Deletion of PIK3C3/Vps34 in sensory neurons causes rapid neurodegeneration by disrupting the endosomal but not the autophagic pathway

Generation of Mice with Sensory Neuron-Specific Deletion of Pik3c3 Gene.

Initial characterization revealed that Pik3c3 is generally expressed at low levels in most cells and moderately expressed in sensory neurons (Fig. S1 A and B). To study the in vivo function of Pik3c3, we generated a Pik3c3 conditional null mutant allele in which the ATP binding domain of the kinase is flanked by LoxP sites (Pik3c3^flox/flox mice) and thus, can be deleted in the presence of Cre recombinase (Fig. 1A and Fig. S1). The resulting mutant gene, if expressed, would encode a truncated and functionally inactive protein. To selectively delete Pik3c3 in sensory neurons, we generated Advillin^Cre/+ knockin mice to express Cre specifically in these neurons (the Cre line will be described elsewhere) (27). [Advillin^Cre/+; Pik3c3^flox/flox] mice (hereafter designated as Pik3c3-cKO) appear normal at birth. However, around postnatal day 5–6 (P5–P6), Pik3c3-cKO mice begin to drag their hindlimbs and develop increasing difficulties in coordinating movement and maintaining body postures (Fig. 1A). All mutant mice die within 2 wk of age.

Using immunofluorescence against the N-terminal domain of PIK3C3, which would detect the truncated protein if it is translated, and fluorescent in situ hybridization to probe Pik3c3 expression, we found that PIK3C3 protein as well as mRNA diminished from all dorsal root ganglion (DRG) sensory neurons in Pik3c3-cKO mice at birth (Fig. 1B). Deletion of PIK3C3 did not affect the level of its binding partner Beclin1 (Fig. S1G). We also examined the levels of PI3P, the product of PIK3C3 in these neurons. PI3P recruits Fab-1/YOBT/Vac1/EEA1 (FYVE) domain-containing proteins to vesicular compartments and thus, can be visualized by transfecting sensory neurons with a mVenus-2xFYVE fusion construct (mVenus is a monomeric bright variant of GFP). In control neurons, mVenus-2xFYVE labeled punctate structures, presumably endosomes, residing in the cell body and along the axons (Fig. 1C Left), whereas in Pik3c3-deleted neurons, mVenus-2xFYVE signals were weak and diffuse (Fig. 1C Right) (n > 20), suggesting that PI3P was depleted and PIK3C3 is the main enzyme producing PI3P in sensory neurons.

Differential Degeneration of Small- and Large-Diameter Pik3c3-Deficient Sensory Neurons.

We next assessed the effect of Pik3c3 deficiency on neuronal survival and apoptosis. At birth, although Pik3c3 has been deleted in all sensory neurons, we found no differences in the number of neurons or apoptotic cells between mutant and control mice (Fig. S2D). At P5, apoptosis increased, and about 27% of sensory neurons were lost in the mutant (Fig. S2 C and D). By P9, cell death continued, and only 44% of neurons remained in the mutant (Fig. S2 C and D). These results indicate that PIK3C3 deficiency causes a progressive, yet rapid degeneration of mature sensory neurons.

Interestingly, using molecular markers for different types of sensory neurons, we found that the numbers of TrkB- or TrkC-positive large-diameter touch or proprioceptive neurons were significantly reduced at P5/P6 and were further depleted at P9 in Pik3c3-cKO mice (Fig. 2 A and E). By contrast, the number of TrkA-expressing small-diameter neurons had only a statistically insignificant reduction at P9, and cRet-positive neurons (including both small- and large-diameter neurons) had a moderate decrease in mutant mice at this age (Fig. 2 A and E). Notably, the expression of all these receptors in newborn (P0) Pik3c3-cKO mice was indistinguishable from the expression in controls (Fig. S2A), indicating that Pik3c3-deleted TrkB/C-positive large neurons degenerated much faster than TrkA/cRet-expressing small neurons.

We also examined the axonal integrity of mutant sensory neurons. Anti-calcitonin gene-related peptide (CGRP) and isolectin B4 (IB4) (28) staining revealed that axonal projections from small-diameter neurons to the dorsal horn appeared intact in Pik3c3-cKO mice at P9 (Fig. 2B), whereas anti-vGluT1 staining (29) showed a dramatic reduction of the large-diameter axon termini in the mutant mice (Fig. 2B, four arrows). To further confirm these antibody staining results, we crossed the conditional mutant mice with a Cre-reporter line in which human placenta alkaline phosphatase (hPLAP) can be induced by Cre recombinase to label axons (30, 31). PLAP staining revealed that axonal projections from all sensory neurons develop normally in Pik3c3-cKO mice and show a wild-type pattern at birth (Fig. S2B). However, in P9 dying mutants, the proprioceptive sensory axon projections to the ventral spinal cord had almost completely degenerated (Fig. 2C, arrowheads). Touch-sensory innervations in layers III/IV were also significantly reduced (Fig. 2C, arrows). In contrast, no apparent changes were observed in the small-diameter innervations to the superficial layers (I/II) of the spinal cord in the same animals (Fig. 2C, block arrows). Electron microscopic analyses of sciatic nerves from P9 mutant mice also showed numerous degenerative pathological changes in large-diameter myelinated axons (Fig. 2D Lower, arrows) but mostly intact and normal appearance of small-diameter unmyelinated axons (Fig. 2D Upper).

This differential degeneration of large- versus small-diameter neurons was not a result of differential expression or deletion of the Pik3c3 gene. All types of sensory neurons express comparable levels of PIK3C3 in wild-type animals (Figs. S3 and S4B). In Pik3c3-cKO, the loss of PIK3C3 protein and mRNA happens equivalently in all DRG neurons (Figs. S3 and S4B). Furthermore, mVenus-FYVE labeling showed a similar diffuse cytoplasmic pattern in both small- and large-diameter mutant neurons (Fig. S4A). Note that expression of mVenus-2xFYVE did not alter the general morphology of sensory neurons (Fig. S4C).

Progressive and Differential Accumulation of Ubiquitin-Positive Aggregates in PIK3C3-Deficient Sensory Neurons.

Antiubiquitin (Ub) staining revealed evenly diffuse signals in control neurons. In contrast, brightly stained Ub-positive aggregates/punctae were found in subsets of Pik3c3-deficient sensory neurons at P6 (Fig. S5A). About 69.6% of TrkB- and 53.2% of TrkC-expressing neurons had numerous punctate Ub-positive aggregates in their cell bodies (Fig. S5B). In contrast, only 13.2% of TrkA- and 11.9% of cRet-expressing neurons contained Ub-aggregates at this stage (Fig. S5B). However, small neurons eventually did accumulate Ub-aggregates at P9 (48.2% of TrkA- and 45.8% of cRet-expressing neurons contained Ub-aggregates at this stage). These data suggest that TrkB/C expressing large-diameter neurons accumulate ubiquitinated aggregates faster and/or earlier in response to Pik3c3 deletion than small-diameter neurons, which may be one cause of their faster degenerations.

Two Distinct Ultrastructural Phenotypes Caused by Pik3c3 Deletion.

Previous cell-culture studies showed that blocking the function of PIK3C3 (using drug or antibody) resulted in the accumulation of abnormal or enlarged endosomes (32, 33). Using anti-Rab5 and anti-Rab7 staining to visualize endosomes, we found that the number of Rab5-positive puntae and Rab7 signal intensity increased significantly in Pik3c3-deficient neurons, which is consistent with previous studies (Fig. S5C). On semithin sections, mutant sensory neurons either contained numerous large vacuoles (Fig. S5C, arrow) or appeared darkly stained by toluidine blue (Fig. S5C). To further investigate these two distinct phenotypes, we performed transmission electron microscopy (EM). Control DRG neurons uniformly showed few vesicles inside the cell body (vesicles occupying only 0.52% of cytosolic area), and on average, they contained 1.65 lysosomes per 100 μm² cytosolic area (Fig. 3A Upper Left and Lower Left). By contrast, one class of mutant neurons (14%) accumulated numerous large vesicles or vacuoles (range = 200 nm to 2 μm in diameter; occupies more than 7.50% of cytosolic area) in their cell bodies and had a small increase in the number of lysosomes (4.55 lysosomes/100 μm²) (Fig. 3A Upper Center and Lower Center). Another class of Pik3c3-mutant sensory neurons (79%) contained fewer numbers of vesicles (2.48% of cytosolic area) but was instead filled with lysosomes (or lysosome-like electron-dense organelles) for a more than 15-fold increase in lysosome numbers (24.8 lysosomes/100 μm²) (Fig. 3A Upper Right and Lower Right). This phenotype resembles the phenotypes observed in cathepsin B and L knockout mice where lysosome-like organelles also accumulated (34). A small numbers of neurons (7%) have intermediate phenotypes.

Lysosomes Are Accumulated in Small- but Not in Large-Diameter PIK3C3-Deficient Sensory Neurons.

The drastic increases in lysosome-like organelles in some Pik3c3-cKO neurons at P9 were further confirmed using immunofluorescence against lysosomal markers lysosome associated membrane protein 2 (LAMP2) (35), and the increase happened between P6 and P9 (Fig. S5 D and E). Because of the fact that LAMP2 immunoactivity was lost after in situ hybridization (using TrkA/B/C and cRet probes), we used antibodies to label different types of sensory neurons and performed two-color immunofluorescence. We found that NF200- (a marker for all large-diameter myelinated neurons) (36) or parvalbumin- (a marker for TrkC-expressing sensory neurons) (37) positive Pik3c3-mutant neurons all expressed low levels of LAMP2 (Fig. 3B). By contrast, the majority of CGRP-positive (they correspond mostly to TrkA-expressing small-diameter neurons) Pik3c3-deficient neurons showed intense anti-LAMP2 staining at P9 (Fig. 3B). These results strongly supported the idea that PIK3C3-deficient large-diameter myelinated neurons are vacuole-filled and rapidly degenerating, whereas mutant small-diameter unmyelinated neurons are lysosomes-filled and slowly degenerating; vacuoles are more detrimental to neurons than the accumulation of lysosomes.

Lack of Autophagy Is Not the Main Reason Underlying the Rapid Degeneration of Pik3c3-Deficient Sensory Neurons.

PIK3C3 is involved in both autophagy and endosomal pathways, and therefore, we asked which pathway is primarily responsible for the neurodegeneration observed in sensory ganglia. To specifically disrupt autophagy pathways in sensory neurons, we used the same Cre driver (Advillin^Cre/+) to delete Atg7, an autophagy-specific gene that is essential for the extension and completion of autophagosomes, in sensory neurons. We generated [Advillin^Cre/+; Atg7^flox/flox] mice (designated as Atg7-cKO) and found that these mice were born normal. In the first 6 months after birth, homozygous Atg7-cKO mice appeared indistinguishable from wild-type (Advillin^Cre/+) and heterozygous [Advillin^Cre/+; Atg7^flox/+] animals. By contrast, Pik3c3-cKO mice died within 2 wk of birth. At 7 months of age, Atg7-cKO mice began to show an abnormal hindlimb-clasping phenotype and had somewhat reduced mobility (Fig. S6D). Between 7 and 9 months of age, Atg7-cKO mice gradually developed frequent tremors, difficulties in movement, and stiffly twisted tails (Fig. S6D). Below, we describe the analyses of the neuronal degeneration and subcellular phenotypes of Atg7-cKO mice, which were completely distinct from those of Pik3c3-cKO mice.

We first analyzed the expression of markers for different types of sensory neurons in the Atg7-cKO mice. At neonatal stages (P5–P7), the expressions of TrkA, cRet, TrkB, and TrkC were essentially the same between control and Atg7-cKO mice (Fig. S6A). In addition, axonal projections from all types of sensory neurons into the spinal cord in Atg7-cKO mice showed normal patterns (Fig. S6C). These data indicated that the gross development of sensory neurons was not affected by the lack of autophagy. In 9-month-old Atg7-cKO mice with visible behavioral deficits, we observed elevated apoptotic cell death in DRG (Fig. S7A). Importantly, the numbers of sensory neurons expressing any of the four markers (TrkA, cRet, TrkB, and TrkC) were all significantly decreased at this stage in mutants (Fig. 4A and Fig. S6B), suggesting that all sensory neurons had undergone degeneration. This is in contrast to the differential degeneration of TrkB/C- versus TrkA/cRet-positive neurons observed in Pik3c3-cKO mice.

Staining for axon projections into the spinal cord from aged control and Atg7-cKO animals revealed that both CGRP and vGluT1 signals were markedly weaker in mutant mice (Fig. 4B), indicating that axons of both small- and large-diameter neurons had degenerated. Note that IB4 staining in aged Atg7-cKO mice displayed strong ectopic signals in the dorsal medial region of the spinal cord (Fig. 4B). On careful examination, this strong IB4 staining inside the spinal cord significantly colocalized with the astrocyte marker GFAP and partially colocalized with the microglia marker CD11 (to a lesser extent), but it did not coincide with the neuronal markers neurofilament (NFM) or peripherin (Fig. S7B). Together, these data suggest that the ectopic IB4 signals represent activated astrocytes and microglia cells, perhaps appearing in response to the loss of neurons or degeneration of axons.

Between 7 and 9 months of age, strong Ub signals occupying large areas within the cytosol became apparent in all Atg7-deficient sensory neurons but not in control neurons (Fig. S7A). This is consistent with the role of autophagy in preventing the formation of protein aggregates. However, the levels and distribution patterns of lysosomal markers did not change in Atg7-deleted neurons compared with those in controls (Fig. S7A). Again, this is different from the significant increase in LAMP1/2 levels in the Pik3c3-deficient small-diameter neurons.

Finally, EM analysis of DRG neurons from 9-month-old mice showed the presence of very large inclusion bodies in all Atg7-mutant sensory neurons (Fig. 4C Right, stars) but none in control neurons (Fig. 4 C Left and D Left). The inclusion bodies contained both proteinous aggregates and occasionally, small membranous substances (Fig. 4D Center). A subset of Atg7-mutant neurons also contained numerous electron-dense organelles that might be dysfunctional lysosomes or lipofuscin bodies. Some of these dense organelles contained fibril-like materials (Fig. 4D Right), but we do not know the nature of these fibrils at present. Taken together, these findings show that the degeneration of sensory neurons caused by the lack of autophagy is both qualitatively different and phenotypically distinct from that caused by the loss of PIK3C3. Thus, the rapid and differential degenerative phenotypes observed in Pik3c3-cKO neurons most likely resulted from defects in the endosomal/endocytic but not autophagic pathways.

PIK3C3-Independent Autophagosome Formation Pathway in Small-Diameter Sensory Neurons.

Emerging evidence suggests the existence of PIK3C3-independent autophagy initiation pathways (38–40). To examine this possibility, we first crossed Pik3c3-cKO mice with the GFP-LC3 transgenic mice (41). In control DRG neurons, GFP-LC3 gave diffuse cytosolic fluorescent signals, whereas in Pik3c3-cKO neurons, we observed numerous bright GFP-LC3 punctae, suggesting the formation of autophagosomes in mutant sensory neurons (Fig. S8A). Costaining DRG sections with GFAP or CD11 showed that LC3 punctae did not colocalize with either astrocytes or microglial cells in mutant DRG (Fig. S8B). By performing two-color colocalization studies both in vivo and in vitro, we found that GFP-LC3 punctae are present selectively in small-diameter neurons, such as those positive for CGRP, but they are mostly absent from large-diameter NF200 or parvalbumin-positive neurons (Fig. 5A and Fig. S9 A and B), suggesting that this potential noncanonical autophagy initiation process selectively occurred in Pik3c3-deficient small-diameter sensory neurons.

The second method to detect autophagy is Western blot analysis of LC3-II. DRG lysates from P0, P5, and P9 control and mutant animals were analyzed (Fig. 5B). There were low basal levels of autophagy in control DRG at all ages examined. In contrast, significantly elevated levels of LC3-II were clearly detected in mutant DRG (at all ages examined), suggesting the extensive formation of autophagosomes despite the lack of PIK3C3.

Because functional lysosomes are required for the completion of the autophagy degradation processes (42, 43) and lysosomes may be dysfunctional in Pik3c3-deleted small sensory neurons, the increase in the level of LC3-II (Fig. 5B) may reflect the accumulation of undegradable autophagosomes and LC3-II isoforms (44, 45). To rule out this possibility, we performed an autophagy flux assay. Briefly, in cells with continued de novo autophagosome formation, the amount of LC3-II protein was kept at a constant low level, because autophagosomes are continuously degraded by lysosomes. Thus, inhibiting lysosome function will result in a significant increase in the amount of LC3-II proteins. We dissociated DRG from control and Pik3c3-cKO mice at P0 and cultured the neurons for 12 h in the presence or absence of the lysosomal inhibitor Chloroquine (7). We found that Chloroquine induced accumulation of LC3-II in both control (LC3II/LC3I ratio increased from 35.3% to 100.3%) and Pik3c3-mutant DRG cultures (LC3II/LC3I ratio increased from 52% to 99.4%) (Fig. 5C). This result further supported the hypothesis that Pik3c3-deficient neurons initiated a noncanonical autophagosome formation process. The autophagy flux assay could not be performed at later stages (P5–P9) after the subcellular pathology had appeared, because no mutant sensory neurons could survive the dissociation and in vitro culture at these later stages. Finally, EM analyses confirmed the presence of numerous double- or multi- membrane–encircled autophagosomes and autolysosomes in the lysosome-filled Pik3c3-deleted neurons that are small-diameter neurons (Fig. 5D).

PIK3C3-Independent Autophagy Still Requires ATG7.

We next generated [Advillin^Cre/+; Pik3c3^flox/flox; Atg7^flox/flox] double conditional knockout mice (designated as Pik3c3/Atg7-dKO) and performed Western blot analysis of LC3 in DRG lysates at P0 (Fig. 5E). Levels of LC3-II were clearly elevated in Pik3c3-cKO DRG compared with DRG isolated from either compound heterozygous [Advillin^Cre/+; Pik3c3^flox/+; Atg7^flox/+] or Atg7-cKO mice. Importantly, this increased amount of LC3-II in Pik3c3-cKO DRG returned to background level in Pik3c3/Atg7-dKO mice. This result indicated that the conversion of LC3-I to LC3-II during the PIK3C3-independent autophagosome formation process still requires ATG7. It suggests that the downstream molecular players after the initiation of the PIK3C3-independent autophagy are conserved between the canonical and noncanonical pathways (at least in sensory neurons).

Pik3c3-cKO, Atg7-cKO, and Pik3c3/Atg7-dKO Mice.

The following PCR primers are used to distinguish between wild-type Pik3c3 and Pik3c3^flox allele: A1, 5′-GGCCACCTAAGTGAGTTGTG-3′; A2, 5′-GAAGCCTGGAACGAGAAGAG-3′; A3, 5′-ATTCTGCTCTTCCAGCCACTG-3′. The PCR primers to detect the Pik3c3⁻ (deleted) allele are L1, 5′- AACTGGATCTGGGCCTATG-3′; L2, 5′-GAAGCCTGGAACGAGAAGAG-3′; L3, 5′-CACTCACCTGCTGTGAAATG-3′. Female Pik3c3^flox/flox mice were bred with male Advillin^Cre/+; Pik3c3^flox/+ mice to generate Advillin^Cre/+; Pik3c3^flox/flox mice (designated as Pik3c3-cKO mice). Atg7-cKO and Pik3c3/Atg7-dKO mice are generated accordingly. All experiments were conducted according to protocols approved by Duke University Institutional Animal Care and Use Committee.

Dissociated DRG Neuron Cultures and Transfection.

Dissociated DRG neuron cultures and transfection were performed as previously described (47). The mVenus-2xFYVE construct was generated similarly as described (48), and it consists of two copies of FYVE finger from mouse Hrs fused at the N termini with mVenus (49).

In Situ Hybridization, Immunofluorescence, TUNEL Staining, EM, Western Blot, and Autophagy Flux Assay Were Performed According to Standard or Previously Described Protocols.

Details can be found in SI Materials and Methods.

Quantitative Analysis.

Quantitative analyses were done with standard Student t test. Details can be found in SI Materials and Methods.