The Structure of Evolutionary Theory

by Stephen Jay Gould

  Genomes do not consist only of genes. Sequences located between and also within gene boundaries, accounting for a large portion of the ge­nomes of higher Eucarya, are not being addressed in a similar manner, partly due to the widespread opinion that these sequences are without function . . . We propose to name all identifiable structures represented by a nucleic acid sequence (DNA or RNA) as “nuons.” A nuon can be a gene, intergenic region, exon, intron, promotor, enhancer, terminator, pseudogene, short or long interspersed element ... or any other retroelement, transposon, or telomer — in short, any unit from a few nucleotides to thousands of base pairs in length.

  Proceeding upwards, aggregates of genes can also function as units of se­lection — including, as prominent agents in evolution, chromosomes (Nei, 1987), and organelles and bacterial plasmids within cells (Eberhard, 1980, 1990).

  Organismic selection generally works with great effectiveness in suppress­ing “revolts” to organismic integrity by differential proliferation of elements from within (see Buss, 1987; Leigh, 1991). Most of the characteristic prop­erties of genomic organization and embryological development — from Ham­ilton's “gavotte of the chromosomes” in meiosis, to such phenomena as germ line sequestration and maternal determination in embryogenesis — may have evolved largely to suppress suborganismic selection, thereby assuring the in­tegrity of multicellular organisms. Meiosis itself presumably evolved to place one copy of each gene “in the same [gametic] boat,” thus converting organ­isms, rather than genes, into a primary unit of selection by the Musketeer's criterion of “all for one.” But, once achieved, meiosis must be actively guarded by organismal selection against destabilizing drivers and distorters — all to preserve what Leigh (1991, p. 258) calls “the genome's common inter­est in honest meiosis.”

  Nonetheless, the evolutionary literature abounds with cases, both “classic” and new, of meiotic drivers, chromosomal segregation distorters, and other phenomena that favor the plurifaction of individual genes or sequences (in­cluding entire chromosomes) within the genome or population of genomes — usually with negative consequences for organismal selection above. Perhaps such cases must be relatively rare in nature, and only prominent in our litera­ture for their intriguing oddity and exceptional status in the light of organ­ismal selection's usual power to suppress such “outlaws.”

  Driving genes and chromosomes use a variety of devices to increase their relative representation by suborganismal selection. Some, including the clas­sic t-allele of house mice (Lewontin, 1970), cause dysfunction in sperm car­rying the nondriving homologue; others, like the supernumerary chromosomes [Page 692] of rye, segregate preferentially into functional gametes. Werren (1991, p. 393) attributes the interest generated by these cases to implications for the hierarchical model of selection: “Driving chromosomes are of general interest in population genetics as examples of 'selfish' or 'parasitic' genetic elements. Such elements challenge the concept of the individual genome as a 'coopera­tive' unit because they gain a transmission advantage relative to the rest of an individual's genome but are often detrimental to individual organisms.”

  Werren (1991; Werren, Nur, and Wu, 1988; Werren and Beukeboom, 1993) has also discovered and developed one of the most elaborate and inter­esting cases of suborganismal selection, a testimony to the complexity of in­teraction among levels of selection as well. In the parasitoid wasp Nasonia vitripennis, a supernumerary chromosome called PSR (paternal sex ratio) has evolved “an extreme and unusual form of transmission drive” (Werren, 1991, p. 392). This chromosome, carried in sperm, induces supercondensation of all male chromosomes (except itself) into a chromatin mass before the fertil­ized egg's first mitotic division. These chromosomes are then eliminated, while PSR survives. Since wasps are haplodiploid, this elimination converts an egg that would have become a diploid female into a haploid male (with PSR). This procedure obviously gives PSR a selective advantage in transmis­sion drive because males produced from fertilized eggs will always transmit this unpaired chromosome.

  Just as obviously, organismic selection must oppose PSR, lest the en­tire population become both male and extinct. Werren (1991) had modeled conditions of maximal opposition from organismal selection. Subdivision of populations will be most effective in producing increased competition among PSR males, with reduced availability of females. But the story becomes even more complicated because suborganismal competition against PSR has also evolved by at least two devices that bias the sex ratio in a female direction (Werren and Beukeboom, 1993): (1) a maternally transmitted bacterium, called son-killer, that prevents the development of unfertilized (male) eggs; and (2) a cytoplasmically inherited agent of unknown structure and origin, that induces female wasps to produce nearly 100 percent daughters (called MSR, for maternal sex ratio). The possibilities introduced by haplodiploidy surely influence this variety and complexity in competing selection among suborganismic units — so stories this elaborate may not be common in nature. Still, as students of teratology in anatomy have always argued, we test and il­lustrate general rules by studying such cases at the limits.

  But the main weight of gene selection in nature — the category that estab­lishes a high relative frequency for the phenomenon — probably resides in cases that are synergistic with, or orthogonal to, organismic selection, and therefore not opposed by this powerful, conventional mode. Any genetic ele­ment that can propagate itself within the genome, either by iteration in tan­dem or by duplication and transposition to other chromosomes, works thereby as a vehicle of its own relative increase — and therefore as an agent of positive Darwinian selection at the genic level. If this propagation encounters no resistance at some other level (particularly by the watchful organism) — [Page 693] either because bodies don't “notice” the increase (at least while the number of genic copies remains “within bounds”), or because higher-level selection also benefits from such differential genic proliferation (if, for a hypothetical exam­ple, an X-driving chromosome helped to generate the female bias that inter-demic selection also favored) — then genic selection can be quite rapid and powerful. This general phenomenon, perhaps of great importance in evolu­tion, has acquired the unfortunate name of “selfish DNA,” as designated in two seminal papers, representing independent and simultaneous discovery, and published back to back in Nature in 1980 (Orgel and Crick, 1980; Doolittle and Sapienza, 1980). These authors proposed that such genic selec­tion, orthogonal (at first) to organismal selection, might account for most of the middle-repetitive DNA — some 15 to 30 percent of the genome in humans and Drosophila, and usually existing as tens to a few hundred copies per se­quence, with copies often widely dispersed among several chromosomes.

  Other hypotheses might explain this phenomenon, particularly as a poten­tial organismic need for enhanced levels of any products ultimately made by any gene. (In a purely organismal view, all genes may be able to proliferate, but not to fix their multiple copies unless organismic selection favors the in­crease. However, the two levels might also act synergistically, with genic drive evolving only in some genes, and for Darwinian benefit at this basal level, but with proliferation then enhanced by positive organismic selection upon bod­ies carrying more copies).

  The “selfish DNA” hypothesis includes an attractive feature, rooted in the hierarchical theory of selection, for explaining stabilization of copy number at tens to hundreds, rather than an ultimately suicidal proliferation to inevita­ble death of the organism and all gene-individuals contained therein. Genie selection may begin in the orthogonal mode, as initial increases impose no consequences upon the phenotype. But organisms must eventually take no­tice, if only for the energetic drain, and presumed slowing of ontogenetic de­velopment, imposed by replication of so many unneeded copies with every cellular division. Original orthogonality must therefore eventually yield to a situation of genic selection contrary to organismal interest. At this point, neg­ative selection at the organismic level should stabilize and limit further in­crease
— the presumed explanation, within the theory, for the intermediary copy number of middle-repetitive DNA.

  Although I regard the hypothesis of “selfish DNA” as powerful, probably correct in many cases, and therefore as our best argument for substantially important selection at the genic level, two features in its initial promotion dis­tress me because they embody (without conscious intent, I assume) the persis­tent parochialism of organismic bias, even among those who explicitly pro­mote the hierarchical alternative. First, consider the unfortunate choices of names. Proliferating genic elements have generally been called “outlaws,” “renegades,” or “parasites”; and the general phenomenon entered our litera­ture as the hypothesis of “selfish DNA.” Orgel and Crick imposed a double whammy of opprobrium in the title of their original article: “Selfish DNA: the Ultimate Parasite.” The only reason that I can imagine for such derogatory [Page 694] terms resides in the unstated (and probably unconscious) notion that benefits for organisms define the ultimate goal and purpose of evolution as a general phenomenon. Thus, anything that can evolve, but either hurts the or­ganism actively, or even just manages to sneak past organismal scrutiny, must be designated as selfishness, nastiness, or even usurpation — as promoted by some reprobate object that would place its own propagation above the gen­eral good of evolution.

  Surely, we must reject such parochial thinking and terminology. Propa­gating genic elements should not be described as parasites or renegades; nor can they be defined as “selfish” in any meaningful or general sense. Rather, propagating genes follow the Darwinian imperative at their own level, and therefore act as any good Darwinian agent “should” — that is, to increase their own representation within their own environment, the genome in this case. As Darwinians, we should honor their pluck in such a difficult endeavor (for organisms do tend to be watchful and suppressive), rather than heaping derogatory terms upon them. Such genes could only be deemed “selfish,” “parasitic,” etc., from a false and limited perspective that values the organism alone as an agent of evolutionary success. After all, we don't call a peacock selfish for evolving such a beautiful tail, and thus limiting the geological lon­gevity of the species.

  To fully embrace the hierarchical model, a concept that marks a fundamen­tal shift in theory, not just an interesting new wrinkle upon an unaltered con­cept of nature's basic construction, we must reconceptualize all of evolution, and revise both our worldview, and our language, accordingly.

  Second, even in terms of our conventional focus on organisms, genic selection may provide crucial and indispensable flexibility for evolution of any substantial organismic novelty, including features conventionally placed in our most vaunted category of “increasing complexity.” The general argument has become traditional in evolutionary theory (since the pioneering book of Ohno, 1970), and represents a solution to the following, otherwise dis­abling, paradox: Organismal selection on the earth's original prokaryotic biota might have constructed an optimal cell, “mean and lean” as could be, with a single copy of each gene to make, in the best possible way, one product indispensable for cellular success and propagation. But how could such an in­flexible organism ever change beyond minor adjustment to altered environ­mental circumstances? As Ohno wrote (1970): “from a bacterium only nu­merous forms of bacteria would have emerged.” But duplicated copies can provide requisite redundancy, permitting one copy to manufacture the needed product, while others become free to change — and to add new functions, thus providing a potential route to increasing complexity.

  But if selection only works at the organismic level, and our “mean and lean” bacterial prototype has attained an optimal configuration, what pro­cess provides evolution with the multiple copies needed for flexible addition of functions? We gain nothing from noting that duplications provide later blessings, since evolution cannot operate for the benefit of unknown and unpredictable [Page 695] futures, unless our basic view of scientific causality needs funda­mental revision, and the future can determine the present.

  Hierarchical selection provides the most promising exit from this substan­tial paradox: multiple copies cannot originate for future organismic benefit, but they can evolve by present genic selection! (Later exaptive utilization in the generation of organismal complexity illustrates the important historical principle that reasons for origin must be sharply separated from current util­ity — see Chapter 11 for extensive discussion. Evolution continually recycles, in different and creative ways, many structures built for radically different initial reasons.) In 1970, Ohno wrote with great prescience: “The creation of a new gene from a redundant copy of an old gene is the most important role that gene duplication played in evolution.”

  Thus, if duplication requires genic selection in many or most cases, then the first level of evolution's hierarchy not only operates with respectable relative frequency, but even provides an indispensable boost for generating the sum-mum bonum of our deepest prejudices — the complex organism, with even­tual evolution of a single strange mammalian species endowed with a unique capacity for self-reflection, but occupying an isthmus of a middle state, a good vantage point for looking down with thanks to duplicating genes, and up with awe to a tree of life that could generate such an interesting and acci­dental little twig.

  THE CELL-INDIVIDUAL I speak here not of free-living unicells (where cell and organism represent the same unit of evolutionary individuality), but of cells that generally house full genomes and form the environment of genes at the level below, while also serving as parts and building blocks of multi-cellular organisms at the level above. From our limited viewpoint as highly complex metazoans built by intricate and integrated programs of embryo-logical development, we tend to neglect this intermediary level of differential cellular proliferation (not just to build bigger organs in the somatic environ­ment, for such a process yields no evolutionary reward in competition with other cells for representation in future generations, but rather to gain prefer­ential access to the germ line, and thus to achieve evolutionary success by positive selection at the cell level). We neglect this subject because positive se­lection now so rarely occurs at this level in complex metazoans — and for a reason continually emphasized in this chapter: the effectiveness of multi-cellular organisms in suppressing the differential propagation of subparts as a necessary strategy for maintaining functional integrity, the definitive property of individuality at the organismal level.

  This suppression has been so effective, while the consequences of failure remain so devastating, that human organisms have coined a word for the cell lineage's major category of escape from this constraint, a name with power to terrify stable human organisms beyond any other threat to integrity and persistence — cancer. I suspect that we would learn much more about this large class of diseases (mistakenly viewed by most of the public as a single entity) if [Page 696] we treated the subject in evolutionary terms as a historical result of the cell's initial capacity, retained from its phylogenetic past as an entire organism, for differential proliferation over other cells (formerly competitors as separate organisms, not compatriots as components of other organs). Of course, mod­ern human cells that escape this constraint do themselves no ultimate good, for they have no access to the germ line, and their unrestrained growth even­tually eliminates both their own lineage and the entire surrounding organism. To this extent, the organism's general strategies do eventually prevail, follow­ing an initially successful assault by a cell lineage. But what a pyrrhic victory! Nonetheless, the double effectiveness of a virulent cancerous cell lineage — crowding out in place and distant metastasis to other locations in the body — recalls the more “benign” strategies of other successful evolutionary plurifiers within a constrained space (genie proliferation by tandem duplication and transposition; budding off of new demes and “capture” of existing demes by immigration and transformation).

  If selection at the level of cell lineages now plays only a minor role in most groups of multicellular org
anisms, we should not view this hierarchical level as intrinsically impotent, but rather as historically suppressed in “the evolu­tion of [multicellular] individuality,” to cite the title of Leo Buss's seminal book on this intriguing subject (Buss, 1987). In Buss's terminology selection upon cells must now unfold in the “somatic environment,” where suppres­sion reigns in the service of organismic integrity, whereas such selection once occurred in the “external environment,” where unicellular organisms could experience the full independence and competitive range of Darwin's world. (In fact, since most organisms on earth remain unicellular — see Gould, 1996a, on the persistence of the bacterial mode throughout the history of life — this transition has never occurred for the vast majority of organisms on earth.)

  This cellular level therefore provides our best demonstration that the cur­rent evolutionary hierarchy in styles of individuality arose both historically and contingently, and not with necessity as a timeless, predictable, invari­ant consequence of natural law. Levels have surely been added sequentially through time, as Buss has emphasized. If life began with naked replicators at the genic or subgenic level, then these earliest times for life may have featured, uniquely for this initial interval, the property that strict Darwinians have tried so hard to impose upon our richer world of modern life — selection at one level only. The evolution of cells led to a tripartite hierarchy that charac­terized most of life's 3.5 billion year history, and still regulates the majority of earthly organisms: genes, cells, and clones. The evolution of sexual reproduc­tion added species, while the complex processes that constructed the multi­cellular individual then added the organism (the body that encloses cells and cell lineages).