|Home | About | Journals | Submit | Contact Us | Français|
Little is known about how individual olfactory receptor neurons (ORNs) select, from among many odor receptor genes, which genes to express. Acj6 (Abnormal chemosensory jump 6) is a POU-domain transcription factor essential for the specification of ORN identity and Or (odor receptor) gene expression in the Drosophila maxillary palp, one of the two adult olfactory organs. However, the mechanism by which Acj6 functions in this process has not been investigated. Here we systematically examine the role of Acj6 in the maxillary palp and in a major subset of antennal ORNs. We define an Acj6 binding site by a reiterative in vitro selection process. The site is found upstream of Or genes regulated by Acj6, and Acj6 binds to the site in Or promoters. Mutational analysis shows that the site is essential for Or regulation in vivo. Surprisingly, a novel ORN class in acj6 adults is found to arise from ectopic expression of a larval Or gene, which is repressed in wild-type via an Acj6 binding site. Thus Acj6 acts directly in the process of receptor gene choice; it plays a dual role, positive and negative, in the logic of the process, and acts in partitioning the larval and adult receptor repertoires.
Animals possess many odor receptor genes, and many functionally distinct olfactory receptor neurons (ORNs). Yet in animals ranging from mammals to fruit flies, individual ORNs express only one or a few odor receptor genes (Ache and Young, 2005). How do ORNs select which receptors to express?
Drosophila contains two olfactory organs, the antenna and the maxillary palp. Each contains sensilla that house the dendrites of ORNs. The most numerous antennal sensilla are the basiconic sensilla, which subdivide into large and small subtypes (Shanbhag et al., 1999). Physiological analysis of the antenna and maxillary palp has identified 35 functional ORN classes, housed in 17 sensillum types in stereotyped combinations (Clyne et al., 1997; de Bruyne et al., 1999, 2001; Elmore et al., 2003). The Drosophila larva contains a distinct olfactory organ with 21 ORNs (Stocker, 2008).
The fly genome contains 60 Or (Odorant receptor) genes (Clyne et al., 1999a; Vosshall et al., 1999). Receptor-to-neuron maps were established by physiological and molecular analysis (Goldman et al., 2005; Hallem et al., 2004; Couto et al., 2005; Fishilevich and Vosshall, 2005). The maxillary palp contains three types of sensilla, pb1, pb2, and pb3, each of which contains two ORNs, i.e. pb1A, pb1B, pb2A, pb2B, pb3A, and pb3B (Figure 1A). A single receptor gene has been mapped to each of these six ORN classes, except that two have been mapped to pb2A (Goldman et al., 2005). The larval olfactory organ expresses a subset of Or genes that overlaps partially with the adult Or repertoire.
The first gene shown to be required for proper Or expression was acj6 (abnormal chemosensory jump 6), identified in a behavioral screen (McKenna et al., 1989). In acj66, the maxillary palp does not contain ORNs with the response spectra of pb1A, pb2A, pb2B, or pb3A; by contrast, pb1B and pb3B appear normal (Clyne et al., 1999b). Surprisingly, a novel ORN class, pb2C, was observed. Further analysis showed that acj6 encodes a POU domain transcription factor (Clyne et al., 1999b), and that some Or genes depend on acj6 for their expression (Clyne et al., 1999a; Goldman et al., 2005). These results did not reveal whether Acj6 acts directly on Or genes or indirectly.
Here we extend our analysis of Acj6 regulation to include all Or genes of the maxillary palp and all ORNs of the large antennal basiconic sensilla. Through an in vitro selection procedure we define an Acj6 binding site, find that it lies upstream of Acj6-sensitive Or genes, and determine by mutational analysis that it is required for Acj6-mediated regulation. pb2C is found to derive its response spectrum from the ectopic expression of a larval Or gene, revealing an unexpected role for Acj6 in repression of larval odor receptors in the adult. The negative regulation is mediated by an Acj6 binding site. Thus Acj6 exerts a dual role in receptor gene choice, positive and negative, adding a degree of freedom to the remarkable process by which each of 60 odor receptors is constrained to a discrete subset of neurons.
Drosophila stocks were cultured at 25°C. Wild-type flies were Canton-S unless otherwise stated. All DNA constructs were sequenced and then injected into w1118 embryos. More than three transgenic lines were obtained and analyzed for each transgene. Each GFP reporter line was doubly heterozygous for UAS-mCD8-GFP unless otherwise indicated. In all experiments with acj6, the acj66 allele, a null allele, was used.
The GST-Acj6 expression construct was a gift from Dr. S. Certel and contains the coding region (nucleotides 131-1577) of acj6 isoform (1, 4) (Certel et al., 2000). GST::Acj6 fusion protein expression was induced in the E. coli strain BL21 at room temperature, and the fusion protein was purified using glutathione-sepharose 4B following the manufacturer’s instructions (GE Life Sciences).
SELEX was performed using methods previously described (Wey and Schafer, 1996) with some modifications (Supplementary Fig. S1). Double-stranded oligonucleotides were generated by annealing the template random oligonucleotides CAGGTCAGTTCAGCGGATCCTGTCG(N)16 GAGGCGAATTCAGTGCAACTGCAGC, to a 25-base pair primer complementary to the 3′ invariant sequence (GCTGCAGTTGCACTGAATTCGCCTC) and then filling in with Klenow DNA polymerase. Unannealed oligonucleotides were removed using MicroSpin™ G-25 columns (Amersham).
For the initial round of selection, 2 ng random oligonucleotides were combined with 0.2 ug GST::Acj6 fusion protein in a 50 ul binding system that contained the following components: 0.5% Nonidet P-40, 20 mM Tris/HCl (pH 8.0), 100 mM NaCl, 1 mM EDTA, 4 ug poly (dI-dC). The binding reaction proceeded at 4°C for 30 min with gentle rocking. Protein-oligonucleotide complexes were precipitated with 10 ul glutathione-sepharose beads pre-washed and equilibrated in the binding buffer with 0.5% (w/v) skim milk. After three washes with the binding buffer, the beads were resuspended in 100 ul elution buffer (10 mM Tris·Cl, pH 8.5), and heated at 85°C for 10 min to release bound oligonucleotides, a fraction (10 ul) of which was then amplified by PCR for 20 cycles. The PCR product was column-purified and a fraction (2 ng) was used for the next round of selection. Control binding reactions without GST::Acj6 protein were performed with each selection to ensure that nonspecific binding of oligonucleotides to the sepharose beads did not yield a visible band after PCR amplification. For the fifth and sixth rounds, oligonucleotides were incubated with GST::Acj6 protein bound to glutathione-sepharose beads; ~ 1ug of Acj6 protein was used. After six rounds of selection and amplification, PCR products were cloned and 26 clones were randomly selected and sequenced.
The GST moiety of the GST-Acj6 fusion protein was cleaved with PreScission™ protease (GE) in a buffer containing 50 mM Tris·Cl (pH 8.0), 150 mM NaCl, 1mM EDTA, 1mM DTT, and 0.01% NP-40 “substitute” (Sigma-Aldrich) according to the manufacturer’s instructions. 5′-end biotin-labeled complementary oligonucleotides and non-labeled competitor sequences (W. M. Keck facility at Yale) were annealed in a buffer containing 10 mM Tris·Cl (pH 8.0), 1 mM EDTA, and 50 mM NaCl. The annealed double-stranded oligonucleotide probes were purified from a native polyacrylamide gel and run through a Microspin G-50 column (GE). The sequence of the Or85e probe and its cold competitor was 5′-TTTTAAAAACTCATTAATGCATAAAATCAT-3′; the Or85e mutated competitor sequence was 5′-TTTTAAAAACTTCTAGAAAGTTAAAATCAT-3′ (mutated nucleotides are underlined). Binding sites in other Or genes were also tested in competition assays. The Or46a sequence was 5′-ATCGTCGGCTAAATAATTCAGTATTTTTGC-3′; the mutant Or46a sequence was 5′-ATCGTCGGCATCTCTAGACCGTATTTTTGC-3′. The Or45b sequence was 5′-AACTAGTACGCATACTTGATCATATATATA-3′; the mutant Or45b sequence was 5′-AACTAGTAGCTGCCGGATCCCATATATATA-3′; the “switched binding site” sequence Or45bMOr46a was 5′-AACTAGTACTAAATAATTCAGATATATATA-3′. The amounts of protein, probes, and competing sequences are indicated in the figure legend. Protein was diluted in 2.5 mM Hepes (pH 7.6) buffer containing 0.1M KCl, 1.25 mM MgCl2, 0.01 mM EDTA, 1% glycerol, and 1 mg/ml BSA, when needed. The binding reactions included 20 mM Tris·Cl, 100 mM KCl, 2 mM DTT, 1 ug poly (dG-dC) (Sigma), 5% glycerol, 0.05% NP-40, 1.25 mM MgCl2, and 1% (w/v) skim milk and were incubated at room temperature for 15 minutes. The resulting complexes were then resolved on a 5% native polyacrylamide gel in 0.5 X TBE, 2.5% glycerol at 4°C and detected with Pierce Chemiluminescent nucleic acid detection module according to the manufacturer’s protocol.
Transgenic flies carrying wild-type Or85e-GAL4 (3.1 kb of upstream sequences), Or42a-GAL4 (4.1 kb), and Or46a-GAL4 (1.9 kb) were described previously (Goldman et al., 2005). For Or85eM-GAL4, the Or85e promoter regions upstream and downstream of an Acj6 site (at −85 bp) were cloned separately with primers containing an XbaI site in place of the Acj6 site. The upstream KpnI-XbaI fragment was ligated with the downstream XbaI-NotI fragment and then inserted into pG4PN (Dobritsa et al., 2003), resulting in an 8-nucleotide substitution within the Acj6 binding site (aaaaaCTCATTAATGCAtaaaa was mutated to aaaaaCTTCTAGAAAGTtaaaa). Similar strategies were used to construct Or46aM-GAL4 (tcggcTAAATAATTCAGgtatt was mutated to tcggcATCTCTAGACCGgtatt) and Or42aM310-GAL4 (gaatcATCATTTATTAAtggaa was mutated to gaatcGGATCCTTATAAtggaa). Overlap extension PCR (Higuchi et al., 1988) was used to introduce additional mutations in the Acj6 sites at −249 bp and −102 bp of Or42a. In Or42aM310-249-GAL4, (gggctTTCATTCTTTAGgtcat) at −249 bp was mutated to (gggctTTCACCGTTTAGgtcat). In Or42aM310-102-GAL4, (cgcctATGAATTATCTAattga) at −102 bp was mutated to (cgcctATGACCGATCTAattga). Or42aM310-249-102-GAL4 bears mutations in all three Acj6 sites.
Wild type Or45b-GAL4 was described by Kreher et al., 2005. Overlap extension PCR was used to mutate the Acj6 site at −392 bp (tagtaCGCATACTTGATcatat) to (tagtaGCTGCCGGATCCcatat) in Or45b (M)-GAL4. In Or45b (MOr46a)-GAL4, this site was replaced by (tagtacTAAATAATTCAGatat), the Acj6 site in the Or46a promoter.
Digoxigenin-labeled RNA probes were hybridized to whole maxillary palps attached to the proboscis, or to antennal sections, and were detected with alkaline phosphatase conjugated anti-digoxigenin antibody and its substrates BCIP/NBT or Fast Red tablets (Roche) as previously described (Clyne et al., 1999a; Goldman et al., 2005).
We extended our analysis of Or genes in an acj66 background to include the entire repertoire of maxillary palp receptor genes. Of the seven genes, five depend on acj6, as determined by in situ hybridization: their expression was not detected in the acj66 maxillary palp (Figure 1B). The genes that depend on Acj6 include the gene that is normally expressed in the pb1A ORN (Or42a), both of the genes that are expressed in pb2A (Or33c and Or85e), the gene expressed in pb2B (Or46a) and the gene expressed in pb3A (Or59c). Or71a and Or85d appear to be expressed normally in acj66. Results for some of these genes have been reported earlier (Clyne et al., 1999a; Goldman et al., 2005).
To identify an Acj6 binding site, we used an in vitro selection procedure, SELEX (systematic evolution of ligands by exponential enrichment; Supplementary Fig. S1). In essence, a pool of random oligonucleotides was allowed to bind to Acj6 protein, and oligonucleotides that bound were eluted, amplified by PCR, and allowed to bind again. Six cycles of enrichment were thus carried out for oligonucleotides that bind Acj6, after which oligonucleotides from the enriched population were sequenced. A consensus binding site was defined from the sequenced oligonucleotides, and a positional weight matrix was established (Figure 2A, Supplementary Fig. S2).
We scanned the DNA sequences upstream of each of the seven maxillary palp Or genes for Acj6 binding sites. We focused on the 500 nucleotides upstream of each translation start site, because regions of this length were sufficient to confer faithful expression of a reporter gene in the case of previously analyzed maxillary palp Or genes (Ray et al., 2007).
Of the seven Or genes, five contained upstream Acj6 sites, and those were the five genes that depend on Acj6 (Figure 2B and Supplementary Fig. S3). Specifically, all of these five genes contained Acj6 sites with scores greater than 7 (scores are defined in Supplementary Fig. S2). If a score of 6 is set as a threshold, then the other two genes have no sites in this region. Moreover, the two genes that did not depend on Acj6 had no sites in a more extensive upstream region extending 2 kb from the translation start site.
We tested whether Acj6 binds to the predicted Acj6 sites in Or promoters using electrophoretic mobility shift assays (EMSAs). We first tested an Acj6 site located 85 bp upstream of the translational start site of Or85e, and found that Acj6 protein, expressed in bacteria, indeed bound to it (Figure 2C, lanes 1–4). The amount of binding depended on the amount of Acj6 added to the binding reaction. At the highest levels of Acj6, two intense bands could be observed, of which the lower and higher molecular weight bands are likely to represent binding of the site to an Acj6 monomer and an Acj6 dimer, respectively. Highly cooperative homodimerization has previously been reported for POU domain proteins (Rhee et al., 1998).
To test whether binding of Acj6 to the Or85e site is sequence-specific, we asked whether various sequences could compete with the site for binding to Acj6. We found that an excess of the unlabeled Or85e site, or Acj6 sites from two other genes, Or46a and Or45b (described below) competed effectively; however, these sites did not compete when mutated (Figure 2C). The simplest interpretation of these results is that Acj6 binds directly to the Acj6 site in a sequence-specific fashion.
The correlation between the presence of Acj6 sites and Acj6-dependence suggested a role for the sites in Or gene regulation. To test the possibility of a functional role for these sites, we mutated them upstream of three genes, Or85e, Or46a, and Or42a. We examined the effects of these mutations on the expression of Or promoter-GAL4 drivers that have previously been shown to drive faithful expression of a UAS-GFP reporter gene (Goldman et al., 2005).
Upstream of Or85e we mutated the site at −85 bp (Figure 3A). We note that positions are determined with respect to translational start sites, but the transcriptional start sites as determined by 5′RACE are within 50 bp of the predicted translational start sites in all cases examined, including Or85e and Or46a (Ray et al., 2007). We found that mutation of this −85 bp site in the Or85e-GAL4 driver abolished expression, as determined by loss of GFP expression (Figure 3A) (≥20 maxillary palps examined for each of four independent transgenic mutant lines).
Upstream of Or46a lie two Acj6 sites that overlap. We mutated both simultaneously and saw no remaining expression of the GFP reporter in most maxillary palps (Figure 3B). A small fraction (~20%) of maxillary palps showed staining of a few cells (n=2.3±0.7 cells, n= 13 maxillary palps), a much smaller number than is observed when a wild-type promoter is used to drive expression (n=15.1±0.6 cells, n=8 maxillary palps).
Upstream of Or42a lie three Acj6 sites, at positions −310, −249, and −102. When we mutated the site at −310, there was a small reduction (p<0.0001; ANOVA) in the number of GFP-labeled cells in the maxillary palp compared to the control (Figure 3C). When we mutated both the site at −310 and the site at −102, there was a major reduction in labeling. Mutation of the sites at both −310 and −249 also led to a strong reduction in labeling. Mutation of all three sites virtually eliminated labeling.
These results show that Acj6 sites are essential for the normal expression of each of three Or genes tested. The results also show that multiple Acj6 sites are functional upstream of Or42a.
Although the ORNs of the maxillary palp are particularly convenient to study, the great majority of ORNs in the fly’s olfactory system are located in the antenna. We wanted to examine the extent to which acj6 is required for the specification of ORN identity and receptor gene expression in the antenna.
We carried out an analysis of ORN responses in the antenna of acj66. We systematically measured the responses of all ORNs in the large basiconic sensilla using a panel of diagnostic odorants. There are three types of large basiconic sensilla: ab1 contains four ORNs, ab1A, ab1B, ab1C, and ab1D; ab2 contains two ORNS, ab2A and ab2B; ab3 contains two ORNs, ab3A and ab3B.
In acj66, ab1A and ab1B were severely affected (Figure 4A). We found no ORNs with the response spectra of these ORN classes. We did, however, observe sensilla containing ORNs with response spectra very similar to those of ab1C and ab1D: one ORN responded only to CO2, as expected of ab1C, and another responded only to methyl salicylate, as expected of ab1D (de Bruyne et al., 2001). In these sensilla we did not observe the spontaneous action potentials of large amplitudes that are characteristic of ab1A and ab1B, suggesting that these ORNs are either absent or electrically silent.
We observed a second type of sensillum in the acj66 antenna that contained an ORN with a response spectrum very similar to that of ab2A (Figure 4A). In this sensillum type we observed a second ORN that exhibited spontaneous action potentials but that did not respond to any tested odorant. Likewise, we observed a third type of sensillum containing an ORN similar to ab3A, paired with an ORN that showed spontaneous action potentials but no odor-evoked activity. We note that among the four classes of odor-responding ORNs in the acj66 antenna, some yielded responses of smaller magnitude than in the wild-type control, but the overall response profiles of these four ORN classes appeared very similar to those of wild-type in each case.
The simplest interpretation of these results is that ab1A, ab1B, ab2B, and ab3B all depend critically on acj6 for the acquisition of their odor-sensitivities (Figure 4B). It is possible that acj6 plays a subtle role in the differentiation of the other ORN classes, but acj6 does not seem to be required by these other ORNs in the process by which they select a receptor gene.
We next tested the expression of Or genes in the acj66 antenna (Figure 5A and data not shown). We found by in situ hybridization that the Or genes that are expressed in the ab1A, ab1B, ab2B and ab3B ORNs, Or42b, Or92a, Or85a, and Or85b, respectively, were not detectable in acj66. As a control we examined the wild-type antenna and detected expression of all four genes. As an additional control we tested the expression of Or59b and Or22a, which are expressed in ab2A and ab3A, respectively. Both of these Or genes were found to be expressed in both acj66 and wild-type. In summary, the lack of detectable odor responses from four ORN classes of the acj66 antenna correlates with a lack of expression of the receptor genes that these ORNs express in wild-type.
Having identified four antennal Or genes that depend on Acj6 for their expression, we asked whether they contained Acj6 sites. All four contain Acj6 sites within 1 kb of the predicted ATG translation start site (Figure 5B); all of the Acj6 sites are within or very near the 500 bp region upstream of this ATG. We also considered the receptor genes of ORNs that are present and show approximately normal response spectra in acj66 (ab1C, ab1D, ab2A, and ab3A). ab1C expresses two Gr (Gustatory receptor) genes in wild-type, Gr21a and Gr63a (Jones et al., 2007; Kwon et al., 2007), neither of which contains an Acj6 site within 1 kb of the translation start site. ab1D and ab3A express Or10a and Or22a, respectively, which do not contain Acj6 sites. ab2A expresses Or59b, which contains an Acj6 site despite its independence of Acj6. In summary, in large basiconic sensilla all Acj6-dependent receptor genes contain Acj6 sites and most Acj6-independent receptor genes lack them.
We previously reported the surprising finding of a novel ORN class in the acj66 maxillary palp (Clyne et al., 1999b). Sensilla containing either pb2A or pb2B were absent in the mutant, but sensilla containing a novel ORN class, pb2C, were detected. The response spectrum of pb2C could not be explained in terms of a linear amplification or summation of the response spectra of other maxillary palp ORN classes, and the simplest interpretation was that either pb2A or pb2B was transformed into pb2C.
We recently noticed that one receptor, Or45b, which had been characterized during the course of a systematic functional analysis of larval odor receptors (Kreher et al., 2005, 2008), exhibited a response spectrum similar to that observed for pb2C. We therefore compared directly the responses of pb2C to the responses conferred by Or45b in the “empty neuron” system, an in vivo expression system (Dobritsa et al., 2003; Hallem et al., 2004). We found the odor response profiles to be very similar (Figure 6A), suggesting that Or45b, a larval Or gene, might be ectopically expressed in the maxillary palp of acj66.
We carried out an in situ hybridization with an Or45b probe to the maxillary palp and found expression in acj66, but not wild-type (Figure 6B). The simplest interpretation of these results is that Acj6 represses Or45b in the adult maxillary palp.
To investigate the mechanism of Acj6-mediated repression, we first examined the upstream DNA sequences of Or45b and found two Acj6 sites, at −499 and −392 (Figure 6C, top left). The gene also contains two other regulatory sequences that are found upstream of Or genes expressed in the maxillary palp: there are two Dyad-1 elements, CTA(N)9TAA, at −301 and −55, and an Oligo-1 element, CTTATAA, at −193. Dyad-1 is required for Or expression in the maxillary palp, and Oligo-1 both enhances expression of Or genes in the maxillary palp and represses their expression in the antenna (Ray et al., 2007).
The Acj6 site at −392 is in fact capable of binding Acj6: it competed with the positively acting site in Or85e for binding to Acj6 in gel shift analysis (Figure 2C). To test the hypothesis that Acj6 represses Or45b in the maxillary palp via the Acj6 site, we mutated the site at −392. (Numerous efforts to ablate the site at −499 were unsuccessful.) The loss of the site at −392 resulted in expression in maxillary palps in a wild-type (acj6+) background, consistent with a role for this site in repression of Or45b [Figure 6C; Or45b (M)]. Thus loss of either Acj6 or an Acj6 site allowed expression of Or45b in the maxillary palp. We conclude that Acj6 acts negatively on Or45b via the Acj6 site at −392.
Finally, since other Acj6 sites act positively, we wondered whether the negative effect of the Acj6 site at −392 upstream of Or45b was due to its context, e.g. on other proteins that are bound in the vicinity. To test this possibility, we replaced the Acj6 site at −392 with the Acj6 site that acts positively at −346 upstream of Or46a [Figure 6C, Or45b (MOr46a)]. We found that the Acj6 site that promotes expression upstream of Or46a did not promote expression upstream of Or45b in the maxillary palp of wild-type. The lack of Or45b expression depended on repression by Acj6: Or45b was expressed in the maxillary palp of acj66 animals carrying the Or46a site upstream of Or45b (Figure 6C). One possible interpretation of these results is that Acj6 binds to the Or46a site when it is upstream of either Or46a or Or45b, but the mode of regulation that it exerts depends on other proteins that are bound in the vicinity (Figure 6D).
How individual ORNs select individual odor receptors to express is a remarkable problem in molecular neurobiology. Most ORN classes in the fly select one or a small number of receptor genes from a large family of 60 Or genes. The choice dictates the odor response spectrum of the ORN and, consequently, its contribution to the coding of olfactory information.
The simplest model for receptor gene choice in the fly is that it is governed by a combinatorial code of transcription factors (Ray et al., 2007; Ray et al., 2008; reviewed by Fuss and Ray, 2009). Although there is evidence to support this model, much remains to be learned about the elements of the code and the logic by which it operates. Distinct combinations of six transcription factors could in principle dictate the specification of 26=64 distinguishable ORN classes, a number comparable to the number of Or genes and of the same order as the number of ORN classes. Thus for factors A–F, one ORN class might require factors A, B, and D and be specified by the combination (A+B+C−D+E−F−), while another class might be specified by the combination (A−B−C+D+E−F+). In this simple model, each factor would be required for the specification of half of the 64 ORN classes. Alternatively, a much larger number of such factors could act, each in a small number of classes. In the 14 ORN classes for which data are now available, six in the maxillary palp and eight in the antenna, Acj6 is required for the specification of eight, approximately half. Thus we have found evidence that at least one transcription factor acts broadly across the ORN repertoire.
One aspect of the problem of receptor gene choice is that each ORN must express one or a small number of receptor genes; another aspect is that each ORN must not express all the other receptor genes. In principle, each aspect of the problem could be governed by separate sets of factors. We have found here that Acj6 plays a role in both positive and negative regulation of Or genes. This duality of Acj6 function imparts economy to the process, an economy that may be critical in a process that controls such a large number of Or genes in a large number of ORN classes. Allowing individual factors to act in two aspects of the problem reduces the number of required factors.
Another example of dual regulation by a single transcription factor was found in the differentiation of photoreceptor subtypes in the Drosophila visual system (Tahayato et al., 2003). Otd, a homeodomain transcription factor, activates the rhodopsin genes rh3 and rh5 but represses rh6 in other photoreceptors, through a common binding site. The modes of regulation are likely to depend on the availability of a coactivator and a corepressor of Otd. An example of dual regulation in a chemosensory system is provided in C. elegans by the UNC-3 Olf/EBF protein (Kim et al., 2005). This protein represses a number of genes, including the odr-10 olfactory receptor gene, in the ASI chemosensory neuron, and it activates other chemoreceptor genes. UNC-3 was shown to repress multiple genes by direct binding to sites in the promoters, but it may activate the expression of ASI-specific genes through an indirect mechanism.
Drosophila contains two distinct olfactory systems, one in the larva and one in the adult. The two systems differ in morphology and developmental origins, and they operate under different conditions: the larva burrows in semi-solid medium whereas the adult walks and flies through air. Both olfactory systems, however, depend on members of the Or gene family. Some members of the Or family are larval-specific, others are adult-specific, and some are expressed in both (Kreher et al., 2005; Fishilevich et al., 2005; Couto et al., 2005). By what mechanism are members of the same gene family partitioned into two different systems?
We have found that Acj6 acts in the partitioning mechanism. It is required to restrict at least one member, Or45b, to the larval system: when acj6 is mutated, Or45b is ectopically expressed in the adult. Or45b contains in its upstream region two regulatory elements, Dyad-1 and Oligo-1, which are found adjacent to all or most maxillary palp Or genes and are required for their expression in the maxillary palp. We do not know if the Dyad-1 and Oligo-1 elements play a role in the expression of Or45b in larvae. These observations, taken together, suggest the possibility that Or45b may have shifted over evolutionary time from a maxillary palp gene to a larval gene. If so, the shift is likely to have occurred before the divergence of D. melanogaster and D. pseudoobscura tens of millions of years ago, since the odor response spectra of the maxillary palp ORNs in these two species are similar (Ray et al., 2008).
Using an in vitro DNA binding assay we defined a positional weight matrix and a consensus Acj6 binding site. Only one Acj6 site has been experimentally identified and tested previously, in a study of sequences upstream of the gene encoding choline acetyltransferase, Cha (Lee and Salvaterra, 2002). These Cha sequences were scanned for motifs similar to those found for the mammalian POU-IV transcription factors, and of seven identified, one was found to bind Acj6 protein in an EMSA assay. It does not show close similarity to our site (matrix score=3.4) and its activity in vivo was not evaluated.
We have found a strong correlation between the presence of the Acj6 site that we have identified and Acj6-dependence in vivo. Moreover, we have shown directly, in vivo, that these sites are required for normal expression of flanking Or genes. These results are consistent with a direct role for Acj6 in mediating Or expression, as opposed to an indirect role via another transcription factor.
We have shown that one of the Acj6 sites, the Or45b site, acts in negative regulation. Its score based on the positional weight matrix is intermediate among the sites that act in positive regulation, which is consistent with our finding from a binding-site-swap experiment that repression of Or45b depends not on the specific sequence of an Acj6 site but on its flanking sequences--the Or46a site also mediated repression when placed upstream of Or45b.
In several cases Acj6 sites lie within 8 bp of conserved sequence elements that are required for appropriate Or expression (Ray et al., 2007; Ray et al., 2008, Supplementary Fig. S4). An Acj6 site also lies very near a putative binding site for Pdm3, another POU domain transcription factor that regulates ORN development and that interacts genetically with Acj6 (Tichy et al., 2008). The proximity of these regulatory elements is likely to reflect the interaction of Acj6 and its cofactors. Identification of the cofactors of Acj6 will further contribute to our understanding of the combinatorial regulation of specific Or expression.
Acj6 is likely to have a number of interesting transcriptional regulatory targets in addition to odor receptors and choline acetyltransferase. It may regulate other molecules that are essential for neuronal signaling: while some ORNs in acj66 exhibited spontaneous action potentials but did not respond to odorants, other ORNs showed no electrical activity, consistent with a loss of both odor receptors and other components.
Acj6 is also likely to regulate molecules required for synaptic connectivity. Acj6 is required for axon targeting specificity of a subset of ORN classes, some of which require Acj6 autonomously and some non-autonomously (Komiyama et al., 2004). The second-order neurons of the Drosophila olfactory system, the projection neurons, also require Acj6 for both normal dendritic targeting and axon terminal arborization (Komiyama et al., 2003). Moreover, the role of Acj6 extends beyond the olfactory system: acj6 mutants have defects in retinal axon targeting and synapse selection as well (Certel et al., 2000).
Our definition of an Acj6 binding site may now facilitate the identification of some of its transcriptional targets. There are many sites in the genome with matrix scores >7, and analysis of the adjacent coding regions may be useful in identifying novel components required for synaptic connectivity and other processes.
We thank Sarah Certel for GST-Acj6 fusion protein and its expression constructs, Scott Kreher for UAS-Or45b transgenic flies, W. van der Goes van Naters for help with electrophysiology, Chih-Ying Su for critical reading of the manuscript, and Anandasankar Ray for discussion. This work was supported by grants from the National Institutes of Health Grant and a McKnight Investigator Award to J.R.C.
Publisher's Disclaimer: This PDF receipt will only be used as the basis for generating PubMed Central (PMC) documents. PMC documents will be made available for review after conversion (approx. 2–3 weeks time). Any corrections that need to be made will be done at that time. No materials will be released to PMC without the approval of an author. Only the PMC documents will appear on PubMed Central -- this PDF Receipt will not appear on PubMed Central.