INTRODUCTION

The antennae of silk moths are well known for their exquisite sensitivity to pheremonal odorants (1-3). Early reports demonstrated that the males of the wild silk moth Samia cynthia could locate a sex pheromone source over 2 miles away within several hours of their release (4). Studies of the silk moth Bombyx mori suggested that a single pheromone molecule was sufficient to activate olfactory neurons in the antenna (5). In insects, odors are detected by sensilla, small hair-like structures arrayed along the antennae. The sensilla are hollow, fluid-filled cuticular structures that contain the receptor cilia of olfactory neurons. Small holes penetrate through the wall of a sensillum, permitting entry of odor molecules; odorant-binding proteins are then thought to transport the odor molecules through the fluid-filled lumen to receptor proteins in the receptor membranes of the olfactory neurons (3, 6-9).

In searching for proteins that might support olfactory mechanisms, we characterized the protein profile of olfactory neuron receptor membranes of the wild silk moth Antheraea polyphemus. The morphology of the A. polyphemus antenna permits the relatively easy isolation of olfactory sensilla in a manner yielding olfactory receptor cilia as the only cellular component (10). This preparation is free of other parts of the olfactory neurons as well as of nonneuronal cells of the antenna, and it was previously used to identify a pheromone-binding membrane protein using a radiolabeled photoaffinity analog of the A. polyphemus sex pheromone (10). This protein co-migrated with bovine serum albumin on SDS gels (around 67 kDa) and appeared to be uniquely expressed in antennal tissue and to be a principal component of the olfactory cilia membrane.

We have now used this A. polyphemus sensilla preparation to isolate and clone a 67-kDa membrane protein that is a major component of the ciliary receptor membranes of olfactory neurons; we have named this protein Snmp-1, or sensory neuron membrane protein-1. Molecular analyses suggest that Snmp-1 mRNA is uniquely expressed in olfactory neurons and that the protein is localized to the cilia, dendrites, and soma but is absent from the axons. Snmp-1 expression appears to initiate with the developmental appearance of the olfactory neurons, and it increases significantly late in antennal development, coincident with the acquisition of olfactory function. Sequence analysis indicates that Snmp-1 is homologous with the CD36 family of proteins, an as yet small family of receptor-like membrane proteins with identified members in vertebrates, arthropods, and nematodes. The few vertebrate CD36 proteins characterized thus far are thought to have two transmembrane domains and to interact with proteinaceous ligands (11-16). Snmp-1 is the first member of the CD36 family isolated from neural tissue.

EXPERIMENTAL PROCEDURES

A. polyphemus silk moths were obtained as diapausing pupae from D. Bantz of Racine, Wisconsin, or collected in Wisconsin as wild males lured to caged females emitting sex pheromone. Pupae were held in diapause at 4 °C and allowed to develop to adults by incubation at 27 °C. Antennae from wild males were excised upon capture onto dry ice and were stored (70 °C) for subsequent processing.

Olfactory sensilla were collected from 800 A. polyphemus antennae (wild caught males). Antennal branches were isolated and processed in batches (40 antenna equivalents) as described previously (10). Dried sensillar material was recovered (200 mg), and debris was thoroughly removed under a microscope. Branch material was collected for later mRNA isolation and cDNA library construction. Frozen sensilla were lyophilized, suspended in homogenization buffer (10 ml of TEM-P; 10 mM Tris-HCl, 10 mM EGTA, 1 mM MgCl₂, pH 7.0, 0.1 mM phenylmethylsulfonyl fluoride) using a Dounce homogenizer and transferred to a 30-ml Corex centrifuge tube. Sensilla suspension was sonicated (5 min, Beckman microprobe, setting 45, in ice/water) and centrifuged three times (2,500 rpm; Sorvall HB-4 rotor; 10-ml resuspensions in TEM-P). The pooled supernatants were ultracentrifuged (35,000 rpm, Sorvall TH641 swinging bucket rotor, 4 °C, 90 min), the supernatant was collected, and the pellet was resuspended (TEM-P) and ultracentrifuged a second time. The final pellet was suspended in H₂O with brief sonication and stored in aliquots (70 °C). For electrophoresis, samples were lyophilized, dissolved in SDS sample buffer, and denatured as described previously (10).

For direct sequencing, proteins were electrophoretically transferred (Bio-Rad Trans-Blot 170-3910) as described by Vogt et al. (18), either to polyvinylidene difluoride membranes (Immobilon-P, Millipore Corp.; Refs. 19 and 20) or to glass fiber filters (Whatman GF/C) modified by soaking in trifluoroacetic acid (21). For transfer to GF/C filters, SDS was first removed by washing the gel in four changes of 2% Nonidet P-40 in H₂O (200 ml each wash); the gel was then rinsed in H₂O (three times) and transferred in 1% acetic acid overnight at 80 V and 4 °C. GF/C and polyvinylidene difluoride blots were stained in Coomassie Blue and destained in acetic acid/methanol (18). Stained bands were excised and sequenced using an Applied Biosystems model 470A gas phase microsequencer equipped with a model 120A on-line phenylthiohydantoin detector (by Dr. Kenneth Williams and Kathy Stone, Protein and Nucleic Acid Chemistry Facility, Yale University School of Medicine). Three blots and subsequent sequence analyses were performed; repetitive yields ranged from 91.2 to 91.8%. The 67-kDa protein yielded the N-terminal sequence MLLPKPLKYAAIGGGVFVFGILIGXVIFPV. The lack of ambiguity in the obtained sequences indicated that these sequences derived from the abundant protein shown (Fig. 1A) rather than from a minor co-migrating protein.

All RNA isolations utilized the acid guanidinium thiocyanate-phenol-chloroform method (22, 23); tissue was initially ground under liquid nitrogen in a mortar and pestle in the presence of guanidinium thiocynate solution. cDNA template was synthesized using cloned Moloney murine leukemia virus RNase H reverse transcriptase (Life Technologies, Inc.), following recommended protocols and including 400 units of RNasin (Promega) and 5 µg of total RNA in a 40-µl reaction.

For PCR, degenerate primers were designed against four regions of the derived N-terminal amino acid sequence; degeneracy was reduced by making four primers to each region (Table I). PCR (100 µl) used Taq DNA polymerase (3 units; Promega), supplied buffer containing Triton and 1.5 mM magnesium, 2 mM dNTP, 2 µCi of [³²P]dCTP, 50 pmol of each primer, and 1 µl of cDNA from the above reaction. PCR was performed on a Perkin-Elmer thermocycler under oil overlay: 3 min 94 °C followed by 35 cycles at 94 °C (30 s), 52 °C (1 min), and 72 °C (3 min). PCR products were analyzed by 8% PAGE in TBE (24); products were visualized by autoradiography on x-ray film (Kodak X-Omat).

Table I. Oligonucleotides used for PCR Oligonucleotides 1-4 are sense; 5-16 are antisense. All oligonucleotide sequences are 5-3 under amino acid sequences. N-terminal  M  L  L  P  K  P 1 ATGCTNCTNCCNAATCC 2 ATGCTNTTRCCNAARCC 3 ATGTTRCTNCCNAARCC 4 ATGTTRTTRCCNAARCC C-terminal G  I  A  A  Y  K 5 CCDATRGCRGCRTAYTT 6 CCDATYGCRGCRTAYTT 7 CCDATRGCYGCRTAYTT 8 CCDATYGCYGCRTAYTT C-terminal I  G  F  V  F  V 9 ATNCCRAARACRAARAC 10 ATNCCRAARACRAAYAC 11 ATNCCRAAYACRAARAC 12 ATNCCRAAYACRAAYAC C-terminal V  P  F  I  V 13 ACRGGRAADATRAC 14 ACYGGRAADATYAC 15 ACYGGRAADATRAC 16 ACRGGRAADATYAC

A "best" primer pair was determined based on the PCR results and used to amplify a 65-base pair product; primers were constructed with EcoRI sites: TTTTGAATTCATGTTRTTRCCNAARCC (sense) and TTTTGAATTCATNCCRAATACTAAYAC (antisense). The resulting PCR product was digested with EcoRI, gel-purified (8% PAGE), ligated into M13 vector, and sequenced (dideoxynucleotide method; Sequenase kit, U. S. Biochemical Corp.).

A random primed adult male antennal cDNA library was constructed in  Zap II (Stratagene) from mRNA isolated from the antennal branch material (described above) employing the RNase H method. cDNA (300 ng) was synthesized (described above) from 4 µg of mRNA (selected four times on oligo(dT)-cellulose) and 0.8 µg of random hexamer primer. Prior to ligation, cDNA larger than 800 bp was gel-purified and precipitated with excess glycogen to yield 36 ng of cDNA. The resulting library had a titer of 300,000 plaque-forming units.

The library (180,000 plaque-forming units) was screened in XL1 Blue bacteria (Stratagene) following recommended protocols. Probe was a 32-base synthetic oligonucleotide (see Fig. 2) end-labeled with ³²P ATP using a T4 polynucleotide kinase (24) and used at 5 × 10⁶ cpm/ml. Nylon membrane (ICN) lifts were prehybridized (15 h) and hybridized (24 h) in 5 × SSC (24), 5 × Denhardt's solution, 0.1% SDS, and 0.2 mg/ml salmon sperm DNA at 50 °C without formamide. Following hybridization, membranes were washed twice in 2 × SSPE (24), 0.1% SDS at room temperature, once in 2 × SSPE, 0.1% SDS at 50 °C, and once in 0.2 × SSPE, 0.1% SDS at 50 °C, all for 20 min. Membranes were exposed to Kodak X-Omat film. Positive plaques were eluted into 0.5 ml of SM buffer (24) and rescreened at low density. Final positive clones were subjected to plasmid rescue following recommended protocols (Stratagene) yielding 24 cDNA clones in pBluescript II (SK⁺). Following PCR analysis using the degenerate primer pairs 4-10 and 4-13 (Table I), a clone designated RP11 (random prime 11) was chosen for further analysis.

snmp-1(RP11) was sequenced as double-stranded DNA by the dideoxynucleotide termination method (Sequenase kit, U. S. Biochemical Corp.). The entire clone was sequenced in both sense and antisense directions following exonuclease deletions from both 5 and 3 directions (ExoIII/mung bean nuclease deletion system, Stratagene).

RNA was in vitro transcribed from the full-length RP11 cDNA clone linearized with KpnI, following protocols described below. Transcribed RNA (2 µg) was in vitro translated to protein using nuclease-treated rabbit reticulocyte lysate (Promega) in the presence of [³⁵S]methionine (ICN Tran³⁵S-label), subjected to SDS-PAGE, and visualized by fluorography (17).

Proteins with sequence homology to the deduced amino acid sequence of snmp-1 were identified by using the NCBI BLAST network server (25); additional analysis of BLAST identified homologues employed the FASTA algorithm (26, 27). Snmp-1 and CD36 family member amino acid sequences were aligned with the Clustal W multiple alignment program (28) for all group and pairwise comparisons.

Snmp-1 RNA probe used in the Northern blot analyses and in situ hybridizations was in vitro transcribed from the snmp-cr subclone (base pairs 1-1578, Fig. 2). Snmp-cr was prepared by PCR amplification of the coding region using primers specific to the extreme 5 and 3 ends; product was gel-purified and ligated into pCR-Script vector (Stratagene). Digoxigenin-incorporated RNA probes (antisense and sense) were synthesized from linearized plasmid using T7 or T3 RNA polymerase (Stratagene) following recommended protocols (Boehringer Mannheim) and in the presence of 40 units of RNasin (Promega). For in situ hybridization studies, RNA was alkaline-degraded to an approximately 160-base length (29).

RNA probe for the ribonuclease protection assays (RPAs) was transcribed from the snmp-p subclone (base pairs 563-890, Fig. 2). snmp-p was prepared by PstI removal of a 1,850-base pair fragment from the snmp-1(RP11) clone. Snmp-p was linearized with NcoI and transcribed to produce a 408-base (327 bases of snmp-p + 81 bases of vector) ³²P-labeled antisense RNA probe using the Maxiscript RNA transcription kit (Ambion) following recommended protocols (specific activity  3 × 10⁸ cpm/µg). Full-length transcript was purified by PAGE (5% acrylamide, 8 M urea) and autoradiography.

The control probe used in the Northern blot analysis derived from a Manduca sexta 18 S rRNA clone (GenBank^TM accession number U88190[GenBank]) in pBluescript (Stratagene). DNA was amplified by PCR using M13 forward and reverse primers in the presence of 2 mM dATP, dGTP, and dCTP, 0.6 mM dTTP, and 0.3 mM digoxigenin-dUTP (Boehringer Mannheim). Amplification was performed as follows using an Idaho Technology thermocycler: 2 min of denaturation at 94 °C; 35 cycles at 94 °C (15 s), 50 °C (30 s), and 74 °C (25 s).

For each sample, 5 µg of poly(A)-enriched mRNA (Fig. 5A) or 10 µg of total RNA (Fig. 5C) was electrophoresed on a 1% agarose gel containing formaldehyde (24) and electrophoretically transferred (Trans-Blot Cell; Bio-Rad) in 1 × TAE (24) onto nylon membrane (HyBond-N, Amersham Corp.). Membranes were prehybridized for 2.5 h at 68 °C (5 × SSC, 0.1% N-lauroylsarcosine, 2 × Denhardt's solution, 0.02% SDS, 100 µg/ml herring sperm DNA) and hybridized with 25 ng/ml digoxigenin-labeled snmp-cr probe under the same conditions in a solution containing 50% formamide. To show that equivalent amounts of RNA were present in all lanes (Fig. 5C), M. sexta rRNA DNA probe was hybridized to the portion of the blot containing target rRNA and processed separately; hybridization was at 50 °C but at otherwise identical conditions to the snmp-cr blot. Both blots were washed at hybridization temperature in 0.1 × SSC, 0.1% SDS; probes were visualized by luminous detection (Lumiphos 530; Boehringer Mannheim) on x-ray film (Kodak, X-Omat).

For each sample, 10 µg of total RNA was hybridized with ³²P-labeled snmp-p(NcoI) probe (8.7 × 10⁴ cpm; 14-16 h; 42 °C; 20 µl of hybridization solution (80% formamide, 100 mM Na⁺-citrate (pH 6.4), 300 mM Na⁺-acetate, 1 mM EDTA)) following recommended protocols (RPA II kit, Ambion). Unhybridized RNA was degraded by incubation in 200 µl of supplied RNase digestion mixture (containing RNase A and T1) for 30 min at 37 °C. The resulting protected RNA fragments were precipitated and visualized following PAGE (5% acrylamide, 8 M urea) on x-ray film exposed 6-12 h at 70 °C.

All histology was done on newly emerged adult tissue; hybridization protocols were modified from Byrd et al. (29). Antennae were partially dissected and fixed by perfusion and subsequent incubation in 2% paraformaldehyde in phosphate-buffered saline (PBS), overnight on ice, dehydrated to 70% methanol, and stored at 20 °C. For sectioning, tissue was transferred to 70% ethanol, dehydrated though a graded series of ethanol and toluene, and incubated in melted paraffin (Periplast+) for 2-4 h before being embedded in plastic molds. Longitudinal sections and cross-sections (10 µm) were transferred to water drops on slides coated with albumin (Albumin Fixative, EM Diagnostic Systems). Slides were dewaxed in xylene, and sections were treated with Proteinase K (5 µg/ml in PBST (PBS, 0.1% Tween 20) for 15 min at room temperature), followed by acetic anhydride (0.25% in 100 mM triethanolamine for 5 min at room temperature); two washes with glycine/PBS (2 mg/ml glycine, 5 min/wash) were applied between treatments. Sections were prehybridized overnight at 42 °C (0.6 M NaCl; 10 mM Tris, pH 7.5; 2 mM EDTA; 1 × Denhardt's solution; 50 µg/ml herring sperm DNA; and 50 µg/ml tRNA) and hybridized with 100 ng/ml digoxigenin-labeled snmp-cr probe or zebrafish odor receptor probe 1A (negative control, Ref. 29) under the same conditions but in the presence of 50% formamide. Sections were washed in hybridization solution (minus Denhardt's solution, DNA, tRNA, and probe) progressively diluted with PBST. Hybridized probe was enzymatically detected by phosphatase reaction with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium following recommended protocols (Boehringer Mannheim, Ref. 29). Slides were coverslipped with Aquamount mounting medium (Lerner Laboratories) and photographed with differential interference contrast optics.

Polyclonal antiserum (rabbit) was prepared against recombinant Snmp-1. The snmp-1 coding region was cloned into the expression vector pET 15b (Novagen) and transfected into DE3 (pLysS) cells. Cells were grown to a 0.6 optical density (600 nm) at 37 °C, induced with isopropyl-1-thio--D-galactopyranoside (1.0 mM final concentration), and cultured at 27 °C for 4 h. Centrifuged bacterial pellet from a 50-ml culture yielded approximately 1 mg of recombinant Snmp-1 protein. Induced 59.9-kDa protein was isolated from the insoluble fraction of bacterial lysates by a His tag affinity column (Novagen) followed by SDS-PAGE. Gel slices were homogenized through syringes with 10 mM Tris (pH 7.0) and an equal volume of Freund's complete adjuvant and were used to immunize a rabbit by subcutaneous injection (University of South Carolina Institute for Biological Research and Technology Antibody Facility). Preabsorbed antiserum was prepared by incubation with DE3 bacterial lysate. Bacteria were lysed in 50 mM Tris, pH 7.9; 2 mM EDTA; 0.1% Triton X-100; 250 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride; 100 ng/ml lysozyme; 30 °C, 30 min. The lysate was sheared by passage through a syringe (23 g), incubated with total antiserum (1:5000 serum dilution; 4 °C, overnight), and centrifuged (16,000 × g, 4 °C, 20 min). The resulting supernatant was used as antiserum for Western blot analysis.

For Western blot analysis, bacterial lysate (20 µl) containing expressed Snmp-1 and A. polyphemus antennal branch homogenates (2 antennal equivalents) were subjected to SDS-PAGE (10%), transferred to nitrocellulose membrane (Trans-Blot Cell, Bio-Rad; BA-S NC, Schleicher & Schuell), and incubated with Snmp-1 preabsorbed antiserum (36 h, 4 °C). The membrane was then incubated with goat IgG horseradish peroxidase conjugate (ICN; 1:5000, 2 h, room temperature) and stained (VIP substrate; Vector). Preimmune serum was used under otherwise identical conditions as a negative control. All antibody treatments included 3% nonfat dry milk in PBST as a blocking agent, and all washes were in PBST.

For immunocytochemistry, tissue sections were prepared as described above for in situ analysis. Sections were dewaxed with toluene and incubated with total Snmp-1 antiserum (1:1000, 15 h, 4 °C) followed by goat IgG horseradish peroxidase conjugate (ICN; 1:100, 2 h, room temperature) and stained with VIP substrate (Vector) following recommended protocols. For a negative control, sections were incubated with preimmune serum under identical conditions. All washes and antibody treatments included 3% nonfat dry milk in PBS as a blocking agent. Slides were coverslipped with Permount (Fisher) and photographed using bright field or differential interference contrast optics. Antisera were immunohistochemically active at dilutions to 1:10,000.

All x-ray film and photographic images were digitized and processed using Adobe Photoshop and printed using a Kodak ColorEase dye sublimation printer.

RESULTS

Membranes of olfactory cilia were isolated from 800 antenna of wild caught A. polyphemus males collected in Wisconsin. SDS-PAGE analysis indicated an approximately 67-kDa protein as a major component of this membrane preparation (Fig. 1A), here named Snmp-1. N-terminal amino acid analysis yielded a 30-amino acid sequence. PCR primers based on this sequence were used to amplify cDNA derived from leg and antennal mRNA; an antenna-specific product of the expected size was cloned and sequenced and then verified to encode the N-terminal amino acid sequence. A 32-base oligonucleotide was synthesized and used to probe a cDNA library.

A random primed cDNA library was constructed from antennal mRNA and screened with the N-terminal 32-base oligonucleotide probe; 24 clones were obtained. Two clones were chosen and sequenced in their entirety. Both clones yielded matching sequences; each contained the N-terminal amino acid sequence obtained from the 67-kDa protein and identical in-frame stop codons following base position 1575. The cDNA sequence of one of these Snmp-1 clones identified as RP11 (random prime clone 11) is presented along with its translated amino acid sequence (Fig. 2).

The snmp-1(RP11) cDNA consists of 2,726 nucleotides and contains 1,575 nucleotides encoding a protein of 525 amino acids with a derived molecular mass of 59,917 Da. In vitro translation of RNA transcribed from full-length snmp-1(RP11) clone revealed a protein of about 59 kDa by SDS-PAGE (Fig. 1B). Antiserum generated against 59.9-kDa bacterial expressed Snmp-1 protein recognized a 67-kDa protein isolated from antennal extracts; preimmune serum showed no detectable immunoreactivity against antennal proteins (Fig. 1C). The approximately 7-kDa difference between cloned and native proteins may be due to post-translational glycosylation. Sequence analysis indicates four possible N-linked glycosylation sites within a generally hydrophilic region of the protein (Fig. 2). Identification of the translational start site is based on the start methionine in the N-terminal amino acid sequence obtained from the purified 67-kDa protein. No additional ATG was observed 5 to this site in the cDNA sequence, although only 27 5 nucleotides were present in the analyzed snmp-1(RP11) clone. Support for the stop codon position and size of open reading frame comes from the observed molecular weight of in vitro translated and bacterial expressed protein (Fig. 1, B and C) as well as the presence of 23 additional in-frame stop codons downstream from the predicted termination site.

A data base homology search using the NCBI BLAST network server identified five candidate SNMP-1 homologous proteins; probability values range from 10²⁷ to 10⁴³, where a value 0.05 is considered statistically significant (30). FASTA PRDF statistical comparison of Snmp-1 with these five proteins generated a similarly significant value of 10²⁷ to 10³⁰ (26, 27). These presumptive homologues compose the CD36 receptor family: human CD36 (31); human CLA I (32); rat LIMP II (12); Drosophila emp (14); nematode cm08 h8 (GenBank^TM accession number M89468[GenBank]); and hamster SR-BI (33). Fig. 3 shows an alignment of SNMP-1 with these five proteins. Overall identities range from 20.6 to 24.9% (Table II). However, two amino acid domains show much higher amino acid conservation. A 44-amino acid domain (residues 63-107, 31.8-45.5%, Table II) contains a putative N-linked glycosylation site conserved in five of six proteins including SNMP-1, and a 92-amino acid domain (residues 273-365, 29.3-44.6%, Table II) contains four cysteine residues conserved in all proteins and an additional cysteine conserved in five of six proteins including SNMP-1 (see Figs. 2 and 3). Fig. 4 presents hydropathy plots of Snmp-1 and the CD36 proteins; two prominent hydrophobic domains near the N and C termini are conserved in all proteins. A third hydrophobic domain is evident in Snmp-1 (Figs. 2 and 4); this domain is relatively short but unique to Snmp-1.

Table II. Percentage of amino acid identities between SNMP-1 and CD36 family members Percentage of identities obtained from pairwise alignments are shown. Amino acid numbers (63-107, 273-365) correspond to SNMP-1 positions. See "Results" for identities of CD36 family members. Protein Overall 63-107 273-365 % Emp 24.9 45.5 29.3 CD36 20.6 31.8 44.6 LIMP II 22.3 41.0 31.5 SR-BI 24.4 45.5 37.0 CLA I 23.8 45.5 38.0

Tissue localization of Snmp-1 mRNA was examined by Northern blot analysis using the snmp-cr probe and by RPA using the snmp-p probe. Northern blot analysis (Fig. 5A) indicates that snmp-1 mRNA is approximately 6 kilobase pairs and is present in both male and female antennae, but it could not be detected in thoracic ganglia, leg, or head mRNAs (Fig. 5C). snmp-1 mRNA appeared much more abundant in male antennae than in female antennae. Furthermore, the snmp-1 probe hybridized to a doublet of mRNA (Fig. 5A), suggesting the presence of multiple forms of snmp-1 mRNA in the antenna.

Tissue localization by RPA (Fig. 5B) showed the 327-base fragment of the snmp-p probe was protected from degradation by male and female antennal mRNAs; probe was not protected by mRNAs derived from head, thoracic ganglia, or leg tissue (data not shown). Protection by male antennal mRNA appeared much greater than by female antennal mRNA. Interestingly, the probe was degraded to a discrete number of products in both male and female antennal incubations in addition to the expected full-length product of probe-minus-vector sequence (327 bases). This complex degradation pattern also suggests that there are multiple snmp-1 mRNA species; partial but consistent degradation of probe would occur if related but different mRNA targets offered incomplete protection of the probe from nuclease treatment. However, this pattern could also result from secondary structural anomalies in a single RNA target. Probe incubated alone, with antennal RNA (data not shown), or with yeast RNA, but not treated with nuclease (Fig. 5B; Probe), did not show the characteristic degradation pattern, indicating that degradation did not result from some activity intrinsic to the incubation conditions. Separate RPAs performed on antennal RNA isolated from 12 wild caught individuals showed a consistent pattern of degradation (data not shown), suggesting that this heterogeneity was not the result of allelic variation among animals, although it could result from the constant presence of two stable alleles among this population.

The developmental time course of snmp-1 mRNA expression during adult development was examined by RPA (Fig. 5D). Animals were staged using morphological characteristics that were easily visualized though the cuticle; these characteristics were initially determined and correlated to day number by charting a group of 20 individual males through their 14 day development. snmp-1 mRNA was first detected at very low levels on day 6, increased gradually to day 11, and then increased dramatically by day 13.5. snmp-1 mRNA continues to be expressed well into the adult stage as evidenced by the cloning of snmp-1 from mRNA derived from wild caught adults and the Northern blot analysis.

The A. polyphemus antenna resembles a feather in form; branches are arrayed along a central stalk, and olfactory sensilla are organized in several parallel rows running the length of opposite sides of each branch (34). In situ hybridization studies showed Snmp-1 mRNA localized to specific cells within the male antenna (Fig. 6, A and B). Hybridization to sectioned branches visualized stained cells arranged on opposite sides of the branch lumen and associated with olfactory sensilla. These cells can be identified as somata of olfactory neurons by their small and round shape, by their depth in the epithelium, and by their positions to the side of their corresponding cuticular hair (35). Hybridization with a non-insect probe under identical conditions showed no staining (Fig. 6C).

Immunocytochemical studies showed Snmp-1 protein to be localized in the somata, dendritic neck, and cilia of the olfactory neurons (Fig. 6, D-G); incubation with preimmune serum under identical conditions showed no staining (Fig. 6H). No staining was observed in the axons or nerve bundles that occupy the lumen of the antennal branches (Fig. 6D). Snmp-1 antigen appeared throughout the somata but concentrated toward the distal (dendritic) side as indicated by the intensity of staining (Fig. 6E). Interestingly, Snmp-1 antisera appeared to recognize only a single olfactory neuron per sensillum, although there are typically two or three olfactory neurons innervating each sensillum (36). In one survey of 22 sections, 217 sensilla were observed attached to antennal branches; 98 sensilla clearly contained only one stained dendritic process entering a cilium (Fig. 6F), and no sensillum contained multiple stained dendritic processes. Similarly, in a survey of well over 100 isolated sensilla, all but one showed a single stained olfactory cilium within the hair shaft (Fig. 6G); one isolated sensillum was observed with what appeared as two stained cilia. It is worth noting that the sensillum diameter is approximately 6 µm, while the sections we made were 10 µm thick; thus, the majority of our sensilla observations were made on whole, unsectioned sensilla and did not miss neuronal processes that might be lost through sectioning.

DISCUSSION

Snmp-1 is a prominent 67-kDa membrane protein from the olfactory cilia of the silk moth A. polyphemus, which appears to be antenna-specific. Histological analysis indicates that snmp-1 mRNA is expressed in antennal olfactory neurons and that Snmp-1 protein is localized in the cilia, dendrites, and soma of these neurons. Developmental studies suggest a low level of snmp-1 mRNA expression beginning at about 45% of development but increasing significantly toward the end of adult development. This correlates with the developmental expression of other antennal proteins that are thought to play direct roles in odor processing such as the M. sexta odorant-binding proteins (OBPs) and odorant-degrading enzymes (Refs. 7, 9, 37, and 38). The expression of OBPs and odorant-degrading enzymes coincides with the onset of odorant-dependent electrical activity in the antenna (37, 39, 40). The neuronal localization of Snmp-1 protein in the antenna and the expression of Snmp-1 mRNA late in adult development coincident with functional maturation of the antenna argue that Snmp-1 plays an important role in silk moth olfaction.

Immunocytological analysis indicated that Snmp-1 antigen was present in only one neuron per sensillum in the male antenna. This is particularly curious because there are typically two or three neurons present in each sensillum, all closely related, having derived from a common sensory mother cell (34). Furthermore, each neuron differs in its response to specific pheromone components (3, 41, 42). The staining of only one neuron per sensillum may indicate that Snmp-1 is a molecular marker of that sibling neuron, thus suggesting some unique role for that neuron. Snmp-1(RP11) may also represent only one of several Snmp-1 homologues, where other Snmp-1 homologues are uniquely expressed in the other sibling neurons. The presence of multiple Snmp-1 homologues was suggested in both our Northern blot and RPA experiments. Northern blot analysis (Fig. 5A) revealed a doublet of similar sized bands, and RPA studies showed a consistent and complex pattern of probe degradation (Fig. 5, B and D). Both results are consistent with the possibility of multiple Snmp-1 forms resulting from homologous genes or alternative splicing of a single gene.

Data base analysis indicates that Snmp-1 is homologous with the CD36 family of membrane proteins, which includes four vertebrate members (CD36, SR-BI, LIMP II, and CLA I), the arthropod emp gene, and the nematode cm08 h8 gene (43). Recently an emp homologue was identified in the Anopheles malaria mosquito (a-emp; GenBank^TM accession number U43499[GenBank]). This homology appears to extend to the three-dimensional structure of these proteins as well. Hydropathy analysis of Snmp-1 and several members of the CD36 family indicate that the proteins contain two prominent hydrophobic regions of substantial length (about 20 amino acids) located in the N and C termini (Fig. 4). This structure has led to suggestions that the N- and C-terminal regions are transmembrane domains (12, 14, 16, 31). The large intervening hydrophilic domain is thought to be extracellular, in part supported by numerous putative N-linked glycosylation sites within this region; one of these sites is conserved in five of the six proteins analyzed including Snmp-1 (Fig. 2). A deletion mutant of CD36 lacking the C-terminal hydrophobic region was shown to bind thrombospondin, suggesting that the hypothesized C-terminal transmembrane domain may not be critical for function, at least for CD36 (44). Hydropathy analysis indicates a unique Snmp-1 hydrophobic region around position 140; the length of this region is relatively small for a putative transmembrane region (about 10 amino acids) but may contribute to the tertiary structure of the presumptive extracellular domain of this protein.

Two mammalian members, CD36 and SR-BI, have been implicated as receptor molecules. Possible ligands for CD36 include thrombospondin, long chain fatty acids, and collagen types I and II; ligands for SR-BI include acetylated low density lipoprotein (13, 33, 45-47). Roles for these proteins as receptors may include mediators of cell-cell interaction or internalization of certain ligands (low density lipoproteins) (33, 48-51). CD36 protein dimerization has been suggested by Trezzini et al. (52) and observed by Huang et al. (53); dimerization could provide sufficient three-dimensional transmembrane structure for effective signal transduction. Immunoprecipitation studies in human platelet cells indicate that CD36 associates in a dimerized state with several classes of tyrosine kinases (53, 54). These speculations derive solely from studies of the vertebrate members; the insect and nematode members remain largely uncharacterized and may possess very different functions. Snmp-1 is the first member of this family identified in nervous tissue and its function may be quite different than those of the currently described nonneuronal homologues.

The antennae of A. polyphemus have a primary function of detecting odors, both sex pheromones and plant volatiles. The assumption that genes uniquely expressed in these antennae have an olfactory role led to the earlier identification of insect OBPs and odorant-degrading enzymes (55). Similarly, Snmp-1 appears to be uniquely expressed in antennal neurons, suggesting that it, too, has an olfactory role. The presence of Snmp-1 mRNA in both male and, to a lesser extent, female antennae suggests that the role of Snmp-1 might not be restricted to the pheromone-sensitive neurons from which it was isolated. However, it is possible that females contain a subpopulation of neurons sensitive to specific pheromone components for the purpose of monitoring their pheromone release and that female Snmp-1 is restricted to such neurons.

Homology between Snmp-1 and CD36 implies that Snmp-1 could have a receptor-like role within the olfactory system. One possible function of Snmp-1 is that of a pheromone receptor protein. Snmp-1 has several features one might expect of a sex pheromone receptor, including antennal specificity, receptor membrane localization, increased expression in the adult, and possible heterogeneity. A previous study using a photoaffinity analog of the A. polyphemus sex pheromone identified a pheromone-binding membrane protein of similar size and tissue distribution (10); Snmp-1 may be that same protein. Several reports suggest that chemoreception in arthropods is mediated through second messenger responses, implicating the presence of a G-protein-coupled receptor system (56-59). Snmp-1 is clearly not homologous with the seven-transmembrane G-protein-coupled receptors identified as olfactory receptors in vertebrates (see Ref. 60) or with the candidate chemosensory receptors identified in the nematode C. elegans (61, 62). Nevertheless, it is possible that Snmp-1 represents a novel class of odor receptor present in insects.

Snmp-1 expression initiates following neurogenesis (34), increasing significantly toward the end of adult development and continuing into the adult stage. Snmp-1 may have more than one role corresponding to the developmental phase of its expression. The early phase of Snmp-1 expression coincides with development and growth of the olfactory neurons and suggests a possible Snmp-1 role in the mediation of cell-cell interactions or cytoskeleton-membrane interactions. A role of Snmp-1 in cell-cell interaction would be consistent with the proposed function of CD36 in mediating interactions between red blood cells and endothelial cells in humans (33, 48, 63). CD36 is a receptor for thrombospondin, an extracellular matrix protein involved in the outgrowth of chick retinal neurons (51, 63, 64). The late phase of increased Snmp-1 expression corresponds with the acquisition of functional maturity of the olfactory neurons and suggests some role for Snmp-1 in odor detection. Other molecules of the moth olfactory neurons show a similar biphasic pattern of expression: the appearance of acetylcholine and the expression of the acetylcholine-synthesizing enzyme choline acetyltransferase and the acetylcholine-degrading enzyme acetylcholine esterase initiate early around the time of neurogenesis and continue into the adult when they function in neurotransmission (65). Such early expression of neurotransmitters has been proposed to play significant roles in neuronal development (66-68).

An intriguing possible function of Snmp-1 may involve the OBPs that are also expressed in olfactory sensilla. The extracellular OBPs are small, water-soluble proteins thought to transport hydrophobic odors through the aqueous fluid to receptor proteins in the membranes of olfactory cilia (6, 8, 9, 40, 55, 69, 70). Snmp-1 may interact with odor-OBP complexes, stabilizing them at the membrane surface and thus enhancing the delivery of odor molecules to nearby receptor proteins. Such protein-protein interaction would be consistent with properties of the several CD36 family proteins, such as the interactions of low density lipoprotein with CD36 and SR-BI. While the specific function of Snmp-1 has yet to be determined, the observations made in this study suggest strongly that it has an important role in olfactory function and odor detection. Further characterization of Snmp-1 and isolation of additional Snmp-1 homologues in A. polyphemus as well as other moth species will add insight into this expanded neuronal role of CD36-type proteins.