Evolution of Mammalian KELL Blood Group Glycoproteins and Genes (KEL): Evidence for a Marsupial Origin from an Ancestral M13 Type II Endopeptidase Gene

KELL is a member of the M13 family of type II neutral endopeptidases, which functions as a blood group antigen in human and animal populations. KELL amino acid sequences and structures and KEL gene locations were examined using bioinformatic data from several mammalian genome projects. Mammalian KELL sequences shared 55-99% identity, as compared with 21-31% sequence identities with other M13-like family members. Four predicted N-glycosylation sites were conserved among the mammalian KELL proteins examined. Sequence alignments, key amino acid residues and conserved predicted secondary and tertiary structures were also studied, including active site residues, predicted disulfide forming Cys residues, cytoplasmic, transmembrane and extracellular sequences and KELL C-terminus amino acid sequences. Mammalian KEL genes usually contained 18 or 19 coding exons on the direct strand. Transcription factor binding sites within the human KEL promoter may regulate transcription within erythroid cells. Phylogenetic analyses examined the relationships and potential evolutionary origins of the mammalian KEL gene with six other vertebrate neutral endopeptidase M13 family genes. These suggested that KEL originated in an ancestral marsupial genome from a gene duplication event of a neutral endopeptidase M13-like gene. *Corresponding author: Roger S Holmes, Eskitis Institute for Drug Discovery, School of Biomolecular and Physical Sciences, Griffith University, Nathan, QLD 4111, Australia, Tel: +61 7 37356008; E-mail: r.holmes@griffith.edu.au Received April 14, 2013; Accepted July 17, 2013; Published July 22, 2013 Citation: Holmes RS (2013) Evolution of Mammalian KELL Blood Group Glycoproteins and Genes (KEL): Evidence for a Marsupial Origin from an Ancestral M13 Type II Endopeptidase Gene. J Phylogen Evolution Biol 1: 112. doi:10.4172/2329-9002.1000112 Copyright: © 2013 Holmes RS. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.


Introduction
KELL blood group glycoprotein (KELL, also called CD antigen 238; EC 3.4.24.-) is one of at least seven members of the M13 family of neutral endopeptidases, which are zinc-containing type II transmembrane enzymes [1][2][3][4]. KELL blood group glycoprotein contains several antigens that are highly immunogenic and serve as the third most effective system in triggering an immune reaction, after the ABO and Rh blood groups [5][6][7]. KELL is also a single-pass transmembrane protein which is linked to the XY protein, found in red blood cell (RBC) membranes [8,9], and serves as an endothelinconverting enzyme, an endopeptidase which cleaves 'big' endothelin-3 to form an active vasoconstrictor peptide [10,11].
The gene encoding KELL (KEL in humans; Kel in mice) is highly expressed in erythroid tissues, but also in other tissues, including testis, heart, spleen and skeletal muscle [3,9]. The structures of the human KEL and the mouse Kel genes have been reported, containing 18 or 19 exons of DNA encoding KELL sequences [9,10,27,28]. The molecular basis of the major human KELL antigens has been determined, which result from KEL point mutations and single amino acid substitutions [29]. At least 30 KELL antigens are reported [27,30,31], including the highest frequency KEL polymorphism in human populations, designated as K 1 / K 2 , due to a C→T substitution in exon 6 causing a 193Thr to 193Met substitution and disruption of an N-glycosylation site [32]. Other high incidence KELL antigens have been reported, including KALT, which is sensitive to treatment of RBCs by trypsin [27,33]. A 'null' human KEL phenotype (designated as K 0 ), which abolishes KELL expression in erythroid tissues, resulting from a mutation at the splicing exon 3 donor site [33], while other K 0 alleles have been reported in low frequency in a Chinese population [34]. Knock-out (Kel-/Kel-) (KO) mice lacking RBC KELL glycoprotein exhibited changes in red cell ion transport and some mild motor dysfunction, have provided a useful model to study the biochemical and physiological roles for this protein [31]. RBC Gardos channel activity, which normally functions as a potassium chloride cotransport and calcium-activated potassium channel, and assists with maintaining RBC hydration status [34], was increased in KO erythrocytes, although lacking endothelin-converting endopeptidase activity. Antibodies generated in the circulation system in response to KELL antigens are usually immunoglobulin G, which may result in severe hemolytic transfusion reactions and hemolytic disease of the fetus and newborn [5,7]. Recent studies have suggested that these reactions may occur not only as a result of immune RBC destruction, but also by the suppression of erythropoiesis by anti-KELL-K 1 antibodies, which can lead to severe anemia in the fetus or new born [35,36]. This paper reports the predicted gene structures and amino acid sequences for several mammalian KEL genes and proteins, the predicted structures for mammalian KELL proteins and the structural, phylogenetic and evolutionary relationships for these genes and enzymes with those for six other vertebrate M13 neutral Type II endopeptidase gene families. The results suggest that the mammalian KEL gene arose from the duplication event of an ancestral mammalian M13 Type IIlike endopeptidase gene, with the appearance corresponding to the emergence of marsupial mammals during evolution.

Mammalian KEL gene and KELL protein identification
BLAST studies were undertaken using web tools from the National Center for Biotechnology Information (NCBI) [37]. Protein BLAST analyses used mammalian KELL and six other vertebrate M13 neutral endopeptidase amino acid sequences previously described [1,2,9,23,[38][39][40][41][42][43] (Table 1 and 2). BLAT analyses were subsequently undertaken for each of the predicted KELL amino acid sequences and other M13 neutral endopeptidase-like genes, using the UC Santa Cruz Genome Browser with the default settings to obtain the predicted locations for each of these vertebrate genes, including predicted exon boundary locations and gene sizes (Table 1 and 2) [44]. The AceView website was used to obtain structures for the major human KEL and mouse Kel transcripts [28].

Predicted structures and properties of mammalian KELL proteins
Predicted secondary and tertiary structures for human and other mammalian KELL proteins were obtained using the SWISS-MODEL web-server [43], and the reported tertiary structure for human ECE1 complexed with phosphoramidon [20] (PDB:3dwbA), with a modeling residue range of 60-713 for human KELL. Molecular weights, N-glycosylation sites and predicted transmembrane [46], cytosolic and extracellular sequences for mammalian KELL and related vertebrate M13 Type II endopeptidase proteins were obtained using Expasy web tools (http://au.expasy.org/tools/pi_tool.html). Identification of conserved domains for mammalian KELL proteins was made using NCBI web tools [47] (http://www.ncbi.nlm.nih.gov/Structure/cdd/ wrpsb.cgi).

Comparative human (KEL) gene expression
The genome browser (http://genome.ucsc.edu) was used to examine GNF Expression Atlas 2 data, using U133A and GNF1H expression chips for the human KEL gene (http://biogps.gnf.org) [48]. Gene array expression 'heat maps' were examined for comparative gene expression levels among human and mouse tissues, showing high (red); intermediate (black); and low (green) expression levels.

Phylogeny studies and sequence divergence
Phylogenetic analyses were undertaken using the http://phylogeny. fr platform [49]. Alignments of mammalian KELL sequences, six other vertebrate M13 neutral endopeptidase sequences and a nematode (Caenorhabditis elegans) M13 endopeptidase-like sequence were assembled using PMUSCLE [50] (Table 1 and 2). Alignment ambiguous regions were excluded prior to phylogenetic analysis, yielding alignments for mammalian KELL sequences with other vertebrate neutral endopeptidase and nematode (Caenorhabditis elegans) M13like sequences. The phylogenetic tree was constructed using the maximum liklihood tree estimation program PHYML [51].

Alignments of mammalian KELL amino acid sequences
The deduced amino acid sequences for gibbon (Nomascus leucogenys), horse (Equus caballus) and rat (Rattus norvegicus) KELL proteins are shown in Figure 1, together with previously reported sequences for human [2] and mouse KELL proteins [9] (Table 1). Alignments of human with other mammalian sequences examined were between 55-98% identical, suggesting that these are members of the same gene family, whereas comparisons of sequence identities of mammalian KELL proteins with human ECE1 and MME proteins exhibited lower levels of sequence identities: 27-31% and 19-22%, respectively, indicating that these are members of distinct M13 Type II endopeptidase-like gene families ( Table 1).
The amino acid sequences for mammalian KELL proteins contained between 713 (for human KELL isoform 'b') and 784 (for rabbit KELL) residues, whereas most mammalian KELL sequences contained either 732 amino acids (for human KELL isoform 'a'), or 713 residues (for human KELL isoform 'b') ( Figure 1; Table 1). Previous studies have reported several key regions and residues for human and mouse KELL proteins (human isoform 'b' KELL amino acid residues were identified in each case). These included an N-terminus cytoplasmic tail (residues 1-28), followed by a hydrophobic transmembrane twenty residue segment, which may anchor the enzyme to the plasma membrane [2]. These N-terminal cytoplasmic tail and transmembrane regions revealed a high degree of amino acid sequence conservation ( Figure 1). This region was further characterized by conserved 'book-end' residues for mammalian KELL, Arg27-Arg/Trp28 at the N-terminal end, and Tyr48 at the C-terminal end of the membrane anchoring segment, which may contribute to the membrane spanning properties. Figure  2 provides an alignment of the cytoplasmic N-terminal sequences for several mammalian KELL sequences, including human KELL 'a' and 'b' isoform sequences. These contained identical transmembrane and endopeptidase domains, but with N-terminal cytoplasmic domains of different lengths (residues 1-47 and 1-24 for the human 'a' and 'b' isoforms, respectively). This explains the differences in amino acid composition for these human KELL proteins (732 and 713 amino acids, respectively). Figure 2 also shows N-terminal cytoplasmic domain sequences for other mammalian cytoplasmic domain KELL sequences, including rabbit KELL, which contained an extended N-terminal cytoplasmic domain sequence of 99 amino acid residues.
Residues 49-713 of the human KELL sequence were identified using bioinformatics as a large peptidase family M13 endopeptidaselike domain involved in the proteolysis of peptides in the body [47].  This C-terminal region is predicted to be localized extracellularly, and to contain an active M13-like endopeptidase sequence capable of metabolizing physiologically active peptides. Four (horse and pig KELL sequences) to eight (mouse KELL sequence) N-glycosylation sites were predicted for the mammalian KELL sequences examined, which shared 16 distinct N-glycosylation sites ( Figure 1; Table 3). Most mammalian KELL sequences exhibited at least four common N-glycosylation sites, designated as sites 1, 10, 13 and 16 (Table 3). It is relevant to note that a high frequency human KEL polymorphism (site 10, Table 3) is responsible for removing an N-glycosylation site, resulting in the synthesis of antigens K 1 and K 2 with distinct antigenic properties [27,32]. This lends support to a significant role being played by at least this N-glycosylation site in contributing to the structure and antigenic properties of this protein.
Twelve active site residues were conserved for the mammalian KELL sequences examined (Figure 1; Supplementary Table 1s; Figure  1), which are deduced from the M13 peptidase domain active site residues reported for the NCBI domain studies [47]. These included an active site catalytic residue (Glu563); an active site proton donor (Asp619); 3 residues involved in binding the active site Zinc (His562, His566 and Glu615), deduced from 3D studies of a related enzyme, MME [14]; and 7 other active site residues, also deduced from the MME tertiary structure, namely Asn521, Ala522, Ile599, Tyr658, Ala659, His675 and Arg681 (Figure 1; Supplementary Table 1s). Of particular significance, however, was the amino acid substitution observed for the KELL 'active site' Glu563 residue, which was replaced by a basic amino acid (lysine) for the rat and rabbit KELL sequences. This is likely to have   Table 1 for sources of KELL sequences; *shows identical residues for KELL subunits; : similar alternate residues; . dissimilar alternate residues; predicted cytoplasmic residues are shown in red; predicted transmembrane residues are shown in blue; N-glycosylated and potential N-glycosylated Asn sites are in green; endopeptidase active site residues are shown in red; three predicted Zinc-binding residues are also shown in red; conserved Cys53, forming a disulfide bond with XY protein is shown in pink; conserved Cys residues that align with MME disulfide bond forming Cys residues are in blue; predicted α-helices for vertebrate KELL are in shaded yellow, and numbered in sequence from the start of the predicted transmembrane domain; predicted β-sheets are in shaded green, and also numbered in sequence from the N-terminus; bold underlined font shows residues corresponding to known or predicted exon start sites; exon numbers refer to human KEL gene exons; note the three major domains identified as cytoplasmic (N-terminal tail); transmembrane (for linking KELL to the cell membrane); and extracellular domains. Amino acids are color coded: yellow for proline (P); S (serine); green for hydrophilic amino acids, S (serine), Q (glutamine), N (asparagine), and T (threonine); brown for glycine (G); light blue for hydrophobic amino acids, L (leucine), I (isoleucine), V (valine), M (methionine), W (tryptophan); dark blue for amino acids, T (tyrosine) and H (histidine); purple for acidic amino acids, E (glutamate) and D (aspartate); and red for basic amino acids, K (lysine) and R (arginine); the N-terminus (cytoplasmic) domain is identified; human (a) and (b) refer to major KELL isoform sequences.
a major impact on the endopeptidase activity for these proteins and may reflect a 'null' phenotype for rat and rabbit KELL proteins, similar to that observed for the human K 0 KELL variant phenotype [33]. Most of the other residues in the active site and C-terminal KELL regions were predominantly conserved for many of the mammalian KELL proteins examined, which may reflect strong functional roles for these sequences as well (Figure 1).
There were several conserved Cys residues among the mammalian KELL sequences examined, including Cys53 (Figure 1), which forms a heterodimeric disulfide bond with the XK red cell membrane protein, a multipass transport protein [8]. Ten other Cys residues were conserved among the mammalian KELL sequences examined, which aligned with 10 of 12 Cys residues reported as forming disulfide bonds for the human MME sequence [53]. These included the following potential disulfide bond positions for human KELL: Cys58-Cys63; Cys81-Cys698; Cys89-Cys663; Cys136-Cys391; and Cys591-Cys729, which suggests that mammalian KELL proteins contain five disulfide bonds in similar positions to those reported for five of the six human MME disulfide bonds [53], in addition to the disulfide bond linking KELL with the XK red cell membrane protein [8].
Alignments of the C-terminal sequences for human, opossum and tasmanian devil (Sarcophilus harrisii), KELL proteins are shown in Supplementary Figure 1s, together with the corresponding KEL 3' nucleotide sequences. In contrast to all other mammalian C-terminal KELL sequences examined, the opossum KELL sequence contained three additional amino acids (Leu-Ser-Ala) to the C-terminal Trp residue observed for all other mammalian KELL sequences. It is likely that mutations in this region of the ancestral opossum KEL sequence may have caused a shift of the stop codon downstream, with a corresponding extension of the C-terminal opossum KELL sequence.

Predicted secondary and tertiary structures for mammalian KELL proteins
Predicted secondary structures for human, mouse, gibbon, horse and rat KELL sequences were examined, particularly for the extracellular sequences (Figure 1), using the known structure reported for an M13 family peptidase, human ECE1 [20]. α-Helix and β-sheet structures were identical in each case, with 28 α-helices and 9 β-sheet structures being observed. Of particular interest were α-helices 23, 25,   The structure for the human ECE1 subunit and the predicted structure for the human KELL subunit are based on the reported structure for human ECE1 [20], and obtained using the SWISS MODEL web site based on PDB 3dwbA http:// swissmodel.expasy.org/workspace/ The rainbow color code describes the 3-D structures from the N-(blue) to C-termini (red color); predicted α-helices, β-sheets and proposed active site cleft are shown.
26 and 27, which contained several predicted active site residues for human KELL. A predicted tertiary structure for human KELL is shown in Figure 3, together with human ECE1 [20]. The tertiary structure of the extracellular domain (residues 60-713) for human KELL was similar to that described for human ECE1 [20]. Twenty-eight α-helices and nine β-sheet structures were observed, which is similar to that described for the predicted secondary structure for this enzyme. In addition, two major domains for these enzymes were observed, that enclose a large cavity previously shown to contain the enzyme's active site. The more C-terminal of these two domains has been shown to have a fold similar to that of thermolysin which contains the active site residue, whereas the other domain may serve to control access of substrates to the active site [20]. Overall, the predicted human KELL structure closely resembled that reported for human ECE1 [20], and MME [14]. Previous modeling studies have reported that the human KELL protein contains two globular domains, consisting mostly of α-helices, with the domain situated closest to the transmembrane sequence containing both the N-and C-terminal sequences and the active site [54]. In addition, the outer domain of the protein contained all of the amino acids involved in forming at least 30 human KELL antigenic sites. Table 1 summarizes the predicted locations for mammalian KEL genes based upon BLAT interrogations of several mammalian genomes, using the reported sequences for human [2] and mouse [9], KELL and the predicted sequences for other mammalian KELL proteins and the UC Santa Cruz genome browser [44]. The mammalian KEL genes examined were transcribed on the minus strand. Figure 1 summarizes the predicted exonic start sites for human, mouse, gibbon, horse and rat KEL genes, based on the KELL 'b' isoform, with each having 18 coding exons, in identical or similar positions to those predicted for the human KEL gene. Figure 4 shows the predicted structures for the major human and mouse KEL transcripts. In each case, two major transcripts were observed, including the reference sequences (NM_000420 and NM_032540), which were 2811 and 2531 bps in length, with extended 5'-and 3'-untranslated regions (UTR), for the human and mouse KEL transcripts. The two major human KEL transcript isoforms, designated as 'a' and 'b' , encode proteins with distinct N-terminal amino acid sequences, and contain 732 and 713 amino acids with 19 and 18 coding exons, respectively (Table 1; Figure 2) [28]. The human KEL promoter region does not contain a TATA box, but has potential transcription factor binding sites for GATA-1 and Sp1 [29], as well as several other KEL promoter transcription factor binding sites (Table 4). Of particular significance to KEL gene regulation for erythroid cell development are the following sites: NF-E2: a transcriptional factor essential for erythroid maturation and differentiation, which also participates in the transcriptional activation of the mammalian beta-globin gene locus [55]; GATA1: a transcriptional activator localized in the promoters of the globin gene family and other erythroid-specific genes [56]; and EVI1: a transcription activator which is essential for the proliferation and maintenance of hematopoietic stem cells [57]. It would appear that the KEL gene promoter is well endowed with gene regulatory sequences, which may contribute to the high levels of KEL expression in mammalian erythroid cells, and to the maintenance of this expression during development.

KEL tissue expression
Supplementary Figure 2 presents 'heat maps' showing comparative KEL expression in various human tissues obtained from GNF Expression Atlas 2 Data using U133A and GNF1H (human) chips (http://genome.ucsc.edu; http://biogps.gnf.org) [48]. These data supported high level tissue expression for human KEL, particularly in bone marrow (BM) erythroid cells, which is consistent with previous reports for this gene [5][6][7]. Another study reported that the KELL protein is expressed in nonerythroid tissues, including testis and skeletal muscle, although KEL transcripts were also reported in several other tissues [3]. Supplementary Figure 2 supports this wider KEL tissue expression, particularly for testis and fetal liver. Overall, human KEL and mouse Kel tissue expression levels were 0.5-0.8 times the average level of gene expression which supports a key role for this protein as an endopeptidase, but with a highly specific erythroid tissue expression profile [5][6][7]28].

Phylogeny and divergence of mammalian KEL and other vertebrate M13 type II endopeptidase genes and proteins
A phylogenetic tree ( Figure 5) was calculated by the progressive alignment of 15 mammalian KELL amino acid sequences with several other vertebrate M13 type II endopeptidase-like sequences, which was 'rooted' with the worm (Caenorhabditis elegans) membrane metallopeptidase (NEP22) sequence Table 1 and 2). The phylogram showed clustering of the KELL sequences into groups consistent with their evolutionary relatedness, as well as groups for each of the vertebrate M13 type II endopeptidase families, which were distinct from the worm NEP22 sequence. These groups were significantly different from each other (with bootstrap values >90). This data suggests that the KEL gene and KELL protein have appeared early in marsupial and eutherian mammalian evolution, for which a proposed common ancestor for these genes may have predated or coincided with the appearance of marsupial mammals during vertebrate evolution. The KEL gene (and KELL protein) was apparently absent from monotreme (platypus: Ornithorhynchus anatinus), bird (chicken: Gallus gallus), reptile (lizard: Anolis carolensis), amphibian (frog: Xenopus tropicalis) and fish (zebrafish: Danio rerio) genomes, whereas other vertebrate M13 type II endopeptidase genes (MME, MMEL1, ECE1, ECE2, ECEL1 and PHEX) were present throughout vertebrate evolution ( Figure 5) [57]. In addition, these results suggested that the mammalian KEL gene evolved, following a gene duplication event of an ECE-like gene, given that ECE1, ECE2 and KEL appear to be more closely related to each other, than to other M13 type II endopeptidase genes. This is consistent with a recent study of the evolution of vertebrate ECEL1 genes and proteins [57].

Conclusions
This study suggests that mammalian KEL genes and encoded KELL proteins represent a distinct gene and protein family of M13 Type II neutral endopeptidase-like proteins which share conserved sequences, and active site residues with those reported for other related M13 Type II endopeptidases, MME, MMEL1, ECE1, ECE2, ECEL1 and PHEX, previously studied [19][20][21][22]25,58]. KELL has a distinctive property among these M13 Type II neutral endopeptidases in serving as a major RBC antigen and blood group within human populations [2,[5][6][7]. In addition, KELL is capable of processing 'big' endothelin-3, generating endothelin-3 (ET-3), which is a potent bioactive peptide capable of acting as a vasoconstrictor [10]. KELL 'null' variants, however, exist naturally in human populations without apparent major health impact [31,34,35]. Moreover, the mammalian active site Glu563 residue is substituted in rat and rabbit KELL proteins by a lysine residue (Supplementary Table 1s), which may reflect a more substantial role for KELL as a RBC blood group antigen rather than as an active endopeptidase.  Figure 5: Phylogenetic tree of mammalian KELL amino acid sequences with representative vertebrate MME, MMEL1, ECE1, ECE2, ECEL1 and PHEX sequences.
The tree is labelled with the endopeptidase-like name and the name of the animal and is 'rooted' with the worm (Caenorhabditis elegans) NEP22 sequence, which was used to 'root' the tree. Note the 5 major groups corresponding to the MME and MMEL1; PHEX; ECEL1; KEL; and ECE1 and ECE2 gene families. A genetic distance scale is shown. The number of times a clade (sequences common to a node or branch) occurred in the bootstrap replicates are shown. Only replicate values of 90 or more, which are highly significant, are shown with 100 bootstrap replicates performed in each case. A proposed sequence of gene duplication events is shown arising from an ancestral endopeptidase-like gene.
KELL is encoded by a single gene (KEL), among the marsupial and eutherian mammalian genomes studied, which is highly expressed in human and mouse erythroid cells and usually contained 18 or 19 coding exons on the negative strand. Several transcription factor binding sites were localized within the human KEL gene promoter region, including GATA1, EVI1 and NFE2, which are essential in regulating erythroid cell differentiation and maintenance of expression, and may contribute significantly to the high level of gene expression in bone marrow erythroid cells.
Predicted secondary and tertiary structures for vertebrate KELL proteins showed a strong similarity with other M13 type II endopeptidaselike proteins. Several major structural domains were apparent for mammalian KELL proteins, including a N-terminal cytoplasmic tail of varying lengths between human KELL isoforms and other mammalian KELL sequences; a transmembrane domain which anchors the enzyme to the cell membrane; and an extracellular domain, containing the active site (including a zinc binding site), which is responsible for endopeptidase activity; and four conserved N-glycosylation sites, which may contribute significantly to the antigenic properties for this blood group protein. Several KELL endopeptidase domain cysteine residues were conserved among mammalian sequences, including Cys53 (see Figure 1), which forms a heterodimeric disulfide bond with the XK red cell membrane protein, a multipass transport protein [8]. Ten other Cys residues were also conserved which aligned with 10 of 12 Cys residues forming disulfide bonds for the human MME sequence [53]. This suggests that mammalian KELL proteins may contain five disulfide bonds in similar positions to those reported for five of the six human MME disulfide bonds [53]. Alternatively, these Cys residues may serve other KELL structural or functional roles.
A phylogenetic study used 15 mammalian KELL sequences with a range of other vertebrate M13 type II endopeptidase sequences, which indicated that the KEL gene had appeared early in marsupial mammalian evolution, prior to or coincident with the appearance of marsupial mammals, and existed as a distinct gene family within this group, together with the MME and MMEL1; ECE1 and ECE2; ECEL1; and PHEX gene groups. Moreover, the study indicated that the KEL gene may have originated from a gene duplication event of an ancestral mammalian ECE-like gene, in common with the ECE1 and ECE2 genes, which originated much earlier in vertebrate evolution, as both of these genes have been reported in bird, reptile, amphibian and fish genomes [57].