Primary cilia signaling promotes axonal tract development and is disrupted in Joubert Syndrome Related Disorder models

Summary

Appropriate axonal growth and connectivity are essential for functional wiring of the brain. Joubert Syndrome Related disorders (JSRD), a group of ciliopathies in which mutations disrupt primary cilia function, are characterized by axonal tract malformations. However, little is known about how cilia-driven signaling regulates axonal growth and connectivity. We demonstrate that the deletion of related JSRD genes, Arl13b and Inpp5e, in projection neurons leads to de-fasciculated and mis-oriented axonal tracts. Arl13b deletion disrupts the function of its downstream effector Inpp5e and deregulates ciliary-PI3K/AKT signaling. Chemogenetic activation of ciliary-GPCR signaling and cilia-specific optogenetic modulation of downstream second messenger cascades (PI3K, AKT, AC3) commonly regulated by ciliary signaling receptors induce rapid changes in axonal dynamics. Further, Arl13b deletion leads to changes in transcriptional landscape associated with dysregulated PI3K/AKT signaling. These data suggest that ciliary signaling acts to modulate axonal connectivity and that impaired primary cilia signaling underlies axonal tract defects in JSRD.

eTOC Blurb

Guo et al. show that primary cilia-driven signaling regulates growth cone dynamics and axonal tract development. Ciliary signaling receptor activation and associated changes in signaling cascades (PI3K, AKT, AC3) and transcriptional landscape affect axons. Disrupted ciliary signaling following mutations in Arl13b or Inpp5e lead to axonal tract malformations in JSRD.

Graphical Abstract

Introduction

Primary cilia are present on virtually all cell types and, when defective, can lead to ciliopathies, occurring on the order of ~1:10,000 live births in the general population. A diagnostic feature of many ciliopathies is axonal tract defects (Green et al., 2018; Guadiana et al., 2013; Guemez-Gamboa et al., 2014; Hildebrandt et al., 2011; Reiter and Leroux, 2017; Sarkisian and Guadiana, 2015; Parisi, 2019; Poretti et al., 2007). Appropriate axonal growth and pathfinding are essential for wiring the developing brain (Chedotal and Richards, 2010; Engle, 2010). Axonal tract phenotypes in ciliopathies imply a role for primary cilia in axonal development and connectivity, but it is unclear how cilia-driven signaling emanating from near the neuronal soma acts to regulate axonal growth and connectivity.

Joubert Syndrome Related disorders (JSRD), a prominent collection of ciliopathies, are autosomal recessive, congenital disorders with a variety of neurological symptoms including developmental delay, intellectual disabilities (ID), abnormal respiratory rhythms, hypotonia, ataxia, oculomotor apraxia, and mirror movement synkinesis (Parisi, 2019). The diagnostic feature of JSRD is the axonal tract malformation called the molar tooth sign (MTS), which presents as thickened, elongated, and horizontally orientated superior cerebellar peduncles (SCPs) that fail to decussate in the midbrain (Brancati et al., 2010; Juric-Sekhar et al., 2012; Parisi, 2019; Romani et al., 2013; Sattar and Gleeson, 2011; Senocak et al., 2010). In addition to SCP malformation, absent or reduced decussation of the corticospinal tract (CST), corpus callosum (CC), and central pontine tracts are frequently observed in JSRD patients. Consistent with these axonal anomalies and the resultant changes in brain wiring, JSRD patients are often diagnosed with intellectual disabilities and autism spectrum disorder (ASD) (Marley and von Zastrow, 2012; Novarino et al., 2011; Reiter and Leroux, 2017).

Causative mutations for JSRD have been identified in 30+ cilia-related genes (Reiter and Leroux, 2017), implying a role for primary cilia in axon growth and guidance. To understand the role of primary cilia in axonal development and connectivity in the context of JSRD, we examined the function of two interrelated JSRD genes ARL13B (JBTS8, a small GTPase) and INPP5E (JBTS1, a phosphoinositide 5-phosphatase), respectively, in axonal development. ARL13B complexes with INPP5E, localize to primary cilia and both are essential for ciliary function. ARL13B is known for regulating ciliary signaling receptor (e.g., GPCRs, PDGFRα, Smo) localization, whereas INPP5E controls the phosphoinositide composition of ciliary membrane and thus receptor trafficking, localization, and downstream signaling cascade activation (e.g. PI3K/AKT) (Barral et al., 2012; Bielas et al., 2009; Cantagrel et al., 2008; Caspary et al., 2007; Garcia-Gonzalo et al., 2015; Mariani et al., 2016; Nachury et al., 2010; Phua et al., 2017; Umberger and Caspary, 2015). Deletion of Arl13b or Inpp5e impairs the ability of the primary cilium to function as a signaling hub, thus providing excellent molecular models to examine how cilia driven signaling regulate axon growth and connectivity in normal and JSRD brains.

Here, using neuron specific mouse genetic models of Arl13b and Inpp5e, related human mutations, and chemo-genetic and opto-genetic manipulation of primary cilia signaling, we demonstrate that ciliary signaling lies at the heart of appropriate patterns of axon tract development and connectivity.

Results

Aberrant axonal pathway development in Arl13b mutants

To evaluate the effect of primary cilia in axonal development, Nex-Cre line was used to delete Arl13b in post mitotic cortical projection and deep cerebellar neurons (Goebbels et al., 2006; Su et al., 2012). These projection neurons are ciliated (SFig.1). The Nex-Cre line enables Cre recombinase-mediated deletion of Arl13b allele in these neurons from E13.5 (Higginbotham et al., 2012). Cre-dependent Tau^mGFP or Ai9 reporter lines were used to visualize axons with GFP or mRFP, respectively (Hippenmeyer et al., 2005; Madisen et al., 2010). Deletion of Arl13b did not affect the survival of cortical projection or deep cerebellar neurons (SFig.2).

SCP axons of the deep cerebellar nuclei (DCN) extend toward and decussate in the caudal half of the midbrain, pass below and lateral to the red nucleus in the rostral midbrain, innervate the red nucleus, and then extend rostrally to the ventrolateral nucleus (VLN) of the thalamus (Fig.1A–B). The initial outgrowth of SCP from cerebellum is diffuse in Arl13b mutants (Arl13b^lox/lox; Nex-Cre; Tau^mGFP) (SFig.3A–D). Decussating SCP axons exiting cerebellum in controls (Arl13b ^lox/+; NexCre; Tau^mGFP) turn rostrally at 91±2.4° angle, whereas in mutants, SCP turns with a significantly wider angle (118±2.8°; SFig.3E–F), suggesting a less coalesced and horizontally oriented SCP projection. The decussating Arl13b-deficient SCP axonal bundles are wider, loosely fasciculated, and contain misoriented fibers (Fig.1C–F; SCP Width [axial]: Control, 291±17μm, Arl13b cKO, 355±12μm; n=6 brains per group). To examine the projection pattern of SCP, AAV5-Synapsin-tdTomato virus (Klapoetke et al., 2014) was unilaterally injected into the DCN of control and Arl13b-NexCre mutant mice to label DCN axons (Fig.1B, G, H, O). Compared to control, the decussation of mutant SCP is wider (Fig.1I, J, P) and the DCN axonal projection into RN (Fig.1K, K’, L, L’, Q) and VLN is significantly reduced (Fig.1M, N, R).

Figure 1. Deletion of Arl13b leads to SCP axonal tract defects.

(A) Schematic of SCP tract in sagittal (left), axial (middle), and coronal (right) sections. (B) Schematic of SCP axonal projection labeling by unilateral injection of AAV5-Synapsin-tdTom into the DCN. (C-D) GFP labeling shows “x” shaped SCP tract (outlined in white dashed line) in Arl13b^lox/+; NexCre; Taustop (C) and Arl13b^lox/lox; NexCre; Taustop (D) axial sections. (E, F) GFP labeling show SCP tract in Arl13b^lox/+; NexCre; Taustop (E) and Arl13b^lox/lox; NexCre; Taustop (F) in coronal sections (P1). White arrow (E) points to SCP decussation in control. Arrow (F) points to diffused, misoriented axons of SCP in Arl13b^lox/lox; NexCre; Taustop brains. (G, H) Coronal cerebellar sections show tdTom⁺ DCN neurons in Arl13b^lox/+; NexCre (G) and Arl13b^lox/lox; NexCre (H) brains (P60). DCN is outlined with dotted lines. (I, J) tdTom⁺ SCP decussation is wider in mutants. White dashed lines outline the width of SCP tract. (K, L) tdTom⁺ SCP axons innervate the RN. (K’, L’) Higher magnification of the RN region (white box) in K and L. (M, N) The target projection pattern of SCP into VLN of the thalamus. (O) Quantification of tdTom fluorescence intensity in the DCN indicates no significant differences. (P) Quantification of XSCP width. (Q, R) Quantification of fluorescence density of tdTom⁺ DCN axonal terminals in RN (Q) and VLN (R) indicates reduced projection in Arl13b cKO brains. Data shown are mean±SEM (n=4 brains per group). *P<0.05 (Student’s t-test). DCN, deep cerebellar nuclei; SCP, superior cerebellar peduncle; RN, red nucleus; VLN, ventrolateral nucleus; XSCP, decussation of SCP. Scale bar, 425μm (C, D); 142μm (G, H), 238μm (I, J), 832μm (K, L), 960μm (M, N). See also SFig. 3.

To examine other JSRD relevant axonal tracts disrupted in Arl13b mutants, we analyzed the corticospinal tract (CST). At P1, pyramidal decussation of CST is markedly disrupted (Fig.2A–E). Clearly decussating pyramidal axons were evident in control (Fig. 2B, D, arrowheads), but were significantly reduced in Arl13b^lox/lox; NexCre brains (Fig. 2C, E, arrows). Consistent with this, the density of axons, post decussation, innervating the distal cervical spinal cord, is reduced in Arl13b-NexCre mutants (Fig.2F [arrowhead], G [arrow]). Together, these results suggest that the NexCre deletion of Arl13b in postmitotic cortical neurons leads to disrupted CST formation and decussation.

Figure 2. Deletion of Arl13b leads to CST and CC axonal tracts defects.

(A) Schematic of CST tract in sagittal section. Dashed lines indicate where the CST decussation is and where B-C and F-G were imaged, respectively. Gray box indicates where D and E were imaged in cleared brains. (B, C) GFP⁺ CST tract decussation at lower medulla of Arl13b^lox/+; NexCre; Taustop (B, arrowhead) and Arl13b^lox/lox; NexCre; Taustop (C, arrow) brains (P1). (D, E) CST decussation in cleared Arl13b^lox/+; NexCre; Ai9 (D) and Arl13b^lox/lox; NexCre; Ai9 (E) brains. Arrowhead (D) and arrow (E) point to the CST fascicles post decussation. (F, G) Sagittal sections of Arl13b^lox/+; NexCre; Ai9 (F) and Arl13b^lox/lox; NexCre; Ai9 (G) (P1) show reduced tdTom⁺ CST tract fascicles post decussation and the reduced extension of CST axons into the dorsal spinal cord in mutants (G [arrow]; compare to control [F, arrowhead]). (H) Schematic of CC tract in coronal section. Dashed rectangles indicate where I-P were imaged. (I, J) Arl13b^lox/+ (I) and Arl13b^lox/lox (J) embryos (E15) were electroporated with pCAG-Cre and pCAG-GFP plasmids. GFP labeled callosal axons in CC were analyzed at P1. Brackets indicate the width of CC. (I’, J’) CC area highlighted in I and J. (I”, J”) Representative GFP⁺ axonal growth cone of callosal neurons at the midline. Arrowheads point to longer filopodia in mutants. (K-P) pCAG-Cre and pCAG-GFP electroporated Arl13b^lox/+ (K, M, O) and Arl13b^lox/lox (L, N, P) embryos (E15) were analyzed at P21. GFP labeling show disorganized CC axons in Arl13b^lox/lox brains (L) electroporated with pCAG-Cre when compared to control (K). (M-P) Reduced axonal branching is evident in ipsilateral layer 5 (yellow asterisk, M, N) and contralateral layers 2/3 and 5 (white asterisks, O, P) in Arl13b^lox/lox; Cre brains. (Q, R) Quantification of GFP fluorescence intensity in the ipsilateral (Q) and contralateral (R) cortical wall. Data shown are mean±SEM (n=4 brains per group). *P<0.05 (Student’s t-test). Scale bar, 135μm (B-G); 95μm (I, J); 1200μm (K, L); 415μm (M, N); 623μm (O, P).

To assess the effect of Arl13b deletion in callosal projection neurons, we performed unilateral in utero electroporation (IUE) at E15.5 with CAG-Cre and CAG-EGFP plasmids leading to Arl13b deletion and GFP labeling of callosally projecting neurons of layers 2/3 (Fig.2H). Callosal axons were analyzed at P1, as they cross the cortical midline to form the corpus callosum (CC) (Fig.2I–J). Compared to controls, Arl13b-deficient axons are loosely organized, poorly fasciculated, and form a wider CC tract (Fig.2I, I’, J, J’; midline width of CC tract: control (95.2±1.2μm), mutant (140.2±4.4μm), Student’s t-test, p =6.2E-5, n=4 brains). The midline crossing growth cones of Arl13b-deficient neurons also display longer filopodial projections compared to control neurons (Fig.2I”, J”[arrowhead]); mutant: 5.63±0.22μm, control: 2.31±0.27μm, Student’s t-test, p =0.001, n=16 neurons). At P21, as CC axons reach their targets, control CC axons extensively branch and innervate ipsilateral layer 5 and contralateral layers 2/3 and 5 (Fig.2M, O[asterisk]). In contrast, Arl13b-deficient CC displayed dis-organized axons (Fig.2K–L), as well as reduced axonal branching and innervation of target layers (Fig.2N[asterisk], P[asterisk], Q, R). Lastly, decussation defects were also evident in the pontine nuclei. Transverse pontine tegmental axons do not decussate normally and are misoriented (SFig.3G–H).

Thickened superior cerebellar peduncles that show disorganized decussation, perturbed pyramidal axon decussation, and transverse pontine fiber defects are consistent with the axonal phenotypes seen in JSRD. Our observations suggest that Arl13b deletion in projection neurons disrupts axonal tract development in a manner consistent with human JSRD phenotype. Further, these results suggest that disruption of primary cilia signaling through the deletion of Arl13b disrupts the appropriate formation and connectivity of major axon tracts in the brain, including SCP, CST, and CC. We therefore sought to explore the relationship between cilia signaling and axonal development.

Deletion of Arl13b leads to altered axon growth behavior

To examine developmental changes in patterns of axon growth in Arl13b-deficient neurons, we isolated and cultured Ai9⁺ DCN neurons from control and Arl13b^lox/lox; Nex-Cre; Ai9 (Arl13b cKO) brains. Evaluation of different stages of neuronal polarization (Banker, 2018) after initial plating suggests that axon-dendritic polarization (SFig.4) or axon formation and outgrowth are not affected in the absence of Arl13b, but fewer and shorter lateral axonal branches were seen in Arl13b-deficient neurons (Fig.3A [arrowheads], B–C). Live imaging indicates that control axonal branches arise along the main axonal shaft (Fig.3D [arrowheads], E). Consistent with the reduced axonal branching observed, mutant DCN neuron axons showed significantly reduced formation and dynamics of axonal branching protrusions (Fig.3D–E). These results suggest defects in the initiation and extension of axonal branching in Arl13b-deficient neurons.

Figure 3. Deletion of Arl13b leads to axonal growth defects in vitro.

(A) Growth of DCN neurons dissociated from Arl13b^lox/+; NexCre; Ai9 and Arl13b^lox/lox; NexCre; Ai9 (cKO) brains (P2) were analyzed at different time points as indicated. Red arrowheads point to the primary axonal branching points. (B, C) Quantification of axonal outgrowth at 3 days in vitro (DIV). Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (D) Axonal filopodia dynamics of DCN neurons from Arl13b^lox/+; NexCre; Ai9 (Ctrl) and Arl13b^lox/lox; NexCre; Ai9 (Arl13b cKO) brains. Red arrowheads point to representative axonal filipodia at time 0sec. Blue arrowheads point to newly formed axonal filopodia at time 60sec. (E) Quantification of axonal lateral branch density and half lifetime. Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (F) Representative axonal growth cones of DCN neurons from Ctrl and Arl13b cKO brains. Red arrowheads point to aberrantly long filopodia extensions. (G) Quantification of axonal growth cone (GC) area, axonal filopodia length, and growth cone filopodia/lamellipodia ratio. Data shown are mean±SEM (n=28 cells). *P<0.05 (Student’s t-test). (H) Schematic of the microfluidic chamber assay for the expression of Pcdh17-EGFP along axon-axon contacts. (I, J) Expression of Pcdh17-GFP along the BFP⁺/tdTom⁺ axon-axon contact sites in DCN neurons of Ctrl (I) and Arl13b cKO (J) brains. White brackets indicate BFP⁺/tdTom⁺ axon-axon contact regions. White arrows point to Pcdh17 expression at axon-axon contact sites. (K) Quantification of Pcdh17-EGFP⁺ axon-axon contacts. (L) Effect of human ARL13B mutations on axon growth. DCN neurons of Arl13b^lox/+; NexCre and Arl13b^lox/lox; NexCre were transfected with mCherry, mCherry-ARL13B (wild type or different human mutations), or mV358A mutant and analyzed after 3 days. (M) Quantification of primary axonal length and branch number. Data shown are mean±SEM. *P<0.05 (One-way ANOVA: F_8,216 =24.11; p=4.7E-33; post-hoc p_{[lox/+ vs. lox/lox}=6.0E-14; post-hoc p_{[lox/+ vs. lox/lox-ARL13B}=0.09; post-hoc p_{[lox/+ vs. lox/lox-G75R]}=5.6E-10; post-hoc p_{[lox/+ vs. lox/lox-R79Q]}=6.8E-09; post-hoc p_{[lox/+ vs. lox/lox-W82X]}=9.5E-09; post-hoc p_{[lox/+ vs. lox/lox-Y86C]}=4.8E-10; post-hoc p_{[lox/+ vs. lox/lox-R200C]}=2.1E-10; post-hoc p_{[lox/+ vs. lox/lox-V358A]}=2.5E-11; 25 neurons). Scale bar, 28μm (A), 8μm (D, F), 4.5μm (J, K), 30μm (L). See also SFig. 5.

We also observed reduced dendritic outgrowth of mutant DCN neurons in vitro (Fig.3A; total dendritic length _[DIV3]: control (274.1±12.9μm), mutant (116.5±10.0μm), Student’s t-test, p=2.34E-11; # of primary dendrites_[DIV3]: control (5.7±0.3), mutant (3.1±0.2), Student’s t-test, p=2.58E-08, n =16 _[Control], 17 _[mutant] neurons).

Compared to control DCN neuron growth cones that show characteristic lamellipodial veils in between dynamic filopodial extensions (t_1/2 [half lifetime]=24.4± 1.3s), Arl13b-deficient growth cones are smaller but have numerous longer filopodial protrusions (Fig.3F–G [arrowheads]) with significantly reduced dynamics (t_1/2=118.5± 3.3s, Student’s t-test, p<0.05, n=12 cells). Further, Arl13b null (Arl13b^−/−) embryonic cortical neurons also show reduced axon branching and disrupted growth cone morphology (SFig.5). Axons of Ift88 null (Ift88^−/−) cortical neurons without primary cilia were similarly affected (SFig.5). Together, these in vitro observations suggest that in the absence of appropriate primary cilia signalling, axonal growth cone dynamics and the filopodia to lamellipodia balance are altered.

The fasciculation defects in vivo and the axonal branching and growth cone defects in vitro in Arl13b-deficient axons suggest that communication and interactions between axons as they branch and extend as fiber tracts may have been perturbed in cilia mutant neurons. Intercellular recognition through adhesion molecules such as protocadherins are essential for axon growth and fasciculation. Within the protocadherin family, Protocadherin-17 (Pcdh17) is highly expressed in cortical and DCN neurons. Recruitment of Pcdh17 to axon-axon contacts is thought to facilitate the fasciculation and collective extension of the developing SCP axons (Hayashi et al., 2014; Hoshina et al., 2013). We therefore examined if primary cilia signaling is necessary to recruit Pcdh17 to points of axonal contacts in vitro. Control or Arl13b-deficient cortical neurons labeled with either BFP or tdTomato/Pcdh17-EGFP were plated onto different sides of microfluidic chambers, allowing for the establishment of direct axon-axon contacts between BFP⁺ and tdTom⁺ neurons (Fig.3H). Compared to control neurons that show expression of Pcdh17 at axonal points of contacts, Arl13b-deficient neurons show reduced recruitment of Pcdh17 to axon-axon contacts (Fig.3I–K). These observations suggest that disrupted ciliary signaling in Arl13b-deficient neurons can impair axon-axon recognition/ adhesion signals necessary for the organized growth of axonal tracts.

Multiple mutations (p. R79Q, p. W82X, p. Y86C, p. R200C, p. G75R) in ARL13B are known to cause JSRD (Cantagrel et al., 2008; Juric-Sekhar et al., 2012; Miertzschke et al., 2014; Parisi, 2019; Rafiullah et al., 2017; Thomas et al., 2015). To test whether these JSRD mutant alleles affect axonal growth, we cre-inducibly expressed them in Arl13b-deficient DCN neurons (Guo et al., 2017). Wild-type human ARL13B rescued the branching defects in the Arl13b-deficient DCN neurons (Fig. 3L–M). However, JSRD causing variants R79Q, G75R, W82X, Y86C, and R200C failed to do so consistent with their known disruption of Inpp5e ciliary targeting (Fig.3L–M; (Humbert et al., 2012). These results suggest that human mutations in Arl13b disrupt axonal development. Further, a mouse nonciliary form of functional Arl13b (V358A; Higginbotham et al., 2012) also did not rescue the axonal branching defects in Arl13b-deficient DCN neurons (Fig.3L–M), suggesting that the localization and activity of Arl13b within the primary cilium is crucial to its role in the appropriate regulation of axonal growth pattern.

Collectively, these observations reveal that neurons deficient in ciliary Arl13b display reduced axonal branching, aberrant growth cone morphology, altered filopodia/lamellipodial balance, and impaired axon-axon adhesive contacts.

Arl13b-Inpp5e pathway

To define the signaling mechanisms underlying Arl13b mediated axonal dynamics, we first examined inositol polyphosphate-5-phosphatase (Inpp5e) as Inpp5e is absent in Arl13b-deficient cilia indicating Arl13b is necessary for Inpp5e ciliary targeting (Humbert et al., 2012). Inpp5e hydrolyzes the 5-phosphate of PtdIns(4,5)P2 and PtdIns(3,4,5)P3 (PIP2 and PIP3), which act as docking phospholipids necessary for ciliary localization of signaling receptors and activation of signaling cascades. Importantly, mutations in INPP5E lead to JSRD and axonal tract defects (Bielas et al., 2009; Conduit et al., 2012; Garcia-Gonzalo et al., 2015; Jacoby et al., 2009). These observations suggest that Inpp5e acts as an effector of Arl13b. The deficiency of ciliary Inpp5e, due to either deletion of Arl13b or Inpp5e, may compromise the Arl13b-Inpp5e signaling pathway necessary for axonal tract development.

We therefore examined SCP, CST, and CC tract formation and organization in Inpp5e-NexCre (Inpp5e cKO) brains. Similar to Arl13b-NexCre mutants, Inpp5e-NexCre mutant SCP tracts are diffuse, wider, and loosely fasciculated (Fig. 4A–D; SCP Width (axial): Control, 285±19μm, Inpp5e cKO, 349±23μm; P<0.05 (Student’s t-test); n=6 brains per group). Mutant CST tract displays reduced pyramidal decussation (Fig.4E–G). Inpp5e-deficient callosal axons show significantly reduced branching and target innervation when compared to controls (Fig.4H–M). Further, reduced axonal branching and aberrant growth cone morphology similar to that of Arl13b-deficient neurons were seen Inpp5e-deficient DCN neurons in vitro (Fig.4N–Q). Reduced dendritic growth is also evident in Inpp5e mutant DCN neurons in vitro (Fig. 4N; total dendritic length _[DIV3]: control (292.1±23.4μm), mutant (101.5±8.3μm), Student’s t-test, p=1.28E-13; # of primary dendrites_[DIV3]: control (5.3±0.6), mutant (2.9±0.2), Student’s t-test, p=1.52E-09, n=16 neurons).

Figure 4. Deletion of Inpp5e leads to axonal defects similar to Arl13b cKO.

(A) Schematic of SCP tract in sagittal sections. (B- D) GFP labeling in Inpp5e^lox/+; NexCre; Taustop and Inpp5e^lox/lox; NexCre; Taustop brains (P1) show the trajectory of SCP tract in sagittal (B), axial (C), and coronal (D) sections. (B) Diffused outgrowth of SCP in Inpp5e cKO when compared to control. Dashed lines (C) outline the SCP decussation in axial sections. Yellow arrowheads in C and D point to the diffused, wider, and disrupted decussation of SCP in Inpp5e mutants. (E) Schematic of CST tract in sagittal sections. Dashed lines indicate where F and G were imaged, respectively. (F) Altered decussation of TdTom⁺ CST in Inpp5e^lox/lox; NexCre; Ai9 brains. (G) Reduced GFP⁺ CST tract fascicles post decussation and reduced extension of CST axons into the dorsal spinal cord in mutants (arrow; compare to control [arrowhead]). (H) Schematic of CC tract in coronal sections. Outlined areas indicate where J and K were imaged. (I-K) Inpp5e^lox/+ and Inpp5e^lox/lox embryos (E15) were electroporated with pCAG-Cre and pCAG-GFP plasmids and analyzed at P21. (I) Low magnification images of GFP labeled CC tract. (J-K) Reduced axonal branching is evident in ipsilateral layer 5 (yellow asterisk) and contralateral layers 2/3 and 5 (white asterisks) in Inpp5e^lox/lox brains electroporated with pCAG-Cre. (L-M) Quantification of GFP fluorescence intensity (arbitrary units) along the cortical wall of ipsilateral side (L) and contralateral side (M). Data shown are mean±SEM (n=4 brains). *P<0.05 (Student’s t-test). (N) Dissociated DCN neurons from Inpp5e^lox/+; NexCre; Ai9 (Ctrl) and Inpp5e^lox/lox; NexCre; Ai9 (Inpp5e cKO) brains were analyzed at different time points as indicated. Red arrowheads point to the primary axonal branching points. (O) Quantification of axonal outgrowth (3 DIV). Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (P) Representative axonal growth cones of DCN neurons from Ctrl and Inpp5e cKO brains. Red arrowheads point to aberrantly long filopodial extensions. (Q) Quantification of axonal growth cone area, filopodia length, filopodial half lifetime, and growth cone filopodia/lamellipodia ratio. Data shown are mean±SEM (n=24 cells). *P<0.05 (Student’s t-test). CB, cerebellum; DCN, deep cerebellar nuclei; SCP, superior cerebellar peduncle; GC, growth cone. Scale bar, 205μm (B), 150μm (C, D), 600μm (F, G), 1215μm (I), 240μm (J), 340μm (K), 25μm (N), 5μm (P).

These common defects in Arl13b and Inpp5e-deficient neurons suggest that both can regulate neuronal development from the cilium. We therefore set out to delineate the Arl13b-Inpp5e regulated signaling mechanisms, which when perturbed in cilia, could affect axonal development.

Arl13b-Inpp5e pathway converges on ciliary-PI3K/AKT signaling

PI(4,5)P2 and PI(3,4,5)P3, the main substrates of Inpp5e downstream of Arl13b, are major second messengers of receptor tyrosine kinases (RTK), GPCR, and Shh pathways (Bielas et al., 2009; Conduit et al., 2012; Garcia-Gonzalo et al., 2015; Oude Weernink et al., 2007; Phua et al., 2017). Several RTKs, GPCRs and Smo receptors selectively accumulate in neuronal primary cilia (Christensen et al., 2012; Hilgendorf et al., 2016; Reiter and Leroux, 2017). Upon activation, GPCR, Shh and RTK signaling lead to the activation of phosphatidylinositol 3-kinase (PI3K) and the production of PIP3, which in turn activates AKT to orchestrate a broad spectrum of cellular functions, including axonal outgrowth and growth cone dynamics (Hakim et al., 2012; Henle et al., 2011; Manning and Cantley, 2007; Manning and Toker, 2017; Oude Weernink et al., 2007). Consistent with the role of Inpp5e that hydrolyzes PIP3, deletion of Inpp5e leads to enrichment of PIP3 at primary cilia and elevated whole cell PI3K-AKT signaling activity (Dyson et al., 2017; Phua et al., 2017; Plotnikova et al., 2015). To test the hypothesis that ciliary Arl13b-Inpp5e signaling converges onto the production of ciliary PIP3 and activation of AKT to regulate axonal development, we first examined the level of PIP3 in Arl13b and Inpp5e-deficient neuronal cilia. We expressed PIP3 biosensors RFP-PH_AKT or mCh-PH_GRPI in DCN neurons from control, Arl13b cKO, and Inpp5e cKO brains (Idevall-Hagren et al., 2012). Compared to control, a significant increase in the levels of PIP3 was observed in Arl13b and Inpp5e-deficient neuronal primary cilia (SFig. 6A–E). We found a similar increase of PIP3 level in Arl13b null cortical neurons (SFig. 6F–K). These observations suggest that deletion of Arl13b or Inpp5e leads to increased ciliary PIP3. To test whether the increased PIP3 levels lead to elevated activity of AKT, we examined the phosphorylation level of AKT, and its downstream effectors glycogen synthase kinase 3 (GSK3), ribosomal protein S6 kinase (S6K), and transcription factor cyclic AMP-response element binding protein (CREB), in control and Arl13b null or Inpp5e deficient (Inpp5e^rdg; Su et al., 2012) brains. A significant increase of p-AKT, p-GSK, p-S6, and p-CREB is evident in both Arl13b and Inpp5e deficient brains (SFig. 6L–M), suggesting an elevated global level of AKT activity follows the loss of Arl13b or Inpp5e. Taken together, these results reveal that deficiency of Arl13b-Inpp5e in neurons lead to increased levels of ciliary PIP3 as well as elevated global AKT activity.

Optogenetic/Chemogenetic manipulations of ciliary-GPCR and ciliary-PI3K/AKT signaling change axonal behavior

We therefore set out to further delineate the role of cilia-regulated PI3K/AKT pathway on axonal development. First, to evaluate the effect of acute activation of ciliary signaling receptors upstream of PI3K on axons, we focused on ciliary-GPCR signaling that can activate the PI3K signaling pathway (Fig.5A; Spangler and Bruchas, 2017). We employed a chemogenetic approach using DREADDs (Designer Receptors Exclusively Activated by Designer Drugs) to control GPCR signaling specifically in neuronal cilia (Guo et al., 2017). DREADDs are mutant GPCRs (M3-muscarinic receptors) that have been molecularly evolved to be activated by the small molecule clozapine N-oxide (CNO). Our approach entails cilia-specific expression of a DREADD, hM3D_q, that couples to G_q and can induce Ca²⁺ activation when activated by CNO (Armbruster et al., 2007; Dong et al., 2010). Cilia-targeting-sequence [CLVCCWFKKSKTRKIKP] (Follit et al., 2010) was use the drive the ciliary localization of hM3D_q (Cilia-hM3D_q). We have previously shown that Cilia-hM3D_q triggers cilia-specific GPCR activation and related Ca²⁺ signaling (Guo et al., 2017). The expression of Cilia-hM3D_q did not significantly change the length of primary cilia (Cilia-GFP: 2.72±0.10μm, Cilia-hM3D_q: 2.64±0.11μm, Student’s t-test, p=0.56, n=14 neurons). Cilia-hM3D_q and Ca²⁺ sensor RFP-GECO1.0 were co-expressed in DCN neurons. Upon 10μM CNO treatment, filopodial protrusions, correlated with Ca²⁺ elevation (t_1/2=8.34±3.13s, n=16 cells), start to emerge at the distal end of axons (Fig.5B–D). In contrast, control DMSO treatment did not change axonal growth cone dynamics or Ca²⁺ level (Fig.5B–D). Further, ACIII (Adenylate Cyclase III), a member of type III adenylate cyclase family that can potently induce cyclic AMP (cAMP), downstream of GPCR signaling activation, is selectively expressed in neuronal primary cilia. Optogenetic activation of a blue light-sensitive, cilia-targeted type III adenylyl cyclase ([Cilia-bPAC]; (Jansen et al., 2015; Stierl et al., 2011) led to cAMP induction and rapid filopodial protrusions (Fig.5E–F) similar to Cilia-hM3D_q activation (Fig.5F–G). The expression of Cilia-bPAC did not significantly change the length of primary cilia (Cilia-GFP: 2.62±0.13μm, Cilia-bPAC: 2.96±0.12μm, Student’s t-test, p=0.51, n=12_[Cilia-GFP], 16_[Cilia-bPAC] neurons). In contrast, inhibition of ciliary ACIII activity with a modified, constitutively active, cilia-specific G_i-protein-coupled receptor, GPR88* (Siljee et al., 2018) in Arl13b-deficient neurons leads to opposite effects on filopodia/lamellipodial balance. GPR88* rescued the growth cone and filopodia phenotypes of the Arl13b-deficient neurons, reducing filopodial protrusions and increasing lamellipodial area (Fig. 5H–I). These results suggest that the activity of ciliary GPCRs can regulate axonal growth cone morphology and dynamics.

Figure 5. Chemogenetic or optogenetic manipulation of ciliary GPCR signaling leads to rapid changes in axonal growth cone dynamics.

(A) Schematic depicting optogenetic/chemogenetic approaches to manipulate different inter-related ciliary signaling pathways (GPCRs, ACIII, PI3K, AKT, Inpp5e). (B) Time-lapse imaging of control DCN neuron axon expressing Cilia-hM3D_q and RFP-GECO1.0 (pseudocolored) treated with 10μM CNO (left panel). Yellow arrowheads point to the filopodial extensions. Kymograph of the changes in RFP-GECO1.0 fluorescence intensity at the distal axonal tip (white arrow at T0) upon CNO or DMSO treatment. (C) Axonal RFP-GECO1.0 fluorescence intensity before and after CNO or DMSO treatment (n=16 cells). Data shown is mean±SEM. Relative intensity of RFP (F/F₀) at peak [CNO vs. DMSO]: p= 1.2E-8, Student’s t-test. (D) Quantification of changes in the growth cone morphology before and after CNO or DMSO treatment. Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (E) Schematic of bPAC mediated, blue light-dependent cAMP production in primary cilia. (F, G) Changes in the growth cone morphology before and after ciliary activation of bPAC in control DCN neurons expressing Cilia-bPAC and pCAG-BFP. t’ in panel F includes 20 seconds to move from growth cone (pre-light activation) to primary cilia, blue light photo-activation of primary cilia (3 mins.) and 20 seconds to move back to growth cone (post-light activation). Data shown are mean± SEM (n=24 neurons). *P<0.05 (Student’s t-test). (H) Inhibition of ciliary ACIII activity with modified GPR88* in Arl13b-deficient neurons. Live imaging of growth cone dynamics in Arl13b^lox/+; NexCre; Ai9, Arl13b^lox/lox; NexCre; Ai9, and Arl13b^lox/lox; NexCre; Ai9+ DIO-AAV-Flag-GPR88* DCN neurons. Inset shows expression of FLAG tagged GPR88* in cilium. (I) Quantification of growth cone (GC) morphology. Data shown are mean±SEM. *P<0.05 (One-way ANOVA: F_{2,42[GC area]} =32.0; p=3.6E-9; post-hoc p_{[GC area, lox/+ vs. lox/lox]}= 6.9E-06; post-hoc p_{[GC area, lox/+ vs. lox/lox-GPR88*]}=0.01; F_{2,42[# of GC filopodia]}=36.6; p=6.2E-10; post-hoc p_{[# of filopodia, lox/+ vs. lox/lox]}=1.1E-04; post-hoc p_{[lox/+ vs. lox/lox-GPR88*]}=0.001; F_{2,42[L of GC filopodia]}=50.4; p=6.9E-12; post-hoc p_{[L of GC filopodia, lox/+ vs. lox/lox]}=1.8E-07; post-hoc p_{[L of GC filopodia, lox/+ vs. lox/lox-GPR88]}=0.07; n=15 neurons). Scale bar, 4μm (B); 8.5μm (F, H).

To address how the PI3K/AKT signaling, emanating from primary cilia, exerts its impact on axons, we generated optogenetic tools to gain precise, spatio-temporal control of ciliary-PI3K/AKT signaling in neurons and evaluated if selectively activating this signaling pathway in cilia can change axonal behavior (Fig.6A). The blue light activated, CIBN/CRY2 dimerization opto-system was used to selectively recruit PI3-kinase, phosphatidylinositol 5-phosphatase modules (5-Ptase module of Inpp5e), and AKT to cilia-membrane (Gradinaru et al., 2010; Idevall-Hagren et al., 2012; Katsura et al., 2015). This optogenetic system is based on blue-light (488nm)-induced dimerization between plant proteins, cryptochrome 2 (CRY2) and the transcription factor CIBN (Idevall-Hagren et al., 2012). CIBN is expressed selectively in cilia using an established ciliary targeting sequence and is anchored to the ciliary plasma membrane through myristoylation (Su et al., 2013). CRY2 is fused to various second messenger modules. Blue light-induced dimerization between CRY2 and the ciliary membrane anchored CIBN (Cilia-CIBN) leads to the recruitment of these modules to cilia. The expression of Cilia-CIBN did not significantly change the length of primary cilia (Cilia-GFP: 2.93±0.22μm, Cilia-CIBN: 2.89±0.13μm, Student’s t-test, p =0.82, n=14_[Cilia-GFP], 15_[Cilia-CIBN] neurons).

Figure 6. Optogenetic manipulation of ciliary PI3K/AKT signaling leads to rapid changes in axonal growth cone dynamics.

(A) Schematic of optogenetic recruitment of PI3-kinase to the cilium and evaluation of changes in axonal growth cone. Untagged (CRY2-iSH2) or mCherry tagged iSH2 (mCh-CRY2-iSH2) was used. PIP3 sensor mCh-PH_GRPI was not included in experiments where mCh-CRY2-iSH2 was used. (B-C) Expression of Cilia-CIBN-GFP and mCh-CRY2-iSH2 (B) or mCh-PH_GRPI (C) in the primary cilia of control DCN neurons before and after ciliary blue light activation (upper panels, B-C). Lower panels (B-C) illustrate the dynamics and kymographs of the ciliary mCh-CRY2-iSH2 (B) and mCh-PH_GRPI (C) intensity before and after ciliary blue light illumination (n=16 cells). (D-E) Changes in the growth cone morphology before and after ciliary recruitment of PI3-kinase in two different DCN neurons expressing Cilia-CIBNGFP and CRY2-iSH2. Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (F) Schematic of optogenetic recruitment of AKT to the ciliary membrane. (G) Expression of Cilia-CIBN-GFP and mCh-CRY2-AKT in the primary cilia of control DCN neurons before and after ciliary blue light activation (upper panels). Lower panels illustrate the dynamics and kymograph of ciliary mCh-CRY2-AKT intensity before and after ciliary blue light activation (n=16 cells). (H-I) Changes in the growth cone morphology before and after ciliary recruitment of AKT in two different DCN neurons expressing Cilia-CIBN-GFP and CRY2-AKT, and mCh-PH_GRPI. Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (J) Schematic of optogenetic recruitment of 5pase_INPP5E to the ciliary membrane. (K) Expression of Cilia-CIBN-GFP and mCh-CRY2–5pase_INPP5E in the primary cilia of DCN neurons. (L) Dynamics and kymograph of PIP3 sensor mCh-PH_GRPI intensity in primary cilia after ciliary recruitment of 5pase_INPP5E (n=16 cells). (M-N) Changes in the growth cone morphology before and after ciliary recruitment of 5pase_INPP5E in DCN neurons expressing Cilia-CIBN-GFP and mCh-CRY2–5pase_INPP5E. Data shown are mean±SEM (n=24 neurons). *P<0.05 (Student’s t-test). (O-P) Changes in fluorescence intensity of PIP3 reporters, RFP-PH_AKT (O) and mCh-PH_GRPI (P), in axonal growth cone (white arrow) after ciliary activation of PI3-kinase. (Q, R) Dynamics and kymograph of mCh-PH_GRPI intensity in primary cilia (Q) and axonal growth cone (R) after ciliary activation of AKT (n=16 cells). t’ in panels D, H, M, O, P, R includes 20 seconds to move from growth cone (pre-light activation) to primary cilia, blue light photo-activation of primary cilia (3 mins.) and 20–23 seconds to move back to growth cone (post-light activation). Scale bar, 1.5μm (B, C, G, K, O, P, R); 2.5μm (D, H, M). See also SFig. 7.

We first examined if activating ciliary-PI3 kinase and generating PIP3 at the ciliary membrane will modulate axonal behavior. To do this, we generated cilia targeted CIBN (Cilia-CIBN-GFP), and co-expressed it with CRY2-iSH2, in which the inter-SH2 (iSH2) region of the p85α regulatory subunit of class I PI3-kinases is fused with CRY2, in DCN neurons (Idevall-Hagren et al., 2012). iSH2 constitutively binds to the endogenous catalytic p100α-subunit of PI3-kinase and with blue light stimulation the subunits are recruited along with iSH2 domain to the ciliary membrane (Fig.6A). In dark condition, CRY2-iSH2 is mostly enriched at the cytosol (Fig.6B). However, upon a train of 200ms blue laser (488nm) illumination solely focused on primary cilium, CRY2-iSH2 is rapidly recruited into the Cilia-CIBN-GFP expressing cilium (t_1/2=9±2.5s, n=16 cells) (Fig.6B). Co-expression the PIP3 reporter mCh-PH_GRPI with CRY2-iSH2 in these experiments allows simultaneous photo-activation and monitoring of PIP3 production. An increase of PIP3 within cilia was observed (t_1/2=93±14s, n=16 cells) following ciliary photo-activation (Fig.6C), confirming that ciliary recruitment of PI3 kinase is able to produce PIP3 in the ciliary membrane. We then analyzed the effect of ciliary PI3 kinase activation and PIP3 generation on axonal growth cones. The axonal growth cone dynamics were monitored before and after ciliary PI3 kinase activation in control DCN neurons. Before photo-activation, growth cones display characteristic lamellipodial webbing and short filopodial extensions (Fig.6D). Following blue light activation, growth cones changed their morphology and dynamics (24/29 cells), retracting lamellipodia while extending numerous aberrantly longer filopodial like protrusions, within 30 seconds of blue light illumination (Fig.6D–E). In contrast to the short-lived filopodial protrusions before photo-activation with a half-life time t_1/2=24 ±1.4s (n=24 cells), the filopodial protrusions formed after photo-activation are highly stable over the imaging duration (20 minutes). We confirmed that these axonal changes are not due to photo-toxicity induced by the photo-activation, since no changes in PIP3 level or growth cone dynamics were detected upon blue light illumination in neurons expressing only Cilia-GFP/CRY2-iSH2 and no Cilia-CIBN (SFig.7A–B). Further, ciliary photo-activation with non-blue light, or blue light illumination at comparably sized, non-ciliary regions in the neuronal soma did not change growth cone dynamics in neurons expressing Cilia-CIBN/CRY2-iSH2 or CIBN/CRY2-iSH2, respectively (SFig.7C–F). These results suggest that activation of ciliary PI3 kinase and the resultant production of ciliary PIP3 can rapidly induce changes in neuronal growth cone morphology and dynamics.

PI3-kinase activation and PIP3 production triggers AKT activation. To test if ciliary activation of AKT can induce similar changes in axonal dynamics, the kinase domain of AKT fused with CRY2 (mCh-CRY2-AKT; Katsura et al., 2015) was co-expressed with Cilia-CIBN-GFP in DCN neurons to enable blue light-induced AKT recruitment and activation at the ciliary membrane (Fig.6F). Similar to PI3-kinase activation, mCh-CRY2-AKT showed rapid recruitment within the cilia upon ciliary blue light illumination (t_1/2=31±5.4s, n=16 cells) (Fig.6G), followed by axonal growth cone changes, including reduced growth cone area and extension of filopodial protrusions (Fig.6H–I).

In contrast to the enhancement of ciliary PIP3 production in the above experiments, we also tested the effect of ciliary recruitment of 5ptase_INPP5E to reduce ciliary PIP3 production (Fig.6J). mCh-CRY2–5ptase_INPP5E, containing the 5-Ptase catalytic module of Inpp5e necessary to hydrolyze PIP3 and PIP2, fused with CRY2, was co-expressed with Cilia-CIBN-GFP (Fig.6K). Ciliary blue light activation of these neurons induced a reduction of ciliary PIP3 (Fig.6L) (t_1/2=21±3.4s, n=16 cells), which is followed by retraction of filopodia and expansion of lamellipodia webbing (n=24 cells) (Fig. 6M, N). Ciliary activation of iSH2, AKT, or Inpp5e did not significantly change cilia length (Cilia length: iSH2_[before]=2.81±0.1μm, iSH2_[after]=2.78±0.1μm; AKT_[before]=2.93±0.2μm, AKT_[after] =2.85±0.2μm; Inpp5e_[before]=2.87±0.3μm, Inpp5e_[after]=2.88±0.1μm; Student’s t-test, p>0.05, n=16 cells/group). Together, these optogenetic manipulations of the ciliary-PI3K signaling reveal that this pathway has the capacity to rapidly and bi-directionally modulate axonal growth cone dynamics.

What are the signaling mechanisms that enable locally activated, ciliary-PI3K signaling to spread to axons? Recent studies show that AKT activation is sustained by a positive feedback loop mediated by PI3K activation and actin polymerization (Katsura et al., 2015). Feedback networks involving kinase-dependent cascades (e.g. PI3K and MAPK signaling) are known to self-perpetuate and rapidly spread local signaling over long distances along the plasma membrane to reach distant cellular targets (Fivaz et al., 2008; Katsura et al., 2015; Kholodenko, 2006; Sawano et al., 2002; Welf et al., 2012). To monitor the axonal changes in PIP3, post ciliary-PI3 kinase activation, axonal PIP3 reporters (RFP-PH_AKT or mCh-PH_GRPI) were imaged before and after ciliary activation of PI3-kinase. Both RFP-PH_AKT and mCh-PH_GRPI showed a gradual increase in the growth cones post activation, indicating lateral propagation of PI3K signaling (Fig.6O–P). Further, we noticed the ciliary activation of AKT is followed by a gradual increase of PIP3 in primary cilia and growth cones (Fig.6Q–R). These data demonstrate that activated PI3K-AKT signaling in cilia can spread signal to axonal growth cones. Consistent with a role for positive feedback loops mediated by PI3K activation and actin polymerization in the propagation of such signaling, inhibition of actin polymerization with low dosage of Latrunculin B known to block AKT-PI3K feedback loop or inhibition of PI3-kinase activity with LY294002 abolished the increase of PIP3 at the growth cone and growth cone morphological changes induced by ciliary PI3K/AKT activation (SFig.7G–H).

Ciliary signaling regulated transcriptional landscape during axonal development

In addition to inducing rapid, local changes on axonal behavior, ciliary signaling may also impact transcriptional programs necessary for axonal development (Green and Mykytyn, 2014; McIntyre et al., 2012; Mukhopadhyay and Rohatgi, 2014; Wang et al., 2015). To gain further insights into the cilia signaling regulated transcriptional landscape during neuronal growth and connectivity, we performed RNAseq analysis of cortices dissected from E12.5 Arl13b^−/− and littermate control embryos. This analysis identified 135 differentially expressed genes, 130 upregulated and 5 downregulated, in Arl13b^−/− (SFig.8A). Gene Ontology (GO) analysis suggest that Arl13b deficiency leads to changes in the transcriptional landscape that converge onto biological processes including neural development/differentiation, cell adhesion, cytoskeleton organization, signal transduction, and metabolism (SFig.8B). Moreover, in support of our observation that PI3K/AKT signaling is a central driver of cilia signaling mediated changes in axonal development, IPA (Ingenuity pathway analysis) gene-gene interaction network analyses suggest that AKT is at the hub, inter-linking differentially expressed genes in Arl13b mutants (SFig.8C). This analysis of global transcriptional changes following Arl13b deletion further suggests a role for aberrantly activated PI3K/AKT and related signaling in the manifestation of axonal developmental changes in cilia mutants.

Discussion

Axonal tract defects are a common feature in many human ciliopathies, with JSRD exhibiting the most consistent and distinct axonal malformations such as MTS and abnormal CST organization. Here, using two, projection neuron-specific, JSRD mouse models, and cilia-specific, optogenetic and chemogenetic signaling approaches, we demonstrate that primary cilia driven signaling regulates proper growth cone dynamics and axonal development. Collectively, these results indicate that ciliary receptor activation and the resultant changes in convergent signaling cascades (e.g. PI3K, AKT, AC3) can effectively modulate of axonal behavior. These studies reveal a new cellular mechanism vital for the formation of neuronal connectivity in the mammalian brain and provide new insights into JSRD pathology.

Axonal defects related to JSRD

During the formation of the connectome in the brain, axons are guided by tropic and trophic gradients of chemoattractant or chemorepellent cues toward their target. Axons travel in large tracts or fascicles and follow stereotyped pathways to reach synaptic targets. Group extension of axons towards appropriate targets is necessary for neuronal wiring in the brain (Fame et al., 2011). Our observations show that while the axonogenesis or extension of axons are not affected in cilia mutants, the branching, fasciculation, crossing of axons and thus their accurate final projections are disrupted in cilia mutants.

Both a motile growth cone at the tip of the extending axon that properly responds to guidance cues and axon branching that enables axon sorting and fasciculation are crucial to ensure the fidelity of axonal tract development (Cioni et al., 2018; Nishikimi et al., 2013; Raper and Mason, 2010; Wang and Marquardt, 2013; Yu and Bargmann, 2001). Growth cone motility relies on the interplay between filopodia that sense environmental cues for pathfinding, and lamellipodia that support persistent axonal outgrowth (Mattila and Lappalainen, 2008; Mejillano et al., 2004; Vitriol and Zheng, 2012). Filopodia are necessary to initiate axon branches (Dent, 2004; Winkle et al., 2016). Branching along the axonal shafts enable projecting axons to connect with additional targets en route. Once reaching their appropriate final target, axons establish connections by extending terminal axon arbors (Gibson and Ma, 2011; Kalil and Dent, 2014). We found that cilia mutant neurons show altered filopodia vs. lamellipodia balance and disrupted branching. Arl13b or Inpp5e-deficient neurons grow a long primary axon with significantly reduced lateral axonal branches and terminal arbors at targets, thus failing to sufficiently innervate their targets to form appropriate neuronal connections. These observations suggest that disruption of primary cilia mediated modulation of axonal growth cone and branching dynamics contributes to the axon-axon interaction and axonal path finding defects in cilia mutants. However, the nature of ciliary modulation of axonal response to guidance cues such as Netrin, Slit, Semaphorin, Ephrin, and Wnt during axonal pathfinding remains to be fully elucidated.

Patients with every ARL13B and INPP5E mutation suffer from intellectual disabilities and psychomotor delay. ARL13B^G75R mutant patients also suffer from epilepsy. While these neurological phenotypes clearly suggest circuit malfunction, apart from major axonal pathway defects detectable in MRI, other deficits such as neuronal dendritic disruptions are yet to be defined in genotyped JSRD patient brain tissue. Future studies focused on examining how these disease-causing mutations affect the interactions between ARL1B and INPP5E, as well as the interactome of ARL13B would be essential to further dissect the Arl13b-Inpp5e driven signaling mechanisms underlying JSRD.

Primary cilium as a signaling node during axonal development

Primary cilium, with a volume ~ 1/1000 of the soma, possesses diverse signaling machineries, including RTKs (e.g. IGFR, EGFR, PDGFR-α, Trk receptors p75^NTR and TrkB) and GPCRs (e.g. Smo, Frizzled receptor FZD3, Sstr3, 5HT6, NPY2R, NPY5R, D2, MchR1, GPR161), their downstream effectors (e.g. PI3 kinase, Inpp5b, Inpp5e, Ocrl, AKT, CaMKIIβ, AC3, 5 and 6), and second messengers (e.g. PIP3, PIP2, Ca²⁺ and cAMP), thus providing a unique environment where signaling components are highly concentrated in a small cellular domain to facilitate efficient signaling activation and crosstalk (Arellano et al., 2012; Armato et al., 2011; Choi et al., 2011; Christensen et al., 2012; Conduit et al., 2012; DeCaen et al., 2013; Delling et al., 2013; Franco et al., 2014; Green and Mykytyn, 2014; Guo et al., 2015; Hilgendorf et al., 2016; Jacoby et al., 2009; Mukhopadhyay and Rohatgi, 2014; Nichols et al., 2013; Omori et al., 2015; Phua et al., 2015; Puram et al., 2011; Schou et al., 2015; Zhu et al., 2009). However, due to the lack of methods to exert precise control of cilia-specific signaling events, the exact cellular and physiological impact of the signaling emanating from primary cilia on the developing neurons and neural circuit formation remained poorly understood. To overcome this limitation, we employed a combination of mouse genetics, chemogenetic and optogenetic approaches to enable the interrogation of ciliary signaling in the developing neurons. Collectively, our observations suggest that altered ciliary PI3K/AKT signaling homeostasis underlies the aberrant axonal growth cone dynamics in cilia mutant neurons and that signals emanating from cilia have the capacity to remotely regulate axonal growth cone behavior changes over long distance, via a yet-to-be defined modes of propagation.

Propagation of ciliary signaling in neurons

Deficiency of Inpp5e, the main inositol polyphosphate-5-phosphatase within cilia, disrupts the ciliary membrane phosphoinositide (PIP/PIP2/PIP3) composition and results in dysregulation of Shh and PI3K/AKT signaling (Chavez et al., 2015; Conduit et al., 2012; Dyson et al., 2017; Eramo and Mitchell, 2016; Garcia-Gonzalo et al., 2015). Our observations show that Inpp5e deficiency caused by Arl13b or Inpp5e deletion leads to globally elevated PIP3 level and PI3K/AKT activity in developing neurons, supporting the notion that dysregulated ciliary signaling can propagate throughout the whole neuronal cell body to manifest pathogenic changes in normal state.

Once activated, PI3K/AKT signaling employs downstream kinase cascades and second messengers (e.g. Ca²⁺, cAMP, IP3, PIP3) to amplify and propagate signals for cell-wide, long-distance signaling (Civelli, 2012; Fruman et al., 2017; Gavi et al., 2006; Lemmon and Schlessinger, 2010). We show that ciliary GPCR signaling induces axonal Ca²⁺ wave down to the growth cone. Previous studies suggest that Shh-Smo signaling, possibly via primary cilia, act to down regulate cAMP levels and protein kinase A activity in commissural neurons to allow Sema3 mediated repulsion of axons at midline crossing (Parra and Zou, 2010). The exact mechanisms that enable ciliary signaling to trigger axonal Ca²⁺ wave and its integration with axonal PI3K/AKT signaling waves are yet to be fully elucidated. Nevertheless, our observations indicate that ciliary signaling can engage multiple convergent pathways to enable signal propagation across the neuronal cell body.

Primary cilia signaling is likely to influence neuronal behavior via both rapid local signaling cascade changes as well as by triggering changes in transcriptional programs underlying neuronal growth and connectivity. Considering the time window it takes for long projecting axons to grow, navigate, and connect with their distant targets (i.e., several days), primary cilia signaling driven local changes may facilitate rapid fine tuning of axonal behavior, whereas transcriptional program changes may facilitate permissive conditions for axon growth, navigation, and target recognition. PI3/AKT signaling, in addition to locally changing axonal behavior, is also known to directly regulate transcriptional networks important for axonal development (Chen et al., 2014; Graef et al., 2003; Kwon et al., 2006; Nguyen and Di Giovanni, 2008; Manning and Toker, 2017; Sánchez-Alegría et al., 2018). It is thus conceivable that PI3K/AKT signaling serves as a converging node to integrate ciliary receptor signaling necessary for both rapid axonal tract growth modulation and long-term neuronal developmental program regulation.

This study points to the tantalizing possibility that neuronal primary cilia may customize a “ciliary signaling signature” that elicits cilia-specific impact on neuronal events such as calcium waves, cytoskeletal rearrangements, adhesion, and transcriptional regulation necessary for proper axonal tract development and wiring. In spite of not being required for extension of axons, cilia appear to exert a crucial modulatory role in appropriate axonal wiring. Cilia signaling, from locales where the neuronal soma are, may enable fine tuning of projection neuron axonal behavior, long distances away, necessary to establish the final axonal connectivity. Cilia may thus serve as modulators that enable coordination of signaling between neuronal origin and target (e.g., DCN and thalamus or cortical projection neuron and spinal cord) necessary to establish precise patterns of connectivity. Consistent with this, our observations show that neuronal cilia regulate axonal growth cones, at the tip of axonal projections, whose functions are necessary for precise wiring of the neural circuitry. Future efforts to decipher the communication networks between neuronal cilia and their cellular environment or between axons and cilia, in the context of brain development, will facilitate a holistic view of how primary cilia receive and convey environmental signals to modulate brain formation and organization. Nevertheless, our studies delineate the significance of cilia driven signaling in axonal tract development and connectivity in the developing brain, thus defining a hitherto undefined cell biological mechanism fundamental for neuronal circuit formation in the brain and whose dysregulation may underlie axonal tract defects in Joubert Syndrome Related Disorders.

STAR Methods

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to, and will be fulfilled by the lead contact, E. S. Anton (anton@med.unc.edu).

Experimental Model and Subject Details

Mice

This study is based on data from mice at various developmental stages (embryonic days 16, 18, postnatal days 1, 7, 30, and 60) and includes both males and females. Mice were cared for according to animal protocols approved by the University of North Carolina. Arl13b or Inpp5e was conditionally inactivated in cortical projection neurons and deep cerebellar nuclei by crossing Arl13b^Lox/Lox mice (Arl13b^tm1Tc, MGI: 4948239) (Su et al., 2012) or Inpp5e^Lox/Lox (Inpp5e^tm1Ssch, MGI:4360185) (Jacoby et al., 2009) with Nex-Cre mice (Goebbels et al., 2006). Further, transgenic mice carrying the Tau-Lox-STOP-Lox-GFP cassette (Hippenmeyer et al., 2005) or Ai9 reporter (Madisen et al., 2010) were bred into the Arl13b^Lox/+; Nex-Cre and Inpp5e^Lox/+; Nex-Cre lines. Cre recombination of the Tau-Lox-STOP-Lox-mGFP or Ai9 transgene leads to GFP or mRFP expression, respectively, in the processes of all Cre-positive neurons. Arl13b^Lox/+; Nex-Cre or Inpp5e^Lox/+; Nex-Cre littermates served as controls. Arl13b null mice (Arl13b^−/− or Arl13b^hnn; MGI:3578151), Ift88 null mice (Ift88^tm1.1Bky, MGI: 3710186), Inpp5e^M2 mice (MGI: 5296378) were generated as described in Caspary et al., 2007, Haycraft et al., 2007, and Su et al., 2012, respectively. All lines were genotyped as previously described (Caspary et al., 2007; Su et al., 2012; Higginbotham et al., 2013). Location of different anatomical regions of the mouse brain were annotated using the mouse brain atlas (Paxinos and Franklin, 2001).

DNA constructs

CIBN-GFP, CRY2-iSH2, mCh-CRY2-iSH2, mCherry-CRY2–5Ptase_INPP5E, CRY2–5Ptase_INPP5E, RFP-PH_AKT, and mCh-PH_GRPI (Idevall-Hagren et al., 2012) are gifts from Dr. Pietro De Camilli, Yale University. The cilia targeted CIBN-GFP was generated by replacing R-GECO1.0 in IA-R-GECO1.0 with CIBN-GFP. IA-R-GECO1.0 (Su et al., 2013) is a gift from Dr. Takanari Inoue, John Hopkins University. CRY2-venus-AKT1 (Katsura et al., 2015) is a gift from Dr. Takeaki Ozawa (the University of Tokyo). mCh-CRY2-AKT1 was generated by replacing 5Ptase_INPP5E with AKT1 in mCherry-CRY2–5Ptase_INPP5E. The cilia-targeted bPAC (cilia-bPAC) (Jansen et al., 2015) is a gift from Dr. Dagmar Wachten (Center of Advanced European Studies and Research, Bonn, Germany). pAAV-DIO-FLAG-GPR88* (Siljee et al, 2018) is a generous gift from Dr. Christian Vaisse (The University of California, San Francisco). Pcdh17-GFP (Hayashi et al., 2014) is a generous gift from Dr. Masatoshi Takeichi (RIKEN Center for Developmental Biology). The Cre-inducible adeno-associated virus (AAV) vectors expressing ARL13B variants and mCherry were generated as described previously (Guo et al., 2017). The AAV5-Synapsin-tdTomato virus (Klapoetke et al., 2014) was obtained from UNC viral vector core.

CLARITY imaging

CLARITY experiments were performed based on protocols available at: http://wiki.claritytechniques.org/index.php/Solutions. Briefly, Arl13b^Lox/lox-NexCre-Ai9 and control pups (P1) were perfused with ice cold 4% PFA. Brains with spinal cords attached were dissected and post fixed in 4% PFA overnight at 4 °C, followed by incubation in 30ml hydrogel solution for 2 days at 4 °C. Following polymerization of the hydrogel monomers, samples were incubated with clearing solution for 2 weeks at 37 °C. Cleared samples were washed with PBST (0.1% TritonX in 1X PBS) buffer for 2 days and stored in PBST at 4 °C. For CLARITY imaging, samples were incubated with 85% glycerol for 1h, placed in a glass bottom FluoroDIsh chambers (World Precision Instruments, Inc.) filled with 85% glycerol and imaged using a Zeiss780 confocal microscope equipped with 10x objectives.

Microfluidic chamber assay

Microfluidic chambers (Higginbotham et al., 2012) were attached to laminin coated glass coverslips. 5μl containing 200K cells from DCN of P3 Arl13b^Lox/+; NexCre or Arl13b^Lox/Lox; NexCre brains electroporated with BFP (Addgene #127348) plasmids were seeded onto one side of the microfluidic chamber. 24 hours later, the same number of cells from P3 Arl13b^Lox/+; NexCre; Ai9 or Arl13b^Lox/Lox; NexCre; Ai9 brains electroporated with Pcdh17-EGFP were seeded onto the opposite chamber. Cells were maintained at 37°C/5% CO2 and 10μl of new media (DMEM/N2/B27) was added every 24 hours. After 48 hours, cells were fixed in 4% paraformaldehyde and analyzed for the expression pattern of Pcdh17-EGFP at BFP⁺/tdTom⁺ axon-axon contacts.

In utero electroporation

Lateral ventricles of Arl13b^Lox/+ and Arl13b^Lox/Lox, or Inpp5e^Lox/+ and Inpp5e^Lox/Lox (E15.5) embryos were electroporated as described previously (Guo et al., 2015; Higginbotham et al., 2012). Briefly, 1μl of CAG-Cre (Addgene #13775) and CAG-GFP (Addgene #16664) plasmid DNA (2 μg/μl) was injected into the lateral ventricles of E15.5 brains and electroporated using 5 pulses at 30 V for 50 ms at 950 ms intervals through the uterine wall using a BTX ElectroSquarePorator (ECM 830). Embryos were then allowed to develop in vivo and analyzed at P1 or P21. Plasmids used for in utero electroporation were prepared using the EndoFree Plasmid kit (Qiagen).

Primary Neuronal Culture and transfection

Deep cerebellar nuclei (P2) or dorsal cortices (E15) were dissected in Hank’s buffered salt solution (HBSS), dissociated in HBSS containing papain (Worthington) and DNase I (100 mg/ml, Sigma) for 10 min at 37C, washed in HBSS three times and manually triturated in Neural basal medium supplemented with DNase I. Neurons from E11.5 or 12.5 cortices were similarly processed. Neurons were transfected using Amaxa™Nuleofector™ II kit (Lonza) or Invitrogen Lipofectamine 2000 reagent according to the manufacturer’s protocol. 1.0 × 10⁴ cells were plated per 35 mm glass bottom dish (MatTek) coated with poly-L-lysine /Laminin (Sigma) and cultured for 1–4 days in Neurobasal-A /5% B27/1% GlutaMax/1% P/S media.

Chemogenetic manipulation of ciliary signaling and axonal imaging

Cultured DCN neurons were transfected with R-GECO1.0 and Cilia-hM3D_q and imaged at an acquisition rate of 0.5 Hz using a Zeiss 780 live cell confocal microscope. The effects of CNO or DMSO on axonal Ca²⁺ and growth cone behavior were assessed in neurons expressing Ciliary hM3D_q by recording upto 20min before and 20min after addition. GECO1.0 fluorescence intensity values were normalized against background and the average baseline values were measured prior to CNO or DMSO addition. Image analysis was performed using Zeiss LSM Image Browser or ImageJ software.

Optogenetic manipulation of ciliary-PI3K/AKT, ciliary-Adenylyl cyclase signaling and axonal imaging

For ciliary-PI3K activation, cultured DCN neurons were transfected with Cilia-GFP-CIBN, mCh-CRY2-iSH2, or CRY2-iSH2. For ciliary-AKT activation, DCN neurons were transfected with Cilia-GFP-CIBN, mCh-CRY2-AKT1, or CRY2-venus-AKT1. For ciliary-Inpp5e_5ptase activation, DCN neurons were transfected with Cilia-GFP-CIBN and mCh-CRY2–5ptase-Inpp5e. RFP-PH_AKT or mCh-PH_GRPI were used as PIP3 sensors. For ciliary-bPAC activation, DCN neurons were transfected with mCh-Sstr3-bPAC and pcDNA3.1-BFP. A train of 200ms blue laser light (488nm) focused on primary cilia for upto 3 minutes was used to induce dimerization between CRY2 and CIBN within primary cilia (Idevall-Hagren et al., 2012). Growth cone morphology and dynamics were imaged with 561nm laser (mCh or mRFP) at an acquisition rate of 0.5 Hz upto 20min before and 20min after photo-activation using a Zeiss 780 live cell confocal microscope. The imaging positions and parameters for primary ciliary light activation and growth cone imaging was setup using the multi-position, time series, and photo bleaching applications in ZEN software.

Immunohistochemistry

Cerebral cortical sections and neurons were immunolabeled as previously described (Guo et al., 2017; Schmid et al., 2003; Weimer et al., 2008; Yokota et al., 2007) with the following primary antibodies: anti-GFP (chicken, 1:1000; Abcam, ab13970), anti-RFP (rabbit, 1:500; Rockland, cat. nr. 600–401-379), anti-ACIII (rabbit polyclonal, 1:100; Santa Cruz, sc-56855), anti-Tau (mouse 1:1000, Millipore, MAB3420), anti-Map2 (chicken polyclonal, 1:1000, Abcam, ab5392), Immunoreactivity was detected by incubation with appropriate AlexaFluor 488 or Cy3-conjugated secondary antibodies (Invitrogen and Jackson ImmunoResearch). AlexaFluor 488, 647 or Cy3-conjugated (Invitrogen and Jackson ImmunoResearch) secondary antibodies were used. Nuclei were counterstained with DAPI (Sigma).

Tissue Culture and transfection

Mouse embryonic fibroblast (MEF) cells were derived from wild type and Arl13b^−/− E12.5 embryos as previously described (Higginbotham et al., 2012). MEF cells and IMCD3 (authenticated by American Type Culture Collection [ATCC]) were grown in DMEM/ F12 media supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S) at 37°C/5% CO₂. Cells were allowed to reach 80% confluency and ciliogenesis was induced by serum starvation for 24 hours. Cells were then transfected using Invitrogen Lipofectamine 2000 reagent according to manufacture’s instructions.

Immunoblotting

Cortices of Arl13b null and control embryos (E12) or cultured cells were harvested in 200 μl cell lysis buffer (Cell signaling, #9803) with protease and phosphatase inhibitor cocktail (Thermo Scientific). The lysates were briefly sonicated and centrifuged at 14,000 rpm for 20 min at 4°C. Protein concentrations were determined using the protein assay reagent (Bio-Rad). Samples were mixed with 3X SDS sample buffer and boiled for 5 min. Samples were then resolved in SDS-polyacrylamide gels, transferred to nitrocellulose, and incubated with following primary antibodies: rabbit anti-pAKT [Thr308] (1:1000, Cell signaling, #9275), rabbit anti-pGSK3β (1:1000, Cell signaling, #9331), rabbit anti-pS6 (1:1000, Cell signaling, #4858), rabbit anti-AKT (1:1000, Cell signaling, #9272), mouse anti-β-actin (1:1000, Abcam, ab8226). Immunoblots were developed using horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG, followed by detection with enhanced chemiluminescence.

RNA-seq

RNA was isolated from E12.5 control and Arl13b^−/− cortices with the RNeasy plus mini kit (Cat. #74134, Qiagen). RNA yield and quality was determined with a Nanodrop 1000 Spectrophotometer (Thermo Scientific). RNA quality was further assessed by an Agilent Bioanalyzer 2100 to obtain a RNA integrity number (RIN). RIN values exceeding 8 were used for sequencing. RNA samples were used to generate and barcode cDNA libraries using the Nugen ovation RNAseq Plus Kit at the UNC High Throughput Sequencing Facility. Pool of 6 samples were sequenced in a single lane in a HiSeq 4000 sequencer (Illumina) using 50 bp paired-end reads.

Image analysis

All images were analyzed using ImageJ (NIH) and Zeiss Zen. Kymographs and fluorescence intensity of Ca²⁺ sensors RFP-GECO1.0 or IA-GFP-GECO1.0, and PIP3 sensors RFP-PH_AKT or mCh-PH_GRPI were quantified using ImageJ (Guo et al., 2017). ROIs located in cilia, soma, or growth cones of neurons were selected and the mean value of fluorescence intensity of ROIs (after background subtraction) over time was plot profiled. For the turning angle measurement of SCP, lines were drawn along the SCP tract pre and post turning and the turning angle was quantified using the angle tool in Image J.

RNA-seq data analysis

Reads were filtered for quality using the FASTX toolkit’s fastq_quality_filter (http://hannonlab.cshl.edu/fastx_toolkit/index.html) with 90% of the bases having a minimal quality score of 20. Sequencing adaptors were trimmed using cutadapt. Reads were aligned to the mm9 reference genome using STAR. Transcript abundance was estimated using Salmon (Patro et al., 2017). DESeq2 was used to normalize read depth and calculate differential gene expression. Genes with a Benjamini-Hochberg (B&H) FDR adjusted p-value < 0.05, fold change > 2 or <−2 were considered significant. GO term annotation of gene lists was performed using ToppGene. GO terms with a B&H FDR adjusted p-value < 0.05 were considered significant. Lists of significant GO terms were clustered using REVIGO. Ingenuity Pathway Analysis (IPA) [https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis] was used for pathway analysis, novel network identification, and upstream regulator analysis.

Quantification and Statistical Analysis

General statistical analysis

GraphPad or Excel was used for data analysis. Two-tailed Student’s t test and ANOVA were performed using GraphPad. No a priori power analyses were performed, but sample sizes were similar to those described in previous related studies (Guo et al., 2015; Guo et al., 2017; Higginbotham et al., 2012, 2013). All experiments were independently repeated for at least 3 or more times. All data are expressed as means ± standard error of the mean (SEM). Statistical details, including p values, are indicated in text or figure legends.

Data and Code Availability

The published article includes all datasets generated or analyzed during this study. RNA-seq data were deposited in the GEO database (GSE120310).

Highlights.

• Chemogenetic and or optogenetic activation of primary cilia alters axonal behavior.

• Ciliary activity modulates axonal growth cones and filopodial/lamellipodial balance.

• Arl13b-Inpp5e activity in cilia facilitates axonal tract formation and targeting.

• Disrupted ciliary signaling contributes to axonal tract malformations in JSRD.