|Home | About | Journals | Submit | Contact Us | Français|
Selected vulnerability of neurons in Huntington’s disease (HD) suggests alterations in a cellular process particularly critical for neuronal function. Supporting this idea, pathogenic Htt (polyQ-Htt) inhibits fast axonal transport (FAT) in various cellular and animal HD models (mouse and squid), but the molecular basis of this effect remains unknown. Here we show that polyQ-Htt inhibits FAT through a mechanism involving activation of axonal JNK. Accordingly, increased activation of JNK was observed in vivo in cellular and animal HD models. Additional experiments indicate that polyQ-Htt effects on FAT are mediated by the neuron-specific JNK3, and not ubiquitously expressed JNK1, providing a molecular basis for neuron-specific pathology in HD. Mass spectrometry identified a residue in the kinesin-1 motor domain phosphorylated by JNK3, and this modification reduces kinesin-1 binding to microtubules. These data identify JNK3 as a critical mediator of polyQ-Htt toxicity and provides a molecular basis for polyQ-Htt-induced inhibition of FAT.
Huntington’s disease (HD) is an autosomal dominant, adult-onset neurodegenerative disease1. Patients with a single mutant huntingtin (Htt) gene develop symptoms in midlife with 100% penetrance. The mutation is a toxic gain of function, because loss of a single gene has no phenotype and the Htt-null is embryonic lethal2. The disease-causing mutation is expansion of a CAG trinucleotide repeat encoding a polyglutamine (polyQ) tract and polyQ-expansions in other genes may also lead to adult-onset neurodegeneration, but pathogenic mechanisms remained uncertain2. Several lines of evidence indicated that pathogenesis in Huntington’s disease include inhibition of fast axonal transport (FAT)3. Although consequences of FAT inhibition in neurons are well established, mechanisms underlying inhibition of FAT by pathogenic Htt (polyQ-Htt) were unknown. Proposed mechanisms included wild-type Htt (WT-Htt) loss of function4, physical blockade by polyQ-Htt aggregates5,6, sequestration of motor molecules4, 6, and misregulation of FAT3, 7, 8.
We found no evidence for direct interactions of conventional kinesin and cytoplasmic dynein (CDyn) with WT-Htt or polyQ-Htt in vivo. However, analysis of HD models showed that polyQ-Htt, but not WT-Htt increased cJun N-terminal kinase (JNK) activity. Pharmacological and peptide inhibitors of JNK prevented inhibition of FAT by polyQ-Htt. Surprisingly, only a subset of JNK isoforms inhibited FAT, with neuron-specific JNK3 selectively mimicking effects of polyQ-Htt. In vitro phosphorylation and mass spectrometry studies showed that JNK3, but not ubiquitously expressed JNK1, phosphorylated Ser176 in the kinesin heavy chain (kinesin-1, KHC) motor domain. Consistent with this location, phosphorylation of kinesin-1 by JNK3 inhibited kinesin binding to microtubules and translocation along axons. Our data indicate that polyQ-Htt inhibits FAT by a mechanism involving axonal JNK3 activation and phosphorylation of kinesin-1.
PolyQ-Htt inhibits FAT in various experimental systems, including Drosophila4, 5, neuroblastoma cells9, 10, and isolated squid axoplasm8, but the molecular basis of inhibition was undetermined3. Interactions have been reported between exogenously overexpressed Htt and conventional kinesin4 or various subunits of cytoplasmic dynein (CDyn)9-11. Alternatively, polyQ-expansion was proposed to affect Htt function as a scaffolding protein for molecular motors9, 12. However, interactions between endogenous WT-Htt and molecular motors were not evaluated. We tested interactions by immunoprecipitation and subcellular fractionation (Fig. 1), as previously described13, 14. To avoid overexpression-related artifacts, we used brain tissue from 14 month-old homozygous HdhQ109 knock-in mice and age-matched controls, which express polyQ-Htt or WT-Htt at endogenous levels15. At this age, both polyQ-Htt-derived nuclear inclusions and insoluble aggregates are found in the brains of HdhQ109 knock-in mice15. Conventional kinesin is a heterotetramer composed of two heavy chains (kinesin-1, KHC) and two light chains (KLC)13. Antibodies recognizing kinesin-1 (H2) effectively immunoprecipitated kinesin-1 from detergent-soluble brain lysates independent of genotype (Fig. 1A). Kinesin-1 antibodies co-immunoprecipitated KLCs13, 14, but failed to co-immunoprecipitate either WT-Htt or polyQ-Htt. Similarly, antibodies against DIC co-immunoprecipitated DHC16, but did not immunoprecipitate WT-Htt or polyQ-Htt (Fig. 1A). Conversely, anti-Htt antibodies immunoprecipitated Htt from both wild type and homozygous HdhQ109 knock-in mouse brain lysates, but no kinesin-1, KLC, DIC nor DHC could be detected in Htt immunoprecipitates. To detect substoichiometric amounts of Htt associated with conventional kinesin or CDyn, we performed three rounds of immunoprecipitation, sufficient to nearly deplete mouse brain lysates of kinesin-1 (Fig. 1B) or DIC (Supplemental Fig. 1B). As in previous studies, marked reductions in kinesin-1 and KLC levels occurred with each immunoprecipitation cycle13, 14. However, Htt levels remained unchanged after three immunoprecipitation cycles, regardless of genotype.
PolyQ-Htt was reported to sequester molecular motors in detergent-insoluble aggregates when overexpressed4, 6. To evaluate this at endogenous levels, brain lysates from wild type and homozygous HdhQ109 knock-in mice were fractionated into detergent-soluble and insoluble fractions, and partitioning of Htt and molecular motors analyzed by immunoblot (Supplemental Fig. 1). WT-Htt and polyQ-Htt levels were comparable in detergent soluble and insoluble fractions, but the bulk of kinesin-1, DHC and DIC were detergent-soluble. Molecular motor levels were similar for wild type and homozygous HdhQ109 knock-in mice. Thus, WT-Htt and polyQ-Htt expressed at endogenous levels do not interact with molecular motors; so FAT inhibition associated with polyQ-Htt must result from a different mechanism.
Effects of WT-Htt and polyQ-Htt on FAT were assayed in isolated squid axoplasm8. Perfusion of WT-Htt showed no effect (Fig. 2A), but perfusion of polyQ-Htt at a concentration 100-1000 times lower than kinesin-13, 8 dramatically inhibited both anterograde (kinesin-dependent) and retrograde (CDyn-dependent) FAT (Fig. 2B). In axoplasm, anterograde FAT of membrane-bounded organelles (MBOs) depends primarily upon conventional kinesin17, which is regulated by phosphorylation7, 18, 19. Multiple kinase activities may be deregulated in Huntington’s disease20-22 and inhibition of FAT by pathogenic androgen receptor (polyQ-AR) depends on JNK activation in axoplasm7. This led us to evaluate the role of JNK in polyQ-Htt-induced inhibition of FAT. Axoplasmic JNK activity was assessed in vitro using recombinant GST-cJun, a well-characterized JNK substrate23 (Fig. 2C). Incubation with axoplasm lysates produced a time-dependent increase in cJun phosphorylation at JNK sites (serines 63/73), reflecting basal JNK activity in axoplasm. Pharmacological properties of endogenous axoplasmic JNK were validated with two JNK inhibitors (SP600125 and JIP peptide)23. SP600125 is a pharmacological inhibitor of JNK, showing >20-fold selectivity for JNK over a range of kinases24. JIP peptide (JNK inhibitor I) is a 20-amino acid peptide derived from a JNK-binding protein, which selectively inhibits JNK, but not p38 activity25. Both SP600125 and JIP peptide dramatically inhibited cJun phosphorylation, suggesting common pharmacological properties between squid and mammalian JNKs (Fig. 2C).
As with polyQ-AR7, co-perfusion of axoplasm with polyQ-Htt and either SP600125 (500nM, Fig. 2D) or JIP peptide (100 nM, Fig. 2E) blocked effects of polyQ-Htt on FAT, suggesting that polyQ-Htt inhibition of FAT requires JNK activity. In cultured cells, pharmacological inhibitors of histone deacetylase (HDAC) were reported to rescue defects in FAT induced by polyQ-Htt by increasing tubulin acetylation26. However, coperfusion of polyQ-Htt with the HDAC6 inhibitor BC-6-25 (10μM) (Compound 227) did not prevent inhibition of FAT by polyQ-Htt in axoplasm (Fig. 2F). Thus, inhibition of FAT by polyQ-Htt involves activation of endogenous axonal JNK.
Vesicle motility and biochemical assays indicated that JNK activity mediates polyQ-Htt inhibition of FAT. JNK activation involves phosphorylation by upstream mitogen-activated protein kinase kinases (MKKs, typically MKK4/MKK7)28. Effects of polyQ-Htt on JNK phosphorylation were evaluated in cellular and animal Huntington’s models, using antibodies against total JNK (pan-JNK, phosphorylation-independent) or against phosphorylated, catalytically active JNKs (pJNK). Effects of polyQ-Htt expression on JNK activation were evaluated qualitatively after transient transfection (Fig. 3A). NSC34 neuroblastoma cells were transfected with WT-Htt (Q18) or polyQ-Htt (Q56)29. Immunoblots show exogenous Htt constructs were expressed at levels similar to endogenous Htt. Pan-JNK and pJNK antibodies recognized two major immunoreactive bands (p46 and p54) corresponding to mammalian JNKs in NSC34 cell lysates and striatal mouse tissue (Fig. 3A, C)30. Expression levels for both pan-JNK immunoreactive bands were comparable in cells expressing WT-Htt and polyQ-Htt. However, pJNK antibodies showed a clear increase in p54 immunoreactivity in cells expressing polyQ-Htt (Fig. 3A), consistent with previous reports20, 22, 31.
Differential activation of p54 and p46 JNKs led us to evaluate their identity using antibodies selective for JNK1, JNK2 and JNK3. Specificity was validated by immunoblotting against recombinant GST-tagged JNKs, as well as brain lysates from mice with ablation of individual JNK genes30 (Fig. 3B). Anti-JNK1, 2 and 3 antibodies selectively recognized their target JNK (Fig. 3B), showing no immunoreactivity against brain lysates from mice deleted for that JNK gene (Fig. 3B). We examined JNK expression in striatal tissue from 8 month-old wild-type, heterozygous and homozygous HdhQ109 mice (Fig. 3C). HdhQ109 knock-in mice are asymptomatic at this age15. Phosphorylation of Akt at Ser473 was not affected (Fig. 3C) and pan-JNK immunoreactivity was comparable, regardless of genotype. However, pJNK immunoreactivity was significantly greater in lysates from heterozygous and homozygous HdhQ109 knock-in mice, relative to wild-type (Fig. 3C). Cortical lysates gave similar results (data not shown). P54 and p46 were differentially activated as observed in transfected cells. P54 showed greater increases in pJNK immunoreactivity than p46 in HdhQ109 mice relative to wild type in both striatum (Fig. 3E) and cortex (not shown). Consistent with autosomal dominance of HD, JNK activation increased in both heterozygous and homozygous HdhQ109 mice, but activation was greater in homozygotes than heterozygotes.
Immunoreactivity corresponding to p54 and p46 in NSC34 and striatal lysates mainly corresponded to JNK2/3 and JNK1, respectively (Fig. 3D). An activation ratio (Fig. 3E) was defined as the ratio of pJNK/JNK immunoreactivities for JNK1 (p46) or JNK2/3 (p54). Both JNK1 and JNK2/3 activation ratios were elevated in heterozygotes and homozygotes relative to wild-type brain. JNK1 activity in wild-type mice was 1.097±0.146 (mean ± SEM), whereas JNK1 activity was 1.527±0.061 in homozygotes and 1.342±0.133 in heterozygotes HdhQ109 mice (n = 4 in all cases). Differences in JNK1 activity between wild-type and HdhQ109 homozygotes was significant at p≤0.05 in a pooled t-test, but failed to reach significance between wild-type and HdhQ109 heterozygotes. JNK2/3 activity for wild-type mice was 0.242±0.012 (mean ± SEM), whereas JNK2/3 activity was 0.590±0.055 for homozygotes and 0.440±0.035 for heterozygotes (n = 4 in all cases). JNK2/3 activity increased in both homozygote and heterozygote HdhQ109 mice relative to WT mice and differences were significant at p≤0.01 in a pooled t-test. Both JNK1 and JNK2/3 activities were increased by polyQ-Htt, but polyQ-Htt had a greater effect on JNK2/3 than JNK1.
JNK2/3 is selectively activated in neurons by stress in the presence of constitutive JNK1 activity32, 33. Based on differential activation of JNKs (Fig. 3A–B), effects of specific JNK isoforms on FAT were evaluated. SB203580 differentially affects JNK1 and JNK2/3 in vitro24. At 10μM concentration, SB203580 preferentially inhibits JNK2/3; JNK1 inhibition requires a 10-fold higher concentration23. Axoplasmic JNK activity was almost eliminated by 100μM SB203580, but significant JNK activity remained with 10μM SB203580 (Fig. 4A). However, 10μM SB203580 completely prevented inhibition of FAT by polyQ-Htt (Fig. 4B), suggesting that polyQ-Htt effects involved JNK2 or JNK3. Significantly, 10μM SB203580 also prevented inhibition of FAT by polyQ-AR7. JNK isoforms were distinguished using biochemical approaches. JSAP1 is a scaffold protein regulating JNK activation34. Four JSAP1 isoforms were identified (a,b,c and d), all containing a 17-amino acid JNK binding domain (JBD) homologous to the JIP peptide (Fig. 4C). A 31 amino acid insert in JSAP1c/JSAP1d selectively reduces binding to JNK334. As a result, JSAP1d binds JNK3 with lower affinity than JSAP1a, and inhibits JNK3 less efficiently34. Recombinant JSAP1a (amino acids 115–233) and JSAP1d (amino acids 115–264) polypeptides were expressed and purified, then co-perfused at 5μM with polyQ-Htt (Fig. 4D). Perfusion of JSAP1a and JSAP1d alone had no effect on FAT (not shown). As with JIP peptide (Fig. 2E), co-perfusion of polyQ-Htt and JSAP1a completely prevented inhibition of FAT by polyQ-Htt (Fig. 4D). In contrast, JSAP1d reduced, but did not eliminate polyQ-Htt effects on FAT (Fig. 4E), suggesting JNK3 mediates polyQ-Htt effects in FAT.
Effects of specific JNK isoforms on FAT were evaluated in isolated axoplasm (Fig. 5). The activity of recombinant JNKs was normalized in vitro using c-Jun as substrate (data not shown). JNK1 (200nM) did not affect FAT (Fig. 5A), consistent with reports of constitutive JNK1 activity in neurons32, 33. JNK2 (100nM) decreased anterograde FAT rates slightly (1.25 μm/sec, compared to 1.6 μm/sec for control buffer; P≤ 0.01, two-sample t-test) (Fig. 5B), but did not affect retrograde FAT (Fig. 5D). As with polyQ-Htt perfusion (Fig. 5D), JNK3 (100nM) inhibited both directions of FAT (Fig. 5C). Quantitative analysis showed that JNK3 inhibited both anterograde (0.9 μm/sec; compared to 1.6 μM/sec in control buffer) and retrograde (1.25 μm/sec, compared to 1.4 μm/sec in control buffer) FAT at p≤ 0.01, two-sample t-test, suggesting that JNK3 is principally responsible for polyQ-Htt effects on FAT.
JNK3 phosphorylated kinesin-1, but not KLCs7. Using liquid chromatography tandem mass spectrometry (LC/MS/MS) we mapped residues in kinesin-1 phosphorylated by JNK3 (Fig. 6 and Supplemental Fig. 2). Kinesin-1 constructs encompassing the first 584 amino acids of kinesin-1C (KHC584) were phosphorylated in vitro with JNK3 (Fig. 6B)7 and compared to non-phosphorylated KHC584. Phosphorylated polypeptides were digested with trypsin directly for LC/MS/MS analysis or for phosphopeptide enrichment by IMAC35 and LC/MS/MS analysis (Supplemental Fig. 2). Identification of mass spectrometry-generated peptides used the SEQUEST algorithm. KHC584 identity was confirmed by identification of multiple kinesin-1C peptides (coverage of 72%) (Supplemental Fig. 2). Residual phosphorylation of several peptides was detected (not shown), but using analysis of protease cleavage sites, MS spectra, cross correlation (X-corr) as well as delta correlation (dCn) values, a unique peptide corresponding to amino acids 173-188, was identified with unequivocal evidence of phosphorylation by JNK3 (Fig. 6A, and Supplemental Fig. 2). Phosphatase treatment reduced the phosphorylated peak intensity compared to neighboring peaks (not shown). This JNK3-phosphorylated peptide was within the kinesin-1c motor domain and contained two serines (Fig. 6A). LC/MS/MS allowed precise mapping of the specific serine residue carrying the phosphorylation. Tandem mass spectrometry analysis by collision-induced dissociation (CID) indicated that phosphorylation occurs on Ser176, but not Ser175 (Fig. 6 and Supplemental Fig. 2). Presence of a proline residue adjacent to Ser176 (a hallmark of JNK substrates) supported identification of Ser176 as the JNK3 phosphorylation site. Mass spectrometric analysis of JNK1-phosphorylated KHC584 did not reveal phosphorylation at this site (Supplemental Fig. 3), consistent with differential effects of JNK1 and JNK3 on FAT (Fig. 5A, C). To confirm ESI-LC/MS/MS analysis, site-directed mutagenesis substituted alanine for either Ser175 or Ser176 of KHC584 (KHC584-S175A and KHC584-S176A). Bacterially-expressed mutant proteins were purified (Fig. 6B), then phosphorylated in vitro with JNK3, and 32P incorporation analyzed by Phosphorimaging. Phosphorylation of KHC584-S176A, but not KHC584-S175A, was reduced relative to wild-type KHC584 by≈30% (KHC584-WT, Fig. 6B). Ser176 is conserved among squid, mouse and human kinesin-1s (Supplemental Fig. 4A). Levels of 32P incorporated in KHC584-S176A were consistent with mass spectrometry findings showing partial phosphorylation of KHC584in vitro at non-conserved residues outside the motor domain (not shown). Ser176 phosphorylation was also detected in kinesin-1 immunoprecipitated from mouse brain (not shown). Together, these data identified Ser176 as a major kinesin-1 residue phosphorylated by JNK3.
Effects of polyQ-AR on FAT involved JNK activation, increased kinesin-1 phosphorylation and reduced microtubule binding of kinesin-17. Ser176 localizes to loop 8, a sequence within the motor domain of kinesin-1 implicated in kinesin-1 interaction with microtubules (Supplemental Fig. 4B)36, suggesting its phosphorylation might affect kinesin-1 binding to microtubules. NSC34 cells were transiently transfected with WT-Htt, or polyQ-Htt (see Fig. 3A). Total kinesin-1 levels were comparable for untransfected, WT-Htt, and polyQ-Htt-expressing cells (Fig. 7A). However, microtubule-binding assays using the non-hydrolysable ATP analog AMP-PNP7 revealed a reduction in kinesin-1 binding to microtubules in polyQ-Htt-expressing cells relative to WT-Htt-expressing cells (Fig. 7A). Similar effects were observed when using immortalized striatal cell lines derived from wild type and HdhQ109 mice (not shown). This was consistent with reduced kinesin-1 binding to microtubules in cells expressing polyQ-AR7.
Recombinant KHC584 was incubated with 100μM radiolabeled 32P-γATP in the presence or absence of active JNK3 (Fig. 7B), then incubated with purified microtubules in the presence of AMP-PNP (2 mM) or 2 mM ATP (to control for non-specific pelleting, not shown). Microtubule pellets (P) and corresponding supernatants (SN) were analyzed by immunoblot to visualize non-phosphorylated KHC584 and autoradiography to visualize JNK3-phosphorylated KHC584. In the presence of AMP-PNP, kinesin-1 forms a rigor structure with microtubules13. Virtually all non-phosphorylated KHC584 partitioned with microtubules by immunoblot (Fig. 7B), but a large fraction of JNK3-phosphorylated KHC584 remained in the SN (Fig. 7B), suggesting that phosphorylation of kinesin-1 by JNK3 inhibits kinesin-1 binding to microtubules. The presence of JNK3-phosphorylated KHC584 in P fractions was consistent with residual phosphorylation at residues outside the motor domain. JNK-3 phosphorylation of kinesin-1 reduced binding to microtubules with AMP-PNP by ≈50% (Supplemental Fig. 5).
Effects of S176 phosphorylation on kinesin-1 motility in vivo were assessed by expressing GFP-tagged, truncated kinesin-1 constructs in cultured hippocampal neurons (Fig. 8). When expressed in hippocampal cells, a GFP-tagged kinesin-1 construct (KHC560-GFP -WT) selectively translocates to axon distal ends and accumulates, but not in dendrites, indicating that this motor translocates preferentially along axonal microtubules37. KHC560-GFP-WT translocation is so efficient that little or no fluorescence can be detected within cell bodies or along axons37. To evaluate whether S176 phosphorylation affects kinesin translocation efficiency, we compared localization of phosphorylation-mimicking construct KHC560-GFP-S176E and its unphosphorylatable counterpart KHC560-GFP-S176A to that of KHC560- GFP-WT by quantitative fluorescence microscopy (Fig. 8A–F). Much less phosphomimetic KHC560-GFP-S176E construct accumulated at axonal tips than the KHC560-GFP-WT construct (56 ± 2% vs. 87 ±3%, respectively, mean ± SEM; T-test, p < 0.0001) (Fig. 8G), with more KHC560- GFP- S176E fluorescence in cell bodies and faint staining of neurites (Fig. 8C), the expected pattern if a significant fraction of phosphomimetic kinesin-1 were distributed like a soluble protein. Slightly less KHC560-GFP-S176A accumulated at axon tips (79% ± 2%) compared to KHC560- GFP-WT, but accumulations were still significantly higher than KHC560-GFP-S176E (t-test, p<0.0001) (Fig. 8G). Thus, a mutation of Ser176 mimicking phosphorylation reduces efficiency of processive kinesin-1 translocation in cultured neurons.
Despite ubiquitous expression of Htt, neuronal cells are uniquely vulnerable to the toxic gain-of-function in polyQ-expansion, suggesting that processes critical for proper neuronal function and survival are selectively altered in HD3. Axonal transport is essential for neurons, due to their large size and complex functional architecture 3. Significantly, alterations in kinesin-1 function and regulation are associated with various human neuropathologies3, 7, 38-40.
Several reports link FAT inhibition to pathogenesis in HD4, 5, 8 10 and other polyQ-expansion diseases3, 7. For example, overexpression of polyQ-Htt in cultured cells10 and Drosophila4, 5 leads to reduced transport and accumulation of vesicular cargos within axons. PolyQ-Htt and polyQ-AR, two otherwise unrelated proteins, similarly inhibit FAT vesicle motility in isolated axoplasm independently of changes in gene transcription or aggregate formation7, 8. Such observations implicate FAT defects in HD pathogenesis, which would partially explain the selective vulnerability of neurons in HD and other polyQ-expanded diseases3.
Several models were proposed to explain the inhibition of FAT induced by polyQ-Htt. Some require physical interactions between Htt and conventional kinesin41 or CDyn subunits9, 10, 12. PolyQ-Htt aggregates were proposed to sequester molecular motors4 6, or physically block FAT5, 42. Alternatively, Htt mutations were suggested to interfere with normal functions of Htt as a scaffold linking molecular motors to transported cargoes9, 10 11, or to affect the subunit composition and function of molecular motors12. However, axoplasm experiments indicated that polyQ-Htt inhibited FAT at concentrations 100-1000 fold lower than molecular motors, suggesting activation of enzymatic activities3, 7, 8.
Cell fractionation and immunoprecipitation evaluated interactions between Htt and molecular motors in brain tissue from wild-type and homozygous HdhQ109 mice. Regardless of genotype, similar levels of full-length Htt were found in detergent-soluble (S) and detergent-insoluble (I) fractions. In contrast, KHCs, KLCs, DIC and DHC levels were much higher in soluble fractions, suggesting that most molecular motors are not sequestered by polyQ-Htt. Unlike experiments involving Htt overexpression, immunoprecipitation failed to detect interactions between endogenous Htt and kinesin41 or CDyn12 (Fig. 1). Detection of KLCs in KHC- and DHC in DIC-immunoprecipitates confirmed that physiologically relevant protein complexes were preserved13, 14. Such observations are inconsistent with models in which polyQ-Htt inhibits FAT through direct interactions with molecular motors.
Another model proposed that pathogenic polyQ-Htt activates neuronal regulatory pathways that affect FAT3. Kinesin-1 and CDyn are regulated by phosphorylation of specific motor subunits7, 18. Kinase activities that regulate FAT, including GSK-343; Akt44, and JNK20, 22, 45 are altered in Huntington’s and polyQ-AR activates JNK7. JNK inhibitors blocked polyQ-Htt effects on FAT and JNK activation was documented in polyQ-Htt-expressing cells and HdhQ109 mice (Fig. 3). JNK activation was more pronounced in homozygous HdhQ109 mice than heterozygous littermates, but increased in both, consistent with autosomal dominant inheritance of HD. JNK activation in presymptomatic HdhQ109 mice indicated that this change represents an early event in HD.
Mammals have three JNK genes (JNK1, JNK2 and JNK3)46. Analysis of JNK in HdhQ109 mice and polyQ-Htt-transfected cells indicated that polyQ-Htt increased JNK2/3 activity more than JNK1 (Fig. 3), consistent with reports showing constitutive JNK1 activity and selective JNK2/3 activation in neurons with stress32, 33. This prompted us to examine roles for specific JNK isoforms mediating polyQ-Htt effects on FAT. SB203580 inhibits JNK2/3 at 10μM, but JNK1 inhibition requires ≥100μM23. Significant JNK activity (presumably JNK1) remains in axoplasms treated with 10μM SB203580 (Fig. 4A), but 10μM SB203580 prevented polyQ-Htt effects on FAT (Fig. 4B), suggesting that polyQ-Htt effects on FAT involve JNK2/3. JSAP1 polypeptides further supported a role for JNK3, because JSAP1a, which inhibits all JNKs, preserved FAT effectively, but JSAP1d, with a 31- amino acid insert that selectively interferes with JNK3 binding34 did not (Fig. 4D–E). To evaluate the role of individual JNKs on FAT, effects of active recombinant JNK1, JNK2 and JNK3 on FAT were evaluated in axoplasm. Active JNK1 had no effect on FAT (Fig. 5A), consistent with constitutive JNK1 activity in axons32, 33 (Fig. 4A). JNK2 produced a selective, moderate effect on anterograde FAT, but no effect on retrograde FAT. In contrast, JNK3 effects on FAT paralleled those of polyQ-Htt by inhibiting both anterograde and retrograde FAT. Although JNK2 may contribute, JNK3 appears to primarily mediate polyQ-Htt effects on FAT. Notably, JNK3 is preferentially expressed in nervous tissues47 and polyQ-Htt pathology is restricted to neurons.
To provide a molecular basis for FAT inhibition by JNK3, relevant phosphorylation sites were identified. Mass spectrometry mapped JNK3 phosphorylation of kinesin-1 to S176, a conserved residue in the motor domain (Fig. 6A, Supplemental Fig. 2–3). S176 was also modified in kinesin-1 immunoprecipitated from mouse brain (not shown). Mutagenesis studies confirmed that JNK3 phosphorylated Ser176 (Fig. 6B). JNK1 did not phosphorylate kinesin-1 at Ser176 (Supplemental Fig. 3), consistent with unique substrate preferences for JNK348. Ser176 was conserved among squid, mouse and human kinesin-1s (Supplemental Fig. 4A), validating the squid axoplasm model. Collectively, these data demonstrate that JNK3 activity is elevated in brains of HdhQ109 mice and axoplasm perfused with polyQ-Htt (Fig. 3–4); that JNK3 phosphorylates kinesin-1 at Ser176 (Fig.6 and Supplemental Fig. 2–3), and that Ser176 modification is sufficient to inhibit trafficking of kinesin-1 (Fig. 8). Antibodies selective for this epitope are under development and should allow us to quantify the fraction of neuronal kinesin-1 modified by JNK3 in HD brains.
Based on the Ser176 position within the motor domain36, we evaluated effects of JNK3 phosphorylation on kinesin-1 binding to microtubules. JNK3-treated recombinant kinesin-1 and kinesin-1 in polyQ-Htt-transfected cells exhibited reduced binding to microtubules (Fig. 7). Further, pseudophosphorylation at Ser176 dramatically decreased kinesin-1 translocation in vivo (Fig. 8), indicating that Ser176 phosphorylation interferes with kinesin-1 function.
Our results suggest that polyQ-Htt inhibits FAT by activation of axonal JNKs. Additionally, JNK activation by polyQ-Htt is also consistent with changes in transcription49, and JNKs can be pro-apoptotic30 (Supplemental Fig. 6). Significantly, JNK3, and not JNK1, activity mediated polyQ-Htt effects on FAT, providing a partial explanation for selective neuronal vulnerability in HD. JNK1 and JNK2 are ubiquitously expressed, whereas JNK3 is selectively expressed in brain and testes47. Identification of kinesin-1 as a novel JNK3 substrate provided a molecular basis for changes in FAT consistent with neuron-selective pathology of polyQ-Htt.
In sum, neuronal JNK3 is activated by polyQ-Htt. JNK3 in turn phosphorylates kinesin-1, decreasing its ability to bind microtubules and move cargoes in FAT (Fig. 8 and Supplemental Fig. 6). Identification of Ser176-phosphorylated kinesin-1 in normal mouse brain suggests a role of JNK3 in a normal pathway for regulating kinesin and delivering specific MBO cargoes to particular neuronal domains. Increased activation of this pathway by polyQ-Htt would lead to reductions in kinesin-based transport, deficits in synaptic and axonal function, and eventual dying-back neuronal degeneration3. Inhibition of JNK3 sufficed to prevent inhibition of FAT due to polyQ-Htt. These results suggest that elevation of JNK activity and consequent inhibition of FAT may represent primary pathogenic events in Huntington’s disease. Further studies should illuminate molecular components mediating JNK activation by polyQ-Htt. Regardless; inhibition of JNK activity to protect FAT represents a promising therapeutic target for treatment of Huntington’s disease.
The following antibodies were used: anti-KHC (H2 clone, Chemicon)13, anti-pan KLCs (63-90 clone, Chemicon)13, anti-dynein intermediate chain (Santa Cruz #13524), anti-dynein heavy chain (Santa Cruz #9115), anti-phospho Akt (44-622G, Biosource), anti-Htt (2166, Chemicon), anti-pan-JNK (Upstate #06-748), anti-JNK1 (Pharmingen #554268), anti-JNK2 (Cell Signaling #4672), anti-JNK3 (Cell Signaling #2305), anti-phospho JNK (Cell Signaling #9251), anti-GST (Sigma #G7781), anti-phospho c-Jun (Santa Cruz #16312), anti-tubulin (DM1A, Sigma). All antibodies were characterized for specificity against the tissues used (see Supplemental Figure 7). The following secondary antibodies were used: Jackson 111-035-045 HRP-conjugated goat anti-rabbit, Jackson 115-035-146 HRP-conjugated goat anti-mouse IgG, and Jackson 805-035-180 HRP-conjugated bovine anti-goat IgG.
SB203580, SP600125 (JNK inhibitor II), okadaic acid, staurosporine, K252a, 50nM okadaic acid, microcystin and c-Jun were obtained from Calbiochem. The HDAC inhibitor BC-6-2527 was a generous gift from Alan Kozikowski (UIC). Inhibitor stocks were prepared in DMSO. Active JNK kinases were from Upstate (JNK1 #14-327, JNK2 #14-329, and JNK3 #14-501). His-tagged KHC584 protein constructs were expressed in BL21-Codon Plus (DE3) E. coli (Stratagene), and purified using Talon beads (Clontech). Trx-tagged JSAP1 polypeptides were expressed and purified as described before34.
Proteins were separated by SDS-PAGE on 4-12% Bis-Tris gels (NuPage minigels, Invitrogen), using MOPS Running Buffer (Invitrogen) and transferred to PVDF using Towbin buffer supplemented with 10% (v/v) methanol (90 minutes at 400mA using Hoeffer TE22 apparatus). Immunoblots were blocked with 1% (w/v) non-fat dried milk in TBS (25 mM Tris pH 7.2, 2.68 mM KCl, 136.8 mM NaCl). Membranes were incubated with primary antibodies overnight at 4° C in 1% IgG free-BSA (Jackson Immunoresearch), and washed four times with 0.1% Tween-20 in TBS. Primary antibody binding was detected with HRP-conjugated anti-mouse, anti-rabbit or anti-goat antibodies (Jackson Immunoresearch), and visualized by chemiluminescence (ECL, Amersham). Quantitative immunoblotting was performed as before39.
Mouse brains from wild-type and HdhQ109 knock-in mice were homogenized in lysis buffer (LB; 25 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, and mammalian protease inhibitor cocktail (Sigma, 1/100 dilution)]. Lysates were centrifuged twice for 5 minutes at 55,000 rpm (163,640 gmax) using a TLA 100.3 rotor (Beckman Instruments, Palo Alto, CA). The resulting supernatant fractions were precleared using a mixture of Protein G agarose beads (Pierce), and non-immune mouse IgG-conjugated Sepharose beads (Jackson Immunoresearch) for 1 hour at RT. 400 μg of each precleared brain lysate were brought to 1 ml with LB, and incubated with 5μg of the appropriate antibody plus 10μl of Protein G agarose beads at 4° C for 3hs. Immunocomplexes were recovered by centrifugation (3000 gmax for 30 seconds), washed four times with 1 ml LB, once with 50mM HEPES pH 7.4, and resuspended in Laemmli buffer. For immunodepletion experiments (Fig. 1B and supplemental Fig. 1B), precleared mouse brain lysates were subject to three cycles of immunoprecipitation with H2 or DIC antibodies as above. After each immunoprecipitation cycle, a 50μl aliquot of each supernatant was saved for immunoblot analysis, as before13, 14.
Axoplasm was extruded from giant axons of the squid Loligo pealii (Marine Biological Laboratory) as described previously7, 8. Recombinant proteins, peptides or inhibitors were diluted into X/2 buffer (175 mM potassium aspartate, 65 mM taurine, 35 mM betaine, 25 mM glycine, 10 mM HEPES, 6.5 mM MgCl2, 5 mM EGTA, 1.5 mM CaCl2, 0.5 mM glucose, pH 7.2) supplemented with 2-5 mM ATP and 20 μl added to perfusion chambers. Preparations were analyzed on a Zeiss Axiomat with a 100X, 1.3 n.a. objective, and DIC optics. Hamamatsu Argus 20 and Model 2400 CCD camera were used for image processing and analysis. Organelle velocities were measured with a Photonics Microscopy C2117 video manipulator (Hamamatsu).
Expression vectors encoding amino acids 1–548 of human Htt with an N-terminal FLAG epitope tag were translated in vitro as described previously8, and used for vesicle motility assays. Expression vectors encoding amino acids 1-969 of human Htt (wild type (Q16) and pathogenic (Q46)) with an N-terminal FLAG tag were used for transfection studies (Fig. 3A and and7A7A)29. His-tagged KHC584-S176A, KHC584-S176A, and GFP-tagged KHC560 mutant constructs were generated using the Quick-site mutagenesis kit (Stratagene).
For experiments on kinase activity in axoplasm (Figs. 2C and and4A),4A), 3- 4 axoplasms were triturated in 280μl of KB buffer (KB: 10mM Hepes pH 7.4, 10mM MgCl2, 1mM DTT) minus ATP. Reactions (50μl) were incubated at RT and started by adding ATP (50μM). Aliquots were collected at various time points up to 20 minutes and analyzed by immunoblotting.
Mouse brain tissue and cultured cells were homogenized in ROLB buffer [10mM Hepes pH7.4, 0,5% Triton X-100, 80mM β-glycerophosphate, 50mM Sodium Fluoride, 2mM Sodium Orthovanadate, 100nM Staurosporine, 100nM K252a, 50nM Okadaic acid, 50nM Microcystin, 100mM Potassium Phosphate and mammalian protease inhibitor cocktail (Sigma)]. Lysates were clarified by centrifugation, and protein concentration determined using BCA kit (Pierce). All mouse use was under protocols approved by the Institutional Animal Care and Use Committee at UIC and followed NIH guidelines for vertebrate animal use.
NSC34 cells were plated at 10,000 cells/cm2 on 10 cm tissue culture dishes for biochemical studies and grown on DMEM containing 1mM glutamine and 10% Fetal Bovine Serum (Hyclone). When cells reached 60-70% confluency, media was replaced by DMEM plus 1mM glutamine, and cells transfected with WT-Htt and polyQ-Htt constructs (1-969)29 using Lipofectamine 2000 (Invitrogen), as previously described39. JNK activation was evaluated 24h after transfection.
NSC34 cells were transfected with WT-Htt and polyQ-Htt constructs as described above. After 24 hours, cells were scraped in BRB80 buffer [80 mM PIPES pH 7, 1 mM MgCl2, 1 mM EGTA, 1 μM staurosporine, 1 μM K252a, 50 nM okadaic acid, 200 nM microcystin, 1/100 dilution of phosphatase inhibitor cocktail II (Calbiochem), and1/100 dilution of mammalian protease inhibitor cocktail (Sigma)] at 4°C, and homogenized using a 30g syringe (500μl of BRB80/10 cm dish). Lysates were centrifuged at 14,000 rpm in a tabletop centrifuge for 5 minutes, and centrifuged at 55,000 rpm in TLA100.3 rotor for additional 5 minutes. 700 μg of total protein from clarified supernatants were incubated with 0.2 mg of taxol-stabilized microtubules and 2.5 mM AMP-PNP (Sigma) for 30 min at 37 °C. The mixture above was loaded on top of a cushion made of 20% sucrose in BRB80 buffer, and centrifuged for 15 minutes at 50,000 rpm using a Beckman TLA-100.3 rotor at room temperature. Microtubule-enriched pellets were resuspended in Laemmli buffer and analyzed by immunoblotting. For microtubule binding experiments (Fig. 7B), 4μl aliquots of in vitro-phosphorylated KHC584 (prepared as above) were incubated with 64μg of taxol-stabilized microtubules in BRB80 buffer and either 2mM AMP-PNP or 2mM ATP (100μl total volume). After 30 minutes incubation, these mixtures were centrifuged for 15 minutes at 50,000 rpm over a cushion of 20% in BRB80 in a Beckman TLA-100.3 rotor at 37° C. Pellets and supernatants were collected, and the levels of kinesin in each fraction quantified by phosphorimager scanning of radioactivity or with fluorescently-labeled secondary antibodies.
Accumulation of KHC560-GFP-WT and KHC560-GFP-S176E constructs in neurites of cultured hippocampal neurons was evaluated as before37.
All experiments were repeated at least 3 times. Unless otherwise stated, the data was analyzed by ANOVA followed by post-hoc Student-Newman-Keul’s test in order to make all possible comparisons. Data was expressed as mean ± SEM and significance was assessed at p<0.05 or 0.01 as noted.
Brains from wild-type (12 months old) and HdhQ109 knock-in (14 months old) were homogenized in 3mls of HB buffer (50mM Hepes pH 7.4, 150 mM NaCl, 1% Triton X-100) and centrifuged 2X 2,500g for 10 minutes, and 1X at 7,000g for 10 minutes to pellet debris and nuclei. Clarified lysates were spun at 150,000g for 20 minutes, in a Beckman TLA 100.3 rotor at 55,000 RPM. The pellets and supernatant represented detergent insoluble and detergent-soluble fractions, respectively. Detergent insoluble pellets were resuspended in HB. Equal amounts of protein from each fraction were analyzed by immunoblotting.
In vitro phosphorylation experiments (40 μl volume) were performed by incubating KHC584 protein constructs (3 μM) with 0.1μM JNK3 (Upstate) or 0.2μM JNK1 (Upstate) in HEM buffer (50 mM HEPES, 12 mM MgSO4, 100 μM ATP pH 7.4), as previously described7
Phosphorylated KHC584 protein was subjected to in solution trypsin digestion. Briefly, dried samples were resuspended in 30 mM Hepes, 30 mM NaF in the presence of 1 μg of trypsin (Sigma, proteomics grade) and incubated at 37° C overnight (16-18h). The resulting tryptic peptides were later resuspended in buffer A (5% acetonitrile, 0.4% acetic acid, 0.005% of HFBA in water) for mass spectrometry analysis or prepared for IMAC35. Samples were analyzed by liquid chromatography on line with a LTQ (a two-dimensional ion trap) instrument equipped with a commercial nanospray source (Thermo Finnigan, San Jose, CA). Samples were automatically loaded by a microautosampler (Famos, LC Packings, Sunnyvale, CA) onto an 11-cm × 100μm fused silica capillary column packed with reverse C18 material (5-μm, 100-Å Magic beads, Michrom Bioresources, Auburn, CA). The solvent system was delivered by an HP1100 pump (Agilent Technologies, Palo Alto, CA). Samples were analyzed by performing full scan followed by tandem MS or MS/MS or MS2 of the 5 most intense ions (top 5) by collision induced dissociation (CID). Sample loading, solvent delivery, and scan function were controlled with Xcalibur software (Thermo Finnigan). Liquid chromatography tandem mass spectrometry files were searched using SEQUEST algorithm against a generated database containing KIF5C_Rat and KIF5C_Mouse, among others proteins. SEQUEST search parameters included mass tolerance of ±1.5 Da, trypsin specificity, and differential search for serine, threonine and tyrosine phosphorylation. The generated data set was filtered using INTERACT50 based on the following criteria: delta correlation, dCn: 0.1, X correlation value for +1 peptides ≥ 1.8, +2 ≥ 2.2 and +3 ≥ 3.2.
The authors would like to thank Drs Marcy MacDonald and Marian DiFiglia for knock-in mice and huntingtin constructs, respectively, and Bin Wang for excellent technical assistance. This work was supported by 2007/2008 MBL summer fellowship to GM; an HDSA grant to GM; NIH grants MH066179 to GB; and ALSA, Muscular Dystrophy Association, and NIH (NS23868, NS23320, NS41170) grants to STB.
Author Contributions: GAM and STB performed transport and biochemistry experiments and wrote the manuscript; YY, SLP, AK, KL performed transport and biochemistry experiments; KY provided and characterized recombinant JIP; BB, ETC, CB and DH performed Mass Spectrometry studies; and CFH and GB did GFP-kinesin experiments. All authors reviewed and edited the manuscript