Connexin 40, a Target of Transcription Factor Tbx5, Patterns Wrist, Digits, and Sternum

Abstract

Haploinsufficiency of T-box transcription factor 5 (TBX5) causes human Holt-Oram syndrome (HOS), a developmental disorder characterized by skeletal and heart malformations. Mice carrying a Tbx5 null allele (Tbx5^+/Δ) have malformations in digits, wrists, and sternum joints, regions where Tbx5 is expressed. We demonstrate that mice deficient in connexin 40 (Cx40), a Tbx5-regulated gap junction component, shared axial and appendicular skeletal malformations with Tbx5^+/Δ mice. Although no role in skeleton patterning has been described for gap junctions, we demonstrate here that Cx40 is involved in formation of specific joints, as well as bone shape. Even a 50% reduction in either Tbx5 or Cx40 produces bone abnormalities, demonstrating their crucial control over skeletal development. Further, we demonstrate that Tbx5 exerts in part its key regulatory role in bone growth and maturation by controlling via Cx40 the expression of Sox9 (a transcription factor essential for chondrogenesis and skeleton growth). Our study strongly suggests that Cx40 deficiency accounts for many skeletal malformations in HOS and that Tbx5 regulation of Cx40 plays a critical role in the exquisite developmental patterning of the forelimbs and sternum.

Holt-Oram syndrome (HOS), an intriguing human dominant disorder characterized by upper limb malformations and congenital heart disease, is caused by haploinsufficiency of a member of the T-box family of transcription factors, TBX5 (6, 23, 29). While structural heart malformations occur in approximately 85% of Holt-Oram patients, wrist and digit malformations are universally found, and are required for a clinical diagnosis (7, 8, 23, 37). Some skeletal manifestations of HOS are visible on inspection, including a finger-like or absent thumb, foreshortened arms, and sloping shoulders that poorly abduct or adduct. Radiologic examination often reveals many more malformations in bones of the wrist (e.g., misshapen scaphoid and carpal fusion), deformation of the humeral head, maldevelopment of the ulna and radius, and abnormal sternum configuration. The low incidence of HOS (estimated at 1 per 100,000; 16) and limited understanding of Tbx5 downstream target genes have, however, hindered full investigation of this T-box transcription factor in skeletal development.

We previously generated a conditional mutant allele of the Tbx5 gene in mice (13). Homozygous Tbx5-null mice die early in embryogenesis (before embryonic day 10.5 [E10.5]) due to severe cardiac developmental defects. Heterozygous Tbx5^+/Δ mice genetically recapitulate dominant human mutations that cause HOS (7, 13, 23). Tbx5^+/Δ mice are live born but have a variety of cardiac abnormalities. In the heart, Tbx5 haploinsufficiency results in reduced expression of genes encoding connexin 40 (Cx40), atrial natriuretic peptide (13), and presumably other genes required for normal heart development. Genes regulated by Tbx5 in the developing forelimbs or the consequences of their altered expressions remain unknown.

Cx40 (also known as Gja5) is a member of a family of over 20 highly related genes encoding the intercellular channel-forming proteins found in gap junctions (for reviews, see references 18 and 45). Intercellular channels allow the direct movement of ions and small molecules between the cytoplasm of adjacent cells. Selective expression of connexins in specialized cells of the heart is known to contribute to cardiac development and is essential for normal adult cardiac electrophysiology. In the cardiac conduction system, Cx40, Cx43, and Cx45 (20, 27, 33, 42) are essential for normal cardiac electrophysiology. Cx43 and Cx45 are involved in production of bone matrix proteins in osteoblastic cells, but none of them have been involved in the patterning of the skeleton (28, 36). While it has not been ppreviously recognized to participate in skeleton formation, we hypothesized that altered expression of Cx40 in Tbx5^+/Δ mice might contribute to upper limb and thoracic skeletal malformations found in HOS. Fortunately, Cx40-deficient mice (42) are viable and provide a useful tool for defining the role of Cx40 in skeletal formation.

To test the role of Cx40 in Tbx5-mediated skeletal formation, we characterized Tbx5 and Cx40 expression in bones and examined the skeletons of Tbx5^+/Δ, Cx40^+/−, Cx40^−/−, and compound Tbx5^+/Δ Cx40^+/− mice. Our study reveals a requirement for connexin 40 in segmentation and patterning of particular axial and appendicular bones. Due to the similar skeletal phenotypes of Tbx5, Cx40, Sox9, and transforming growth factor β (TGF-β) signaling-deficient mice, we further investigated the possible involvement of TGF-β family members and transcription factor Sox9 in these mispatterning events. Taken together, our data suggest that Tbx5 regulates exquisite patterning of the digits, wrist bones, and sternum via Cx40-containing gap junctions.

MATERIALS AND METHODS

In situ hybridization and histology.

Newborn mice and E12.5 to E17.5 embryos were dissected free of the uterine muscle and studied. E0.5 was defined as noon on the day postcoitus when a vaginal plug was detected. Embryos were removed, washed once in phosphate-buffered saline (PBS), fixed overnight in 4% paraformaldehyde in PBS, and either frozen (−20°C, 100% methanol) or kept in 70% ethanol (room temperature) prior to paraffin embedding. Embedded sections were stained with hematoxylin and eosin or used for in situ hybridization, performed as described previously (3). Plasmids containing Cx40, Tbx5, collagen X (ColX), and Sox9 sequences were used as templates for digoxigenin (DIG)-labeled riboprobes, which were produced according to the manufacturer's specifications (Roche). DIG-RNA probes were hybridized overnight at 70°C and incubated with anti-DIG-AP Fab fragment, and signal was detected using Nitro Blue Tetrazolium and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside substrates (Roche). Analyses of whole-mount specimens was complemented by in situ hybridization of paraffin wax tissue sections using Cx40, Tbx5, ColX, Sox9, PTHrP, Ihh, bone morphogenetic proteins (BMPs), Gdf5, and BmprIB radioactive probes labeled with [³⁵S]UTP according to previously described protocols (41).

Skeletal preparations.

Skeletal preparations (43) were obtained as follows. After removal of external tissues from embryos or adult mice, specimens were washed in PBS and fixed in 95% ethanol (EtOH) for 5 days. Samples were then stained in Alcian Blue solution (75% EtOH, 20% glacial acetic acid, 15 mg/100 ml Alcian Blue 8GX [Sigma]) for 2 days, fixed for 5 days in 95% EtOH, and stained with Alizarin Red solution (250 ml KOH [2%] with 12.5 mg Alizarin Red S [Sigma]) for 2 additional days. The stained skeletons were dissected and bones measured using the NIH Image Program. Comparisons were made between age-matched mutant and wild-type animals. No skeletal differences were observed between hybrid mice (129SvEv and C57BL/6 backgrounds) and inbred mice with the 129SvEv or C57BL/6 backgrounds.

Immunoblot analysis.

Tissues from E13.5 embryo mice were homogenized using a hand tissue grinder (25 strokes) in lysis buffer (0.5% NP-40, 50 mM Tris-HCl [pH 8.0], 200 mM NaCl, 20 mM NaF, 20 mM β-glycerophosphate, 1 mM dithiothreitol, protease inhibitor cocktail [Roche], phosphatase inhibitor cocktail [Sigma]). Lysates were incubated on ice for 15 min and then centrifuged at 10,000 × g. Protein concentration was determined by the Bradford method (Bio-Rad). Lysates (50 μg) were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis on a 4 to 20% gel (Pierce), transferred onto Hybond-P membranes (Amersham), and probed using antibodies (Cell Signaling) against mouse Smad4, phospho-Smad1 (Ser463/465), Smad5 (Ser463/465), Smad8 (Ser426/428), and phospho-Smad2 (Ser465/467) or using TGF-β2 (Santa Cruz Biotechnology) antibody. The ECL blotting system (Pierce) was used to visualize proteins.

Cx40^−/− (homozygous null) mice (C57BL/6 background) (42) were bred to Tbx5^+/Δ mice (129 SvEv background) (13). Intercrosses of the fertile F₁ progeny (Tbx5^+/Δ Cx40^+/−) produced wild-type (Tbx5^+/+ Cx40^+/+), single-mutant (Tbx5^+/Δ Cx40^+/+, Tbx5^+/+ Cx40^+/−, and Tbx5^+/+ Cx40^−/−), and compound-mutant (Tbx5^+/Δ Cx40^−/−) mice. Genotypes were determined by separate PCRs using previously published protocols (13, 42).

RESULTS

Tbx5 and Cx40 expression in the skeleton.

Tbx5 is expressed throughout the forelimb mesenchyme from the earliest stages of development, a pattern that is consistent with a role in limb bud initiation (1, 21, 31, 38). Later in development (E15.5), when mesenchymal cells have migrated to sites of skeleton formation and begun to aggregate into condensed blocks, Tbx5 expression was observed in the periarticular surfaces of the sternum and wrist bone perichondrium (Fig. 1A and E).

FIG. 1.

Skeletal expression of Tbx5 and Cx40 in wild-type (Wt) and Tbx5^+/Δ mice. (A to D) In situ hybridization with antisense ³⁵S-labeled Tbx5 (A and B; black) and Cx40 (C and D; red) riboprobes to the sternum and ribs from E15.5 wild-type (A and C) and Tbx5^+/Δ (B and D) embryos. Note Tbx5 (A and B) and Cx40 (C and D) expression in the perichondrium surrounding sternal bands and where ribs contact the sternum (asterisk). Cx40 was also expressed in the rib perichondrium. (E and F) In situ hybridization of ³⁵S-labeled Tbx5 and Cx40 probes to wild-type wrist sections demonstrated colocalization of Cx40 and Tbx5 expression (dIV, digit IV; H, hamate; L, lunate). Hybridization with ³⁵S-labeled sense probes revealed no signal in wild-type and Tbx5^+/Δ mice (data not shown). (G to J) Whole-mount in situ hybrid-ization to E12.5 wild-type (G and I) and Tbx5^+/Δ (H and J) embryos using digoxigenin-labeled antisense Cx40 riboprobe. Note Cx40 is expressed in the wild-type sternal band (G, arrows), where Tbx5 is also detected (not shown), but its expression is decreased in the sternal band of Tbx5^+/Δ embryos (H, arrows). Similarly, Cx40 is expressed in the developing scapula and forelimb region (I, arrow) of wild-type embryos but its expression is diminished in Tbx5^+/Δ embryos (J, arrow).

Cx40 expression was identified in the developing ribs, migrating sternal band (Fig. 1G, arrows), prescapula (Fig. 1I, arrow), and fore (wrists, Fig. 1E) and hind limbs (not shown). Within the axial skeleton Tbx5 and Cx40 were coexpressed in the sternal perichondrium (Fig. 1C and D, asterisk), but only Cx40 was expressed in the rib perichondrium. In the appendicular skeleton, both Tbx5 and Cx40 colocalized to the forelimbs and carpal bones (Fig. 1E and F and data not shown), structures that are most often malformed in HOS. Cx40 expression was less abundant than Tbx5 expression in these bones (compare Fig. 1E and F).

In comparison to wild-type embryos, heterozygous Tbx5^+/Δ embryos exhibited significantly reduced cardiac Cx40 expression (compare Fig. 1G and H; 13). Cx40 expression was also decreased in the migrating sternal band and prescapula regions (compare Fig. 1G with H and I with J) of day E12.5 Tbx5^+/Δ embryos. By day E15.5, Cx40 expression in wild-type and Tbx5^+/Δ embryos were not discernibly different by in situ hybridization (Fig. 1C and D).

To investigate whether abrogated expression of Cx40 contributes to the skeletal malformations that characterize Tbx5 haploinsufficiency, we first examined skeletal morphologies of Tbx5-deficient (13) and Cx40-deficient (42) mice. Remarkably, these mutant mice shared a variety of malformations in the bones of the wrist, digit, and sternum.

Shared skeletal malformations in Tbx5^+/Δ and Cx40-deficient mice.

Like humans, mice have eight carpal bones that provide exquisite flexibility of the wrist (25). Six of these (trapezium, lunate, scaphoid, hamate, capitate, and trapezoid) (Fig. 2A) showed a variety of malformations in Tbx5^+/Δ mice (Fig. 2B). More than 70% of the mutant wrists contained fused bones, with the scaphoid and the trapezium involved in almost all fusions (>95%) (Fig. 2D). In addition to the wild-type (absence of fusion) formation, four different fusion patterns of carpal bones were found in Tbx5^+/Δ mice (Fig. 2B): trapezium-scaphoid (Tm-S), scaphoid-lunate (S-L), trapezium-scaphoid-lunate (Tm-S-L), and absence of trapezium. The frequency and laterality of these malformations were determined from evaluations of 76 Tbx5^+/Δ, 58 Cx40^+/−, and 74 Cx40^−/− wrists (Fig. 2D). Carpal fusions were not symmetrical in the left and right wrists of the same animal, but no significant difference was observed in the severity of phenotype in the right versus left (not shown).

FIG. 2.

Carpal bone and digit abnormalities in Tbx5^+/Δ and Cx40 mutant mice. (A to D) Skeletal preparations of carpal bones from adult mice of the indicated genotypes stained with Alcian Blue and Alizarin Red, which delineate cartilage and bone, respectively. (A) Schematic representation of carpal bones from a wild-type (Wt) mouse identifying six major carpal bones: H, hamate; C, capitate; Tm, trapezium; Td, trapezoid; S, scaphoid; L, lateral cuneiform. (B) Ensemble of carpal bone patterns observed in Tbx5^+/Δ wrists. Each schematic representation identifies a carpal bone fusion, indicated in grey. Tm-S fusions occurred most frequently. One animal had no Tm (arrow). (C) Carpal bone fusions observed in Cx40^+/− and Cx40^−/− wrists. Schematic representations of fused bones (green) demonstrate three fusions involving the trapezium (Tm): Tm-Td, Tm-C, and Tm-Td-C (U shaped). (D) Distribution of carpal fusions observed in wrists of mutant mice. The histograms represent the combined frequency of fusions observed in both left and right wrists because no differences between frequencies of left and right wrists were observed. Note that fusions involve the Tm in 95% of Tbx5^+/Δ (grey) and 100% of Cx40 mutant (green) mice. Fusions in Cx40 mutant mice often involved the trapezoid (64 to 94%) and rarely involved the capitate (12 to 50%) bones. Fusions in Tbx5^+/Δ mice always (100%) involved the scaphoid and rarely (11%) the lunate bones. In compound Tbx5^+/Δ Cx40^+/− mice, the Tm fused to both proximal and distal carpal bones (not shown). (E to G) Altered phalanges and metacarpal bones in mutant and wild-type digits. (E) Schematic representation of metacarpal bones and phalanges elongated in Tbx5^+/Δ (grey striped) or Cx40 (green) mutant mice compared to the wild type (clear). m, metacarpal bones; pp, proximal phalanges; ip, intermediate phalanges; dp, distal phalanges. (F) Metacarpal bone (m) and proximal phalanges of digit I from adult Tbx5^+/Δ mice are markedly elongated compared to age-matched wild-type littermate. (G) Alcian Blue- and Alizarin Red-stained bone preparations from paws of mutant and wild-type mice. The metacarpal bone and proximal phalange of digit one exhibit a premature ossification center in newborn Tbx5^+/Δ mice (denoted by arrow) compared to wild-type and Cx40^−/− mice (flanking). (H to J) Sox 9 expression in wild-type, Tbx5^+/Δ, and Cx40^−/− digits at E13.5 as assessed by in situ hybridization. Note more proximal and higher expression of Sox9 (darker) in the phalangeal region of Tbx5^Δ/+ (I) and Cx40 (J) mutant mice compared to their wild-type littermate (H).

Analyses of carpal bone structure in Cx40^+/− and Cx40^−/− mice also revealed abnormalities. The patterns were identical in heterozygous and homozygous null mice (Fig. 2C), but the frequency was lower (28%) in Cx40^+/− mice than in Cx40^−/− mice, where over 70% had malformations (Cx40^+/− versus Cx40^−/−, P < 0.0001). As in Tbx5^+/Δ mice, Cx40-deficient mice had a variety of carpal bone fusions. The trapezium was involved in all Cx40^−/− fusions (n = 68) with the trapezoid and/or capitate (Fig. 2D). Four carpal bone formation patterns were observed: wild type (absence of fusion), trapezium-capitate (Tm-C), trapezium-trapezoid (Tm-Td), and trapezium-trapezoid-capitate (Tm-Td-C).

Digit structures of Tbx5^+/Δ and Cx40-deficient (Cx40^+/− and Cx40^−/−) mice were abnormal. In the mouse, digit I, like the human thumb, consists of a metacarpal bone (m) and proximal (pp) and distal (dp) phalanges, while digits II to V consist of a metacarpal bone and three phalanges (proximal, intermediate [ip], and distal) (Fig. 2E). The metacarpal bones of Tbx5^+/Δ, Cx40^+/−, and Cx40^−/− mice from all five digits are elongated compared to the wild type (stripes in Fig. 2E; Table 1). Metacarpals (digits II to V) were approximately 15 to 30% and 10 to 25% longer in Tbx5^+/Δ and Cx40-deficient mice, respectively, than in wild-type bones (Table 1). The lengths of Cx40^−/− and Cx40^+/− metacarpals were not significantly different from one another, but two intermediate phalanges in Tbx5^+/Δ mice and three intermediate phalanges in Cx40-deficient mice were elongated (Fig. 2E; Table 1).

TABLE 1.

Morphometric analyses of wild-type, Tbx5^+/Δ, Cx40^+/−, Cx40^−/−, and Tbx5^+/Δ Cx40^+/− mice

Bone or parameter	Wild type	Tbx5+/Δ	P vs wta	Cx40+/−	P vs wt	Cx40−/−	P vs wt	Tbx5+/Δ Cx40+/−	P vs wt
nc	10	12		14		14		12
Digitsb
Intermediate phalange
II	6.87 ± 0.8	7.41 ± 0.27	0.04	7.70 ± 0.53	0.01	7.94 ± 0.77	0.00	7.97 ± 0.54	0.00
III	8.25 ± 0.56	8.43 ± 0.37	0.38	8.97 ± 0.24	0.00	9.14 ± 0.69	0.00	8.99 ± 0.43	0.00
IV	7.97 ± 0.71	8.49 ± 0.33	0.03	8.79 ± 0.33	0.00	8.90 ± 0.72	0.00	8.62 ± 0.43	0.03
V	6.01 ± 0.72	5.86 ± 0.37	0.57	6.17 ± 0.32	0.48	6.05 ± 0.60	0.87	6.30 ± 0.44	0.25
Proximal phalange
I	3.08 ± 0.63	6.55 ± 0.44	0.00	3.62 ± 0.57	0.04	3.56 ± 0.57	0.07	6.99 ± 0.37	0.00
II	11.20 ± 1.0	11.42 ± 0.54	0.53	11.55 ± 0.83	0.35	11.91 ± 0.82	0.07	12.40 ± 0.61	0.00
III	12.24 ± 0.85	12.41 ± 0.84	0.64	12.93 ± 0.51	0.02	13.27 ± 0.89	0.01	13.32 ± 0.45	0.00
IV	12.09 ± 0.81	12.42 ± 0.59	0.28	12.94 ± 0.49	0.00	13.00 ± 0.87	0.02	12.87 ± 0.54	0.01
V	9.77 ± 1.23	10.17 ± 0.72	0.37	9.35 ± 0.63	0.45	9.99 ± 0.7	0.59	10.59 ± 0.36	0.04
Metacarpal
I	2.77 ± 0.43	7.95 ± 0.76	0.00	3.91 ± 0.37	0.00	3.40 ± 0.56	0.01	9.06 ± 0.57	0.00
II	16.19 ± 1.16	19.25 ± 0.96	0.00	17.77 ± 0.87	0.00	18.27 ± 2.66	0.03	19.79 ± 0.87	0.00
III	19.97 ± 1.09	22.53 ± 0.83	0.00	21.49 ± 0.82	0.00	21.75 ± 1.29	0.00	23.41 ± 0.92	0.00
IV	18.68 ± 1.29	21.09 ± 0.80	0.00	19.62 ± 0.94	0.04	19.99 ± 1.11	0.01	21.92 ± 1.3	0.00
V	11.58 ± 0.98	14.94 ± 1.54	0.00	12.86 ± 0.44	0.00	12.84 ± 0.72	0.00	15.63 ± 1.37	0.00
n	36	20		12		16		13
Sternumd
2-pc manubrium	None	3 (15)		1 (8.3)		None		7 (54)
Sternebral fusion(s)
S2-S3-S4	None	2 (10)		None		None		None
S3-S4	None	11 (55)		6 (50)		10 (62.5)		10 (83)
S3-S4-X	None	4 (20)		None		None		1 (10)
Xiphoid
Bifurcated	None	19 (95)		None		None		12 (100)
Protuberant	None	10 (50)		6 (50)		9 (56)		3 (25)
n	16	12		14		16		13
Forelimbb
Scapula	59.3 ± 1.68	55.81 ± 1.30	0.00	60.48 ± 1.58	0.10	61.48 ± 1.17	0.00	58.25 ± 1.48	0.17
Scapular area	241 ± 52	306 ± 17	0.00	290 ± 12	0.00	265 ± 46	0.21	309 ± 13	0.00
Humerus	72.13 ± 2.18	68.48 ± 1.97	0.00	70.55 ± 2.61	0.08	71.14 ± 3.05	0.28	68.89 ± 2.18	0.00
Ulna	80.07 ± 3.98	77.89 ± 2.95	0.14	82.13 ± 1.75	0.09	83.53 ± 3.09	0.01	80.87 ± 2.38	0.69
Radius	66.61 ± 2.9	63.53 ± 1.91	0.00	66.77 ± 1.77	0.86	69.08 ± 2.36	0.01	64.57 ± 2.23	0.03

^aP vs wt, P value for comparison to wild type.

^bResults (bone length × 100/femur length) are reported as average percent ± standard deviation.

^cn, number of bones measured.

^dA variety of abnormalities of the sternum were observed (see Fig. 3). Abbreviations: 2-pc, two-piece; S3-S4, fusion between S3 and S4 sternebrae; X, xiphoid process. Values in parentheses are percentages.

The metacarpal bones of digit I in both Tbx5^+/Δ mice and Cx40-deficient mice were longer than the wild type (Fig. 2E; Table 1), while Tbx5^+/Δ metacarpal and proximal bones were significantly (2- to 2.5-fold) longer than those of wild-type mice (Fig. 2F). This striking increase in metacarpal and proximal phalange length of digit I occurred in 100% of adult Tbx5^+/Δ mice (n = 452) (Fig. 2F) and corresponded to the “finger-like thumb” observed in Holt-Oram patients (37). We suspect that the higher level of expression of Tbx5 in digit I versus digits II to V in wild-type animals (assessed by whole-mount in situ hybridization [data not shown]) contributed to the uniform and marked elongation of digit I in Tbx5^+/Δ mice. The mineralization (assessed by Alizarin Red staining) of digit I in Tbx5^+/Δ mice was also abnormal. At birth, digit I of wild-type, as well as Cx40 mutant, mice had one mineralization center (located in the distal phalange) while digit I in Tbx5^+/Δ mice had three mineralization centers (distal and proximal phalanges and metacarpal bone) (Fig. 2G, arrows).

Analyses of the remaining eight phalanges from digits II to V showed elongation of two (intermediate phalanges of digits III and IV) in Tbx5^Δ/+ mice and three additional in Cx40^+/− or Cx40^−/− mice (proximal phalanges of digits II and IV and intermediate phalange of digit III) (Fig. 2E and Table 1).

Sox9, a transcription factor implicated in skeletogenesis is required for chondrogenesis and expressed within precartilaginous condensations and condensing progenitor cells (47). We examined Sox9 expression in Tbx5^+/Δ and Cx40^−/− embryos and found no significant differences in expression levels compared to the wild type at either E12.5 or E14.5 (data not shown). However, at E13.5, when Sox9 expression is restricted to the tip of the digits in normal mice, levels remained high in both mutant mice in the distal cells of the digital ray (Fig. 2H to J).

Examination of the axial skeleton revealed sternal patterning abnormalities in both mutant mice. The sternum is a segmented structure (Fig. 3A; see reference 25) consisting of the manubrium (M, rostral end), four sternebrae (designated S1, S2, S3, and S4), and the caudal xiphoid process (X). Although the features were absent in wild-type mice, 95% of Tbx5^+/Δ mice, 60% of Cx40^+/− mice, and 67% of Cx40^−/− mice had abnormally shaped thoracic rib cages and shorter sterna (Fig. 3A and B; n = 20, 12, and 16, respectively). Histological examination (hematoxylin-and-eosin-stained, paraffin-embedded sections; not shown) and skeletal staining of newborn sternum show remarkably similar phenotypes in both Tbx5- and Cx40-deficient mice (Fig. 3C and Table 1). Those malformations are the result of mispatterning events in the mutant sterna.

FIG. 3.

Sternal defects in Tbx5^+/Δ, Cx40-deficient, and compound mutant mice. (A) Schematic representation (left) of normal sternal bones (M, manubrium; S1 to -4, sternebra 1 to 4; X, xiphoid). Skeletal preparations of wild-type (Wt) and Tbx5^+/Δ specimens (E15.5) demonstrate that the Tbx5^+/Δ sternum is shortened, consistent with abnormalities in chondrogenesis. (B) Newborn Cx40^+/− mice exhibit comparable sternal bone abnormalities to Tbx5^+/Δ mice. Note reduced sternal length results from fusion of the third and fourth sternebrae (asterisk). (C) The S3-S4 junction in newborn wild-type mice is well delineated and contains two components. Tbx5^+/Δ S3-S4 junctions show incomplete fusion (red arrow) of two hemisternebrae and disappearance of the S3-S4 junction due to ossification. Cx40^+/− and Cx40^−/− mice also have abnormal S3-S4 junctions with complete or asymmetric loss of the S4 ossification center (yellow head arrow). (D) Protuberance of the xiphoid process is evident in the lateral view of the rib cages of Tbx5^+/Δ and Cx40^−/− mice. (E) Additional ossification center detected at the rostral part of the manubrium in newborns may account for the two-piece bone. (F) Abnormal manubrium in adult Cx40^+/−, Tbx5^+/Δ, and Tbx5^+/Δ Cx40^+/− mice. Note the presence of an additional joint resulting in a two-piece manubrium only in mutant mice (shown here for Tbx5^+/Δ).

At birth, wild-type S4 contained one mineralization center, formed from the fusion of two centers located on both sides of the ventral midline. Both mutant mice S4 maintained two immature mineralization centers (arrow in Fig. 3C) or have an incomplete fusion of mineralization centers in S3 and S4 (Fig. 3C, arrows). The aberrant and asymmetric mineralization of sternebrae associated with multiple misalignments of rib pairs contributed to the malformation and shortening of sterna (Fig. 3C contains a variety of examples).

Further abnormalities were observed at both the rostral and caudal ends of Tbx5^+/Δ and Cx40-deficient sterna. Over 50% of Tbx5- and Cx40-deficient animals (Fig. 3D; Table 1) exhibited a protuberant xiphoid process that was often so profound as to be externally visible in live mice (Fig. 3D; Table 1). The rostral structures of the sterna in 15% of Tbx5^+/Δ or 8% of Cx40^+/− mutant mice were composed of two bones developing from two mineralization centers (Fig. 3E) rather than the normal single manubrium found in adult wild-type mice (Fig. 3F).

Distinct skeletal malformations of Tbx5- and Cx40-deficient mice.

The major bones of the forelimb, humerus, ulna, and radius, while abnormal in both Tbx5^+/Δ and Cx40^−/− mice, were shorter in Tbx5^+/Δ mice but longer in Cx40^−/− mice compared to wild-type animals (Table 1). We examined the growth plate of the humerus by in situ hybridization using a probe for type X collagen. The hypertrophic zone, which contains terminally differentiated chondrocytes, was larger in Tbx5^+/Δ than in wild-type (Fig. 4A and B versus C and D) or Cx40-deficient mice (not shown), suggesting that premature maturation of cells accounts for the foreshortened bone. Because PTHrP and Ihh are two major genes involved in the maturation of the growth plate, we examine their expression levels in the mutant mice. No modification of expression was detected between the different genotypes (assessed by in situ hybridization [data not shown]).

FIG. 4.

Skeletal phenotypes specific to Tbx5^+/Δ mutant mice. (A to D) In situ hybridization of collagen X probe in the epiphyseal growth plate of the humerus identified an expanded hypertrophic chondrocyte zone in Tbx5^+/Δ (C and D) compared to wild-type (A and B) mice or Cx40-deficient mice (not shown). ³⁵S-labeled collagen X probe was hybridized to sections, counterstained with toluidine blue, and imaged under bright-field (A and C) and dark-field (B and D) illumination. (E to F) The two-piece scaphoid due to a novel joint (arrow) found in 17% of Tbx5^+/Δ mice occurred independently of fusion. (G) Severe sternebral fusion (S2-S3-S4) in Tbx5^+/Δ. (H to J) Degrees of bifurcation observed in Tbx5^+/Δ (I and J) but not Cx40 mutant mice compared to the wild type (H). (K) Western blot analyses of TGF signaling in wild-type, Tbx5^+/Δ, and Cx40^−/− mice. Expression of TGF-β2 was similar in Tbx5^+/Δ, Cx40, and wild-type mice. While Smad2 and Smad4 protein levels were also comparable, phosphorylated Smad2 was selectively reduced in Tbx5^+/Δ mutant mice. Phosphorylated Smad1 (P-Smad1), -5, and-8 remained unchanged among the genotypes.

Seventeen percent of Tbx5^+/Δ mice had an abnormal scaphoid wrist bone, while none of the Cx40-deficient mice displayed this phenotype. The abnormal scaphoid is composed of two distinct bones rather than one single bone (Fig. 4E versus F). Differences in the shape of the scapula were also observed in both mutant mice compared to the wild type. However, the abnormal scapula morphologies were dissimilar in Tbx5^+/Δ and Cx40-deficient mice (Table 1). As noted above, the sterna from Tbx5- and Cx40-deficient mice present similar mispatterning of the sternebrae S3-S4, even though multiple fusions of sternebrae (S2-S3-S4 and S3-S4-X) were occasionally observed in the Tbx5^+/Δ mice (Fig. 4G; Table 1). Furthermore, the cartilaginous and ossified components of the Tbx5^+/Δ xiphoid process were split medially, while the wild-type and Cx40-deficient xiphoid processes were completely fused (Fig. 4H to J).

Several studies (for reviews, see references 17, 22, and 49) indicate that BMPs regulate joint development in the bones that exhibited abnormal fusions in Tbx5^+/Δ and Cx40-deficient mice (24, 44). We therefore examined expression of BMP family members Gdf5, BMP5-6-7, and BMPrIB (Alk6) by radioactive and whole-mount in situ hybridization in Tbx5^+/Δ and Cx40-deficient mice. Age-matched mutant and wild-type mice had comparable amounts of these RNA species (data not shown).

Protein levels of various components of the TGF signaling pathway in Tbx5^+/Δ and Cx40-deficient E13.5 mice were assessed by Western blot analyses (Fig. 4K). Total protein levels of TGF-β₂ and Smad1, -2, -3, -4, -5, or -8 levels were not altered in forelimb extracts from Tbx5^+/Δ mice (Fig. 4K and not shown). However, the level of phosphorylated Smad2 was significantly less in extracts derived from Tbx5^+/Δ mice compared to wild-type or Cx40-deficient mice (Fig. 4K). In contrast, no significant changes in the levels of phosphorylated Smad1, -5, and -8 levels were observed.

Rib malformations were evident only in Cx40-deficient mice, because Cx40, but not Tbx5, is expressed in these bones (Fig. 1A and C). Bifurcation of the first rib pair (Fig. 5B) and ectopic growth of a cervical pair of ribs (designated 1′, which resulted in partial or complete duplication of this set of ribs) (Fig. 5C and D) were detected in approximately 18% of Cx40^+/− mice and 4% Cx40^−/− mice. This ectopic growth/bifurcation of a cervical segment suggests that Cx40 participates in the normal segmentation of the rib cage.

FIG. 5.

Skeletal phenotypes specific to Cx40 mutant mice. (A to D) Abnormal ribs were found in Cx40 mutant mice but not in wild-type mice (A). Variation of phenotype in Cx40^+/− mice ranged from bifurcation of the first rib cartilage (B) to C5 elongation (C, arrow) to formation of an additional 1′ rib (D). (E and F) Lower-limb malformations in Cx40 mutant mice. Fusion of the lateral cuneiform (lc) and navicular (na) anklebones in Cx40^−/− (F) and Cx40^+/− (not shown) mice compared to the wild type (Wt) (E). There is also delayed ossification (red) in multiple anklebones of Cx40 mutant mice.

The hind limbs (which do not express Tbx5) of Cx40-deficient mice revealed anklebone abnormalities (Fig. 5E versus F). Fusion of the lateral cuneiform (lc) and navicular (na) bones was found in 25% of heterozygous Cx40^+/− mice and 100% of Cx40^−/− mice. Delay in navicular bone mineralization was evident in all Cx40-deficient mice (Fig. 5E versus F).

Phenotypes of compound Tbx5^+/Δ Cx40^+/− mutant mice.

To better understand the interaction between Tbx5 and Cx40 in the morphogenesis of the skeleton, Tbx5^+/Δ and Cx40^−/− mice were interbred to generate compound mutants. Tbx5^+/Δ Cx40^+/− were then intercrossed, and the genotypes of 201 of their offspring identified 25 wild-type, 77 Cx40^+/−, 51 Cx40^−/−, 16 Tbx5^+/Δ, and 31 Tbx5^+/Δ Cx40^+/− mice. Interestingly, only one Tbx5^Δ/+ Cx40^−/− mouse was found, and no Tbx5-null mice were live born, consistent with our previous report (13). In this mixed genetic background, a substantial fraction of Tbx5^+/Δ (22%) and compound mutants, Tbx5^+/Δ Cx40^+/− (69%) and Tbx5^+/Δ Cx40^−/− (98%), died within the first week of life.

At birth, all Tbx5^+/Δ Cx40^+/− mice had normal weight and body habitus. Skeletal studies showed the phenotypes described for either Tbx5^+/Δ or Cx40-deficient mice or demonstrated a slightly exacerbated phenotype. For example, the metacarpal phalanges of Tbx5^+/Δ Cx40^+/− were longer compared to those seen in the parental mutants (most notably in digit I). Furthermore, the compound mutant mice also had slight elongation of digit II and V proximal phalanges (Fig. 2E; Table 1). Similarly, Tbx5^+/Δ Cx40^+/− mice had a higher frequency (54%) of two-piece manubrium than either Tbx5^+/Δ (15%) or Cx40^+/− (8%) mice (Table 1). Only in the bones of the forelimb (foreshortened in Tbx5^+/Δ mice and elongated in Cx40^+/− mice) did the compound heterozygous mice express a phenotype that was an intermediate of the two parental phenotypes (Table 1).

DISCUSSION

Haploinsufficiency of transcription factor TBX5 universally causes upper limb and axial skeletal defects in Holt-Oram syndrome and in Tbx5^+/Δ mice. We demonstrate that some of these skeletal bone malformations result from altered expression of a single protein, Cx40, a gap junction subunit. Tbx5^+/Δ, Cx40-deficient, and compound Tbx5^+/Δ Cx40-deficient mice share common defects in wrist, sternal, and forelimb digit bones. In addition, Tbx5 and Cx40 independently regulate forelimb bone length, while Cx40, but not Tbx5, participates in hind limb development. These data provide the first evidence that Cx40 is critical for endochondral bone development and defines Tbx5-Cx40 signaling as an important regulatory pathway for development of the axial and appendicular skeleton. Cx40 is the first connexin family member implicated in skeleton patterning.

Tbx5 functions early and late in limb development.

During embryogenesis, Tbx5 is expressed early in limb field development. Both its temporal-spatial expression and data accrued from targeted mutagenesis studies indicate Tbx5 is essential for the early induction and maintenance of Fgf10, which initiates the forelimb bud (1, 21, 31, 38). In addition to these early roles, our data indicate Tbx5 functions at later stages in limb development and segmentation, after mesenchymal cells have aggregated and condensed into primordial bones. The location of Tbx5 expression in the perichondrium of developing bone, the expansion of the hypertrophic zone in Tbx5-deficient bones, and subsequent malformations found in Tbx5 mutant mice all support our conclusion that this molecule participates in chondrocyte maturation and bone length. Further, the particularly prominent expression in the periarticular perichondrium of forelimb and sternal bones defines a unique role for Tbx5 in joint formation.

Few transcription factors, including Dlx5 and Runx2 (or Cbfa1), expressed in the perichondrium of developing bones have been identified (19, 48). These transcription factors are expressed in the perichondrium adjacent to the diaphysis of the cartilage models and in the maturing prehypertrophic and hypertrophic chondrocytes. The complementary pattern of Dlx5 and Runx2 expression with Tbx5 suggests the hypothesis that these transcription factors work in a synergistic manner to regulate growth and maturation along the length and at the ends of endochondral bone.

We assume that Tbx5 haploinsufficiency leads to multiple changes in gene expression. While Cx40 probably mediates some Tbx5^+/Δ phenotypes, other phenotypes, such as xiphoid and forelimb bone maturation, are independent of Cx40 (Fig. 6). Further, Tbx5^+/Δ mice have lower levels of phospho-Smad2 (Fig. 4K) than wild-type or Cx40-deficient mice, suggesting that the BMP/TGF-β signaling pathway mediates some Tbx5^+/Δ phenotypes.

FIG. 6.

Model of Tbx5 and Cx40 regulation. Cx40 and other genes, regulated by Tbx5, produce normal patterning of specific bones. In many instances, Tbx5 activates Cx40 expression and hence determines the proper patterning and shape of sternal and carpal bones and forelimb digit length. Different transcription factors (potentially Tbx4) regulate Cx40 expression in hind limbs and other bones. Tbx5 haploinsufficiency directly or indirectly (hatched arrow) misregulates downstream genes, causing abnormal skeleton formation. Tbx5 haploinsufficiency also modulates other gene products independent of Cx40, such as receptor-activated Smad2 (phosphorylated Smad2 [P-Smad2]) levels.

Tbx5, via gap junction protein Cx40, specifies articulation of autopod and sternum.

Our study discovered an unrecognized and critical role for Cx40 in developing endochondral bone that in large part overlaps Tbx5 functions in skeletal morphogenesis. Analyses of Tbx5^+/Δ and Cx40-deficient forelimbs (Fig. 2) each showed trapezium bone fusion (95% and 100% of mutant mice, respectively), as well as elongated metacarpal bones and phalanges. Unusual sternal malformations (Fig. 3), including a two-piece manubrium and a protuberant xiphoid process, were also common to Tbx5^+/Δ and Cx40-deficient mice. We demonstrate that normal levels of Tbx5 and Cx40 are required for the formation of the same bones, suggesting that they act in the same pathway. Of course, other possibilities exist. For example, they might act in parallel independent pathways that are both required for the development of these specific bones. Furthermore, we have not demonstrated that Tbx5 directly regulates Cx40 expression in bone. Nevertheless, given the data indicating direct activation of the Cx40 promoter by Tbx5 and markedly reduced Cx40 expression in Tbx5^+/Δ mice (13), as well as colocalized expression of Cx40 and Tbx5 in the perichondrium of developing sternum and forelimb bones, we prefer the model that Tbx5 directly regulates Cx40 in the wrists, sternum, and digits, as it does in the heart. A parallel relationship exists for another T-box and connexin gene pair: Tbx2 regulation of Cx43 expression (10, 11).

Because multiple connexins (Cx43, Cx45, and now Cx40) and T-box gene family members are implicated in skeletal morphogenesis, we propose that the Tbx and connexin proteins comprise a general molecular network to specify bone shape and structure. Several lines of experimental data independently support this model. First, different connexin molecules are required for distinct types of bone formation; Cx43 deficiency disrupts both endochondral (producing truncation, nicking, and splitting of the limb skeleton in chicks; 32) and intramembranous bone formation (delayed ossification, osteoblast dysfunction, and craniofacial abnormalities; 28). In contrast, our data indicate Cx40 functions primarily in endochondral bone formation. Second, expression of a particular connexin in some bones is required for morphology and presumably physiologic function. That is, while there may be functional redundancy between some connexin family members (34, 36), there are also exquisite bone-specific requirements for particular connexins, including Cx43 in cranial-facial bones (36) and, as evidenced here, Cx40 in the wrist and sternal bones. Patterns of connexin expression change in osteoblasts subjected to sheer stress (14), a finding that further indicates bone structure and function is in part provided through these molecules. Third, important skeletal malformations are produced by all known human T-box gene mutations (reviewed in reference 35); TBX5 mutations cause upper limb and sternum malformations (6, 29); TBX3 mutations cause ulnar bone and lateral digit abnormalities (5, 46); mutations in TBX4 are linked to an autosomal dominant disorder called small-patella syndrome (9); TBX22 mutations cause cleft palate with ankyloglossia (12). Mice engineered to carry these defects recapitulate human skeletal malformations (13, 15, 30). Collectively, these studies support a model in which bone identity is in part specified by Tbx regulation of connexin molecules. Subtle variation (twofold) in Tbx expression causes significant changes in connexin levels, defines a mechanism for fine control of gap junction composition, therein tailoring chemical signals for bone growth and maturation. We speculate that regulation of connexin gene expression by T-box transcription factors has coevolved. That is, the primordial T-box gene apparently regulated one or more connexin genes and as the connexin gene family expanded, the regulation by T-box transcription factors has been conserved.

The precise role of gap junctions in bone development remains uncertain. Gap junctions, comprised of different connexins, appear to facilitate selective transfer of particular ions and small molecules between cells, a process that is of potential importance in the development and maintenance of distinct bone structure and physiology. Because Ca²⁺ has been implicated as a critical signaling molecule in bone morphogenesis and because Cx40 plays a critical role in the cardiac conduction system, where cellular Ca²⁺ also plays a critical role in the signaling process, we are tempted to speculate that connexins mediate their activity in bone by altering Ca²⁺ homeostasis.

In autopod and axial skeletal structures, we suggest that Tbx5 level controls Cx40 expression (Fig. 6, green) in a direct and/or indirect manner. Mutation in either Tbx5 or Cx40 genes reduces Cx40 expression and results in anomalous joint formation and erroneous bone length. In the immature forelimb digits, reduced levels of either Tbx5 or Cx40 prolonged expression of Sox9 (Fig. 2H to J), a transcription factor linked to type II collagen and aggregan expression during chondrocyte differentiation (for reviews, see references 4 and 40). Altered expression of Sox9, either directly or indirectly, may contribute to skeletal malformations in Tbx5-deficient mice. Selective inactivation of Sox9 null mice, produced by Col2a-1Cre transgene-mediated recombination, produces partial fusion of mesenchymal condensations that contribute to abnormal joint structure (26). Changes in the temporal expression of Sox9 might similarly contribute to joint malformations in Tbx5^+/Δ and Cx40-deficient mice, as well as modify the ossification in Tbx5^+/Δ forelimb digits and Cx40^+/− anklebones. Conditional Sox9 null newborn mice resulting from inactivation by the Prx1-Cre transgene had very short limbs but appeared otherwise normal (2). However, they primarily died in the immediate postnatal period from respiratory distress due to the absence of a sternum (2), as do Tbx5-Prx1-Cre mice (38).

Tbx5-independent Cx40 regulation.

We also identified Cx40-dependent but Tbx5-independent abnormalities (Fig. 6, blue). Notably, bones of the ankle were altered in Cx40-deficient mice (particularly lateral cuneiform and navicular); molecules other than Tbx5, which is not expressed in the hind limbs, must regulate the segmentation of the ankle (Fig. 5). A prediction of the T-box regulation of connexin model is that another T-box molecule directs Cx40 expression in the ankles. Given the pattern of expression of Tbx4 (39), this is a good candidate to assume hind limb regulation of Cx40. In a few bones, including the scaphoid, xiphoid, and humerus, the phenotypes of Cx40-deficient mice and Tbx5^+/Δ mice were not the same. Presumably, in these bones Tbx5 regulates expression of other genes that modulate bone growth independently of Cx40.

Taken together, our data define a specific role for a Tbx5-Cx40 regulatory cascade in skeletal morphogenesis. These studies also suggest a more general mechanism by which transcriptional regulation is linked to structural proteins that are critical in the development of bone structure and function. Further, haploinsufficiency of other transcription factors, such as Nkx2.5, Gata4, and Tbx1, that cause a variety of congenital malformations like Tbx5, may alter the expression of only a few target proteins. Identification of these few target proteins may eventually allow novel therapeutics involving replacement of target proteins.