Toward a catalog for the transcripts and proteins (sialome) from the salivary gland of the malaria vector Anopheles gambiae

Malaria affects 200 million people worldwide and causes approximately 1.5 million deaths every year (Fauci,2001). The disease is caused by Plasmodium parasites transmitted by the blood-sucking mosquito A. gambiae, known as the major vector of malaria in sub-Saharan Africa(Collins and Paskewitz, 1995). The parasite has a complex life cycle in the vector, where it becomes infective after invasion of the salivary gland(Cerami et al., 1992; Touray et al., 1992). Plasmodium is transmitted via the bite of an infected mosquito, which releases the sporozoite stage into the skin (Sidjanki and Vandeberg, 1997) together with saliva(Huribut, 1966). Saliva not only operates as a carrier to deliver the sporozoite into the host(Krettli and Miller, 2001),but also contains a number of pharmacologically active molecules which counteract host defenses that are triggered by blood feeding(Ribeiro, 1987). Although both genome surveys (Kappe et al.,2001; Carlton et al.,2001; Janssen et al.,2001) and genome projects(Gardner et al., 1998; Bowman et al., 1999) for Plasmodium spp. have been conducted, only more recently has systematic sequencing of A. gambiae genes been envisaged(Adam, 2001; Balter, 2001).

In an attempt to reveal the complexity of A. gambiae salivary glands, a high-throughput approach designed to identify a large number of cDNAs in the gland of this mosquito has been employed. Remarkably, only approximately 15% of our cDNA isolates match A. gambiae sequences previously reported (Arcà et al.,1999); many of the remaining clusters have unknown functions. Generation of a set of A. gambiae salivary cDNAs, in addition to the Plasmodium genome currently available, may provide indispensable tools for the systematic and comprehensive analysis of molecules that may play an active role in the pathogenesis of malaria.

All water used was of 18 MΩ quality and was produced using a MilliQ apparatus (Millipore, Bedford, MA, USA). Organic compounds were obtained from Sigma Chemical Corporation (St Louis, MO, USA) or as stated otherwise.

A. gambiae gambiae Giles mosquitoes were reared under the expert supervision of Mr André Laughinghouse. Insectary rooms were kept at 26±0.5°C, with a relative humidity of 70% and a 16 h:8 h light:dark photoperiod. Adult female mosquitoes used in the experiments were aged 0-7 days, took no blood meals, and were maintained on a diet of 10% Karo syrup solution. Salivary glands from adult female mosquitoes were dissected and transferred to 20 μl Hepes saline (HS; NaCl 0.15 mol l^-1, 10 mmol l^-1 Hepes, pH 7.0) in 1.5 ml polypropylene vials in groups of 20 pairs of glands in 20 μl of HS or as individual glands in 10 μl of HS. Salivary glands were kept at -75°C until needed.

A. gambiae salivary gland mRNA was isolated from 80 salivary gland pairs from adult females at days 1 and 2 after emergence using the Micro-FastTrack mRNA isolation kit (Invitrogen, San Diego, CA, USA). The polymerase chain reaction (PCR)-based cDNA library was made following the instructions for the SMART cDNA library construction kit (Clontech, Palo Alto,CA, USA). A. gambiae salivary gland mRNA (200 ng) was reverse transcribed to cDNA using Superscript II Rnase H-reverse transcriptase(Gibco-BRL, Gaithersburg, MD, USA) and the CDS III/3′ PCR primer(Clontech) for 1 h at 42°C. Second-strand synthesis was performed through a PCR-based protocol using the SMART III primer (Clontech) as the sense primer and the CDS III/3′ primer as antisense primer. These two primers create SfiIA and SfiIB sites at the ends of the nascent cDNA. Double-strand cDNA synthesis was done on a Perkin Elmer 9700 Thermal cycler(Perkin Elmer Corp., Foster City, CA, USA) using Advantage Klen-TaqDNA polymerase (Clontech). PCR conditions were the following: 94°C for 2 min; 19 cycles of 94°C for 10 s and 68°C for 6 min. Double-strand cDNA was immediately treated with proteinase K (0.8 μg μl^-1) for 20 min at 45°C and washed three times with water using Amicon filters with a 100 kDa cutoff (Millipore). Double-strand cDNA was then digested with SfiI for 2 h at 50°C. The cDNA was then fractionated using columns provided by the manufacturer (Clontech). Fractions containing cDNA of more than 400 base pairs (bp) in size were pooled, concentrated, and washed three times with water using an Amicon filter with a 100 kDa cutoff. The cDNA was concentrated to a volume of 7 μl. The concentrated cDNA was then ligated into a Lambda TriplEx2 vector (Clontech), and the resulting ligation reaction was packed using Gigapack Gold III from Stratagene/Biocrest (Cedar Creek, TN, USA) following the manufacturer's specifications. The obtained library was plated by infecting log-phase XL1-Blue cells (Clontech) and the amount of recombinants were determined by PCR using vector primers flanking the inserted cDNA and visualized on a 1.1% agarose gel with ethidium bromide(1.5 μg ml^-1).

The A. gambiae salivary gland cDNA library was plated to approximately 200 plaques per Petri dish (150 mm diameter). The plaques were randomly selected and transferred to a 96-well polypropylene plate containing 100 μl of water per well. The plate was covered and placed on a gyratory shaker for 1 h at room temperature. The phage sample (5 μl) was used as a template for a PCR reaction to amplify random cDNA. The primers used for this reaction were sequences from the TriplEx2 vector and were named PT2F1(5′-AAG TAC TCT AGC AAT TGT GAG C-3′), which is positioned upstream from the cDNA of interest (5′ end), and PT2R1 (5′-CTC TTC GCT ATT ACG CCA GCT G-3′), which is positioned downstream from the cDNA of interest (3′ end). Platinum Taq polymerase (Gibco-BRL) was used for these reactions. Amplification conditions were: 1 hold at 75°C for 3 min, 1 hold at 94°C for 2 min, and 33 cycles at 94°C for 1 min,49°C for 1 min, and 72°C for 1 min and 20 s. Amplified products were visualized on a 1.1% agarose gel with ethidium bromide. The concentration of double-stranded cDNA was measured using Hoechst dye 33258 on a Flurolite 1000 plate fluorometer (Dynatech Laboratories, Chantilly, VA, USA). PCR reactions(3-4 μl) containing between 100 and 200 ng of DNA were then treated with exonuclease I (0.5 units μl^-1) and shrimp alkaline phosphatase(0.1 units μl^-1) for 15 min at 37°C and 15 min at 80°C on a 96-well PCR plate. This mixture was used as a template for a cycle-sequencing reaction using the DTCS labeling kit from Beckman Coulter Inc. (Fullerton, CA, USA). The primer used for sequencing (PT2F3) is upstream from the inserted cDNA and downstream from primer PT2F1. The sequencing reaction was performed on a Perkin Elmer 9700 thermocycler. Conditions were 75°C for 2 min, 94°C for 4 min, and 30 cycles of 96°C for 20 s,50°C for 20 s and 60°C for 4 min. After cycle-sequencing the samples,a cleaning step was done using the multiscreen 96-well plate cleaning system(Millipore). The 96-well multiscreening plate was prepared by adding a fixed amount (manufacturer's specification) of Sephadex-50 (Amersham Pharmacia Biotech, Piscataway, NJ, USA) and 300 μl of deionized water. After 1 h of incubation at room temperature, the water was removed from the multiscreen plate by centrifugation at 750g for 5 min. After partially drying the Sephadex in the multiscreen plate, the whole cycle-sequencing reaction was added to the center of each well, centrifuged at 750g for 5 min, and the clean sample was collected on a sequencing microtiter plate (Beckman Coulter Inc.). The plate was then dried on a Speed-Vac SC 110 model with a microtiter plate holder (Savant Instruments Inc, Holbrook, NY, USA). The dried samples were immediately resuspended with 25 μl of deionized ultrapure formamide (J. T. Baker, Phillipsburg, NJ,USA), and one drop of mineral oil was added to the top of each sample. Samples were either sequenced immediately on a CEQ 2000 DNA sequencing instrument(Beckman Coulter Inc.) or stored at -30°C.

Raw sequences originating from the DNA sequencer were assigned one of five letters in their result: ATCG for identified nucleotide bases, and N when the sequencer program could not call a base. Usually the beginning and ends of the sequences have a higher proportion of N calls. Sequences also contain primer and vector sequences used in library construction. For this reason, raw sequences were treated by a program written in VisualBasic 6.0 (VB) (Microsoft Corp., Redmond, WA, USA) as follows. (i) Sequences were analyzed in their first 80 bp for groups of four Ns, and, if found, the block of four Ns closer to position 80 was used to trim the raw sequence from this 5′ N-rich region. (ii) For sequences longer than 110 bp, windows of 10 bp were screened for the occurrence of four or more Ns above position 100. The positive window with the smallest position value was used to trim the sequence from the 3′ N-rich region. Sequences thus trimmed and having more than 10 % N content were discarded. (iii) Good quality and trimmed sequences were then searched for occurrence of the primers used in library construction (the SMART III primer as well as the CDS/R primers). A moving window the size of the primer was searched on the sequence for matches with the primer sequence. If more than a 70 % match was obtained, or if a contiguous match longer than 50 %of the length of the primer was observed, the sequence was trimmed at the beginning or end of the window, depending on the expected position of the primer. This simple algorithm avoided errors due to spurious insertions. (iv)The trimmed sequence was `polished' by removing any trailing N residues. The sequence final N content was assessed, as well as its AT content and length. The final sequence was written to a FASTA-format file containing in its definition line the actions taken by the program.

To obtain information on the possible role of the cDNA sequences, the FASTA file containing all the stripped sequences was blasted against the GenBank nonredundant protein database (NR) from the National Center for Biotechnology Information (NCBI) using the standalone BlastX program found in the executable package at ftp://ftp.ncbi.nlm.nih.gov/blast/executables/(Altschul et al., 1997). The NR database as well as the cumulative updates were regularly downloaded,uncompressed with GUNZIP (found at www.gzip.org/), and formatted for Blast program use with the FormatDB program(executables also found at ftp://ftp.ncbi.nlm.nih.gov/blast/executables/) with the help of a program written in PERL code (software found at www.activeperl.com). NCBI sequences are indicated in this manuscript by their accession number as gi|XXXX where XXXX is a unique identifier number. The resulting file was parsed, and the best match was incorporated in the FASTA definition line after the delimiter|. The sequences were next submitted to the standalone program RPSBlast (Altschul et al., 1997) and searched against the Conserved Domains Database(CDD) (found at ftp://ftp.ncbi.nlm.nih.gov/pub/mmdb/cdd/), which includes all Pfam (Bateman et al.,2000) and SMART (Schultz et al., 2000) protein domains. The RPSBlast result file was parsed as above and the best match incorporated also into the FASTA definition line of the sequence. When all sequences of a particular cluster were blasted against the NR protein database (using the BlastX program), the best protein match was searched for the species from which the NR database sequence originated. If the species was not A. gambiae, or no matches to the NR were found,the cluster was marked as representing a novel A. gambiae sequence(indicated by Y under the column marked N (novel) in Table 1). All cluster sequences that gave a match to an A. gambiae protein sequence were further individually inspected to verify whether the cDNA sequence represented nearly the same information translated as the protein match or a closely related but different protein. In this latter case, a Y would also be added to the results in Table 1 in the N (novel)column for the row of the cluster in question.

The FASTA file containing all sequences was clustered by first blasting(using the BlastN program) each sequence against the formatted database file using a Blast cutoff score of 1E^-60. The resulting file was used to join in a single cluster all sequences that shared at least one common sequence in the BlastN result. Thus, if sequence A had a 1E^-60match to B and B had a similar match to C, the three sequences would be joined even if A had a less meaningful score in relation to C. The clustering program also made individual FASTA-formatted files for each cluster, sorted in descending order of sequence size. When these files contained two or more sequences, they were used as input for the sequence alignment program CLUSTALW(Higgins et al., 1996), which was called automatically by the clustering program. CLUSTAL alignment files were thus created for each cluster having two or more sequences. This clustering program was also written in VB. Finally, a program was written in VB that combines all the results to create Table 1 of this paper, except for the Function column. The output of this program is imported into a Microsoft Excel spreadsheet. In the supplemental material, Table 1 includes hyperlinks to the best NR protein match in the NCBI site, all FASTA files for each individual cluster, CLUSTAL alignment files for each cluster, when available,and the FASTA file for the whole database. Each cluster was individually analyzed for the probable function of its translation product and assigned a`probably secreted', `probably housekeeping' or `indeterminate' function. This decision was based on the best match to the NR protein database and related sequences as searched online at the NCBI site(www.ncbi.nlm.gov) and on the SMART and/or Pfam matches, including searches of the nature of the domains by online searches of the respective sites.

A portion (4 μl) of the lambda phage containing the cDNA of interest was amplified using the PT2F1 and PT2R1 primers (conditions as described above). The PCR samples were cleaned using the multiscreen-PCR 96-well filtration system (Millipore). Cleaned samples were sequenced first with PT2F3 primer and subsequently with custom primers. Primer selection for complete sequence of selected full-length cDNA was also assisted by a program (written in VB) that identified unique primer sites within the sequences. To assemble the sequences, the previously known sequence was blasted against the new sequence using the standalone program b12seq found with the executable package provided at the NCBI ftp site mentioned above. After identifying the regions of overlap, the two sequences were joined. The program attempted to locate a poly(A) region by using a 12-bp window in which 11 A residues would constitute a poly(A) string. If no poly(A) was found, a new set of primers would be found to continue extension of the cDNA. The program also generates CLUSTAL alignments of all sequences and produces a consensus output and the three possible translations of this unidirectionally cloned RNA. The final alignment is adjusted by hand. If necessary, the original tracings of the DNA sequencer are reviewed for critical base calls. The translated sequences are submitted as a FASTA file to the SIGNALP server (at http://www.cbs.dtu.dk/services/SignalP/) (Nielsen et al., 1997),which responds by e-mail: indicating whether a signal peptide exists and its location. A program written in VB interprets this SIGNALP result file and removes the signal peptide, if it is predicted to exist, to create a mature protein sequence. Molecular masses using average molecular masses for C, H, O,N, P and S are calculated for all protein sequences, as are pI based on reduced proteins, following the pKa for amino acids within proteins as indicated before (Altland,1990; Bjellqvist et al.,1994). This program, combined with the program generating Table 1 of this paper, produced an output that can be read by the spreadsheet program Excel to produce Table 2 in this paper. In the supplemental material, available on request, hyperlinks are given to all proteins.

A precast 16% polyacrylamide gel was used and run in Trisglycine-SDS buffer. Alternatively, a NU-PAGE 12% Bis-Tris gel, 1 mm thick (Invitrogen),was used and run in MOPS buffer according to the manufacturer's instructions. To estimate the molecular mass of the samples, SeeBlue™ markers from Invitrogen (myosin, bovine serum albumin, glutamic dehydrogenase, alcohol dehydrogenase, carbonic anhydrase, myoglobin, lysozyme, aprotinin and insulin,chain B) were used. Salivary gland homogenates were treated with SDS (2%) or NU-PAGE LDS sample buffer (Invitrogen) without reducing conditions. 20 pairs of homogenized salivary glands per lane (approximately 20 μg protein) were applied when visualization of the protein bands stained with Coomassie Blue was required. For amino-terminal sequencing of the salivary proteins, 20 homogenized pairs of glands were electrophoresed and transferred to polyvinylidene difluoride (PVDF) membrane using 10 mM CAPS, pH 11.0, 10%methanol as the transfer buffer on a Blot-Module for the Xcell II Mini-Cell(Invitrogen). The membrane was stained with Coomassie Blue in the absence of acetic acid. Stained bands were cut from the PVDF membrane and subjected to Edman degradation using a Procise sequencer (Perkin-Elmer Corp.). To find the cDNA sequences corresponding to the amino acid sequence, obtained by Edman degradation of the proteins transferred to PVDF membranes from PAGE gels, we wrote a search program (in VB) that checked these amino acid sequences against the three possible protein translations of each cDNA sequence obtained in the mass sequencing project. This program was written using the same approach utilized in the BLOCKS (Henikoff and Henikoff, 1994) or PROSITE(Sibbald et al., 1991)databases. The program is very useful when mixed sequence information occurs,for example, amino-terminal sequences deriving from a mix of equal peptides. In this case, two different cDNA sequences may be unambiguously found as matches.

Statistical tests were performed with SigmaStat version 2.0 (Jandel Software, San Rafael, CA, USA). Kruskal—Wallis ANOVA on ranks was performed, and multiple comparisons were done by the Dunn method. Dual comparisons were made with the Mann—Whitney rank sum test.

A literature search indicates that the salivary gland of Anopheles gambiae contains a number of putative proteins including apyrase,5′ nucleotidase, lyzozyme, members of the D7 family of proteins, nine so-called A. gambiae hypothetical proteins (HP), and eight so-called A. gambiae salivary gland (SG) proteins(Arcà et al., 1999). Most of these proteins are translations of expressed sequence tags (EST) or full-length clones; accordingly, they have not been unambiguously characterized as salivary molecules. In an attempt to improve our understanding of the complexity of the proteins and transcripts expressed in A. gambiae salivary glands, we have performed SDS-PAGE and a cDNA library using, respectively, the proteins and mRNA from this same tissue.

Fig. 1 shows the pattern of separation of A. gambiae salivary protein by SDS-PAGE stained by Coomassie Blue. The gel shows relatively low tissue complexity, with approximately 15 clearly visible stained bands and many others lightly stained. To identify these proteins, they were transferred to PVDF membranes and the bands cut from the membrane and submitted to Edman degradation. Amino-terminal information was successfully obtained for many of these bands,and they were identified as SG6 (approx. 10 kDa apparent molecular mass), A. gambiae D7-related proteins 1-3 (approx. 10-14 kDa apparent molecular mass), similar to Glossinia morsitans antigen 5 (approx. 30 kDa), similar to Aedes aegypti D7 (approx. 33 kDa), similar to A. aegypti 30 kDa allergen (approx. 36 kDa), HP 8 (CB1, 44 kDa), similar to HP 9 (bB2, 46 kDa) and herein called HP 9-like, SG1-like 2 (approx. 48 kDa),putative 5′ nucleotidase (approx. 64 kDa) and SG1 (approx. 105 kDa). Although the predicted translation products of some of these proteins have been reported (Arcà et al.,1999), amino acid sequencing has not been performed before. Edman degradation for other bands was attempted unsuccessfully, either because the protein's amino terminus was blocked, or because PTH-amino acids could not be reliably identified.

To complement the data generated by SDS-PAGE and identify potentially novel molecules in the salivary gland of A. gambiae, a cDNA library was constructed and hundreds of independent clones randomly 5′ sequenced. When a cluster analysis of all 691 sequences from this library was performed at e^-60, 251 independent clusters were organized. Subsequently, clusters were blasted against the nonredundant (NR) and protein motifs databases. Signal peptides were predicted by submission of the cluster sequences to the SignalP server, allowing the identification of putative secretory (S) and housekeeping (H) cDNA. A comprehensive diagram depicting the steps used for generation of the data is shown in Fig. 2. The results are presented as Tables 1,2,3,4. The electronic versions of the tables, available on request, also contain: (i)columns with hyperlinks to the best match of the NR database, (ii) links to NR matches found for the cluster, (iii) matches to the conserved domain database(CDD), (iv) FASTA-formatted files for each cluster, and (v) CLUSTAL alignments of each cluster having two or more sequences.