Genetically Accessing Neurons via Binary Expression Systems

Using the transposable P element to introduce transgenes into Drosophila has revolutionized the genetic study of Drosophila [1]. The P element-mediated transformation with the yeast transcription activator GAL4 revealed that the expression of the foreign gene GAL4 can be controlled by local genomic elements, such as enhancers [2]. Introducing both GAL4 and its binding sequence UAS (Upstream Activating Sequence) into flies led to the invention of the most used binary expression system, GAL4/UAS [3], in which the transgene following UAS was expressed only in the cells expressing GAL4, thus allowing targeted expression of effector genes simply by crossing a GAL4 line with a UAS line (Fig. 1A).

Open in a separate window

Fig. 1

Genetic targeting of transgene expression in Drosophila. A GAL4/UAS binary expression system. Expression of the transgene (indicated by dotted region) following the UAS is dictated by GAL4, whose expression is under the control of a regulatory sequence such as a promoter. B GAL4/UAS/GAL80 allow a set-difference expression of transgene. The UAS-transgene is expressed in cells expressing GAL4 controlled by promoter 1 but not GAL80 (which binds to GAL4 and inhibits its action on UAS) controlled by promoter 2. C split-GAL4 allows an intersectional expression of transgene. The transgene following the UAS is expressed in cells expressing both split-GAL4 parts, the GAD and DBD, which are under the control of promoter 1 and promoter 2, respectively. D FLP/FRT adds another layer of control over a binary expression system such as GAL4/UAS. The UAS-transgene is expressed in cells expressing both GAL4 and FLP, which stochastically excise the FRT-stop cassette to allow the effector gene to be expressed.

GAL4 lines were previously constructed with the GAL4 coding sequence following a minimal promotor (enhancer trap) [4–6] or a defined promoter (enhancer fusion) [3] integrated into the genome with the P element at random loci. Expression of the enhancer trap GAL4 depends on the local cis-regulatory sequences such as enhancers. Recently, the community has been increasingly drawn to newer collections of driver lines such as the Janelia FlyLight [7, 8] and Vienna Tiles (VT) lines [9] that are generated with a modular design: optimized vectors, defined cis-regulatory sequences, and characterized genomic loci for PhiC31-mediated targeted integration. The expression patterns of most FlyLight and VT lines are visualized and searchable, which makes it easier to find useful drivers expressing in brain regions of interest and greatly facilitates the dissection of neural circuits.

Binary expression systems other than GAL4/UAS are also frequently used to allow independent expression of different transgenes in different neuron populations. The LexA/LexAop system introduces both bacterial transcription repressor LexA’s DNA binding domain fusing the viral transcription factor VP16 acidic activation domain (VP16AD) or the GAL4 trans-activating domain (GAD) or human p65 activation domains (p65AD) and LexA binding sequence LexAop into flies [8, 10]. The QF/QUAS system borrows the transcriptional activator QF and its binding sequence QUAS from Neurospora crassa [11, 12].

Spatiotemporally Refining Genetic Targeting

There are about 100, 000 neurons in the fly nervous system. Only in rare cases do drivers label sparse or even a single pair of neurons. Genetic strategies have been developed to narrow down the expression of drivers and make it possible to study brain functions at single-cell resolution.

One strategy takes advantage a repressor of the transcription activator. GAL80 binds to the GAD and inhibits GAL4 action on UAS, thus independently targeting GAL4 and GAL80 allows expression of the UAS-transgene in neurons expressing GAL4 but not GAL80, achieving a narrowed set-difference expression pattern [13] (Fig. 1B). elav-GAL80 [14] represses the expression of GAL4 pan-neuronally and is often used to ascertain that GAL4-driven expression of the effector, in neural tissue indeed, is responsible for the observed effect; tsh-GAL80 [15] can mask the expression of GAL4 in most ventral nerve cord (VNC) neurons, and is useful for differentiating neurons in the brain from those in the VNC in some cases. The temperature-sensitive and ubiquitously expressed GAL80 (tubP-GAL80^ts) is used to achieve temporal control over transgene expression by transferring the restrictive 18°C to permissive 30°C to relieve GAL8 repression, e.g., to have transgene expressed only during the adult stage to circumvent developmental effects [3, 16]. In the QF/QS/QUAS tertiary expression system, suppression of QF by its repressor QS can be relieved by feeding quinic acid to flies, allowing spatiotemporal control of the transgene expression [11].

Another strategy takes advantage of reconstitution of a functional transcription factor from its split parts. Split-GAL4 fuses the complementary DNA-binding domain (DBD) and transcription-activation domain GAD to each of a pair of heterodimerizing leucine zippers (Zip⁻ and Zip⁺), to allow them to reconstitute functional GAL4. Transgenes under UAS control are expressed only in cells expressing both DBD and GAD (and thus reconstituted GAL4). Targeting the DBD and GAD independently produces intersectional expression of two drivers [17] (Fig. 1C). The orthogonal modular design of constructs for FlyLight facilitates the generation of split-GAL4 lines with the same regulatory sequences and thus mostly reproducible expression patterns of GAL4 lines. Recently, as more split-GAL4 lines have been characterized [18], systematic dissection of neuronal subtypes in various complex brain areas such as the lamina [19], lobula complex [20], mushroom bodies output neurons [21], and the sexually dimorphic P1 neurons [22] has become feasible. A split-LexA system that independently targets the DNA-binding LexA and trans-activating VP16AD has also been developed to allow intersectional expression [23].

Combining split-GAL4 and GAL80 can further narrow down the expression of transgenes. However, while GAL80 represses GAD-mediated transcription, it does not effectively inhibit split-GAL4 that uses p65AD or VP16AD as transcriptional activators. Killer Zipper (KZip⁺), a dominant-negative repressor of split-GAL4 that disrupts the formation of the split-GAL4 heterodimer, has been introduced to allow targeted expression in neurons expressing both parts of split-GAL4 but not GAL80 [24].

Stochastic Labeling by Site-Specific Recombination

Recombination-mediated excision of functional cassettes in the transgene adds a separate layer of control over the binary expression systems. The yeast recombinase Flippase (FLP) catalyzes mitotic recombination between the FRT (flippase recognition target) sequences on homologous chromosomal loci. FLP/FRT is introduced into Drosophila [25, 26] for the induction of stochastic labeling by inserting FLP-excisable FRT cassettes into effector constructs such as the UAS-transgene and targeted recombination. Directed flanking FRTs can generate “flip out” excision of the flanked cassette while oppositely-directed flanking FRTs can generate inversion of the flanked cassette. An FRT-flanked cassette with a transcriptional terminator (FRT-stop) inserted after UAS and just before the transgene is often used to “flip in” the transgene in neurons with FLP, which excises the FRT-stop cassette, and narrows down expression of the transgene in neurons intersectionally defined by both independently-targeted GAL4 and FLP [27] (Fig. 1D). For example, FLP under control of the brain-restricted Otd promoter Otd-FLP has been used in combination either with tubP > GAL80 > (> indicates FRT site) or with UAS > stop > genes allows effector transgenes to be expressed in the brain [28–30]. Heat-shock-inducible FLP (hs-FLP) allows control over the timing and volume of recombination events which eventually results in varying degrees of labeling. Other site-specific recombination systems have also been introduced into Drosophila, such as Cre/LoxP from bacteriophage [31], KD/KDRT (KD-recombinase Target Recognition) from Kluyveromyces drosophilarum, R/RSRT from Zygosaccharomyces rouxii, and B3/B3RT (B3 recombinase Target Recognition) from Zygosaccharomyces bisporus [32].

MARCM (Mosaic Analysis with a Repressible Cell Marker) uses hs-FLP/FRT and the ubiquitously-expressed GAL80 to induce mosaic cells with transgene expression in targeted tissue without transgene expression [13]. MARCM affords sparse expression of reporters or effectors in a targeted neuron population, and allows study of the morphology and development as well as the behavioral functions of neural circuits at single-cell resolution [33].

FINGR (Flippase-induced Intersectional GAL80/GAL4 Repression) combines the FRT-stop with GAL80 to allow FRT-dependent “flip-in” of GAL80 to repress GAL4 expression with a collection of enhancer-trap Flippases [34] and allows more restricted targeting in neurons with GAL4 but without FLP expression.

Monitoring Neuronal Activity with Genetically-Encoded Ca²⁺ Indicators

Intracellular Ca²⁺ is a widely-used indicator for neuronal activity. Various genetically-encoded indicators have been developed, based on fluorescent proteins and Ca²⁺-binding proteins such as Calmodulin (CaM), allowing activity to be monitored in a targeted neuronal population.

Monitoring Transient Ca²⁺ in Neurons

Transient Ca²⁺ can be monitored using GCaMP, a single molecule GFP-based Ca²⁺ probe with the CaM and its receptor, the M13 fragment from myosin light chain kinase, attached to the C and N termini of a circularly-permutated EGFP (cpEGFP) [56]. Ca²⁺-binding to CaM induces a conformational change of CaM-M13 and thus the cpEGFP, resulting in a change in fluorescence intensity. It is widely used to image neuronal responses [57]. The latest version, GCaMP6, can even detect single action potentials in the soma [58]. Subcellularly-targeted GCaMP allows the compartmentalized imaging of Ca²⁺. Syt::GCaMP, which attaches GCaMP6 to the C terminus of Syt1, then locates to and detects Ca²⁺ influx specifically in presynaptic terminals [59].

Transient Ca²⁺ can also be monitored using Twitch-2C, a ratiometric fluorescent Ca²⁺ reporter that utilizes ‘Twitch’ sensors based on the C-terminal domain of Opsanus troponin C and fluorescence resonance energy transfer (FRET) [60]. A conformational change mediated by Ca²⁺ binding alters the FRET between two fluorophores (mTurqoise2 and cpCitrine174), resulting in a change in the ratio of cyan to yellow fluorescence, which can be monitored by ratiometric imaging [61].

Red versions of GECIs, such as R-GECO (based on mApple) [62] and RCaMPs (based on mRuby) [63], of which the recent versions jRCaMP1a, b or jRGECO1a [64] have sensitivity comparable to GCaMP6, provide an additional channel of imaging to allow dual-color imaging of separate neuronal populations when used together with GCaMP to simultaneously monitor two populations of neurons [65].

GECIs can be used for neuron imaging in acutely dissected live nervous system or head fixed animal. Flyception allows more temporarily precise real-time imaging in freely-moving flies through an imaging window with the head cuticle removed in a specially-designed arena [66].

Capturing Neuronal Ca²⁺ Activity in a Light-Defined Time-Window

CaMPARI (calcium-modulated photoactivatable ratiometric integrator) takes advantage of the photoconvertible green fluorescent protein (FP) that irreversibly converts to a red fluorescent species upon illumination with violet light. The construct attaches a circular permutation of a photoconvertible FP to CaM and M13 peptide to allow irreversible conversion from green to red fluorescence with an increase in Ca²⁺ and the simultaneous presence of UV light (405 nm). It is used to label the increased activity of targeted neurons during behavior by simultaneous UV photoconversion [67]. CaMPARI has recently been used to identify clock neuron classes that respond to changes in temperature [68], and verify the gustatory response of interneurons involved in bitter processing in affixed adult flies [69], and to monitor the response of dopamine neurons to odor [70] and octopaminergic neurons to behavioral interruption training in freely-behaving larvae [71].

Transcriptional Reporter for Ca²⁺ level

CaLexA (Ca²⁺-dependent nuclear import of LexA) takes advantage of a Ca²⁺-responsive transcription factor, nuclear factor of activated T cells, and fuses it to VP16AD and mutant LexA that lacks a nuclear localization signal. Sustained neuronal activation promotes nuclear transport of the LexA chimera and drives transgene expression under the control of LexAop [72].

TRIC (transcriptional reporter of intracellular Ca²⁺) combines split GAL4 parts with CaM and its targeting peptide CaMKII to reconstitute functional GAL4, contingent upon the Ca²⁺ level. By fusing GAD to CaM and DBD to CaMKII, Ca²⁺-binding facilitates the association of CaM with CaMKII and thus the reconstitution and nuclear transport of functional GAL4. Thus, the expression of reporter genes under GAL4/UAS control reflects a sustained neuronal activation [73].

CaLexA and TRIC are both useful for reporting chronic neuronal activation [74, 75]. A real-time CaLexA-LUC assay combining UAS-CaLexA with LexAop-LUC has revealed the activity of circadian neurons in intact freely-moving flies [76]. A similar Tric-LUC has also been developed [77].

Optogenetics Controls Neuronal Activity with Light

Optogenetics tools use light-sensitive ion channels to achieve manipulation of neuronal activity with high temporal resolution. Neurons can be activated by blue light with Channelrhodopsin-2 (ChR2), a blue light (470 nm)-sensitive nonselective cation channel derived from the green alga Chlamydomonas reinhardtii [95], However, blue light has a visual effect on flies and does not penetrate the cuticle well. The side-effects can be circumvented by using red light. ReaChR (Red-activatable Channelrhodopsin) is derived from ChR1 in Chlamydomonas reinhardtii with a red-shifted spectral sensitivity that allows neuronal activation by red light (590 nm–630 nm) [96]. CsChrimson, derived from red light (590 nm)-responsive ChR1 in the red alga Chlamydomonas noctigama, has been introduced into Drosophila to activate neurons with red light [97].

The Cl⁻ channel Halorhodopsin from the archaebacterium Natronomonas pharaonis (NpHR) is able to suppress neuronal firing upon activation by yellow light (570 nm) [98–100]. The theta anion Channelrhodopsins from algal Guillardia (GtACRs) are have been introduced into flies for optogenetic silencing of neurons by cyan (GtACR1, at 515 nm) or blue light (GtACR2, at 470 nm) [101].

Thermogenetics Controls Neuronal Activity with Temperature

Neuronal function can be modulated by temperature in the poikilotherm Drosophila with temperature-sensitive molecules. Neuronal transmission can be temporally blocked by temperature-sensitive Shibire^ts1 at temperatures above 29°C [88]. Neurons can be activated by the thermosensitive cation channel TrpA1 if shifted from the restrictive 22°C or below to above 25°C [94].

Tracing Neuronal Processes

Neuronal projection of sparsely labeled neurons can be easily visualized by fluorescent reporter. However, neuronal projection within a neural plexus is difficult to discern, which can be relieved by the PA-GFP (photoactivatable GFP), a GFP variant that greatly increases its fluorescence with 488 nm excitation light after intense 413 nm radiation. With targeted expression of PA-GFP in neurons [103], light treatment at specific loci harboring part of the neuronal processes can convert the PA-GFP to its high-fluorescence form and diffusion of the converted GFP allows the projection pattern of a subset of targeted neurons to be highlighted [92, 93]. Although it is invasive to achieve precise light treatment, PA-GFP allows higher spatial specificity than that permitted by genetic targeting, and can reveal connections to discrete brain regions using a broadly-expressed PA-GFP [104].

Trans-neuronal Tracing

To reveal the circuitry in which neurons perform functions, it is necessary to map the downstream and upstream neurons. The relationship between two populations of neurons can be grossly addressed by simultaneously expressing different markers in each to see if overlap or contiguity exists in their expression patterns. To systematically map trans-synaptic connected neurons, new strategies have been developed to label the downstream neurons of target neurons.

trans-Tango allows anterograde trans-synaptic tracing, based on the Tango assay in which a synthetic signaling pathway consisting of two fusion proteins converts the activation of a cell-surface receptor such as G protein-coupled receptor to reporter expression via site-specific proteolysis. With trans-Tango, ligands presented at presynaptic sites in targeted neurons (labeled by UAS-myrGFP) trans-synaptically activate receptors and release membrane-tethered QF to drive expression of the QUAS-controlled reporter in postsynaptic neurons via ubiquitously-expressed Tango that uses the QF/QUAS to drive a marker (QUAS-mtdTomato) in the receptor cells [105] (Fig. 3a). This has been used to trace the APDN-TuBusup-EB circuit that regulates sleep-wake arousal [65, 106], and the connections revealed by trans-Tango can be functionally verified with UAS-driven P2X2 and QUAS-driven GCaMP [65].

Open in a separate window

Fig. 3

Methods for anterograde trans-synaptic tracing. A Trans-Tango allows anterograde tracing based on pan-neuronally-expressed Tango of human glucagon (hGCG) and presynaptically targeted expression of the glucagon. The trans-Tango is composed of two chimera proteins: pan-neuronally expressed hGCGR::TEVcs::QF tethers the transcription activator QF to the membrane via linking it to the human glucagon receptor (hGCGR) with a linker containing the cleaving site for N1a protease from the tobacco etch virus (TEVcs); hArr::TEV allows recruitment of TEV to activated hGCGR via human β-arrestin2, and one QUAS reporter. Overall, GAL4/UAS targeted expression of hGCG::hICAM1::dNRXN1 presents hGCG at presynaptic sites via the synaptic protein Neurexin1 (dNRXN1) and extracellular domain of human cell adhesion molecule (ICAM1), and activates hGCGR in postsynaptic neurons and recruits hArr::TEV to cleave QF off the membrane to translocate into the nucleus and initiates reporter expression under QUAS control. B TRACT allows anterograde tracing of neuronal circuitry by ligand-induced intramembrane proteolysis. TRACT is composed of CD19 presynaptically localized with nSyb in neurons targeted by LexA/LexAop, and the pan-neuronally-expressed fusion protein nlgSNTG4 that tethers transcription factor GAL4 and CD19 antibody ID3 to the synaptic sites of cell membranes via the intracellular domain of neuroligin (NLGN), the Notch regulatory region (NRR) and transmembrane domain of Notch. The presynaptic ligand CD19 binding to postsynaptic receptors ID3 partially unfolds the NRR to allow for cleavage of the S2 site by endogenous metalloproteases, and further allows cleavage (S3) by γ-secretase-mediated intramembrane proteolysis of the intracellular domain containing GAL4 which then translocates to the nucleus to activate the transcription of reporter genes such as GFP under UAS control.

TRACT (TRAnsneuronal Control of Transcription) can also label anterograde trans-synaptic neurons based on ligand-induced intramembrane proteolysis. Drosophila n-Synaptobrevin (n-Syb)-fused artificial ligands in LexA-labeled neurons are presented on the presynaptic membrane, and the trans-synaptic interaction with an artificial receptor induces intramembrane proteolysis in post-synaptic neurons to liberate a membrane-bound GAL4 variant, and ultimately drives the expression of a UAS-controlled reporter [107] (Fig. 3B). In principle, combined use of TRACT and trans-Tango allows simultaneous labeling of the primary (UAS-driven reporter) and secondary (QUAS-riven reporter) postsynaptic neurons of a LexA driver.

Revealing Potential Synaptic Contacts Between Neuronal Populations

As synapses are beyond the resolution of the ordinary light microscope, apparent contacts of neurons are not necessarily synaptic connections, so genetic methods have been developed to aid in marking putative synaptic connections.

GRASP (GFP Reconstitution Across Synaptic Partners) makes use of the trans-cellular complementation of split GFP parts (spGFP1-10 and spGFP11) to detect the membrane proximity of two cells. The spGFPs, neither of which is fluorescent alone, are separately expressed in the outer membranes of two populations of cells. If two adjacent cells are in enough proximity, the spGFP could reconstitute into a functional fluorescent GFP, revealing potential synaptic contacts (Fig. 4A). The initial version of GRASP used CD4 to present spGFPs to all the outer membranes of cells [108]. To improve the specificity of synaptic labeling, one of the two spGFPs was anchored to synaptic molecules to target synaptic sites. spGFP1-10 was fused to the transmembrane synapse protein Neurexin for better synaptic targeting [109]. Fusing one of the spGFPs to the C terminus of n-Syb (syb:GRASP) allows directional and activity-dependent GRASP labeling, since the presynaptic vesicle membrane n-Syb can only present the spGFP to the synaptic cleft after vesicle fusion to the cell membrane [110]. Recently, t-GRASP (targeted GFP Reconstitution Across Synaptic Partners) tested different pre-synaptic and post-synaptic/dendritic proteins for targeting both spGFPs, and further improved the synaptic specificity [111].

Open in a separate window

Fig. 4

Synaptic labeling. A GRASP detects membrane proximity permitting synaptic contacts between two neurons by the fluorescence of a reconstituted functional GFP from complementary spGFPs expressed in each of the neurons. B STaR allows Brp-fused tags with V5 expressed in FLP-mediated stochastically-labeled neurons (with myr-tdTomato) to mark the active zone, mimicking the endogenous pattern of Brp in a cell-type-specific fashion.

STaR (Synaptic Tagging with Recombination) allows inducible tagging of synaptic proteins to visualize synapses in the processes of sparsely-identified neurons, to aid in identifying synapses for specific neurons using ordinary fluorescent light microscope. It is hardly possible to discern dense synaptic puncta in neuronal processes of small size using antibodies against synaptic proteins, which are usually broadly expressed. STaR uses bacterial artificial chromosomes to harbor an engineered genomic region of synaptic protein with a tag sequence separated by a recombination recognition sequence flanked by a stop sequence that can be excised by recombinase, so the induced expression of tagged synaptic proteins display a level and spatiotemporal pattern similar to that of the endogenous synaptic protein. STaR of Brp, a component protein of presynaptic active zones, with a V5 tag, allows FLP-mediated stochastic expression of the V5-tagged Brp to label active zones [112] in defined neurons that are labeled with the LexAop-driven transgene with a co-translating 2A-LexA cassette [113] downstream of the V5 tag (Fig. 4B). Using GFP instead of V5 also allows labeling of active zones of specific neurons in live animals, and can be used to monitor structural synaptic plasticity [114–116]. STaR also allows co-labeling of presynaptic protein and postsynaptic receptors, thus revealing the contact sites of synaptic partners [112].

This review was supported by the National Natural Science Foundation of China (6531000063, 31571093, 31622028, 31471063, and 31671074), the Science Foundation of Jiangsu Province of China (BK20160025), and Fundamental Research Funds for the Central Universities, China (2242016R20028 and 2017FZA7003).