This study represents the starting point for a systematic collection of detailed 3D expression patterns across key stages of mouse embryonic development where data can be readily cross-compared and compiled in a searchable database. We present the first description of the data illustrating the quality and resolution achieved, the capacity for cross comparison and the limitations. To present a manageable amount of data we show only the expression of the genes encoding the Wnt signalling molecules (19 genes) and the Fzd receptors (10 genes) at a single stage of development; embryonic day (E) 11.5, Theiler stage (Ts) 19. For cross-comparison of the expression patterns we focus on the developing forelimb bud, an important site of Wnt regulated morphogenesis, to document and illustrate the potential of a comprehensive set of 3D data for these signalling molecule and receptor genes. The stage was chosen as a point just prior to overt cellular differentiation when the bud is composed of multiple territories that need to be defined in context of their gene expression patterns.

Probes representing each of the 19 Wnt genes and 10 Fzd receptor genes were hybridised to a minimum of 10 Ts19 embryos in at least two independent experiments, however some more difficult patterns were generated in up to six experiments. A minimum of two specimens for each gene were scanned using OPT and the pattern reconstructed in 3D, and many were reconstructed up to 10 times to optimise parameters and check for variability. Movies showing 3D volume representations of the whole embryo pattern for each of the 29 genes can be viewed on a dedicated website; http://www.tcd.i/Zoology/research/WntPathway/. Fig. 1 shows still images of a selection of the patterns comparing an external view of the 3D OPT data (left) with the original photograph of the hybridised embryo (right). A range of staining intensities was generated (not shown) and tested. The staining intensities selected are deliberately low as staining which is too dark to transmit any light is not suitable for OPT imaging (acts as an optical barrier in multiple light paths and dampens autofluorescence, which is used to visualise the morphology of the specimen). However we have found that sites of low level expression, even if not clearly visible to the naked eye in the lightly stained original specimens are nevertheless picked-up by OPT and are clearly visible in the reconstructions; the sensitivity of the scanning technique on cleared tissue is greater than visual inspection of the original specimen and we have represented all of the expression sites visible in more intensely stained specimens.

We conclude that using carefully stage-matched specimens and selecting appropriate staining intensities, comparable 3D representations of expression patterns can be generated quickly. In light of our experience in this study, we recommend at least two independent hybridisations using 4–5 stage-matched embryos for each probe/stage, selecting a minimum of two structurally sound and appropriately stained embryos for scanning.

The images shown in Fig. 1 are external, projection views of the 3D data generated by OPT, mimicking the original whole-mount preparation but with the capacity to reveal deeper internal staining. The data can be viewed and analysed in any orientation or section plane as required. Fig. 2 illustrates multiple ways in which the Ts19 Fzd1 expression data can be viewed, showing external projection views of the 3D representations from different directions (Fig. 2A–D) and virtual sections in different planes. The tissue-level resolution (isotropic voxel dimension of ~10 μm) allows expression to be assigned to a particular tissue or organ. Fig. 2E–K shows the raw data scanned using visible light representing the expression sites (high grey-level) against background tissue (low grey-level) and Fig. 2E′–K′ shows merged data from two-channel scans using visible light (pseudocoloured in red), showing the expression sites, and autofluorescence (pseudocoloured in green), showing tissue morphology. Merged data can also be visualised as external views of the 3D object (see Fig. 5, column A for examples). These methods emphasise different aspects of the expression pattern. Simple grey-level representation shows subtle details of the pattern such as differences in level of intensity and very fine spatial restriction that is sometimes obscured in images that use strong pseudocolouration. In contrast, pseudocoloured images often more clearly show the morphological context of expression.

Fig. 2 represents the data for a single gene (Fzd1) at a single stage of development (Ts19). Expression is evident in the dorsolateral mantle layer of the midbrain (Fig. 2H and H′); the lateral telencephalon (Fig. 2I and I′); a lateral stripe in the hindbrain just posterior to the midbrain hindbrain boundary (Fig. 2A, H and H′); the eye (Fig. 2A), localised in the anterior walls of the lens vesicle and ventral optic stalk (not shown); the frontonasal process around the nasal pits (Fig. 2B, F and F′), and complex patches in the anterior maxillary and posterior mandibular processes and 2nd branchial arch (Fig. 2A); the body wall around the heart (Fig. 2F′); the myotome and extending into the body wall (Fig. 2G′); a lateral body wall domain extending between the fore and hindlimbs (Fig. 2B); and in the limb, in the AER, patches in the limb mesenchyme, particularly anterior, and in patches in the dorsal aspect at the base of the limb (Fig. 5), particularly around the anterior forelimb (Fig. 2A). The selected sections show a subset of the full pattern visible in Fig. 2A–D; all possible sections and sites can be viewed in the 3D reconstruction available through EMAGE, for example using Edinburgh Mouse Atlas Project (EMAP) software (MA3DView and MAPaint http://genex.hgu.mrc.ac.uk/MouseAtlasCD/intro.html). This is the most comprehensive capture of the Fzd1 expression pattern. Previous expression data for Fzd1 showed expression in the thymus primordium at Ts20 (Bleul and Boehm, 2001), of which there was no indication in the present study yet at Ts19, and at earlier stages (Borello et al., 1999) in a number of sites corresponding to some of those reported here.

Tables 1 and 2 in Supplementary data list sites of expression for each gene indicating newly described sites and added detail. The amount of pre-existing data varies but even for the most thoroughly described patterns additional aspects have emerged. For example the well documented domain of Wnt1 at the midbrain/hindbrain boundary (Bally-Cuif et al., 1992; Dymecki and Tomasiewicz, 1998; Wilkinson et al., 1987) is no longer throughout the dorsoventral extent of the neural tube at Ts19 but, while strong in the dorsal midline, is absent from the ventral floor and is restricted to the marginal zone in basal and alar territories (Fig. 4B). Wnt4 expression has been extensively described in the developing kidney (Stark et al., 1994) for example but much less well described in the CNS where we detected expression in different territories depending on the anterior–posterior level. It is throughout the ventricular zone of the forebrain and midbrain, strongest in the telencephalic vesicles and lowest in the midbrain (Figs. 4C and and1C).1C). In the preotic hindbrain, expression is restricted to a very localised region of the ventral marginal floor plate (Fig. 4C, arrow). From just posterior to the otic vesicle expression was detected along the length of the neural tube, in different dorsoventral domains depending on the anterior–posterior level. From anterior to the base of the forelimb there are two peaks of expression in a broad ventricular territory, one in dorsal and one ventral to the midpoint. In more posterior positions a single dorsal domain is seen (Fig. 4D). Wnt5a expression has also been extensively described in the literature (Gavin et al., 1990; Grove et al., 1998; Yamaguchi et al., 1999) particularly in the limb, branchial arches and forebrain (Fig. 1D). In addition Fig. 4E shows asymmetric expression in the mesenchyme to one side of the midgut in the umbilical hernia. The CNS also showed additional spatial restriction of Wnt5a expression in a thick ventricular band in the ventral midbrain and a more restricted ventricular zone dorsally (Fig. 1D). Fine ventricular staining is seen throughout the midbrain/hindbrain boundary and expression becomes again more intense in a broad ventricular zone in the anterior hindbrain (Fig. 4F). At the level of the otic vesicle within the hindbrain there is a very broad ventricular domain, excluding the floor plate, with territories of different dorsoventral levels (Fig. 4G). In the most posterior part of the hindbrain Wnt5a is restricted to two patches just dorsal to the floor plate (Fig. 4H).

One generalisation that could be made from our survey is that Fzd genes tend to show more extensive and less tightly defined expression domains than Wnt genes (Fig. 1). Among Fzd receptor genes, Fzd8 and Fzd10 show particularly striking patterns; Fzd8 in the future sites of skeletal muscle and Fzd10 in the dorsal CNS and progress zones of the limb (Fig. 1).

Most of the Wnt and Fzd genes have multiple sites of expression in different systems, but there are notable exceptions. Wnt8b is specific to the forebrain where the pattern has been well described previously at Ts17 (Richardson et al., 1999; Theil et al., 2002). Here we show extra details of the pattern in 3D at Ts19 in the cortical hem and choroid invagination (http://www.tcd.i/Zoology/research/WntPathway/), and an additional domain in symmetrical stripes in the diencephalon (Fig. 3H), similar to expression of the human orthologue (Lako et al., 1998). The paralogous gene Wnt8a also shows very restricted expression, in this case in the lens epithelium of the developing eye (Fig. 4I) and otic vesicle. The Wnt9 paralogues also show restricted expression at this stage. Relatively little has been published on the expression of Wnt9 genes but Wnt9a has been reported in the limb mesenchyme in the position of perspective joints (Guo et al., 2004) and Wnt9b in the facial region (Lan et al., 2006). While we saw expression of Wnt9b in very localised points in the facial ectoderm, at the edges of the nasal pit and at the maxillonasal groove (Fig. 4K), no expression of Wnt9a was detected in the limb at this stage. Novel sites of expression however were detected for Wnt9a in the medial telencephalon, and the midbrain hindbrain boundary (not shown), and for Wnt9b along the entire length of the mesonephric duct (Fig. 4L and M).

Eighteen of the twenty-nine Wnt and Fzd genes are expressed in localised domains within the developing forelimb bud at Ts19; 11 Wnts and 7 Fzds (Tables 1 and 2). In addition Wnt16 is expressed in limb mesenchyme around the future skeletal elements in the positions of the future elbow and digit joints by Ts20 (not shown) and Fzd3 is expressed throughout the limb mesenchyme, slightly elevated in the proximal anterior region. 3D movies of the limb patterns are available at http://www.tcd.i/Zoology/research/WntPathway/, Table 2 summarises the expression in named limb subdomains and Fig. 5 shows comparisons of the raw expression data in the left forelimb. For comparison, Fig. 5 also shows the localisation of differentiating cartilage stained with Alcian blue. For an overview of the 3D patterns, the left column in Fig. 5 shows external views of volume representations of the 3D data with expression domains represented in blue. These are stills from the 3D movies (http://www.tcd.i/Zoology/research/WntPathway/). The angles viewed are different for each gene and were chosen to best show the pattern rather than for comparison across genes. The remaining columns show closely matched 2D virtual sections through the 3D data in two midpoint longitudinal planes (from dorsal to ventral and from anterior to posterior) and three transverse planes. While 2D views do not show all aspects of the 3D patterns, these show the possibility of comparing data across specimens generated in this way.

Aspects of the expression of several Wnt genes in the limb were previously undescribed. For example analysis of the complete set of serial virtual sections through the specimens revealed expression of Wnt2 and Wnt4 in localised patches of proximal limb mesenchyme (Fig. 5, column B) that were not visible from external views of whole-mount preparations. These domains appear to overlap each other, are adjacent to a patch of Wnt11 expression, do not coincide with Alcian blue stained tissue and are overlapping with expression of Fzd4, Fzd6, Fzd8, Fzd9 and Fzd3 and close to sites of Fzd1 and 7.

Wnt6 is clearly expressed in the AER (Fig. 5, columns B and C) in addition to surface ectoderm, where it is restricted to ventral ectoderm in the handplate (Fig. 5, column E) but extends more dorsally in the limb shaft (Fig. 5, columns F and G). Five Wnt genes are therefore expressed in the AER (5a, 6, 10a, 10b and 11) at this stage of development. Not all are expressed throughout the AER but have different boundaries of expression within the distal ectoderm with Wnt11 only detected in the most distal part of the AER, Wnt10a also more restricted than Wnt10b and skewed toward anterior. Three Fzd members show elevated expression in adjacent cells; Fzd1 and Fzd10 in the AER itself, Fzd1 also in adjacent surface ectoderm, Fzd4 and 10 in the progress zone with Fzd4 skewed toward the posterior, Fzd9 is at a low level in the mesenchyme of the future digit elements. Fzd3 is at a low level throughout the mesenchyme.

Most of the AER expressing Wnts are either expressed throughout the distal most limb (5a) or are ectoderm specific (10a, 10b and 6), except Wnt11 which is expressed in complex territories within the mesenchyme along its proximo-distal extent, not adjacent to the AER. From whole-mount views the expression appears to be at the “core” of the limb bud (Christiansen et al., 1995) but on section analysis it is clear that the expression is excluded from the very centre of the limb and is more subectodermal. A comparison with Alcian blue staining shows that the domains do not overlap with forming cartilage. Comparison with other Wnt and Fzd expression patterns shows that it overlaps partially with expression of Fzd1 and Fzd8.

In addition to the AER, Wnt10a and Wnt10b are also expressed in localised patches of ectoderm in the proximal limb. Both are expressed in ventral ectoderm at the base of the limb where Wnt10a extends more distally along the limb shaft. Wnt10b is most strongly expressed in an anterior patch on the ventral limb bud base, with a lower level patch just posterior to the midline (see 3D movie of the cropped limb http://www.tcd.i/Zoology/research/WntPathway/). The ectoderm at the base of the limb bud is in fact a site of expression of multiple Wnt and Fzd genes although the exact distribution and the foci of most intense patches of expression are different for each (Fig. 5, left column and 3D movies http://www.tcd.i/Zoology/research/WntPathway/). Wnts 2, 3, 4, 6, 7A, 7B, 10A and 10B, Fzd6 and Fzd7 are all expressed in the ventral ectoderm with Wnt7b and Wnt10b most intense toward the anterior aspect. Table 2 highlights also that multiple genes are expressed in a complex pattern within the mesenchyme that does not correlate to any know morphological territories. This is true for Wnt11, Wnt5a in the proximal bud, where it overlaps with Alcian blue staining, Fzd1, where it is most intense in patches in the anterior shaft, Fzd4, with three discrete patches in limb mesenchyme- in the posterior distal region of the progress zone and dorsal and ventral patches, anterior to the midline, surrounding the future skeletal elements (Alcian blue) in the limb shaft, Fzd8, with a pattern suggestive of the position of future muscle masses (a pattern somewhat complementary to Fzd4), Fzd6 and Fzd9, which show lower level but more extensive patterns, highest dorsally. Fzd7 shows particularly elevated expression in a ring in the proximal limb shaft, surrounding but significantly more peripheral than the Alcian blue stained cells.

One of the best-studied Wnt genes in the context of limb development is Wnt7a which has been shown to be important in signalling from the dorsal ectoderm and D/V patterning of the limb (Parr et al., 1998; Parr and McMahon, 1995; Yang and Niswander, 1995). Here we see that at Ts19 the dorsal ectoderm expression of Wnt7a is only found in the distal bud; in proximal regions expression is localised in ventral ectoderm. The paralogous gene Wnt7b is also expressed in the ventral ectoderm, extending more distally than 7A and is not expressed in the distal dorsal ectoderm so that transverse sections mid way along the handplate show Wnt7a in the dorsal and Wnt7b in the ventral ectoderm (Fig. 5, E column).

Wnt9a has been reported in the region of future skeletal joints in the mesenchyme of E11.5 mouse limbs (Guo et al., 2004) but we record no such expression at Ts19. This may be a staging issue since Guo et al. also report expression of Wnt16 in the territory of future joints at E11.5 which we record only very lightly stained in one specimen at Ts19 but clearly visible by Ts21. Nevertheless, we did not record any Wnt9a expression in the limbs of Ts20 embryos. It is important to note the dynamism of many of these patterns where we see differences in the expression of some genes among embryos that could be classified as Ts19. For example in very late Ts19 and Ts20 limbs Wnt10b and Wnt11 are no longer expressed in the AER. This underlines the importance of careful stage matching in comparative studies.

Two of the Fzd genes (2 and 5) did not show expression in the limb buds, while Fzd3 expression was not localised and Fzd7 and 9 showed widespread expression in the mesenchyme. Only Fzd9 and Fzd10 expression was previously reported at Ts19 in the mouse limb bud (Nunnally and Parr, 2004; Wang et al., 1999) while expression of Fzd1, Fzd3, Fzd6, Fzd7 and Fzd9 was briefly noted in the limb buds of earlier embryos (Borello et al., 1999). As reported by Wang et al. (1999), we observed Fzd9 expression in the mesenchyme of the future digits in the distal limb bud and additionally in dorsal mesenchyme of the limb shaft. We observed very similar expression of Fzd10 in the progress zone as was observed by Nunnally and Parr (2004), with expression distributed uniformly in a broad band of distal mesenchyme and ectoderm. In contrast expression analysis in the chick embryo has shown localised expression in the polarising region (Kawakami et al., 2000). It is possible that localised Wnt activity within the polarising region might be controlled at a level other than localisation of a specific receptor in the mouse. Alternatively, Wnt activity may be conveyed by a different Fzd. In this context we note elevated Fzd4 expression in the posterior region of distal limb mesenchyme. Fzd3 and Fzd7 transcripts are also present in this region although again not localised to the polarising region.

The selected descriptions compiled here reveal a number of new characteristics of Wnt and Fzd gene expression patterns. For example in the limb there are areas where multiple genes are expressed; hotspots of Wnt and Fzd gene transcription. These include very specific individual patterns within the surface ectoderm at the base of the limb, some highest in the anterior and complex patterns in the mesenchyme (Table 2). It is clear that the patterns do not relate simply to the known signalling centres in the limb supporting the concept of local interactions operating throughout the limb field. In some cases these interactions contribute to the demonstrated activities of signalling centres, for example Wnt7a in dorsal ectoderm (Parr and McMahon, 1995) and from the data here perhaps Fzd4 (or other more widely expressed Fzd genes) in the polarising region. Others may be involved in local interactions superimposed on or integrated with signals from previously defined signalling centres.

This represents a core set of data that can be built upon through the addition of other interacting gene transcripts such as those encoding other Wnt pathway components (currently under assembly), components of pathways that cross regulate Wnt signalling and ultimately linking into wider networks. Compiled with a much larger set of data in this way, this work will contribute to finding patterns that identify networks and modules of activity so that such modules can be examined functionally in the mouse and compared across species to gain an understanding of the evolution of molecular modules. The present study also allows cross gene comparison and has generated a resource which is suitable for a 3D data base that can be searched spatially. 3D recording allows aspects of the pattern, such as a very localised spot or the peak of a complex gradient, to be revealed. Such features cannot be fully represented in 2D, for example the concentration of Fzd1 transcripts in the anterior limb mesenchyme and the shape and distribution of patches of Wnt2, Wnt4 and Fzd4 in the limb (Fig. 5), because they cannot be appreciated from single 2D sections. Another example of such a feature is seen in the external view of Wnt7a limb expression (Fig. 5, Column A) where a slightly anterior skew within the dorsal ectoderm domain is discernable in the distal region, a feature not detectable in any of the sections.

For the best OPT reconstruction data the intensity of colourometric stain should be moderate (as described in results). Also a low level of background staining of the tissue (so the tissue appears vaguely pink, Fig. 1) was found to be helpful in viewing OPT data captured in the visible channel alone as this allows the tissue context to be just visible when the full spectrum of grey level data is viewed (e.g. Fig. 2). The protocol was therefore optimised to produce such data. Staining components were diluted to 175 μg/ml 4-nitro blue tetrazolium chloride and 62.5 μg/ml 5-bromo-4-chloroindolyl-phosphate as a standard staining solution and staining was allowed to develop slowly with careful monitoring. For developing of very strong signals the above solution was diluted up to 1/10. Different intensities of staining were tested for each gene to ensure maximum capture of the data.

Embryos were physically sectioned (20–30 μm) using a Bright Model OTF Cryostat either prior to or following ISH. Embryos were fixed in 4% Paraformaldehyde overnight and embedded in 1.5% agarose, 5% sucrose. Trimmed blocks were equilibrated in 30% sucrose solution until they sank (usually overnight) and slowly frozen over a dry ice bath. About 20–30 μm sections were collected on BDH superfrost + slides and stored at −20 °C until hybridisation. In situ hybridisation to sections was carried out largely as described in Moorman et al. (2001) except proteinase K treatment was 10 μg/ml in 50 mM Tris, 5 mM EDTA for 5 min, the post-proteinase K fixation was in 4% paraformaldehyde, sections were dehydrated through an ethanol series prior to addition of probe, there was no pre-hybridisation and coverslips were used over the hybridisation solution, hybridisation was at 55 °C overnight, post-hybridisation washes were 2× 20–30 min in 50% formamide, 2× SSC, 65 °C; 3× 10 min in 2× SSC, 65 °C; 3× 10 min in 0.2× SSC, RT; 2× 10 min in TNT (100 mM Tris, pH 7.5, 150 mM NaCl, 1% Tween-20) at RT. The blocking solution for immunological detection of probe was Maleic acid buffer with 3% blocking powder (Roche).

This work was supported entirely by Science Foundation Ireland (Programme Award 02/IN1/B267). We would like to thank several members of the Edinburgh Mouse Atlas Project for helpful discussion, advice, software and assistance with its use, particularly Richard Baldock, Bill Hill and Peter Stevenson. We thank S. Venkataraman and Jeff Christiansen for help in entering the data in EMAGE. We thank Gareth Brady for technical assistance and Harris Morrison for help and advice in running OPT. We thank several researchers who provided plasmid DNA for the production of expression probes (Table 3).