Fig. 1. Differential forelimb morphology in mice and bats. (A) An adult mouse, Mus musculus. (B) An adult bat, Carollia perspicillata. Digits are numbered from anterior (I) to posterior (V). Bat digits are elongated compared with mouse digits (Inset) and maintain (more ...)

Current data argue that interdigital cell death is largely regulated by bone morphogenetic protein (Bmp) signaling. Bmps are expressed in the interdigit regions during mouse and chick limb development, and inhibition of Bmp signaling suppresses cell death (1–7). Although Bmps are expressed similarly in developing webbed duck feet and in the free-toed chick, cell death and Bmp targets such as Msx2 are restricted to the distal region of duck feet. Duck feet show significant expression of Gremlin, a Bmp inhibitor, in the interdigit region, which appears to restrict the action of Bmps to the distal portion of the duck foot (8).

To determine the molecular mechanism underlying the maintenance of the interdigital tissue in bat wings, we compared the expression of Gremlin, Bmps, and their downstream targets during development of the bat forelimb (where there is little cell death) and the hindlimb (where there is significant apoptosis, resulting in free digits). We also discovered a unique domain of Fgf8 expression in the bat forelimb. To explore the importance of these signals in the regulation of interdigital cell death in the bat limb, we devised an ex vivo culture system to test the effects of manipulation of these signals. Our data demonstrate that, in the bat wing, inhibition of Bmp signaling and activation of Fgf signaling cooperate to prevent interdigital cell death.

Bmp Expression in Bat Forelimbs and Hindlimbs. We examined the expression of Bmp genes before (stage 16) and at the start of (stage 17) hindlimb interdigit regression (9). At stage 16, Bmp2 is expressed throughout the hindlimb interdigits (Fig. 2C) but is restricted distally during regression of the mesenchyme (Fig. 2D). At early stages in the forelimb, Bmp2 expression is strongest in interdigit III–IV, with lower levels in anterior and posterior interdigits (Fig. 2A). At stage 17, Bmp2 is strongest flanking the tips of digits III–V, with lower levels in the posterior interdigits (Fig. 2B). The bat Bmp2 expression differs from Bmp2 expression patterns in mice (10) and birds (2, 3, 11), where Bmp2 is expressed throughout most of each interdigit in both forelimbs and hindlimbs.

Fig. 2. Bmp pathway gene expression in developing bat limbs. Analysis of Bmp signaling components in Carollia forelimbs and hindlimbs is shown. (A–D) Bmp2 expression in forelimbs (A and B) and hindlimbs (C and D) at stage 16 (A and C) and stage 17 (B (more ...)

In contrast to Bmp2, Bmp4 shows little interdigit expression in forelimbs and hindlimbs but is localized within the distal mesenchyme, apical ectodermal ridge (AER), and digit tips and flanking the digits dorsally and ventrally, possibly in the presumptive tendons (Fig. 2 E–H). Bmp4 expression appears to be reduced or absent in bat forelimb and hindlimb interdigits compared with chicks and ducks (2, 3, 11) but is similar to Bmp4 patterns in mice (10). Bmp7 expression in bat forelimbs initiates as strong expression in every interdigit, as well as anterior to digit I and posterior to digit V (Fig. 2I). Later, the strongest expression is observed flanking the forelimb digits, with lower levels in the interdigits (Fig. 2J). In the hindlimbs, Bmp7 expression is found in the proximal interdigit and subjacent to the AER at stage 16 (Fig. 2K), similar to chicks and ducks but different from mice, where Bmp7 is expressed in all except the most distal interdigit (10). During hindlimb interdigit regression, Bmp7 is most strongly expressed flanking the digits and in distal interdigit tissue (Fig. 2L). Despite some differences in bat Bmp expression compared with mice and birds, Bmp2 and Bmp7 are expressed in both the forelimb and the hindlimb, although bat hindlimb interdigits will undergo regression, whereas bat forelimb interdigits will form the chiropatagium.

Msx genes, which are downstream targets of Bmp signaling, are expressed in interdigits before and during regression in chicks and mice and appear to play a role in interdigital apoptosis (12, 13). In ducks, interdigital cell death in the webbed foot is correlated with the restriction of Msx2 to the distal edge of the limbs (14). If repression of Bmp targets is a common mechanism to restrict cell death and generate webbed limbs, Msx gene expression in bat wings should reflect this repression and hence be absent or greatly reduced compared with bat hindlimbs. We examined the expression of Msx1 and Msx2 by using a cross-reactive monoclonal antibody as well as a mouse Msx2 riboprobe. Correlating well with the early domains of Bmp7 gene expression, we found Msx genes to be highly expressed throughout the interdigits in both forelimbs and hindlimbs before and during regression (Fig. 2 M–P). Thus, in Carollia, Msx expression domains are similar to the patterns observed in chick and mouse limbs but contrast with the pattern in the duck webbed foot. Although Msx activity is implicated in interdigital cell death, expression of Msx RNA or protein is apparently not an accurate indicator of whether interdigital apoptosis will occur.

Bat Wings Express Gremlin and Display Unique Expression of Fgf8. In the webbed duck foot, proximal expression of the Bmp antagonist Gremlin is thought to prevent Bmp-mediated cell death and restrict this activity to the distal region of the interdigits (8). We examined the expression of Gremlin to test whether it might play a similar role in the retention of interdigital webbing in bat wings. By stage 16, Gremlin is highly expressed in the two anteriormost interdigits in the forelimb and at lower levels in the posterior forelimb, whereas it is expressed proximally in all interdigit regions of the hindlimb (Fig. 3A and C). At stage 17, Gremlin is expressed in all interdigits of the forelimb, but it is largely excluded from the most distal portions of the limb (Fig. 3B). By this time, the interdigital expression in the hindlimb is lost, and Gremlin is largely found flanking the digits and at their tips (Fig. 3D). Because Gremlin is a potent Bmp inhibitor, the expression in the forelimb interdigits places Gremlin in a position to block Bmp-mediated cell death. Intriguingly, Msx gene expression in bat wings suggests that Gremlin is not able to inhibit all Bmp signaling in the forelimb interdigits or that another factor activates Msx expression there.

Fig. 3. Fgf signaling and Gremlin expression in bat limbs. (A–D) Gremlin expression in forelimbs (A and B) and hindlimbs (C and D) at stage 16 (A and C) and stage 17 (B and D). Roman numerals in A and C indicate digit number. (E–H) Fgf8 expression (more ...)

Fgfs have been proposed to be survival signals for a variety of tissues (15), and Fgf application has been shown to transiently block cell death in chick limb interdigits (16–19). In mice, expression of a gain-of-function allele of Fgf4 results in persistent webbing between the digits (20). Similarly, human mutations that activate Fgf receptors block cell death between the digits (21). Together, these experiments suggested that Fgf signaling might be a good candidate for repressing cell death in bat wing interdigits. Thus, we examined the expression of a key Fgf gene in the limb, Fgf8. As in mice (22) and chicks (23), Fgf8 is expressed throughout the AER of bat forelimbs and hindlimbs (Fig. 3 E and G; data of earlier stages not shown). However, Fgf8 is also expressed in limb mesenchyme at early stages and in the forelimb during the time when interdigit regression should occur (Fig. 3 E and G; data of earlier stages not shown). The early mesenchyme expression is similar to that reported in axolotls (24) but is strikingly different from other amniotes. Fgf8 was expressed at high levels in the interdigit region of bat forelimbs, most strongly in the posterior two interdigits (Fig. 3E) and at lower levels in hindlimb interdigits. In contrast to axolotl, Fgf8 is maintained in bat forelimb interdigits at the time of interdigital regression and lost in the bat hindlimb. (Fig. 3 F and H). The interdigit expression of Fgf8 might also account for Msx expression in the forelimb, because Fgfs have been shown to induce Msx2 expression in avian interdigit tissue (19). The interdigit expression of Fgf8 in the forelimb interdigits was largely mirrored by Sprouty2 (Spry2) (Fig. 3 I–L), which is activated in response to Fgf signaling (25). This differential expression of Fgf8 led us to hypothesize that Fgf signaling in the bat forelimb interdigit may act to specifically maintain the survival of bat wing interdigital tissue.

Functional Disruption of Fgf Signaling and Enhancement of Bmp Signaling Induces Interdigital Cell Death in Bat Wings. Comparative gene expression served as a starting point for understanding how the bat interdigital webbing is maintained. In other systems, coapplication of Fgf and a Bmp inhibitor has been shown to significantly suppress interdigital apoptosis (19). To functionally test the role of Fgf and Bmp signaling on bat wing interdigital cell death, we performed the converse experiment. We implanted into bat forelimb interdigit regions beads soaked in a Fgf receptor inhibitor (SU5402) and Bmp protein (Fig. 4A and B). The limbs were then cultured and subsequently assayed for cell death by an active caspase-3 antibody or by TUNEL. Analysis of freshly dissected uncultured forelimbs shows there is little cell death in the bat forelimb interdigit region, although there is cell death in the AER. We quantified the results of our culture experiments and found a highly significant increase in cell death compared with controls (P = 0.0003; n = 17; Fig. 4 C and D). The few active caspase-3-positive mesenchyme cells in the control limb (Fig. 3C) are a result of culturing the limbs. Thus, the combination of decreased Fgf signaling and increased Bmp signaling is able to cause the regression of the bat forelimb interdigital tissue. Single treatments with Bmp- or SU5402-soaked beads did not result in significantly increased cell death compared with controls. Taken together with the expression pattern data, our studies suggest that retention of the interdigital tissue in the bat wing is due, at least in part, to inhibition of Bmp signaling (possibly through Gremlin expression) and activation of Fgf signaling by means of a previously uncharacterized domain of Fgf8.

Fig. 4. Functional analysis of Fgf and Bmp signaling in bat limbs. (A–D) Control (A and C) and Bmp- and SU5402-treated (B and D) cultured bat limbs (stage 16 late) after a 23-h incubation. (C and D) Active caspase-3 immunofluorescence on longitudinal (more ...)

Here, we reveal a previously uncharacterized mechanism that is used in the bat wing to prevent apoptosis and interdigital regression (Fig. 4E). In the bat wing, inhibition of Bmp and activation of Fgf signaling are required to prevent interdigital cell death. Specifically, we show that high levels of Gremlin expression in the forelimb interdigits were not reflected by a difference in Msx expression in forelimbs and hindlimbs. In addition, increasing the levels of Bmps in interdigits was not sufficient to induce cell death, suggesting an additional mechanism for maintenance of interdigital tissue in the bat wing. The strong and maintained expression of Fgf8 in bat forelimb interdigits is intriguing and suggestive of an evolutionarily unique role in the maintenance of interdigital webbing. Experimentally, we show that perturbation of this balance by increasing Bmp and decreasing Fgf signaling results in extensive forelimb interdigital apoptosis. We therefore propose that formation of the webbed bat forelimb requires Fgf signaling in the forelimb interdigit domains acting in combination with reduced Bmp signaling (Fig. 4E).

The evolution of flight in bats is a matter of conjecture. The paucity of intermediate forms in the fossil record has made it difficult to ascertain the order of events that led to flight in the order Chiroptera. One hypothesis is that bat ancestors first glided by means of the lateral wing membranes (plagiopatagium), similar to flying squirrels (subfamily Pteromyinae), flying lemurs (order Dermoptera), and marsupial gliders (genus Petaurus). This evolutionary change would have required an outgrowth of tissue from the flank of the body or limbs, and the molecular mechanisms regulating plagiopatagium initiation and development in any of these groups are not known. In bats, this tissue forms in a very different manner than the chiropatagium. The plagiopatagium grows out from the flank starting at stage 14 (9). At stage 17, the plagiopatagium has begun to connect to the proximal portion of digit V of the forelimb. Fusion of the plagiopatagium to digit V is not complete until late stage 19/early stage 20. Because the outgrowth of the plagiopatagium is not due to blocking apoptosis, we assume that another mechanism regulates its morphogenesis.

The development of the chiropatagium depends partly on the retention of early interdigit tissue. Our data suggest that the modulation of Bmp and Fgf signaling plays a critical role in this process and may have been involved in the evolution of the wing membrane, with one of the key events being acquisition of a unique expression domain of Fgf8 signaling. Our data provide molecular insight into the evolution of powered flight in bats. Retention of interdigital webbing in the bat would have been an important step to developing true powered flight and likely contributed to the success of this widespread and diverse order.

Another step in the evolution of bat wings was the elongation of the forelimb digits to support the wing membrane. Flying lemurs possess both plagiopatagia and webbing between the digits but lack the elongated digits found in bat wings. It would be interesting to determine whether a similar mechanism for interdigital retention operates in flying lemurs. In bats, the length of the forelimb digits appears to be controlled by an increase in Bmp activity within the cartilage (26). It is exciting that the modulation of both of these aspects of bat wing development depends partly on changes in Bmp signaling: increased Bmp in the digits and reduced Bmp in the interdigits. Thus, changes in the temporal expression, spatial expression, and levels of expression of key developmental regulators such as Bmp and Fgf appear to be important in driving the evolution of vertebrate limbs.

Animal Use and Experimental Embryology. Carollia embryos were collected from wild-caught, pregnant females on the island of Trinidad. Limbs were removed from stage 15–18 embryos (9), and beads soaked in control or experimental solutions were implanted in the third and fourth interdigit regions of the left and right limbs from the same embryo. The limbs were cultured for 18–72 h and then fixed, photographed, and assayed for cell death either by immunofluorescence using an active anti-caspase-3 antibody (Promega, Madison, WI) or TUNEL assay (Roche, Indianapolis, IN).

Preparation of Beads. Affi-Gel Blue or formate-derivatized AG1-X2 beads (Bio-Rad, Hercules, CA) were used as carriers for the administration of Bmps or SU5402 (Calbiochem, Nottingham, U.K.), respectively. Beads with a diameter ranging between 50 and 150 μm were selected, depending on the stage of the embryo. The beads were washed in PBS or DMSO and then incubated for 1 h at room temperature in different concentrations of growth factor. Human recombinant Bmp-4 and Bmp-2 (R & D Systems, Minneapolis, MN) were used interchangeably at concentrations of 0.3 and 0.5 mg/ml, respectively (1). Control beads were incubated in 4 mM HCl plus 0.1% BSA. SU5402 was used at a concentration of 4 mg/ml diluted in DMSO (19), and control beads were soaked in DMSO alone.

Gene Expression Analysis. In situ hybridization was performed on whole-mount specimens by using digoxygenin-labeled RNA probes derived from bat or mouse sequences. Using degenerate primers (5′-TCTyTAACCTCAGCAGCATCC-3′ and 5′-CCCCTCyACyACCATCTCCTG-3′), we cloned a 784-bp fragment of the Bmp4 gene from Carollia genomic DNA and a 467-bp fragment of the Carollia Gremlin gene (5′-GGAATTCAAAGGkTCCCAAGGwGCC-3′ and 5′-TGCGGCCGCrTCGATGGATATGCAACG-3′). These sequences have been submitted to GenBank (accession nos. DQ855011 and DQ855012). We cloned a 410-bp fragment of Carollia Bmp7 from RNA extracted from a stage-13 embryonic head by using the primers that are described in ref. 26. We used mouse riboprobes, which also recognize the bat transcripts, to test for the expression of Fgf8 (27), Bmp2 (28), Msx2 (29), and Sprouty2 (25). The monoclonal antibody 4G1 (which recognizes Msx1 and Msx2), developed by Thomas M. Jessell and Susan Brenner-Morton, was obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA).

Cell Death Assays and Statistical Analysis. Active caspase-3 expression was assayed by an antibody that recognizes the cleaved active form of caspase-3 (Promega). The TUNEL assay was performed as recommended by the manufacturer (Roche). Apoptotic cell numbers were quantified by counting the percentage of dying cells in a 100-μm square distal to the implanted beads in the interdigital space between digits IV and V. Using contralateral limbs as a control and the Wilcoxon signed rank test, the average difference between Bmp- or SU5402-treated and control-treated limbs was not significant (n = 6, P = 0.345 and n = 7, P = 0.6121, respectively), whereas limbs treated with Bmp and SU5402 beads did show a statistically significant (n = 17, P = 0.0003) increase in the percentage of dying cells.

We thank Chris Cretekos, Karen Sears, Irene Zohn, and Kathryn Anderson for support, discussions, and comments on the manuscript; Simeon Williams for field assistance; the Department of Life Sciences, University of the West Indies (St. Augustine, Trinidad) and, particularly, Dr. Indira Omah-Maharaj for assistance and use of departmental facilities during the course of the fieldwork; and the Wildlife Section, Forestry Division, Ministry of Agriculture, Land, and Marine Resources (currently in the Ministry of Public Utilities and the Environment) of the Republic of Trinidad and Tobago for providing required collecting and export licenses. This work was supported by a National Research Service Award fellowship (to S.D.W.), the National Science Foundation (R.R.B.), and the National Institutes of Health (L.A.N.). L.A.N. is an Investigator of the Howard Hughes Medical Institute.