The Specialist-Generalist Spectrum: Evolutionary Drivers of Diet Breadth in Nature and Microbial Models

Diet breadth is a specific case in foraging theory and relates also to the broader issue of niche width (Futuyma and Moreno 1988). In nature we see populations or species that, relative to others, are extreme specialists or generalist feeders, with the bulk of organisms falling somewhere in the middle of the spectrum. For example, Koala bears (which technically are not bears but rather marsupials) eat Eucalyptus leaves and even show clear preferences among species of Eucalyptus. In contrast, their closest living relatives, wombats, are also herbivores but have a relatively broad diet including grasses, bark, roots, and sedges.

Rabbits and Virginia opossums are relative omnivores. An analogous dietary spectrum applies among free-living microbes. For instance, Sulfolobus, discussed previously, is a specialist at oxidizing elemental sulfur. Methylococcus lives on methane. In contrast, various species of Pseudomonas can grow on any one of many dozens of carbon sources (Clarke 1982; MacLean and Bell 2003). Sulfolobus is the Koala bear of the microbial world, and Pseudomonas is the opossum.

Since generalists have the obvious advantage of a greater range of food sources from which to choose than do specialists, the interesting question is why do populations respond to heterogeneity by evolving specialists? In early work that shaped the field, MacArthur and Connell (1966; see their Chap. 3) examined theoretically the foraging efficiency of generalists versus specialists with respect to the extent of environmental heterogeneity or ‘grain.’ Environments can be classified as 'fine-grained' or 'coarse-grained' relative to organism foraging depending on whether the resources are consumed in the proportion in which they occur (fine-grained) or selectively, e.g., where the organism selects one because of its prevalence in preference to others (coarse-grained). Grain is discussed in detail in Chap. 7 where we consider the environment. The short answer from their model is that in a coarse-grained environment specialists will have an advantage over generalists at harvesting the resource of their choice. In a fine-grained environment, the outcome may favor either but, where the resources are quite similar, the generalist will theoretically win (see related work by Levins 1968; and by MacArthur and Levins 1964; MacArthur and Pianka 1966).

The above results were interpreted by MacArthur, Levins, and Connell, among many others subsequently, within the context of the well-known adage ‘a jack-of-all-trades is master of none.’ While this maxim has intuitive appeal, at least in regard to human activities, it does not necessarily apply in ecology. The limitations on generalism usually are ascribed to either the unavoidable interference in performing different tasks equally well (inability to maximize competing functions simultaneously) or costs of maintaining characters specifically for reducing environmental variation in fitness (Kassen 2002). For example, the genetic tradeoffs in a parasite to have many hosts, or alternatively to become highly adapted to a particular host (see Sidebar and later section, Phylogeny), may drive the evolution of specialists (Chappell and Rausher 2016).

However, there is some contradictory evidence to the assumption that a generalized phenotype embodies costs constraining optimal performance (e.g., in tests of performance as a function of body temperature, a ‘jack-of-all-temperatures may also be a master of all’; Huey and Hertz 1984; see also Reboud and Bell 1997; Remold 2012). Thus, apparently, traits promoting performance in one environment can do so in others as well. Tests of the prediction of costs or trade-offs generally have produced ambiguous or at best nuanced results (recall earlier discussion under Optimal Digestion of experiment by Brown et al. 1998; for caveats and updated semantics and status of generalist/specialist theory, see Kassen 2002; Jasmin and Kassen 2007a, b; Remold 2012). The widespread assumption that being a generalist entails costs has been examined in Kassen’s review (2002) as well as by Remold (2012).

Kassen noted that a trade-off between fitness (performance) and breadth of adaptation in only one of the four cases considered. It should also be recognized that species may come to be associated with particular resources for reasons other than competition or feeding preference (for amplification and examples, see Futuyma and Moreno 1988).

With respect to specialization, the common assumption is that species or populations showing differential performance across environments do so because of trade-offs. These in turn are usually explained mechanistically as arising by antagonistic pleiotropy whereby a mutation providing for better performance in one environment is deleterious in another (Cooper and Lenski 2000; Kassen 2002; MacLean and Bell 2002). However, specialization can result from mutation accumulation in which, by genetic drift, various mutations arising in the environment where they are neutral prove harmful in another. Two populations may also independently adapt to their alternative environments by accumulating alleles that are advantageous in one while being neutral in the other (Elena and Lenski 2003; Zhong et al. 2009). Thus, there are three mechanisms for producing specialists. By analyzing data from a long-term field study of aphid parasitoids, Straub et al. (2011) found that specialists were more abundant than generalists on their shared hosts (as predicted by the ‘jack-of-all-trades is master of none’ postulate) but that the generalist fitness cost depended not on the number of utilized host species per se, but rather on their taxonomic breadth.

The generalist/specialist issue can be approached in rigorous, if somewhat artificial, experiments with microbial systems in vitro. These offer control of both the genetic makeup of the populations and the environmental variables. Dykhuizen and colleagues have done such studies for many years (e.g., Dykhuizen and Davies 1980; Dykhuizen 1990, 2016; Zhong et al. 2004, 2009) and their work is illustrative of the approach. In the 1980 study they defined the generalist as a strain of E. coli that could use both of the disaccharides lactose and maltose (lac⁺, mal⁺), whereas the specialist could metabolize maltose only (lac^-, mal⁺) because the genes for uptake and catabolism of lactose had been deleted. As a technical point, the construction of the strains involved genetic engineering of an initial (specialist) strain containing a deletion of the lactose operon that removed all of the proteins involved with lactose breakdown but no others.

The generalist was then created by transferring through phage transduction (see Chap. 2) only the genes for lactose metabolism into the specialist from a donor strain, thus constituting the lac+, mal+ generalist phenotype in a genetic background otherwise identical to the specialist. While various other strains were constructed and tested in these experiments, a key result is that when cultured separately on maltose, the specialist grew about 7% faster than the generalist, consistent with predictions. In direct competition experiments in chemostat culture under energy-limited conditions on mixtures of lactose and maltose, the strains generally coexisted, with the final ratio depending on the percent lactose in the medium.

As argued by the authors, this implies that coexistence in nature would occur under certain conditions, e.g., where the specialist uses the abundant resource in common while the resource unique to the generalist is rare, or if the selection differential on the shared resource is appreciable. Extrapolation from this system to nature is constrained by the experimental conditions, wherein chemostat culture presents to E. coli a constant, homogeneous (fine-grained) energy-limited environment, whereas in nature E. coli in the gut faces a markedly heterogeneous feast-or-famine life and an even more heterogeneous survival stage outside the body (see, e.g., Koch 1971). Nevertheless, these and similar experiments are a powerful and useful intermediary step for hypothesis-testing between theoretical models and experiments in nature.

In the older literature there are numerous studies similar to the experiments above. Typically these were conducted by comparing progeny carrying introduced point mutations (aux- otrophs) against fully functional (prototrophic) parents. The underlying hypotheses usually were that the metabolically simpler auxotroph ‘specialist’ would have a selective advantage because it would not carry the additional biosynthetic steps of what is in effect the proto- troph ‘generalist.’ Data from numerous experimental systems involving mutants support the assertion. Zamenhof and Eichhorn (1967) compared various nutritionally deficient mutants and fully functional (back mutant) strains of Bacillus subtilis in continuous (chemostat) culture in liquid media. One experiment demonstrated that a histidine requiring mutant (his^-) was competitively superior in dual culture to its histidine nonrequiring, spontaneous back- mutant (his+).

In experiments with other mutants the authors went on to show, first, that dispensing with an earlier rather than a later biosynthetic step gave an auxotrophic carrier a selective advantage and, second, that a de-repressed strain producing the final metabolite (tryptophan) in quantity was at a selective disadvantage compared with the normal repressed strain. Both outcomes are consistent with a conclusion of advantageous metabolic streamlining in specialists.

The most obvious interpretation of the foregoing is that the prototroph continues to synthesize at least some of the metabolite in question, thereby incurring energy costs, despite the availability of the chemical in the medium. These functions are completely blocked in the auxotroph. Evidently the feedback inhibition and gene repression mechanisms are incomplete in the prototroph. Zamenhof and Eichhorn went on to speculate that it may have been through energetic savings that parasitism evolved, once nutritious ‘media’ in the form of macroorganisms became available to support fastidious variants of fully functional, free-living microbes. Interestingly, the issue of operon repression was taken up in detail in subsequent experiments by Dykhuizen and Davies (1980).

They showed that on the lactose-limited medium a lac constitutive mutant (producing the lactose enzyme whether or not lactose is present; operating as essentially a generalist) replaced the lac+ specialist by enabling the constitutive to subsist on lower levels of lactose than that required for derepression of the operon in the lac+. They further demonstrated that equilibrium of lac+ and lac constitutive cells was established in the presence of both maltose and lactose, implying that the lac constitutive cells were less efficient in using maltose than the lac+ cells. This was interpreted not in terms of energy conservation or a diauxie, but rather ‘resource interference’—a circumstance when organisms are not as efficient at using a resource when it is combined with others as when alone. Here, lactose use appeared to interfere with maltose use, which they attribute likely to lack of space in the bacterial membrane for permeases of both sugars.

Work extending the above findings and incorporating genomic techniques to study E. coli adaptation to sugars over hundreds of generations has been done by Zhong et al. (2004, 2009). Here, growth-limiting concentrations of either lactulose (rather than lactose), methyl- galactoside, or a 72:28 mixture of the two were used. Across multiple experiments and isolates they describe eight unique gene duplications and 16 unique deletions in a genomic pattern that was consistent. The lac duplications and mgl mutations usually occurred in different backgrounds, producing specialists for the respective sugar, but not to the other. Interestingly, although growth in the mixed sugars provides conditions for generalists to evolve, they rarely did, rather growth was dominated by specialists. These experiments did not test the mutation accumulation hypothesis for specialist evolution (such mutants would have been eliminated by the experimental design) and of the remaining two (antagonistic pleiotropy and independent specialization) antagonistic pleiotropy was inferred to be the mechanism most consistent with the results.

Finally, similar studies with nutrient generalist or specialist populations considered the ability of E. coli to evolve simultaneously in response to distinct selection pressures over extensive periods, i.e., 6000 generations of culture. Satterwhite and Cooper (2015) tested whether replicate populations (generalists) conditioned in an environment of two resources (varying presentations of lactose and glucose) have essentially the same fitness as replicate populations grown either in glucose alone or lactose alone (specialists). They found that for the first 4000 generations, generalists were usually as fit in the individual resources as were the specialists in those resources, but this period of cost-free adaptation was followed subsequently by a cost of adaptation for all generalists. Whether such costs might eventually diminish is an open question.