Abstract
The 185/333 gene family is highly expressed in two subsets of immune cells in the purple sea urchin in response to immune challenges. The genes encode a surprisingly diverse set of transcripts, which is a function of the variable presence or absence of blocks of shared sequences, known as elements that generate element patterns. Diversity is also the result of a significant level of point mutations. Together, variable element patterns and single nucleotide polymorphisms result in many unique transcripts. The 185/333 genes only have two exons, with the variable element patterns encoded entirely within the second exon. The diversity of the gene family may be the result of frequent recombination among the 185/333 genes that generates a mosaic distribution of element sequences among the genes. A comparative analysis of the sequences for the genes and messages from individual sea urchins indicates that these two sequence sets have largely different nucleotide sequences and appear to use different element patterns. Furthermore, the nucleotide substitution patterns between genes and messages reveal a strong bias toward transitions, particularly cytidine to uridine conversions. These data are consistent with cytidine deaminase activity and may represent a novel form of immunological diversification in an invertebrate immune response system.
Molecular diversity among immune response proteins has been observed in nearly all organisms, including vertebrates, invertebrates, and plants (reviewed in Refs. 1, 2). In addition to the well-understood mechanisms in vertebrate somatic gene rearrangement of the Ig gene family (3), novel approaches to increasing immune system diversity have been identified in cyclostomes and a surprising number of invertebrate species that do not have mammal-like adaptive immune functions (2, 4). Variable lymphocyte receptor genes are composed of multiple cassettes of leucine-rich repeats that are somatically assembled through a process that employs short sequence repeats that flank the cassettes (5). This assembly is mediated by a putative AID4 (activation-induced cytidine deaminase)/APOBEC (apolipoprotein B mRNA editing catalytic component)-like cytidine deaminase (6). The freshwater snail, Biomphalaria glabrata, which is an intermediate host for the human parasite Schistosoma mansoni, expresses fibrinogen-related proteins (FRePs) in response to parasite infection (7, 8, 9). FReP genes undergo low-frequency somatic diversification using a small set of source genes and point mutations (10). In amphioxus, five gene families encode the Ig domain-containing variable chitin-binding proteins (VCBP) (11). Each locus has many alleles that yield high levels of protein diversity among individuals. In insects, a novel system exists for generating immune diversity using the Down syndrome cell adhesion molecule (Dscam) gene, which also encodes Ig domains (12, 13, 14). Dscam is a single-copy locus in which 4 of the 24 exons exhibit substantial duplication that results in differential exon arrays. Although all of the exons in the arrays are transcribed, all but one of each exon variant are spliced out through mutually exclusive alternative splicing. The number of exons in the Dscam arrays, plus the combinatorial nature of the splicing, yields a potential transcript repertoire of 38,016 unique transcripts from a single gene (15). There are ∼16,000 unique Dscam transcripts expressed in mosquito hemocytes that encode proteins putatively involved in microbial phagocytosis (12). In higher plants, a large number of disease resistance genes are organized genomically in gene clusters, an arrangement that promotes diversification through sequence exchange via unequal crossovers and gene conversion (1, 16). The presence of environmental or pathogen stressors increases the frequency with which these diversification events occur in plants (17, 18, 19). In all, these studies suggest that mechanisms to diversify immune genes have evolved independently in multiple species, underscoring the role that this diversity putatively plays in defending the host against complex pathogen pressures.
The genome of the purple sea urchin, Strongylocentrotus purpuratus, has provided insight into immunological diversification mechanisms (20, 21). This species has long been a model for molecular and developmental biology in part based on its phylogenetic position at the base of the deuterostome clade. Sea urchins are members of the echinoderm phylum, which is the sister phylum to chordates. The echinoid class of echinoderms includes the sea urchins and sand dollars, while the other classes include sea cucumbers, brittle stars, sea stars, and crinoids or sea lilies. The close phylogenetic relationship between echinoderms and chordates makes investigations of the sea urchin immune system evolutionarily relevant. Inspection of the S. purpuratus genome has identified a previously unknown level of immune system complexity (20, 21). The genome contains homologs of the complement cascade components (reviewed in Ref. 22), as well as a number of large gene families that encode TLRs, NOD-like receptors (NLRs), and scavenger receptor cysteine-rich (SRCR) gene families (20). In vertebrates, these genes encode proteins that interact with pathogens either directly or indirectly. Many of the sea urchin immune gene families exhibit significant expansion in the S. purpuratus genome compared with their vertebrate counterparts. The sea urchin TLR gene family is composed of 222 gene models, compared with between 1 and 20 genes found in genomes of other animals (23). It has been suggested that the multiplicity of this gene family may increase the spectrum of recognizable pathogens (20) or enhance the specificity with which pathogens may be recognized (24). Similarly, there are 203 NLR gene models and 1095 SRCR domains within the S. purpuratus genome, as compared with ∼20 NLR genes and 81 SRCR domains in humans (20). Thus, the sea urchin, which occupies a unique phylogenetic position as an invertebrate at the base of the deuterostome clade, contains a complex immune system that is characterized in part by a number of large gene families.
The 185/333 gene family is another major component of the purple sea urchin immune response (25, 26, 27, 28, 29). Originally identified as 60% of the expressed sequence tags (ESTs) isolated from coelomocytes (immune cells; Table I⇓) following challenge with LPS (27), the 185/333 sequences are very unusual. They are homologous to only two previously uncharacterized sea urchin sequences, and may be specific to echinoids (Refs. 25, 27 and D. A. Raftos, unpublished observation). Optimal alignment of the 185/333 sequences requires the insertion of large gaps, which define blocks of sequence known as “elements”. The presence and absence of these elements define “element patterns” (see Fig. 1⇓ and Table I⇓) (27, 29). Elements range in size from 12 to 357 nucleotides, and each element is variable in sequence. Additionally, the 185/333 sequences have six different types of repeats that are present in both tandem and interspersed orientations. The significant diversity in the 185/333 sequences is primarily the result of the element patterns, and secondarily the result of point mutations.
Element patterns found in the genes and messages. Element patterns identified in genes are indicated by circles; patterns identified in messages are indicated with squares. The animal (1, 2, 4) from which each element pattern was isolated is indicated. A consensus sequence that includes all possible elements is shown at the top. The arrowhead below element pattern E2 indicates the position of the stop codon in the element pattern variant E2.1 (28 ). The subtype of elements Ex15 and Ex25 is indicated by the letters in the boxes. B9 is a newly described element pattern that is associated with intron ε (not shown) (26 ). The intron (not shown) is located between the leader (L) and the first element.
185/333 terminology
Although 185/333 expression has been observed in immunoquiescent sea urchins, there is a striking increase in expression following immune challenge with either whole bacteria (30), LPS (27), dsRNA, or β-1,3-glucan (28). In response to challenge with various pathogen-associated molecular patterns (PAMPs) and sham-injected injury controls that received artificial coelomic fluid, there is a change in the dominant element patterns of the messages and an increase in the number of minor patterns (28). Furthermore, different suites of 185/333 proteins have been observed by two-dimensional electrophoresis following challenge with LPS vs peptidoglycan (50). These data suggest that the sea urchin may be able to discriminate among different PAMPs by modulating the specific 185/333 genes expressed, although the mechanisms are unknown.
The function of the 185/333 proteins is currently unknown. The predicted proteins have a leader sequence, a glycine-rich region, a conserved RGD motif, a histidine-rich region, and do not contain any cysteine residues (27, 29). The proteins localize to cytoplasmic vesicles of two subsets of coelomocytes, small phagocytes, and polygonal cells, and despite the apparent lack of transmembrane region, they are also present on the surface of small phagocytes (31). The 185/333 proteins are associated with small phagocytes that are present in small cellular clots, which likely lead to syncytia formation (31). Preliminary data suggest that they may bind bacterial cell surfaces (our unpublished observations). Significant increases in protein and transcript expression following immune challenge, the observed diversity in the genes (25, 26), transcripts (28, 29), and proteins and the 185/333 proteins are expressed primarily by the immune cells suggest an immunological role (31). Early work on diverse immune response systems in other invertebrates, such as the VCBPs in protochordates (11) and FRePs in snails (7), similarly did not demonstrate function. However, like the FRePs and VCBPs, the general characteristics of the 185/333 system strongly suggests that that 185/333 proteins are involved in immune function.
The diversity among the 185/333 genes may have been driven by pathogen pressure. Detailed analysis of the repeats suggests that the evolution of the genes and the source of the diversity may be the result of a high rate of gene recombination (25). No recombination hotspots have been identified, and results suggest that recombination may occur at any point along the gene. Molecular clock analysis of the sequences suggests that the current 185/333 genes recombine rapidly and are the result of a recent diversification of the gene family. This frequent recombination generates mosaic or hybrid genes from parent sequences, which are speculated to be the basis for the broad and diversifying 185/333 protein repertoire.
The 185/333 sequences are extremely diverse. From 872 gene and message sequences isolated from 16 animals, 477 unique coding regions (CRs) with 51 distinct element patterns have been identified (26, 28, 29). Previous work, which analyzed messages and genes independently, has shown that all of the genes and half of the messages have CRs with stop codons in one of three positions in the terminal element (26, 29). The remaining 185/333 messages encode truncated proteins due to either a frame shift leading to missense sequence and a premature stop codon or a single nucleotide polymorphism (SNP) that introduces a premature stop codon (28).
In the present study, we have compared the 185/333 genes to the messages from three animals to understand the means by which the 185/333 system diversifies. Notably, we found that, although the repertoires of both genes and messages were diverse (26, 28, 29), very few genes and messages from individual sea urchins had identical sequences. That the gene and message sequences from individual animals differed so strikingly led us to analyze further the differences that characterized these two sets of sequences. Both the sequences of the elements and the types of element patterns in the genes were different from those in the messages. Despite the sequence differences, the gene from which each message was most likely transcribed was estimated based on the fewest number of changes between the two sequences. Results of this analysis suggested that most of the messages were the product of one gene per individual. The rest of the 185/333 genes were either expressed at very low levels or were not expressed in response to the immune challenges that have been employed. Detailed sequence comparisons between the genes and their assumed transcripts indicated a strong preference for cytidine to uridine substitutions, a change that is consistent with cytidine deaminase activity (32). Overall, the data indicate that the 185/333 gene and message sequences are different and suggest an unknown mechanism of posttranscriptional RNA editing that may be a previously unidentified mechanism for generating invertebrate immunological diversity.
Materials and Methods
Sequences and animals
The 185/333 gene sequences from animals designated 2, 4, and 10 employed in this study are available from GenBank (accession nos. EF607618–EF607793) (26). Genes from animal 1 were isolated and sequenced as in Ref. 28 and are reported here with GenBank accession nos. EU401669–EU401677 (supplemental File 1).5 Genes from animals 1, 2, and 4 were isolated from genomic DNA (gDNA) obtained from coelomocytes; genes from animal 10 were from sperm-derived gDNA (26). Transcript sequences from animals 1, 2, and 4 were described in Ref. 28 . The GenBank accession numbers of the 185/333 mRNA sequences can be found in supplemental File 2. From animal 1, 185/333 transcripts were isolated before and after challenge with LPS. 185/333 transcripts were isolated before and after sham injections with artificial coelomic fluid to measure the injury response in animal 4. Animal 2 was the source for 185/333 messages isolated before and after three separate challenges with LPS, β-1,3-glucan, and dsRNA (28).
Sea urchins 1, 2, 4, and 10 were outbred individuals that were obtained and maintained as described in Ref. 33 . Briefly, sea urchins were collected from the nearshore waters off the coastline of southern California by SCUBA divers and maintained in the marine aquarium facility at George Washington University. There is an estimated 4% difference between individual S. purpuratus animals (34), and it is unlikely that animals 1, 2, 4, and 10 were related. Based on the size of the animals, we estimated that they were likely more than 5 years old, and perhaps as old as 50 years (21, 35).
Bioinformatics
Sequences were aligned according to the cDNA-based alignment (26, 28, 29). BioEdit (36) was used to manipulate and optimize the alignments. PERL scripts were used to calculate diversity scores and diversity characteristics (26, 29). Diversity scores are a measure of entropy (37) and are a function of the number and frequency of nucleotides in each position within the alignment. The maximum possible score occurs given an equal distribution of each of the four nucleotides among the sequences. Diverse sequences have higher diversity scores; conserved sequences receive a diversity score of 0. The diversity score of an alignment is the average score of each of the alignment positions. The gene from which each message was transcribed was estimated using a PERL script. Sequences from each gene and message were compared in a pairwise fashion to calculate the number of changes required to generate each message from each gene. Nucleotide substitutions were scored as 1, and gaps were scored as 10 regardless of the length of the gap. For each message, the gene with the lowest score was considered to be the most likely source. Messages for which the lowest gene score was >50 were considered to be “orphans” with no known source gene. This cutoff was chosen based on a histogram of the lowest gene scores (supplemental File 3).
Results
The 185/333 sequences contain six types of repeats (25, 27) that facilitate at least two different alignments (26, 29). Sequences can be aligned according to the locations of gaps used for the cDNAs, which has been called the cDNA-based alignment (Fig. 1⇑) (29), or based on the locations of the repeats as the primary determinant in defining the element borders (26). Comparisons between the two alignments did not indicate significant differences with regard to the level of sequence diversity, the number of elements or element patterns, or the identification of gene recombination (25). Because the transcripts have been analyzed previously using the cDNA-based alignment (28, 29), that alignment is used here.
Unique sequences
In previous studies, 185/333 genes were cloned from gDNA isolated from three animals (animals 1, 2, and 4) and messages were isolated from the same three animals before and following challenge with various PAMPs (26, 28). The sequences of the genes and mRNAs were characterized separately in previous work (26, 28); however, the results presented here are the first comparison between the genes and messages from the three animals. The sequences were analyzed to determine how many unique sequences (those that did not have identical nucleotide sequence with any other gene or message) were found among the genes and the messages, and how many unique messages were present before and after immune challenge to facilitate subsequent comparisons between the messages and the genes.
The genes and messages from animals 1 and 4 were largely unique, both with respect to messages isolated before vs after challenge and when genes and messages were compared. The majority (62–86%) of the messages collected from animals 1 and 4 before and after immune challenge were unique (Table II⇓). All but two of the 185/333 genes from animals 1 and 4 had unique CRs (4-1520 and 4-1532; supplemental File 1). None of the CRs from the genes isolated from animal 1 matched identically to any of the messages. Animal 4 had CRs from two genes that were identical to six messages (Table II⇓ and supplemental File 1) (26). The genes and messages from animals 1 and 4, however, constituted smaller datasets than those from animal 2, which was more exhaustively sampled. It was therefore possible that the lack of identity between the gene and message sequences was due to incomplete sampling rather than a true discord between the sequences. Based on the larger datasets, a more thorough analysis of the genes and messages from animal 2 was undertaken.
185/333 messages isolated before and after different challenges have different sequences than the 185/333 genes
Animal 2 was challenged separately with three PAMPs (LPS, β-1,3-glucan, and dsRNA) over a period of about 4 years (28). Unique messages constituted 56–95% of the messages sequenced, depending on the immune challenge (Table II⇑). Although no identical messages were found both before and after challenge with the same PAMP, a few identical messages were identified multiple times during the different immune challenges of animal 2. The suites of messages expressed before and after challenge were both equally diverse and were also largely different from one another. There were 51 unique CRs of 53 cloned genes from animal 2 (Tables II⇑ and III⇓ and supplemental File 1) (26). Comparison between the sequences of the genes and messages indicated that only five sequences were common to both genes and messages (Table III⇓). The extensive analysis of animal 2 indicated that not only was a single animal capable of expressing at least 89 unique 185/333 transcripts under immunoquiescent and postchallenge conditions, but that these transcripts were distinct from the relatively large number of unique alleles characterized from animal 2. This surprising result, that the 185/333 messages generally differed from the genes in individual sea urchins, prompted further investigation into the specific differences that characterized the two sets of sequences.
Few 185/333 genes and messages have identical coding regions
Element patterns
Gene element patterns.
Changes in transcript element patterns have been used to infer changes in 185/333 gene expression before vs after challenge with PAMPs (28). Similarly, this approach has been used here to compare message element patterns to gene element patterns for three individual sea urchins to determine whether the two data sets only differed in sequences as described above, or whether the differences extended to element patterns. For animal 1, six element patterns were found among nine genes (Table IV⇓) including E2, which was the most common element pattern identified from the messages (28, 29), and D1, which was most common from the genes (26). In addition, two new element patterns, B9ε and B3δ, were observed among the genes cloned from animal 1 (Supplemental File 1). The 53 genes from animal 2 had 14 different element patterns, while animal 4 had 10 element patterns from 29 genes (Table IV⇓). There were fewer element patterns than unique sequences because some genes with the same element pattern had different sequences. In total, 21 element patterns were identified from 92 unique 185/333 genes from three animals.
Element patterns of the 185/333 genes and messages
Message element patterns.
From the 29 unique messages isolated from animal 1, three element patterns were identified (Table IV⇑). The messages from animal 2 had 12 different element patterns from 89 unique messages, and animal 4 had 10 element patterns from 43 distinct messages. Element patterns D1 and E2 were the only patterns shared among all three animals. Similar to the genes, there were many messages that shared element patterns but had different sequences. Although numerous element patterns were identified from genes and messages, it was surprising that very few of the patterns were shared between the two sets of sequences. From animal 1, only patterns E2 and D1 were observed in both genes and messages. Five element patterns were shared between genes and messages for animals 2 and 4 (Table IV⇑). The message repertoires were characterized by major and minor patterns (Table I⇑) (28). The major patterns were common to both genes and messages and were present in messages isolated from multiple animals following various antigenic challenges. It has been suggested that certain element patterns may be common among individual sea urchins (e.g., 01, D1, and E2, which have been identified in genes and messages of every animal analyzed) but that each individual may also maintain a repertoire of less common, minor patterns (26, 28). When the frequencies of the element patterns were considered, however, it was apparent that the two groups of sequences were also different. For example, 81% of messages isolated following various Ag challenges were either E2 or an E2 variant, as compared with 9% of genes. Therefore, although the major element patterns were represented in both genes and messages, their relative frequencies differed considerably. Furthermore, the minor element patterns were less likely to be shared.
Variant element patterns.
Previous analysis of the 185/333 transcripts showed that a number of variant element patterns had premature stop codons (28). The most common example was variant E2.1, which had the same elements as E2 messages, but included a SNP in element Ex10 that altered a histidine codon to a stop. In addition to E2.1, 11 other element pattern variants were identified. Half of the element patterns from nine sea urchins had messages with frame shifts and early stop codons that encoded truncated proteins (28). No variant element patterns, including E2.1, have been observed among the sequenced 185/333 genes (26), suggesting that variant element patterns are a feature unique to the 185/333 messages.
Sequence diversity of genes and messages
The sequence diversity of the 185/333 system has been well documented (26, 27, 28, 29) and is thought to stem, at least in part, from frequent recombination among members of the 185/333 gene family (25). Diversity of genes and messages has been illustrated through 1) entropy-based diversity scores of both the full-length sequences and the individual elements that are calculated using the frequency of each nucleotide present in an alignment position (37), 2) the percentage of variable positions, 3) the number of SNPs that are present only in a single sequence (unshared polymorphisms), and 4) the average number of unique element sequences (26). To characterize the sequence differences between the genes and messages from individual animals, the attributes listed above were calculated for the two datasets separately and in combination for each animal both as complete alignments and as individual elements.
Alignment diversity.
Diversity scores and variable positions: Results from these analyses supported the previous observation that full sequences of both the genes and messages from individual animals were diverse (Fig. 2⇓A) (26, 28). Diversity scores for the alignments of either genes or messages ranged from 0.0397 to 0.2745 (messages from animals 1 and 4, respectively; Fig. 2⇓A). However, when the gene and message sequences were combined into a single alignment, the diversity score was consistently higher than would be expected given the diversity scores of gene and message alignments alone (Fig. 2⇓A). Similarly, although between 5.8% and 16.7% of the alignment positions were variable when genes and messages were analyzed as distinct groups (messages from animals 1 and 2, respectively; Fig. 2⇓B), when the two groups were combined, the percentage of variable positions increased. These data suggested that, although animals maintained diverse gene and message repertoires, the two sets of sequences were different from one another.
Diversity characteristics of the genes and messages. A, Diversity of both full-length sequences and elements of genes and messages is similar. Diversity scores (26 ,29 ,37 ) were calculated from alignments of the full-length 185/333 genes and messages, as well as the average diversity of individual elements from the genes and messages. Diversity of the full-length gene sequences is indicated by bars with black diagonal lines, messages by bars with black checkered boxes, and the two datasets combined as solid black bars. The average diversity of the elements from the genes is indicated by bars with gray diagonal lines, messages as bars with gray checkered boxes, and the combined data sets as solid gray bars. The t tests indicated that the diversity of the elements isolated from genes vs messages was not statistically different (p > 0.05). B, The number of variable positions and average number of unshared SNPs per sequence are similar among genes and messages. The percentage of positions that are variable is indicated with black bars. The average number of unshared SNPs per sequence is shown with gray bars. Diversity values calculated using the genes only are shown with diagonal lines and messages only are shown with checkered boxes. Diversity characteristics calculated using the gene and message sequences together are indicated with solid bars. The percentage of variable positions or unshared SNPs in genes and messages were not statistically different, as determined using two-tailed t tests (p > 0.05). C, Less than 37% of element sequences are common to 185/333 genes and messages. The total number of unique element sequences from the genes (black circles) and messages (gray circles) and the number of element sequences common to both genes and messages were calculated and are shown in the intersection of the circles. The total number of element sequences and the percentage shared are indicated.
Unshared polymorphisms: The number of SNPs observed in single sequences was calculated for genes and messages from each animal, as well as for sequences from all animals combined, to assess the frequency of random point mutations. An increased frequency of random point mutations present in the messages may suggest nondirected posttranscriptional editing. Results show that there were fewer unshared SNPs per sequence in animal 2 compared with animals 1 and 4 (Fig. 2⇑B). However, the low numbers of sequences collected from animals 1 and 4 artificially inflated the numbers of unshared SNPs. Single messages with novel element patterns from animals 1 and 4 were the source of 92% and 75% of the unshared SNPs, respectively. When the outlier messages were removed from the data for animals 1 and 4, reanalysis indicated that results were similar to the results for animal 2 (not shown). When the sequences from animal 2 were analyzed, there was no difference in the number of unshared SNPs between the genes and messages. Overall, there was no significant difference in the number of unshared SNPs between genes and messages, suggesting that random posttranscriptional modifications do not appear to be altering the 185/333 messages.
Element diversity.
Diversity of individual element sequences was also analyzed to circumvent the problem of the gaps introduced to compensate for the variations in element patterns (26, 29). Diversity scores and the number of unique sequences were calculated for each element (Fig. 2⇑, A and C). In agreement with previous observations (26, 29), the average diversity scores of the elements were lower than the scores of the full-length alignment because element alignments eliminated the problems introduced by the large gaps required to align full-length sequences. No significant difference was observed between the genes and messages with regard to the diversity scores or number of unique element sequences. When the element sequences from genes and messages were compared, an average of 28% were shared among both genes and messages (Fig. 2⇑C). Thus, although some element sequences were shared among genes and messages, variations in element patterns plus variations in element sequences resulted in very few shared full-length sequences.
Matching messages to genes
Although very few of the message sequences matched exactly to genes from within individual animals (Table III⇑), it was of interest to identify from which gene each message was most likely transcribed. To this end, messages were compared in a pairwise manner to genes from each individual animal and scored for substitutions and gaps. For each message, the gene with the lowest score was considered to be the most likely source (Table V⇓). Of the 245 messages analyzed, 221 were most likely the products of only 24 genes (Table V⇓). The remaining 24 messages included 6 that were too short (<200 nucleotides) for meaningful analysis, and 18 for which the optimal source gene had a score of >50 (supplemental File 3), and therefore they were considered to be orphans and derived from genes that had not been cloned and sequenced. For each animal, most of the messages appeared to have been transcribed from a single gene (Table V⇓). For example, 70% of the messages from animal 2 were most similar to a single E2 gene. Similarly, 54% of the messages from animal 4 and 94% of messages isolated from animal 1 were also likely the products of E2 genes. From animals 1 and 2, no messages were identical to the genes from which most of the messages were most likely transcribed (genes 1-1517 and 2-063, respectively; Table V⇓). In contrast, four messages from animal 4 that were isolated following sham injection were identical to gene 4-1550 (element pattern E2; supplemental File 1), which was the most likely source for most of the messages from animal 4 (Tables II⇑ and IV⇑). Previous work predicted that gene expression changed in response to challenge with PAMPs and was the source of sequence changes in the messages (28). However, more detailed analysis of the data presented herein predict that a large majority of the messages isolated both before and after challenge were most likely transcribed from a single gene. This unexpected result, given the gene family size (25, 29) and that was observed in each of the three animals, prompted us to characterize further the differences between the genes and messages.
The majority of the genes were likely transcribed from a few genes
All of the messages from animal 2 that were most likely transcribed from gene 2-063 were isolated either before or after challenge with either dsRNA or β-1,3-glucan; none was isolated either before or after challenge with LPS. All 10 of the messages collected before LPS challenge were classified as orphans (Table V⇑). Of the 12 messages isolated following LPS challenge, 3 were identical to known 185/333 genes (Table III⇑), 5 were likely the product of gene 2-107, and the remaining 4 were similar in sequence to four different genes (Table V⇑). Most of the messages isolated before and after dsRNA and β-1,3-glucan challenge were derived from gene 2-063 (Table V⇑). Similarly, most of the messages isolated before and after injury from animal 4 and LPS challenge from animal 1 were likely derived from single genes. Therefore, despite the extraordinary diversity observed within both the gene family and transcript repertoire of single animals, it appears that only a small fraction of the genes were highly expressed under the immunological challenges employed.
To better characterize the specific sequence differences between the genes and messages, each message was compared with the gene from which it was most likely transcribed, and the number of each type of nucleotide substitution was counted (Fig. 3⇓). The overwhelming majority of the changes (73%) were transitions (either a purine substituted for a purine or a pyrimidine in place of a pyrimidine). In all, the predicted changes between the genes and messages had a 3:1 transition/transversion ratio. This is much higher than the expected 1:2 ratio that would occur if transitions and transversions (a purine substituted for a pyrimidine, or vice versa) occurred randomly (there are four possible transitions and eight possible transversions). Notably, 30% of the substitutions resulted from a cytidine in the gene and a uridine in the message. This percentage of cytidine to uridine transitions is significantly higher (p < 0.001) than expected given typical RNA polymerase activity, and is consistent with the activity of cytidine deaminases (32).
Nucleotide differences between the genes and messages suggest a strong cytidine to uridine preference. Messages were compared with the genes from which they were most likely transcribed, and the number of each type of nucleotide substitution was determined. The size of each box indicates the percentage of total substitutions for three analyses: (A) when both genes and messages were included, (B) when only those messages were included that were likely transcribed from gene 2-063, or (C) when only those messages that were not transcribed from gene 2-063 were included. The size of the boxes is scaled to percentage, as shown at the bottom of the figure. Missing boxes indicate that those transitions or transversions did not occur. Gray boxes indicate transversions; black boxes indicate transitions.
185/333 sequence diversity occurs throughout the length of the sequence; there are no hypervariable regions (26, 29). Likewise, the differences between the genes and messages were located throughout the 185/333 sequence, with no obvious pattern of changes (Fig. 4⇓). However, there were certain positions within the alignment in which most of the messages were different from their corresponding gene (Fig. 4⇓). Three categories of positions showed different numbers of messages with altered nucleotides. The top category included 17 positions in which more than 110 messages (of 207 total) differed from the genes from which they were most likely transcribed. Positions in the medium category were such that between 24 and 50 of the messages differed from their corresponding genes. The bottom category was defined as those positions in which <15 of the messages were different than the genes. Therefore, although differences between the gene and message sequences occurred throughout, certain positions appeared to be altered more frequently than others.
Locations of variant nucleotide positions in messages relative to their corresponding gene. For each alignment position, the number of messages with a different nucleotide than the gene from which they were most likely transcribed is indicated by bars. The number of messages with substitutions relative to the corresponding gene for each position could be categorized into three levels: low (<12), medium (25–50), and high (>110), as indicated by the gray shading. The short bars below the x-axis indicate the locations of the element borders. A total of 217 variant nucleotide positions were observed. For most of these positions (89 positions), a single message was different from its corresponding gene. However, for 17 positions, >100 messages differed from their corresponding genes.
Diversity of sperm- vs coelomocyte-derived genes
Because the differences between genes and messages were heavily biased toward cytidine to uridine transitions, we investigated whether the discrepancy between the gene and message sequences was the result of an unknown mechanism generating somatic diversification that occurred specifically in coelomocytes. 185/333 genes have been analyzed from coelomocyte gDNA (animals 1, 2, and 4), as well as gDNA isolated from sperm (animal 10) (26). The diversity of the genes isolated from these two tissue types was analyzed to address this question (supplemental File 4). No significant differences were observed in the average diversity scores for elements or for full-length sequences, nor for the percentage of variable positions for genes isolated from either tissue source. To understand further the relationship between genes isolated from sperm and coelomocytes, pairs of putative homologous genes were identified via the pairwise comparison technique that was used to correlate gene and message sequences. Analysis of the difference between pairs of genes revealed a slight transition bias (a 5:4 transition/transversion ratio; data not shown) in which most changes were either adenine to guanine or guanine to adenine. The strong preference for cytidine to uridine transitions identified in comparisons between genes and messages (Fig. 4⇑) was not observed when gene repertoires from sperm and coelomocytes, as well as genes from different individuals, were compared. The diversity of the members of the 185/333 gene family was not elevated in genes isolated from coelomocyte gDNA and suggested that a coelomocyte-specific somatic diversification mechanism may not occur.
Discussion
The data presented herein indicate that the 185/333 genes and messages isolated from single animals are largely different and suggest a putative posttranscriptional editing mechanism that favors transition substitutions. In addition to sequence differences, the genes and messages share the same major element patterns, but have different minor element patterns. Although the sequence diversity within the message repertoire is similar to that of the genes, the increased diversity of both full-length and element alignments of genes plus messages compared with separate alignments indicates that the sequences of the genes and messages differ. Despite the large size and diversity of the gene family, it appears that only a few genes are the source of most of the messages. This implies that most of the sequenced 185/333 genes may not be not expressed under the immunological challenges that have been employed to date.
The results presented herein rely on accurate sequences of the 185/333 messages and genes. Because the 185/333 sequences were amplified by PCR before cloning, and were analyzed by cycle sequencing, previous studies have investigated two possible sources of errors that would artificially inflate sequence diversity: Taq-induced errors (38) and template switching (39, 40). Given the conditions used to amplify and clone the 185/333 messages, the error rate of the Taq polymerases was previously analyzed by amplification, cloning, and sequencing a single 185/333 cDNA, which was determined to be one error per 18 clones (28). Similarly, the error rate for the cloned and sequenced 185/333 genes was determined to be a maximum of one error in 15 clones (26). The error rate of the cycle sequencing method was tested by sequencing a single clone multiple times and was determined to be <0.005% (26). Therefore, it is unlikely that the disparity observed between the gene and message sequences is driven by Taq-induced errors alone.
Template switching is another possible source of error, and it is a process by which artificial, recombinant DNA molecules are generated during PCR amplification when multiple homologous templates are present in the reaction (39, 40). Because the 185/333 gene family and transcript repertoire are both composed of pools of very similar sequences, it was important to determine the role that template switching might play during the amplification of these sequences. The conditions for PCR in which the 185/333 genes were originally amplified were mimicked, and the rate of template switching for a pair of clones was assessed using pairs of nested primers and measuring how frequently their orientation with respect to each other switched in the amplicons that were generated (26). Template switching occurred in <1 in 1100 amplicons, which would be undetectable in the ∼100 clones isolated from each animal. Consequently, template switching is unlikely to be a source of sequence diversity for the set of 185/333 genes and messages analyzed herein and in previous studies (25). Because the rates of Taq-induced misincorporation and template switching are below thresholds that would affect the sequence diversity, this indicates that the observed differences among sequences presented herein are reliable.
The size of the 185/333 gene family has been estimated to be between 80 and 120 alleles using three different approaches: quantitative PCR with gDNA as the template (26), a statistical estimation based on the number of genes sequenced from three animals vs the frequency that duplicate genes were identified (25), and a rough analysis of numbers of 185/333+ bacterial artificial chromosomes (BACs), including estimates of the average size of the BAC inserts, genome coverage by the BAC library, average size of the 185/333 genes, and distance between clustered genes (our unpublished results). If each allele has a unique sequence, only about half of the alleles from animal 2 have been sequenced (53 unique alleles) (26). It is possible that the genes from which most of the sequenced messages in animal 2 were transcribed have not been cloned, and that this may account for the discord between the gene and message sequences. However, based on the total number of different 185/333 sequences isolated from animal 2, this hypothesis may not be feasible. In addition to the 53 unique genes cloned from animal 2, 89 additional unique transcripts were identified, for a total of 135 unique sequences, which is more than the estimated number of alleles in the 185/333 gene family (25, 29). If about half of the 185/333 genes from animal 2 have been isolated, then there was an ∼50% possibility that the sequenced genes were the source of most of the messages. Given that eight unique E2 genes were isolated from animal 2 (26), and if the genes were cloned in the same proportion in which they exist in the genome, we can estimate that there may be between 12 and 18 E2 alleles in an individual animal (based on a gene family size of 80–120 alleles and a range of 40–60% chance that the correct gene was isolated). However, 54 unique transcripts with either E2 or E2.1 element patterns were sequenced, which is many more than the estimated number of E2 alleles. Although posttranscriptional editing could increase the sequence diversity of the messages, it is not likely to change the element patterns of either the genes or the messages. Thus, that more unique E2 transcripts were identified than the estimated number of E2 genes within an individual sea urchin supports the hypothesis of posttranscriptional diversification of the messages rather than purely gene-encoded diversity.
Generating the diversity
Cytidine deaminase activity.
Analysis of the differences between the messages and the genes from which they were likely transcribed shows a strong bias toward transitions, particularly cytidine to uridine substitutions. This type of change is often the result of deamination reactions that modify cytidine by a hydrolytic deamination at the C4 and C6 positions of the purine base (reviewed in Ref. 41). The most commonly known form of cytidine to uridine editing occurs in the vertebrate mRNA that encodes apolipoprotein B and changes a glutamine (CAA) to a stop (UAA), which truncates the protein (42, 43) (reviewed in Ref. 32). This editing reaction is mediated in part by the APOBEC-1 protein (44), which associates with other proteins, including the APOBEC complementation factor, to form an editing complex known as the editosome (41, 45). Three mammalian proteins are related to APOBEC-1, with the most interesting from the view point of immune function being AID. AID is involved in generating Ig diversity through somatic hypermutation and class switch recombination (reviewed in Ref. 46), and it has also been implicated in the assembly of cyclostome variable lymphocyte receptor genes (6). Additionally, all major lineages of higher plants exhibit cytidine to uridine editing in mitochondrial and chloroplast mRNAs and tRNAs (47). The mechanisms for mRNA editing in plants remain largely unknown (41).
Analysis of the sea urchin genome reveals a few annotated homologs of the RNA editing proteins (21). Several cytidine deaminases have been annotated, but phylogenetic analysis suggests that these proteins are not homologous to either AID or members of the APOBEC enzyme family (20). BLAST searches of the sea urchin EST sequences with the mouse APOBEC-1 sequence (GenBank accession no. NM_031159.3) did not result in any significant matches, suggesting that homologs to APOBEC-1, if they exist within the S. purpuratus genome, are not commonly or highly expressed. In contrast, there are sea urchin gene models that encode APOBEC complementation factor-like proteins (SPU_ 011837 and SPU_011875). Because the RNA editing proteins are small and rapidly evolving, they may be too divergent for phylogenetic analysis of the sequences to suggest a function (20). Given that the 185/333 messages appear to differ from the genes both before and after immune challenge suggests that this diversification may be constitutive, rather than a specific response to immune challenge, as in AID. If constitutive RNA editing is involved in this system, this suggests that a non-AID-like deaminase may be involved. Thus, although some of the mRNA editing components have been identified in the sea urchin, it is not clear whether any have deaminase-like function.
Low-fidelity polymerases.
In addition to possible RNA editing activity, other proteins may be involved in diversification of the 185/333 messages during transcription. One such enzyme identified in the sea urchin genome is the TdT/polymerase μ (TdT/Polμ; SPU_009980) homolog (20, 24). In higher vertebrates, TdT/Polμ is involved in Ig diversification and has been implicated in low-fidelity DNA replication (48, 49). Although no significant matches to TdT/Polμ have been identified among the currently available sea urchin ESTs, expression may be restricted to specific times or tissues or may be below detection.
The diversity of 185/333 genes isolated from coelomocyte gDNA is comparable to the diversity of sperm-derived 185/333 genes, which suggests that deaminases do not somatically diversify the 185/333 genes. Despite the fact that variability among S. purpuratus individuals is estimated to be 4% (21, 34), comparison of the genes isolated from sperm vs coelomocyte gDNA from different animals did not reveal a cytidine to uridine transition bias similar to that observed between the genes and messages from single animals. Based on the data presented herein, message diversification appears to be the result of an unknown posttranscriptional RNA editing mechanism that, in many cases, alters specific nucleotides that change specific codons resulting either in the incorporation of a different amino acid into the encoded protein or in a stop. Changes also include small insertions/deletions that alter the reading frame and result in missense amino acid incorporation typically leading to a stop. It is of note that half of the messages encode truncated proteins (28) of which at least some are expressed (Dheilly et al., submitted), suggesting that the benefits of diversification through this method may be limited by the introduction of premature stop codons that may alter protein function. Alternatively, given that as many as 55% of the E2 messages are altered to the E2.1 variant with a single early stop codon (28), it is possible that the N-terminal, glycine-rich region of the truncated protein may have a function that is independent of the histidine-rich C terminus.
Conclusions: two levels of diversity
The diversity of the 185/333 genes may be, in part, the result of frequent recombination among the genes (25). This recombination may be facilitated by a variety of repeats within the genes, by di- and tri-nucleotide repeats that flank the genes and are present in the short intergenic region, and by the close linkage among the genes within the genome (26). This putative propensity for recombination has been proposed as a major diversification system acting within the 185/333 gene family to generate a large number of unique genes from a limited number of element sequences. As a result, the members of the 185/333 gene family have a mosaic or patchwork appearance of elements. It is therefore unexpected that a second level of diversification may be involved in this system in the form of edited messages (and perhaps poor fidelity transcription), and that, despite maintaining a large and diverse gene family, only a few of the genes may be transcribed. Future analysis of the cis-promoters, predictions of binding sites for transcription factors, and the regulation of 185/333 gene expression may show that most of the genes may not be expressed, but may only function as a source for gene recombination. The 185/333 system in the sea urchin therefore represents an intriguing example of invertebrate immunological diversity that appears to be generated not only at the genomic level, but also posttranscriptionally by acting on the mRNA sequences.
Acknowledgments
We thank Dr. David Raftos and the anonymous reviewers for input that resulted in improvements to this manuscript.
Disclosures
The authors have no financial conflicts of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This research was supported by funding from the National Science Foundation (MCB-0424235) (to L.C.S.) and a Weintraub Fellowship from the George Washington University (to K.M.B.).
↵2 Current address: College of Osteopathic Medicine, Touro University, Henderson, NV 89014.
↵3 Address correspondence and reprint requests to Dr. L. Courtney Smith, 340 Lisner Hall, 2023 G Street NW, Washington, DC 20052. E-mail address: csmith{at}gwu.edu
↵4 Abbreviations used in this paper: AID, activation-induced cytidine deaminase; APOBEC, apolipoprotein B mRNA editing catalytic component; CR, coding region; EST, expressed sequence tag; gDNA, genomic DNA; FReP, fibrinogen-related protein; NLR, NOD-like receptor; PAMP, pathogen-associated molecular pattern; Polμ, DNA polymerase μ; SNP, single nucleotide polymorphism; SRCR, scavenger receptor cysteine-rich; VCBP, variable chitin-binding proteins.
↵5 The online version of this article contains supplemental material.
- Received July 22, 2008.
- Accepted October 13, 2008.
- Copyright © 2008 by The American Association of Immunologists
References
In this issue
