Extensive Diversity of Ig-Superfamily Proteins in the Immune System of Insects

See allHide authors and affiliations

Science  16 Sep 2005:
Vol. 309, Issue 5742, pp. 1874-1878
DOI: 10.1126/science.1116887


The extensive somatic diversification of immune receptors is a hallmark of higher vertebrates. However, whether molecular diversity contributes to immune protection in invertebrates is unknown. We present evidence that Drosophila immune-competent cells have the potential to express more than 18,000 isoforms of the immunoglobulin (Ig)-superfamily receptor Down syndrome cell adhesion molecule (Dscam). Secreted protein isoforms of Dscam were detected in the hemolymph, and hemocyte-specific loss of Dscam impaired the efficiency of phagocytic uptake of bacteria, possibly due to reduced bacterial binding. Importantly, the molecular diversity of Dscam transcripts generated through a mechanism of alternative splicing is highly conserved across major insect orders, suggesting an unsuspected molecular complexity of the innate immune system of insects.

Immunoglobulin (Ig)-domain-containing proteins constitute the largest repertoire of surface receptors in animals and serve many functions in molecular recognition, cell adhesion, and signaling. Most striking is the exceptional diversity of antigen-specific receptors of the adaptive immune system in higher vertebrates, which depends on somatic gene rearrangement and clonal selection. However, somatic rearrangement of highly diverse immune receptors has been considered to exist in a relatively small number of animal species restricted to the jawed vertebrates (1, 2).

We previously identified a single Drosophila Dscam gene as a member of the Ig superfamily and initially characterized its essential function in neuronal wiring (3). Gene organization of Dscam comprises clusters of variable exons flanked by constant exons (Fig. 1A) (3). Although mechanistically entirely different from somatic rearrangements, alternative splicing of the Dscam gene combines constant and variable exons by mutually exclusive splicing and potentially generates as many as 19,008 different extracellular domains. Therefore, it is conceivable that a large protein isoform repertoire with the potential for recognizing diverse ligands and epitopes could be generated (3). To explore this, we undertook a comparative and functional analysis of Dscam expression in immune-competent cells of flies and other insects.

Fig. 1.

Dscam isoform expression in hemocytes and fat body tissue. (A) Gene and protein structure of Drosophila Dscam. Mutually exclusive alternative splicing occurs for exon clusters 4, 6, 9, and 17 (3). (B and C) Alternative exon expressions in brain and hemocytes (B) or brain and fat body cells (C). Microarrays were hybridized with Cy3- and Cy5-labeled cDNA (9), and one representative array from slides that contain triplicate arrays is shown. Spots corresponding to exon 9 sequences are indicated by a box (white line) and reveal predominant expression in brain tissue. (D) Summary of expression data for all alternative exons. Every exon is represented by a single colored square: moderate to high expression (green), low expression (yellow), and very low or no expression (red). For quantitative evaluation and definition of thresholds, see (9). Four experiments for hemocytes or fat body cells and eight for brain tissue were analyzed and presented as adjacent columns. Average values are listed separately in last column (marked by asterisks). Total number of exons expressed is given at bottom. Restricted usage of exon 9 is consistent with expression profiling studies comparing different neuronal cell types (12) as well as photoreceptor neurons and S2 cells (24).

Fat body cells and hemocytes (i.e., insect blood cells) constitute important cells of the insect immune system (4-6). Most proteins in insect hemolymph, the insect equivalent of blood serum, are produced in fat body cells, which also secrete antimicrobial peptides that constitute an important component of the humoral immune defense (7). In contrast, hemocytes are involved in cellular defense strategies such as phagocytosis and wound repair (8).

In situ hybridization of tissue from third instar Drosophila larvae with a Dscam-specific probe (9) revealed Dscam expression in fat body cells (fig. S1). For a comparison of Dscam expression in immune and neural tissue we isolated mRNA from larval hemocytes, fat body and brain tissue (9). Hemocyte-specific green fluorescent protein (GFP) expression allowed for the purification of hemocytes by fluorescence activated cell sorting (FACS) (fig. S3) (9). Reverse transcription polymerase chain reaction (RT-PCR) analysis and sequencing of ∼50 cDNAs revealed that the majority of Dscam mRNAs in hemocytes, fat body, and brain contain unique exon 4 to 6 combinations (table S1) (9).

For a global assessment of alternative splicing in different cell types we used custom-made oligo-arrays (9). Microarrays contained specific 50-mer oligo-probes for all alternatively spliced exons. Dscam mRNA sequences were amplified by RT-PCR, and cDNAs were fluorescently labeled and hybridized to the microarrays (9) (Fig. 1, B to D). We found that 59 of the 60 alternative exon 4 and exon 6 sequences were expressed in all three cell types. In brain tissue, 32 exon 9 sequences were expressed (Fig. 1D). However, only a subset (total of 14) was expressed in fat body cells and a slightly different subset (total of 15) was expressed in hemocytes (Fig. 1D). On the basis of relative expression amounts, we estimate that 80 to 90% of all Dscam mRNAs in hemocytes and fat body contain exon 9.6, 9.9, 9.13, 9.30, or 9.31, demonstrating that exon 9 splice variants in fat body cells and hemocytes are distinct from those found in brain (Fig. 1D and fig. S2B). Considering all of the alternative exons detected (12 for exon 4, 47 for exon 6, 16 for exon 9, and 2 for exon 17), we calculate that this potentially allows for the generation of more than 18,000 diverse receptor isoforms in fat body cells and hemocytes (Fig. 1) (9).

We raised antibodies against extracellular (D-ex1, D-ex2) and intracellular (D-cy) domains of Dscam (Fig. 2A) (9). All antibodies recognized an ∼210-kD endogenous form of Dscam in extracts from cultured S2 cells (Fig. 2, B to D), a cell line thought to be derived from embryonic hemocytes and shown to share many characteristics with hemocytes (10, 11). A 210-kD form of Dscam was also confirmed in purified larval hemocytes (Fig. 2, C and D, and fig. S3) (9), fat body tissue, and at comparatively high amounts in brain (Fig. 2, D and E).

Fig. 2.

Dscam protein is expressed in immune-competent cells. (A) Proteins recognized by antibodies to Dscam D-cy, D-ex-1, and D-ex-2. (B) Dscam protein ∼210 kD in S2 cells and in brain tissue (Brain) lysates. RNAi treatment (DscamRNAi) in S2 cells selectively depletes expression of the 210-kD band compared with untreated (no RNAi) or control-treated (LacZ-RNAi) cells. (C) Hemocytes isolated from independent transgenic lines (UAS-Line1 and UAS-Line2) of larvae that express dsDscam-RNA(9) (RNAi knock-down) have strongly reduced Dscam protein amounts compared with control (Wt Control, lanes 3 and 5). (D) Full-length Dscam protein (∼210 kD) is present in fat bodies (three fat bodies per lane), purified hemocytes (3 × 104 cells per lane) (9), and brain (one brain per lane). S2 cells (1 × 104 cells per lane) treated with control dsRNA (LacZ-RNAi) or Dscam dsRNA (DscamRNAi) show specificity of Dscam detection. (E) Dscam protein was immunoprecipitated (IP) from fat body (FB) or brain (Br) by using D-ex2 or D-cy. Western blot analysis detects 3 Dscam protein forms (⚫) in fat bodies (9). (F) Dscam protein IP from lysates of S2 cells (cells) or conditioned media (media) using D-cy (lanes 1 to 4) or D-ex1 (lanes 5 to 8). Western blot analysis using D-cy (top) or D-ex1 (bottom) indicate presence of shorter protein forms in S2 cells (lane 5). The smaller ∼170-kD band (top arrowhead) likely represents a Dscam protein lacking a cytoplasmic domain. A truncated Dscam protein of ∼160 kD (bottom arrowhead) is only detected in S2 cell-conditioned media lane 7, marked with asterisk). Dscam RNAi–treated samples (+) and LacZ RNAi–treated control samples (–) served as controls. (G) Brain tissue (Br) or larval hemolymph (Hl) were immunoprecipitated with D-ex1 or D-cy. Western blot analysis reveals a truncated Dscam protein of ∼160-kD in hemolymph (lane 4, marked with asterisk) not detected with D-cy (lane 2).

Immunoprecipitations from fat body extracts revealed three Dscam forms (Fig. 2E), possibly representing truncated forms generated by proteolytic cleavage. Unexpectedly, we found that S2 cell-conditioned medium contained a soluble Dscam protein of ∼160 kD (Fig. 2F) and that a secreted Dscam protein of the same molecular weight was also present in hemolymph serum (Fig. 2G).

Liquid chromatography and tandem mass spectrometry (LC-MS/MS) directly confirmed that S2 cells secrete Dscam isoforms (9) (table S2 and fig. S4). Coverage of the secreted forms by the identified peptides amounts to more than 50% of the entire extracellular part of Dscam (fig. S4A). Importantly, some of the identified peptides confirmed the presence of alternatively spliced sequences, including five Ig2 sequences and at least 12 Ig3 sequences (table S2 and fig. S4, B and C). In agreement with the expression profiling of exon 9 (Ig7) sequences (Fig. 1), we identified three distinct Ig7 domains (i.e., Ig76, Ig79, and Ig713), which correspond to the most abundantly expressed exon 9 sequences (Fig. 1 and fig. S2B).

Considering the protein sequencing results (table S2 and fig. S4), the presence of secreted Dscam in hemolymph (Fig. 2), and the large pool of diverse Dscam mRNAs in fat body cells or hemocytes (Fig. 1D), it is possible that thousands of Dscam isoforms circulate in the hemolymph of Drosophila.

We next sought to determine whether Dscam proteins are functionally required in immune-competent cells. However, because animals with homozygous amorphic mutations in Dscam die as embryos (3, 12), it was not possible to directly test this in null mutant animals. Nevertheless, we were able to purify GFP-labeled hemocytes from Dscam mutant larvae that carry a transallelic combination of hypomorphic (Dscam39) and amorphic (Dscam20) mutations (13). Immunoblotting showed that Dscam20/Dscam39 animals have a strong overall reduction in protein concentration (Fig. 3A). One important function of hemocytes is the ingestion of bacterial pathogens by phagocytosis (8). We therefore challenged wild-type and Dscam-deficient hemocytes with heat-killed fluorescently labeled Escherichia coli and determined the number of hemocytes containing fluorescent bacteria [phagocytic index (Fig. 3B and fig. S5)] (9). Normal hemocytes exhibited highly efficient phagocytosis, and 85 to 90% had taken up bacteria after 10 min (Fig. 3B). In contrast, only 55% of Dscam mutant cells had taken up bacteria (Fig. 3B).

Fig. 3.

Dscam is required in hemocytes for efficient phagocytosis and binds to E. coli. (A) Western blot analysis of larval brain tissue from control (+/+) or Dscam39/Dscam20 larvae reveals strong reduction of Dscam protein concentration. (B) Phagocytosis assays (9) were performed on hemocytes isolated from Dscam39/Dscam20 mutant larvae. Significant reduction (marked by double asterisks) in phagocytosis was found in Dscam-deficient hemocytes but not wild-type (+/+) hemocytes. (C) Assays performed on Dscam-deficient hemocytes (Dscam-RNAi). Hemocytes from three independent UAS lines showed a significant reduction (P < 0.01) compared with wild-type (Wt) or ds-LacZ RNA-treated hemocytes (LacZ-RNAi) (9). (D) Treatment of S2 cells with neutralizing antibody to Dscam significantly (P < 0.01) decreases phagocytosis. S2 cell treatment with either Schneider media vehicle control (control) or the pre-immune serum (pre-immune) served as controls. Error bars in (B) to (D) indicate SEM. (E and F) Dscam isoforms show binding to E. coli bacteria (DH5α strain). Binding of Fc-tagged recombinant proteins (E) or control antibodies (F) was detected by FACS (9) using Alexa–Fluor (Molecular Probes, Eugene, OR) 488 conjugated to protein A. (E) Dscam-7.27.25 isoform (blue) and Dscam-7.27.13 isoform (green) show binding when compared with the control sample incubated with protein A Alexa–Fluor 488–conjugated only (gray). Dscam-1.30.30 (red) shows no binding above background. (F) Fc peptides (yellow) and an unrelated mouse antibody (Anti-IgG, green) exhibited no binding and were used as negative controls.

To investigate more directly the possible role of Dscam in immune defenses, we addressed three questions: First, is Dscam cell autonomously required for phagocytosis in hemocytes? Second, can antibodies that specifically bind extracellular Ig domains of Dscam acutely interfere with phagocytosis? Third, can Dscam isoforms directly bind to pathogens?

We used expression of double-stranded RNA [i.e., RNA interference (RNAi)] to suppress Dscam expression in transgenic flies (9) (Fig. 3C). A hemolectin promoter region–GAL4 fusion, termed Hml-GAL4, was used for activating expression exclusively in embryonic and larval hemocytes (14, 15). Hemocytes with Dscam-specific knock-down showed a substantially reduced rate of phagocytosis, with less than 60% of the cells containing bacteria (Fig. 3C). This partial inhibition may reflect RNAi-mediated knock-down in only a subset of the highly heterogeneous cell population of larval hemocytes (8, 16). We therefore examined S2 cells, which represent a less heterogeneous cell population also capable of phagocytosis (Fig. 3D), and used antibodies to Dscam to block Dscam function (9). We reasoned that the short application of antibodies against Dscam, in contrast to continuous RNAi, may be less likely to influence general hemocyte characteristics or development (8, 16). Treatment of S2 cells with polyclonal α-Dscam serum D-ex1 resulted in a 30% decrease in the phagocytic index (Fig. 3D). It is possible that the α-Dscam antibody may not directly block Dscam-bacteria interactions or may have additional indirect influences on the process of phagocytosis. However, the reduction of phagocytosis is consistent with the loss-of-function in vivo analysis (Fig. 3, B and C) and in vitro binding studies presented below. Taken together, partial but significant reduction in phagocytosis could be achieved by genetic inhibition of expression in hemocytes and by blocking Dscam protein interactions.

We next tested by flow cytometry whether different Dscam isoforms are capable of binding directly to bacteria. Validity of a standard binding assay was tested by using a polyclonal antibody that specifically recognizes E. coli epitopes (fig. S6) (9), and the same assay was used to test binding of different recombinant Dscam isoforms (Fig. 3, E and F, and fig. S6). All isoforms contained C-terminal Fc tags (9), which were used for detection with fluorescently labeled protein A (9). Isoforms are designated by the combination of alternative variable Ig domains (9). Dscam-1.30.30-Fc and Dscam-7.27.25-Fc contain all of the extracellular domains, whereas Dscam-7.27.13-Fc contains only the N-terminal 9 Ig plus the first FNIII domain. We found that Dscam-7.27.25-Fc and Dscam-7.27.13-Fc could bind to live DH5α E. coli bacteria (Fig. 3E and fig. S6). Binding of Dscam-7.27.13-Fc to E. coli (Fig. 3E) suggests that the 10 N-terminal domains containing all three variable Ig domains are sufficient for binding. In contrast, binding of isoform Dscam-1.30.30-Fc to E. coli is barely detectable (Fig. 3E and fig. S6) and therefore similar to Fc peptides alone or control Ig domains containing antibodies to heavy chain (mouse) (Fig. 3F). It is possible that lack of binding of Dscam-1.30.30-Fc is unique to just this isoform and may not generally reflect the presence of distinct pools of binding and nonbinding isoforms. Therefore, it remains an important task to examine in future studies binding properties of other isoforms. Importantly, the molecular basis of Dscam binding to bacteria is presently unknown, and an assessment of binding specificity will crucially depend on the identification of potentially distinct epitopes on bacteria.

Although we do not know the detailed molecular basis of Dscam function in immune-competent cells, our results are consistent with the possibility that Dscam acts as a signaling receptor or co-receptor during phagocytosis. In addition, binding of Dscam isoforms to bacteria may reflect the possibility that diverse secreted Dscam isoforms are involved in opsonizing invading pathogens in the hemolymph.

Comparative genomic analysis of Dscam-like sequences show high conservation of orthologous Dscam genes in Diptera and Hymenoptera orders (17). To explore Dscam expression and alternative splicing in other insect orders, we examined Dscam gene structure and expression in the flour beetle Tribolium castaneum (Coleoptera) and the silk moth Bombyx mori (Lepidoptera) (18). Orthologous genes were identified in both species (9), and all Dscam-like domains were found to be highly conserved (Fig. 4 and fig. S7). We confirmed the expression of alternative Dscam isoforms by cloning and characterizing 32 cDNAs from Tribolium RNA (Tr-Dscam) (Fig. 4B) (9). Alternatively spliced mRNA segments of Tr-Dscam matched corresponding Ig2, Ig3, and Ig7 segments of Drosophila Dscam (Fig. 4). RT-PCR and sequencing of Dscam mRNA extracted from fat body tissue of Tribolium larvae (9) revealed nine different isoform sequences (out of 16 cDNAs) (Fig. 4B). These results suggest that expression of diverse Dscam isoforms in immune-competent fat body cells is conserved among highly diverged insect species.

Fig. 4.

Conservation of Dscam diversity in insects. (A) Simplified phylogenetic tree indicating estimated number of species. Ord ers with less than 7000 species are indicated by gray lines (18). Alternative splicing is present in all holometabolous insect orders. Abbreviations are as follows: T. castaneum, T.c.; Apis mellifera, A.m.; Bombyx mori, B.m.; Anopheles gambiae, A.g.; and D. melanogaster, D.m. Table lists number of exons identified. (B) Alignment (25)of Ig3 sequences (N-terminal half) of Tribolium. Twenty-eight alternative exons are predicted to encode Ig3 sequences or verified by RT-PCR from embryonic (e) or fat body (fb) tissues (9). Numbers of cDNAs are listed on right side. Gray shading indicates conserved amino acids. (C) Sequence alignment of the Ig domains 2 and 3 of D. melanogaster, Anopheles gambiae, B. mori, Apis mellifera, and T. casteneum. Despite the large evolutionary distances, C-terminal halves of Ig2 and Ig3 are nearly identical. A highly conserved Ig-domain scaffold embeds alternatively spliced sequences, which are species-specific and show little conservation. Variable N-terminal half of Ig2 and Ig3 reveal distinct high variability regions (hatched rectangles). Location of variability hot spots is conserved across species. Note that the locations of sequence variability in Dscam Ig domains is distinctly different from variability in complementary determining regions of V-type Ig domains of mammalian immune receptors. Ig2 domains are most variable at the end of exon 4, likely to correspond to the C′ strand of the Ig-fold. Ig3 domains are most variable at the middle of exon 6, possibly corresponding to the A and A′ strands, similar to Amphioxus VCBP sequence variability (20).

Our study provides evidence for a potentially extensive repertoire of thousands of Ig-domain-containing proteins in immune-competent cells of insects, which represent an estimated 60% of metazoan species (19). Recently, novel and diverse receptor sequences have been identified in jawless vertebrates (lamprey), protochordates (Amphioxus), and mollusks (freshwater snail) (20-22). It has also been reported that a large class of scavenger receptors (with an estimated 1200 scavenger receptor cysteine-rich domains) are expressed in putative immune effector cells (coelomocytes) of echinoderms (19). Similarly, immune responses of crustaceans apparently use an extensive set of diverse antimicrobial peptides (23). Although most animals have not acquired adaptive immunity, this apparently broad conservation of receptor diversity strongly suggests important functions, and future studies will have to further address whether the presence of diverse immune receptors in invertebrates increases the effectiveness of immune responses of individual animals. Alternatively, given the relative short life span of many invertebrates, it may be that immune receptor diversity is less important ontogenetically but rather enhances the adaptive potential of animal populations to changing environmental and pathogenic threats.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

Tables S1 and S2

References and Notes

References and Notes

View Abstract

Navigate This Article