Abstract
The frameshift hypothesis is a widely accepted model of bird wing evolution. This hypothesis postulates a shift in positional values, or molecular-developmental identity, that caused a change in digit phenotype. The hypothesis synthesized developmental and paleontological data on wing digit homology. The "most anterior digit" (MAD) hypothesis presents an alternative view based on changes in transcriptional regulation in the limb. The molecular evidence for both hypotheses is that the MAD expresses Hoxd13 but not Hoxd11 and Hoxd12. This digit I "signature" is thought to characterize all amniotes. Here, we studied Hoxd expression patterns in a phylogenetic sample of 18 amniotes. Instead of a conserved molecular signature in digit I, we find wide variation of Hoxd11, Hoxd12, and Hoxd13 expression in digit I. Patterns of apoptosis, and Sox9 expression, a marker of the phalanx-forming region, suggest that phalanges were lost from wing digit IV because of early arrest of the phalanx-forming region followed by cell death. Finally, we show that multiple amniote lineages lost phalanges with no frameshift. Our findings suggest that the bird wing evolved by targeted loss of phalanges under selection. Consistent with our view, some recent phylogenies based on dinosaur fossils eliminate the need to postulate a frameshift in the first place. We suggest that the phenotype of the Archaeopteryx lithographica wing is also consistent with phalanx loss. More broadly, our results support a gradualist model of evolution based on tinkering with developmental gene expression.