The TGFβ type I receptor TGFβRI functions as an inhibitor of BMP signaling in cartilage.

The type I TGFβ receptor TGFβRI (encoded by Tgfbr1) was ablated in cartilage. The resulting Tgfbr1 Col2 mice exhibited lethal chondrodysplasia. Similar defects were not seen in mice lacking the type II TGFβ receptor or SMADs 2 and 3, the intracellular mediators of canonical TGFβ signaling. However, we detected elevated BMP activity in Tgfbr1 Col2 mice. As previous studies showed that TGFβRI can physically interact with ACVRL1, a type I BMP receptor, we generated cartilage-specific Acvrl1 (Acvrl1 Col2 ) and Acvrl1/Tgfbr1 (Acvrl1/Tgfbr1 Col2 ) knockouts. Loss of ACVRL1 alone had no effect, but Acvrl1/Tgfbr1 Col2 mice exhibited a striking reversal of the chondrodysplasia seen in Tgfbr1 Col2 mice. Loss of TGFβRI led to a redistribution of the type II receptor ACTRIIB into ACVRL1/ACTRIIB complexes, which have high affinity for BMP9. Although BMP9 is not produced in cartilage, we detected BMP9 in the growth plate, most likely derived from the circulation. These findings demonstrate that the major function of TGFβRI in cartilage is not to transduce TGFβ signaling, but rather to antagonize BMP signaling mediated by ACVRL1.

Growth plate cartilage supports bone elongation and provides an essential platform for the generation of articular joints (1). Cartilage is generated by chondrocytes, which can be classified into different subtypes depending on their location and differentiation status (1). These subtypes form distinct zones in the growth plate. The pool of resting chondrocytes at the distal ends of developing long bones supplies cells for building growth plate cartilage. Resting chondrocytes differentiate into columnar chondrocytes, which have a high proliferative rate and stack together into columns. Columnar cells further differentiate into prehypertrophic chondrocytes, which cease proliferation and enlarge. These cells further differentiate into hypertrophic chondrocytes, which have a larger cell size and eventually undergo apoptosis or transdifferentiate into osteoblasts (2). Eventually, the growth plate is replaced by calcified bone.

TGFβ and bone morphogenetic protein (BMP) signaling components play important roles in cartilage and joint formation and maintenance (3–7). TGFβs 1 to 3 bind to complexes containing the type I receptor TGFβRI and the type II receptor TGFβRII. Ligand binding triggers phosphorylation of TGFβRI, enabling activation of the intracellular mediators SMADs 2 and 3 in the canonical TGFβ pathway. TGFβs also signal through activation of kinases such as MAP kinase, TAK1, JNK, and RhoA. BMPs are a diverse group of ligands that share homology to TGFβs 1 to 3, but bind to distinct receptor complexes consisting of type I (ACVRL1 [ALK1], ACVR1 [ALK2], BMPR1A [ALK6], and BMPR1B [ALK3]) and type II receptors (ACTRIIA, ACTRIIB, and BMPRII) (8). These complexes activate the intracellular SMADs 1/5/8 in the canonical pathway, as well as noncanonical pathways overlapping with those activated by TGFβ. In addition to shared noncanonical pathways, there is cross talk between TGFβ and BMP pathways at multiple levels (3, 9, 10).

Although the potent in vitro chondrogenic activity of TGFβs is well documented (11), the function of TGFβ pathways in cartilage in vivo is unclear. TGFβRII is the type II receptor for TGFβs 1 to 3. Tgfbr2 mice exhibit axial skeletal defects but have no alterations in appendicular elements (12). Similarly, loss of Smad3, Smad2, or both in chondrocytes leads to no or only subtle growth plate phenotypes at birth (13, 14). These studies may not reflect all TGFβ pathway activity in committed chondrocytes. Targeting SMADs 2 and 3 leaves noncanonical pathways intact. Studies based on loss of TGFβRII may also underestimate TGFβ pathway activity; in cranial neural crest cells, TGFβs activate TGFβ signaling in the absence of TGFβRII (15). Furthermore, ligands other than TGFβs 1 to 3 (e.g., GDF8, GDF10, and GDF11) can activate signaling through TGFβRI in complexes containing ACTRIIA or ACTRIIB (16–18).

These findings raise the possibility that TGFβRI acts independently of TGFβRII in cartilage. A study of Tgfbr1 mice showed that TGFβRI plays a critical role in formation of the perichondrium adjacent to the growth plate (19). Whether TGFβRI has a role in growth plate chondrocytes has not been investigated. TGFβRI is also of interest because it enables TGFβs to activate BMP signaling through a mechanism involving association of the BMP receptor ACVRL1 with TGFβRI/TGFβRII complexes (20, 21). TGFβ signaling through ACVRL1/TGFβRI/TGFβRII complexes augments BMP signaling in vascular cells (21). Whether such complexes exist in other tissues, and their impact on BMP signaling, is unclear.

Using mice lacking TGFβRI in cartilage (Tgfbr1;Col2-Cre [Tgfbr1]), we find that TGFβRI is required to maintain the pool of resting chondrocytes in the growth plate. Unexpectedly, the primary function of TGFβRI is not to activate TGFβ pathways. Rather, Tgfbr1 mice exhibit elevated BMP activity mediated by ACVRL1. Mechanistically, loss of TGFβRI triggers the formation of ACVRL1/ACTRIIB complexes in the growth plate. These have high affinity for the circulating BMP ligand BMP9, which we detected in growth plate cartilage. Thus, these studies indicate that the major function of the type I TGFβ receptor in growth plate cartilage is to prevent the promiscuous activation of ACVRL1-mediated BMP signaling by the circulating ligand BMP9.

Results

TGFβRI Is Expressed in Growth Plate Cartilage during Embryonic Development.

We examined when and where TGFβRI is expressed to understand its potential roles in the growth plate. Tgfbr1 mice were used as negative controls for antibody specificity and confirmed that TGFβRI is efficiently deleted (). At embryonic day 13.5 (E13.5), TGFβRI is expressed in intervertebral discs, and at lower levels in chondrocytes within vertebral bodies (). TGFβRI is expressed primarily in the perichondrium of appendicular elements at E13.5 but is present at low levels in chondrocytes (). TGFβRI expression in appendicular growth plates is increased by E16.5 () and persists at least through E18.5 (). Expression is strong at E16.5 and E18.5 in the columnar zone (CZ) but is absent from the hypertrophic zone (HZ) (). At E16.5, resting chondrocytes in the developing posterior tibial condyle exhibit higher TGFβRI levels than do those in the anterior condyle ( [yellow and white boxes] and ). At E18.5, TGFβRI protein levels are high in resting zone (RZ) chondrocytes throughout the growth plate, and in the anterior and posterior condyles ().

Severe Chondrodysplasia in Tgfbr1 Mutant Neonates.

Analysis of Tgfbr1 mice demonstrated a major role for TGFβRI in formation of the perichondrium (19). However, because Dermo1-cre is expressed in many mesodermal-derived tissues, whether TGFβRI had a direct function in chondrocytes could not be determined. We therefore generated Tgfbr1;Col2-Cre (Tgfbr1) mice to delete TGFβRI specifically in chondrocytes. The resulting Tgfbr1 mice exhibited perinatal lethality. The precise cause of death is unknown, but mutants exhibit major skeletal defects. At E13.5 and E14.5, there are no patterning defects in Tgfbr1 mice, but mutants have smaller vertebrae and occipital bones; no obvious defects in appendicular elements are seen at these stages (Fig. 1 ). At E16.5, defects in Tgfbr1 axial elements persist; mutants develop abnormal spinal curvature and kinks in the ribs at the chondro-osseous junction along with shortened sternebrae (Fig. 1). There are no defects in appendicular patterning at E16.5, but long bones are shorter by this stage (Fig. 1 ).

Fig. 1.

Skeletal malformation in Tgfbr1 mice. (A–D) Cleared skeletal preparations of E12.5, E13.5, E16.5, and E18.5 control and Tgfbr1 mutant embryos. Lateral views of littermates demonstrating that spine malformations can be detected in Tgfbr1 mutants by E13.5 and become more severe at E18.5 (n = 5). Blue shows Alcian blue staining of cartilage. Red shows Alizarin red staining of bone. Control = Tgfbr1;Tgfbr1 = Tgfbr1;Col2a1Cre; black arrows in A–C denote the reduced sizes of cervical vertebral elements in mutants; black arrowheads in A–D denote the smaller, flatter lumbar vertebral bodies in Tgfbr1 mice. Green arrow in C denotes the locations of kinks in rib elements. Green arrowhead in C denotes the shortened sternebrae in Tgfbr1 mutants. (E–H) Lateral views of E16.5 (E and F) and E18.5 (G and H) forelimbs (E and G) and hindlimbs (F and H), demonstrating the short humerus, femur, and tibia in Tgfbr1 mutants compared with control littermates (n = 5). Black arrowhead points out the malformation of scapulae; black arrow denotes the dislocation of elbow joints; green arrow points out the small tibial condyles in E18.5 Tgfbr1 mutant mice. (I) Magnified views of the defects in E18.5 scapulae and elbow joints in Tgfbr1 mutant mice shown in G. Black arrowhead points out the malformation of scapulae and black arrow denotes the dislocation of elbow joints in Tgfbr1 mutant mice. (J) Magnified views of the defects in E18.5 knee joints in Tgfbr1 mutant mice shown in H. Black and blue double-headed arrows measure the length of control and Tgfbr1 mutant femoral cartilage, respectively. Green and white bars measure the widths of the neck of femoral cartilage of control and Tgfbr1 mutants, respectively. Red and yellow bars measure the widths of the neck of tibial cartilage of control and Tgfbr1 mutants, respectively. The white arrow points out the flattened posterior tibial condyle in mutants; the green arrow shows the smaller anterior tibial condyle in mutants. (K and L) Quantification of the total lengths of tibiae in E16.5 and E18.5 control and Tgfbr1 mutant embryos. n = 5 mice. Columns indicate average length of cartilage elements. Error bar shows SD. *P < 0.05; Student’s 2-tailed t test. (M and N) Quantification of the length of tibial and femoral cartilages at knee joints in E18.5 control and Tgfbr1 mutant embryos. n = 5 mice. Columns indicate average length of cartilage elements. Error bar shows SD. *P < 0.05; Student’s 2-tailed t test. (O and P) Quantification of the widths of tibial and femoral cartilages at knee joints in E18.5 control and Tgfbr1 mutant embryos. n = 5 mice, Columns indicate average length of cartilage elements. Error bar shows SD. *P < 0.05; Student’s 2-tailed t test.

At E18.5, the severity of skeletal defects in the Tgfbr1 mice continues to progress. Tgfbr1 mice have shorter statures (E16.5: Mut/Ctrl, 87.2% ± 2%; E18.5: Mut/Ctrl, 73.9% ± 3%) and shorter tibiae than controls (Fig. 1 ). E18.5 Tgfbr1 mice have bent scapulae, kinks at the distal ends of humeri, and dislocation of elbow joints (Fig. 1 ); these defects are not present at E16.5 (Fig. 1). The condyles and entheses of tibiae are also smaller in mutants at E18.5 (Fig. 1 ). The proximal tibial cartilage is no different in length (Fig. 1), but is wider in Tgfbr1 mice than in control littermates at E18.5 (Fig. 1), and both the length and width of distal femoral cartilage are larger in Tgfbr1 mice (Fig. 1 ). As these defects are not seen in Tgfbr2 mutants (12) or Smad2/Smad3 mice (13), TGFβRI must transduce its effects in chondrocytes in combination with a different type II receptor, and through a pathway other than the canonical TGFβ pathway.

Chondrocyte Proliferation and Differentiation Defects in Tgfbr1 Mutant Growth Plate Cartilage.

We performed histological analyses to investigate mechanisms underlying the Tgfbr1 phenotype. The RZ is a sparsely dividing reservoir of cells that gives rise to the actively proliferating cells of the CZ (1). Tgfbr1 mice have shorter RZs and longer CZs (Fig. 2 ) at E16.5 and E18.5 (Fig. 2 ), although the lengths of HZ and the total growth plate do not differ in Tgfbr1 and control littermate mice (Fig. 2 ). Immunostaining for proliferating cell nuclear antigen (PCNA) revealed that levels of proliferation are higher in the CZ and RZ in Tgfbr1 mice (Fig. 2 ). These results suggest that loss of Tgfbr1 leads to premature conversion of slowly dividing RZ cells into rapidly proliferating CZ cells and depletion of the RZ (Fig. 2 ). Terminal deoxynucleotidyltransferase deoxyuridine triphosphate nick-end labeling (TUNEL) assay was performed to test whether increased apoptosis also contributes to the shorter RZ and smaller tibial condyles in Tgfbr1 mice. No TUNEL-positive cells are seen in Tgfbr1 RZ and in condyles; strong TUNEL staining is present only in the terminal HZ and adjacent perichondrium (Fig. 2). No differences are observed in Tgfbr1 and control littermate mice in these areas (Fig. 2).

Fig. 2.

Increased proliferation in Tgfbr1 mutant growth plates. (A and B) Sections of E16.5 and E18.5 proximal tibias were stained with safranin O, showing reduced RZ and expanded CZ. Heights of different growth plate zones are indicated by double arrows, and the zone boundaries are marked by lines. Blue dashed line is the demarcation between the CZ and pre-HZ. All images are aligned to this boundary. Top yellow lines demarcate the approximate boundary between the RZ and CZ. Bottom yellow lines demarcate the boundary between the HZ and zone of ossification (chondro-osseous junction). Yellow double-headed arrows indicate the extent of the CZ. Green double-headed arrows demarcate the extent of the RZ. Blue double-headed arrows demarcate the extent of the HZ. Three litters containing mice of all of the genotypes shown in the figures were examined. At least 5 mice per genotype were examined. The images shown are from littermates. Black arrow highlights absence of the PCL enthesis. (C–J) Quantification of the lengths of different zones and whole growth plates. Data are expressed as means of length + SD (n = 5 mice). *P < 0.05; Student’s 2-tailed t test compared with control Tgfbr1. (K) Immunostaining for PCNA (red) in E16.5 proximal tibia. RZ (green brackets) and CZ (yellow brackets). All mice are littermates. (L) Quantitation of the percentage of PCNA-positive cells in distinct zones. The percentage of PCNA-positive cells in the RZ and CZ was quantified. Data are expressed as means of percent PCNA-positive + SD (n = 4 mice). *P < 0.05 compared with control Tgfbr1 mice. (M) TUNEL (green) staining in E16.5 proximal tibia. All mice are littermates. White arrowheads point to perichondrium. Red dotted lines indicate chondro-osseous junction. (N) Quantitation of the percentage of TUNEL-positive cells at chondro-osseous junctions. Data are expressed as means of percent TUNEL-positive + SD (n = 4 mice). There are no significant differences between Tgfbr1 and control Tgfbr1 mice.

Increased BMP Signaling in Growth Plate Cartilage.

TGFβRI can trigger multiple downstream pathways, including SMAD2/3-mediated canonical signaling, TAK1-mediated noncanonical signaling, and SMAD1/5/8-mediated BMP signaling (8, 9, 20, 21). Immunostaining showed that there is a significant decrease in the percentage of p-SMAD2/3–positive cells in Tgfbr1 growth plates (Fig. 3 ). However, Smad2 and Smad3 mutants do not have any of the defects seen in Tgfbr1 growth plates (13), suggesting that the decrease in p-SMAD2/3 is not a major contributor to the Tgfbr1 phenotype. Similarly, TAK1 is required in growth plate cartilage to activate proliferation (22). In palatal mesenchyme, TGFβ2 activates TAK1 independently of TGFβRII through TGFβRI (15). However, immunostaining revealed no difference in pTAK1 expression in Tgfbr1 mice (Fig. 3 ).

Fig. 3.

Increased BMP signaling activity in Tgfbr1 mutant growth plates. All images are E16.5 proximal tibias. Quantification data are expressed as means + SD (n = 3 mice). *P < 0.05; Student’s 2-tailed t test. (A and B) Immunostaining for phosphorylated SMAD2 and SMAD3 showing decreased protein levels in Tgfbr1 mutant growth plates. (B) Quantification of pSMAD2/3 protein level in posterior condyles (yellow boxes), anterior condyles (white boxes), and CZs (red boxes) of growth plate. (C and D) Immunostaining for phosphorylated TAK1 (pTAK1) showing no change of expression levels in Tgfbr1 mutant growth plates. (D) Quantitation of pTAK1 protein level in the posterior condyles (yellow boxes), anterior condyles (white boxes), and CZs (red boxes) of growth plate. (E and F) Immunostaining for phospho-SMAD1/5/8 protein showing increase levels of activated BMP signaling in Tgfbr1 growth plates. (F) Quantitation of pSMAD1/5/8 protein level in the posterior condyles (yellow boxes), anterior condyles (white boxes), and CZs (red boxes) of growth plate.

Multiple studies indicate that BMP signaling triggers proliferation of RZ and CZ chondrocytes (23–25). Because the growth plate defects in Tgfbr1 mice are consistent with elevated BMP signaling and cannot be explained by decreased activation of TGFβ pathways, we considered the possibility that BMP signaling is altered. Immunostaining for pSMAD1/5/8 revealed an increased percentage of pSMAD1/5/8-positive cells in the RZ and CZ of Tgfbr1 mice (Fig. 3 ). The increased pSMAD1/5/8 levels do not arise through increased mRNA expression of key BMP, TGFβ, or GDF ligands, BMP type I or type II receptors, the BMP inhibitor NOGGIN, or the BMP signaling mediators SMAD1, SMAD5, and SMAD8 ().

Loss of Acvrl1 Rescues Growth Plate Defects in Tgfbr1 Mutant Mice.

The above findings indicated that TGFβRI acts to inhibit BMP signaling in growth plate cartilage. In considering potential mechanisms, we focused on the type I BMP receptor ACVRL1 (ALK1). ACVRL1 has been studied almost exclusively in the context of the vasculature, because heterozygous loss-of-function mutations in Acvrl1 lead to hereditary hemorrhagic telangiectasia in humans (26). However, Acvrl1 is expressed in articular cartilage (27, 28). ACVRL1 can interact physically with TGFβRI and TGFβRII to enable TGFβ ligands to activate BMP signaling (21). According to this model, loss of TGFβRI would lead to decreased BMP signaling, opposite to the elevated BMP signaling observed in Tgfbr1 growth plates. We therefore speculated that an alternative function for TGFβRI could be to inhibit BMP signaling through ACVRL1. This might occur if TGFβRI prevents formation of highly active (ACVRL1/ACTRIIA or ACVR1/ACTRIIB) complexes by sequestering ACVRL1 into less active complexes containing TGFβRI.

ACVRL1 is expressed in all zones of the growth plate (Fig. 4). To test the role of ACVRL1 in cartilage and whether there is a genetic interaction between Tgfbr1 and Acvrl1, we generated Acvrl1;Col2-Cre (Acvrl1) and Acvrl1/Tgfbr1 double-mutant Acvrl1;Acvrl1;Col2-Cre (Acvrl1/Tgfbr1) mice. Cartilage-specific deletion of ACVRL1 was confirmed by immunohistochemistry (IHC) (Fig. 4). Acvrl1 mice are apparently normal at E18.5 (Fig. 4 ). Histological analysis showed that Acvrl1 mice have normal growth plates (Fig. 4 ), and no changes in pSMAD1/5/8 levels were detected (Fig. 4 ). Strikingly, however, simultaneous loss of Acvrl1 and Tgfbr1 rescues most of the skeletal defects seen in Tgfbr1 mice (Fig. 4 ). The length of the CZ (Fig. 4 ) and pSMAD1/5/8 levels (Fig. 4 ) are restored to normal in Acvrl1/Tgfbr1 mice. However, the rescue of Tgfbr1 defects is not complete. Vertebral bodies remain smaller than control littermates in Acvrl1/Tgfbr1 mice (Fig. 4 ; arrowheads), and the enthesis defects are not fully rescued (Fig. 4 ; arrows). These defects could reflect a role for TGFβRI in transducing TGFβ pathways, and/or interactions between TGFβRI and type I BMP receptors other than ACVRL1 in these regions. Nonetheless, these results demonstrate that Acvrl1 is not necessary for growth plate formation, but interacts genetically with Tgfbr1, and that TGFβRI antagonizes ACVRL1 function in the growth plate.

Fig. 4.

Rescue of the Tgfbr1 mutant phenotype in Acvrl1/Tgfbr1 double mutants. All images are E18.5. (A and B) Immunofluorescent staining showing ACVRL1 protein presence in proximal tibial growth plates of control mice (A) and lack of staining in Acvrl1 mutant mice (B) verifies antibody specificity and cartilage-specific knockout. Red, α-ACVRL1 staining; blue, DAPI. (C–F) Skeletal preps showing no alterations in Acvrl1 mice, and rescue of the chondrodysplasia in Acvrl1/Tgfbr1 mice compared with Tgfbr1 mice. Arrowheads point to the vertebral body (VB); note that there are no defects in Acvrl1 VB, but both Acvrl1/Tgfbr1 and Tgfbr1 mice have smaller VB compared with control mice. n = 5/genotype. (G–J) Sections through proximal tibia growth plates showing restoration of the highs of the CZ and the sizes of posterior condyle in Acvrl1/Tgfbr1 mice. Blue brackets indicate the high of CZ. Black arrow points to the rescued tibia condyle in Acvrl1/Tgfbr1 mice. Blue arrow points to the defective of posterior tibial condyle in Acvrl1/Tgfbr1 mice. n = 3/genotype. (K–N) pSMAD1/5/8 levels appear to be restored to normal in Acvrl1/Tgfbr1 mice. n = 3/genotype.

TGFβRI Associates with ACVRL1 and ACTRIIB to Inhibit ACVRL1/ACTRIIB Complex Formation.

The genetic studies demonstrated that a key function for TGFβRI is to constrain ACVRL1 signaling in the growth plate. ACVRL1 normally complexes with ACTRIIA/B to transduce BMP9/10-activated SMAD1/5/8 BMP signaling (29). TGFβRI complexes with ACTRIIA/B transduce GDF8/11-activated SMAD2/3 TGFβ signals (16, 18). Therefore, competition between ACVRL1 and TGFβRI for ACTRIIA and/or ACTRIIB provides a potential explanation for the elevated BMP signaling in Tgfbr1 mice and its rescue in Acvrl1/Tgfbr1 double mutants. We performed coimmunoprecipitation (co-IP) in ATDC5 chondrocytes under serum-free conditions to identify endogenous TGFβRI interactions with type I and type II BMP receptors. Among the type I BMP receptors, TGFβRI associates primarily with ACVRL1 (Fig. 5 ). A low level of association of TGFβRI with ACVR1, the type I receptor most closely related structurally to ACVRL1 (8), was detected; almost no association with BMPR1A or BMPR1B was found. Proximity ligation assay (PLA) confirmed that the observed interaction between TGFβRI and ACVRL1 occurs in vivo (). Thus, the genetic interaction between ACVRL1 and TGFβRI is correlated with a physical interaction between these receptors in chondrocytes. We next considered interactions with type II receptors. We focused on TGFβRII because TGFβRI/TGFβRII complexes transduce canonical signaling by TGFβs 1 to 3, and on ACTRIIA and ACTRIIB because these type II receptors can complex with either ACVRL1 or TGFβRI (3, 16, 18, 29). Among the tested type II receptors, TGFβRI associates exclusively with ACTRIIB under serum-free conditions (Fig. 5 ). The lack of association between TGFβRI and TGFβRII is consistent with the lack of a growth plate phenotype in Tgfbr2 mice (12).

Fig. 5.

TGFβRI associates with ACVRL1 and ACTRIIB to block ACVRL1/ACTRIIB complex formation. (A) TGFβRI associates most with ACVRL1. Lysates prepared from ATDC5 cells were subjected to IP with an anti-TGFβRI antibody cross-linked to magnetic beads. Immunoprecipitated proteins were then probed with anti-ACVRL1, anti-ACVR1, anti-BMPR1A, and anti-BMPR1B antibodies. (B) Quantification of protein levels shown in A. Three biological replicates were analyzed. The average input protein level is normalized to 1. Immunoprecipitated protein level (IP-TGFβRI) was normalized using input and plotted relative to input + SD (n = 3). Significance was established using 2-way ANOVA and t test. *P < 0.013. (C) Of the tested type II BMP receptors, TGFβRI associates almost exclusively with ACTRIIB. Immunoprecipitated proteins were probed with anti-TGFβRII, anti-ACTRIIA, and anti-ACTRIIB antibodies. Note that ACTRIIB, but not TGFβRII or ACTRIIA, could be coimmunoprecipitated with TGFβRI. (D) Quantification of protein levels shown in C. Three biological replicates were analyzed as described in B. *P < 0.017. (E) Western blot images confirm that TGFβRI protein is deleted by Crispr-cas9 in Tgfbr1 mutant (Tgfbr1) ATDC5 cells, and that these cells exhibit elevated basal pSMAD1/5/8 levels. (F and G) The Western blot images of TGFβRI (F) and pSMAD1/5/8 levels (G) were quantified using ImageJ. Values of protein level were normalized using GAPDH and are plotted relative to control + SD (n = 3 biological replicates). Significance was established using Student's 2-tailed t test. *P < 0.05. (H) Lysates from control or Tgfbr1 cells were subjected to IP with an anti-ACVRL1 antibody. Immunoprecipitated proteins were then probed with anti-TGFβRII, anti-ACTRIIA, and anti-ACTRIIB antibodies. In control cells, ACVRL1 associates with ACTRIIA. In Tgfbr1cells, ACVRL1 associates with both ACTRIIA and ACTRIIB. (I) Quantification of protein levels shown in E. Three biological replicates were analyzed. Immunoprecipitated protein level was normalized using input and plotted relative to input + SD (n = 3). Significance was established using Student's 2-tailed t test. *P < 0.01. Note that there is a significant increase of ACTRIIB/ACVRL1 association in Tgfbr1 cells compared with control cells. (J) Lysates from control or Tgfbr1 cells were subjected to IP with an anti-TGFβRII, anti-ACTRIIA, or anti-ACTRIIB antibody. Immunoprecipitated proteins were then probed with an anti-ACVRL1 antibody. In control cells, ACVRL1 associates with ACTRIIA. In Tgfbr1 cells, ACVRL1 associates with both ACTRIIA and ACTRIIB. (K) Quantification of protein levels shown in G. Three biological replicates were analyzed as described in I. *P < 0.05. Note that there is a significant increase in ACTRIIB/ACVRL1 association in Tgfbr1 cells compared with control cells.

The association of TGFβRI with ACTRIIB in chondrocytes is consistent with potential competition between ACVRL1 and TGFβRI for this type II receptor. We generated Tgfbr1 mutant ATDC5 cells (Tgfbr1) using Crispr-Cas9 systems to test this possibility. The removal of TGFβRI was confirmed by Western blot (Fig. 5 ). Elevated pSMAD1/5/8 levels were seen in Tgfbr1 cells even under serum-free conditions (Fig. 5 ), confirming that loss of TGFβRI leads to elevated BMP signaling in chondrocytes. Co-IP assays were performed to determine how loss of TGFβRI impacts association of ACVRL1 with type II receptors. In control cells, ACVRL1 associates primarily with ACTRIIA (Fig. 5 ). In Tgfbr1 cells, ACVRL1 associates with both ACTRIIA and ACTRIIB (Fig. 5). Thus, loss of TGFβRI significantly increases the interaction of ACVRL1 with ACTRIIB (Fig. 5). The increased formation of ACVRL1/ACTRIIB complexes is not caused by increased ACTRIIB protein levels because there is a similar level of input of ACTRIIB between control and Tgfbr1 cells (Fig. 5). There is no change in ACVRL1/TGFβRII or ACVRL1/ACTRIIA complexes in Tgfbr1 cells compared with the control (Fig. 5). Furthermore, there is a significant increase in ACVRL1/ACTRIIB interaction in Tgfbr1 cells compared with control cells after normalization to ACVRL1 input levels (Fig. 5 ). No changes in levels of ACVRL1/TGFβRII or ACVRL1/ACTRIIA complex formation were seen in Tgfbr1 cells (Fig. 5 ), suggesting that the increased formation of ACVRL1/ACTRIIB complexes is responsible for the elevated pSMAD1/5/8 levels in Tgfbr1 cells.

Loss of TGFβRI Increases Basal BMP Activity and BMP9 Responsiveness in Chondrocytes.

The above findings demonstrated increased formation of ACVRL1/ACTRIIB complexes in Tgfbr1 cells. To test whether the ACVRL1/ACTRIIB complex has a specific function in mediating elevated BMP signaling in Tgfbr1 cells, cells were treated with BMP9, which acts exclusively through ACVRL1 or ACVR1 complexed to ACTRIIA or ACTRIIB, but shows significantly higher affinity for ACTRIIB than for ACTRIIA (29). TGFβ1 and BMP2 were used as controls because TGFβ1 can trigger BMP signaling through ACVRL1/TGFβRI/TGFβRII complexes (27). BMP2 activates BMPR1A and BMPR1B, and to a lesser extent ACVR1, but not ACVRL1 (8, 30–32). 12×SBE-Luc reporter assays showed that Tgfbr1 cells have higher basal BMP activity (2×) than control cells (1×), and increased responsiveness to BMP9 (5.8× vs. 3.3×) (significance confirmed by ANOVA and t test) (Fig. 6). In contrast, responsiveness to BMP2 was slightly lower in Tgfbr1 (2.6×) compared with control cells (3.2×) (no significance by ANOVA and t test) (Fig. 7). Consistent with the previously demonstrated requirement for TGFβRI in TGFβ-mediated BMP pathway activation (21), TGFβ1 stimulates BMP signaling in control, but not in Tgfbr1 cells. Taken together, these results indicate that loss of TGFβRI increases sensitivity to BMP9. Since the affinity of BMP9 for ACVRL1/ACTRIIB complexes is several orders of magnitude higher than for ACVRL1/ACTRIIA complexes (29), our results suggest that the increased formation of ACVRL1/ACTRIIB receptor complexes (Fig. 5) may be responsible for the increased responsiveness to BMP9 in Tgfbr1 cells.

Fig. 6.

Loss of TGFβRI leads to elevated basal BMP activity and BMP9 responsiveness in chondrocytes. Cells were transfected with 12×SBE-Luc reporter (77) and Renilla luciferase control reporter. Relative 12×SBE-Luc activity shows the normalized level of 12×SBE-Luc/Renilla Luc. Assays were repeated 2 times. n = 3/technical repeats in each assay. Error bars show SE. Significance was established using 2-way ANOVA and t test. *P < 0.05/n (n = number of groups). (A) Elevated basal BMP activity and BMP9 responsiveness in chondrocytes. The basal BMP activity in Tgfbr1 cells is more than 2-fold higher than in control cells (columns 1 and 5). TGFβ1 treatment (2 ng/mL) increases BMP activity in control cells but does not affect Tgfbr1 cells (columns 2 and 6). BMP2 (100 ng/mL) and BMP9 (100 ng/mL) have similar activities in control cells at the tested concentrations (columns 3 and 4). BMP9 induces higher BMP activity than BMP2 in Tgfbr1 cells (columns 7 and 8). (B) NOGGIN does not block elevated basal BMP activity in Tgfbr1 cells. Elevated BMP activity in Tgfbr1 cells compared with control cells (columns 1 and 5). BMP2 activity in control and Tgfbr1 cells (columns 2 and 6). NOGGIN inhibits BMP2 activity in both control and Tgfbr1 cells (columns 3 and 4, and columns 7 and 8). NOGGIN does not block elevated basal BMP activity in Tgfbr1 cells (columns 5 and 8). (C) BMP9 antibody does not block elevated basal BMP activity in Tgfbr1 cells. BMP reporter activity is activated by BMP9 in control ATDC5 cells (columns 1 and 2). α-BMP9 antibody blocks exogenous BMP9 action in control and Tgfbr1 cells (columns 3 and 7). Tgfbr1 cells exhibit elevated basal BMP activity (columns 1 and 5). Tgfbr1 cells respond to BMP9 (columns 5 and 6). α-BMP9 antibody has no effect on elevated basal BMP activity in Tgfbr1 cells (columns 5 and 8). (D) ACVRL1-Fc does not block elevated basal BMP activity in Tgfbr1 cells. BMP reporter activity is activated by BMP9 in control ATDC5 cells (columns 1 and 2). ACVRL1-Fc blocks exogenous BMP9 action in control and Tgfbr1 cells (columns 3 and 7). Tgfbr1 cells exhibit elevated basal BMP activity (columns 1 and 5). Tgfbr1 cells respond to BMP9 (columns 5 and 6). ACVRL1-Fc has no effect on elevated basal BMP activity in Tgfbr1 cells (columns 5 and 8). (E) ACTRIIB-Fc does not block elevated basal BMP activity in Tgfbr1 cells. BMP reporter activity is activated by BMP9 in control ATDC5 cells (columns 1 and 2). ACTRIIB-Fc blocks exogenous BMP9 action in control and Tgfbr1 cells (columns 3 and 7). Tgfbr1 cells exhibit elevated basal BMP activity (columns 1 and 5). Tgfbr1 cells respond to BMP9 (columns 5 and 6). ACVRL1-Fc has no effect on elevated basal BMP activity in Tgfbr1 cells (columns 5 and 8). (F) ACVRL1 antibody blocks elevated basal BMP activity in Tgfbr1 cells. BMP reporter activity is activated by BMP9 in control ATDC5 cells (columns 1 and 2). Anti-ACVRL1 antibody does not block BMP9 action in control cells (columns 2 and 4). Tgfbr1 cells exhibit elevated basal BMP activity (columns 1 and 5). Anti-ACVRL1 antibody blocks elevated basal BMP activity in Tgfbr1 cells (columns 5 and 7). Anti-ACVRL1 antibody restores BMP9 activity to control level in Tgfbr1 cells (columns 2, 4, and 8). (G) Loss of ACVRL1 kinase activity restores BMP signaling to normal level in the Tgfbr1 cells. Elevated BMP activity in Tgfbr1 cells compared with control cells (columns 1 and 3). Mutation of threonine-196 to valine near the canonical start of the ACVRL1 kinase domain (Acvrl1) using Crispr-Cas9 system in the Tgfbr1 cells (column 5); BMP activity in Tgfbr1/Acvrl1 double-mutant cells is restored to normal (columns 1 and 5). BMP9 activity in control and Tgfbr1 cells (columns 2 and 4). BMP9 activity in Tgfbr1/Acvrl1 double-mutant cells is lower than Tgfbr1 cells and restored to control levels (columns 2, 4, and 6). (H) Anti-ACTRIIB antibody blocks elevated basal BMP activity in Tgfbr1 cells. BMP reporter activity is activated by BMP9 in control ATDC5 cells (columns 1 and 2). Anti-ACTRIIB antibody deceases BMP9 action in control cells (columns 2 and 4). Tgfbr1 cells exhibit elevated basal BMP activity (columns 1 and 5). Anti-ACTRIIB antibody blocks elevated basal BMP activity in Tgfbr1 cells (columns 5 and 7). Anti-ACTRIIB antibody decreases BMP9 activity by 2-fold in Tgfbr1 cells (columns 6 and 8) but does not completely restore BMP9 activity to control levels (columns 4 and 8).

Fig. 7.

BMP9 protein is present in growth plate cartilage. Immunofluorescence analysis of WT, Tgfbr1, and Bmp9 mice to detect the presence of BMP9 in cartilage. WT and Tgfbr1 sections are from E18.5 embryos, and Bmp9 sections are from P0 newborn pups. α-BMP9 staining is red. Nuclei are stained with DAPI (blue). n = 3 mice. (A and B) In WT and Tgfbr1 mice, BMP9 protein is localized in bones as expected. However, BMP9 also presents in the cartilage in these 2 mouse lines. Arrow points to a blood vessel. C, cartilage; B, bone; M, muscle. (C) The specificity of anti-BMP9 antibody was validated by using Bmp9 mice. No positive signal was detected in Bmp9 mice. Arrow points to a blood vessel. C, cartilage; B, bone; M, muscle.

As discussed, basal BMP activity is elevated in Tgfbr1 ATDC5 cells even under serum-free conditions. To test whether the elevated basal activity is due to production of an endogenous ligand, the BMP inhibitor NOGGIN (blocks BMPs 2, 4, 5, and 7, and GDFs 5 to 7) (33) was added to the cells followed by reporter assay. The elevated basal BMP activity in Tgfbr1 cells is not affected by NOGGIN; NOGGIN activity was confirmed by its ability to block the effects of recombinant BMP2 activity (Fig. 6). NOGGIN does not inhibit BMPs 9 or 10, the ligands for ACVRL1 (34). To test whether the elevated basal BMP activity in Tgfbr1 cells is due to production of BMP9 or BMP10 and signaling through ACVRL1/ACTRIIB complexes, anti-BMP9 antibody, ACVRL1-Fc, or ACTRIIB-Fc was added to the cells. Anti-BMP9 antibody blocks BMP9 function (35, 36). ACVRL1-Fc and ACTRIIB-Fc bind to BMP9 and BMP10 (29, 37). While anti-BMP9 antibody, ACVRL1-Fc and ACTRIIB-Fc inhibit the effects of exogenous BMP9, they do not affect the elevated basal activity (Fig. 6 ). These results demonstrate that the increased BMP output is not due to increased expression of ligands inhibited by NOGGIN (BMPs 2, 4, 5, and 7, or GDFs 5 to 7), or BMPs 9 and 10.

Anti-ACVRL1 antibody was used to test whether the increased basal and BMP9-induced BMP signaling is mediated directly by ACVRL1 in Tgfbr1 cells. In control cells, anti-ACVRL1 has no impact (Fig. 6 , columns 3 and 4). However, in Tgfbr1 cells, anti-ACVRL1 blocks both elevated basal BMP activity and responsiveness to BMP9 (Fig. 6 , columns 7 and 8). The basis for the lack of effect of anti-ACVRL1 in control cells is unclear but may be a consequence of sequestration of ACVRL1 by TGFβRI (Fig. 5) into inactive signaling complexes. According to this scenario, the observed responsiveness to BMP9 in control cells may be mediated by ACVR1 rather than ACVRL1. Nonetheless, these results show that ACVRL1 is responsible for increased BMP activity in Tgfbr1 cells. To confirm this, the kinase domain in ACVRL1 was mutated by replacing threonine-196 with valine using Crispr-Cas9 in Tgfbr1 cells. Loss of ACVRL1 kinase activity rescues both basal BMP activity and responsiveness to BMP9 in Acvrl1K/Tgfbr1 cells (Fig. 6), confirming that activation of ACVRL1 causes increased BMP signaling in Tgfbr1 cells.

Finally, we tested the potential involvement of ACTRIIB in the increased basal and BMP9-induced BMP signaling in Tgfbr1 cells. In control cells, anti-ACTRIIB has no impact on basal BMP signaling but deceases BMP9-induced activity (Fig. 6 , columns 3 and 4). In Tgfbr1 cells, anti-ACTRIIB blocks elevated basal BMP activity (Fig. 6 , columns 1 and 7). In addition, anti-ACTRIIB blocks most (69%) of BMP9-induced BMP activity (Fig. 6 , columns 6 and 8). The remaining responsiveness to BMP9 is higher than the control level (Fig. 6 , columns 4 and 8). It is unclear whether this is due to only partial blockade of ACTRIIB by the antibody, or to involvement of other type II BMP receptors. Nonetheless, the results indicate that ACTRIIB plays the major role in elevated basal activity and BMP9 responsiveness in Tgfbr1 cells, supporting the hypothesis that loss of TGFβRI increases BMP signaling through increased formation of ACVRL1/ACTRIIB complexes.

BMP9 and ACVRL1/ACTRIIB Complexes Are Present in Cartilage of Tgfbr1 Mice.

The finding that ACVRL1 mediates elevated BMP signaling in Tgfbr1-deficient chondrocytes was unexpected because neither ACVRL1 nor its ligands BMP9 and BMP10 have been implicated in growth plate formation. BMP9 and BMP10 are not expressed in normal cartilage (38, 39). In accordance, we did not detect Bmp9 and Bmp10 RNA in control or Tgfbr1 cartilage by RT-PCR and qPCR (). However, BMP9 and BMP10 circulate in blood (2 to 6 and 0.2 to 2 ng/mL, respectively) in mice (40, 41) at high enough levels to activate ACVRL1. IHC was performed to test whether BMP9 from the circulation or elsewhere can penetrate growth plate cartilage. BMP9 is detected in control and Tgfbr1 tissues, with high levels in blood vessels, bone, and muscle, and low levels in cartilage (Fig. 7 ); the specificity of the BMP9 antibody is confirmed by the absence of staining in Bmp9 tissue (37) (Fig. 7). These results suggest that BMP9 in cartilage may defuse from nearby vascularized tissues and can potentially activate ACVRL1 in cartilage.

Our in vitro data showed an increase in ACVRL1/ACTRIIB complex formation in Tgfbr1 cells, and that these complexes cause elevated BMP activity in Tgfbr1 cells. We therefore tested whether we could detect increased ACVRL1/ACTRIIB complex formation in cartilage of Tgfbr1 mice using PLA. These complexes are rarely detected in control cartilage but are present in Tgfbr1 cartilage ().

Discussion

The studies described here demonstrate that TGFβRI is an essential regulator of growth plate chondrocyte proliferation and differentiation. However, TGFβRI, the requisite type I receptor for signal transduction by TGFβs 1 to 3, does not exert its primary function in growth plate chondrocytes via activation of TGFβ pathways. Several lines of evidence support this conclusion. First, the strong Tgfbr1 phenotype contrasts with the near absence of growth plate phenotypes in embryos and neonates of Tgfbr2 or Smad2;Smad3 mice (12, 13). Second, the Tgfbr1 phenotype is consistent with elevated BMP activity: Tgfbr1 mice have a shorter RZ and a longer CZ, associated with an increased proportion of proliferating cells in both regions. This is a hallmark of BMP signaling in chondrocytes (6, 42, 43). Third, pSMAD1/5/8 levels are elevated in Tgfbr1 growth plates, and SMADs 1/5/8 play essential roles in activating chondrocyte proliferation in vivo (42). Fourth, pSMAD1/5/8 levels and the majority of the growth plate defects are restored to normal in Acvrl1/Tgfbr1 mice.

In addition to the appendicular growth plate phenotype documented here, Tgfbr1 mice exhibit axial defects. The axial phenotype appears to be similar to that in Tgfbr1 mice (19). Tgfbr1 is highly expressed in intervertebral discs and perichondrium (19). Col2Cre is expressed in the sclerotome before the specification of separate lineages for vertebral cartilage and intervertebral discs (44). Since both Col2Cre and Dermo1Cre would target these tissues, the axial defects in Tgfbr1 mice most likely arise through a role for TGFβRI in perichondrium and intervertebral discs. Additional studies employing other genetic models would be needed to define cartilage-intrinsic roles for TGFβRI in vertebral elements. Furthermore, TGFβRI is more widely expressed in posterior condylar tibial cartilage than in anterior condylar tibial cartilage (). Consistently, the posterior condyle is smaller than the anterior condyle in Tgfbr1 mice. These findings establish Tgfbr1 as a gene that regulates the differential development of condyles. Evidence for a role for TGFβ signaling in the formation of tendon and ligament insertions has been presented previously. Tgfbr2 mice lack tibial medial condyle and deltoid tuberosity formation (45). However, Tgfbr2 mice do not show such phenotypes (12), suggesting that the defects in Tgfbr2 mice are not caused by cells derived from committed chondrocytes. The Tgfbr1 defects are distinct from those in Tgfbr2 mice in that Tgfbr1 mice have both anterior and posterior condyle defects instead of absence of the medial condyle, and Tgfbr1 mice have normal tuberosities. In addition, none of the defects seen in Tgfbr1 mice is found in Tgfbr2 mice (12). This suggests that Tgfbr1 regulates posterior and anterior condyle formation in cartilage, but probably not through canonical TGFβ signaling. Scx+;Sox9+ cells in the TGFβRI-expressing regions of the condyles give rise to ligaments and the ligamentous junction (46). Future studies would thus be of interest to elucidate the role of TGFβRI in the differential development of condyles, ligaments, and associated structures, and to understand the basis for the differing condylar phenotypes in Tgfbr1 and Tgfbr2 mice.

Our studies unveiled that the excess BMP signaling in Tgfbr1 mice is mediated by ACVRL1. A role for TGFβRI as an inhibitor of ACVRL1 function has not been reported previously. Little is known about ACVRL1 function in cartilage, but TGFβRI and ACVRL1 have opposing functions in chondrocytes in vitro (28). The absence of a discernable phenotype in Acvrl1 mice indicates that this receptor is dispensable for cartilage development. The loss of TGFβR1 leads to a deleterious gain of function of ACVRL1 signaling in cartilage. This raises the question of the normal function of Acvrl1 in cartilage. Although we have not detected an obvious phenotype in Acvrl1 mice at birth, we cannot rule out subtle defects. Examination of adult mice may reveal such changes. Furthermore, it is possible that loss of ACVRL1 is compensated for by the closely related type I BMP receptor ACVR1 (ALK2). Like ACVRL1, ACVR1 is activated by BMPs 9 and 10 (8). Acvr1 mice are viable and exhibit only mild appendicular defects (47). Removing both Acvrl1 and Acvr1 in cartilage might provide evidence for compensatory roles in cartilage.

ACVRL1 has been studied most extensively in the vasculature. Acvrl1 mice die at midgestation due to major defects in angiogenesis, and Acvrl1 mice and humans exhibit hereditary hemorrhagic telangiectasia, a condition characterized by vascular malformations (26). TGFβRI and ACVRL1 interact physically in endothelial cells, but the functional output is different from the one shown here in cartilage; in endothelium, ACVRL1 antagonizes TGFβRI signaling, and TGFβRI/ACVRL1/TGFβRII complexes activate BMP signaling (21, 28, 48), whereas in cartilage, TGFβRI antagonizes ACVRL1. It is likely that TGFβRI and ACVRL1 have context-specific mutually antagonistic functions. Other studies show that ACVRL1 can act independently of TGFβRI in endothelial cells (49, 50). Studies in multiple cell types would be of interest to understand the basis for the cell type-specific responses to the TGFβRI:ACVRL1 balance.

The functions of TGFβRI and ACVRL1 in adult articular cartilage are unclear. Although Acvrl1 mice exhibit no developmental defects, cartilage-specific loss of ACVRL1 might impact the maintenance of articular cartilage in adults. ACVRL1 is expressed in articular chondrocytes (27, 28), and the ACVRL1/TGFβRI ratio increases in aged and osteoarthritic articular cartilage due to decreased TGFβRI expression (28). It will therefore be important to test whether Acvrl1 mice are resistant to age-induced osteoarthritis. If so, this could lead to therapeutic options for treatment of osteoarthritis based on inhibition of ACVRL1 activity. A recent study showed that postnatal deletion of TGFβRI in cartilage leads to a strong osteoarthritis phenotype, suggesting direct effects in articular cartilage (51). However, TGFβRI was deleted in young mice, and therefore both growth plate and articular cartilage were targeted. Nonetheless, clear alterations in gene expression were noted in articular cartilage. Expression of Prg4, a gene expressed in superficial zone articular chondrocytes and essential for maintenance of boundary lubrication (52–54), is greatly reduced upon postnatal loss of TGFβRI. It will be of great interest to elucidate the mechanism by which TGFβRI mediates Prg4 expression (51). Prg4 can be induced by TGFβ (51, 55, 56), but this may be independent of the canonical pathway involving TGFβRII based on the observation that postnatal loss of TGFβRII in cartilage is protective against osteoarthritis (57). Taken together with studies implicating ACVRL1 in the progression of osteoarthritis (58–60), it is likely that TGFβRI will exert some functions in articular cartilage through direct activation of TGFβ pathways and others via inhibition of BMP signaling through antagonism of ACVRL1. Experiments to test this are underway.

None of the defects in Tgfbr1 mice is seen in Tgfbr2 mutants (12). In accordance, we found that TGFβRII interacts only minimally with TGFβRI and ACVRL1 in chondrocytes; in normal cartilage, TGFβRI and ACVRL1 interact primarily with ACTRIIB and ACTRIIA, respectively. TGFβRI prevents ACVRL1/ACTRIIB complex formation, as demonstrated by the de novo formation of these complexes in Tgfbr1 chondrocytes and Tgfbr1 growth plates. Our data indicate that ACVRL1/ACTRIIB complexes are responsible for the elevated BMP signaling in Tgfbr1 growth plates (). We speculate that elevated BMP activity in vivo arises at least in part because ACVRL1/ACTRIIB complexes bind BMP9 with higher affinity (300×) than ACVRL1/ACTRIIA complexes (29). Cartilage-specific deletion and overexpression of ACTRIIB in normal and Tgfbr1 mice would provide a genetic test of this model.

BMPs 9 and 10 are the only ligands that activate ACVRL1 at physiological levels (30) and are potent inducers of chondrogenesis in vitro (61). Tgfbr1 cells show greatly enhanced sensitivity to BMP9. Neither BMP9 nor BMP10 has been implicated previously in skeletal development. BMP9 is produced in the liver and is present in the circulation at high enough levels to activate BMP signaling (36, 58, 62). BMP10 is predominantly expressed in the heart and is in the circulation, but at lower levels than BMP9 (40, 63). Although neither BMP9 nor BMP10 is produced in cartilage () (38, 39), BMP9 protein is present in growth plate chondrocytes. We speculate that the BMP9 in the growth plate is derived from the circulation, as high levels were detected in blood vessels and tissues close to the growth plate. Although the mechanisms are unknown, even large molecules can diffuse through dense cartilage extracellular matrix (64).

We consistently observed elevated basal BMP signaling in Tgfbr1 cells. The elevated basal signaling is due to activity of ACVRL1 because it was blocked by a mutation that inactivated ACVRL1 kinase activity. However, we were unable to find evidence that the basal activity is ligand dependent. Experiments were performed under serum-free conditions; NOGGIN, anti-BMP9 antibody, ACVRL1-Fc, or ACTRIIB-Fc did not block the increased basal signaling. NOGGIN binds to BMPs 2 and 4 to 7 and GDFs 5 to 7 at the concentrations used here, but does not block BMP9 and BMP10 (33, 65, 66). ACVRL1-Fc blocks BMP9 and BMP10. ACTRIIB-Fc blocks multiple ligands including activins, GDF8, and GDF11 (67, 68). We cannot rule out an effect due to increased levels of an untested ligand, nor can we rule out an effect mediated by altered levels of expression of BMP receptors. qRT-PCR experiments revealed that levels of Bmpr1a are reduced in Tgfbr1 chondrocytes (). Although this would typically be expected to reduce BMP signaling, Antebi et al. (69) showed that altering relative levels of BMPs receptors can greatly alter responsiveness to specific ligand pairs, enabling some pairs to switch from an antagonistic to an additive mode. An analysis of responsiveness of Tgfbr1 chondrocytes to multiple pairs to BMP ligands will be needed to assess how loss of TGFβRI impacts signaling by multiple ligands through each BMP receptor complex. It is conceivable that the elevated basal BMP activity in Tgfbr1 cells involves a ligand-independent effect of ACVRL1. Forced expression of ACVRL1 increases basal BMP activity (30, 70, 71). Fluid flow can trigger ligand-independent activation of BMP signaling in vascular tissues (72), and/or can greatly increase sensitivity of ACVRL1 to BMP9 (73). As cartilage experiences mechanical strain in vivo, it will be of interest to test how mechanical forces impact BMP signaling through ACVRL1 in cartilage.

In summary, contradictory to the paradigm, the major role of TGFβRI in growth plate chondrocytes is not to transduce TGFβ signaling, but rather to block ACVRL1-mediated BMP signaling. It is possible that TGFβRI exerts its effects by blocking ACVRL1-mediated BMP signaling in tissues other than growth plate cartilage.

Materials and Methods

Mouse Lines and Breeding.

All procedures involving animals were approved by the Institutional Animal Care and Use Committee of University of California, Los Angeles. Mice were housed in an Association for Assessment and Accreditation of Laboratory Animal Care-accredited facility in accordance with the Guide for the Care and Use of Laboratory Animals (74). This study was compliant with all relevant ethical regulations regarding animal research. E12.5–E18.5 embryos from C57BL6 backgrounds were used. Both sexes were used as sex was not determined in embryos. To genotype Tgfbr1-flox mice, 3 primers were used: 5′-ATGAGTTATTAGAAGTTGTTT-3′, 5′-ACCCTCTCACTCTTCCTGAGT-3′, and 5′-GGAACTGGGAAAGGAGATAAC-3′, with 55 °C annealing temperature, yielding 350-bp wild-type (WT) and 150-bp flox-inserted mutant bands (75). To genotype Acvrl1-flox mice, 2 primers were used: 5′-CCTGGACAGCGACTGTACTAC-3′ and 5′-GCCCCATTGCTCTCCTCAAAC-3′, with 68 °C annealing temperature, yielding 450-bp WT and 350-bp flox-inserted mutant bands (50). To genotype Col2a1-Cre mice, 2 primers were used: 5′-TGCTCTGTCCGTTTGCCG-3′ and 5′-ACTGTGTCCAGACCAGGC-3′, with 58 °C annealing temperature, yielding a 800-bp band (44).

IHC and TUNEL Assay.

Limb tissues were dissected and fixed in 4% paraformaldehyde in PBS. They were then decalcified with Immunocal (Decal Chemical Corporation) for 3 d at 4 °C, embedded in paraffin, and cut at a thickness of 7 μm. Paraffin sections were deparaffinized and rehydrated by passage through xylene and 100%, 95%, and 70% ethanol. Samples were treated with 1 mg/mL hyaluronidase for 30 min at 37 °C. Sections were blocked with 5% goat or donkey serum, in TBS for 1 h at room temperature and incubated with 1:200 diluted primary antibody at 4 °C overnight. Slides were washed 3 times in TBST for 10 min each and incubated at room temperature for an hour in secondary antibody (1:500) conjugated with fluorescent dyes (Cell Signaling; goat anti-rabbit red, R37117; goat anti-rabbit green, R37116; goat anti-mouse red, A11032) in 1% goat or donkey serum TBST. Slides were washed 3 times in PBST for 10 min each and stained in DAPI (Sigma; D9453) at 0.1 μg/mL for 10 min. Lastly, the slide was mounted in 10 μL of Fluoro-Gel (Electron Microscopy Sciences; 1798510). Primary antibodies included the following: α-TGFβRI (Abcam; 31013), α-ACVRL1 (Santa Cruz; 19546), α-pSMAD1/5/8 (Cell Signaling; 9511S), α-pSMAD2/3 (Cell Signaling; 3108S), α-PCNA (Abcam; 2586S), α-TAK1(Cell Signaling; 4508S), and α-BMP9 (R&D; AF3209). Apoptotic cells were detected by in situ TUNEL assay using the In Situ Cell Death Detection Kit (Sigma; 11684795910) following the manufacturer's instructions. All experiments were repeated on sections from at least 3 embryos of each genotype. All comparisons were between littermates.

Skeletal Preparation and Histology.

Whole-mount skeletal preparations (n = 5) were performed as described (33). For histological analyses, paraffin sections were produced from E13.5, E16.5, E18.5, and postnatal day 0 (P0) mice. Safranin-O staining was performed as described (76). Heights of RZ, CZ, and HZ were measured directly from images (n = 4) taken from each of 4 mice per genotype, and significance was evaluated using Student’s 2-tailed t test.

RNA Isolation and qPCR.

Limb joint cartilages isolated from E17.5 mice were frozen in liquid nitrogen and smashed to powder. Cartilage powder was then fixed in TRIzol. Total RNA was isolated by the phenol-chloroform method and converted to cDNA. The cDNA was amplified and quantified using SYBR Green reagent (Sigma) in a Stratagene Mx3005P qPCR System (Thermo Scientific). Primer sequences are shown in .

Co-IP.

Immunoprecipitations (IPs) were performed using Pierce Crosslink Magnetic IP/Co-IP kit (Thermo Scientific; 88805). To obtain the lysate solution, confluent ATDC5 cells in 150-mm dishes were fixed in 1% formaldehyde at room temperature for 10 min followed by quenching with 125 mM glycine at room temperature for 5 min. Fixed cells were washed with cold PBS twice. Cells were collected by scraper and centrifuged at 750 × g for 4 min and then resuspended in PBS supplemented with protease inhibitors mixture (Thermo Scientific). Cross-linking was performed on the suspended cells using disuccinimidyl suberate (Thermo Scientific) with a final concentration of 2.5 mM. The reaction was allowed to proceed for 30 min at room temperature before quenching with Tris with a final concentration of 20 mM. After centrifugation at 750 × g for 4 min, cell pellet was resuspended in 500 μL of IP Lysis Buffer from the kit supplemented with protease inhibitors. Lysate was sonicated in an ice-water bath for cell disruption. After IP was performed with antibodies listed in . The pull-down products were analyzed by Western blot.

Western Blot Analysis.

Protein samples were heated in Laemmli buffer at 95 °C for 5 min. Thirty to 50 μg of protein was subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes. After blocking, the membranes were probed with appropriate primary antibodies and then incubated with specific horseradish peroxidase-conjugated secondary antibodies (Bio-Rad). Immunodetection was performed using enhanced chemiluminescence consistent with the manufacturer’s protocol (Thermo Fisher Scientific). The band intensities were quantified and normalized to the corresponding controls. The antibodies used are listed in .

DNA Transfection and Reporter Assay.

ATDC5 cells were plated in 48-well plates at a density of 2.0 × 104 cells per well in Dulbecco’s modified Eagle’s medium (DMEM) (Corning) containing 10% fetal bovine serum (FBS) and 1% streptomycin/penicillin. After 1 h, the cells were transfected in quadruplicates with PolyJet reagent. In total, 0.05 μg of Renilla plasmids (Addgene) and 0.1 μg of 12× SBE-Luc (77) plasmids with 0.75 μL of PolyJet reagent were added per well. The cells were grown overnight at 37 °C with CO2. Twenty-four hours posttransfection, medium was replaced with serum-free DMEM containing 200 μg/mL ascorbic acid and 1% streptomycin/penicillin. After 3 h of starvation, cells were treated with various ligands, Fc-proteins, and antibodies: TGFβ1, 5 ng/mL (R&D 240-B); BMP2, 200 ng/mL (R&D; 355-BM); BMP9, 100 ng/mL (Biolegend; 553102); NOGGIN, 400 ng/mL (R&D; 1967-NG); anti-BMP9, 400 ng/mL (R&D; MAB3209); ACVRL1-Fc, 400 ng/mL (R&D; 370-AL); ACTRIIB-Fc, 400 ng/mL (R&D; 3725-RB); anti-ACVRL1, 400 ng/mL (Santa Cruz; 19546); and anti-ACTRIIB, 400 ng/mL (Santa Cruz; 25453). Twenty-four hours posttreatment, growth media was removed. Then, the cells were washed with PBS and lysed in 0.1 mL of Passive Lysis Buffer (PLB). Dual-luciferase reporter assay was performed using the Promega kit (E1910) in a FLUOstar Omega system (BMG Labtech) following the manufacturer’s instructions. Data are presented as ratios of Luc/Renilla activity.

CRISPR-Cas9 Gene Editing.

Oligos that target Tgfbr1 were synthesized and linked to lentiCRISPRv2 (1-vector system) plasmid as described (78). Oligo sequences for Tgfbr1 were as follows: 5′-CACCGCCTCTTCATTTGGCACACGG-3′ and 5′-AAACCCGTGTGCCAAATGAAGAGGC-3′. ATDC5 cells were seeded in 6-well plates at a density of 1 × 106 cells per well in DMEM containing 10% FBS and 1% penicillin/streptomycin. After 24 h, the cells were transfected with 1 µg of lentiCRISPRv2 plasmid and 3 μL of Lipofectamine (SignaGen) per well. After 2 d of transfection, Tgfbr1 mutant cells (Tgfbr1) were selected with puromycin at 4 μg/mL. Mutation of Tgfbr1 was confirmed by PCR. The PCR used were 5′-GCTTCGTCTGCATTGCACTT-3′ and 5′-AGGTGGTGCCCTCTGAAATG-3′. WT allele has a 126-bp PCR product, and Tgfbr1 deletion-mutant allele has no PCR product. Loss of TGFβRI protein was confirmed by Western blot using anti-TGFβRI antibody (Abcam; ab31013). For generating the Acvrl1 kinase mutation, ATDC5 cells were transfected with RNA oligos (CRISPR RNA [crRNA] and transactivating crRNA [tracrRNA]) and single-stranded DNA oligonucleotides (ssODNs). crRNA was designed as 5′-UGGUGCAGAGGACGGUAGCUGUUUUAGAGCUAUGCU-3′ and ordered from Integrated DNA Technologies (IDT); tracrRNA was ordered from IDT; ssODN was designed as 5′-GCAGCGGCTCGGGGCTCCCCTTCTTGGTGCAGAGGGTGGTAGCTCGGCAGGTTGCGCTGGTAGAGTGTGTGG-3′. RNA oligos (crRNA and tracrRNA) were mixed in equimolar concentrations to a final duplex concentration of 50 μM. The mix was heated at 95 °C for 5 min and allowed to cool to room temperature (15 to 25 °C). For each sample, crRNA:tracrRNA (50 μM working solution) was diluted in their respective buffers to a final volume of 6 μL each. Tgfbr1 cells were transfected with 6 μL of crRNA:tracrRNA (50 μM), 3 μL of ssODN (100 μM), 1 μL of pLentiCas9-Blast plasmid (0.1 μg/μL; Genscript), and 4 μL of Lipofectamine (SignaGen) per well (24-well plate). After 2 d of transfection, Acvrl1 mutant cells were screened by single-cell colony culture and PCR. PCR primers targeting the Acvrl1 mutant allele were designed as 5′-CTAGGACTTCCTGGACAGCG-3′ and 5′-CAGCCCGTTGAGTGCTTACC-3′. Mutation in Acvrl1 was confirmed by sequencing.

PLA.

PLA was performed directly after IHC following manufacturer’s instructions (Sigma-Aldrich; Duolink PLA Technology). After blocking, sections were incubated with goat anti-ACVRL1 (5 μg/mL; Santa Cruz; 15946), rabbit anti-TGFβR1 (5 μg/mL; Abcam; 31013), and rabbit anti-ACTRIIB (5 μg/mL; Santa Cruz; 25453) antibodies overnight at 4 °C, followed by incubation with secondary antibodies conjugated with PLA probe at 37 °C for 1 h as recommended by manufacturers. Then, ligation and amplification were performed (Duolink detection kit; orange, 555 nm). Finally, mounting medium with DAPI was used.

Statistical Analysis.

Statistical comparisons were made between 2 groups using Student’s t test; P values of <0.05 were considered significant. For multiple comparisons, we performed 2-way ANOVA (2 factor with replication) analysis, followed by post hoc tests that uses t test to compare 2 groups and controls the familywise error rate by Bonferroni correction of P values (0.05/n, n = number of groups).