Dendrite regeneration mediates functional recovery after complete dendrite removal

Abstract

Unlike many cell types, neurons are not typically replaced if damaged. Therefore, regeneration of damaged cellular domains is critical for maintenance of neuronal function. While axon regeneration has been documented for several hundred years, it has only recently become possible to determine whether neurons respond to dendrite removal with regeneration. Regrowth of dendrite arbors has been documented in invertebrate and vertebrate model systems, but whether it leads to functional restoration of a circuit remains unknown. To test whether dendrite regeneration restores function, we used larval Drosophila nociceptive neurons. Their dendrites detect noxious stimuli to initiate escape behavior. Previous studies of Drosophila sensory neurons have shown that dendrites of single neurons regrow after laser severing. We removed dendrites from 16 neurons per animal to clear most of the dorsal surface of nociceptive innervation. As expected, this reduced aversive responses to noxious touch. Surprisingly, behavior was completely restored 24 hours after injury, at the stage when dendrite regeneration has begun, but the new arbor has only covered a small portion of its former territory. This behavioral recovery required regenerative outgrowth as it was eliminated in a genetic background in which new growth is blocked. We conclude that dendrite regeneration can restore behavior.

Graphical Abstract

Introduction

Neurons can be damaged in a variety of ways from acute trauma in car accidents and excitotoxicity during stroke to long-term accumulation of misfolded proteins in neurodegenerative disease. One important repair process that helps maintain a functional nervous system through time is axon regeneration, which occurs in response to loss of part or all of the axon. Axon regeneration is initiated by signals generated at the site of injury that are transmitted to the cell body to lead to transcriptional reprogramming into a growth state (Mahar and Cavalli, 2018; Rishal and Fainzilber, 2010). Functional recovery occurs if the axon can reach and reinnervate an appropriate target. Axon regeneration occurs in the vertebrate periphery as well as in invertebrates (Brace and DiAntonio, 2017; Byrne and Hammarlund, 2017; Hao and Collins, 2017; Liu et al., 2011; Rasmussen and Sagasti, 2017) and represents an evolutionarily conserved process that contributes to lifelong nervous system function.

While axon regeneration is broadly conserved in bilaterian animals and widely studied, very little is known about repair capacity of the receiving end of neurons: dendrites. Dendrite damage has been documented in rodent models of clinically important scenarios including traumatic brain injury (Gao et al., 2011; Sword et al., 2013) and ischemic stroke (Murphy et al., 2008; Zhu et al., 2017). However, it is very difficult to track individual neurons over time after these large-scale disruptions to determine whether dendrites regenerate. In a more defined prick injury model, dendrites could be seen growing into the damaged region of cortex suggesting the possibility that dendrite regeneration might contribute to recovery after central nervous system damage in mammals (Paveliev et al., 2016). Single cell responses to dendrite injury have been tracked in the zebrafish spinal cord, demonstrating that individual vertebrate neurons can reactivate dendrite growth in response to injury (Stone et al., 2022). Dendrite regrowth after injury has been documented in most detail in invertebrate model systems including Drosophila and C. elegans (see below). It therefore seems that dendrites can initiate a growth program after injury, but it is not known whether this growth can contribute to restoration of function and thus truly be part of a regenerative process.

In order to test whether dendrite regeneration can restore neuronal function, we needed a model system in which a specific behavior could be triggered by a defined set of neurons, and in which dendrites could be damaged independently from axons. As core mechanisms of axon regeneration are conserved from invertebrates like C. elegans and Drosophila to vertebrates including zebrafish and mice (He and Jin, 2016; Winter et al., 2022), but dendrite injury assays are much better established in invertebrates, we focused on these systems. In C. elegans, dendrite regeneration has been studied primarily in PVD neurons, cells that extend large dendrite arbors over the body wall (Oren-Suissa et al., 2010; Tsalik et al., 2003) and sense harsh touch (Way and Chalfie, 1989). These neurons primarily repair dendrite breaks using plasma membrane fusion to join disconnected regions (Brar et al., 2022; Oren-Suissa et al., 2017), although they can also initiate regenerative growth (Brar et al., 2022). As neurite fusion is not seen in many other animals, including vertebrates, we preferred to use a model system in which this mechanism does not contribute to dendrite repair. Drosophila sensory neurons that innervate the body wall have also been used to study the response to dendrite injury. In this case, dendrite severing leads to outgrowth of new dendrites (Song et al., 2012; Stone et al., 2014) as recently documented in zebrafish (Stone et al., 2022).

Drosophila non-ciliated sensory neurons, termed dendritic arborization neurons, are grouped into four categories in larvae based on branching complexity (Grueber et al., 2002). Each of these classes corresponds to distinct sensory function. For example, the simplest neurons, Class I, are proprioceptive (Hughes and Thomas, 2007) and respond to folding of the body wall during crawling (He et al., 2019; Vaadia et al., 2019). Class III neurons respond to gentle touch (Yan et al., 2013), and the most complex Class IV neurons are nociceptive and can help larvae avoid parasitoid wasps (Hwang et al., 2007). Class I, III and IV neurons have all been tested for dendrite regeneration. In all cases complete removal of dendrites using laser microsurgery resulted in regrowth of the dendrite arbor (Stone et al., 2014; Thompson-Peer et al., 2016). However, whether these regenerated arbors can elicit behavior remains untested. Two studies have tested whether regenerated dendrites are electrically active. The first characterized electrical activity of larval class III dendrites before and after injury using a mechanical poking assay. Recordings were made from neurons in larval fillet preps. In control neurons action potentials were elicited by the probe, and after dendrotomy, this activity was lost. After three days of regeneration, stimulation again elicited action potentials, however the sensitivity of the cells was reduced leading to the conclusion that the cells did not regain normal function (Thompson-Peer et al., 2016). The second study was performed on Class IV neurons in the adult abdomen. While dendritic arborization neurons have been studied more extensively in larvae, a subset of these cells survives pupariation and remodels to innervate the adult body wall (Shimono et al., 2009). Dendrites were removed from a Class IV neuron in young adult flies using laser microsurgery and allowed to regrow for seven days. In fillet preps made at this point, the exposed dendrites fired with the same frequency in response to a puff of acid solution in neurons with uninjured or regenerated arbors (DeVault et al., 2018). From these studies, it seems likely that regenerated dendrite arbors can sense stimuli, but whether their function is sufficient to trigger a behavioral output has not been tested.

Larval Class IV dendritic arborization neurons are a good model system for assessing dendrite function. Noxious heat above 39°C and harsh touch elicit a characteristic rolling behavior (Tracey et al., 2003) and Class IV neurons are necessary and sufficient for triggering this behavior (Hwang et al., 2007). The body wall is completely covered by dendrites of Class IV neurons with three cells innervating each hemisegment (Grueber et al., 2002). Dendrite regeneration assays for these cells use lasers to remove dendrites from a single neuron (Stone et al., 2014; Thompson-Peer et al., 2016). As each cell innervates a relatively small fraction of the body wall, it is unclear whether a stimulus applied to the outside of the animal could be easily targeted to the territory of an individual neuron. We optimized our laser microsurgery so that we could remove dendrites from most of the Class IV ddaC neurons that innervate the dorsal surface of the larva. This scale of injury allowed us to use a manually directed heat probe to elicit aversive behavior. We found that, compared to control animals, animals tested four hours after large scale dendrite removal had deficits in nociception. 24 hours after injury, when these neurons had initiated regeneration, nociceptive behavior was similar to that in control animals. In a regeneration-deficient genetic background, nociceptive function was not restored. Taken together, this data indicates that dendrite regeneration can restore neuronal function in vivo.

Results

Animals can recover from laser removal of dendrites from most ddaC neurons.

To test whether we could pair a behavioral assay with dendrite injury, we focused on Class IV dendritic arborization neurons that are responsible for nocifensive responses in larvae. These cells can be activated by touching a heated probe to the Drosophila epidermis, where the Class IV cell bodies and dendrites are anchored (Hwang et al., 2007; Tracey et al., 2003). For this assay to be useful for monitoring function after dendrite removal, we needed to be able to target a specific region on the cuticle that was innervated by injured neurons. Standard injury assays for Class IV neurons involve removal of dendrites from a single cell, which innervates too small an area to be easily touched with a heated probe. We therefore decided to try to perform surgery on multiple neurons in a single animal. The dorsal surface of the animal is innervated by ddaC (dorsal dendritic arborization C) Class IV neurons (Grueber et al., 2002). Eight pairs of these cells cover the dorsal surface behind the head (Figure 1A). We previously removed dendrites from most of these neurons to generate RNA libraries for sequencing (Nye et al., 2020), so we optimized dendrite removal from all 16 neurons. Animals with fluorescently labeled Class IV neurons were mounted on a slide and immobilized by taping a coverslip on top of them. The base of each ddaC dendrite was positioned at the focus point of a pulsed UV laser to sever it, and then the animal was immediately removed from the slide and returned to media. After optimizing efficiency of mounting and severing, most animals survived the large-scale laser surgery. When animals were imaged at 24h after injury, new dendrite arbors were seen growing from the cell bodies in most cases (Figure 1B). Growth was qualitatively similar in these large-scale injuries to growth at 24h after single neuron injury (Nye et al., 2020). To further establish that simultaneous injury of many neurons elicited growth through similar pathways to single cell injury, we assayed growth 24h after this procedure in a genetic background that blocks dendrite regrowth after single cell injury. The exocyst, an eight subunit complex required for polarized secretion (Lepore et al., 2018; Polgar and Fogelgren, 2018), is strongly required for dendrite regeneration (Swope et al., 2022). Expression of an RNAi hairpin targeting one of the exocyst subunits, Sec6, strongly reduced dendrite regrowth in this large-scale assay (Figure 1D), consistent with results from single neuron dendrite removal (Swope et al., 2022). We also tested whether animals could tolerate removal of axons from the same set of neurons as we wished to use this as a control. After axon removal (Figure 1C), dendrite arbors simplified as previously described after single axon injury (Chen et al., 2012), and cells became dimmer reflecting reduced expression of the ppk promoter, but cells and animals survived. We concluded that global removal of dendrites from ddaC neurons was a good platform for determining whether regenerated dendrites could function to initiate behavior.

Figure 1.

Laser severing of dendrites or axons of 16 ddaC neurons.

(A) A dorsal overview of a larva expressing CD4-tdGFP in class IV neurons before injury. (B) Dendrites of 16 ddaC neurons in an animal expressing the control RNAi hairpin were severed and allowed to regenerate for 24 hours; HPD stands for hours post dendrotomy. (C) Axons of 16 ddaC neurons were severed and the animal was imaged 24 hours later; HPA stands for hours post axotomy. Red arrows indicate severed and degenerating axons, green arrows indicate intact axons in the head. (D) Sec6 RNAi hairpins were expressed in Class IV neurons and dendrites were severed from 16 ddaC neurons and imaged 24 hours later as in (B).

Heat probe touch to the dorsal surface of Drosophila larvae initiates aversive behaviors

To be able to assess how dendrotomy affected the response to heat probe touch, we first needed to characterize behavior in control animals after dorsal touch of a heat probe. Reported response time and frequency vary between labs likely due to variation between probes and how and where they contact the animal (Follansbee et al., 2017; Neely et al., 2010; Tracey et al., 2003). The classic response to stimulation with noxious heat is a full roll along the body axis, starting with curling the head and tail towards each other (Figure 2A, 1^st image on left) then starting to roll upside down (Figure 2A, 2^nd and 3^rd images), finishing by returning upright (Figure 2A, 4^th and 5^th images). Most studies using this assay use a temperature between 39° and 46°C, with higher temperatures typically inducing a faster, more robust response (Follansbee et al., 2017; Honjo et al., 2016; Tracey et al., 2003), so we elected to use a 46°C probe for our study (detailed explanation in Materials and Methods). To perform the nociceptive assay, we touched larvae on the dorsal surface with the 46° probe for up to 10 seconds. As expected, 67 of 67 uninjured larvae rolled within 10 seconds (Figure 3B), with 75% of larvae rolling within 2 seconds (Figure 3C).

Figure 2.

Larval movements in response to 46° probe.

(A) Sequence of larval movement during a stereotyped roll. The tail and head curl in (first image on left), and the body rotates in the opposite direction (2^nd, 3^rd, and 4^th images) until right side up again (5^th image). (B) Example images of aversive behaviors quantified in figures 3 and 4. These include: rapid and simultaneous contraction of the head and/or tail (1^st image); repetitive backwards crawling (2^nd image); lifting and flailing of the head or tail (3^rd and 4^th images); or an incomplete roll (last image).

Figure 3.

Class IV dendrites can regenerate and regain nociceptive capability.

(A) An example neuron expressing control RNAi before injury (left), 4h post dendrotomy (HPD) (middle), and after 24h of regeneration (right). Red boxes indicate extent of dendrite arbor size at each time point. (B) The latency to roll from the onset of stimulus is graphed for uninjured, dendrite cut 4 and 24 hours after injury and axon cut 6 and 24 hours after injury. Any larvae that did not roll were not included in this graph but are summarized in the table above. (C) The latencies in (B) were binned into <2s, 2–4s, 4–6s, and 6–10s to compare proportions for each condition. (D) % of larvae displaying any immediate aversive reaction to the stimulus is graphed for each condition. Numbers in bars represent numbers of larvae. ***, p<.001 with Mann-Whitney test.

While optimizing this assay, we noticed additional behaviors in larvae that did not roll instantaneously. While the corkscrew roll behavior is widely used to test for pain circuit function, few papers describe or quantify other nociceptive behavior in response to heat probe touch. However, other behaviors including writhing and bending have been described in global heat plate assays (Chattopadhyay et al., 2012), in response to parasitic wasps (Hwang et al., 2007) or direct activation of Class IV neurons (Burgos et al., 2018).

The most common additional aversive behaviors we observed were flailing (Figure 2B, 1^st and 2^nd images) or contraction of either or both head and tail segments (Figure 2B, 3^rd images). Rapid crawling or incomplete rolls were also common (Figure 2B, 4^th and 5^th images). Often larvae that were not rolling exhibited more than one of these motions in a clear attempt to escape from the heat probe. Importantly, none of these behaviors are ever seen during normal locomotion and are different from “light touch” responses that have been characterized in other studies. Light touch responses include pausing feeding, hesitation, or turning away and slowly retreating from the stimulus (Kernan et al., 1994). While activation of other types of sensory neurons can increase the likelihood Class IV neuron activation will lead to rolling, they cannot elicit aversive behavior without Class IV neurons (Ohyama et al., 2015).

Since these immediate behaviors we observed were clearly stimulus-induced and likely to be mediated by Class IV neurons, we decided to group them all as “aversive reactions.” We included an immediate roll in this category. 66 of 67 uninjured larvae either rolled or showed one of these behaviors immediately upon thermal stimulation (Figure 3D). In control animals, this instant response was typically followed by rolling after a variable delay (Figure 3B).

ddaC dendrite removal causes deficits in aversive behavior that are restored 24h after injury.

Having developed large-scale dendrite injury assays and characterizing aversive heat probe responses, we could pair the two to explore functional regeneration. After dendrites are severed, they start to degenerate within a few hours (Figure 3A, middle), and by 24h have regenerated into a significant area of their original space (Figure 3A, right). Thus, we chose to begin testing function during the degeneration stage four hours after injury and after regeneration initiated 24 hours after injury.

Four hours after large-scale injury, the majority of larvae (57 of 73) could still roll after stimulation (Figure 3B), although the latency to roll was significantly longer than control (Figure 3B and C). Fewer than 25% of injured larvae rolled in 2 seconds or less, compared to 75% of uninjured larvae (Figure 3C). Larvae that did not roll within 10 seconds were not included in latency quantifications.

That larvae with dendrites removed from dorsal Class IV neurons still exhibit nocifensive rolling is surprising as silencing Class IV neurons has been shown to eliminate rolling in 80% of animals (Hwang et al., 2007). One explanation for this result could be that the cell bodies and/or axons of Class IV neurons can initiate the response to heat. To address this possibility, we severed axons of dorsal ddaC neurons and assayed nociception 6h (during degeneration) and 24h (after degeneration, before regeneration) after injury. None of the larvae with severed axons rolled in response to the stimulus either at 6h or 24h. This result suggests that bald Class IV cell bodies with intact axons retain some ability to sense noxious heat. Quantification of immediate aversive responses showed that without dendrites ddaC neurons lose most, but not all, ability to respond immediately to a heated probe (Figure 3D) four hours after injury. Having established that dendrite removal resulted in a substantial behavioral deficit, although not complete elimination of function, we had a baseline to determine whether function improved during regeneration.

We began by assaying function 24 hours after injury, at a time when regeneration had initiated, but was not complete. At this time, latency to roll was very similar to that in uninjured animals (Figure 3B and C). Similarly, immediate aversive behavior was largely recovered (Figure 3D). In contrast, no recovery was seen in animals in which axons had been removed (Figure 3B–D) supporting the idea that the behaviors we recorded required ddaC neurons to activate them, and that the change from 4 hours to 24 hours after dendrite removal was due to improvement of ddaC signaling.

Regeneration-deficient neurons do not recover function.

One alternate explanation for improvement of nociceptive function 24 hours after dendrite injury could be that neurons recover function as acute tissue damage is resolved without requiring new outgrowth. To address this possibility, we performed injury and functional assessment in a regeneration-deficient background. We recently showed that the exocyst complex is strongly required for dendrite regrowth in class IV neurons (Swope et al., 2022). We used a Sec6 RNAi line that results in a strong reduction in dendrite regeneration (Swope et al., 2022). Uninjured Sec6 RNAi neurons have similarly complex dendrites to control neurons (Figure 4A and (Swope et al., 2022)), and larvae display similar rolling and immediate aversive reaction sensitivity to the 46°C probe (Figure 4B–D). As previously described, dendrite regeneration is almost completely abrogated after removal of all dendrites from a single ddaC neuron (Figure 4A and (Swope et al., 2022). Similarly, there was very little dendrite regrowth after removal of dendrites from the 16 dorsal ddaC neurons (Figure 1D).

Figure 4.

Sec6 RNAi neurons do not regenerate or regain nociception.

(A) An example Sec6 RNAi-expressing neuron before injury (left), 4h post dendrotomy (HPD) (middle), and after 24h of regeneration (right). Red boxes indicate extent of dendrite arbor size at each time point. (B) Latency to roll from stimulus onset is graphed for uninjured, dendrite cut 4 and 24 hours after injury, and axon cut 6 and 24 hours after injury. Any larvae that did not roll are not graphed but are summarized in the table above the graph. (C) The latencies to roll were binned into categories, <2s, 2–4s, 4–6s, and 6–10s, to compare proportions for each condition. (D) % of larvae displaying any immediate aversive reactions to the stimulus are graphed. Numbers in bars represent numbers of larvae. ***, p<.001 with Mann-Whitney test.

Four hours after dendrite removal latency to roll was increased as in control animals (Figure 4C and D), and rapid aversive responses were also reduced (Figure 4D). Unlike control animals, there was no recovery of either response 24 hours after injury (Figure 4B–D). We conclude that that regenerative growth after dendrite injury is required for functional recovery.

Discussion

Here, we have shown that after large-scale removal of dendrites from dorsal class IV neurons, larvae lose much of their nociceptive sensitivity. 24 hours of regeneration is sufficient to largely restore nociception and its behavioral outputs. In a regeneration-deficient background, nociception did not improve after injury. Previous electrophysiological work demonstrated that regenerated class III and IV dendrites can sense their native noxious stimuli and fire action potentials (DeVault et al., 2018; Thompson-Peer et al., 2016); we now show that class IV neurons can sense stimuli and send signals strong enough to elicit a behavioral phenotype in vivo, supporting that they have regained their initial capacity for nociception.

The two studies that assayed electrophysiological responses of neurons after dendrite injury used much longer recovery times than we found necessary. As behavior was completely restored 24 hours after injury, we did not try longer timepoints as there was no room for improvement. In the study on larval Class III neurons, cellular responsiveness was measured three days after dendrite injury by applying gentle touch to a larval fillet prep. While cells could still respond, action potentials were less frequent, and it was concluded that function of regenerated dendrites is different than uninjured ones (Thompson-Peer et al., 2016). In our study, we do not know whether electrical activity of neurons is identical to those of uninjured cells when our assay is performed, indeed it would be surprising if it was as the dendrite arbor is much smaller. Critically, however, behavioral responses are restored. In the study on Class IV neuronal electrical activity, neurons were injured in adults and assayed 7 days later. In these adult cells regeneration takes longer to initiate and no response to acid stimulus was seen 24 hours after dendrite removal, but firing frequency was identical to uninjured cells 7 days after injury. Thus, while not all regenerated dendrite arbors may be identical to uninjured counterparts, these data all suggest that regenerated dendrite arbors respond to appropriate stimuli and can function.

One somewhat surprising aspect of our study was that function is recovered so rapidly, while regenerative regrowth is far from complete. It may be that nociceptive neurons have evolved to respond to stimuli even with incomplete arbors. In a previous screen to identify genes required for nociception, several of the genetic backgrounds that increased time to respond reduced dendrite arbor complexity, while several that reduced time to respond increased complexity (Honjo et al., 2016). However, there were also counter examples where increased sensitivity to stimulus was associated with reduced complexity. Moreover, some of the genotypes that reduced both complexity and sensitivity likely affected more than cellular shape. For example, Lis1 was in this group (Honjo et al., 2016). It is a dynein subunit required for normal dendrite shape (Liu et al., 2000). However, as a dynein subunit, it also plays a central role cargo transport into dendrites. As dendritic microtubules have minus-end-out orientation in Drosophila (Stone et al., 2008), all cargoes including sensory channels, are predicted to use dynein for transport into dendrites. Thus, it is difficult to link dendritic shape changes in Lis1 RNAi neurons to reduced sensitivity. However, a previous study used an infrared laser to heat regions of the dendrite arbor or only the cell body, and this localized stimulus was sufficient to cause global calcium transients and neuronal firing (Terada et al., 2016). This study is therefore consistent with our finding that some function remains after complete dendrite removal, and that regeneration of a small arbor could restore normal sensitivity to a large heat probe.

While nociceptive responses are primarily due to function of Class IV neurons, silencing other classes of dendritic arborization neurons can slightly reduce behavioral output after noxious stimulation (Hwang et al., 2007). Is there some possibility that changes in other cells mediate reduction in response and then subsequent recovery in our large-scale injury paradigm? Although our laser surgery approach is tightly focused on cells of interest, we may sometimes damage dendrites or axons of other neurons in the vicinity, or glial or epithelial cells. It is possible that off-target responses could mediate reduction in function 4 hours after injury. However, it is unlikely that changes in surrounding cells mediate functional recovery at 24 hours after injury. The first argument against this is that no recovery occurs after axon injury. However, one caveat here is that the site of damage and number of cut sites is different for axon and dendrite injury. A much stronger argument is the lack of recovery in Sec6 RNAi neurons. In this experiment Sec6 is knocked down by expressing hairpin RNAs only in Class IV neurons, so surrounding cells should be normal. These normal surrounding cells are not, however, sufficient to increase responsiveness to heat 24 hours after injury as no improvement is seen in the Sec6 neuronal RNAi animals. This is consistent with a previous study that shows other sensory neurons can potentiate the response to Class IV neuron stimulation but cannot themselves cause nocifensive responses (Ohyama et al., 2015). We conclude that the behavioral recovery we observe is due to growth of the Class IV neurons themselves, rather than any resolution of injury in nearby cells.

Most studies on dendrite regeneration in invertebrate systems have used sensory dendrites as they are optically accessible near the surface of the animal. Drosophila sensory dendrites share many features with canonical post-synaptic dendrites, including presence of minus-end-out microtubules, ribosomes, and branching near the cell body (Hill et al., 2012; Rolls and Jegla, 2015; Stone et al., 2008). However, they are not post-synaptic and so do not need to reconnect to other neurons during regeneration. Post-synaptic motor neurons in the zebrafish spinal cord were recently shown to regrow after injury (Stone et al., 2022), but reintegration into a circuit was not analyzed. It will be extremely interesting, and technically challenging, to expand functional studies on dendrite regeneration to other types of neurons, including those that receive input from other neurons.

Materials and methods

Drosophila stocks

The two RNAi lines used in this study, γTub37c #25271 and Sec6 #105836, were obtained from the Vienna Drosophila Resource Center (https://stockcenter.vdrc.at/control/main). We have used γTub37c #25271 RNAi as a control in many of our studies as γTub37c is not expressed in somatic tissues (Wiese, 2008) like neurons, and it was used for the control here. The transgenic tester line we used was ppk:Gal4, ppk:EGFP, ppk:CD4-tdGFP, UAS:dicer2-nlsBFP/TM3, Tb-RFP, and was previously used and described in our study on the role of the exocyst in regeneration (Swope et al., 2022). Females of the tester line were crossed to males of the RNAi lines, and non-tubby offspring were used.

Larval mounting for imaging and neuron selection

Larvae have 20 ddaC neurons (10 on left side, 10 on right); two in each body segment. We chose to sever axons or dendrites from 8 neurons on each side, sparing the neurons in the head to help larva survive this laborious injury paradigm. Larvae were mounted on a glass slide on top of a thin pad of dried agar, then immobilized with a coverslip taped on top. Larvae were not anesthetized.

Image acquisition and analysis

Zeiss microscopes (Thornwood, NY) were used to acquire images in this study. The following systems were used:

• LSM800 inverted confocal on an Axio Observer Z1 stand and equipped with GaAsP detectors and a Zeiss Plan-APOCHROMAT 63x DIC (oil, 1.4NA) objective;

• LSM800 upright confocal on an Axio Imager.Z2 stand and equipped with GaAsP detectors and Zeiss Plan-APOCHROMAT 63x DIC (oil, 1.4NA) and Zeiss Plan-APOCHROMAT DIC (UV) VIS-IR 40x (oil, 1.3NA) objectives;

• Zeiss Axio Imager.M2 widefield equipped with an AxioCam 506 mono camera and Zeiss Plan-APOCHROMAT 63x DIC (oil, 1.4NA), Zeiss EC Plan NEO FLUAR 40x (oil, 1.3NA) objectives.

Uninjured neurons and 24 hours post dendrotomy (HPD) were imaged with the LSM800 upright with a 40x objective; 4 HPD neurons were imaged with the LSM800 inverted with a 63x objective. Images in Figure 1 were taken using a 40X objective and stitching with up to up to 20×5 tiles. Images in Figures 3 and 4 were taken as 2×2 or 3×3 tiles. For these tiled images, occasionally larvae moved slightly during imaging, so apparent line breaks in representative images are artifacts of movement, not of injury.

Class IV injury assays

Larvae were mounted on a glass slide, dorsal side up, as straight as possible to allow access to all dorsal neurons. A MicroPoint UV pulsed laser (Oxford Instruments) mounted to a Zeiss Axio Imager.M2 was used to sever individual dendrites or axons from neurons. The tester line we used has both a membrane-bound fluorescent protein (CD4-tdGFP) and cytoplasmic fluorescent protein (EGFP), and when a cut is incomplete, cytoplasmic EGFP rapidly diffuses back into the cut site. When a cut is complete, cytoplasmic EGFP does not flow back into this site, and there is a stark black space in between the severed ends of neurites. Because of this, we confirmed successful cuts at time of cutting rather than re-mounting larvae at a later time point to confirm the cuts were successful. Images in Figures 3 and 4 are high resolution images of single injured neurons for clarity.

Nociceptive assay

We used a HAKKO FX-888D soldering iron as our probe. The soldering iron was equipped with a conical 0.125mm tip and the wire sensor of a K-type thermocouple (meter-depot DM6802A+) )was attached around the base of the soldering tip to monitor temperature of the probe. When the soldering iron was set to 52°C the thermocouple measured a temperature of 46°C and this setting was used in all of the experiments. To perform the nociceptive assay, we removed larvae from their food and washed them gently in water. We then put 3–4 larvae on a black plastic pad on a dissecting scope platform with enough water that they could crawl normally, but not enough that they were floating and had movement dictated by surface tension. We let larvae crawl for 5–10 seconds before performing the assay to ensure that their movement was normal (important for injured larvae). To test their nociception, we touched the very tip of the heated soldering iron to the dorsal side of the larvae as gently as possible for up to 10 seconds. As soon as larvae rolled, the stimulus was removed; if larvae did not complete a full roll in 10 seconds of stimulation, the stimulus was removed and categorized as a “no roll”. We recorded a video of each larva from which we calculated the latency to roll and whether the larva displayed an aversive reaction to the stimulus.

For graphing this data, only larvae that rolled had their latency included. All non-rolling larvae were excluded from latency data but were included in aversive reaction graphs (Figure 3D, Figure 4D).

Statistical analysis

Statistics were performed and graphs made with GraphPad Prism. An unpaired Mann-Whitney U-test was performed to compare conditions. Significance is shown as *** (p<.001). All error bars represent standard deviation.

Highlights.

Drosophila larvae were used to test whether neurons can regrow functional dendrites

All dendrites from dorsal nociceptive neurons were removed

After dendrite removal, heat probe induced escape behavior was reduced

24 hours of dendrite regeneration was sufficient to completed restore escape behavior

No recovery occurred when dendrite regeneration was genetically blocked