Development of bat flight: Morphologic and molecular evolution of bat wing digits

Forelimb Digits of Fossil and Modern Bats Are Highly Similar.

First, we compared the wing elements of the earliest known fossil (Fig. 1a) and modern bats (Fig. 1b) by using a morphometric analysis of length and width data from the forelimb bones of several fossil bats and representative species from many modern bat families. Independent regressions of the lengths of the metacarpals and phalanges of the third, fourth, and fifth digits (the primary supportive elements of the wing membrane) on a proxy for body size [principal component (PC)1] show that digits from extinct bats are not proportionally different from those of modern bats (Fig. 1c). As expected, PC1 was highly and evenly positively correlated with many variables (data not shown), suggesting that it is highly correlated with body size and, therefore, is an appropriate body-size proxy. These findings indicate that the bat wing digit proportions have not changed substantially during the past 50 million years of evolution.

Initial Digit Condensations Are Similar in Proportion in Embryonic Bats and Mice.

To study when during ontogeny the changes in bat wing digit elongation occur, we examined embryonic stages of development and compared the bat forelimb with both the bat hind limb and mouse forelimb. Mice are thought to be similar in morphology to the ancestor of bats, and much is known about mouse limb development (21). The early cartilage condensations and segmentation (i.e., joint) patterns of the bat embryo (stage 16; Fig. 2a), as revealed by Alcian blue staining, are relatively similar in size and position to those of a comparably aged mouse embryo [embryonic day (E)12.5 or Carnegie stage 16; Fig. 2b] until bat stage 20. Examination of skeletal preparations of bat embryos from progressively later stages indicates that bat forelimb digits do not begin their rapid elongation relative to those of mouse until stage 20 (Fig. 2d).

Later-Stage Bat Wing Digits Have a Greatly Enlarged Hypertrophic Zone and Increased Proliferation Relative to Mice.

To study the histological changes that may underlie the elongation of the bat wing digits, we examined the developing cartilage at the cellular level. These analyses reveal that the major difference in the morphology of the bat growth plate is a great expansion of the hypertrophic zone of differentiated chondrocytes in the metacarpals and phalanges of the third, fourth, and fifth digits, relative to mouse (Fig. 2c). In mice, the metacarpal hypertrophic zone comprises, at its maximum, ≈12% of the growth plate (Fig. 2c). At stages 18 and 19, bat forelimb metacarpals have a similar percentage of the growth plate occupied by the hypertrophic zone (Fig. 2 c and d). Strikingly, by stage 20, the relative size of the hypertrophic zones in the third, fourth, and fifth bat metacarpals have increased to >30% of the bat growth plate (Fig. 2d). This abrupt increase in the extent of the bat hypertrophic zone corresponds with the onset of exponential elongation in the bat forelimb digits (Fig. 2d). In addition to increased chondrocyte hypertrophy, bat forelimb digits also display increased rates of proliferation within the growth plate at stage 20 relative to bat hind limbs and mouse forelimbs of comparable stages, as revealed by phospho-histone H3 (Ser-10) Ab staining (Fig. 3a–c).

Changing the Levels of BMP Can Alter Embryonic Bat Digit Length.

To test the ability of bat forelimb digits to respond to gain or loss of Bmp signaling, we cultured the bat embryonic handplate in control media or media supplemented with either Bmp2 or Noggin protein. Cultures containing Bmp2 caused a significant increase in metacarpal length compared with the contralateral bat forelimb cultured in control media (average increase, 239 μm; P = 0.003). Also, Bmp2 increased the relative proportion of the growth plate composed of the hypertrophic zone compared with controls (average increase, 16.9%; n = 10, P = 0.028). In contrast, Noggin treatment resulted in significantly shorter metacarpals (average decrease, 183 μm; P = 0.015) and a significantly smaller hypertrophic zone compared with controls (average decrease, 6.8%; n = 12, P = 0.046).

Bat Forelimb Digits Express Higher Levels of Bmp2 than the Digits of Either Bat Hind Limbs or Mouse Forelimbs.

To assess whether the bat exhibits a different pattern of Bmp expression relative to mouse, we performed immunofluorescence and semiquantitative RT-PCR to assay Bmp protein localization and gene-expression levels, respectively. In mice, Bmp2, Bmp4, and a known downstream component of Bmp signaling (phospho-Smad proteins 1/5/8) are expressed at moderate levels in the perichondrium of the metacarpals and metatarsals with weaker levels within the hypertrophic zone (Bmp2) and immediately adjacent areas (Bmp4) (17–18) (Fig. 3 g and k). We detected no major differences in the general protein localization of Bmp2/4 and phospho-Smad in the bat forelimb metacarpals (Fig. 3 d, e, and i) and bat hindlimb metatarsals (Fig. 3 f and j) at stages 18, 19, 20, 21, or 22 or mouse forelimb metacarpals (Fig. 3 g and k) at E13.5, E14, E14.5, E15, and E15.5. Strikingly however, both Bmp and phospho-Smad proteins are more intense and continuous in the perichondrium of stage-20 bat forelimb metacarpals (Fig. 3 d, e, and i) than they are in the homologous elements of the bat hind limb (Fig. 3 f and j) or mouse forelimb (Fig. 3 g and k) of comparable stages. We confirmed this difference in expression level by semiquantitative RT-PCR performed on stage-20 bat and mouse metacarpals. By using 18S rRNA as a control, we found that Bmp2 expression is significantly higher (≈31%; P = 0.04) in bat metacarpals relative to mice (Fig. 3h). This increase was confirmed by using β-actin as an additional control (≈35% increase in Bmp2 expression in bat metacarpals). To determine whether this difference was specific to Bmp2, we also performed semiquantitative RT-PCR for Bmp4 and Bmp7. Our results with 18S rRNA and β-actin as controls indicate that Bmp4 and Bmp7 expression levels are not significantly increased in bats relative to mice (data not shown). Also, we examined the expression patterns of several genes that are known to be associated with the maturation of chondrocytes within the growth plate (i.e., Ihh, Pthrp, Fgfr1, Fgfr2, and ColX) by using RNA in situ hybridization, and we did not observe any notable differences between bat and mouse digits (data not shown). These data demonstrate a significant and specific increase in Bmp2 expression as well as phosphorylation of Smad proteins, the Bmp signal transducers, which also serve as a read-out of active Bmp signaling, in the bat forelimb digits.

Adult Skeletal Morphometrics.

To evaluate the digit dimensions from fossil and modern bats, we performed a morphometric analysis. Measurement data of the lengths (distal to proximal and parallel to axis of the diaphysis) and widths (at the diaphysis midpoint) of all long bones of the forelimb and hindlimb (i.e., humerus, radius, ulna, femur, tibia, fibula, metacarpals, metatarsals, and phalanges) were obtained from osteological specimens of adult bats housed primarily at the Field Museum of Natural History (Chicago) (extant bats) and the American Museum of Natural History (AMNH; New York) (extinct bats), as well as from published photographs (8). In total, 10 extant and four extinct bat species were measured. The extant bats species span a broad phylogenetic range and were as follows: Rousettus egyptiacus, Glossophaga soricina, Nycteris macrotis, Rhinopoma hardwickei, Hipposideros commersoni, Pteronotus parnellii, Natalus stramineus, Balantiopteryx io, Eptesicus fuscus, and Tadarida brasiliensis. The extinct bat species are were as follows: Icaronycteris index (AMNH specimen nos. 125000 and 39501), Hipposideros sp. (AMNH specimen nos. 10019 and 10020), Archaeonycteris trigonodon [original Senckenburg Museum (Frankfurt) 80/1379; ref. 8], Paleochiropteryx tupaidodon (cast AMNH specimen no. 107679, original Senckenburg Museum Messel Fossil Locality 10; ref. 8). Measurements were taken three times and averaged to minimize the effect of measurement error. For extant bats, three specimens were measured and the results averaged. Measurements of <150 mm were taken with Mitutoyo absolute digimatic calipers, and measurements from 150 mm to 30 cm were taken with Fowler vernier calipers. Linear measurements were log-transformed before analysis to standardize their variances.

Linear regression analyses, with the lengths of the third, fourth, and fifth metacarpal as the dependent variables and PC1 (as a proxy for body size) as the independent variable, were performed to compare the relative lengths of the metacarpals of extinct and extant bats. Only the data from extant bats were used to generate regression lines. This approach allows for comparison of the observed digit lengths of extinct bats with the expected digit lengths of modern bats of a comparable size. PC1 was determined from a PC analysis (PCA) that was performed on the correlation matrix that was obtained by using the combined length and width data of all forelimb elements of these groups. PCA is a multivariate statistical technique that summarizes the variation that is present in a data set on a series of orthogonal axes, or PCs (24). PC1 summarizes most of the variation in the data, and as a result, it is commonly highly correlated with body size. If so, PC1 can be used as an appropriate proxy for regression analysis.

Embryos.

C. perspicillata makes an excellent model taxon because of its great abundance in the wild (25), the existence of procedures for maintaining and breeding it in the laboratory (26), and its recent use in several developmental studies (27–29). C. perspicillata embryos were collected from wild-caught, pregnant females captured on Trinidad. Bat embryos were dissected, staged (30), and compared with WT C3H mouse embryos at a similar stage in development (mice were staged according to the Carnegie staging system).

Limb Cultures.

Bat metacarpals from embryos of various stages (19–22) were stripped of skin and muscle and cultured for 4 days in BGJb medium (GIBCO/BRL), antibiotic/antimycotic (Life Technologies, Grand Island, NY), and 0.1% BSA at 37°C with 5% CO₂ (19, 31). Cultures were supplemented with 500 ng/ml recombinant human Bmp2 (R & D Systems) or 500 ng/ml recombinant mouse Noggin protein (R & D Systems). The right limb from a given embryo was compared with the left limb from the same embryo that was cultured in unsupplemented media.

Tissue Processing.

For skeletal analysis, embryos at stages <20 were fixed overnight in Bouin's solution and stained for acidic glycosaminoglycans (cartilage) with Alcian blue; older embryos were fixed overnight in 4% paraformaldehyde and stained for cartilage and calcium (bone) with Alcian blue and Alizarin red, respectively. At least two bat and five mouse embryos from all stages 16–24 were examined. For histology, at least three bat and three mouse embryos from stages 18–22 were fixed overnight in 4% paraformaldehyde, and the limbs were embedded in paraffin, sectioned at 10 μm, stained in hematoxylin/eosin, and analyzed.

In situ hybridization was performed on embryos that were fixed overnight in 4% paraformaldehyde, embedded in paraffin, sectioned at 10 μm, and processed by using digoxigenin-labeled RNA probes. In all cases, mouse riboprobes with known functions in the mammalian growth plate (i.e., Ihh, Pthrp, and Bmp2, etc.) also recognized bat transcripts as shown by comparison of expression in various embryonic tissues. Immunohistochemistry and immunofluorescence were performed on frozen 10-μm sections with Abs against Bmp2/4 (Santa Cruz Biotechnology), phosphorylated Smad 1/5/8 (Cell Signaling Technology, Beverly, MA), and phosphorylated histone H3 (Ser-10) (Cell Signaling Technology). In the immunohistochemical experiments, hematoxylin was used as a counterstain. At least two bat and two mouse embryos from stages 18–22 were analyzed by immunofluorescence and in situ hybridization.

Semiquantitative RT-PCR.

To quantify Bmp2, Bmp4, and Bmp7 transcripts in bat and mouse, we performed two separate semiquantitative RT-PCR analyses, one analysis using 18S rRNA and the other analysis using β-actin RNA as a control. Metacarpals were dissected from comparably aged bat (stage 20) and mouse (E14.5, equivalent to Carnegie stage 20) embryos; samples from three bats and three mice were used. RNA was extracted (RNeasy kit; Qiagen, Valencia, CA), and cDNA was generated for each sample with the SuperScript III First-Strand Synthesis system for RT-PCR (Invitrogen).

To generate conserved primers for Bmp2, we first performed RT-PCR to amplify fragments of bat Bmp2 by using E90 bat testes mRNA (GenBank accession no. DQ279784). Degenerative primer sequences were obtained by aligning Bmp2 sequences from human (Homo), mouse (Mus), chick (Gallus), frog (Xenopus), and fish (Danio). The bat fragments were sequenced and aligned with mouse sequences to find conserved regions to generate the following primers (which span a 407-bp region) that are homologous to both bat and mouse: GCCTGCAGCAGCCAACTTG (sense) and CAGTCATTCCACCCCACATC (antisense). We followed the same procedure to generate conserved primers for Bmp4 (GenBank accession no. DQ279782) and Bmp7 (GenBank accession no. DQ279783). The conserved Bmp4 primers (which span a 429-bp region) are as follows: CTCATCACACGACTACTGGAC (sense) and GCAGTAGAAGGCCTGGTAGC (antisense). The conserved Bmp7 primers (which span a 410-bp region) are as follows: CAGGGCTTCTCCTACCCCTAC (sense) and TGACCACCCAGTGGTTGCTGG (antisense).

Semiquantitative RT-PCR using 18S rRNA as a control was performed with a QuantumRNA Universal 18S Internal Standard kit (the primers of which isolate a 315-bp fragment) according to the manufacturer's protocol (Ambion, Austin, TX) and a primer/competimer ratio of 2:8 and 42 PCR cycles.

To perform semiquantitative RT-PCR using β-actin RNA as a control, we generated conserved primers for β-actin (GenBank accession no. DQ279785) as described for Bmp2. The conserved primers homologous to both bat and mouse β-actin (spanning a 569-bp region) are as follows: CCATCCTGCGTCTGGACCTG (sense) and ACGATGGAGGGGCCGGACTC (antisense). Because β-actin is expressed at a much higher level than Bmp2 within metacarpal tissues, we used differing cycles (20 for β-actin and 55 for Bmp2) to be within the linear range. We used the following PCR conditions: 94°C for 20 sec, 55°C for 30 sec, and 72°C for 45 sec. PCR products were run on 1.5% agarose gels, and the resulting bands were quantified by using quantity one 1D analysis software (Bio-Rad). Significance of differences between bat and mouse levels were evaluated by using Mann–Whitney U tests (32).