Compartmentalized Calcium Transients Trigger Dendrite Pruning in Drosophila Sensory Neurons

See allHide authors and affiliations

Science  21 Jun 2013:
Vol. 340, Issue 6139, pp. 1475-1478
DOI: 10.1126/science.1234879

Dendritic Pruning

During metamorphosis, Drosophila sensory neurons eliminate their dendritic trees, but axons and soma remain intact. Kanamori et al. (p. 1475, published online 30 May) demonstrate that compartmentalized calcium transients in dendrites function as the spatiotemporal cue for pruning of unwanted branches. Such a localized calcium signal, induced by a local elevation of branch excitability, activates calcium-dependent proteinases and eventually causes branch death.


Dendrite pruning is critical for sculpting the final connectivity of neural circuits as it removes inappropriate projections, yet how neurons can selectively eliminate unnecessary dendritic branches remains elusive. Here, we show that calcium transients that are compartmentalized in specific dendritic branches act as temporal and spatial cues to trigger pruning in Drosophila sensory neurons. Calcium transients occurred in local dendrites at ~3 hours before branch elimination. In dendritic branches, intrinsic excitability increased locally to activate calcium influx via the voltage-gated calcium channels (VGCCs), and blockade of the VGCC activities impaired pruning. Further genetic analyses suggest that the calcium-activated protease calpain functions downstream of the calcium transients. Our findings reveal the importance of the compartmentalized subdendritic calcium signaling in spatiotemporally selective elimination of dendritic branches.

Nervous system development relies on a balance between progressive and regressive events (1, 2). After progressive events such as axon and/or dendrite outgrowth and synapse formation, neurons refine their connections through regressive events, such as pruning of axons and dendrites (13). In dendrite pruning, neurons selectively eliminate excessive or inappropriate connections that formed during development. Thus, proper dendrite pruning depends on local activation of the machinery needed to eliminate unnecessary dendritic branches; however, our understanding of locally acting mechanisms involved in this process remains incomplete.

Drosophila dendrite arborization sensory neurons are classified into four subtypes (I to IV) on the basis of their dendritic morphology, and class IV dendrite arborization (C4da) sensory neurons undergo dendrite pruning during early metamorphosis (37). C4da dendrites are typically severed at the proximal regions ~5 hours after puparium formation (APF), followed by fragmentation and clearance of the severed branches by 14 hours APF (4, 5). Previous studies have demonstrated that the ubiquitin-proteasome system and the caspase activity act to execute the pruning process (57), but it remains unclear what mechanisms locally determine which dendritic branches are to be eliminated.

Calcium signaling is often utilized for the spatiotemporal regulation of neural development (2, 3). To monitor changes in intracellular Ca2+ ion concentration ([Ca2+]i), we expressed the genetically encoded calcium (Ca2+) indicator GCaMP3 selectively in C4da neurons using the Gal4/UAS binary system (which exploits the yeast Gal4 transcriptional activator and the upstream activating sequence) (fig. S1) (8). We continuously monitored [Ca2+]i levels in single C4da neurons from 1 to 8 hours APF and found no apparent [Ca2+]i changes in soma and axons (9). However, we observed robust Ca2+ transients in groups of dendrites, which typically started 2 to 4 hours APF (Fig. 1, A to C, and movie S1). The dendritic Ca2+ transients appeared with a frequency of 0.88 ± 0.66 min−1 on average (n = 36), and the mean change in fluorescence (∆F/F0) of these events was 499.1 ± 162.5% (99 events, n = 10). Once Ca2+ transients commenced in a particular group of dendrites, transients continued to occur in the same dendrites. Furthermore, patterns, as well as amplitudes, of Ca2+ transients observed in individual dendritic groups were not correlated to each other (Fig. 1D and movie S1), which suggested that the Ca2+ transients are compartmentalized into each subdendritic region. Hereafter, we designate the groups of dendritic branches with the compartmentalized Ca2+ transients as “branch units.” Dendritic arbors in single neurons were typically subdivided into 4 to 5 branch units, which were mainly formed in dendrites distal from the secondary and from the tertiary branches (fig. S2).

Fig. 1 Compartmentalized Ca2+ transients occur in C4da dendrites during early metamorphosis.

(A and B) Live confocal image (A) and corresponding schematic trace (B) of a C4da neuron 2 hours APF labeled with mCD8 fused with red fluorescent protein (mCD8-RFP). Dendritic branches corresponding to branch units 1 and 2 are shown in magenta and green, respectively. On the schematic trace, i, ii, iii, iv, and v correspond to the places where we measured the calcium dynamics [see legend of (D)]. (C) Ratiometric images of GCaMP3 fluorescence at three time points (t0, t1, and t2). Each time point is indicated with arrowheads in (D). (D) Time traces of GCaMP3 fluorescence in five different regions, the locations of which are indicated in (B) (fractional change in fluorescence, ∆F/F0). A time-lapse movie corresponding to the blue time bar is available as movie S1. Scale bar, 50 μm.

An example of long-term simultaneous monitoring of [Ca2+]i and dendritic arbor structures by GCaMP3 and membrane-bound red fluorescent protein (RFP), respectively, is shown (Fig. 2). The first branch unit started its Ca2+ transients 2 hours APF, when no obvious pruning was detected (Fig. 2, C and D, and movie S2). The second and third branch units started their Ca2+ transients around 3.5 and 4.5 hours APF, respectively, when all the dendritic branches seemed still intact (Fig. 2E and movies S3 and S4). Five hours APF, the first branch unit underwent fragmentation after a continuous Ca2+ influx, whereas the other two still continued their Ca2+ transients, and their dendritic arbors seemed intact (Fig. 2F). Similarly, the second and third branch units were eliminated after continuous Ca2+ influx about 6.5 and 7.5 hours APF, respectively (Fig. 2, G and H). Thus, dendrites are pruned in the temporal order in which they experienced Ca2+ transients, the mean duration of which was 140.5 ± 43.6 min (n = 11). The location and timing of the occurrence of Ca2+ transients correlated perfectly with those of dendrite pruning (n ≥ 50 neurons), which indicated that the Ca2+ transients in branch units spatially and temporally predict the dendritic branches to be pruned within several hours.

Fig. 2

The occurrence of Ca2+ transients spatially and temporally correlates with dendrite pruning. (A) Live confocal image of a C4da neuron 0 hour APF. (B) Three branch units are visualized with ratiometric images of GCaMP3 fluorescence. (C) Long-term time traces of GCaMP3 fluorescence in branch unit 1 (blue), 2 (green), and 3 (magenta). Time-lapse movies S2 to S4 correspond to the light blue arrowheads. (D to H) Live confocal images and corresponding schematic traces at five different time points [indicated with arrowheads in (C)]. Schematic traces of each branch unit are shown in the same colors as in (C). Scale bars, 100 μm.

Neuronal Ca2+ signaling involves an intricate interplay between Ca2+ influx across the plasma membrane and Ca2+ release from the endoplasmic reticulum (ER) stores (10). We next conducted a candidate screen to identify the Ca2+ influx pathways responsible for generating the Ca2+ transients (tables S1 and S2). Both knockdown and knockout of ca-beta, the auxiliary β subunit of the voltage-gated Ca2+ channels (Cavβ) (11), resulted in a significant reduction in the amplitude and frequency of Ca2+ transients (Fig. 3, A and B; table S1; and fig. S3). These defects in ca-beta knockout clones [generated by the mosaic analysis with a repressible cell marker (MARCM) technique (12)] were fully rescued by restoring ca-beta expression in C4da neurons (Fig. 3B), which confirmed the cell-autonomous function of ca-beta. Furthermore, Ca2+ transients were completely abolished in neurons lacking functional copies of both ca-α1D and cacophony (cac), the Drosophila homologs of the α1 subunit Cav1 (L-type) and Cav2 genes (P/Q-, N-, and R-type), respectively, whereas the frequencies were only partially reduced in the single-mutant clones of ca-α1D or cac (Fig. 3, A and B). Thus, the voltage-gated Ca2+ channels (VGCCs) are responsible for generating the compartmentalized Ca2+ transients.

Fig. 3

VGCCs are responsible for generating Ca2+ transients. (A) Long-term time traces of GCaMP3 fluorescence in wild-type (top), ca-beta10 (middle), and cacHC129; ca-α1Dx10 (bottom) neurons between 1 and 8 hours APF. For each genotype, green and magenta time traces represent the GCaMP3 fluorescence changes in square framed regions 1 and 2, respectively. (B) Quantification of the frequency of Ca2+ transients. Means ± SEM. (C) A schematic of the photostimulation experiment. Neurons are photostimulated in the blue square region. (D) Changes in the threshold blue-light intensities to induce dendritic Ca2+ responses. Data points from the same neurons are connected with lines. Dotted lines, median values. *P < 0.05; **P < 0.001; analysis of variance (ANOVA), Dunnett’s post hoc test (B); paired t test (D).

What mechanisms can induce Ca2+ transients in such a compartmentalized manner? One possibility is that intrinsic dendritic excitability increases locally in individual branch units, which thereby sensitizes the VGCCs-mediated Ca2+ influx. To test this, we designed an experiment to optogenetically assess and manipulate dendritic excitability (Fig. 3C; see Methods in the supplementary materials for details). Briefly, we expressed the channelrhodopsin-2 (ChR2), a cation channel gated by blue light (13) and the Ca2+ indicator RGECO1 (14) in C4da neurons, and measured the threshold of the blue light intensity required to elicit dendritic Ca2+ influxes, which we confirmed to be dependent on the VGCCs opening in response to ChR2 activation (fig. S4). One hour APF, the threshold intensity was defined to be 1.1 mW mm−2 (median, n = 13 neurons) (Fig. 3D, left). Typically, 2 to 3 hours APF, branch excitability was elevated in local dendrites, as the threshold intensities were reduced by a factor of three in a group of dendrites (a median of 0.44 mW mm−2) (Fig. 3D, right) but not in other branches (fig. S4). In all cases, the dendritic branches with elevated excitability commenced “spontaneous” Ca2+ transients and were eventually eliminated within 3 hours (fig. S4), which suggested that the dendritic branches with elevated excitability correspond to branch units. Thus, our data indicate that intrinsic excitability of dendritic branches is locally elevated at the early pupal stage, which presumably promotes the robust VGCC-mediated Ca2+ influx in branch units. The increment of the branch excitability is regulated in part by the ecdysone signaling in C4da neurons, as cell-autonomous suppression of the ecdysone signaling blocked the occurrence of Ca2+ transients (fig. S5).

To test whether Ca2+ transients are required for dendrite pruning, we examined the pruning phenotypes in VGCCs mutant neurons. In MARCM clones, homozygous for ca-beta and doubly homozygous for cac and ca-α1D, dendrite pruning was significantly compromised (Fig. 4A), although there were no obvious defects in development of larval dendrites (fig. S6). No apparent defects in microtubule destabilization or dendritic severing at proximal regions were observed in ca-beta mutant neurons (fig. S7), which suggested that Ca2+ transients seem to be dispensable for these events. The pruning defects in ca-beta mutants were largely rescued by restoring the expression of ca-beta in C4da neurons (Fig. 4A). To inhibit the VGCC-mediated Ca2+ influx specifically during metamorphosis, we temporally expressed a hyperpolarizing Kir2.1 potassium channel (15) in C4da neurons by using the flip-out technique, and we found that in the flip-out clones generated by a heat shock treatment at 24 hours before the onset of metamorphosis, dendrite pruning was significantly inhibited (fig. S8). Thus, the VGCC-mediated Ca2+ influx is required in metamorphosis for dendrite pruning.

Fig. 4

Ca2+ transients are required for dendrite pruning. (A) Quantification of the pruning defects in the VGCCs mutant MARCM clones. Box plots indicate the median (white line), 25th, and 75th percentiles (box); the data range (whiskers); and outliers (circles). Outliers are defined as data points greater than the 75th percentile of all data points plus 1.5 times the interquartile range. (B) Quantification of the pruning defects in the calpain mutants. *P < 0.05; ***P < 0.001. NS, not significant.

Local caspase activation in dendrites is reported to be important for removal of C4da dendrites (6, 7). Given that caspase activation was predominantly detected after 7 hours APF, which is after Ca2+ transients were observed in most C4da neurons (fig. S9), caspases might function downstream of the Ca2+ signaling. However, comparable levels of caspase activation were detected in ca-beta mutant dendrites (fig. S10). In addition, we also found no obvious defects in the occurrence of the dendritic Ca2+ transients in neurons mutant for the initiator caspase dronc (fig. S11). It is thus likely that the caspase is not the major effector of the Ca2+ signaling and that the Ca2+ transients and the caspase activity function cooperatively to promote dendrite pruning. Consistently, pruning defects were synergistically enhanced in ca-beta; dronc double-mutant neurons compared with those in ca-beta or dronc single-mutant neurons (fig. S12).

To gain insight into the molecular mechanism underlying how Ca2+ transients promote dendrite pruning, we examined the pruning phenotypes after knockout or knockdown of individual genes whose functions are controlled by the intracellular Ca2+ (table S3). Among the 17 genes tested, dendrite pruning was partially, but significantly, attenuated in the Ca2+-activated protease calpain calpA mutants, and this attenuation was largely rescued by restoring the expression of calpA in C4da neurons (Fig. 4B). The pruning defects in calpA mutants were significantly enhanced by a mutation in calpB, the other member of the two fly homologs of the conventional calpain, which resulted in pruning defects comparable to those in VGCC mutants (Fig. 4B). We tested for genetic interactions between calpains and VGCCs and found that pupae heterozygous for both ca-beta and two calpain genes showed significant pruning defects, whereas each one of them separately was normal (Fig. 4B), which suggested that the calpains function downstream of Ca2+ transients to promote dendrite pruning.

We have shown that Ca2+ transients occur locally in dendrites to trigger dendrite pruning in Drosophila sensory neurons. These compartmentalized Ca2+ transients are, to our knowledge, the earliest events that occur in the dendritic branches that are destined to be pruned. Although many of previous studies have focused on pathological roles of the calpains (16), our findings suggest that, during developmental dendrite pruning, the calpain activities are developmentally regulated and are likely compartmentalized in unnecessary dendrites. We speculate that such controlled and localized calpain activations are accomplished by an intrinsic property of neurons—that is, by local increase in branch excitability to induce local Ca2+ transients. Compartmentalized changes in dendritic branch excitability are also observed in many different types of mammalian neurons (17, 18). It will be intriguing to examine whether compartmentalized Ca2+ transients and subsequent calpain activations are part of a conserved mechanism to trigger dendrite pruning. Future studies deciphering the role of dendritic Ca2+ transients will provide insights into the regulation of dendrite pruning.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 to S3

References (1939)

Movies S1 to S4

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank Y. N. Jan, C. Montel, M. Miura, L. L Looger, K. Deisseroth, Bloomington Stock Center, and Kyoto Stock Center for reagents; Riken BSI-Olympus collaboration center for measuring the blue-light intensity; W. Sugano and A. Kawasaki for technical assistance; S. Nakanishi and J. Z. Parrish for helpful comments on the manuscript. This work was supported by grants-in-aid for Science and Technology from the Japanese Government Ministry of Education, Culture, Sports, Science, and Technology (MEXT), the Strategic Research Program for Brain Science, CREST, and Takeda Science Foundation.
View Abstract


Navigate This Article