Katanin-mediated severing generates microtubules during neurite outgrowth

TL;DR

The study shows that microtubules in growing nerve cells are created through severing, a process aided by the Katanin enzyme and two other proteins.

Contribution

The study reveals a novel role for Katanin and two microtubule minus-end proteins in severing microtubules during neurite outgrowth.

Findings

Microtubules in developing neurites are generated through severing events that maintain microtubule polarity.

Katanin, PTRN-1/CAMSAP, and NMTN-1/WDR47 are required for microtubule severing in C. elegans PVD neurons.

Impaired severing reduces microtubule numbers and disrupts endosome and mitochondrial trafficking.

Abstract

Neurons contain a polarized and staggered microtubule array that is important for supporting intracellular transport. How these microtubules are generated and oriented during development is not fully understood. Here we have taken advantage of the low microtubule density in the C. elegans PVD neuron to observe in-vivo microtubule dynamics during neurite outgrowth. We found that individual microtubules are added to outgrowing neurites via microtubule severing events, which generate new microtubules while maintaining existing polarity. We further show that the Katanin enzyme complex is specifically required for severing. Surprisingly, both PTRN-1/CAMSAP and NMTN-1/WDR47, two microtubule minus-end proteins, are also required for microtubule severing. These two proteins colocalize together, and confine Katanin-mediated severing activity towards the distal neurite growth cone. When microtubule severing is impaired, the resulting microtubule array contains fewer microtubules, impacting the trafficking of various endosomes and mitochondria. Together, our findings demonstrate that microtubule severing is an important mechanism to build the microtubule array in developing neurons.

Figures

e](#F1),[f](#F1)). In the remaining cases, new plus-ends polymerized after severing (blue arrows in [Fig. 1e](#F1),[f](#F1)). Most of the newly generated microtubules were short-lived and completely depolymerized from their plus-end within 5 minutes. However, about 11% (11/100) of new microtubules persisted for at least five minutes until the end of the imaging window, suggesting that severing can contribute to the stabilized microtubule array found in the mature neuron ([Fig. 1e](#F1),[g](#F1)). Quantification of the severing location showed that the severing events were primarily localized n

a](#F2),[b](#F2)). The *mei-1(lof)* allele we used in this study is a temperature sensitive allele that has been previously shown to severely reduce embryo viability at 25°C (restrictive temperature)^[39](#R39),[40](#R40)^. We found that the microtubule severing defects in the PVD posterior dendrite were as strong at 20°C as at 25°C, indicating that the dosage requirement for embryonic lethality is different from that of microtubule severing in PVD (Extended Data Fig. 2a). Since temperature had no effect on the microtubule severing phenotype in PVD between 20°C and 25°C, we referred to this *m

b,c). Additionally, the few microtubule minus-ends that were present were still decorated with PTRN-1 and NMTN-1 puncta, indicating that the minus-end binding property of both proteins was independent of severing activity ([Fig. 3e](#F3) and Extended Data Fig. 3b,c). Meanwhile, in *nmtn-1(lof)* mutants, we found that mScarlet::FLPon::PTRN-1 signal lost its punctate pattern and became cytoplasmic ([Fig. 4a](#F4),[b](#F4) and Extended Data Fig. 3b). In *ptrn-1(lof)* mutants, mScarlet::FLPon::NMTN-1 still localized to the distal dendrite, but the individual puncta became less discrete ([Fig. 4c](

a,b), early endosomes (Extended Data Fig. 4c,d), and late endosomes (Extended Data Fig. 4e,f). Together, these data indicate that microtubule severing contributes to the generation of microtubules in the posterior dendrite and normal microtubule content is critical to efficient organelle transport.

a](#F5)). At this stage, higher order dendrites have also developed. While many higher-order dendrites contain microtubules, there are notable gaps in signal and many of the quaternary dendrites lack microtubules ([Fig. 5a](#F5)). To interrogate changes in the microtubule array, we performed time-lapse imaging of GFP::TBA-1 and used the number of dynamic plus-ends to calculate microtubule density and assess density defects in severing mutants. Indeed, in the severing deficient *mei-1(lof)*, *katb-1(lof)*, *ptrn-1(lof)*, and *nmtn-1(lof)* mutants, we observed ~20–50% reduction of microtubule de

a](#F6)). RAB-10-positive vesicles exhibited bouts of anterograde or retrograde movements separated by pauses of variable durations ([Fig. 6c](#F6)). We measured both movement speed and pause time in wild-type and severing mutants. The transition between pause and movements requires a motor cargo complex gaining access to a nearby microtubule, which led us to hypothesize that a change in microtubule density could affect pause duration. Comparing mutants to the wild-type control, while there were some changes in the anterograde and retrograde speed of RAB-10-positive vesicles, there was not an

a](#F7)). We performed time-lapse imaging of GFP::TBA-1 and were able to observe individual microtubule dynamics as well as microtubule severing in the axonal growth cone ([Fig. 7b](#F7)). In wild-type animals, severing frequency in HSN axon was noticeably lower compared to the PVD posterior dendrite, suggestive of cell type specific regulations. Nonetheless, HSN axonal severing frequency was significantly lower in *mei-1(lof)*, *katb-1(lof)*, *ptrn-1(lof)*, and *nmtn-1(lof)* mutants compared to wild-type controls ([Fig. 7b](#F7),[c](#F7)). These results argue that the MEI-1/Katanin mediated m

Full text

Introduction

Neuronal microtubule arrays are essential for neuronal development and function. In axons, microtubules are organized uniformly plus-end out^1^. In contrast, invertebrate dendrites contain exclusively minus-end out microtubules while vertebrate dendrites contain both minus-end out and plus-end out microtubules^1^. The differing microtubule polarities in axons versus dendrites is a major determinant of neuronal polarity through regulating intracellular transport and neurite growth^2–6^.

Another hallmark of neuronal microtubules is that they form a continuous staggered array. In most cells, microtubules arrays are organized by a microtubule organizing center (MTOC)^7^. The MTOC acts as a hub for microtubule nucleation and is where many microtubule minus-ends are localized. However, in neurons, microtubule minus-ends are not localized to any single location in the cell and are instead distributed throughout the axons and dendrites. Thus, the molecular mechanisms that generate neuronal microtubule arrays must build arrays of the correct polarity while also distributing polymers throughout the neuronal processes.

As previously mentioned, one way to build microtubule arrays is through microtubule nucleation at MTOCs. During cell division, centrosomes act as MTOCs, organizing the mitotic spindle that facilitates proper division. However, many differentiated cells, neurons included, inactivate their centrosomes post-division and utilize other structures as MTOCs^7,8^. In neurons, it has been shown that Golgi^9–11^ and endosomes^12–14^ can act as non-centrosomal MTOCs. Additionally, microtubules themselves can serve as a template for additional microtubule nucleation through the Augmin/HAUS complex^15–18^. Despite their distinct mechanisms, all these structures contribute to generating polarized microtubules in axons and dendrites. However, little is understood about how microtubule minus-ends are subsequently distributed along neurites with the correct directionality to form the staggered array after nucleation.

It has been proposed that prior to centrosome inactivation, centrosome derived microtubules are severed and transported into the axon^19,20^. If these short fragments are stabilized and correctly sorted, they can act as seeds for future polymerization. This mechanism addresses the need to distribute microtubules throughout developing neurites. Indeed, imaging experiments have showed that microtubule transport/sliding occurs in growing dendrites in dissociated cultured Drosophila neurons^21–23^. However, how short microtubule fragments are initially generated is not well understood, as there is a lack of direct evidence specifically supporting the severing of centrosomal microtubules. Nonetheless, microtubule severing more generally has been proposed as a potential mechanism to regulate microtubule arrays in neurons.

All three known microtubule severing enzymes, Katanin, Spastin, and Fidgetin, are expressed in neurons and have been implicated in neurodevelopment^24,25^. However, the exact phenotypes that arise from microtubule severing are both enzyme- and cell-dependent. In Drosophila, Spastin and Katanin mutants show defects in the development of neuromuscular junctions^26,27^. However, their effect on the underlying microtubule organization differs as Spastin mutants contain fewer microtubule bundles^26^ while Katanin mutants contain more microtubules^27^. Also in Drosophila, Katanin has been shown to negatively regulate dendritic arborization^27^, while a Katanin-like protein is responsible for promoting arborization^28^. Adding to the complexity, the same Katanin-like protein promotes microtubule disassembly later in neuronal development during dendritic pruning^29^. Spastin’s positive regulation of microtubule content has also recently been demonstrated in iPSC derived neurons where it amplifies microtubule dynamics to facilitate the distribution of the presynaptic cargoes^30^. Despite the vast array of evidence pointing to severing enzymes regulating microtubule organization and neuronal development, the lack of direct, in vivo observation of microtubule severing events has fundamentally hampered our understanding of the location and physiological importance of microtubule severing in neurons and the molecular mechanisms that regulate microtubule severing enzymes.

Here, we recorded microtubule dynamics in the outgrowing neurites of the C. elegans PVD neuron in vivo. The single microtubule resolution in this system allowed us to directly observe microtubule severing events. We identified Katanin as the corresponding severing enzyme and further showed that the severing activity is targeted to distal neurites by PTRN-1/CAMSAP and NMTN-1/WDR47. Unexpectedly, these two previously recognized minus-end protecting proteins create new minus-ends by recruiting severing activity. Furthermore, we show that microtubule severing increases microtubule content in dendrites, which is important for supporting proper intracellular transport and distribution of multiple organelles.

Results

To understand how microtubule arrays are established in neurites during development, we used the C. elegans PVD neuron, a sensory neuron that contains two major dendritic arbors, the anterior and posterior dendrite, and a single unbranched axon (Fig. 1a). The PVD neuron begins to develop during the L2 stage, where the anterior and posterior primary (1°) dendrites first extend from the soma before the higher order dendrites are established (Fig. 1a). During neurite outgrowth, microtubules fill the 1° dendrites. The anterior dendrite contains minus-end out microtubules while the posterior dendrite and axon contain plus-end out microtubules^31^ (Fig. 1b). A single dendritic MTOC is positioned behind the anterior dendrite growth cone and is responsible for generating the minus-end out microtubule array in the anterior dendrite^13,14^. To understand how the plus-end out microtubule arrays in PVD’s axon and posterior dendrite are built, we directly visualized microtubules during neurite outgrowth using transgenically expressed GFP::TBA-1, the main α-tubulin isotype in PVD, and performed time-lapse imaging. In the cell body, individual filamentous structures can be observed growing and shrinking, indicating that GFP::TBA-1 readily incorporates into microtubules (Video S1). We have previously shown that GFP::TBA-1 incorporates into most, if not all, microtubules based on comparisons between fluorescence and electron microscopy analyses^6,13,32^. GFP::TBA-1 signal can also be detected throughout the posterior dendrite (Video S1). Kymograph analysis of this region showed that there are a small number of microtubules distributed throughout the posterior dendrite (Fig. 1c). These microtubules form a staggered, partially tiled array with 1–2 microtubules at any given location. Due to the low number of microtubules, the plus- and minus- ends of individual microtubules can be unambiguously identified. The plus-ends exhibit characteristic polymerization and depolymerization dynamics (magenta lines) while the minus-ends are relatively stable (cyan lines) (Fig. 1c). With this single microtubule resolution, we were able to directly interrogate how the plus-end out microtubule array in the posterior dendrite is being built. Unlike in the anterior dendrite, we did not observe any MTOC activity. With an active MTOC, we would expect microtubule plus-ends polymerizing in both the anterior and posterior directions, however, most plus-ends point posteriorly towards the growth cone (Fig.1c). Additionally, microtubule transport/sliding events were extremely rare in our recordings and unlikely to play a role in supplying microtubules to the outgrowing dendrite. Interestingly, we found clear examples of microtubule severing events (one microtubule giving rise to two separate microtubules) in many of our timelapses (yellow arrow in Fig. 1d and Video S2). Severing events were identified by the creation of a new plus-end (dashed magenta line) and new minus-end (dashed cyan line) (Fig. 1d). Furthermore, we generated a strain with endogenously tagged GFP::TBA-1 and used the FLP-FRT system to selectively label microtubules in PVD^33,34^. Kymograph analyses showed that individual microtubules could be similarly identified as in the transgenic GFP::TBA-1 strain. Microtubule severing events could also be observed at a similar frequency compared to the transgenic strain (Extended Data Fig. 1a,b).

The immediate consequence of a microtubule severing event is the production of a new microtubule plus- and minus-end. The new plus-ends are dynamic, either polymerizing or depolymerizing, while the new minus-ends are relatively stable. About 75% of the nascent plus-ends depolymerized immediately after severing, consistent with in vitro severing events, most likely due to the dominant GDP-tubulin species at the severing site prior to the events (purple arrows in Fig. 1e,f). In the remaining cases, new plus-ends polymerized after severing (blue arrows in Fig. 1e,f). Most of the newly generated microtubules were short-lived and completely depolymerized from their plus-end within 5 minutes. However, about 11% (11/100) of new microtubules persisted for at least five minutes until the end of the imaging window, suggesting that severing can contribute to the stabilized microtubule array found in the mature neuron (Fig. 1e,g). Quantification of the severing location showed that the severing events were primarily localized near the growth cone, with >90% of all events occurring within 5–30μm from neurite tip (Fig. 1e,h). To our knowledge, this is the first direct in vivo observation of microtubule severing in a neurite.

MEI-1/Katanin is responsible for severing microtubules in PVD

Previous in vitro studies using reconstitution systems demonstrated that microtubules are severed by a family of AAA-ATPases which include Katanin, Spastin, and Fidgetin^35–38^. The C. elegans orthologs are MEI-1, SPAS-1, and FIGL-1, respectively. To determine which enzyme was responsible for microtubule severing in PVD neurites, we examined microtubules dynamics in mei-1, spas-1 and figl-1 loss-of-function mutants. Of these three mutants, only mei-1(lof) mutants showed a severe reduction in microtubule severing, with many worms having no observable events within the 8.33 minute timelapse window (Fig. 2a,b). The mei-1(lof) allele we used in this study is a temperature sensitive allele that has been previously shown to severely reduce embryo viability at 25°C (restrictive temperature)^39,40^. We found that the microtubule severing defects in the PVD posterior dendrite were as strong at 20°C as at 25°C, indicating that the dosage requirement for embryonic lethality is different from that of microtubule severing in PVD (Extended Data Fig. 2a). Since temperature had no effect on the microtubule severing phenotype in PVD between 20°C and 25°C, we referred to this mei-1 allele as a “loss-of-function” (lof) mutation throughout the paper. Transgenic expression of MEI-1 using a PVD specific promoter in mei-1(lof) mutants significantly increases the frequency of microtubule severing, resulting in partial rescue of the severing defects (Fig. 2b). Comparatively, spas-1(lof) and figl-1(lof) mutants showed normal frequencies of microtubule severing (Fig. 2a,b). Together, this evidence suggests that MEI-1/Katanin acts cell-autonomously in PVD to sever microtubules during neurite extension.

Katanin is a heterodimeric protein complex consisting of the catalytic AAA-ATPase subunit, Katanin p60, and a regulatory subunit, Katanin p80^24,41^. In C. elegans, the Katanin p60 ortholog is MEI-1. Katanin p80 has three paralogs in worms, MEI-2, F47G4.4, and F47G4.5^42^. MEI-2 has been shown to localize MEI-1 to the meiotic spindle during C. elegans meiosis^42,43^, while the function of the other two paralogs has not been studied. In this study, we describe a function for F47G4.4 and so going forward will refer to it as “KATB-1”, named after the vertebrate Katanin p80 ortholog, KATNB1. To test the function of the three Katanin p80 paralogs, we obtained a previously described temperature sensitive mei-2 allele and deletion mutants for katb-1, and F47G4.5 and examined the severing phenotype in all three mutants. Similar to the mei-1(lof) allele, the mei-2 temperature sensitive allele showed embryonic viability defects at the restrictive temperature^40,42,43^. However, hatched mutants grown at the restrictive temperature throughout development showed normal levels of microtubule severing (Fig. 2a,b and Extended Data Fig. 2a). F47G4.5(lof) mutants also showed normal severing frequency. In contrast, katb-1(lof) mutants completely lacked microtubule severing, mimicking the mei-1(lof) mutants (Fig. 2a,b). Taken together, these results show that MEI-1/Katanin p60 and its regulatory partner, KATB-1/Katanin p80, function together in PVD to sever microtubules. Additionally, these data suggest that in C. elegans, different regulatory subunits target MEI-1/Katanin in specific biological contexts.

Microtubule severing is regulated by PTRN-1/ CAMSAP and NMTN-1/WDR-47

Within the developing PVD neurites, the microtubule severing activity is enriched in the distal neurite, suggesting additional regulatory mechanisms. Previous studies have shown that Katanin can interact with CAMSAPs to reduce the length of CAMSAP-decorated microtubule minus-ends^44,45^, while WDR-47 binds to CAMSAPs to antagonize this activity^46^. PTRN-1/CAMSAP is a microtubule minus-end binding protein that stabilizes the minus-end of non-centrosomal microtubules^47–49^. NMTN-1/WDR-47 is a WD-40 repeat protein that localizes along microtubules and is important for neuronal migration^50,51^. We examined ptrn-1(lof) and nmtn-1(lof) deletion mutants for microtubule severing in PVD. Surprisingly, both ptrn-1(lof) and nmtn-1(lof) mutants showed strongly reduced levels of microtubule severing (Fig. 3a,b). As with the mei-1(lof) and katb-1(lof) mutants, few to no severing events could be observed in the ptrn-1(lof) and nmtn-1(lof) mutants within the time-lapse window. Cell type specific expression of wild-type PTRN-1a in ptrn-1(lof) mutants increased severing frequency and partially rescued the severing defects. Similarly, cell type specific expression of wild-type NMTN-1b in nmtn-1(lof) mutants partially rescued the severing frequency phenotype (Fig. 3b). These data suggest that both PTRN-1/CAMSAP and NMTN-1/WDR-47 are required for microtubule severing in PVD, contrasting the antagonistic functions of Katanin and WDR-47 in mammalian cell cultures^46^.

To better understand how PTRN-1/CAMSAP and NMTN-1/WDR-47 regulate MEI-1/Katanin-mediated microtubule severing, we utilized the FLP-FRT system to endogenously tag both proteins in PVD specifically^33,34^. Neither of the endogenous fluorescence tags affected the frequency of microtubule severing (Extended Data Fig. 3a). We then performed dual time-lapse imaging to simultaneously visualize the subcellular localization of PTRN-1 or NMTN-1 together with microtubule dynamics. Kymograph analysis showed that both mScarlet::FLPon::PTRN-1 and mScarlet::FLPon::NMTN-1 formed puncta. The kymographs also allowed us to track how new minus-ends were generated by severing events (yellow arrows in Fig. 3c,d). Strikingly, many PTRN-1 (38.7% [36/93]) and NMTN-1 (32.8% [39/119]) puncta formed before minus-ends were generated by severing, indicating that they were not recruited to existing minus-ends (yellow arrowheads in Fig. 3c,d and yellow column in Fig. 3e). Instead, PTRN-1 and NMTN-1 puncta predicted severing sites and defined the localization of newly generated minus-ends. Supporting this idea, in mScarlet::FLPon::PTRN-1 animals, 84.2% (32/38) of observed severing events colocalized with a PTRN-1 puncta, while in mScarlet::FLPon::NMTN-1 animals, 92.9% (39/42) of observed severing events colocalized with a NMTN-1 puncta (Fig. 3f). In all cases of colocalization, PTRN-1 and NMTN-1 puncta remained associated with the newly generated minus-ends after a severing event. We also observed PTRN-1 (30.1% [28/93]) and NMTN-1 (34.5% [41/119]) puncta localized to existing microtubule minus-ends (orange arrowheads in Fig. 3c,d and orange column in Fig. 3e). These likely represent microtubules that were severed and stabilized prior to the imaging experiments, consistent with their known functions as minus-end proteins. Together, these results suggest that PTRN-1/CAMSAP and NMTN-1/WDR-47 define the location of microtubule severing. Once severing generates a nascent minus-end, these two proteins can then bind and stabilize the minus-end.

To validate this surprising finding, we studied the interdependence of PTRN-1 and NMTN-1’s localization and MEI-1 activity. If PTRN-1 and NMTN-1 are simply minus end binding proteins, their punctate localization should be lost or reduced in the mei-1 mutants, due to the reduced production of microtubule minus-ends. If these two proteins act upstream to recruit MEI-1, their punctate localization should be independent of MEI-1 activity. Indeed, in mei-1(lof) and katb-1(lof) mutants, mScarlet::FLPon::PTRN-1 and mScarlet::FLPon::NMTN-1 still formed discrete puncta in the posterior dendrite, consistent with the idea that PTRN-1 and NMTN-1 functions upstream of MEI-1 (Fig. 4 a–d). Kymograph analysis in mei-1(lof) and katb-1(lof) mutants also revealed that the majority of PTRN-1 and NMTN-1 puncta did not colocalize with any specific microtubule behavior in the absence of the Katanin complex (Extended Data Fig. 3b,c). Additionally, the few microtubule minus-ends that were present were still decorated with PTRN-1 and NMTN-1 puncta, indicating that the minus-end binding property of both proteins was independent of severing activity (Fig. 3e and Extended Data Fig. 3b,c). Meanwhile, in nmtn-1(lof) mutants, we found that mScarlet::FLPon::PTRN-1 signal lost its punctate pattern and became cytoplasmic (Fig. 4a,b and Extended Data Fig. 3b). In ptrn-1(lof) mutants, mScarlet::FLPon::NMTN-1 still localized to the distal dendrite, but the individual puncta became less discrete (Fig. 4c,d), which was particularly evident from the kymograph analysis (Extended Data Fig. 3c). These data suggest that the localization and function of PTRN-1 and NMTN-1 are interdependent.

Previous studies have showed that CAMSAP and WDR-47 directly interact with each other^46,51,52^. To further test their relationship, we created dual-color endogenous labels and performed time-lapse imaging of GFP::FLPon::PTRN-1 and mScarlet::FLPon::NMTN-1. The PTRN-1 and NMTN-1 puncta are indeed strongly colocalized (Fig.4 e,f). Furthermore, within the PVD posterior dendrite, the majority of the puncta are localized distally towards the dendritic tip, mimicking the distribution of microtubule severing events (Fig. 4g). Together, these data further support a model where PTRN-1/CAMSAP and NMTN-1/WDR-47 mutually recruit or stabilize each other to form puncta at distal neurites. These puncta recruit MEI-1/Katanin p60 and KATB-1/Katanin p80 to sever and generate new microtubules in the distal neurite.

Microtubule severing increases microtubule density in the posterior dendrite

To understand the functional importance of microtubule severing, we examined microtubule arrays in mature dendrites. In L4 stage animals, microtubules form a continuous, tiled array in the posterior primary dendrite (Fig. 5a). At this stage, higher order dendrites have also developed. While many higher-order dendrites contain microtubules, there are notable gaps in signal and many of the quaternary dendrites lack microtubules (Fig. 5a). To interrogate changes in the microtubule array, we performed time-lapse imaging of GFP::TBA-1 and used the number of dynamic plus-ends to calculate microtubule density and assess density defects in severing mutants. Indeed, in the severing deficient mei-1(lof), katb-1(lof), ptrn-1(lof), and nmtn-1(lof) mutants, we observed ~20–50% reduction of microtubule density in the posterior dendrites (Fig. 5b,c). On the contrary, in spas-1(lof), figl-1(lof), mei-2(of), and F47G4.5(lof) mutants, in which microtubule severing was normal, no such change in microtubule density was observed (Fig. 5c). These data phenocopy the reduction in microtubule density seen during neurite outgrowth (Extended Data Fig. 2b) and are consistent with the idea that microtubule severing contributes to the generation of microtubules in neurites. However, the presence of staggered plus-ends even in severing mutants also shows that microtubule severing is not the only way to generate microtubules. We suspect many microtubules originate from the soma with their plus-ends growing into the posterior dendrites.

Cargo localization and transport are impaired in the absence of microtubule severing

One of the main functions of microtubule arrays is to support long-range, intracellular cargo transport. To determine if the reduction in microtubule density observed in the severing mutants caused cargo trafficking defects, we studied the distribution of several cargos in the posterior primary dendrite of L4 stage animals. The cargo markers we used included RAB-10 (recycling endosomes), RAB-7 (late endosomes), RAB-5 (early endosomes), and TOMM-20 (mitochondria), though we focused primarily on RAB-10 because it is a marker for a type of recycling endosome that is required for DMA-1 trafficking and PVD dendrite growth and maintenance^31,53^. Using an endogenous cell-specific GFP::FLPon::RAB-10 marker, we found that RAB-10 vesicles were consistently localized along the primary dendrites and were also present in the secondary and tertiary branches (Fig. 6a). Interestingly, compared to wild-type, there was an increase in the number of RAB-10 vesicles in the posterior primary dendrite in each of the microtubule severing mutants including mei-1(lof), katb-1(lof), ptrn-1(lof), and nmtn-1(lof) mutants (Fig. 6a,b).

To determine why more RAB-10 vesicles accumulate in dendrites, we performed time-lapse recordings of GFP::FLPon::RAB-10 in the proximal region of the posterior dendrite (orange dashed box in Fig. 6a). RAB-10-positive vesicles exhibited bouts of anterograde or retrograde movements separated by pauses of variable durations (Fig. 6c). We measured both movement speed and pause time in wild-type and severing mutants. The transition between pause and movements requires a motor cargo complex gaining access to a nearby microtubule, which led us to hypothesize that a change in microtubule density could affect pause duration. Comparing mutants to the wild-type control, while there were some changes in the anterograde and retrograde speed of RAB-10-positive vesicles, there was not an overall trend among all severing mutants (Extended Data Fig. 4g). Interestingly, RAB-10-positive vesicles consistently paused longer in all severing mutants compared to wild-type (Fig. 6d). Additionally, the frequency of pauses was also greater in the severing mutants. (Fig. 6e). The increased pause duration in severing mutants is consistent with the reduced microtubule content in their posterior dendrites. In other words, when there are fewer microtubules in the neurite, it is less likely that the motor-cargo complex can “latch” onto a microtubule and initiate movements. If this hypothesis is correct, we anticipate that the severing mutants might also have trafficking defects of other cargos because access to microtubules is universally required for cargo movements.

To test this, we examined the distribution of three additional cargos, TOMM-20 (mitochondria), RAB-5 (early endosomes) and RAB-7 (late endosomes). Strikingly, we observed an increase in cargo density in the posterior dendrite of mei-1(lof) mutants for all three cargos: mitochondria (Extended Data Fig. 4a,b), early endosomes (Extended Data Fig. 4c,d), and late endosomes (Extended Data Fig. 4e,f). Together, these data indicate that microtubule severing contributes to the generation of microtubules in the posterior dendrite and normal microtubule content is critical to efficient organelle transport.

Microtubule severing occurs in the HSN axon, and is similarly regulated by MEI-1/Katanin, PTRN-1/ CAMSAP, and NMTN-1/WDR-47

The microtubule polarity in the PVD posterior dendrite is atypical in having a plus-end out microtubule array, similar to the polarity in axons. To determine if microtubule severing also occurs in axons, we recorded microtubule dynamics in outgrowing HSN axons. The HSN neuron is a motor neuron that innervates vulval muscle cells to control egg laying. The HSN cell body localizes in the midbody region, sending a long axon ventrally and anteriorly along the ventral nerve cord towards the head (Fig. 7a). We performed time-lapse imaging of GFP::TBA-1 and were able to observe individual microtubule dynamics as well as microtubule severing in the axonal growth cone (Fig. 7b). In wild-type animals, severing frequency in HSN axon was noticeably lower compared to the PVD posterior dendrite, suggestive of cell type specific regulations. Nonetheless, HSN axonal severing frequency was significantly lower in mei-1(lof), katb-1(lof), ptrn-1(lof), and nmtn-1(lof) mutants compared to wild-type controls (Fig. 7b,c). These results argue that the MEI-1/Katanin mediated microtubule severing functions in axons and suggest that microtubule severing may play a more general role in increasing microtubule content in plus-end out processes.

Discussion

Neuronal microtubules form acentrosomal, staggered arrays that fill the dendrite and axon. Here, we show that microtubule severing creates new microtubules and increases overall microtubule content in outgrowing neurites with plus-end out microtubule polarity, including in the PVD posterior dendrite and HSN axon. Importantly, microtubule severing achieves this while also maintaining polarity. Comparatively, MTOC-based mechanisms that increase microtubule content require additional regulation to bias nucleation and polymerization in the right direction. In PVD’s anterior dendrite, the minus-end out microtubule array is built by an endosome based MTOC^13,14^. During neurite outgrowth, this MTOC produces a bidirectional array, with plus-end out microtubules extending toward the growth cone, and minus-end out microtubules growing towards the cell body. Biased transport of the MTOC along the plus-end out microtubules keeps the MTOC close to the dendrite tip throughout neurite extension and prevents long-term stabilization of plus-end out microtubules. For neurites like the axon and posterior dendrite, which only contain plus-end out microtubules, this MTOC based mechanism cannot supply microtubules to the distal neurite without generating microtubules with opposite polarity. Another nucleating mechanism is Augmin-mediated nucleation, which inherently maintains polarity by nucleating microtubules from the side of an existing microtubule template^15–18^. Augmin/HAUS was first shown to amplify microtubule content in axons of mammalian neurons^15^, but has additionally been shown to regulate microtubule arrays in the dendrites of mammalian^16^ and Drosophila^17,18^ neurons. Interestingly, the C. elegans genome does not encode an Augmin/HAUS homolog, suggesting that there might be additional mechanisms to generate the plus-end out microtubule in neurites. Based on our data, we propose that microtubule severing is another major contributor to amplify the microtubule content in neurites.

We further identified the molecular players of microtubule severing in neurons. We showed that Katanin, Patronin/CAMSAP and NMTN-1/WDR47 are required for the microtubule severing both in the PVD dendrite and in HSN axons. Interestingly, all three molecules are conserved in fly, mouse and human, suggesting that similar microtubule severing might occur in the neurons of flies and vertebrates. Indeed, genetic studies in Drosophila have shown that Katanin mutants have disorganized axonal microtubules and defective synaptic boutons^27^. Similarly, a recent study demonstrated that the severing enzyme Spastin is necessary to increase microtubule polymerization at developing synapses in human neurons^30^. The limitation of these studies (and others) has been a lack of direct observation of microtubule severing events. Here we performed time-lapse imaging of fluorescently tagged tubulin and directly observed microtubules being severed in a neuronal process in a live animal. To our knowledge, the only other systematic in vivo observation of microtubule severing has been made in plant cells^54–57^.

Katanin is a heterodimeric protein that contains a p60 ATPase subunit and a p80 regulatory subunit^24,41^. C. elegans has 3 paralogs of the p80 regulatory subunit, mei-2, F47G4.4, and F47G4.5^24,42^. In this study, we found that in the PVD neuron, the previously uncharacterized F47G4.4 (KATB-1) is the specific Katanin p80 subunit that regulates Katanin-mediated microtubule severing. Of the three paralogs, KATB-1 is most similar to vertebrate Katanin p80, KATNB1, containing both the WD40 and con80 domains. Meanwhile, mei-2 and F47G4.5 possess only the con80 domain^24,42^. Interestingly, we observed that neuronal phenotypes are more sensitive to disruption of mei-1 than other mei-1 dependent processes. This could be explained by the cell specificity of mei-2 (functioning in oocyte meiosis) versus katb-1 (functioning in neurite outgrowth). In the context of meiosis and regulation by mei-2, the mei-1 allele used in this study is temperature sensitive, becoming non-functional only at the restrictive temperature, while retaining function at the permissive temperature. In contrast, in the PVD neuron, regulation of mei-1 by katb-1 is more sensitive and the mei-1 mutant allele impairs microtubule severing regardless of temperature.

We have also described novel functions for CAMSAP and WDR-47. Previous studies have demonstrated that both CAMSAP and WDR-47 directly interact with each other and function in conjunction with Katanin^46,51,52^. In cultured hippocampal neurons, WDR-47 protects CAMSAP from Katanin, preventing Katanin-mediated shortening of microtubule minus-ends^46^. In mouse primary cilia, CAMSAP, WDR-47, and Katanin, are all required to make the central pair of microtubules^52^. Based on the known functions of CAMSAP and WDR-47, the authors proposed that CAMSAP and WDR-47 stabilized microtubules created by Katanin. However, the lack of direct observations of microtubule formation and stabilization leaves open alternative explanations. Different from previous studies, we found that PTRN-1/CAMSAP and NMTN-1/WDR-47 function upstream of MEI-1/Katanin, and are each required for microtubule severing. PTRN-1/CAMSAP and NMTN-1/WDR-47 form puncta at future severing sites prior to microtubule severing and their localization are not affected by Katanin, indicating that they recruit Katanin to the microtubule for severing. New microtubule ends are generated after severing, and both proteins remain associated with the new minus-end, consistent with their previously described functions of protecting microtubule minus-ends^46–49,51^.

Our analyses of microtubule content showed reduced number of microtubules in severing mutants, indicating that severing is indeed an important mechanism to produce microtubules in neurites. However, a staggered microtubule array was still able to form by the L4 stage in the severing mutants, suggesting that there is an alternative mechanism to distribute microtubule minus-ends in the PVD posterior dendrite. Nonetheless, we could not observe any evidence of a MTOC or sliding of microtubules in both wild-type or mutant animals.

The remaining microtubule array in the severing mutants may explain why gross neuronal morphology is normal. However, previous studies of Drosophila severing enzyme mutants showed that DA sensory neurons showed abnormal dendritic arbors indicating that microtubule severing might affect dendrite morphology in certain contexts^27,28^. Additionally, growth of the higher order PVD dendrites has been shown to be driven primarily by actin dynamics and regulation of a ligand/receptor complex^58^, which may further downplay the necessity of the microtubule array with regards to PVD dendritic morphology.

Despite having normal dendrite morphology, the severing mutants showed wide-spread trafficking defects in membrane organelles. Across every cargo we interrogated, we found that density of static organelles was higher in the posterior dendrite of mei-1 mutants compared to wild-type. Timelapse imaging of RAB-10 positive cargos revealed that the increase in cargo density is caused by prolonged pauses between bouts of movements. Cargos tend to pause at the end of microtubules, where they must switch to another track to continue transport^32^. The overall reduction in microtubule content would lead to reduced microtubule overlap, increasing the time required for paused cargos to engage a new microtubule. The fact that we observed similar increases in static RAB-10, RAB-5, and RAB-7 endosomes and mitochondria in the posterior dendrite strongly argues that the defect is due to a common mechanism, such as reduced microtubule content. Together, our results suggest that microtubule severing contributes to the formation of neuronal specific microtubule arrays in the neurite and is important for intracellular transport of various membrane organelles.

Methods

C. elegans strains

C. elegans strains were maintained at 20°C on NGM plates with OP50 E. coli as food^59^.

Transgenic animals were generated by microinjection of DNA plasmid (1.5–2.5 ng/μl) together with a Podr-1::GFP co-injection marker (60–75 ng/μl).

Transgenic GFP::TBA-1 was expressed in a tba-1(ok1135) mutant background. The tba-1(ok1135) allele is a null that eliminates endogenous TBA-1 production, promoting incorporation of the transgenic GFP::TBA-1 and increasing signal. tba-1 null animals do not have any obvious defects, most likely due to the redundancy of tubulin isotypes. Strains used are listed in Supplementary Table 1.

Molecular biology

All plasmids in this study were generated using isothermal assembly with commercially synthesized overlapping oligonucleotides^60^. The pSM vector was used as the backbone for all generated plasmids. Constructs used are listed in Supplementary Table 2

Genome editing using CRISPR/Cas9

Endogenous fluorophore insertions, base pair changes, or deletions were generated by microinjection of CRISPR/Cas9 protein complexes. Injection mix is as follows: Cas9 at 1.525 μM, tracrRNA (IDT) at 4.5 μM, crRNA (IDT) at 4.5 μM, and repair template at 50–100 ng/μL. Repair templates were generated using PCR and melted before adding to injection mix^61^. dpy-10(cn64) crRNA (gctaccataggcaccacgag) and repair template (cacttgaacttcaatacggcaagatgagaatgactggaaaccgtaccgcATgCggtgcctatggtagcggagcttcacatggcttcagaccaacagcct) were used as a co-injection marker. F2 animals were screened for the desired edit, and successful knock-ins were confirmed using PCR followed by sequencing. Strains were then outcrossed prior to use for experiments. Guide crRNAs used to generate CRISPR/Cas9 strains are listed in Supplementary Table 3.

C. elegans synchronization and staging

To image PVD posterior dendrite during outgrowth, stage L2 animals were synchronized and staged. To image mature PVD posterior dendrite, stage L4 animals were synchronized and staged. C. elegans animals were synchronized by bleaching gravid adults in a hypochlorite solution followed by several M9 washes. Resulting embryos were kept in M9 and allowed to hatch overnight at room temperature, resulting in a population of L1 arrested worms. These worms were transferred to seeded OP50 NGM plates and kept at 20°C for 22–26 hours to produce a synchronized L2 population. To produce synchronized L4 worms, arrested L1s were plated and kept at 20°C for 44–48 hours. Exact culture timings varied based on genotype. For temperature sensitive assay, worms were similarly synchronized at room temperature, but once transferred to OP50 NGM plates, were kept at either 20°C or 25°C.

C. elegans imaging

Before imaging, L2 worms were anesthetized using 1mM levamisole in M9 buffer and mounted on 5% circular agarose pads in a 35mm Mattek glass bottom dish. All worms were imaged at room temperate and within two hours of mounting. In all cases, the microscope and image settings (laser power, exposure time, gain) used were identical for all genotypes across the experiment.

Dynamic imaging was performed using an inverted Zeiss Axio Observer Z1 microscope equipped with a Yokogawa CSU-X1 spinning disk, Hamamatsu EM-CCD digital camera, and a Plan-Apochromat 100x/1.4 NA objective or 63x/1.4 NA objective. Stage L2 animals were imaged using the 100x objective, while L4 animals were imaged using the 63x objective. The microscope was controlled by MetaMorph Microscopy software. Dynamic imaging was done at a single focal plane and was manually adjusted if the focal plane shifted. TBA-1 dynamics were recorded at 2 frames/second. For co-imaging experiments, recordings were taken at 1 frame/second. Rab-10 transport dynamics were recorded at 5 frames/second. PTRN-1, NMTN-1, and mitochondria localization were imaged on a Zeiss LSM980 Airyscan 2 system using a Plan-Apochromat 63×1.4 NA objective. Cargo localization (Rab-5, Rab-7, and Rab-10) was imaged using a 3i spinning disk system, which consisted of an inverted Zeiss Axio Observer Z1 microscope equipped with a Yokogawa CSU-W1 spinning disk, Prime 95B Scientific CMOS camera, and a C-Apochromat 63×/1.2 NA water-immersion objective. The microscope was controlled by 3i Slidebook software.

Image analysis

Images were processed using FIJI (ImageJ). Image registration was performed using the HyperStackReg plugin^62^. Kymographs were made using the KymographBuilder plugin.

Severing frequency was calculated by counting the number of severing events observed per kymograph and controlling for total length of dendrite analyzed and imaging duration. Subsequent values were then reported as “# events per 100 μm per minute”.

Microtubule density in Figure 5 was calculated by counting the number of dynamic plus-ends over the length of the kymograph and reported as “# microtubules per 100 μm”. Microtubule density in Extended Data Figure 2 was similarly determined, but due to the transient nature of microtubules during this developmental period, the microtubule number was determined by averaging between the first and last 10 seconds of the timelapse. Subsequent values are reported as “# microtubules per 10 μm”.

Puncta/cargo density was determined by first thresholding the imaging and only counting signal above the threshold. The threshold for each puncta/cargo differed but remained consistent for every genotype across the experiment. After thresholding, the Process->Find Maxima function in ImageJ was used to distinguish abutting puncta/cargo.

Rab-10 dynamics were quantified by drawing line segments along cargo tracks on kymographs. The coordinates of each line segment endpoint were then used to calculate the distance, duration, and speed of each individual movement bout. Individual bouts were then sorted into anterograde, retrograde, and pausing events. Pausing events were bouts with speed <0.05 μm per second. Average values per cargo track was reported.

Statistics

Statistical comparisons were performed using Prism10 (Graphpad). All imaging data were replicated in at least two independent imaging sessions.

Supplementary Material

This is a list of supplementary files associated with this preprint. Click to download.

SupplementaryInformation.docx

SupplementaryTables.xlsx

SupplementaryVideo1.avi

SupplementaryVideo2.avi

ExtendedDataFigures.docx