Normal Dendrite Growth in Drosophila Motor Neurons Requires the AP-1 Transcription Factor

Drosophila Stocks, Culturing, and Genetics

Flies were reared on standard corn meal medium at 25°C and were grown in 12 h light-dark cycles. The C380-GAL4 line has been described previously (Budnik et al., 1996) and expresses in most motor neurons, plus other unidentified neurons. The Cha-GAL80 (Choline-acetyl transferase promoter driven GAL4) line was obtained from Dr. Toshihiro Kitamoto, and has been shown to suppress GAL4 activity in cholinergic neurons (Kitamoto, 2002). The eve^RRK-GAL4 line and related eve^RRA-GAL4 line (RRA-GAL4) were obtained from Miki Fujioka and expresses in aCC and RP2 motor neurons (Fujioka et al., 2003). UAS-GFP, UAS-FLP, act-FRT-GAL4, and kayak alleles were obtained from the Drosophila stock collection in Bloomington. UAS-Fbz, UAS-Jbz, UAS-Fos, and UAS-Jun have been described previously (Eresh et al., 1997) and were obtained from Marianne Bienz. UAS-eag(DN) flies were from Dr. Ralph Greenspan (Broughton et al., 2004), UAS-Sh(DN) (Mosca et al., 2005), and UAS-Sh[act] (EKO) (White et al., 2001) were from Dr. Haig Keshishian and UAS-Fos-RNAi was from Dr. Dirk Bohmann (Uhlirova and Bohmann, 2006). A recombinant including both UAS-Sh(DN) and UAS-eag(DN) was made and termed “Electrical Knock In (EKI)”. UAS-nod-lacZ flies were obtained from Dr. Y.N. Jan (Clark et al., 1997) and UAS-Kinesin-GFP flies were made by Patty Estes in the Ramaswami laboratory.

Immunohistochemistry, Imaging, Sholl Analysis, and 3D Reconstruction

Tissue Culture Techniques and High K⁺ Protocols

Cell cultures were prepared as described previously (Kraft et al., 1998) with a few modifications. All dissection equipment was soaked in 70% EtOH for 10 min before dissection. Larvae (C380-GAL4, UAS-CD8GFP; ChaGAL80 crossed with either Oregon-R to obtain labeled control motor neurons or with various transgenic strains) were selected and sterilized by placing the animal in 70% EtOH for 1 min. The larvae were rinsed three times in sterile water before the CNS was removed in sterile Drosophila medium (Schneider’s medium + 10% fetal bovine serum + 50 µg/mL insulin). The cephalic lobes were discarded and only the thoracic plus abdominal ganglia were used in cultures. The ganglia were micro-dissected to remove ventral lying GFP positive neurons that have not been identified. The ganglia were enzymatically treated in liberase for 1 h at room temperature (10 µL 2× liberase stock/1 mL Rinaldini’s saline [800 mg of NaCl + 20 mg of KCl + 5 mg of NaH₂PO₄*H₂O + 100 mg of NaHCO₃ + 100 mg of glucose + 100 mL of H₂O]). The ganglia were rinsed twice in 1 mL Drosophila Medium, triturated with a fire polished glass Pasteur pipette, further triturated with an Eppendorf tip and plated on Con-A/Laminin coated dishes in a 100 µL bubble. The cells were plated at a density of one micro-dissected ganglion/dish. The dishes were flooded 24 h later with 1 mL Drosophila Medium and imaged 24 h after flooding. The cells were maintained in a 25°C, humidified incubator.

The culture dishes were made by drilling an 8-mm hole in 35-mm Petri dishes (Corning 430165). A 12-mm round coverslip (Bellco 1943–10012) was then glued under the Petri dish with Sylgard 184 to create a well. The dishes were placed under UV light for 2 h to sterilize. The well was coated with 70 µL of ConA (Sigma C-2010) + laminin (Becton-Dickinson 354232) + milliQ H₂O and incubated for 2 h at 37°C. Dishes were then rinsed with 6 mL sterile H₂O and used within 1 month.

The motor neurons in the high K⁺ experiments were fed medium with a 10 mM increase in [K⁺] from 22 mM in normal Drosophila culture medium to 32 mM by the addition of KCl. This is predicted to depolarize the neurons by about 10 mV. Control cells for the high K⁺ experiments were fed medium with the osmolarity adjusted to match the treated cells. As for all experimental manipulations, treated and control cultures were paired such that cells obtained from the same dissociation were divided into treated and control dishes. This minimized variability due to animals, dish coating, media and the dissociation procedure itself. Similarly, comparisons among neurons of different genotypes involved cultures that had been prepared together.

Cycloheximide was administered to cultures to inhibit protein synthesis (Mizel and Bamburg, 1976) by adding 1 × 10⁻² mg/mL to the Drosophila medium. Plectreurys Toxin (PLTX-II, 5 × 10⁻² µM, Alomone Labs) were administered to the cultures as previously described (Jiang et al., 2005). The pharmacological agent was administered to neurons after 24 h in vitro, and the cells were imaged 24 h later. A live cell assay was performed on cycloheximide and high K+ exposed cells to test their viability using the LIVE/DEAD Viability/Cytotoxicity Kit (Molecular Probes, L-3224). The culture media was washed with 500 mL of PBS and then placed in 4 M EthD-1 solution in PBS and incubated at room temperature for 20–40 min.

To selectively visualize motor neurons in culture, the C380-GAL4 driver (Budnik et al., 1996; Sanyal et al., 2003) was combined with a choline-acetyl transferase (cha) GAL-80 construct (Kitamoto, 2002) to suppress GAL4 activity in cholinergic neurons. This restricted GAL4 expression to a small population of dorsal motor neurons [Fig. 1(B)] and a few unidentified ventral neurons that were micro-dissected out prior to cell culture preparation. To determine which specific dorsal motor neurons were labeled with GFP, rhodamine-phalloidin staining of body wall muscles was used to visualize the peripheral innervation pattern of these neurons [Fig. 1(A)]. Muscles 1–4, 6, 7, 9, 10, 13, 15, 16, 19, and 20 were innervated in each hemisegment. We conclude that MN1-Ib (aCC; ipsilateral axon with contralateral dendritic tuft), MNISN-1s (RP2; ipsilateral axon), and MNSNb/d-1s (RP5; contralateral axon) are labeled in this line based on their central morphology and prior identification of motor neuron muscle innervation patterns (Choi et al., 2004). The strongest GFP expression was consistently observed in MNISN-1s, which is also known as RP2. For analysis, fluorescent living cells, not in contact with other neurons, were imaged at 2 days in vitro and the total neurite length, total branch number, and higher order branching patterns measured through digital reconstruction of the neurons [Fig. S1(G,H)]. Because one thoracicoabdominal ganglion was dissociated and plated per culture dish, there was a maximum of about 60 labeled motor neurons per dish. In practice, about 20 labeled motor neurons were present in each culture dish.

Motor neurons maintained a glutamatergic identity in culture, as they stained with a Drosophila vesicular glutamate transporter antibody [Fig. 1(E,F)], α-VGlut (Daniels et al., 2004). They were also polarized since the microtubule minus-end directed reporter, nod-LacZ (Clark et al., 1997), and the plus-end directed kinesin-GFP (P. Estes and M. Ramaswami, unpublished), fusion proteins were transported selectively when expressed by UAS transgenes in cultured motor neurons [Fig. 1(C,D)]. In greater than 90% of the neurons, kinesin-GFP was localized within a single primary process (i.e., a neurite that emerged directly from the cell body), whereas the remaining primary processes were marked by nod-lacZ. Similar segregation of Kinesin and Nod fusion proteins, to axons and dendrites respectively, is observed in vivo within Drosophila sensory neurons (Clark et al., 1997), mushroom body Kenyon cells (Lee et al., 2000), motor neurons (Sanchez-Soriano et al., 2005), and also within cultured Kenyon cells (Kraft et al., 2006).

Analysis of Cultured Motor Neurons

The living cells were imaged with a Hamamatsu c4742-96 camera using Wasabi software version 1.4 on a Nikon inverted scope at 60×. Both phase/contrast and fluorescent images were obtained for each cell. The motor neurons had to meet the following criteria in order to be included in the analysis: alive and healthy with no signs of lifting off the dish, bright GFP signal representing a high level of UAS expression, and isolation from neurite interactions with other neurons. The Tiff image files were analyzed by tracing the fluorescent image by hand on a Wacom tablet PC. The parameters of total neurite length, total branch number, higher order branching patterns, and cell body circumference were recorded using Simple PCI software. All data were plotted and analyzed using SigmaPlot (Jandel Scientific) and Microsoft Excel. Statistical significance was determined using unpaired t-tests; an asterisk represents p < 0.05. All data are represented as mean +/−SEM.

Electrophysiology

Cultures were prepared as described from third instar larvae. Whole cell current or voltage clamp recordings were obtained from GFP-labeled motor neurons with patch pipettes that were pulled from borosilicate glass (Friedrich and Dimmock, Millville, NJ). Similar results were obtained using different driver lines, although as indicated in the text, the baseline level of excitability varied consistently among lines. Therefore, for each of the transgenic manipulations described, comparisons are made with the appropriate driver line as control. An Axopatch 1D amplifier, Digidata 1320 interface, and PClamp software (Axon Instruments, Sunnyvale, CA) were used for data acquisition and analysis. The extracellular solution included (mM): NaCl, 140; KCl, 3; CaCl₂, 1; MgCl₂, 4; HEPES, 5. The intracellular pipette solution (modified from Jiang et al., 2005), included: potassium gluconate, 120; NaCl, 20; CaCl₂, 0.1; MgCl₂, 2; HEPES, 10; EGTA 1.1. Whole cell recordings from RP2 motor neurons in vivo were obtained using previously published methods (Choi et al., 2004). The intracellular solution also included a low concentration of rhodamine-dextran, which allowed positive identification of motor neuron identity based upon dendritic morphology.

Direct Measurement of Motor Neuron Dendritic Fields in the Central Nervous System

We adapted a strategy used previously in developmental studies to reproducibly label individual or small subsets of motor neurons optically isolated from their neighbors in the CNS.

The eve-GAL4^RRK line expresses in aCC and RP2 neurons in the dorsal medial motor neuron cluster (Fujioka et al., 2003). The expression of this GAL4 line is strongest in embryos, and by larval stages a single copy of this transgene does not drive GFP above levels of detection. We combined this GAL4 line with three other transgenes; UAS-Flp (a source of the recombinase, flippase), an actin promoter separated from a GAL4 by a spacer DNA fragment flanked by two FRT sites (target sites for the Flp recombinase), and a UAS-mCD8-GFP reporter [Fig. 2(A)]. Larval central nervous systems from our test line showed a mosaic pattern of GFP expression, owing to chance flip-out events in aCC and RP2 neurons during early embryonic stages [Fig. 2(B)]. RP2 and aCC neurons can be easily distinguished based on their dendritic field [Fig. 2(C,D)] (Choi et al., 2004) and peripheral innervation [Fig. 2(E–H)]. We selected segments where RP2 dendrites (owing to their simpler morphology) can be imaged cleanly without conflicting labeling from other neurons in adjoining segments [Fig. 2(B), I; see materials and methods]. After generating a series of images representing the entire thickness of the motor neuron using laser scanning confocal microscopy, we reconstructed the dendritic field using AMIRA (Voltex function). We then created an external bounding surface that tightly enclosed all the dendrites [Fig. 2(I)]. The internal volume contained in this shell was consistent across samples and sensitive enough to detect small changes in dendritic outgrowth (roughly 10%). These results demonstrate the utility of this simple technique which allows for the measurement of dendritic fields in 3D space.

A key advantage of this scheme is that it allows the expression of transgenes of interest in selectively labeled motor neurons. Thus, Drosophila Fos (dFos, kayak) expression via this “flipout” system causes robust overexpression of Fos in all GFP-labeled neurons [Fig. 2(J–L), arrowheads]. This allowed us to test the effect of altering protein function in targeted neurons without changing the surrounding milieu [compare Fos staining intensity in neurons overexpressing AP-1 (arrowhead) with neuron displaying endogenous lower levels of Fos (arrow) in Fig. 2(K)]. This is particularly relevant since loss-of-function mutations in both Fos and Jun often result in pleiotropic phenotypes and are typically lethal. Selectively altering protein function in these neurons largely overcomes these experimental limitations.

AP-1 Dependent Transcription Modulates Dendrite Growth

AP-1 is normally expressed by larval motor neurons [note detectable nuclear staining marked by an arrow in non-GFP expressing neurons in Fig. 2(K)]. To test if Drosophila motor neuron dendrites are regulated by AP-1 dependent transcription, we asked how dendritic growth was affected following inhibition or overexpression of AP-1 components (Fos and Jun) in RP2 neurons. Figure 3(A) summarizes this set of experiments and shows that AP-1 positively regulates dendrite growth in vivo. Although overexpression of AP-1, but not Fos or Jun alone, increased dendrite volume by 92%, inhibition of AP-1 by expressing a dominant-negative Fos or Jun (the bZip domain of either dFos or dJun; Fbz or Jbz) significantly reduced dendrite volume as compared to control animals. The nervous systems remained functional in these animals. Chronic pan-neuronal expression of Fbz or Jbz did not noticeably affect action potential production or synaptic transmission in the CNS (see supplementary data and Fig. S2).

To compare our method of volume measurement with well established techniques for assessing neuronal branching, we performed Sholl analysis (Sholl, 1953) on images gathered from three test genotypes, wild-type controls, animals expressing wild-type AP-1 protein in RP2 neurons, and those expressing dominant-negative Jun (Jbz) protein [Fig. 3(C)]. These measurements suggest that although maximum projections of these motor neurons do not allow easy reconstruction of dendrite branches, we are conservatively able to detect significant differences between test genotypes. Thus, expression of wild-type AP-1 proteins in these neurons led to an expansion of the dendritic field [Fig. 3(D)]. Conversely, AP-1 inhibition using the bZip only dominant-negative Jbz construct produced fewer dendrites.

As shown in Figure 3(A), established and newly isolated loss-of-function alleles of dFos show significantly reduced RP2 dendritic fields. It is noteworthy that both kay² and kay⁰⁸²³² are lethal alleles and affect motor neuron dendrites when heterozygous over a wild-type Fos allele [Fig. 3(A)]. kay¹²⁷¹⁰, though inserted close to the Fos gene, is not expected to affect Fos expression since the insertion is far upstream of the Fos transcription start site. Finally, expression of a validated Fos-RNAi construct (Uhlirova and Bohmann, 2006) in RP2 neurons also results in dendritic fields that are 41% of controls [Fig. 3(A)]. Taken together these results strongly suggest that endogenous AP-1 mediated transcription influences dendritic morphology in motor neurons.

The cell autonomy of AP-1 function for regulating dendritic plasticity, indicated by in vivo analyses of mosaic animals, is supported by parallel studies in vitro, which additionally allow direct measurements of branch number in AP-1 perturbed neurons. Results from overexpressing or inhibiting AP-1 in cultured motor neurons are consistent with in vivo findings Figure 3(B). Thus, AP-1 increased total neurite length by 80% while Fbz strongly reduced neurite length. Expression of Jbz did not reduce total neurite length significantly in this assay. We speculate that this is probably due to either insufficient inhibition achieved with the C380-GAL4; Cha-GAL-80 system that was used to drive expression for cell culture experiments as compared to the strong actin-GAL4 in vivo, or due to a weaker dominant-negative effect of the Jbz protein. Both branch number and total length were increased by AP-1 perturbation, suggesting AP-1 dependent changes in growth and complexity of neurons in vitro [Fig. 3(B), gray bars]. These results suggest that effects of AP-1 on dendrite growth derive from autonomous action in motor neurons, and are not dependent for instance, on altered retrograde signaling from postsynaptic cells.

Experimental Strategies to Influence Neural Activity in Motor Neurons In Vivo and In Vitro

AP-1 expression is commonly induced by increased neuronal activity. To develop models of activity driven dendrite growth and test if AP-1 is required in this paradigm, we used composite GAL4 lines mentioned earlier (RN2-flipout, RRA-GAL4, and C380-Gal4; ChaGal80) to drive transgenic modifiers of neural excitability or activity in GFP labeled motor neurons in vivo and in vitro. We attempted to increase motor neuron activity by coexpressing both Shaker and ether-a-go-go dominant-negative transgenes (UAS-Sh(DN),UAS-eag(DN) which reduce potassium currents (Broughton et al., 2004; Mosca et al., 2005). The Shaker dominant-negative transgene causes increased branching of motor axon terminals when expressed in an ether-a-go-go mutant background (Mosca et al., 2005), as previously reported in Shaker/ether-a-go-go double mutants (Budnik et al., 1990). For our experiments we used a recombinant chromosome termed “EKI” (for electrical knock-in), which included both transgenes.

Identified motor neurons showed increased excitability in EKI larvae. Motor neurons recorded in situ from third instar larvae express sodium, potassium and calcium currents (Worrell and Levine unpublished observation; 2007, SFN Abstract). The resting membrane potentials of EKI and control motor neurons were not different (EKI: −61.8 mV; n = 8; Control: −61.5 mV; n = 8). Neither control nor EKI expressing motor neurons produced action potentials spontaneously during recordings. Upon injection of depolarizing current, however, TTX-sensitive action potentials were produced. Motor neurons in EKI larvae had similar action potential thresholds but significantly higher action potential frequency in response to the injection of positive current, as compared to the same identified (RP2 and aCC) motor neurons in control larvae [Fig. 4(C,D)]. The postspike hyperpolarization was smaller in EKI larvae, often resulting in spike doublets. In addition, the characteristic delay to first spike, which is a consistent feature of RP2 (Choi et al., 2004), was reduced in EKI larvae. RP2 and aCC motor neurons expressing two copies of the “electrical knock-out” transgene (EKO) (White et al., 2001), encoding a modified Shaker channel lacking the inactivation domain, had similar resting membrane potentials to control motor neurons (EKO: −59.7 mV; n = 7; Control: −59.6 mV; n = 5). The action potential thresholds of EKO motor neurons were also similar to control. In general, fewer action potentials were evoked with the injection of depolarizing current into EKO expressing motor neurons, but the differences from control were not significant [Fig. 4(A,B)]. Note that different driver lines were used for the EKI and EKO motor neuron physiology experiments to facilitate GFP labeling (see Figure legend).

To confirm that the effects of EKI and EKO transgene expression reflected intrinsic motor neuron excitability rather than altered synaptic drive, whole cell recordings were also obtained from isolated larval motor neurons in cell culture. The resting membrane potentials of the control motor neurons were somewhat depolarized and variable in vitro, relative to those measured in vivo, probably reflecting damage inflicted by recording from the fragile cells. The mean resting potential for control motor neurons was −27 mV (range, −20 to −37 mV; n = 18). The resting potentials cultured motoneurons of transgenic animals were similar to control (EKI: mean, −21 mV; range −20 to −35 mV, n = 11; EKO: mean, −23 mV; range −10 to −44 mV, n = 15). In voltage-clamp experiments cultured motor neurons from control and EKI larvae expressed potassium currents at 2.5 and 4 days in vitro [Fig. 5(A)]. In both control and EKI motor neurons unclamped, TTX-sensitive events were observed after 4 days in vitro, with the frequency of these events being higher in EKI motor neurons [Fig. 5(A), arrows]. As with control motor neurons, no action potentials were produced spontaneously in EKI motor neurons at 4 days. Unlike control, however, in which none of the motor neurons produced action potentials when depolarized in current clamp mode at day 2.5, 33% of the EKI motor neurons recorded at 2.5 days (n = 24) already produced action potentials. By day 4 in vitro, 91% (n = 11) of the EKI motor neurons were capable of spiking as compared to 60% (n = 18) of the control cells. By contrast, EKO caused the expression of large voltage-dependent potassium currents in motor neurons [Fig. 5(A)]. None of the EKO-expressing motor neurons produced action potentials at 2.5 days in vitro, and only 27% (n = 15) were capable of spiking at 4 days.

Neural Activity Stimulates Motor Neuron Dendrite Growth and Requires AP-1 Function

Given that motor neuron excitability could be influenced as predicted by transgene expression, we next tested whether the effects of AP-1 on dendritic growth in vivo could be mimicked by EKI expression. The dendritic field volumes of RP2 motor neurons were significantly increased in vivo in EKI larvae [Fig. 6(A)]. By contrast, expression of EKO did not influence dendritic growth significantly [Fig. 6(A)]. To help confirm that the effects of EKI expression reflected a change in intrinsic cell excitability, parallel experiments were performed in cell culture. Motor neurons cultured from EKI larvae displayed increased growth and branching in vitro [Fig. 5(B)]. As with the cells measured in vivo, however, EKO expression did not alter growth or branching of cultured motor neurons [Fig. 5(B)].

An alternative strategy that is commonly used to depolarize cultured neurons is to increase the potassium concentration in the culture medium (Aizawa et al., 2004; Green et al., 2007). Larval motor neurons in cultures that were exposed to elevated extracellular K⁺ for 24–48 h in vitro (10 mM increase in [K⁺]), remained viable [supplementary Fig. S1(A–F)] and increased neurite length and branching significantly as compared to control cells maintained for 48 h in normal medium [Fig. 5(B–D)]. Administration of the translational inhibitor cycloheximide to motor neuron cultures reduced both basal and induced levels of neurite growth, indicating the necessity for protein synthesis in motor neuron growth and branching in vitro [Fig. 5(C)]. It is likely that the induced growth reflected the dendritic compartment of motor neurons in vitro. Whereas the number of primary processes arising from the cell body was increased, the number of neurons with more than one kinesin-labeled primary process was not. Moreover, there was no increase in the total length of kinesin labeled primary processes and their higher-order sub-branches, whereas the total length of the non-kinesin labeled processes increased with high K⁺ exposure [Fig. 5(D)].

Results reported in the previous sections demonstrate separately that dendritic growth in Drosophila motor neurons is responsive to neural activity and requires AP-1 function. Is the activity-driven component of dendrite growth dependent on AP-1? A model in which AP-1 is a major mediator of activity-dependent growth makes three predictions. First, neural activity dependent growth should be suppressed by AP-1 inhibition; second, elevation of AP-1 activity might promote growth even under depressed neural activity and third, AP-1 or its constituents, Fos and Jun might be rapidly regulated following changes in neural activity. We describe experiments below that test and confirm these predictions.

To ascertain if AP-1 is required for activity-dependent growth of motor neuron dendrites, we inhibited AP-1 function by coexpressing UAS-Fbz in neurons expressing eag (DN) and Sh (DN) channel subunits (EKI). Under normal conditions, EKI results in dramatic growth of dendrites in vivo, but AP-1 inhibition completely abolished this activity-induced growth [Fig. 6(A)]. Indeed, dendritic volume was statistically indistinguishable in EKI-Fbz and Fbz animals (Fbz: 61% ± 6 and EKI-Fbz: 68% ± 9, as compared to wild-type controls). The observed genetic epistasis of Fbz indicates that AP-1 function is essential for neural activity induced dendritic growth. In vitro experiments using high potassium stimulation also supported this conclusion [Fig. 6(B)]. Fbz blocked high-K⁺ driven dendritic growth. Jbz, a weaker inhibitor of AP-1, had a similar, though smaller effect on depolarization-induced growth. These results are consistent with the idea that neural activity stimulates AP-1 dependent transcription to promote dendrite growth.

Our working model also predicts that experimentally induced increases in neural activity may have no further effect on dendritic growth in AP-1 overexpressing neurons. To test this prediction, we asked how high potassium stimulation affected dendritic growth in AP-1 overexpressing motor neurons. Figure 6(B) shows that adding high potassium to AP-1 expressing motor neurons in vitro did not cause any additional growth. Thus, our data argue that AP-1 overexpression stimulates a physiologically relevant pathway normally recruited for activity-dependent dendritic expansion.

A potential mechanism by which neural activity might lead to increased AP-1 activity is through increased calcium entry. To test if increased calcium currents are responsible for dendritic growth seen in EKI expressing neurons, we inhibited voltage-gated calcium channels using Plectreurys toxin, PLTX. PLTX has been shown previously to inhibit some insect voltage-gated calcium channels (Jiang et al., 2005). We chose EKI over highK stimulated cultures for these experiments since we could dependably observe depolarization mediated changes in motor neurons both in vitro and in vivo. When added to normal cultures, PLTX inhibited neurite growth consistent with a role for calcium currents in mediating growth under basal conditions. Interestingly, PLTX completely prevented EKI mediated neurite outgrowth [Fig. 6(C)]. This suggests that elevation of calcium currents in EKI neurons is required for increased neurite outgrowth in vitro. We next tested if AP-1 expression can bypass the need for increased neural activity. Indeed, when AP-1 expressing motor neurons were treated with PLTX in vitro, they demonstrated robust outgrowth comparable to untreated AP-1 neurons (AP-1: 180%; AP-1+PLTX: 173% as compared to controls) [Fig. 6(C)]. Further, neurons in the same culture dish not expressing AP-1 (distinguishable due to lack of GFP expression) displayed reduced growth, similar to PLTX treated control cells. These results indicate that even under conditions of reduced calcium entry, AP-1 dependent transcription is capable of generating a genomic response that drives dendrite growth. Taken together, these results strongly suggest that EKI mediated depolarization of motor neurons increases calcium entry through PLTX-sensitive calcium channels, likely followed by the induction of signaling mechanisms that finally enhance AP-1 activity leading to dendrite growth. The essential elements of this model are depicted in Figure 6(F).

To explore potential mechanisms by which neural activity might influence AP-1 function, we tested levels of Fos protein in stimulated versus control neurons. We focused on Fos because it is a well-studied immediate early gene and our experiments consistently showed more severe effects of Fos inhibition both in vivo and in vitro. We raised and tested an antibody against a Drosophila Fos derived peptide. This antibody faithfully detects increased Fos protein achieved through GAL4 driven Fos overexpression. As shown in Figure 6(D), increased nuclear Fos is detected in dorso-medial motor neurons in the larval ventral ganglion (arrow) as expected from C380-GAL4 driven expression of UAS-Fos. Similarly, we were able to detect a band of the predicted molecular weight in western blots from adult head protein extracts. In these experiments, we also confirmed increased expression of Fos when AP-1 was expressed pan-neuronally using the C155-GAL4 driver. Next, using this antibody, we tested if acute induction of neural activity alters Fos levels in motor neuron nuclei. We stimulated cultured motor neurons with high potassium containing saline and measured Fos immunoreactivity following 0, 4, and 20 h of treatment. Figure 6(E) shows rapid induction of Fos within 4 h. Prolonged treatment with high potassium saline did not cause an increase in Fos staining, presumably indicating that protein levels return to baseline with maintained stimulation. These results support a model where neural activity promotes dendrite growth by upregulating Fos.

Motor Neuron Dendritic Growth Is Activity-Dependent

In the absence of prior demonstration, it was important to initially establish that motor neuron dendrites in Drosophila display robust changes in response to neural activity. We used both pharmacological and genetic manipulations to alter neural activity in vivo and in vitro. As shown in Figure 5(C,D), maintaining cultured motor neurons in high K⁺ medium enhanced the growth of neurites that have the characteristics of dendrites. Expression of dominant-negative subunits of the potassium channels eag and Sh, increased excitability and caused increased dendrite growth both in vivo and in vitro (Fig. 4–Fig 6). These experiments suggest that a conserved mechanism of plasticity operates in these motor neurons to regulate dendritic growth. Additionally, by developing the first assay for activity-stimulated dendritic growth in Drosophila, we now enable a wide range of experiments to further dissect underlying signaling and cell biological pathways.

It is worth noting that the cell culture approach allowed the comparison of acute depolarization (high potassium) with the more chronic alteration of excitability caused by the expression of eag and shaker dominant-negative transgenes. Despite the interesting possibility of compensatory changes following long-term transgenic manipulations, the functional properties of larval motor neurons in vivo were stably altered by reducing potassium channel function. By contrast, the effects of elevated potassium reflect an acute manipulation after cells are placed in culture. In spite of these differences the effects of EKI expression on cell growth in vitro were similar to those induced by elevated potassium. The lack of an effect on outgrowth by EKO expression may reflect a cellular compensatory mechanism. Although effects on cell physiology were detected, the difference in excitability measured in vivo was not significant. Two alternatives are that there is a basal level of normal growth that is not activity-sensitive, or that reduced activity had an influence on growth or branching that our analysis did not detect.

We suggest that the depolarization by high potassium or the increased excitability caused by EKI expression would lead to an increase in calcium influx through voltage-dependent channels (Berke and Wu, 2002; Jiang et al., 2005; Peng and Wu, 2007). The voltage-dependent properties of Eag and Shaker potassium channels (Wu et al., 1983), suggests that EKI expression would not necessarily be expected to alter the resting membrane potential. Even without altering the resting membrane potential, however, the reduction in voltage-dependent potassium currents might allow increased calcium influx in response to spontaneous depolarization or calcium waves (Jiang et al., 2005). This would be augmented, in vivo, by the higher action potential frequency evoked by depolarization of motor neurons in EKI larvae as compared to wild type. It is also possible that the expression of eag-DN alters a modulatory function of the potassium channel subunit (Zhong and Wu, 1993; Sun et al., 2004). Consistent with this hypothesis, we find that inhibiting PLTX-sensitive calcium channels prevents EKI mediated neurite outgrowth. A similar dependence of neuronal growth on calcium channels has been demonstrated convincingly in the Drosophila giant neuron culture system (Peng et al., 2007). Ultimately, it will be important to establish the specific parameters of activity and calcium flux that are essential for modulating the intracellular signals that mediate growth plasticity (Fields et al., 1997; Zhao et al., 2007).

Dendrite Growth Is AP-1 Dependent

We have shown previously that AP-1 regulates synaptic plasticity at Drosophila motor terminals (Sanyal et al., 2002, 2003). Our present results uncover a role for AP-1 in regulating dendritic growth. Reduction of AP-1 activity decreases, while enhancement of AP-1 increases, dendritic outgrowth in vivo and in vitro (see Fig. 3). These conclusions are further strengthened by the fact that known and predicted loss-of-function alleles of kayak (DFos) also reduce dendrite volume significantly, when present in heterozygous combinations with wild-type alleles. It is to be noted that these mutations in Fos are homozygous lethal at early developmental stages, and transgenic strategies for tissue specific genetic manipulations are especially useful in these contexts. Additionally, multiple Fos isoforms have been reported, and using a dominant negative construct offers a good way to inhibit these multiple Fos proteins (Giesen et al., 2003). In conclusion, the observations that (a) Fos is normally detectable in motor neuron nuclei, (b) Fbz, Jbz and Fos-RNAi inhibit and AP-1 increases dendrite growth, (c) loss-of-function alleles of kayak decrease dendrite volume, and (d) a variety of controls validate the genetic perturbations used in our study, strongly suggest that AP-1 functions physiologically to control dendrite growth.

Activity-Dependent Dendritic Plasticity Requires AP-1

To test if neural activity driven growth requires AP-1 dependent transcription, we inhibited AP-1 function in a background of elevated neural activity. Figure 6(C) indicates that neural activity induced dendrite growth is completely abolished in vivo by the coexpression of Fbz. Further, in vitro experiments with High K⁺ induced dendrite growth show that expression of either Fbz or Jbz substantially reduces the extent of depolarization induced growth. Our results also indicate that the extent of overgrowth seen through AP-1 induction cannot be increased any further by High K⁺ induced depolarization [Fig. 6(B)]. Furthermore, calcium channel inhibition by PLTX toxin does not preclude AP-1 driven neurite growth, suggesting that AP-1 functions downstream of neural activity and calcium entry to enhance neurite growth [Fig. 6(D)]. These results are consistent with the idea that AP-1 is a major contributor to activity-induced plasticity of dendrites.

Finally, to ascertain the mechanism by which neural activity might influence AP-1 function, we tested expression of Fos protein following acute induction of activity in cultured motor neurons. As depicted in Figure 6(E), Fos protein levels are increased by ~35% as compared to uninduced controls. This induction is rapid, occurring within 4 h, but transient, as Fos levels return to baseline during 20 h of stimulation. Although modest, this change in cellular Fos levels is consistent with previous observations (Morgan et al., 1987) and suggests, though it does not prove, a model where neural activity recruits AP-1 possibly through synthesis of new Fos protein, to promote dendrite growth. Hence, our results confirm that AP-1, positively regulates plasticity in postsynaptic compartments, as demonstrated earlier for presynaptic terminals (Sanyal et al., 2002), and establishes it as a key component in activity-dependent neuronal plasticity.

A key goal for future studies must be to determine how neural activity is translated into transcription factor activity (Fields et al., 1997; Zhao et al., 2007). Calcium dependent signaling is arguably the most important pathway for the activation of both CREB and Mef2A, two transcription factors that influence plasticity in opposite ways (Redmond et al., 2002; Flavell et al., 2006; Shalizi et al., 2006), as well as AP-1 (Zhao et al., 2007). A second question is that of interaction between transcription factors. Transcriptional regulation of Fos by CREB (and vice versa) has been described in some detail, but the temporal sequence in which these proteins function or their relative importance in plasticity have not been assayed rigorously in the same preparation (Liu and Graybiel, 1996; Kang et al., 2001; Sanyal et al., 2002; McClung and Nestler, 2003). Finally, further experiments to identify the downstream targets of these transcription factors, will enable the description of a composite transcription factor network that mediates protein synthesis during long-term neural plasticity.