We present further evidence for an interactive mechanism in the formation of the spaced pattern of heterocysts in Anabaena . The evidence comes from experiments which are an extension of those described earlier, in which filaments of the alga are broken near to a proheterocyst. We argue that a proheterocyst depends upon neighbouring vegetative cells for the removal of an inhibitory substance: when the proheterocyst is deprived of these supporting vegetative cells it will be forced to regress. We showed earlier that such regressions do occur in early proheterocysts when a filament is broken on one side only. We now find that advanced pro-heterocysts can be made to regress when double breakages are performed to leave small fragments containing the proheterocysts. The probability of a proheterocyst regressing is correlated with its stage of development and with the size of the fragment: the smaller the fragment, the more advanced is the stage at which regression will occur. To formulate this we have defined developmental stages in terms of ultrastructure and compiled the results of a diversity of breakage operations with the cells at these specified stages. Certain compounds affect the spacing of the heterocyst pattern, causing it to become wider or narrower. These compounds have the predicted effect upon regression frequencies, up-holding our assumption that regressions express an underlying competitive mechanism.
Filaments of Anabaena have a spaced pattern of differentiated cells called heterocysts, which is maintained as a filament grows by the regular determination of new heterocysts. By following the growth of every cell in a filament, we have identified proheterocysts (prospective heterocysts) at their earliest appearance, and described the sequence of events in the formation of the pattern. The determination of proheterocysts obeys 2 rules: (1) that there are inhibitory zones around pre-existing heterocysts, and (2) that only the smaller daughter of a division can become a heterocyst (all divisions are asymmetrical). There are, however, certain conditions in which these rules are over-ridden, where a pattern consisting of groups of consecutive proheterocysts is seen which resolves into a normal discrete pattern. This process is highly suggestive of interaction between developing cells. We have tested this hypothesis in normal growth conditions by breaking filaments near to early proheterocysts, on the assumption that this will cause a build-up of inhibitory effect of the cell upon itself. It is found that these cells regress, losing their differentiated character and dividing. We therefore propose an interactive model for pattern formation in Anabaena.