Early Embryonic Gene Expression Profiling of Zebrafish Prion Protein (Prp2) Morphants

Background The Prion protein (PRNP/Prp) plays a crucial role in transmissible spongiform encephalopathies (TSEs) like Creutzfeldt-Jakob disease (CJD), scrapie and mad cow disease. Notwithstanding the importance in human and animal disease, fundamental aspects of PRNP/Prp function and transmission remains unaccounted for. Methodology/Principal Findings The zebrafish (Danio rerio) genome contains three Prp encoding genes assigned prp1, prp2 and prp3. Currently, the second paralogue is believed to be the most similar to the mammalian PRNP gene in structure and function. Functional studies of the PRNP gene ortholog was addressed by prp2 morpholino (MO) knockdown experiments. Investigation of Prp2 depleted embryos revealed high mortality and apoptosis at 24 hours post fertilization (hpf) as well as impaired brain and neuronal development. In order to elucidate the underlying mechanisms, a genome-wide transcriptome analysis was carried out in viable 24 hpf morphants. The resulting changes in gene expression profiles revealed 249 differently expressed genes linked to biological processes like cell death, neurogenesis and embryonic development. Conclusions/Significance The current study contributes to the understanding of basic Prp functions and demonstrates that the zebrafish is an excellent model to address the role of Prp in vertebrates. The gene knockdown of prp2 indicates an essential biological function for the zebrafish ortholog with a morphant phenotype that suggests a neurodegenerative action and gene expression effects which are apoptosis related and effects gene networks controlling neurogenesis and embryo development.


Introduction
The molecular mechanisms responsible for prion pathogenesis share similarities with a wider group of neurodegenerative disorders. The genetic tractability of zebrafish (Danio rerio) makes it a suitable model for addressing the fundamental molecular mechanisms underlying neurodegenerative disorders such as prion disease [1], Alzheimer disease [2] and Parkinson's disease [3].
In the prion theory of Prusiner [1] transmissible spongiform encephalopathies are proposed to be caused by a proteinaceous infectious agent, a pathological variant of the normal cellular prion protein, PRNP scrapie (PRNP Sc ; also named PrP Sc ), which acts as a template for the conversion of normal cellular prion protein PRNP C (also named PrP C ) to PRNP Sc . There are however, major unresolved issues in prion research relating to the more fundamental biological questions of the prion gene function and mechanisms for transmission of the disease. PRNP is a glycosylated glycosylphosphatidylinositol (GPI) anchored membrane protein, but its function, cellular distribution and polarized sorting in the cell remain to be definitively established. Also the functional relationship with its binding partner(s) remains unaccounted for. A list of PRNP C candidate binding partner proteins together with the major proposed functions for PRNP C itself has been summarized in a review by Westergard et al. [4], including protection against apoptosis and oxidative stress, copper ions uptake and binding, synaptic signal transduction and as linker to the extracellular matrix.
The PRNP gene is phylogenetically conserved, though teleost genomes harbor multiple prp paralogues due to extra genome duplication. Of the three zebrafish PRNP genes, prp3 is the more divergent variant [5], whereas prp1 and prp2 more closely resemble the mammalian variant with prp2 most likely being the closest ortholog [6]. Miesbauer et al. [7] demonstrated that PrP-related proteins from zebrafish are complexly glycosylated and contain a glycosylphosphatidylinositol anchor. It appears that the nomenclature for prion protein variants and the corresponding genes can be confusing (Table S1). In this paper we have chosen to use prp1, prp2, prp3 and Prp1, Prp2, Prp3 for the zebrafish genes and proteins respectively. For the mammalian homolog we denote the normal variant gene and protein PRNP and PRNP (PrP C ) respectively. The misfolded pathological PRNP scrapie variant is termed PRNP Sc (PrP Sc ).
Knockout mice lacking the neuronal cell surface PRNP protein show normal development and behavior [8]. On the contrary, prp1 mRNA knockdown in zebrafish results in a strong embryonal phenotype, characterized by absence of embryonic cell adhesion and arrested gastrulation [9]. Similarly, prp2 morphant zebrafish also developed a lethal phenotype but at a later larval stage at 7 days post-fertilization (dpf) [9]. In the same study it was also demonstrated that both zebrafish Prp and mouse PRNP mRNAs could rescue the knockdown phenotype, indicating an evolutionary conserved Prion protein function.
The apparent difference between the mouse gene knockout and zebrafish models suggests that important aspects of the Prion protein biology may be revealed in the zebrafish model, using morpholino (MO) mediated translation blocking [10]. Aiming at uncovering the function of zebrafish prp2 gene, two MOs (prp2-MO1 and prp2-MO2) for targeting the prp2 mRNA were designed and injected into one cell stage zebrafish embryos. The resulting phenotypes were characterized and microarray studies were conducted on 24 hpf zebrafish embryos microinjected with the most potent of the two MOs, prp2-MO2. In order to confirm the microarray data and demonstrate that both MOs showed the same specificity, quantitative real-time PCR (qRT-PCR) was performed on mRNA of selected genes from both prp2-MO1 and prp2-MO2 injected embryos.

Results
Optimization of the two MOs specific for the 59UTR of GenBank sequence AJ620614.1 (Fig. 1) injections revealed that 0.2 mM prp2-MO2 resulted in 71% mortality at 24 hpf and 74% at 48 hpf. In order to reduce the mortality, 0.15 mM prp2-MO2 was chosen because it led to a mortality rate below 50%. Phase contrast microscopy of Prp2 morphants revealed a clear phenotype with midbrain and hindbrain developmental defects as compared to the control group ( Fig. 2 B-D). In order to further characterize the phenotypic effects, wholemount immunofluorescence was carried out using antibodies against Prp2 and HNK-1. Fluorescence microscopy analysis of the wild type embryos at 24 hpf unveiled low but significant levels of Prp2 proteins in the telencephalon, eye and trigeminal ganglion (Fig. 3A). Investigation of HNK-1 showed strong staining to the trigeminal ganglion and neuronal extensions (Fig. 3B). In the morphants no Prp2 activity could be detected and the Zn12 antibody visualized altered trigeminal ganglion morphology and reduced number of peripheral neurons in the head (Fig. 3C). Also, nuclear staining with DAP1 revealed apoptotic cells with hyper-condensed nuclei in the trigeminal ganglion, eye, olfactory placode, and in the midbrainhindbrain area (Fig. 3D).
Microarray analysis using a 16k oligonucleotide library revealed that 249 genes were significantly differentially expressed in the prp2-MO2 samples (Table 1). IPA analysis identified and mapped 120 of the 249 genes as being genes with mammalian ortholog identifiers (Table S2). The remaining 129 genes were either zebrafish specific or yet non-annotated genes with hypothetical or unknown function (Table S3). The most upregulated gene was cyclin G1 (CCNG1) with a 4.2 fold change. Leucine rich repeat neuronal 1 gene (LPRN1) was the most downregulated gene with a minus 2.46 fold change. The majority of the differentially expressed genes were encoding enzymes or transcription regulators with their location in the cytoplasm or nucleus, respectively. The IPA organization of differentially expressed genes into biological function and disease clusters included ''cell death'', ''embryonic development'' and ''nervous system development and function'' (Fig. S1, Table 2). The ''cell death/apoptosis'' related gene cluster was the largest sub-cluster with a total of 32 genes, of which 5 were related to apoptosis of neural cells: TP53, TP73, and BAX (upregulated) and SNPR1 and SNCB (downregulated, Fig. 4A, Table S2). The cluster of ''embryonic development'' (Fig. S1, Table 2) contained 11 genes, of which IPA could join 9 to each other in a network of sub-clusters in which SHH was the only downregulated gene and BMP4 had the highest score of the upregulated (Fig. S2). Both SHH and TP53 were included in one of the sub-clusters directly involved in the developmental process of embryonic cell lines. Analysis of genes related to ''nervous system development and function'' (Fig. S1, Table 2) revealed 5 subfunctional clusters involved in the developmental processes. Among these clusters, development of brain, forebrain, neuroepithelial cells, neurons and the central nervous system (CNS) were present (Fig. 4B). Here TP73, AKTIP and BMP4 (upregulated) with SHH (downregulated) play a central role with TP53 acting on CNS and BAX on neuron development (Fig. 4B). In the cluster ''nervous system development and function'', IPA connected 13 of the differentially expressed genes to each other in a network of different sub-function clusters. The most prominent network was the sub-function cluster ''neurogenesis'', which contained TP53, BMP4, MEF2C and BAX (upregulated) together with SOX11, GFAP, ACHE and SHH (downregulated) (Fig. S2). Also, the ''differentiation of neuroglia'' and ''migration of neurons'' harbored 5 genes each.
In order to explicitly investigate whether the differentially expressed genes in the prp2-MO2 injected embryos could be directly associated with prion gene (PRNP/prp2) functions, a customized IPA network with prion and prion associated genes was generated and found to include 18 upregulated and 13 downregulated genes, where the mammalian prp2 ortholog PRNP has a direct relation to MMP2 and an indirect relation to BAX and TP53 (Fig. 5). However, PRNP (prp2) itself and some of the other key molecules of the predicted network are not a part of the differentially expressed genes submitted from the microarray dataset ( Fig. 5) but showed change in expression by qRT-PCR ( Fig. 6; See below and discussion for further clarification).
Validation of the microarray results by qRT-PCR showed a high Pearson's correlation coefficient (0.998) between the microarray and qRT-PCR data (Fig. 6). Further, the corresponding   qRT-PCR analysis of the prp2-MO1 morphant transcriptome, using the same 14 gene probe sets, resulted in a comparable pattern of gene expression as the above mentioned results (Pearson's correlation coefficient 0.92). Hence, all of the qRT-PCR expression data strongly support the assumption of a gene specific action by the two prp2 specific MOs used. In addition, no change in prp1 gene expression was observed by qRT-PCR analysis of the control and/or the two Prp2 morphants (data not shown).

Discussion
In the current study, we have targeted prp2 mRNA translation using two different targeting MOs, prp2-MO1and prp2-MO2. The former had a significantly weaker effect on mortality which may be attributed to its different binding site on the prp2 mRNA ( Fig. 1). A difference in prp2-MO induced mortality compared to what Malaga-Trillo reported may also be explained by different morpholinos [9]. Quantification of gene expression in the morphants prp2-MO1 and prp2-MO2 by qRT-PCR ( Fig. 6) implies that the observed changes in gene expression are reliable both with respect to technical quality of the microarray data per se and with respect to gene target specificity of the prp2-MO2 chosen for the microarray experiment. Investigations of the corresponding A) The number of significantly differentially expressed genes for prp2-MO2 injected 24hpf zebrafish embryos. B) The number of differentially expressed IPA mapped genes (mammalian homologs) and unmapped by IPA (no mammalian homologs found). Arrows represent up-and down-regulated genes respectively. doi:10.1371/journal.pone.0013573.t001 Table 2. List of IPA biological function clusters for differentially expressed genes in 24hpf prp2-MO2 injected zebrafish morphant embryos.  prp1 mRNA levels by qRT-PCR showed unaltered levels of this paralogue in the morphants, which suggests that the Prp1 and 2 proteins have gained different main functions in zebrafish. The lethal effect of prp2 knockdown is in agreement with the proposal of an apoptosis-protective role of the PRNP C protein [11], as also confirmed by hyperdense nuclei observed in DAP1 stained morphants (Fig. 3D). The 24 hpf prp2-MO2 morphant transcriptome (Table S2 and S3) also reveals a cluster of 39 genes involved in cell death, of which 32 were in the sub-cluster of apoptosis (Fig. S1and 4A). Cyclin G1 (CCNG1) the gene with the highest level of upregulated expression in our experiment is known to control TP53 and TP73 which are involved in cell cycle control and induction of apoptosis [12], two processes which are augmented in the present study. Furthermore, TP53 and TP73 together with S1PR, BAX and SNCB directly influence apoptosis of cortical neurons, neuroepithelial, neuroblastoma, neural   precursor and nervous tissue cells (Fig. 4A). The altered expressions of these genes with their role in apoptosis of neural cells [13,14] support a putative role for Prp2 in the control of apoptosis of neural cells in zebrafish development. For example, the SNCB protein have been reported to decrease apoptosis of TSM1 neurons [13] and S1PR-null mice show a dramatic increase in apoptosis with a decrease in mitosis in the developing nervous system [14]. Both the observed phenotype morphology (Fig. 2B and C) and impaired neurogenesis of the trigeminal ganglion and peripheral neurons ( Fig. 3C and D) of the morphants can be correlated to the observed changes in gene expression, clustered to functions associated with embryo development and development of brain, forebrain and central nervous system (Fig. 4B and S2). Two of these key molecules, sonic hedgehog (SHH) and bone morphogenetic protein 4 (BMP4), were both downregulated. SHH acts as a signal molecule in many tissues including the midline structures in the brain, spinal cord and thalamus by the zona limitans intrathalamica [15][16][17]. SHH also has a reported ability to act on axonal guidance [18] as well as on neurogenesis where it influences proliferation of neuroblasts and granule cell precursors [19,20] (Fig. S2). Similar to SHH, BMP4 is involved in many biological processes including negative control of neurogenesis in olfactory epithelial cells and in brain/forebrain development [21]. However, since both SHH and BMP4 are involved in many biological processes, exemplified by their presence in 11 and 9 respectively of the 14 IPA defined biological function clusters of this study (Table 2), further investigations are needed to pinpoint more specific effect-relationships. In addition to SHH and BMP4, several other genes associated with the processes of neurogenesis, SOX11, GFAP and ACHE, were downregulated in prp2 knockdown embryos (Fig. S2). SOX11 is suggested to function in the developing nervous system while lack of glial fibrillary acidic protein (GFAP) gene function in knock-out mice increases the outgrowth of axons from neurons of the spinal cord [22] (GFAP is central in an IPA cluster related to neurological disorders. See below and Fig. S3). ACHE has been reported to be essential for normal dendrite and axon formation in hippocampal neurons and may function in excitatory synapse development, plasticity and remodeling [22][23][24]. The present microarray and qRT-PCR analyses demonstrate an upregulation of the BAX gene, which is a pro-apoptotic protein required for neurotrophin-deprived neuronal apoptosis [25]. Overall, prp2 knockdown in zebrafish embryos suggests impaired neurogenesis ( Fig.'s 2, 3 and 4B) which is in concert with the proposed function of PRNP [26,27]. IPA PRNP network analysis was applied to the differentially expressed genes from the prp2-MO2 24 hpf transcriptome and revealed several direct and indirect connections to PRNP (Fig. 5). The postulated connection between PRNP and the observed upregulation of the pro apoptotic genes BAX and TP53 may indicate an apoptosis protective role of Prp2, as proposed by Aguzzi et al [11] and in agreement with the pronounced mortality effect observed in this study. The proposed IPA network includes further nine connections to PRPN, heparin and phosphatidylserine together with 6 genes that were not present in our dataset but together associated with over 30 of the differentially expressed genes (Fig. 5). With the microarray technology used (a 16k library for coverage of 20-25 k genes), it is obvious that not all gene products affected by the MO treatment are included in the presented data set. Next generation sequencing (RNA-seq) will provide a powerful tool which could be used to uncover a magnitude of more details from the complete transcriptome, including all transcripts, alternative splice isoform variants and microRNAs, to allow a deeper understanding of gene networks involved in prion disease.
It is conceivable that the used prp2-MO gene knockdown strategy used here mimics the loss of the normally folded prion protein (PRPN/Prp2), which is part of prion induced pathogenesis. It does however not resemble an increase in misfolded prion protein and its aggregation into high molecular weight fibers which characterize prion disease. In a study of sporadic Creutzfelt-Jakob disease (sCJD), 79 upregulated and 275 downregulated genes were identified [28]. The major alterations in sCJD brain samples included upregulation of the genes encoding immune and stress-response factors and elements involved in cell death and cell cycle, as well as prominent downregulation of genes encoding synaptic proteins. There was no obvious common pattern of differentially expressed genes between the respective clusters or gene lists and thus providing no evidence for that the loss of the normally folded prion protein function would be a major cause of prion induced pathogenesis [28]. Of 230 genes in the sCJD gene list, 4 are however common with the Prp2 morphants: GFAP, phosphoglycerate dehydrogenase (PHGDH), SWAP-70 protein (SWAP70) and TP53. SWAP70, PHGDH and TP53 are included in the IPA apoptosis cluster (Fig. 4A and B). When extending the comparisons between differentially expressed genes to include two mouse scrapie models [29,30] only 1 gene (GFAP) is in common for all 4 studies and two when comparing the 3 mammalian (GFAP, ABCA1). GFAP is proposed by IPA to be indirectly linked to PRNP and neurogenesis ( Fig. 5; S2). Comparing the prp2 morphant with the mouse scrapie model [29] revealed 5 genes in common: GFAP, CD9 antigen (CD9), solute carrier family 14 member 1 (SLC14A1), heat shock 22kDa protein 8 (HSPB8) and syndecan 4 (SDC4). An IPA cluster analysis could link all the 5 genes to neurological disorders (Fig. S3). CD9 and HSPB8 included in the IPA apoptosis cluster (Fig. 4A). When comparing each mouse study alone against the sCJD data [29,30] the overlapping gene numbers are 11-14 which is not far from the 4-5 for the zebrafish-sCJD or zebrafishmouse, taken into consideration the fact that only 120 of 249 of the zebrafish genes have known mammalian homologs (Table 1). With this background it is fair to state that the zebrafish prp2 morphants provide a good model to shed light on the normal biological function of the prion protein. Our gene expression study is based on depleting the mammalian prion protein ortholog from the developing zebrafish embryo and has been able to link the zebrafish prp2 gene to biological processes including cell death, embryo development and neurogenesis. Although not many genes overlap when comparing 3 mammalian scrapie related transcriptomes with the prp2 gene knockdown zebrafish transcriptome, it is interesting to see one gene which is linked to a variety of neurological disorders in common for all 4 studies (Fig. S3). Many genes associated with these processes have been uncovered but still a more distinct mechanism of action for Prp2 remains to be found. We conclude that the data from this study, together with the paper of Malaga-Trillo et al. [9], place zebrafish as a highly relevant model to address the still largely unanswered questions of the molecular mechanisms underlying prion protein function, aggregation and the corresponding disease pathology.

Zebrafish and prp2-MO microinjections
Newly fertilized embryos were obtained from zebrafish reared in Alestrom Zebrafish Lab http://zebrafish.no/ (http://zebrafish. no) according to husbandry standards as described in Alestrom Lab Standard Operating Procedures. As an AAALAC (http:// www.aaalac.org) accredited laboratory, all activities in the unit are subject for evaluation by an IACUC. In all of the experiments presented, only zebrafish embryos of age up to 24 hpf were used.
For titration of dose response, two MOs specific for the 59UTR of GenBank sequence AJ620614.1 (Fig. 1) were injected in single cell embryos at the following concentrations: 0.1, 0.15, 0.5 and 1mM. The mortality rate for negative control embryos (non-injected and mock injected with 16 Danieu solution) was below 20% (data not shown).

Method for immobilizing embryos and microscopy
Prp2 morphants larvae were embedded in CyGEL Sustain TM (Biostatus Limited, Leicestershire, UK) and investigated for phenotype using a Nikon AZ 100 fitted with Nikon Digital Sight DS Ri1 camera. Morphants and wild type embryo were fixed in 4% PFA and transferred to methanol for immunofluorescence.

Whole mount immunofluorescence
Wild type and morphant embryos were dechorinated and fixed in 4% PFA and then dehydrated in 70% methanol. Fluorescent immunohistochemistry was carried out on rehydrated and permeabilized embryos (10 min 1% Triton X-100, Sigma) after a 2 hr blocking in 5% dry milk dissolved in 16Phosphate Buffered Saline Tween 20 (PBST; 16 PBS, 0.1% Tween). Antibodies applied included the HNK-1 specific Zn12 [31] (Developmental studies Hybridoma Bank, IA, USA) and a rabbit serum against Prp2 [32]. The former was used at a 200 fold dilution in 2% dry milk with 1% DMSO in 16 PBST, whereas the latter was diluted 1006. After over night incubation at 4uC with the primary antibodies, the embryos were washed thoroughly in 16PBST and incubated with Alexa conjugated secondary antibodies diluted 2006 (Invitrogen, Ca, USA) for 2 hours. Final 16 PBST washes was carried out before glycerol mounting and microscopy. All images were captured using a Zeiss Axioplan Z1 fitted with the Apotome system for structured illumination microscopy. The images were post processed deconvolved and visualized in 3D using Zeiss Axiovison software.

RNA isolation
Total RNA from injected control (16 Danieau solution) and MO injected 24 hpf zebrafish embryo was isolated for microarray and qRT-PCR analysis. Briefly, pools of 50 embryos were treated with 1 mL of Trizol reagent (Invitrogen, CA, U.S.A). Homogenization was carried out using MagNA Lyser Beads (Roche Diagnostics Gmbh, Mannheim, Germany) and total RNA isolation was performed according to the TRIzol manufacturer (Invitrogen), followed by a 15 min DNaseI treatment (Qiagen Hilden, Germany) at 25uC. Further RNA purification was conducted using RNAeasy mini kit (Qiagen). After purification, the samples were eluted in 50 mL of RNase-free water and aliquoted for microarray and qRT-PCR analyses. RNA yield and integrity were determined using a NanoDrop ND-1000 instrument

Microarray strategy
The total RNA from zebrafish embryo morphant pools was compared to total RNA from Danieau-injected control pools. The replicate embryo pools were analyzed in two sets of dye-swaps making a total of four hybridizations. Normalized microarray datasets did not result in significantly altered gene expression for zebrafish housekeeping gene b-actin2 (172 probes in our arrays). Like other microarray gene expression profiling studies, we see a reproducible high correlation between microarray and qPCR data [33].

Data deposition
Microarray data have been submitted to the European Bioinformatics Institutes Array Express accessible through experiment accession number E-MEXP-2365.

Linear RNA amplification, target labeling and hybridization
One mg total RNA was linearly amplified and labeled, using an Amino Allyl MessageAmp TM II aRNA amplification kit (Ambion TX, U.S.A). Five mg of the resulting amplified RNA (aRNA) from the MO-treated and control groups were labeled either with Cy3-dUTP or Cy5-dUTP (GE Healthcare CT, U.S.A). The labeled targets were examined for amplification yield and incorporation efficiency by measuring the aRNA concentration at 260 nm, Cy3 incorporation at 550 nm and Cy5 incorporation at 650 nm using a Nanodrop ND-1000. One to 5 mg of each labeled target aRNA were mixed with 9 mL of 256 fragmentation buffer (Agilent Technologies) and the final volume adjusted to 225 mL with RNAse free H 2 O followed by incubation for 30 min at 60uC. The hybridization solution was prepared by adding 220.5 mL of 26 hybridization buffer (Agilent) and 4.5 mL (10 mg/mL) sonicated herring sperm DNA (Promega WI, U.S.A) to the labeled target aRNA. Microarray slides were pre-hybridized at 42uC, 60 min using 0.1% Bovine Serum Albumin Fraction V, 56SSC and 0.1% SDS. Hybridization was performed at 60uC for 16 hours using Agilent gasket slides G2534-60003, hybridization chamber and oven according to the manufacturer (Agilent Technologies). Microarray slides were then washed 36 5 min in 0.56 SSC and 0.01% SDS with the first wash at 42uC and the next two at room temperature. Finally, slides were washed 3 times at room temperature with 0.066 SSC and dried immediately by centrifuging at 1000 rpm for 1 minute.

Zebrafish oligonucleotide library and microarray construction
The Zebrafish OligoLibrary TM (Compugen, Ontario, Canada) used was composed of 16,399 (16k) 65-mer oligonucleotide probes, originally representing 16,228 clusters predicted to be representing unique genes plus 171 positive control probes. All probe sequences were selected from the 39 UTR in order to cover a maximum number of splice variants. The oligonucleotides were printed on CMT UltraGAPS slides at the Norwegian Microarray Consortium (NMC, www.microarray.no ) with a Microgrid II (BioRobotics, Oslo). Probe annotation and information about Unigene ID, Unigene description, Gene symbol, GO ontology and human and mouse homologue ID and descriptions were obtained with the Unigene build release, using the Genome institute of Singapore Unigene & Gene Ontology Annotation Tool Genome institute of Singapore (http://giscompute.gis.a-star.edu.sg/,govind/unigene_ db/). GenBank accession numbers were used as a query. In the present study annotations were made using Danio rerio Build #107-58.

Pathway analysis and functional profiling
For pathway and biological function analysis of significantly differentially expressed genes, the Ingenuity Pathway Analysis (IPA) program was used. The results were analyzed using the Unigene & Gene Ontology Annotation Tool for identification of mammalian (human, mouse and rat) orthologs before IPA software analysis. The output results comprise gene networks, pathways and functional clusters, where given scores and p-values are based on numbers of uploaded genes in the cluster or network and the size of cluster or network in the IPA knowledge database.
Fisher's exact test was used to determine the statistical significance and the likelihood of random clustering to a pathway. Scores of 2 or higher have at least a 99% likelihood of not being generated by chance alone. Mammalian orthologs from IPA are displayed using human homolog identifiers whereas zebrafish genes and proteins are annotated according to standard zebrafish nomenclature guidelines (Table S1).

Quantitative real-time PCR
The accuracy of the microarray results were validated by analyzing 14 differentially expressed genes by qRT-PCR using the same samples of total RNA. The intron spanning qRT-PCR primers were designed using ''Universal Probe Library Assay Design Center/ ProbeFinder version 2.40'' ((http://www.roche-applied-science. com/sis/rtpcr/upl/adc.jsp). Primer sequences are summarized in Table 3. Template cDNAs were produced from 1 mg total RNA using Superscript III Reverse Transcriptase according to manufacturer (Invitrogen). qRT-PCR analysis was carried out on a LightCycler 2.0 instrument using LightCyclerHFastStart DNA MasterPLUS SYBR Green I (Roche). All reactions were performed in triplicate with 0.5 mL of the respective primers (10 pmol) and 2.0 mL enzyme mix in a total reaction volume of 10 mL. The PCR conditions were: 95uC for 10min followed by 40 cycles of 10 s at 95uC, 10 s at 60uC and 6 s at 72uC. Melting curve analysis and agarose gel electrophoresis was carried out to confirm PCR products. Data analysis with crossing point (Cp) was determined by use of the maximum-second-derivative function in the LightCyclerHSoftware version 4.0 (Roche). Relative gene expression levels were obtained using beta-actin as the internal reference gene and efficiency correction with external standards. The mean gene expression values for each group (exposed versus control) were compared and p-values calculated using unpaired Student's t-test.

Data analysis and statistical methods
Microarray slides were scanned using a GenePix 4000B (Axon instrument, Foster City, CA, U.S.A) with threshold level just before saturation of several spots. Raw data generated from GenePix were imported into the Bioconductor package LIMMA and corrected for background. For within array and between array normalization, print tip Loess and scale was used. An empirical Bayes moderated t-test [34][35][36][37] was applied to detect differentially expressed genes across treated and control samples. P-values were corrected for multiple testing using Benjamini-Hochberg (BH) method [38] and p-values,0.05 were selected as differentially expressed genes. The generated gene list was further filtered for genes with low intensity and with small changes in expression. In the averaged normalized MA-Plot, the majority of genes were clustered between M-values of 60.4 (fold change 61.3) and selected to the threshold criteria for the differentially expressed gene list. Another threshold criterion was A-values,6, representing empty or weak intensity spots, which were excluded from the final gene list. Relative gene expression values (qRT-PCR) for each target gene were expressed as mean fold change obtained from 3 replicate samples. Correlation between microarray and qRT-PCR data for 14 genes was analyzed using Pearson productmoment correlation coefficient. Figure S1 IPA generated functional clusters of differentially expressed genes. IPA generated functional clusters of differentially expressed genes in 24 hpf zebrafish prp2 morphant (prp2-MO2) embryos. All clusters have p-values,0.05 and are ranked by decreasing p-values. Corresponding p-values are illustrated in Table 3. The numbers in the pie diagram corresponds to the numbers of genes with significantly different expression in each cluster. The clusters appear in the same order in the diagram as in the ID pane.