Development of the Split Gal4 system
To implement the strategy of restricting transgene expression in vivo
by independently targeting transcription factor domains, we developed tools that could be deployed in the fruit fly, Drosophila melanogaster
, in which the binary Gal4-UAS system for transgene expression is already well-established (, Brand and Perrimon, 1993
). This system includes several hundred transgenic fly lines designed to express reporter and effector transgenes under the control of the “upstream activating sequence” recognized by the Gal4 DBD. We therefore sought to develop a system () in which the Gal4 DBD was paired with an AD potent enough to drive transgene expression at levels high enough to permit manipulation of cellular function. Previous studies have indicated that the primary Gal4 AD (i.e. “region II,” Ma and Ptashne, 1987b
) coupled with the Gal4 DBD might promote transcription at levels that are only a fraction of those produced by intact Gal4 (Ma and Ptashne, 1987a
). We therefore also tested the more potent AD from the Herpes Simplex Virus 1 transcription factor, VP16.
The ternary Split Gal4 System improves upon existing binary expression systems in restricting transgene expression
To optimize transcriptional activity, we fused the Gal4 DBD and the ADs to heterodimerizing leucine zippers, varying the zipper type, site of fusion (N- or C-terminus), and length of the polyglycine spacer between the zipper and the transcription factor domains (see Table S1 for details). The constructs that produced maximal transcriptional activity when co-transfected into Drosophila
SL-2 cells are shown schematically in . These optimal constructs incorporated synthetic leucine zippers based on the chicken B-Zip family member Vitellogenin Binding Protein (Moll et al., 2001
). The sequences of these zippers do not match those of Drosophila
B-Zip family members (Fassler et al., 2002
) and were selected for their strong heterodimerizing and low homodimerizing potentials. Transcriptional activities of the optimal constructs, expressed pairwise in transfected cells and measured in terms of the relative enzymatic activity of a UAS-ß -galactosidase reporter, were 52% and 84% that of intact Gal4 for the Gal4DBD-Gal4AD and the Gal4DBD-VP16AD pairs, respectively (). The individual optimized constructs, which we refer to as “hemidrivers,” displayed little to no transcriptional activity when transfected alone.
Optimized Split Gal4 constructs efficiently drive UAS-reporter gene expression in vitro and in vivo in the Drosophila nervous system
In vivo characterization of the Split Gal4 System
We evaluated the efficacy of the optimized hemidrivers in vivo in three steps, increasing restriction of transgene expression at each step. In Step 1, we created and crossed transgenic lines that expressed each hemidriver throughout the nervous system. This permitted evaluation of each hemidriver’s ability to promote neuronal transgene expression when used in concert with complementary hemidrivers. In addition, it permitted the assessment of possible side effects of each hemidriver’s individual expression. In Step 2, we created lines that expressed each hemidriver in a small set of identified neurons that expresses Crustacean Cardioactive Peptide (CCAP). Crosses of these lines to complementary, pan-neuronal hemidriver lines from Step 1 permitted rigorous evaluation of each hemidriver’s ability to faithfully restrict transgene expression to a set of neurons. In Step 3, we created lines that overlapped in expression pattern with the lines created in Step 2, but only within distinct subset of the CCAP-expressing neurons. This permitted validation of the Split Gal4 system as a tool for mapping the functional identities of defined subsets of neurons. The results obtained at each step of evaluation are discussed in the sections below, with the headings indicating whether the promoters used to express the hemidrivers (P1 and P2) drive transgene expression in a common set of neurons (P1 equals P2), in an included subset (P1 includes P2), or in the intersection of two distinct sets (P1 overlaps P2).
Step 1: Pan-neuronal expression using the Split Gal4 system (P1 equals P2)
Transgenic fly lines that expressed each hemidriver panneuronally were made using the promoter of the elav
gene (Yao and White, 1994
), and are referred to as elavGal4DBD
, and elavVP16AD
. These lines were viable and healthy, suggesting that expression of the transcription factor constructs throughout the nervous system is not deleterious in itself. In addition, the individual hemidrivers appeared transcriptionally inactive in vivo,
just as they were in vitro.
The progeny of crosses of each hemidriver line to a line carrying a UAS-EGFP or UAS-EYFP reporter transgene did not exhibit obvious nervous system fluorescence at any stage of development (data not shown). This was true even for homozygous animals containing two copies each of the single hemidrivers and the UAS-EGFP transgenes (), although at long exposure times limited fluorescence was observed in animals expressing two copies of elavGal4DBD
(), suggesting that this hemidriver may have weak transcriptional activity on its own.
In contrast, pairwise crosses of elavGal4DBD and elavGal4AD lines in the presence of a UAS-reporter yielded progeny with bright fluorescence specifically expressed in the central nervous system (CNS; ). We designate crosses in which DBD and AD hemidrivers are paired, with the intersection symbol, ∩ . All Gal4DBD ∩ elav Gal4AD crosses, yielded viable progeny that expressed the reporter transgenes in the nervous system, in some cases at levels approaching that of intact Gal4 (). Nervous system fluorescence was strong enough to be observed by epifluorescence under a dissection microscope as well as by confocal microscopy from the late embryonic period through adulthood.
Crosses in which the elavGal4DBD and elavVP16AD hemidrivers were paired yielded viable adult progeny with even stronger nervous system fluorescence (), though, as noted in Experimental Procedures, many of these crosses were lethal. Lethality was independent of reporter gene expression and has not been observed in any other crosses, including those that drive expression in all cholinergic neurons, which comprise much of the fly nervous system, or in broad patterns defined by VP16AD enhancer-trap lines (see below). Overall, our observations strongly indicate that toxicity is not a general feature of VP16AD and Gal4DBD co-expression in neurons, and may instead result from disruption of function in a small number of cells variably represented in the expression pattern of the elav promoter.
Step 2: The Split Gal4 system can faithfully target gene expression to a subset of neurons (P1 includes P2)
To confirm the ability of the Split Gal4 system to restrict transgene expression to a subset of cells within the pattern dictated by a promoter of interest, we first sought to limit UAS-EGFP expression within the nervous system to an identified set of neurons. To this end, we generated a second set of hemidrivers (CCAPGal4DBD
, and CCAPVP16AD
) using the promoter for the CCAP gene, which has been demonstrated previously (Park et al., 2003
) to be active in approximately 50 neurons of the larval central nervous system (). We then performed pairwise crosses, combining complementary transcription factor domains expressed under the control of the elav and CCAP promoters, and monitored the pattern of expression of a UAS-EGFP reporter transgene in the progeny.
The Split Gal4 system restricts UAS-transgene expression to the intersection of the expression patterns of two promoters
∩ CCAP Gal4AD
crosses resulted in UAS-EGFP expression exclusively within the more restricted CCAP pattern (), consistent with the reconstitution of transcriptional activity in only cells expressing both transcription factor domains. The reciprocal crosses, namely CCAPGal4DBD
, gave similar results (data not shown), and in both cases expression within the CCAP-expressing neurons (NCCAP
) was confirmed by double-labeling with an anti-CCAP antibody (Fig. S1A–C). In most crosses, some expression was seen in small numbers of non-CCAP-immunoreactive neurons. This lack of fidelity was similar in exent to that observed with the CCAP-Gal4 driver line created by Park et al. (2003)
, the expression pattern of which includes non-CCAP-expressing neurons in some animals (see legend, ). With both Gal4 drivers and Split Gal4 hemidrivers, “position effects” resulting from the action of genomic enhancers adjacent to the site(s) of transgene insertion are thus likely to perturb the transcriptional activity of the CCAP-promoter.
We determined the frequency of reporter gene expression in individual CCAP-immunopositive neurons in multiple preparations using the CCAPGal4AD line with the greatest fidelity to create a “consensus expression pattern” as shown in . Comparison of this pattern with the corresponding consensus expression pattern for the CCAP-Gal4 driver line () reveals only slight deviations, with the Split Gal4 cross failing to label two CCAP-expressing neurons in the final abdominal segment.
Gal4 transcriptional activity is regulated in yeast by the Gal80 repressor, which binds to a motif included within the Gal4AD (Ma and Ptashne, 1987a
). As shown in Figure S2, Gal80 also efficiently suppresses reporter transgene expression in elavGal4DBD
∩ CCAP Gal4AD
crosses when co-expressed under control of the CCAP promoter. The Gal4AD-based implementation of the Split Gal4 system can thus be used in conjunction with the increasing number of transgenic Gal80-expressing fly lines, further extending the range of application of the technique. In particular, coupling the Split Gal4 technique to the TARGET system (McGuire et al., 2003
), which exploits a temperature-sensitive Gal80 mutant to temporally regulate Gal4 activity, should permit conditional, as well as spatially restricted, expression of genes of interest.
Restricted expression of the UAS-EGFP reporter within NCCAP was also achieved with the VP16AD construct. The most faithful expression pattern was observed in CCAPGal4DBD ∩ elavVP16AD crosses, when expression of the Gal4DBD domain was anchored to NCCAP ( and Fig. S1D–F). Interestingly, the reciprocal crosses (elavGal4DBD ∩ CCAP VP16AD) resulted in expansion of the expression pattern to include numerous neurons outside of NCCAP with all CCAPVP16AD lines tested (Fig. S1G–I). This infidelity of expression presumably derives from ectopic VP16AD expression and may be the result of a cryptic enhancer in the CCAPVP16AD construct. In any case, it does not affect applications in which precise targeting of the VP16AD hemidriver is unnecessary, such as those using enhancer-trap lines discussed in Step 3. However, the possibility of imprecise targeting of this hemidriver must be considered if defined expression of VP16AD is desired.
To further investigate the selectivity of transgene expression driven by Split Gal4 hemidrivers,, and to confirm the utility of the Split Gal4 system in manipulations of neuronal function, we targeted expression of the cell death gene reaper
∩ CCAP Gal4AD
∩ elav VP16AD
crosses. As shown in , expression of UAS-reaper
under the control of the elav-Gal4 driver, results in early developmental lethality with almost all animials dying at the embryonic or early larval stages. In contrast, reaper
expression under control of the CCAP-Gal4 driver results in the two developmental phenotypes first described by Park et al (2003)
: animals either die as pupae with morphological defects characteristic of head eversion failure, or survive to adulthood without expanding their wings (UW; ). Immunohistochemical examination of animals from such crosses shows that reaper
expression by CCAP-Gal4 typically kills most, but not all, of the CCAP-expressing neurons during development (), whereas control crosses show no neuronal mortality ().
Genetic ablation of NCCAP neurons using the Split Gal4 system causes pupal lethality and wing expansion deficits that correlate with expression strength
As expected, expression of UAS-reaper using the Split Gal4 system led to pupal lethality and wing expansion deficits at levels that correlated with the extent of NCCAP ablation. In addition, the results better define the different efficacies of the two implementations of the Split Gal4 system. Expression of a single copy of the reaper transgene in NCCAP using the Split Gal4 system with Gal4AD (elavGal4DBD ∩ CCAP Gal4AD) was without effect, but expression of two copies caused wing expansion deficits in 90% of progeny, with little concomitant pupal lethality (). On average, expression of two copies of UAS-reaper resulted in the death of approximately two-thirds of the NCCAP neurons (). The Split Gal4 system, implemented with Gal4AD, thus clearly drives reaper expression less potently than intact Gal4 and the absence of the pupal lethal phenotype indicates that the neurons responsible for it were not killed or were killed in insufficient numbers even by two copies of the UAS-reaper transgene. In contrast, expression of a single copy of the reaper transgene in CCAPGal4DBD ∩ elav VP16AD crosses resulted in nearly complete pupal lethality () and the death of almost all NCCAP neurons (), indicating that this implementation of the Split Gal4 system drives reaper expression more potently than intact Gal4. These results are consistent with the levels of EGFP reporter gene expression seen using the two implementations of the system (see ). Control crosses expressing single hemidrivers and one or two copies of the reaper transgene produced progeny without pupal mortality or significant wing expansion deficits (). Overall, our results indicate that the Split Gal4 system drives reaper expression in NCCAP at levels sufficient to ablate some or all of these neurons, with VP16AD providing considerably more potent transgene expression than Gal4AD.
Step 3: The Split Gal4 system can target gene expression to the intersection of two overlapping expression patterns (P1 overlaps P2)
The greatest utility of the Split Gal4 system lies in its potential to limit transgene expression to the intersection of two distinct, but overlapping expression patterns. We describe here an example of this application using two defined promoters. The next section describes a second example using hemidrivers made with undefined promoters (i.e. enhancer-trap lines) to drive gene expression at the intersection of two patterns.
The Gal4AD, with its high degree of fidelity when expressed under a specific promoter, is particularly useful when both the AD and Gal4DBD hemidrivers must be faithfully targeted. We used the CCAPGal4AD
in conjunction with a Gal4DBD hemidriver made with the choline acetyltransferase (Cha) promoter (Salvaterra and Kitamoto, 2001
) to selectively drive transgene expression in cholinergic NCCAP
neurons. Preliminary immunohistochemical studies indicated that at least some NCCAP
neurons were cholinergic, and to provide a baseline for assessing the fidelity of ChaGal4DBD
-driven expression, we determined the consensus pattern of Cha promoter activity within NCCAP
immunohistochemically (). As indicated in , certain CCAP-immunopositive neurons, such as the bilaterally-represented pairs in the brain (), are consistently double-labeled in Cha-Gal4>UAS-EGFP preparations. In the ventral nerve cord, the more-weakly immunopositive member of the pair of CCAP-expressing neurons in hemisegments A1–A4 is also often EGFP-labeled, as shown in the examples of , while the strongly immunopositive member of this pair is never EGFP-labeled. In general, we observed considerable variability in the pattern of overlap, suggesting extensive variability of the Cha-Gal4 expression within NCCAP
, but the consensus labeling pattern (), derived from nine double-labeled preparations, shows that 18 of the 48 NCCAP
neurons are found within the Cha-Gal4 pattern in greater than a third of the preparations, while thirteen other neurons are found double-labeled at lower frequency.
A subset of NCCAP neurons lies within the expression pattern of the choline acetyltransferase (Cha) promoter, as identified by Cha-Gal4>UAS-EGFP labeling
shows the corresponding results obtained with the Split Gal4 system, in which UAS-EGFP expression was driven by ChaGal4DBD ∩ CCAPGal4AD (). Whole mounts of the CNS from these animals were stained with anti-CCAP antibody () to positively identify each EGFP-labeled neuron, and a consensus expression pattern was determined from the examination of multiple animals (). As with the Cha-Gal4 driver, and therefore unsurprisingly, there was considerable variability in the expression pattern, but in all preparations almost all EGFP-expressing neurons were also CCAP-immunopositive (), and there was good correspondence in the two consensus expression patterns (compare with ). In particular, the neurons most frequently observed in the Split Gal4 pattern were the CCAP-immunopositive neurons of the brain (, arrowheads) and specific hemisegmentally-represented neurons in the ventral nerve cord (, box). Just as observed with the Cha-Gal4 driver, the latter neurons in hemisegments A1–A4 corresponded to the weakly CCAP-immunopositive neuron of a pair (, and I, box), both of which are equally well labeled by CCAP-Gal4 ().
Targeting the cholinergic subset of NCCAP, using complementary hemidrivers made with the Cha and CCAP promoters
Restriction of gene expression to arbitrary subsets of cells within a group of interest using VP16AD enhancer-trap hemidrivers
Another strategy for restricting expression within a pattern of interest, particularly useful when there are no known promoters with overlapping expression patterns, uses enhancer-trap lines (Bellen et al., 1989
) to express the complementary hemidriver. Enhancer-trap constructs are made with a minimal promoter and express in “arbitrary” patterns dictated by local enhancer elements near the site of transgene integration into the genome. Because the fidelity of expression of the enhancer-trap construct is immaterial, the VP16AD, with its potential for driving transgene expression at high levels, is useful for this application. We therefore made enhancer-trap lines with the VP16AD construct (ETVP16AD
) and selected those with expression patterns that overlapped with NCCAP
. The broad expression patterns of two such lines (ETVP16AD
-N4 and ETVP16AD
-N6), revealed by crosses to elavGal4DBD
, are shown in , respectively. The restricted patterns of expression within NCCAP
(), generated by crosses of these lines to CCAPGal4DBD
, are shown in , respectively. As is evident from their consensus expression patterns (), the two enhancer-trap hemidrivers limit expression to two nearly mutually exclusive subsets of NCCAP
. ETVP16AD-N4 crosses to CCAPGal4DBD
drive expression almost exclusively in neurons of the subesophageal and thoracic ganglia, while those of ETVP16AD
-N6 drive expression primarily in the brain and abdominal ganglia.
Split Gal4 enhancer-trap lines can be used to target distinct subsets of neurons within NCCAP
Dissection of neural network function using the Split Gal4 system
Having identified enhancer-trap lines capable of selectively expressing transgenes in discrete subsets of NCCAP, we next used them to probe the functional identities of the neurons within this group. As indicated above, NCCAP plays a critical role in wing expansion in the early adult and we wished to determine which neurons within this group might be necessary and/or sufficient for this process. We therefore ablated the subsets of NCCAP neurons included in the ETVP16AD-N4 and ETVP16AD-N6 expression patterns using UAS-reaper, and correlated the patterns of the surviving NCCAP neurons with the wing expansion phenotype.
We used anti-CCAP immunostaining of the nervous system to assess the patterns of surviving NCCAP
neurons (). Because the CCAP-expressing neurons apoptose shortly after eclosion (Draizen et al., 1999
), it was necessary to evaluate the patterns of the surviving NCCAP
neurons in pharate adults, that is, at a stage just before eclosion and wing expansion. Therefore, we could not compare wing expansion with the surviving NCCAP
neurons in individual animals. To help identify these neurons, which differ somewhat in number and position from those of the larval stage, we double-labeled the preparations with an antibody to the hormone bursicon (), which is expressed in a subset of NCCAP
, including 14 neurons in the abdominal ganglion (). Ablations performed using the ETVP16AD
-N4 hemidriver primarily affected neurons of the subesophageal and thoracic ganglia (, black circles), as expected from the expression pattern in third-instar larvae (). Six abdominal ganglion neurons that appear in the expression pattern only during pupal development (data not shown), and which do not express bursicon, were also ablated. In contrast, ablations performed using the ETVP16AD
-N6 hemidriver mostly spared the neurons of the thoracic and subesophageal ganglia, but eliminated, on average, 12 of the 30 abdominal ganglion neurons, including nine that express bursicon (). Interestingly, only ablations performed with the ETVP16AD
-N4 hemidriver yielded animals that failed to expand their wings (). This effect was not fully penetrant, but because there were slight variations in the patterns of cell death from animal to animal (see , legend), it is likely that the 15% of animals that expanded their wings normally corresponded to those in which some small number of critical neurons were not killed.
Genetic ablation of neurons within NCCAP using the Split Gal4 system identifies anatomically and functionally distinct subsets of this network
The two lines therefore identify functionally, as well as anatomically, distinct subsets of NCCAP. The set of neurons surviving the ablation with ETVP16AD-N6 (, red) is clearly sufficient to support wing expansion, and the set surviving ablation with ETVP16AD-N4 (, red) is insufficient for wing expansion. Conversely, some or all of the neurons ablated in crosses using ETVP16AD-N4 (, black circles) must be necessary for wing expansion. If we remove from this latter group the small number of neurons that are clearly unnecessary, because they are ablated without effect in ETVP16AD-N6 crosses (e.g. the middle pair of subesophageal neurons), we are left with a minimal subset of neurons, some or all of which must be necessary for wing expansion (, red).
This “critical subset” represents part of a larger group of previously identified neurons, and our results with the Split Gal4 system confirm and extend an earlier model of NCCAP
function. Previously, we showed that neurons within NCCAP
act as a small network with output and regulatory functions, both of which are necessary for wing expansion (Luan et al., 2006
). The critical subset identified here (, red filled) definitively implicates several neurons within the broad group previously identified as candidate regulatory neurons as required for wing expansion (, blue circles). However, as can be seen in , the critical subset cannot be sufficient for wing expansion, since it does not include any of the 14 bursicon-expressing neurons that comprise the output group (green circles). As shown in , ablation of up to nine of the bursicon-expressing neurons with ETVP16AD
-N6 does not compromise wing expansion, suggesting that some or all of the remaining five bursicon-expressing neurons are necessary. Alternatively, it is possible that the output function is distributed, and that various subsets of the output group are sufficient to promote wing expansion. Although further work with other enhancer-trap lines and other effectors will be required to elucidate the individual functional identities of neurons within NCCAP
, this example illustrates how the Split Gal4 system can be used to systematically subdivide and define the functional elements of a neuronal circuit.
The Split Gal4 technique introduced here is a general and versatile method for targeting transgene expression to subsets of neurons within a pattern of interest. Because it uses the Gal4DBD, it is compatible with the hundreds of UAS-effector lines already available in Drosophila. Its implementation using the Gal4AD also makes it compatible with the growing number of Gal80 lines and permits an additional level of spatial or temporal control. The two implementations of the technique presented have complementary strengths. The Gal4AD, although weaker than the VP16AD, can drive effector transgene expression at levels sufficient to manipulate neuronal viability and function, and can be targeted with high fidelity when used with specific promoters. The VP16AD, as we have shown, is better-suited for use in enhancer-trap constructs where its lower promoter fidelity is unproblematic and its higher transcriptional activity permits even weak effector transgenes to be used at single-copy dosages. Although expression of the VP16AD together with the GalDBD may be toxic in a small population of neurons, as evidenced by the general lethality of panneuronal expression, such toxicity is clearly not the rule, as we have not observed it in any other application, including applications using VP16AD enhancer trap lines with broad expression in the nervous system ().
We have demonstrated the utility of the Split Gal4 technique in the field of neuronal circuit analysis, where we expect it will find fertile application. It should be possible to use the technique not only to subdivide known neuronal groups, but also to perform screens analogous to current Gal4 enhancer-trap screens to identify as yet unknown neuronal substrates of physiological processes or behaviors. Unlike current screens, however, it should be possible to use the Split Gal4 technique iteratively, to successively refine the identification of neuronal substrates within a group of interest once it is identified, as outlined in Figure S3. The level of resolution attainable with the Split Gal4 system may vary depending on the homogeneity of gene expression in the cell group of interest, but given the high degree of neuronal specialization normally observed, this is unlikely to be a major limitation. We also anticipate that the usefulness of the Split Gal4 system introduced here will extend beyond the nervous system to other tissues and other fields of research. Finally, while we have implemented the Split Gal4 technique in Drosophila, there is no reason it cannot be implemented in other genetic model organisms, such as zebrafish or mice, where similar transcriptional systems for the expression of foreign genes exist.