Neurons survive simultaneous injury to axons and dendrites and regrow both types of processes in vivo

Abstract

Neurons extend dendrites and axons to receive and send signals. If either type of process is removed, the cell cannot function. Rather than undergoing cell death, some neurons can regrow axons and dendrites. Axon and dendrite regeneration have been examined separately and require sensing the injury and reinitiating the correct growth program. Whether neurons in vivo can sense and respond to simultaneous axon and dendrite injury with polarized regeneration has not been explored. To investigate the outcome of simultaneous axon and dendrite damage, we used a Drosophila model system in which neuronal polarity, axon regeneration, and dendrite regeneration have been characterized. After removal of the axon and all but one dendrite, the remaining dendrite was converted to a process that had a long unbranched region that extended over long distances and a region where shorter branched processes were added. These observations suggested axons and dendrites could regrow at the same time. To further test the capacity of neurons to implement polarized regeneration after axon and dendrite damage, we removed all neurites from mature neurons. In this case a long unbranched neurite and short branched neurites were regrown from the stripped cell body. Moreover, the long neurite had axonal plus-end-out microtubule polarity and the shorter neurites had mixed polarity consistent with dendrite identity. The long process also accumulated endoplasmic reticulum at its tip like regenerating axons. We conclude that neurons in vivo can respond to simultaneous axon and dendrite injury by initiating growth of a new axon and new dendrites.

Graphical Abstract

Introduction

Most neurons use dendrites to receive signals and axons to send them, and function is lost if either type of process is damaged. As neurons typically cannot be replaced in mature nervous systems, repair pathways have evolved to allow neurons with damaged axons or dendrites to regain function. Axon injury and repair have been studied for over one hundred years, and some aspects of the axon injury response have been elucidated. After axons are severed, the portion of the axon distal to the cut site degenerates (Gerdts et al., 2016), and an injury signal is sent to the cell body (Abe and Cavalli, 2008; Rishal and Fainzilber, 2014). In many neuron types, including those in the vertebrate peripheral nervous system (PNS), and in invertebrates like Drosophila and C. elegans, this injury signal re-initiates axon growth (He and Jin, 2016; Mahar and Cavalli, 2018). In the vertebrate central nervous system (CNS), some neurons initiate growth in response to axon injury, either from the cut stump (Kerschensteiner et al., 2005) or from collateral branches (Kerschensteiner et al., 2004). However, other CNS neurons fail to grow either because they do not mount the initial injury response or because their axons encounter an inhibitory environment (Curcio and Bradke, 2018; Liu et al., 2011). In vertebrate and invertebrate neurons that regenerate, the dual leucine zipper kinase (DLK) is a central component of the axon injury signal (Hammarlund et al., 2009; Shin et al., 2012; Xiong et al., 2010; Yan et al., 2009) that acts upstream of transcriptional changes that trigger regeneration in permissive cells (Mahar and Cavalli, 2018).

Although much less studied and much less understood, it has recently been demonstrated that- at least in Drosophila- neurons can initiate a program of dendrite regeneration after partial or complete removal of the dendrite arbor (DeVault et al., 2018; Song et al., 2012; Stone et al., 2014; Thompson-Peer et al., 2016). Although an injury-induced growth response, dendrite regeneration seems to use different molecular machinery than axon regeneration. It does not require DLK or its downstream transcription factors to initiate the injury response (Stone et al., 2014). Comparison of axon and dendrite regeneration in the same Drosophila cell type also revealed that axon regeneration requires endoplasmic reticulum (ER) localization to the growing axon tip, while dendrite regeneration does not (Rao et al., 2016). Further confirming this distinction between the two types of regeneration, dendrite regrowth requires the tyrosine kinase Ror, while axon regeneration does not (Nye et al., 2020). Thus, neurons seem to possess two distinct regeneration programs: one for axons and one for dendrites.

Some types of injury or stress are expected to damage axons and dendrites at the same time. For example, traumatic brain injury (TBI) can cause axon breakage followed by degeneration (Hill et al., 2016; Strich, 1956), and can also induce dendrite degeneration (Gao and Chen, 2011). Similarly, ischemic stroke can cause dendrite damage (Murphy et al., 2008; Risher et al., 2010) in addition to axon damage (Hinman, 2014). In the case of stroke, axon sprouting (Carmichael et al., 2017) and dendrite plasticity (Brown et al., 2007) are observed during recovery, but it is unclear whether these derive from neurons that were damaged in the initial injury, or are part of a response by surrounding neurons.

If neurons survive the initial trauma to axons and dendrites, the next question is whether they can launch the axon and dendrite regeneration programs together. There are some hints that axon regeneration may antagonize dendrites. In some vertebrate neurons, axon injury causes retraction of dendrite arbors (Leung et al., 2011; Linda et al., 1992; Sumner and Watson, 1971; Wang et al., 2002) suggesting that axon injury responses including regeneration antagonize dendrite maintenance or growth. Similar retraction of dendrite arbors after axon injury has been observed in Drosophila neurons (Chen et al., 2012). From a more hypothetical perspective, both axon and dendrite regeneration require large-scale addition of membranes to the cell, and so regrowth of one type of neurite might well be expected to reduce the resources available for growth of the other type.

There are two scenarios where the outcome of simultaneous axon and dendrite damage has been established. One is in the generation of primary neuronal cultures. Preparation of cortical or hippocampal rodent cultures involves dissociation of embryonic neurons, which removes axons and dendrites. The remaining cell body is cultured with ample growth factors, and can successfully regrow axons and dendrites (Kaech and Banker, 2006). In this context of extra growth factors, neurons do survive and after an unpolarized phase axons are specified and grow, followed a few days later by dendrites (Bradke and Dotti, 2000; Dotti et al., 1988). The second example is amphid neurons in C. elegans, which have a sensory cilium rather than a branched dendrite arbor. These neurons are not capable of regenerating the sensory ending after it is removed, but simultaneous damage to this region and the axon stimulates extra outgrowth compared with axon injury alone (Chung et al., 2016). In this case, the extra outgrowth is not dependent on the standard DLK pathway.

Key unresolved issues include whether neurons with branched dendrites can survive simultaneous loss of axons and dendrites in vivo, and if they do survive, will they be able to regenerate both types of neurites? To address these issues we performed controlled injuries on the ddaE neuron in Drosophila larvae as this cell type has well-characterized responses to separate axon and dendrite injury. The ddaE neuron is a sensory neuron with a branched dendrite arbor that is positioned on the dorsal surface of Drosophila larvae and is a member of the simplest class of dendritic arborization neuron, Class I (Grueber et al., 2002). It is responsible for helping to coordinate movement (Hughes and Thomas, 2007) and signals in response to folding of the epidermis during larval crawling (He et al., 2019; Vaadia et al., 2019). Although it is a sensory neuron, ddaE has distinct axons and dendrites that are polarized in many of the same ways as mammalian central neurons. For example, axons have plus-end-out microtubule polarity, while branched dendrites contain minus-end-out microtubules (Hill et al., 2012; Stone et al., 2008), and dendrites contain protein synthetic machinery like ribosomes (Hill et al., 2012), while axons have little. Most importantly here, injury of ddaE axons triggers the conserved DLK/JNK signaling pathway (Rao and Rolls, 2017; Stone et al., 2014; Stone et al., 2010), which is used to initiate axon regeneration in many other neuron types from C. elegans (Hammarlund et al., 2009) to mice (Shin et al., 2012). Injury of dendrites does not initiate DLK signaling in these cells, and DLK, JNK and downstream transcription factor fos do not play a role in dendrite regeneration (Stone et al., 2014). In these neurons, laser-induced severing of the axon beyond approximately 50 microns from the cell body elicits classic axon regeneration from the severed stump (Rao and Rolls, 2017; Stone et al., 2012). In contrast, injury close to the cell body (within 20μm, see examples in Figure 1) converts a dendrite into a regenerating axon (Rao and Rolls, 2017; Stone et al., 2010), as described in mammalian neurons (Gomis-Ruth et al., 2008). In this study we severed axons close to the cell body as it allows length of the regenerating axon to be measured. ddaE neurons also regenerate after removal of one or all dendrites, and recapitulate the same number of branch points that the arbor contained before injury (Stone et al., 2014; Thompson-Peer et al., 2016).

Figure 1: DLK pathway activation after axon, dendrite and combined injury.

A. An overview of a Drosophila larva expressing mCD8-RFP, and 6xmCherry under 221-Gal4, and GFP driven from the puc locus is shown in the top left panel. The box indicates a dorsal neuron cluster. The other panels show examples of ddaE neurons injured at different sites. White arrows indicate cut sites. B. Cell bodies of ddaE neurons expressing puc-GFP and red cell shape marker are shown immediately after injury and 24h later. Green arrow indicates the nuclear puc-GFP used to measure DLK pathway activation. The blue arrow points to new neurites growing out from the cell body. C. Quantification of puc-GFP change from 0 hours post injury to 24 hours post injury in the indicated injury paradigm is shown. N for each condition is indicated within its column on the graph, error bars are graphed as the symmetric lower half of standard deviation. Asterisks denote statistical significance <0.001 as determined by unpaired two-tailed t-test.

Using this model system, we first investigated whether neurons of intact animals could survive simultaneous axon and dendrite injury. Surprisingly, we found that mature neurons survived removal of the axon and all dendrites and were able to initiate the conserved axon injury response pathway. We then tested whether neurons could regrow axons and dendrites over a longer timescale. We found that both types of neurites grew at the same time, and observed no reduction in growth compared to axon or dendrite regeneration alone.

Results

Neurons survive simultaneous axon and dendrite injury and initiate injury response signaling

To test whether ddaE neurons could survive injury to axons and dendrites at the same time, we performed different combinations of laser injuries in whole, intact Drosophila larvae. 221-Gal4 was used to express membrane-bound (UAS-mCD8-mRFP1) and soluble (UAS-6XmCherry) red fluorescent proteins (RFP) allowing for optimal visualization of cell shape and detection of severing. Incomplete cutting of an axon or dendrite could be detected by rapid recovery of mCherry into the region. Sites of axon and dendrite injury are indicated in Figure 1A. Images of neurons were acquired immediately after injury (0 h) and after 24 h. Between time points animals were returned to normal growth conditions. As expected, ddaE neurons survived removal of the axon or all dendrites (Figure 1B). Survival was also observed in all cases when axon severing was paired with removal of all but one dendrite (ax+den-1). When the axon and all dendrites (ax+den), are removed all that remains is a free cell body beneath epidermal cells. Normally dendritic arborization neurons are attached to the body wall, in part by integrin-mediated interactions between the dendrites and epidermal cells (Han et al., 2012; Kim et al., 2012; Yang and Chien, 2019). After stripping the neuron of all dendrites and the axon, the cell body tended to move. An example of this movement is evident in the 0 h cell body image (Figure 1B) from the double image of the axon stump captured in sequential frames. Some cell bodies drifted away into the body cavity, where we could no longer see them. By 24 h after injury, denuded cell bodies that remained in place had processes emerging from the cell body (Figure 1B).

To determine whether axon injury signaling occurred normally when axon injury was paired with dendrite injury, we used a reporter of the DLK/JNK pathway. Puckered (puc) is a MAP kinase phosphatase that is transcriptionally upregulated after axon injury (Xiong et al., 2010). We used a GFP insertion in the puc gene (Morin et al., 2001) that results in GFP accumulation in the nucleus after axon injury (Stone et al., 2014) as a readout of DLK/JNK activation. As previously reported, axon- but not dendrite- injury resulted in increased puc-GFP signal in the nucleus (Figure 1B and 1C). Similar increases were observed when axon injury was paired with dendrite injury (Figure 1B and 1C) indicating that dendrite injury does not interfere with axon injury signaling. We conclude that ddaE neurons can survive injury to both axons and dendrites, and even removal of all neurites. Moreover, axon injury signaling is similarly initiated when axons are injured alone or in conjunction with dendrites.

Neurons can regrow long unbranched and short branched processes at the same time

The ability of neurons to survive simultaneous axon and dendrite injury prompted us to investigate whether they could regenerate both types of neurites over longer time frames. We began by monitoring growth after leaving a single dendrite (ax+den-1) as this condition allowed neurons to remain anchored to the epidermal cells and left a dendrite as a substrate for conversion to an axon. Control axotomy and dendrotomy experiments, where either the axon or all dendrites were removed (middle panels of Figure 1A), were performed for comparison. In this study, unlike most of our previous studies, we killed the neighboring class IV ddaC neuron by aiming the pulsed UV laser at the cell body. The Gal4 driver used in this study results in some expression in the ddaC cell, whose dendrites overlap those of ddaE. In some cases, tracing regenerating ddaE neurites is confounded by overlapping ddaC dendrites, and this problem is eliminated by killing ddaC. After severing the ddaE axon close to the cell body in conjunction with ddaC ablation, a dendrite was converted to a regenerating axon as previously described (Figure 2A and (Stone et al., 2010)). Note one difference from our previous studies is the temperature at which animals were incubated between axon injury and monitoring outgrowth. In previous studies animals were incubated at 20C and imaged 96h after injury, while here animals were incubated at 25C and imaged after 72h. Development and growth occurs faster at 25C than 20C and so the two different time points are similar, though not identical. Under the conditions used in this study, new outgrowth from the tip of the original dendrite averaged 180 microns by 72h after injury (Figure 2D). Dendrite regrowth after removal of the entire dendrite arbor was also similar to that previously reported for these cells (Stone et al., 2014; Thompson-Peer et al., 2016), and the arbor regrew to the same complexity as before injury (Figure 2B and 2E). Uninjured ddaE neurons do not add new branches during the time window used (Stone et al., 2014; Sugimura et al., 2003), so branch addition represents a response to injury.

Figure 2: Analysis of neurite regrowth after different types of injury.

A–C. 221-Gal4 was used to express mCD8-mCherry and EB1-GFP in ddaE neurons. Example images of mCD8-mCherry are shown at 0 hours to 72 hours post injury. Red arrows indicate cut sites, green arrows point to new branches and the star indicates the tip of the growing axon. D. Quantification of tip growth in the axotomy and ax+den-1 injury paradigms is shown. See methods for a description of how measurements were made. E. Quantification of branchpoint addition in the axotomy, dendrotomy, and ax+den-1 paradigms. N for each condition is indicated within its column on the graph, error bars are graphed as the symmetric lower half of standard deviation. Asterisks denote statistical significance <0.001 as determined by unpaired two-tailed t-test.

Having recapitulated earlier studies that showed ddaE neurons could regenerate axons and dendrites separately, we injured axons and dendrites simultaneously, while leaving one dendrite intact (ax+den-1). The remaining dendrite initiated growth from its tip (Figure 2C) and was able to extend a similar length as neurons with only axon injury (Figure 2D). At the same time new branch points were added near the cell body (Figure 2C). The final number of branch points was similar to that when the axon was injured alone. Although, many additional branch points had to be added to reach this point when the comb dendrite was removed with the axon (Figure 2E). We conclude that ddaE neurons can not only survive contemporaneous axon and dendrite injury, but can extend a long neurite as well as add short neurite branches at the same time, suggesting that both axons and dendrites can regenerate simultaneously.

One result that we had not anticipated was the increase in dendrite branching in response to axon injury (Figure 2E). New branches typically initiated from existing dendrites and were quite short (Figure 2A). Similar excess branching was observed when axon injury way paired with removal of one dendrite (Figure 2E). As the addition of new dendrite branches in response to axon injury was unexpected, we considered several explanations. One possibility was that induction of growth by axon injury could result in membrane addition to dendrites if the axon was unable to extend. Support for this idea was provided by a negative association between new axon length and branch number (Figure 3A–3D). We also considered the possibility that tissue damage caused by ddaC ablation could increase dendrite branching. Some support for this possibility was obtained from experiments in which ddaC neurons were ablated in the absence of ddaE axon injury. In this case a small increase in ddaE dendrite branch number was observed (Figure 3E–3G). When the percent increase was compared to that of ddaE neurons without ddaC ablation, there was a trend (p-value 0.06) towards having a slightly higher increase in branching with ddaC ablation. Thus, there may be a minor effect of adjacent injury on branching, and likely some relationship between low amount of axon outgrowth and higher levels of branching.

Figure 3: The relationship between increases in ddaE branching and axon regrowth or ddaC ablation.

A and B. Representative images of ddaE neurons 0 hours and 72 hours after the indicated injury type. Neurons expressed EB1-GFP and mCD8-mCherry, but only the mCherry channel is shown. Red arrows indicate cut sites. C.D. Graphs showing the relationship between branchpoint addition and axon outgrowth in the indicated injury type. E.F. Representative images showing the morphology of uninjured ddaE neurons at 0 and 72 hours with and without ddaC ablation. G. Quantification of branchpoint addition in uninjured neurons from 0 to 72 hours following with and without ddaC ablation. N for each condition is indicated within its column on the graph, error bars are graphed as the symmetrical lower half of standard deviation.

New processes with axonal and dendritic microtubule polarity are generated after complete neurite removal

We next wished to determine whether neurons could regrow neurites after complete removal of the axon and all dendrites, and if so, to probe identity of the regrowing processes. While we knew that axons could survive at least 24h after removal of all neurites (Figure 1), we hypothesized that regrowing all neurites from an isolated cell body would be more challenging than using a remaining dendrite as a platform for regeneration. To our surprise, a single long process as well as shorter branched processes emerged from denuded cell bodies (Figure 4A). The long processes grew, on average, over 300 microns (Figure 4B), which is comparable to the amount of regrowth following axon removal alone (Figure 2D). The absolute amounts of growth are difficult to compare across these two injury paradigms as conversion of a dendrite to a new axon results in growth of a pre-existing neurite from the tip. As the animal grows larger during larval development, the pre-existing part of the neurite will also expand as it does in uninjured animals. We therefore normalize growth from pre-existing neurites to account for body expansion and get a better estimate for outgrowth from the tip (see Methods). When all neurites are removed, the new process grows de novo and so the measurement is an absolute value of outgrowth from the cell body. Denuded neurons also regrew branched processes, and as with dendrite removal alone, the pre-injury complexity was recapitulated by 72h (Figure 4A and 4C). Thus, even after removal of all neurites, cells were able to regrow a single long neurite as well as shorter branched ones.

Figure 4: Analysis of ddaE neurons after removal of all neurites.

A. ddaE neurons expressing EB1-GFP and mCD8-mCherry were subjected to complete neurite removal at 0h; images of mCD8-mCherry are shown. Cut sites are indicated with red arrows. Neurons were imaged 4h later (middle row) to confirm severing. Neurons were imaged again at 72h to monitor regeneration. Green arrows indicate new dendrite-like growth. Asterisks indicate the extending tip of the regenerating axon-like neurite. B. Quantification of absolute growth of the axon-like neurite during the 72 hours post injury is shown. C. Quantification of branchpoint addition during the 72 hours post injury is shown. D. Kymographs of EB1-GFP time series were generated from the new dendrite-like and axon-like processes 72h after injury. E. Quantification of microtubule polarity at 72h after injury in the new dendrite-like and axon-like processes after denuding, and in regenerating dendrites after dendrotomy is shown.

Based on morphology, we hypothesized that the long neurite was a regenerating axon and the shorter, branched ones were new dendrites. One of the most universal features that distinguishes axons and dendrites is microtubule polarity. In all species where it has been examined, the polarity of axonal microtubules is plus-end-out, meaning the more rapidly growing plus ends are oriented away from the cell body. In contrast, dendrites contain a significant population of minus-end-out microtubules (Baas and Lin, 2011; Rolls and Jegla, 2015). Mature ddaE neurons have axons with greater than 95% plus-end-out polarity and dendrites with greater than 90% minus-end-out polarity (Stone et al., 2008). To test whether polarity was reestablished after removal of all neurites, we assayed microtubule polarity by tracking movement of EB1-GFP. EB1 is recruited to microtubule plus ends as they grow (Akhmanova and Steinmetz, 2015; Jiang and Akhmanova, 2011), and direction of growth can be used to map microtubule polarity in neurons (Stepanova et al., 2003). Consistent with the long neurites having axonal identity, EB1-GFP comets moved almost exclusively away from the cell body in plus-end-out orientation (Figure 4D and 4E). The shorter branched neurites had comets moving in both directions (Figure 4D and 4E) and an overall polarity of about 60% minus-end-out microtubules, which is similar to new dendrites that are generated after removal of ddaE dendrites (Figure 4E). This difference in microtubule polarity supports the idea that both axons and dendrites can be correctly specified after removal of all neurites.

The endoplasmic reticulum concentrates at the regenerating axon tip in all injury paradigms

Thus far our data suggested that axon and dendrite regeneration could occur simultaneously after damage to both types of processes. To further test neurite identity, we assayed the distribution of an endoplasmic reticulum (ER) protein during different types of regeneration. Although the ER forms a continuous network throughout neurons, it becomes particularly concentrated at tips of regenerating axons, but not regenerating dendrites (Rao et al., 2016). One marker that can be used to track this distribution is Rtnl1-GFP (Rao et al., 2016). We expressed this tagged protein together with the plasma membrane marker mCD8-mCherry and analyzed its distribution after different injury combinations. As expected after axon removal, Rtnl1-GFP concentrated at tips of regenerating axons (Figure 5A). Compared to the base, regions near the tip of the growing neurite were about two-fold brighter. The remaining dendrites had roughly constant Rtnl1-GFP fluorescence along their length (Figure 5D). Similar fluorescence patterns were observed after axon injury was paired with removal of all, or all except one, dendrites (Figure 5B–D). The long neurite typically had a region of bright Rtnl1-GFP accumulation, while the only accumulations in the shorter, branched processes were slight increases at dendrite branch points, similar to those seen at branch points of dendrites that were not growing (Figure 5A). This difference in distribution of Rtnl1-GFP between the long neurite and short branched ones is consistent with them taking on axonal and dendritic identity respectively. We conclude that neurons respond to simultaneous axon and dendrite injury by initiating both regeneration pathways concurrently.

Figure 5: Rtnl1-GFP enriches at the tip of the growing axon-like neurite, but not the shorter dendrite-like neurites.

A–C. ddaE neurons expressing mCD8-mCherry and Rtnl1-GFP were injured at time 0 and imaged 72 hours later. Images of neurons after different types of injury from the 72 hour timepoint are shown. Asterisks denote the tip of the long axonal process. D. Quantification of the enrichment of Rtnl1-GFP in tips of regenerating neurites in the indicated injury paradigms. N for each condition is indicated within its column on the graph, error bars are graphed as the symmetrical lower half of the standard deviation.

Discussion

Using model neurons that have been shown to regenerate axons and dendrites individually, we determined that they could survive and regenerate after complete removal of all neurites. The neurons were able to reinitiate outgrowth of two different types of neurites at the same time. Moreover, coupling dendrite injury with axon injury did not reduce the amount of axon outgrowth by the cell or impair axon injury signaling. The two types of regenerating neurites were classified as axons and dendrites based on microtubule polarity and ER accumulation. We conclude that mature, functional neurons can recover from complete axon and dendrite removal in vivo and reinitiate polarized outgrowth of axons and dendrites.

After complete neurite removal the regenerating axon was identical to regenerating axons converted from dendrites after proximal axotomy, with plus-end-out microtubule polarity and ER concentrated near the growing tip. However, regenerating dendrites did not attain minus-end-out polarity. During embryonic development ddaE neurons have a mixed polarity phase, and this resolves to about 90% minus-end-out in the large dorsal comb dendrite during larval stages (Feng et al., 2019; Hill et al., 2012). During dendrite regeneration in large ddaC neurons, microtubule polarity is initially mixed, but by 24h after dendrite removal is mostly minus-end-out (Feng et al., 2019; Stone et al., 2014). One explanation for the lack of resolution to minus-end-out in regenerating ddaE dendrites is the difference in branch organization in the regenerated dendrite arbors compared to the ddaE comb dendrite or ddaC dendrites where polarity was previously analyzed. In the comb dendrite and ddaC dendrites, many branch points are found along each dendrite. The regenerated ddaE dendrites tend to emerge in more of a star shape close to the cell body and are quite short. As dendrite branch points are key control points for uniform microtubule polarity (Mattie et al., 2010), it is possible the different branching pattern could impact the ability to reinforce polarity.

In other neuron types, including ddaC neurons, simplification of dendrites is observed after axon injury (Chen et al., 2012). This response was not observed in ddaE neurons (Figure 2), and instead short branches were added to dendrites. The lack of retraction could be due to the lack of plasticity of ddaE dendrites compared to ddaC. During larval life, the large, complex arbor of ddaC undergoes branch addition and retraction, with a net increase in branching (Sugimura et al., 2003). In contrast, the shape of ddaE is fixed very early in larval life (Stone et al., 2014; Sugimura et al., 2003). If axon injury shifts the balance of retraction and addition in ddaC, net simplification could occur. With no ongoing dynamics in ddaE, more active removal of branches would need to take place. The branch addition we observed in ddaE seemed related to two different phenomena: first, an inverse relationship with the amount of axon outgrowth, and second, additional damage due to the ablation of a neighboring cell.

The ability of ddaE neurons to survive complete removal of all neurites is similar to that of embryonic cortical and hippocampal neurons dissociated from brains and recovered in culture. In culture, media is enriched to promote survival. Here the cell body has to rely on endogenous growth factors. ddaE neurons are normally closely associated with the epidermis and receive signals from epidermal cells during development (Li et al., 2016). It is possible that epidermal cells could increase growth factor production to support the cell after injury. Alternatively, ddaE neuron survival may not depend on trophic factors from surrounding cells.

The temporal pattern of ddaE outgrowth after complete denuding is slightly different from that of primary neuronal cultures or ddaE neurons during development. Cultured neurons first grow short unspecified processes, one of which becomes the axon at day two or three. The axon then grows, while the unspecified processes wait for several more days before initiating growth as dendrites (Craig and Banker, 1994; Dotti et al., 1988). During embryonic development, ddaE neurons also send out axons well before dendrite growth is initiated (Feng et al., 2019; Hill et al., 2012). After removal of all neurites, however, dendrite regrowth initiated rapidly, and all neurons had short branched processes like those seen in Figure 1B at 24h after injury. Thus, unlike primary neuronal cultures or embryonic ddaE neurons, ddaE neurite regrowth does not proceed through sequential polarization with a first axon-only growth phase followed by emergence of dendrites.

Overall, one striking aspect of the response to axon and dendrite removal was its robustness. Despite axon regeneration requiring reinitiation of large-scale outgrowth, it was not dampened by simultaneous regrowth of dendrites. This indicates that even though axon regeneration is itself impressive, it does not saturate the growth capacity of the neuron. Indeed, the ability of ddaE neurons to regenerate at all is somewhat surprising. These neurons function primarily during larval life where they respond to epidermal folding as the larva crawls (He et al., 2019; Vaadia et al., 2019). In some specific segments ddaE neurons die during pupariation, while in others they survive, remodel and are observed in young adults; all are gone by the end of the first week of adult life (Shimono et al., 2009). Despite the relatively short utility of these neurons during the lifespan, they are able to survive and regenerate when they are injured.

If even transient ddaE neurons can survive complete neurite removal and initiate polarized regeneration, it seems likely that this capacity exists in other neuron types as well. It will be extremely interesting to probe other cell types in invertebrates, and also vertebrates. So far, however, it has not even been possible to determine whether vertebrate neurons can regenerate in vivo after dendrite removal. Vertebrate neuron cell bodies are typically protected under bone and so controlled injuries to dendrites will require overcoming technical hurdles. However, survival of neurons after simultaneous axon and dendrite damage could play a role in the plasticity that enables improvement after stroke, and perhaps also TBI.

Methods

Fly culture

All flies used for this study were maintained at 25C on cornmeal agar media with a per liter composition of: 4.5g agar, 15.5g yeast, 25.9g sucrose, 51.7g dextrose, 85.8g cornmeal, 4ml of propionic acid and 6ml tegosept. Flies used for experiments in this study were moved to fresh media every 24 hours to facilitate selecting appropriate age offspring, and the media with eggs was aged for 48h at 25C to generate larvae for imaging.

Fly lines

Two different tester lines were used in this study. For the puc-GFP assay, the tester line was UAS-mCD8-RFP; 221-Gal4, puc-GFP. All other assays used the 221-Gal4, UAS-mCD8-mCherry tester line. Virgin females were collected from each tester line and crossed to the appropriate males for each experiment. Males from the following lines were used: 20xUAS-6xmCherry, UAS-EB1-GFP, and UAS-Rtnl1-GFP. Information about each line is in the Key Resources Table.

puc-GFP assay

Virgin females from the UAS-mCD8-RFP; 221-Gal4, puc-GFP tester line were crossed to males from the 20xUAS-6xmCherry line. Whole, live two-day old larvae were selected and mounted for imaging on a plain glass slide. A 22×40 mm coverslip was placed over the larva and held down with tape to prevent larval movement. A Zeiss inverted LSM800 equipped with an Andor MicroPoint UV pulsed laser was used for neurite severing and imaging. All images were acquired using GaAsP detectors. For GFP quantification, cells were imaged with a Zeiss 63× 1.4NA oil immersion objective set to a zoom of 3, a dwell time of 2.06 microseconds per pixel, 2x averaging, a pinhole of 43 microns and a 512×512 image size. The ddaE neurons were subjected to the indicated type of axon and/or dendrite damage, and a post injury z-stack image was acquired. Larvae were then incubated at 25°C on fly media for 24 hours before being re imaged. For quantification, the brightest 3 slices from each Z stack were manually aligned and combined using the maximum projection feature of ImageJ. Quantification shows fold change in nuclear GFP intensity from 0 hours post injury to 24 hours post injury.

Regeneration assays

Virgin females from the 221-Gal4, UAS-mCD8-mCherry tester line were crossed to males from the UAS-EB1-GFP line. Whole, live two-day old larvae were mounted on a glass slide with dried 3% agarose pad. A 22×40 mm coverslip was taped down on top of the animal with enough pressure to hold it still, but not damage it. Laser injury was performed on a Zeiss widefield AxioImager.M2 equipped with an Andor Micropoint UV pulsed laser. Sites of axon and dendrite injury in the ddaE neuron for each of the paradigms are shown in Figure 1A. Class IV ddaC neurons were killed by concentrating the pulsed UV laser at the nucleus to eliminate overlapping dendrites that might make it difficult to trace ddaE regrowth. After injury, mounted animals were moved to a Zeiss LSM 800 upright microscope and imaged using a 63× 1.4 NA oil objective and GaAsP detectors. After imaging, larvae were returned to fresh Drosophila food and kept at 25°C until they were remounted for follow up imaging at 72h. For the ax+den condition larvae were also imaged at 4h to ensure the cell was completely denuded.

Quantification of axon regeneration

Average tip growth for the regenerating axon in axotomy alone was calculated by measuring the length of the dendrite that initiated tip growth immediately after severing and again at 72hrs. Another dendrite that did not initiate tip growth was also measured to account for normal developmental growth. Tip growth beyond normal growth (regeneration) was calculated using the following formula:

$${\text{Regeneration}}_{\text{tip}\text{growth}}={\text{Length}}_{\text{new}\text{axon}}-\left[{\text{Length}}_{0\text{h}}/\left({\text{Length}}_{0\text{h control dendrite}}/{\text{Length}}_{72\text{h control dendrite}}\right)\right]$$

Average tip growth for the regenerating axon in ax+den-1 was quantitated using the same formula except the average ratio of Length _{0h control dendrite} / Length _{72h control dendrite} from the axotomy alone experiments was used to account for normal developmental growth since the comb dendrite where a control dendrite is normally found has been removed in this condition. Average tip growth for the regenerating axon in ax+den was quantitated by measuring total outgrowth from the cell body of the longest neurite.

Quantification of dendrite regeneration

Dendrite regeneration was quantitated by counting all visible branch points in the dendrite arbor at 0h and 72h. Average number of branch points at each time point was plotted.

Microtubule polarity assay and kymograph generation

Microtubule polarity was assayed in regenerating neurites after ax+den injury and regenerating dendrites after dendrotomy by tracking the movement of EB1-GFP comets either towards or away from the cell body. EB1-GFP videos to determine polarity were captured at 1 frame per 1.17, 0.93 or 1.86 seconds. Analysis was done using Image J software and only comets that were visible in three consecutive frames were scored. Kymographs were made using the Multi Kymograph Image J plugin, using a line width of 1.

Rtnl1 localization assay

Virgin females from the 221-Gal4, UAS-mCD8-mCherry tester line were crossed to males from the UAS-Rtnl1-GFP line and regeneration assays were performed as described above. Rtnl-GFP intensity in the regenerating axon/axonal processes from the axotomy only, ax+den-1, and ax+den conditions were measured using the previously described method (Rao et al., 2016). For regenerating axon/axonal processes, the brightest 10μM within 100μM regions at the tip and base were used for analysis. For the regenerating dendritic process, the brightest 10μM within 20μM regions at the tip and base of the longest regenerating dendritic process were used. Intensity measurements were performed using ImageJ (Schindelin et al., 2012).

KEY RESOURCES TABLE

Reagent or resource	Source	Identifier
Experimental Models: Organisms/Strains
221-Gal4	Wesley Grueber	Gift
Puc-GFP	FlyTrap Project	G00426, Morin et al. PNAS 2001
UAS-mCD8-RFP	Bloomington Drosophila Stock Center	BL27398
UAS-6xmCherry	Bloomington Drosophila Stock Center	BL52268
UAS-RtnM-GFP	Rolls Lab	Rao et al. Mol Biol Cell 2016
UAS-EB1-GFP	Tadashi Uemura	Gift
UAS-mCD8-mCherry	Bloomington Drosophila Stock Center	BL27392
Software and Algorithms
ImageJ	https://imagej.net/Fiji	Schindelin et al. Nat Meth 2012

Highlights.

A Drosophila model neuron survives simultaneous axon and dendrite removal in vivo.

Axon- and dendrite-like neurites can regenerate at the same time.

Axon and dendrite identity can be re-established in mature neurons in vivo.