The mechanisms of morphogenesis are not well understood, yet shaping structures during development is essential for establishing correct organismal form and function. Here, we examine mechanisms that help to shape the developing face during the crucial period of facial primordia fusion. This period of development is a time when the faces of amniote embryos exhibit the greatest degree of similarity, and it probably results from the necessity for fusion to occur to establish the primary palate. Our results show that hierarchical induction mechanisms, consisting of iterative signaling by Sonic hedgehog (SHH) followed by Bone morphogenetic proteins (BMPs), regulate a dynamic expression pattern of Shh in the ectoderm covering the frontonasal (FNP) and maxillary (MxP) processes. Furthermore, this Shh expression domain contributes to the morphogenetic processes that drive the directional growth of the globular process of the FNP toward the lateral nasal process and MxP, in part by regulating cell proliferation in the facial mesenchyme. The nature of the induction mechanism that we discovered suggests that the process of fusion of the facial primordia is intrinsically buffered against producing maladaptive morphologies, such as clefts of the primary palate, because there appears to be little opportunity for variation to occur during expansion of the Shh expression domain in the ectoderm of the facial primordia. Ultimately, these results might explain why this period of development constitutes a phylotypic stage of facial development among amniotes.
Morphogenesis of the face is a complex process that involves iterative signaling between epithelial and mesenchymal cells. These interactions establish domains of directional growth that ultimately sculpt the face. An important morphogenetic event in facial development is the fusion of the maxillary (MxP), lateral nasal (LNP) and medial nasal (MNP) portion of the frontonasal process (FNP) to form the primary palate of the upper jaw. We have recently determined that a phylotypic stage exists, precisely around the time of fusion of the facial primordia (Young et al., 2014). At this time, the faces of amniote embryos exhibit the highest degree of morphological resemblance because, independent of the final adult form, they all pass through a developmental stage in which the facial buds collide and fuse to form the face (Young et al., 2014).
Signaling by Sonic hedgehog (Hu et al., 2003; Hu and Marcucio, 2009b; Young et al., 2010), Fibroblast growth factor (Richman et al., 1997; Szabo-Rogers et al., 2008,, 2009; Griffin et al., 2013), Bone morphogenetic protein (Francis-West et al., 1994; Duprez et al., 1996; Ashique et al., 2002a,,b; Foppiano et al., 2007) and Wnt (Brugmann et al., 2007; Geetha-Loganathan et al., 2009; Reid et al., 2011) all contribute to facial morphogenesis. Our work has focused primarily on the role of the frontonasal ectodermal zone (FEZ) during development of the upper jaw (Hu et al., 2003; Hu and Marcucio, 2009b). In amniotes, the FEZ is located in the stomodeal ectoderm and comprises a domain of Shh-expressing cells that line the roof of the mouth and a domain of Fgf8-expressing cells that covers the dorsal region of the FNP. Once Shh expression begins in the FEZ, Fgf8 expression becomes restricted to the epithelium lining the nasal pits. Together, these and other molecules regulate the proximo-distal extension and dorso-ventral polarity of the upper jaw. However, the role of the FEZ on directing growth of the facial primordia during prominence fusion is unknown.
In this study, we have determined the relative movements of the MNP, LNP and MxP during their fusion. These tissue displacements are regulated by a series of epithelial-mesenchymal interactions, involving the FEZ and the neural crest mesenchyme. Signaling by the SHH and BMP pathways regulates the changing expression domain of Shh in the FEZ, which coordinates the directional growth of the facial buds required for the formation of the primary palate.
Integrated movement of the MNP toward the LNP and MxP
Our first goal was to assess the major trajectories of tissue displacement during development of the primary palate. Three-dimensional (3D) shape analysis of the face during normal morphogenesis demonstrates how the MNP, LNP and MxP follow orchestrated movements that result in distal contact and ultimately fusion. Specifically, when applied to an early-stage embryo, the shape vector associated with age shows that the MxP primarily grows in an anterior direction, whereas the MNP grows in a caudal (downward or posterior) direction with little initial anterior outgrowth (Fig. 1; supplementary material Movie 1). The globular process of the FNP and the lower margins of the nasal pits next meet at the distal tip of the MxP, an event associated with subsequent fusion of the prominences. After fusion, the avian FNP undergoes rapid expansion, associated with the derived proportions of the premaxilla and upper beak (Young et al., 2014).
Shh expression in the facial ectoderm is dynamic
In both chick (Gallus gallus) and mouse (Mus musculus) embryos, Shh is expressed in the surface cephalic ectoderm covering the roof of the stomodeum once neural crest cells arrive in the region (Hu and Marcucio, 2012). The Shh expression domain initially forms a boundary with ectodermal cells expressing Fgf8, but, once Shh is expressed, Fgf8 is quickly downregulated and transcripts are restricted to the ectoderm of the nasal pits (Hu et al., 2003). In chicks, this occurs around Hamburger–Hamilton stages (HH) 20-21, and in mice, this occurs between embryonic day (E) 9.5 and E10.5 (Hu and Marcucio, 2009b). In both species, after Fgf8 becomes restricted to the nasal pit, the Shh expression domain expands laterally into the maxillary process, the globular process of the FNP and into the nasal pit (Fig. 2A-D,F-I; n=12). After fusion, Shh expression is only detectable in the roof of the mouth (Fig. 2E,J; n=12).
SHH and BMP signaling work in series to regulate the expansion of Shh
The goal of our work was to assess the molecular mechanism(s) underlying the expansion of Shh expression in the surface ectoderm, and to determine the effect that this changing expression has on morphogenesis of the upper jaw. Our previous research has shown that both SHH and BMP signaling is required for the initial expression and maintenance of Shh expression in the FEZ (Foppiano et al., 2007; Hu and Marcucio, 2009a; Marcucio et al., 2011). Therefore, we wanted to assess the role of these two signaling pathways during expansion of the Shh expression domain.
To test the ability of SHH signaling to regulate Shh expression we performed gain- and loss-of-function experiments in chick embryos. First, we implanted beads soaked in SHH-N into one side of the FNP of chick embryos at HH 20 and assessed Shh expression 24 h later (∼HH 23/24). Shh expression was strongly upregulated in the globular domain of the FNP on the treated side compared with the untreated side (Fig. 3A,B; n=12), and this was accompanied by increased cell proliferation (Fig. 3C-E). Controls, implanted with beads soaked in BSA showed no change in gene expression (Fig. 6A; n=10). In addition to changes in Shh expression, we also observed that Bmp2, 4 and 7 were significantly upregulated in these embryos [Fig. 3F; n=8; see also Hu and Helms (1999); Foppiano et al. (2007); Chong et al. (2012)].
Next, we electroporated a dominant-negative Ptc (PtcΔLoop2) into the ectoderm covering one side of the FNP at HH 15 (Fig. 4; n=35). We assessed SHH signaling in treated embryos using qPCR, and determined that, 48 h after electroporation (∼HH 23/24), Ptc and Gli1 expression levels were significantly downregulated (Fig. 4I; n=6), as was expression of Shh itself (Fig. 4I; n=8). Control embryos, electroporated with HSP-lacZ (Fig. 4A,B), showed no change in gene expression (Fig. 4C; supplementary material Fig. S1). The decreased Shh expression on the treated side was apparent by in situ hybridization (Fig. 4D; n=21). We observed similar outcomes when we blocked SHH signaling using the antagonist cyclopamine (supplementary material Fig. S2; n=30), and we did not observe evidence of apoptosis in the mesenchyme after blocking SHH with cyclopamine (supplementary material Fig. S3B,C; n=9). Additionally, Fgf8 expression was not restricted to the nasal pits after blocking SHH signaling, thereby suggesting that SHH signaling is required for downregulation of Fgf8 expression across the midline of the FNP (Fig. 4E,F; n=9). This agrees with previous publications (Cordero et al., 2004; Foppiano et al., 2007; Abzhanov et al., 2007). We observed no change in Wnt9b expression compared with normal embryos (Fig. 4G,H; n=9). Bmp2 exhibited no change, whereas Bmp4 and Bmp7 were downregulated (Fig. 4J; n=6).
As we observed changes in Bmp expression levels associated with altered Shh expression, we wanted to assess the extent to which BMP signaling acts in concert with SHH signaling to regulate the expansion of Shh expression in the facial ectoderm. We blocked BMP signaling by implanting a bead soaked in Noggin protein into the mesenchyme on one side of the FNP of chick embryos at HH ∼15/16. We then examined Shh expression 48 h later. We observed a significant decrease in Shh expression levels (Fig. 5E, qPCR; n=6). Whole-mount in situ hybridization confirmed that Shh expression was lost in the ectoderm on the treated side of the embryos (Fig. 5A-C; n=9), and there was a lack of outgrowth on that side of the embryo (Fig. 5B,C).
Our data suggest that SHH and BMP signaling work together to regulate the expansion of Shh expression within the facial ectoderm. We next wanted to determine whether SHH and BMP work in parallel or in succession to control expression of Shh in the ectoderm. First, we blocked BMP signaling by implanting a bead soaked in Noggin into one side of the FNP (HH ∼15/16), and 24 h later we implanted a bead soaked in SHH-N. We observed that the Shh expression domain on the experimental side was absent (Fig. 5D; n=32), indicating that, after blocking BMP signaling, SHH was not able to restore the Shh expression pattern in the facial ectoderm. Next, we blocked SHH signaling by electroporating PtcΔLoop2 into the ectoderm at HH ∼15/16, and 24 h later we placed a bead soaked in BMP4 into the mesenchyme on the experimental side. After 24 h we assessed Shh expression and determined that BMP signaling was able to restore Shh expression in the facial ectoderm (Fig. 6D; n=9).
SHH signaling directs growth of the facial primordia during fusion of the primary palate
The expansion of the Shh expression domain into the globular process and nasal pit epithelium has not been described before, and we wanted to understand the functional significance of the expanded Shh expression domain on facial morphogenesis. We therefore repeated the loss-of-function experiments using PtcΔLoop2 and cyclopamine described above, and allowed embryos to develop to later stages for analysis. In all cases, we observed hypoplasia of the lateral edge of the FNP, a failure of the globular process to contact the lateral nasal and maxillary processes, and a cleft of the primary palate (Fig. 7; n=7; supplementary material Fig. S4; n=7). The treated side was smaller than the contralateral side (Fig. 7B versus Fig. 7D), and this hypoplasia was associated with decreased cell proliferation (supplementary material Fig. S2E,F; n=8) but not apoptosis (supplementary material Fig. S3; n=8).
In addition to the loss-of-function experiments, we also activated SHH signaling in the globular process and assessed the morphological outcomes at later time points. We observed that the direction of growth of the lateral regions of the FNP was altered (supplementary material Fig. S5; n=6). Instead of arching laterally to extend and touch the lateral nasal and maxillary processes, the FNP grew rostrally and did not meet with the other facial primordia, and the treated side was longer and broader than the contralateral side. The result of this aberrant growth was the formation of a cleft of the primary palate (supplementary material Fig. S5; n=6). In this case, the morphological changes were associated with increased cell proliferation and the affected region was larger (Fig. 3; n=8).
Morphogenesis is a complex process in which 3D shape is generated de novo during embryogenesis. Although many advances have been made regarding the signaling interactions that control histogenesis, how these signaling interactions regulate morphogenesis are not well studied. A recent report indicated that, during morphogenesis of the upper jaw, SHH signaling participates in regulating the balance between proliferation and apoptosis, which is required for proper growth of the facial primordia and for fusion and dissolution of the epithelial seam during fusion of the facial primordia (Kurosaka et al., 2014). The entire process of fusion is molecularly complex, and is regulated in part by the interactions of Pbx, Wnt, Irf6 and p63 (Thomason et al., 2010; Ferretti et al., 2011). Here, we focused on the mechanisms that regulate the changing pattern of expression of Shh changes during fusion of the facial primordia. Our results show that an iterative series of signals between the ectoderm and mesenchyme of the face orchestrate the changing expression pattern of Shh in the FEZ. This changing pattern of Shh regulates the morphogenetic movements that lead to the formation of the primary palate by coordinating the shape changes of the MNP, LNP and MxP in order for fusion of the primordia to occur.
Mechanisms of morphogenesis
The nature of the morphogenetic movements that generate the shape changes during fusion of the primary palate is unknown. Previous models have relied on location of differential zones of proliferation as mediators of shape change and morphogenesis (Ede et al., 1975; Wu et al., 2006). Our work here supports a role for changes in proliferation driving the morphogenetic process, but simple changes in the rate or location of proliferating cells is unlikely to underlie all of the directional growth that is observed during facial development. Morphogenesis of epithelially derived structures results from cellular changes that involve shape changes and polarized activities (Guillot and Lecuit, 2013). A recent study on hair follicles has shown that movements of epithelial cells, but not cell proliferation, were responsible for shaping the follicle (Ahtiainen et al., 2014). The relevance of directed cell behaviors in morphogenesis is not restricted to the epithelium. Studies on the limb have shown that non-directional growth alone is unable to explain the shape changes the limb bud experiences during development. Rather, polarized cell activities, such as cell movement in the mesenchyme, appear to comprise the emergent properties that generate shape (Boehm et al., 2010; Hopyan et al., 2011). Results from work in our laboratory indicate that polarized cellular behaviors associated with cell movement are associated with facial shape (Li et al., 2013). In a multi-primordia organ like the upper jaw, the influence of these cellular behaviors on interactions among the facial primordia would also contribute to overall jaw morphology (Linde-Medina and Newman, 2014). Together, these results suggest that cell movements generate forces within the mesenchymal tissues that drive morphogenetic processes. Further work is required to directly test this hypothesis in facial morphogenesis.
SHH and BMP form a signaling loop in the craniofacial complex
SHH and BMP signaling are essential for proper development of the face. BMP signaling is directly or indirectly involved in the regulation of the onset of Shh expression in the FEZ at HH 20 (Foppiano et al., 2007). Subsequently, SHH from the ectoderm appears to regulate the expression patterns of BMPs in the adjacent mesenchyme (Hu and Marcucio, 2009b). Our data show that, in turn, BMP signaling in combination with SHH itself regulate the expansion of the Shh domain into the epithelium of the nasal pits, maxillary processes and globular processes. However, our experiments do not reveal whether these regulatory interactions are direct. That is, whether SHH directly activates the Bmp4 and Bmp7 loci, and whether BMPs directly activate the Shh locus, are not known. Previously, a bead soaked in SHH protein was shown to activate expression of Bmp4 in the mesenchyme of the FNP (Hu and Helms, 1999). Moreover, work by other investigators has revealed that Bmp2 has a Gli2 response element that mediates Shh induction of Bmp2 expression in osteoblasts (Zhao et al., 2006) and in the limb bud (Vokes et al., 2008). We are currently exploring the transcriptional interactions of these pathways during development of the FNP. Nonetheless, our results contradict previous research in which it was proposed that the ability of Noggin to block BMP signaling was responsible for the regulation of Shh expression in the facial ectoderm (Ashique et al., 2002a; Lana-Elola et al., 2011). These studies did not take into account that the Shh expression domain changes over time and that it eventually occupies new ectodermal regions during the directional growth of upper jaw primordia, a developmental event in which SHH signaling comes to play.
Our data demonstrate that SHH and BMP signaling act in succession to regulate the dynamic expression of Shh in the facial epithelium during morphogenesis. However, variation in SHH and BMP signaling does not appear to alter the positioning of the Shh expression domains. For example, gain-of-function experiments using SHH intensified Shh expression in the epithelial domains but did not lead to ectopic Shh expression or an expansion of the primary domain. This suggests that the ectodermal domains that express Shh acquire a competence prior to the onset of Shh expression and that the regulation of the spatio-temporal configuration of Shh expression in the FEZ is separate and distinct from the regulation of the level of Shh expression. For example, Shh expression in the unique domains of ectoderm is probably programmed by tissue-specific regulatory regions within the Shh locus. However, enhancer elements that drive Shh expression in the facial ectoderm remain unknown (Epstein et al., 1999,, 2000; Jeong and Epstein, 2003; Jeong et al., 2006,, 2008; Marcucio et al., 2011; Anderson et al., 2014).
This finding is in contrast to the observed pattern of Shh expression in the rugae (a series of epithelial ridges on the palate), where a Turing-type mechanism of activator-repressor interactions regulates the sequential onset of Shh expression in each rugal domain (Economou et al., 2012). In the palatal epithelium, ectopic Shh expression is observed when the activator-repressor mechanism is disrupted, as, for example, when a rugal domain is excised, eliminating the inhibitory molecules secreted by these cells. In this scenario, the expression domains arise from emergent inductive processes. If a similar mechanism functioned in the FEZ, then, by changing the steepness of the activator-repressor gradient in gain- and loss-of-function experiments, new Shh expression domains would appear, due to altered threshold responses within the ectoderm. Instead, we observed that the intensity of Shh expression changed, but the location of the domains remained unaltered. In this case, the inductive mechanisms underlying the new Shh expression domains appear to obey hierarchical principles as previously described (Salazar-Ciudad et al., 2003).
Interestingly, a similar phenomenon is observed during the initial formation of the FEZ. We determined that SHH signaling within the brain (Marcucio et al., 2005; Hu and Marcucio, 2009a) and BMP signaling within the FNP (Foppiano et al., 2007) and the neural crest cells (Hu and Marcucio, 2012) act together to induce Shh expression in the ectoderm, thus giving rise to the formation of the FEZ. However, in our experiments, we never observed ectopic expression of Shh in the ectoderm, suggesting that the signals and cells appear to trigger a gene expression pattern that is inherent within the ectoderm itself. To date, the mechanisms underlying this phenomenon are unknown.
Relationship of shape to normal and diseased phenotypes
The inability to establish ectopic domains of Shh expression or deviate significantly from the anatomical configuration of the FEZ during formation of the primary palate reflects the constraint placed on facial shape during this period of development. We have shown in a large comparative analysis of the development of the amniote face that a phylotypic stage exists around the time of palatal fusion (Young et al., 2014). At this time, facial shapes among amniotes converge in order for the primary palate to form, and after that, the faces are able to diverge into the array of morphologies present in adult amniotes. Our results suggest that spatial stability of the Shh expression domain in the facial ectoderm is a central feature of this constraint on variation during primary palate formation. Therefore, understanding how the Shh expression domain is stabilized is crucial for unraveling the mechanistic basis for the minimization of morphological variation both within and among species during the formation of the face. Our results show that, whereas the level of Shh expression is dynamically regulated by SHH and BMP signaling, the spatial domain of the FEZ is not. Instead, other pathways or mechanisms must act to stabilize the spatial domain of the FEZ during the formation of the face. This mechanistic separation of the spatial domain of FEZ expression and regulation of the level of Shh expression in the FEZ helps to explain the reduction in morphological variation during facial prominence fusion. Our results have revealed that levels of SHH and BMP signaling are important for proper development. For example, when we activated or suppressed the SHH pathway the resulting embryos produced clefts due to either hypo- or hyperplasia of the affected primordia. This agrees with our previous work, showing that shape is highly associated with successful fusion of the facial primordia. We propose that the resilience of the FEZ spatial expression domain contributes to the reduced shape variance observed at the time of prominence fusion among amniotes, as well as the high resilience to clefting phenotypes during development, but above or below a threshold of SHH signaling shapes can be generated that produce clefts.
In summary, we have determined that Shh expression in the facial ectoderm is dynamic. The expression domain unfolds in a very precise way as development proceeds, and the changing pattern of expression is regulated by a series of SHH and BMP signaling events. The role of SHH in promoting fusion of the facial primordia is essential for proper facial morphogenesis (Kurosaka et al., 2014). Understanding the precise mechanisms by which the Shh expression domain is positioned within the ectoderm will be a key step in understanding the molecular mechanisms that regulate morphogenesis of the upper jaw.
MATERIALS AND METHODS
Chick embryos were incubated at 39°C until either collection or experimentation. For FEZ analysis, chick embryos were harvested at HH 22, HH 25 and HH 26, and mouse embryos were harvested at E10.5, E11 and E11.5. Alternatively, chick embryos were prepared for surgery as described (Hu and Marcucio, 2011). All embryos were fixed in 4% paraformaldehyde, dehydrated in a graded series of methanol and stored at −20°C until analysis. The experiments conform to institutional and national regulatory standards that are applied for analysis of vertebrate embryos.
Preparation and placement of beads
Affi-Gel Blue beads (70-140 mesh, 100-200 μm diameter; Bio-Rad) were rinsed four times in PBS and manually selected for uniform size. Beads were soaked in 0.1% bovine serum albumin (BSA), recombinant human BMP4 (4 μg/ml with 0.1% BSA; R&D Systems), recombinant human Noggin-N (250 μg/ml with 0.1% BSA; R&D Systems) or recombinant SHH-N (400 μg/ml with 0.1% BSA; Ontogeny) for ≥1 h at 37°C. Prior to initiating experiments, we performed a dose-response experiment for BMP4 to determine a dose that did not induce apoptosis. Protein-soaked beads were stored at 4°C for a maximum of 1 week. Beads were placed into the right side of the FNP by accessing the face as described (Hu and Marcucio, 2011).
PtcΔLoop2 [see Briscoe et al. (2001), concentration 3 μg/μl] or HSP-lacZ (concentration 3 μg/μl) were injected beneath the head at HH 15/16 and electroporated into the ectoderm using an Intracel TSS20 Ovodyne Electroporator (10 V, 5 pulses, pulse width=20).
We collected chicken embryos (n=17) spanning the time period of initial facial morphogenesis (84-160 h of incubation), from initial prominence outgrowth through primary palate fusion. We scanned each embryo in a µCT 40 micro-CT scanner (Scanco) at a resolution of 12-20 µm. Reconstructed image stacks were converted into 3D objects in Amira 5.4 (Visage Imaging) and surfaces were landmarked in Landmark Editor 3.0 (Institute for Data Analysis and Visualization) (n=20). We next imported the raw coordinates into MorphoJ (Klingenberg, 2011), and performed a Procrustes superimposition to minimize differences in orientation and scale. We next performed a multivariate regression of the Procrustes coordinates on age (hours of incubation), and utilized the resulting age-shape vector to project a representative early stage embryo (84 h) to 160 h of development.
BrdU labeling and TUNEL
Twenty minutes prior to collection, 200 nl of bromodeoxyuridine (BrdU) labeling reagent (Invitrogen) was injected into the vitelline vein. Embryos were fixed, embedded and sectioned as described. BrdU-labeled cells were visualized on paraffin sections; detection of BrdU incorporation was assessed by immunohistochemistry following the manufacturer's instructions (Invitrogen). BrdU-incorporated cells were counted (Olympus) and the percentage of proliferating cells in the medial portion of the FNP was determined. Student's t-test was performed to determine statistical significance (P<0.05). A terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed according the manufacturer's instructions (Roche).
In situ hybridization
In situ hybridization was performed on paraffin sections and whole embryos as described (Albrecht et al., 1997). Subclones of chick Shh, Wnt9b, Fgf8 and mouse Shh were linearized to transcribe digoxygenin-labeled antisense riboprobes. For whole mounts, tissues were hybridized with 0.5-1 μg/ml digoxygenin-labeled cRNA probes, and, after washing, were incubated with an alkaline phosphatase-conjugated anti-digoxygenin antibody (Boehringer). NBT/BCIP substrate (Roche) was used for color detection. For in situ hybridization on sections, tissues were hybridized with 35S-labeled cRNA, washed, air-dried and coated with liquid emulsion (Kodak). The in situ hybridization signal was visualized as described previously (Albrecht et al., 1997).
Total RNA preparation and qPCR
The FNP of embryos was dissected and subjected to 2 mg/ml dispase digestion on ice for 20 min. Neural ectoderm was separated from the surface ectoderm and mesenchyme using tungsten needles. Ectoderm and mesenchyme from individual embryos were used to create mRNA. RNA samples were isolated using an mRNeasy Kit (Qiagen). cDNA synthesis used Invitrogen Superscript III following the manufacturer's instructions. qPCR was performed using a Bio-Rad CFX96 real-time PCR machine. qPCR primers for specific genes are listed in supplementary material Table S1. Relative gene expression was calculated based on the 2−ΔCt method. ΔCt was calculated between each target gene and Gapdh as control. At least three biological replicates were prepared for all analyses and Student's t-test was used to assess significance.
PtcΔLoop2 was a gift from Gary Struhl. We would like to thank all of the members of the Orthopaedic Trauma Institute, especially Gina Baldoza for constant support and Marta Linde for critical comments on the manuscript.
D.H. and R.S.M. designed the research and conceived the experiments. D.H. performed all experimental manipulations. D.H., X.L., Y.X. and N.Y. collected the data. All authors analyzed the data and contributed to writing the paper.
This research was funded by the National Institutes of Health/National Institute of Dental and Craniofacial Research (NIH/NIDCR) [5R01-DE019638 to R.S.M. and B.H.; 2R01-DE018234 to R.S.M.]. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.