigiosin, does not have any direct function, but rather is simply a byproduct of other processes. The relationship between cell density and pigment production has led some scientists to propose that this pigment is a secondary metabolite associated with "quorum sensing" gene expression patterns. Quorum sensing is a means by which bacterial cells communicate. Each cell releases signaling molecules into the environment, which other cells of the same species can detect. As the density of the population increases, so does the abundance of the signaling molecules. These signaling molecules in turn regulate gene expression in such a way that when a certain abundance of signaling molecules (and thus cells) exists, the cells alter their gene expression to change their metabolic strategy accordingly. This allows bacterial populations to maximize their efficiency in dealing with a rapidly changing environment, potentially even cooperatively coordinating their metabolic processes. In this model, as the Serratia, population size increases, their gene expression pattern changes accordingly. One of the consequences of this is that some new metabolic pathways are initiated, a byproduct of which is the prodigiosin pigment.

While this model may adequately explain the relationships between pigment intensity and population size as well as nutritional state, the change in pigment expression with temperature remains puzzling. Perhaps it is noteworthy that the optimal growth range for Serratia marcescens is 25-37C, and it is above this temperature range that pigment production ceases. It is possible that the loss of pigment is associated with mild temperature stress. Alternatively, it may be that temperature also plays a role in regulating the expression of the prodigiosin pathway, although some scientists assert that the evidence suggests the decreased pigmentation is a physiological rather than genetic response. Obviously more research is necessary before a satisfying answer will be found.

Incidentally, Haddix and Werner (2000) used these properties of Serratia marcescens to propose classroom experiments which illustrate changes in gene expression with environmental changes. The experiments they propose seem interesting, educational, and fairly straightforward. Their paper can be seen and downloaded at
http://papa.indstate.edu/amcbt/volume_26/v26-4p3-13.pdf.