Fuzzy wuzzy was a bear. Andy Panda, too - giant panda related to bear, not raccoon

Stephen Jay Gould

We revel in our persistent failure to resolve certain key arguments, because all the fun is in the fighting. Whatever shall we do if Lite's moguls ever resolve the ''tastes great-less filling'' controversy, thereby reconverting the field of American TV advertising to its former unmitigated wasteland? Thus, in a perverse sort of way, we actually regret the solution of some old scientific puzzles. Many people reacted with just this tinge of sadness to the recent announcement (Nature, Sept. 12-18) that pandas really are bears after all, not raccoons. The proof involved an array of independent techniques, all using the latest methods of molecular biology. Genes don't lie; the issue is closed.

Somehow, I couldn't share this pleasure tinged with disappointment. I appreciated the Nature study for its decisive elegance, but it told me nothing I didn't already know. I have never doubted that giant pandas are modified bears. D. Dwight Davis proved this link, and debunked the alternativetie to raccoons, in our century's greatest work of comparative anatomy (The Giant Panda: a morphological study of evolutionary mechanisms). I read Davis's monograph when it first appeared during my initial year as a graduate student. He convinced me then, and his arguments remain as conclusive today. Most of my colleagues in the ''classical'' subjects of systematics and morphology hold Davis's work in equally high regard and do not doubt his principal conclusion.

The evidence of molecules has prevailed in public approbation, where earlier data of bones and muscles, equally persuasive in this case, did not. In asking why two equally impressive proofs had such different receptions, we must address some important and troubling questions about the nature of science and the styles of its translation from technical literature to public consciousness.

When Pere Armand David, the great French explorer-priest, acquired the Western world's first giant panda in 1869, he never doubted its evident affinity with bears. He even placed his specimen within the bear genus Ursus, christening it as a new species, Ursus melanoleucus. Although later continental naturalists considered the panda sufficiently distinct to merit its own genus (Ailuropoda), few ever doubted the basic affinity with bears--a tradition that has prevailed ever since among non-English-speaking naturalists.

In England, however, St. George Mivart identified the giant panda as an aberrant raccoon. (Mivart was a fine anatomist, known best for a critique of natural selection that Darwin himself took more seriously than any other challenge.) Mivart's opinion prevailed in English and American circles until Davis published his monograph--an interesting example of parochialism in science and the development of theories along national lines.

The great bear-raccoon debate would never have arisen if the deck of related carnivores hadn't contained a prominent joker--the so-called lesser panda, Ailurus fulgens. The giant panda looks so much like a bear that its sole existence would never have provoked such squabbling. But many naturalists insisted that the giant panda must be a genealogical sister to the much smaller lesser panda. And the lesser panda does, indeed, look very much like a raccoon. Thus, if the giant and lesser pandas form a coherent group, and if the lesser panda has affinities with raccoons, then we're forced to deny the commonsense link of giant panda to bear and assert the counter intuitive tie with raccoons.

Davis's assortment of this mixed deck was simple, and spot on (as the genetic data now affirm). He persuaded himself, first of all, that giant pandas could only be modified bears. His analysis (summarized below) involved an incredibly detailed 300-page scrutiny of every bump, bone, nerve, and blood vessel, accompanied by meticulous description, incisive analysis, and some of the most beautifully detailed (and colored) anatomical drawings ever published (most by Ms. H. E. Givans, who maintains a lively interest in natural history and has greatly enjoyed Davis's vindication; Davis himself died of lung cancer in 1965 at age 56).

Davis concluded, ''Every morphological feature examined indicates that the giant panda is nothing more than a highly specialized bear.'' He then recognized the simi- larities of giant and lesser panda as confusing results of independent evolution for common function, not as signs of genealogical relationship; they are, he proclaimed, ''convergences resulting from similar functional requirements; they are not similarities resulting from common ancestry.'' He regarded the taxonomic position of the lesser panda as ambiguous, but allied this species with raccoons: ''Its morphology resembles that of the Procyonidae [raccoons] more closely than that of any other family.'' In short, giant pandas are bears, lesser pandas may be raccoons, and the two pandas aren't close genealogical relatives.

The new genetic data include four independent tests (reported byStephen J. O'Brien, William G. Nash, David E. Wildt, Mitchell E. Bush, and Raoul E. Benveniste), each yielding the same an- swer--and each affirming Davis's conclusions. O'Brien and his colleagues began with DNA-DNA hybridization, an elegantly simple technique that combines all the single-copy DNA of two species by linking one strand of the double helix of species one with the complementary strand of species two. The more tightly the two strands bind, the more similarity in total genetic composition, and the closer the genealogical relationship. The second technique works from the opposite direction--detailed similarities of specific genes, rather than overall binding of the entire genetic program. For 50 genes O'Brien and his colleagues derived the same evolutionary tree that DNA-DNA hybridization had established. (This analysis didn't proceed by tedious and expensive sequencing of DNA bases, but indirectly by studying electrical properties of the proteins coded by these genes.) The third technique exploits the uncanny ability of vertebrate immune systems to rec- ognize (and reject) foreign tissue; degree of immune reactivity should measure evolutionary dif- ference--because proteins of closely related forms should be better accepted as ''self,'' those of more distant species decisively rejected. The immunological tree of the pandas, constructed about ten years ago by Vincent Sarich and others, matches exactly the hybridization and gene-distance schemes determined by O'Brien and his colleagues.

All these criteria affirm Davis's arguments (see evolutionary tree, page 44). One limb of the carnivore tree split into a bear and a raccoon lineage more than 40 million years ago. Raccoons and lesser pandas occupy one branch (with a point of division so soon after the bear-raccoon split that we can't be entirely sure about raccoon affinities for the lesser panda). The giant panda, however, didn't split from the bear lineage until much later (no more than 30 million years ago, and as little as 10 million by some estimates). The giant panda lies firmly on the branch of bears, the lesser panda, with less assurance, on the raccoon limb.

A fourth criterion--number and form of chromosomes--uses geneticinformation of a different kind to reach the same conclusion. Giant pandas have 42 chromosomes, most bears 74--a difference that seems, at first glance, to speak against close affinity. But most panda chromosomes are long, with their joining points (or centromeres) in the middle. Each of these long, ''jointed'' chromosomes is the product of fusion between two short chromosomes with joining points at or near the ends. Bear chromosomes are short and acrocentric (joining points at end)--and, you guessed it, two bear chromosomes can often be identified as a pair that produced one of the long panda chromosomes. So pandas probably evolved from bear ancestors by a relatively simple process of chromosomal fusion.

On the other hand, only two chromosomes of the lesser pan- da have recognizable counterparts in either giant pandas or bears. However, 14 lesser panda chromosomes show clear ho- mology (common structure due to descent) between lesser pandas and raccoons--although ten of these can be identified in most non-bear carnivores. Thus, lesser pandas and rac- coons share a general pattern found in most members of their order, while both giant pandas and bears have evolved a strikingly new arrangement, with the adde wrinkle of secondary fusion in giant pandas. The chromosomal evidence supports phylogenies based on genetic similarity: giant pandas are modified bears; less- er pandas may be close to raccoons.

But how do we establish relationships on evolutionary trees, andwhy does the modern evidence of molecules command assent, while classical studies of bones and tissues rarely prevail? It may seem easy at first glance: just tote up the similarities between any two groups; the more alike, the closer the evolutionary relationship. But similarities come in two basic and contrary forms, one reflecting evolutionary relationship, the other confounding it. Hence, our task is more complex: we must learn to distinguish the two kinds of similarity, identify and exclude the confounding style, and base evolutionary trees only on the confirming style.

Similarity due to descent from a common ancestor indicates evolutionary relationship; it's called homology. I am typing with finger bones used by a mouse to run or an aardvark to dig into termite mounds. The bones are simliar in form and position (despite differences in function) because all three species inherited them from the common ancestor of terrestrial vertebrates--and they record our evolutionary affinity (whereas different bones in the fin of a salmon indicate another lineage of descent). Similarity due to independent evolution for common function in separate lineages confounds evolutionary relationship; it's called analogy. The ichthyosaur, a marine descendant of terrestrial reptiles, evolved a dorsal fin and a bilobed tail with uncanny external resemblance to similar features in fishes. Such devices are essential for proper balance and locomotion in water, and several lineages have evolved them independently. But an ichthyosaur's evolutionary affinities are with bronto sauruses, not with sharks. In short, analogy must be identified and eliminated, homology recognized and used in the construction of evolutionary trees. This principle is easy enough to state, but how do we know a homology when we see one? In practice, the fundamental distinction may be very difficult--and the traditional debates of post-Linnaean natural history center upon this dilemma.

We do have one reliable guide-- easy to state as a generality but ever so hard to place into practice (until recently). Homology is the soul of deep resemblance in a world built by genea- logical connections of evolutionary history. Analogies may be striking externally, but they're always superficial. From a distance, most of us couldn't tell an ichthyosaur from a fish. But a fine study of details betrays the reptile in the ichthyosaur; the complex pathways of history cannot be erased. The ichthyosaur's ''fin'' bones are curiously modified digits of terrestrial ancestors, not the rays of fishes; the vertebral column bends downward into the tail, not upward or centrally, as in all fishes. When similarities are sufficiently complex, numerous, and independent, they must record homology. The dictates of function aren't so stringent that two separate lineages can evolve thousands upon thousands of quirky, independent similarities. Such deep resemblance must record common descent.

The basic principle for recognizing homology is therefore easy to formulate: find enough complex, independent similarities. But how to do it in practice? And now we come to the nub of our issue--and to an understanding of why we value molecular data so highly.

The classical data of natural history are form, physiology, and behavior. Unfortunately, these are the very aspects of life most subject to analogy and least likely to meet the stringent requirement of strict homology. The shortcomings of morphology as a potential guide to homology are three: First, form is the central breeding ground of convergence (analogous similarity). Common function inscribes itself in the morphology of organisms. Second, external forms that appear complex can often reflect simple causes with multiple effects. Slow down the rate of development just a bit, and a large suite of juvenile characters may appear in adults (a process called neoteny). If we counted all these similarities one by one, we might be fooled into thinking that we had resemblance sufficiently complex to identify homology--whereas the underlying cause might be simple and easily repeatable. Third, form may not be complex enough to affirm homology. This can be particularly acute in paleontology, where evidence may be confined to simple hard parts.

For these three reasons, classical morphology often fools us andbecomes a relatively poor source for detailed evidence about homology (though it remains a wonderful and indispensable ground for any number of fascinating evolutionary problems involving function and developmental history and their interaction with phylogeny).

The virtue of molecular evidence as the best criterion for evolutionary trees is now simple to grasp, but it's almost always misstated and misunderstood. Most people assume that genes are best because they are somehow more basic, more fundamental, more really what it's all about--while morphology is merely external and derivative. As a morphologist, nothing drives me more quickly to fury and frenzy than this ridiculous argument. The genes and forms of organisms are two fundamental phenomena of biology, co-equal in interest and import but different in the kind of information they convey.

Genes are the best guides for evolutionary trees, and morphol- ogies usually don't work well in detail--but not because genes are intrinsically superior as biological objects. Genes work better for one simple reason: they happen to embody the kind of evidence needed to record homology, while morphology does not. Remember the criterion: enough, sufficiently complex, independent sim- ilarities. The DNA program of organisms contains just this kind of information--millions of nucleotides building thousands of genes, each with a pattern so intricate that no two creatures could ever evolve detailed similarity of sequence independently. Molecular techniques are revolutionizing systematics (the science of classification) because we've finally found the kind of evidence we always knew we needed to recover homology. We may correctly regard the panda problem as resolved because the similarities recorded by O'Brien and his colleagues are too intricate and dense to reflect anything but homology.

Where does this statement leave D. Dwight Davis and his monograph of 1964? Was he merely lucky in his precisely correct judgment based upon classical morphology, or just a man with sufficiently good intuition that he could extract the right message from messy and imperfect data? Nothing of the sort. Davis proved that giant pandas are bears as surely as O'Brien confirmed it with molecules. Genes, as I argued above, aren't intrinsically preferable by nature. They usually serve as the best guides to homology because they embody the necessary criteria of number, complexity, and independence. Morphology often fails to display these criteria, but when it does, morpholo- gy works as well and as surely as genes. The greatness of Davis's monograph lies in its unprecedent- ed detail and its clear recognition of criteria. Davis understood brilliantly what homology required, and what morphology could and could not do. And he fashioned the tools to extract homology from data rarely used t their full promise. (Davis's personal tragedy must reside in his failure to persuade his colleagues. We are so accustomed to inconclusive results from morphology that we failed to recognize a genuine proof when it appeared before us. Since the proof lay in 300 pages of excruciating detail, the ordinary pressures of hurried lives didn't help Davis's case.)

Davis milked from morphology the very data that homology requires. Above all, he studied bones, muscles, nerves, and organs in such detail that he approached the requisite complexity yielded more easily by molecules (while most classical studies focus upon single parts or single types of evidence and may therefore fall prey to analogy). The list of panda-bear similarities became too great to record anything but homology.

More important, he understood exactly what aspects of morphologyconfound homology, and he strove to grasp their operation in pandas. He began with a simple tabulation of the differences between bears and pandas. The list became long as Davis elaborated his voluminous evidence. Its length alone might have threatened the homology of bears and pandas. But Davis then sought to understand his list not as a superficial toting, but as an expression of general principles in growth and evolution.

Davis recognized the key difference between morphology and molecules: morphology is an integrated system built by general rules of growth, not a set of independent items. If he could resolve his long list of differences as correlated consequences of a few changes in growth, then the closeness of pandas and bears would be affirmed. In other words (and in a brilliant stroke of argument), Davis proposed to use a feature of morphology that usually confounds homology--the potentially simple basis of apparent complexity--to understand the long list of bear-panda differences as multiple consequences of but a few underlying differences. The stunning success of this effort permitted Davis to see homology under the veil of an apparently impressive, but actually small, set of differences.

Davis correctly identified the key to this long list--a functional shift of panda ecology from ancestral carnivory to their present odd life (for members of the order Carnivora) as nearly exclusive consumers of bamboo. This shift entailed three major evolutionary changes--and Davis could reduce his list of bear-panda differences largely to the complex consequences of these few alterations. First, bamboo is much harder to chew and digest than meat. Pandas underwent a major change in dental and facial anatomy, increasing the size of grinding teeth, and greatly strengthen- ing the bones and musculature of face and jaws. Since morphology is an integrated system, not a set of independent structures, these changes ramified throughout the body. Selection upon a growth field that could strengthen bones and muscles in the face produced, for example, a set of correlated changes in shoulders and forelimbs--yielding the panda's characteristic ''unbalanced'' shape of overemphasized forequarters and ambling gait. If selective pressures for locomotory efficiency had been great, these non- adaptive side consequences of facial strengthening might have been eliminated, but pandas, in their bamboo forests, do not pursue active prey, need not escape enemies, and live surrounded by appropriate food. Thus, Davis concluded, selection pressures didn't arise to suppress the nonadaptive consequences of facial enlargement.

Second, pandas evolved a curious device to manipulate bamboo--a sixth digit (The Panda's Thumb of my recent book), built from the wrist's radial sesamoid bone. Davis proved that this change involved little more than simple enlargement of the sesamoid, but also entailed a set of conse- quences that gives a false impression of bear-panda differences if not properly reduced to a single change in growth pattern. Finally, pandas increased the parts of their brains that mediate these changes in feeding and food gathering.

In short, by understanding the long list of bear-panda differences as correlated consequences of only a few alterations in growth, Davis proved the deep similarity between bear and panda, and managed to recognize homology in the usually difficult data of classical morphology. He concluded that only half a dozen or so basic changes in growth might convert a bear to a panda--a result amply corroborated by genetic similarities only recently tabulated. The identification of homology requires a clear view of criteria, an attention to detail, and an understanding of proper evidence. Homology falls easily from DNA, but a master of morphology could find it just as well in the classical data of comparative anatomy. We end with some epitomizing doggerel, expressing a higher truth about pluralism in method, respect for proper canons of evidence, and correct taxonomic affinity.

Homology

Fuzzy Wuzzy was a bear--

Andy Panda, too.

D. Dwight Davis knew.

COPYRIGHT 1986 Discover
COPYRIGHT 2004 Gale Group