Picture an incredibly large city with all its lights on.
Every home, street, gym, cinema, restaurant, and shop in this city has its own lighting systems that enables people to go about their daily lives. When viewed from above, these lights collectively make the entire city glow.
Now imagine that the city’s electrical system becomes fraught with faults, such that every home, street, and building gradually begins to lose power. Over time, it becomes increasingly difficult for citizens to do their washing, drive through the city, and deliver value at their workplace. Eventually, the entire city suffers a complete power outage, wholly disabling the city and all its people.
Now realize that each cell within the brain corresponds to an individual light bulb and that the entire brain represents the city. When the brain is functioning normally, neurons send signals using action potentials to carry out tasks, such as recalling names and memories. However, neurodegenerative diseases (particularly dementias) gradually cause neurons in the brain to “turn off”, just like the city’s lights. Indeed, as the number of these dysfunctional neurons increases, the brain’s overall functioning weakens. This ultimately results in tremendous struggle for affected individuals in thought, memory recall, and the general ability to complete daily activities. However, whilst an electrical circuit can always be repaired or replaced, the damage inflicted by neurological disorders is permanent.
Whilst the incurrence of irreparable deterioration to one’s brain may strike some as an “unfortunate yet extremely rare” prospect, in reality neurodegenerative diseases affect an estimated 80 million people worldwide — individuals who suffer from incurable and debilitating conditions, ranging from memory loss and apathy to anxiety and depression.
Tau Pathology
Tau pathology appears prominently in Parkinson’s disease (PD) and Parkinson’s disease dementia (PDD), is strongly correlated to functional deficits in Alzheimer's disease brains, and makes a presence in dementia with Lewy bodies, where studies have correlated it with cognitive decline and α-synuclein pathological burden. These sources give the impression that multiple pathologies may act additively to influence disease progression and that underlying risk factors for one pathology may confer risk for additional pathologies.
The figure above illustrates the abnormal accumulation of tau protein in neuronal cell bodies (denoted by the arrow) and neuronal extensions (denoted by the arrowhead) in the neocortex of a patient who died with Alzheimer’s disease, where the black bar (bottom right) corresponds to 25 microns.
Recent post-mortem neuropathology studies have demonstrated that patients with severe clinical AD and PDD exhibit elevated levels of pathological tau in an expanded set of brain regions. In fact, the increase pattern of observed tau — beginning in the locus coeruleus, then transentorhinal and entorhinal cortex, before moving through the hippocampus and cortical regions — is indicative of the pathology “spreading”. For instance, pathological tau from human brains injected into non-transgenic (NTG) mice can be internalized into nearby neurons, initiating misfolding and hyperphosphorylation of endogenous mouse tau in a prion-esque fashion, and resulting in the inspection of tau pathology in more regions of the mouse brain connected to the injection site.
However, despite evidence attesting to the very occurrence of tau pathology spread, the exact process through which it does so remains a matter of debate; this disagreement derives from the difficulty in disambiguating the contribution of neuroanatomical connectivity, intrinsic neuronal vulnerability, and spatial proximity of regions. Recent studies based on mathematical modeling of the human brain, though, suggest that anatomical connectivity serves as a strong predictor of brain atrophy or general pathology patterns in neurodegenerative diseases.
Introducing Taupath
In order to clarify how neuroanatomical connections, spatial proximity, and regional vulnerability contribute to the spread of tau pathology through the brain in Alzheimer’s disease and other tauopathies, a methodology for reproducibly quantifying tau pathology can be developed; in this work, researchers seeded tau pathology in 134 regions of the mouse brain via an intracranial injection over 9 months. For simplicity, the computational and network modeling of tau spread are dubbed Taupath.
Special Thanks
This project is a replication of the following paper, the authors of which I owe a debt of gratitude for providing the code and data used, carrying out the experiments described, and informing the corresponding theory.
Eli J. Cornblath et al. (2021). Computational modeling of tau pathology spread reveals patterns of regional vulnerability and the impact of a genetic risk factor. Science.
Quantitative Immunohistochemistry to Evaluate Tau Pathology Spread
In order to investigate mechanisms underlying the spread of tau pathology, researchers used a seed-based model of tauopathy in which AD brain–derived tau was injected into NTG mice, potentially inducing the misfolding of endogenous tau into hyperphosphorylated tau inclusions. By leveraging biochemical sequential detergent extraction of gray matter from AD patient brains, an enriched fraction of paired helical filament (PHF) AD tau was procured.
The schematic above depicts the process through which AD brains with a high burden of tau pathology went through sequential extraction of tau PHFs, where “PBS” and “HS” correspond to phosphate-buffered saline and high salt (respectively).
This method of extraction yielded a final purity of 22.6 to 35.7% tau. The purified tau fraction was found to retain the pathogenic conformation present in human disease and induce the misfolding of tau in mice without the overexpression of tau.
In order to elucidate the means through which tau pathology spreads through the brain, PHF tau from individual extractions in the hippocampus and overlaying cortex was injected into NTG mice.
In order to capture the temporal dynamics of tau pathology spread, the mice’s brains were captured at 1, 3, 6, and 9 months post-injection and then sectioned. Representative sections were ultimately selected and stained for phosphorylated tau pathology throughout the brain.
As described, the figure above outlines the injection (left) and euthanization (middle) of the mice, alongside the sectioning of their brains (right).
Next, in order to quantify pathology as a percentage of each area occupied by pathology, 194 areas from 134 anatomical regions on the selected sections were manually annotated by researchers; these annotations were based on the Allen Brain Atlas (ABA) Common Coordinate Framework, with small sub-regions with minimal pathology grouped together to minimize error and annotation time.
The figure below summarizes the selection of representative sections and annotation of 194 regions for each brain (top), where the scale bar corresponds to a length of 1 mm; in order to reduce selection bias, a second set of nearby sections was annotated in a similar manner (amounting to 16,684 annotations in total), and the average for each region was used in subsequent analyses:
The enlarged image of the annotated supramammillary nucleus (SUM) is shown (left) with the inclusions stained for pS202/T205 tau. Annotating this image (right) allows for the automated quantification of the percentage of area occupied with pathology in specific regions of the brain, and overlaying an analysis mask on the initial image demonstrates this, where the scale bar corresponds to a length of 100 µm.
Quantitative Pathology Mapping Reveals Dynamic Patterns of Tau Pathology Spread over Time
Since imaging of tau pathology in these mice can only be done after death, all data are pseudo-longitudinal (meaning different time points are represented by distinct groups of mice). In spite of this, the observed tau pathology patterns were highly reproducible across cohorts and time points, and were found to continue in a quasi-linear fashion over time in most regions once pathology seeded a region, suggesting similarity in the process of pathology formation across the brain. Moreover, the lack of a plateau in most regions implies that this model is recapitulating the early disease process, prior to the saturation of pathology and profound neuron death.
However, while the process of pathology formation was similar across regions, the delay between injection and pathology formation showed drastic variation. The following figures within this section depict the percentage of area occupied with tau pathology plotted as a function of time, with the green arrows denoting the time point with initial substantial pathology accumulation, data represented as means ± SEM, and the scale bar corresponding to a length of 50 µm.
First, regions at the injection sites (or with high connectivity to the injection sites) developed pathology by 3 MPI. The quantitative pathology plot (left) and images (right) in the figure above demonstrate the ipsilateral supramammillary nucleus’s accumulation of pathology by 3 MPI:
The figure below depicts that additional regions — including CA1 (field CA1 of the hippocampus), iCA3 (field CA3 of the hippocampus), iENTl (entorhinal area, lateral), cDG (dentate gyrus), and cVTA (ventral tegmental area) — showed a similar onset of pathology at 3 MPI:
Second, other regions showed greater delay, with minimal pathology exhibited before 6 MPI. Indeed, the quantitative pathology plot (left) and images (right) in the figure below show that ipsilateral perirhinal area (iPERI) accumulated pathology by 6 MPI:
The figure below portrays that additional regions—including iMS (medial septal nucleus), iPAR (parasubiculum), cCA1 (field CA1 of the hippocampus), cECT (ectorhinal area), and iPRN (pontine reticular nucleus) — showed a similar onset of pathology at 6 MPI:
Third, another group of regions did not develop pathology until 9 MPI. As presented in the quantitative pathology plot (left) and images (right) in the image below, the ipsilateral accessory olfactory bulb (iAOB) was observed to accumulate pathology by 9 MPI:
The figure below illustrates that additional regions—including cPAR (parasubiculum), cPRE (presubiculum), cVISC (visceral area), and cCA2 (field CA2 of the hippocampus)—showed a similar onset of pathology at 9 MPI:
Finally, some regions never developed substantial pathology during the course of the study.
While the aforementioned pathology patterns in individual regions are informative, the many overall patterns throughout the brain are best observed as a heatmap overlaid on the mouse brain. For instance, at 1 MPI, minimal tau pathology was found outside of the injection sites, while at 3 MPI, more pathology had accumulated ipsilateral to the injection site including hippocampal and entorhinal regions. Furthermore, by 6 MPI, pathology had spread to the contralateral hippocampus and associated cortical areas, and ultimately by 9 MPI, the pathology’s continued spread had affected more rostral regions, although certain rostral cortical and contralateral thalamic regions show minimal pathology.
The figure above depicts regional pathology measures plotted on anatomical maps as a heatmap, with the color blue representing minimal pathology, the color white representing moderate pathology, the color red representing substantial pathology, and the color gray representing the absence of tau pathology in white matter regions.
Collectively, these patterns attest to the spread of tau pathology throughout the brain, in a process potentially involving transmission along neuroanatomical connections or other regional factors.
Predicting Tau Pathology Spread by Diffusion through the Neuroanatomical Connectome
Although a substantial body of literature evinces that tau can be released from neurons in an activity-dependent manner and internalized by other neurons, engendering the spread of tau pathology throughout the brain, the exact nature of how this spread occurs remains unclear. The current main hypotheses are that tau pathology spreads between anatomically connected regions of the brain, between physically contiguous regions, or to selectively vulnerable populations of neurons. Since these mechanisms of spread are not mutually exclusive, each of these putative routes were examined to assess their contribution to the observed pathology spread in mice.
Researchers evaluated the ability of a network diffusion model with neuroanatomical connectivity (specifically, brain-map.org) as a scaffold for the prediction of empirical measures of tau pathology over time. Researchers used ABA regions at the injection sites (including iDG, iCA1, iCA3, iRSPagl, and iVISam) as seed regions for the initiation of this model.
Direct connectivity of these regions was observed at the highest level in hippocampal, septal, and entorhinal regions of the brain, and direct connectivity to the injection site was highly correlated with tau pathology measures in these very regions. The model posited that tau spread from the injection site along anatomical connections both in retrograde and anterograde directions, and that the final amount of regional pathology was determined by a weighted sum of these two independent processes.
Indeed, this bidirectional anatomical connectivity model weakly predicted tau pathology at 1 MPI, likely due to minimal spread at this particular point in time, and strongly predicted pathology at 3, 6, and 9 MPI.
The figure above presents predictions of lag tau pathology (x axis) from spread models based on retrograde and anterograde anatomical connections, plotted against log actual regional tau pathology values (y axis) at 1, 3, 6, and 9 MPI. The solid lines in the graphs represent lines of best fit, and the shaded ribbons represent the 95% prediction intervals, and the r and P values for the Pearson correlation between model-fitted values and observed are noted on each plot.
Moreover, this particular bidirectional anatomical connectivity model outperformed a model in which tau spread was proportional to the Euclidean distance between regions. The content of the figure below mirrors that of the image above, with one notable exception: predictions of log tau pathology from spread models are based on Euclidean distance, rather than retrograde and anterograde anatomical connections. In this figure, the five outlying regions in the upper right denote the injection sites, the inclusion or exclusion of which had no measurable effect on the model fit:
Furthermore, rewiring the network to disrupt the connectome but preserve basic network properties — including in-degree and out-degree connectivity — was found to reduce model performance.
In order to further validate the anatomical connectivity model, researchers ensured that the model’s performance was specific to the choice of the experimental injection site, compared with 500 randomly chosen sets of five regions with a mean spatial proximity similar to that of the experimental injection sites.
In fact, the following figure depicts an evaluation of 500 alternate combination of five seed sites for their ability to predict tau pathology spread at 1, 3, 6, and 9 MPI, in order to evaluate the specificity of the seed sites to predict the pathology spread pattern. As depicted below, usage of the actual five injection regions (denoted by the black diamonds) produced among the best fits at all time points:
These results validate the model’s specificity to the experimental injection site. The performance of models using random sites may be partially explained by a generalized additive model relying on three variables: in-projection similarity between actual and alternate seeds, out-projection similarity between actual and alternate seeds, and Euclidean distance between actual injection sites and alternate seed sites.
Finally, the out-of-sample performance of the bidirectional anatomical connectivity model was thoroughly evaluated compared to three other models in which pathology spread was based on either Euclidean distance, anterograde spread alone, or just retrograde spread.
In order to compare the distributions of out-of-sample fits between each of the four models, 500 train-test splits of the mice used were generated. Model parameters in the training set — including diffusion rate constants and regression weights for anterograde and retrograde spread — were then obtained, before model performance in the test set at each time point was evaluated via the spatial Pearson correlation coefficient between the tau pathology (as observed in the test set) and the predicted pathology (as estimated by each model). This analysis validated the bidirectional model’s superiority to both the anterograde and retrograde models, and established that all three connectivity-based models were superior to the Euclidean distance model by 6 and 9 MPI.
The figure below presents distributions of model fits in 500 held-out samples using Euclidean, anterograde, retrograde, and bidirectional models, where fit differences were analyzed by pairwise two-tailed nonparametric tests for different models. All connectivity models outperformed Euclidean distance after 3 MPI, and a bidirectional model was observed to outperform either a retrograde or anterograde model on their own:
Finally, to better gauge whether the inclusion of Euclidean distance could improve the bidirectional model’s predictivity, a model integrating bidirectional connectivity and Euclidian distance was derived and subsequently evaluated. It was found that the bidirectional-alone model performed at least as well as the combined model at all time points. To further capture of the contribution of anterograde and retrograde connections to the bidirectional model, the diffusion rate constant and standardized β from the model were estimated, ultimately deriving a stronger contribution of retrograde connectivity to the spread model in both measures.
These findings strongly indicate that both anterograde and retrograde spread along anatomical connections independently contribute to the propagation of tau pathology, with the latter being predominant.
Model Residuals to Estimate Regional Vulnerability
Other theories posit that tau pathology spread is mediated by intrinsic neuronal vulnerability. Although this theory is less easily parsed for tauopathies than for α-synucleinopathies where dopaminergic neurons appear especially vulnerable, the existing network model of tau pathology spread could be (and was) repurposed to infer regional vulnerability to tau pathology spread.
These estimates were made by averaging the residuals from the bidirectional anatomical connectivity model over 3, 6, and 9 MPI, with 1 MPI being excluded because of the low model predictivity and dissimilarity to residuals from other points in time. While septal, mesencephalic, and caudal cortical regions had greater regional vulnerability, amygdalar, thalamic, and rostral cortical nuclei were more resilient.
In order to given an average regional vulnerability to tau pathology, the residuals between actual tau pathology levels and pathology levels predicted by the bidirectional model were averaged for each region over time and across hemispheres. In the following figure, these average values are plotted as an anatomical heatmap:
Researchers then tested whether Mapt (the gene providing instructions for making a protein called tau) expression from ABA in situ hybridization data was associated with vulnerability. Despite fairly broad expression across the brain — including most gray matter and white matter regions — Mapt showed no association with regional vulnerability estimates; this observation suggests that tau expression is not a major limiting factor in the spread of tau pathology.
Seeking to further investigate gene expression patterns associated with regional vulnerability, a genome-wide search was conducted for genes whose expression patterns (as quantified by the ABA) were spatially similar to the regional vulnerability measure — a search which found that many gene expression patterns were correlated with regional vulnerability.
After adjusting the false discovery rate (FDR) to q < 0.05, researchers found that 20 genes exhibited expression patterns with a statistically significant spatial correlation with relative regional vulnerability to tau pathology.
In the figure above, gene expression patterns that were statistically significantly associated with regional vulnerability are plotted, with heatmap values indicating the similarity of these gene expression patterns to each other. As seen above, the genes cluster into two groups, one of which consisting of 8 genes positively associated with regional vulnerability, and the other comprising 12 genes associated with resilience (in other words, negatively associated with regional vulnerability), as denoted by the brackets at the bottom of the plot.
Indeed, plotting expression against vulnerability of individual genes demonstrates that this association cannot be attributed to outliers.
The figure below plots normalized relative regional vulnerability as a function of normalized Elovl5 expression (left) and normalized Inpp1 expression (right), where the solid line represents the line of best fit, the shaded ribbons represent the 95% prediction intervals, and the r and P values for the Pearson correlation between vulnerability and gene expression are noted on each plot:
Based on these findings, gene expression patterns harbor considerable potential as parameters of regional vulnerability in the computational modeling of tau pathology spread.
In Silico Seeding from Alternate Sites
Beyond inferring mechanisms of tau pathology spread through network modeling, the value of the aforementioned validated network models can be extended through the generation of predictions of tau spread patterns from alternate injection sites, assuming that the rates of spread alongside the contributions of anterograde versus retrograde spread are the same for those injection sites.
For instance, due to the lateral location of the entorhinal cortex (one of the earliest sites with tau pathology outside of the brainstem in humans), the site is difficult to inject reproducibly in mice. However, in silico modeling of tau pathology spread from this injection site shows a more lateralized spreading pattern — one largely affecting hippocampal and parahippocampal regions, and much later, contralateral regions as well.
The following two figures use the diffusion rate constant estimated by model fitting to empirical tau pathology spread to gauge the distinct spreading patterns that arise after injection into alternate sites into two target areas. In both figures, estimated spread is plotted as a heatmap in anatomical space, with blue denoting regions with minimal estimated pathology and red denoting regions with elevated pathology.
In the first case, the entorhinal area represents the injection site, as illustrated in the figure above.
In order to compare tau pathology spread to existing literature and data modeling α-synuclein pathology spread, researchers also modeled a caudoputamen injection site, as illustrated in the second figure below:
While tau pathology is predicted to spread to some conserved regions (such as the substantia nigra and frontal cortical regions), more engagement of thalamic and mesencephalic regions and less engagement of contralateral regions is exhibited by tau pathology than with α-synuclein pathology.
Altered Tau Pathology Patterns of LRRK2-G2019S Mice
The quantitative pathology-network modeling approach already established in NTG mice enables assessment of the dynamics of tau pathology spread. In order to further explore the value of this approach, similar analysis was performed in mice expression the G2019S mutation in LRRK2 (dubbed LRRK2-G2019S mice going forwards); this mutation is considered the most common cause of familial Parkinson’s disease, alongside a common risk factor for idiopathic PD. Although these mice show similar symptoms to idiopathic PD, 21 to 54% patients lack the hallmark α-synuclein Lewy bodies harbored by patients with idiopathic PD. Crucially, prevailing literature indicates that most mice carrying the LRRK2 mutation exhibit tau pathology.
In order to quantify alterations in tau pathology distribution and spread related to LRRK2, quantitative analysis of pathology in LRRK2-G2019S mice was performed. First, it was deduced that in the absence of pathological tau injection, LRRK2-G2019S mice did not accumulate detectable tau pathology up to 12 months of age. However, following pathological tau injection, LRRK2-G2019S mice were found to accumulate tau pathology in similar regions as NTG mice, demonstrating that anatomical connectivity represents the constraint on overall spreading.
The figure above presents regional pathology measures for LRRK2-G2019S mice plotted on anatomical scaffolds as a heatmap, with the color blue representing minimal pathology, the color white representing moderate pathology, and the color red representing substantial pathology.
However, despite this particular observed similarity, clear differences in the regional distribution of tau pathology in LRRK2-G2019S mice were observed.
The figure above plots the fold change between NTG and LRRK2-G2019S mice on anatomical maps as a heatmap, with the color blue representing regions with higher pathology in NTG mice and the color red representing regions with higher pathology in LRRK2-G2019S mice.
In some regions, such as the injected iDG and highly connected iSUM, tau pathology is almost identical in NTG and LRRK2-G2019S mice; by contrast, other regions exhibited elevated levels of pathology in LRRK2-G2019S mice, particularly those requiring extended periods of time to display this pathology.
The figure above plots the percentage of area occupied with tau pathology as a function of time for four different regions (iSUM, cCA1, iAOB, and iDG), demonstrating four distinct patterns of pathology propagation in NTG and LRRK2-G2019S mice. Some regions (such as the iSUM and iDG) exhibited similar tau pathology progression in both NTG and LRRK2-G2019S mice. By contrast, although similar at 1 and 3 MPI, the other regions (the cCA1 and iAOB) showed enhanced pathology in LRRK2-G2019S mice, particularly at 9 MPI. The representative images of the regions plotted on the left were stained for p-tau and are positioned directly adjacent to the plots demonstrating the pathology patterns (right), where the scale bar corresponds to a length of 50 µm.
These results indicate that the G2019S mutation in LRRK2 has some particular effect on the nature of tau pathology spread.
The LRRK2-G2019S Genetic Risk Factor’s Effect on Network Dynamics of Tau Pathology Spread
To further elucidate which parameters of tau pathology spread can be altered, spread in LRRK2-G2019S mice was assessed using network modeling. As the overall quantitative pathology pattern in the prior section suggests, tau spread can be accounted for in LRRK2-G2019S mice by anatomical connectivity. Akin to NTG mice, tau pathology spread in LRRK2-G2019S mice is only moderately well fit at 1 MPI, but shows improved fit at 3, 6, and 9 MPI, confirming that anatomical connectivity is indeed a major factor in driving tau pathology spread in LRRK2-G2019S mice.
The figure above depicts predictions of regional log tau pathology (x axis) in LRRK2-G2019S mice from spread models based on retrograde and anterograde anatomical connections, plotted against log actual regional tau pathology values (y axis) at 1, 3, 6, and 9 MPI.
Researchers subsequently sought to rationalize the regional differences in pathology in LRRK2-G2019S mice. To this end, the relationship of estimated regional vulnerability in NTG mice to the difference in pathology in LRRK2-G2019S mice was first assessed.
A negative correlation between NTG vulnerability and the difference in pathology at 1 and 3 MPI was observed, attesting to a shift in pathology in LRRK2-G2019S mice from vulnerable regions to more resilient regions at these time points. However, at 6 and 9 MPI, the relationship between NTG vulnerability and pathology difference was observed to have diminished. Given the low levels of pathology in NTG and LRRK2-G2019S mice at 1 and 3 MPI, the inverse relationship between the difference in pathology in LRRK2-G2019S mice and NTG vulnerability may reflect early changes in the directionality of movement in the brain, rather than a shift in the vulnerability of regions.
The figure above presents plots of the NTG vulnerability measure against log G2019S/NTG pathology, which are characterized by a negative correlation between the two measures — one most pronounced at the earlier time points of 1 and 3 MPI. In this figure, the solid lines represent the lines of best fit, the shaded ribbons represent the 95% prediction intervals, and the r and P values for the Pearson correlation between G2019S/NTG pathology ratio and NTG vulnerability are noted on the plots.
Researchers next endeavored to determine whether the difference in regional pathology distribution in LRRK2-G2019S mice was related to a difference in spring in anterograde or retrograde directions. In order to infer the mechanisms of network spread affected by the LRRK2-G2019S mutation, the bidirectional models were fit on bootstrapped samples of NTG and LRRK2-G2019S mice to obtain the distributions of model parameters; these notably comprise the diffusion rate constants and regression weights that gauge the relative importance of anterograde and retrograde spread.
Researchers observed no statistically significant difference in the overall fit of NTG and LRRK2-G2019S mice over time. This finding indicates that the network model in question adequately captures tau pathology spread in LRRK2-G2019S mice.
The figure above illustrates distributions of model fit (Pearson r) for fitting data on bootstrapped samples of mice, which conclusively show that NTG and G20 did not differ in model fit.
Furthermore, the overall diffusion rate constant along anterograde connections was found to be no different than that along retrograde connections, indicating that the rates of anterograde and retrograde spread are comparable.
The figure above displays the distributions of diffusion rate constants, which reveal greater inter-sample variability in retrograde compared to anterograde constants. The figure further confirms that LRRK2-G2019S and NTG did not differ in diffusion rate constants.
Finally, it ought to be highlighted that the contribution of anterograde and retrograde connections did show differences over time, insofar as anterograde spread was deemed less important and retrograde spread was deemed more important in accounting for tau spread patterns in LRRK2-G2019S mice.
The figure above spotlights that anterograde and retrograde betas differed between NTG and LRRK2G2019S mice, with LRRK2-G2019S preferentially spreading in the retrograde direction. In the figure’s box plots, the box edges represent the 25th and 75th percentiles, the middle line corresponds to the median, the whiskers extend from the box edges to the most extreme data point that is at most 1.5 times greater than the interquartile range (IQR), and data beyond the end of the whiskers are plotted as individual dots.
These findings indicate that the LRRK2-G2019S mutation could lead to increased shunting of misfolded tau into the already-predominant retrograde pathways — an interpretation which partially explains the differences in pathology patterns observed in the mice.
Closing Thoughts
This work demonstrates how the usage of an interdisciplinary approach bridging quantitative pathology and network analysis enables the prediction of tau pathology patterns via linear diffusion through the anatomical connectome, with modulations of that spread by regional vulnerability.
By leveraging such an approach, it can be shown that tau pathology spreads from a given initial injection site through the brain via neuroanatomical connectivity, and that this spread is preferential in a retrograde direction, despite being constrained by anatomical connectivity. It is further substantiated that this spread can modulated by a genetic risk factor for PD.
Through the provision of a framework for understanding neuropathological progression in tauopathies, this work paves the path to the development of viable therapeutic treatment for an assortment of neurodegenerative diseases.
Special thanks once again to Eli J. Cornblath, Howard L. Li, Lakshmi Changolkar, Bin Zhang, Hannah J. Brown, Ronald J. Gathagan, Modupe F. Olufemi, John Q. Trojanowski, Danielle S. Bassett, Virginia M. Y. Lee, and Michael X. Henderson—the authors of the paper Computational modeling of tau pathology spread reveals patterns of regional vulnerability and the impact of a genetic risk factor.
