Dropviz: the mouse brain atlas, Saunders et al. 2018.

[ME/CFS Science Blog]

I had a go at matching genetic data on ME/CFS with the gene expression data from Dropviz.

This is the paper that discovered the eccentric medium spiny neuron (discussed here). It's a mouse brain atlas with RNA data from 0.69 million cells. The main publication is Saunders et al. 2018. but there's also a website with visualization and an option to download the data.

DropViz

Paolo Maccalini already tested this dataset in his paper using the meta-analysis of DecodeME and the Million Veterans Program (MVP) data. Using the level 2 classification of Dropviz he found 6 cell types that passed the bonferonni-correction as shown in Table A1.

https://s4me.info/threads/biologica...encing-of-me-cfs-2026-maccallini-et-al.50225/

This uses the FUMA website approach of grouping the data per dissection and controlling for the average gene expression of that dissection. That's fine but it also means that the results for a cell type are influenced by what else is in the dissection.

As a supplementary to this, I want to do analysis that ignores the dissection groupings and controls for the average of the entire dataset, similar to what we did with the Human Brain Atlas. Luckily the FUMA website provides a dataset that groups all dissections and has the data already in the right format. You can download these here:

Functional Mapping and Annotation of Genome-wide association studies

 

[ME/CFS Science Blog]

Level 1
Here's what I got for the level 1 classification which has 88 cell types. The results show again that only the neuron type reaches statistical significance.



It's also clear that the signal isn't confined to one brain region. The signals in the cerebellum and substantia nigra look a bit weaker though.

 

[ME/CFS Science Blog]

Level 2
Level 2 has 565 cell types. I've tried to deduce some info form the markers they have. For example: Gad1Gad2' = GABAergic, while Slc17a7 or Slc17a6 = Glutamatergic. There were significant hits for both.



Here are the ID's of the 13 cell types that reached bonferonni-significance:


Using the annotation file on the Dropviz website, it seems that the eccentric medium spiny neurons had cluster IDs of 13-1, 13-2, 13-3, 13-4, and 13-5. One of these (STR.Neuron_Gad1Gad2_Drd1-Otof-Sorcs1.13_2) was among the 13 significant cell types but the others didn't really stand out.

 

[ME/CFS Science Blog]

Here are the ID's of the 13 cell types that reached bonferonni-significance:

Here's some more info on these cell types by matching them to the annotation file:

ID	Brain Region	Common name	cell_cluster
GP.Neuron_Gad1Gad2-Th_Adora2a-Th.3_9	Globus Pallidus	Striatum, Th+ SPN	GABAergic
STR.Neuron_Gad1Gad2_Drd1-Cxcl14.10_5	Striatum	dSPN, lateral striatum	GABAergic
HC.Neuron_Slc17a7_C1ql2-Penk.4_2	Hippocampus	Dentate Principal cells	glutamatergic
HC.Neuron_Slc17a7_Pvrl3-Grm5.6_2	Hippocampus	CA3 Principal cells	glutamatergic
HC.Neuron_Slc17a7_C1ql2-Cck.4_1	Hippocampus	Dentate Principal cells	glutamatergic
STR.Neuron_Gad1Gad2_Drd1-Nefm.10_4	Striatum	dSPN	GABAergic
PC.Neuron_Slc17a7_Syt6-Sla.1_4	Posterior Cortex	Layer6a	glutamatergic
FC.Neuron_Gad1Gad2_Synpr-Pcdh11x.1_6	Frontal Cortex	Neuron.Gad1Gad2.Synpr-Pcdh11x	GABAergic
GP.Neuron_Gad1Gad2_Adora2a.3_3	Globus Pallidus	Striatum, Indirect spiny projection neuron iSPN	GABAergic
GP.Neuron_Gad1Gad2_Drd1-Nefm.3_1	Globus Pallidus	Striatum, dSPNs, neurofilament state+	GABAergic
STR.Neuron_Gad1Gad2_Drd1-Otof-Sorcs1.13_2	Striatum	eccentric SPN, dSPN-like markers	GABAergic
STR.Neuron_Gad1Gad2_Adora2a-Nefm.11_1	Striatum	iSPN	GABAergic
PC.Neuron_Slc17a7_Syt6-Nefm.1_8	Posterior Cortex	Layer 6a	glutamatergic

One thing that stands out is that the cells from the Globus Pallidus actually also point to the striatum. They are all from cluster 3 and the Saunders et al. 2018 paper notes about this:

Resolving Neuron Types within the Basal Ganglia
....
To define GPe neuron types, we screened markers of global clusters and subclusters for expression in the GPe (GP/NB dataset). GPe neurons were present in cluster 2 (n = 11,103 cells), one of three neuron clusters (Figure 6A). Cluster 1 contained cholinergic neurons (n = 437 cells), and cluster 3 contained neurons of the adjacent striatum and basolateral amygdala (n = 9,847 cells).

So much like the Human Brain Atlas I think there's a potential link here to medium spiny neurons (not just the eccentric ones) in the striatum.

 

[ME/CFS Science Blog]

So much like the Human Brain Atlas I think there's a potential link here to medium spiny neurons (not just the eccentric ones) in the striatum.

Using conditional analysis, three cell types remain as independent:

- the striatal inhibitory medium spiny neurons,
- the dentate principal cells in the hippocampus
- the layer 6a cells in the posterior cortex.

 

[ME/CFS Science Blog]

If I understand correctly layer 6A form a connection between the cortex and thalamus but perhaps also the claustrum and other inner structures.
Layer 6A Pyramidal Cell Subtypes Form Synaptic Microcircuits with Distinct Functional and Structural Properties - PMC

So one hypothesis for ME/CFS could be a disrupted network of cortico-striatal communication which reminded me of this hypothesis from Peter Behan and Abhijit Chaudhuri from the year 2000:

We have introduced and defined the concept of central fatigue for the latter disorders. We have further proposed, with supportive neuropathological data, that central fatigue may occur due to a failure in the integration of the limbic input and the motor functions within the basal ganglia affecting the striatal–thalamic–frontal cortical system.

Fatigue and basal ganglia - PubMed

 

[Simon M]

I appreciate all the thoughts, analysis and research that is going into your work.

I'm also losing the thread a bit (that's my brain, not your explanations.). Is this gene expression work highlighting tissue and cell types in mice and human where DecodeME (etc) highlighted genes are differentially expressed?

And do these expression atlases cover all tissue or are they particularly focused on the brain?

 

[ME/CFS Science Blog]

Is this gene expression work highlighting tissue and cell types in mice and human where DecodeME (etc) highlighted genes are differentially expressed?

It uses tools like MAGMA that convert SNP associations with ME/CFS into gene associations with ME/CFS.

It doesn't only use the SNPs that reach significance but all the results for the millions of SNPs across the genome. It uses their p-value and combines them into a p-value for a gene. It does this by taking all the SNPs that were inside a gene (or within a close window around that gene that you can specify such as 30kb) and combining their p-value into one for the gene. It does this while taking the correlation among SNPs (the LD taken from a reference genome) into account as well as gene size, gene density and minor allele frequency.

There's quite some uncertainty whether those nearby genes are the relevant ones for ME/CFS but because it's done for so many, it tends to pick up on the most important patterns. So what you end up with is a list of genes and a p-value reflecting their association with ME/CFS. You can then use that to see how well it matches gene expression in a particular tissue or cell type.

And do these expression atlases cover all tissue or are they particularly focused on the brain?

Some cover all tissues and when we did those the results for ME/CFS strongly pointed to the brain, and in particular towards neurons in the brain. One to check out is the Finucane et al. 2018 paper which picked out the relevant cell types for many diseases. Blood or immune' for RA and Lupus, central nervous system for schizophrenia, pancreas for diabetes, etc. When ME/CFS is added, it also points to the CNS.



Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types, 2018, Finucane et al. | Science for ME

Same story with other databases with cell types from across the body such e.g. as DESCARTES. Immune cell data from ImmGen didn't show a signal. So that's why it's interesting to focus on brain atlases such as this one. An overview of all theses results is available in this thread:

Post in thread 'Overview of RNA datasets for tissue and cell type enrichment analysis'

Saturday at 6:17 PM

ImmGen data
The Finucane 2018 paper also included an older version of the dataset on immune cells from the Immunological Genome (ImmGen) project.
Immunological Genome Project

Using LDSC, I got similar results to trafalmadorian97, with the top hit reaching a p-value around 0.005 for 292 cell types tested, not really close to significance.

Further discussed in this Thread

• ME/CFS Science Blog

• 

 

[Simon M]

It doesn't only use the SNPs that reach significance but all the results for the millions of SNPs across the genome. It uses their p-value and combines them into a p-value for a gene. It does this by taking all the SNPs that were inside a gene (or within a close window around that gene that you can specify such as 30kb) and combining their p-value into one for the gene. It does this while taking the correlation among SNPs (the LD taken from a reference genome) into account as well as gene size, gene density and minor allele frequency.

There's quite some uncertainty whether those nearby genes are the relevant ones for ME/CFS but because it's done for so many, it tends to pick up on the most important patterns. So what you end up with is a list of genes and a p-value reflecting their association with ME/CFS. You can then use that to see how well it matches gene expression in a particular tissue or cell type.


Some cover all tissues and when we did those the results for ME/CFS strongly pointed to the brain, and in particular towards neurons in the brain. One to check out is the Finucane et al. 2018 paper which picked out the relevant cell types for many diseases. Blood or immune' for RA and Lupus, central nervous system for schizophrenia, pancreas for diabetes, etc. When ME/CFS is added, it also points to the CNS.

View attachment 33038

Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types, 2018, Finucane et al. | Science for ME

Same story with other databases with cell types from across the body such e.g. as DESCARTES. Immune cell data from ImmGen didn't show a signal. So that's why it's interesting to focus on brain atlases such as this one. An overview of all theses results is available in this thread:

Post in thread 'Overview of RNA datasets for tissue and cell type enrichment analysis'

Saturday at 6:17 PM

ImmGen data
The Finucane 2018 paper also included an older version of the dataset on immune cells from the Immunological Genome (ImmGen) project.
Immunological Genome Project

Using LDSC, I got similar results to trafalmadorian97, with the top hit reaching a p-value around 0.005 for 292 cell types tested, not really close to significance.

Further discussed in this Thread

• ME/CFS Science Blog

• 

Thanks for those explanations.

It's striking how ME/CFS stands out, along with schizophrenia, with one tissue type dominating so much.

I'm trying to understand how this creates a vulnerability, an increased risk of ME/CFS , which is what GWAS measure. That's easier to understand with immune issues which more directly links with increased risk of a triggering infection.

Just having a link to fatigue doesn't seem enough. Could it be that there is an increased risk of locking into a pathological brain regulatory loop? Maybe there are two problems, to explain how these differences generate the symptoms of the illness, especially PEM. But also to explainthe increased risk of triggering the disease.

 

[ME/CFS Science Blog]

I'm trying to understand how this creates a vulnerability, an increased risk of ME/CFS , which is what GWAS measure. That's easier to understand with immune issues which more directly links with increased risk of a triggering infection.

Just having a link to fatigue doesn't seem enough. Could it be that there is an increased risk of locking into a pathological brain regulatory loop? Maybe there are two problems, to explain how these differences generate the symptoms of the illness, especially PEM. But also to explainthe increased risk of triggering the disease.

Wrote more about this here:
The symptom signaling theory of ME/CFS involving neurons and their synapses | Science for ME

Don't see the issue with triggering versus maintaining the disease being two separate problems. Suppose the issue is overfiring of a neural network involved in fatigue and other sickness behavior symptoms. The genes involved in creating and regulating that network will likely show up in GWAS because they increase the risk of that pathological state. So mechanisms and risk would be largely the same?

 

[paolo]

I'm trying to understand how this creates a vulnerability, an increased risk of ME/CFS , which is what GWAS measure. That's easier to understand with immune issues which more directly links with increased risk of a triggering infection.

Notably, the increased risk of developing ME/CFS after an infection may be similar in scale to the one observed for psychiatric diseases. I considered these tow studies: Chang 2023 for ME/CFS and Köhler-Forsberg 2018 for mental disorders. Yet, mental disorders are not considered post-infectious diseases.

I think it would be useful to obtain a dataset like the ones used in the above studies, and perform the same calculations for ME/CFS, psychiatric diseases, and autoimmune diseases. Same methods, same dataset.

 

[paolo]

Level 2
Level 2 has 565 cell types. I've tried to deduce some info form the markers they have. For example: Gad1Gad2' = GABAergic, while Slc17a7 or Slc17a6 = Glutamatergic. There were significant hits for both.

Did you use the meta-GWAS for these analyses or DME_1? DME_1 alone reaches bonferroni significance in the L2 DropViz cell-type analysis.

 

[ME/CFS Science Blog]

Did you use the meta-GWAS for these analysis or DME_1?

The meta-GWAS (DME_1 + MVP)