Bridging the gap between post mortem microscopic and in vivo macroscopic connectivity

Leegaard et al’s article is a worthy summary of recent efforts to map the brain’s connections at various levels of detail and the benefits that we would realize from cross-validating them. Here is one example:

Using the same approach, Axer et al. (2011b) show how 3-D [polarized light imaging] derived fiber orientation vectors can subsequently be used as a basis for high-resolution tractography of fiber tracts, potentially suitable for bridging microscopic and macroscopic connectome representations. The importance of correlating various non-invasive MRI derived measurements to cellular-level morphological data is also emphasized by Annese (2012), presenting the perspective that whole-brain histological maps (Figures 1E,F) created using large-scale digital microscopy spanning several histological modalities will support the analysis and interpretation of MRI-based connectivity studies.

They also have an awe-inspiring figure showing off the results of many different new techniques.

A, B = diffusion MRI in vivo; C, D = detailed fiber architecture via 3-d polarized light imaging (ex vivo); E, F = digital histology of tract stained for myelin; G = combined optogenetics and fMRI; H = knife edge scanning microscopy; I = GUI of the Human Connectome Project; J = data mining to reconstruct hippocampal connections; K = connectivity-based cortical parcelation; L, M = GUI of the Connectome Viewer; N = structural network motifs; O = connectome matrix of the rat brain; P = connectome matrix of the rat hippocampus; Q = a model of connectivity in the developing tapdole spinal cord; doi: 10.3389/fninf.2012.00014 


Leergaard TB, Hilgetag CC and Sporns O (2012) Mapping the connectome: multi-level analysis of brain connectivity. Front. Neuroinform. 6:14. doi: 10.3389/fninf.2012.00014

Morphology is probably not destiny in CA1

Interesting paper by Knafo et al from a couple of months ago that I’ve been meaning to talk about. This group previously made a synthetic peptide (called FGL) that mimics the active site of a cell adhesion molecule involved in neurite outgrowth (NCAM).

When they subcutaneously inject some amount of this molecule into rats, they learn to navigate a water maze faster. So, cognitive enhancement. Not bad.

Down to mechanism. They fix rat hippocampal sections and investigate a variety of neural morphology measures. No significant differences, although the synapse density looks pretty close to significance to me.

left = representative em images showing synapses, right = differences on various types of synapses

Instead, it seems like this peptide is exerting its effects by recruiting more AMPA receptors to the postsynaptic densities of excitatory CA1 cells. Probably their strongest evidence for this is that, in hippocampal slice cultures, more GFP-tagged AMPA receptors are delivered to the synapse 24-36 hrs after the addition of the peptide.

rectification index = ratio of whole-cell clamped responses at -60 and +40 mV; Inf. = cells infected with GFP (green); FGL = memory-enhancing synthetic peptide

So this is an example where neural connectivity, by itself, seems quite unable to explain all the physiology and behavior in their experiments. Of course, it’s possible that concomitant with the functional (but not structural) changes in the hippocampus, there are correlated structural (but not functional) changes elsewhere in the rat’s brain. But that seems less parsimonious.


Knafo S, Venero C, Sánchez-Puelles C, Pereda-Peréz I, Franco A, et al. (2012) Facilitation of AMPA Receptor Synaptic Delivery as a Molecular Mechanism for Cognitive Enhancement. PLoS Biol 10(2): e1001262. doi:10.1371/journal.pbio.1001262

Linking genetic pathways with regional brain changes due to Alzheimer’s

The ADNI has collected longitudinal MRI scans and genotypes from ~ 100 individuals with AD, ~200 with MCI, and ~150 healthy eldery controls, making it quite the boon for Alzheimer’s research. In a paper that caught my eye, Silver et al took this data set and did some cool things with it.

First, they did some genotype filtering, such as removing SNPs with a frequency of less than 10% for the less common allele. This means that both alleles for each SNP they include will be relatively common, which ensured that their subsequent regression would have adequate sample sizes for each group. Next, they map the SNPs to genes, and then they map the genes to genetic pathways.

schematic illustration; unfilled squares/circles = genes/SNPs that do not map to any pathways/genes; orange = SNP that maps to more than one pathway;

To integrate the brain data, the researchers calculated the change from baseline in each voxel in scans separated by 6 and 18 months. They then did a voxel-wise ANOVA on the rate of change to determine which voxels show differences between AD and healthy controls.

What I find amazing is that even using a conservative correction for multiple comparisons, they still found that 148,023 out of 2,153,231 (7%) of the voxels showed a difference between AD cases and healthy controls. That shows you the extent of AD damage; see below for the specific regions most affected.

top = log p-value distribution for differences in rate of change between AD patients and healthy controls; bottom = final set of voxels that passed the p-value threshold;

Finally, they estimate which SNPs and pathways have the strongest associations with changes in AD-affected brain regions. The top three pathways they blame are the chemokine pathway, the jak-stat pathway, and the tight junction pathway.

The first two of these are related to cytokine signaling and thus continue to emphasize the role of inflammation in AD progression. Tight junction proteins have also been associated with AD to explain the loss in BBB integrity. So, although the AD picture remains messy, studies and data sets like this should help.


Silver M, et al. 2012 Identification of gene pathways implicated in Alzheimer’s disease using longitudinal imaging phenotypes with sparse regression. arXiv:1204.1937v1