Transcriptomes and pathways associated with infectivity, survival and immunogenicity in Brugia malayi L3

General properties of the B. malayi Version 2 array (BmV2 array)

The oligonucleotide elements on the BmV2 array represent ~86% of currently annotated B. malayi gene models defined by Ghedin et al [13]; 11,975 of 15,412 B. malayi sequences correspond to 9,798 unique annotated gene models, and the remaining 3,437 sequences that do not match annotated gene models represent genes that have not yet been sequenced or annotated in the Brugia genomic DNA. The BmV2array elements included all genes in the Wolbachia genome and ESTs from O. volvulus and W. bancrofti with little homology to known B. malayi sequences. A homology search of the 18,104 sequences from which the oligos were derived vs. the NCBI non-redundancy (NR) database revealed that ~69% (12, 423/18,104) had significant homology to publicly available known or unknown proteins from other species (e-05) [See additional file 1]. Of 15,412 Brugia sequences, 9,824 sequences had best matches to 8,038 unique C. elegans genes with RNAi information; 37% of these genes had RNAi phenotypes [See additional file 2].

Functional assignments for elements on the microarray were performed using InterPro, KEGG, and GO analysis. The complete results are reported in supplemental materials [protein domain matches in additional file 3; pathways in additional file 4]. The top 20 most abundant protein domain matches in the oligoarray are presented in additional file 5. Based on InterPro protein domain matches, 48% of the products of genes on the array mapped to one or more gene ontology terms [14, 15]. These GO associations are available online through the Amigo viewer http://nematode.net/cgi-bin/amigo/go_brugia_malayi_microarray_v2/go.cgi?session_id=94439521169750740719

Overview of differences in gene expression by L3 type

A complete list of elements for the array with oligonucleotide sequences and hybridization data are available online through the link: http://nematode.net/Microarray/index.php. The expression profile analysis included elements with hybridization signals above background for at least one condition in pair-wise comparisons (L3i versus L3c and L3ir versus L3c). A total of 774 genes exhibited significant differential expression in pair-wise comparisons (≥ 2 fold differences with P-values < 0.01) [See additional file 6]. 353 of these genes (~46%) were up-regulated in L3i relative to L3c (the "L3i gene set"); 266 (~33%) were up-regulated in L3c (the "L3c gene set", with 244 genes up relative to L3i and 22 up relative to L3ir); 165 (~21.3%) were L3ir-up-regulated relative to L3c (the "L3ir gene set') (Figure 1). The L3i gene set contains genes required for L3 survival in mosquitoes and for infection of mammalian hosts. The L3c gene set represents genes induced by culture conditions that mimic the mammalian environment, and the L3ir gene set represents genes induced by irradiation. Interestingly, some genes exhibited up-regulated expression in more than one condition (Figure 1).

We were surprised to see that the L3i list was the largest of the three L3 gene sets. However, this was consistent with the finding that more Ancylostoma canium L3 (AcL3) genes had reduced transcription after serum stimulation (equivalent to our L3c) relative to non-serum stimulated AcL3 (equivalent to our L3i) [16, 17]. This result suggests that the L3i is not simply a quiescent or resting stage. L3i are actively preparing for their transition into the mammalian host by expressing proteins that are needed for processes like invasion and immune evasion and proteins for survival (stress resistance and antioxidants).

As expected, irradiation repressed gene expression globally; the L3ir gene set was the smallest of the three L3 gene sets. Of 165 L3ir-up-regulated genes, 75 (45%) genes were also up-regulated in L3i relative to L3c (Figure 1). Most of these genes (54/75) were expressed at higher levels in L3i than in L3ir, and only 4 (intermediate filament, collagen family, and Gly-rich proteins) had higher expression in L3ir; 17 were expressed at similar levels in both conditions [See additional file 6].

We previously reported partial conservation of L3i expression patterns in L3ir based on an EST analysis with qRT-PCR confirmation data [5]. The current analysis has confirmed and extended this observation. Several collagen genes and other so-called immunogenic genes (Gln-rich, spx-1), were over-expressed in L3ir relative to L3c [See additional file 6]. Persistent expression of important targets of protective immunity may explain why L3ir are better inducers of anti-L3 immunity than L3i, which are likely to change gene expression to resemble L3c with reduced expression of these proteins shortly after infection.

Distinct functions of known genes associated with different L3 gene sets

The L3i gene set contains genes associated with immune evasion, stress resistance and infectivity

Several classes of known genes were highly expressed in the L3i gene set as shown in Table 4. Immunomodulatory genes are important for invasion and establishment of infection in mammalian hosts [23], and prior studies have suggested that L3i modulate the innate immune system [6]. Microarray experiments showed that live BmL3 induced functional alterations of epidermal Langerhans cells (LC) and diminished their ability to present antigens to T-cells. Thus, L3 may actively suppress host immune responses to the parasite immediately after invasion [24]. Our findings suggest a molecular basis for such immune evasion, because a number of L3i up-regulated genes encode immune-modulators. These molecules may suppress local immunity to promote parasite survival just after invasion. For example, filarial ALT proteins encoded by the Bm-alt gene family modulate cytokine-induced signaling in immune cells and may render incoming L3 more resistant to IFN-γ induced killing by macrophages, as shown in a transgenic L. mexicana system [25]. Filarial cystatins are well-described pathogenicity factors that down-regulate T-cell proliferation and induce anti-inflammatory cytokine responses [26, 27]. The cystatin protein encoded by Bm-cpi-2 interferes with antigen presentation by inhibiting multiple cysteine protease activities including the key enzyme asparaginyl endopeptidase [28]. Bm-spn-2, a member of the B. malayi serpin family, has been reported to impair granulocyte function by inhibiting two neutrophil enzymes, cathepsin G and neutral elastase [29].

The L3i gene list includes a broad spectrum of anti-oxidant and detoxification gene families such as superoxide dismutase (Bm-sods) [30], thioredoxin peroxidase (Bm-tpx-2) [31] glutathione S-transferase (Bm-gst) [32, 33], and aldo/keto reductase family proteins [34]. Since non-feeding filarial L3 are protected by a resistant cuticle, increased transcriptional activity of these genes may be needed for detoxifying endogenous compounds such as metabolites of fatty acids [35]. Fatty acids appear to be an important energy source in L3i as shown below. These enzymes also protect L3 from reactive oxygen species (ROS) generated in energy production process (oxidative phosphorylation) [36]. In addition, L3i are preparing to cope with oxidative stress upon host entry. Prior studies have shown that filarial larvae (L1 and L3) are more susceptible to killing than adult worms via antibody-dependent cellular cytotoxicity in vitro. This killing is believed to depend on reactive oxygen and nitrogen intermediates produced by activated myeloid cells [37, 38] L3 must be able to resist oxidative stress to survive in mammalian hosts.

The L3i gene set also includes several proteases that may be important for invasion and molting. For example, filarial cathepsin L-like proteases (Bm-CPL-1,-4 and -5) are believed to be important for larval migration and molting [39–41]. L3i express a protease (BMC12240), which is analogous to a matrix metalloprotease in the canine hookworm (Ac-mtp-1) that is highly expressed in non-activated AcL3 and involved in skin penetration [17, 42]. Thus, our results support the concept that proteases involved in tissue migration are over-expressed in helminth stages that infect mammalian hosts [43].

The L3i gene set also includes genes that are believed to be involved in pathogenesis, parasitism and stress resistance. For example, the highly expressed Bm-val-1 encodes a venom allergen-like protein that is related to genes in A. caninum that encode Ancylostoma secreted-proteins (ASPs) [44]. ASPs belong to the pathogenesis-related proteins superfamily (PRPs) [45]; they are associated with the transition to parasitism in hookworms and highly expressed in non-activated AcL3 [16, 17, 46]. It is interesting that the O. volvolus homologue of Bm-val-1 (Ov-ASP-1) has been reported to induce anti-L3 immunity in an animal model [47].

The L3i gene set also includes a unique small heat shock protein (sHSP) of B. malayi that was represented by two ESTs on the array (BMC02059 and TC3642). This gene was highly expressed in L3i relative to L3c, and its over-expression was confirmed by qRT-PCR (data not shown). This sHSP is developmentally regulated and inducible in B. malayi [48]. Small heat shock proteins are among the most highly expressed dauer-specific transcripts in C. elegans, and they have a demonstrated role in longevity and stress resistance [49–51].

The last two categories of proteins highly expressed in L3i were surface or cuticular proteins and highly immunogenic proteins. Genes encoding cuticular collagens accounted for 19% of L3i up-regulated known genes (47/249). Their dominant presence is understandable, as the nematode cuticle protects the worms from environmental stress. Cuticular collagens are also needed for the development of the next stage of the parasite (L4) after invasion. C. elegans homologs of these genes have RNAi phenotypes of developmental delay such as larval arrest (Lva), larval lethal (Lvl) (BMC0572), and slow growth (Gro) (BMC12441) [See additional file 6]. Collagen genes are over-expressed just before molting in C. elegans [51]. The L3i gene set also contains a highly expressed ground-like family member (grl-4) [52] (4083.m00057). Grl-4 is believed to function during the dauer-to-L4 molt in C. elegans [51]. It would be interesting to test whether this gene is also required for molting in filarial worms. The highly immunogenic proteins, such as Bm-VAL, intermediate filament, and microfilaria surface-associated proteins may be targets of protective immunity [28, 53]. Filarial SPX-1 and Gln-rich protein have been used as diagnostic antigens [54, 55]. These proteins have also been suggested as targets of protective immunity to Brugia and Ascaris [56, 57]. Brugia larval allergen (gene 14323.m00082) was differentially recognized by IgE antibody from putatively immune endemic normal humans; this protein induces specific IgE antibody responses in animals (authors' unpublished data).

The L3ir gene set contains irradiation responsive (IR) genes and highly immunogenic genes that are also up-regulated in L3i

The L3ir gene set includes genes involved in DNA repair and cuticle regeneration after radiation. Genes reported to be IR in other systems were also up-regulated in L3ir as shown in Table 5. These included cytidine deaminase, uridine phosphorylase, aquaporin and finger proteins [58–60]. Notably, various collagen gene families were also over-represented, and these accounted for more than 30% of the total and 50% of L3ir known genes, respectively. Collagen is a principle component of the nematode cuticle that is secreted by the underlying hypodermis, and collagen synthesis is often increased after radiation injury in animals [61, 62]. The nematode cuticle protects the parasite from environmental insults, and it may be an important target of immune responses [63, 64]. The collagen gene Ovcol-1 of O. volvulus, a homologue of the B. malayi collagen gene Bmcol-2 (BMC12441, TC2798) up-regulated in L3ir, is preferentially recognized by immunoglobulin G3 (IgG3) from putatively immune individuals [65]. Cuticular collagens of nematodes are very different from vertebrate collagens, and they are potential targets of protective immune responses [66]. In addition to collagens, the L3ir gene set contains highly immunogenic secretory and membrane proteins; most of these are also present in the L3i gene set. It is interesting that these genes were more highly expressed in L3ir (which induce protective immunity) than in L3c (which do not). It is possible that the enhanced immunogenicity of L3ir is partially due to their persistent high level expression of L3i proteins that are normally down-regulated shortly after L3 enter the mammalian host. This could prolong and enhance priming of the host immune system to critical protective antigens.

The L3c gene set contains genes that are involved in growth and molting

Overview of pathways and GO analysis

KEGG and InterPro analyses were performed to functionally classify and categorize genes up-regulated in each L3 type. The complete KEGG pathways identified for each gene set are listed in additional file 8. Significant protein domain matches for each gene set (P < 0.05) are shown in additional file 9. The GO associations for whole L3 gene set are available online through the Amigo viewer http://nematode.net/cgi-bin/amigo/go_brugia_malayi_L3/go.cgi?search_constraint=terms&action=replace_tree&session_id=91575331171470308620. Differences in pathways and GO terms inferred by transcriptional expression profiling were observed among different L3 types. It is important to note that the RNA levels do not correlate with protein level in all cases; therefore, such differences are needed to be confirmed using appropriate methods.

Altered pathways and significant GO terms in L3i gene set

KEGG pathways that were significantly enriched in the L3i gene set are shown in Table 8. An overview of the main up-regulated pathways for energy metabolism in L3i is presented in Figure 2. Developmentally arrested and non-feeding infective L3 were expected to have low energy production. We were surprised to see that the most enriched pathway mapped in the L3i gene set was energy metabolism (oxidative phosphorylation (ko00190)). This suggests that energy production and consumption might be high in L3i. High energy consumption might be needed in part to detoxify endogenous metabolic products or host proteins. For example, the detoxification reactions catalyzed by GSTs and dehydrogenases/reductases are metabolically costly [35]. Energy is also required for invasion as filarial L3 actively migrate immediately after entering the host through subcutaneous tissues en route to lymphatic vessels [69]. However, additional studies will be needed to test these hypotheses. The microarray data show that BmL3i probably use oxidative phosphorylation and the TCA cycle to produce and store energy through aerobic cellular respiration in the form of ATP as shown in Figure 2. All of the enzymes needed for electron transport in the oxidative phosphorylation pathway are up-regulated in L3i. This explains the association of the significant GO term "mitochondria" with these enzymes [See additional file 10]. This also explains high expression of enzymes to detoxify ROS such as superoxide dismutase and catalase (Table 4) in L3i, since complexes I and III of the respiratory chain generate these ROS [36]. NADH and succinate needed for ATP production can be generated in the citric acid cycle (TCA, Ko00020). Genes encoding three TCA enzymes (malate dehydrogenase, succinate dehydrogenase, and isocitrate dehydrogenase) were over-expressed by a factor of 4 in L3i relative to L3c. Acetyl-CoA, the first molecule entering the TCA, is produced through fatty acid β-oxidation; acetyl-CoA C-acyltransferase (K00632, E2.3.1.16, fadA), the enzyme catalyzing the final step of β-oxidation, was highly expressed in L3i [See additional file 8]. This is consistent with the finding in dauer stage of C. elegans [51, 70]. This suggests that the metabolism of non-feeding BmL3 is adapted to utilize internal energy reserves, predominantly fatty acids. However, there may be the alternate methods to produce Acetyl-CoA in BmL3. For example, Acetyl-CoA might be produced through the pyruvate dehydrogenase complex, as genes encoding two enzymes in the complex were up-regulated; Acetyl-CoA also could be formed by acetate and coenzyme A, as acetyl-coenzyme A synthetase (k01895, E 6.2.1.1) was also over-expressed in L3i. We also found up-regulation of genes encoding enzymes for pyruvate metabolism (ko00620) and gluconeogenesis/glycolysis (ko00010) in L3i including phosphoenolpyruvate carboxykinase (PEPCK) and triose-phosphate isomerase [See additional file 11]. This is also consistent with results reported for dauer larvae in C. elegans [[35, 49], and [51]].

Pathway	ko	P_value	Metabolism
Oxidative phosphorylation	ko00190	0.0003	Energy Metabolism
Citrate cycle (TCA cycle)	ko00020	0.0045	Carbohydrate Metabolism
Antigen processing and presentation	ko04612	0.0114	Immune System

GO analysis indicated that genes up-regulated in L3i were significantly involved in four biological processes (namely generation of precursor metabolites and energy GO:0006091; establishment of localization GO:0051234; proteolysis GO:0006508 and coenzyme metabolic process GO:0006732), and associated with three molecular functions (structural constituent of cuticle GO: 0042302; catalytic activity GO: 0003824; and enzyme regulator activity GO: 0030234) [See additional file 10]. This is consistent with the results of the KEGG analysis. BmL3 generate metabolites and energy by oxidative phosphorylation (GO:0006119). Coordinately, genes involved in the molecular function "electron carrier activity" (GO:0009055) are over-expressed in L3i. Most of the genes involved in the molecular function "structural constituent of cuticle" (GO: 0042302) encoded collagens (cuticular collagen 34 and 14 and 2). Interestingly, these collagens were also mapped into the biological process "phosphate transport" (GO: 0006817), and some of these genes were nematode – (BMC09088) or filarial-specific (BMC06834, TC2884). C. elegans homologs of up-regulated collagens in L3i had RNAi phenotypes such as embryonic lethal (Emb, BMC05026; BMC12506), developmental delay, and larval lethal (Lvl) (BMC05172) [See additional file 7]. These B. malayi collagens merit further study as potential vaccine targets.

Several enzyme activities such as oxidoreductase, hydrolase, and kinases were obviously increased in L3i. These molecular functions were well related to parasite survival (oxidoreductase), parasite invasion and development (hydrolase), and energy production (kinase). The increased enzyme regulator activity (GO:0030234) was mainly for enzyme inhibitor activity (GO: 0004857), especially increased activities of serine-type endopeptidase inhibitor (serpin) and a cysteine protease inhibitor (cystatin). In addition, the products of 14965.m00429 and BMC02136, up-regulated in L3i, were mitochondrial ATPase inhibitors (ATP family protein) associated with down-regulation of nucleotide metabolic processes (GO:0045980, biological process). These molecules might contribute to developmental arrest of L3i in mosquitoes through negative regulation of nucleotide metabolism. GO analysis identified the mitochondrion (GO:0005739) and extracellular region (GO:0005576) as significant cellular localizations for L3i up-regulated genes.

Altered pathways and GO terms in L3ir and L3c

The statistically enriched gene ontology list for L3ir genes shared some features with the L3i list [See additional file 12]. However, most of the L3ir up-regulated genes were located in cytoplasm (GO:0005737) in contrast to extracellular region (GO:0005576) for L3i genes. L3ir genes had fewer significant GO terms than L3i or L3c. GO terms related to catalytic activity and enzyme regulatory activity were not enriched in the L3ir gene set.

The only KEGG pathway mapped in the L3c gene set was "ribosome for protein translation". In contrast, L3c had a long list of GO terms [See additional file 13]. However, most of the up-regulated genes were involved in biological processes of protein synthesis and growth such as translation (GO:0006412) and cell development (GO:0048468). Coordinately, the molecular functions related to these biological events were also enriched such as "structural constituent of ribosome" (GO: 0003735), "binding activities of nucleic acid binding" (GO:0003676), and "nucleotide binding" (GO:0000166)). In contrast to L3i, most genes up-regulated in L3c appear to be intracellular (GO:0005622) in various compartments such as organelle (GO:0044446) and microtubule cytoskeleton (GO:0015630).

Additional comparisons of BmL3 gene expression results with those reported for C. elegans and hookworm larvae

We compared gene expression data from BmL3 with data from infective larvae of A. caninum (Ac) and L3 dauer larvae of C. elegans. Datu et al identified 602 mRNAs differentially transcribed between non-activated- and activated-Ac L3 by serum [17]. We compared transcription profiles of the non-activated AcL3 and mosquito derived BmL3 (L3i). The genes coding cytochrome C oxidase, genes encoding PRPs family member ASP, and alpha-crystallin domain containing proteins (small heat shock proteins except hsp-12.6) were up-regulated in both non-activated AcL3 and BmL3i. However, genes encoding catalytic proteases such as cysteine, serine and metalloprotease were up-regulated in BmL3i relative to BmL3c, which contrasts with finding that these genes were up-regulated in activated AcL3 relative to non-activated AcL3. In addition, the lack of similarity between activated AcL3 and BmL3c was observed in gene-by-gene comparisons. For example, PRPs and cysteine protease were up-regulated in activated AcL3 but not in BmL3c. Also BmL3i actively expressed genes involved in invasion and immune evasion, while genes that encode such PRPs were activation-associated in AcL3. The biological significance of these differences remains to be determined. However, it is important to note that AcL3 are free living organisms while BmL3 are parasitic in mosquitoes.

Consistency with prior studies and confirmation by real-time qRT-PCR

The fact that the findings from this study are consistent with previous reports tends to validate our gene expression profiles. For example, prior studies have shown those BmL3i genes such as cathepsin L-like proteases-5 (Bm-cpl-5), serine protease inhibitor-1 (Bm-spn-1), and several Bm-alt family genes are highly expressed in BmL3i [41, 55, 80]. Irradiation responsive genes in the BmL3ir gene set such as cytidine deaminase, uridine phosphorylase and collagens are also up-regulated after irradiation in other organisms [58–60, 62]. In addition, expression profiles were confirmed by real-time qRT-PCR for selected gene candidates as shown in additional file 14.

Parasite materials

BmL3 were isolated by dissection from Aedes aegypti mosquitoes that had been fed 14 days earlier on microfilaremic jirds as previously described [83]. Infective larvae were washed three times with RPMI-1640 (GIBCO-BRL, Grand Island, NY and immediately snap frozen on dry ice and stored at -80°C as mosquito derived infective larvae (L3i) as previously described [5]. Parasite cultures were performed as previously described [84]. Briefly, 2000–2500 L3i were incubated in 5 ml complete NI-medium [NCTC 135 (Sigma Chemical Co., St Louis, MO) and IMDM (GIBCO-BRL, Life Technologies, Grand Island, NY) (1:1) containing 2 mM glutamine, 200 U/ml streptomycin, 200 ug/ml penicillin, 0.5 ug/ml amphotericin B, and 10% fetal calf serum in 15 ml conical polystyrene tubes (Costar, Cambridge, Mass) at 37°C in a 95% air, 5% CO₂ atmosphere for two days. Cultured worms (L3c) were recovered after 2 days, frozen immediately, and stored at -80°C. For irradiated L3 (L3ir), L3i were placed in 2 ml of culture medium in a 5 ml polystyrene tube and exposed to a laboratory certified ¹³⁷ Cesium source (J.L. Shepherd and Associates, San Fernando, CA) at a dose of 75 krad and then cultured for 2 days as described above for L3c.

RNA isolation and probe preparation

Total RNA from L3i, L3c and L3ir parasites was prepared using TRIzol (GibcoBRL, Life Technologies) as previously described [5]. cDNA was synthesized from 5–7 ug total RNA from each sample using 3DNA capture sequence primers (3DNA Array 350 Detection system, Genisphere, Hatfield, PA) and SuperScript II Reverse Transcriptase (Gibco BRL, Gaithersburg, MD) for each probe according to standard protocols. cDNA was concentrated by Microcon YM-100 filter (Millipore) and either used immediately or stored at -80°C. cDNA was synthesized from two different RNA samples from each L3 type. A two-step protocol was used for hybridization (3DNA Array 350 Detection system, Genisphere, Hatfield, PA) as previously described [85]. Each experiment consisted of a pair-wise competitive hybridization of cDNA samples (L3i/L3c and L3ir/L3c) with reciprocal dye-flip replicates. Since biological replicates and dye-flip replicates were tested, a total of four DNA microarrays were used for each comparison of two types of cDNA.

Microarray fabrication

The BmV2array contains 18,104 elements derived from B. malayi (15,412), Onchocerca volvulus (O. volvulus, 1,016), Wuchereria bancrofti (W. bancrofti, 872) and Wolbachia (wBm, 804 genomic elements). The Brugia elements represent all B. malayi non-redundant gene models annotated in the course of the B. malayi genome project http://www.tigr.org/tdb/e2k1/bma1/bma1.shtml and ESTs from the TIGR Gene Indices (B. malayi v. 5.0 http://compbio.dfci.harvard.edu/tgi/tgipage.html) that did not match B. malayi gene models (which may represent un-annotated genes). Other elements represented O. volvulus and W. bancrofti ESTs that differ significantly from known B. malayi genes. The array also displays oligos designed on each wBm gene model. ArrayOligoselector http://arrayoligosel.sourceforge.net/[86] was used according to default parameters to choose unique 65 mers for each gene model and EST. The oligos were screened to avoid cross-hybridization to non-redundant human, bacterial and viral data nucleotide databases http://www.tigr.org/software/grid.shtml, and synthesized by standard methods by Illumina (San Diego, CA). The oligonucleotides (50 nM in 3× SSC with 0.75 M betaine) were printed in duplicate on MWG Epoxy slides (MWG Bioteche Inc, High Point, NC) by a locally constructed linear servo-arrayer (after the DeRisi model, http://derisilab.ucsf.edu/).

Data processing and analysis

Slide scanning and image analysis were performed as described previously (85). Briefly, slides were scanned immediately after hybridization on a ScanArray Express HT Scanner (Perkin Elmer, Boston, MA) to detect Cy3 and Cy5 fluorescence at 543 and 633 nm, respectively. The scanner produces green Cy3 and red Cy5 16-bit TIFF image files and extracts intensities from the scanner images for both dyes. The resultant values were background subtracted and Lowess [86, 87] normalized by using GeneSpring version 6.1 software (Silicon Genetics, Redwood City, CA). Twenty percent of the data were used to calculate the Lowess fit at each point. Genes with fold differences ≥ 2 and P < 0.01 based on pair-wise comparisons (L3i/L3c, L3c/L3ir) were considered to be differentially expressed. The best P-value obtained with high PMT scans and its accompanying arithmetic ratio was reported.

Annotation and Functional Assignments

A similarity search was performed using WU-TBLASTX https://blast.wustl.edu[88] with the 18,104 elements on the chip as queries versus the NCBI NR database (downloaded on Nov 14, 2006). Homologies were reported with a probability of 1E-05 or better. To identify cases where Brugia homologs in C. elegans have been surveyed for knockout phenotypes using RNA interference (RNAi) [18, 19, 21], Wormpep BLAST matches were cross referenced to a list of 19, 793 C. elegans genes with available RNAi information (Version 156). For each Brugia sequence, only the highest-scoring C. elegans match was considered.

Default parameters for InterProScan v13.1 were used to search against the InterPro database [89]. Gene ontology (GO) was assigned and displayed graphically by AmiGO with default parameters and ontology data released on March 15, 2007 [15]. For each GO term, its enrichment in a subset group (such as the L3i group) was measured over background using a hypergeometric analysis with the P-value cutoff e-05. Less informative ontology terms, including those at level 4 or higher for Biological Process or Molecular Function, and those at level 2 or higher for Cellular Component, were removed from the enrichment list.

An E-value cut-off of 1.0e^-10 reported by WU-BLASTP against Genes Database Release 39.0 from Kyoto Encyclopedia of Genes and Genomes was used for pathway mapping; the top match and all the matches within a range of 30% of the top BLAST score that met the cut-off were accepted as valid KEGG associations [90, 91]. A hypergeometric analysis (measuring the coverage of KEGG Orthology or KO gene groups for each KEGG pathway compared to the complete gene set on the V2 chip) was implemented to identify enriched pathways for each gene set [92].

Validation of transcription by real-time qRT-PCR

Reverse transcription real-time PCR was used for the validation of microarray data as preciously described [85]. The gene coding for histone (BMC12346) was verified and used as endogenous control as previously reported. The sequences of all of the primers used in the real-time PCR are listed in additional file 16.