Development of the mouse heart begins in late gastrulation when mesodermal-derived cardiac progenitor cells generate a rudimentary heart tube that subsequently undergoes looping, trabeculation, and compartmentalization (Buckingham et al.,2005). In addition, neural crest cells emanating from the dorsal folds of the neural tube migrate to the cardiac microenvironment via the pharyngeal arches and OFT, and subsequently play essential roles in heart morphogenesis (Harvey,2002). These complex cardiac cell behaviors are controlled by cytokines and growth factors such as Fgf8 and Wnt5a (Abu-Issa et al.,2002; Schleiffarth et al.,2007), intracellular signaling effectors including Pinch1 and Fak (Liang et al.,2007; Vallejo-Illarramendi et al.,2009), and cell-cell adhesion proteins belonging to the Notch family (Jain et al.,2010). Genetic ablation of many of these components in mice leads to developmental pathologies that are similar to those that form in humans with congenital heart defects (Srivastava,2006).
Extracellular matrix (ECM)-regulated adhesion and signaling pathways are also important for normal heart development. For example, genetic deletion of both the EIIIA and EIIIB splice variants of fibronectin leads to severe cardiovascular phenotypes and early embryonic lethality (Astrof et al.,2007). Mammalian cells adhere to ECM proteins via integrin cell surface receptors that control intracellular signaling pathways and cytoskeletal dynamics (Hynes,2009). Many integrins are expressed in cardiac cells, including those derived from the neural crest, and are essential for normal development (Thiery,2003). Antibodies that neutralize αv or β1 integrins, for example, induce widespread neural crest defects during avian embryogenesis (Bronner-Fraser,1986; Delannet et al.,1994). Additionally, combined genetic deletion of αv and α5 integrins leads to abnormal cardiovascular development (van der Flier et al.,2010), likely owing to defective cardiac neural crest cell growth and survival (Mittal et al.,2010).
Here we have used molecular genetic strategies to determine functions for integrin αvβ8 and its intracellular signaling protein, Band 4.1B, during embryogenesis. The cytoplasmic tail of β8 integrin interacts directly with the C-terminal domain of Band 4.1B (McCarty et al.,2005a); however, cooperative in vivo functions for β8 integrin and Band 4.1B have not been explored. We show that most mouse embryos genetically null for β8 integrin and Band 4.1B die by mid-gestation due to abnormal heart formation and function that may be related to defects in the cardiac neural crest. These data are the first to establish in vivo functional links between αvβ8 integrin and Band 4.1B during embryogenesis and suggest that these components, or the pathways they regulate, may be altered in congenital heart diseases.
To analyze in vivo functions for β8 integrin and Band 4.1B during embryogenesis, mice homozygous null for the Band 4.1B gene (4.1B−/−) were interbred with mice harboring a β8 integrin heterozygous null allele (β8+/−). The resulting F1 compound heterozygotes (β8+/−;4.1B+/−) were born in the expected Mendelian ratios, and were viable and fertile with no obvious phenotypes (data not shown). Therefore, we interbred β8+/−;4.1B+/− animals to produce β8+/−;4.1B+/+ breeding pairs for generating wild type control (+/+; +/+) or β8−/− (β8 integrin single knockout) progeny. Alternatively, 4.1B−/− (4.1B single knockout) or β8−/−;4.1B−/− (double knockout) littermates were generated by interbreeding β8+/−;4.1B−/− males and females. Genotypes of all mice were confirmed by PCR-mediated gene amplification using DNA isolated from ear or tail snips (see Supp. Fig. S1, which is available online).
As shown in Table 1, analysis of pups within 8 hr after birth (P0) revealed higher than expected Mendelian ratios of wild type (n=43/139, or 31% vs. the expected 25%) and 4.1B single knockouts (n=60/224, or 27% vs. the expected 25%). Slightly lower than expected numbers of β8−/− mice (n=28/139, or 20% vs. the expected 25%) were found alive, indicating a small percentage of embryonic or early neonatal lethality. One hundred percent of the β8 integrin single knockouts found alive at birth displayed intracerebral hemorrhage (Supp. Fig. S2) similar to what has been reported previously (Zhu et al.,2002; Mu et al.,2008; Mobley et al.,2009). All β8−/− pups that were monitored beyond P0 (n=16) survived the neonatal period but none lived beyond P40 (Supp. Fig. S3), likely owing to hemorrhage-related neurological defects such as hydrocephaly and lower limb paresis (data not shown).
Table 1. Genotype Distribution of Wild Type, Single Knockout, and Double Knockout Mice at E10.5, E11.5, and P0
aTo generate wild type (+/+; +/+) and β8 integrin single knockouts (β8−/−), β8+/−; 4.1B+/+ mice were interbred. To generate 4.1B single knockouts (4.1B−/−) and double knockouts (β8−/−;4.1B−/−), β8+/−;4.1B−/− mice were interbred. The expected Mendelian ratios for all genotypes at the various developmental ages are 25%. Note that approximately 60% of double knockout embryos die between E10.5 and E11.5.
Interestingly, of 224 pups born from β8+/−;4.1B−/− intercrosses, only 23 double knockouts were found alive (10% vs. the expected 25%), indicating 60% embryonic lethality in double knockout mice (Table 1). Like β8 integrin single knockouts, double knockout pups also displayed severe intracerebral hemorrhage (Supp. Fig. S2). All double knockouts monitored beyond P0 (n=18) survived the neonatal period but died by P40 (Supp. Fig. S3). In contrast, 4.1B single knockouts (60/224 or 27%) generated from β8+/−; 4.1B−/− intercrosses developed normally and survived into adulthood (Table 1 and Supp. Fig. S3). 4.1B+/− intercrosses also yielded expected Mendelian ratios of 4.1B−/− progeny (25%) that did not display obvious phenotypes, similar to what we have reported previously (Yi et al.,2005). Lastly, β8+/−;4.1B−/− mice did not display obvious neonatal or adult abnormalities (data not shown), and at birth they were found in slightly higher than expected Mendelian ratios due to lethality in double knockout littermates.
To identify the time of embryonic lethality in double knockouts, we intercrossed mice and analyzed F1 embryos at specific developmental ages. As summarized in Table 1, wild type (34/163, or 21%), β8−/− (35/163, or 22%), and 4.1B−/− (43/182, or 24%) embryos analyzed at E10.5 were found in expected Mendelian ratios with no obvious developmental abnormalities (Fig. 1A–C, E–G). We also found the expected Mendelian distribution of double knockout embryos at E10.5 (42/182, or 23%). Analysis of these 42 viable double knockout embryos, however, revealed 11 with noticeable growth retardation and hypovascularity (Fig. 1D). Six of the 11 double knockout embryos also displayed obvious pericardial edema (Fig. 1H). Two additional E10.5 double knockout embryos were dead, as determined by the absence of a detectable heartbeat and widespread necrosis (data not shown).
Analysis of double knockout embryos at E11.5 revealed significantly reduced viability (Table 1). Unlike wild type and single knockouts, which were found in slightly higher than expected Mendelian ratios, we detected 68% fewer than expected double knockout embryos (6/78 or 8% vs. the expected 25%). The remaining 40% of double knockouts that were viable at E11.5 showed focal intracerebral hemorrhage similar to β8 integrin single knockouts (Supp. Fig. S2). The lower percentages of viability detected for double knockout embryos at E11.5 (8%) is similar to the reduced numbers of double knockout pups found alive at P0 (10%), revealing that nearly all death in double knockouts occurs between E10.5 and E11.5.
To further characterize the embryonic vasculature, sagittal sections through E10.5 embryos were stained with H&E or an anti-laminin antibody to visualize vascular basement membranes. Wild type (Fig. 2A, E) and 4.1B single knockouts (Fig. 2C, G) displayed normal blood vessel morphologies in the developing neural tube. Blood vessels within the neural tube of β8 single knockouts, however, showed distended, sinusoidal-like morphologies (Fig. 2B, F). Similar to β8 integrin single knockouts, double knockout embryos also displayed abnormal blood vessel morphologies in the neural tube (Fig. 2D, H). These data are consistent with prior reports showing that αvβ8 integrin in the neuroepithelia controls normal blood vessel growth and sprouting in the CNS (McCarty et al.,2005b; Proctor et al.,2005).
Yolk sacs from E10.5 wild type, single knockout, and double knockout embryos were also immunostained with anti-CD31 antibodies to reveal extraembryonic blood vessels. Elaborate vascular networks were detected in wild type and single knockout yolk sacs (Fig. 3A–C, top panels); however, many double knockout yolk sacs contained blood vessels with abnormal morphologies and reduced sprouting (Fig. 3D, top). In addition, H&E-stained histological sections taken from wild type and single knockout yolk sacs revealed close juxtaposition between endoderm and mesoderm, with blood vessels containing nucleated erythrocytes (Fig. 3A–C, bottom panels). In contrast, H&E-stained double knockout yolk sacs displayed compressed endodermal cytoarchitecture with blood vessels displaying reduced diameters and containing few circulating erythrocytes (Fig. 3D, bottom) consistent with their hypovascular appearance (Fig. 1D, H).
These phenotypes suggested cooperative functions between αvβ8 integrin and Band 4.1B in the yolk sac vasculature; therefore, we examined αvβ8 integrin and Band 4.1B protein expression by immunoblotting detergent-soluble lysates prepared from E10.5 yolk sacs. β8 integrin protein expression was not detected in yolk sacs, although αv integrin and Band 4.1B proteins were expressed (Supp. Fig. S4A). In contrast, detergent-soluble lysates from E10.5 hearts revealed robust expression αvβ8 integrin and Band 4.1B proteins (Supp. Fig. S4B). These data suggest that systemic blood vessel pathologies in double knockouts are likely secondary consequences of other cardiovascular-related defects.
We next examined heart morphologies in wild type, single knockout and double knockout embryos by staining sections with H&E. Atrioventricular cushions and conotruncal cushions displayed normal morphologies in E10.5 wild type (Fig. 4A) and single knockout embryos (Fig. 4B, C). Most double knockout embryos, however, displayed endocardial cushion hypotrophy in the OFT (Fig. 4D). The conotruncal cushions and OFT mesenchyme are formed, in part, by cardiac neural crest cells that express smooth muscle markers; therefore, we immunostained wild type, single knockout and double knockout embryos with anti-SMAα-actin. In comparison to wild type (Fig. 4E, I) and single knockout embryos (Fig. 4F, G, J, and K), reduced SMAα-actin expression was seen in the myocardium and OFT mesenchyme in double knockouts (Fig. 4H, L). Immunostaining hearts with a second smooth muscle marker, desmin, also revealed diminished expression in the myocardium and OFT of double knockouts (Supp. Fig. S5).
Since heart morphogenesis involves contributions from cardiac neural crest cells (Snarr et al.,2008), we analyzed neural crest patterning in E10.5 wild type, single knockout, and double knockout embryos by whole-mount immunostaining with an anti-neurofilament antibody 2H3 (Dodd et al.,1988). As shown in Figure 5A–C, we detected an elaborate network of neurofilament-positive cells in wild type (n=6), β8−/− (n=4), and 4.1B−/− (n=7) embryos; however, most double knockout embryos showed obvious patterning defects. The trigeminal ganglion in double knockouts was smaller than the controls and displayed shortened projections into the ophthalmic, maxillary, and mandibular placodes (Fig. 5D). In addition, neurofilament-positive cells in the pharyngeal arches were absent in most double knockout embryos. We also detected abnormal neurofilament-positive projections in the trunks of double knockout embryos (Supp. Fig. S6). These widespread defects were detected in three of five E10.5 double knockout embryos analyzed, with the remaining double knockouts displaying apparently normal spatial patterning of neurofilament-expressing cells.
In this report, we have used gene knockout strategies in mice to analyze functional links between β8 integrin and the Band 4.1B cytoskeletal adapter protein during embryogenesis. We show that ablation of both β8 integrin and Band 4.1B causes developmental cardiovascular phenotypes and lethality by mid-gestation; hence, these results are the first evidence for cooperative signaling between β8 integrin and Band 4.1B in vivo.
The partially penetrant phenotypes that we detect in double knockout embryos are likely related to strain variation. Prior reports have shown strain-dependent effects on β8−/− phenotypes; for example, on a C57BL6/129S4 heterogeneous background the majority of β8−/− mice are reported to die at mid-gestation probably due to placental defects (Zhu et al.,2002). We have shown, however, that nearly 100% of β8−/− mice on a mixed C57BL6/129S4/CD1 background are born alive (Mobley et al.,2009), indicating that there may be strain-specific genetic modifiers of the placental phenotype. Indeed, we did not detect placental defects in single or double knockout embryos analyzed in this study (data not shown). Interestingly, mice genetically null for TGFβ1, an ECM protein ligand for αvβ8 integrin (Mu et al.,2002; Annes et al.,2004), also develop strain-dependent phenotypes with highly variable patterns of lethality (Bonyadi et al.,1997).
Proper morphogenesis of the heart, and in particular the OFT, is dependent on the cardiac neural crest (Stoller and Epstein,2005). Given that other mutant mouse models that display similar OFT phenotypes have cardiac neural crest defects (Conway et al.,2003), it is enticing to speculate that αvβ8 integrin and Band 4.1B function to promote cardiac neural crest migration to the OFT and possibly other regions of the developing heart. Alternatively, αvβ8 integrin-Band 4.1B adhesion and signaling pathways may promote cardiac neural crest cell growth and survival within to the heart microenvironment. We have not detected increased apoptosis in double knockout embryos (data not shown); however, in most E10.5 double knockout embryos, we detect abnormal neurofilament-positive projections into the pharyngeal arches, suggesting that β8 integrin and Band 4.1B may cooperatively promote cardiac neural crest cell migration (Fig. 5). It should also be noted that proper development of the OFT also involves mesoderm-derived endocardial cells (Acloque et al.,2009); therefore, αvβ8 integrin and Band 4.1B may function in more than one cardiac cell type, e.g., those derived from the local mesoderm and distal neural crest.
The cardiovascular phenotypes that we report herein develop in the absence of both β8 integrin and Band 4.1B gene expression, indicating that these two proteins may act in parallel and/or redundant signaling pathways. For example, Band 4.1B and possibly other Band 4.1 family members may modulate signaling pathways downstream of αvβ8 integrin and other cell surface receptors. In this scenario, functions for Band 4.1 proteins would likely remain unaltered in the absence of β8 integrin. Interestingly, lethal cardiovascular phenotypes have been reported in mice that lack both αv and α5 integrin genes (van der Flier et al.,2010), raising the intriguing possibility that Band 4.1B may serve as an intracellular “signaling nexus” for multiple integrins. The C-terminal domain of Band 4.1B binds directly to the β8 integrin cytoplasmic tail (McCarty et al.,2005a), potentially enabling the Band 4.1B N-terminal FERM domain to interact with the cytoplasmic domains of other integrin subunits or other transmembrane proteins.
The neural crest also contributes to proper development of the palate (Smith and Tallquist,2010). Interestingly, αv−/− and β8−/− mice develop a cleft palate, although with variable penetrance (Bader et al.,1998; Zhu et al.,2002). Various reports have demonstrated that TGFβ3, a ligand for αvβ8 integrin, plays essential roles in palate development (Kaartinen et al.,1995; Proetzel et al.,1995). In addition, selective ablation of TGFβ receptors or Smads in neural crest cells results in cleft palate formation and impaired morphogenesis of the OFT (Kaartinen et al.,2004; Wurdak et al.,2005; Wang et al.,2006; Nie et al.,2008). Therefore, αvβ8 integrin and Band 4.1B may cooperatively function to regulate TGFβ activation in the neural crest. We are currently using various cell type–specific gene ablation approaches to investigate integrin-mediated TGFβ activation and signaling via Band 4.1B in neural crest development.
Generation and characterization of 4.1B−/− mice has been described elsewhere (Yi et al.,2005). To generate the various single and double gene knockouts analyzed in this study, 4.1B−/− mice (C57BL6/129S4 background) and β8+/− mice (C57BL6/129S4/CD-1 background) were crossed and F1 progeny were interbred. The genotypes of all mice were determined using PCR-based methods described previously (Zhu et al.,2002; Yi et al.,2005) (Supp. Fig. S1). The primer sequences used for β8 integrin gene amplification are as follows: 5′-ATTATCTGGTTGATGTGTCAGC-3′, 5′-GGAGGCATACAGTCTAAATTGT-3′, 5′-AGAG GCCACTTGTGTAGCGCCAAG-3′, and 5′-AGAGAGGAACAAATATCCTTCCC-3′. The primer sequences used for amplification of the 4.1B gene are as follows: 5′-CGCACCTGGTGCATGACC-3′, 5′-CGCCACCGTCTGAGCAGC-3′, and 5′-GCACG TTTGGTAGCAGTTCCC-3′. Embryo staging and analyses were performed by timed matings, with noon on the plug date defined as E0.5.
Antibodies, Immunoblotting, and Immunohistochemistry
E10.5 wild-type yolk sacs and hearts were lysed in 50 mM Tris, pH7.4, 150 mM NaCl, 1% NP40, 2.5 mM EDTA containing a cocktail of protease and phosphatase inhibitors (Roche, Mannheim, Germany). Detergent-soluble lysates were resolved by SDS-PAGE and then immunoblotted with anti-4.1B, anti-β8 integrin, and anti-αv integrin rabbit polyclonal antibodies as described previously (McCarty et al.,2005a,b; Mobley et al.,2009; Tchaicha et al.,2010). The HRP-conjugated goat anti-rabbit IgG used for immunoblotting was purchased from Jackson ImmunoResearch (West Grove, PA).
Embryos and yolk sacs were fixed in ice-cold 4% PFA/PBS for 12–16 hr and embedded in paraffin or Tissue-Tek OCT freezing compound (Sakura Finetek, Torrance, CA). The following antibodies used for immunohistochemistry were purchased from commercial sources: rabbit anti-laminin (1:100, Sigma, St. Louis, MO), rabbit anti-desmin (1:200, Abcam, Cambridge, CA), mouse anti-SMAα-actin (1:500, Sigma, St. Louis, MO), biotinylated swine anti-rabbit IgG (1:250, DAKO, Carpentaria, CA) and biotinylated horse anti-mouse IgG (1:250, Vector Laboratories, Burlingame, CA). Embryo sections were then analyzed using a Zeiss Axio Imager Z1 microscope.
Whole-Mount Embryo Immunostaining
Embryos and yolk sacs were fixed in a mixture of methanol:DMSO (4:1) for 2 hr at 4°C and stored in 100% methanol at −20°C. The tissues were rehydrated, bleached in 3% H2O2 in PBS for 2 hr, and blocked in PBSMT (2% non-fat milk powder, 0.1% Triton X-100 in PBS) for 2 hr at room temperature. Embryos or yolk sacs were then incubated with mouse anti-neurofilament 2H3 (3 μg/ml, University of Iowa Hybridoma Bank) or rat anti-CD31 (1:100, BD Biosciences, San Jose, CA) diluted in PBSMT overnight at 4°C. The samples were then washed five times in PBSMT at room temperature for 1 hr each and incubated with HRP-conjugated goat anti-rat IgG or goat anti-mouse IgG (1:500, Jackson ImmunoResearch, West Grove, PA). Embryos were washed five times in PBSMT at room temperature for 1 hr each and in PBT (0.2% BSA, 0.1% Triton X-100 in PBS) overnight at 4°C. The samples were developed using DAB chromagen kit (Vector Laboratories, Burlingame, CA), cleared in 100% glycerol, and photographed under a dissection microscope.
We thank Mohammad Hossain for breeding and genotyping experimental mice and Dr. S. Paul Oh (University of Florida) for insightful comments on the manuscript.