CD36 homolog divergence is responsible for the selectivity of carotenoid species migration to the silk gland of the silkworm Bombyx mori.

Dietary carotenoids are absorbed in the intestine and delivered to various tissues by circulating lipoproteins; however, the mechanism underlying selective delivery of different carotenoid species to individual tissues remains elusive. The products of the Yellow cocoon (C) gene and the Flesh (F) gene of the silkworm Bombyx mori determine the selectivity for transport of lutein and β-carotene, respectively, to the silk gland. We previously showed that the C gene encodes Cameo2, a CD36 family member, which is thought to function as a transmembrane lipoprotein receptor. Here, we elucidated the molecular identity of the F gene product by positional cloning, as SCRB15, a paralog of Cameo2 with 26% amino acid identity. In the F mutant, SCRB15 mRNA structure was severely disrupted, due to a 1.4 kb genomic insertion in a coding exon. Transgenic expression of SCRB15 in the middle silk gland using the binary GAL4-UAS expression system enhanced selective β-carotene uptake by the middle silk gland, while transgenic expression of Cameo2 enhanced selective lutein uptake under the same GAL4 driver. Our findings indicate that divergence of genes in the CD36 family determines the selectivity of carotenoid species uptake by silk gland tissue and that CD36-homologous proteins can discriminate among carotenoid species.

fl esh), containing mainly ␤ -carotene ([+ C , F ] in Fig. 1B, C ). Conversely, the middle silk gland of the F mutant, homozygous for the recessive + F allele, has a defect in the cellular uptake of ␤ -carotene, resulting in the formation of light yellow cocoons containing mainly lutein ([ C , + F ] in Fig. 1B, C ). The double mutant of the C and F genes, which is homozygous for both the + C allele and the + F allele, produces white cocoons containing limited amounts of lutein and ␤ -carotene ([+ C , + F ] in Fig. 1B, C ). Identifi cation and comparison of the molecular identities of the products of the C gene and the F gene is therefore expected to provide molecular genetic insights into the selectivity of transport of carotenoid species to target tissues by circulating lipoproteins.
Carotenoids are absorbed in the intestine of the caterpillar, transferred to the hemolymph lipoprotein, lipophorin, and ultimately, accumulate in the middle silk gland, resulting in the formation of golden yellow cocoons containing both lutein and ␤ -carotene (10)(11)(12). By studying spontaneous mutants obtained over the 4,000-year history of sericulture ( 12,13 ), it was revealed that lipophorin-mediated lutein and ␤ -carotene transport to the middle silk gland are controlled by the Yellow cocoon ( C ) gene ( 14,15 ) and the Flesh ( F ) gene ( 16 ), respectively. The middle silk gland of the C mutant, homozygous for the recessive + C allele, has a defect in the cellular uptake of lutein, resulting in the formation of cocoons of creamier yellow (also called , + F ] genotypes are the same as those shown in a previous report ( 17 ). Scale bar: 1 cm.
Analysis of 789 BF1 individuals by SNP genotyping indicated that 9 individuals had been misjudged by their cocoon colors; 8 BF1 individuals whose cocoons were judged to be white were in fact heterozygous for the SNP markers F-090703-03, F-100222-3-3, 06-013, F-100222-7-2, F-100222-8-1, and 06-011; and 1 BF1 individual whose cocoon had been judged to be fl esh-colored was in fact homozygous for these markers (supplementary Table II). We therefore omitted these 9 individuals from Fig. 2 for the sake of simplicity. Even if these 9 individuals were taken into consideration, logarithm 10 of the odds (LOD) scores for the SNP markers Bm_scaf111-1, F-100222-3-3, 06-013, and Bm_scaf111-4 were calculated to exceed 200, strongly supporting the linkage between these SNP markers and the F locus.
To determine the sequence of BAC clone 055_A23, derived from the genomic DNA of strain p50, shotgun sequencing data obtained with a Roche 454 DNA sequencer by Hokkaido System Science Ltd. (Sapporo, Hokkaido, Japan) were combined with sequence data from the silkworm genome database KAIKObase (URL: sgp.dna.affrc.go.jp/KAIKObase/) ( 20 ).

Quantifi cation of transcripts by quantitative PCR
Total RNA was isolated from tissues as described previously ( 17 ). Single-stranded cDNAs from various tissue samples were synthesized from total RNAs with Superscript III reverse transcriptase (Invitrogen, Carlsbad, CA) and oligo-dT primer and were then treated with RNaseH (Takara, Kyoto, Japan). Quantification of transcripts was carried out by quantitative PCR using the cDNAs as templates with LightCycler FastStartDNA Master-PLUS SYBR Green I reagent (Roche, Mannheim, Germany) and a LightCycler DX400 thermocycler (Roche).
Serial dilutions of plasmids containing the cDNA sequences were used as copy-number standards. Transcript levels of the genes other than rpL3 were normalized to the level of the rpL3 transcript in the same samples, as described previously ( 21 ).
For SCRB15 , the cDNA sequence containing the full-length coding sequence from the F allele was determined from strain c44 ( F / F ). First, a partial SCRB15 cDNA fragment was amplifi ed The C gene product was previously identifi ed as Cameo2 , a transmembrane protein-encoding gene belonging to the CD36 family, which is found in species ranging from mammals to insects ( 17 ). Some CD36 family proteins, such as the mammalian SR-BI and the fruit fl y NinaD, have been implicated in cellular uptake of carotenoids (3)(4)(5)(6)(7). SR-BI also functions as a lipoprotein receptor and facilitates cellular uptake of cholesteryl ester from high-density lipoproteins ( 18 ).
In this study, we obtained genetic evidence that the F gene encodes SCRB15, another member of the CD36 family.

Silkworms
Nontransgenic silkworm strains used in this study were preserved in the silkworm stock center at Kyushu University, Fukuoka, Japan, and the Genetic Resources Conservation Research Unit of the National Institute of Agrobiological Sciences, Ibaraki, Japan. The larvae were reared on mulberry leaves. The transgenic strains were produced and maintained in the Transgenic Silkworm Research Unit, National Institute of Agrobiological Sciences, Ibaraki, Japan. Larvae were reared on an artifi cial diet derived from mulberry leaves until the fourth instar, and on mulberry leaves during the fi fth instar. The fi rst days corresponding to the developmental stages of the fourth to fi fth larval ecdysis, and wandering, a characteristic behavior with enhanced locomotory activity just prior to the commencement of the spinning of cocoons, were designated as V0 and W0, respectively. In this article, we designate the strain w06 with an F /+ F genotype as "w06F" for clarity.

Mapping of the F gene using single nucleotide polymorphisms
In a previous crossing, during positional cloning of the C gene ( 17 ), we used strain c11 that is homozygous for the dominant C allele (wild-type allele) of the C gene and strain number 925 that is homozygous for the recessive + C allele (mutant allele) of the C gene. As strain c11 and number 925 were also homozygous for the dominant F allele (wild-type allele) of the F gene and the recessive + F allele (mutant allele) of the F gene, respectively, the BF1 progeny produced cocoons in which the appearance was divided into the following four colors according to the combination of the alleles of the C and the F genes: i ) deep yellow, derived from the genotype [ C , F ] ( C/+ C , F/+ F ); ii ) light yellow, derived from the genotype [ C , + After the visual assessment of the cocoon colors, genomic DNA was extracted from pupae in cocoons and used to map the C gene. In the BF1 individuals, there were no discrepancies between the single nucleotide polymorphism (SNP) marker pattern and cocoon color phenotype for the C gene. Subsequently, we used the BF1 progeny with the fl esh-colored [+ C , F ] and white [+ C , + F ] cocoons to map the F gene. We did not use [ C , F ] and [ C , + F ] because rigorous discrimination between individuals with these genotypes was diffi cult.
For F gene mapping, 16 SNP markers previously reported on chromosome 6, and 11 new SNP markers were used. The PCR primers for the SNP markers are listed in supplementary Table I or were published previously ( 19 ). PCR products were treated with ExoSAP-It (USB Corp., Cleveland, OH) and subjected to direct sequencing. sequence from strain e09 (+ F /+ F ) was identical to that from strain w06 (+ F /+ F ).

Northern blotting analysis
A 32 P-labeled riboprobe was synthesized from a plasmid containing both the 3 ′ part (621 bp) of the coding sequence and the 5 ′ part (39 bp) of the 3 ′ -UTR of SCRB15 . No repetitive sequence was found in the insert of this plasmid. Total RNA was electrophoresed on 1% agarose gels containing formaldehyde and was then transferred onto Hybond N + membrane (GE Healthcare UK, Buckinghamshire, England). Hybridization was performed with Ultrahyb (Ambion, Austin, TX).

SCRB15 genomic sequence analysis
As a part of the SCRB15 genomic sequence, including exons 4-6, was absent from the publicly available silkworm genome sequence data ( 22 ), we determined the sequence of genomic fosmid clone RO0299-D04, derived from strain p50 by a shotgun method ( 23,24 ), to reveal the overall genomic structure of SCRB15 .
To compare the genomic sequence of exon 6 between the F and the + F alleles, the genomic sequence of each of strains c44 ( F / F ), w06 (+ from the middle silk gland via RT-PCR with the primer pair Primer15-1 (5 ′ -GATAAGAACGTACACTATCGCATGG-3 ′ ) and Primer15-2 (5 ′ -TCGAACAGTTTTGGATCAGCATCC-3 ′ ), both of which were designed based on the predicted SCRB15 fragment sequence (BGIBMGA013438 in the China gene model in the KAIKObase). The sequence of amplifi ed fragments was determined by direct sequencing. Subsequently, four more primers, were designed based on the determined partial sequence.
The 5 ′ -and 3 ′ -ends of the SCRB15 sequence were obtained via rapid amplifi cation of cDNA ends (RACE) using a SMART RACE cDNA amplifi cation kit (Clontech, Mountain View, CA) with Primer15-10 for the fi rst 5 ′ -RACE product, Primer15-8 for the nested 5 ′ -RACE product, Primer15-11 for the fi rst 3 ′ -RACE product, and Primer15-13 for the nested 3 ′ -RACE product. The determined sequence was then combined to obtain the full-length SCRB15 cDNA sequence from strain c44 ( F / F ).
SCRB15 cDNA sequences from strain w06 (+ F /+ F ) were amplifi ed from the middle silk gland via RT-PCR with two primer pairs, Primer15-5 (5 ′ -GATATATGAATTT TCGCAAAA G AGA C T A AG-3 ′ ) for the 5 ′ -UTR and Primer15-12 (5 ′ -TCAATG CAGTC AAT C-C TTAC-3 ′ ) for the 3 ′ -UTR, and Primer15-23 (5 ′ -ATAGGG AA-GC CGGAGTAATGAGGTA-3 ′ ) and Primer15-22 (5 ′ -CAGA CTC-AGGGTGGAGTCGAGTAT-3 ′ ) for the coding exon, and directly sequenced. The SCRB15 cDNA sequence from strain w06 (+ F /+ F ) was identical to that from strain c44 ( F / F ) except for a splicing out of exon 6 or a 1.4 kb insertion into exon 6 (for details, see Results). The SCRB15 cDNA sequence from strain e09 (+ F /+ F ) was amplified via RT-PCR with the primer pair Primer15-5 and Primer15-12, and directly sequenced. The SCRB15 cDNA  ( 20 ). Sequence of BAC clone 055_A23 was determined in this study to fi ll the gap between Bm_scaf111 and Bm_scaf78. No sequences homologous to Cameo2 were found in 055_A23. SCRB11-15 were named in previous reports ( 17,49 ).  Schematic SCRB15 mRNA structure from strain c44, which is homozygous for the F allele. Exon boundaries were determined based on the genomic sequence database (for exons 1-3 and 7-11) ( 20 ) and our own genomic sequence data (for exons for 30 min. A cocoon (6-14 mg) was cut into small pieces of less than 1 mm width using scissors; these pieces were then transferred into a glass tube with 2.0 ml of DMSO, mixed, and sonicated at 50-60°C for 30 min. After collection of the extracts, residual cocoon pieces were extracted twice more with 2.0 ml of DMSO, and the resulting three extracts were pooled. After filtration through a polyvinylidene difluoride membrane, the extracted carotenoid composition was analyzed by highperformance liquid chromatography (HPLC). A reverse-phase column (DOCOSIL SP100 [6 × 250 mm]; Senshu Kagaku, Tokyo, Japan) was used under the following conditions: mobile phase, isocratic solvent, 70% methanol, and 30% ethyl acetate; room temperature; fl ow rate, 2 ml/min; and detection, 445 nm. Carotenoid standards were purchased from Sigma (St. Louis, MO).

Data deposition
The cDNA sequence of SCRB11 from c44 ( F / F ), and of that from w06 (+ F /+ F ), the cDNA sequence of SCRB12 from c44 ( F / F ), and of that from w06 (+ F /+ F ), the cDNA sequence of SCRB15 from strain c44 ( F / F ), the sequence of the insertion in exon 6 of SCRB15 from strain w06 (+

Mapping of the F gene on chromosome 6 of the silkworm
To specify a candidate genomic region for the F gene, we performed genetic linkage analysis using SNP markers ( 19 ) and the B. mori genome sequence ( 22 ). Using 58 BF1 individuals, the F locus was roughly mapped on chromosome 6, on which the F gene lies ( 30,31 ). The F -linked region was narrowed to an end of chromosome 6, bounded by the SNP marker 06-011 ( Fig. 2A ).
For Southern blotting analysis, a digoxigenin-labeled DNA probe for SCRB15 exon 6 was amplifi ed from a cDNA-containing vector by PCR using the primer pair Primer15-25 and Prim-er15-20. The hybridization and detection procedures with this probe were described previously ( 25 ).

Silkworm transgenesis
Transgenic SCRB15 expression using the binary GAL4/upstream-activating sequence (UAS) system ( 26 ) was performed as described previously ( 17 ). To construct the effector vector, pBacMCS[UAS-SCRB15-3xP3-EGFP], SCRB15 was amplifi ed by RT-PCR from the middle silk gland of strain c44 ( F / F ) with Primer15-17 (5 ′ -ATGCACTAGTTTTTCGCAAAAGAGACTAAG-3 ′ ) for the 5 ′ -UTR and Primer15-14 (5 ′ -ATGCACTA GTCAATGC-AGTCAATCCTTACA-3 ′ ) for the 3 ′ -UTR, both of which contain an Spe I site. The amplifi ed fragment was digested with Spe I and ligated into a pBacMCS[UAS-3xP3-EGFP] vector ( 27 ), which was previously digested with Bln I. For the effector strains, the effector construct and the helper plasmid pHA3PIG ( 28 ) were injected into preblastoderm embryos of the strain w1-pnd-925, a nondiapausing strain with the phenotype of yellow hemolymph and white cocoons ( 17 ). The existence of the transgene was identifi ed by the fl uorescence in the eye, EGFP for UAS-SCRB15 and DsRed2 for Ser1-GAL4 ( 29 ). Individuals exhibiting colorless hemolymph in the larval stage, which had colorless silk glands and produced white cocoons, were not used to examine the color phenotype or carotenoid composition of the middle silk gland and cocoons.

Analysis of carotenoid composition
Larval hemolymph was collected into an Eppendorf tube, frozen in liquid nitrogen, and stored at -70°C until use. Thawed hemolymph was centrifuged at 800 g for 5 min to remove hemocytes. Twenty microliters of the supernatant was transferred into a glass tube with 3.0 ml of dimethyl sulfoxide (DMSO), mixed, and sonicated at 50-60°C for 30 min. The middle silk gland was frozen in liquid nitrogen and broken into fi ne pieces; some of these pieces (20- No Eco RI or Nde I recognition sites were found within the insertion. Predicted Xba I fragment sizes were 11.6 kb for the F allele and 7.7 kb (weaker) and 5.3 kb (stronger) for the + F allele. Predicted Eco RI fragment sizes were 10.9 kb and 12.3 kb for the F and + F alleles, respectively. Predicted Nde I fragment sizes are 6.4 kb and 7.8 kb for the F and + F alleles, respectively. (I) Genotyping of SCRB15 by genomic PCR in multiple strains of B. mori and multiple individuals of B. mandarina , the putative wild ancestor of B. mori ( 33 ). Details of the geographical location of the sampling area of B. mandarina are as described previously ( 50 ). The different band from strain c11 obtained with Primer-25 and Primer-24 was an artifact unrelated to SCRB15 that was amplifi ed with Primer-24 as both sense and antisense primers (data not shown).  ( 36 ), are indicated by asterisks and bold type, respectively. (B) A neighbor-joining tree for SCRB15 and other homologs from insects and mammals generated using MEGA5 software ( 51 ). SCRB15 and Cameo2 are highlighted. The fi rst two characters of the gene names represent their species: Bm , B. mori ; conserved and did not show insertions, deletions, or premature stop codons, but they demonstrated four nonsynonymous substitutions in SCRB11 {i.e., Y19N [from tyrosine at amino acid position 19 in strain c44 ( F / F ) to asparagine in strain w06 (+ F /+ F )], I116N, D305E, and V454I}. In contrast, RT-PCR revealed that the SCRB15 coding sequence lengths were signifi cantly different between these chromosomes ( Fig. 4A , B ). Sequencing of the smaller PCR product from strain w06 (+ F /+ F ) ( Fig. 4B , arrow) revealed the absence of 201 base pairs in the coding sequence ( Fig.  4C ). These 201 base pairs perfectly corresponded to SCRB15 exon 6. Sequencing of the larger PCR product ( Fig. 4B , arrowhead) identifi ed an insertion of 1,400 base pairs in exon 6 ( Fig. 4D ). Although the insertion was not homologous to known transposable elements ( 32 ), a duplication of eight nucleotides was found at the insertion site (highlighted in Fig. 4D ), which could be a sign of an insertion event. Northern blotting analysis ( Fig. 4E ) supported the structural difference in SCRB15 between the chromosome bearing the F allele and that bearing the + F allele and revealed a decrease in SCRB15 expression in strains with the + F allele, consistent with the results of quantitative RT-PCR analysis ( Fig. 3D ).
The genomic sequence of SCRB15 was also compared between the F and + F allele ( Fig. 4F-H ); this revealed that the same insertion occurred in the genomic sequence of exon 6 of the + F allele. Thus, likely due to this genomic insertion, SCRB15 exon 6 was either spliced out ( Fig. 4C ) or transcribed with the insertion ( Fig. 4D ), resulting in the absence of 67 amino acids or an unusual stop codon contained in the insertion sequence, respectively, in the F mutant.
Genotyping of SCRB15 in multiple strains of B. mori and multiple individuals of B. mandarina , the putative wild ancestor of B. mori ( 33 ), indicated that the insertion occurred in every + F allele-bearing individual from different strains of B. mori , but it was absent in B. mandarina , suggesting a common and single origin of the + F allele ( Fig. 4I ). This origin could be associated with the previous categorization of SCRB15 (termed BGIBMGA013438) into the 354 protein-coding genes of the silkworm that represent good candidates for artifi cial selection during silkworm domestication, identifi ed from a whole-genome analysis of approximately 16 million SNPs ( 34 ).

Bioinformatic characterization of the SCRB15 sequence
SCRB15 was predicted to encode a 57.6 kDa protein composed of 504 amino acids; glycosylation at fi ve N-glycosylation consensus sites was expected to increase the molecular weight of the protein ( Fig. 5A ). Its amino acid identity with Cameo2 was 26%. Identical amino acid residues between SCRB15 and Cameo2 were dispersed over the entire Next, novel primer sets were designed and fi ne mapping was performed with 789 BF1 individuals. During the fi ne mapping, a BAC genomic DNA clone, 055_A23, which bridges the gap between the two scaffolds, Bm_scaf111 and the Bm_scaf78, was sequenced since the F -linked region appeared to occur in this area. Consequently, the F -linked region was narrowed down to a 499 kb region between two SNP markers, Bm_scaf111-1 and F-100222-3-3, both of which were on the same scaffold, Bm_scaf111 ( Fig. 2B  and supplementary Table III). Twenty-six genes were predicted to lie within the narrowed region of genomic DNA, based on the China gene model of the silkworm genome database, KAIKObase ( 20 ) ( Fig. 2C ). Among these genes, four (i.e., SCRB14 , SCRB11 , SCRB15 , and SCRB12 ) belong to the CD36 gene family; that is, they are paralogs of Cameo2 . The function of the F gene is similar to that of the C gene, despite its unique carotenoid species selectivity. Therefore, we considered these four genes to be possible F gene candidates. No other Cameo2 paralogs, other than SCRB13 , were found within the genomic sequence of chromosome 6, including the BAC clone 055_A23 ( Fig. 2C ).

Expression analysis of Cameo2 paralogs near the F locus
Cellular uptake of ␤ -carotene in the middle silk gland, associated with the F gene, occurs in a spatiotemporally controlled manner ( 10 ) ( Fig. 3A , B ). To examine its relationship to ␤ -carotene uptake, we examined the expression profi les of the F gene candidates with quantitative RT-PCR. SCRB13 , which lies outside of the F -linked region, was also included in the analysis for comparison. SCRB11 , SCRB15 , and SCRB12 exhibited a middle silk gland-specifi c expression pattern in fi nal instar larvae possessing the F allele, the dominant wildtype allele of the F gene ( Fig. 3C ). Expression levels of these three genes in the middle silk gland of the strain with the F allele were higher in the mid-stage of the fi nal instar when ␤ -carotene was extensively acquired by the middle silk gland ( Fig. 3D ). In the posterior portion of the middle silk gland, within which ␤ -carotene mainly accumulates, the highest expression of these three genes was noted ( Fig. 3E ). In contrast, SCRB14 and SCRB13 expression levels were not specifi c to the middle silk gland ( Fig. 3C ) and were not well correlated with developmental changes in ␤ -carotene uptake by the middle silk gland ( Fig. 3D ).

Genomic insertion in a coding exon severely affects the SCRB15 mRNA structure in the F mutant
Next, the coding sequences of SCRB11 , SCRB15, and SCRB12 were compared between the chromosome containing the F allele and that containing the + F allele by RT-PCR and sequencing. SCRB11 and SCRB12 were well Dm , D. melanogaster ; Ag , Anopheles gambiae ; Dp , Danaus plexippus ; Hs , Homo sapiens ; and Mm , Mus musculus . Numbers at nodes indicate the percentage of bootstrap values of 10,000 replicates. Groups are based on previous reports ( 39,40 ). Although 21 homologs were found in the gene database of Danaus plexippus ( 38 ), only 11 of those more than 368 amino acids long were included in this tree; the other 10 homologs were shorter than 288 amino acids. DPGLEAN04114 of Danaus plexippus (accession number: EHJ78189) is composed of 1,801 amino acids, while the other CD36 homologs are approximately 500 amino acids. DPGLEAN04114 consists of an aldehyde oxidase-homologous sequence at the N-terminus and a CD36-homologous sequence, closely related to SNMP1, at the C terminus. Both aldehyde oxidase ( 52 ) and SNMP1 (53)(54)(55) are implicated in pheromone detection in insect olfactory neurons. We next focused on the region of pigmentation in the middle silk gland of the last larval instar. As mentioned before, pigmentation in the larvae bearing the F allele commences in the posterior part and spreads into the middle of the gland ( Figs. 3A and 6C ) ( 10 ), which likely refl ects the expression pattern of SCRB15 ( Fig. 3E ) and migration of liquid silk toward the anterior part of the silk gland. In contrast, pigmentation of the transgenic larvae, in which the SCRB15 expression was driven by Ser1-GAL4 , commenced from the middle part and spread into the posterior part of the gland ( Fig. 6D ). This pattern was concordant with the property of the Ser1-GAL4 driver as monitored using the UAS-EGFP gene ( 29 ). On the other hand, the pigmentation pattern in larvae bearing the C allele commences from the middle part of the gland but spreads into the posterior region only minimally ( Fig. 6E ) ( 10 ), which likely refl ects the middle-specifi c expression pattern of Cameo2 ( 17 ). In the transgenic larvae, in which Cameo2 expression was driven by Ser1-GAL4 , pigmentation commenced from the middle and spread into the posterior part ( Fig. 6F ) ( 17 ), which is similar to what was observed in the SCRB15 transgenic larvae but not larvae bearing the F or C allele. Overall, the pigmentation seemed to occur where SCRB15 or Cameo2 were expressed. These data suggest that carotenoid uptake by SCRB15 or Cameo2 do not require cofactors of which the expression is restricted to limited regions in the middle and posterior parts of the middle silk gland. It should be noted that the carotenoidbinding protein (CBP), which is an obligate cofactor of the products of both the C gene and the F gene for carotenoid transport into the middle silk gland ( 10,12 ), is expressed from the middle to the posterior part of the middle silk gland ( 17,41 ).
Carotenoid content of larval hemolymph, middle silk gland, and cocoons of the transgenic lines were analyzed by HPLC with a reverse-phase column ( Fig. 6G-J ). A decrease in ␤ -carotene, but not lutein content, was observed in hemolymph of transgenic larvae expressing SCRB15 ( Fig. 6H ). This may be due to selective uptake of ␤ -carotene by the middle silk gland ( Fig. 6I ). Cocoons produced by the transgenic larvae expressing SCRB15 selectively accumulated ␤ -carotene ( Fig. 6J ). The ␤ -carotene quantities did not reach the same level as in a strain bearing the F allele, which is consistent with observations regarding the color intensity of the cocoons ( Fig. 6B ).
sequence, but there were fewer in the central section (residues 215-289 in SCRB15). TMHMM software ( 35 ) analysis suggested that SCRB15 comprises a large extracellular loop, anchored to the plasma membrane on each side by transmembrane domains adjacent to short cytoplasmic N-terminal and C-terminal domains. The putative molecular mass, transmembrane topology, and multiple glycosylation sites of SCRB15 are common to members of the CD36 family ( 36 ). The theoretical isoelectric point deduced from the primary amino acid sequences was somewhat different for the predicted extracellular loop of SCRB15 (pH 7.8) and that of Cameo2 (pH 5.0). While our previous analysis using the SignalP 3.0 program suggested that the fi rst N-terminal helices of some of CD36 family proteins, including Cameo2, represented a signal peptide ( 17 ), analysis using the SignalP 4.0 program ( 37 ), a new software version designed to discriminate between signal peptides and transmembrane regions, suggested that the fi rst N-terminal transmembrane helices of SCRB15 and Cameo2 are anchored to the membrane and are not signal peptides.
A phylogenetic tree of CD36 protein family members from mammals and insects, including the monarch butterfl y Danaus plexippus , whose whole genome was recently reported ( 38 ), was constructed based on their primary amino acid sequences ( Fig. 5B ). As shown in previous studies ( 39,40 ), the insect proteins could be divided into three groups, while mammalian proteins formed a single, distinct group. SCRB15 and Cameo2 were somewhat removed from each other and fell into groups 1 and 2 , respectively. SCRB11-15 formed a Lepidoptera-specifi c clade in group 1, suggesting that a small expansion may have occurred in the Lepidopteran lineage.

Enhancement of ␤ -carotene selective uptake into the middle silk gland by transgenic expression of SCRB15
To verify the function of SCRB15 as a product of the F gene, we examined the restoration of ␤ -carotene accumulation in the middle silk gland after transgenic expression of the SCRB15 gene in a strain with the phenotype of the F mutant, using the binary GAL4/UAS system ( 26 ). Following transgenic expression of SCRB15 in the middle silk gland by the Ser1-GAL4 driver ( 29 ), pigmentation was observed in the middle silk gland ( Fig. 6A ). Cocoon pigmentation was also restored, although the color was not as intense as those of some nontransgenic strains ( Fig. 6B ). . The stage of the hemolymph and the middle silk gland was at one or two days prior to W0, except for strains c10 (W0), c43 (W0), and w06 (W0). Statistical signifi cance (* P < 0.05; ** P < 0.01; *** P < 0.00005) was analyzed using Student t -test. The cocoon color of strain w06 (+ C /+ C , + F /+ F ) is shown in Fig. 1C . The late and bigger silk gland images in (C) and (E) are the same as those in Fig. 1C . Scale bar: 1 cm.
From the enhancement of selective ␤ -carotene uptake into the middle silk gland and cocoons, we conclude that the F allele encodes SCRB15.

DISCUSSION
Regardless of the essential roles of carotenoids in animals, handling of carotenoids in the body requires special processes due to its water insolubility. The molecular events involved in the transport of any carotenoid from the gut lumen to the gut epithelial cells, to the blood, and to its target cells are not yet understood. In B. mori , several mutants with altered cocoon colors are defective in one of the steps involved in the transport of carotenoids from the midgut lumen to its target tissue (the middle silk gland). We have been characterizing carotenoid transport at the molecular level using the B. mori cocoon color mutants ( 12,17 ).
In this study, we identifi ed the F gene responsible for control of the selective cellular uptake of ␤ -carotene as SCRB15 , following the identifi cation of the C gene that controls selective cellular uptake of lutein ( 17 ). A homology search revealed that SCRB15 belongs to the CD36 gene family, which includes Cameo2 , encoded by the C gene. Functional differentiation after gene duplication therefore likely facilitated differences in selectivity for carotenoid species. To date, it has been shown that members of the CD36 family are critical for a large variety of biological activities in species from mammals to insects: these activities include angiogenesis, viral and bacterial recognition by the immune system, transport of lipids from cholesteryl ester to fatty acids, and taste and olfactory perception ( 42 ). Our study indicates that this family is also critical for discrimination of chemical structures that differ only marginally, such as carotenoids in the silkworm ( Fig. 1 ).
A recent report indicated that the mammalian CD36 was involved in both lycopene and lutein uptake by adipocytes ( 43 ). CD36 homologs may not always discriminate among carotenoid species strictly.
Carotenoids are carried in the hemolymph by lipophorin, a high-density lipoprotein in insects, which contains both ␤ -carotene and lutein ( 11 ). As lipophorin contains both carotenoids in its hydrophobic core, transfer of ␤ -carotene or lutein from lipophorin to the middle silk gland occurs at the interface between hemolymph and cell membrane of the middle silk gland. Consistent with the emerging view, SCRB15 and Cameo2 are transmembrane proteins expressed in the middle silk gland. We expect that SCRB15 and Cameo2 function as noninternalizing lipophorin receptors that facilitate selective uptake of carotenoids, as shown for SR-BI for cellular uptake of cholesteryl ester from high-density lipoproteins ( 18 ).
Expression profi les of SCRB11 and SCRB12 were similar to that of SCRB15 ( Fig. 3C ). Although the SCRB15 mutation is apparently a main cause of ␤ -carotene defi ciency of the F mutant, it could be possible that SCRB11 and SCRB12 are involved in the cellular uptake of ␤ -carotene. Furthermore, this study does not exclude the possibility that the four nonsynonymous substitutions in SCRB11, observed between strain c44 ( F / F ) and strain w06 (+ F /+ F ), were a cause of the modest rescue in ␤ -carotene accumulation by transgenic expression of SCRB15 ( Fig. 6 ).
␤ -carotene is absorbed into the midgut and transferred into the hemolymph irrespective of which allele of the F gene is present. Considering the genomic disruption of a coding exon in the + F allele ( Fig. 4 ) and the middle silk gland-specifi c expression pattern ( Fig. 3C ), SCRB15 may not be essential for absorption of ␤ -carotene in the midgut. Thus, the ␤ -carotene transport system in the midgut is likely to be different from that in the middle silk gland.
Selective transport of lipids other than carotenoids also occurs in insects. For example, diacylglycerol is transported by lipophorin to the fl ight muscle to provide an energy source for fl ight ( 44 ). As CD36 family members are engaged in cellular uptake of lipids other than carotenoids in mammals, such as fatty acids or cholesteryl ester ( 42 ), it is interesting to consider the possibility that these molecules in silkworm are also involved in the transport of lipids other than carotenoids. Identifi cation of SCRB15 as the F gene casts a spotlight on this possibility for group 1 ( Fig. 5B ), since the function of the group 1 insect proteins remains largely unknown. It should be noted that although the SCRB13 expression was barely detected in the tissues used in this study ( Fig. 3C-E ), an SCRB13 EST clone (accession number: AK385087) was found in the corpora allata in the larvae, where juvenile hormone, a sesquiterpenoid hormone that is critical for insect development, is synthesized and secreted ( 45 ).
The molecular mechanism underlying the specifi c carotenoid selectivity of SCRB15 and Cameo2 is a subject for further study. While it is unknown whether the interaction between carotenoids and these proteins is direct or indirect (i.e., needing unidentifi ed cofactors), the coding sequence of SCRB15 should contain key residues facilitating the selectivity for ␤ -carotene that would not be present in Cameo2, as the selectivity for carotenoids was observed by transgenic expression using the same Ser1-GAL4 driver ( Figs. 6G-J ). Because production of transgenic larvae is not suitable for high-throughput studies, in vitro reconstitution of selective carotenoid uptake by SCRB15 and Cameo2, such as has been achieved for cellular cholesteryl ester uptake by SR-BI (46)(47)(48), is required to examine the selectivity mechanism by point mutagenesis or analysis of chimeric genes of SCRB15 and Cameo2.