To identify mucins we have used a method (PTSpred) to predict PTS/mucin domains where we analyzed both predicted proteins as well as genomic sequences translated in all six possible reading frames. We have previously applied this method to analyze the puffer fish Fugu rubripes 2. We have here used that method to analyze human and chicken proteins as well as genomic sequences. We have also taken advantage of the fact that all mucins (MUC7 and MUC20 excluded) contain either von Willebrand D (VWD) or SEA domains. Thus, we have analyzed proteins with Pfam models of the VWD and SEA domains and Genewise 24 was used to screen genomic sequences using the same models.

We considered a protein sequence to be a potential mucin if it contained at least one PTS domain as well as a VWD or SEA domain. Such candidates were further evaluated by phylogenetic analyses of SEA/VWD domains. A protein was considered a strong mucin candidate only in case the phylogenetic analysis supported a relationship between its SEA or VWD domain(s) to those of previously characterized mammalian mucins.

To test the efficiency of our computational methods we first analyzed available human protein sequences as well as the human genome assembly (for details see Materials and Methods). All human mucins, except MUC7 and MUC20, contain PTS domains as well as either VWD or SEA domains. In summary, our methods successfully identified all of these previously known mucins. In addition, we identified MUC19 1525 that was not known at the time we carried out this work. With the help of EST, mRNA, protein and genome sequences we were also able to reconstruct more complete and accurate human mucin protein sequences and elucidate gene structures (T. Lang et al., unpublished). These results show that our computational methods are reliable in terms of mucin gene predictions and that rigorous analysis of available sequence information is necessary in order to derive reliable predictions of gene and protein sequences.

We have now analyzed the chicken genome for mucin genes making use of the assembled genomic sequence and the proteins predicted by ENSEMBL 26. The genome assembly used in this work is expected to be approximately 90% complete. We have used methods described above for screening of the human and F. rubripes genomes. Most of the VWD and SEA domains identified in searches with hmmer and Genewise could be attributed either to mucins or to other previously known human proteins containing these domains. The predicted chicken mucin genes were characterized by a variety of bioinformatics tools and comparisons with known mucin genes and proteins from other species. For instance, all the sequences of the known mucin genes were aligned to the chicken genome. In this way, we could not only identify the human homologues, but also obtain a more complete sequence and understanding of the predicted chicken mucin. For more information on our current assembly of chicken mucins genes, including a comparison to the ENSEMBL predictions, the reader is referred to our mucin web site 27. A summary of our current inventory of mucins in man, mouse, chicken and the fish Fugu rubripes is shown in Table 1.

A total of eight strong mucin candidates in chicken were identified, all with PTS domains and with either VWD (5) or SEA (3) domains. Analysis of the proteins with VWD domains revealed Muc2, Muc5ac, Muc5b, and Muc6 genes that are located in a cluster on chicken chromosome 5 and are discussed in more detail below. In addition we found a homologue to Muc4 on chromosome 9. We observed that a protein predicted by ENSEMBL contained the major portion of Muc4, including a part of the PTS domain followed by AMOP, VWD, EGF1, EGF2, and TM domains characteristic for the human MUC4 6. The missing N-terminal part, including a signal sequence and the major part of the PTS domain, was reconstructed from the genome sequence. The resulting protein sequence is partially identical to a protein previously described as Muc4-related (Genbank: XP_426704.1) 22. The VWD domain of the human MUC4 is unusual in that it lacks cysteines, and this is also true for the chicken Muc4. The sequences of the human and chicken VWD domains are similar (Fig. 1) and taken together our information about the chicken protein strongly supports an orthologous relationship to human MUC4.

An analysis of predicted proteins with SEA domains identified chicken Muc13 and Muc16 homologues as well as a weak Muc1 candidate. Muc13 is described further below. The chicken Muc16 protein, previously identified by Duraisamy et al 22, is encoded on chromosome 28 and has a PTS domain followed by at least four SEA domains. The assignment as Muc16 based on phylogenetic analysis (Fig. 2) agrees with previous results 22 and is also consistent with the fact that human MUC16 is the only mucin known to have multiple SEA domains 8.

Two different proteins with SEA domains related to human MUC1 were identified. One of these were previously analyzed and it was proposed that it is more closely related to a heparin sulfate proteoglycan than to mammalian Muc1 22. The other MUC1-related protein was analyzed here. However, it did not convincingly cluster with the SEA domains of other Muc1 proteins ('unknown' in Fig. 2). Furthermore, the SEA domain of this protein is preceded by a PTS domain, but a transmembrane domain characteristic of MUC1 could not be identified. Finally, the N-terminal region of this putative mucin gene cannot be analyzed due to a gap in the genomic sequence. Therefore, it is not possible at this stage to predict the existence of a chicken Muc1.

We expect the predicted genes to be bona fide mucin genes because of the strong similarity to mucins from other species with respect to protein sequence, protein domain structure as well as gene structure. In general, it is difficult to distinguish between bona fide genes and pseudogenes. However, an analysis of available chicken ESTs provides evidence of expression for a majority of mucins genes that we have identified (Table 1). Thus, only in the case of Muc4, Muc16 and for the protein distantly related to Muc1 we were not able to find a corresponding EST sequence. The absence of EST support is not conclusive, as the available chicken EST data is not expected to be comprehensive.

At the same time it must be pointed out that there are limitations to our approach. We are not able to effectively identify mucins that are lacking VWD and SEA domains, mainly because that PTSpred will give rise to a number of false positive sequences. In addition, we might fail to detect mucin candidates because genome assemblies are incomplete, particularly with respect to mucin genes, and because of limitations in gene prediction procedures.

Five VWD-containing proteins were found within a region of chromosome 5, covering 12 million bases. The domain structure of the proteins on chromosome 5 suggested that this region has an organization similar to the human MUC2/5AC/5B/6 gene cluster. The relative gene order and polarity was identical to the corresponding human mucins as shown in Fig. 3. Thus, the Muc6 mucin is positioned next to and in the opposite direction to Muc2, Muc5ac and Muc5b.

The domain structures of the individual chicken gel-forming mucins were analyzed and the results are shown in Fig. 4. Typically, these mucins have three VWD domains followed by alternating PTS and CysD domains, and at the C-terminal end a cysteine-knot (CK) domain. The Muc2 and ovomucin proteins have an additional VWD domain. The chicken Muc2 ortholog was identified as the protein most similar to the human MUC2. However, the central part of the predicted molecule contained at least three CysD and four PTS domains, whereas the human protein only has two CysD and two PTS domains. A gap in the genomic sequence precludes further comparison and a conclusion as to the differences. The chicken Muc5ac and Muc5b proteins have a similar domain structure with central repeated CysD and PTS domains as in the human orthologues. However, chicken Muc5B lacks the C-terminal VWD domain in contrast to the human orthologue. The chicken Muc5ac genomic sequence has a large gap in its 3'-end preventing further comparison. Also for Muc6 the domain structure is identical to the human orthologue (Fig. 4), but a gap in the 3' genomic sequence makes it impossible to compare this region.

All VWD domains identified by a screen of the chicken genome with Genewise were also compared to previously known VWD domains using BLAST and ClustalW. The phylogenetic tree from a ClustalW analysis is shown in Fig. 1. Interestingly, all the mucin VWD domains are clustered in a characteristic manner based on their position in the mucins as we have previously shown for Fugu rubripes mucins 2. The different VWD domains numbered 1–4 in Fig. 1 are clearly homologous such that the chicken VWD-1 is most closely related to the human VWD-1, etc. The grouping of the human and chicken VWD domains strongly supports our assignment of the chicken mucins as Muc2, Muc5ac, Muc5b and Muc6, respectively.

When the chromosome 5 locus of chicken is compared to the corresponding locus in human, the most notable difference is the presence of one additional gene in the chicken. The predicted protein contains four VWD domains organized as for the gel-forming mucins (Fig. 4). This protein was recently cloned by Watanabe et al 23 and referred to as the α-subunit of ovomucin. An additional subunit called β has also been described 23. Both subunits are abundant in egg white and are responsible for its gel-like properties. However, from the results presented here it is obvious that the β-subunit of ovomucin is an orthologue to the human MUC6. In the following, the α-subunit of ovomucin is simply referred to as ovomucin as this protein is specific to the chicken mucin locus. Ovomucin has a similar domain structure as the other proteins in the cluster except that it does not contain the PTS domain characteristic of mucins (Fig. 4).

Interestingly, from the phylogenetic tree in Fig. 1 it seems that the VWD domains of ovomucin are more deeply branched than Muc2/5ac/5b, suggesting that this ovomucin is a more ancient protein. It will be interesting to further study this issue by identifying homologues of the gel-forming mucins in other species. Preliminary results suggest that there are ovomucin homologues in X. tropicalis and in the fishes F. rubripes, T. nigroviridis and D. rerio (zebrafish). However, ovomucin is not present in man and rodents. The tree in Fig. 1 also seems to suggest that Muc6 is more deeply branched than the other gel-forming mucins of the same locus, raising the possibility that this protein is the ancestral form of the Muc2/5ac/5b/Muc6 proteins.

A gene encoding the chicken Muc13 orthologue was identified on chromosome 7. The protein has an N-terminal signal sequence, followed by one PTS, one SEA, one transmembrane domain and a cytoplasmic tail. There is a gap in the genome assembly encoding the PTS domain and therefore the full sequence of this domain cannot be predicted. However, the PTS domain is composed of at least 12 repeats, each 20 amino acids in length (Fig. 5).

Typical for the PTS domains of previously known mucins are that these are built from tandem repeats that often show a remarkable length polymorphism (VNTR, variable number of tandem repeats) 128. The mechanism and functional significance of this variability in length is currently not known, but there are several indications that such variation is associated with disease. The allele length of MUC1 has been linked to susceptibility to Helicobacter pylori infection and gastric cancer 2930. Furthermore, it has recently been suggested that the allele length of MUC1 influences the expression of tumor associated carbohydrate antigens and possibly also the aggressiveness of gastric cancer 31.

For all previously described mucins, including the human MUC13, the PTS domain is found within a single large exon. However, the chicken Muc13 PTS domain is encoded by multiple exons. There is a chicken EST (Genbank accession AJ452523) that gives support to this conclusion. As with most other mucins, the chicken Muc13 PTS domain contains repeats. It is interesting to note that the sequences encoded by the exons are nearly identical, i.e. the sequence encoded by one exon corresponds to a repeat unit of the PTS domain (Fig. 5).

The chicken Muc13 tandem repeats thus have a different genomic organization as compared to higher animals. An analysis of zebrafish proteins (unpublished) identified a Muc13 homologue (Fig. 2) with its gene encoding the PTS domain divided into several exons. These findings suggest that this organization of the PTS domain represents an ancestral design of the vertebrate Muc13 gene and perhaps of other mucins.

The genomic organization of the Muc13 gene raises the possibility that a variation in length of the PTS domain may be accomplished not only by recombination events, as is the case for the human MUC1 polymorphism, but also by a regulation of splicing of mucin domain exons. This will allow a length variation not only between individuals, but also within one and the same individual.

The HEG (heart of glass) gene was first identified in zebrafish where it regulates the growth of the heart 32. Three isoforms are generated, one of which is predicted to be transmembrane and two secreted. Homologues have been identified in vertebrates, including man, and we now identified a chicken homologue. We observed that this protein is encoded by a gene adjacent to the Muc13 gene on chromosome 7 and these two genes have the same polarity. Synteny between chicken and man extends even beyond these two genes in both directions. Analysis of the human and mouse genomes shows that the HEG and Muc13 genes are organized in the same manner in these animals. Although Muc13 has an SEA domain which is absent in HEG, there are interesting similiarities between the two proteins as both have transmembrane, EGF and PTS domains. This observation together with the fact that the genes are in the same locus suggest an evolutionary relationship.

The results of our inventory of mucins in vertebrates are summarized in Table 1. In our previous study of F. rubripes mucins, we concluded that this vertebrate has a set of gel-forming mucins comparable to those of man and rodents. Further analysis of F. rubripes, Tetraodon nigroviridis, and Danio rerio as well as Ciona intestinalis suggest that some of the mucins are quite different to the higher vertebrate mucins (in preparation). In the chicken however, we found obvious homologues of the primate and rodent Muc2, Muc5ac, Muc5b, and Muc6. They are homologous both with respect to sequence of the VWD domains as well as to their localization and direction in the gene cluster. Interestingly, the chicken cluster contains an additional gene encoding ovomucin, found in egg white. This gene seems to be present also in frogs and fishes (Lang, T., et al., unpublished observation), but has disappeared during the development of mammals and might not be needed by animals where the fertilized egg is developed within the female body. A more detailed study of the phylogenetic distribution of ovomucin will probably give more clues as to its evolutionary history.

Whereas fish, chicken, and man have a reasonably similar set of gel-forming mucins, many of the transmembrane type mucins are missing in chicken and fish. In particular, this is true for fish as only genes encoding Muc13 and a MUC1-like protein was identified in F. rubripes. In the chicken, also the transmembrane mucins Muc4 and Muc16 were identified. Still the transmembrane mucins homologous to the human MUC3, MUC12, and MUC17 seem to be missing in chicken as well as in fishes. These mucins might therefore be a more recent development in vertebrates. A more detailed view of the phylogeny of these proteins will be crucial to better understand the evolution of mucins. Therefore, it is necessary to carry out a careful inventory of mucins in more animals.

As sources for protein and genomic sequences we used UCSC 33, Ensembl 26, NCBI 34, and Celera 35. Protein domain profiles were from Pfam 36. We made use of the ENSEMBL version of the chicken genome, 27.1d. The genomic DNA sequence had 111864 contigs, with a total of 1,08 × 10⁹ bases.

For identification of PTS domains we made use of PTSpred that can be used to search both DNA and protein sequences 2. To identify Pfam domains we used for protein sequences hmmpfam of the hmmer package 37 and for nucleotide sequences Genewise 24. Transmembrane domains were identified by TMHMM 38 and signal sequences by SignalP 39. For exon prediction Genscan 40 was used. Alignments of proteins and DNA were done by BLAST, ClustalW 41 or programs of the GCG package (GCG, Madison, WI). The repetitive nature of the PTS domains was analyzed with Dotplot of the GCG package. In house Perl scripts were used for additional tasks.

Two different sets of proteins were considered, on the one hand, proteins predicted by ENSEMBL and on the other hand, proteins predicted by ab initio methods. When analyzing the ENSEMBL proteins the PTS domain, VWD and SEA domain analysis identified 146, 53 and 26 proteins, respectively. Analysis of the corresponding set of proteins predicted by ab initio methods resulted in 78 PTS, 52 VWD and 17 SEA domain candidates. Ten different proteins had both PTS and either VWD or SEA domains. Two were identified as related to otogelin. The remaining eight proteins were identified as Muc1 (weak candidate), Muc2, Muc4, Muc5ac, Muc5b, Muc6, Muc13, and Muc16 and are described under Results and Discussion. We also used Genewise to scan the chicken genome sequence but that analysis did not result in any additional strong mucin candidates.