Border formation in a Bmp gradient reduced to single dissociated cells

Abstract

Conversions of signaling gradients into sharp “all-or-none” borders are fundamental to tissue and organismal development. However, whether such conversions can be meaningfully reduced to dissociated cells in culture has been uncertain. Here we describe ultrasensitivity, the phenomenon equivalent to an all-or-none response, in dissociated neural precursor cells (NPCs) exposed to bone morphogenetic protein 4 (Bmp4). NPC ultrasensitivity is evident at the population and single-cell levels based on Msx1 induction, a well known Bmp target response, and occurs in the context of gene expression changes consistent with Bmp4 activity as a morphogen. Dissociated NPCs also display immediate early kinetics and irreversibility for Msx1 induction after brief Bmp4 exposure, which are attractive features for initial border formation. Relevance to border formation in vivo is provided by Bmp4 gain-of-function studies in explants and evidence for single-cell ultrasensitivity in normal and mutant Bmp gradient contexts in the developing forebrain. Together, these studies demonstrate relatively simple, robust, and reducible cell-intrinsic properties that contribute to developmental border formation within a signaling gradient.

Reducibility of complex biological phenomena to dissociated cells in culture is a central precept of experimental biology. In neuroscience, examples include the axon potential (1) and, specific to the cells of interest for this study, the timing and production of cell lineages by embryonic cortical neural precursor cells (NPCs) (2). Single-cell reducibility is also established for morphogens (“form-generating” substances), which act to specify different cell fates according to concentration not only in intact tissues, but also on dissociated, non-reaggregated precursor cells (3, 4). However, it is unclear whether the fundamental process of converting graded morphogen activity into sharp borders (5) can be similarly reduced. Some of the strategies used to make tissue borders, such as transcriptional cross-repression (6, 7), may not require cell–cell interactions and are conceptually reducible. However, all-or-none changes that correspond to a border in vivo have not been formally described in dissociated cells exposed to small changes in morphogen concentration.

Our studies focus on the border between embryonic cerebral cortex and the telencephalic dorsal midline (cortex–midline border), which is regulated by Bmp signaling (8, 9) and a Bmp activity gradient (10). This border fails to form properly in a mouse model of human holoprosencephaly involving genetic ablation of the roof plate (10), the CNS dorsal midline source of Bmps (11, 12). Roof plate ablation results in curtailed Bmp production (13), reduced midline Bmp activity, and a flattened Bmp activity gradient in the dorsal telencephalon (10). Associated midline defects include abrogated expression of Msx1 (13), a well known and ancient Bmp target gene (14–16). Although Bmp signaling is required for telencephalic dorsal midline development, it remains uncertain whether Bmps act as formal morphogens in this system—i.e., whether they can induce two or more cell fates at different concentration thresholds.

Here we show that formation of an Msx1 expression border by graded Bmp signals can be reduced to single dissociated cells in vitro. We first use explants to demonstrate Bmp4 and NPC sufficiency to recapitulate thresholded Msx1 expression in the telencephalic dorsal midline. Using dissociated NPCs, we then describe ultrasensitive Msx1 induction with small increments of Bmp4 concentration, which occurs in the context of Bmp4-driven gene expression changes consistent with a morphogen response. These findings provide evidence for cell-intrinsic properties that can contribute to border formation without requiring tissue-level interactions.

Results

Threshold Msx1 Induction in Explants.

Normal Msx1 expression does not follow the Bmp activity gradient in linear fashion; rather, expressions of native Msx1 mRNA, protein, and an engineered nuclear lacZ (Msx1-nlacZ) fusion reporter (17) are confined to the midline at embryonic day 10.5 (E10.5) (Fig. 1B) and are maintained stably thereafter [supporting information (SI) Fig. 5] (18). Msx1 induction is therefore well modeled as a Bmp gradient-to-threshold conversion in vivo (Fig. 1C).

Fig. 1.

Confirmations of graded Bmp activity and NPC threshold conversion in dorsal forebrain explants. (A) Coronal schematics of the E10.5 telencephalon (light blue) and Msx1-expressing dorsal midline (dark blue). (B) Graded Bmp activity (pSmad1/5/8) and three thresholded Msx1 readouts at the E10.5 dorsal midline. Msx2 is more broadly expressed than Msx1 in this region (50), accounting for the larger expression domain detected with the Msx1/2 antibody. Right highlights expression in neural tissue for clarity. (Scale bar: 0.1 mm.) (C) Models of Msx1 threshold conversion with and without exogenous Bmp4 point sources (beads), which predict dose-, distance-, and orientation-dependent ectopic Msx1 induction. The endogenous Bmp gradient profile is based on previous measurements (10). (D) Explant and bead schematic. (E and F) Ectopic Msx1 mRNA and nlacZ inductions (arrowheads) around Bmp4-soaked, but not BSA-soaked, blue Affigel (E) or clear heparin acrylic beads (F) in E10.5 explants cultured for 2 days. Findings confirm predictions described in C and establish Bmp4 and cortical NPC sufficiencies for threshold conversion in explants. [Scale bars: 1 mm (low power) and 0.2 mm (high power).] c, cortex; d, diencephalon; dm, dorsal midline; ee, epidermal ectoderm; l, lateral ganglionic eminence; m, medial ganglionic eminence; me, mesenchyme.

If this model of threshold Msx1 conversion were true, Bmp gain-of-function studies should display three features: (i) concentration dependence (e.g., beads with more Bmp4 should induce more Msx1 when introduced at similar positions within the endogenous gradient), (ii) position dependence (beads with identical Bmp4 should induce more Msx1 when placed closer to the midline Bmp source), and (iii) orientation dependence (Msx1 induction around individual beads should be higher toward the Bmp source). These three features were observed in dorsal forebrain explants (Fig. 1D) using two different bead types and Msx1 readouts (Fig. 1 E and F). Although linear Msx1 activation above threshold could also yield dose- and position-dependent effects, linear activation would be difficult to reconcile with oriented Msx1 induction around individual beads. Taken together with in vivo findings, these explant data therefore support the threshold Msx1 induction model. The apparent absence of Msx1 orientation dependence in previous cortical explant–bead studies (8) suggests that inclusion of the dorsal midline is required to maintain the endogenous Bmp gradient. In addition to confirming graded Bmp activity in the dorsal forebrain, these findings demonstrate the sufficiencies of exogenous Bmp4 and cortical NPCs to cause and effect Msx1 threshold conversion, respectively.

Population Level Ultrasensitivity.

Given these sufficiencies, we formally examined dissociated cortical NPCs for ultrasensitivity, the phenomenon analogous to threshold activation within a concentration gradient (19–21). Ultrasensitivity describes “switch-like” responses that reach maximal levels over a small range of stimulus concentration, resulting in sigmoidal dose–response curves rather than the hyperbolic ones of typical Michaelis–Menten reactions (Fig. 2A). Ultrasensitivity is often coupled with bistability (also known as hysteresis) or irreversibility, which maintain on-states after stimulus reduction or removal, respectively (Fig. 2B). Ultrasensitivity and bistability/irreversibility are general principles in all-or-none cell fate decisions (22–24) and are used to describe border formation within morphogen gradients in vivo and in silico (19–21). Although observed in a few cell culture contexts (25, 26), ultrasensitivity of dissociated cells to a morphogen has not been described. The degree of ultrasensitivity is numerically captured by the Hill number (nH), with nH > 1 indicating ultrasensitivity and a benchmark nH of 2.8–3.0 for the cooperative binding of oxygen to hemoglobin (27).

Fig. 2.

NPC ultrasensitivity and irreversibility in dissociated cultures at the population level. (A and B) Schematic Michaelis–Menten, ultrasensitive, bistable, and irreversible response curves. (C–F) Hill plots of real-time qRT-PCR data from E10.5 Msx1-nlacZ (D), E12.5 Msx1-nlacZ (E), and E12.5 wild-type (F) cortical NPC cultures. Native and nlacZ-containing Msx1 transcript inductions display significant ultrasensitivity in response to Bmp4 (nH = 2.4–3.8), whereas Tgif induction is non-ultrasensitive (nH = 0.3 using data points up to 128 or 256 ng/ml Bmp4). Tgif up-regulations at the two highest Bmp4 doses were significant by t test (P < 0.05). (G) Real-time qRT-PCR data from E10.5 Msx1-nlacZ cells for other midline genes. Unlike endogenous Msx1 transcripts, Ttr is up-regulated at intermediate Bmp4 concentrations, whereas both Ttr and Lmx1a display suppression at higher Bmp4 concentrations associated with maximal Msx1 induction, consistent with a morphogen effect. Ttr up-regulations at 8 and 16 ng/ml Bmp4 were significant by t test (P < 0.05). (H–J) Induction kinetics, wild-type cortical NPCs, 50 ng/ml Bmp4. Msx1 up-regulation is detected by 15 min, is statistically significant by 2 h, and is stable for at least 5 days in the continuous presence of Bmp4. (K and L) Washout studies, wild-type cortical NPCs, 50 ng/ml Bmp4. Maximal Msx1 induction requires only brief Bmp4 exposure (no more than 15 min) and does not require Bmp4 thereafter, thus displaying irreversibility. Error bars show standard errors. t tests: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Using the wide linear dynamic range of real-time quantitative RT-PCR (qRT-PCR), we found that cortical NPCs from E10.5 Msx1-nlacZ embryos exhibited significant ultrasensitivity to Bmp4 (Fig. 2D) (nH = 3.6 and 2.4 for native and nlacZ-containing Msx1 transcripts, respectively). Similar degrees of ultrasensitivity were detected in E12.5 cultures (nH = 3.3–3.8) (Fig. 2 E and F). Specificity to Msx1 was confirmed by the non-ultrasensitive induction of Tgif (nH = 0.3) (Fig. 2F). [Note that Msx1 and Tgif were the only two midline genes, among six examined, that were significantly induced in dissociated cortical NPCs by Bmp4 (10).] Two factors argue against a significant cell selection bias to these data: (i) real-time qRT-PCR data reflect per-cell averages after normalization to cyclophilin A levels, one of the two best internal references among four screened (13); and (ii) potential proapoptotic Msx1 effects would artificially reduce rather than increase nH values.

To examine whether ultrasensitive Msx1 induction occurs as part of a morphogen response, we examined expression of Transthyretin (Ttr) and Lmx1a, which differentially mark two fates in the telencephalon dorsal midline: (i) choroid plaque at the immediate midline (Msx1+/Ttr−/Lmx1a−) and (ii) choroid plexus epithelium bilaterally (Msx1+/Ttr+/Lmx1a+) (Fig. 1) (10, 12, 13). At intermediate Bmp4 concentrations, Ttr was significantly up-regulated in E10.5 NPCs, whereas both Ttr and Lmx1a displayed suppression at higher Bmp4 concentrations (Fig. 2G). Previous E12.5 culture studies did not reveal Lmx1a regulation but did yield similar Msx1 and bimodal Ttr response curves (10), suggesting that Bmp4 can act as a morphogen to promote choroid plexus epithelial and choroid plaque fates at intermediate and high concentrations, respectively.

Time-course and washout studies were then performed to examine kinetics and stability of Msx1 induction. Consistent with an earlier description of Msx1 as an immediate early Bmp response gene (28), Msx1 induction was rapid, with significant up-regulation by 2 h (Fig. 2I). Msx1 levels were also stable in the continuous presence of suprathreshold Bmp4 (Fig. 2J). Strikingly, maximal Msx1 induction required only brief Bmp4 exposure (no more than 15 min), and, unlike the rapid decay of classic immediate early genes, Msx1 induction was stable thereafter (Fig. 2L). Thus, Msx1 induction was not only ultrasensitive, but also displayed irreversibility (Fig. 2B).

Single-Cell Ultrasensitivity.

Cells can be ultrasensitive at the population or single cell level, but not necessarily both (22, 29). With increasing morphogen concentration, single-cell ultrasensitivity should result in more all-or-none expression within individual cells rather than graded differences in expression level (Fig. 3A). To examine this, we studied pSmad and Msx1 expression in cortical NPCs using immunocytochemistry and cytochemistry. Whereas nuclear pSmad increases were graded and relatively homogeneous with increasing Bmp4 concentration, inductions of Msx1/2 (Fig. 3C) and Msx1-nlacZ proteins (SI Fig. 6A) were clearly more all-or-none in character. In both sets of cells, faint staining could also be detected, particularly at intermediate Bmp4 concentrations, consistent with Msx1 induction not being a perfect switch. At suprathreshold Bmp4 concentrations, X-Gal staining within individual cells generally became more intense with increasing Bmp4 concentrations, suggesting an additional graded regulation of Msx1 levels in vitro. This correlates well with Msx1/2 and Msx1-nlacZ staining intensities being highest at the immediate midline in vivo (Fig. 1B).

Fig. 3.

NPC ultrasensitivity in dissociated cultures at the single-cell level. (A) Schematics of graded and ultrasensitive responses at the single-cell level. (B) Culture schematic. (C) Double label immunocytochemistry of E12.5 wild-type cortical NPCs. Msx1/2 induction (red) is ultrasensitive compared with pSmad (green), which is graded and more homogeneous. Similar graded pSmad findings were seen in four independent cultures. (Scale bar: 0.05 mm.) (D) Percentage of X-Gal-positive cells in E12.5 Msx1-nlacZ cortical NPCs (SI Fig. 6). Msx1-nlacZ induction occurs equally well in cells that are completely isolated compared with those in contact with other cells. Errors bars show standard errors from two independent cultures.

To examine single-cell behavior more selectively, we separately scored Msx1-nlacZ positivity in cells (SI Fig. 6A) that were either completely isolated or in contact with other cells. This resulted in two remarkably similar curves (Fig. 3D). Msx1 induction therefore occurs efficiently in isolated cells and is not obviously regulated by cell contact, further highlighting the cell-intrinsic nature of this response. (Note that, although the line plots in Fig. 3D have sigmoidal upslopes, these percent positivity data do not inform ultrasensitivity; e.g., the percent positivity curve for the schematized graded response in Fig. 3A would be highly sigmoidal.)

Evidence for Single-Cell Ultrasensitivity in Intact Tissues.

We then sought evidence in vivo and in explants that would be consistent with NPC-intrinsic ultrasensitivity at the single-cell level—namely, Msx1-positive cells surrounded by negative neighbors. Such evidence was found near the normal Msx1 expression border in vivo (Fig. 4A), in cortical NPCs surrounding Bmp4-coated beads in explants (Fig. 4B), and in the midline region of roof-plate-ablated embryos with reduced midline Bmp activity (Fig. 4C). The roof plate ablation findings were particularly striking, in that the few Msx1-expressing cells in the context of a reduced and flattened Bmp activity gradient (10) expressed high levels that were comparable to wild-type midline cells, whether clustered or widely separated. Although migration or movement of Msx1+ cells could account for some of these observations, this seems inadequate to explain the sparse and widely separated Msx1-expressing cells in roof-plate-ablated mutants. Thus, although other mechanisms cannot be formally excluded, these findings suggest ultrasensitive and cell-intrinsic induction of Msx1 in vivo.

Fig. 4.

Evidence for single-cell ultrasensitivity in intact tissues. (A) Msx1/2-positive NPCs (arrows) surrounded by negative cells near the E10.5 cortex–midline border. [Scale bars: 0.1 mm (low power) and 0.05 mm (high power).] (B) Scattered Msx1-nlacZ-positive NPCs (arrows) amid negative cells surrounding a Bmp4-soaked heparin acrylic bead in E10.5 explants cultured for 2 days. B Right Inset is inverted and thresholded for clarity. Dashed lines, beads; arrowhead, endogenous Msx1-nlacZ-expressing dorsal midline domain. [Scale bars: 0.2 mm (low power) and 0.05 mm (high power).] (C) Scattered Msx1 transcript-positive NPCs (arrowheads) amid negative cells in the E10.5 dorsal midline region after roof plate ablation. The pSmad1/5/8 gradient, which is reduced and flattened compared with normal (10), remains continuous across the cortex–midline border. C Right highlights neural expression for clarity. [Scale bars: 0.1 mm (low power) and 0.05 mm (high power).] (D) Model using NPC-intrinsic ultrasensitivity and irreversibility to initially convert the Bmp gradient into a crude Msx1 border, which is then refined by additional mechanisms into a sharp cortex–midline boundary. cx, cortex; dm, dorsal midline.

Discussion

Ultrasensitive Msx1 induction appears to be involved in the earliest stages of border formation within the Bmp gradient: (i) Msx1 induction in vivo occurs early and precedes the acquisition of other midline fates such as cell death and choroid plexus (8, 13, 30); (ii) in other systems, Msx1 acts as an early positional determinant rather than later cell fate specifier (14, 15); (iii) the Bmp-Msx/msh pathway in neurectoderm patterning is phylogenetically ancient (the “ontogeny recapitulates phylogeny” argument), apparently predating the counteracting ventral signaling pathways of Dorsal and Sonic hedgehog (Shh) in invertebrates and vertebrates, respectively (16); and (iv) Msx1 induction is rapid and cell-intrinsic, enabling its deployment before slower translation-dependent mechanisms. As a form of cell “memory” that reflects a history of suprathreshold stimulus exposure rather than ongoing stimulation (4, 31), irreversibility for Msx1 would complement ultrasensitivity to establish and maintain the cortex–midline border. Msx1 stability and brevity of the Bmp4 stimulus requirement would also obviate the need for steady-state gradients in vivo, which are often assumed for computational purposes but are not necessarily realistic (32, 33). Cell-intrinsic irreversibility could also account for the continued expression of Msx1 into adulthood (18), when Bmp signaling is presumably much lower. Collectively, these findings lead to a model wherein ultrasensitivity results in discrete, all-or-none Msx1 decisions at the single-cell level that rapidly sculpt a crude cortex–midline border, which is subsequently refined by cell–cell and gene network interactions that remain to be defined (Fig. 4D).

The mechanism(s) underlying ultrasensitive Msx1 induction also remain unknown but are constrained by current findings. The apparently linear relationship between Bmp4 concentration and nuclear pSmad1/5/8 level resembles that described for activin and pSmad2/3 in cultured Xenopus blastula cells (34). Thus, whereas pSmad is required for Msx1 transactivation (35, 36), pSmad and Msx1 levels are not linearly correlated. Switch-like Msx1 induction therefore differs from more linear morphogen concentration-dependent responses (7), and its mechanistic basis must be downstream of Bmp signaling and nuclear pSmad levels. It is important to note, however, that pSmad may still be directly involved in the ultrasensitivity mechanism [e.g., cooperative pSmad binding or interaction with other transcription factors (TFs) at the Msx1 promoter]. In addition, regulation of baseline Msx1 transcriptional repression or mRNA stability are possibilities with the requisite short timescales. Mechanisms with longer timescales could then be used to refine the initial Msx1 border, such as the competition and feed-forward border mechanisms used by decapentaplegic (dpp), the Bmp4 orthologue and morphogen in Drosophila development (37, 38). Tissue-level mechanisms, such as sorting and migration, may also be used (6, 7).

The need for ultrasensitive Msx1 induction in border formation may relate to the unique means by which Msx1 regulates transcription. As a transcriptional repressor (14, 15), Msx1/msh has been implicated in cross-repressive interactions that are commonly used to establish and/or maintain borders (6). These interactions often involve homeodomain-containing TFs, as observed in the vertebrate spinal cord (39). However, despite being a homeodomain TF, Msx1 acts primarily via protein–protein interaction rather than DNA binding (14, 40) to repress transcription by “squelching” (41). With homeodomain TFs of the Dlx, Lim, and Pax families, Msx1 initiates homeodomain–homeodomain interactions that prevent DNA binding by either TF (42, 43). Squelching by Msx1 therefore results in biochemical cross-inhibition between homeodomain TFs but precludes transcriptional cross-repression requiring DNA binding. In such cases where cell-intrinsic ultrasensitivity involving conventional cross-repression cannot be used to make or sharpen gene expression borders, ultrasensitive induction of a transcriptional squelcher would be a logical, and perhaps more essential, solution for making an initial border. Squelching also means that Msx1 function in any given cell depends on the other TFs expressed, which may explain the significant context dependence of Msx/msh pathway functions in patterning, proliferation, death, and differentiation (14, 15), and perhaps the disparate roles of Bmp signaling in stem cells in general (44).

Methods

Mice.

For timed pregnancies, noon of the vaginal plug date was designated day 0.5; developmental stages were confirmed by embryo crown–rump measurement. The following mice and matings were used: β-actin-Cre [ACTBCre, FVB/NTgN(ACTB-Cre)2Mrt; Jackson Laboratories] (45) with Gdf7-XstopX-DTA (Gdf7DTA) (46) for roof plate ablation; Msx1^nlacZ (17) with CD1 (Charles River) for expression analysis; and CD1 for wild-type studies. Genotypes were determined by PCR genotyping of tail DNA for Cre and DTA (47) or by X-Gal staining of limb buds (Msx1^nlacZ allele). All animal studies were carried out according to Institutional Animal Care and Use Committee guidelines.

In Situ Hybridization, Immunohistochemistry, and Immunocytochemistry.

These were performed and imaged as described (10, 13) with one modification: for immunohistochemistry, some tissues were fixed for 1 h at 4°C in 4% paraformaldehyde, 0.1% saponin, and 0.25 M sucrose to improve nuclear antigen fixation. Immunocytochemistry was performed as described (48) with one modification: cells were fixed with 4% paraformaldehyde with 0.1% saponin, 5 mM MgCl₂, 10 mM EGTA, and 4% sucrose in PBS for 15 min at room temperature. A mouse Msx1 EST (IMAGE clone 903377, GenBank accession no. AA518368) was used as a riboprobe template. The following antibodies were used: pSmad1/5/8 (rabbit polyclonal antibody against human Smad5 phosphopeptide, recognizes double phosphorylated Smad1, Smad5, and Smad8; 1:40 dilution; catalog no. AB3848; Chemicon) and Msx1/2 (mouse monoclonal antibody against chick Msx1/2; 1:3,500 dilution; 4G1; Developmental Studies Hybridoma Bank, University of Iowa).

X-Gal Histochemistry and Cytochemistry.

X-Gal histochemistry was performed as described (13). For X-Gal cytochemistry, cells on laminin-coated coverslips were fixed with 4% paraformaldehyde and 0.25% glutaraldehyde, 2 mM MgCl₂, and 5 mM EGTA for 10 min at room temperature. Coverslips were then rinsed three times, 30 min each, with 2 mM MgCl₂ in PBS. X-Gal staining was done overnight at room temperature and stopped with three PBS rinses. Coverslips were then dehydrated through graded alcohols to xylene and mounted with Permount.

Explant Cultures.

Explants were prepared and processed as described (10). One hour after plating whole dorsal forebrain explants ventricular surface up on Whatman membranes, Affigel blue gel beads (Bio-Rad) or heparin acrylic beads (Sigma), soaked in 10 μl of 100 μg/ml recombinant human Bmp4 (R & D Systems) or BSA, were placed onto explants by using pulled and flame-polished microcapillary pipettes and a mouth aspirator. Beads were rinsed briefly in PBS before placement. Explants were processed 24–36 h after bead placement for whole-mount in situ hybridization or X-Gal staining.

Dissociated Cortical NPC Cultures.

Mouse NPCs were isolated from E10.5 or E12.5 embryo cortex as described (10, 48). Briefly, skin and mesenchymal layers were manually removed from dissected telencephalic vesicles. Vesicles were incubated in 0.05% trypsin with 0.02% EDTA and 0.2% BSA in HBSS for 20 min at 37°C. Trypsinization was stopped by an equal volume of 1 mg/ml soybean trypsin inhibitor (Sigma) in HBSS. Tissue digests were dissociated by using several rounds of trituration with fire-polished Pasteur pipettes. Cells were washed once with 0.2% BSA in HBSS and plated at 50,000 cells per milliliter on laminin-coated coverslips in media (49) with 20 ng/ml EGF, 10 ng/ml FGF2 (R & D Systems or PeproTech), and 2 μg/ml heparin (Sigma). Bmp4 was added 24 h after initial plating. For washout experiments, Bmp4-containing media was aspirated and washed three times with fresh media containing EGF, FGF2, and heparin. RNA was harvested and column-purified (Bio-Rad) at designated times after initial Bmp4 addition.

Imaging.

Digital image capture of tissue sections, explants, and dissociated cells was performed as described (10, 48). For comparative studies (e.g., mutant versus littermate control sections, coverslips in different Bmp4 concentrations), processing steps and assays were carried out in parallel, and images were captured by using identical camera settings and image enhancements. Parallel image enhancements were limited to levels, brightness, and contrast in Photoshop. Fluorescent images were taken in eight-bit monochrome mode (Spot RT), then converted to full RGB color and overlaid in Photoshop. For panels highlighting neural expression, color images were inverted to grayscale, and nonneural tissues were manually traced with the Photoshop pen tool and filled in with off-white (RGB 245:245:245) pinlight at 100% opacity. Schematics and figure labeling were done in Illustrator.

Real-Time qRT-PCR.

RNA preps, cDNA syntheses, PCR quality controls, experimental runs, and statistical methods were performed as described (10, 13). All primers and amplicons used in this study (SI Table 1) were verified by amplification efficiency testing, melting curve analysis, and agarose gel electrophoresis; all experimental amplicons were also verified by sequencing and BLAST analysis. All cDNA samples were validated for reverse transcription reaction efficiency and minimal genomic DNA contamination by using 18S primers (cDNA/genomic target ratio > 10⁵) and run in duplicate or triplicate for 40 cycles. Primers for cyclophilin A (CYPA) or 18S rRNA, the two best internal references among four screened on embryonic cortical tissues and cells (13), were included for normalization in all experimental runs (Fig. 2G normalized to 18S; all others normalized to CYPA).

Quantitative Methods.

Real-time qRT-PCR data (Fig. 2) were calculated and plotted as described (10). Data were plotted as normalized −ΔCt values (normalized to 0 and plotted such that up-regulation is positive and down-regulation is negative) to accurately account for control sample variation, which is inaccurate for ΔΔCt values (control samples normalized to either 0 or 1 before statistics). For Msx1/2 immunocytochemistry (Fig. 3C), monochrome images were acquired and thresholded in parallel; Msx1/2- and Hoechst-positive counts were performed manually. For X-Gal cytochemistry, counts were manually recorded during microscopic viewing by using only slight differential interference contrast optics to increase X-Gal detection sensitivity. All counts were blinded. Means and standard errors (two-sample, two-tailed t tests) were derived from the following numbers of independent cultures: two to four (Fig. 2), three (Fig. 3C and SI Fig. 6C), and two (Fig. 3D and SI Fig. 6 A and B, except one set for the 128 ng/ml Bmp4 sample). Isolated/contacted cell counts for 0, 4, 8, 16, 32, 64, 128, and 256 ng/ml Bmp4 samples in Fig. 3D and SI Fig. 6B: 80/5,080, 170/2,120, 110/2,940, 135/1,350, 95/3,300, 110/1,890, 210/1,100, 120/2,765; cell counts for 0, 1, 4, 8, 16, 32, and 64 ng/ml Bmp4 in SI Fig. 6B: 340, 370, 390, 375, 255, 105, and 235. Hill curve fits were generated by using the standard three-parameter Hill equation built into the SigmaPlot program, R = R_max × X^nH/(EC₅₀^nH + X^nH), where R_max is the maximal response, nH is the Hill coefficient, and EC₅₀ is the concentration at which half-maximal response occurs. Graphs and curve fits were produced in SigmaPlot or Kaleidagraph using the same equation.