Microtubule organization of vertebrate sensory neurons in vivo

Abstract

Dorsal root ganglion (DRG) neurons are the predominant cell type that innervates the vertebrate skin. They are typically described as pseudounipolar cells that have central and peripheral axons branching from a single root exiting the cell body. The peripheral axon travels within a nerve to the skin, where free sensory endings can emerge and branch into an arbor that receives and integrates information. In some immature vertebrates, DRG neurons are preceded by Rohon-Beard (RB) neurons. While the sensory endings of RB and DRG neurons function like dendrites, we use live imaging in zebrafish to show that they have axonal plus-end-out microtubule polarity at all stages of maturity. Moreover, we show both cell types have central and peripheral axons with plus-end-out polarity. Surprisingly, in DRG neurons these emerge separately from the cell body, and most cells never acquire the signature pseudounipolar morphology. Like another recently characterized cell type that has multiple plus-end-out neurites, ganglion cells in Nematostella, RB and DRG neurons maintain a somatic microtubule organizing center even when mature. In summary, we characterize key cellular and subcellular features of vertebrate sensory neurons as a foundation for understanding their function and maintenance.

Introduction

Dorsal root ganglion (DRG) neurons mediate the majority of touch sensory input for vertebrates. The cell bodies of DRG neurons reside in ganglia adjacent to the spinal cord and extend neurites into the spinal cord and into the periphery where they innervate the skin (Lumpkin et al., 2010; Nascimento et al., 2018). Sensory endings embedded in the skin receive and integrate information and transmit it centrally to interneurons in the spinal cord. The sensory endings thus function like many dendrites as the input region of the neuron. Despite these neurons having been originally characterized in the 1800s, with thousands of papers published since, it is still unclear if DRG sensory endings possess any of the cellular features that distinguish dendrites from axons.

Whether DRG sensory endings have cellular features of axons or dendrites is more than an academic or semantic question. Cargo is transported into axons and dendrites using different machinery (Harterink et al., 2018; Kapitein et al., 2010; Zheng et al., 2008), and disruption of neuronal transport has been linked to several types of neurodegenerative disease (Guillaud et al., 2020; Prior et al., 2017). In addition, injured dendrites do not use the same machinery as axons to sense damage and initiate regeneration (Stone et al., 2014), so determining whether DRG sensory endings have axonal or dendritic features could inform our understanding of how they regenerate.

Efficient cargo transport and injury sensing are both particularly important for long-term DRG function. To sense the environment, their endings need to be positioned such that they are exposed and vulnerable. They are thus particularly vulnerable, and failure to maintain sensory endings leads to peripheral neuropathy. In 2016 alone 130,000 lower extremity amputations were performed due to downstream effects of sensory neuron impairment in diabetes. In order to understand sensory ending maintenance and regeneration, it is critical to know whether this compartment shares the basic features that underlie transport and injury responses with axons or dendrites.

The stereotyped neuron presented in biology classes resembles a motor neuron, and defines dendrites as being arborized and post-synaptic, and axons as singular and unbranched until reaching their terminus, where they are presynaptic and signal to the downstream cell (Alberts et al., 2007). Even 60 years ago this model was recognized as insufficient to describe common cell types including touch sensing neurons that have large arborized receptive fields that are not post-synaptic (Bodian, 1962). Bodian proposed that despite not being post-synaptic, sensory endings in the skin should be considered dendrites. In this framework, dendrites generate signals in the skin and the axon begins as sensory neurites converge and exit the skin to bundle into the nerve taking signals towards the central nervous system. At the time this model was proposed, few intracellular differences between axons and dendrites were known. With the ability to visualize intracellular features, additional differences between axons and dendrites emerged and can be brought to bear on the nature of sensory endings. For example, now we know that dendrites contain ribosomes, Golgi outposts and microtubules oriented with minus ends distal to the cell body (minus-end-out), all of which are rare in axons (Craig and Banker, 1994). Subcellular features have been examined in invertebrate branched sensory endings; Drosophila mechanosensory neurons with branched arbors possess dendritic cellular signatures including minus-end-out microtubules (Stone et al., 2008), Golgi outposts (Ye et al., 2007) and ribosomes (Hill et al., 2012), while their axons projecting to the central nervous system do not. Sensory endings of vertebrate DRGs have not been examined in the same way.

To develop specialized neurites with distinct contents, cargoes need to be differentially distributed from the major site of synthesis in the cell body to their site of function. In many neurons, microtubule polarity plays a central role in the ability to direct cargoes to axons and dendrites. Axonal microtubules have microtubules with their dynamic plus ends pointed away from the cell body, while dendrites have mixed orientation in vertebrates or minus-end-out orientation in invertebrates (Baas and Lin, 2011; Rolls and Jegla, 2015). Thus, the presence of minus-end-out microtubules distinguishes dendrites from axons and allows the minus end-directed motor dynein to deliver cargo specifically to dendrites (Kapitein et al., 2010). While the role of microtubule polarity in dendrite identity is easiest to conceptualize in multipolar neurons in which axons and dendrites emerge separately from the cell body, loss of dynein also specifically affects dendrites in unipolar neurons in Drosophila (Liu et al., 2000). While microtubule layout has only been examined in a couple of unipolar neuron types in flies, the primary neurite that emerges from the cell body has plus-end-out microtubule polarity, as does the axon. The dendrites that branch from the primary neurite, in contrast, have minus-end-out polarity (Stone et al., 2008). Thus, dendrites with minus-end-out microtubules can emerge from plus-end-out processes, and dynein-mediated cargo transport is essential to their identity. We therefore hypothesized that the sensory endings of DRG neurons could have mixed or minus-end-out microtubule polarity despite originating from an axon.

In primary cultures, adult DRG neurons typically generate one to four neurites that have plus-end-out axonal microtubule polarity and receive similar cargoes (Gumy et al., 2017; Moughamian et al., 2013). However, sensory endings do not differentiate in vitro, restricting assessment of their microtubule polarity to in vivo settings. In vivo assessment of microtubule polarity in peripheral nerves containing sensory neuron projections has been conducted. Trigeminal neurons, functional equivalents of DRGs that innervate the head region, have plus-end-out microtubule polarity in peripheral nerves to the cornea (Topp et al., 1994). The branched endings that innervate the cornea were not, however, examined, and we are not aware of studies examining microtubule polarity in sensory endings of DRG neurons. To our knowledge, microtubule organization has only been described in sensory endings of one vertebrate cell type at a single timepoint early in development. Rohon-Beard (RB) neurons perform similar functions to DRG neurons in embryonic and larval fish and amphibians, including zebrafish (Ribera and Nusslein-Volhard, 1998; Spitzer, 1984). Their cell body sits in the spinal cord and elaborate branched sensory endings innervate the skin (Katz et al., 2021; Palanca et al., 2013). In zebrafish these cells die after several weeks (Williams and Ribera, 2020) as DRG sensory endings take over skin innervation (Rasmussen et al., 2018). Analysis of microtubule organization showed that peripheral RB arbors have plus-end-out microtubule polarity in embryos one day after fertilization (Lee et al., 2017), but later timepoints were not reported.

Two primary methods have been used to assess microtubule polarity in neurons. The first method was based on the observation that under specific buffer conditions free tubulin polymerizes onto the sides of microtubules in sections prepared for electron microscopy to form visible “hooks” whose curvature reflects microtubule orientation (Heidemann and McIntosh, 1980). This method was used to determine that axonal microtubules in a variety of neuron types have uniform plus-end-out polarity (Baas and Lin, 2011) and that dendrites in cultured hippocampal neurons have mixed polarity (Baas et al., 1988). However, this method is quite cumbersome and would be difficult to apply to sensory endings in the skin as they form a complex network in which it would be difficult to determine orientation of neurites in sections. The second method for polarity assessment relies on live imaging of proteins including end-binding proteins 1 and 3 (EB1 and EB3) that bind to polymerizing microtubule plus ends (+TIPs) (Baas and Lin, 2011). It was first used to confirm microtubule polarity in cultured hippocampal neurons (Stepanova et al., 2003) and has since been applied to a variety of vertebrate and invertebrate neurons in vivo (Goodwin et al., 2012; Kleele et al., 2014; Maniar et al., 2012; Rolls et al., 2007; Stone et al., 2020; Stone et al., 2008; Yau et al., 2016). This +TIP-tracking method seemed most promising to use for probing polarity of sensory endings.

As +TIP tracking requires live imaging to assess microtubule polarity, optical accessibility was a key determinant for selection of a vertebrate in which to image sensory endings. While mice are a popular vertebrate model, their fur and high autofluorescence in the skin make them a poor choice for optical interrogation of fine sensory endings. In contrast, zebrafish are a vertebrate model organism with excellent transparency and low fluorescent background at early life stages, and quite amenable to imaging into adulthood if an appropriate mutant background is selected (Antinucci and Hindges, 2016; Moss et al., 2013; White et al., 2008). In this study we therefore used live imaging of +TIPs in zebrafish to evaluate microtubule polarity of vertebrate sensory endings.

Results

Rohon-Beard (RB) sensory endings maintain axonal microtubule polarity at maturity

RB neurons in embryonic and larval zebrafish have been used extensively for live imaging (Andersen et al., 2011; Andersen and Halloran, 2012; Lee et al., 2017; Liu and Halloran, 2005; Paulus et al., 2009; Wang et al., 2012). RB neurons begin central axon outgrowth 15–17h after fertilization (Andersen et al., 2011; Kuwada et al., 1990), and peripheral axon outgrowth initiates shortly afterwards (Andersen et al., 2011). To complement analysis of microtubule organization in the peripheral arbor at 24h post fertilization (Lee et al., 2017), we wished to observe the polarity of the endings all the way to their distal tip when the arbors are mature and stable.

To assess microtubule polarity in RB sensory endings, we expressed a fusion of GFP to mouse EB3, which has previously been used in zebrafish to track growing plus ends (Distel et al., 2010). The Isl1[ss] promoter, which expresses in RB neurons (Higashijima et al., 2000; Palanca et al., 2013), was used to drive expression of EB3-GFP (Isl1[ss]:EB3-GFP). To generate animals in which individual RB peripheral arbors could be observed without overlap from adjacent cells, the EB3-GFP expression plasmid was injected at the single cell stage and embryos with sparse, well resolved neurons were selected for imaging (Figure 1A). Early larval zebrafish lack gills and have low oxygen requirements so they can be immobilized for imaging by embedding anesthetized animals in agarose. We used this mounting strategy to image EB3-GFP comet movements in RB neurons at 24–96 hours post fertilization (Figure 1B and C). Direction of comet movement (comets moving away from the cell body were scored as plus-end-out) was assessed in kymographs generated from the videos (Figure 1D and E).

Figure 1. Zebrafish RB neurons have uniform plus-end-out microtubule polarity.

A. Schematic illustration of zebrafish egg injection. B. Maximum projection of a time-series image of a 1dpf zebrafish RB neuron expressing EB3-GFP controlled by Isl1[ss]. C. Maximum projection of a time-series image of a 4dpf zebrafish RB neuron expressing EB3-GFP D. Graph displaying quantification of comet direction in the central and peripheral neurites of RB neurons at timepoints from 1dpf to 4dpf; number on column is the number of comets counted for that condition. E. Sample kymographs of RB neurons expressing EB3-GFP at 1 and 4dpf.

As expected based on the previous analysis of microtubule growth in RB neurons 24 hours post fertilization (Lee et al., 2017), EB3-GFP comets moved away from the cell body at tips of plus-end-out microtubules at this time point (Figure 1D and E and Movie 1). To determine whether the central axon also has plus-end-out microtubule polarity as has been assumed, we assayed EB3-GFP behavior in RB neurites within the spinal cord. These comets also moved away from the cell body (Figure 1D). Thus, at 24 hours post fertilization both neurites have the same microtubule polarity.

During development of Drosophila sensory arbors, dendrites have plus-end-out microtubule polarity during the very first stages of outgrowth (Feng et al., 2019) and then minus-end-out microtubules are added as the arbors develop (Feng et al., 2019; Hill et al., 2012). The mature minus-end-out polarity is not established until the neurons have been in place and functioning for several days (Hill et al., 2012). We therefore hypothesized that RB peripheral neurite polarity might change over several days of maturation. We imaged EB3-GFP in peripheral and central neurites daily up to 96h. At each timepoint the central neurite was plus-end-out and the peripheral arbor was also plus-end-out all the way out to the tips (Figure 1D).

While RB neurons are still present two to three weeks post fertilization (Palanca et al., 2013; Williams and Ribera, 2020), cell death during larval development (Williams and Ribera, 2020) combined with expression in the replacement DRG neurons made labeled RB neurons easier to find in transgenic animals than in sparse mosaic animals. We therefore generated a transgenic zebrafish line expressing LexA under the control of the a previously characterized 305bp fragment (−1036: −731) of the pufferfish p2×3–2 promoter (Palanca et al., 2013), and EB3-GFP driven by a 4xLexOP, which we will refer to as P2rx3a>EB3-GFP. Melanocytes and iridophores migrate into the skin to absorb and scatter light and make imaging of whole animals more difficult as they mature. We therefore generated the P2rx3a>EB3-GFP line in a casper (mifta^w2/mifta^w2, roy^a9/roy^a9) mutant background, which lacks both cell types (White et al., 2008). To further optimize our ability to image neuron structure and microtubule behavior, we created composite transgenic lines that included other LexA/LexOP transgenes with complementary properties. For earlier timepoints, because Isl1[ss] turns on earlier than p2×3–2 (Palanca et al., 2013) and to include a cell-shape marker, we bred P2rx3a>EB3-GFP to a casper line containing Isl1[ss]>tdTomato to generate a double transgenic fish line. Isl1[ss] drives expression transiently, so for later timepoints we generated a separate line by breeding P2rx3a>EB3-GFP to casper P2rx3a>mCherry. For both combinations, we selected the brightest fish expressing both green and red markers in each generation for imaging (Figure 2). We confirmed that EB3-GFP expression in transgenic lines yielded similar results as transient expression with the Isl1 promoter by imaging larvae three days post fertilization (Figure 2C and H and Movie 2).

Figure 2. Zebrafish RB neuron microtubules remain plus-end-out for the life of the cell.

A. Overview of 3dpf zebrafish embryo bearing transgenic insertions of P2rx3a>EB3-GFP, P2rx3a>mCherry and Isl1[ss]>tdTomato showing the location and abundance of RB neurons at 3dpf. Enlarged image shows exit of neurites from the cell bodies in the spinal cord. B. Higher magnification image showing the morphology of RB neuron sensory endings. Transgenes are not expressed uniformly in all neurons so some cells are labeled red, others green and some both. C. A representative kymograph generated from RB sensory endings is shown. D. Overview of the posterior portion of a 21dpf zebrafish showing the position of DRG neuron cell bodies, the morphology of the peripheral nerves and the strong reduction in visible RB neurons relative to 3dpf. E. higher magnification region showing a rare visible RB neuron in a 21dpf zebrafish. F. image of a quantified portion (blue box) of RB sensory neurite in a 21dpf zebrafish. G. Sample kymograph of RB sensory neurite in a 21dpf zebrafish neuron expressing EB3-GFP H. Graph showing quantification of comet direction in 3dpf and 21dpf RB sensory neurites expressing EB3-GFP; number on column is the number of comets counted for that condition

Using this combination of transgenic lines, we were able to image RB peripheral arbors up to three weeks post fertilization. At this timepoint, viability is compromised when animals are fully embedded in agarose, so we selectively anchored the tail to the coverslip and imaged this region of the animal. As at earlier timepoints, the entire sensory arbor had plus-end-out microtubule polarity (Figure 2G and H and summarized in Table 1). Thus, RB neurons contain only plus-end-out microtubules in central and peripheral neurites throughout their life.

Table 1. Summary of microtubule polarity results.

The table summarizes the different cells, developmental times, and regions from which movies of EB3-GFP were acquired.

Cell type, location and age	microtuble polarity
RB centrally projecting axon
1dpf	plus-end-out
2dpf	plus-end-out
3dpf	plus-end-out
4dpf	plus-end-out
RB peripherally projecting axon	plus-end-out
1dpf	plus-end-out
2dpf	plus-end-out
3dpf	plus-end-out
4dpf	plus-end-out
21dpf	plus-end-out
DRG central neurite (proximal to cell body) 4–5 weeks post fertilisation	plus-end-out
DRG peripheral neurite (proximal to cell body)4–5 weeks old	plus-end-out
DRG peripheral neurite nerve at surface 8–11 months old	plus-end-out
DRG sensory endings at surface 8–11 months old	plus end out

Zebrafish DRG neurons retain bipolar morphology into adulthood.

Because RB neurons are present only early in development, we also wished to examine microtubule organization in DRG neurons that mediate most somatosensation in mature vertebrates. RB and DRG neurons have several significant structural differences. RB peripheral neurites do not bundle into axons prior to entering the skin (O’Brien et al., 2012), while DRG peripheral neurites form nerves that are ensheathed in glia. In adult fish these nerves bring the neurites to the base of the scale, and bundles are also found on the surface of mature scales (Rasmussen et al., 2018). Free sensory endings emerge from these bundles to innervate the skin on the surface of the scale (Rasmussen et al., 2018); RB sensory endings innervate the immature skin prior to scale formation.

Another structural difference that could impact microtubule organization and trafficking is the arrangement of neurites as they exit the cell body. RB neuron cell bodies are located within the spinal cord and typically have ascending and descending central axons that emerge from either side of the cell body. Subsequently the peripheral neurite exits the cell body either directly or from the base of one of the central axons (Andersen et al., 2011; Andersen and Halloran, 2012; Wang et al., 2012). The peripheral axon is larger than either central one, and so is easier to visualize. In most cases at both 3 dpf (Figure 2A) and 21 dpf (Figure 2E), the cell body maintains two exits: one for a central axon and one dominated by the much larger peripheral neurite. An occasional neuron with a single visible initial process was also observed (asterisk in Figure 2A). Rodent DRG neurons also initially grow separate central and peripheral axons from the cell body (Takahashi and Ninomiya, 1987), but later the cell body seems to move aside and a single neurite root is formed (Takahashi and Ninomiya, 1987). Eventually a single fairly long primary neurite emerges from the cell body and gives rise to the central and peripheral neurites in a morphology termed pseudounipolar (Takahashi and Ninomiya, 1987). The description of DRG neurons as pseudounipolar originated in the nineteenth century (Nascimento et al., 2018; Takahashi and Ninomiya, 1987), but the organization of the dorsal root ganglia and resident neurons has not been investigated in zebrafish. We therefore took advantage of the ability to image inside transparent zebrafish to track cellular morphology of neurons in the DRG at different ages.

Very early stages of DRG development have been visualized in zebrafish embryos using the neuroG promoter and two neurites were seen emerging from the cell body as in RB neurons (McGraw et al., 2008). To determine whether zebrafish neurons would transition to a pseudounipolar shape as described in larger mammals, we imaged the DRG in 5 week old animals (Figure 3A–E). Surprisingly, most neurons in the DRG maintained two completely separate neurites (Figure 3D), and as well as projecting differentially to either the spinal cord or periphery, these neurites were distinguished by caliber. Central axons were almost always much thinner than peripheral neurites (Figure 3D). Both central and peripheral neurites had plus-end-out microtubule polarity (Figure 3E). We were able to identify one neuron with a single neurite root exiting the cell body (Figure 3D asterisk) indicating that this morphology is possible, even though less common than true bipolarity. As this raised the possibility that DRG neuron morphology in 5 week animals might still be transitional, we also imaged DRGs in older fish. Despite much thicker tissue, we were able to image the posterior most DRG clusters in 3 month old, mating age fish (Figure 3G) using two-photon excitation (Figure 3H, 3I). The weaker Z axis discrimination of two-photon microscopy relative to confocal made it difficult to visualize distinct centrally projecting neurites in the middle of the DRG cluster (Figure 3I), however isolated cells along the edge of the cluster could be clearly seen to be bipolar (Figure 3H, 3I). As in younger fish, the central axon was distinguished from the peripheral neurite by its thinness (Figure 3I). We conclude that in zebrafish most RB and DRG neurons maintain bipolar morphology at maturity, although pseudounipolar cells are also present at low frequency.

Figure 3. Most zebrafish DRG neurons retain bipolar morphology throughout development.

A. Brightfield overview of 5 week old casper zebrafish. B. Confocal image of the posterior region of the zebrafish spine showing the spinal cord, DRG neurons and their nerves. C. higher magnification image showing DRG neurons cell bodies and their centrally and peripherally projecting neurites. D. Airyscan super-resolution images of DRG neurons expressing EB3-GFP, showing clear centrally and peripherally projecting neurites originating from separate locations on the cell body marked pink and blue respectively. Asterisk indicates rare neuron with a cell body projection from which both neurites emerge. E.F. Representative kymographs from the centrally and peripherally projecting neurites respectively of 5-week-old zebrafish DRG neurons expressing EB3-GFP G. Brightfield overview of 3 month old zebrafish. H. 2 photon image of the posterior DRG cluster of a 3 month old zebrafish expressing EB3-GFP I. Higher magnification image of a DRG neuron at the edge of the cluster. Central and peripheral neurites are indicated with pink and blue lines respectively

Adult DRG peripheral neurites have plus-end-out microtubules in their receptive field.

Having shown that the architecture and microtubule arrangement of RB and DRG neurons is similar near the soma, we wished to analyze more peripheral regions of DRG neurons, including the free sensory endings that extend into the epidermis over the scale surface (Rasmussen et al., 2018). In order to characterize microtubule behavior in mature DRG neurons, we used our P2rx3a>EB3-GFP, P2rx3a>mCherry transgenic casper mutant fish line. The presence of the mCherry was especially helpful in visualizing fine sensory endings.

For imaging, we selected 8–10 month old zebrafish at least 25mm in length (Figure 4A) that are mature adults. In order to avoid potential artifacts from hypoxia or other complicating factors, we intubated the fish using a flow cell based on a previously published design (Xu et al., 2015), but adapted for use with an inverted confocal microscope (Figure S1). This gave us the ability to mount the anesthetized fish on the confocal microscope for hour-long imaging sessions, and recover them afterwards if necessary.

Figure 4. Zebrafish DRG sensory endings have uniform plus-end-out microtubule polarity throughout their entire arbor.

A. Brightfield overview of an 8–10 month-old adult zebrafish. B. Diagram of a single DRG sensory neuron innervating a posterior scale. C. Confocal z-stack of a scale innervated by DRG sensory neurons expressing mCherry and EB3-GFP D. False-colored time series z-stack image of the surface of a scale on an 8–10 month-old zebrafish expressing EB3-GFP with successive timepoints batched into colors progressing from the blue to red end of the spectrum, showing a series of rainbows with their red end oriented towards the end of the neurite, reflecting a plus end out microtubule polarity. E. Quantification of microtubule polarity from movies of EB3-GFP in DRG nerves and endings in 8–10 month-old zebrafish; number on column is the number of comets counted for that condition F.G. Representative kymographs of EB3-GFP in the nerve and sensory endings respectively. H. Kymograph generated from a movie of EB3-GFP in a sensory ending in which both growing minus ends (blue) and plus ends (pink) are visible.

We collected EB3-GFP time-lapse movies from neurites bundled within nerves as well as from free sensory endings on the surface of the scale (Figure 4). Although it was difficult to resolve individual neurites in the nerve, all EB3-GFP comets moved in the same direction, away from the cell body, indicating uniform plus-end-out polarity. Similar plus-end-out polarity was observed in the defasciculated and branched free endings on the scale (Figure 4D–G). In single plane timeseries movies we were able to track direction of comet movement and used this strategy for most of our analysis. However, in any individual movie only a short section of neurite was in focus. To more completely visualize microtubule growth in sensory endings, we acquired time series Z-stacks that clearly show EB3-GFP comets moving from the nerve to the tips of sensory endings (Figure 4D and Movie 3). To determine whether other parameters of microtubule behavior might be regionally different in DRG peripheral neurites, we analyzed plus end growth speed from these movies. While speed varied somewhat from fish to fish, we did not observe regional differences (Figure S2). We conclude that even in adult zebrafish, microtubules have plus-end-out microtubule polarity in sensory endings in the skin (Figure 4B).

To further compare microtubule organization in mature DRGs to that in larval RB neurons, we wanted to visualize microtubule minus end behavior. Minus ends in mammalian axons have been difficult to visualize, but are thought to be capped by short stretches of CAMSAP2 (Yau et al., 2014) or organized by nucleation sites recruited to the side of pre-existing microtubules by the HAUS/augmin complex (Cunha-Ferreira et al., 2018; Sanchez-Huertas et al., 2016). In fine neurites in vivo, however, we previously reported that microtubule minus ends do not seem to be capped and instead grow persistently. We characterized this minus end growth most completely in dendritic sensory endings in Drosophila, where we could show that it was important for organization of the microtubule network (Feng et al., 2019). We observed a similar phenomenon in RB sensory endings in zebrafish (Feng et al., 2019). While difficult to detect in more optically challenging adult zebrafish, we did find some EB3-GFP spots that moved in the opposite direction, and slower than plus ends (Figure 4H). We conclude that microtubule minus ends are capable of growth in mature DRG sensory endings. Thus, overall, microtubule organization is similar in RB and DRG neurons.

Sensory neurons maintain a somatic hotspot of microtubule organization.

Neurons with well-defined axons and dendrites that have distinct arrangements of microtubules inactivate the central somatic microtubule organizing center (MTOC) as they mature (Lindhout et al., 2021). In contrast, sea anemone neurons with two or three axon-like neurites maintain a somatic MTOC (Stone et al., 2020). To determine whether somatic microtubules in vertebrate neurons with only axon-like processes are organized around a MTOC as they mature, we analyzed somatic microtubules at different developmental time points in zebrafish sensory neurons.

To establish the best strategy to detect somatic MTOCs, we imaged EB1-GFP behavior in mature Drosophila sensory neurons that do not use a single MTOC (Nguyen et al., 2011) and Nematostella tripolar neurons that maintain a single active somatic MTOC (Stone et al., 2020) and Movie 4). By comparing different types of image analysis methods, we found that integrating EB1-GFP fluorescence along a line throughout a timelapse movie worked well to easily distinguish the two types of somatic microtubule organization (Figure 5 A and B). The presence of a somatic organizer resulted in a strong peak of EB1-GFP fluorescence, while only a broad peak spanning the thick region of the cell was present in the Drosophila neuron (Figure 5A and B). In the Drosophila neuron, we used a centriole protein, fzr, to determine where to place the line to integrate the fluorescence signal as EB1-GFP itself could not be used to identify a peak.

Figure 5. Zebrafish sensory neurons maintain a somatic hotspot of microtubule organization.

A. Brightfield image of a 3^rd instar drosophila larva and image of a ddaE neuron cell body expressing EB1-GFP and fzr-RFP are shown. The blue line indicates the region used to quantify EB1-GFP intensity from a sum projection of a timeseries. The output for fzr-RFP (red) and EB1-GFP (green) is shown in the graph. B. A brightfield image of a 4–6 tentacle stage Nematostella vectensis polyp and an image of a tripolar ganglion cell expressing EB1-GFP are shown. The blue line was used as the region for summing intensity through the time course and the output is shown in the graph. C. A brightfield image of a 14dpf zebrafish and images of DRG neuron cell bodies expressing EB3-GFP are shown together with lines used to quantify EB3-GFP fluorescence. The graphs show intensity along the line summed through the timecourse. D.-G. Brightfield overviews and images of RB neuron cell bodies expressing EB3-GFP are shown from different age fish. In each a line was used to generate summed intensity graphs.

We imaged DRG neurons when the p2×2–2 promoter became active and they began to express EB3-GFP, and found that they displayed hot-spots of EB3-GFP fluorescence similar to that in the Nematostella neuron (Figure 5C). We could not image DRG neuron cell bodies at sufficient resolution to determine whether the MTOC persisted throughout the life of the cell. We therefore used RB neurons, which we showed above have similar microtubule organization and somatic structure to DRG neurons, to image somatic microtubule organization throughout the cellular lifespan.

In order to visualize RB neurons in the first week after fertilization, we crossed the P2rx3a>EB3-GFP, P2rx3a>mCherry double transgenic casper line to the P2rx3a>EB3-GFP, Isl1>tdTomato double transgenic casper line and selected the brightest embryos displaying both markers for imaging. For later timepoints we in-crossed the P2rx3a>EB3-GFP, P2rx3a>mCherry casper line, and selected the brightest embryos with both markers for imaging. Timepoints were taken every day after fertilization for the first two weeks, and again at three weeks post fertilization. For embryos, EB3-GFP comets could be seen radiating from one point in the soma (Movies 5 and 6). At later time points, the analysis strategy developed by comparing Drosophila and Nematostella cell bodies enabled detection of a hotspot of EB3-GFP fluorescence. At all timepoints a peak of fluorescence, like that in the Nematostella neurons, was apparent. Example neurons at different ages are shown (Figure 5D–G). Even at three weeks, when RB neurons are rare and DRG neurons much more prevalent (Figure 2D and E), a clear EB3-GFP hotspot was present. For each zebrafish condition at least 4 cells were analyzed with the exception of the rare 21dpf RB neurons, where 3 were analyzed. All cells had clear summed EB3-GFP peaks similar to the examples shown. This analysis suggests that vertebrate sensory neurons maintain an active somatic MTOC even when mature.

Discussion

To address the long-standing question of whether branched receptive arbors of vertebrate sensory neurons have cellular dendritic features, we used live cell imaging to probe microtubule organization in two types of zebrafish neurons. We found that both RB and DRG neurons have central and peripheral axons with uniform plus-end-out microtubule polarity. Even the branched sensory endings that innervate the skin and receive and integrate signals have the axonal plus-end-out polarity, and this arrangement is maintained at all stages of maturity. In addition to showing that a key defining feature of dendrites, the presence of minus-end-out microtubules, is absent from receptive regions of sensory neurons, we characterized some additional aspects of their microtubule and cellular organization. Mature polarized neurons in mammals (Stiess et al., 2010) and Drosophila (Nguyen et al., 2011) do not use a centrosomal MTOC to organize somatic microtubules. However, the recent finding that Nematostella neurons with only axon-like neurites maintain an active somatic MTOC (Stone et al., 2020) raised the possibility that vertebrate sensory neurons with exclusively axon-like neurites might do the same. Indeed, we found that somatic microtubule organization in RB and DRG neurons more closely resembles that in unpolarized Nematostella neurons than Drosophila polarized neurons (Figure 5). The presence of a somatic MTOC does not, however, mean that microtubules are all anchored at this point. The observation that growing minus ends are present in RB neuron (Feng et al., 2019) and DRG neuron (Figure 4H) sensory endings confirms that microtubule minus ends are present far from the somatic MTOC.

In addition to characterizing key aspects of microtubule organization in vertebrate sensory neurons in vivo, we also examined how their neurites emerge from the cell body. Invariably, mature DRG neurons are described as pseudounipolar, meaning that they have a single neurite originating from the cell body that branches to give rise to the central and peripheral axons. However, this description relies on analysis of relatively few animal species. We therefore took the opportunity to expand our understanding of the basic cellular morphology of sensory neurons to another animal. Surprisingly, we found that, while pseudounipolar RB and DRG neurons were present, they were outnumbered at all life stages by true bipolar cells that had small central axons and much larger peripheral axons originating directly from the cell body. Our work provides important cellular context for understanding vertebrate sensory neuron biology.

One particularly intriguing aspect of vertebrate sensory neuron cell biology is how central and peripheral axons become structurally and functionally distinct when they have the same microtubule polarity. Nematostella neurons with similar microtubule polarity, all neurites plus-end-out, do not sort different cargoes to each neurite; instead all neurites have both pre- and post-synaptic structures (Stone et al., 2020). Indeed, there are some hints that polarized sorting may not be particularly efficient in DRG neurons. The receptor that senses noxious heat, VR1, is not only found in sensory endings, but also localizes to projections within the spinal cord (Tominaga et al., 1998). However, it is clear that RB and DRG central and peripheral axons do have critical differences. For example, their caliber is clearly different as they emerge from the cell body or primary neurite (Figure 3) and during development they are influenced differentially by specific transcription factors and centrosome position (Andersen et al., 2011; Andersen and Halloran, 2012; Tanaka et al., 2011). Developing central and peripheral axons also respond to different signals (Liu and Halloran, 2005), suggesting different receptors may be present on the two processes. Mature central and peripheral axons are differentiated by their response to injury. While severing of the DRG peripheral axon elicits transcriptional reprogramming and regeneration, severing the central axon does not (Liu et al., 2011; Rishal and Fainzilber, 2010) suggesting that injury sensing machinery is differentially distributed. How might proteins be differentially targeted to the two axons? One possibility is that there is no specific targeting, only specific retention based on different interactions of the two neurites with surrounding cells. Another, not mutually exclusive possibility, is that other microtubule features, for example post-translational modifications that can steer some kinesins specifically into axons of polarized mammalian neurons (Tas et al., 2017), distinguish the peripheral and central axons and guide cargoes selectively to each.

The lack of any region of the sensory arbor with minus-end-out microtubules separates vertebrate sensory neurons from invertebrate ones. Invertebrate sensory neurons with branched receptive fields in Drosophila and C. elegans have minus-end-out microtubules in at least part of the arbor (Harterink et al., 2018; Rolls et al., 2007; Stone et al., 2008). In contrast, we show that two types of vertebrate sensory neurons do not. As most neurons in bilaterian animals have axons and dendrites with distinct microtubule polarity (Rolls and Jegla, 2015) we hypothesize that this is the most efficient way to generate two different types of neurites. However, this efficiency may be counterbalanced by other factors in specific scenarios. One such example might be the PVD neuron in C. elegans. It has a branched sensory dendrite arbor (Tsalik et al., 2003) that has minus-end-out microtubules in the large anterior process and plus-end-out microtubules in the smaller, but morphologically similar, posterior process (Harterink et al., 2018). Both processes emerge from a single neurite exiting the cell body. While this arrangement means that the two dendrite regions have different microtubule organization, it allows for a continuous parallel microtubule array throughout the main trunk of the dendrite. In the case of vertebrate sensory neurons, the long distance between the cell body near the spine and sensory endings in the skin may make kinesin-based anterograde transport along a uniform microtubule array important. Axonal kinesins are optimized for fast, processive transport (Soppina et al., 2014), and so this may make the tradeoff with reduced sorting efficiency to neurites with different microtubule organization worthwhile. While this rationale applies primarily to axons in the nerve, reorganizing microtubule polarity in the sensory endings may be either mechanistically difficult or of little benefit for cargo distribution.

The three neuron types in which microtubule organization has been found to be similar in all neurites, Nematostella tripolar ganglion neurons (Stone et al., 2020), RB neurons and DRG neurons (Figure 5), all seem to maintain a distinct somatic MTOC. While more examples would be helpful to determine the significance of this observation, one possibility is that a centrosomal MTOC is used to establish and maintain plus-end-out polarity in all neurites. Microtubules would not need to remain attached to the MTOC for this to work. The HAUS/augmin complex can recruit nucleation sites to the sides of existing microtubules (Petry et al., 2013), and plays a role in axonal microtubule organization (Cunha-Ferreira et al., 2018; Sanchez-Huertas et al., 2016). While this arrangement has been described as generating branched microtubule arrays, the majority of nucleation sites generate new microtubules in parallel to existing ones (Petry et al., 2013) and so could propagate polarity from a plus-end-out arrangement emerging from the MTOC throughout the neurites.

Methods

Drosophila maintenance and imaging

Drosophila were maintained at 25C on cornmeal agar fly food as described (Feng et al., 2019). To generate larvae for somatic microtubule imaging virgins from the line 221-Gal4, UAS-EB1-GFP (Stone et al., 2008) were crossed to males from the line UAS-fzr-mRFP (Basto et al., 2006). Adults were transferred into a fresh food vial every 24 hours, and larva that were 72–96h old were selected for imaging. Larvae were mounted under number 1.5 glass coverslips secured with tape, and a 63× 1.4NA oil immersion objective was used for imaging at 1.27 seconds per frame.

Nematostella vectensis maintenance and imaging

Nematostella were maintained in glass bowls in 12 g/L Instant Ocean and fed brine shrimp daily. Animals were kept in the dark at room temperature. Animals used for spawning were >6months in age and females were kept in separate bowls in order to control timing of fertilization of the eggs. They were exposed to light for 10–12 hours using a standard fluorescent light box to initiate spawning. Egg sacs from female bowls were removed and mixed with water from bowls containing males for 20–30 minutes within in one hour of spawning to achieve synchronized fertilization. Fertilized embryos were then isolated from the egg sacs by gently agitating in 12g/L Instant Ocean containing 2% L-cysteine, pH 8.5. Embryos were then rinsed 3–5X with 12g/L Instant Ocean and immediately transferred to Falcon 353007 60mm polystyrene dishes for injection. Embryo injections were carried out as previously described (Stone et al., 2020). Animals were injected with a plasmid containing the Nematostella EB1-GFP and mCherry controlled by the pan-neuronal elav promoter (NvElav-EB1-GFP-2A-mCherry) and screened for expression of the NvEB1-GFP transgene 1–2 weeks post injection using an AxioZoom.v16 microscope.

For imaging, whole, live Nematostella polyps at the 4–6 tentacle stage were anesthetized in 2% urethane in 12g/L Instant Ocean until tentacles were completely extended. They were then mounted on a metal slide with a 12mm hole in the center as described (Stone et al., 2020). The underside of the hole was covered with an air permeable membrane and the top side of the hole was covered with a piece of nylon mesh. Polyps were transferred to the top of the nylon mesh in a droplet of 12g/L Instant Ocean containing 2% urethane and covered with a 12mm round coverslip. Animals were imaged on a Zeiss LSM 510 confocal microscope. NvEB1-GFP videos were taken with a 63× 1.4NA oil immersion objective and captured at 1 frame per 2.88 s.

Zebrafish lines, maintenance and rearing

The following zebrafish lines were used in this paper (see table of reagents for full formal genotypes). Lines were not necessarily homozygous for transgenes, but were propagated by selecting the brightest fish expressing both markers in each generation.

P2rx3a>EB3-GFP, P2rx3a>mCherry, mifta^w2/mifta^w2, roy^a9/roy^a9

P2rx3a>EB3-GFP, Isl1[ss]>tdTomato, mifta^w2/mifta^w2, roy^a9/roy^a9

Wild type AB (for Isl1[ss]:EB3-GFP injection and imaging)

Adult zebrafish were maintained in an Aquaneering habitat at a density of 3 fish per liter, sex separated and mated weekly, fed brine shrimp twice daily, given a 14h light 10h dark light cycle, and water was kept at 28.5C with a pH of 7.5. Young fish were raised in glass bowls in a 28.5C incubator with a 14h light 10h dark light cycle and received once daily 50% water exchanges of 28.5C Blue Water (0.6g/L Instant Ocean salt, 0.01mg/l methylene blue) until day 5, when their bowls were transferred from the incubator to the room in which the Aquaneering colony was kept. Animals were fed rotifers from 5 dpf to 3 months. Brine shrimp were added to their diet from 2 weeks of age. At 3 weeks post fertilization the fish were transferred from bowls into the Aquaneering habitat and maintained on a slow drip. All experiments were approved by the Institutional Animal Care and Use Committee, including procedures for pain management and euthanasia. For adult experiments male and female fish were used. For larval experiments no distinction was made due to the lack of sex chromosomes and the lack of apparent sexual differentiation at early developmental timepoints. Adults used in experiments ranged from 8 to 10 months old.

Overview of microscopy

Brightfield overview images of younger fish, the drosophila larvae, and the nematostella polyp were generated using a Zeiss Axiocam 105 color camera mounted on a Zeiss Stemi 305 dissection scope. Brightfield overview images of 3 month and older fish were obtained using a Google Pixel 2 camera phone. Confocal images of fish injected with Isl1[ss]>EB3-GFP plasmid were obtained using a Zeiss LSM 780 with a 20x air objective. All other zebrafish confocal fluorescence images were generated using an inverted Zeiss Axio Observer.Z1 LSM 800 with Airyscan confocal microscope outfitted with a 10× 0.3 NA air objective, a 20× 0.8 NA air objective and a 40× 1.2 NA water objective, each of which was used depending on the level of detail and field of view required for the experiment. All 2-photon images were generated on a Leica SP8 confocal microscope with a Spectra-Physics Insight X3 using a 25× 1.0 NA 2.6mm WD water objective. For details of Nematostella and Drosophila EB1-GFP imaging see above.

Live imaging of adult zebrafish

Imaging of adult fish was performed in a flow cell adapted from a previous design (Xu et al., 2015) by adding a hole in the bottom what was sealed with a coverslip to adapt it for an inverted microscope (Figure S1). The flow cell was carved from single block of aluminum by the Penn State physics department machine shop. Pumping force was provided by a Rainin Dynamax RP-1 peristaltic pump. A closed recirculating loop was constructed by connecting the pump, flow cell, and fluid reservoir with polyethylene tubing. Flow rate was calibrated to ~6ml/minute of 0.016% MS-222 in 0.6g/l Instant Ocean. Fish were immobilized with 4% low melt agar. Time series Z-stacks were generated by selecting fast scan settings such that 5 focal planes with ~2um separation could be imaged with a cumulative frame time of 3.45 seconds. This allowed for imaging of EB3-activity in a >100μm*100μm*10μm volume that contained the fine sensory endings. To generate a movie focal planes at each time point were merged in ImageJ via the “maximum projection” function.

Imaging of non-adult transgenic fish lines

Fish younger than 2 weeks were fully embedded in 1.5% low melting agar in a 4 well Ibidi μ-slide. Fish older than 2 weeks were anesthetized, and bathed in enough liquid to raise level several millimeters over the fish into a well of the μ-slide. Only the tail was immobilized by submerging the tip of a micropipetter filled with 45C 1.5% low melting agar directly over the tail region and carefully squirting molten agar onto the tail before it solidified. This left the gills free and facilitated recovery of the fish. All fish used in the data set successfully recovered after imaging before being euthanized.

Fish DRG EB3-GFP speed analysis

EB3-GFP videos of fish DRG neurons were analyzed using the “multiple kymograph” feature of ImageJ. Areas of the neurites were selected that were clearly part of the nerve innervating that portion of the scale, as well as the terminal sensory endings. Additionally, we analyzed the thin neurites before the final branching event that defined them as terminal, and a medium thickness region that while not part of the main nerve, may still have contained more than one neurite. Speed was determined manually by scoring the resultant kymographs and comparing displacement along the X and Y axis over the run of the comet that was captured.

Somatic MTOC analysis

Time series images were collected for each of the cells in question. A line was drawn through the brightest EB region for all cells except those in Drosophila where fzr-mRFP was used to guide the line. ImageJ was used to perform a “sum slices” projection on the EB video, which integrates all EB activity into a single frame, creating a high dynamic range image with the brightest point reflecting the spot with greatest EB activity. A line was then drawn through the brightest point and ImageJ’s “plot profile” function was used to generate the output data. Lines used are indicated on each of the example images.

Microinjection of zebrafish eggs with Isl1[ss]:EB3-GFP and live imaging

Zebrafish eggs at the one-cell stage were placed in a 1.5% agarose injection mold. A solution of DNA (10 – 50 ng) mixed with phenol red was prepared and loaded onto an injection needle. Injections were performed on a Zeiss upright microscope (Carl Zeiss) using a Narishige M-152 micromanipulator (Narishige Group). Injected eggs were transferred to a petri dish with artificial seawater and a few drops of methylene blue and incubated at 28.5 degree Celsius. After a few hours, unfertilized eggs were discarded using a pipette. The following day, embryos were screened for fluorescence.

Embryo mounting was carried out as described previously (O’Brien et al., 2009). Embryos were treated with 5% phenylthiourea (PTU) in Ringer’s solution to inhibit pigment formation. GFP-positive zebrafish embryos at 1 day post fertilization (dpf) were dechorionated with forceps and anesthetized in 0.02% tricaine. Embryos at 2, 3 and 3 dpf were directly anesthetized in 0.02% tricaine. A mounting chamber was prepared by fixing a Teflon ring on to a cover slip with vacuum grease. Anesthetized embryos were dropped into 1% low melting agarose solution, maintained at 42 degree Celsius in a heating block. An embryo was then transferred to the mounting chamber along with a drop of agarose. Using a probe, the embryo was positioned correctly for imaging. After the agarose hardened; the chamber was filled with PTU Ringers solution with 0.02% tricaine. The top of the ring was then greased and closed with a glass slide. See imaging overview section for microscopy details.

Plasmid construction

pEX_P2rx3a.E1b:LexA-VP16.Sv40.4xLexOP.E1b:EB3-GFP.Sv40 plasmid has been described (Feng et al., 2019).

Construction of plasmid expressing EB3-GFP under the Isl1[ss] promoter: All cloning steps were performed using the Multisite Gateway system (ThermoFisher Scientific) and Tol2kit (Kwan et al. 2007). Dr. Reinhard Koster provided the vector pCSEB3-GFP containing the human EB3 protein tagged with GFP (Distel et al. 2010). The EB3 fragment was amplified from pCSEB3-GFP using the forward primer 5’-GGGGACAAGTTTGTACAAAAAAGCAGGCTGCCACCATGGCTGTCAATGTGTACTCCAC-3’ and 5’-GGGGACCACTTTGTACAAGAAAGCTGGGTAGTACTCGTCCTGGTCTTCCTG-3’ as the reverse primer and then recombined with pDONR221 to generate a middle entry clone. The Isl1[ss] promoter (Sagasti et al., 2005) was amplified using the forward primer, 5’-GGGGACAACTTTGTATAGAAAAGTTGAACAAAAGCTGGAGCTCCACC-3’ and 5’-GGGGACTGCTTTTTTGTACAAACTTGTCTTTCAGGAGGCTTGCTTC-3’ as reverse primer and was recombined with pDONRP4-P1R to generate a 5’ entry vector. All entry clones and the destination vector pDestTol2pA2 were then subjected to an LR reaction to create the final expression vector containing Isl1[ss]>EB3-GFP, which was sequenced to confirm the presence of all three inserts.

Supplementary Material

1. Figure S1. Chamber for imaging adult zebrafish.

A. A diagram of the intubation and imaging chamber is shown. B. An adult fish mounted for imaging is shown on the microscope. The objective can be seen through the coverslip under the tail. The thin intubation tube is positioned in the mouth to flow water over the gills.

2. Figure S2. Regional analysis of plus end growth speed in DRG peripheral neurites.

A. An image of a scale with fluorescently labeled DRG neurons is shown. The regions used for analysis of EB3-GFP speed are shown. B. Growth speeds generated from kymographs of EB3-GFP are plotted in different regions of the DRG peripheral neurite. Each color represents a different fish. Between fish differences overwhelmed any regional variation.

3. Movie 1. EB3-GFP dynamics in 4dpf zebrafish.

RB neurons were imaged in embryos that had been injected with Isl1[ss]:EB3-GFP. Comets at tips of growing microtubule plus ends can be seen moving towards neurite ends, which point towards the lower right of the movie.

4. Movie 2. EB3-GFP dynamics in 3 dpf transgenic P2rx3>EB3-GFP zebrafish.

RB neurons were imaged in transgenic zebrafish that express EB3-GFP. The cell body is out of frame at the top of the movie.

5. Movie 3. EB3-GFP in DRG sensory endings in an adult zebrafish.

A EB3-GFP z stack timeseries movie was acquired and each z stack was compressed into a single frame in the movie. The scale edge is at the lower right of the movie.

6. Movie 4. EB1-GFP in tripolar ganglion cells in Nematostella vectensis.

EB1-GFP was expressed in Nematostella neurons under control of the elav promoter. A hotspot of microtubule generation can be seen within the cell body.

7. Movies 5 and 6. EB3-GFP in cell bodies of RB neurons in 1dpf zebrafish.

Time series movies of RB neuron cell bodies in 1dpf transgenic P2rx3>EB3-GFP zebrafish show comets emerging from a hotspot.

Highlights.

Vertebrate sensory endings have uniform axonal microtubule polarity in the skin.

Mature zebrafish dorsal root ganglion (DRG) neurons are primarily bipolar.

Zebrafish sensory neurons maintain a somatic microtubule organizing center.