Increasing hierarchical complexity throughout the history of life: phylogenetic tests of trend mechanisms

—The history of life is punctuated by a number of major transitions in hierarchy, defined here as the degree of nestedness of lower-level individuals within higher-level ones: the combination of single-celled prokaryotic cells to form the first eukaryotic cell, the aggregation of single eukaryotic cells to form complex multicellular organisms, and finally, the association of multicellular organisms to form complex colonial individuals. These transitions together constitute one of the most salient and certain trends in the history of life, in particular, a trend in maximum hierarchical structure, which can be understood as a trend in complexity. This trend could be produced by a biased mechanism, in which increases in hierarchy are more likely than decreases, or by an unbiased one, in which increases and decreases are about equally likely. At stake is whether or not natural selection or some other force acts powerfully over the history of life to drive complexity upward. Too few major transitions are known to permit rigorous statistical discrimination of trend mechanisms based on these transitions alone. However, the mechanism can be investigated by using ‘‘minor transitions’’ in hierarchy, or, in other words, changes in the degree of individuation of the upper level. This study tests the null hypothesis that the probability (or rate) of increase and decrease in individuation are equal in a phylogenetic context. We found published phylogenetic trees for clades spanning minor transitions across the tree of life and identified changes in character states associated with those minor transitions. We then used both parsimonyand maximum-likelihood-based methods to test for asymmetrical rates of character evolution. Most analyses failed to reject equal rates of hierarchical increase and decrease. In fact, a bias toward decreasing complexity was observed for several clades. These results suggest that no strong tendency exists for hierarchical complexity to increase. Jonathan D. Marcot* and Daniel W. McShea. Department of Biology, Duke University, Box 90338, Durham, North Carolina 27708-0338. E-mail: [email protected] and [email protected] *Present address: Paleontology Section, University of Colorado Museum, UCB 265, Boulder, Colorado 80309 Accepted: 8 December 2006 Organisms are hierarchical, consisting of nested associations of individuals among individuals (Wimsatt 1974, 1994; Salthe 1985; Valentine and May 1996). Free-living eukaryotic cells are composed of former prokaryotic cells, multicellular individuals are composed of eukaryotic cells, and colonies are composed of multicellular individuals. In evolution, it is clear that the hierarchical maximum—the degree of nestedness of the hierarchically deepest organism in existence—has increased over time, most obviously in the transitions from prokaryote to eukaryotic protist, to multicellular individual, and then to colony (McShea 2001a). The trend is significant for several reasons. First, these transitions have led to evolutionary novelties, allowing organisms to take advantage of previously unexploited resources (Knoll and Bambach 2000), which in turn have led to spectacular evolutionary radiations such as those of the protists and metazoans. Also, hierarchy is one of two components in the biological revolutions that Maynard Smith and Szathmáry (1995) have called the ‘‘major transitions’’ in evolution (the other being change in the way information is transmitted from generation to generation). Further, a trend in hierarchy may be the main source of the widespread intuition that complexity, in some sense, increases in evolution (McShea 1996). And finally, this is currently one of the best documented macroevolutionary trends. Indeed, to our knowledge, it is the only trend in a well-defined feature of organisms that has been formally documented at the largest scale, life as a whole over its entire history (McShea 2001a). Body size obviously increases as well (Bonner 1988), at least partly as a consequence of the hierarchy trend (because higher-level individuals tend be larger 183 HIERARCHICAL EVOLUTION FIGURE 1. Schematic representations of two alternative mechanisms underlying the increase in maximum hierarchical structuring. Each triangle represents the diversification of a group at a given level. (Differences in shading among triangles are for visual clarity and have no other significance.) Arrows show transitions from one major level to an adjacent one. Black, right-pointing arrows show major upward transitions (increases), i.e. the origin of groups at higher levels from ancestors at lower levels, and gray, left-pointing arrows show major reversals or downward transitions (decreases). For example, the bold arrow in A shows the origin of a group of highly individuated multicellular organisms from a single-celled ancestor; this particular transition marks the advent of (highly individuated) multicellularity, and therefore corresponds to a rise in the maximum for life as a whole. (The dotted line in A represents the overall trajectory of the maximum.) In A, all arrows point right, meaning that change is biased, with increases more probable than decreases. In B, change is unbiased, meaning that increases and decreases are equally probable. than lower-level ones). And there are other promising candidates for longest-term trends, such as energy intensiveness (Vermeij 1987) and absolute fitness (Van Valen 1984). But only the hierarchy trend has been formally demonstrated at this scale. The trend is well known, but the processes or mechanisms underlying it are not. In particular, the question of whether hierarchical change is biased toward increasing nestedness has been central in the discussion of possibilities (McShea 1996, 2001b; Gould 1996). If the trend were the result of biased change, evolutionary increases would be more probable than decreases. For hierarchy, this would mean that, for example, the number of instances in which a solitary multicellular organism evolved to form a colonial individual outnumbered those in which the members of a colony returned to a solitary existence in the course of their evolution. And the same would be true for other transitions across the hierarchy spectrum. Figure 1A shows a hypothetical trend resulting from such biased change. In contrast, if change were unbiased, increases and decreases would be equally likely, for example if solitary arises from colonial just as often as the reverse. A trend occurs because, at the lowest level, further decrease is blocked by a boundary, a lower limit on hierarchical structure. For organisms, the natural candidate for a lower limit is the level of the prokaryotic cell. Bacteria did not arise as associations of lowerlevel living organisms, so far as we know, nor has any lower-level organism arisen from any of a bacterium’s components. If so, then from a starting point at the bacterial level, hierarchical structure could only have increased. (Even if bacteria are not the lowest level—perhaps it is viruses, if they are alive—presumably some lower limit exists.) In this scenario, the trend is simply an indirect result of the increase in species diversity, producing a diffusive or passive spread of species away from the hierarchy lower limit. Figure 1B shows a trend produced by unbiased change. With hierarchy on the horizontal axis, the vertical dashed line on the left marks the putative lower limit, or what has also been called a ‘‘left wall’’ (Gould 1996). At issue in the biased-unbiased distinction is whether or not this fundamental trend has 184 JONATHAN D. MARCOT AND DANIEL W. MCSHEA been powered in some sense—whether there are forces at work in evolution tending to drive hierarchy up. Biased change would be expected if, for example, organisms with greater hierarchical depth enjoyed some selective advantage (Knoll and Bambach 2000). Perhaps hierarchical organization is favored because of the opportunities it offers for division of labor, through differentiation of component lower-level individuals (e.g., Bonner 1988). There are other possibilities. It might be that losses of hierarchical structure are prevented by constraints or opposed by selection. The suggestion has been made that lower-level individuals tend, with time, to become integrated and more dependent on the upper-level individual they constitute, making a return to a solitary existence more difficult (Szathmáry and Maynard Smith 1995; McShea 2002). Or it could be that hierarchy is actually driven upward by constraints of some kind, that some unknown feature of the developmental process tends to produce more increases than decreases. No precise constraint-driven mechanism has been articulated (although see Salthe 1993), but it cannot be ruled out. (A constraint-driven mechanism has been offered for complexity in another sense, number of part types [McShea 2005a,b].) The various rationales for biased change might seem so reasonable, so compelling, as to make the alternative, an unbiased mechanism, seem improbable. Unbiased change might also seem unlikely for empirical reasons. For the highest-level transition, multicellular individual to colonial organism, reversals are known (e.g., several clades of bees [see Wcislo and Danforth 1997; Danforth et al. 2003]). But few, if any, reversals are known at the lower levels. Single-celled protists have combined to produce multicellularity at least 13 times (Bonner 1998). But there are no widely accepted cases of single-celled protists arising from multicellular individuals, at least not from the most highly individuated ones like land plants and metazoans (but see McShea 2001b). Also, bacteria have joined to produce a well-individuated higher-level organism, the eukaryotic cell, just once in the history of life. And so far as we know, the reverse has never occurred. No solitary bacterium has ever arisen from a eukaryotic cell. Nevertheless, an unbiased mechanism is plausible. The main reason is that the modern prokaryotic and protistan fauna is not well known, and therefore we cannot dismiss the possibility that major downward transitions have occurred but have not yet been discovered. In other words, it could be that the protists arising from a hierarchical reduction of multicellular eukaryotes, and the prokaryotes arising from a hierarchical reduction of protists, are out there, waiting to be recognized for what they are by molecular or other phylogenetic analyses. Further, relatively few major upward transitions are known, and therefore few downward transitions would need to be discovered to offset the upward ones, to produce the balance predicted by an unbiased mechanism. (For further discussion, see McShea 2001b.) Given the limited number of major transitions known, testing for unbiased versus biased change is not straightforward. One solution is to devise a higher-resolution hierarchy scale on which we can measure not only the jumps from one major level to another, but smaller, incremental changes in hierarchy as well. Such a scale would capture not only Maynard Smith and Szathmáry’s ‘‘major transitions,’’ but ‘‘minor transitions’’ in hierarchical evolution as well (McShea 2001b). We explain such a scale in the next section. It reveals minor transitions in hierarchy to be frequent enough to make statistical comparison of numbers of increases and decreases meaningful, and to make empirical investigation of the underlying trend mechanism possible. Then, we use a phylogenetic approach to analyze the patterns of minor transitions in hierarchical structure in a number of clades, distributed across the hierarchy spectrum, to test the null hypothesis that hierarchical change is unbiased. The higher-resolution scale also enables us to pose higher-resolution questions about trend mechanism. In particular, it could be that driving forces are absent in certain minor transitions but present in others. And we test for this as well. We must point out that the distinction between unbiased and biased hierarchical 185 HIERARCHICAL EVOLUTION change is a crude one, and also that it does not cover the entire range of possible trend mechanisms. First, Sterelny and Griffiths (1999) and Knoll and Bambach (2000) have raised the possibility that hierarchical evolution is unbiased, on the whole, but that diffusion toward higher levels is blocked at various points on the hierarchy scale by strong (but ultimately penetrable) ‘‘right walls,’’ representing the difficulties involved in achieving the next-level hierarchical organization. For example, they argued that hierarchical evolution of early prokaryotes was blocked by a right wall, finally breached only after more than a billion years of prokaryotic experimentation by the evolution of the eukaryotic cell (from an association of prokaryotes). Thus, biased and unbiased change mechanisms are not necessarily mutually exclusive of one another, and both may be acting at different levels within the hierarchical spectrum. Second, the unbiased-biased distinction would miss the mechanistic action, so to speak, if the trend were the result of species selection, rather than passive diffusion or any upward drive (Wright 1967; Wagner 1996). For example, it could be that greater hierarchical structure is associated with higher speciation rates, producing an increasing trend without any asymmetry in rates of hierarchical change at all. Higher-level organisms have a greater depth of parts within parts, and the resulting combinatorics could give them greater potential for combining parts in different ways, to produce a greater variety of adaptive designs (McShea and Changizi 2003). We cannot test any of these alternative mechanisms with the present data, and therefore this study should be understood as an attempt to address only a subset of the possibilities—a first pass at understanding the mechanism underlying the hierarchy trend. Materials and Methods