Hagfish were sedated by immersion for 15 min in 0.5 g/liter MS222 (Sigma) buffered to pH 7 before i.p. injection with an antigen/mitogen mixture in 0.5 ml of hagfish PBS (per liter, 28 g of NaCl/0.2 g of KCL/1.44 g of Na₂HPO₄/0.24 g of KH₂PO₄, pH 7.4, 1 osmol). The mixture contained 10⁹ live Escherichia coli TG1 bacteria, 10⁹ sheep erythrocytes (Colorado Serum, Denver) and 100 μg each of phytohemagglutinin and pokeweed mitogen (Sigma). Immune stimulation was repeated at weekly intervals and, 4 days after the fourth stimulation, blood was collected with a syringe from the tail blood sinus and diluted 1:1 with hagfish PBS containing 30 mM EDTA. Buffy coat leukocytes collected after 5-min centrifugation at 50 × g were sorted by their light-scatter characteristics as described (8, 9) by using a MoFlo cytometer (Cytomation, Ft. Collins, CO).

Hagfish VLR. Inshore hagfish Eptatretus burgeri VLR homologs were identified by using lamprey VLR as blast queries against the database of expressed sequence tags from leukocyte RNA of unstimulated animals nos. 7 and 8 (16). Clones with significant matches were sequenced on both strands, 64 VLR-A and 15 VLR-B cDNA clones. For the Pacific hagfish, unseparated blood cells and buffy coat leukocytes from three unstimulated individuals (nos. 1-3 and 6) and buffy coat leukocytes from two immunostimulated animals (nos. 4 and 5) were used for extraction of blood genomic DNA and leukocyte RNA. Extraction of RNA was with TRIzol Reagent (Invitrogen), and PolyA RNA was selected with the Dynabeads mRNA purification Kit (Dynal, Lake Success, NY). First-strand cDNA synthesis was primed with 20 pmol of HgVLRA.F1 (Table 2, which is published as supporting information on the PNAS web site) for VLR-A or HgVLRB.F1 for VLR-B, using the SuperScript III First-Strand cDNA Synthesis kit (Invitrogen), and the products were column-purified (QIAquick PCR purification, Qiagen, Valencia, CA). VLRs were then PCR-amplified by using Expand High Fidelity PCR (Roche Applied Science, Indianapolis) from the cDNA or from genomic DNA, in 50-μl reactions containing 1 μl each of the sets of forward and reverse primers (F1 or F2 and R1 or R2) at 10 pmol/μl, 5 μl of 10× buffer, 36.25 μl of double distilled water, 5 μl of cDNA or genomic DNA (250 ng), and 0.75 μl of the polymerase. Reactions were amplified by using one cycle of 94°C, 1 min; then 35 cycles of 94°C, 30 sec; 58°C, 30 sec; 72°C, 1 min; and a final 7-min elongation at 72°C. Products were column-purified, cloned in pCRII-TOPO (Invitrogen), and the inserts were sequenced. For the Pacific hagfish, 109 VLR-A RT-PCR clones were sequenced (four contained in-frame stop codons; not shown), and 36 genomic mature VLR-A amplicons (two contained in-frame stop codons). For VLR-B, 37 RT-PCR clones were sequenced (one contained an in-frame stop codon) and 38 genomic mature VLR-B amplicons (four contained in-frame stop codons; not shown). Liver genomic DNA from Inshore hagfish no. 9 (16) was used for PCR cloning and sequencing mature VLRs, four mature VLR-A amplicons (two contained in-frame stop codons), and three mature VLR-B amplicons.

Nonparasitic Lamprey VLR. First-strand cDNA was synthesized as above by using the reverse primer VLR_3UT.R (Sea lamprey 3′ UTR primer, Table 2). For the American brook lamprey, the forward primer was Slit.F (Sea lamprey 5′ UTR primer) and for the Northern brook lamprey, LRR_N.F1 (another Sea lamprey 5′ UTR primer). In total, 13 unique VLR clones of the American brook lamprey and seven of the Northern brook lamprey were sequenced.

Bacterial Artificial Chromosome (BAC) Libraries and Clones. An Inshore hagfish BAC library (20) was screened by PCR by using VLR primers as above (F1 or F2 and R1 or R2). The Pacific hagfish BAC library (VMRC23) was constructed from EcoRI partial digests of erythrocyte DNA from a single specimen in the vector pCCBACE1 (Epicentre Technologies, Madison, WI). This library consists of ≈184,000 recombinants and encompasses ≈5× coverage of the hagfish genome. The entire library was screened by hybridization with 5′ and 3′ VLR-A and -B probes, and positive clones were authenticated by PCR. One BAC for each VLR type from the Pacific and Inshore hagfish was sequenced at ≈10× coverage and assembled into contigs (Macrogen, Seoul, Korea). In case of incomplete sequence of the inserts, only portions containing the gVLR and LRR cassettes were included with uncaptured gaps in the contigs: Eb_gVLR-A, 43,362 bp; Eb_gVLR-B, 92,072 bp; Es_gVLR-A, 81,648 bp; and Es_gVLR-B, 76,730 bp.

Transcripts of hagfish VLR are abundant in lymphocyte-like cells but not in myeloid cells or erythrocytes sorted by their light-scatter characteristics. VLR-A transcript levels were ≈3-fold higher than VLR-B levels in blood leukocyte samples (not shown). Both VLR types of the Pacific hagfish are highly heterogeneous (Fig. 2), exhibiting variable numbers of the 24-residue LRR modules and pronounced LRR sequence diversity. Comparable diversity was observed for VLR-A (n = 66) and -B (n = 18) sequences from Inshore hagfish (Fig. 5, which is published as supporting information on the PNAS web site). Interestingly, five clusters of two to four VLR-A clones that were identical or differed by only one to two residues were found among the 40 transcripts from hagfish no. 5 (marked by asterisks in Fig. 2 A), which was given four weekly injections of an antigen and mitogen mixture. The finding that 30% of the VLR-A transcripts from this hagfish consisted of clusters of related sequences may indicate clonal expansion of VLR-A-bearing lymphocytes. The clones with one to two amino acid substitutions could reflect additional VLR diversification through somatic hypermutation. The elucidation of the implied role for clonal expansion and diversification in hagfish immunity will require additional experiments.

The dataset of unique sequence Pacific hagfish VLR-A (n = 130) reveals two to six copies per transcript of the 24-residue LRRs (n = 527; average, four). In the VLR-B dataset (n = 69), there are one to six copies of the 24-residue LRRs (n = 195; average, 2.8), whereas in the set of 129 Sea lamprey VLR (ref. 19; GenBank accession nos. AY577943-AY578059), there were one to nine copies of 24-residue LRRs (n = 325; average, 2.5). We then analyzed separately in a neighbor-joining phylogenetic tree the individual components of these VLR, except for LRRNT and LRRCT that were too diverse among the species for reliable alignment (Table 1; 328 LRR1 domains, 328 CP domains, and 1,047 single domains of the 24-residue LRRs). The clusters were nearly exclusively of the same type and species origin, i.e., Pacific hagfish VLR-A, VLR-B, or Sea lamprey VLR clustering. There were no instances of identical LRR domains between the different VLR types. However, a large portion of the LRR1 and CP domains within hagfish VLR-A and lamprey VLR clusters were identical (Table 1). In contrast, the LRR1 domains in hagfish VLR-B were 98% unique; the sets of 24-residue LRRs also consisted predominantly of unique sequences: 97% were unique in hagfish VLR-B, 90% in VLR-A, and 83% in the Sea lamprey VLR. This remarkably high degree of diversity is especially remarkable, given that consensus sequences derived for each of the 24-residue LRR types share at least 10 framework residues.

Hagfish VLR Genes. Genomic organization of the Pacific and Inshore hagfish VLR loci was determined from sequences of large insert genomic clones isolated from BAC libraries, one BAC for each VLR type (Fig. 3). Only one copy of each of the gVLRs was identified in hagfish genomes (not shown). The sequences and organization of the loci are nearly identical in both species and fairly conserved between gVLR-A and -B. Hagfish gVLR begin with a 5′ UTR followed by two coding regions (Fig. 4A). As in the Sea lamprey gVLR, the 5′ UTR is split by an intron, 6.4 kb long in the Pacific hagfish gVLR-A and 220 bp long in gVLR-B. The first coding region in the hagfish gVLR encodes the signal peptide and an LRRNT domain in gVLR-A and only residues 1-13 of the 23-residue signal peptide in gVLR-B. Next, there are short intervening sequences of 171 and 105 bp for gVLR-A and -B, respectively. The second coding region consists of the 3′ end of LRRCT and the C terminus, as in the Sea lamprey gVLR, except that the lamprey region coding for the 5′ end of LRRCT is missing. The hagfish gVLR are compact, 671 bp from start-to-stop codons in gVLR-A and 410 bp in gVLR-B.

The hagfish gVLR loci harbor cassettes encoding diverse LRR motifs located ≈20-40 kb downstream from the germline genes (Fig. 3). In the VLR-A locus, there is a cassette encoding six to eight terminal residues of a diverse CP domain and a 5′ LRRCT that includes a four-residue identical overlap with the gVLR-A 3′ LRRCT. Farther downstream, there is a cassette of two diverse LRRs positioned in reverse orientation relative to the gVLR-A and then an inverted incomplete 5′ LRRCT. In the gVLR-B locus, there is a cassette encoding residues 7-23 of the signal peptide and a 5′ LRRNT, then a diverse CP domain and 5′ LRRCT, one inverted LRR and, farther downstream, another inverted LRR cassette consisting of the 12-terminal and 8-proximal residues of LRRs. No other diverse LRR modules were identified in flanking DNA spanning ≈50 kb upstream and ≈70 kb downstream from the gVLRs. However, diverse LRR elements likely exist elsewhere in the genome to provide missing components of the mature VLR genes identified in samples of genomic PCR amplicons from lymphocyte-like cells, 35 unique mature VLR-A and 38 VLR-B sequences from two animals (Fig. 2). Thus the hagfish mature VLR genes must be assembled through somatic recombination, as is the case for lamprey, but the mechanism for the recombination process is still unknown.

Germ-line VLR genes in hagfish lymphocyte-like cells are actively transcribed before gene rearrangement. PCR amplicons of VLR-A germ-line transcripts are ≈0.7 kb long and ≈0.5 kb long for VLR-B (Fig. 4B, RT-PCR; the position of PCR primers is indicated in Fig. 4A), whereas the larger amplicons correspond to transcripts from the rearranged mature VLR genes, ≈1.1 and ≈0.8 kb for VLR-A and -B, respectively. The corresponding PCR amplicons from blood genomic DNA are ≈0.7 kb for the germ-line genes and ≈1.1 kb for the mature VLR-A and -B genes (Fig. 4B, genomic PCR). In transcripts from germ-line and mature VLR genes, the 5′ intron is spliced out to yield RT-PCR products shorter than the corresponding genomic PCR amplicons (see VLR-B in Fig. 4B; gVLR-A amplicons do not include the 6.4-kb intron). However, the intervening sequences between the coding exons are retained in the germ-line transcripts, because they lack consensus eukaryotic splice sites. The germ-line transcription may be required for gVLR rearrangement, as is the case in mammalian antibody class switch recombination, for which germ-line switch region transcription is obligatory (22, 23).