From zebrafish heart jogging genes to mouse and human orthologs: using Gene Ontology to investigate mammalian heart development.

Generation of the list of zebrafish heart jogging genes

A list of 30 zebrafish genes that affect heart jogging was compiled using a variety of approaches. Twelve zebrafish proteins were identified as they were already annotated to the ‘heart jogging’ GO terms, the remaining 18 proteins were then identified using the ZFIN (http://zfin.org/) Site Search, with the search phrase 'heart jogging', and filtering using the 'Expression/Phenotypes' category. This search retrieves figures from papers that have ‘heart jogging’ in the figure legend, and thus are likely to be describing specific zebrafish genes (and proteins) involved in this process. Many of these genes had not yet been curated with GO terms. Each of the papers identified in this way were reviewed; of the 23 zebrafish genes identified in these papers five (Bmpr1aa, Tbx1, unm_hu119, unm_hu202, unm_hu304) were eliminated, as none of these papers provided experimental evidence for the involvement of these genes in heart jogging. This left 30 zebrafish proteins with strong evidence for a role in the heart jogging process (Table 1). The experimental evidence describing the association of each gene to the process of heart jogging was manually reviewed, to ensure consistent criteria were applied.

Generation of the list of human and mouse ‘jogging ortholog’ genes

The HUGO Gene Nomenclature Committee Comparison of Orthology Predictions (HCOP) search tool (http://www.genenames.org/cgi-bin/hcop.pl) was used to identify the closest possible human and mouse ortholog for each of the 30 zebrafish genes. HCOP displays predictions from 11 homology prediction tools, including EnsemblCompara, Homologene and Inparanoid³⁴. For all but one gene, southpaw, HCOP returned human or mouse homologs for the zebrafish genes. The lack of a close mammalian ortholog of southpaw was confirmed with a UCSC BLAT analysis against the human and mouse genomes³⁵. BLAST analysis³⁶ showed that the closest possible human and mouse homolog for the zebrafish southpaw gene was Nodal (33% identity). Indeed, both southpaw and nodal are specifically expressed in the left lateral plate mesoderm^5,37 and knockdown of murine Nodal in this region leads to a disruption of cardiac asymmetry, as does injection of southpaw morpholinos, suggesting a functional orthology between southpaw and Nodal^5,37. However a reciprocal HCOP search showed that the zebrafish genes nodal-related 1 and 2 are the closest orthologs of human NODAL. Hence we have not included a human or mouse ortholog for zebrafish southpaw (Table 1). Three pairs of zebrafish paralogs (bmpr2a/bmpr2b; foxj1a/foxj1b; nipbla/nipblb) have a single corresponding ortholog in human and mouse. Therefore, there are 26 human and 26 mouse orthologs to the 30 zebrafish genes identified as relevant to zebrafish heart jogging (Table 1).

Gene ontology annotation

The human ‘jogging ortholog’ genes were fully manually annotated, by an experienced GO curator³⁸. Individual PubMed queries were run for each gene using the approved human gene symbol and filtering on ‘human’. To achieve full annotation, all of the relevant publications (a total of 232) containing unique functional data for each gene were annotated, regardless of the specific biology described in each paper. This approach enabled consistent annotation of all experimental data relating to each gene, thus ensuring an unbiased overview of any common processes associated with these genes. In addition, the GO term ‘heart looping’ was associated with a ‘jogging ortholog’ human gene if dextrocardia or situs inversus totalis phenotypes had been associated with a mutation in the gene, in order to follow the generally agreed view that leftward heart looping will have resulted in these phenotypes².

Functional enrichment analysis

The Mouse Genome Informatics functional enrichment tool VLAD (VisuaL Annotation Display; http://proto.informatics.jax.org/prototypes/vlad-1.0.3/) was used to look for overrepresentation of GO terms in each gene list relative to the whole genome of the organism. The annotation datasets used for the analysis were zfin (4th March 2013), goa_human (5th March 2013) and mgi (7th March 2013) for the zebrafish, human and mouse analyses respectively, and the ontology dataset used was dated 10th March 2013. The query gene lists (as UniProt IDs) were pasted into the ‘Query Set’ field, the ‘Universe Set’ field was left blank (to specify all genes in species specific annotation file) and the ‘Display Settings’ options selected were ‘pruning threshold’:3 and ‘collapsing threshold’:6. No evidence codes were excluded from the analyses. For this analysis the total number of genes (universe set size) having annotations in the biological process ontology were 14,577, 30,441 and 24,813 for zebrafish, human and mouse respectively. In line with common practice, when using functional analysis tools, enriched GO terms with 1 or 2 associated query genes were excluded from the final results table.

Creation of an ‘early heart development’ mouse gene list

A list of 103 mouse genes likely to play a role in early heart development was created by combining gene lists derived from three sources: The Mouse Genome Informatics Mammalian Phenotype Ontology browser http://www.informatics.jax.org/searches/MP_form.shtml³⁹, the QuickGO browser http://www.ebi.ac.uk/QuickGO/⁴⁰ and the ‘jogging ortholog’ gene list described above (see Mousegenelist.csv in Data File). The Mammalian Phenotype Ontology browser was queried for genotypes annotated with the terms ‘abnormal direction of heart looping’, ‘situs inversus totalis’, ‘dextrocardia’ and ‘mesocardia’, creating a list of 180 genotypes with an associated gene. Due to the multiple phenotypes associated with each of these genotypes only 58 genes were identified through this approach, and of these only 5 overlap with the 26 ‘jogging ortholog’ genes. Thirty-five genes were identified by filtering on the GO term ‘determination of heart left/right asymmetry’ and its child terms, the evidence code IMP (Inferred by Mutant Phenotype), and the mouse taxon. Of these only two are also present in the ‘jogging ortholog’ gene lists and 11 are present in the phenotype gene list. Twenty-six mouse ‘jogging ortholog’ genes were added to this combined gene list, and any duplicated genes were removed.

Annotation of the zebrafish heart jogging genes and the human ‘jogging ortholog’ genes

Thirty zebrafish genes were annotated to the GO term ‘heart jogging’ or one of its child terms based on experimental data from the literature (Table 1). Human and mouse orthologs of these genes were identified, as described in the Methods section, resulting in a list of 26 mammalian ‘jogging orthologs’.

The human ‘jogging ortholog’ genes were then fully annotated with GO terms based on published experimental data. All manual annotations to the human, mouse and zebrafish genes can be visualized with the QuickGO Gene Ontology browser http://tinyurl.com/humanortholog, http://tinyurl.com/mouseortholog and http://tinyurl.com/zebrafishgenes.

Functional enrichment analysis

The zebrafish heart jogging gene list and the human and mouse ‘jogging ortholog’ gene lists were analysed using the VLAD enrichment tool. This identified 155 biological process GO terms that were significantly enriched in the zebrafish (see Human_data.csv in Data File), 431 in the human (see Human_data.csv in Data File) and 402 in the mouse (see Mouse_data.csv in Data File) gene lists. The enriched GO terms from all three species were grouped into five biological areas: Development, Patterning, Cellular Process, Signalling and Movement. The relative enrichment of key GO terms from each area was compared across all three species (see Biological_process_summary.csv in Data File; summarized in Table 2).

Enrichment of heart development terms. As expected there was a significant enrichment of developmental process terms in all three gene lists, including an enrichment of the GO term ‘heart development’. However, there was also enrichment of terms such as ‘renal system development’ and ‘nervous system development’, indicating the role of these proteins in regulating the development of a range of organ systems and tissues. These data analyses also show an enrichment of terms describing specific, but universal, cellular processes, such as signalling and regulation of transcription (Table 2). These terms represent essential aspects of development, but are grouped discretely due to their roles in many other biological processes.

‘Pattern specification’, described in GO as a ‘developmental process that results in the creation of defined areas or spaces within an organism to which cells respond and eventually are instructed to differentiate’ and several of its more specific child terms (such as ‘specification of symmetry’), were also enriched in all three gene lists. Within the symmetry ontology, the GO term ‘determination of heart left/right asymmetry’ is annotated to all 30 genes in the zebrafish jogging gene list, however, it is only associated with 8 and 4 jogging ortholog genes in the human and mouse respectively. Of the 97 zebrafish genes associated with ‘determination of heart left/right asymmetry’ 31% are also present in the zebrafish jogging gene list. In contrast, only 11% of the human and 8% of the mouse genes associated with this term are also ‘jogging orthologs’. These results confirm an overlap in the functional role of the zebrafish jogging genes and the human and mouse orthologs in the determination of heart left/right symmetry. However, this relatively low level of overlap may reflect the limitations of model organism and human research in this area, rather than a lack of functional conservation of these genes.

In addition, there were some differences in the developmental terms that were enriched between species. For example, the GO terms ‘vasculature development’ and ‘sensory organ development’ are enriched in both the human and mouse ‘jogging ortholog’ gene lists (Table 2), but neither of these processes are enriched in the zebrafish jogging ortholog genes. This difference may reflect the type of experiments zebrafish are used for, rather than reflecting a difference between zebrafish and mammals in the genes required for these developmental processes.

Enrichment of cilia terms. Terms in the cellular component organization or biogenesis ontology were enriched across all three gene lists (Table 2 and Biological_process_summary.csv in Data File). Specifically there was an enrichment of terms describing ‘cilium morphogenesis’ and ‘protein complex assembly’ (Figure 1). Within each of these, some more specific terms were enriched, for example the human and mouse ‘jogging ortholog’ gene lists were enriched for the term ‘axonemal dynein complex assembly’, whilst the zebrafish and human gene lists showed an enrichment of the term ‘cilium assembly’.

Figure 1. View of relationships between enriched terms from the cellular component organization or biogenesis ontology.

The GONUTs view of relationships between enriched terms from the cellular component organization or biogenesis ontology⁸⁴. The grey arrows are used where a term has an ‘is a’ relationship to its parent term, the blue arrows indicate a ‘part of’ relationship. The bars below each GO term indicates which of these terms are enriched in the zebrafish ‘jogging’ gene list, and the human and mouse ‘jogging ortholog’ gene lists.

Terms such as ‘regulation of cell projection organization’ were also enriched in the human and mouse ‘jogging ortholog’ gene lists. ‘Regulation’ terms have a ‘regulates’ relationship with the relevant processes; for example the term ‘positive regulation of cell projection organization’ has a ‘positively_regulates’ relationship to the term ‘cell projection organization’. In GO an important benefit of building a DAG, rather than a flat-list of controlled vocabulary terms, is that relationships can be used to make inferences from one term to another. However, the VLAD enrichment tool does not automatically create a transitive relationship between ‘regulation’ terms and the processes or functions they regulate. Consequently genes annotated to a ‘regulation’ term will not be associated with the regulated process term (unless there is an independent annotation to the process term). It is also important to recognise that it can be difficult for a curator to choose between annotating to the biological process itself, or to the term describing the regulation of that biological process, based on the published experimental data. Therefore, to get a full picture of the genes involved in a process, including the genes that regulate the process, it is necessary to combine the genes annotated to GO terms describing both the ‘process’ and the ‘regulation of the process’. For example, 10 zebrafish, 12 human and 14 mouse genes within the ‘jogging ortholog’ gene lists are annotated to either ‘cell projection organization’ or ‘regulation of cell projection organization’ (or children of these terms). This represents 33%, 40% and 47% of these zebrafish, human and mouse ‘jogging’ gene lists respectively, indicating that the process of cell projection organization is an important function for this group of genes. In addition, many of the ‘jogging’ and ‘jogging ortholog’ genes annotated to ‘cell projection organization’ terms have also been annotated to the cellular component term ‘cell projection’ (7, 11, 13, genes in zebrafish, human and mouse, respectively, see Data File). The enrichment of the biological process term ‘cell projection organization’ and cellular component term ‘cell projection’ within these gene lists is consistent with the key role of cilia located in the node/Kuppfer’s vesicle to determine heart left/right asymmetry in all three species.

Enrichment of cell migration terms. Cell migration also plays a key role in the establishment of the heart cone, heart jogging and heart looping^9,10 and enrichment of the GO term ‘cell migration’ is seen in the ‘jogging’ gene lists of all three species (Table 2 and Biological_process_summary.csv in Data File). Lenhart et al. (2013)⁹ identified FoxH1, spaw, Bmp4, Lefty2 and Has2 as essential to the asymmetric cell migration that leads to heart jogging. However, our literature review suggests that some genes may have functions in both cilia assembly, within the Kuppfer’s vesicle, and cell migration. For example, thymocytes from Foxj1 transgenic mice display defective migration⁴¹, whereas Foxj1-null mice are defective in ciliogenesis⁴². Similarly, in zebrafish, Fzd2 has been shown to play a role in cilium assembly²² as well as pancreatic insulin-cell migration⁴³. Consequently, further investigations into the role of these genes in heart jogging cell migration may provide further insight into this process.

Co-annotation of heart development associated genes

In order to investigate the contribution of individual genes in the multiple processes associated with early heart development we created human and mouse heart development gene lists and examined the associated GO biological processes terms. A list of 103 mouse genes with roles in early heart developmental processes was created by merging the three gene lists created using the Mouse Genome Informatics phenotype browser, the QuickGO browser as well as the ‘jogging ortholog’ gene list (Mousegenelist.csv in Data File).

GO captures a range of biological processes that a single gene is involved in. By comparing the overlap between the GO terms associated with specific gene lists it is possible to see what cellular mechanisms are likely to be contributing to the various heart developmental processes. Using the QuickGO browser, genes in the zebrafish ‘heart jogging’ gene list, which were associated with the GO terms ‘heart looping’, ‘signal transduction’, ‘cell migration’ and ‘cell projection organization’ (and all child terms, including ‘regulation’ terms), were downloaded, as well as the genes associated with these terms that were also present in the mouse ‘early heart development’ gene list (Mousegenelist.csv in Data File).

In the zebrafish ‘heart jogging’ gene list a similar proportion of the genes have the potential to play a role in cell projection organisation (10 genes), cell migration (8 genes) and signal transduction (13 genes) (Figure 2A). In the list of 103 mouse genes that are associated with early heart development, either by phenotype, annotation or homology to the zebrafish ‘heart jogging’ gene list, 82 have been annotated to the GO term heart looping. In contrast to the zebrafish ‘jogging’ gene list, signal transduction appears to play a major role in the mouse early heart development, with 27 genes associated with both signal transduction and heart looping, whereas only 18 and 9 genes, respectively, are associated with cell migration and cell projection organization (Figure 2B). These results fit well with what is known about these gene lists. The zebrafish ‘jogging’ gene list defines a group of genes whose functions are required very early in heart development, when the role of cilia in symmetry breaking initiates the heart jogging process. Whereas, in the mouse ‘early heart development’ gene list the genes included have roles in heart looping, which is developmentally later event than heart jogging. Therefore, although the initial events associated with breaking of left-right symmetry are represented within this gene list, the genes involved in the later process of ensuring the complex looping of the heart tube, through controlled signalling and cell migration, contribute to a large proportion of this list.

Figure 2. Venn diagrams describing the multiple roles of genes associated with heart development.

Venn diagrams showing the overlap between the GO terms associated with A) the zebrafish ‘heart jogging’ gene list (30 genes) and B) the mouse combined heart development gene list (103 genes).

Human disease phenotypes associated with the ‘jogging ortholog’ genes

While annotating the 26 human ‘jogging ortholog’ genes we noticed that almost half of these genes have not been associated with a specific disease phenotype (Table 3). However, of the 26 genes examined, mutations in 14 had been associated with a disease phenotype, a fifth of which were ciliopathies. Dextrocardia or situs inversus totalis (reversal or mirroring of the major visceral organs) was associated with 6 of the human ‘jogging ortholog’ genes. Location of the heart on the right side (rather than the left) is generally agreed to be the result of left-handed, instead of right-handed looping of the heart tube in early embryogenesis². The association of these ‘jogging ortholog’ genes with heart looping defects confirm that there is conserved functional homology between at least some of these orthologous zebrafish and human genes in the very early stages of heart development, which lead to the initial heart asymmetry. All four of ciliopathy-associated ‘jogging orthologs’ were also described as associated with situs inversus totalis, confirming the conserved role of these genes in the cilia within the symmetry determining left-right organizer.