Research Article

Axonopathy and Transport Deficits Early in the Pathogenesis of Alzheimer's Disease

See allHide authors and affiliations

Science  25 Feb 2005:
Vol. 307, Issue 5713, pp. 1282-1288
DOI: 10.1126/science.1105681

Abstract

We identified axonal defects in mouse models of Alzheimer's disease that preceded known disease-related pathology by more than a year; we observed similar axonal defects in the early stages of Alzheimer's disease in humans. Axonal defects consisted of swellings that accumulated abnormal amounts of microtubule-associated and molecular motor proteins, organelles, and vesicles. Impairing axonal transport by reducing the dosage of a kinesin molecular motor protein enhanced the frequency of axonal defects and increased amyloid-β peptide levels and amyloid deposition. Reductions in microtubule-dependent transport may stimulate proteolytic processing of β-amyloid precursor protein, resulting in the development of senile plaques and Alzheimer's disease.

Axons and axonal transport exhibit prominent defects in a wide variety of neurological diseases. These defects often manifest as axonal swellings or spheroids, which correspond to axonal enlargements and aberrant accumulations of axonal cargos and cytoskeletal proteins (1). Molecular motor proteins propel axonal cargoes to and from presynaptic terminals along microtubule tracks and are thus crucial to understanding the role of impaired axonal transport in the pathogenesis of neurological diseases. A number of observations suggest that axonal transport may fail during the progression of Alzheimer's disease (AD) (2). We tested for axonal defects that are diagnostic of transport deficits and that might be related to early stages in the pathogenesis of AD.

AD is a common neurodegenerative disorder characterized by progressive cognitive deterioration and severe synaptic and neuronal loss. Pathological hallmarks of this disease—neurofibrillary tangles, neuropil threads, and senile plaques—are potentially linked to alterations of the axonal compartment (3). Neurofibrillary tangles and neuropil threads are related to the abnormal phosphorylation of the microtubule-associated protein tau and its dislocation from axons to presynaptic terminals and somatodendritic compartments. Senile plaques are composed of dystrophic neurites, some but not all of which are embedded in a matrix of extracellular amyloid. Some dystrophic neurites correspond to axonal swellings, which often contain abnormal accumulations of axonal cargos and tau (4).

Familial AD (FAD) mutations in β-amyloid precursor protein (βAPP) and in presenilin genes (PS1 and PS2) alter the production of amyloid-β peptides (Aβs), the major constituents of senile plaques, suggesting that proteolytic processing of βAPP into Aβs plays a central role in AD pathogenesis (5). In neurons, βAPP and its proteolytic machinery, PS1 and β-site βAPP cleaving enzyme (BACE), undergo kinesin-I–mediated fast anterograde axonal transport (69), during which βAPP may undergo proteolysis into Aβ-bearing intermediates (10) and Aβs (7, 11). Increased Aβ levels (11, 12) and senile plaques (13, 14) in the axon-enriched white matter of mouse model and human AD brains suggest aberrant Aβ generation, or degradation, in axons. Reduced brain white matter in βAPP-deficient mice (15) also indicates that βAPP plays a role in axonal structure and function. Similar phenotypes have been found in a mouse model of AD (16) and in human AD (17), suggesting that axonal failure is an important part of AD. Overexpression of βAPP in Drosophila causes axonal transport defects, which are markedly enhanced by ordinarily benign reductions in the amount of kinesin molecular motor proteins (18). Finally, βAPP and Aβs accumulate in axonal swellings that resemble dystrophic neurites of senile plaques but instead form during brain aging (19) or after traumatic brain injury (20). These observations suggest a link between axonal transport of βAPP, aberrant Aβ generation, and AD pathology but do not untangle cause and effect relationships.

Early axonal defects in mouse AD models and human AD. To test whether FAD mutations of βAPP cause axonal defects, we analyzed a well-studied mouse AD model, Tg-swAPPPrp (21), in which aberrant production of human Aβs results in deposition of senile plaques (22, 23). Neuronal loss in the nucleus basalis of Meynert (NBM) is an invariable feature of AD (24), and alterations of cortical NBM projections are documented in aged primates (25), mouse models of AD (26, 27), and human AD (28). We therefore examined fibers of the NBM, which provide the major cholinergic input to the cerebral cortex.

The organization of the NBM revealed by choline acetyltransferase (ChAT) staining was indistinguishable between 4- and 20-month-old wild-type and Tg-swAPPPrp littermates. Unexpectedly, axonal varicosities in Tg-swAPPPrp exhibited substantial variation in size and morphology. Generally, varicosities correspond to en passant synaptic boutons, are regularly spaced, and have relatively constant diameters. In contrast, varicosities in Tg-swAPPPrp mice were often unusually large and irregularly spaced (Fig. 1A). Some displayed bizarre morphologies or lacked staining in their center, leaving only a ChAT-immunoreactive (IR) silhouette of the varicosity (Fig. 1B). Similar varicosities were occasionally encountered in the NBM of aged Tg-swAPPPrp mice stained for phosphorylated tau (phospho-tau). To assess the microtubule-based transport machinery, we stained ChAT-IR fibers for phosphorylated high-molecular-weight neurofilament protein (phospho-NF-H) and kinesin light-chain subunits of kinesin-I (KLC) (Fig. 1C). Unlike varicosities, colocalization between ChAT, phospho-NF-H, and KLC in axonal swellings in Tg-swAPPPrp was almost complete and very intense.

Fig. 1.

Identification of axonal defects in mouse models and in human AD. (A) ChAT-IR fibers [(a) to (e)] with varicosities (arrows) of unusual size and morphology [(d) and (e)] in Tg-swAPPPrp mouse NBM (scale bar, 10 μm). (B) Some varicosities exhibited ChAT-IR only within their perimeter, giving them a “ghost” appearance (normal varicosities, black arrows; “ghost” varicosities, green arrows; perimeter of “ghost” viscosities in (a) is traced in red in (b); scale bar, 10 μm). (C) Projections of stacks of optical sections of varicosities [(a) to (e)] and swellings [(f) to (j)] in the NBM, demonstrating accumulation of phospho-NF-H (pNF-H) and KLC in ChAT-IR swellings (scale bar, 10 μm). WT, wild-type. (D) Increased number of ChAT-IR fibers with varicosities of diameter larger than 3.0 μm (swellings) in the NBM of 4- (P < 0.005) and 20- (P < 0.05) month-old Tg-swAPPPrp mice (100 fibers per NBM, n = 4 and 5, respectively), compared with wild-type mice (100 fibers per NBM, n = 4 and 5, respectively). (E) Density of ChAT-IR fibers is reduced in the NBM of 20-month-old Tg-swAPPPrp (n = 5) but not wild-type (n = 5) mice (P < 0.05). (F) Swellings [(a) and (b)] and spheroids [(c) and (d)] in the cortex and hippocampus of 19- to 21-month-old Tg-swAPPPrp mice (n = 6) stained with Bielschowsky's silver (scale bar, 10 μm). (G) ChAT- (a) and phospho-tau-IR (b) swellings (arrows) in the NBM of Braak II human AD (scale bar, 10 μm). (H) Increased number of ChAT-IR fibers with swellings in NBM in Braak group I–III (n = 3, P < 0.05) but not Braak group IV–VI (n = 3), compared with Braak group 0 (n = 3) human brains. Some error bars are too small to be visible. (I) Increased number of phospho-tau-IR swellings in NBM in Braak groups I–III (n = 6, P = 0.07) and IV–VI (n = 3, P < 0.05) compared with Braak group 0 (n = 3) human brains.

We quantified NBM ChAT-IR fibers by scoring coded samples for varicosities with diameters significantly larger than average in 4- and 20-month-old wild-type and Tg-swAPPPrp mice (21). ChAT-IR varicosities of diameters larger than 3.0 μm were found almost exclusively in Tg-swAPPPrp mice (Fig. 1D) and were paralleled by significant decreases in ChAT-IR fiber density in the NBM in 20-month-old Tg-swAPPPrp mice compared with the wild type (Fig. 1E). To test whether abnormal varicosities exist in other mouse AD models, we examined Tg-sw/lonAPPThy1 mice and found a significant increase in ChAT-IR varicosities of diameters larger than 3.0 μm in 2.5-month-old Tg-sw/lonAPPThy1 mice compared with wild-type littermates (fig. S1).

To explore whether comparable swellings existed in other brain regions relevant to AD, we scored coded (21) Bielschowsky's silver-stained brain sections from 19- to 21-month-old wild-type and Tg-swAPPPrp mice. Cortices and hippocampi of Tg-swAPPPrp, but not wild-type, mice harbored frequent large swellings that resembled the ChAT-IR swellings in the NBM (Fig. 1F). Although morphologically similar to dystrophic neurites, ∼50% of all swellings were found in brain areas devoid of amyloid (fig. S2).

To test for similar defects in AD, we probed for changes in ChAT- and phospho-tau-IR fibers in the human NBM. Brains were grouped for analysis into Braak AD stages 0, I–III, and IV–VI. We found a subset of ChAT- and phospho-tau-immunoreactive (phospho-tau-IR) fibers with swellings of morphology and diameters resembling those observed in the mouse model of AD (Fig. 1G). Differences in ChAT-[Krustal-Wallis (K-W) test, P = 0.021] (21) and phosphotau-IR (K-W test, P = 0.00021) swellings between the Braak groups were significant. Compared to Braak group 0 (Fig. 1H), ChAT-IR swellings were common in Braak group I–III (up to 40 to 50% of all ChAT-IR fibers) but not in Braak group IV–VI. In contrast, the increase in phospho-tau-IR swellings observed in Braak group I–III became significant only in Braak group IV–VI, compared with Braak group 0 (Fig. 1I). The density of phospho-tau-IR fibers was also significantly increased in Braak group IV–VI but not in Braak group I–III, compared with Braak group 0 (fig. S3). Thus, axonal swellings occurred long before detectable amyloid deposition and, unexpectedly, accumulated excessive amounts of neurofilaments and kinesin-I. Formation of ChAT-IR swellings may represent an early event in AD pathogenesis, preceding or triggering the onset of known pathological changes such as aberrant tau phosphorylation.

Axonal swellings contain abnormal accumulations of organelles and vesicles. To evaluate the contents of axonal swellings, we stained NBM samples for ChAT and processed them for electron microscopy (Fig. 2A). We observed 5 to 7 ChAT-IR swollen profiles per section from 4-month-old Tg-swAPPPrp mice and none from wild-type mice. Swellings contained large numbers of organelles and vesicles, were not myelinated, were not associated with postsynaptic densities, and frequently exhibited diameters larger than 3 μm (Fig. 2B). Similar profiles devoid of ChAT-IR were also observed (Fig. 2C). Tg-swAPPPrp mice averaged 8 to 10 such profiles per section, in contrast to wild-type mice in which 2 to 3 were found per section at most. Swellings consisted of haphazardly arranged vesicles, mitochondria, sporadic multi-lamellar bodies, and vacuoles. Some were entirely filled by mitochondria, and most contained dense bodies that resembled those found in axons in early stages of Wallerian degeneration (29) and in dystrophic neurites embedded in amyloid in AD (30). A subset of swellings encountered exclusively in the Tg-swAPPPrp mice displayed characteristics of axonal degeneration, including electron-dense granular axoplasm and a considerable amount of axoplasmic debris of tubular appearance (Fig. 2D).

Fig. 2.

Ultrastructure of axonal swellings in a mouse model of AD. (A) ChAT-IR cell bodies (b) and fibers [arrow in (a); the cell body and fiber boundaries are emphasized with red dotted lines in (c)] in the NBM. Scale bar, (a) 25 μm, (b) and (c) 1 μm. (B) ChAT-IR fibers in 4-month-old Tg-swAPPPrp [(b) to (d), n = 3] but not wild-type [(a), n = 3] mice had large diameters and contained conspicuous amounts of organelles and vesicles (scale bar, 1 μm). (C) Fibers of large diameters displaying an abnormal accumulation of vesicles, organelles, and dense bodies (arrows), but lacking ChAT-IR, in Tg-swAPPPrp mice [(a) and (b); scale bar, 0.5 μm]. (D) Degenerative changes in fibers of Tg-swAPPPrp but not wild-type NBM [(a) and (b); scale bar, 1 μm].

Formation of axonal swellings is enhanced by reduction in kinesin-I. Aberrant accumulation of kinesin-I, organelles, and vesicles within axonal swellings is suggestive of impaired axonal transport. To test whether impairing anterograde axonal transport enhances the formation of axonal swellings, as is seen in Drosophila models of AD (18, 31), we reduced the genetic dosage of KLC1 by 50% in Tg-swAPPPrp mice (figs. S4 and S5). Axonal swellings that were independent of amyloid deposits exhibited a significant increase in average number per brain in Tg-swAPPPrp;KLC1wt/KLC1null compared to Tg-swAPPPrp;KLC1wt/KLC1wt mice(Fig.3, A and B). This increase appeared region-specific and was most prominent in select cortical regions (Fig. 3C). There was no significant difference in the average number of axonal swellings per amyloid deposit between Tg-swAPPPrp;KLC1wt/KLC1wt and Tg-swAPPPrp;KLC1wt/KLC1null mice (fig. S6).

Fig. 3.

Reduction in kinesin-I promotes formation of axonal swellings. (A) Swellings and spheroids (arrows) in Bielschowsky's silver stained sensory cortices of 19-month-old Tg-swAPPPrp;KLC1wt/KLC1wt (c) and Tg-swAPPPrp; KLC1wt/KLC1null (d) but not WT;KLC1wt/KLC1wt (a) or WT;KLC1wt/KLC1null (b) mice (scale bar, 25 μm). (B) Quantification of axonal defects. Nineteen- to 21-month-old Tg-swAPPPrp;KLC1wt/KLC1null mice (n = 6) had significantly increased numbers of swellings per brain (P < 0.05) compared with Tg-swAPPPrp;KLC1wt/KLC1wt mice (n = 7). Swellings in 19- to 21-month-old WT;KLC1wt/KLC1wt (n = 4) or WT;KLC1wt/KLC1null (n = 3) brains were rare. (C) Increased numbers of swellings upon deletion of one copy of KLC1 was region specific: Sensory, visual, and entorhinal cortex, but not cingulate cortex or dentate gyrus, had significant increases in the number of swellings (P < 0.05). (D) Kymographs representative of in vivo transport of APPYFP in KLC1wt/KLC1wt [(a), n = 9] and KLC1wt/KLCnull [(b), n = 14] hippocampal cultures (scale bar, 10 μm). (E) Altered percentage of APPYFP particles traveling in anterograde and retrograde directions in KLC1wt/KLCnull (n = 292, P < 0.005) compared to KLC1wt/KLC1wt (n = 193, P < 0.005) hippocampal cultures.

We directly tested and confirmed that a relatively benign reduction in kinesin-I is sufficient to impair axonal transport of βAPP. We transiently transfected neuronal cultures harvested from KLC1wt/KLC1wt and KLC1wt/KLC1null mouse hippocampi with wild-type βAPP linked to yellow fluorescent protein (APPYFP). In KLC1wt/KLC1null, we observed a significant reduction in the percentage of APPYFP particles undergoing anterograde transport and an increase in the percentage of APPYFP particles undergoing retrograde transport compared with KLC1wt/KLC1wt primary hippocampal cultures (Fig. 3, D and E, and movies S1 and S2). There was no difference in the percentage of stalled APPYFP particles. Similar results were obtained in Drosophila (figs. S7 and S8 and movies S3 to S5). Thus, a phenotypically benign 50% reduction in kinesin-I subunits was sufficient to produce marked impairment in anterograde transport of βAPP accompanied by enhanced formation of axonal swellings, independently of amyloid deposition.

Reduction in kinesin-I increases Ab generation and its intraneuronal accumulation. Although full-length βAPP levels were indistinguishable between Tg-swAPPPrp; KLC1wt/KLCwt and Tg-swAPPPrp;KLC1wt/KLCnull mice (Fig. 4, A and B), levels of Aβs as well as Aβ42/Aβ40 ratios were consistently elevated in Tg-swAPPPrp;KLC1wt/KLCnull mice (Fig. 4, C and D). These increases in the Aβ42/Aβ40 ratio were not a result of genetic background effects and were comparable in magnitude to well-established Aβ42/Aβ40 ratio increases caused by an FAD mutant PS1 transgene (fig. S9). General effects of axonal structure and microtubule organization on Aβ42/Aβ40 ratios were ruled out by analyzing a deletion of the low-molecular-weight neurofilament subunit (NF-L) (fig. S9).

Fig. 4.

Reduction in kinesin-I increases levels of brain Aβs. (A) Immunolabeled Western blot showing that full-length βAPP levels were similar, whereas KLC1 was specifically reduced in 15- to 17-month-old Tg-swAPPPrp;KLC1wt/KLCnull (n = 7) compared to Tg-swAPPPrp;KLC1wt/KLCwt (n = 7) mice. (B) Quantitation of full-length βAPP from Western blots as in (A) in 11- to 24-month-old Tg-swAPPPrp;KLC1wt/KLCwt (n = 15) and Tg-swAPPPrp; KLC1wt/KLCnull (n = 17) mice. (C) Eleven- to 12-, 15- to 17-, and 19- to 24-month-old Tg-swAPPPrp;KLC1wt/KLC1null mice (n = 2, 7, and 8, respectively) had increased levels of brain Aβs, determined by enzyme-linked immunosorbent assay (21), compared with Tg-swAPPPrp;KLC1wt/KLC1wt mice (n = 2, 7, and 6, respectively; P < 0.05). No Aβs were detected in WT;KLC1wt/KLCwt (n = 5) or WT;KLC1wt/KLCnull (n = 3) mice. (D) Eleven- to 12- and 15- to 17-month-old Tg-swAPPPrp;KLC1wt/KLCnull mice (n = 9) had increased Aβ42/Aβ40 ratios when compared with the Tg-swAPPPrp;KLC1wt/KLCwt (n = 9) mice (P < 0.05).

Aβs may accumulate intracellularly in AD brains, possibly before extracellular amyloid deposition (32). We asked if this feature was present in Tg-swAPPPrp and whether it could be enhanced by KLC1 reduction. Antibodies against the C terminus of βAPP and against the N terminus of the Aβ sequence of βAPP primarily stained the somatodendritic compartment throughout the brain, with the latter showing a discrete staining of the corpus callosum (fig. S10). The staining pattern with an antibody against the C terminus of Aβ40 corresponded to that observed for βAPP and was indistinguishable between 17- to 22-month-old Tg-swAPPPrp; KLC1wt/KLC1wt mice and Tg-swAPPPrp; KLC1wt/KLC1null mice. However, an antibody directed against the C terminus of Aβ42 exhibited significantly increased staining within cell bodies in the sensory and entorhinal cortices, but not in the hippocampi or dentate gyri, of 17- to 22-month-old Tg-swAPPPrp; KLC1wt/KLC1null mice compared with Tg-swAPPPrp;KLC1wt/KLC1wt mice (Fig. 5, A and B). Evidence that KLC1 reduction caused accumulation of Aβ42 in axons came from the finding that human Aβ42 and Aβ42/Aβ40 ratios were increased in the sciatic nerves of 11- to 17-month-old Tg-swAPPPrp;KLC1wt/KLC1null compared with Tg-swAPPPrp; KLC1wt/KLC1wt mice. Aβ42/Aβ40 ratios were also significantly higher in the sciatic nerves than in the brains of Tg-swAPPPrp;KLC1wt/KLC1null but not Tg-swAPPPrp;KLC1wt/KLC1wt mice (Fig. 5C). Thus, Aβ42 formation could be enhanced by the reduction of KLC1, such that it accumulates intraneuronally in Tg-swAPPPrp mice at an age that precedes or coincides with the onset of amyloid deposition (23).

Fig. 5.

Aberrant intraneuronal accumulation of Aβs upon reduction of kinesin-I. (A) Staining for Aβ42, counterstained with 4′,6-diamidino-2-phenylindole (DAPI), in 19-month-old sensory [(a) and (b)] and entorhinal [(c) and (d)] cortices of the Tg-swAPPPrp;KLC1wt/KLC1null [(b) and (d)] and Tg-swAPPPrp;KLC1wt/KLC1wt [(a) and (c)] mice (scale bar, 25 μm). (B) Analysis of integrated optical densities (IOD) for βAPP and Aβ in the sensory cortex, entorhinal cortex, pyramidal layer of the CA1, and granular layer of the dentate showed that 17- to 22-month-old Tg-swAPPPrp;KLC1wt/KLC1null (n = 5) mice had greater Aβ42-IOD values in sensory and entorhinal cortices than Tg-swAPPPrp;KLC1wt/KLC1wt (n = 5) controls (P < 0.05). (C) Increased sciatic nerve versus brain Aβ42/Aβ40 ratio in 11- to 17-month-old Tg-swAPPPrp;KLC1wt/KLC1null (n = 3) compared with Tg-swAPPPrp;KLC1wt/KLCwt (n = 3) mice (P < 0.05).

Reduction of kinesin-I enhances amyloid deposition. To determine whether the increased Aβ42/Aβ40 ratio observed upon reduction of KLC1 correlated with accelerated or enhanced amyloid deposition, we examined senile plaques in 18- to 24-month-old Tg-swAPPPrp;KLC1wt/KLCwt and Tg-swAPPPrp; KLC1wt/KLCnull mice (21) (Fig. 6A). The average number of amyloid plaques visualized with thioflavine S and Bielschowsky's silver was significantly increased in Tg-swAPPPrp; KLC1wt/KLCnull compared with Tg-swAPPPrp;KLC1wt/KLC1wt brains (Fig. 6, B and C). Average diameters and volumes of Bielschowsky's silver-stained senile plaques were also significantly increased in Tg-swAPPPrp;KLC1wt/KLCnull compared with the Tg-swAPPPrp;KLC1wt/KLCwt mice (fig. S11). The most marked increase in the number of plaques was visualized by staining for C termini of Aβ40 and Aβ42. Diffuse extracellular Aβ immunoreactivity was found almost exclusively in Tg-swAPPPrp;KLC1wt/KLCnull brains (fig. S12), suggesting that reduction in KLC1 also promotes the formation of noncongophilic and nonfibrillar deposits. The average number of thioflavine-S–positive senile plaques in 11- to 12-month-old Tg-swAPPPrp;KLC1wt/KLCnull mice was comparable to that in 17- to 19- or 20- to 24-month-old Tg-swAPPPrp;KLC1wt/KLCwt controls (Fig. 6D and fig. S13). This number increased in aged mice and far exceeded the number of amyloid deposits in older Tg-swAPPPrp;KLC1wt/KLCwt mice. Enhanced amyloid deposition in Tg-swAPPPrp;KLC1wt/KLC1null mice was most prominent in the same cortical regions as the swellings not associated with amyloid (Fig. 6E). Moreover, a significant correlation between the number of swellings not associated with amyloid and the number of senile plaques was observed in all transgenic mice examined (fig. S14), suggesting that swellings precede and participate in the formation of amyloid plaques.

Fig. 6.

Reduction in kinesin-I accelerates and enhances amyloid deposition. (A) Brain sections from Tg-swAPPPrp (b), but not wild-type (a), mice showed thioflavine S–positive (TS) senile plaques (Nissl-stained visual cortex in blue; scale bar, 50 μm) surrounded by a corona of dystrophic neurites (arrows; scale bar, 25 μm) when stained with Bielschowsky's silver (c). Plaques were also stained for the C terminus of Aβ40 and Aβ42 or against its N terminus, 6E10 [(d) and (e); scale bar, 10 μm); Aβ40 immunoreactivity is visible in the projection (arrow) from the CA1 layer (d). (B) Enhanced senile plaque deposition in sensory cortices (scale bar, 100 μm) of 21-month-old Tg-swAPPPrp;KLC1wt/KLC1null [(e) to (h)] compared with Tg-swAPPPrp;KLC1wt/KLC1wt [(a) to (d)] mice. (C) Increase in average number of senile plaques visualized by four independent stainings in 18- to 24-month-old Tg-swAPPPrp;KLC1wt/KLC1null (n = 8) compared with Tg-swAPPPrp;KLC1wt/KLC1wt mice (n = 7, P < 0.05). (D) Significant overall difference (P < 0.05) in age of onset and progression of amyloid deposition in 11- to 24-month-old Tg-swAPPPrp;KLC1wt/KLC1wt (n = 10) and Tg-swAPPPrp;KLC1wt/KLC1null (n = 13) mice assayed by thioflavine S. (E) Regional differences in amyloid deposition between 18- to 24-month-old Tg-swAPPPrp;KLC1wt/KLC1wt (n = 7) and Tg-swAPPPrp;KLC1wt/KLC1null (n = 8, P < 0.05) mice.

Discussion. Our findings suggest that axonal blockages that interfere with axonal transport occur early in the course of AD and in mouse models of AD. Morphologically, these blockages resembled axonal swellings, a hallmark of axonal injury, and dystrophic neurites embedded in amyloid within senile plaques of AD (4, 30). Unlike dystrophic neurites, which are primarily associated with amyloid deposits in AD (33), axonal swellings preceded amyloid and other disease-related pathology in a mouse AD model by at least a year and were found in brain areas lacking amyloid deposition in the early stages of human AD. Thus, axonal swellings do not form in response to amyloid deposition. Rather, axonal swellings may represent precursors of some form of dystrophic neurites. Reports of gross white matter changes in mice transgenic (16) or deficient (15) for βAPP and in human AD (17) are consistent with this proposal.

Axonal transport deficits were previously found during late stages of AD (34). Our data suggest that axonal transport deficits also play an early, and potentially causative, role in AD. For example, direct relevance to AD of axonal phenotypes caused by mutant βAPP overexpression in Drosophila and mice comes from human studies suggesting that overexpression of βAPP is responsible for the AD pathology observed in Down syndrome (35). Similarly, our samples of early-stage human AD brains exhibited swellings in cholinergic axons of the NBM akin to those observed in mouse AD models. The observation that reduction of kinesin-I dosage causes an increase in the proportion of βAPP undergoing retrograde transport directly correlates with the accumulation of Aβs in cell bodies and amyloid deposition in the vicinity of cell bodies of the entorhinal cortex within the perforant pathway. These findings also suggest that enhanced amyloidogenesis in the entorhinal cortex is not the result of amyloid deposition by afferents terminating in this region.

Together, the chronological and genetic relationships between reduced axonal transport, axonal swellings, Aβ generation, and amyloid deposition suggest strong mechanistic ties among these events. Axonal swellings could form because of impaired axonal transport and promote aberrant Aβ generation. If aberrant Aβ generation occurs locally at sites of blockage, then amyloid deposition may occur as a result of focally increased secretion of Aβs or lysis of Aβ-enriched axonal swellings. Either, or both, of these processes may provide an appropriate substrate for the formation of senile plaques. Indeed, accumulation of Aβs is observed at sites of axonal damage after traumatic brain injury in the central nervous system (20, 36). Alternatively, upon encountering an axonal swelling or blockage, vesicles containing βAPP or Aβs might undergo retrograde transport to the somatodendritic compartment, where aberrant Aβ generation and/or amyloid deposition could occur. This process could be enhanced by genetic reduction of kinesin-I, which we found increased the probability that an APP vesicle entered the retrograde transport pathway. βAPP is also proposed to play a critical role in linking kinesin-I to axonal vesicles (6, 7). If βAPP processing to Aβs occurs at axonal (or dendritic) blockages and causes kinesin-I release from vesicles and reduction of axonal transport (7), then blockages may lead to local stimulation of βAPP processing, which in turn would cause additional vesicle stalling and further local stimulation of βAPP processing. This proposed sequence of events would generate an autocatalytic spiral in which processes leading to axonal blockages and Aβ production become mutually stimulatory. A vicious cycle of axonal blockages and aberrant Aβ generation also provides a rational explanation for early synaptic loss in AD. Such a mechanism would be critical in AD linked to polymorphisms in KLC1 (37) as well as in sporadic AD, which could potentially be initiated by likely differences in age-dependent declines in axonal transport among humans.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5713/1282/DC1

Materials and Methods

Figs. S1 to S14

References and Notes

Movies S1 to S5

References and Notes

View Abstract

Navigate This Article