Molecular Evolution of Vertebrate Goose-Type ... - Springer Link

J Mol Evol (2003) 56:234–242 DOI: 10.1007/s00239-002-2396-z

Molecular Evolution of Vertebrate Goose-Type Lysozyme Genes David M. Irwin,1 Zhiyuan Gong2 1

Department of Laboratory Medicine and Pathobiology, Banting and Best Diabetes Centre, Faculty of Medicine, University of Toronto, 100 College Street, Toronto, Ontario, Canada M5G 1LG 2 Department of Biology, National University of Singapore, Singapore Received: 8 April 2002 / Accepted: 23 September 2002

Abstract. We have found that mammalian genomes contain two lysozyme g genes. To better understand the function of the lysozyme g genes we have examined the evolution of this small gene family. The lysozyme g gene structure has been largely conserved during vertebrate evolution, except at the 5¢ end of the gene, which varies in number of exons. The expression pattern of the lysozyme g gene varies between species. The fish lysozyme g sequences, unlike bird and mammalian lysozyme g sequences, do not predict a signal peptide, suggesting that the encoded proteins are not secreted. The fish sequences also do not conserve cysteine residues that generate disulfide bridges in the secreted bird enzymes, supporting the hypothesis that the fish enzymes have an intracellular function. The signal peptide found in bird and mammalian lysozyme g genes may have been acquired as an exon in the ancestor of birds and mammals, or, alternatively, an exon encoding the signal peptide has been lost in fish. Both explanations account for the change in gene structure between fish and tetrapods. The mammalian lysozyme g sequences were found to have evolved at an accelerated rate, and to have not perfectly conserved the known active site catalytic triad of the bird enzymes. This observation suggests that the mammalian enzymes may have altered their biological function, as well.

Correspondence to: D.M. Irwin; email: [email protected]

Key words: Lysozyme g — Gene duplication — Protein structure — Protein function — Accelerated evolution

Introduction Goose-type lysozyme (lysozyme g) was first identified as an anti-bacterial enzyme found in the egg whites of various bird species (Dianoux and Jolle`s 1967; Canfield and McMurry 1967). Lysozyme g was found to differ from the well-characterized chicken-type lysozyme (lysozyme c) in molecular weight, amino acid composition, and enzymatic properties, indicating that these two enzymes were the products of two different genes (Arnheim and Steller 1970; Arnheim et al. 1973). While initially found in the egg whites of geese and relatives, lysozyme g has been detected by antibodies in other tissues of geese (Hindenburg et al. 1974), and within egg whites from many species of birds (Prager et al. 1974). The amino acid sequences of lysozymes g from goose, swan, and ostrich (Simpson et al. 1980; Schoentgen et al. 1982; Simpson and Morgan 1983) demonstrated that they had no, or very limited, similarity to lysozyme c sequences. With the determination of the three-dimensional structures of chicken-type (Blake et al. 1965) and goose-type (Gru¨tter et al. 1983) lysozymes, as well as bacteriophage lysozyme (Matthews and Remington 1974), it was suggested that all of these lysozymes might have a common evolutionary origin (Gru¨tter et al. 1983; Weaver et al. 1985). The divergence of these three families of lysozymes, though, is sufficiently ancient

235

that they have essentially no amino acid sequence similarity. Despite an assumed ancient origin, there has been little experimental evidence for the existence of lysozyme g outside of some bird families (see Prager and Jolle`s 1996). Recently, however, a lysozyme g cDNA and gene were characterized from the Japanese flounder, indicating that this gene is not unique to birds (Hikima et al. 2001). Japanese flounder lysozyme g has approximately 50% amino acid sequence identity with lysozyme g sequences from birds. Comparison of the chicken and flounder lysozyme g gene structures (Nakano and Graf 1991; Hikima et al. 2001) suggested that important differences exist near the 5¢ end of the chicken and founder genes, including the portions that code for the N-terminal regions of the proteins. The 5¢ end of the chicken lysozyme g gene is composed of three exons, whereas that of the flounder gene is composed of two exons, and the flounder gene predicts a shorter N-terminus for the protein. The remainder of the predicted flounder lysozyme g amino acid sequence is well conserved, including all putative active site residues (Hikima et al. 2001). Recombinant Japanese flounder lysozyme g was shown to have lysozyme activity (Hikima et al. 2001). As fish and birds have lysozyme g genes, parsimonious reasoning says that the common ancestor of birds and mammals must have possessed this gene. We therefore hypothesized that mammalian genomes must posses a lysozyme g gene. We searched the nearly complete human and mouse genomes for lysozyme g-like sequences and found that there are two putative lysozyme g genes in these mammalian genomes. Both of the lysozyme g genes are expressed in humans and mice, although at low levels. Analysis of the evolution of the known lysozyme g gene sequences suggests that this enzyme has changed function several times during vertebrate evolution. Materials and Methods Molecular Databases Lysozyme g sequences in GenBank were identified by searching the nt (non-redundant), EST (expressed sequence tags), pat (patent), and HTGS (high throughput genomic sequence) databases maintained by the National Center for Biotechnology Information (NCBI, http://www.ncbi.nlm.nih.gov), as well as the EBI Ensembl database (http://www.ensembl.org/). These databases were searched initially using the chicken lysozyme g cDNA and predicted protein sequences (Nakano and Graf 1991). Subsequent searches used sequences that were found in our sequences or with sequences generated by the assembly (see below) of predicted genes. Additional genome database resources (e.g., UniGene, and human genome databases) maintained by the NCBI were also searched for sequences and expression information.

Sequence Alignments and Phylogenetic Analysis Genomic sequence (e.g., human lysozyme g, accession number AC079447), and EST sequences, that encodes lysozyme g-like sequences were obtained from GenBank or Ensembl, and subsequent manipulations were done with MacDNASIS, version 3.7 (Hitachi, San Bruno, CA). EST sequences were managed with MacDNASIS to identify overlapping sequence and predict longer cDNA sequences. Comparison of EST/cDNA sequences to genomic sequences was used to predict potential gene structures. Predicted exons were used to infer putative full-length cDNA sequences and predict potential lysozyme g protein sequences. Potential promoter sequences were identified using the Neural Network Promoter Prediction program implemented at http://www.fruitfly.org/seq_tools/promoter.html. RepeatMasker (Smit and Green 2002; http://ftp.genome.washington.edu/RM/RepeatMasker.html) was used to identify the presence and types of repetitive sequences within each lysozyme g genomic DNA sequence. Alignments of human, mouse, and rat lysozyme g genomic regions were generated with PipMaker (Schwartz et al. 2000; http://nog.cse.psu.edu/pipmaker/). For phylogenetic analysis, DNA and protein sequence alignments were generated with Clustal W (Thompson et al. 1997). Aligned sequences were edited to remove all ambiguously aligned DNA or amino acid residues prior to phylogenetic analysis. Phylogenies of DNA and protein sequences were estimated using PAUP*4.0b10 (Swofford 2002), using both parsimony, with the PROTPARS matrix, and distance methods.

Expression of Lysozyme g Genes EST clones representing the two human lysozyme g genes (accession numbers AA954529 and AA843569 for g1 and g2, respectively) were obtained from GenomeSystems (St. Louis, MO). The cDNA inserts were labeled with 32P and hybridized to human multiple tissue expression arrays (MTE2, Clontech, Palo Alto, CA), according to the manufacture’s instructions. Hybridized membranes were exposed to phosphoimage screens and analyzed using a STORM 840 (Molecular Dynamics, Sunnyvale, CA).

Results Identification of Human Lysozyme g Genomic Sequences With the near completion of the human (International Human Genome Sequencing Consortium 2001) and mouse genomes, it has become possible to investigate whether these genomes contains genes that had previously only been identified in other species. We searched the publicly-available draft human genome, translated into all possible reading frames, for sequences that showed similarity to the predicted chicken lysozyme g amino acid sequence. These searches resulted in the identification of the Bacterial Artificial Chromosome (BAC) clone RP11-111H13 (accession number AC079447.4) from the nt database and two incomplete BAC sequences (RP11-38C17 and RP11-192D3, accession numbers AC092587.2 and AC092002.2, respectively) from the htgs database. All three of these BAC sequences have been assembled as overlapping

1 236

BAC clones within a single contig on chromosome 2 of the human genome, and thus there appears to be only one genomic locus with similarity to chicken lysozyme g in the human genome. However the chicken lysozyme g sequence has similarity to two different sequences, approximately 30 kb apart, within the BAC sequence AC079447. This suggests that there are two genes on human chromosome 2, which we have named lysozyme g1 and lysozyme g2. To better characterize these human lysozyme glike genes, we searched the nt, est, and patent databases of GenBank for sequences identical to the human lysozyme g genomic sequences. The nt database did not contain any sequences with identity to any of the potential human lysozyme g-like sequences. In contrast, the dbEST contained 17 and four sequences showing near identity to the two lysozyme g-like genomic regions, lysozyme g1 and g2, respectively. None of the EST sequences possessed both lysozyme g1 and g2 sequences. The two sets of ESTs similar to lysozyme g are also represented in the UniGene database. The lysozyme g1 genomic region contains sequences identical to those found in UniGene clusters Hs.164589 (16 ESTs), whereas the lysozyme g2 sequence are found within UniGene cluster Hs.250729 (1 EST). The patent database yielded eight additional sequences similar to lysozyme g1 and none identical to lysozyme g2. Structure of the Human Lysozyme g1 Gene Using sequences identified in the dbEST and pat databases, we could predict potential mRNAs, and gene structures for both human lysozyme g genes. For the human lysozyme g1 gene, all ESTs and patent sequences overlapped with no evidence for any large insertions or deletions within any sequence. The EST and patent mRNA sequences are spread over eight non-contiguous regions on chromosome 2 spanning approximately 20 kb of genomic sequence. Only one of the 25 human lysozyme g1 EST and patent sequences predicted the 5¢ most exon, while nine other sequences had their 5¢ ends within the next exon. The putative splice acceptor sequence of the second potential exon does not agree well with consensus acceptor sequence, as it lacks a poly-pyrimidine sequence. RepeatMasker results indicate that the 5¢ most predicted exon sequence is composed entirely of a FLAM_C SINE repetitive DNA element. These observations suggest that the longest lysozyme g1 cDNA sequence is an aberrant transcript, and that the typical 5¢ end of the lysozyme g1 gene is in the next exon; if so, then the human lysozyme g1 gene is composed of only seven exons (Fig. 1). We searched the genomic sequences upstream of the first exon (in Fig. 1) and found that it contains a potential promoter (i.e., TATA) sequence. The sequences at the

Fig. 1. Structure of vertebrate lysozyme g genes. The thin line represents introns and flanking sequences, whereas the boxes represent exons. Introns and exons are not to scale. Open boxes are untranslated regions, light shaded boxes encode predicted signal peptides, and solid boxes encode the predicted mature lysozyme g. Predicted or inferred splicing patterns are shown above each gene. The phase of the introns interrupting the coding region is shown in each intron. Phase I indicates after the first base of a codon, while phase O is between codons. The second exon of the mammalian g1 gene is only found in humans.

ends of each of the remaining exons are consistent with them being intron-exon junctions, as an AG sequence preceded and a GT sequence followed each exon. The sequence of the full-length lysozyme g1 gene is predicted to be about 17 kb. The coding sequence (Fig. 2) was found to span exons 3 through 7 (Fig. 1), with the ATG start codon in exon 3 and a stop codon in exon 7. A potential polyadenylation signal (AATAAA) was found near the 3¢ end of exon 7. Structure of the Human Lysozyme g2 Gene Following analysis similar to that described above of the lysozyme g1 gene, the four EST sequences encoding the human lysozyme g2 gene were found to span approximately 11 kb of genomic DNA and suggest the existence of only five exons. Unlike the human lysozyme g1 gene, large insertions, and deletions were necessary to align all four lysozyme g2 EST sequences, suggesting that some of the ESTs either represent alternatively spliced or unprocessed mRNAs. The longest ESTs, W25819, and W26875, predict five and four exons, respectively. All the intron-exon boundaries predicted by these ESTs contain appropriate mRNA splice signals. The EST AA843569 is collinear with the genomic sequence encoding exon 4, as well as the flanking intronic sequences. As there is no evidence for mRNA splicing to generate this EST it suggests that it is either an unprocessed mRNA or genomic DNA contamination. The EST AW418634 also shows no sign of mRNA splicing, and appeared to contain only exon 5 and flanking intron sequences of the lysozyme g2

237

Fig. 2. Alignment of lysozyme g amino acid sequences. A multiple sequence alignment of deduced and predicted lysozyme g sequences (for sources see text). Amino acid sequences are shown in single letter code. Dashes (-) represent gaps that have been inserted to maximize alignment. Question marks (?) represent missing (or uncertain) amino acid sequence. The arrow above the human g1 and human g2 sequences indicates inferred locations of signal peptidase cleavage from mammalian lysozyme g1 and g2, respectively. The sequences are numbered from the N-terminus of the mature protein (for secreted bird and mammalian sequences) or

from the N-terminus of the entire protein (fish sequences). The zebrafish g1 sequence is a partial sequence based on two EST sequences. The catfish sequence is predicted to be full-length, but is only predicted from EST sequences. The rat sequences are derived from a draft BAG sequence and do not predict the N-terminus of rat g1. Positions of introns in the mammalian, chicken, and flounder genes are indicated by the open triangles below the sequence. Lines joining cysteine residues indicate the deduced disulphide bridges in bird lysozyme g sequences. The three active site residues in bird lysozyme g are marked by *.

gene. Likewise this EST is likely either a portion of an unprocessed mRNA or DNA contamination. A predicted cDNA constructed using the inferred exon sequences was used to predict the lysozyme g2 amino acid sequence. This predicted amino acid sequence of lysozyme g2 was similar to other lysozyme g sequences, except that it diverged greatly after amino acid 114, showing no similarity to the well-conserved C-terminal 40 amino acids of bird or human g1 lysozymes. BLASTX searches with either bird lysozyme g or human lysozyme g1 protein sequences detected the presence of a genomic sequence downstream of exon 5 that had similarity to the C-terminal portion of lyso-

zyme g. Examination of the corresponding DNA sequences suggests that this sequence could be spliced to exon 5 to generate a sixth exon that would now predict a lysozyme g2 protein sequence similar to other lysozyme g sequences across its full length (Fig. 2). These observations suggest that the lysozyme g2 gene is about 13 kb in length and contains six exons, the last five of which contain protein coding sequence (Fig. 1). Mouse Lysozyme g Sequences Searches of the nearly complete mouse genome (Ensembl mouse genome assembly, v1) identified two

1 238

genomic sequences with similarity to the human lysozyme g genes. The two genomic sequences appear to represent the mouse orthologs of the human lysozyme g1 and g2 genes. Both sequences are located on mouse chromosome 1, and are closely linked, predicted to be about 30 kb apart in the mouse genome, and organized in a head-to-tail fashion. To better characterize the two mouse genes, the mouse nr and dbEST databases were searched with the two putative mouse lysozyme g gene sequences. A cDNA sequence (accession AK009014) similar to the mouse lysozyme g1 gene (the mouse sequence with greatest similarly to the human g1 gene sequence) was found in the NCBI nt database, as well as 46 EST sequences in dbEST. The mouse lysozyme g1 sequences were also represented in UniGene as cluster Mm.46417. The mouse lysozyme g2 gene (which has greater similarity to the human g2 gene sequence) was only represented by two mouse ESTs (AA738825 and BB571255). The mouse lysozyme g1 cDNA and lysozyme g2 ESTs confirmed the coding exons and allowed prediction of the non-coding exons (Fig. 1). All of the mouse lysozyme g1 EST sequences overlapped with the fulllength cDNA sequence, suggesting that they are all products of the same gene. Interestingly, comparison of the human and mouse lysozyme g1 DNA sequences suggested that the mouse lysozyme g1 mRNA does not include exon 2 (see Fig. 1). Only three of the 46 ESTs span this missing exon sequence, and all of these sequences appear to support the full-length cDNA sequence. Searches of the mouse genomic sequence, guided by PipMaker (Schwartz et al. 2000), identified within the mouse lysozyme g1 gene a sequence that was similar to the human lysozyme g1 exon 2. The potential mouse lysozyme g1 exon 2 sequence was found to lack consensus splice signals, and thus could not be part of the lysozyme g1 mRNA. The mouse lysozyme g1 ESTs and cDNA predict a 5¢ end similar to that of the human lysozyme g1 gene (see Fig. 1). A sequence similar to the FLAM_C SINE at the 5¢ end of one of the human lysozyme g1 sequences is not found in the mouse genome, again suggesting that the human FLAM_C SINE containing sequence is an aberrant transcript. The two mouse lysozyme g2 EST sequences represent the 5¢ and 3¢ ends of this genes transcript and do not overlap. The predicted structure of the mouse lysozyme g2 (see Fig. 1) is identical to the human lysozyme g2 gene, including the 3¢ end of the human lysozyme g2 gene that is not represented by any EST or cDNA sequence.

genomes, the rat BAC sequence possessed two lysozyme g-like sequences. Prediction of coding sequences within the genomic sequence suggested that the rat also possesses lysozyme g1 and g2 genes (Fig. 2). Searches of EST and other databases failed to identify any EST or cDNA sequences for the two rat lysozyme g genes. The non-coding portions of the two rat lysozyme g genes were predicted based on alignments with the mouse and human lysozyme g genes, generated using PipMaker (Schwartz et al. 2000). While the first non-coding exons of both the lysozyme g1 and g2 genes could be identified in our alignments, the exon containing the start codon of lysozyme g1 could not be found. This suggests that either exon 2 of the rat lysozyme g1 sequence is not present within the draft sequence or that the sequence could not be recognized as it has been deleted or evolved to such an extent that similarity has been lost.

Rat Lysozyme g Sequences

Expression of Lysozyme g Genes

Searches of other databases for lysozyme g-like sequences resulted in the identification of a draft sequence of a rat BAC (accession AC095781) that had similarity to lysozyme g. Like the human and mouse

Despite the fact that lysozyme g was first found in birds about 35 years ago, the existence and expression of a mammalian lysozyme g gene has not been characterized until now. A search of mammalian EST

Other Vertebrate Lysozyme g Sequences Lysozyme g cDNA and gene sequences have previously been reported from the chicken (Nakano and Graf 1991) and the Japanese flounder (Hikima et al. 2001). To determine whether there are any additional lysozyme g-like sequences from other species present in molecular databases, we searched them using BLASTN and BLASTX using chicken, flounder, and human lysozyme g sequences. Our BLAST searches identified EST sequences from flounder, catfish, and zebrafish. The flounder ESTs were all very similar to the full-length founder cDNA and gene sequence (Hikima et al. 2001), suggesting that they were all products of the same gene. A total of three catfish ESTs were identified (BE468929, BE469696, BE470418) and were found to overlap and predict a full-length open reading frame (Fig. 2). As this fulllength is predicted from ESTs there remains a possibility of a few errors in this sequence. Two overlapping zebrafish ESTs (accession AI396784 and AI385012) were identified, although these represent 5¢ and 3¢ sequences of the same clone. These sequences only predict a partial reading frame for lysozyme g, which is named zebrafish lysozyme g1 in Fig. 2. Searches of our own zebrafish EST database (Zeng and Gong 2002) identified an EST that differed from the one found in GenBank described above. We completed the sequence of our EST clone, enabling us to predict a full-length open reading frame, which is named zebrafish lysozyme g2 (Fig. 2).

239

databases identifies few ESTs similar to lysozyme g genomic sequences suggesting that these genes are expressed at low levels. Of the 17 ESTs for the human lysozyme g1, gene only five were isolated from normal tissue libraries (adult kidney and fetus), while the remaining were isolated from a manipulated cell line (Arthersys RAGE library). Only four human lysozyme g2 ESTs were identified, two from retinal cDNA libraries, one from testis, and one from a tumor tissue. To complement this EST survey, we hybridized lysozyme g1 and g2 cDNAs to a panel of mRNAs from 61 normal adult tissues, seven fetal tissues, and eight cell lines. We detected expression of the lysozyme g1 gene in only adult kidney. None of the fetal mRNA samples, including fetal kidney, were positive for lysozyme g1 expression, suggesting that the fetal lysozyme g1 ESTs were derived from a tissue not represented on the mRNA array. No significant expression of the lysozyme g2 gene was detected in the mRNA array, although the filter did not include mRNA from retina. These results suggest that a major site of lysozyme g1 gene expression is the kidney, with probably another fetal tissue also expressing the gene. The lysozyme g2 must be expressed at very low levels, or a restricted number of cells, to explain the few number of ESTs in the GenBank and the lack of detectable signal in our mRNA arrays. We also used the EST databases to survey expression of the lysozyme g1 and g2 in mouse. A total of 45 ESTs for mouse lysozyme g1 were found in GenBank. Like the full-length mouse lysozyme g1 cDNA, all of the ESTs were from a mouse tongue cDNA library. The large number of tongue ESTs suggests that the lysozyme g1 gene is expressed at a relatively high level in mouse tongue, approximately 0.35% of the 12,123 reported mouse tongue ESTs. No mouse lysozyme g1 ESTs were identified from kidney, despite the large number of mouse kidney-derived ESTs (80,000) in the EST database. As described above only two ESTs similar to lysozyme g2 were found in mouse. The two mouse lysozyme g2 ESTs were isolated from independently constructed skin cDNA libraries. Discussion Structure of Lysozyme g Genes A common ancestry of lysozyme g, lysozyme c, and bacteriophage lysozyme was hypothesized based upon limited regions of similar three-dimensional structures in the three families, despite the sequences lacking significant similarity (Matthews et al. 1981; Weaver et al. 1985). The characterization of the chicken lysozyme g gene (Nakano and Graf 1991) demonstrated that the exon-intron structure of this gene was different from that predicted by exon-the-

ory-of-genes, if the lysozyme g and lysozyme c had a common ancestor (Weaver et al. 1985). With the isolation and characterization of flounder (Hikima et al. 2001) and mammalian (this work) lysozyme g genes, we can determine whether the lysozyme g gene structure has been conserved within vertebrates. As shown in Fig. 1, the number of exons for the lysozyme g gene varies, from a minimum of five in flounder to a maximum of seven for human g1. The chicken and remaining mammalian lysozyme g genes each have six exons. The difference in number of exons in these orthologs is due to changes in the numbers of exons that encode the 5¢ untranslated region of the mRNAs. The human g1 gene has three exons containing 5¢ untranslated sequence, whereas the flounder gene only has one (see Fig. 1). Within the coding sequence, all intron-exon junctions appear to be in homologous locations and interrupt the coding region in the same phase (i.e., position within a codon, see Figs. 1 and 2). The size of the second coding exon generates the greatest variability in protein coding lengths, and contains a portion of the amino acid sequence that poorly aligns (see Fig. 2). The upstream exon that encodes that start codon has from four (flounder) to 64 (chicken) coding bases. The downstream (fully coding) exon varies in size from 78 (flounder) to 141 (mammalian g2) bases. Although the sizes and number of noncoding exons vary within vertebrates, the structure of the lysozyme g gene has been largely maintained within vertebrates (see Figs. 1 and 2). Unfortunately, our new observations do not help resolve the question concerning the potential common ancestry of lysozyme g, lysozyme c, and bacteriophage lysozyme. Expression of Lysozyme g Genes Lysozyme g was initially discovered as the lysozyme enzyme present in the egg whites of geese (Dianoux and Jolle`s 1967; Canfield and McMurry 1967). Lysozyme g has been found to be present at varying levels in the egg whites of some other species of birds (Prager et al. 1974), and it is the major lysozyme in a few species, such as swan and ostrich (Simpson et al. 1980; Schoentgen et al. 1982). In addition to egg whites, lysozyme g enzyme has been detected at low levels in other tissues of the goose (Hindenburg et al. 1974). The first lysozyme g gene to be characterized was that of chicken, where it was found to be expressed in non-adherent bone marrow cells and some cells present in lung (Nakano and Graf 1991). No expression of the lysozyme g gene was found in the oviduct of chicken, in contrast to the lysozyme c gene which is expressed at high levels (Nakano and Graf 1991). Thus, within birds the pattern of expression of lysozyme g varies between species.

1 240

Only recently have a gene and cDNA for lysozyme g been characterized from a non-avian species, the Japanese flounder (Hikima et al. 2001). In contrast to the restricted expression patterns found in birds, lysozyme g in was found to be expressed in all flounder tissues examined, although at varying levels (Hikima et al. 2001). Lysozyme g expression was up-regulated in some tissues (e.g., blood and intestine) upon infection of flounder with bacteria (Hikima et al. 2001). The lysozyme g ESTs identified in catfish and zebrafish were isolated from tissues that express lysozyme g in the flounder, suggesting that similar expression patterns are seen in these species of teleost fish. In mammals, the expression of lysozyme g has not been previously reported. We now show that the genomes of three mammalian species contain two lysozyme g genes. Expression surveys using multiple tissue arrays demonstrate that neither of the human lysozyme g genes is widely or highly expressed. In our survey of 61 normal adult and eight fetal tissues we could detect expression of the lysozyme g1 gene in only adult kidney. Expression of the lysozyme g2 gene could not be detected in the surveyed adult and fetal tissues. Expression of both human genes could be inferred from EST database searches, where a limited number of ESTs have been identified from normal tissues (e.g., kidney), suggesting that both genes are expressed at low levels and in a tissue restricted fashion. EST expression data suggest, like within birds, that there is variation in site or level of expression of lysozyme g genes within mammals. Interestingly the two tissues where mouse lysozyme g ESTs have been found, tongue and skin, were not include in the human multiple tissue array; thus, it is possible that these genes may be expressed in those tissues in humans, although they have not been found in the EST database. A relatively large number (45) of lysozyme g1 tongue-derived ESTs were identified in the mouse. This observation suggests that in the mouse, lysozyme g1 is expressed in tongue at a relatively high level (0.35% of the reported mRNA transcripts). None of the human lysozyme g1 ESTs is derived from the tongue. Despite the low number of human tongue ESTs (