- Email: [email protected]
Cell type-specific transcriptional regulation of the drosophila FMRFamide neuropeptide gene
Neuron,
Vol. 10, 279-291,
February,
1993, Copyright
0 1993 by Cell Press
Cell Type-Specific Transcriptional Regulation of the Drosophila FMRFarnidei Neuropeptide Gene Lynne E. Schneider,* Marie S. Roberts, and Paul H. Taghert Department of Anatomy and Neurobiology Washington University School of Medicine Saint Louis, Missouri 63110
Summary We have used IacZreporter gene constructs to study the promoter/enhancer regions of the Drosophila FMRFamide neuropeptide gene in germ line transformants. FMRFamide is normally expressed in - 60 diverse neurons of the larval CNS that represent - 15 distinct cell types. An 8 kb FMRFamide DNA fragment (including 5 kb of 5’ upstream sequence) was sufficient to direct a pattern of /acZ expression that mimicked nearly all spatial aspects of the normal pattern. This result indicates that the cell-specific regulation of FMRFamide expression is largely generated by transcriptional mechanisms. Reporter gene expression was lost from selected cell types when smaller fragments were tested, suggesting that multiple control regions are included in the FMRFamide promoter. One region (a 300 bp fragment from -476 to -162) acted as an enhancer for 1 of the - 15 FMRFamide-positive cell types, the 012 neurons. These results suggest that, in the mature nervous system, the complex pattern of FMRFamide neuropeptide gene expression derives from the activity of discrete, cell type-specific enhancers that are independently regulated. Introduction In the mature nervous system, many neurons possess cell-specific transmitter phenotypes. The establishment of these neuronal properties during development requires the restriction of gene expression to specific neurons. We are interested in the mechanisms that underlie the developmental restriction of neuropeptide transmitter phenotypes. We are using the Drosophila FMRFamide neuropeptide gene as a model because its expression is restricted to a small though heterogeneous group of neurons. This expression is stereotyped such that individual neurons can be followed throughout development. The organization and expression of the Drosophila FMRFamide gene have been extensively characterized (Nambu et al., 1988; Schneider and Taghert, 1988, 1990; Chin et al., 1990; Schneider et al., 1991). This information provides the basis for an in vivo analysis of neuropeptide expression and function using the molecular genetic techniques that are available in Drosophila.
*Present address: Carnegie Institute of Washington, ment of Embryology, 115 West University Parkway, Maryland 21210.
DepartBaltimore,
The developmental expression of the FMRFamide gene and its single transcript has been characterized using antibodies and in situ hybridization. Expression is restricted to the CNS. In the CNS of the larva, the pattern consists of -60 neurons that are distributed throughout the brain and ventral nerve cord (Schneider et al., 1991). Following metamorphosis, there is an increase in the number of neurons expressing the gene to a total of - 120 (O’Brien et al., 1991; L. Schneider, E. Sun, D. Garland, and P. Taghert, submitted). Based on position and axonal projections, these neurons represent - 15 discrete cell types including large neuroendocrine cells and interneurons associated with the ventral ganglion, central brain, and sensory neuropils. We do not know whether these neurons participate in a common functional circuit or whether they subserve independent roles in the animal. Likewise, very little information is available concerning the roles of Drosophila FMRFamide peptides. Studies in other insects, however, have suggested some possible functions. In locusts, for example, FMRFamide and related peptides modulate skeletal musclecontractions in the leg(EvansandMyers,1986). In theflyCalliphora,some but not all endogenous FMRFamide-like neuropeptides induce fluid secretion from isolated salivary glands (Duve et al., 1992). We presume that the endogenous FMRFamide-like peptides of Drosophila play similar roles. To study the functions of FMRFamide peptides in thesedifferent neuron types, we hopefirst todefinethe molecular mechanismsthat underliecell type-specific expression and then to perturb FMRFamide expression in specific neurons. Previous studies of cell-specific gene expression have demonstrated that transcriptional regulation results from the combination of multiple promoter/enhancer elements. For example, the Drosophila dopa decarboxylase gene (Ddc), which is involved in the synthesis of the transmitters dopamine and serotonin, is expressed in a pattern of -150 neurons (Hirsh, 1989). Promoter deletion and in vitro mutagenesis studies have suggested that the pattern of Ddc gene expression is controlled by multiple cis-acting DNA elements (Scholnick et al., 1983,1986; Beall and Hirsh, 1987; Johnson et al., 1989). Similarly, in the fushi tarazu, even-skipped, and sevenless genes of Drosophila, numerous cis-regulatory elements act in concert to produce complex patterns of expression (Hiromi and Gehring, 1987; Goto et al., 1989; Botwell et al., 1989, 1991). This report describes the identification of c/s regulatory regions of the FMRFamide gene that are involved in the generation of the cell-specific pattern of FMRFamide neuropeptide expression in the mature nervous system. We present evidence for three principal conclusions. First, an 8 kb fragment containing upstream and intragenic regions of the FMRFamide
NellrOll 280
gene contains sufficient regulatory information to direct the appropriate pattern of transcription (and therefore peptide expression) to the normal complement of -15 neuronal cell types. Second, separate regions of this fragment are required for expression by different neuronal cell types. Third, neuropeptide gene expression in a single cell type, optic lobe (OL) visual system neurons, is produced by a small DNA region that displays the properties of a cell typespecific enhancer. Results The Drosophila FMRFamide gene is expressed only in the CNS, and within the larval CNS, expression is restricted to -60 neurons that represent - 15 distinct cell types. To analyze this stereotyped pattern of expression, we designed a study of FMRFamide gene regulation with three specific goals. First, we wanted to determine how much FMRFamide DNA is necessary to produce a wild-type pattern of expression. Second, we wanted to determine the distribution of the regulatory regions that underlie cell-specific FMRFamide expression. Finally, we wanted to determine whether the regulatory regions that are necessary for gene expression in different cell types are shared or distinct. We addressed these questions using three different types of /acZexpression vectors: a translational fusion vector, a transcriptional fusion vector, and an enhancer-detector vector. The first formed fusion FMRFamide-IacZ RNAs and proteins; the second formed fusion FMRFamide-/acZRNAs, but lacked FMRFamide protein sequences; the third did not contain FMRFamide sequences in either the RNAs or proteins formed. These vectors and the FMRFamide genomic regions that we inserted into them are diagrammed in Figure 1. We introduced each of these constructs into the Drosophila genome using P element transformation and, for each, established stable transformant lines (Spradling and Rubin, 1982). For clarity, we have confined our descriptions to the CNS of late larval and adult stages. In addition, for each of the constructs described below, IacZ expression is compared with FMRFamide expression as revealed by in situ hybridization and by an antiserum raised against the FMRFamide precursor (Schneider et al., 1991, submitted; O’Brien et al., 1991). /acZ Expression in the pWF3 Construct Closely Resembled the Endogenous Pattern of FMRFamide Gene Expression In the first /acZconstruct, pWF3, we hoped to include all of the regulatory regions that are necessary for appropriate expression of FMRFamide RNA and protein. Therefore, we began with a large (-8 kb) genomic fragment that contained 5 kb of upstream sequence, exon I (106 bp), the entire intron (- 3 kb), and the first 69 bp of exon II. This fragment was inserted, in frame, into the /acZgene of the translational fusion vector pCaSpeR-Bgal (Figure 1; Thummel et al., 1988).
The presumed translational start site of the FMRFamidetranscript isencoded bythefirstthree nucleotides of exon II (Schneider and Taghert, 1990), thus the fusion protein contains the first 23 amino acids of the FMRFamide precursor. There is only one detectable transcript from the FMRFamide gene (Schneider and Taghert, 1990), and the sequences around the transcription start site are contained, unaltered, within the pWF3 construct. Therefore, we assume that the fusion gene initiated from the same site. In addition, there are no ATG triplets in exon I and only short open reading frames within the intron (Schneider and Taghert, 1990; M. S. R. and P. H.T., unpublished data). Therefore, we assume that translation of the fusion protein initiated at the site that most closely matches the consensus sequence (nucleotides l-3 of exon II). We initially isolated two independent transformants and, subsequently, generated four more by mobilizing one of the original inserts into new genomic sites (Robertson et al., 1988). Due to different fixation conditions required to visualize FMRFamide and B-galalctosidase (B-gal) immunoreactivity, we did not directly compare staining with the two antibodies in the same animals. We therefore assigned neuronal identities and made comparisons, based on the position, size, morphology, and number of immunoreactive neurons in each group. All six of the pWF3 transformants had a similar pattern of /acZ expression, which was nearly indistinguishable from those of FMRFamide RNA and protein. Certain features of the IacZ pattern within the nervous system were possibly ectopic; these are described in a following section. We did not determine the extent of /acZ expression in nonneuronal tissues of transformed animals. The pWF3 pattern included most of the major cell types that express the FMRFamide gene (schematized in Figure 6). The most prominent FMRFamide neurons are the large neuroendocrine neurons (TV and Tva) that are located in each .of the thoracic neuromeres, and these could be identified in both larvae (Figure 2A) and adults (Figure 3A) of pWF3 animals. Similarly, in the suboesophageal neuromeres, the large SE2 interneurons and the ventral Sv cells were B-gal immunoreactive in both larvae (Figure 2A) and adults (see Figure 5A). In the brain, numerous B-gal-positive neurons corresponded to FMRFamide-positive neurons in larvae (Figures 4A and 4B) and adults (Figure 5A). For example, the OL2 pair of neurons lies between the central brain and optic ganglia; in pWF3 lines, these cells expressed /acZ (Figure 5A), as did adjacent smaller neurons (called OL2,), as described below. The OLI and 3, MP2, and Aldm cells were the only major cell types that expressed FMRFamide but not B-gal immunoreactivity in pWF3 lines. Although the overall intensity of B-gal immunoreactivity varied in the six lines, presumably due to the effect of chromosome position on the level of expression, it was typically higher than the endogenous gene. For example, neurons that normally express the FMRFamide gene at low levels (e.g., the Td and A8
Neuropeptide
Gene
Regulation
281
FMRFamide Locus Exon I
Exon II I
1
I
Translational Fusions
pWF6 El
Exon I
Exon II (AA l-23)
Iacz
pWF7 El
Exon I
Exon II (AA 1-2.3)
hi!
pWF4 El
Transcriptional Fusion
Exon I t-4500 bp)
Adh-AUG
\J
f I R
lacz
J
I 1 P SPBR
Enhancer Testing
Figure
1. Diagrammatic
Representation
of the
DNA
Constructs
Used
to Test FMRFamide
Gene
Regulation
At the top of the figure, the positions and sizes of the two exons are indicated. Below the gene schematic, the three translational fusion constructs, one transcriptional fusion construct, and two enhancer-testing constructs are shown in register with respect to FMRFamide gene sequences. The black boxes indicate FMRFamide exons (or portions thereof); solid lines indicate upstream or intronic FMRFamide sequences. Gray boxes indicate the /acZgene sequences. In pWF4, pWF8, and pWF11, Adh-AUG refers to sequences containing the initiator methionine from the alcohol dehydrogenase gene. The /acZ gene in pWF8 and pWFl1 begins with sequences constituting a minimal promoter and cap site from the hsp70 gene. The distances between FMRFamide sequences and the hs43 sequences in the enhancer-detecting constructs are not to scale. All numbers in parentheses refer to base pair sequences with the first nucleotide of exon I called +I. The white gene and the P element terminal sequences that flank all constructs are included but not shown. B, BamHI; C, Clal; E, Spel; C, Bglll; P, Pstl; R, EcoRI; S, Sall; X, Xbal.
were clearly visible in pWF3 flies (Figures 4B and4C). Whilethe higher level of expression probably reflected the greater stability of the IacZ fusion product, the relative levels of IacZand FMRFamide expression among the -15 cell types were the same. For
neurons)
example, theTv, SE2, and OL2 neurons had the strongest staining with both FMRFamide RNA probes, antiFMRFamide antibodies, and anti-IacZ antibodies; other specific groups had correspondingly intermediate or low levels of staining with all reagents.
Neuron 282
Figure
2. Expression
of /acZ in the Larval
Nervous
System
of pWF Transformant
Lines
B-Gal immunoreactivity is shown in whole-mount nervous systems from late third instar larvae: anterior is toward the top, and the ventral aspect of the ventral ganglion is shown in each photograph. The TV and associated smaller neurons (including Tva) are B-gal immunoreactive in pWF3 (A) and pWF7 lines (6); only the TV cell is immunoreactive in pWF4 (C) lines and only the smaller cells in pWF6 lines (D). Staining in SE2 neurons is present only with the largest construct, pWF3 (A), whereas staining in Sv neurons is present with the pWF3 (A), pWF7 (B), and pWF6 (D) constructs. The dark images at the midline of the pWF4 tissue are the thoracic neurohaemal organs (out of focus), in which immunoreactive terminals of the TV neurons are present; see also Figures 4E and 4F. Note the differences in staining intensity; pWF3 lines typically display the strongest levels of B-gal-like immunoreactivity. Abbreviations for cell names are explained in the text. Bar in (C), 100 pm for all panels.
Figure
3. Expression
of /acZ by theTv
and Tva Neurons
in the Ventral
Ganglion
of the Adult
Nervous
System
of pWFTransformant
Lines
Staining in the paired TV neurons is seen in all three thoracic neuromeres (long arrows), whereas paired Tva neurons are typically seen in only the second thoracic neuromere (short arrows). TV and Tva are both B-gal positive in pWF3 (A) and pWF7 (C) lines. Only the TV cell is B-gal positive in pWF4 lines (B), and only the Tva cell is B-gal positive in pWF6 (D). Neither cell type is immunoreactive in pWF8 or pWFl1 (data not shown). Bar in (C), 100 pm for all panels.
Neuropeptide 283
Figure
Gene
Regulation
4. Photomicrographs
Showing
/acZ Expression
in Larval
and Adult
Nervous
System
of pWF Transformant
Lines
(A and B) pWF3, (C and D) pWF6, and (E and F) pWF4. In (A)-(D), anterior is toward the top of the photograph; in (E) and (F) anterior is to the left, and dorsal is toward the top. In (A) and (B), lad expression is shown in two focal planes of the same larval preparation; the same is true for (E) and (F). kc2 expression in pWF3 lines is similar to endogenous FMRFamide expression, as shown in the dorsal region of the brain (A) and the ventral brain and dorsal nerve cord (B). pWF6 /acZ expression also mimics many aspects of FMRFamide expression, shown in theadult ventral nerve cord (C)and dorsal brain (D). In theventral ganglion of pWF4 transformants, ladexpression is restricted to the TV neurons (E), but within these cells it is found throughout the entire cell body, axons, and terminals ([F], within the segmental neurohaemal organs, called transverse nerves FNI-31) to a greater degree than in the other lines. Bar in (E), 150 pm for (A)-(D) and 75 nm for (E) and (F).
The greater stability of the g-gal fusion protein may also explain our observation that P-gal immunoreactivity in adult-specific neurons appeared at an earlier time during adult development than did FMRFamide hybridization signals or immunoreactivity. For example, FMRFamide expression was first detected in the adult-specific Tva and OL2 neurons during the prepupal stages of adult development (O’Brien et al., 1991). In contrast, in pWF3 animals we could first detect
P-gal immunoreactivity larvae (Figure 2A).
in these
neurons
in wandering
/acZ Expression by Different Cell Groups Correlated with Different Regions of the FMRFamide Gene To begin a dissection of the regulatory regions of the FMRFamide gene, we generated three constructs in which we made deletions of the FMRFamide se-
Figure
5. /acZ Expression
in the Adult
Brain
of pWF Transformants
All photographs were taken at approximately the same ventral focal plane. Anterior is and (D) pWFll.The SE2 neurons express lacZonly in the pWF3 lines. TheOL2and smaller the central brain and optic ganglia, are visible in all except the pWF6 lines. In pWFl1 positive; diffuse staining in the overlying tissue is likely an artifact. Note that the lacZ as intense as that produced by the much larger pWF3 translational fusion construct.
quences represented in pWF3. pWF7 and pWF6 are also translational fusion constructs. They contain the exon I, intron, and exon II sequences that are present in pWF3, but pWF7and pWF6 include only922 bp and 162 bp, respectively, of upstream sequence. pWF4 is atranscriptional fusion construct thatwas made using the pCaSpeR-AUG-Bgal vector (Thummel et al., 1988): pWF4contains -4.5 kb of’upstream sequence, a portion of exon I (including the transcription start site), and the translation initiation site from the Drosophila Adh gene. In flies transformed with each of these constructs a portion of the endogenous pattern of FMRFamide expression was lost. To simplify the discussion of these results, we present the changes in expression in three principal FMRFamide neuronal cell types: the Tv/Tva neurons, the SE2 neurons, and the OL2 neurons. The two TV neurons are neuroendocrine cells that are present in each of three thoracic neuromeres. They express the FMRFamide gene throughout the life of the fly, beginning in late embryogenesis. In the adult-specific Tva neuron, which is located adjacent to TV in the second thoracic neuromere, FMRFamide expression is first detected in early pupal stages. Both neurons innervate the segmental neurohaemal or-
toward the top. (A) pWF3, (B) pWF4, (C) pWF6, OL2, neurons (arrowheads), which lie between lines, only the OL2 and OL2, neurons are B-gal expression in the OL2 group of pWFl1 lines is Bar in (C), 100 pm for all panels.
gans. Despite their common positions and axonal targets, laczexpression in thesetwocell types correlated with different genomic fragments. In pWF3 and pWF7 lines (n = 6), TV and Tva were both B-gal immunoreactive (Figures 2A and 2B; Figures 3A and 3C). These constructs include the sequences from base pairs -922 to +69 of exon II. In eight lines transformed with the pWF6 construct (which contains the sequences from base pairs -162 to +69 of exon II), /acZ was expressed in the Tva, but not theTv neurons (Figure 2D; Figure 3D). In contrast, the .pWF4 construct (n = 5: one original and four mobilized transformant lines) produced expression in the TV neurons only (Figure 2C; Figure 3B). This construct contains 4.5 kb of upstream DNA plus 50 bp of exon I. Thus, TV expression required sequences between base pairs -922 and -162, and Tva expression required sequences between base pairs -162 and +69 of exon II (see Figure 7). The analysis of these transformants also permitted the definition of regulatory regions required by the SE2 neurons. These neurons are unpaired midline interneurons of the suboesophageal neuromeres that project widely throughout the brain and ventral ganglion. We detected /acZ expression in this group in
Regulation
the pWF3 lines (Figure 29; Figure 5A), but not in pWF4 (Figure2C; Figure 5B), pWi7(Figure 2B), or pWF6 lines (Figure 2D; Figure 5C). The SE2 neurons were the only major cell group whose expression was specific to pWF3 lines; theonly DNA region that is unique to the pWF3 construct is an -450 bp fragment at the Send of the 5 kb upstream region. Therefore, the presence of this 450 bp region was necessary for expression in the SE2 neurons. By similar analysis, DNA regions required for /acZ expression by the prominent OL2 (and OL2,) neurons of the brain were mapped to the same sequences as appeared necessary for the TV neurons, i.e., sequences between base pairs -922 and -162.OL2 neurons were B-gal positive in lines pWF3 (Figure 2A; Figure 5A), pWF4 (Figure 5B), and pWF7 (data not shown), but not in pWF6 (Figure 5C).
Ectopic Aspects of /acZ Expression in pWF Transformants Some aspects of IacZ expression were ectopic with respect to established patterns of FMRFamide expression (Schneider et al., 1991; O’Brien et al., 1991). First, in pWF3 and pWF6 lines, we often observed a B-galpositive cell (V,) in a position where we had seen FMRFamide expression only in rare cases. Second, in some lines we observed B-gal immunoreactivity in neurons where we had never seen FMRFamide expression. The ectopic neuron that appeared most consistently was the X cell, present in all pWF3 and many pWF6 and pWF7 lines (Figure 6). Other novel cell types included neurons that were slightly stained and that appeared less often. In general, the pWF6 lines displayed the most ectopic expression, although most of these lines also shared the pattern that is shown in Figure 6. For example, four of eight pWF6 lines had /acZ expression in dispersed perineural glial cells of the optic lobes and central brain of adults. This ectopic staining was patterned and reproducible. Another pWF6 line displayed ectopic B-gal staining in clusters of 2-8 cells at thedorsal andventral midlineofthe brain and in small cell clusters throughout the ventral ganglion. Finally, B-gal expression in some cell types was ectopic by virtue of the large numbers of B-gal-immunoreactive neurons. For example, in pWF3 lines there were -60 OL2, neurons in each hemisphere (Figure 5A), whereas with anti-FMRFamide antibodies we had sometimes detected 2-5 small neurons in each hemisphere. We do not know whether these examples of ectopic expression represent anomalous expression due to position effects or previously undetected aspects of FMRFamide gene expression.
A Cell Type-Specific and OL2, Neurons Comparisons transformant the FMRFamide distinct cell
Enhancer
for the OL2
of the /acZ patterns lines suggested that gene are necessary types. To determine
from the different different regions of for expression in whether these re-
gions contain enhancers that are also sufficient to dictate cell type-specific gene expression, we tested two DNA regions for their ability to drive expression from a heterologous promoter; for these constructs, we used the Casper hs43-IacZenhancer-detector vector. The pWF8 enhancer construct contained sequences from +66 bp of exon I to +I248 bp of the intron (Figure 1). The experiments described above implicated this as part of the region necessary for expression by several cell types (see Figure 7). The eight pWF8 lines did not produce any B-gal immunoreactivity in the CNS, and therefore this DNA fragment appears insufficient to regulate FMRFamide expression by itself. The pWF11 enhancer construct contained the FMRFamide gene fragment from base pairs -476 to -162 upstream to the transcription start site. This -300 bp fragment is included in the region that was implicated for expression by the TV, SP4, OL2, and OL2, neurons. The fragment was inserted in an orientation opposite to that found within the FMRFamide gene. Eight of nine pWF11 lines had B-gal immunoreactivity exclusively in the OL2 and OL2, neurons (Figure 5D); the ninth line had no detectable /acZ expression. These results indicate that this DNA fragment possesses the properties of an enhancer (Maniatis et al., 1987): it can influence expression of a reporter gene by a heterologous promoter, and it acts in an orientation-independent manner. The activity of this enhancerwas highly selective: within the nervous system, it regulated gene expression in one cell type. The OL2 and OL2, neurons expressed /acZ at high levels in both pWF3 and pWF11 lines (compare Figures 5A and 5D). In contrast, in the pWF4and pWF7lines (both of which contained more FMRFamide regulatory DNA than did pWF11 lines), IacZexpression in the OL2 neurons was weak (Figure 5B), and only a few OL2, neurons were stained (Figures 5B and 5D). FMRFamide expression,asassayed byantibodystainingand in situ hybridization, is typically strong in OL2 neurons;weak antibody staining is observed in only a few OL2, neurons.
Discussion Previously, we mapped FMRFamide gene expression using both antibody staining and in situ hybridization (Schneider et al., 1991; O’Brien et al., 1991). We have now shown that an 8 kb fragment of the FMRFamide gene (pWF3) is sufficient to direct /acZ expression in vivo with a pattern and intensity that are nearly indistinguishable from that of the endogenous gene. This correspondence suggests that the pattern and level of FMRFamide gene expression are largely determined by transcriptional control mechanisms. The lossof B-gal immunoreactivityfrom specificcell types following transformation with the deletion constructs suggests that control regions within the FMRFamide promoter are distributed throughout the 8 kb of flanking and intragenic sequences (Figure 7). The distributed organization of required elements is con-
I
Anti-proFMRFamide Immunoreactivity
I
l-l-3v
rl
pWF3
Exon I (-so00 bp)
BR
L
PSPBR
n
Exon II (AA l-23) b
E
IacZ J
GC
0
pWF4
a‘ v l-l-3*
Exon I
Ad%-AUG
64500 bp)
lacZ J
Ic I R
Figure
P SPBR
6. Summary
of FMRFamide
and
/acZ
Expression
in Transformant
Lines
Each pair of drawings represents the mature larval and adult CNS, respectively. For transformant lines, the summary drawings are shown on the right, with the DNA constructs on the left. Only those neuron groups that stained consistently are included. The constructs are positioned in register with respect to FMRFamide gene sequences. Data on endogenous proFMRFamide expression (first panel) are taken from a study using enriched setum antibodies that recognize an epitope on the deduced proFMRFamide precursor (Schneider et al., submitted). The nomenclature for naming cell types derives from White et al. (1986) and denotes the relative position of the ceil
Neuropeptide 207
Gene
Regulation
B
III
pWF6
Exon II MA l-23)
Exon I f-162
bp)
1acZ J
4 PBR
E
CC
SPl
I
I'
pWF7
, . Sl-3V Tl-3V
Exon II (AA l-23)
Exon I
PSPBR
E
1acZ
c c
_
. --
. . RX) TVs Td
v
GC
r-l
pWFl1
-476 -162
bp
A&-AUG
bp ““I
body (e.g., SPI, superior TI cells: thoracic ventral, of FMRFamide expression,
1
T
protocerebrum 1; O!&!, optic lobe 2). We add lower case letters to indicate positional detail (e.g., TV, Td, and dorsal, and lateral cells, respectively). The X cell is shown in parentheses because it does not match the map although it was observed consistently in pWF3, pWF6, and pWF7 lines.
Neuron 288
Figure 7. A Schematic That Illustrates Positions of DNA Regions within FMRFamide Promoter That Control Specific /acZ Gene Expression Sufficient
R
No Activity E Exon
I
II
-5000 bp
sistent with previous studies of tissue-specific gene expression, which have suggested that transcription is regulated by multiple modular enhancers (Tamura et al., 1985; Garabedian et al., 1986; Hammer et al., 1987; Grosfeld et al., 1987; Hiromi and Gehring, 1987; Fisher and Maniatis, 1988). There have been comparatively fewer studies of cell type-specific gene expression within a single tissue. This type of regulation is particularly relevant to the nervous system because of the great diversity of neuronal phenotypes. In many cases, gene fusion constructs and deletions within normal regulatory regions have produced novel patterns of reporter gene expression in the nervous system (e.g., Swanson et al., 1985; Low et al., 1986; Kondoh et al., 1987; Hebener et al., 1989; Scholnick et al., 1991). Thus, altering the spatial arrangment of regulatory regions can create ectopic or extremely restricted patterns of gene expression. These results support the hypothesis that restriction of gene expression (and thus the generation of diverse neuronal phenotypes) results from the
Exon rl
II
the the Cell-
The FMRFamide gene is diagrammed in the centerofthefigurewiththetwoexonsindicated as black boxes. Restriction site abbreviations aredefined in the legend to Figure 1. The fragment that directs the full pattern of expression starts at the -5000 BamHl site (B) and ends at the exon II Clal site (C). Below the gene, the dotted lines and corresponding drawings summarize the results of the deletion analysis. These regions are necessary for /acZ expression in the indicated neurons. Above the gene, the results of the enhancer constructs are summarized. The 300 bp fragment lying 162 bp 5 to exon I is sufficient to direct expression of the reporter gene by a heterologous promoter in the OL2 neurons.
combinatorial activity of overlapping sets of enhancer elements. In vivo analyses of promoters for several neuronal genes have begun to define the mechanisms underlying phenotypic complexity. Highly restricted patterns of reporter gene expression have been attained using regulatory fragments of neural-specific genes. For example,Young et al. (1990) reported that oxytocin gene expression was closely mimicked in mice bearing a 5.2 kb DNA fragment from the oxytocinlvasopressin locus. In the case of the sevenless tyrosine kinase gene, Botwell et al. (1991) found that a 475 bp intragenie fragment was sufficient to drive the complex pattern of cell-specific gene expression in the developing retina. The organization of regulatory elements that underlie more widespread patterns of gene expression has also been described. Deletion analysis of the Drosophila choline acetyltransferase promoter revealed the presence of multiple regulatory elements, whose removal led to the absence of reporter gene expression in subsets of the normal pattern (Kita-
Neuropeptide 289
Gene
Regulation
moto et al., 1992). For the Ddc gene of Drosophila, deletion and mutation analyses have suggested that its expression by any single cell type depends on cell type-specific enhancer elements and a common, tissue-specific regulatory element that is required for expression by all Ddc neurons (Bray et al., 1988; Johnson et al., 1989; Hirsh, 1989). Similar to many of these studies, we found that the FMRFamide gene also contains a broadly distributed set of regulatory regions that appears to control different subsets of the cellular pattern. One of these regions, from base pairs -476 to -162, can act as a cell type-specific enhancer for a single neuronal type, the OL2 group. This result extends previous studies of geneexpression in the mature-nervous system: within a large regulatory domain that controls gene expression by many neuronal cell types, we have defined a small region that is sufficient for expression by a single cell type. The 300 bp OL2 enhancer region may contain more than one regulatory element. A sequence comparison between D. melanogaster and D. virilis revealed that this fragment contains four distinct sequences that display at least 80% sequence conservation (Taghert and Schneider, 1990). These could represent four independently regulated regions, or they may represent cis-acting elements that act in some combinatorial fashion. Regardless of its complexity, it is clear that this -300 bp fragment is not required for expression in all FMRFamide-expressing neurons because the construct that did not contain this region (pWF6) still expressed the fusion gene in many FMRFamidepositive neurons (Figure 6). While complex patterns of cell-specific gene expression are commonly thought to reflect the action of overlapping sets of regulatory elements that act combinatorially (e.g., Struhl, 1991), a simpler, noncombinatorial model may be sufficient to explain our results. We propose that the regulatory elements of the FMRFamide gene are autonomous enhancers, many of which are capable of directing expression to individual cell types (see Figure 7). Accordingly, the diverse neurons that share this specific neuropeptide phenotype do not share a common transcriptional control. Instead, each cell type possesses an independent mechanism to regulate FMRFamide gene expression; the parallel regulation of these distinct enhancers in differentcell types results in the full pattern of FMRFamide expression. This model is similar to onesthatdescribetheactivationof the pair-rulegenes hairy and eve in individual stripes across the Drosophila blastoderm (Howard et al., 1988; Harding et al., 1989; Howard and Struhl, 1990; Pankratz et al., 1990). This model has many important implications for the regulation of FMRFamide gene expression. For example, the use of independent enhancers could provide a very high degree of flexibility in the evolutionary modification of gene expression at the level of single cell types. A possible example of such evolutionary modification is illustrated by the TV and Tva neurons.
Although these two cells share a common cell body position, axonal target, and neuroendocrinefunction, different regions of the FMRFamide promoter are required for their respective expression (Figure 3). Furthermore, both the TV and Tva cells are present and morphologically differentiated in larval stages (D. Zitnan and P. Taghert, unpublished data), but only the TV neuron has detectable levels of FMRFamide expression at this time. In contrast, both the TV and Tva neurons in D. virilis are FMRFamide immunoreactive in larval stages (Taghert and Schneider, 1990). This subtle difference between closely related species may result from the modified use of independent, cell type-specific enhancers. There are many current limitations to this model that will require further investigation. First is the assumption that the other FMRFamide regulatory elements behave in a cell-specific manner, as does the -300 bp OL2 enhancer region. To support this conclusion, other gene regions that have been implicated, by deletion analysis, to control cell-specific gene expression (see Figure 7) will have to be tested for their ability to drive a heterologous promoter. Second, this study did not address possible functions of the FMRFamide gene outside of the -8 kb in pWF3. A third limitation of the model is that these data are derived from three different types of vectors. The different vectors could affect /acZ expression due to differences in, for example, posttranscriptional events that are not normal components of FMRFamide gene expression. We think it is unlikely that the use of different vectors substantially affected reporter expression because, in general, our results were consistent between constructs. For example, only those constructs that contained the sequences from base pairs -922 to -162 showed expression in the TV, SP4, and OL cells, regardless of the vector type. This suggests that the major differences between reporter gene expression from the different constructs were due to the presence or absence of specific FMRFamide DNA sequences. While these studies have demonstrated transcriptional control underlying FMRFamide expression in the Drosophila CNS, we have focused only on qualitative, but not quantitative, features of reporter gene expression. In transformants bearing the largest construct examined, the typical levels of /acZ expression were often stronger than in comparable cell types of transformants bearing smaller test fragments (see Figure 3; Figure 5). This suggests that other DNA regions that were not well defined by these studies (but that lie within the 8 kb tested) probably contribute to the general level of transcription in many individual cell types. An analogous conclusion was reached in the study of the Ddc promoter by Johnson et al. (1989). It is notable, however, that the 300 bp fragment tested in pWF11 lines produced /acZ expression in the OL2 group that was typically as strong as that found in animals bearing the largest construct (compare Figures 5A and 5D).
In summary, we have begun to define the organization of DNA regulatory regions that are responsible for the cell-specific pattern of FMRFamide neuropeptide gene expression. The broad spatial distribution of regulatory elements and the autonomy of the OL2 enhancer suggest that distinct enhancer elements direct FMRFamide gene expression to specific neuronal cell types. These results provide a basis for further definition of the enhancer elements within the FMRFamide gene and of the factors that interact with such elements. Experimental
Procedures
Design of DNA Constructs Theconstructsused inthisstudyarediagrammedinFigurel.The pWF3 construct was generated in three steps. First, the genomic clone Mt-8 (Schneider and Taghert, 1990) was digested with Pstl and Clal, and the fragment that begins at base pair -162 upstream of exon I and ends at base pair +69 of exon II was subcloned into the Pstl and Clal sites of the Bluescript vector (Stratagene). Second, this subclonewas digested with BamHl (BarnHI sites were present at +50 bp in exon I and in the polylinker 5’to exon I), and the -5 kb,BamHl fragment from Mt-8 (base pairs -5000 to f50 bp of exon I) was then ligated into this site. A recombinant with the correct orientation was selected by restriction enzyme analysis. Third, the entire insert was removed by Xbal-Clal digestion, Xbal were linkers added, and then ligated into the Xbal site of the pCaSpeR-Bgal vector (Thummel et al., 1988). The pWF6 and pWF7 constructs were generated from the pWF3 construct. pWF3 was digested with Pstl to remove sequences from base pairs -5000 to -162. The pWF6 construct was then created by religating the remaining portion of the construct. The pWF7 construct was generated by ligating the -940 to -162 Pstl fragment back into the,unique Pstl site of pWF6. The orientation of this fragment in the pWF7 construct was determined by restriction enzyme analysis. The construct pWF4 was made by’ligating the 4.5 kb EcoRlBamHl genomic fragment of phage Mt-8 (base pairs -4500 to +50 of exon I) into EcoRI-BamHI-digested pCaSpeR-AUG-Bgal (Thummel et al., 1988). The constructs pWF8 and pWFl1 were generated using the Casper hs43-laGvector (V. Pirrotta). The pWF8 construct contained the EcoRI-,Spel fragment from base pair +70 of exon I to 1250 of the intron. pWFl1 DNA was made by inserting the SallPstl fragment (base pairs -492 to -162) into Casper hs43-/a& as a Sall-Spel fragment; the Spel site was derived from the polylinker of the Bluescript plasmid into which this genomic fragment had previously been subcloned. The pWFl1 insert is in reverse orientation with respect to the /acZ gene. All DNA constructswereconfirmed byrestrictiondigestand partialsequence analysis. Generation of Transformant lines Transformant lines were generated as described by Spradling (1986). The constructs were purified twice on cesium chloride gradients, precipitated in ethanol, and resuspended in 0.1 mM phosphate buffer containing 5 mM KCI. DNAs were injected at concentrations of 300-800 ng/ul for transformation constructs and 100-400 nglul for helper plasmid (~25.7~‘). The host strain was yw67c23. Once established, individual transformant lines were maintained as homozygotes, or over appropriate balancer chromosomes. P-Cal lmmunocytochemistry Central nervous systems were dissected as whole mounts and fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS). Fixation for 2 hr at room temperature or overnight at 4°C gave comparable results. Tissues were rinsed in PBS containing
0.3% TritonX-lOO(PBS/TX)and blocked in PBSiTXcontaininglO% horse serum. The tissues were then stained for one to two nights at 4OC in mouse anti-B-gal (Promega) diluted in PBS/TX containing 3% horse serum and 0.001% sodium azide. The primary antibody was visualized either by using an horseradish peroxidase-conjugated anti-mouse antibody or with the Vector elite kit (Vector Labs). We found that two neuronal cell types of the background strain (~w6’~~~) were strongly stained by anti-B-gal antibodies when the immunohistochemical reactions were processed with Vectastain kit reagents. These neurons happened to be FMRFamide-positive neurons, the SPI and LPI cells of the brain (data not shown). Such endogenous activity was not evident when B-gal-like immunoreactivitywas revealed with a horseradish peroxidase-conjugated secondary antibody. We therefore scored the presence or absence of these two cell types, in various transformant lines, using the simpler staining protocol, and relied on themoresensitiveVectastainmethodforthoseneuronsinwhich such endogenous B-gal-like immunoreactivity was not detectable.
We wish to thank Carl Thummel, Vince Pirrotta, Yash Hiromi, and Steve Crews for sending vectors; John Majors and Jay Hirsh for advice; and John Merlie, Martha O’Brien, Karen O’Malley, Josh Sanes, and Dusan Zitnan for their comments on the manuscript. This work was supported by a grant from the National Institutes of Health (#21749) to P. H. T. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisemenf” in accordance with 18 USC Section 1734 solely to indicate this fact. Received
August
11,1992;
revised
December
7, 1992.
References Beall, C. J., and Hirsh, J. A. (1987). glial expression of the Drosophila Genes Dev. 7, 510-520.
Regulation of neuronal and dopa decarboxylase gene.
Botwell, D. D. L., Kimmel, B. E., Simon, M. A., and Rubin, G. M. (1989). Regulation of the pattern of sevenless expression in the developing Drosophila eye. Proc. Natl. Acad. Sci. USA 86,62456249. Botwell, D. D. L., Lila,T., Michael, W. M., Hackett, D., and Rubin, G. M. (1991). Analysis of the enhancer element that controls expression of sevenless in the developing Drosophila eye. Proc. Natl. Acad. Sci. USA 88, 6853-6857. Bray, S. J., Johnson, W. A., Hirsh, J., Heberlein, U., and Tjian, T. (1988). A c&acting element and associated binding factor required for CNS expression of the D. melanogaster dopa decarboxylase gene. EMBO J. 7, 177-188. Chin, A., Reynolds, E., and Scheller, R. H. (1990). Organization and expression of the Drosophila FMRFamide-related prohormone gene. DNA Cell Biol. 9, 263-271. Duve, H., Johnson, A. H., Sewell, J. C., Scott, A. G., Orchard, I., Rehfield, J. F., and Thorpe, A. (1992). Isolation, structure and activity of -Phe-Met-Arg-Phe-NH2 neuropeptides (designated calIiFMRFamides) from the blowfly Calliphora vomitoria. Proc. Natl. Acad. Sci. USA 89, 2326-2330. Evans, P. D., and Myers, FMRFamide and related 126403-422.
C. M. (1986). The modulatory peptides on locust muscle.
actions of J. Exp. Biol.
Fisher, J. A., and Maniatis, T. (1988). Drosophila Adh: a promoter element expands the tissue specificity of an enhancer. Cell 53, 451-461. Garabedian, M. J., Shepherd, B. M., and Wensink, P. C. (1986). A tissue-specific transcription enhancer from the Drosophila yolk protein 1 gene. Cell 45, 859-867. Goto, T., Macdonald, P., and Maniatis, T. (1989). Early and periodic patterns of even skipped expression are controlled
late by
Neuropeptide 291
Gene
Regulation
distinct regulatoryelementst$at Cell 57, 413-422.
respond
todistinctspatialcues.
Crosfeld, F., van Assendelft, G. B., Greaves, D. R., and Kollias, G. (1987). Position-independent, high-level expression of the human B-globin gene in transgenic mice. Cell 57, 975-985. Hammer, R. E., Krumlauf, R., Camper, S. A., Brinster, R. L., and Tilghman, S. M. (1987). Diversity of alpha-fetoprotein gene expression in mice is generated by a combination of separate enhancer elements. Science 235, 53-58. Harding, K., Hoey,T., Warrior, R., and Levine, M. (1989). Autoregulatory and gap gene response elements of the even-skipped promoter of Drosophila. EMBO J. 8, 1205-1212. Hebener, J. F., Cwikel, B. J., Hermann, H., Hammer, R. E., Palmiter, R. D., and Brinster, R. E. (1989). Metallothionein-vasopressin fusion gene expression in transgenic mice. J. Biol. Chem. 264, 1884418852. Hiromi, Y., and Gehring, W. J. (1987). Regulation the Drosophila segmentation gene fushi tarazo. Hirsh, J. (1989). Molecular genetics of dopa biogenic amines in Drosophila. Dev. Cenet.
and function of Cell 50,963-974.
decarboxylase 10, 232-238.
and
Howard, K. R., and Struhl, G. (1990). Decoding positional mation: regulation of the pair-rule gene hairy. Development 1223-1231.
infor110,
Howard, K. R., Ingham, P., and Rushlow, specific alleles of the Drosophila segmentation Dev. 2, 1037-1046. Johnson, W. A., and POU-domain protein expression in specific 470.
Hirsh, J. (1990). Binding to a sequence element dopaminergic neurons.
C. (1988). Regiongene hairy. Genes of a Drosophila regulating gene Nature 342,467-
Schneider, L. E., and Taghert, P. H. (1990). Organization pression of the Drosophila FMRFamide neuropeptide Biol. Chem. 265, 68906895. Schneider, L. E., O’Brien, hybridization analysis of tide gene. I. Restricted stages. J. Comp. Neural.
and exgene. J.
M. A., and Taghert, P. H. (1991). In situ the Drosophila FMRFamide neuropepexpression in embryonic and larval 304, 608-622.
Scholnick, S., Morgan, B. A., and Hirsh, J. (1983). The cloned dopa decarboxylase gene is developmentally regulated when reintegrated into the Drosophila genome. Cell 34, 37-45. Scholnick, S., Bray, S. J., Morgan, B. A., McCormick, C. A., and Hirsh, J. (1986). Distinct central nervous system and hypoderm regulatory elements of the D. melanogasterdopa decarboxylase gene. Science 234, 998-1002. Scholnick, S., Caruso, P. A., Kleminic, J., Mastick, G. S., Mauro, C., and Rotenberg, M. (1991). Mutations within the Ddc promoter alter its neuron-specific pattern of expression. Dev. Biol. 746, .423-437. Spradling, A. C. (1986). P element-mediated transformation. Drosophila: A Practical Approach, D. B. Roberts, ed. (Oxford: Press), pp. 175-196. Spradling, A. C., and Rubin, P elements into Drosophila 278, 341-347. Struhl, K. (1991). patterns. Neuron
Mechanisms 7, 177-181.
G. M. (1982).Transposition germ line chromosomes. for
diversity
in gene
In IRL
of cloned Science expression
Swanson, L. W., Simmons, D. M., Arrija, J., Hammer, R., Brinster, R., Rosenfeld, M. G., and Evans, R. M. (1985). Novel developmental specificity in the nervous system of transgenic animals expressing growth hormone fusion genes. Nature 377,363-365.
Johnson, W.A., McCormick, C.A., Bray, S. J., and Hirsh, J. (1989). A neuron-specific enhancer of the Drosophila dopa decarboxylase gene. Genes Dev. 3, 676-686.
Taghert, P. H., and Schneider, L. E. (1990). Inter-specific comparison of a Drosophila neuropeptide gene encoding FMRFamiderelated neuropeptides. J. Neurosci. 70, 1929-1942.
Kitamoto, T., Ikeda, K., and Salvaterra, P. M. (1992). Analysis of cis-regulatory elements in the 5’ flanking region of the Drosophila melanogaster choline acetyltransferase gene. J. Neurosci. 12, 1628-1639.
Tamura, T., Kunert, C., and Postlethwait, J. (1985). Sex- and cellspecific regulation of yolk polypeptide genes introduced into Drosophila by P-element-mediated gene transfer. Proc. Natl. Acad. Sci. USA 82, 7000-7004.
Kondoh, H., Katoh, K., Takahashi, Y., Fujisawa, H., Yokohama, M., Kimura, S., Katuki, M., Saito, M., Tatsuji, N., Hiramoto, Y., and Okada, T. S. (1987). Specific expression of the chicken deltacrystallin gene in the lens and the pyramidal neurons of the pyriform cortex in transgenic mice. Dev. Biol. 720, 177-185.
Thummel, C. S., Boulet,A. M., and Lipshitz, H. D. (1988). Vectors for Drosophila P element-mediated transformation and tissue culture transfection. Gene 74, 445-456.
Low, M. J., Lechan, R. M., Hammer, R. E., Brinster, R. L., Habener, 1. F., Mandel, G., and Goodman, R. H. (1986), Gonadotrophinspecificexpressionof metallothioneinfusion genes in pituitaries of transgenic mice. Science 237,1002-1004.
White, K., Hurteau, T., and Punsal, P. (1986). FMRFamide-like immunoreactivity in Drosophila: and distribution. J. Comp. Neural. 247, 430-438. Young, W. S. I., Reynolds, K., Shepard, Castel, M. (1990). Cell-specific expression in transgenic mice. J. Neuroendocrinol.
Neuropeptide development
E. A., Gainer, H., and of the ratoxytocin gene 2, 917-925.
Maniatis, T., Goodbourn, S., and Fischer, J. A. (1987). Regulation of inducibleand tissue-specific gene expression. Science 236, 1237-1245.
Note
Nambu, J. R., Murphy-Erdosh, C., Andrews, P. C., Feistner, and Scheller, R. H. (1988). Isolation and characterization Drosophila neuropeptide gene. Neuron 7, 55-61.
We have recently found using X-Gal histochemistry that ectopic /acZ expression in the CNS of pWFl1 lines can also be detected in dispersed, small cells of the optic lobes (J. Liu and P. Taghert, unpublished data).
G. J., of a
O’Brien, M. A., Schneider, L. E., and Taghert, P. H. (1991). In situ hybridization analysis of the Drosophila FMRFamide neuropeptide gene. II. Constancy in the cellular pattern of expression during metamorphosis. J. Comp. Neurol. 304, 623-633. Pankratz, M. J., Jlckle, H. (1990). direct pair-rule the Drosophila
Seifert, E., Gerwin, N., Billi, B., Nauber, U., and Gradients of Krtippel and knirps gene products gene stripe patterning in the posterior region of embryo. Cell 67, 309-317.
Robertson, H. M., Preston, C. R., Phillis, R. W., Johnson-Schlitz, D. M., Benz, W. K., and Engels, W. R. (1988). A stable source of P element transposase in Drosophila melanogaster. Genetics 778, 461-470. Schneider, L. E., and Taghert, P. H. (1988). isolation and characterization of a Drosophila gene that encodes multiple neuropeptides related to Phe-Met-Arg-Phe-NHZ (FMRFamide). Proc. Natl. Acad. Sci. USA 85,1993-1997.
Added
in Proof
Vol. 10, 279-291,
February,
1993, Copyright
0 1993 by Cell Press
Cell Type-Specific Transcriptional Regulation of the Drosophila FMRFarnidei Neuropeptide Gene Lynne E. Schneider,* Marie S. Roberts, and Paul H. Taghert Department of Anatomy and Neurobiology Washington University School of Medicine Saint Louis, Missouri 63110
Summary We have used IacZreporter gene constructs to study the promoter/enhancer regions of the Drosophila FMRFamide neuropeptide gene in germ line transformants. FMRFamide is normally expressed in - 60 diverse neurons of the larval CNS that represent - 15 distinct cell types. An 8 kb FMRFamide DNA fragment (including 5 kb of 5’ upstream sequence) was sufficient to direct a pattern of /acZ expression that mimicked nearly all spatial aspects of the normal pattern. This result indicates that the cell-specific regulation of FMRFamide expression is largely generated by transcriptional mechanisms. Reporter gene expression was lost from selected cell types when smaller fragments were tested, suggesting that multiple control regions are included in the FMRFamide promoter. One region (a 300 bp fragment from -476 to -162) acted as an enhancer for 1 of the - 15 FMRFamide-positive cell types, the 012 neurons. These results suggest that, in the mature nervous system, the complex pattern of FMRFamide neuropeptide gene expression derives from the activity of discrete, cell type-specific enhancers that are independently regulated. Introduction In the mature nervous system, many neurons possess cell-specific transmitter phenotypes. The establishment of these neuronal properties during development requires the restriction of gene expression to specific neurons. We are interested in the mechanisms that underlie the developmental restriction of neuropeptide transmitter phenotypes. We are using the Drosophila FMRFamide neuropeptide gene as a model because its expression is restricted to a small though heterogeneous group of neurons. This expression is stereotyped such that individual neurons can be followed throughout development. The organization and expression of the Drosophila FMRFamide gene have been extensively characterized (Nambu et al., 1988; Schneider and Taghert, 1988, 1990; Chin et al., 1990; Schneider et al., 1991). This information provides the basis for an in vivo analysis of neuropeptide expression and function using the molecular genetic techniques that are available in Drosophila.
*Present address: Carnegie Institute of Washington, ment of Embryology, 115 West University Parkway, Maryland 21210.
DepartBaltimore,
The developmental expression of the FMRFamide gene and its single transcript has been characterized using antibodies and in situ hybridization. Expression is restricted to the CNS. In the CNS of the larva, the pattern consists of -60 neurons that are distributed throughout the brain and ventral nerve cord (Schneider et al., 1991). Following metamorphosis, there is an increase in the number of neurons expressing the gene to a total of - 120 (O’Brien et al., 1991; L. Schneider, E. Sun, D. Garland, and P. Taghert, submitted). Based on position and axonal projections, these neurons represent - 15 discrete cell types including large neuroendocrine cells and interneurons associated with the ventral ganglion, central brain, and sensory neuropils. We do not know whether these neurons participate in a common functional circuit or whether they subserve independent roles in the animal. Likewise, very little information is available concerning the roles of Drosophila FMRFamide peptides. Studies in other insects, however, have suggested some possible functions. In locusts, for example, FMRFamide and related peptides modulate skeletal musclecontractions in the leg(EvansandMyers,1986). In theflyCalliphora,some but not all endogenous FMRFamide-like neuropeptides induce fluid secretion from isolated salivary glands (Duve et al., 1992). We presume that the endogenous FMRFamide-like peptides of Drosophila play similar roles. To study the functions of FMRFamide peptides in thesedifferent neuron types, we hopefirst todefinethe molecular mechanismsthat underliecell type-specific expression and then to perturb FMRFamide expression in specific neurons. Previous studies of cell-specific gene expression have demonstrated that transcriptional regulation results from the combination of multiple promoter/enhancer elements. For example, the Drosophila dopa decarboxylase gene (Ddc), which is involved in the synthesis of the transmitters dopamine and serotonin, is expressed in a pattern of -150 neurons (Hirsh, 1989). Promoter deletion and in vitro mutagenesis studies have suggested that the pattern of Ddc gene expression is controlled by multiple cis-acting DNA elements (Scholnick et al., 1983,1986; Beall and Hirsh, 1987; Johnson et al., 1989). Similarly, in the fushi tarazu, even-skipped, and sevenless genes of Drosophila, numerous cis-regulatory elements act in concert to produce complex patterns of expression (Hiromi and Gehring, 1987; Goto et al., 1989; Botwell et al., 1989, 1991). This report describes the identification of c/s regulatory regions of the FMRFamide gene that are involved in the generation of the cell-specific pattern of FMRFamide neuropeptide expression in the mature nervous system. We present evidence for three principal conclusions. First, an 8 kb fragment containing upstream and intragenic regions of the FMRFamide
NellrOll 280
gene contains sufficient regulatory information to direct the appropriate pattern of transcription (and therefore peptide expression) to the normal complement of -15 neuronal cell types. Second, separate regions of this fragment are required for expression by different neuronal cell types. Third, neuropeptide gene expression in a single cell type, optic lobe (OL) visual system neurons, is produced by a small DNA region that displays the properties of a cell typespecific enhancer. Results The Drosophila FMRFamide gene is expressed only in the CNS, and within the larval CNS, expression is restricted to -60 neurons that represent - 15 distinct cell types. To analyze this stereotyped pattern of expression, we designed a study of FMRFamide gene regulation with three specific goals. First, we wanted to determine how much FMRFamide DNA is necessary to produce a wild-type pattern of expression. Second, we wanted to determine the distribution of the regulatory regions that underlie cell-specific FMRFamide expression. Finally, we wanted to determine whether the regulatory regions that are necessary for gene expression in different cell types are shared or distinct. We addressed these questions using three different types of /acZexpression vectors: a translational fusion vector, a transcriptional fusion vector, and an enhancer-detector vector. The first formed fusion FMRFamide-IacZ RNAs and proteins; the second formed fusion FMRFamide-/acZRNAs, but lacked FMRFamide protein sequences; the third did not contain FMRFamide sequences in either the RNAs or proteins formed. These vectors and the FMRFamide genomic regions that we inserted into them are diagrammed in Figure 1. We introduced each of these constructs into the Drosophila genome using P element transformation and, for each, established stable transformant lines (Spradling and Rubin, 1982). For clarity, we have confined our descriptions to the CNS of late larval and adult stages. In addition, for each of the constructs described below, IacZ expression is compared with FMRFamide expression as revealed by in situ hybridization and by an antiserum raised against the FMRFamide precursor (Schneider et al., 1991, submitted; O’Brien et al., 1991). /acZ Expression in the pWF3 Construct Closely Resembled the Endogenous Pattern of FMRFamide Gene Expression In the first /acZconstruct, pWF3, we hoped to include all of the regulatory regions that are necessary for appropriate expression of FMRFamide RNA and protein. Therefore, we began with a large (-8 kb) genomic fragment that contained 5 kb of upstream sequence, exon I (106 bp), the entire intron (- 3 kb), and the first 69 bp of exon II. This fragment was inserted, in frame, into the /acZgene of the translational fusion vector pCaSpeR-Bgal (Figure 1; Thummel et al., 1988).
The presumed translational start site of the FMRFamidetranscript isencoded bythefirstthree nucleotides of exon II (Schneider and Taghert, 1990), thus the fusion protein contains the first 23 amino acids of the FMRFamide precursor. There is only one detectable transcript from the FMRFamide gene (Schneider and Taghert, 1990), and the sequences around the transcription start site are contained, unaltered, within the pWF3 construct. Therefore, we assume that the fusion gene initiated from the same site. In addition, there are no ATG triplets in exon I and only short open reading frames within the intron (Schneider and Taghert, 1990; M. S. R. and P. H.T., unpublished data). Therefore, we assume that translation of the fusion protein initiated at the site that most closely matches the consensus sequence (nucleotides l-3 of exon II). We initially isolated two independent transformants and, subsequently, generated four more by mobilizing one of the original inserts into new genomic sites (Robertson et al., 1988). Due to different fixation conditions required to visualize FMRFamide and B-galalctosidase (B-gal) immunoreactivity, we did not directly compare staining with the two antibodies in the same animals. We therefore assigned neuronal identities and made comparisons, based on the position, size, morphology, and number of immunoreactive neurons in each group. All six of the pWF3 transformants had a similar pattern of /acZ expression, which was nearly indistinguishable from those of FMRFamide RNA and protein. Certain features of the IacZ pattern within the nervous system were possibly ectopic; these are described in a following section. We did not determine the extent of /acZ expression in nonneuronal tissues of transformed animals. The pWF3 pattern included most of the major cell types that express the FMRFamide gene (schematized in Figure 6). The most prominent FMRFamide neurons are the large neuroendocrine neurons (TV and Tva) that are located in each .of the thoracic neuromeres, and these could be identified in both larvae (Figure 2A) and adults (Figure 3A) of pWF3 animals. Similarly, in the suboesophageal neuromeres, the large SE2 interneurons and the ventral Sv cells were B-gal immunoreactive in both larvae (Figure 2A) and adults (see Figure 5A). In the brain, numerous B-gal-positive neurons corresponded to FMRFamide-positive neurons in larvae (Figures 4A and 4B) and adults (Figure 5A). For example, the OL2 pair of neurons lies between the central brain and optic ganglia; in pWF3 lines, these cells expressed /acZ (Figure 5A), as did adjacent smaller neurons (called OL2,), as described below. The OLI and 3, MP2, and Aldm cells were the only major cell types that expressed FMRFamide but not B-gal immunoreactivity in pWF3 lines. Although the overall intensity of B-gal immunoreactivity varied in the six lines, presumably due to the effect of chromosome position on the level of expression, it was typically higher than the endogenous gene. For example, neurons that normally express the FMRFamide gene at low levels (e.g., the Td and A8
Neuropeptide
Gene
Regulation
281
FMRFamide Locus Exon I
Exon II I
1
I
Translational Fusions
pWF6 El
Exon I
Exon II (AA l-23)
Iacz
pWF7 El
Exon I
Exon II (AA 1-2.3)
hi!
pWF4 El
Transcriptional Fusion
Exon I t-4500 bp)
Adh-AUG
\J
f I R
lacz
J
I 1 P SPBR
Enhancer Testing
Figure
1. Diagrammatic
Representation
of the
DNA
Constructs
Used
to Test FMRFamide
Gene
Regulation
At the top of the figure, the positions and sizes of the two exons are indicated. Below the gene schematic, the three translational fusion constructs, one transcriptional fusion construct, and two enhancer-testing constructs are shown in register with respect to FMRFamide gene sequences. The black boxes indicate FMRFamide exons (or portions thereof); solid lines indicate upstream or intronic FMRFamide sequences. Gray boxes indicate the /acZgene sequences. In pWF4, pWF8, and pWF11, Adh-AUG refers to sequences containing the initiator methionine from the alcohol dehydrogenase gene. The /acZ gene in pWF8 and pWFl1 begins with sequences constituting a minimal promoter and cap site from the hsp70 gene. The distances between FMRFamide sequences and the hs43 sequences in the enhancer-detecting constructs are not to scale. All numbers in parentheses refer to base pair sequences with the first nucleotide of exon I called +I. The white gene and the P element terminal sequences that flank all constructs are included but not shown. B, BamHI; C, Clal; E, Spel; C, Bglll; P, Pstl; R, EcoRI; S, Sall; X, Xbal.
were clearly visible in pWF3 flies (Figures 4B and4C). Whilethe higher level of expression probably reflected the greater stability of the IacZ fusion product, the relative levels of IacZand FMRFamide expression among the -15 cell types were the same. For
neurons)
example, theTv, SE2, and OL2 neurons had the strongest staining with both FMRFamide RNA probes, antiFMRFamide antibodies, and anti-IacZ antibodies; other specific groups had correspondingly intermediate or low levels of staining with all reagents.
Neuron 282
Figure
2. Expression
of /acZ in the Larval
Nervous
System
of pWF Transformant
Lines
B-Gal immunoreactivity is shown in whole-mount nervous systems from late third instar larvae: anterior is toward the top, and the ventral aspect of the ventral ganglion is shown in each photograph. The TV and associated smaller neurons (including Tva) are B-gal immunoreactive in pWF3 (A) and pWF7 lines (6); only the TV cell is immunoreactive in pWF4 (C) lines and only the smaller cells in pWF6 lines (D). Staining in SE2 neurons is present only with the largest construct, pWF3 (A), whereas staining in Sv neurons is present with the pWF3 (A), pWF7 (B), and pWF6 (D) constructs. The dark images at the midline of the pWF4 tissue are the thoracic neurohaemal organs (out of focus), in which immunoreactive terminals of the TV neurons are present; see also Figures 4E and 4F. Note the differences in staining intensity; pWF3 lines typically display the strongest levels of B-gal-like immunoreactivity. Abbreviations for cell names are explained in the text. Bar in (C), 100 pm for all panels.
Figure
3. Expression
of /acZ by theTv
and Tva Neurons
in the Ventral
Ganglion
of the Adult
Nervous
System
of pWFTransformant
Lines
Staining in the paired TV neurons is seen in all three thoracic neuromeres (long arrows), whereas paired Tva neurons are typically seen in only the second thoracic neuromere (short arrows). TV and Tva are both B-gal positive in pWF3 (A) and pWF7 (C) lines. Only the TV cell is B-gal positive in pWF4 lines (B), and only the Tva cell is B-gal positive in pWF6 (D). Neither cell type is immunoreactive in pWF8 or pWFl1 (data not shown). Bar in (C), 100 pm for all panels.
Neuropeptide 283
Figure
Gene
Regulation
4. Photomicrographs
Showing
/acZ Expression
in Larval
and Adult
Nervous
System
of pWF Transformant
Lines
(A and B) pWF3, (C and D) pWF6, and (E and F) pWF4. In (A)-(D), anterior is toward the top of the photograph; in (E) and (F) anterior is to the left, and dorsal is toward the top. In (A) and (B), lad expression is shown in two focal planes of the same larval preparation; the same is true for (E) and (F). kc2 expression in pWF3 lines is similar to endogenous FMRFamide expression, as shown in the dorsal region of the brain (A) and the ventral brain and dorsal nerve cord (B). pWF6 /acZ expression also mimics many aspects of FMRFamide expression, shown in theadult ventral nerve cord (C)and dorsal brain (D). In theventral ganglion of pWF4 transformants, ladexpression is restricted to the TV neurons (E), but within these cells it is found throughout the entire cell body, axons, and terminals ([F], within the segmental neurohaemal organs, called transverse nerves FNI-31) to a greater degree than in the other lines. Bar in (E), 150 pm for (A)-(D) and 75 nm for (E) and (F).
The greater stability of the g-gal fusion protein may also explain our observation that P-gal immunoreactivity in adult-specific neurons appeared at an earlier time during adult development than did FMRFamide hybridization signals or immunoreactivity. For example, FMRFamide expression was first detected in the adult-specific Tva and OL2 neurons during the prepupal stages of adult development (O’Brien et al., 1991). In contrast, in pWF3 animals we could first detect
P-gal immunoreactivity larvae (Figure 2A).
in these
neurons
in wandering
/acZ Expression by Different Cell Groups Correlated with Different Regions of the FMRFamide Gene To begin a dissection of the regulatory regions of the FMRFamide gene, we generated three constructs in which we made deletions of the FMRFamide se-
Figure
5. /acZ Expression
in the Adult
Brain
of pWF Transformants
All photographs were taken at approximately the same ventral focal plane. Anterior is and (D) pWFll.The SE2 neurons express lacZonly in the pWF3 lines. TheOL2and smaller the central brain and optic ganglia, are visible in all except the pWF6 lines. In pWFl1 positive; diffuse staining in the overlying tissue is likely an artifact. Note that the lacZ as intense as that produced by the much larger pWF3 translational fusion construct.
quences represented in pWF3. pWF7 and pWF6 are also translational fusion constructs. They contain the exon I, intron, and exon II sequences that are present in pWF3, but pWF7and pWF6 include only922 bp and 162 bp, respectively, of upstream sequence. pWF4 is atranscriptional fusion construct thatwas made using the pCaSpeR-AUG-Bgal vector (Thummel et al., 1988): pWF4contains -4.5 kb of’upstream sequence, a portion of exon I (including the transcription start site), and the translation initiation site from the Drosophila Adh gene. In flies transformed with each of these constructs a portion of the endogenous pattern of FMRFamide expression was lost. To simplify the discussion of these results, we present the changes in expression in three principal FMRFamide neuronal cell types: the Tv/Tva neurons, the SE2 neurons, and the OL2 neurons. The two TV neurons are neuroendocrine cells that are present in each of three thoracic neuromeres. They express the FMRFamide gene throughout the life of the fly, beginning in late embryogenesis. In the adult-specific Tva neuron, which is located adjacent to TV in the second thoracic neuromere, FMRFamide expression is first detected in early pupal stages. Both neurons innervate the segmental neurohaemal or-
toward the top. (A) pWF3, (B) pWF4, (C) pWF6, OL2, neurons (arrowheads), which lie between lines, only the OL2 and OL2, neurons are B-gal expression in the OL2 group of pWFl1 lines is Bar in (C), 100 pm for all panels.
gans. Despite their common positions and axonal targets, laczexpression in thesetwocell types correlated with different genomic fragments. In pWF3 and pWF7 lines (n = 6), TV and Tva were both B-gal immunoreactive (Figures 2A and 2B; Figures 3A and 3C). These constructs include the sequences from base pairs -922 to +69 of exon II. In eight lines transformed with the pWF6 construct (which contains the sequences from base pairs -162 to +69 of exon II), /acZ was expressed in the Tva, but not theTv neurons (Figure 2D; Figure 3D). In contrast, the .pWF4 construct (n = 5: one original and four mobilized transformant lines) produced expression in the TV neurons only (Figure 2C; Figure 3B). This construct contains 4.5 kb of upstream DNA plus 50 bp of exon I. Thus, TV expression required sequences between base pairs -922 and -162, and Tva expression required sequences between base pairs -162 and +69 of exon II (see Figure 7). The analysis of these transformants also permitted the definition of regulatory regions required by the SE2 neurons. These neurons are unpaired midline interneurons of the suboesophageal neuromeres that project widely throughout the brain and ventral ganglion. We detected /acZ expression in this group in
Regulation
the pWF3 lines (Figure 29; Figure 5A), but not in pWF4 (Figure2C; Figure 5B), pWi7(Figure 2B), or pWF6 lines (Figure 2D; Figure 5C). The SE2 neurons were the only major cell group whose expression was specific to pWF3 lines; theonly DNA region that is unique to the pWF3 construct is an -450 bp fragment at the Send of the 5 kb upstream region. Therefore, the presence of this 450 bp region was necessary for expression in the SE2 neurons. By similar analysis, DNA regions required for /acZ expression by the prominent OL2 (and OL2,) neurons of the brain were mapped to the same sequences as appeared necessary for the TV neurons, i.e., sequences between base pairs -922 and -162.OL2 neurons were B-gal positive in lines pWF3 (Figure 2A; Figure 5A), pWF4 (Figure 5B), and pWF7 (data not shown), but not in pWF6 (Figure 5C).
Ectopic Aspects of /acZ Expression in pWF Transformants Some aspects of IacZ expression were ectopic with respect to established patterns of FMRFamide expression (Schneider et al., 1991; O’Brien et al., 1991). First, in pWF3 and pWF6 lines, we often observed a B-galpositive cell (V,) in a position where we had seen FMRFamide expression only in rare cases. Second, in some lines we observed B-gal immunoreactivity in neurons where we had never seen FMRFamide expression. The ectopic neuron that appeared most consistently was the X cell, present in all pWF3 and many pWF6 and pWF7 lines (Figure 6). Other novel cell types included neurons that were slightly stained and that appeared less often. In general, the pWF6 lines displayed the most ectopic expression, although most of these lines also shared the pattern that is shown in Figure 6. For example, four of eight pWF6 lines had /acZ expression in dispersed perineural glial cells of the optic lobes and central brain of adults. This ectopic staining was patterned and reproducible. Another pWF6 line displayed ectopic B-gal staining in clusters of 2-8 cells at thedorsal andventral midlineofthe brain and in small cell clusters throughout the ventral ganglion. Finally, B-gal expression in some cell types was ectopic by virtue of the large numbers of B-gal-immunoreactive neurons. For example, in pWF3 lines there were -60 OL2, neurons in each hemisphere (Figure 5A), whereas with anti-FMRFamide antibodies we had sometimes detected 2-5 small neurons in each hemisphere. We do not know whether these examples of ectopic expression represent anomalous expression due to position effects or previously undetected aspects of FMRFamide gene expression.
A Cell Type-Specific and OL2, Neurons Comparisons transformant the FMRFamide distinct cell
Enhancer
for the OL2
of the /acZ patterns lines suggested that gene are necessary types. To determine
from the different different regions of for expression in whether these re-
gions contain enhancers that are also sufficient to dictate cell type-specific gene expression, we tested two DNA regions for their ability to drive expression from a heterologous promoter; for these constructs, we used the Casper hs43-IacZenhancer-detector vector. The pWF8 enhancer construct contained sequences from +66 bp of exon I to +I248 bp of the intron (Figure 1). The experiments described above implicated this as part of the region necessary for expression by several cell types (see Figure 7). The eight pWF8 lines did not produce any B-gal immunoreactivity in the CNS, and therefore this DNA fragment appears insufficient to regulate FMRFamide expression by itself. The pWF11 enhancer construct contained the FMRFamide gene fragment from base pairs -476 to -162 upstream to the transcription start site. This -300 bp fragment is included in the region that was implicated for expression by the TV, SP4, OL2, and OL2, neurons. The fragment was inserted in an orientation opposite to that found within the FMRFamide gene. Eight of nine pWF11 lines had B-gal immunoreactivity exclusively in the OL2 and OL2, neurons (Figure 5D); the ninth line had no detectable /acZ expression. These results indicate that this DNA fragment possesses the properties of an enhancer (Maniatis et al., 1987): it can influence expression of a reporter gene by a heterologous promoter, and it acts in an orientation-independent manner. The activity of this enhancerwas highly selective: within the nervous system, it regulated gene expression in one cell type. The OL2 and OL2, neurons expressed /acZ at high levels in both pWF3 and pWF11 lines (compare Figures 5A and 5D). In contrast, in the pWF4and pWF7lines (both of which contained more FMRFamide regulatory DNA than did pWF11 lines), IacZexpression in the OL2 neurons was weak (Figure 5B), and only a few OL2, neurons were stained (Figures 5B and 5D). FMRFamide expression,asassayed byantibodystainingand in situ hybridization, is typically strong in OL2 neurons;weak antibody staining is observed in only a few OL2, neurons.
Discussion Previously, we mapped FMRFamide gene expression using both antibody staining and in situ hybridization (Schneider et al., 1991; O’Brien et al., 1991). We have now shown that an 8 kb fragment of the FMRFamide gene (pWF3) is sufficient to direct /acZ expression in vivo with a pattern and intensity that are nearly indistinguishable from that of the endogenous gene. This correspondence suggests that the pattern and level of FMRFamide gene expression are largely determined by transcriptional control mechanisms. The lossof B-gal immunoreactivityfrom specificcell types following transformation with the deletion constructs suggests that control regions within the FMRFamide promoter are distributed throughout the 8 kb of flanking and intragenic sequences (Figure 7). The distributed organization of required elements is con-
I
Anti-proFMRFamide Immunoreactivity
I
l-l-3v
rl
pWF3
Exon I (-so00 bp)
BR
L
PSPBR
n
Exon II (AA l-23) b
E
IacZ J
GC
0
pWF4
a‘ v l-l-3*
Exon I
Ad%-AUG
64500 bp)
lacZ J
Ic I R
Figure
P SPBR
6. Summary
of FMRFamide
and
/acZ
Expression
in Transformant
Lines
Each pair of drawings represents the mature larval and adult CNS, respectively. For transformant lines, the summary drawings are shown on the right, with the DNA constructs on the left. Only those neuron groups that stained consistently are included. The constructs are positioned in register with respect to FMRFamide gene sequences. Data on endogenous proFMRFamide expression (first panel) are taken from a study using enriched setum antibodies that recognize an epitope on the deduced proFMRFamide precursor (Schneider et al., submitted). The nomenclature for naming cell types derives from White et al. (1986) and denotes the relative position of the ceil
Neuropeptide 207
Gene
Regulation
B
III
pWF6
Exon II MA l-23)
Exon I f-162
bp)
1acZ J
4 PBR
E
CC
SPl
I
I'
pWF7
, . Sl-3V Tl-3V
Exon II (AA l-23)
Exon I
PSPBR
E
1acZ
c c
_
. --
. . RX) TVs Td
v
GC
r-l
pWFl1
-476 -162
bp
A&-AUG
bp ““I
body (e.g., SPI, superior TI cells: thoracic ventral, of FMRFamide expression,
1
T
protocerebrum 1; O!&!, optic lobe 2). We add lower case letters to indicate positional detail (e.g., TV, Td, and dorsal, and lateral cells, respectively). The X cell is shown in parentheses because it does not match the map although it was observed consistently in pWF3, pWF6, and pWF7 lines.
Neuron 288
Figure 7. A Schematic That Illustrates Positions of DNA Regions within FMRFamide Promoter That Control Specific /acZ Gene Expression Sufficient
R
No Activity E Exon
I
II
-5000 bp
sistent with previous studies of tissue-specific gene expression, which have suggested that transcription is regulated by multiple modular enhancers (Tamura et al., 1985; Garabedian et al., 1986; Hammer et al., 1987; Grosfeld et al., 1987; Hiromi and Gehring, 1987; Fisher and Maniatis, 1988). There have been comparatively fewer studies of cell type-specific gene expression within a single tissue. This type of regulation is particularly relevant to the nervous system because of the great diversity of neuronal phenotypes. In many cases, gene fusion constructs and deletions within normal regulatory regions have produced novel patterns of reporter gene expression in the nervous system (e.g., Swanson et al., 1985; Low et al., 1986; Kondoh et al., 1987; Hebener et al., 1989; Scholnick et al., 1991). Thus, altering the spatial arrangment of regulatory regions can create ectopic or extremely restricted patterns of gene expression. These results support the hypothesis that restriction of gene expression (and thus the generation of diverse neuronal phenotypes) results from the
Exon rl
II
the the Cell-
The FMRFamide gene is diagrammed in the centerofthefigurewiththetwoexonsindicated as black boxes. Restriction site abbreviations aredefined in the legend to Figure 1. The fragment that directs the full pattern of expression starts at the -5000 BamHl site (B) and ends at the exon II Clal site (C). Below the gene, the dotted lines and corresponding drawings summarize the results of the deletion analysis. These regions are necessary for /acZ expression in the indicated neurons. Above the gene, the results of the enhancer constructs are summarized. The 300 bp fragment lying 162 bp 5 to exon I is sufficient to direct expression of the reporter gene by a heterologous promoter in the OL2 neurons.
combinatorial activity of overlapping sets of enhancer elements. In vivo analyses of promoters for several neuronal genes have begun to define the mechanisms underlying phenotypic complexity. Highly restricted patterns of reporter gene expression have been attained using regulatory fragments of neural-specific genes. For example,Young et al. (1990) reported that oxytocin gene expression was closely mimicked in mice bearing a 5.2 kb DNA fragment from the oxytocinlvasopressin locus. In the case of the sevenless tyrosine kinase gene, Botwell et al. (1991) found that a 475 bp intragenie fragment was sufficient to drive the complex pattern of cell-specific gene expression in the developing retina. The organization of regulatory elements that underlie more widespread patterns of gene expression has also been described. Deletion analysis of the Drosophila choline acetyltransferase promoter revealed the presence of multiple regulatory elements, whose removal led to the absence of reporter gene expression in subsets of the normal pattern (Kita-
Neuropeptide 289
Gene
Regulation
moto et al., 1992). For the Ddc gene of Drosophila, deletion and mutation analyses have suggested that its expression by any single cell type depends on cell type-specific enhancer elements and a common, tissue-specific regulatory element that is required for expression by all Ddc neurons (Bray et al., 1988; Johnson et al., 1989; Hirsh, 1989). Similar to many of these studies, we found that the FMRFamide gene also contains a broadly distributed set of regulatory regions that appears to control different subsets of the cellular pattern. One of these regions, from base pairs -476 to -162, can act as a cell type-specific enhancer for a single neuronal type, the OL2 group. This result extends previous studies of geneexpression in the mature-nervous system: within a large regulatory domain that controls gene expression by many neuronal cell types, we have defined a small region that is sufficient for expression by a single cell type. The 300 bp OL2 enhancer region may contain more than one regulatory element. A sequence comparison between D. melanogaster and D. virilis revealed that this fragment contains four distinct sequences that display at least 80% sequence conservation (Taghert and Schneider, 1990). These could represent four independently regulated regions, or they may represent cis-acting elements that act in some combinatorial fashion. Regardless of its complexity, it is clear that this -300 bp fragment is not required for expression in all FMRFamide-expressing neurons because the construct that did not contain this region (pWF6) still expressed the fusion gene in many FMRFamidepositive neurons (Figure 6). While complex patterns of cell-specific gene expression are commonly thought to reflect the action of overlapping sets of regulatory elements that act combinatorially (e.g., Struhl, 1991), a simpler, noncombinatorial model may be sufficient to explain our results. We propose that the regulatory elements of the FMRFamide gene are autonomous enhancers, many of which are capable of directing expression to individual cell types (see Figure 7). Accordingly, the diverse neurons that share this specific neuropeptide phenotype do not share a common transcriptional control. Instead, each cell type possesses an independent mechanism to regulate FMRFamide gene expression; the parallel regulation of these distinct enhancers in differentcell types results in the full pattern of FMRFamide expression. This model is similar to onesthatdescribetheactivationof the pair-rulegenes hairy and eve in individual stripes across the Drosophila blastoderm (Howard et al., 1988; Harding et al., 1989; Howard and Struhl, 1990; Pankratz et al., 1990). This model has many important implications for the regulation of FMRFamide gene expression. For example, the use of independent enhancers could provide a very high degree of flexibility in the evolutionary modification of gene expression at the level of single cell types. A possible example of such evolutionary modification is illustrated by the TV and Tva neurons.
Although these two cells share a common cell body position, axonal target, and neuroendocrinefunction, different regions of the FMRFamide promoter are required for their respective expression (Figure 3). Furthermore, both the TV and Tva cells are present and morphologically differentiated in larval stages (D. Zitnan and P. Taghert, unpublished data), but only the TV neuron has detectable levels of FMRFamide expression at this time. In contrast, both the TV and Tva neurons in D. virilis are FMRFamide immunoreactive in larval stages (Taghert and Schneider, 1990). This subtle difference between closely related species may result from the modified use of independent, cell type-specific enhancers. There are many current limitations to this model that will require further investigation. First is the assumption that the other FMRFamide regulatory elements behave in a cell-specific manner, as does the -300 bp OL2 enhancer region. To support this conclusion, other gene regions that have been implicated, by deletion analysis, to control cell-specific gene expression (see Figure 7) will have to be tested for their ability to drive a heterologous promoter. Second, this study did not address possible functions of the FMRFamide gene outside of the -8 kb in pWF3. A third limitation of the model is that these data are derived from three different types of vectors. The different vectors could affect /acZ expression due to differences in, for example, posttranscriptional events that are not normal components of FMRFamide gene expression. We think it is unlikely that the use of different vectors substantially affected reporter expression because, in general, our results were consistent between constructs. For example, only those constructs that contained the sequences from base pairs -922 to -162 showed expression in the TV, SP4, and OL cells, regardless of the vector type. This suggests that the major differences between reporter gene expression from the different constructs were due to the presence or absence of specific FMRFamide DNA sequences. While these studies have demonstrated transcriptional control underlying FMRFamide expression in the Drosophila CNS, we have focused only on qualitative, but not quantitative, features of reporter gene expression. In transformants bearing the largest construct examined, the typical levels of /acZ expression were often stronger than in comparable cell types of transformants bearing smaller test fragments (see Figure 3; Figure 5). This suggests that other DNA regions that were not well defined by these studies (but that lie within the 8 kb tested) probably contribute to the general level of transcription in many individual cell types. An analogous conclusion was reached in the study of the Ddc promoter by Johnson et al. (1989). It is notable, however, that the 300 bp fragment tested in pWF11 lines produced /acZ expression in the OL2 group that was typically as strong as that found in animals bearing the largest construct (compare Figures 5A and 5D).
In summary, we have begun to define the organization of DNA regulatory regions that are responsible for the cell-specific pattern of FMRFamide neuropeptide gene expression. The broad spatial distribution of regulatory elements and the autonomy of the OL2 enhancer suggest that distinct enhancer elements direct FMRFamide gene expression to specific neuronal cell types. These results provide a basis for further definition of the enhancer elements within the FMRFamide gene and of the factors that interact with such elements. Experimental
Procedures
Design of DNA Constructs Theconstructsused inthisstudyarediagrammedinFigurel.The pWF3 construct was generated in three steps. First, the genomic clone Mt-8 (Schneider and Taghert, 1990) was digested with Pstl and Clal, and the fragment that begins at base pair -162 upstream of exon I and ends at base pair +69 of exon II was subcloned into the Pstl and Clal sites of the Bluescript vector (Stratagene). Second, this subclonewas digested with BamHl (BarnHI sites were present at +50 bp in exon I and in the polylinker 5’to exon I), and the -5 kb,BamHl fragment from Mt-8 (base pairs -5000 to f50 bp of exon I) was then ligated into this site. A recombinant with the correct orientation was selected by restriction enzyme analysis. Third, the entire insert was removed by Xbal-Clal digestion, Xbal were linkers added, and then ligated into the Xbal site of the pCaSpeR-Bgal vector (Thummel et al., 1988). The pWF6 and pWF7 constructs were generated from the pWF3 construct. pWF3 was digested with Pstl to remove sequences from base pairs -5000 to -162. The pWF6 construct was then created by religating the remaining portion of the construct. The pWF7 construct was generated by ligating the -940 to -162 Pstl fragment back into the,unique Pstl site of pWF6. The orientation of this fragment in the pWF7 construct was determined by restriction enzyme analysis. The construct pWF4 was made by’ligating the 4.5 kb EcoRlBamHl genomic fragment of phage Mt-8 (base pairs -4500 to +50 of exon I) into EcoRI-BamHI-digested pCaSpeR-AUG-Bgal (Thummel et al., 1988). The constructs pWF8 and pWFl1 were generated using the Casper hs43-laGvector (V. Pirrotta). The pWF8 construct contained the EcoRI-,Spel fragment from base pair +70 of exon I to 1250 of the intron. pWFl1 DNA was made by inserting the SallPstl fragment (base pairs -492 to -162) into Casper hs43-/a& as a Sall-Spel fragment; the Spel site was derived from the polylinker of the Bluescript plasmid into which this genomic fragment had previously been subcloned. The pWFl1 insert is in reverse orientation with respect to the /acZ gene. All DNA constructswereconfirmed byrestrictiondigestand partialsequence analysis. Generation of Transformant lines Transformant lines were generated as described by Spradling (1986). The constructs were purified twice on cesium chloride gradients, precipitated in ethanol, and resuspended in 0.1 mM phosphate buffer containing 5 mM KCI. DNAs were injected at concentrations of 300-800 ng/ul for transformation constructs and 100-400 nglul for helper plasmid (~25.7~‘). The host strain was yw67c23. Once established, individual transformant lines were maintained as homozygotes, or over appropriate balancer chromosomes. P-Cal lmmunocytochemistry Central nervous systems were dissected as whole mounts and fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS). Fixation for 2 hr at room temperature or overnight at 4°C gave comparable results. Tissues were rinsed in PBS containing
0.3% TritonX-lOO(PBS/TX)and blocked in PBSiTXcontaininglO% horse serum. The tissues were then stained for one to two nights at 4OC in mouse anti-B-gal (Promega) diluted in PBS/TX containing 3% horse serum and 0.001% sodium azide. The primary antibody was visualized either by using an horseradish peroxidase-conjugated anti-mouse antibody or with the Vector elite kit (Vector Labs). We found that two neuronal cell types of the background strain (~w6’~~~) were strongly stained by anti-B-gal antibodies when the immunohistochemical reactions were processed with Vectastain kit reagents. These neurons happened to be FMRFamide-positive neurons, the SPI and LPI cells of the brain (data not shown). Such endogenous activity was not evident when B-gal-like immunoreactivitywas revealed with a horseradish peroxidase-conjugated secondary antibody. We therefore scored the presence or absence of these two cell types, in various transformant lines, using the simpler staining protocol, and relied on themoresensitiveVectastainmethodforthoseneuronsinwhich such endogenous B-gal-like immunoreactivity was not detectable.
We wish to thank Carl Thummel, Vince Pirrotta, Yash Hiromi, and Steve Crews for sending vectors; John Majors and Jay Hirsh for advice; and John Merlie, Martha O’Brien, Karen O’Malley, Josh Sanes, and Dusan Zitnan for their comments on the manuscript. This work was supported by a grant from the National Institutes of Health (#21749) to P. H. T. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisemenf” in accordance with 18 USC Section 1734 solely to indicate this fact. Received
August
11,1992;
revised
December
7, 1992.
References Beall, C. J., and Hirsh, J. A. (1987). glial expression of the Drosophila Genes Dev. 7, 510-520.
Regulation of neuronal and dopa decarboxylase gene.
Botwell, D. D. L., Kimmel, B. E., Simon, M. A., and Rubin, G. M. (1989). Regulation of the pattern of sevenless expression in the developing Drosophila eye. Proc. Natl. Acad. Sci. USA 86,62456249. Botwell, D. D. L., Lila,T., Michael, W. M., Hackett, D., and Rubin, G. M. (1991). Analysis of the enhancer element that controls expression of sevenless in the developing Drosophila eye. Proc. Natl. Acad. Sci. USA 88, 6853-6857. Bray, S. J., Johnson, W. A., Hirsh, J., Heberlein, U., and Tjian, T. (1988). A c&acting element and associated binding factor required for CNS expression of the D. melanogaster dopa decarboxylase gene. EMBO J. 7, 177-188. Chin, A., Reynolds, E., and Scheller, R. H. (1990). Organization and expression of the Drosophila FMRFamide-related prohormone gene. DNA Cell Biol. 9, 263-271. Duve, H., Johnson, A. H., Sewell, J. C., Scott, A. G., Orchard, I., Rehfield, J. F., and Thorpe, A. (1992). Isolation, structure and activity of -Phe-Met-Arg-Phe-NH2 neuropeptides (designated calIiFMRFamides) from the blowfly Calliphora vomitoria. Proc. Natl. Acad. Sci. USA 89, 2326-2330. Evans, P. D., and Myers, FMRFamide and related 126403-422.
C. M. (1986). The modulatory peptides on locust muscle.
actions of J. Exp. Biol.
Fisher, J. A., and Maniatis, T. (1988). Drosophila Adh: a promoter element expands the tissue specificity of an enhancer. Cell 53, 451-461. Garabedian, M. J., Shepherd, B. M., and Wensink, P. C. (1986). A tissue-specific transcription enhancer from the Drosophila yolk protein 1 gene. Cell 45, 859-867. Goto, T., Macdonald, P., and Maniatis, T. (1989). Early and periodic patterns of even skipped expression are controlled
late by
Neuropeptide 291
Gene
Regulation
distinct regulatoryelementst$at Cell 57, 413-422.
respond
todistinctspatialcues.
Crosfeld, F., van Assendelft, G. B., Greaves, D. R., and Kollias, G. (1987). Position-independent, high-level expression of the human B-globin gene in transgenic mice. Cell 57, 975-985. Hammer, R. E., Krumlauf, R., Camper, S. A., Brinster, R. L., and Tilghman, S. M. (1987). Diversity of alpha-fetoprotein gene expression in mice is generated by a combination of separate enhancer elements. Science 235, 53-58. Harding, K., Hoey,T., Warrior, R., and Levine, M. (1989). Autoregulatory and gap gene response elements of the even-skipped promoter of Drosophila. EMBO J. 8, 1205-1212. Hebener, J. F., Cwikel, B. J., Hermann, H., Hammer, R. E., Palmiter, R. D., and Brinster, R. E. (1989). Metallothionein-vasopressin fusion gene expression in transgenic mice. J. Biol. Chem. 264, 1884418852. Hiromi, Y., and Gehring, W. J. (1987). Regulation the Drosophila segmentation gene fushi tarazo. Hirsh, J. (1989). Molecular genetics of dopa biogenic amines in Drosophila. Dev. Cenet.
and function of Cell 50,963-974.
decarboxylase 10, 232-238.
and
Howard, K. R., and Struhl, G. (1990). Decoding positional mation: regulation of the pair-rule gene hairy. Development 1223-1231.
infor110,
Howard, K. R., Ingham, P., and Rushlow, specific alleles of the Drosophila segmentation Dev. 2, 1037-1046. Johnson, W. A., and POU-domain protein expression in specific 470.
Hirsh, J. (1990). Binding to a sequence element dopaminergic neurons.
C. (1988). Regiongene hairy. Genes of a Drosophila regulating gene Nature 342,467-
Schneider, L. E., and Taghert, P. H. (1990). Organization pression of the Drosophila FMRFamide neuropeptide Biol. Chem. 265, 68906895. Schneider, L. E., O’Brien, hybridization analysis of tide gene. I. Restricted stages. J. Comp. Neural.
and exgene. J.
M. A., and Taghert, P. H. (1991). In situ the Drosophila FMRFamide neuropepexpression in embryonic and larval 304, 608-622.
Scholnick, S., Morgan, B. A., and Hirsh, J. (1983). The cloned dopa decarboxylase gene is developmentally regulated when reintegrated into the Drosophila genome. Cell 34, 37-45. Scholnick, S., Bray, S. J., Morgan, B. A., McCormick, C. A., and Hirsh, J. (1986). Distinct central nervous system and hypoderm regulatory elements of the D. melanogasterdopa decarboxylase gene. Science 234, 998-1002. Scholnick, S., Caruso, P. A., Kleminic, J., Mastick, G. S., Mauro, C., and Rotenberg, M. (1991). Mutations within the Ddc promoter alter its neuron-specific pattern of expression. Dev. Biol. 746, .423-437. Spradling, A. C. (1986). P element-mediated transformation. Drosophila: A Practical Approach, D. B. Roberts, ed. (Oxford: Press), pp. 175-196. Spradling, A. C., and Rubin, P elements into Drosophila 278, 341-347. Struhl, K. (1991). patterns. Neuron
Mechanisms 7, 177-181.
G. M. (1982).Transposition germ line chromosomes. for
diversity
in gene
In IRL
of cloned Science expression
Swanson, L. W., Simmons, D. M., Arrija, J., Hammer, R., Brinster, R., Rosenfeld, M. G., and Evans, R. M. (1985). Novel developmental specificity in the nervous system of transgenic animals expressing growth hormone fusion genes. Nature 377,363-365.
Johnson, W.A., McCormick, C.A., Bray, S. J., and Hirsh, J. (1989). A neuron-specific enhancer of the Drosophila dopa decarboxylase gene. Genes Dev. 3, 676-686.
Taghert, P. H., and Schneider, L. E. (1990). Inter-specific comparison of a Drosophila neuropeptide gene encoding FMRFamiderelated neuropeptides. J. Neurosci. 70, 1929-1942.
Kitamoto, T., Ikeda, K., and Salvaterra, P. M. (1992). Analysis of cis-regulatory elements in the 5’ flanking region of the Drosophila melanogaster choline acetyltransferase gene. J. Neurosci. 12, 1628-1639.
Tamura, T., Kunert, C., and Postlethwait, J. (1985). Sex- and cellspecific regulation of yolk polypeptide genes introduced into Drosophila by P-element-mediated gene transfer. Proc. Natl. Acad. Sci. USA 82, 7000-7004.
Kondoh, H., Katoh, K., Takahashi, Y., Fujisawa, H., Yokohama, M., Kimura, S., Katuki, M., Saito, M., Tatsuji, N., Hiramoto, Y., and Okada, T. S. (1987). Specific expression of the chicken deltacrystallin gene in the lens and the pyramidal neurons of the pyriform cortex in transgenic mice. Dev. Biol. 720, 177-185.
Thummel, C. S., Boulet,A. M., and Lipshitz, H. D. (1988). Vectors for Drosophila P element-mediated transformation and tissue culture transfection. Gene 74, 445-456.
Low, M. J., Lechan, R. M., Hammer, R. E., Brinster, R. L., Habener, 1. F., Mandel, G., and Goodman, R. H. (1986), Gonadotrophinspecificexpressionof metallothioneinfusion genes in pituitaries of transgenic mice. Science 237,1002-1004.
White, K., Hurteau, T., and Punsal, P. (1986). FMRFamide-like immunoreactivity in Drosophila: and distribution. J. Comp. Neural. 247, 430-438. Young, W. S. I., Reynolds, K., Shepard, Castel, M. (1990). Cell-specific expression in transgenic mice. J. Neuroendocrinol.
Neuropeptide development
E. A., Gainer, H., and of the ratoxytocin gene 2, 917-925.
Maniatis, T., Goodbourn, S., and Fischer, J. A. (1987). Regulation of inducibleand tissue-specific gene expression. Science 236, 1237-1245.
Note
Nambu, J. R., Murphy-Erdosh, C., Andrews, P. C., Feistner, and Scheller, R. H. (1988). Isolation and characterization Drosophila neuropeptide gene. Neuron 7, 55-61.
We have recently found using X-Gal histochemistry that ectopic /acZ expression in the CNS of pWFl1 lines can also be detected in dispersed, small cells of the optic lobes (J. Liu and P. Taghert, unpublished data).
G. J., of a
O’Brien, M. A., Schneider, L. E., and Taghert, P. H. (1991). In situ hybridization analysis of the Drosophila FMRFamide neuropeptide gene. II. Constancy in the cellular pattern of expression during metamorphosis. J. Comp. Neurol. 304, 623-633. Pankratz, M. J., Jlckle, H. (1990). direct pair-rule the Drosophila
Seifert, E., Gerwin, N., Billi, B., Nauber, U., and Gradients of Krtippel and knirps gene products gene stripe patterning in the posterior region of embryo. Cell 67, 309-317.
Robertson, H. M., Preston, C. R., Phillis, R. W., Johnson-Schlitz, D. M., Benz, W. K., and Engels, W. R. (1988). A stable source of P element transposase in Drosophila melanogaster. Genetics 778, 461-470. Schneider, L. E., and Taghert, P. H. (1988). isolation and characterization of a Drosophila gene that encodes multiple neuropeptides related to Phe-Met-Arg-Phe-NHZ (FMRFamide). Proc. Natl. Acad. Sci. USA 85,1993-1997.
Added
in Proof