PI31 expression is neuroprotective in a mouse model of early-onset parkinsonism

Significance

Mutations in FBXO7/PARK15 cause early-onset parkinsonism in human patients. Using both Drosophila and conditional mouse KO models we demonstrate that FBXO7 deficiency leads to PI31 destabilization, impaired synaptic proteasome transport, and tau hyperphosphorylation. Transgenic expression of PI31 in neurons prevents neurodegeneration, greatly improves motor performance and synaptic integrity. It also extends lifespan up to fourfold in full-body Fbxo7 knockout mice. These findings reveal that PI31 is a key neuroprotective factor acting downstream of FBXO7. They also suggest that stimulating local proteasome activity at synapses can suppress synaptic dysfunction in neurodegenerative diseases. Therefore, our current study provides both fundamental insights into synaptic protein homeostasis and a framework for developing therapies to treat early-onset parkinsonism and related disorders.

Abstract

Neurodegenerative diseases present one of the most significant global health challenges. These disorders are defined by the accumulation of abnormal protein aggregates that impair synaptic function and cause progressive neuronal degeneration. Therefore, stimulating protein clearance mechanisms may be neuro-protective. Variants in FBXO7/PARK15 cause Parkinsonian Pyramidal Syndrome, an early-onset parkinsonian neurodegenerative disorder in humans, and inactivation of this gene in mice recapitulates many phenotypes seen in patients. The proteasome regulator PI31 is a direct binding partner of Fbxo7 and promotes local protein degradation at synapses by mediating fast proteasome transport in neurites. PI31 protein levels are reduced when the function of Fbxo7 is impaired. Here we show that restoring PI31 levels in Fbxo7 mutant fly and mouse strains prevents neuronal degeneration and significantly improves neuronal function, health, and lifespan. Notably, Fbxo7 inactivation in mouse neurons causes hyperphosphorylation of tau, and this was suppressed by transgenic expression of PI31. Our results demonstrate that PI31 is a crucial biological target through which Fbxo7 deficiency drives pathology. Therefore, targeting the PI31-pathway may represent a promising therapeutic approach for treating neurodegenerative disorders.

Most age-related neurodegenerative diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD), and amyotrophic lateral sclerosis (ALS) are defined by the accumulation of abnormal protein aggregates (1–11). These deposits are thought to directly or indirectly reflect the presence of neurotoxic protein aggregates. However, the precise mechanisms underlying the accumulation of pathognomonic proteins and the nature of their toxicity have remained elusive (4, 12–14). Although these conditions affect different types of neurons, they all begin with alterations in synaptic efficacy that are associated with synaptic and dendritic spine pathology, pointing to synaptic failure as a common and early disease mechanism (15–17). Moreover, the various pathological aggregates and deposits in these diseases contain poly-ubiquitinated (poly-Ub) proteins, indicating that these proteins were tagged for destruction but escaped proteasome-mediated degradation (18, 19). One possible explanation is that insufficient local proteasome activity at synapses initiates the formation of poly-Ub aggregates. The ubiquitin–proteasome system (UPS) is the primary mechanism by which cells degrade unwanted, damaged, and potentially toxic intracellular proteins (4, 20–23). New proteins are both synthesized and locally degraded by the proteasome at synapses, and this process is critical for synaptic function and plasticity (24–31). Bingol and Schuman originally demonstrated that proteasomes are rapidly recruited to synapses by active transport (31). The proteasome-binding protein PI31 (Proteasome Inhibitor of 31kD) serves as an adapter to directly couple 26S proteasomes to motor proteins and mediate fast transport of proteasomes in axons and dendrites (32). Although PI31 was originally identified based on its ability to inhibit the hydrolysis of small peptides by the 20S proteasome core particle in vitro, it was subsequently shown to promote protein breakdown in vivo in organisms as diverse as plants, yeast, fruit flies, and mice (33–37). PI31 stimulates the assembly of 19S and 20S particles into 26S proteasomes, and it is required for proteasome transport in axons and dendrites (32, 38). Importantly, disruption of this process impairs synaptic protein homeostasis and causes progressive neuronal dysfunction and degeneration in Drosophila and mice (32, 35). Moreover, reduced activity of PI31 is directly linked to human disease. Biallelic variants in both PMSF1, the gene coding for PI31, and FBXO7, a conserved direct binding partner of PI31, cause early-onset neurodegenerative disease (39–44). Depending on the severity of variants, neurological symptoms in humans range from perinatal lethality with neurological manifestations to early-onset PD/parkinsonism (39–47). Fbxo7 binds directly to PI31, and its inactivation leads to proteolytic cleavage and reduced levels of PI31 (Fig. 2A) (33, 48, 49). This suggests that loss of Fbxo7 may cause disease by reduction of PI31 function. In this case, increased PI31 protein levels would be expected to alleviate Fbxo7 phenotypes. Here we demonstrate that, like PI31, nutcracker (ntc), the Drosophila ortholog of Fbxo7, is required for fast axonal transport of proteasomes. In addition, we show that transgenic PI31 overexpression can rescue this phenotype. Furthermore, increasing PI31 protein levels in Fbxo7 mutant mice suppresses neuronal degeneration and greatly extends animal health and lifespan. Thus, targeting this pathway may represent a promising strategy to treat neurodegenerative disease.

Fig. 2.

Motor neuron-specific inactivation of Fbxo7 resembles loss of PI31. (A) Inactivation of Fbxo7 reduces PI31 protein levels. Western blot analysis revealed that loss of Fbxo7 reduces PI31 protein levels in various tissues. SC is spinal cord. Control n = 3, Fbxo7 knockout n = 3. (B) Conditional inactivation of PI31 in adult mice leads to a dramatic reduction in thymus and testis size and weight. Tamoxifen injected PI31^fl/fl (control) n = 4 to 9, PI31^fl/fl;Ubc^CreERt2/+ n = 5 to 10 mice. Statistical significance determined by the Student t test. Data are represented by a grouped scatter plot with mean and error bars (STDEV). ***P Value < 0.001, ****P Value < 0.0001, ns = nonsignificant. (C) Fbxo7^fl/fl;Hb9^Cre/+ mice have reduced weight and developed kyphosis, similar to PI31^fl/fl; Hb9^Cre/+ mice. Two 3-mon-old mice of each genotype, Fbxo7^fl/fl;Hb9^Cre/+ and Fbxo7^fl/fl control, are shown. (E−I) Micrographs depict the innervation of triangularis sterni muscles by spinal motor neurons. Loss of Fbxo7 results in axonal swellings (white arrowheads), especially at synapses and their vicinity, and axonal sprouting. (H and I) Accumulation of large P62 aggregates at NMJs of Fbxo7^fl/fl;Hb9^Cre/+ mice is indicative of proteotoxic stress (H and I are Insets marked by white boxes in G). (J and K) Quantification of axonal swellings and axon sprouting. (Scale bars in D and E are 100 μm, in F to I they are 20 μm.) Statistical significance was defined by the Student t test. Data are represented by a grouped STDEV. **P Value < 0.01. Fbxo7^fl/fl control n = 3, Fbxo7^fl/fl;Hb9^Cre/+ n = 4.

Results

Elevated PI31 Can Compensate for the Conditional Loss of nutcraker/FBXO7 in Drosophila Motor Neurons.

Inactivation of nutcracker (ntc), the Drosophila ortholog of Fbxo7, causes motor defects, dramatically shortens lifespan, and is associated with mitochondrial defects that resemble classic Pink1/parkin phenotypes (50). To examine a requirement of ntc for axonal transport of proteasomes, we conducted live imaging of proteasome movement in axons using a Prosβ5-RFP reporter, a proteasome subunit fused to a red fluorescent protein (32, 51). Expression of the reporter was driven by R94G06-GAL4 for highly specific expression in a subset of motor neurons. Inactivation of ntc in compound heterozygous larva, ntc^f07259/ms771, significantly reduced motility of proteasomes in axons (Fig. 1 A and B and Movies S1 and S2). Both anterograde and retrograde movements of proteasomes were severely reduced, whereas the number of stationary proteasome particles was increased (Fig. 1 A and B and Movies S1 and S2). These phenotypes are very similar to the ones seen in PI31^−/− and LC8 dynein light chain homolog dDYNLL1/ctp^−/− motor neurons (32, 51). Next, we tested whether increasing PI31 levels could rescue the ntc^f07259/ms771 proteasomal transport deficit by adding either one or two copies of a UAS-PI31 transgene. Indeed, two copies of the UAS-PI31 transgene rescued ntc^f07259/ms771 anterograde proteasome transport and improved retrograde transport while decreasing the number of stationary particles (Fig. 1 A and B and Movie S3). Large swellings and aggregates positive for Prosβ5-RFP were observed in ntc^f07259/ms771 larval axons (Fig. 1 C and D). This phenotype was also rescued by two copies of the UAS-PI31 transgene.

Fig. 1.

ntc is required for fast axonal proteasome transport in Drosophila motor neurons, and this defect can be rescued by increased expression of PI31. (A and B) Loss of ntc results in significantly reduced anterograde and retrograde transport of proteasomes in axons of motor neurons, and this is evident by the increase of stationary particles. Raising PI31 levels rescues the transport phenotype. (A) Representative kymographs of Prosβ5-RFP motility in motor neuron axons of wild-type (wt), ntc^f07259/ms771, ntc^f07259/ms771 UAS-PI31/+, and ntc^f07259/ms771 UAS-PI31/ UAS-PI31 larvae. Prosβ5-RFP and PI31 expression was driven using the motor neuron-specific R4G06-Gal4. Kymographs were generated from 60 s live-imaging experiments. The x-axis represents distance, and the y-axis represents time. The directions of anterograde and retrograde movement are indicated on top of the kymographs. Stationary particles appear as vertical lines, whereas motile particles appear as diagonal lines on kymographs. (B) Quantification of proteasome movement. Shown are the percentages of Prosβ5-RFP particles that moved in anterograde or retrograde direction or appeared stationary in axons of wt n = 6, ntc^f07259/ms717 n = 6, ntc^f07259/ms771 UAS-PI31/+ n = 6, and ntc^f07259/ms771 UAS-PI31/ UAS-PI31 n = 6 Drosophila. Mean and error bars (SEM). One-way ANOVA with Tukey’s honesty significant different post hoc test. (C and D) Large swellings and aggregates (4 to 10 µm) positive for Prosβ5-RFP were present in motor neuron axons of ntc^f07259/ms771 but not in wt axons. Increased PI31 levels reduce the formation of the swellings and aggregates in ntc^f07259/ms771 axons, and two copies of the UAS-PI31 transgene completely rescued the phenotype. (C) Representative images of wt n = 5, ntc^f07259/ms771 n = 6, ntc^f07259/ms771 UAS-PI31/+ n = 7, and ntc^f07259/ms771 UAS-PI31/ UAS-PI31 n = 5, axons. (D) Quantification of Prosβ5-RFP axonal swellings and aggregates in motor neurons of wt, ntc^f07259/ms771, ntc^f07259/ms771 UAS-PI31/+ and ntc^f07259/ms771 UAS-PI31/ UAS-PI31 binned by size. SEM. Significant test ANOVA Kruskal–Wallis *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Transgenic PI31 Expression Can Suppress Neurological Phenotypes in Fbxo7 Null Motor Neurons.

Inactivation of Fbxo7 causes site-specific, inactivating proteolytic cleavage and reduced levels of PI31 protein (33) (Fig. 2A). To further explore the idea that PI31 and Fbxo7 are functionally linked, we compared phenotypes resulting from inactivation of PI31 to those of Fbxo7 and found them to be very similar. Like Fbxo7 mutant mice, induction of ubiquitous loss of PI31 in PI31^fl/fl;Ubc^CreERt2 adult mice have a dramatic decrease in testis and thymus size with similar histopathology (Fig. 2B and SI Appendix, Fig. S1) (52, 53). Moreover, inactivation of Fbxo7 in motor neurons in PI31^fl/fl;Hb9^Cre/+ mice produced phenotypes that strongly resemble those we previously reported for PI31 inactivation in motor neurons (Fig. 2 C−K) (35). Like for PI31^fl/fl;Hb9^Cre/+ mice, Fbxo7 ^fl/fl;Hb9^Cre/+ mice were born with the expected Mendelian ratio and initially showed no gross motor dysfunction. 4 wk after birth, a difference in weight, kyphosis (Fig. 2C), and motor problems became apparent that progressively increased with age. Using whole-mount fluorescent confocal imaging to examine motor innervation of the triangularis sterni muscle reveals a loss of synapse architecture (Fig. 2 D−G), axonal swellings (Fig. 2 E and J), axon sprouting (Fig. 2 E and K), and large p62 granules in the NMJ and its vicinity (Fig. 2 H and I), very similar to PI31^fl/fl;Hb9^Cre/+ mice (35, 54). Therefore, motor neuron-specific loss of either Fbxo7 or PI31 produces nearly identical phenotypes.

Because inactivation of Fbxo7 reduces the levels of PI31 (Fig. 2A), we considered that at least some of the neuronal pathology seen in Fbxo7 ^fl/fl;Hb9^Cre/+ may be caused by decreased PI31 function. If so, ectopic expression of PI31 should be able to at least partially rescue Fbxo7 deficiency. To test this hypothesis, we generated transgenic mice that conditionally express a Flag-tagged form of PI31 (TgPI31) (SI Appendix, Fig. S2A). Since the Hb9^Cre allele is not viable as a homozygous, we used the Chat^Cre driver to increase the number of animals with desired genotypes. Chat^Cre has a broader expression pattern than Hb9^Cre, which includes spinal motor neurons along with all other cholinergic neurons. As expected, both physiological and cellular phenotypes were more severe compared to Hb9^Cre. To validate that TgPI31 is expressed and functional, its ability to rescue PI31 mutant mice was tested (SI Appendix, Fig. S2 B−H). The TgPI31 construct expressed protein of expected size and was able to completely rescue PI31^fl/fl;Chat^Cre mutant mice, demonstrating that the transgenic protein is biologically active (SI Appendix, Fig. S2 B−H). Next, we crossed TgPI31 to Fbxo7 ^fl/fl;Chat^Cre mice to test for possible rescue. Fbxo7 ^fl/fl;Chat^Cre were born at expected Mendelian ratios and appeared similar to their control littermates in the first 3 wk of age (Movie S4). However, starting at 4 wk of age, mutant mice failed to gain weight and by the age of 8 wk had to be euthanized due to poor health (Fig. 3 A and B). All animal work was performed as required by the United States Animal Welfare Act and the National Institutes of Health's policy to ensure proper care and use of laboratory animals for research, and under established guidelines and supervision by the Institutional Animal Care and Use Committee (IACUC) of The Rockefeller University. Protocol was approved by Dr. Engin Ozertugrul PhD M.Ed. Mice were housed in accredited facilities of the Association for Assessment of Laboratory Animal Care (AALAC) in accordance with the National Institutes of Health guidelines. Again, the phenotypes of Fbxo7 ^fl/fl;Chat^Cre mice were very similar to PI31^fl/fl;Chat^Cre mice at both the cellular and behavioral levels. In particular, the synaptic architecture of the NMJ in Fbxo7 ^fl/fl;Chat^Cre mice was highly abnormal, and neurons showed pathological hallmarks, such as axonal swelling and sprouting (Fig. 3 C, D, F, G, I, and J). Whereas Fbxo7 ^fl/fl;Chat^Cre developed a progressive motor neuron disease, Fbxo7 ^fl/fl;TgPI31;Chat^Cre mice were practically indistinguishable from control animals and did not develop any detectable motor dysfunction at 8 wk (Fig. 3 A and B and Movies S5 and S6). Micrographs of muscle innervation confirmed that NMJs of transgenically rescued animals were intact and appeared similar to that of normal controls (Fig. 3 E, H, I, and J). We conclude that expression of PI31 in Fbxo7 null motor neurons can suppress axonal degeneration and preserve neuronal function and organismal health.

Fig. 3.

Transgenic expression of PI31 rescues Fbxo7 knockout pathology. (A and B) Fbxo7^fl/fl;Chat^Cre animals are significantly smaller in size and develop kyphosis. Transgenic expression of PI31 in these mice, Fbxo7^fl/fl;Chat^Cre;TgPI31/+, suppressed these phenotypes and yielded animals that were indistinguishable from normal control littermates. (A) 8-wk old Fbxo7^fll+;Chat^Cre control, Fbxo7^fl/fl;Chat^Cre;TgPI31/+ and Fbxo7^fl/fl;Chat^Cre mice. (B) Fbxo7^fl/fl;Chat^Cre lose weight, while transgenic expression of PI31 in Fbxo7^fl/fl;Chat^Cre;TgPI31/+ mice restored body weight to that of control littermates. Quantification of mouse body weight. Mean and error bars (STDEV). (C−J) Loss of Fbxo7 in cholinergic neurons results in axonal swellings (white arrowheads) and axonal sprouting. These phenotypes were rescued to control levels in mice with TgPI31. (C−H) Confocal images of triangularis sterni muscle innervation by spinal motor neurons. (Scale bars in C−E 100 μm, in F−H 20 μm.) (I and J) Quantification of axonal swellings and axon sprouting. AU is arbitrary units. Statistical significance was determined by the Student t test. Data are represented by a grouped STDEV. *P Value < 0.05, **P Value < 0.01. Control n = 4, Fbxo7^fl/fl;Chat^Cre n = 4, Fbxo7^fl/fl;Chat^Cre;TgPI31/+ n = 6.

Transgenic Expression of PI31 Suppresses Phenotypes of Panneuronal and Whole-Body Inactivation of Fbxo7.

Variants of FBXO7 which cause human disease have primarily received clinical attention because they impair the function and survival of neurons in the central nervous system (CNS). To test whether PI31 can protect against the neurological phenotypes resulting from the loss of Fbxo7 in all neurons, we use the panneuronal specific Cre transgene Actl6b^Cre (BAF53b^Cre) (55). Using this driver allows us to simultaneously inactivate Fbxo7 and drive TgPI31 expression in all neurons (Fig. 4) (55). Conditional loss of Fbxo7 using this driver results in phenotypes very reminiscent of full-body inactivation of Fbxo7 knock-out mice, [Fbxo7^{tm1b(EUCOMM)Hmgu} (Fbxo7 ^KO/KO (tm1b)] and Fbxo7^{tmd(EUCOMM)Hmgu} (Fbxo7 ^KO/KO (tm1d); see SI Appendix, Fig. S3 for a description of Fbxo7 alleles) in their lack of weight gain after post-natal day 11 (P11), mortality, and motor deficit (43, 48). After P5 the body weight of Fbxo7 ^fl/fl; Actl6b^Cre mice is retarded and on average mice die after 23.28 d (median is 23 d) when provided wet food starting at P14 (Fig. 4 A and B). A single copy of TgPI31 in Fbxo7 ^fl/fl; Actl6b^Cre/+ mice increases body weight and extends median life to 42 d. On P20 Fbxo7 ^fl/fl; Actl6b^Cre/+ mice show a quicker latency to fall on the rotarod demonstrating a motor deficit (Fig. 4C). Unlike Fbxo7 ^KO/KO (tm1b) mice, when locomotion of P20 Fbxo7 ^fl/fl; Actl6b^Cre/+ mice are evaluated by open field assay they show an increase in total distance traveled (Fig. 4D). This is more consistent with the hyperactive behavior reported with the conditional loss of Fbxo7 in glutamatergic forebrain neurons (56). However, though the Fbxo7 ^fl/fl; Actl6b^Cre/+ mice travel further than the control mice, they show a decrease in their vertical activity, that is rearing and jumping, consistent with a deficit in hind limb function and a lack of balance when rearing (Fig. 4E). We found that all these phenotypes are suppressed by conditional expression of PI31 in Fbxo7 ^fl/fl; TgPI31/+; Actl6b^Cre/+ mice (Fig. 4 C−E). As reported for Fbxo7 ^KO/KO (tm1b) mice we also saw an increase in astrogliosis in P18 Fbxo7 ^fl/fl; Actl6b^Cre/+ mouse brains, which is also repressed by exogenous PI31 expression (Fig. 4F). Thus, PI31 can suppress phenotypes of Fbxo7 loss autonomously in the CNS as well as in peripheral neurons.

Fig. 4.

Elevated expression of PI31 suppresses phenotypes of panneuronal inactivation of Fbxo7. (A) Upon panneuronal inactivation of Fbxo7 (Fbxo7^fl/fl; ActI6b^Cre/+) mice stop gaining weight after postnatal day 11 (P11). TgPI31 expression increases body weight of these Fbxo7^fl/fl; ActI6b^Cre/+ mice. Mouse weight (grams) over time (days) of labeled genotypes—Fbxo7^fl/fl (control), Fbxo7^fl/fl; TgPI31/+, Fbxo7^fl/fl; ActI6b^Cre/+ and Fbxo7^fl/fl; ActI6b^Cre/+;TgPI31/+. Mean with error bar (STDEV). (B) Mice with panneuronal loss of Fbxo7, Fbxo7^fl/fl; ActI6b^Cre/+ mice, die around the age of weaning. TgPI31 doubles the life of Fbxo7^fl/fl; ActI6b^Cre/+ mice. Kaplan–Meier survival curves show median life of 23 d for Fbxo7^fl/fl; ActI6b^Cre/+ mice n = 29, and 42 d for Fbxo7^fl/fl; ActI6b^Cre/+;TgPI31/+ mice n = 30, significance P < 0.0001. (C) Mice with panneuronal loss of Fbxo7 show a quicker latency to fall on the rotarod assay then their control littermates. TgPI31 rescues this deficiency. Graphed is seconds to fall of labeled genotypes—Fbxo7^fl/fl (control) n = 30, Fbxo7^fl/fl; TgPI31/+ n = 28, Fbxo7^fl/fl; ActI6b^Cre/+ n = 18 and Fbxo7^fl/fl; ActI6b^Cre/+;TgPI31/+ n = 30. (D and E) In the open field assay P20 Fbxo7^fl/fl; ActI6b^Cre/+ mice appear hyperactive as shown by the greater distance traveled (D), and have difficulty rearing as shown by a vertical activity deficit (E). TgPI31 suppresses hyperactivity and vertical deficit of mice with panneuronal loss of Fbxo7. Labeled genotypes—Fbxo7^fl/fl (control) n = 28, Fbxo7^fl/fl; TgPI31/+ n = 21, Fbxo7^fl/fl; ActI6b^Cre/+ n = 18 and Fbxo7^fl/fl; ActI6b^Cre/+;TgPI31/+ n = 21. For (C−E) mean with error bar (STDEV) and one-way ANOVA with the Tukey post hoc test; ns, not significant, **P < 0.01, ***P < 0.001, ****P < 0.0001. (F) Mice with panneuronal loss of Fbxo show increased GFAP staining astrocytes, suggesting induction of astrogliosis. TgPI31 prevents the induction of astrogliosis. GFAP staining for glia in the region of the caudate putamen with axons of dopaminergic neurons (tyrosine hydroxylase positive) of P18 mice, with and without the panneuronal driver ActI6b^Cre and TgPI31 as labeled. Confocal images from two mice of each genotype are shown. Signal in white is GFAP, green is tyrosine hydroxylase, and blue is nuclei (Hoechst 33342), and (Scale bar, 20 µm.)

Loss of Fbxo7 leads to aberrant function of cells other than neurons. For example, specific conditional ablation of Fbxo7 in myelinating glia, oligodendrocytes, and Schwann cells can cause neuronal degeneration (57). Having demonstrated the ability of PI31 to suppress Fbxo7 loss autonomously in neurons, we next tested whether PI31 could suppress Fbxo7 loss in the whole animal. To do this mice with a constitutively expressing PI31 transgene (TgPI31cs) were generated. Our lab previously demonstrated that loss of PI31 results in lethality during late embryogenesis in mice (35). To test whether TgPI31cs could rescue PI31 knock out mice (PI31^KO/KO) (Psmf1^{tm1d(EUCOMM)Hmgu}), PI31^KO/+; TgPI31cs/+ mice were crossed to PI31^KO/+; TgPI31cs/+ mice to generate PI31^KO/KO mice with one or two copies of TgPI31cs. TgPI31cs rescues the embryonic lethality of PI31^KO/KO mice, and the mice born appear normal. At P40, PI31^KO/KO mice with TgPI31cs perform as well as their w.t. littermates on the rotarod (Fig. 5A).

Fig. 5.

Expression of PI31 suppresses phenotypes of whole-body Fbxo7 null mice. (A) Whole-body inactivation of PI31 in PI31^KO/KO mice causes death in late embryogenesis, and TgPI31cs rescues this lethality. PI31^KO/KO;TgPI31cs mice appear normal after birth and at P40 perform equally well on the rotarod as wild type (w.t.) littermates. No significant difference was seen by the Student t test, P = 0.782, with n = 5 for PI31^KO/KO;TgPI31cs mice and n = 5 for control w.t. mice. (B) TgPI31cs quadruples life of Fbxo7^KO/KO mice. Kaplan–Meier survival curves show median life of 22 d for Fbxo7^KO/KO n = 14, median life of 45 d for Fbxo7^KO/KO;TgPI31cs/+ (one copy) n = 30, and median life of 89 d for Fbxo7^KO/KO;TgPI31cs (two copies) n = 18. Significance P < 0.0001. (C) After postnatal day 11 (P11) Fbxo7^KO/KO mice stop gaining weight. Fbxo7^KO/KO mice with TgPI31cs gain weight longer and attain greater body weight then Fbxo7^KO/KO mice. Mouse weight (grams) over time (days) of labeled genotypes—Fbxo7^fl/fl (control), Fbxo7^KO/KO, Fbxo7^fl/fl;TgPI31/+, Fbxo7^KO/KO;TgPI31cs/+ and Fbxo7^KO/KO;TgPI31cs. Mean with error bar (STDEV). (D and E) In the open field assay P18 Fbxo7^KO/KO mice appear hyperactive, as shown by the greater distance traveled (D). In addition, they have difficulty rearing, as shown by a vertical activity deficit (E). By P20 motor function in most Fbxo7^KO/KO mice had deteriorated to a point where hyperactivity was masked. TgPI31cs suppresses both the hyperactivity and vertical deficit of FBXO7^KO/KO mice. Labeled genotypes—Fbxo7^fl/fl (control) n = 11, Fbxo7^KO/KO n = 8, Fbxo7^KO/KO;TgPI31cs/+ n = 8 and Fbxo7^KO/KO;TgPI31cs/(+) (one or two copies of TgPI31cs) n = 20. Mean with error bar (STDEV) and one-way ANOVA with the Tukey post hoc test; ns is not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

We then tested whether the TgPI31cs transgene could rescue Fbxo7 knock out mice (Fbxo7^{tm1d(EUCOMM)Hmgu}, Fbxo7^KO/KO (tm1d), SI Appendix, Fig. S3). The phenotypes observed for the Fbxo7^KO/KO (tm1d) were as reported for the FBXO7^KO/KO (tm1b) (48). After P5 the Fbxo7^KO/KO (tm1d) mice do not gain much body weight (with wet food provided daily starting at P14). On average male mice were 4.8 g by P18 versus control mice which weigh 9 g by P18 (Fig. 5B). Fbxo7^KO/KO (tm1d) mice expire on average at 21.7 d (median is 22 d) (Fig. 5C). No Fbxo7^KO/KO (tm1d) mice live past P24. A single copy of TgPI31cs allow male Fbxo7^KO/KO (tm1d) mice to reach an average body weight of 7 g by P18, which increases to 9.3 g by P28 while control mice average 17.5 g by P28 (Fig. 5B). These mice maintain this body weight till the final days of life which average 46.6 d (median is 45 d) (Fig. 5 B and C). One copy of TgPI31cs doubled the life of Fbxo7^KO/KO (tm1d) mice. When mice were crossed to allow for two copies of TgPI31cs, two distinct populations of rescued Fbxo7^KO/KO (tm1d) mice could be distinguished. One population matches the Fbxo7^KO/KO (tm1d) mice with a single copy of TgPI31cs described above. The second population, presumably those with two copies of TgPI31cs, show greater suppression of the Fbxo7^KO/KO (tm1d) phenotypes. Thus, by P28 mice with two copies of TgPI31cs can be easily distinguished from those carrying a single copy. Male mice of this population reached an average of 14.8 g by P56 versus 20.4 g for control mice. They maintain this weight till their final days of life which average 87.6 d (median is 89 d) (Fig. 5 B and C). Thus, two copies of the constitutive TgPI31 quadrupled the life of Fbxo7^KO/KO (tm1d) mice. This demonstrates a dosage sensitivity for PI31 suppression of loss of Fbxo7^KO/KO (tm1d) phenotypes. Finally, Fbxo7^KO/KO (tm1d) mice were tested for locomotion with the open field assay and like the panneuronal conditional mice they demonstrate an increase in total distance traveled and a deficit in vertical activity (Fig. 5 D and E). This hyperactivity and vertical activity deficit are also suppressed by either one or two copies of TgPI31cs (Fig. 5 D and E).

Tau, a Biomarker for Neurodegenerative Diseases, Is Hyperphosphorylated in Brains of Fbxo7 Null Mice and This Phenomenon Is Suppressed by Transgenic Expression of PI31.

Reduced function of either Fbxo7 or PI31 results in early-onset parkinsonism that is accompanied by additional neurological and cognitive symptoms (39–41, 43, 44). Variants of PI31 have also been linked to AD in a genome wide association study (42). The microtubule-associated tau protein becomes excessively phosphorylated in various neurodegenerative diseases and serves as a biomarker for driving these diseases (58). Therefore, we examined the state of the tau protein in Fbxo7 null brains compared to controls. First, we used the Tau-1 antibody which only recognizes its epitope when it is not phosphorylated on Ser195, Ser198, Ser199, Ser202, and Thr205 (59, 60). Using Tau-1 a significant decrease in staining was seen in Fbxo7 null neurons versus controls (Fig. 6A). This result is consistent with a change of tau phosphorylation in response to the loss of Fbxo7 function. To examine this possibility further, we performed westerns for tau with extracts from P20 Fbxo7^KO/KO (tm1d) brains using a Ser199-Ser202 phosphorylation-dependent tau antibody (Thermo Fisher 44-768G) along with an antibody recognizing total tau protein (Encor MCA-5B10) (Fig. 6 G−I). This analysis reveals a clear increase in Ser199 and Ser202 phosphorylated tau in Fbxo7^KO/KO (tm1d) brain extracts (Fig. 6 G and H). Significantly, tau hyperphosphorylation is suppressed in Fbxo7^KO/KO (tm1d) by TgPI31cs transgenes (Fig. 6 G and H). Evidence of increased tau phosphorylation is also seen in Fbxo7 null neurons with the AT8 antibody which recognizes tau phosphorylation on Ser202 and Thr205 (SI Appendix, Fig. S4). We conclude that inactivation of Fbxo7 in neurons leads to hyperphosphorylation of tau, and this may be one mechanism contributing to pathologies in Fbxo7 deficiency. Remarkably, modest elevation of PI31 protein levels by transgenic expression suppressed the hyperphosphorylation of tau. Once again, we note decreased protein levels of both endogenous and transgenic PI31 upon inactivation of Fbxo7, consistent with our earlier results that PI31 undergoes site-specific proteolytic cleavage and degradation when its direct binding to Fbxo7 is disrupted (Fig. 6 C−F) (33). Notably, endogenous PI31 is stabilized by increased expression of transgenic PI31 in Fbxo7^KO/KO (tm1d) mice (Fig. 6 C and D), possibly by titrating a rate-limiting factor for PI31 degradation. Collectively, these results suggest that decreased levels of PI31 due to impaired function of Fbxo7 play a prominent role for driving pathology and the hyperphosphorylation of tau, and that moderately elevated levels of PI31 are neuroprotective in this model.

Fig. 6.

Loss of Fbxo7 leads to hyperphosphorylation of Tau. (A) Panneuronal inactivation of Fbxo7 in Fbxo7^fl/fl; ActI6b^Cre/+ mice causes decreased antibody staining against dephosphorylation-dependent Tau-1 epitope. This is rescued by the TgPI31 transgene. Tau-1 staining of tau in the region of the caudate putamen with axons of dopaminergic neurons (tyrosine hydroxylase positive) in the brains of P18 mice, with and without the panneuronal driver ActI6b^Cre and TgPI31 as labeled. Images from two mice of each genotype are shown. Signal in green is Tau-1 (MAB3420) staining of tau, red is tyrosine hydroxylase, and blue are nuclei (Hoechst 33342), (Scale bar, 20 µm.) (B) Immunoblot for Fbxo7 confirms KO status of mice. (C−E and F) Loss of Fbxo7 leads to loss of both endogenous PI31 and exogenously expressed PI31. This indicates that the decrease in PI31 is post-translational. Interestingly, transgenic PI31 expression stabilizes endogenous PI31, suggesting that the PI31 degradation rate is limited. (G−I) Loss of Fbxo7 in Fbxo7^KO/KO mice leads to an increase in phosphorylation of tau in whole brain extracts, which is suppressed by transgenic expression of PI31. This increase in phospho-tau levels is statistically significant versus all other genotypes. All other pairings are statistically insignificant. The GAPDH panel in G is identical to the one shown in B because the blot was stripped and reprobed. (B, C, and F) Western blot analysis of whole brain extracts of P20 mice with genotypes labeled TgPI31cs (Fbxo7^fl/fl;TgPI31cs), TgPI31cs/+ (Fbxo7^fl/fl;TgPI31cs/+), control (Fbxo7^fl/fl), Fbxo7^KO/KO, Fbxo7^KO/KO;TgPI31cs/+ and Fbxo7^KO/KO;TgPI31cs. Except for the TgPI31cs homozygous mouse in the first lane, extracts for three mice of each genotype are shown. (D−F, H, and I) Densitometry of western blots of indicated mouse genotypes. n = 3 TgPI31cs (Fbxo7^fl/fl;TgPI31cs), n = 3 control (Fbxo7^fl/fl), n = 5 Fbxo7^KO/KO, n = 6 Fbxo7^KO/KO;TgPI31cs/+ (one copy Tg) and n = 4 Fbxo7^KO/KO;TgPI31cs (two copies Tg). Mean with error bar (STDEV) and one-way ANOVA with the Tukey post hoc test; ns, not significant, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Discussion

Variants in FBXO7 have been identified in human patients with familial PARK15, a rare form of early-onset parkinsonism with pyramidal tract involvement (39, 40, 45, 47). Here we show that transgenic overexpression of the PI31 proteasome regulator is neuroprotective in Fbxo7 animal models for this disease. In particular, moderately increased expression of PI31 was able to suppress neurodegeneration in both Drosophila and mouse Fbxo7 null mutants, even in full knockout animals. This effect was surprisingly strong as 2 copies of a PI31 transgene extended the viability of mice in which Fbxo7 was inactivated in all cells (“Fbxo7 whole body KO”) approximately four-fold: Whereas Fbxo7 null animals died around P22, it was extended to over 89 d in animals with two PI31 transgene copies (Fig. 5B). Similarly striking effects were seen for the suppression of weight loss and motor performance (Figs. 4 and 5). It is worth noting that levels of transgenic PI31 protein were rather modest and comparable to endogenous protein levels (Fig. 6 C and D−F). In Fbxo7^KO/KO mice with one copy of the transgene, total PI31 was 82% of endogenous PI31 in control mice. Moreover, rescue of Fbxo7 phenotypes was very sensitive to transgene copy number, revealing a pronounced dosage-sensitivity (Fig. 5). It is worth noting that modest levels of PI31 overexpression (2 to 5-fold) were well tolerated in both Drosophila and mice, with no obvious deleterious effects detectable throughout animal lifespan.

The extent to which loss of Fbxo7 function can be rescued by PI31 expression is surprising for several reasons. First, because Fbxo7 is the substrate recognition module of an SCF E3 ubiquitin ligase complex it would be expected to have a considerable number of substrates for ubiquitylation (61). Second, mutations in Fbxo7 were reported to affect the stability, function, and interactions with several key PD-related proteins (45, 47). For example, it was proposed that Fbxo7 directly interacts with PINK1 and Parkin to stimulate mitophagy, although the physiological relevance of this proposed mechanism remains to be established (50, 62–65). Finally, an earlier report analyzing Fbxo7 mouse mutants also showed that loss of Fbxo7 reduces PI31 protein levels, but the authors concluded that proteasomal defects were not mediated by PI31 (48). However, their conclusions were based on whole cell assays for proteasome activity, which would have missed the localized defects in protein homeostasis at synapses that we previously reported (32, 35). The ability to strongly attenuate Fbxo7 mutant phenotypes with exogenous PI31 in both Drosophila and mice clearly demonstrates that PI31 is a key biological target in the pathology mechanism, and that this pathway has been conserved in evolution from insects to mammals (33). Our results also indicate that the reported effects on mitochondrial function and neuro-inflammation may be mediated, at least in part, by PI31 acting down-stream of Fbxo7 (44, 47, 49, 50, 62, 65–69). However, transgenic rescue by PI31 is not complete, even in Fbxo7^KO/KO mice with two copies of TgPI31cs which have an almost 2-fold higher PI31 level than that of endogenous PI31 in control mice. Therefore, while PI31 is a major target in Fbxo7 deficiency, it is likely that Fbxo7 plays other independent functions that contribute to disease symptoms. Finally, a two to fivefold transgenic increase of PI31 protein levels appears to be well tolerated in both Drosophila and mice, since no adverse effects were detectable over the normal lifespan of flies and mice in a WT background.

In addition to its interaction with Fbxo7, recent work also links PI31 directly to human disease. Results from a pedigree analysis showed that PSMF1, the human gene encoding the PI31 protein, can cause early-onset PD/parkinsonism (44). Depending on the severity of variants, these patients suffer from a phenotypic spectrum that ranges from early lethality with arthrogryposis to early-onset PD/parkinsonism and display a range of additional neurological symptoms (44). PI31 CKO mice recapitulate many of these phenotypes, suggesting a remarkable degree of conservation at the molecular, cellular, and functional level of this pathway. Variants in PSMF1 and FBXO7 have also been identified in AD and ALS (42, 44, 47). Interestingly, one patient with the R242H variant in PSMF1 was diagnosed with early-onset parkinsonism, while another patient with the same variant displayed symptoms of AD (42, 44). We propose that this diversity of phenotypes is likely due to genetic background and environment that differentially affect different brain regions in “weak” (hypomorphic) variants, i.e. variants that retain some protein function. According to this model, strong variants cause very early, juvenile onset disease with high penetrance. On the other hand, milder variants predispose aging neurons to disease in combination with other genetic or epigenetic factors, such as cellular stress and excitotoxicity. While considerably more future work is needed to fully understand the disease mechanism, the available data clearly link impaired function of FBXO7 and PI31 directly to human disease, and they point to a critical role of this pathway for local, proteasome-mediated protein degradation at synapses.

PI31 stimulates protein breakdown in vivo by at least two mechanisms. First, PI31 promotes the assembly of 19S and 20S subunits into 26S particles, which are the primary proteolytic machines responsible for the regulated degradation of poly-ubiquitinated intracellular proteins (23, 38, 70). Second, PI31 couples proteasomes directly to microtubule-based motors and thereby mediates the fast bidirectional transport of proteasomes between the cell body and synapses (32). PI31 binds directly to both proteasomes and Fbxo7, and PI31 and Fbxo7 are enriched in axons (SI Appendix, Fig. S5) (32). Inactivation of PI31 blocks proteasome transport, disrupts synaptic protein homeostasis, stimulates aggregate formation, and causes progressive neuronal dysfunction and eventually cell death (32, 35). These observations suggest the following working model. Diminished PI31 activity reduces the availability of active proteasomes at synapses, and this increases the likelihood that poly-ubiquitinated proteins tagged for destruction escape proteasomal degradation and form aggregates. This situation is made worse since oligomers of many aggregation-prone neurotoxic proteins can directly inhibit the proteasome, potentially creating a feedforward loop of increased proteotoxic stress (71–74). Although synaptic protein aggregates can be removed by the autophagy–lysosome pathway (ALP), this requires retrograde transport of autophagosomes since no mature lysosomes are present at synapses (21, 75, 76). Furthermore, the ALP cannot substitute for proteasome-mediated local regulation of synaptic protein levels. Over time, and possibly influenced by other genetic factors, environment, and stress, proteins that escape proteasome-mediated degradation at synapses are expected to progressively impair synaptic function and long-term survival. It should be noted that synapses also contain free neuronal 19S regulatory particles that alter synaptic transmission in a degradation-independent manner (77). While PI31 binds directly to 20S particles and mediates the transport of single-capped 26S proteasomes, we cannot rule out an effect on 19S particles in Fbxo7 and PI31 null neurons. Finally, we also considered the possibility that PI31 may affect uncapped 20S neuronal membrane-bound proteasomes (NMP) which have been implicated in the degradation of damaged and/or unstructured proteins and are thought to regulate neuronal activity (78–81). However, thus far we have not detected any substantial differences in NMPs between WT and Fbxo7 null neurons.

We show that Fbxo7 null mice have hyperphosphorylated tau (Fig. 6), but the underlying mechanism is not clear at this time. A large number of both kinases and phosphatases have been reported to affect the phosphorylation state of tau (58, 82). Moreover, proteotoxic stress can activate the stress-activated protein kinases JNK and p38 which then phosphorylate tau (83). Thus, tau hyperphosphorylation may be the result of proteotoxic stress caused by the loss of Fbxo7 and PI31. Alternatively, it is possible that decreased proteasome activity at the synapse causes increased levels of the CDK5 cofactor p35 which is regulated by the proteasome (84, 85). Since CDK5 kinase has been shown to phosphorylate tau at S199, S202, and Thr205, increased CDK5 activity could explain our findings.

Interestingly, tau stabilizes microtubules in neurons and supports axonal transport of proteins and organelles. The fact that Fbxo7 and PI31 deficiency causes hyperphosphorylation of tau is expected to promote aggregation and tangle formation, which in turn may further negatively impact proteasome transport and synaptic protein homeostasis. A recent post-mortem study of patients with mild motor deficits showed increased phospho-tau antibody AT8 staining that was independent of alpha-synuclein aggregates (86). Therefore, it is possible that tau phosphorylation drives or amplifies disease in PSMF1 and FBXO7 patients.

Our findings have important implications for the possible development of therapeutic strategies to treat age-related neurodegenerative diseases associated with aggregate-prone proteins. Impaired activity of the UPS is well documented in age-related neuronal degeneration, and efforts to stimulate proteasome activity to promote clearance of aggregate-prone pathognomonic proteins have been made (72, 87, 88). Indeed, it has been reported that increasing proteasomes globally in neurons can protect from Alzheimer’s-like pathology in mouse and fly amyloid precursor protein overexpression models (87). However, increasing the local proteolytic capacity at synapses through targeting proteasome transport pathways has not been previously explored. The results presented here suggest that stimulating the PI31 pathway is a promising strategy in this regard. One potential avenue to explore for eventual clinical translation would be adeno-associated viruses (AAV)-mediated “PI31 gene therapy.” AAV-based gene therapy has been clinically successful in mitigating a number of neurological and CNS disorders (89–92). Patients with PMSF1 and FBXO7 deficiency would be obvious candidates to benefit from this approach.

Summary of Methods.

We used Drosophila and mouse models of FBXO7 deficiency to assess the neuroprotective effects of PI31. In flies, axonal proteasome transport was assayed using live imaging of Prosβ5-RFP in motor neurons of nutcracker(FBXO7) deficient flies with and without transgenic PI31. Transgenic mouse lines expressing Flag-tagged PI31, both conditionally and constitutively, were used to test rescue effects in conditional Fbxo7 and full Fbxo7^KO mice. Conditional Fbxo7 knockout mice were generated using several Cre drivers (Hb9^Cre, Chat^Cre, and Actl6b^Cre) to inactivate Fbxo7 in specific neuronal populations. Neuromuscular junction integrity was assessed via immunofluorescence of the triangularis sterni muscle. Behavioral performance was evaluated using rotarod and open field assays. Survival, body weight, motor function, and histopathology for gliosis were compared across genotypes to assess the extent of phenotypic rescue by PI31. Western blotting and immunostaining were employed to analyze PI31 expression and tau phosphorylation in brain tissues. Detailed Materials and Methods can be found in the SI Appendix. Mice were euthanized via CO2 inhalation followed by cervical dislocation. All animal work was performed as required by the United States Animal Welfare Act and the National Institutes of Health's policy to ensure proper care and use of laboratory animals for research, and under established guidelines and supervision by the IACUC of The Rockefeller University. Protocol was approved by Dr. Engin Ozertugrul PhD M.Ed. Mice were housed in accredited facilities of the AALAC in accordance with the National Institutes of Health guidelines.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Representative movie of Prosβ5-RFP mobility in motor neurons of Drosophila third-instar control larvae. Prosβ5-RFP expression from UAS transgene is driven by the motor neuron-specific R4G06-Gal4. The movies were created from stack-images collected at 500 ms intervals and was sped up to 15 fps. The genotype is ntc^f0259 UAS-Prosβ5-RFP R4G06-Gal4/ UAS-Prosβ5-RFP R4G06-Gal4.

Movie S2.

Representative movie of Prosβ5-RFP mobility in motor neurons of Drosophila ntc^f0259/ms771 compound heterozygous third-instar larvae. Note the two large Prosβ5-RFP containing aggregates which do not move. Prosβ5-RFP and PI31 expression from UAS transgenes is driven by the motor neuron-specific R4G06-Gal4. The movie was created from stack-images collected at 500 ms intervals and was sped up to 15 fps. The genotype is ntc^f0259 UAS-Prosβ5-RFP R4G06-Gal4/ ntc^ms771 UAS-Prosβ5-RFP R4G06-Gal4.

Movie S3.

Representative movie of Prosβ5-RFP mobility in motor neurons of Drosophila third-instar larvae. The phenotype seen in Movie S2 is rescued by two copies of the PI31 transgene. Prosβ5-RFP and PI31 expression from UAS transgenes is driven by the motor neuron-specific R4G06-Gal4. The movies were created from stack-images collected at 500 ms intervals and was sped up to 15 fps. The genotype is UAS-PI31/UAS-PI31; ntc^f0259 UAS-Prosβ5-RFP R4G06-Gal4/ ntc^ms771 UAS-Prosβ5-RFP R4G06-Gal4.

Movie S4.

Three week old Fbxo7^fl/fl;Chat^Cre mice are similar to control and Fbxo7^fl/fl;Chat^Cre;TgPI31/+ littermates. No significant size difference or motor dysfunction is detected. Displayed in the movie are two control mice, two Fbxo7^fl/fl;Chat^Cre;TgPI31/+ mice, and one Fbxo7^fl/fl;Chat^Cre mouse.

Movie S5.

Eight week old Fbxo7^fl/fl;Chat^Cre mice exhibit motor dysfunction and have breathing problems compared to control littermates. The mutant mouse marked with an asterisk has significant motoric problems. Exogenous PI31 rescues these phenotypes as demonstrated by the two Fbxo7^fl/fl;Chat^Cre;TgPI31/⁺ mice which are indistinguishable from their two control littermates in weight and motor function. Shown in the movie are the same mice that are displayed in Movie S4: two control mice, two Fbxo7^fl/fl;Chat^Cre;TgPI31/⁺ mice, and one Fbxo7^fl/fl;Chat^Cre mouse.

Movie S6.

Another example for the ability of TgPI31 expression to suppress motor dysfunction in mice lacking Fbxo7 function in motor neurons at eight weeks of age. The mutant mouse marked with an asterisk is motorically impaired. On the other hand, two Fbxo7-cko mice expressing the PI31 transgene were indistinguishable from a control animal.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.