Background

For over 500 million years, mollusks have successfully used a variety shells to populate the world over [1]. Due in part to the simple sheer prevalence of mollusks in past and present environments and their variety of shell formation strategies, these organisms represent the current model from which biomineralization is studied. Facilitating the shell formation in molluscs is the mantle organ. Phylum Mollusca is typically classified by an invertebrate unsegmented body, a mantle and a calcareous shell. The latter two are the subject of this investigation. The shell is internally lined by the mantle, composed of a thin sheath of tissue radiating out to the shell margins. In the case of a bivalve this organ is zootomically divided into two regions: the mantle pallial located proximal to the shell hinge, and the mantle edge situated distal to the hinge [2]. The distal mantle is further characterized by enlargement of the sheath at the shell margin into three terminal folds: the outer fold (OF), middle fold (MF), and the inner fold (IF). They are arranged such that OF is closest to the shell and the IF furthest. The main function of the mantle is recognized as the secretion of organic components necessary for shell biomineralization but it also has other purposes [3]. The mantle has a sensory function and can initiate closure of the valves in response to unfavourable environmental conditions [4]. In addition, the mantle also controls inflow of water into the shell's internal chamber responsible for respiratory and filter feeding purposes. These functions are said to be zone-specific in the greater mantle organ, referencing the IF as muscular, MF as sensory, and the OF as secretory in task [3]. Likewise the mantle edge and mantle pallial are considered principally secretory tissues.

To date, the secretory function of the mantle has been the focus of significant research with regard to biomineralization of the shell [1, 5, 6]. This is especially the case within pearl oyster species, considering pearl cultivation's reliance on mantle tissue. The pearl oyster shell typically consists of an outermost organic layer termed the periostracum, and calcium carbonate oriented in two distinct microlaminates, the outer calcite prismatic layer and the inner aragonite nacreous layer [1]. Evidence in the microstructure of both prismatic and nacreous layers has credited an organic framework as being central to the ordered mineralization [7, 8]. As such, the organic component has been the subject of much research devoted to its extraction and characterization [5]. Primarily these investigations have identified a number of matrix proteins, a subset of which have had their corresponding gene sequence determined. Some of those identified include: nacrein [9], MSI60/MSI31 [10], N66/N14 [11] prismalin-14 [12], and caspartin/calprismin [13]. However, many of the proteins remain to be identified, due in part to insolubility, self-aggregation of the molecules or an unusual resistance to temperature, chemicals and enzymes [5, 14]. More recently, alternative techniques to identify organic matrix proteins have been employed, including the use of expression cDNA libraries generated from mantle tissue screened with antibodies elicited from unfractionated organic matrix [15]. Although this technique has yielded positive identification of matrix proteins, it is largely inefficient and has meant the expense is inhibitory for most laboratories. Moreover, mantle tissue cDNA libraries have been screened with degenerate primers based on the signal peptide sequences of known proteins [11, 16, 20]. High-throughput sequencing and EST microarrays facilitate a comprehensive and inclusive experimental approach in which alterations in the state of entire transcriptomes can be simultaneously assayed. This technology has begun to be applied allowing the large scale investigation of gene products expressed in the mantle tissue with reference to biomineralization and other mantle-associated processes [21–23]. Although gene products identified may not necessarily be incorporated in the shell, this technique would circumvent the aforementioned technology limitations. Additionally it should be noted that a transcriptomic approach would not prejudice against gene products potentially involved in biomineralization but not integrated into the shell.

In order to expedite the elucidation of biological processes associated with the mantle organ this investigation has spatially mapped the differential expression of numerous expressed sequence tags (EST) derived from the mantle of P. maxima using the custom microarray chip PmaxArray 1.0 developed for this investigation.

Results

A Kruskal-Wallis test of the data generated from PmaxArray 1.0 was performed against the five experimental conditions. Outer fold (OF), middle fold (MF), inner fold (IF), ventral mantle (VM) and dorsal mantle (DM) comparisons identified 2012 ESTs of the total 4992 ESTs present on the microarray as statistically differentially expressed in reference to the experimental control (P < 0.001). Hierarchical cluster analysis of these 2012 ESTs grouped them according to similar expression profiles across the conditions. This analysis assisted the selection of four major expression profiles designated clusters A, B, D and E. A sub-cluster of B, termed cluster C was also selected. Clusters of interest were primarily selected based on the likelihood they would be informative in relation to biomineralization characterization. Cluster C was additionally selected due to its extreme difference in expression between the conditions from cluster B as indicated by the colour intensity (Figure 1). A subset of the 2012 ESTs, representing approximately 33% of the corresponding clones, were sequenced and batch blasted against BLASTx (Non-redundant protein sequences nr) and BLASTn (Nucleotide collection nr/nt) databases (Table 1). This subset of the total ESTs identified was deemed sufficient sequence coverage due to redundancy measurements. Many of the smaller cluster's ESTs were sequenced almost in entirety. Sequence alignment software resolved these microarray ESTs to 184 unique sequences. A number of ESTs were selected from each of the five clusters to determine specific local expression in the mantle (Figure 2). These selections were founded on several factors including whether they were either: novel and highly differentially expressed, or share significant homology with annotated genes of interest.

Figure 1
figure 1

Heat map displaying ~2000 P. maxima ESTs significantly differentially expressed among five discrete spatial regions of the mantle organ: inner fold, middle fold, outer fold, ventral mantle and dorsal mantle. ESTs are hierarchically clustered according to their spatial expression profile, the largest of which are labelled A to E. The scale of coloration from red to green indicates expression of the EST relative to the control (equal proportion of all conditions) such that green refers to greater relative expression in the control conditions while red signifies greater relative expression in a spatial treatment.

Full size image
Table 1 List of P. maxima ESTs associated with each cluster
Full size table
Figure 2
figure 2

In situ expression of 17 P. maxima ESTs differentially expressed among mantle regions. Panels are cross-sectional views of the mantle. Block arrows orient images with respect to the shell. DM = dorsal mantle, VM = ventral mantle, IF = inner fold, MF = middle fold, OF = outer fold. Expression is indicated in dark blue and arrow heads, alternative coloration is background. ESTs displaying similar localization are represented by a single example. (A-C) PM077, PM037 and PM041 [GH280037, GH279997, GH280001] expressed in the outer epithelium of the dorsal mantle, terminating immediately at the ventral mantle region. (D-E) PM316 [JG697411] expressed along the inner epithelium of the mantle and the distal outer epithelial of the middle fold. (F) PM317 [JG697412] expressed proximally in the outer epithelium of the middle fold and the inner epithelium of the outer fold. (G-H) PM315 [JG697410] is discontinuously expressed below the epithelium of the inner fold and middle fold. (I-J) PM233, PM234 and PM235 [GH280193, GH280194, GH280195] expressed along the inner epithelium of the outer fold. (K) PM241 [GH280201] expressed proximally in inner epithelium of the outer fold. (L) PM238 and PM239 [GH280198, GH280199] predominantly expressed mid-way along the inner epithelium of the outer fold. (M-O) PM273, PM268, PM280 and PM281 [GH280233, GH280228, GH280240, GH280241] expressed throughout the outer epithelium of the outer fold and ventral mantle, terminating immediately at dorsal mantle region. (P) PM265 [GH280225] expressed as described for (M-N) in addition to expression in the inner epithelium of the outer fold.

Full size image

Cluster A

Cluster A consisted of 225 microarray ESTs typical of the highest relative expression in DM, slightly less expression in VM, and low expression among OF, MF and IF (Figure 1). 197 of the total 225 ESTs were sequenced from which 52 unique sequences were resolved, 21 were contigs and 31 singletons. Putative sequence homologies could only be found for 13 of these ESTs including known shell matrix proteins N14 matrix protein and MSI60 protein. Other noteworthy matches identified are papilin, trypsin inhibitor protein, mantle gene 8 and calconectin (Table 1). A functional domain search of the ESTs significantly similar in sequence with papilin and the trypsin inhibitor protein both revealed tandem Kuntiz trypsin inhibitor domains. The majority of sequences identified from cluster A bear no significant similarity to sequences in public databases, furthermore many of the sequences aligned with poorly described genes and translated proteins. In situ hybridization was able to further resolve the localized expression for three ESTs, including; PM077, PM037 and PM041. These ESTs were chosen because they were among the most highly differentially expressed ESTs in cluster A and sequence similarity searches indicated they were novel. The three ESTs were all detected as expressed in the outer epithelium of the dorsal mantle region (Figure 2A-C). Of particular note is that expression of these ESTs is conspicuously absent at what appears the border of the ventral mantle zone and throughout this region.

Cluster B

Cluster B contained 871 ESTs detailing a relative expression profile as highly expressed in DM, MF, IF, no differential expression in VM, while lowly expressed in the OF in comparison to the control condition (Figure 1). 123 ESTs were randomly selected and sequenced. 68 unique sequences were detected of which 10 resolved as contigs and the remaining 58 singletons. This cluster is principally unannotated; however 15 ESTs are noted for significant sequence similarity to cellular maintenance proteins including: ferritin-like protein, ribosomal proteins, cytochrome oxidase subunits, glutathione peroxidise and radixin (Table 1). In situ hybridization was unable to precisely locate any of these sequences in the mantle tissue potentially due to diffuse expression of the target mRNAs impeding in situ resolution and/or transcript concentrations being outside the range of detection for the in situ hybridization protocol employed in this investigation.

Cluster C

Cluster C is a small sub-cluster of 22 ESTs within cluster B, characterized by relative high expression present in IF, MF compared to low expression in DM, VM, OF (Figure 1). All 22 ESTs were sequenced, condensing into three contigs and two singletons. Sequence analysis revealed no significant sequence similarity to sequences in the public databases. In situ hybridization revealed localized regions of expression of three of the ESTs. PM316 was localised to outer epithelial cells of the MF as well as the inner epithelial cells of the entire mantle organ (Figure 2D-E). PM317 was predominantly expressed in the outer and inner epithelial cells of the ventral sections of folds OF and MF respectively (Figure 2F). PM315 was found to be expressed sub-cutaneously in the IF and MF, specifically appearing interspersed among these regions (Figure 2G-H).

Cluster D

Cluster D is represented by 132 ESTs almost exclusively expressed in the OF mantle region (Figure 1). 129 ESTs were sequenced resolving as 21 unique sequences including 12 contigs and nine singletons. Approximately half of these ESTs show significant sequence homologies, the majority of which align with the family of lysine-rich matrix proteins (KRMP) (Figure 3). Additional matches include tyrosinase and thermoacin (Table 1). Alignment of KRMP deduced amino acid sequences with existing protein family members showed these ESTs were significantly divergent from P. fucata, P. margaritifera and P. maxima (cluster E) KRMP's particularly by a general absence of the C-terminal Gly/Tyr region. Of the cluster D homologs only PM244 did not align with all of the 6 cysteine residues present in the basic region (Figure 3). Local spatial expression of six cluster D ESTs was mapped to the mantle. PM233, PM234 and PM235 had similar patterns of expression, detected on the inner epithelium of the outer fold extending the length of the fold (Figure 2I-J). Conversely, PM241 is expressed only in the proximal most inner epithelial cells of the outer fold (Figure 2K), notably absent in expression of the three preceding ESTs (Figure 2J). PM238 and PM239 indicate a further difference in local expression, observed midway along the inner epithelium of the outer fold (Figure 2L).

Figure 3
figure 3

Alignment of the deduced amino acids of lysine-rich matrix protein family (KRMP). ESTs indentified from this investigation as KRMP similar are: PM277, 275, 276, 281, 280, 278, 258, 279, 260, 268, 233, 226, 244, 236, 234, 235 and 239 [GH280237, GH280235, GH280236, GH280241, GH280240, GH280238, GH280218, GH280239, GH280220, GH280228, GH280193, GH280186, GH280204, GH280196, GH280194, GH280195, GH280199]. Previously sequenced KRMP sequences used in the alignment were obtained from the Genbank database http://www.ncbi.nlm.nih.gov: KRMP-1, 2, 3, 4, 5, 6, 7, 8, 9, 10 and 11 [AAZ95763, AAZ95764, AAZ95765, ABO87297, ABO87298, ABO87299, ABP57445, ABP57446, ABP57447, ABP57448, ABP57449]. Sections A, B and C delimited by vertical lines denote the signal peptide, basic region and Gly/Tyr region respectively as set out by Zhang et al. [Cluster A

The spatial expression profile in cluster A suggests a role associated with the nacreous shell formation of P. maxima. Sudo et al.[10] along with others [26, 27] support this supposition noting a close spatial link between transcript expression in mantle zones and shell microstructure inclusion. Of particular interest within this cluster are PM077 and PM044, as both ESTs possess two tandem KUNTIZ/Bovine pancreatic trypsin domains (KUNTIZ BPT1). PM077 is a significant match to papilin; an extracellular matrix glycoprotein occurring widely from nematodes to humans and known to contain several KUNTIZ domains [28]. Likewise the presence of KUNTIZ domains is expected for PM044 which shares sequence similarity with a pancreatic trypsin inhibitor domain protein. KUNTIZ BPT1 domains are generally regarded as serine protease inhibitors involved in clotting and tissue remodeling [29]. Similarly shell formation is known to involve a number of inhibitory components limiting mineralization. Proteoglycans are one such component, essential to shell formation yet intrinsically inhibit biomineralization [30, 31]. The protease inhibiting domains of PM077 and PM044 may act to maintain the viscous silk gel detailed by Adaddi et al.[32] as necessary for nacre formation. PM077 is expressed in the DM epithelial cells overlying the nacre microstructure in conjunction with the immediate cessation of expression toward the VM zone and prismatic microstructure. Taken together, tissue localization and sequence homologies suggest that PM077 and possibly PM044 are glycoproteins with inhibitory protease activity specific for nacre formation.

ESTs PM037 and PM041 are unannotated however in situ hybridization demonstrated a very specific localization to the epithelium of the DM zone, as already described for PM077. This same distribution of expression is also demonstrated for N14 gene [11, 26] and MSI60 gene [10, 26] both of which code for nacre matrix proteins. The exclusive expression of these two novel ESTs, PM037 and PM041, suggest a role in nacre formation which along with PM077, are the only reported cases of in situ hybridization localizing ESTs to the DM zone since Sudo et al.[10] reported MSI60 gene expression.

Cluster B

This cluster is the largest and most ubiquitous of all the expression profiles identified in this study. ESTs in cluster B display similar expression values across a number of seemingly unrelated mantle tissues. The anatomy and function of the mantle organ is generally considered as follows: OF is secretory (periostracum and shell), MF is sensory, IF is muscular, VM and DM are secretory (shell) [3]. Therefore in the perceived absence of a specialized function uniting these tissues, cluster B most likely represents ESTs involved in general cellular maintenance and regulation rather than shell formation. This proposition is supported by the identification of a number of 'housekee**' genes (HKGs) not seen in any of the other clusters including cytochrome c oxidase, glutathione peroxidase, ezrin/radixin/moesin binding proteins and ribosomal proteins.

Cluster C

This cluster is the smallest and contains ESTs which showed no significant similarities with any reported protein or nucleotide sequences. The in situ hybridization results for ESTs PM317 and PM316 showed association with the periostracal groove in which the outer epithelium of the MF is included. The main function of the periostracal groove is to secrete a glycocalyx coating forming the periostracum. A glycocalyx is a network of polysaccharides that project from cellular surfaces usually secreted by epithelial cells for a range of adhesion functions. The distil expression of PM316 in the MF outer epithelium indicates this EST may code for a glycoprotein incorporated in the mature stages of the outer glycocalyx coating [33]. Similarly, expression of PM317 in the proximal epithelial cells of the periostracal groove may also code a protein involved in glycocalytic coatings and the stepwise construction of the periostracum.

EST PM315 has a peculiar in situ expression pattern in that the transcript is found below the epidermal layer, interspersed throughout the inner region of the MF and the outer region of the IF. Bivalves expose these mantle folds to the external environment [3]. Chemoreceptors, photoreceptors and mechanoreceptors are all usually present in the epidermal layer of these folds in order to elicit closure of the shell valves in response to negative stimulus [34]. Considering PM315 is expressed sub-dermally, it is less likely that this EST has a direct sensory role but rather associated with what appears to be nerve fibres [35], possibly involved in a signal transduction cascade [36].

Cluster D

The expression profile of cluster D ESTs suggests an exclusive role of these genes in the OF tissue, specifically concerned with the inner epithelia. This epithelium forms the bottom half of the periostracal groove, which is a highly dynamic tissue responsible for formation, maturation and extrusion of the complex periostracum layer. The proteinaceous layer functions to seal the extrapallial space and protect the shell from dissolution as well as serve as an initial matrix for mineralization [33]. In situ localization to the inner epithelium of the OF tissue signify a periostracum-related function.

Neuromacin [37] and theromacin [38] are a family of antimicrobial peptides known to occur in a number of invertebrates. These peptides are part of a immediate immune response characterized predominantly by cationic and hydrophobic amino acids [38]. EST PM238 shows a significant sequence similarity to the gene encoding these antimicrobial peptides and its in situ expression profile maps it to where the internal periostracum is formed. Cationic and hydrophobic properties of these peptides [38] are synonymous with the characteristics of the periostracum and water insoluble matrix (WISM) of shells [14, 39]. Specifically, a scenario for PM238 may be that poly-anionic glycoproteins (shell precursors) bind to cationic peptides in the periostracum, effectively anchoring the hydrophilic macromolecules to the hydrophobic WISM. This in turn facilitates active nucleation sites by which microstructure mineralization occurs.

Lysine-rich matrix protein (KRMP) is a family of proteins seemingly specific to mollusks and shell formation of the prismatic design. Cluster D includes seven ESTs significantly similar to the KRMP gene class. Zhang et al.[Microarray Experimental design

Nine animals were sourced and sacrificed as previously described. Mantle tissue from each animal was dissected under a stereomicroscope into outer fold (OF), middle fold (MF), inner fold (IF), ventral mantle tissue (VM) and dorsal mantle tissue (DM) (Figure 4). Animals selected for dissection had similar shell lengths measuring from the hinge to the opposing shell edge 15 cm (+/- 0.9 cm). This selection aided in standardizing the length of mantle tissue dissected. A strip of mantle tissue from the adductor attachment to the mantle edge along this shell growth axis was dissected from the animal and the fold tissues removed at their junctions. The ventral mantle tissue was dissected immediately adjacent to this junction measuring 0.5 cm in length and 1 cm in width. A buffer zone of 1.5 cm in length was used between the ventral mantle dissection and the following dorsal mantle dissection. The dorsal mantle tissue excised from the animal also measured 0.5 cm in length and 1 cm in width. Total RNA was extracted from these tissues as previously described above and pooled across subjects in order to reduce the effect of biological variation. The total number of subjects and arrays required for the pooled experiment to obtain gene expression estimates and confidence intervals comparable to those obtained from a non-pooled experiment is provided by the formula of Kendziorski et al.[51]. The use of nine subjects pooled across a total of three arrays provided the 90% confidence level required. To this effect, equal amounts of total RNA was pooled from the same tissue type from three individuals. This was repeated another two times, totalling nine animals in three separate pools. All the biologically-pooled tissue types were compared against a common reference in which total RNA from all tissues types and all nine animals was equally pooled. Technical variation, which is array-to-array variability, in these microarray experiments was addressed through spot duplication. Two identical grids consisting of each amplified cDNA and including the controls described above were printed onto the left and right sides of each horizontally-orientated array, thus affording spatial separation between duplicate spots, to allow for the normalization of potential hybridization anomalies. As there were five different tissues under investigation, each of which are biologically replicated three times, fifteen PmaxArray 1.0 slides were consumed. Furthermore each slide has a duplicate technical replicate bringing the final total to 30 arrays for the investigation.

Figure 4
figure 4

Schematic diagram depicting a cross-sectional aspect of the P. maxima mantle organ and shell in situ. Mantle tissues used for the comparative differential gene expression analysis are labelled accordingly: inner fold (IF), middle fold (MF), outer fold (OF), ventral mantle (VM) and dorsal mantle (DM). Red lines denote the approximate position that dissections were made separating the five tissues used in the microarray comparative spatial analysis.

Full size image

Microarray hybridization

1 μg of Lucidea universal RNA control (GE Healthcare) was added to 2 μg of pooled total RNA for each tissue type as well as the common reference. The RNA was converted to cDNA then labelled and hybridized to the array using the 3DNA Array 900 MPX expression array detection kit (Genisphere Inc., Hatfield, PA, USA) according to the manufacturer's protocol. Briefly, RNA was reverse transcribed using a random primer combined with an oligo-dT primer. The RNA was then degraded and the cDNA tailed with dTTP followed by ligation to a dendrimer-specific capture oligo (specific for either Cy3 or Cy5). Microarray slides were denatured prior to use by immersion in 95°C MilliQ water for five minutes; the slides were then transferred to 95% ethanol at room temperature for two minutes. Slides were spun dry to reduce streaking at 800 RPM for 2 minutes. The Cy3 and Cy5 "tagged" cDNAs were combined and then hybridized to the array by overnight incubation in a humidity chamber at 65°C using the kit supplied SDS-based buffer and a poly-T-based blocker, as per manufacturer's specifications. The "tagged" cDNA was washed with a series of three SSC-based buffers; the first wash occurred at 65°C for 15 minutes, the other wash steps were carried out at room temperature for 10 minutes each. The slides were spun dry at 800 RPM for two minutes. Fluorescent 3DNA capture reagent (which carries a sequence complementary to the Cy3 and Cy5 tag) was then hybridized to the array using the SDS-based buffer with added Anti-Fade reagent at 65°C for four hours. The fluorescent reagent was then washed as described above for the cDNA hybridization.

Data Analysis

PmaxArray 1.0 slides were scanned using a Genepix 4000B scanner (Axon Instruments, Union City, CA, USA) at 10 μm pixel resolution. ImaGene (BioDiscovery Inc., El Segundo, CA, USA) was used to process raw scanner images and create spot intensity reports, while CloneTracker (Biodiscovery Inc.) generated gene ID map** files and assigned gene identification. Final intensity reports were retrieved as raw spot intensities in tab-delimited files. The data set is deposited in the Gene Expression Omnibus (GEO) database [GSE14303] at the following site: http://www.ncbi.nlm.nih.gov/geo/. Spot intensity reports were imported into data mining software, GeneSight 3.0 (BioDiscovery Inc., El Segundo, CA, USA). Briefly, data was pre-processed and normalized in the following sequence, applying background correction, omitting multiple flagged spots, applying floor correction, omitting low expression spots, calculating ratio values, log-transformation of intensity ratios (base 2), and global LOESS normalization. Ratio data was not normally distributed thus statistical significance among the five tissues were analyzed with a non-parametric, univariate, Kruskal-Wallis test (P < 0.001). Hierarchical cluster analysis was performed among tissues and genes with the Euclidean distance coefficient as distance measure and average linkage.

Sequence Analysis

ESTs from the PmaxArray 1.0 identified as significantly significant (P < 0.001) and representing a cluster of interest, were single pass sequenced from their corresponding clones as detailed previously. Vector and poor quality portions of sequence were trimmed and clustered by sequence alignment into singletons and contigs using Sequencher (Gene Codes Corporation, Ann Arbor, MI, USA). These sequences were compared against public protein and nucleotide databases using the BLASTx and BLASTn tools [52] (E value cut-off < 0.01) from the National Center for Biotechnology Information http://blast.ncbi.nlm.nih.gov/Blast.cgi. Where appropriate sequences were analyzed for protein domains, searched against the Pfam database [53] supplied by the Sanger Institute http://pfam.sanger.ac.uk. Deduced amino acid alignments were performed using the ClustalW tool [54] from the European Bioinformatics Institute http://www.ebi.ac.uk. Signal peptides were predicted for sequences using the Signal P 3.0 program [55] from the Center for Biological Sequence Analysis http://www.cbs.dtu.dk/services/SignalP.

In situ hybridization

RNA anti-sense and sense probes (~400 bp) were generated first by PCR amplifying the EST of interest from cDNA clones using gene-specific primers with T7 and SP6 recognition sequences flanking the 5' end of the primers. 1 μg of the cDNA probe was added to digoxigenin (DIG) RNA-labelling mix (Roche, Penzberg, Germany) as per manufacturer's recommendations for DIG incorporated RNA synthesis. Probes unable to be labelled with DIG were synthesized into unmodified RNA first then non-enzymatically labelled with fluorescein via the Platinum Bright Nucleic Acid Labelling Kit (Kreatech, Amsterdam, Netherlands) according to the manufacturer's protocol. All probes were tested for labelling efficiency using a dot blot technique [50] with the appropriate antibody coupled to alkaline phosphatase.

Mantle tissue was removed from adult P. maxima and immediately fixed in 4% paraformaldehyde for four hours. Fixed tissue was dehydrated through an alcohol series and paraffin wax-embedded. Tissue blocks were sectioned to 7 μm increments. Sections were dewaxed in xylene and rehydrated in an alcohol series in preparation for RNA in situ hybridization. The technique used RNase-free reagents as described by Braissant and Wahli [56] with some modification. Briefly, rehydrated tissues underwent a 2 × 15 minutes wash in PBS with 0.1% active DEPC; 15 minutes equilibration in 5 × SSC; pre-hybridization, two hours at 50°C, in 50% formamide, 5 × SSC, 40 μg/ml salmon sperm DNA; hybridization 4-40 hours at 50°C, with 400 ng/mL of DIG/FLU labelled probe, in 50% formamide, 5 × SSC, 40 μg/mL salmon sperm DNA; washed 30 minutes in 2 × SSC at room temperature; one hour in 2 × SSC at 60°C; one hour in 0.1 × SSC at 60°C; five minutes equilibration in buffer 1 (Tris 100 mM/NaCl 150 mM, pH 7.5); two hours with anti-DIG/FLU antibody, AP-coupled, diluted 1:1000 in buffer 2 [buffer 1 with 0.5% of Blocking Solution (Roche, Penzberg, Germany)]; washed for 2 × 15 minutes in buffer 1; five minutes equilibration in buffer 3 (Tris 100 mM/NaCl 100 mM, pH 9.5); stained overnight in buffer 3 containing 20 μl NBT/BCIP Stock Solution (Roche, Penzberg, Germany); washed in running tap water for 15 minutes; dehydrated in alcohol series; washed in 95% ethanol for three hours; after which slides were mounted with cover slips.