Threespine sticklebacks usually possess a robust pelvic apparatus with attached pelvic spine. However, among those living in freshwater, there has been repeated evolution of pelvic-reduced populations, which lack pelvic spines, and which may thereby become less susceptible to low calcium and to predatory arthropods which grasp the spines (Figure 1) [13]. Chan et al. noted that these fish have pelvis-specific depletion in their expression of Pitx1 [13], a gene expressed in hindlimbs but not forelimbs of many different vertebrates. These authors found that an upstream pelvis CRE is commonly deleted within the Pitx1 gene of pelvic-reduced fish populations. They showed that a 2.5 kilobase DNA fragment including this region from normal fish could, when incorporated in a Pitx1 transgene, be used to rescue normal pelvic structures in pelvic-reduced fish (Figure 1). It appeared that the Pitx1 enhancer may be particularly sensitive to loss due to its location in a fragile, flexible, sub-telomeric region, and that deletions in the pelvic enhancer region are subject to positive selection in multiple natural populations [13]. Change of Pitx1 activity may also account for pelvic reduction in a mammal, the manatee [14].

Bats have forelimb skeletal elements proportionally longer than mice, and this difference develops mainly in late gestation. The ratio of forelimb length to crown rump length is similar for both animals at early stages of limb development but, by adult, becomes about 1:3 for mouse and 2:1 for bat (Carollia perspicillata) [15]. Cretekos et al. noted that the Prx1 gene, expressed in the embryonic forelimb, promotes forelimb bone elongation, and Prx1-null homozygous mice have forelimbs 12.5% shorter than controls at 18.5 days [15]. In particular, these authors found that mice in which an upstream 1 kilobase CRE region of the Prx1 gene is replaced with the orthologous bat sequence have forelimbs that are about 6% lengthened at late gestation, suggesting that mutational changes in the CRE may have been important in the evolution of the bat wing.

Mouse and chicken differ in their vertebral formulae (relative numbers of cervical, thoracic, lumbar and sacral vertebrae). In particular, the chicken has more neck vertebrae and fewer thoracic than mouse. This is due to evolutionary shifts (transpositions) of Hox gene expression boundaries along the axial series of somites [16,17]. Boundaries within neural tissue are correspondingly transposed [18]. Hoxc8 and Hoxa7 are two genes whose endogenous expression boundaries in neural and somitic tissues are shifted posteriorly in chicken relative to mouse.

Belting et al. compared mouse and chicken Hoxc8 CREs in their ability to direct expression of a lacZ reporter gene in transgenic mice [19]. They showed that the mouse 5′-located ‘early enhancer’ CRE directs anterior boundaries of lacZ expression in somitic and neural tissues at positions about four segments more anterior than those directed by the orthologous chicken CRE. These authors concluded that evolutionary transposition of Hoxc8 expression is due to changes in its CRE. The Hoxc8 enhancers from a variety of fish and other vertebrates display a range of mutational changes within putative transcription factor binding sites [19,20,21,22]. Some, though not all, of these provide novel lacZ expression boundaries when tested in the transgenic mouse reporter assay. These additional studies show further that there have been mutational changes within a Hoxc8 CRE during vertebrate evolution, and that these are associated with functional changes in the lacZ reporter assay within transgenic mice.

It might be expected that these findings are reproducible for other Hox genes that show transpositions in their expressions in mouse relative to chicken. However, mouse and chicken Hoxa7 CREs do not differ significantly in the anterior boundaries of neural lacZ expression that they generate in the transgenic mouse reporter assay [18]. This suggests that changes in CREs are unlikely to have been responsible for this difference in Hoxa7 expression in mouse versus chicken. Instead, change in trans-acting factors seems more likely.

McLean et al. used database analysis to identify more than 500 regions of high sequence conservation deleted in humans (hCONDELs) but present in chimp and macaque [23]. One of these flanks the androgen receptor (AR) locus. This region includes a highly conserved CRE which, taken from chimp or mouse, directs lacZ expression in transgenic mouse embryos to two androgen responsive sites: facial vibrissae and the genital tubercle [23]. Mice with a mutation in the AR coding sequence do not have penile spines, structures present in many mammals and thought to increase tactile stimulation. Castrated primates lose their penile spines and castrated mice have shortened facial vibrissae. Both effects are reversed by testosterone administration. Humans do not have sensory vibrissae or penile spines, and this is likely due to evolutionary loss of the AR vibrissae and penile spine enhancer [23]. Strictly speaking, verification requires rescue of this phenotype in human, as performed for pelvic spines in fish, though ethics preclude this.

Expression of the ebony gene generates yellow shade in Drosophila adult cuticle, and its absence in posterior parts of each segment causes a dark, melanic appearance (Figure 2). Rebeiz et al. studied African Drosophila melanogaster populations that vary in the relative widths of yellow and dark bands, where increasing darkness is a derived adaptation to high altitude [24]. They examined GFP reporter transgenes driven by a 5′-located ebony CRE to show that at least five mutations affect ebony expression in the Ugandan populations. These include both new mutations and standing genetic variations. Four of these are single base-pair changes. They suggest that new mutations combine to create an allele of large effect, and that this may be a general feature of CRE evolution, consistent with other evidence that mutations at multiple sites within CREs are responsible for evolutionary changes in gene expression [25,26,27,28]. This may readily be explained, since CREs typically contain multiple transcription factor binding sites, distributed across a few hundred base pairs or more, and each contributes to transcriptional control.

In many species of the melanogaster group, and under control of the Abd-B Hox protein, the posterior abdominal segments of the male (A5 and A6) are strongly pigmented while those in the female are not [26,29]. D. santomea has lost this pigmentation, unlike D. yakuba which occupies lower elevations on the same island habitat [30] (Figure 3). Jeong et al. showed that pigmentation in D. santomea can be partially restored by expression from a D. yakuba tan transgene [30]. Two single base mutations in the 5′-located tan CRE were found responsible for partial loss of pigmentation in D. santomea. The remainder of the loss was attributed to reduction in yellow expression due to changes at other loci that act in trans. The mutant allele in the tan CRE was found to be one of three mutant alleles that have arisen independently in D. santomea populations.

Small ‘hairs’ (microtrichia or trichomes) decorate the dorsal larval surface of most species of the D. melanogaster subgroup, but not D. sechellia (Figure 4). The shavenbaby (svb) encoded transcription factor specifies these dorsal trichomes [31]. svb is not expressed dorsally in D. sechellia due to mutations in the CRE of the svb gene [25]. Frankel et al. identified thirteen substitutions and one deletion in a 500 base pair region of high sequence conservation located within a 5′-positioned CRE of svb [28]. They performed transgene-mediated rescue experiments to show that at least five of the substitutions contributed to trichome development. Mutations at the sites were interactive and not simply additive in their effects. The authors propose, as already mentioned in section 2.2.1, that where the function of a CRE relies on multiple transcription factor binding sites, each with a small effect on expression, evolution may require changes of a large number of such sites to cause a significant phenotypic change. They also suggest that stepwise mutations, as opposed to large scale deletion/addition of an enhancer, may help avoid pleiotropic effects since each individual CRE may mediate expression in more than one spatial domain.

The yellow gene, required for production of black pigment, is expressed uniformly at low levels throughout the D. melanogaster wing, but is also expressed intensely to produce a distal anterior wing spot in D. biarmipes, a closely related species (Figure 5). Gompel et al. showed that an upstream region, the so-called ‘wing element’, regulates yellow gene expression [32]. These authors found that GFP reporter transgenes driven by the D. biarmipes wing element are expressed in D. melanogaster with a pattern rather similar to that of the wing spot in D. biarmipes. They showed that separate sequences in the wing element of D. biarmipes regulate (i) general wing expression, (ii) wing spot. Deletion and mutation studies upon the spot element showed that it contained an activator element (which promotes expression in both anterior and posterior parts of the wing) and repressor elements (which exclude the spot from the posterior wing compartment). These repressor elements, present in D. biarmipes but not D. melanogaster, have gained binding sites for the transcription factor Engrailed, which confine the spot to the anterior compartment. Therefore Engrailed has been co-opted to pattern the location of the wing spot. Gompel et al. suggest that, as a general mechanism for generating wing patterns, random mutation of ancestral CREs (including point and insertional mutations) generate transcription factor binding sites to modify wing pigment patterns by co-opting from the many transcription factors already present in spatially distinct patterns within the wing [32]. This is proposed not only as a mechanism for wing patterns but more importantly as a general mechanism to evolve novel patterns of gene expression for diverse traits in diverse organisms [32].

As already indicated (section 2.2.2), male abdominal-specific pigmentation is under the control of the Abd-B gene. Abd-B protein directly activates yellow to promote melanin formation and pigmentation in abdominal segments A5 and A6 of males.In females, bric à brac (bab), represses this pigmentation. The bab1 gene is regulated by Abd-B and sexually dimorphic doublesex (dsx) proteins which bind to a bab1 intron CRE [29]. The Abd-B/bab/dsx pathway forms a sexually dimorphic switch which was already present in the common ancestor of D. melanogaster and D. willistoni. However, the switch was subsequently utilized to regulate male specific abdominal pigmentation in the D. melanogaster lineage [26]. Williams et al. showed that D. willistoni and D. melanogaster differ in their male-specific pattern (Figure 6) due to changes in the effects of Abd-B and dsx proteins upon the bab1 intron CRE [26]. They examined a variety of GFP reporter transgenes driven by the bab1 intron CRE to show that this is not simply due to change in the number of binding motifs, but also to modification of their overall efficiency by change in the polarity and spacing of existing sites. They describe this as ‘molecular remodelling’ of a pre-existing Abd-B- and dsx-regulated CRE.

As already indicated above, male abdominal-specific pigmentation is under the control of the Abd-B gene. Jeong et al. showed that Abd-B protein activates yellow expression and pigmentation by direct interaction with specific binding motifs within the 5′-located CRE in the yellow gene [29]. They found that GFP reporter transgenes driven by the yellow CRE are only expressed in the male-specific pattern if they include these Abd-B binding sites. In a separate mechanism from that described in sections 2.2.2 and 2.3.2, male-specific pigmentation in D. kikkawai has been lost by mutations which include loss of the Abd-B binding motifs within the yellow gene CRE [29].

Prabhakar et al. showed that a rapidly evolving non-coding DNA element in human (human accelerated conserved noncoding sequence 1, HACNS1) functions as a transcriptional enhancer in mouse embryos, mediating strong expression of a lacZ reporter in the limbuds [27]. Expression is strong in the distal forelimb. In contrast, orthologous DNA elements from chimp and rhesus monkey drive only weak expression in the proximal limb. The authors propose that gain of function in HACNS1 may have influenced the evolution of human limb features, such as hand dexterity and bipedalism, by altering the expression of nearby genes during limb development. Thirteen base substitutions over an 81 base region accounted for the functional difference between chimp and human elements. More recently, it has been suggested that the changes in human HACNS1 have disrupted a repressor function rather than activated a new enhancer [33].

Fezf2 encodes a zinc-finger transcription factor required for molecular specification within layers 5 and 6 of the mammalian neocortex and its corticospinal output. Shim et al. showed that Fezf2 expression is driven by a CRE, named E4 [34]. Mice deleted for E4 showed down-regulation of Fezf2 expression in the neocortex and anatomical changes as earlier reported in Fezf2-deficient mice. Transfection experiments in cultured cells using luciferase reporters driven by E4 showed that it contained at least one functional Sox binding motif. This Sox binding element apparently arose within the E4 CRE at the junction between fish and tetrapods. The authors propose that this event may have contributed to the evolution of the neocortex in mammals and at least some other amniotes [34].

Cdx1 expression in mice utilizes both upstream and intron CREs [35,36]. Homeotic genes including Cdx1 and several anteriorly-expressed Hox genes are directly responsive to retinoic acid (RA), and contain retinoic acid response elements (RAREs) in their CREs. RA is a morphogen in chordates, though not in pre-chordates [37]. In cell culture assays upon transfected Cdx1 reporter constructs, Gaunt and Paul found that both intron and upstream RAREs are functional in mouse DNA, yet only the intron RARE is functional in chicken [38]. Database analysis indicated that the intron RARE first appeared at the transition from amphibians to amniotes (reptiles, birds and mammals), while the upstream enhancer RARE first appeared at the transition from marsupial to eutherian mammals (Table 1) [38].

There is no direct evidence linking acquisition of Cdx1 RAREs to morphological evolution of vertebrate groups, but these authors noted that the main sites of action of Cdx1 were also important sites of distinction between vertebrate types. Cdx1 regulates homeotic patterning of the neck vertebrae in mice, up to and including the atlas/skull level, probably by its effect upon Hox genes [39]. Like most fish, extant amphibian species differ from amniotes in lacking a distinct neck (region between the skull and pectoral girdle). These amphibians possess only one neck vertebra (atlas) and lack the rotation-permitting atlas-axis joint. Fossil evidence indicates that the latter first arose at around the amphibian/amniote split [40,41]. Acquisition of the Cdx1 intron RARE at around this time might therefore have contributed to evolution of the vertebrate neck, including ability to lift and rotate the head in early tetrapods [38]. Cdx1 also regulates patterning of the ureters and Müllerian ducts (precursors to the uterus). Marsupials differ from eutherian mammals in the layout of the ureters and Müllerian ducts. Acquisition of the Cdx1 upstream RARE at the marsupial/eutherian split may therefore have facilitated morphological evolution of the eutherian reproductive tract [38]. These proposed effects of Cdx1 upon morphological evolution remain hypotheses, awaiting experimental verification.

McLean et al. described a conserved region deleted in human (hCONDEL) from a position next to the tumour suppressor gene GADD45G [23]. When taken from chimp or mouse, this DNA region functions as an enhancer to drive lacZ expression in subventricular zones of transgenic mouse embryo forebrain. It contains forebrain-specific p300 transcription factor binding sites. GADD45G normally suppresses the cell cycle. The authors propose that this deletion provides a plausible molecular basis for increased production of particular neuronal cell types in subventricular zone regions, a suggested requirement in evolutionary expansion of the neocortex in primates [23].