Several public databases for "omics" data are available, which cover the expression profile of individual genes in various organisms to a different extent. For our study, we employed the UniGene database, which contains comprehensive coverage of various organisms and a collection of expression data. Each UniGene ID has an attached "Expression Profile", which can be used to search for specific expression sites. To identify genes in a systematic manner, which show a preferential expression in the heart but lack a proper functional annotation, we selected 4 different organisms for our analysis (rat, mouse, human, and chicken) (Figure 1). Applying the rule "Select genes whose rank for 'heart' is under the top 20% across homologues" resulted in the identification of 2,348 mouse genes, which fitted the rule in more than one organism. It should be pointed out that the combination of expression data from different organisms reduced the number of false-positive candidates. Since the expression profile of genes is normally conserved among evolutionary close species, any expression profile that is unique for a single species is likely to indicate faulty expression profiling rather than a biological meaningful result.

After identification of heart-enriched genes, we searched for genes in this group, which lack a proper annotation. We assumed that genes with no or very few publications are "uncharacterized genes". To avoid papers describing large screening efforts (e.g. microarrays, sequencing), which usually do not include a functional analysis, we excluded papers that contained more than 100 GeneIDs (GI). Obviously, this selection is rather strict resulting in exclusion of potentially interesting genes. To address this problem, we added another selection criterion that was based on Medical Subject Headings (MeSH). Inclusion of the MeSH term for "heart" with less than 2 publications resulted in the identification of 1,975 "uncharacterized" mouse genes, which show an expression in the heart. Figure 2 depicts the numbers of heart-enriched genes, which were identified using different combinations of organisms. Selection of genes based on the concurrent expression in the heart in all model organisms massively decreased the number of positive genes yielding 124 heart-enriched genes of which 79 are uncharacterized. Unfortunately, the quality of the expression data is not identical in all model organisms, which resulted in exclusion of genes, which have not been properly accessed but in fact are expressed in the heart. An improvement of the quality of the expression data will reduce the number of false-negative genes in the future.

To demonstrate the efficiency of our method, we searched the Gene Ontology project 24 for the term "heart" and matched the results with our data set that was based on the analysis of the expression data including genes, which lack an annotation. We used 79 GO terms that contain the phrase "heart" and screened the gene names and symbols of the 4 selected model organisms. We found that 32 out of the 79 GO terms contained at least one gene in the organisms studied. The number of genes that were obtained through our method and the coverage of GO terms in percentages of the genes that matched our selection criteria are shown in Additional File 1. For example, about 30% of the genes that carry the GO term "heart development (GO:0007507)" were covered by our method (34% for mouse, 28% for rat, 37% for human, and 37% for chicken). The information about all 43 mouse genes that were extracted through our method using the GO term "heart development" is listed in Additional File 2.

To validate the data obtained by our selection rules and to confirm the database records, RT-PCR experiments using mouse adult organs were performed. Surprisingly, we found that a large number of genes were expressed in multiple organs (Figure 3A). We therefore devised a set of additional rules to exclude ubiquitously expressed genes: (i) Genes must be expressed in less than 50% of total tissues; (ii) Sum of Gene EST must be less than 400; and (iii) At least 25% of Gene EST must be expressed in the target tissue (e.g. heart). Application of these rules efficiently eliminated ubiquitously expressed genes and left genes that were expressed preferentially in the heart (Figure 4B). Interestingly, most of the genes that remained after this selection were also expressed in striated muscles. This result was not unexpected given the fact that skeletal muscles and the heart are both composed of striated muscle cells, which share numerous morphological and physiological properties. Table 1 lists the number of mouse genes that were identified by using the combination of our selection rules. Inclusion of all three additional selection rules left only 30 mouse genes while inclusion of selection rules (i) and (ii) left 920 genes. As shown in Figure 4B, inclusion of selection rules (i) and (ii) did already efficiently exclude ubiquitously expressed genes. Addition of the third selection rule did not improve massively the number of genes with a heart-enriched expression pattern but result in a general decrease of selected genes. The inclusion of the third selection criterion proved to be helpful only for some cases (data not shown).

The expression in the adult heart does not necessarily mean that the corresponding gene is also expressed during early heart development. We therefore chose 20 genes from the genes that emerged from the computational screening and subjected them to whole-mount in situ hybridization (WISH) using E10.5 and E11.5 mouse embryos. At this stage, most organs have either formed or undergo organogenesis. 18 out of the 20 genes showed expression in the heart while 2 genes failed to display a specific expression at these stages (Figure 4 and data not shown). Based on the expression pattern, we distinguished two different groups: Group 1 showed equally strong expression in the heart both at E10.5 and E11.5 (4632412N22Rik, A5530016L24Rik, Klhl31, Fsd2, Lrrc39, and Rnf207); and Group 2 showed an increase of expression at E11.5 compared to E10.5 (Fhod3 and Lmod3).

To obtain initial functional data for the selected genes, we turned to a loss-of-function analysis. Two uncharacterized, heart-enriched genes were chosen: leucine rich repeat containing 39 (Lrrc39) and kelch-like 31 (Drosophila) (Klhl31). Lrrc39 is highly expressed in the adult heart based on the RT-PCR analysis, and also found in eyes, skeletal muscle, brown fat, and weakly in the stomach (Figure 3B). During development, Lrrc39 is expressed at E10.5 in the heart, in somites and in the eye anlagen (Figure 4). Klhl31 showed a slightly different expression pattern. It was expressed in the adult heart, brown fat, skeletal muscle, and testes (Figure 3B). Klhl31 was also detected very strongly in the heart during development at E10.5 by WISH (Figure 4).

We decided to accomplish the loss-of-function analysis in zebrafish embryos because knock-down of genes is relatively easily achieved in zebrafish by injection of antisense morpholinos. In addition, the development of the cardiovascular system in zebrafish embryos can be conveniently monitored by microscopical inspections. The zebrafish homologues of Lrrc39 and Klhl31 were identified using HomoloGene 25 yielding "zgc:112088" and "klhl31" (which are indicated as "zLrrc39" and "zKlhl31" respectively). Morpholinos against "zLrrc39" and "zKlhl31" were directed at their ATG sites.

zLrrc39 morphants showed the first signs of disturbed development at 48 hour post fertilization (hpf). At 108 hpf, zLrrc39 morphants displayed a considerable pericardial edema, which is a characteristic sign of cardiac malfunctions (Figure 5A). All morphants died within 12 hours after this time point. To facilitate detection of developmental defects in the heart, we employed a transgenic zebrafish line "Tg(myl7:EGFP-HRAS)^s883" 26, in which myocardial cells were labeled by GFP. Macroscopical evaluation of injected and control embryos at 48 hpf followed by sectioning revealed a disturbed arrangement of cardiomyocytes in the hearts of zLrrc39 morphants, but not in control embryos, indicating a role of zLrrc39 in morphogenetic processes controlling heart development (Figure 5B, C). In addition, zLrrc39 morphants were characterized by small eyes, malformations of the lower jaw, and distorted muscle structures (Figure 5A). It was interesting to note that the affected organs reflected the expression profile of zLrrc39 during mouse development.

The knock-down of zKlhl31 resulted in a more severe phenotype compared to zLrrc39 essentially abrogating development at 72 hpf. At 48 hpf, zKlhl31 morphants showed cardiac and yolk sac edema as well as a distorted body axis (Figure 5D). Sectioning of the heart of zKlhl31 morphants revealed that ventricle morphogenesis was severely disturbed, which essentially prevented looping of the heart.

The data sources for all datasets used in this study are listed in Additional File 5 including the FTP sites, dates of data retrievals, and version numbers, if applicable. As a starting point, UniGene's "Mm.data" was used to search for the number of mouse GeneIDs (GI) with UniGene IDs. Next, HomoloGene's "homologene.data", was used to search for homologs and their UniGene IDs in rat (Rattus norvegicus; Taxonomy ID 10116), human (Homo sapiens; Taxonomy ID 9606), and chicken (Gallus gallus; Taxonomy ID 9031). The UniGene tissue expression profiles ("Mm.profiles" for mouse, "Rn.profiles" for rat, "Hs.profiles" for human, and "Gga.profiles" for chicken) were screened with the selection rule: "Select genes whose rank for 'heart' is under the top 20% across homologs." At the time of the screening, the number of tissues for each organism and the definition of the top 20% were as follows (organism name, number of tissues, number of tissues that fall under the top 20% category): Mouse, 47, 9; Rat, 25, 5; Human, 45, 9; and Chicken, 18, 4. Selected genes were subjected to the following further selection rules to eliminate ubiquitously expressed genes in mouse: (i) Expressed in less than 50% of total tissues. (ii) Sum of Gene EST must be less than 400. (iii) At least 25% of Gene EST must be expressed in the target tissue. A flowchart of the UniGene in silico screening procedure is given in Figure 1.

To find the number of publications related to an identified gene, Entrez Gene's "gene2pubmed" was used. First, the number of GIs that are listed under each PMID was counted and classified. Publications that include more than 100 GIs were considered as articles that report large screening results (e.g. microarrays, sequencing) and excluded to be counted for the number of publications for each gene. Next, PMIDs whose classification in Medical Subject Headings (MeSH) is under "heart" were searched and cross matched to the number of publications for each gene. If this number was less than 2, the GI (gene) was defined as being "uncharacterized".

Gene Ontology 24 terms for "heart" were searched through AmiGO 50. Next, using these GO terms, gene names and their symbols were extracted from Entrez Gene's "gene2go".

To demonstrate the statistical significance of GO coverage, Fisher's exact test was applied. First, numbers of genes with at least one GO term were computed from Entrez Gene's "gene2go", which resulted in 18,225 genes for mouse, 16,437 for rat, 17,905 for human, and 9,226 for chicken. Then, a Fisher's exact test was applied to 2,348 heart-enriched genes for each GO term and its coverage.

To extract total RNA, adult mice were sacrificed, and washed extensively with ice-cold PBS to remove blood. Isolated organ parts were frozen in liquid nitrogen and homogenized using TRIzol Reagent (Invitrogen). 2 μg of purified RNA was reverse transcribed using SuperScript II First-Strand Synthesis System (Invitrogen) to synthesize the first-strand cDNA by following the manufacturer's protocol. First-strand cDNA was diluted with water to generate a 1:5 dilution. 1 μl of cDNA was used for each PCR reaction. The following thermal profile was used for all PCR experiments: 95°C for 2 minutes, denaturation (94°C) for 30 seconds, annealing (60–64°C) for 30 seconds, extension (72°C) for 1 minute, followed by a final extension (72°C) for 7 minutes. Amplified PCR products were separated in 2% agarose gel and visualized using UV light.

To generate RNA probes, PCR amplified products were purified from the gel and cloned using pGEM-T Easy System (Promega). Antisense RNA probes were labeled with digoxigenin-UTP (Roche), using SP6 or T7 RNA polymerase (Promega). Embryos were fixed in 4% paraformaldehyde (PFA) dissolved in PBS/DEPC-treated water overnight and dehydrated the next day with 25%, 50%, 75%, and 100% methanol. For whole-mount in situ hybridizations, published protocols were used 51 with minor modifications. After staining, embryos were post-fixed in 4% PFA and stored at 4°C until photographs were taken.

The zebrafish strains used in this study were "Bad Nauheim (BNA)", which was originated from a local pet shop and was inbred for several generations, and "Tg(myl7:EGFP-HRAS)^s883" 25. Antisense morpholinos were purchased from BioCat (Clone ID MORPH 1769 for zLrrc39) and GeneTools (zKlhl31). For the morpholino experiments, 3 concentrations were tested (approximately 2–4 nl of morpholino diluted in sterile water/phenol red as an indicator per zebrafish egg): "1:2 dilution (high)", which corresponds to 8.5–17 ng of morpholino; "1:5 dilution (medium)", 3.4–6.8 ng; and "1:10 dilution (low)", 1.7–3.4 ng. For each injection, more than 50 eggs were injected. As a control, sterile water/phenol red without morpholino was used to observe effects caused by the injection.