Dynein Is Required for Polarized Dendritic Transport and Uniform Microtubule Orientation in Axons

The differential distribution of organelles and proteins to distinct compartments within cells is critical to their specialized functions. Proteins and organelles are transported to different subcellular compartments by the MT motors dynein and kinesin. The multi-subunit dynein complex travels towards MT minus ends whereas the majority of kinesins travel towards MT plus ends. Cargo localization depends on motor activity and MT organization¹. In neurons, the signal-sending axons contain MTs that are oriented uniformly plus end-distal, whereas the signal-receiving dendrites have MTs whose orientation is mixed². How might this difference in MT orientation be created? Dynein and kinesin not only move along MTs, but can also transport MTs³. Without the kinesin CHO1/MKLP the orientation of dendritic MTs are uniformly plus end-distal, rather than mixed, raising the possibility that MT motors may regulate MT polarity^{4, 5}. Whether dynein contributes to MT orientation in neurons remains an outstanding question.

Similar to most mammalian neurons, the fly dendritic arborisation (da) neurons have distinct axonal and dendritic compartments^6–9, and their MT organization resembles that in typical mammalian neurons^6,7. In a genetic screen we uncovered mutations in components of the dynein complex, dynein light intermediate chain 2 (dlic2) and dynein intermediate chain (dic, also called short wing), that cause a proximal shift in both organelle distribution and branch position within mutant dendritic arbours. These dynein mutations cause dendritic cargo to be mislocalized to axons and result in mixed orientation of axonal MTs. Our results provide new insight into the function of dynein in neurons, including hitherto unrecognized roles in polarized dendritic targeting and in regulating MT polarity.

A forward genetic screen uncovered mutations in the dynein complex components Dlic2 and Dic causing radical changes in da neuron dendritic arbour patterning (see Supplementary Information, Fig. S1). Mosaic analysis revealed that removing Dlic2 specifically within da neurons resulted in a drastic reduction in dendrite arborisation with greatly reduced receptive field coverage (control: 16,903 ± 3,292 μm; dlic2¹¹⁵⁷: 3,671 ± 681 μm, p<0.001; data represent mean ± standard deviation; n=4; Fig. 1a,b,f; Supplementary Information, Fig. S4, and data not shown). Not only was there a reduction in the total number of branch points (control: 499 ± 124, n=4; dlic2¹¹⁵⁷: 140 ± 29, p<0.01; n=4; Fig. 1g), fewer branches were located distally (Fig. 1b′) while the proximal dendrites branched profusely (Fig. 1b″), revealing a dramatic proximal shift in branch distribution (Fig. 1e). Also, most dlic2¹¹⁵⁷ axons are abnormally thick and many appear to have split, or branched, into multiple neurites a short distance from the soma (Fig. 1h; Supplementary Information, Fig. S2). The multiple neurites of the mutant da neurons bundle with other axons in the nerve, although only one neurite fully extends into the ventral nerve cord (VNC) (Supplementary Information, Fig. S2). These dendritic and axon phenotypes are rescued by neuronal expression of UAS-dlic2-eGFP (Fig. 1d). Likewise, a dic+ transgene rescues the lethality and phenotypes of dic¹²²⁹ animals (Supplementary Information, Fig. S3). In addition, dynein heavy chain (dhc) clones exhibited a similar, although less severe, phenotype (possibly due to differential perdurance of Dhc versus Dlic2; Supplementary Information, Fig. S3), and loss of Lis1, which interacts with dynein, phenocopies dlic2¹¹⁵⁷ (Fig. 1c, and Supplementary Information, Fig. S3). We therefore conclude that these phenotypes result from a loss of dynein function.

Dynein mediates the subcellular localization of Golgi¹⁰. In mammalian and fly neurons, Golgi structures in the form of “outposts” localize to dendrites and influence branching^{8, 11}. Indeed, distal dlic2¹¹⁵⁷ dendritic arbours displayed a nearly four-fold decrease in the number and size of Golgi outposts, which were marked by Mannosidase II-eGFP (ManII-eGFP) (number per 100 μm: control: 7.92 ± 3.11, n=6; dlic2¹¹⁵⁷: 2.00 ± 2.00, p<0.025, n=3; total size per 100 μm: control: 1.12 ± 0.36 μm², n=6; dlic2¹¹⁵⁷: 0.25 ± 0.25 μm², n=3, p<0.01; Fig. 2g,h,k,l). In contrast, there was a significant increase in the number and size of Golgi outposts in the proximal dlic2¹¹⁵⁷ dendritic arbours (number within 30 μm from soma: control: 10.00 ± 3.39, n=5; dlic2¹¹⁵⁷: 15.45 ± 4.18, n=11, p<0.025; total size within 30 μm from soma: control: 3.31 ± 1.08 μm², n=5; dlic2¹¹⁵⁷: 6.00 ± 1.75 μm², n=11, p<0.01; Fig. 2a,b,e,f,k,l). This change in Golgi outpost distribution (reduced distally, increased proximally) thus parallels the change in branch distribution.

In contrast to the scarcity of Golgi outposts in wild type axons⁸ (Fig. 2a,c,k), there was a striking increase in Golgi outpost number in dlic2¹¹⁵⁷ axons (number per 100 μm: control: 2.33 ± 2.71, n=6; dlic2¹¹⁵⁷: 34.49 ± 14.55, n=5, p<0.001; Fig. 2a–d,k). Golgi outpost distribution was similarly altered in the axons of dic¹²²⁹/dic^ts larvae (control: 1.43 ± 1.69 per 100 μm, n=19; dic¹²²⁹/dic^ts: 15.36 ± 7.29, n=11, p<0.001), which survive to 3^rd instar and whose da neurons exhibit similar, though milder, dendrite and axon defects (Supplementary Information, Fig. S3). Live imaging of ManII-eGFP in dic¹²²⁹/dic^ts neurons further revealed Golgi outposts moving from the soma into the axon and travelling anterogradely and retrogradely within the axon (Fig. 2i,j, Supplementary Information, Movies 3 and 4 and Supplementary Table 1). In contrast, in control axons, the Golgi outposts that were occasionally present were stationary, and we did not observe any Golgi outposts moving from the soma into the axon (Fig. 2i). These results suggest dynein actively prevents Golgi outposts from entering axons in wild type neurons.

Dynein also transports endosomes¹², which regulate dendritic membrane supply^{13, 14}. Indeed, disrupting endosomal function alters dendrite morphogenesis, including branch formation¹⁵. Normally, the recycling and early endsomal marker Rab4-RFP and the late endosome and lysosome marker Spinster (Spin)-RFP localize to both axons and dendrites (Fig. 3a,a′,c,c′). In dic¹²²⁹/dic^ts and dlic2¹¹⁵⁷ neurons, Rab4-RFP and Spin-RFP are virtually absent from dendrites (Fig. 3b,b′,d,d′ and Supplementary Information, Fig. S5) but are still present in axons, suggesting that dynein is required for the dendritic localization of endosomes.

We hypothesized that as Dlic2 decreases over time in dlic2¹¹⁵⁷ clones, Golgi outposts and endosomes would fail to travel distally, and new branches would only be added proximally. Indeed, time lapse analysis revealed proximal branch dynamics increased in dlic2¹¹⁵⁷ dendritic arbours and the change in branch dynamics correlated with Golgi outpost position (Supplementary Information, Fig S6), suggesting that machinery (including Golgi outposts) required for branch growth and dynamics accumulate proximally in dynein mutant neurons.

Next we examined the localization of Pickpocket (Ppk), which belongs to a family of conserved degenerin/epithelial sodium channels that likely function as sensory channels¹⁶. Whereas Ppk is normally detectable at low levels in dendrites but not in axons¹⁷ (Fig. 3e,e′), in dlic2¹¹⁵⁷ clones Ppk was present in both dendrites and axons (Fig. 3f,f′).

Besides transporting organelles and proteins, dynein also transports MTs within axons, leading us to investigate if axonal MT organization is impacted by the loss of dynein function. First we used the MT minus end marker Nod-βgal, a chimera comprised of the Nod motor domain fused to the Kinesin1 (Kin1) coiled-coiled domain and β-galactosidase¹⁸. Although full length Nod preferentially binds MT plus ends¹⁹, the Nod-βgal chimera localizes to MT minus ends in multiple cell types and is commonly used as a MT minus end marker¹⁸ (Supplementary Information, Fig. S7). In fly neurons, including da neurons, Nod-βgal localizes specifically to dendrites^{6–9, 18, 20, 21} (Fig. 4a,a′). In dlic2¹¹⁵⁷ clones and in dic¹²²⁹/dic^ts neurons Nod-βgal was still present in dendrites but frequently localized to axons as well (36% of dlic2¹¹⁵⁷ ddaC axons (n=11) and 30% of dic¹²²⁹/dic^ts ddaC axons (n=15) showed strong Nod-βgal signal; Fig. 4b,b′,c,c′). The amount of Nod-βgal in the mutant axons was variable and did not correlate with the formation of ectopic neurites or axonal width. These results are consistent with the finding that Nod-βgal localizes to the axons of fly photoreceptor neurons expressing a dominant-negative form of Glued²², which is a component of the dynactin complex that is necessary for dynein function²³. Over-expressing another dynactin complex member, dynamitin (dmn), also interferes with dynein function²³. ddaC neurons over-expressing dmn exhibited dendritic and axonal morphology defects similar to dlic2¹¹⁵⁷ and dic¹²²⁹/dic^ts, and Nod-βgal was ectopically localized to axons (50% of ppk-Gal4 UAS-dmn axons had strong Nod-βgal signal, n=30; Fig. 4d,d′).

Next we employed the axon specific marker Kin-βgal^{18, 24}, which is comprised of the Kin1 motor and coiled-coiled domains together with β-galactosidase, and found it is localized exclusively to axons in all three different dynein loss-of-function scenarios as in control neurons (Fig. 4e–h). Additionally, pre-synaptic components such as Cystein string protein (Csp), Bruchpilot and Syntaxin localized normally in dic¹²²⁹/dic^ts neurons (data not shown), although there are axonal “cargo jams” indicative of defects in axonal transport (Supplementary Fig. S7).

The ectopic axonal localization of Nod-βgal in dynein loss-of-function neurons suggests a change in MT polarity. To further analyze MT orientation we utilized EB1-GFP, which binds growing MT plus ends and takes on a comet-like appearance as the MT grows. In fly PNS axons EB1-GFP always moves away from the soma (anterograde)⁷, indicating the axonal MTs are orientated with their plus ends-distal, as in mammalian neurons². Similar to control neurons (Fig. 5a,b,e; Table 1 and Supplementary Information, Movie 1), in dic¹²²⁹/dic^ts axons EB1-GFP always moved anterogradely during 2^nd instar (Table 1); however, during 3^rd instar, a third of the EB1-GFP comets travelled retrogradely (Fig. 5c–e; Table 1 and Supplementary Information, Movie 2). Similar abnormal EB1-GFP movements occurred in dlic2¹¹⁵⁷ clones (Supplementary Information, Movie 5) and in neurons over-expressing dmn (40% retrograde, 60% anterograde, n=177; Supplementary Information, Movie 6). The loss of dynein function also affects EB1-GFP movement in the axons of class I da neurons (control: 100% anterograde, n=96; dic¹²²⁹/dic^ts 19% retrograde, 81% anterograde, n=80; Fig. 5e), even though dic¹²²⁹/dic^ts class I axons are not noticeably enlarged, nor do they have ectopic branches. Whereas the loss of dynein function perturbed the orientation of axonal MTs, the orientation of dendritic MTs appeared normal. The direction and rate of EB1-GFP movement in dendrites was comparable between control (rate: 6.49 ± 2.16 μm/min; 96% retrograde, 4% anterograde; n=75) and dic¹²²⁹/dic^ts (rate: 6.37 ± 1.95 μm/min; 98% retrograde, 2% anterograde; n=64). Thus, the loss of dynein activity disrupts the uniform orientation of MTs in axons without causing a detectable effect on the orientation of dendritic MTs.

Proper cellular morphology and function depends on the polarized localization of organelles and proteins to specific subcellular compartments. In this study, we show that dynein plays a crucial role in dendrite arbour patterning and in organizing distinct functional compartments (the axon and dendrites) of a neuron. The position of branches within a dendritic arbour has a key role in determining the inputs a neuron receives from pre-synaptic axons or, in the case of sensory neurons, the local environment. We show here that dynein is necessary for proper positioning of dendritic branches relative to the soma. As a motor, dynein likely influences branch formation by mediating the distribution of cargos that affect branch growth and dynamics. Notwithstanding an overall decrease in dendrite extension and branching in dynein mutants, time-lapse analysis of a few dendrites revealed that they extend normally but have fewer and less stable terminal branches, suggesting that decreased terminal branching is not simply caused by a decrease in dendrite growth. One likely explanation is that “branching machinery” (including Golgi outposts, endosomes and potentially other proteins and/or organelles) that are normally transported distally for dendrite extension and maintenance become trapped in the proximal arbour in the dynein mutant neurons, resulting in decreased distal branching and the formation of ectopic branches close to the cell body.

Without dynein, Golgi outposts and Ppk are present ectopically in axons, revealing a previously unappreciated role for dynein in mediating the dendrite-specific localization of organelles and proteins. One possible explanation for axonal mislocalization is that Golgi outposts and Ppk interact with a MT plus end-directed motor (e.g., kinesin) that transports them into axons in the absence of dynein. Dynein might normally transports such cargo directly to dendrites; alternatively, it is also possible that cargo first enters axons but that dynein counter-acts kinesin and carries this cargo out of axons. Since we never observed Golgi outposts moving from the soma into the axon in wild type neurons, our data favour the former possibility. In contrast to the mislocalization of dendritic protein and organelles, Kin-βgal and proteins destined for the axonal terminal retain their polarized distribution, perhaps to be expected given that kinesin mediates the majority of anterograde axonal transport¹.

In mammalian and fly neurons, axonal MTs are arrayed plus end-distal whereas dendritic MT orientation is mixed^{2, 7}. A long-standing question concerns the mechanism(s) that establish and maintain different MT orientations in axons and dendrites. Loss of dynein function causes the axonal localization of Nod-βgal and retrograde movement of EB1-GFP, indicating that minus end-distal MTs are present in these mutant axons. How might dynein regulate the orientation of axonal MTs? The sliding filament model of axonal MT transport proposes that a subset of dynein in the axon is stationary (via an interaction with stable MTs and/or actin) and that dynein’s motor domain interacts with short MT polymers, propelling plus end-distal MTs down the axon as the motor moves to the MT minus end³. Our in vivo data support the idea that in addition to transporting MTs, dynein functions as a “gatekeeper” to move minus end-distal MTs towards the soma, excluding them from the axon. Neurons lacking functional dynein would still transport MTs, likely via kinesin³, but now minus end-distal MTs would infiltrate the axon. Proximal axons likely have unique properties¹, providing a possible explanation for how minus end distal MTs would be excluded from axons but not dendrites.

Recent studies indicate that the trans Golgi network (TGN), which comprises part of the Golgi outposts^{25, 26}, can also function as a MT organizing center (MTOC)^{27, 28} and influence MT organization. Although it is conceivable that Golgi outposts mislocalized to dynein mutant axons could alter MT polarity, expressing lava lamp^{dominant-negative}, which prevents Golgi from associating with dynein without affecting dynein function²⁹, causes Golgi outposts to mislocalize to axons without altering MT orientation (Ye et al., 2007 and unpublished observations). Moreover, the change in axonal MT orientation is not likely to be simply a consequence of altered axon morphology because the loss of dynein function also alters the MT orientation of class I neuron axons, which appear relatively normal. With our current level of understanding, the model in which dynein acts as a “gatekeeper” is most consistent with our results and the findings of others.