Olfactomedin-1 activity identifies a cell invasion checkpoint during epithelial-mesenchymal transition in the chick embryonic heart

Epithelial mesenchymal transition (EMT) is a cellular process that mediates transition from two-dimensional cellular sheets to a three-dimensional structure (Hay, 1995). EMT is widely utilized by the embryo for gastrulation, neural crest cell formation, heart valve formation and other developmental processes (Hay, 1995; Yang and Weinberg, 2008). In the adult, EMTs occur in the pathologies of both metastasis and organ fibrosis (Thiery et al., 2009). EMT is initiated through a variety of signals including growth factors, inflammatory signals and extracellular proteases (Nieto, 2011; Thiery et al., 2009). A common view is that induction or stabilization of one or more transcriptional repressors such as SNAI1 or ZEB1 leads to loss of cell-cell adhesion and initiates a cascade of processes leading to EMT (Nieto, 2011). The details of this cascade are still being explored but its interruption could lead to epithelial plasticity or a metastable state rather than a complete transition (Klymkowsky and Savagner, 2009). In avian heart valve formation, EMT is mediated by activities of transforming growth factor (TGF) β2 and TGFβ3, but sequential regulation has not been identified in any other EMT (Boyer et al., 1999). We identify here a novel regulator of cell invasion during EMT in the avian heart, Olfactomedin-1 (OLFM1). These findings extend the observation that EMT is mediated by sequential signals and that the ability of a cell to invade a matrix is a regulated element of EMT.

OLFM1 (variant 1) (also known as Noelin-1 and Pancortin-1) is an extracellular matrix protein expressed in the developing heart (reported here), the central nervous system and in a variety of tumor tissues (Kim et al., 2010; Moreno and Bronner-Fraser, 2005; Sun et al., 2006). It is a member of the olfactomedin family of proteins and contains a conserved 250 amino acid olfactomedin domain. Olfactomedin domains are binding sites for protein-protein interaction between olfactomedin family members such as Myocilin and Olfactomedin-3 (Tomarev and Nakaya, 2009). There are four numbered variants of OLFM 1, with variant 1 being the longest isoform (Barembaum et al., 2000). Though called Noelin-1 in the avian species, we use OLFM1 here as the conserved gene name. Avian OLFM1 is conserved across species with 92% amino acid sequence similarity with Xenopus and 85% similarity to murine OLFM1. Mouse and chick transcripts have numbered isoforms, 1–4, generated through alternative splicing. The nomenclature in the human is different as OLFM1 is a homolog of Noelin-1 and Pancortin-1 but human Olfactomedins 2–4 are unique gene products. Alternately spliced forms of OLFM1 in humans are referred to as variants 1–4.

OLFM1 was found to induce additional neural crest cells in the embryo (Barembaum et al., 2000), a process that includes EMT as well as cell differentiation and migration. Because its role in neural crest formation was undefined, we examined OLFM1 during formation of valves in the embryonic chick heart. EMT in the heart is a well-described process based on in vitro analysis by a collagen gel assay (Boyer et al., 1999; Brown et al., 1996; Potts and Runyan, 1989; Tavares et al., 2006). The EMT assay recapitulates the formation of AV valve progenitors from endothelial precursors (Boyer et al., 1999; Markwald et al., 1977; Potts and Runyan, 1989; Yang and Weinberg, 2008).

Immunostaining confirmed synthesis of OLFM1 protein by the myocardium. OLFM1 is an extracellular matrix molecule and might be a component of the tissue interaction in the heart (Barembaum et al., 2000; Moreno and Bronner-Fraser, 2002). Mesenchymal cells are derived from cardiac endothelium in EMT induced by the myocardium (Markwald et al., 1984). Exogenous OLFM1 increased mesenchymal cell numbers and anti-OLFM1 inhibited mesenchyme formation from cardiac explants in the collagen gel assay. Anti-OLFM1 had no effect on endothelial cell separation that identifies the initiation of EMT, but inhibited subsequent invasion of activated endothelial cells. Confirmation of the activity of OLFM1 was shown by similar siRNA perturbation of mesenchyme. Addition of OLFM1 to explanted cultures induces expression of OLFM1 mRNA, indicating that OLFM1 is auto-regulatory. To examine the effects of OLFM1 on mammalian cells, Madin-Darby canine kidney (MDCK) epithelial cells were placed in a transwell invasion assay with TGFβ2, OLFM1 or a combination of the two factors. We found TGFβ2 and OLFM1 to be synergistic in mediating cell invasion. The combination of siRNA targeting OLFM1 and anti-TGFβ showed a similar inhibition of EMT in heart explant cultures. These extracellular factors appear to work in concert to promote cell invasion.

These results define OLFM1 as a regulator of invasion in EMT. This suggests that initiation of cell invasion is a discrete checkpoint in the EMT process where OLFM1, TGFβ and, perhaps other signals, mediate the transition from activation to invasion. The existence of an invasion checkpoint is consistent with previous observations with the neural crest and with the concept of epithelial plasticity, where an incomplete EMT is observed (Barembaum et al., 2000; Thiery et al., 2009). Because OLFM1 is conserved across species and is expressed in a number of tumors, its role in invasion during EMT is likely to extend to a widespread role in both normal development and adult pathology.

Preliminary PCR analysis established that OLFM1 mRNA was expressed in the heart (not shown). To identify the source of OLFM1 in the heart, sections from chick embryo at Hamburger-Hamilton stage 15 (Hamburger and Hamilton, 1951) were stained with a commercial antibody for OLFM1 and counterstained with a cardiac myosin marker (MF-20) and a nuclear marker (DAPI). Though the antigen for antibody production was mouse recombinant OLFM1 protein, the antibody identified a predominant band of ∼50 KDa in chicken tissue, similar to that previously reported (Barembaum et al., 2000) (supplementary material Fig. S1).

OLFM1 staining was observed in the myocardium whereas the endothelium was unstained (Fig. 1). Fig. 1 shows the heart in cross-section through the outflow tract and through a portion of the heart that includes the atrioventricular (AV) canal and common ventricle. The most intense staining was observed in the AV canal portion of the myocardium. OLFM1 was not present in the mesenchyme of the head or in adjacent thoracic tissue. At higher magnification (Fig. 1C), staining was more intense on the endocardial side of the myocardium, the portion of the myocardium that is the site of secretion of cardiac extracellular matrix (ECM) (Kitten et al., 1987). As the inductive stimulus for EMT is derived from the myocardium (Kitten et al., 1987; Krug et al., 1985), this distribution is consistent with OLFM1 as a component of the tissue interaction. OLFM1 expression was also seen in mesenchymal cells of the AV canal after EMT (Fig. 1D, arrows).

OLFM1 was examined in an established assay for embryonic EMT (Runyan and Markwald, 1983). An initial dose-response assay surveyed two markers of mesenchymal cell formation, Fibulin2 (FBLN2) and aldehyde dehydrogenase 1a1 (ALDH1A1) with AV explants (supplementary material Fig. S2). Fibulin2 is a well-established marker of EMT in the heart synthesized by newly formed mesenchyme (Wunsch et al., 1994). ALDH1A1 is a stem cell marker (Ginestier et al., 2007) and a useful measure of mesenchymal cell formation in explant cultures (supplementary material Fig. S3). The expression of ALDH1a1 in the stage 14 chick embryo peaks at 36 hours after dissection and declines after differentiation from mesenchymal cells at 48 hours. Stage 15 AV canal explants were treated with 0, 100, 200, 300 and 400 ng/ml of mouse recombinant OLFM1 for 36 hours and the marker expression was assessed by real time RT-PCR. The dose-response curve was bell-shaped, with the largest response seen at 200 ng/ml for both markers (supplementary material Fig. S4). We found that OLFM1 mRNA in heart explants was regulated by exogenous OLFM1 in a pattern that replicated the mesenchymal markers (Fig. 2A). This suggests that OLFM1 is auto-regulatory. In additional experiments, we treated MDCK epithelial cells with OLFM1 to examine its effects on a mammalian epithelium. ALDH1A1 and FBLN2 proved to be poor markers for EMT in MDCK cells but OLFM1 mRNA was upregulated by OLFM1 protein in these cells. The dose response to OLFM1 was measured by production of OLFM1 mRNA after 24 hours of exposure. The data showed a plateau beginning at 200 ng/ml (Fig. 2B). The difference in response might reflect differences between a developing tissue and a cell line but the 200 ng/ml dose appeared to be useful for both primary explants and the MDCK cell line.

To assess the effects of OLFM1 on endothelial cells, stage 15 (Hamburger and Hamilton, 1951) AV canal explants were cultured on a collagen gels with either exogenous OLFM1 (200 ng/ml) or anti-OFLM1 (10 μg/ml) antibody for 36 hours. Cultures were fixed and photographed to examine morphology and to quantify mesenchymal cells. Control cultures displayed activated endothelial cells dispersed on the gel surface (Fig. 3A) whereas mesenchymal cells were observed within the gel matrix using Hoffman Contrast Optics (Fig. 3E). Cultures treated with OLFM1 displayed a similar endothelial outgrowth and a greater number of mesenchymal cells (Fig. 3B,F). Quantification showed that mesenchymal cells (20 explants per treatment) increased 79% in explants treated with OLFM1 (Fig. 3I).

By contrast, affinity-purified anti-OFLM1 antibody inhibited mesenchymal cell formation in AV canal cultures (Fig. 3G). Unlike reagents that block the initial steps of EMT (Boyer et al., 1999; Mercado-Pimentel et al., 2007), antibody treatment showed normal endothelial cell separation on the cell surface (Fig. 3C). There was a 40% decrease in the number of mesenchymal cells, within the gel matrix, in explants treated with anti-OLFM1 antibody (Fig. 3J). Additional stage 15 AV canal explants were treated with nonspecific IgG antibody as a control. The number of invaded cells in the IgG-treated group did not change compared with the untreated control (Fig. 3J).

To confirm functional antibody specificity, exogenous mouse recombinant OLFM1 and anti-OLFM1 were added together to AV explant cultures. Anti-OLFM1 inhibited mesenchymal cell formation in the presence of exogenous OLFM-1 (Fig. 3H). Cell counts of mesenchyme showed a level of inhibition similar to that for antibody alone (Fig. 3K).

Changes in mesenchymal cell numbers within the gel might result from perturbation of EMT or by loss or gain of cells within the gel by cell proliferation or death. For example, the effects of a canonical Wnt in this assay can be attributed entirely to proliferation or apoptosis of mesenchyme produced by Wnt-specific reagents (Person et al., 2005a).

We examined OLFM1 effects on cell proliferation. Histone H3 staining was employed to identify proliferating cells during EMT. Stained endothelial and mesenchymal cells were counted in 20 AV canal explants to produce an index of proliferation. OLFM1 did not change this index of proliferation compared with untreated controls (Fig. 4A).

A decreased number of mesenchymal cells after anti-OLFM1 antibody treatment might be a result of cell death. AV canals were treated with anti-OLFM1 antibody and apoptotic cells in the endothelia or mesenchyme detected using TUNEL assay. An index of apoptotic cells was obtained from 20 AV canal explants with or without antibody treatment. Apoptosis was not significantly different between controls and treated cultures (Fig. 4B), and there was no evidence of necrotic cell debris in the cultures.

An increase in mesenchymal cell numbers might be observed if mesenchymal cells migrate at a faster rate and enter the gel matrix sooner than controls. Average rates of cell migration within the gel matrix were measured. AV explants were cultured for 18 hours to produce mesenchyme within the gel. Cultures were then treated with OLFM1 (200 ng/ml) or left untreated. Beginning 4 hours after treatment (a time interval selected for early detectable changes in marker mRNA expression), time-lapse images were captured every 3 minutes using an inverted microscope with an incubator stage. Twenty individual cells were tracked for 3 hours in treated and control cultures. Image analysis (Simple PCI software) was used to produce track lengths for each cell. No significant difference in average track length between control and OLFM1-treated cells was observed (Fig. 4C).

Because no differences in proliferation, cell death or migration are seen with OLFM1 or anti-OLFM1 treatment, we conclude that OLFM1 is a regulator of EMT. Because there was no alteration in the initial separation (activation) of endothelium, OLFM1 is principally a regulator of cell invasion from the surface of the gel into the gel matrix.

To examine whether effects of OLFM1 are specific to chick embryo endothelial cells, we tested a mammalian epithelial cell line. MDCK epithelial cells were cultured on Matrigel polymerized within transwell inserts (8 μm pore size filter) for 24 hours with OLFM1, TGFβ2 or both factors in combination (Fig. 5). The lower surface of the filter was stained and cells that transited the gel matrix and filter were counted as epithelial cells that had undergone EMT and invaded. Cells exposed to either OLFM1 or TGFβ2 showed a small but insignificant increase in the number of invaded cells compared with untreated controls. However, there was a synergistic effect between the factors with an approximate 6-fold increase in cell invasion in the presence of both factors. A combination of OLFM1, TGFβ2 and inhibitory OLFM1 antibody showed a number of invaded cells similar to TGFβ2 or control treatment alone (Fig. 5).

There are a number of OLFM family members that share related olfactomedin domains. Although the antibody recognizes a major protein by western blot, it might be recognizing a related family member. Similarly, the OLFM1 protein might mimic a related ligand. Examination of in situ images in the GEISHA database (http://geisha.arizona.edu/geisha/search.jsp?entrez_gene=449625) shows only faint localization of OLFM1 in the chick heart. Of the seven OLFM-related molecules found in the chick, only OLFML2B shows a strong cardiac localization by in situ hybridization (http://geisha.arizona.edu/geisha/search.jsp?db_key_value=30973). We examined the expression of OLFM-related molecules in the embryonic heart by PCR. OLFM1 and OLFM3 were present, although less abundant than OLFML2B (supplementary material Fig. S5). The temporal expression of OLFM1 in embryonic AV canals showed some change but was highest at stage 14 and again at stage 17, when cell invasion begins (supplementary material Fig. S6).

To examine the specificity of OLFM1 function in the chick heart, siRNA specific for OLFM1 (OLFM1 siRNA) was incubated with stage 15 chicken AV canal explants. Explants were cultured on collagen gels to allow mesenchymal cell invasion. There was a significant decrease in mesenchymal cells (25%, P≤0.05) in samples treated with OLFM1 siRNA compared with controls treated with scrambled siRNA (Fig. 6), despite the probable presence of OLFM1 protein in the explants at the time of collection. We extended this experiment to examine whether OLFM1 and TGFβ act in concert in the heart as they do in MDCK cells (Fig. 5). OLFM1 siRNA-treated cultures were incubated with a minimal concentration of pan-TGFβ blocking antibody (1 μg/ml). TGFβ antibody, at this concentration, showed inhibition of mesenchymal cells similar to that of OLFM1 siRNA alone (28%). The combination of OLFM1 siRNA and anti-TGFβ produced an additive inhibitory response consistent with the previous observation in the cell line (Fig. 6).

EMT is an active process, characterized by transcriptional regulation of transcription factors, adhesion molecules and secreted mesenchymal cell products. To characterize the effects of OLFM1 on EMT, AV canals were treated with either OLFM1 or anti-OLFM1 antibody for 36 hours. RNA was then extracted and used to define marker gene expression changes using real time RT-PCR.

During EMT, adhesion molecule transcripts are reduced concomitant with cell separation (Nieto, 2011; Nieto et al., 1994; Peinado et al., 2007). We examined several cell-cell adhesion molecules expressed in the heart. As shown in Fig. 7A, neural cell adhesion molecule 1 (NCAM1) was unchanged with exogenous OLFM1 treatment whereas platelet endothelial adhesion molecule 1 (PECAM1) and cadherin-5 (CDH5) were upregulated. Conversely, all three markers were reduced by blocking antibody. Thus, regulation by OLFM1 is inconsistent with a role in initiation of EMT. Data was graphed on a log scale to show both up- and downregulation by the contrasting treatments. The x-axis is set at 1 and data are normalized to untreated controls. Supplementary material Table S1 shows the numerical values and statistical significance of the real time RT-PCR measurements shown in Fig. 7.

EMT is characterized by the upregulation of several characteristic transcription factors. Although specific roles are not completely understood, the zinc finger transcription factors, Snail1 or Snail2 are thought to initiate EMT through inhibition of cell-cell adhesion transcripts (Comijn et al., 2001; Lincoln et al., 2007; Romano and Runyan, 1999). SNAI2 upregulation is required for avian valvular EMT (Romano and Runyan, 1999). Exogenous OLFM1 increased expression of SNAI1, ZEB1, ZEB2 and SOX9 but reduced expression of SNAI2 (Fig. 7B). Surprisingly, anti-OLFM1 also increased SNAI1 and decreased SNAI2 expression. This pattern of expression, where Snail1 and Snail2 are co-regulated by both ligand and blocking antibody, is inconsistent with early regulation of EMT but might reflect downstream regulatory interactions. OLFM1 and anti-OLFM1 produced contrasting expression of Zeb1 and Zeb2 transcripts with a strong upregulation of Zeb2 by OLFM1. This response is consistent with the Zeb2-null mouse where neural crest cells, expressing Sox10, are induced but unable to migrate from the epithelium (Van de Putte et al., 2003). Effects of OLFM1 and its blocking antibody on Twist1 are also inconsistent with direct regulation of this molecule. Together, the regulation of these EMT transcription factors, as with the cell adhesion molecules above, is inconsistent with initiation of EMT.

EMT is accompanied by reorganization of the cytoskeleton, secretion of extracellular proteins and changes in membrane receptors (Person et al., 2005b). Although TGF-β isoforms are regulators of heart valve EMT (Boyer et al., 1999), TGFβ2 and TGFβ3 are also expressed by migratory mesenchyme (Barnett and Desgrosellier, 2003; Boyer et al., 1999). OLFM1 produced an increase in expression of TGFB3 compared with untreated controls. Anti-OLFM1 decreased expression of both TGFB2 and TGFB3 (Fig. 7C). Fibulin-2 (FBLN2), and Periostin (POSTN), are markers of cardiac mesenchymal cells and showed both an increase in expression with exogenous OLFM1 and a corresponding decrease with anti-OLFM1. These markers reflect either a change in mesenchymal cell numbers or a change in mesenchymal cell differentiation.

VEGF receptors shift expression patterns concomitant with EMT (Stankunas et al., 2010). Vascular endothelial growth factor receptor-2 (VEGFR2 or FLK1) was strongly induced with exogenous OLFM1 whereas VEGFR1 (or FLT1) decreased. Similarly, ALDH1A1 expression was increased with OLFM1 addition (Fig. 7C). However, these markers were unchanged by anti-OLFM1. As the loss of mesenchyme with anti-OLFM1 (Fig. 2J) does not change baseline expression, it is likely that OLFM1 acts on activated endothelial cells (gel surface) to increase these markers as well as mesenchymal cell invasion.

Markers of both cytoskeletal reorganization [RhoA, RhoB, cofilin2 (CFL2) and profilin2 (PFN2)] and proteolytic degradation of the extracellular matrix (PLAU, MMP2, and ADAMTS1) were examined (Fig. 7D). Small changes in PFN2 were potentially consistent with shape change but expression of CFL2, RHOA and RHOB were unaffected by OLFM1 and increased with anti-OLFM1. Importantly, all three extracellular proteases tested, PLAU, MMP2 and ADAMTS1, were strongly upregulated by OLFM1, consistent with their roles in cell invasion.

The observation that OLFM1 mediates cell invasion during EMT supports the concept that there is a point during EMT in the heart when additional signals are required to enable cell invasion. In the absence of this stimulus, activated endothelial cells revert to luminal endothelia. We have previously shown that EMT in the avian heart is mediated by the sequential activities of TGFβ2 and TGFβ3 and that TGFβ3 mediates invasion (Boyer et al., 1999). However, to our knowledge, this isoform switch has not been observed in any other system. The murine heart might utilize a combination of TGFβ2 and bone morphogenetic protein (BMP) but, EMT in this system happens so rapidly in vitro that resolution of specific roles is difficult (Camenisch et al., 2002). The concept of OLFM1-driven invasion is consistent with observations on neural crest cells. Overexpression of OLFM1 produced additional neural crest cells after the first neural crest cells had left the neural tube (Barembaum et al., 2000).

Loss of either TGFβ3 (Boyer et al., 1999) or OLFM1 (this paper) leaves activated, dispersed and fusiform endothelia with limited ability to enter a three-dimensional matrix. The tumor-promoting phorbol ester, PMA, produced similar activated endothelia unable to invade the gel matrix (Runyan et al., 1990). Thus, acquisition of a fusiform or dispersed phenotype is, by itself, not sufficient for cell invasion. These data suggest that there is a checkpoint when additional extracellular signals are required to enable activated cells to invade and complete EMT. Because OLFM1 is secreted by the myocardium, it can be viewed as a component of the inductive stimulus for EMT by the endocardium. In this light, the point at which TGFβ3 and/or OLFM1 act in the heart during EMT is a checkpoint. In the absence of either signal, the ability to invade the underlying matrix is lost or reduced.

In the present study, we employed OLFM1 protein and its corresponding antibody as contrasting treatments and controls. As shown in Fig. 3, the affinity-purified antibody specifically blocks both endogenous and exogenous OLFM1. Where the antibody and protein have contrasting effects on morphology and gene expression, we are confident of a role for OLFM1 as a regulator. In the case of three transcription EMT factors, Snai1, Snai2 and Sox9, OLFM1 protein and antibody have similar, not contrasting, effects (Fig. 7B). Although we are confident that these data are correct because they are based on three independent biological repeats, the basis for these results is unclear. We argue that this result is inconsistent with a role of OLFM1 in the initiation of EMT because Snai2 is upregulated during EMT in the chick heart (Romano and Runyan, 1999) and is downregulated here.

Cell adhesion molecules that should be downregulated as part of the initiation of EMT are upregulated by OLFM1 treatment (Fig. 7A). We speculate that these data might reflect a reversion of activated cells to the endothelial phenotype. In the heart, the entire AV endocardium is activated but only about 10% of endothelial cells complete EMT (Runyan et al., 1990). The eventual return of most endothelial cells to a quiescent state has the developmental advantage of retaining luminal endothelia in the heart. This is consistent with a checkpoint when additional EMT stimulatory and inhibitory signals are received to regulate invasion.

Cancer cells with both epithelial and mesenchymal characteristics have led to the concept of epithelial plasticity or a metastable state (Derynck and Akhurst, 2007). This intermediate state during metastasis might reflect cells that are unable to complete EMT or delayed at a checkpoint in the absence of additional stimuli. Upregulation of invasion markers (including proteases, ECM molecules and Zeb2) by OLFM1 is consistent with a checkpoint to regulate the final cell invasion. Other factors known to regulate proteases or matrix adhesion molecules during EMT might similarly act at invasion rather than at initiation of EMT (Kessenbrock et al., 2010; Yeh et al., 2006). Because we have already seen that TGFβ3 is a component of cell invasion in the chick and that TGFβ1 is recognized as a mediator of EMT (Derynck and Akhurst, 2007), we suggest that OLFM1 and TGFβ act at an invasion checkpoint during EMT.

Addition of exogenous OLFM1 increases the number of mesenchymal cells from cardiac explants. Conversely, an inhibitory antibody reduces mesenchymal cell numbers. Changes in mesenchymal cell numbers might occur through altered EMT, cell proliferation, cell death or rate of migration (Person et al., 2005b). Measures of phosphorylated histone H3 expression, TUNEL-positive cells and cell migration suggest that these alternative mechanisms are unlikely. Together, the data argue that OLFM1 regulates mesenchymal cell numbers by altering EMT. The morphology of activated endothelial cells after anti-OLFM1 treatment suggests that OLFM1 acts late in EMT. Observations of marker gene regulation reinforce this idea. An early element of EMT is a loss of cell-cell adhesion (Nieto, 2011). OLFM1 does not induce a loss of adhesion molecule transcription. The effect of anti-OLFM1 on CDH5 and PECAM1 transcripts might reflect a re-epithelialization of activated endothelium rather than invasion.

Mesenchymal cells in the heart express both FBLN2 and POSTN (Kruzynska-Frejtag et al., 2001; Rongish et al., 1998). Exogenous OLFM1 upregulates mRNA for both markers whereas the antibody inhibits their expression. Similarly, the markers FLT1, FLK1 and ALDH1A1 changed in a manner consistent with mesenchymal cell differentiation or invasion. Unlike the secreted molecules POSTN and FBLN2, the blocking antibody did not affect FLK1 and ALDH1A1. We suggest that loss of expression of secreted markers is specific to mesenchymal cells, whereas markers that remain unaffected by anti-OLFM1 are first expressed in the activated endothelial population and are therefore are unchanged by loss of OLFM1.

TGFβ isoforms are expressed in activated endothelia and newly formed mesenchymal cells as well as being expressed by the myocardium as an inductive signal (Barnett et al., 1994; Boyer et al., 1999; Potts and Runyan, 1989). Altered expression of TGFβ isoforms by both OLFM1 and anti-OLFM1 is consistent with formation of invasive mesenchyme. Expression of TGFβ3 by activated endothelia is auto-regulatory and provides an amplification of EMT (Ramsdell and Markwald, 1997). We found that OLFM1 is similarly auto-regulatory in cardiac explants and in MDCK cells. Auto-regulation of OLFM1 by OLFM1 was also observed in a mouse cardiac mesenchymal cell line and in human glioblastoma cells (R.B.R., unpublished data). Auto-regulation by both TGFβ and OLFM1 might be a useful mechanism to ensure that EMT proceeds to completion. Despite the activities of these two ligands, the majority of endothelial cells do not complete invasion, but revert to differentiated endothelia. This suggests that there are either strong inhibitory signals that prevent completion of EMT or that only a subset of activated endothelial cells receive adequate signals to complete invasion. Our dose-response curve showed that higher levels of OLFM1 inhibited mesenchymal marker expression in the heart. This suggests that OLFM1 might be both stimulatory and inhibitory of EMT. Strikingly, higher levels of OLFM1 had no such inhibitory effect on MDCK cells. The basis for this difference in not understood but might be exploited in the future to understand signal transduction in response to OLFM1.

The observation that MDCK epithelial cells respond to OLFM1, and more strongly to the combination of OLFM1 and TGFβ2, suggests that OLFM1 activity in relation to cell invasion is not unique to the chick. Because TGFβ isoforms are endogenous in heart tissue, MDCK culture provides an opportunity to explore OLFM1 activity with or without TGFβ. OLFM1 has been knocked out in the mouse (Cheng et al., 2007) but the heart phenotype has not been examined. As there are four OLFM genes and another four OLFM-related genes, there is potential redundancy that would obscure the role of OLFM1 in developmental EMTs (Tomarev and Nakaya, 2009). The receptor for OLFM1 has not yet been identified. OLFM4 is a related but distinct gene product associated with metastasis (Grover et al., 2010). Its olfactomedin domain is thought to bind to cells via N-acetylglucosamine residues and cell surface lectins (Grover et al., 2010). Whether the two different OLFMs act similarly is unclear, but EMT could be related to both proteins and they might have similar receptor mechanisms.

In summary, these results show that OLFM1 is a newly recognized mediator of invasion during EMT in the heart. It is coexpressed with TGFβ3, another mediator of cell invasion (Boyer et al., 1999), and appears to regulate a subset of EMT markers, especially proteases and secreted products. Morphology of the endothelia and the expression of EMT markers point directly to a role in the conversion of activated endothelia to invasive mesenchyme. We show that this conversion of the activated epithelial cells into invasive mesenchyme is a specific regulated step that can be seen as a checkpoint in EMT progression. In the absence of the signaling required to pass the checkpoint, cells undergoing EMT either revert to an epithelial phenotype or remain in a poorly invasive metastable state (Klymkowsky and Savagner, 2009). In development, both neural crest and valve progenitors appear to use TGFβ and OLFM1 (Barembaum et al., 2000; Boyer et al., 1999; Duband et al., 1995). Because OLFM1 is a conserved molecule across species, the concept of an invasion checkpoint has implications in adult pathology. Cancer metastasis and organ fibrosis both utilize EMT programs to invade surrounding tissue (Kallergi et al., 2011; Wang et al., 2011; Zhang et al., 2011). Demonstration that MDCK cells respond to OLFM1 in an invasion assay suggests that this checkpoint function is not restricted to the avian embryo. The data demonstrate a synergy between TGFβ and OLFM1 seen in both chick embryo tissue and MDCK cells. We expect OLFM1 to have a similar role in other EMTs. A survey of the Oncomine and GEO databases shows that OLFM1 is upregulated in a number of cancers (Kim et al., 2010; Sun et al., 2006). OLFM1 expression might have a direct relationship to the aggressiveness of specific tumors. Thus, the role of this molecule in human EMT pathology might be as important as in development.

Fertilized White Leghorn chicken eggs were obtained from MacIntire Eggs (San Diego, CA). Eggs were incubated at 37°C for 48–54 hours to obtain embryos at Hamburger and Hamilton stage 15. Using a dissecting stereo microscope, embryos were dissected free of embryonic membranes in 4°C sterile 1× Tyrode’s salt buffer. Embryos were then rinsed in fresh sterile Tyrode’s buffer to dissect out the AV canal segment of the heart using #5 forceps. A video showing the dissection and subsequent assay is available from the corresponding author upon request.

Collagen gels were prepared as previously described (Boyer et al., 1999; Potts et al., 1992). OLFM1 (#4636-NL) and anti-OLFM1 (#AF4636) were obtained from R&D Systems (Minneapolis, MN). AV canal explants were treated with OLFM1 (100–400 ng/μl) or antibody (10 μg/ml) to determine the effects on EMT and molecular regulators of EMT. After treatment, AV canal explants were incubated on the collagen gel for 36 hours. One set of explants was collected and fixed in 4% paraformaldehyde (PFA) for 30 minutes. Cell counts and micrographs were taken to identify morphological differences between treated and untreated cultures. Explants were observed on an inverted microscope equipped with Hoffman Modulation Optics to collect mesenchymal cell counts, or were prepared for cell proliferation assays. Mesenchymal cells were blindly counted by different laboratory members to prevent any bias. Additional explants were extracted to collect mRNA for real time RT-PCR. Some explants were prepared for TUNEL assay or phosphorylated H3 histone staining (#H0412, Sigma-Aldrich, St Louis, MO).

Stage 15 AV canal explants were placed on collagen gels and incubated overnight with 5% CO₂ at 37°C (18 hours). Explants were then treated with OLFM1 (200 ng/ml) for 4 hours. The 4-hour treatment was selected because auto-regulatory expression of OLFM1 can be seen at this time. The interval is consistent with migration assays investigating the effects of HGF on cell migration (Takeuchi et al., 1996). After a 4-hour exposure, explant images were obtained for 3 hours using time-lapse capture on an Olympus (IMT-2) microscope equipped with Hoffman Modulation Optics. Time-lapse images were captured every 3 minutes. To measure mesenchymal cell migration tracks, a centroid was assigned to each cell image and Simple PCI software was used to accumulate the movement over 3 hours. A total of 20 cells were measured for each control and OLFM1 treatment.

Silencer Select Custom Designed siRNA for OLFM1 was obtained from Ambion (#s444361). The siRNA was suspended and diluted in nuclease-free water to obtain a working solution of 500 nM. AV canals were explanted as described above and placed in 500 μl of 1× M199 medium (working solution). The working solution contained a final siRNA concentration of 5 nM. To facilitate to entry of siRNA oligos into the cell, 5 μl of siPORT NeoFX (Ambion) transfecting agent was added to the working solution. The AV canals were incubated in siRNA working solution at 37°C and 5% CO₂ for 45 minutes. These conditions were used as they had been optimized in previous published work (Mercado-Pimentel et al., 2007). After incubation, AV canals were placed on collagen gels to allow EMT and mesenchymal cell migration. The number of invaded cells were counted and compared with results using a scrambled control. The scrambled control used was a 21-oligonucleotide template that did not blast to any gene in the chicken genome, 5′-AGACTGTCGCGTGCTCTGTCC-3′. The siRNA target sequence for the OLFM1 was 5′-GCAUAG-ACCUGACAGAUCATT-3′, which targets the olfactomedin domain of the molecule.

Embryos were collected and rinsed in Tyrode’s solution. Embryos were cryofixed by protocol (Kitten et al., 1987). Briefly, embryos were flash-frozen by dropping them into n-butanol in vials suspended in liquid nitrogen. The n-butanol was kept above freezing by inserting a metal rod at intervals. Frozen embryos were removed with forceps (at liquid nitrogen temperature) and placed in a scintillation vial containing Molecular Sieve beads and frozen 100% ethanol with an overlying layer of liquid nitrogen. The vials were capped loosely and placed in a −80°C freezer for 3–5 days. Subsequently, the embryos were rinsed in 100% ethanol at −20°C and shifted into 100% xylene and embedded in paraffin. Paraffin blocks were sectioned and the sections affixed to glass slides. Slides were de-paraffinized in xylene, hydrated in an ethanol series from 100% ethanol to 100% phosphate buffered saline (PBS). Sections were rinsed in PBS for 10 minutes and blocked for 1 hour at room temperature. Blocking solution contained 1% BSA and 0.1% Tween-20 suspended in PBS. Sections containing AV canal and heart tissue were incubated overnight with primary antibody at 4°C in a humid chamber. Slides were rinsed three times with 1× PBS; Cy5-labeled rabbit anti-sheep IgG was used as a secondary antibody (#072-02-23-06, KPL, Gaithersburg, MD) in a humid chamber for 2 hours at room temperature. The slides were then exposed to DAPI (#D8417, Sigma-Aldrich, St Louis, MO) as a counterstain for nuclei. Vectashield was used as a mounting media (#H-1000, Vector Laboratories, Burlingame, CA) to prevent loss of fluorescent signal. Samples were then examined using a Delta-Vision-Olympus IX70 microscope.

Damaged and fragmented DNA was detected by TdT (terminal deoxynucleotidyl transferase)-mediated dUTP nick-end labeling (TUNEL) assay using the DeadEnd Fluorometric TUNEL System kit (#G3250, Promega, Madison, WI). Explants on collagen gels were treated with 10 μg/ml anti-OLFM1 antibody for 36 hours. The explants were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Tissues were labeled by incubation in 50 μl TUNEL reaction mixture containing the mixture of terminal deoxynucleotidyl transferase and dUTP. The labeling of AV explants was done in the dark at 37°C for 1 hour. Cellular nuclei were stained using DAPI (1:10,000; Sigma) fluorescent stain. Control and treated AV canal explants (n=20), including endothelial outgrowth, were randomly selected for analysis. Approximately 100 cells were counted from each explant. Analysis was performed using a fluorescence microscope and cells were categorized as either apoptotic (green nuclear fluorescence) or normal (no fluorescence). Samples were analyzed using a Delta-Vision Olympus IX 70 microscope.

Proliferating cells were detected by immunostaining for histone H3 phosphorylated at Ser10. The collagen gels with cultured AV explants were treated with OLFM1 (200 ng/ml) for 36 hours. The AV explants were then fixed in 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Polyclonal rabbit antibody for phosphorylated histone H3 was used as the primary antibody at a 1:300 dilution (#H0412, Sigma-Aldrich). After washing with 1× PBS, Alexa Fluor 546 goat anti-rabbit IgG at 1:400 dilution was used as the secondary antibody (Molecular Probes, CA). 1× PBS was used to wash excess secondary antibody and DAPI (1:10,000; Sigma) was used to counterstain cell nuclei. Control and treated AV canal explants (n=20), including epithelial outgrow, were randomly selected for analysis and ∼100 cells were counted from each explant. Analysis was performed using a fluorescence microscope and cells were categorized as either proliferating (red nuclear fluorescence) or normal (no red fluorescence). Samples were then analyzed using a Delta-Vision Olympus IX 70 microscope.

Heads and hearts from stage 17 chicken embryos were collected to obtain a whole-cell lysate using NP-40 buffer (20 mM Tris-HCL, 137 mM NaCl, 10% glycerol, 1% Nonidet P-40, 2 mM EDTA) at 4°C. The lysis NP-40 buffer was supplemented with a Proteinase Inhibitor Cocktail Tablets (Roche, IN) every time before use. Cell lysates were sonicated for 20 seconds and the protein concentration determined using Pierce 660 nm Protein Assay (Thermo Scientific, IL). Equal amounts of head and heart (28 μg) samples were loaded onto a 10% acrylamide gel. Protein was then transferred to nitrocellulose membrane using the Trans-Blot Turbo System (Bio-Rad, CA). OLFM1 protein was detected using anti-mouse Noelin-1 antibody (R&D Systems, MN) as a primary antibody at 1:400 dilution (0.25 μg/ml) overnight at 4°C. The corresponding secondary antibody was used at 1:10,000 dilution at room temperature for 2 hours. Bands were then visualized using the Li-Cor Odyssey infrared imaging system.

Madin-Darby canine kidney (MDCK) epithelial cells (50,000/well) were plated onto transwell culture dishes (8 μm pore size; BD Falcon, NJ) containing 100 μl of solidified Matrigel Matrix (Bioscience, MA). Cells were treated with 200 ng/ml OLFM1 alone, 2 ng/ml of TGFβ2 or a combination of both factors for 24 hours. The membranes were removed from the culture inserts, inverted to display the lower surface and stained with Crystal Violet. Cells that traversed the matrix and membrane were counted and total numbers for each membrane were compared with untreated controls.

Total RNA was extracted from AV canal explants cultured on collagen gel by using TRIZOL reagent (Gibco BRL). RNA was DNase-treated using the TURBO DNA-free kit (Ambion). cDNA was transcribed using the iScript cDNA synthesis kit (Bio-Rad, Richmond, CA). Data normalization against the specific housekeeping genes (GAPDH, LDH, β-actin, and mitochondrial DNA) was problematic due to either the substantial phenotypic change during EMT or to the effects of the treatment. Therefore, total cDNA was used to normalize the quantitative PCR reactions. To sensitively measure cDNA in each reaction, Quanti-iT Oligreen ssDNA reagent (Molecular Probes, CA) was used to measure and aliquot single-stranded cDNA in combination with a fluorometer (Turner Biosystems). Names, UID, numerical data and significance are shown in supplementary material Table S1. Primer sequences for the selected genes are listed in supplementary material Tables S2 and S3.

Adobe Photoshop CS4 Version 11.0.2 software was used to handle and process pictures. The picture processing included changes in brightness, contrast and re-sizing in order to clearly evaluate samples and easily determine differences form control. Picture processing was performed identically on all micrographs in each comparison.

All comparisons employed the Student’s t-test (paired, two-tailed), with P≤0.05 indicating significance compared with a control group. The mesenchymal cell numbers counted in Fig. 3 were for 20 explants in each group. Proliferating and apoptotic cells (Fig. 4A,B) were also counted in 20 explants for each treatment. Track length measurements (Fig. 4C) were performed with tracks of 20 cells for treated and control samples. All real time RT-PCR data are presented as a measure of three, independently collected, pools of 15–20 AV canal explants. Data from each pooled sample were averaged from two or three technical repeats. Error bars in the figures indicate s.e.m. Names, UID, numerical data and significance are shown in supplementary material Table S1.

We would like to thank Carlos Moran-Segura, Sourav Ghosh, Jean Wilson, Paul Krieg, Parker Antin and the Heart Development group at the University of Arizona for useful discussions. We thank Jade Rusche and Jessica Bender for technical support. Thanks also go to Laura Serrano, Michael J. Sullivan and Albert McHenry. We thank Tony Person for first drawing our attention to OLFM1.