Faculty Bibliography

More than 30 neurodegenerative diseases including Alzheimer disease (AD), frontotemporal lobe dementia (FTD), and some forms of Parkinson disease (PD) are characterized by the accumulation of an aggregated form of the microtubule-binding protein tau in neurites and as intracellular lesions called neurofibrillary tangles. Diseases with abnormal tau as part of the pathology are collectively known as the tauopathies. Methylthioninium chloride, also known as methylene blue (MB), has been shown to reduce tau levels in vitro and in vivo and several different mechanisms of action have been proposed. Herein we demonstrate that autophagy is a novel mechanism by which MB can reduce tau levels. Incubation with nanomolar concentrations of MB was sufficient to significantly reduce levels of tau both in organotypic brain slice cultures from a mouse model of FTD, and in cell models. Concomitantly, MB treatment altered the levels of LC3-II, cathepsin D, BECN1, and p62 suggesting that it was a potent inducer of autophagy. Further analysis of the signaling pathways induced by MB suggested a mode of action similar to rapamycin. Results were recapitulated in a transgenic mouse model of tauopathy administered MB orally at three different doses for two weeks. These data support the use of this drug as a therapeutic agent in neurodegenerative diseases.

Tauopathy in the brain of patients with Alzheimer's disease starts in the entorhinal cortex (EC) and spreads anatomically in a defined pattern. To test whether pathology initiating in the EC spreads through the brain along synaptically connected circuits, we have generated a transgenic mouse model that differentially expresses pathological human tau in the EC and we have examined the distribution of tau pathology at different timepoints. In relatively young mice (10-11 months old), human tau was present in some cell bodies, but it was mostly observed in axons within the superficial layers of the medial and lateral EC, and at the terminal zones of the perforant pathway. In old mice (>22 months old), intense human tau immunoreactivity was readily detected not only in neurons in the superficial layers of the EC, but also in the subiculum, a substantial number of hippocampal pyramidal neurons especially in CA1, and in dentate gyrus granule cells. Scattered immunoreactive neurons were also seen in the deeper layers of the EC and in perirhinal and secondary somatosensory cortex. Immunoreactivity with the conformation-specific tau antibody MC1 correlated with the accumulation of argyrophilic material seen in old, but not young mice. In old mice, axonal human tau immunoreactivity, especially at the endzones of the perforant pathway, was greatly reduced. Relocalization of tau from axons to somatodendritic compartments and propagation of tauopathy to regions outside of the EC correlated with mature tangle formation in neurons in the EC as revealed by thioflavin-S staining. Our data demonstrate propagation of pathology from the EC and support a trans-synaptic mechanism of spread along anatomically connected networks, between connected and vulnerable neurons. In general, the mouse recapitulates the tauopathy that defines the early stages of AD and provides a model for testing mechanisms and functional outcomes associated with disease progression.

The risk of developing tauopathic neurodegenerative disease depends in part on the levels and composition of six naturally occurring Tau isoforms in human brain. These proteins, which form filamentous aggregates in disease, vary only by the presence or absence of three inserts encoded by alternatively spliced exons 2, 3, and 10 of the Tau gene (MAPT). To determine the contribution of alternatively spliced segments to Tau aggregation propensity, the aggregation kinetics of six unmodified, recombinant human Tau isoforms were examined in vitro using electron microscopy assay methods. Aggregation propensity was then compared at the level of elementary rate constants for nucleation and extension phases. We found that all three alternatively spliced segments modulated Tau aggregation but through differing kinetic mechanisms that could synergize or compete depending on sequence context. Overall, segments encoded by exons 2 and 10 promoted aggregation, whereas the segment encoded by exon 3 depressed it with its efficacy dependent on the presence or absence of a fourth microtubule binding repeat. In general, aggregation propensity correlated with genetic risk reported for multiple tauopathies, implicating aggregation as one candidate mechanism rationalizing the correlation between Tau expression patterns and disease.

Stress induces aggregation of RNA-binding proteins to form inclusions, termed stress granules (SGs). Recent evidence suggests that SG proteins also colocalize with neuropathological structures, but whether this occurs in Alzheimer's disease is unknown. We examined the relationship between SG proteins and neuropathology in brain tissue from P301L Tau transgenic mice, as well as in cases of Alzheimer's disease and FTDP-17. The pattern of SG pathology differs dramatically based on the RNA-binding protein examined. SGs positive for T-cell intracellular antigen-1 (TIA-1) or tristetraprolin (TTP) initially do not colocalize with tau pathology, but then merge with tau inclusions as disease severity increases. In contrast, G3BP (ras GAP-binding protein) identifies a novel type of molecular pathology that shows increasing accumulation in neurons with increasing disease severity, but often is not associated with classic markers of tau pathology. TIA-1 and TTP both bind phospho-tau, and TIA-1 overexpression induces formation of inclusions containing phospho-tau. These data suggest that SG formation might stimulate tau pathophysiology. Thus, study of RNA-binding proteins and SG biology highlights novel pathways interacting with the pathophysiology of AD, providing potentially new avenues for identifying diseased neurons and potentially novel mechanisms regulating tau biology.

Many visual and auditory neurons have response properties that are well explained by pooling the rectified responses of a set of spatially shifted linear filters. These filters cannot be estimated using spike-triggered averaging (STA). Subspace methods such as spike-triggered covariance (STC) can recover multiple filters, but require substantial amounts of data, and recover an orthogonal basis for the subspace in which the filters reside rather than the filters themselves. Here, we assume a linear-nonlinear-linear-nonlinear (LN-LN) cascade model in which the first linear stage is a set of shifted ('convolutional') copies of a common filter, and the first nonlinear stage consists of rectifying scalar nonlinearities that are identical for all filter outputs. We refer to these initial LN elements as the 'subunits' of the receptive field. The second linear stage then computes a weighted sum of the responses of the rectified subunits. We present a method for directly fitting this model to spike data, and apply it to both simulated and real neuronal data from primate V1. The subunit model significantly outperforms STA and STC in terms of cross-validated accuracy and efficiency.

Natural sounds exhibit complex statistical regularities at multiple scales. Acoustic events underlying speech, for example, are characterized by precise temporal and frequency relationships, but they can also vary substantially according to the pitch, duration, and other high-level properties of speech production. Learning this structure from data while capturing the inherent variability is an important first step in building auditory processing systems, as well as understanding the mechanisms of auditory perception. Here we develop Hierarchical Spike Coding, a two-layer probabilistic generative model for complex acoustic structure. The first layer consists of a sparse spiking representation that encodes the sound using kernels positioned precisely in time and frequency. Patterns in the positions of first layer spikes are learned from the data: on a coarse scale, statistical regularities are encoded by a second-layer spiking representation, while fine-scale structure is captured by recurrent interactions within the first layer. When fit to speech data, the second layer acoustic features include harmonic stacks, sweeps, frequency modulations, and precise temporal onsets, which can be composed to represent complex acoustic events. Unlike spectrogram-based methods, the model gives a probability distribution over sound pressure waveforms. This allows us to use the second-layer representation to synthesize sounds directly, and to perform model-based denoising, on which we demonstrate a significant improvement over standard methods.

This study examines behavioral and physiological measures of pitch matching in cochlear implant (CI) users who have residual hearing in the contralateral ear. Subjects adjusted the frequency of an acoustic tone to match the pitch percept elicited by electrical stimulation in the other ear, when stimulation was alternating across two ears. In general, the selected acoustic frequencies did not line up perfectly with the center frequencies of the analysis bands corresponding to each stimulation electrode. Similar alternating electro-acoustic stimuli were used to record Auditory Evoked Potentials on 8 NH subjects and 3 CI patients. NH subjects were presented with a fixed tone in one ear, while tones in the other ear varied within a few octaves from the fixed tone. CI patients were stimulated with six different audible tones including their pitch-matched tones, while receiving electrical stimulation in the electrode. N1 latency for NH subjects was minimized when the same frequency was presented to both ears. Similarly, N1 latency for CI patients who are able to pitch match was minimized when the tone was at the pitch matched frequency of the stimulated electrode. These results suggest that N1 latency can be a possible objective measure of pitch matching. (Work supported by NIH/NIDCD 1K25DC010834-01;PI:Tan, PSC-CUNY;PI:Martin, and NIH/NIDCD R01-DC03937;PI:Svirsky.).

Alzheimer's disease (AD) and epilepsy are separated in the medical community, but seizures occur in some patients with AD, and AD is a risk factor for epilepsy. Furthermore, memory impairment is common in patients with epilepsy. The relationship between AD and epilepsy remains an important question because ideas for therapeutic approaches could be shared between AD and epilepsy research laboratories if AD and epilepsy were related. Here we focus on one of the many types of epilepsy, temporal lobe epilepsy (TLE), because patients with TLE often exhibit memory impairment, depression and other comorbidities that occur in AD. Moreover, the seizures that occur in patients with AD may be nonconvulsive, which occur in patients with TLE. Here we first compare neuropathology in TLE and AD with an emphasis on the hippocampus, which is central to both AD and TLE research. Then we compare animal models of AD pathology with animal models of TLE. Although many aspects of the comparisons are still controversial, there is one conclusion that we suggest is clear: some animal models of TLE could be used to help address questions in AD research, and some animal models of AD pathology are bona fide animal models of epilepsy.