DISCUSSION

Our models show that a locus will have an advantage if it has an intrinsic transition rate corresponding to the environmental periodicity relevant for this locus. It is clear that an ILC environment would pose a problem if organisms only had a general, genome-wide variation rate. In an ILC environment, an organism which has 'locus-specific responses and memory systems' will have an obvious advantage. It is important to note that the type of response that is adaptive in an ILC environment is neither facultative (short-term stimulus-dependent response), nor constitutive (long-term, stimulus-independent response) but intermediate between these two; it is an intermediate-term response: a response which can be inherited for a while in the absence of the environmental trigger, but not for a very long period. Such a response can be based on a spontaneous, intrinsic transition rate, on induced transition, or on even more complex learning systems such as with heritable behavioral phenotypes. The results of the basic model we have developed, which are illustrated in Figures 1 and 3, show that the relationship between the rate of a spontaneous transition and the periodicity cycle is (see Appendix 1). This agrees with the results obtained by Leigh (Leigh 1970 ). In our basic model, which describes a typical neo-Darwinian evolutionary system, the environment is merely the selective agent; the hereditary variations can be either variations in DNA, epigenetic variations, or behavioral variations. If we know the duration of the environmental cycle for an organism living in an ILC environment we can predict the corresponding optimal . For 'small' n ( ) a corresponding high will often be found, and indicate that the inheritance system underlying the transitions is an interesting non-classical inheritance system, involving either DNA phase variations, EISs, or transmissible behaviors. We therefore predict that the identification of ILC ecological conditions will often lead to the discovery of unusual heredity systems, and that the identification of genes that behave in a non-classical manner, like genes showing very high mutation rates, may facilitate the identification of the relevant ILC ecological conditions that select for them. Another prediction is that experimental altering of the periodicity of an identified ILC environment would result in selection for correspondingly changed variation rates (or changed induction coefficients, if the trait is environmentally induced) in the relevant hereditary or developmental system.

ILC environments often result from interaction between parasites and hosts, and it is therefore not surprising that most phase variations, both those involving DNA variations and those involving EISs, have been described in pathogenic microorganisms. In most cases of which we are aware, n is several times greater than the generation time, but not orders of magnitude greater. Some periodic, intermediate-term, heritable, phenotypic transitions may be the result of EISs affecting gene transcription. In uropathogenic , the phase variation in the expression of pili protein is under methylation control. The switching between the ``on'' and the ``off'' states depends on the methylation pattern of two GATC sites in the gene's regulatory region (Nou et al. 1993). Phase transition involving an EIS is also thought to be involved in the infectious yeast , which can switch between several alternative phenotypes, and form colonies with alternative characteristic forms. One of the best studied types of switching is between white and opaque. This transition involves a dramatic change in the cellular phenotype, which is reflected in the colony's morphology and color. The frequency of switching from white to opaque occurs less frequently than in the opposite direction, and the rate of switching is affected by environmental factors. The mechanism of switching is not clear, but it seems to involve an EIS (the chromatin marking EIS) rather than a DNA sequence change (Soll, Morrow, and Srikantha 1993). It is likely that as with other parasites, switching has evolved as a response to the changing environment presented by the host's defense systems. There are many examples of phase variations in pathogenic bacteria that involve DNA changes brought about by various mechanisms such as recombination, gene conversion, slippage etc. (Robertson and Meyer 1992). As Moxon et al. (1994) have recently argued phase variations in pathogenic bacteria are an adaptation to the unpredictable environment within the host, with the genes that have products directly interacting with the host (``Contingency genes'') evolving very high mutation rates.

The results of Appendix 2, showing that the selection pressure for optimal mutation rates in a fluctuating environment is very high (), which means that strong forces facilitate the evolution of high, locus-specific hereditary variations. We would therefore expect the evolution of such systems to be very common. The existence of several types of different inheritance systems also facilitates the precise modulation of transition rates. A particular inheritance system may be more suitable than others as a response and heredity system, because it has pre-adaptation allowing more easy adjustment of the relevant phenotypic transition rates in a given environment. The selection of transition rates between phenotypic states may underlie heritable phenotypic changes in the life-history of organisms that experience ILC environments, including parasites that exploit several different hosts sequentially, organisms with seasonal polyphenisms, and phenomena such as phase variations in plants (Brink, 1962 ). Phenotypic, behavioral inheritance occurs in organisms exhibiting social learning, and ILC environment may have been an important factor in the evolution of transmitted, socially-learnt behavior (Plotkin and Odling Smee 1979).

The sensitivity of EISs to environmental stimuli, and the rapid switch which may occur from one epigenetic state to another, make the EISs cellular effective response systems as well as effective cellular memory systems. As Figure 5 shows, environmentally induced variations are more effective than ``random'' variations, and ``random'' variations behave as induced variations with a small induction coefficient. Adaptation occurs by accumulating information about the environment and transmitting this information to progeny. The environment is both the inducer of heritable variations and the selective agent. The induction by the environment may be complex and involve several stages. The important point is that the induced state can be transmitted to the offspring; both variations in DNA sequence and epigenetic variations can be induced and transmitted, but epigenetic variations are probably more common, because the environment directly modifies the epigenetic system. In the DNA system, the modified epigenetic state must somehow be transferred to the relevant DNA sequence. This seems to be a more demanding task, although examples of developmentally regulated changes in DNA are not as rare as once believed (for review see Watson et al. 1987 , and vol. 8 issue 12 of Trends in Genetics 1992).

We have argued that in an ILC environment, unicellular organisms can evolve locus-specific transition rates. Locus-specific transitions which are tissue- and stage-specific are also fundamental to the ontogeny of multicellular organisms. The evolution of EISs in unicellular organisms was probably crucial to the evolution of multicellular organisms with complex development, since EISs maintain the determined state of cell lineages. Induced heritable transitions are important in all developing organisms; random transitions may also be important in organisms with regulative development, since in such organisms, they may cause phenotypic heterogeneity which may be the basis for somatic selection (Sachs 1988 ).

Adaptation involving heritable induced variations directly links the process of physiological adjustment, which is a process at the level of the individual, with that of evolutionary adaptation, which is a process at the level of the population. Such a direct link implies a type of ``Lamarckian'' inheritance. The properties of the epigenetic inheritance systems and of the behavioral transmission systems, and the nature of the periodical fluctuations to which many types of organisms are subject, determine random and induced heritable phenotypic transition rates. The existence of heritable phenotypic variations, and the way in which the environment influences the transition from one state to another, requires a re-consideration of "Lamarckian" inheritance, not only in the context of ontogeny, but in phylogeny as well.