Patronin-mediated minus end growth is required for dendritic microtubule polarity

Microtubule plus ends exhibit rapid growth and shrinkage phases, a behavior termed dynamic instability (Mitchison and Kirschner, 1984). This behavior occurs in vitro with pure αβtubulin dimers and in vivo, where it can be regulated by plus end binding proteins and microtubule age (Gardner et al., 2013; Akhmanova and Steinmetz, 2015; Brouhard, 2015). Minus ends also exhibit dynamic instability in vitro, although their growth rate is two to three times slower than plus ends (Mitchison and Kirschner, 1984; Dammermann et al., 2003). In vivo, however, minus ends are generally thought not to contribute to microtubule dynamics or exploration of cellular space (Akhmanova and Steinmetz, 2015).

In cells, the γtubulin ring complex acts as a template and nucleator to allow new microtubules to form. As long as it remains attached at the nucleating or minus end, it acts as a cap to block addition of subunits and prevent depolymerization. Free minus ends can then be generated by severing proteins (Baas et al., 2005; Sharp and Ross, 2012). Two well-established fates for free minus ends are depolymerization and stabilization (Dammermann et al., 2003; Akhmanova and Hoogenraad, 2015). In plants, cortical microtubules that are freed from nucleation sites depolymerize at the minus end while growing at the plus end to generate treadmilling behavior (Shaw et al., 2003; Nakamura et al., 2018). In animal cells, depolymerization at the minus end is an important aspect of mitotic spindle dynamics (Wittmann et al., 2001), and minus end recognition by ASPM (abnormal spindle-like microcephaly associated) is important to help control depolymerization (Jiang et al., 2017). In interphase, most free minus ends seem to be rapidly stabilized by calmodulin-regulated spectrin-associated protein (CAMSAP)/Patronin proteins (Akhmanova and Steinmetz, 2015).

Identification of CAMSAP/Patronin proteins was a major breakthrough in understanding the cellular stability of minus ends. They were the first proteins identified that could directly recognize free minus ends (Akhmanova and Hoogenraad, 2015) and are likely responsible for their resistance to depolymerization in interphase (Dammermann et al., 2003; Jiang et al., 2014; Akhmanova and Hoogenraad, 2015). The repertoire of minus end–binding proteins was recently expanded by the addition of ASPM, which can also recognize minus ends and controls their depolymerization in the spindle (Jiang et al., 2017). It is not known whether this family of proteins also functions in post-mitotic cells like neurons.

Invertebrates have one CAMSAP/Patronin family member, termed Patronin, while vertebrates generally have three, CAMSAP1–3, with some extras where genome duplications have occurred (King et al., 2014). For example, in zebrafish, there are two each of CAMSAP1 and 2 and one CAMSAP3 (King et al., 2014). Drosophila melanogaster Patronin stabilizes minus ends against depolymerization by kinesin-13 in cultured cells and in vitro (Goodwin and Vale, 2010). Mammalian CAMSAPs also protect minus ends from depolymerization by kinesin-13s in vitro (Hendershott and Vale, 2014) and in cells (Jiang et al., 2014). Overall stabilization of minus ends by CAMSAPs and Patronin is agreed upon as a critical function without which the microtubule network in cultured Drosophila cells and mammalian cells including neurons is dramatically destabilized (Goodwin and Vale, 2010; Jiang et al., 2014; Yau et al., 2014).

The relationship between CAMSAPs/Patronin and minus end growth is more complex. In vitro CAMSAP2 and 3, as well as two domains of Patronin, suppress addition of tubulin subunits to the minus end in a concentration-dependent manner (Hendershott and Vale, 2014; Jiang et al., 2014). In contrast, CAMSAP1 tracks minus ends as they grow without altering the rate of subunit addition (Hendershott and Vale, 2014). In cells, CAMSAP1 also tracks growing minus ends, but reduction of CAMSAP1 does not result in any change in microtubule behavior (Jiang et al., 2014). CAMSAP2 has been described as suppressing minus end growth (Hendershott and Vale, 2014) and also as promoting addition of short stretches of microtubule to the minus end (Jiang et al., 2014). Although these two models for CAMSAP2 sound incompatible, they are actually not so different. Minus ends grow slowly in the presence of CAMSAP2 (Hendershott and Vale, 2014; Jiang et al., 2014), and this allows short stretches of CAMSAP2 to become stably associated with the microtubule (Jiang et al., 2014). The stretches are on average 1 μm in control cells and 2 μm when katanin is depleted (Jiang et al., 2014), and so, although they are derived from tubulin subunit addition, this does not result in much net growth at the minus end. In primary neuron cultures, stretches >10 μm of CAMSAP2 along microtubules have been observed, but growth has only been tracked for stretches of about a micron, so it is not clear how the longer stretches arise (Yau et al., 2014). Thus, it is still ambiguous whether extended growth at the minus end occurs in cells and, if so, how it contributes to global microtubule organization.

CAMSAP/Patronin proteins are particularly important in neurons where most, if not all, microtubules are noncentrosomal. In cultured hippocampal neurons, reduction of CAMSAP2, the major family member in this cell type, destabilizes microtubules and reduces dendrite complexity (Yau et al., 2014). Caenorhabditis elegans Patronin is required for maintenance of normal neuronal morphology (Marcette et al., 2014), neuronal microtubule stability (Chuang et al., 2014; Richardson et al., 2014), and axon regeneration (Chuang et al., 2014). Beyond stabilizing microtubules, it is not clear whether Patronin regulates specific aspects of microtubule organization in neurons. In epithelial cells, CAMSAP3 is responsible not only for stability of microtubules but also their polarized arrangement with minus ends concentrated at the apical surface (Meng et al., 2008; Noordstra et al., 2016; Toya et al., 2016). Neuronal microtubules are even more dramatically polarized than epithelial ones, with uniform plus-end-out polarity in axons and mixed or minus-end-out polarity in dendrites (Baas and Lin, 2011). We therefore hypothesized that CAMSAP/Patronin proteins might function to control not only microtubule stability in neurons, but also their polarity.

Drosophila dendrites are strikingly polarized with >90% minus-end-out microtubules (Rolls et al., 2007; Stone et al., 2008). It is conceptually straightforward to imagine plus-end-out processes in which fast-growing microtubule plus ends allow microtubules to populate an extending structure, while more complex models are thought to be required for population of processes with minus-end-out microtubules. In dendrites, local nucleation can generate new minus ends (Ori-McKenney et al., 2012; Nguyen et al., 2014; Yau et al., 2014), and outgrowth of minus ends has not been considered as an alternative way to get minus-end-out microtubules into dendrites. However, in Drosophila neurons, nucleation sites are concentrated at dendrite branch points (Ori-McKenney et al., 2012; Nguyen et al., 2014), so how the terminal dendrite beyond the branch point is populated with minus-end-out microtubules remains a conundrum.

Using live imaging of microtubule dynamics with tagged end-binding (EB) proteins in Drosophila and zebrafish neurons, we identified a population of slow-growing microtubule ends that move in the opposite direction to fast-growing plus ends. In Drosophila dendrites, these slow-moving structures are labeled with Patronin, confirming that they are growing microtubule minus ends. We demonstrate that sustained growth of minus ends requires Patronin, and in turn is important for getting minus-end-out microtubules into distal regions of mature dendrites, as well as into developing and regenerating dendrites.

EB proteins have been used to track microtubule plus end growth in mammalian neurons in culture and in vivo (Stepanova et al., 2003; Kleele et al., 2014), and neurons in vivo in zebrafish (Lee et al., 2017), C. elegans (Goodwin et al., 2012), and Drosophila (Stone et al., 2008). In all axons, EB comets move away from the cell body (Baas and Lin, 2011). Dendrites in mammals contain comets that move in both directions (Stepanova et al., 2003; Yau et al., 2016), while in C. elegans and Drosophila, they move primarily toward the cell body (Stone et al., 2008; Goodwin et al., 2012). The polarity measured with plus end dynamics matches overall polarity of microtubules with one exception: in mammalian neurons, minus-end-out microtubules are slightly underrepresented by dynamic plus ends in dendrites, consistent with minus-end-out microtubules having a longer, more stable region behind the dynamic plus end (Yau et al., 2016; Tas et al., 2017). In Drosophila dendrites, monitoring dynamic plus ends predicts the layout of all microtubules, including stable regions (Stone et al., 2008).

Drosophila ddaE neurons are particularly amenable to imaging neuronal microtubule dynamics in vivo as their cell body, dendrite arbor, and proximal axon are all located near the surface of the animal (Fig. 1 A). These cells are proprioceptive dendritic arborization neurons (Hughes and Thomas, 2007) and have a relatively simple stereotyped dendrite arbor (Grueber et al., 2003b; Fig. 1 D). While imaging growing microtubules with EB1-GFP, we noticed an unexpected class of moving dots. In addition to the fast-moving comets characteristic of microtubule plus ends, we observed dimmer round dots that moved slowly and processively (Fig. 1 E and Video 1). When plotted by speed, the EB1 dots fell into two clear groups (Fig. 1 B). We therefore used speed to categorize the dots, and the cutoff between the two groups was 1.6 μm per minute. The fast and slow dots moved in largely opposite directions, with most slow dots moving away from the cell body and fast ones moving toward it (Fig. 1, C and E–G). We hypothesized that the slow dots could be growing microtubule minus ends based on their similarity to growing minus ends in vitro (Bieling et al., 2007). Consistent with the idea that EB1-GFP was tracking growing plus and minus ends, we occasionally observed a plus end comet initiating from a slow-moving dot (Fig. 1 G and and Video 2). Similar behavior of fast- and slow-moving EB1-GFP puncta was observed in axons, with the exception that the fast-moving dots moved away from the cell body and slow-moving ones toward it (Fig. 1, H and I).

As the possibility that neurons might contain growing minus ends was unexpected, we wished to determine whether the phenomenon was specific to Drosophila or more generally found in neurons. We chose zebrafish for comparison, as they are also amenable to live imaging. Rohon-Beard (RB) neurons are the sensory neurons in larval zebrafish and have previously been used for imaging microtubule growth (Lee et al., 2017). The RB sensory endings are branched and similar to Drosophila dendritic arborization neurons (Fig. 1, J–L) with the exception that their microtubules are organized with plus-end-out polarity (Lee et al., 2017). We expressed EB3-GFP in RB neurons and again saw two types of labeled dots (Fig. 1 M and Video 3). The fast-moving ones did not move quite as quickly as in Drosophila (Fig. 1 N), but were still clearly distinct from a slow-moving pool similar to that seen in Drosophila. For zebrafish, the cutoff used to separate fast and slow dots was 1.3 μm per minute. As expected for plus ends, the fast puncta moved mostly away from the cell body (Fig. 1 O). In contrast, the slow dots moved toward the cell body (Fig. 1 O). Together, these data from Drosophila and zebrafish indicate that EB proteins label two classes of moving structures in neurons. The fast-moving puncta are the microtubule plus ends previously characterized in neurons (Stepanova et al., 2003; Stone et al., 2008; Lee et al., 2017). Both the speed and direction of movement of the slow-moving dots suggested that they could be growing minus ends.

Patronin/CAMSAP family proteins recognize microtubule minus ends (Akhmanova and Hoogenraad, 2015). To determine whether slow-moving dots were minus ends, we coexpressed EB1-TagRFP-T with 3xYFP-Patronin (simplified to YFP-Patronin throughout). EB1-TagRFP-T behaves similarly to EB1-GFP and labels fast- and slow-moving spots. The only difference is that with this fluorophore, the two types of dots have similar fluorescence intensity (FI; Fig. 2 D). When YFP-Patronin was expressed at high levels, it labeled the whole microtubule lattice (Fig. 2 A). We therefore selected a transgenic insertion line in which YFP-Patronin was expressed at low levels (Fig. 2 B). Large immobile puncta that appear and disappear but do not move were still present in some neurons (e.g., arrow in Fig. 2 K), perhaps representing residual lattice labeling or recruitment to some other structure. However, when we focused on moving structures, we were able to detect persistent movement of Patronin dots primarily away from the cell body (Fig. 2, C and D). These mobile Patronin puncta coincided with slow-moving EB1 dots labeled with EB1-TagRFP-T (Fig. 2, C and D), but did not overlap the fast EB1 puncta (Fig. 2, C and D; and Video 4). The localization of Patronin to slow EB1 puncta supports the hypothesis that these are growing minus ends.

To confirm that YFP-Patronin recognizes newly generated minus ends, we used a laser to sever dendrites. The predicted location of new minus ends generated by dendrite severing is indicated with a green arrow in Fig. 2 E. A new spot of YFP-Patronin appeared next to the cut site closer to the cell body as predicted (Fig. 2 F).

We next confirmed that the behavior of YFP-Patronin was similar to that of endogenous Patronin and not a function of overexpression or tag location by using two additional tagged lines. Patronin-Venus has the Venus coding sequence replacing the stop codon of the Patronin gene, so that the genomic Patronin locus encodes a tagged protein (Nashchekin et al., 2016). Patronin-Venus fluorescence could be observed in neuronal cell bodies (Fig. S1, A and B) and coincided with slow-moving dots of EB1-TagRFP-T in the cell body (Fig. S1 A). However, it was too dim to visualize in dendrites. EGFP-Patronin was generated from a large genomic region such that the EGFP coding sequence is inserted at the 5′ end of the Patronin gene and used to generate flies that contain an extra copy of EGFP-Patronin under control of its own regulatory regions (Takeda et al., 2018). This Patronin transgene labeled slow-moving spots in dendrites that were labeled with EB1-TagRFP-T (Fig. 2 G and Fig. S2 D). Thus, similar moving spots were labeled with all three transgenes.

To determine whether either Patronin or EB1 was required to recruit the other protein to growing minus ends, we knocked each down and tracked the other. Because regulation of microtubule growth and dynamics is important in all cells and during mitosis, we used cell type–specific RNAi (Dietzl et al., 2007) to reduce expression of Patronin and EB1 post-mitotically in a subset of neurons. Like fluorescently tagged proteins, hairpin RNAs were expressed with the Gal4-UAS system so that the labeled cells were the ones subject to RNAi knockdown. Expression of a long hairpin RNA (VDRC24551) targeting EB1 completely eliminated EB1-TagRFP-T signal in ddaE neurons (Fig. 2, H and J), but did not eliminate Patronin-YFP accumulating on slow-growing puncta (Fig. 2 K). To confirm that EB1 RNAi also reduced endogenous EB1, we stained larval skins with anti-EB1 antibody (Rogers et al., 2002). EB1 expression was seen in the ddaE neuron and in glia surrounding the cell body (Fig. S1 D). When EB1 RNAi was expressed in the ddaE neuron, the glial staining surrounded a dark region occupied by the cell body (Fig. S1 D), indicating that endogenous EB1 was also very effectively reduced by the RNAi. Two other proteins that contain the calponin homology and EB1 C-terminal domain characteristic of EB proteins are encoded in the Drosophila genome. CG32371 is expressed primarily in the testis and CG18190 in the ovary (Chintapalli et al., 2007), and neither gene has been associated with any phenotypes, so EB1 is most likely the major EB protein in neurons.

Two different Patronin RNAi transgenes were very effective at reducing levels of YFP-Patronin (Fig. 2, H and I; and see Fig. 4 A for target sites of the RNA hairpins). However, slow-moving EB1 dots were still observed in these backgrounds (Fig. 2 K). To confirm that endogenous Patronin was also eliminated under these conditions, we paired Patronin tagged at the endogenous locus with Venus (Nashchekin et al., 2016) with RNAi. Patronin-Venus signal was almost completely eliminated from the ddaE neuron expressing the RNAi hairpin, but not from neighboring neurons (Fig. S1, B and C). Thus, Patronin and EB1 can be effectively reduced by RNAi in neurons, and this reduction does not affect localization of the other protein to growing microtubule minus ends (Fig. 2 L).

In vitro CAMSAP2 and 3 as well as key domains of Patronin suppress minus end growth in a concentration-dependent manner (Hendershott and Vale, 2014; Jiang et al., 2014). We therefore asked whether Patronin also regulated minus end dynamics in neurons of intact animals. To analyze minus end polymerization properties, we acquired 20-min videos of ddaE neurons expressing EB1 fused to TagRFP-T, as this fluorophore is relatively resistant to photobleaching (Shaner et al., 2008). Minus ends were considered to be puncta that moved slower than 1.6 μm per minute, and are labeled in pink in example kymographs (Fig. 3 E). In control dendrites, minus ends polymerized for an average of 240 s (Fig. 3 C), with a quarter of minus end growth events lasting longer than 5 min (Fig. 3 D). Note that the average and number of long runs are underestimates as runs that began or ended outside the 20-min videos were excluded, and these likely are over-represented for long runs. In control neurons, most of the minus end growth was away from the cell body, as in the example shown in Fig. 3 F.

With Patronin knockdown, the duration of minus end polymerization decreased to ∼80 and 60 s using two independent RNAi lines (Fig. 3 C). The persistent polymerization events over 5 min were almost completely abolished in these genotypes (Fig. 3 D). The growth length of microtubule minus ends was also significantly decreased from about 3 μm in control to about 1 μm (Fig. 3, B and E). In addition, the growth rate of microtubule minus ends was slightly increased with Patronin knockdown (Fig. 3 A). Similarly, mild overexpression of YFP-Patronin slowed minus end growth slightly compared with neurons in which minus end speed was measured with EB1-GFP or EGFP-Patronin (Fig. S2). Overall, these data support a role for Patronin in promoting extended periods of minus end growth, and slight suppression of growth rate.

As EB1 also labeled growing minus ends, we tested whether it was required for persistent minus end growth. Although EB1 RNAi effectively eliminated almost all protein (Fig. 2, H and J; and Fig. S1 D), it did not reduce the duration of minus end growth (Fig. S2, A and B). Its reduction did, however, slightly decrease minus end polymerization speed (Fig. S2 A), indicating that EB1 may promote subunit addition at the minus end.

To determine whether Patronin affected other parameters of neuronal microtubules, we examined stable microtubules and plus end growth. To monitor stable microtubules, we stained larval skins for the Drosophila MAP1B orthologue, futsch (Hummel et al., 2000). Futsch staining was slightly reduced, particularly in axons (Fig. S3, A and B), indicating that Patronin is likely required for overall neuronal microtubule stability. The behavior of plus ends was also altered when Patronin was reduced; polymerization speed was moderately increased (Fig. S3, C–E). The faster polymerization could be due to increased levels of free tubulin in neurons with less stable microtubules. Thus, reduction of Patronin has global effects on neuronal microtubules. However, the profound effect on minus end growth together with localization to minus ends suggests that its major role may be to promote minus end polymerization over extended time and distance (Fig. 3 G).

To determine whether minus end growth facilitated by Patronin might impact the overall organization of microtubules in dendrites, we assayed polarity in ddaE neurons with normal and reduced levels of Patronin. Dendrites of control ddaE neurons have around 90% minus-end-out microtubules (Stone et al., 2008; Fig. 4 B and Video 5). Knockdown of Patronin with three independent RNAi lines that target different regions of the transcript (Fig. 4 A) resulted in mixed polarity in the dorsal comb dendrite (Fig. 4, B and D; and Video 5). The polarity changes were stronger than those in kinesin-2 RNAi and mutant backgrounds (Mattie et al., 2010) and γtubulin mutant backgrounds (Nguyen et al., 2014). Despite the strong effect on dendritic polarity, no changes were seen in axons (Fig. 4 F).

To make sure that the polarity we detected using EB1 accurately reflected overall microtubule organization, we severed dendrites with a laser to produce new microtubule ends in stable regions of microtubules. All new plus ends moved toward the cell body as predicted by uniform minus-end-out polarity (Fig. S4, A and B). Similarly, Patronin accumulated only next to the cut site at the distal region of the dendrite still attached to the cell body (Fig. S4 C). Although there was no indication that polarity measured with EB1 did not capture overall microtubule polarity, we probed polarity of stable microtubule regions in control and low Patronin neurons using an assay based on microtubule steering (Mattie et al., 2010). At dendrite branch points, growing microtubules use the polarity of stable regions of microtubules to steer at the junction. When polarity is minus-end-out, this results in turning toward the cell body (Mattie et al., 2010; Fig. S4 D). If the stable regions used as tracks have mixed polarity, this will be reflected in growing microtubules turning in both directions. Indeed, in control neurons, almost all microtubule steered toward the cell body, while Patronin knockdown resulted in steering in both directions (Fig. S4 E), consistent with mixed polarity in these dendrites.

To confirm the RNAi results, we used two different mutant alleles of Patronin. Patronin^EY05252 contains a transposon insertion in its 5′ region (Fig. 4 A) and has been used previously as a hypomorphic allele (Nashchekin et al., 2016), and Patronin^k07433 has not been phenotypically characterized but contains a lethal transposon insertion in an intron of the Patronin gene (Bellen et al., 2004). In animals heterozygous for either allele, microtubule polarity remained unaffected (Fig. 4 C). Homozygous Patronin^EY05252 animals died early in development and so could not be analyzed for neuronal microtubule polarity. Homozygous Patronin^k07433 animals survived into larval stages and had normal microtubule orientation in ddaE neurons, indicating it is likely only a weak loss-of-function allele. Trans-heterozygous animals that contained one copy of each allele survived into larval stages, and dendritic microtubule polarity was mixed (Fig. 4, C and E). Similar results were obtained with Patronin^EY05252 crossed in as the male or female parent (Fig. 4, C and E). In addition to polarity, we assayed minus end behavior in the mutant background. As in Patronin RNAi neurons, the duration and length of individual minus end growth events were dramatically reduced in the mutant background (Fig. S5, A–D). Since the mutant phenotypes were consistent with the RNAi ones, we used RNAi knockdown for the remaining analysis.

ddaE neurons have relatively simple dendrite arbors (Grueber et al., 2003b), with a right angle branching pattern that makes them particularly susceptible to changes in microtubule polarity (Mattie et al., 2010). To determine whether Patronin was also important for microtubule polarity in neurons with more complex dendrite arbors with acute branching angles, we assayed phenotypes in class IV ddaC neurons with reduced Patronin function. Major dendrite branches (as indicated by red dotted lines in Fig. 4 H) were analyzed in these cells. In a control RNAi background, these dendrite regions are predominantly minus-end-out (Fig. 4 J and Video 6). When we decreased Patronin levels using RNAi BL36659, overall microtubule polarity became mixed (Fig. 4 J and Video 6). Interestingly, in the majority of neurons tested, at least one of the major dendrite branches had reversed to predominantly plus-end-out polarity (Fig. 4, I and K). As in ddaE neurons, axonal microtubule polarity remained normal in ddaC cells (Fig. 4 G). Together, these data demonstrate that Patronin regulates dendritic microtubule polarity in Drosophila sensory neurons.

To determine whether Patronin loss also affected the shape of the dendrite arbor, we used a membrane marker to label control and Patronin knockdown ddaC neurons. In RNAi experiments, Gal4 drivers that turn on as neurons are specified were used. This timing means that if Patronin protein is present in precursor cells, the phenotype may be masked in early cellular development. Control and Patronin dendrite arbors did not appear grossly different (Fig. S5 E). Indeed, the total length of the dendrite arbor was similar in both backgrounds (Fig. S5 F), as was branching pattern measured by Sholl analysis (Fig. S5 G). We also examined ddaE neuron shape in trans-heterozygous mutants. While the majority of cells looked normal, about a third had a misplaced axon and an unusual straight process (Fig. S5 H). Because no shape defects were seen with RNAi in the more elaborate ddaC neurons, we hypothesize that the ddaE phenotype may arise from a defect in mitosis or early specification of the ddaE neurons in mutant animals rather than a specific effect on dendrite outgrowth. Overall, the changes in microtubule polarity do not seem to directly result in shape changes.

Before trying to understand how Patronin controlled dendritic microtubule polarity, we wished to eliminate indirect effects in response to stress induced by reduction of a key microtubule regulator. Cytoskeletal disruption or changes in axon transport can cause a neuronal stress response that alters microtubule dynamics through transcription factors fos or FoxO (Massaro et al., 2009; Chen et al., 2012; Nechipurenko and Broihier, 2012; Valakh et al., 2013). Some of these stress responses are mediated by DLK/JNK signaling (Massaro et al., 2009; Chen et al., 2012; Valakh et al., 2013; Marcette et al., 2014) and overlap with signaling initiated by axon severing (Chen et al., 2012). After axon injury, DLK acts through JNK to turn on transcription through fos, and the MAPK phosphatase puc can be used as a reporter for this stress pathway (Xiong et al., 2010; Stone et al., 2014). Neuronal stress can also result in increased microtubule dynamics; in Drosophila and mammalian models of neurodegeneration, as well as after axon severing, more growing microtubule plus ends per unit length are observed (Chen et al., 2012; Kleele et al., 2014). Patronin loss has been shown to activate DLK signaling in C. elegans (Marcette et al., 2014), so we wished to determine whether this kinase pathway was also activated here.

To determine whether reduced Patronin activated a neuronal stress response, we used puc-GFP as a reporter of JNK/fos activation. In control neurons, low levels of puc-GFP were seen in the nucleus of the ddaE neuron (Fig. 5, A and B). Both Patronin RNAi lines resulted in elevated puc-GFP nuclear signal (Fig. 5, A and B). Consistent with activation of a stress response, microtubule dynamics was also elevated (Fig. 5 C). Thus, knockdown of Patronin initiated a stress response similar to that induced by axon injury. As this type of stress response can dramatically affect microtubules (Chen et al., 2012), we wished to separate secondary stress phenotypes from phenotypes that might be directly due to loss of Patronin.

To dampen the stress response, we fed animals two different inhibitors of JNK and DLK and chose concentrations that resulted in puc-GFP signal similar to control levels in Patronin RNAi neurons (Fig. 5, A and B). SP600125 is a JNK inhibitor (Bennett et al., 2001) that has been broadly used, including in Drosophila larvae (Chen et al., 2010). GNE-3511 is a more recently generated DLK inhibitor (Patel et al., 2015). Treatment with GNE-3511 seemed to partially reduce microtubule dynamics in both RNAi backgrounds, and SP600125 results were ambiguous in this assay (Fig. 5 C), consistent with it being slightly less effective at eliminating the puc-GFP signal. Microtubule polarity remained mixed in Patronin RNAi neurons exposed to either drug (Fig. 5, D and E). Like polarity, the reduction in persistent minus end growth was not suppressed by the DLK inhibitor (Fig. 5, F and G). We conclude that reduction of Patronin induces a DLK- and JNK-dependent transcriptional stress response and that increased microtubule dynamics may be secondary to this stress response (Fig. 5 H). In contrast, the effect of Patronin on microtubule polarity and persistent microtubule minus end growth does not depend on DLK and JNK signaling and is therefore likely due to direct function of Patronin on microtubules (Fig. 5 H).

So far, the data suggested that Patronin promotes individual bouts of minus end growth that can continue for minutes and extend the microtubule over 10 μm. In addition, Patronin function was required for the minus-end-out arrangement of dendritic microtubules. γTubulin concentrates at dendrite branch points and is important for the minus-end-out organization of dendritic microtubules (Fig. 6 A; Nguyen et al., 2014). However, most dendrite tips do not contain obvious patches of γtubulin (Fig. 6 A). We therefore hypothesized that minus end growth mediated by Patronin could populate the tips of dendrites beyond branch points with minus-end-out microtubules (Fig. 6 B).

To determine whether Patronin might facilitate growth of minus ends toward dendrite tips, we analyzed microtubule polarity in terminal branches of the ddaE neuron with different levels of Patronin (Fig. 6 C). ddaE dendrite tips have mixed rather than minus-end-out polarity (Fig. 6, D and E; Stone et al., 2008). With Patronin knockdown, the proportion of minus-end-out microtubules in terminal dendrites was reduced (Fig. 6, D and E), and the number of minus ends in the same region was also lower (Fig. 6 E). The effect of Patronin overexpression was also analyzed. Patronin tagged with Venus at the endogenous locus (Nashchekin et al., 2016) was used as a control and to estimate expression levels of EGFP-Patronin (Takeda et al., 2018) and YFP-Patronin. Taking into account the two endogenous copies of untagged Patronin, EGFP-Patronin was estimated to mildly increase overall expression, and YFP-Patronin was estimated to result in about twofold more total Patronin in ddaE (Fig. 6 G and Materials and methods). As expected, the polarity in Patronin-Venus expressing neurons was very similar to neurons expressing only EB1-GFP (Fig. 6, E, F, and H). However, even mild overexpression of Patronin resulted in a dramatic increase in minus-end-out microtubules in dendrite tips (Fig. 6, F and H), suggesting that Patronin levels normally limit minus-end-out polarity microtubules in this part of the cell.

Eventually microtubules extending from their minus end are expected to reach the end of the dendrite. To determine their fate at the tip, we visualized growing minus ends by coexpressing EB1-TagRFP-T and YFP-Patronin and acquired videos of distal dendrites. Out of 19 minus ends that reached the end of a dendrite during imaging, 13 minus ends depolymerized immediately after reaching the tip, while 6 paused for ≥50 s at the tip (Fig. 6, I and J). This pausing behavior is reminiscent of the behavior of EB1-coated plus ends that encounter the cell cortex (Mimori-Kiyosue et al., 2005). Together, these data strongly support the idea that Patronin helps populate distal regions of dendrites with minus-end-out microtubules by promoting growth of the minus end.

Based on the ability of minus end polymerization to help minus-end-out microtubules grow to dendrite tips of mature neurons, we hypothesized that a similar process might take place during development to establish polarity in dendrites. To track microtubule organization during early dendrite development, we expressed EB1-GFP using the 1407-Gal4 (also known as insc-Gal4) driver, which is active in embryonic neuroblasts (Luo et al., 1994). Live imaging of microtubule dynamics in developing dendritic arborization neurons was performed in whole-mounted embryos. Developing neurons were categorized into three main stages based on their dendrite morphology. Stage 0 neurons extended short unbranched processes that were populated by microtubules growing from the cell body with plus ends leading and thus had primarily plus-end-out polarity (Fig. 7, A and B; and Video 7). Stage 1–4 dendrites have been described previously and have mixed microtubule polarity (Hill et al., 2012; Fig. 7, A and B; and Video 7). In these stages, minus ends of microtubules were observed growing into dendrites (Fig. 7 C). At later stages of dendrite development, the arbor was branched, and polarity became predominantly minus-end-out (Fig. 7, A and B; and Video 7). When Patronin levels were reduced with RNAi, overall polarity failed to convert to minus-end-out during dendrite development, and microtubule orientation remained mixed late in embryogenesis (Fig. 7, B and D; and Video 8). Together, these data demonstrate that Patronin is required to establish minus-end-out microtubule polarity during dendrite development, most likely by promoting minus end growth into dendrites.

We have previously shown that after complete dendrite removal, Drosophila dendritic arborization neurons can grow a new dendrite arbor (Stone et al., 2014). The capacity to regenerate dendrites is particularly striking in ddaC neurons, which have large complex arbors. Regeneration is well underway by 24 h after dendrite removal, and the new arbor covers its previous territory by 96 h (Stone et al., 2014). Based on the importance of Patronin in populating dendrites with minus-end-out microtubules during development, we hypothesized that it might also play a role during regeneration. Indeed, RNAi knockdown of Patronin in ddaC neurons reduced the amount of dendrite outgrowth observed 24 h after dendrites were removed with laser microsurgery (Fig. 8, A and B).

During dendrite regeneration, microtubules reset their polarity de novo, and gradually reorganize from mixed orientation to uniform minus-end-out polarity (Stone et al., 2014; Fig. 8 D). To determine whether minus ends were actively growing outward in dendrites at early time points during regeneration, we imaged growing minus ends colabeled with EB1-TagRFP-T and YFP-Patronin. 24 h after dendrite removal, minus ends were readily observed extending into new dendrites (Fig. 8 C and Video 9). To determine whether minus end outgrowth might be functionally important in establishing microtubule polarity in this context, we knocked down Patronin in neurons expressing EB1-GFP and performed dendrite injury. In neurons with reduced Patronin, very few minus-end-out microtubules entered regenerating dendrites, and microtubule polarity remained over 75% plus-end-out at 16 h, 24 h, and 48 h after dendrite injury (Fig. 8, C–E; and Video 10). These data suggest that Patronin facilitates growth of minus-end-out microtubules in regenerating dendrites and that this process is essential for dendrite regrowth and reestablishment of minus-end-out polarity.

In contrast to the traditional view that microtubule plus end growth is exclusively responsible for extending microtubules, we found that in neurons, minus ends undergo extensive growth, with some individual runs lasting ≥10 min. Surprisingly, Patronin facilitates the persistent slow growth of minus ends, allowing them to elongate microns in single growth runs. Moreover, this minus end growth has critical functions in organizing the neuronal microtubule cytoskeleton, including populating growing dendrites and distal regions of mature dendrites with minus-end-out microtubules.

Neuronal DLK/JNK stress signaling is a common response to reduced Patronin in C. elegans (Marcette et al., 2014) and Drosophila (Fig. 5). As this signaling is responsible for some of the phenotypes observed in both systems, it will be important to determine whether any of the phenotypes that have been described in mammalian neurons with reduced CAMSAP2 are downstream of DLK or JNK. Changes in microtubule dynamics are a key output of neuronal stress from expression of proteins that cause neurodegeneration in flies (Chen et al., 2012) and mammals (Kleele et al., 2014), so some of the phenotypes attributed to direct function of CAMSAP2 could be secondary to stress. Our data suggest that in Drosophila, Patronin directly regulates minus end extension and microtubule polarity, but that changes in microtubule dynamics may be secondary to stress signaling.

Persistent minus end growth facilitated by Patronin could be an evolutionary innovation in animals like Drosophila that have mostly minus-end-out dendrites. However, we were also able to visualize growing minus ends that behaved similarly to the ones in Drosophila dendrites in plus-end-out zebrafish sensory endings. If this phenomenon is widespread across evolutionarily distant taxa, then why has it not previously been reported? EB proteins have been used to visualize microtubule plus end growth in neurons since 2003 (Stepanova et al., 2003), but slow-moving dots have not been noted before in neurons. One reason for this is likely improvements in sensitivity and signal-to-noise detection in microscopes used for live imaging. However, upon close inspection, there seem to be occasional minus ends in published EB data; for example, slow dim EB structures are present in kymographs from Drosophila (Fig. 1 A in Ori-McKenney et al., 2012) and mammalian neurons (Figs. 1i and 7b in Sánchez-Huertas et al., 2016; Fig. 4 A in Muhia et al., 2016; Fig. 1 F in Stepanova et al., 2010). In the future, it will be important to investigate behavior of microtubule minus ends in neurons from other species in detail to determine whether minus ends grow for long periods and over substantial distances.

It is particularly interesting to compare minus end growth across species because of the expansion of the Patronin/CAMSAP family in vertebrates. In zebrafish, we were able to observe clear minus end growth, but do not yet know which of the five CAMSAPs (two each of CAMSAP1 and 2 and one CAMSAP3; King et al., 2014) in fish might be present at these growing minus ends. In contrast to CAMSAP2 and 3, which suppress minus end growth, CAMSAP1 tracks minus ends as they grow (Hendershott and Vale, 2014; Jiang et al., 2014), so it is a good candidate. However, knockdown of CAMSAP1 has not been associated with changes in microtubule behavior in immortalized cell cultures (Jiang et al., 2014) and was not detected in neurons in the hippocampus (Yau et al., 2014). In contrast, CAMSAP2 knockdown has dramatic effects on microtubule stability in immortalized cells (Jiang et al., 2014) and neurons (Yau et al., 2014), so it is an open question what family member might be present on the growing minus ends we tracked in zebrafish sensory neurons and how different types of CAMSAPs might cooperate or compete with one another.

The finding that Drosophila Patronin facilitates, rather than reduces, minus end growth in neurons is surprising in light of its previous characterization. Rather than tracking growing minus ends like CAMSAP1, it is thought to dampen minus end growth like CAMSAP2 and 3 (Hendershott and Vale, 2014). One possible explanation for this discrepancy between Patronin behavior in vitro (Hendershott and Vale, 2014) and in neurons is that in specific cell types or contexts, its behavior at minus ends could be regulated. In vitro, the coiled coil domain of Patronin tracks dynamic minus ends, and when the neighboring CKK domain is added, the behavior switches to growth suppression (Hendershott and Vale, 2014). In vivo, perhaps the effect of the CKK domain is modulated to uncover the minus end growth tracking ability of the coiled coil domain.

The neuronal phenotypes from reduction of Patronin are consistent with a role in minus end growth, and we propose that extended periods of minus end growth allow minus-end-out microtubules to grow into dendrites. Consistent with this idea, Patronin reduction resulted in fewer minus-end-out microtubules in distal regions of mature dendrites as well as in developing and regenerating dendrites. However, other mechanisms likely work in parallel with minus end growth to populate dendrites with microtubules.

Sliding of minus-end-out microtubules along plus-end-out ones by kinesin-6 motors was the mechanism first proposed to populate dendrites with minus-end-out microtubules in mixed polarity mammalian neurons (Sharp et al., 1997; Yu et al., 1997, 2000), and the function of kinesin-6 has been confirmed more recently with newer methods (Lin et al., 2012). In Drosophila, these motors could play a role in dendritic microtubule organization early in development when microtubules in dendrites are mixed, but it is harder to imagine how they might function in mature minus-end-out dendrites. Analysis of Pavarotti, one of the Drosophila kinesin-6 motors, confirms that it is unlikely to play a major role in polarity of mature dendrites, and actually reduces sliding of microtubules by kinesin-1 into developing axons (Del Castillo et al., 2015).

Local nucleation of microtubules in dendrites can also help populate dendrites with microtubules and, in contrast to sliding, has been shown to be important in Drosophila (Ori-McKenney et al., 2012; Nguyen et al., 2014) and C. elegans (Harterink et al., 2018). In mammalian neurons, noncentrosomal microtubule nucleation also predominates and is known to be critical for axon growth and microtubule organization (Stiess et al., 2010; Sánchez-Huertas et al., 2016), though the localization and role of nucleation sites in dendrites has not been detailed so far. During C. elegans development, microtubule nucleation operates in parallel to minus end regulation by Patronin (Wang et al., 2015), and both types of minus end regulation may cooccur broadly in different cell types and species. In different contexts, nucleation or Patronin may vary in relative importance. For example, during initial dendrite outgrowth, we did not observe strong morphological defects with Patronin knockdown, but during regeneration, dendrite growth was strongly perturbed. While differences in RNAi knockdown could account for the phenotypic differences, it is also possible that nucleation can compensate more completely for Patronin reduction in development compared with regeneration.

Overall, we have demonstrated that, rather than existing as static structures, microtubule minus ends can grow in Drosophila and zebrafish neurons. Minus end growth is slow, but persistent, and up to 10 μm can be added in a single episode of growth. While both EB proteins and Patronin can bind to minus ends, they do so independently, and Patronin is critical for extended stretches of minus end growth. Without Patronin-mediated minus end growth, minus-end-out microtubules were reduced in dendrite tips and developing dendrites. Moreover, extra Patronin was sufficient to convert dendrite tips from mixed polarity to minus-end-out. Dendrite regeneration was particularly sensitive to Patronin reduction. We conclude that minus end microtubule growth is likely to be broadly important in helping minus-end-out microtubules reach cellular regions that are distal to nucleation sites.

Drosphila were maintained on cornmeal fly food at 20 or 25°C. 1 liter of fly food contains 4.5 g of agar, 25.9 g of sucrose, 51.7 g of dextrose, 15.5 g of yeast, 4 ml of propionic acid, 6 ml of tegosept, and 85.8 g of cornmeal. For most experiments, embryos were collected on 35-mm plates filled with fly food for 24 h, and then caps were aged at 25°C for 3 d. To study developing neurons, fly embryos were collected for 24 h at 25°C on apple juice agar caps supplied with yeast paste. One neuron was imaged per animal. Sex of larvae was not differentiated in this study.

Zebrafish were maintained in an Aquaneeering fish habitat, at a density of three fish per liter, sex separated, with weekly mating intervals. Adult fish were fed brine shrimp twice daily and were kept on a 14-h-light/10-h-dark cycle at 28.5°C at pH 7.5. To generate larvae for experiments, embryos were raised in glass bowls at 28.5°C in 1× Blue Water (0.6 g/l Instant Ocean salt mix and 0.01 mg/l methylene blue), and imaged before requiring food. Sex was not determined as fish have nonchromosomal sex selection that does not take place by the time point of 72 h post-fertilization (hpf). 50% water exchanges were performed with fresh 28.5°C Blue Water every day. All zebrafish experiments were performed after receiving approval from the Institutional Animal Care and Use Committee, and guidelines on euthanasia and pain management were followed.

Drosophila strains were obtained from both the Bloomington Drosophila Stock Center (BDSC) and the Vienna Drosophila Resource Center (VDRC). Information about the strains including identifiers is contained in Table S1.

To generate YFP-tagged Patronin, we first amplified the coding sequence of Patronin from cDNA clone AT18914 (Drosophila Genomics Resource Center collection) using primers: forward, 5′-ATG​GCG​CGC​CAA​TGG​ATG​TCG​AAA​CAC​AGG​AA-3′; reverse, 5′-ATA​GGT​ACC​TTA​GAT​TAC​AAG​CGC​CAT​GTC​T-3′. An AscI digestion site was included in the forward primer and a KpnI one in the reverse primer for cloning purposes. The coding region was then inserted to multiple cloning sites of a modified pUAST vector with three copies of YFP preinserted at the 5′ side of the multiple cloning site (Rolls et al., 2007). Next, this pUAST-3xYFP-Patronin plasmid was injected into fly embryos by BestGene Inc. Standard screening and mapping processes were applied, and fly lines expressing high or low levels of UAS-3xYFP-Patronin (denoted YFP-Patronin) were used.

To generate TagRFP-T–tagged EB1, we first amplified the coding sequence of TagRFP-T from TagRFP-T-EEA1 plasmid (42635; Addgene). The start codon was changed to GTG, and 3× stop codon of TAGTAATGA was added to the 3′ end using primers: forward, 5′-TAT​GGC​CGG​CCT​GCT​AGC​GGT​AGT​GGA​GGT​GTG​GTG​TCT​AAG​GGC​GAA​GAG-3′; reverse, 5′-CTT​CTA​GAT​CAT​TAC​TAC​TTG​TAC​AGC​TCG​TCC​ATG​CC-3′. An FseI site was included in the forward primer and XbaI to the reverse primer for cloning. Previously described pUAST-EB1-tdimer12 (Mattie et al., 2010) plasmid that contains the Drosophila EB1 coding sequence upstream of the tdimer12 coding sequence under control of a 5X UAS sequence was used as the backbone, and the tdimer12 coding sequence was replaced with that of TagRFP-T to make pUAST-EB1-TagRFP-T plasmids. Fly lines were generated as described above. Lines that resulted sharp EB1 comets when crossed with our standard Gal4 drivers were used in this study.

Tester lines used in study included (1) UAS-dicer2; 221-Gal4, UAS-EB1-GFP/TM6, (2) 221-Gal4, UAS-dicer2, UAS-EB1-TagRFP-T/TM6, (3) UAS-YFP-Patronin/CyO-pAct-GFP; 221-Gal4, UAS-dicer2, UAS-EB1-TagRFP-T, (4) 221-Gal4, UAS-dicer2, UAS-EB1-TagRFP-T, puc-GFP/TM6, (5) 477 Gal4, UAS-EB1-GFP/CyO; UAS-dicer2, (6) 1407-Gal4, UAS-EB1-GFP; UAS-dicer2, and (7) UAS-dicer2; ppk-Gal4, UAS-mCD8-GFP. Flies from tester lines were crossed with RNAi animals to generate larvae for experiments. UAS-dicer2 was included in every RNAi experiment to generate greater knockdown efficiency (Dietzl et al., 2007). 221-Gal4 was used to drive transgene expression in class I dorsal dendritic arborization ddaD and ddaE neurons (Grueber et al., 2003a), but only ddaE neurons were used for quantification. 477-Gal4 or ppk-Gal4 was used for class IV ddaC neuron expression (Grueber et al., 2003b). 1407-Gal4 (Luo et al., 1994) drove expression of transgenes in a cluster of neurons including ddaE in embryonic and early larval stages. EB1-GFP or EB1-TagRFP-T was used to study microtubule dynamics. mCD8-GFP was used as a cell shape marker. Patronin mutants EY05252 and k07433 were rebalanced with CyO-pAct-GFP balancer so that mutant larvae could be identified. To control for any maternal effects, we used female flies from EY05252 (marked as k07433/EY05252) as well as female flies from k07433 (marked as EY05252/k07433) to generate trans-heterozygous mutant animals. γTubulin-37C RNAi was used as control for RNAi experiments as γTubulin-37C is a maternal gene that is not expressed in somatic tissues (Wiese, 2008). yw was used as the control background for Patronin mutant experiments. Many fly lines used in this study were acquired from the BDSC or the VDRC. Detailed fly line information can be found in Table S1.

To generate the pEX_P2X3.E1B:LexA-VP16.Sv40.4XLexAop.E1B:EB3-GFP.Sv40 plasmid that was used to mark growing microtubule ends in zebrafish, we performed a Gateway LR reaction with the following entry vectors: p5E: P2X3 enhancer fragment (Palanca et al., 2013); pME: E1B.LexA-VP16.Sv40.4xLexAop.E1B; p3E: EB3-GFP. p3E: EB3-GFP was generated by PCR amplification from pCSEB3-GFP containing the human EB3 protein tagged with GFP (Distel et al., 2010). PCR amplification was performed with nonpriming additions of attB sites for use with a Gateway BP reaction with a 3′ DONR vector (primer sequences: forward, 5′-GGG​GAC​AGC​TTT​CTT​GTA​CAA​AGT​GGC​TGC​CAC​CAT​GGC​CGT​CAA​T-3′; reverse, 5′-GGG​GAC​AAC​TTT​GTA​TAA​TAA​AGT​TGT​TTA​CTT​GTA​CAG​CTC​GTC​CAT​GC-3′). The destination vector used for the LR reaction was pDestTol2pA2. The final product was coinjected with tol2 mRNA into zebrafish casper strain embryos at the one cell stage. An approximate volume of 1–2 nl and 60 ng/ml concentration for both plasmid and RNA was used for injections. The fish with the most labeled neurons were selected to be outcrossed so their offspring could be screened for transgenic founders. A transgenic founder with mostly uniform labeling of the sensory neurons with EB3-GFP at an expression level that enabled imaging of EB comets was selected, and its progeny were used for imaging experiments.

For zebrafish live imaging, larval fish (48 to 96 hpf) were anesthetized in 0.4% tricaine and embedded in 1.5% low melt agarose containing 0.4% tricaine and positioned on a glass coverslip. EB3-GFP comet videos were collected with a 63×/1.4 NA Plan Apochromat oil immersion objective on a Zeiss Imager M2 with a Zeiss AxioCam 506 camera using Zeiss Zen Blue software version 2.3. Overview images were collected with 10×/0.3 NA EC Plan-Neofluar objective or 63×/1.4 NA Plan Apochromat oil immersion objective on an inverted Zeiss LSM800 confocal microscope. To provide a cell shape marker for cell morphology, the P2X3.E1B:LexA-VP16.Sv40.4XLexAop.E1B:EB3-GFP.Sv40 fish described above were crossed to P2X3.E1B:LexA-VP16.Sv40.4xLexAop.E1B:mCherry.Sv40, ISL1.E1B:LexA-VP16.Sv40.4xLexAop.E1B:tdTomato.SV40 fish.

For Drosophila live imaging, 3-d-old larvae were mounted between a dried-agarose–coated microscope slide and a coverslip for immobilization. The larva was immobilized by gentle pressure from the coverslip, which was held down with sticky tape. All imaging was performed at room temperature. To study dendrite microtubule polarity, EB1-GFP videos were recorded using a widefield Zeiss Imager M2 microscope with AxioCam M2 or AxioCam 506 camera under 63× 1.4 NA Plan Apochromat oil immersion objective. A Colibri2 LED illumination system and Zeiss Zen Blue software were used to perform one frame per second, 300-frame recordings (Fig. 1, Fig. 4, and Fig. 5). Videos were exported, inverted, aligned, and analyzed with ImageJ software (https://imagej.nih.gov/ij/). Kymographs were generated with the MultipleKymograph ImageJ plugin. At least two kymographs, one with line width one and another with line width three, were generated from each video. EB1 comet numbers were counted manually from videos and confirmed with analysis of kymographs. As each cell contains limited numbers of growing microtubules, data from all cells were pooled to generate the fraction of comets moving toward or away from the cell body in all neurons. For dendrite branch analysis of ddaC neurons, only thick dendritic branches close to cell body were included for quantification. Dendrite branches were classified as plus-end-out if >90% of EB1-GFP comets move away from the cell body, minus-end-out if >90% of EB1-GFP comets moved toward the cell body, and mixed for intermediate values.

For analysis of dendrite tips, secondary dendrites of ddaE neuron without branches were sampled. A 221-Gal4, UAS-EB1-TagRFP-T tester line was crossed to YFP-Patronin, Patronin-Venus (Nashchekin et al., 2016), or EGFP-Patronin (Takeda et al., 2018), and videos were obtained on a Zeiss LSM800 microscope.

All other imaging experiments were performed with confocal microscopes, including an inverted Zeiss LSM700, an inverted Zeiss LSM800 (both of these were equipped with manual MicroPoint laser systems [Andor Technology] to perform injury), or an upright Zeiss LSM800. Time-lapse recording of EB1-GFP or EB1-TagRFP-T was performed with a 63× 1.4 NA oil objective using at 0.5 frames per second for 5 min, or 20 min for minus end polymerization studies shown in Fig. 3 and Fig. 5. Quantification of minus end growth length, duration, and speed was done based on kymograph measurements. To quantify the density of plus ends, the number of microtubule plus ends in a 200-s video was divided by the length of in-focus dendrite. The values were multiplied by 10 to give the minus end number per 10 μm. Similar measurements were done for minus end density (Fig. 6 E), except the number of minus ends was quantified from 5-min videos. Overview images were generated using the tiling function of Zeiss Zen Blue software.

The kymograph in Fig. S1 A was generated from data acquired using the AiryScan detector and processing on a Zeiss LSM800 upright microscope acquired with a Plan Apochromat 63× 1.4 NA oil objective.

Third-instar larva were dissected and stained as previously described (Nguyen et al., 2011; Tao et al., 2016). Briefly, larvae were dissected in Schneider’s medium followed by fixation in 4% PFA for 30 min. Then larval fillets were washed/permeabilized three times for a total of 20 min in permeabilizing buffer (PBS and 0.2% Triton). Next, larval fillets were incubated in blocking buffer (PBS, 0.2% Triton, 10% normal goat serum, and 2% BSA) at room temperature for 1 h. Larvae were then incubated with primary antibody overnight at 4°C. Five washes with blocking buffer were applied the next day, followed by incubation with secondary antibody for 2 h at room temperature. Final washes with blocking buffer were performed before imaging. For EB1 staining, third instar larvae were collected from crossing tester line UAS-YFP-Patronin; 221-Gal4, UAS-dicer2 to control or EB1 RNAi flies. EB1 antibody (a gift from S. Rogers, University of North Carolina, Chapel Hill, NC) was diluted 1:100 in blocking buffer. Cy5 goat anti-rabbit (Life Technologies) secondary antibody was diluted 1:100 in blocking buffer. For MAP1B/futsch staining, third instar larvae were collected from crossing tester line UAS-mCD8-RFP; 221-Gal4 to control or Patronin RNAi BL36659 flies. 22C10 (Developmental Studies Hybridoma Bank) antibody was diluted 1:10 in blocking buffer. Rhodamine-Red-X-coupled secondary antibody (Jackson ImmunoResearch) was diluted 1:100 in blocking buffer.

To quantify 22C10 (futsch) FI, z-stack images of control or Patronin RNAi neurons were acquired with the same settings. After maximum projection of each imaged neuron, raw FI was measured along dendrites (50 µm from the base of a dendrite) and axons (15 µm from the base of axon) of ddaE neurons. FI of a neighboring ciliated dendrite was used for normalization to account for regional differences in staining efficiency.

To study microtubule polarity in developing neurons, the 1407-Gal4, UAS-EB1-GFP; UAS-dicer2 tester line was crossed to control or Patronin RNAi lines. Embryos were collected from apple juice agar caps supplied with yeast paste for a 24-h window at 25°C. Embryos were carefully transferred to a mesh-bottom chamber using a fine paintbrush and then dechorionated with 50% bleach for 2 min, fully rinsed, and transferred to a 1.5-ml centrifuge tube containing 0.5 ml of heptane. Next, embryos were placed on a coverslip to make a single layer. Immediately after heptane evaporated, embryos were covered with a thin layer of halocarbon oil 27 (Sigma-Aldrich). The coverslip with embryos was then taped to a metal microscope slide with an open window in the middle. Images and videos were recorded as described above with Zeiss LSM800 upright microscope.

To study microtubule polarity in regenerating ddaC dendrites, the 477-Gal4, UAS-EB1-GFP/CyO; UAS-dicer2 tester line was crossed with control or Patronin RNAi fly lines. 3-d-old larvae were collected from cornmeal food caps to perform a dendrite severing assay. A mounted larva was visualized under 63× 1.4 NA oil objective with a LSM700 microscope, and a MicroPoint pulsed UV laser was used to sever all the dendrites at their base. Larvae were then allowed to recover on food caps at 25°C for the times indicated before being assayed for dendrite regeneration and microtubule polarity.

To assess dendrite regeneration flies from the UAS-dicer2; ppk-Gal4, UAS-mCD8-GFP tester line were crossed with control or Patronin RNAi flies. Dendrite removal was performed as above. Injured neurons were reimaged with an LSM800 upright microscope 24 h after injury, and the longest diameter across the regenerating dendrite arbor was measured for each neuron.

To visualize minus end growth in regenerating dendrites, UAS-YFP-Patronin/CyO-pAct-GFP; UAS-EB1-TagRFP-T/TM6 flies were crossed with 477-Gal4 flies. Dendrite removal was performed as above, and EB1-TagRFP-T and YFP-Patronin were recorded at 0.5 frames per second for 5 min.

To probe stable microtubules, images were acquired immediately after severing. UAS-mCD8-RFP; 221-Gal4 flies were crossed with UAS-YFP-Patronin flies, and the comb dendrite of ddaE was cut. To study microtubule polarity adjacent to the severing site, UAS-YFP-Patronin/Cyo-pAct-GFP; 221-Gal4, UAS-EB1-TagRFP-T, UAS-dicer2 flies were crossed to control RNAi flies, and the dendrite was severed.

UAS-YFP-Patronin/CyO-pAct-GFP; 221-Gal4, UAS-dicer2, UAS-EB1-TagRFP-T flies were crossed with control, Patronin RNAi VDRC27654, Patronin RNAi BL36659, or EB1 RNAi VDRC24451 flies. Images of all genotypes were acquired with the same settings. FIs of YFP-Patronin and EB1-TagRFP-T were measured in the cell body after maximum projection of each imaged neuron. Then the experimental FI values were divided by the mean FI of control neurons. Mean and SD were calculated from normalized values.

To test the efficiency of Patronin RNAi against endogenous Patronin protein, we took advantage of the endogenously tagged Patronin-Venus fly line generated by the CRISPR-Cas9 system (Nashchekin et al., 2016). Either ppk-Gal4, UAS-TagRFP-T or ppk-Gal4, UAS-TagRFP-T, UAS-Patronin RNAi 2 flies were crossed to Patronin-Venus flies. Z-stacks of ddaC neurons were obtained on LSM 800 upright microscope with 63×/1.4 NA Plan Apochromat oil immersion objective. FI of ddaC cell body and adjacent nonneuron area were measured on maximum projection of z-stacks. Normalized FIs of ddaC cell body were calculated by subtracting FIs of adjacent nonneuron areas. Each normalized ddaC cell body FI was divided by the mean FI of the control group. Mean and SD of control and Patronin RNAi groups were then plotted.

The tester fly line UAS-dicer2; ppk-Gal4, UAS-mCD8-GFP was crossed with control or Patronin RNAi line VDRC27654. Tiled images were generated with a 63×/1.4 NA Plan Apochromat oil immersion objective on an upright LSM800 microscope using Zeiss Zen Blue software. We used the Simple Neuron Tracer plugin in Fiji to trace and label every dendrite of each ddaC neuron, exported traces data to Excel, and calculated the total dendrite length. The Sholl analysis plugin in Fiji was used to generate dendrite complexity data.

To study ddaE neuron morphology, we used a Patronin trans-heterozygous mutant as described above with EB1-GFP as the cell shape marker. A binary classification of normal or abnormal morphology was applied to a total of 72 neurons imaged from 12 larvae. One exception that was classified as abnormal morphology had no ddaE neuron in the hemisegment.

We used Patronin tagged with Venus, 3xYFP, and EGFP. Images of larvae containing each transgene were acquired with the same settings. To compare intensity across fluorophores, we generated a conversion factor based on percent absorbance at the excitation wavelength, the extinction coefficient, the quantum yield of each florescent protein, and the proportion of their emission within the collection window of 495 to 529 nm. The photophysical properties were acquired from the website http://FPbase.org. Additionally, given that fluorescent protein multimers result in an increase in brightness that is not quite linear with copy number (Shearin et al., 2014; Dunsing et al., 2018), we used a multiplier of 2.25× rather than 3× when estimating the expected brightness of the 3xYFP fusion. The normalization equation and results are shown below. Relative signal per molecule represents the efficiency of emission compared with a perfect fluorophore.(Extinction Coefficient) × (Percent Absorbance at 488 nm) × (Quantum Yield) × (Proportion of Emission between 495 and 529 nm) = Relative Signal per Molecule.

Values calculated for each fluorophore and tagged protein are shown in Table S2. The Patronin-Venus was present at the endogenous locus on one chromosome, and so the baseline Patronin expression level would be equivalent to 2× the brightness of this signal. When converted to relative fluorescent level, the baseline amount of Patronin would be 2 × (0.476), which is then set to 1. The other two tagged Patronin transgenes are at different loci, and so would be added on top of the baseline expression level. These total expression levels (untagged + tagged Patronin) are shown in Fig. 6 G.

We used SP600125 at 1 mM in cornmeal fly food based on other reports (Hong et al., 2012), as well as in a laboratory dosage test. 0.5% DMSO was included for better solubility. 50 mM GNE-3511 was used in cornmeal fly food with 0.5% DMSO. Fly food containing 0.5% DMSO was used as vehicle control. The tester line 221-Gal4, UAS-dicer2, UAS-EB1-TagRFP-T, puc-GFP/TM6 was crossed to control or Patronin RNAi fly lines. After aging larvae for 2 d on standard food caps, animals were transferred to drug or vehicle control food and allowed to grow for another day. 3-d-old larvae were imaged for puc-GFP FI readings, microtubule polarity, and microtubule dynamics.

Statistics were performed with GraphPad Prism 7 software. Statistical methods were chosen based on the data type, and the analysis used in each case is detailed in each figure legend. In all figures where error bars are present, these represent the SD.

Fig. S1 includes a kymograph of Patronin-Venus as well as images showing knockdown of Patronin and EB1. Fig. S2 shows quantitation of minus end behavior using tagged Patronin. Fig. S3 includes images of stable microtubules in control and Patronin knockdown neurons as well as quantitation of plus end behavior in control and Patronin knockdown cells. Fig. S4 shows additional assays to probe behavior of stable microtubules in control and Patronin knockdown neurons. Fig. S5 includes analysis of microtubule behavior in Patronin mutant neurons as well as cell shape in Patronin RNAi and mutant neurons. Video 1 is of a ddaE neuron expressing EB1-GFP to monitor growing microtubule ends. Video 2 shows part of a ddaE dendrite expressing EB1-GFP. Video 3 is of microtubule dynamics in the sensory endings of a zebrafish RB neuron. Video 4 shows EB1-TagRFP-T and YFP-Patronin dynamics side by side in a region of a ddaE dendrite. Video 5 has EB1-GFP dynamics in control and Patronin RNAi ddaE neurons. Video 6 compares EB1-GFP dynamics in control and Patronin RNAi ddaC neurons. Video 7 includes EB1-GFP dynamics in ddaE neurons at different stages of development. Video 8 is of microtubule dynamics in a Patronin RNAi ddaE neuron late in dendrite development. Video 9 shows EB1-TagRFP-T and YFP-Patronin behavior in regenerating dendrites. Video 10 is of microtubule dynamics in control and Patronin RNAi ddaC neurons during dendrite regeneration. Table S1 is a list of reagents used in this study. Table S2 shows calculations of relative expression level for each Patronin transgene used.

We are very grateful to the Bloomington Drosophila Stock Center and Vienna Drosophila Resource Center (National Institutes of Health P40OD018537) for providing fly stocks used in this study. Dr. Wesley Grueber (Columbia University, New York, NY) provided us with Gal4 drivers to label da neurons including ppk-Gal4, 477-Gal4, and 221-Gal4. Ian Hertzler helped to quantitate cell shape. We thank Eric Schreiter and Johnny Saldate for their help piloting zebrafish EB3 imaging experiments during the Marine Biological Laboratory Neurobiology summer course.

Funding for this work came from the National Institutes of Health grants R01GM085115 (M.M. Rolls), R21NS066216 (M.M. Rolls), and R21NS090027 (A. Sagasti and M.M. Rolls).

The authors declare no competing financial interests.

Author Contributions: C. Feng, P. Thyagarajan, M. Shorey, D.Y. Seebold, A.T. Weiner, R.M. Albertson, K.S. Rao, and D.J. Goetschius performed the experiments and analyzed the data. A. Sagasti, C. Feng, M. Shorey, and K.S. Rao generated tools used in the experiments. C. Feng, P. Thyagarajan, and M.M. Rolls made figures. A. Sagasti provided training and supervision of zebrafish work. M.M. Rolls supervised the project and wrote the paper in conjunction with C. Feng.