The release of the whole genome sequences of Cx. pipiens, Ae. aegypti and An. gambiae enabled us to analyze the genomic organization of Vtg genes and examine their evolutionary pattern across mosquito genera. By using a Vtg gene from Ae. aegypti (GenBank accession L41842) as query to BLAST search the Cx. pipiens whole genome sequence, four intact Vtg genes were identified, designated as CpVg1a (GenBank accession NZ_AAWU01017720), CpVg1b (GenBank accession NZ_AAWU01017726), CpVg2a and CpVg2b (GenBank accession AAWU01001936) ("Cp" refers to the first letters of genus and species name; other mosquito Vtg genes in the following text are designated in a similar fashion). Each of these genes contained a different sequence in their 5' promoter regions, indicating they were positioned at different genomic loci. CpVg2a and CpVg2b were clustered together and organized in a "tail-to-tail" orientation with an intergenic region of 5,113 bp. It was not clear whether CpVg1a and CpVg1b were clustered due to limitations in the genome assembly. The two Vtg families CpVg1 and CpVg2 shared 64.8%-65.7% nucleotide identity, while subfamily CpVg1a and CpVg1b shared extremely high nucleotide identity (98.9%), as did CpVg2a and CpVg2b (98.4%) (Table 1).

BLAST searches revealed three Vtg genes in the Ae. aegypti genome sequence (AeVgA, accession AAGE02018519; AeVgB, accession AAGE02009986 and AeVgC, accession AAGE02009985), consistent with previous studies 9 10 21 . Based on the genome sequence data, we determined the organization of these three genes. AeVgB and AeVgC were located in the same super contig (supercont1.191) with an intergenic region of approximately 115 kb (super contig location; AeVgB: 160,790-167,359; AeVgC: 39,434-45,832), while AeVgA was located in a different supercontig (supercont1.477) (VectorBase). These three genes shared 61.4%-88.4% nucleotide identity.

BLAST searches revealed three Vtg genes from whole genome sequence of An. gambiae (ACCESSION AAAB01008880) and non-redundant database in GenBank (ACCESSION AF281078). The former encoded three Vtg genes (AgVg1, AgVg2 and AgVg3), but with partial coding sequence for AgVg1 and AgVg2; the later encoded AgVg1 and AgVg2 with partial sequence for AgVg2. The AgVg2 sequence was partial due to a genome sequence gap, but by combining data from both datasets we obtained intact sequences for AgVg1 and AgVg3 (Figure 1). All three Vtg genes showed extremely high nucleotide identity (98.1%-99.2%) and are encoded as a tandem repeat in the An. gambiae genome (Figure 1).

When we attempted to isolate and clone the 5' promoter region of the Cx. tarsalis Vtg gene, the obtained sequence data lead us to identify four different loci that had high conservation in the coding sequence but unique 5'-promoter regions. On the basis of the Cx. pipiens Vtg gene coding sequence, multiple sets of primers were designed to isolate the intact Vtg genes from Cx. tarsalis. We first cloned the partial coding sequences and then obtained the 5' and 3' end fragments. Specific primers located in the 5' promoter region and 3'coding or 3'UTR regions were designed to amplify the full length genes. Four full length Vtg genes from Cx. tarsalis were isolated, designated as CtVg1a, CtVg1b, CtVg2a and CtVg2b. Sequence analysis indicated that the two Vtg families CpVg1 and CpVg2 shared 64.3%-65.5% nucleotide identity, while subfamily CtVg1a and CtVg1b shared extremely high nucleotide identity (98.1%), as did CtVg2a and CtVg2b (97.0%) (Table 1).

While isolating the 5'-promoter region of the CtVg1b gene, we identified a novel gene 805 bp upstream of the CtVg1b start codon, in the opposite orientation. This small gap between the start codons of the novel gene and CtVg1b suggests that this region may act as a bidirectional promoter. The novel gene contained two putatively functional protein-protein binding domains - a RING-finger domain and MATH domain (Figure 2a) and has homology to the human tripartite motif protein 37 (TRIM37) gene - we therefore refer to this gene as TRIM37-like (T37L) protein. By examining Vtg gene sequence data from other mosquitoes (Cx. pipiens, Ae. aegypti, Ae. polynesiensis, An. gambiae and Oc. triseriatus) we identified T37L homologues upstream of one of the Vtg copies in all species examined (Figure 2). Among mosquitoes, the size of the intergenic region ranged from 769 bp to 1004 bp (Figure 2). The T37L amino acid sequences were highly conserved among mosquito species (Figure 2). RT-PCR showed that in Cx. tarsalis CtT37L was transcriptionally expressed in all developmental stages (Figure 2). Interestingly, the down-stream associated Vtg gene (CtVg1b) is also expressed in all developmental stages 22 . The discovery of conservation of the T37L-Vtg structure in all examined mosquito genomes indicated the Vtg gene in this special organization was likely the ancestral origin of mosquito Vtg genes through gene duplication events.

To determine the orthologous relationship of the Vtg genes from Cx. pipiens and Cx. tarsalis, we used their 5' promoter sequences to calculate identity by BLAST analysis (Table 1). The vertical and horizontal directions in Figure 3 show orthologous and paralogous relationship, respectively. The 5' promoter region of each orthologous gene shared 70.6-85.3% nucleotide identity, while producing no significant match with any other Vtg gene (Figure 3). For example, CtVg1a shared 81.1% nucleotide identity with CpVg1a in their 1.1 Kb 5' promoter region directly upstream of the start codon, but no significant match alignment with that region from the other Vtg genes.

Phylogenetic analysis using the full Vtg coding sequences was used to examine the evolution of Vtg genes among mosquitoes (Figure 4). All analyses gave identical tree morphology. With the exception of the Culex Vg2 group (which represents a duplication event unique to the genus Culex), all other mosquito Vtg genes conformed to three major clades: Aedes/Ochlerotatus, Culex and Anopheles. The Aedes/Ochlerotatus clade is more closely related to the Culex Vg1 clade than the Anopheles clade, which is in agreement with the agreed relationships of these genera. Paralogous copies cluster more closely than orthologues within each clade. Three independent duplication patterns were observed among mosquito genera. In Culex, a divergent duplication event produced the Vg1 and Vg2 genes, which each then underwent additional independent duplications generating four Vtg genes in total. These duplication events occurred prior to the divergence of pipiens and tarsalis within the genus. The absence of Vg2 homologues in non-Culex genera may reflect a unique duplication event in Culex, or less likely, independent gene loss events in Anopheles and Aedes/Ochlerotatus. In Aedes, the early duplication from the ancestral copy generated the VgA/B lineage, which underwent an additional duplication to generate 3 Vtg copies - these duplication events occurred prior to the divergence of the Aedes and Ochlerotatus genera. In Anopheles, Vtg gene duplication events occurred rapidly during tandem repeat generation (Figure 4). Unlike Culex and Aedes/Ochlerotatus, none of the An. gambiae Vtg genes cluster with other members of the genus, suggesting that the rapid tandem repeat generation occurred after the split of gambiae from the other members of the Anopheles genus.

Concerted evolution occurs in multigene families and is expected to generate a higher sequence similarity between paralogues than between orthologues. Statistically significant fragments were identified in paralogues Vg1a and Vg1b in the genus Culex. The predicted conversion fragments include not only coding regions but also non coding DNA such as 5' promoter regions and introns (Table 2).

Though concerted evolution certainly plays an important role in maintaining the high level of mosquito Vtg gene sequence conservation, an alternative explanation for high sequence homogeneity is "birth and death" evolution under strong purifying selection 23 24 25 ]. The two models of evolution could be distinguished by comparing the number of synonymous differences per site (Ps) within and between species. In the presence of strong purifying selection without concerted evolution, DNA sequence differences will be observed primarily at the synonymous sites. If a gene has undergone concerted evolution, sequence differences will not be biased toward synonymous sites. If strong purifying selection is responsible for the observed pattern, intraspecific and interspecific synonymous differences are expected to be similar. If intraspecific synonymous differences are much lower than interspecific difference, this would suggest that concerted evolution is the dominant force. In our dataset, Ps values in some paralogous pairs are much lower than that in orthologous pairs. For example, Ps values ranged from 0.0316 to 0.0525 in the paralogous pairs CtVg1a-CtVg1b and CpVg1a-CpVg1b, but ranged from 0.1809 to 0.1851 in the orthologous combinations (Table 3). The Ps value range of 0.0316 to 0.0525 was far from saturation level (0.4-0.7 in many species) 23 . Moreover, extremely high sequence conservation was observed between paralogous introns (88.5% to 100%), but was lower between orthologous introns (63.3% to 81.5%) in Culex (Table 1) which may not be explained simply by purifying selection. Taken together, the data demonstrated that strong purifying selection is likely not responsible for high sequence identity between gene paralogues within the same family.

To examine the selection pressure between the two paralogous families Vg1 and Vg2 in Culex species, dN/dS values were calculated between Vg1 and Vg2. All pairwise comparisons were significantly less than 1 (p < 0.0001, Z-test), indicating that these Vtg sequences are under purifying selection (Table 4). No significant conversion tracts were detected using GENECONV (data not shown). To test if positive selection was acting on different regions of mosquito Vtg genes, we used WSPMaker http://wspmaker.kobic.kr/, a web tool for scanning and calculating selection pressures in sub-regions of two protein-coding DNA sequences 26 . No dN/dS ratio was significantly greater than 1 (data not shown).

We observed that gene conversion events were variable in the amount of gene sequence involved between mosquito species. While in some species, almost the entire gene sequences have undergone concerted evolution, others had a mosaic pattern of conversion events, where some segments of the genes are homogenized and evolved in concert, while others diverged without gene conversion 27 . We determined that sequence homogenization by gene conversion occurred in the entire gene sequences of Culex Vg1 paralogues, while homogenization occurred only partially in Culex Vg2 paralogues, An. gambiae paralogues, Ae. aegypti paralogues and Oc. atropalpus paralogues (Table 2). The lower number of synonymous differences per site (Ps) found in paralogues of CpVg2a-CpVg2b, CtVg2a-CtVg2b and AgVg1-AgVg3 (ranging from 0.0226-0.0564), suggests that sequence homogenization was not due to strong purifying selection (Table 3). We analyzed dN/dS ratios between the unconverted coding region of the paralogues Vg2a and Vg2b in Culex, and paralogues in An. gambiae, Ae. aegypti and Oc. atropalpus. All unconverted coding regions have likely undergone purifying selection (dN/dS < 1, p < 0.01, Z-test) (Table 5).

Culex tarsalis were from the KNWR strain, which was collected in the Kern National Wildlife Refuge (KNWR) in Kern County, California. Larvae were reared at a standard density of 200 larvae/pan and fed a 1:2:2 blend of fish food, rabbit pellets and bovine liver extract. Adults were maintained on 10% sucrose. Mosquitoes were maintained by autogenous oviposition.

Genomic DNA was isolated from Cx. tarsalis adults using the procedure described in Chen and Li 50 . Total RNA was extracted from female adults using the SV Total RNA Isolation System Kit (Promega) as described by the manufacturer. Two micrograms of RNA was used as template for first strand cDNA synthesis to make 3'RACE-ready cDNA using The SMART RACE cDNA Amplification Kit according to the manufacturer's protocol (Clonetech).

To isolate the full Vtg coding sequences, we began by cloning partial coding sequences. Based on the available An. gambiae, Ae. aegypti and Cx. pipiens vitellogenin sequences, two primers (CpVgCoF1: 5'-CCA-AGA-CCA-TGA-CCG-CCC-TG-3' and CpVgCoR1: 5'-TTT-CCC-ATT-TGG-TTG-GTG-TTG-GG-3') were designed to PCR-amplify the coding sequence near the 5' end. The PCR protocol was 2 min at 94°C, followed by 35 cycles consisting of 94°C for 30 s, 50°C for 30 s, and 72°C for 1 min, and a final 72°C extension for 10 min. PCR products were separated on a 1% agarose gel in 1 × TAE buffer. PCR fragments were cloned into the TOPO TA cloning vector (Invitrogen) and sequenced. Two partial vitellogenin gene (Vg1 and Vg2) sequences were identified.

The 5'-flanking sequences were obtained using the GenomeWalker Universal Kit (Clonetech) according to the manufacturer's protocol. Briefly, genomic DNA was digested by eight blunt-end restriction enzymes (SnaBI, MscI, SmaI, ScaI, EcorV, Dra, PvuII, Stu I), and then adaptor-ligated on both ends. Use adaptor and gene-specific primers, we amplified the 5' flanking regions. Two gene-specific primers were designed for each gene. Vg1: Vg1GSP1 (5'-CTC-CCT-CCA-GAT-GTT-GGT-ACG-GTA-TCC-3') and Vg1GSP2 (5'-TCG-TTG-AAG-TTG-GCG-TAC-TCG-GCT-TGC-3'); Vg2: Vg2GSP1 (5'-AAG-TCC-GTG-CGG-TGA-CCA-TCG-TCC-AGA-3'); Vg2GSP2 (5'-ATA-CTG-GTT-GAA-CTG-AGC-GTA-TTG-AGC-3'). The nested PCR protocol was 2 min at 94°C, followed by 20 cycles consisting of 94°C for 30 s, 67°C for 4 min, and a final 72°C extension for 10 min; the second PCR reaction was identical except that cycles were increased to 30 and used 1 μl 50× dilution of the first PCR product as template. A 10:1 mix of Taq and pfu DNA polymerase was used for all reactions. PCR fragments were TOPO cloned and sequenced as described above.

3'Rapid Amplification of cDNA ends (3'RACE) approach was used to isolate the 3' end of the Vtg genes. Two forward gene-specific primers were designed for each Vtg gene for nested 3'RACE PCR on the basis of 3' coding sequence of the Cx. pipiens Vtg genes. Vg1: CpVg13F1 (5'-ACT-TCC-AGA-ACG-CTG-ACA-CC-3') and CpVg13F2 (5'-TCA-AGA-ACG-GAT-TCA-GCG-AG-3'); Vg2: CpVg23F1 (5'-GAG-AAC-AAC-CAG-CAG-CAC-CT-3') and CpVg23F2 (5'-ATT-GTG-CCG-AAC-GGT-GCT-CA-3'). The nested PCR protocol was 2 min at 94°C, followed by 20 cycles consisting of 94°C for 30 s, 52°C for 30 s and 72°C for 90 s, and a final 72°C extension for 10 min; the second PCR reaction was identical except that cycles were increased to 30 and used 1 μl 50× diluted PCR product as template. PCR fragments were TOPO cloned and sequenced as described above.

Finally, we designed four pairs of gene-specific primers to amplify the four full length Vtg genes, in which forward primers and reverse primers were located in 5' promoter region and 3'UTR region, respectively. Vg1a: Vg1aPF (5'-GAA-CCC-ACC-GAT-TGT-TTA-CG-3') and Vg1aSPR (5'-ATG-CCT-TTG-TAA-ACA-GTT-CC-3'); Vg1b: Vg1bPF (5'-TGG-TGT-TGC-TCT-CAG-ACT-TG-3') and Vg1bSPR (5'-CCA-AAT-TCA-TTG-CTT-TCC-GA-3'); - Vg2a: Vg2aPF (5'-TCA-ACA-ACC-ATC-CCT-TCA-CA-3') and Vg2aSPR (5'-TCC-GTT-ACA-ACC-ATC-TAG-AG-3'); Vg2b: Vg2bPF (5'-AAA-GCA-GCC-TCA-AGC-AAT-CG-3') and Vg2bSPR (5'-GGC-AGT-TGT-ATC-TTC-CAA-GG-3'). The PCR protocol was 2 min at 94°C, followed by 35 cycles consisting of 94°C for 30 s, 50°C for 30 s, and 72°C for 4 min 30 sec, and a final 72°C extension for 10 min. PCR products were cloned and sequenced as described above. The four new isolated Vtg genes from Cx. tarsalis have been deposited in GenBank under accession numbers GU017909-GU017912.

While isolating the 5'-promoter region of the CtVg1b gene, we identified a novel gene (CtT37L - see results) 805 bp upstream of the CtVg1b start codon, in the opposite orientation. We determined the expression profile of CtT37L in Cx. tarsalis by reverse transcriptase PCR. Total RNA was isolated using RNAqueous micro-kit columns (Ambion) from first-instar larvae, fourth instar larvae, pupae (male and female), 24-hour old male adults, and female adults 24, 48, 72, 96 and 120 hours post-emergence. The experiment was replicated 3 times. RNA was extracted from 5 pooled individuals except in the case of first instar larvae where 10 individuals were pooled. Isolated RNA was treated with DNase (Ambion) to eliminate residual DNA contamination. cDNA was generated using 1 μg of DNase-treated RNA for reverse transcription using the Superscript RT for PCR kit (Invitrogen) in 20 μl volumes with Oligo(dT)20 primers. No-RT controls were run for each sample. For RT-PCR 2 μl of cDNA was used as a template for amplification with following primers (5'-3'): CtTF109F: CTT-TTG-GTC-CAC-GAA-GTG-GT; CtTF561R: GCT-TTG-AAG-GAC-AGC-GTT-TC. Control actin primers were: actin1F: ATG-TTT-GAG-ACC-TTC-AAC-TCG-C, actin1R: TAA-CCT-TCR-TAG-ATT-GGG-ACG. PCR conditions were 94°C for 3 min, followed by 30 cycles of: 94°C for 30 sec, 50°C for 30 sec, 72°C for 30 sec, followed by a final extension of 72°C for 10 min. Amplicons were resolved by agarose gel electrophoresis.

The complete Vtg genes were identified by from the available An. gambiae, Ae. aegypti and Cx. pipiens genome sequences, and GENBANK sequence data from Ae. polynesiensis, An. stephensi, An. albimanus and Ochlerotatus atropalpus. Sequence analysis was performed using Blast http://www.ncbi.nlm.nih.gov/blast/ against a non-redundant database. Initial alignments were performed on amino acid sequences using Vector NTI advance 10 (Invitrogen) followed by inference of the nucleotide alignment. Phylogenetic analysis of mosquito Vtg genes was conducted on the amino acid alignment using Maximum Likelihood (ML), Bayesian (B) and neighbour-joining (NJ) algorithms. Bayesian phylogenetic analysis was conducted using MrBayes v 3.1.2 51 . The WAG model of amino acid substitution was selected as the most appropriate using the MCMC sampler to test all fixed rate matrices. The rate variation over sites followed a gamma distribution with a proportion of invariant sites; these parameters were estimated from the data by MrBayes. Analyses were run for 500,000 generations, sampling every 100 generations. The first 25% of generated trees were considered the burnin and discarded. We constructed a 50% majority-rule consensus tree from the remaining trees. Maximum likelihood (ML) phylogenetic analysis was conducted using PHYML 52 with the same parameters estimated for Bayesian analyses. Tree robustness was determined with 100 bootstrap replications. Neighbor-joining analyses were conducted using MEGA v.3.1 53 with 1000 bootstrap replications.

Vtg gene nucleotide alignments were generated as described above. GENECONV was used to detect gene conversion events. Global p-values were calculated based on 1000 permutations of the original data using a BLAST-like search algorithm 28 .

Analyses were conducted using MEGA v.3.1 53 . Pairwise dN/dS ratios and significance values were calculated using the Nei-Gojobori method with a Jukes-Cantor correction for multiple hits. Standard errors were calculated by 1000 bootstrap replications.