The Lhx9-integrin pathway is essential for positioning of the proepicardial organ

Cardiovascular disease is one of the highest causes of mortality and reduced quality of life in the USA (Roger et al., 2012). To better optimize therapeutics to preserve and maintain heart physiology it is pertinent that we understand the mechanisms involved in all aspects of heart development, growth and repair. This will further enable us to target resident cardiac cell populations and gene programs for clinical applications.

The heart initially forms in vertebrates as a bilaminar tube comprising an inner endocardium and outer myocardial layer. At later stages of development, a third epithelial-like sheet layer, the epicardium, is added to the heart. In many vertebrates this layer forms from a dynamic precursor cell population, the proepicardial organ (PEO), which arises as a cluster of cells on the right side of the septum transversum adjacent to the heart (Schulte et al., 2007; Jahr et al., 2008; Männer, 1992; Pombal et al., 2008; Serluca, 2008). Whereas FGF and BMP signaling, as well as the transcription factors Snai1 and Twist, have been implicated in PE induction (Schlueter and Brand, 2009, 2013; Schulte et al., 2007), there is little known about the mechanisms that direct the formation of the PE cell cluster. Once the cluster has formed and is positioned correctly, it is induced to bridge towards the developing heart, directed by BMP signaling from the atrioventricular sulcus (AVS) (Ishii et al., 2010).

The epicardial layer is an essential cardiac component because it provides mitogenic stimulation to the developing myocardium (Kang et al., 2008; Lavine et al., 2005; Li et al., 2011; Pennisi et al., 2003; Stuckmann and Lassar, 2002). In addition, the epicardium is the source of epicardial-derived cells (EPDCs) whereby subpopulations of this layer undergo epithelial-mesenchymal transition (EMT) and invade the underlying myocardium. Here, they differentiate into vital cardiac cells such as smooth muscle cells and pericytes of the vasculature and fibroblasts that contribute to the scaffold and support of the heart (Gittenberger-de Groot et al., 2010; Lie-Venema et al., 2007; Männer et al., 2001; Ratajska et al., 2008; von Gise and Pu, 2012). Current understanding is that EPDC lineage specificity is determined early within the precursor PE cell cluster prior to EMT (Acharya et al., 2012; Guadix et al., 2006; Männer, 1999; Mikawa and Gourdie, 1996; Katz et al., 2012; Wei et al., 2015). In the event of a cardiac infarction or damage in the adult, research has also demonstrated the utility of the epicardium in providing regenerative support and cardiac cells to the damaged tissue (Lepilina et al., 2006; Limana et al., 2011; Winter and Gittenberger-de Groot, 2007; Winter et al., 2009). Therefore, better comprehension of how this source of pluripotent cells is specified and develops is vitally important to understanding mechanisms of cardiac development, disease and repair.

Here, we report that LIM homeobox transcription factor 9 (Lhx9) is essential for the formation and function of the epicardium. We show that loss of Lhx9 in Xenopus leads to defects in the assembly of the PE cluster and subsequently failed migration and spreading of epicardial cells onto the surface of the developing heart. Using a pharmacological small-molecule screen in Xenopus, we ascertained that integrin-paxillin signaling is required for proper PE clustering, which is confirmed by a significant decrease in epicardial integrin alpha 4 (itga4) expression upon loss of Lhx9 function. Given that interactions between cells and the surrounding ECM are vital for epicardial formation and its role in cardiac repair (Kálmán et al., 1995; Nahirney et al., 2003; Pae et al., 2008; Wang et al., 2013; Benesh et al., 2013; Fransen and Lemanski, 1991; Mercer et al., 2013; Rongish et al., 1996; Yang et al., 1995), we further show a novel function for Lhx9 in mediating integrin adhesion mechanisms for correct PE cell clustering. Therefore, Lhx9 is vital for the formation of the epicardium and development of the heart.

Animals lacking a functional epicardium die prematurely and display hemorrhaging hearts with thin ventricular walls (Moore et al., 1998; Lie-Venema et al., 2005; Gittenberger-de Groot et al., 2000; Acharya et al., 2012; Combs et al., 2011). Many vertebrates show a right side clustering of PE cells during epicardial formation (Virágh et al., 1993; Kálmán et al., 1995; Nahirney et al., 2003; Serluca, 2008; Pombal et al., 2008; Jahr et al., 2008). Recently, we have demonstrated that the LIM homeodomain transcription factor family lhx9 is transiently expressed in the Xenopus PEO (Tandon et al., 2013). We therefore conducted a detailed analysis to establish the relationship between Lhx9 expression and formation of the PEO on the septum transversum.

The sequence and structure of Lhx9 has been shown to be highly conserved across vertebrates (Bertuzzi et al., 1999; Retaux et al., 1999; Failli et al., 2000; Bachy et al., 2001; Ottolenghi et al., 2001; Alunni et al., 2007; Oshima et al., 2007; Peukert et al., 2011) and comprises two LIM protein-binding domains and a DNA-binding homeodomain (HD). Consistent with other vertebrates, we identified two isoform variants of lhx9 in Xenopus, one with a full-length HD (lhx9-HD) and one with a truncated HD (lhx9-α) generated by alternative splicing of exon 5 (Failli et al., 2000; Mollé et al., 2004; Fig. S1).

To ascertain the endogenous role of Lhx9 during epicardial development we performed a detailed spatio-temporal analysis of both lhx9 transcripts. Results from these studies show that lhx9-α is initially expressed bilaterally on the septum transversum prior to PEO formation (Fig. 1A,E, Figs S2-S4). In all vertebrates, the lateral confinement of the PEO cells to the right side of the embryo can be observed by the expression of the conserved PEO marker tbx18 (Fig. 1K-N, Fig. S3; Tandon et al., 2013). Consistently, we observed a dramatic restriction of lhx9-α expression to the right side of the septum transversum (stages 39-46) (Fig. 1B-D,F) and subsequently, lhx9-α was downregulated relative to tbx18 as the PEO cells matured and migrate onto the heart (Fig. 1D,N, Fig. S3; Tandon et al., 2013). An identical pattern of expression was obtained with an independent probe recognizing the LIM domains of lhx9-α and lhx9-HD (Fig. 1E,F, Figs S1, S4 and S11). We further observed a highly similar pattern of expression for lhx9-HD in the septum transversum (Fig. 1G-J, Fig. S5); however, lhx9-HD was more weakly expressed and dramatically downregulated upon PE clustering compared with lhx9-α (Fig. 1I, Fig. S5). We also did not observe detectable expression of lhx2, a closely related LIM-HD transcription factor to Lhx9, in the developing epicardium at any stage (Fig. S6) (Viczian et al., 2006).Taken together, these studies demonstrate that lhx9-α and lhx9-HD transcripts have overlapping dynamic expression during epicardial development and lhx9-α is the predominant transcript expressed during PEO formation.

To determine the requirement for Lhx9 in the vertebrate PEO, we depleted Lhx9 in Xenopus laevis using morpholinos (see supplementary Materials and Methods for details). Depletion of Lhx9 function in Xenopus does not impede early embryonic development. However, analysis of the developing heart at tadpole stages revealed a severe epicardial defect. Whereas control embryos displayed PE cells clustering to the AVS on the right side of the heart and a layer of epicardial cells on the ventricle surface (Fig. 2A,C, red arrowheads marking PE cell cluster, tbx18 in situ hybridization), embryos lacking Lhx9 failed to accumulate PE cells to the right-side AVS and displayed a reduced and discontinuous epicardial layer on the ventricular surface (Fig. 2B,D-F, red arrowheads).

Having demonstrated that proper epicardial formation requires Lhx9 function and because lhx9 expression is detected on the septum transversum and PE cluster (Fig. 1A-F, Fig. S2), we assessed whether Lhx9 functions during the earlier stages of epicardial specification. At stage 38, the conserved pan-marker of the epicardial lineage, tbx18, was observed on the septum transversum and deemed indistinguishable between control and Lhx9-depleted embryos (Fig. 2G,H). Therefore Lhx9 is not required for specification of the epicardial lineage. However, by stage 41, when PE cells begin to cluster to the right side of embryos (Fig. 2I, red arrowhead), Lhx9-depleted stage-matched sibling embryos maintained diffuse tbx18 expression bilaterally along the septum transversum (Fig. 2J, green arrowheads, Fig. S9A). By stage 42, the PE cluster had attached to the AVS on the right side of the heart in controls, whereas in Lhx9-depleted embryos there was either no detectable cluster and tbx18 expression persisted on the anterior endoderm/septum transversum or the cluster was mispositioned caudal to the heart (Fig. 2L-N).

Consistently, the spatial distribution of two independent epicardial markers, tcf21 and wt1, were similarly disrupted in Lhx9-depleted embryos. Both tcf21 and wt1 expression mark the punctate cluster of PE cells at stage 41 on the right side of the embryo (Fig. 2O,Q,S, red arrowheads), whereas expression of these epicardial markers is reduced and mislocalized in Lhx9-depleted embryos (Fig. 2P,R,T,U, red arrowheads, Fig. S9B). These data collectively imply that Lhx9 function is crucial for normal clustering of PE cells during early stages of development and in its absence, epicardial formation is disrupted.

The cellular and molecular mechanisms that lead to clustering of the PEO on the right side of the septum transversum remain entirely unknown. To assess whether active cell migration is required for asymmetric PE clustering, we conducted a screen in wild-type embryos with a defined bank of small-molecule inhibitors known to disrupt cell migration and adhesion in Xenopus (Harata et al., 2013; Broders-Bondon et al., 2007). Whereas inhibition of myosin II (using Blebbistatin) resulted in tadpoles with severe pericardial edema (possibly due to a reduced heart beat) and paralysis, we did observe clustered PE cells (marked by tbx18) to the right of the embryo. However, because of the pronounced edema, we were unable to determine attachment to the heart (data not shown). Inhibition of Rac1 activity (using NSC23766 trihydrochloride), which has been extensively shown to disrupt cell migration (Diaz et al., 2014; Huang et al., 2013), did not affect PE cell clustering compared with controls, as depicted by tbx18 expression on the AVS (Fig. 3A,B,G-J). Little tbx18 expression was preserved on the septum transversum in these embryos, implying that PE clustering is not dependent on classical cell migration mechanisms involving actin-myosin interactions and GTPase activity.

Interestingly, exposure to a small molecule that disrupts integrin-paxillin interactions (6-B345TTQ) from stages 38 to 41 when PE cell clustering is occurring, led to a PE cluster positioning defect phenocopying loss of Lhx9 function. Integrin binds to ECM components and induces the recruitment of proteins, such as paxillin, to focal adhesions at the site of cell attachment. These protein complexes then function to reorganize the actin cytoskeleton and are essential for the coordinated adhesion and motility of cells through an ECM environment (Bellis et al., 1995). Paxillin is a multi-LIM domain protein that has been shown to directly and specifically bind to the cytoplasmic tail of Itgα4 to coordinate cell spreading, adhesion and migration (Liu et al., 1999; Alon et al., 2005). Although some clustering of PE cells was evident in cultured tadpoles, we observed mispositioning of the cluster to the caudal side of the heart as well as persistent tbx18 expression on the septum transversum (Fig. 3C-F,K) as seen in Lhx9-depleted embryos (Fig. 2H,J,L,M). This result therefore suggests that integrin-mediated mechanisms play an important role in the correct positioning of the PEO cell cluster in developing embryos.

Our data demonstrate that Lhx9 as well as integrin-mediated cellular functions are essential for the correct clustering and positioning of PE cells. We previously reported lhx9 and itga4 are co-regulated in response to loss of the epicardial transcription factor Tcf21 (Tandon et al., 2013). Based on these observations we hypothesized that Lhx9 acts upstream of itga4 to correctly position the PEO. Consistent with this hypothesis, itga4 expression was detected subsequent to lhx9 in the clustered PE cells from stages 39 to 46 (Fig. 4A-D, red arrowheads, Fig. S3). We thus assessed the expression of itga4 in Lhx9-depleted hearts. Consistent with itga4 acting downstream of Lhx9, itga4 expression was reduced and misplaced in Lhx9-depleted embryos (Fig. 4E,F,O, Figs S8 and S9C). By contrast, expression of integrin β1 (Itgβ1) strongly localized to the epicardium (as marked by tcf21) and was independent of Lhx9 function (Fig. 4G-J′, Fig. S11), therefore enabling its use as a marker of epicardial cells. Owing to the clustering defects observed upon reducing the Itgα4-paxillin interaction biochemically (Fig. 3) and reduced itga4 expression in Lhx9-depleted embryos, we sought to determine whether Lhx9 is responsible for integrin-mediated cell adhesion and migration processes in PE cells. To address this possibility, we assayed for phosphorylated paxillin (Y118), which has been used to demonstrate integrin activation at focal adhesions and shown to modulate cell spreading, migration and invasion (Bellis et al., 1995; Burridge et al., 1992; Turner, 1991; Nakamura et al., 2000; Lewis and Schwartz, 1998; Zaidel-Bar et al., 2007; Sachdev et al., 2009). Our data indicated a significant decrease in expression of phosphorylated Y118 paxillin specifically in the attached PE cluster in Lhx9-depleted embryos compared with controls (Fig. 4K-N′,P). Taken together, these data strongly imply Lhx9 acts through Itgα4 to correctly position the PEO.

Collectively, our data demonstrate a role for Lhx9-Itgα4 in the correct positioning and clustering of the PEO. Our data also suggest that integrin activity at focal adhesions is required for the clustering of PE cells. Previous data has implicated Itgα4 as being essential for epicardial development particularly in regulating PE cell clustering, epicardial adhesion and epicardial migration (Sengbusch et al., 2002; Pinco et al., 2001; Yang et al., 1995; Dettman et al., 2003; Dokic and Dettman, 2006; Hirose et al., 2006; Pae et al., 2008; Bax et al., 2010). Integrin heterodimers have been shown to coordinate cytoskeletal interactions and adhesion with the surrounding ECM environment (Gardiner, 2011; Gehler et al., 2013; Huttenlocher and Horwitz, 2011; Manninen, 2015; Wolfenson et al., 2013), which are pertinent processes for epicardial formation and function (Kálmán et al., 1995; Nahirney et al., 2003).

To determine whether the depleted adhesive properties of the PEO and its mis-positioning have persistent biological consequences in the development of the vertebrate epicardium and heart, we examined the epicardial ECM environment of control and Lhx9-depleted hearts. Using Itgβ1 as a marker of epicardial cells, we analyzed the expression of known Itgα4 ligands and ECM components. Although we observed no discernible alterations in the expression of the Itgα4-Itgβ1 (very late antigen 4, VLA-4) ligand vascular cell adhesion molecule 1 (vcam1, Fig. S10) in Lhx9-depleted embryos, we did detect a decrease in fibronectin within the PE cluster, as well as in the heart, relative to controls (Fig. 5A-E) (Humphries et al., 1995; Wu et al., 1995). This suggests that alterations to integrin function in response to loss of Lhx9 are specific to Itgα4 pathway components. In addition, these data further imply that the adhesive qualities and positioning of the PEO are crucial for the proper deposition of epicardial ECM.

To address this further, we examined the localization of epicardial-ECM markers laminin and β-dystroglycan. In control embryos, we observed weak laminin deposition surrounding attached PE cells (Fig. 5F,F′, white arrowheads) but a continuous smooth layer of laminin was present under the mature epicardial layer (Fig. 5F″, white arrowheads). However, in Lhx9-depleted embryos, when epicardial cells were present on the myocardial surface, we observed an increased accumulation of laminin in the basal portion of attached PE cells (Fig. 6G,G′, white arrowheads) as well as discontinuous deposits within epicardial cells on the surface of the heart (Fig. 5G″, white arrowheads). To confirm these findings, we analyzed β-dystroglycan, a transmembrane glycoprotein that attaches cells to the basement membrane through laminin binding (Ervasti and Campbell, 1993; Klietsch et al., 1993; Smalheiser and Schwartz, 1987). In Lhx9-depleted embryos, β-dystroglycan was reduced or disrupted at the point of contact between the PE cluster and heart surface (as indicated by tcf21) (Fig. 5J-K′) compared with controls (Fig. 5H-I′). Note that β-dystroglycan was prominent in the endocardium of both control and Lhx9-depleted hearts, showing that this defect is specific to the epicardium (Fig. 5I′,K′, yellow arrowheads). Taken together, these results demonstrate PE cell adhesive integrity is required for the proper formation of the epicardial layer.

Collectively, our data are consistent with a role for Lhx9 in the clustering and positioning of the PEO (Fig. 6A,B). Our data further show that the ability of Lhx9 to regulate Itgα4 signaling is vital for correct establishment of clustering, attachment and migration of PE cells to the heart (Fig. 6A,C,D). These signals are, in turn, necessary for the correct deposition of ECM components, thus enabling epicardial cells to form a cohesive layer over the myocardial surface (Fig. 6E,F).

Our data collectively show that Lhx9 is essential for proepicardial positioning and loss of Lhx9 is associated with aberrant deposition of essential ECM components during crucial stages of epicardial clustering and migration. We further observe that loss of Lhx9 correlates with a decrease in integrin-ECM signaling and the ability of PE cells to form cohesive adhesions downstream of integrin activity. Because the PE cluster defect in Lhx9-depleted embryos closely resembles the defect caused by disturbing the Itgα4-paxillin interaction biochemically, we propose and demonstrate how Lhx9 functions to modulate itga4 expression. Taken together, these data establish a role for Lhx9 in integrin-mediated cell adhesion and motility crucial for the proper positioning and clustering of PE cells.

Although most vertebrates show a right-side clustering of PE cells during epicardial formation (Virágh et al., 1993; Kálmán et al., 1995; Nahirney et al., 2003; Serluca, 2008; Pombal et al., 2008; Jahr et al., 2008), it remains to be established whether the sidedness of epicardial formation is essential for its function. In our studies, we further found that alterations in the positioning of the epicardium on the septum transversum is essential for the correct attachment of the epicardium to the heart at the AVS, implying that alterations in PE clustering predispose an embryo to cardiac defects.

The function of Itgα4 during epicardial development has been characterized in mice mutants that displayed absence of the epicardial layer and lack of coronary vessels at the AVS (Yang et al., 1995). This was shown to be due to defects in the migratory process (Sengbusch et al., 2002) as we also report here. Therefore, the role of Itgα4 in epicardial migration appears to be conserved regardless of the mode of epicardial formation, such as cluster bridge in frog or free-floating cysts in mouse. Our data provide new evidence that expression of itga4 is restricted to immature PE cells and is downregulated as the epicardial layer matures, similar to the hematopoietic lineage (Arroyo et al., 1999; Lobb and Hemler, 1994). Using Itgβ1 as a marker of epicardial cells, we analyzed the expression of known Itgα4 ligands and ECM components. We report a downregulation of fibronectin in embryos lacking Lhx9 but observed no alterations in expression of vcam1, the Itgα4-Itgβ1 ligand. Taken together, these findings demonstrate that loss of Lhx9 functions upstream of itga4 and that the function of Lhx9 is specific to Itgα4 function.

Our previous studies identified lhx9 and itga4 as genes upregulated upon Tcf21 loss. The role of Lhx9 in cardiac development to date has been limited to studies on the effect of its persistent expression upon FOG2 ablation (Smagulova et al., 2008). Although these studies provide insight into the effects of potentially prolonged Lhx9 activity, the requirement for Lhx9 has not been addressed. Consistently with our finding with Tcf21, we observe here that depletion of Lhx9 results in decreased itga4 expression, with a sparse and disorganized epicardial layer on the heart. Given their similar patterns of expression in the developing epicardium, this result suggests therefore that Tcf21 and Lhx9 function coordinately to enable the correct clustering and migration of PE cells onto the heart surface through integrin-directed mechanisms. It is of interest to note that Tcf21 and LIM homeodomain transcription factors are both found to be essential for the development of the gonad and craniofacial structures (Moncaut et al., 2012; Harel et al., 2012; Cui et al., 2004; Birk et al., 2000), suggesting that the transcriptional targets of these two genes and the pathways they regulate might unveil a conserved mechanism of regulating cell motility and thus, would have implications in cancer, as well as cardiac and organ development (Vladimirova et al., 2009; Yang et al., 2015; Miller et al., 2014, 2013; Weiss et al., 2013; Zhang et al., 2012; Ye et al., 2012; Richards et al., 2011; Arab et al., 2011).

Whole-mount in situ hybridization (ISH) was carried out as previously described (Tandon et al., 2013; Harland, 1991; Charpentier et al., 2015). The pericardial cavity membrane in late tadpole stage embryos was removed post-fixation to improve resolution. Embryos were processed for gelatin (30 µm) sectioning, as previously reported (Gessert and Kühl, 2009; Tandon et al., 2013; Charpentier et al., 2015). All probes were previously used (Tandon et al., 2013) or generated by PCR cloning from Xenopus embryonic cDNA (Table S1). All phenotypes were quantified and statistically assessed using Fisher's exact or Chi-square tests (GraphPad Prism 6). Figures show representative phenotypes quantified. For spatiotemporal analysis of lhx9, lhx2, tbx18 and itga4 transcripts in whole wild-type embryos (Figs S2-S6), multiple focal plane images were compiled using the auto-blend function (Photoshop CS4, Adobe).

Xenopus embryos were cultured, microinjected and staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1967; Tandon et al., 2013). Xla.Tg(Cardiac-actin:GFP)^Mohun transgenic frogs were used as previously reported (Latinkić et al., 2002; Tandon et al., 2013). The 5′UTR from both short and long X. laevis lhx9 alloalleles was cloned and sequenced to verify design of the translation-blocking morpholino (MOT). A 5-base mismatch MO was used as a control (Table S2). Two transcription-blocking morpholinos were designed against the splice donor site of exon 1 from both lhx9 alloallele genomic loci (MO1, Table S2; Tandon et al., 2012) (GeneTools). Each MO (30 ng) was injected at the one-cell stage, with both loci-targeting MO1 being co-injected (Tandon et al., 2013, 2012). To assess the specificity and efficiency of MOT, 1 ng lhx9-5′UTR-GFP RNA (designed against each genomic loci) was co-injected with the stated concentrations of MO. Stage 11 embryos were collected, lysed and analyzed using western blots, as previously described (Tandon et al., 2012, 2013) using GFP antibody (Clontech, JL8, 632381; 1:10,000) and Shp2 antibody (BD Transduction, 610622; 1:2500) as a loading control (Fig. S7). To assess the efficiency of MO1, embryos were injected at the one-cell stage, and the hearts from 25 embryos collected at stage 42 from each condition. RT-PCR was performed (Superscript II, Invitrogen) for lhx9 with gapdh as a loading control (Table S1; Fig. S7). For further details on MO strategies, see supplementary Materials and Methods.

Embryos were processed for agarose vibratome sectioning (100 µm) as reported (Wallingford, 2010; Tandon et al., 2013; Charpentier et al., 2015). Sections were then washed with PBS+1% Triton X-100, blocked with wash buffer+10% fetal calf serum and incubated with primary antibody as previously reported (Langdon et al., 2012; Tandon et al., 2013; Charpentier et al., 2015). Antibodies included mouse anti-tropomyosin (CH1, DSHB; 1:25), rabbit anti-laminin (Sigma, L9393; 1:100), mouse anti-β-dystroglycan (MANDAG2, DSHB, 7D11; 1:100), rabbit anti-fibronectin (Sigma, F3648; 1:250), rabbit anti-phospho-paxillin pTyr118 (Invitrogen, 44-722G; 1:100) and mouse anti-integrin β1 (8C8, DHSB; 1:100). CH1 was deposited to the DSHB by J. J.-C. Lin, MANDAG2 (7D11) by G. E. Morris and 8C8 by P. Hausen and V. Gawantka. Sections were then incubated in DAPI/PBS (200 ng/ml) and imaged using a Zeiss LSM700 spectral confocal laser scanning microscope, with representative figures compiled using ImageJ (NIH) and Photoshop (Adobe). Image fluorescence, gauged by integrated pixel density, was measured using ImageJ and quantified with standard two-tailed Student's t-test (GraphPad).

Xenopus wild-type embryos were incubated in the appropriate concentration of small molecule or DMSO control in 0.1× Modified Barth's Saline (MBS) from stage 38 to stage 41 at room temperature in the dark. Blebbistatin (B0560, Sigma) was used at 10 µM (Straight et al., 2003), 6-B345TTQ (B7438, Sigma) was used at 5 and 10 µM (Kummer and Ginsberg, 2006; Kummer et al., 2010; Hung et al., 2013) and NSC23766 trihydrochloride (SML0952, Sigma) was used at 5 and 50 µM. Embryos were briefly rinsed in 0.1× MBS, fixed in 4% paraformaldehyde and subjected to in situ hybridization as described.

We would like to thank the faculty of the Microscope Services Laboratory at UNC for help with microscopy imaging. The antibodies against tropomyosin, Integrin β1 and β-dystroglycan were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA.