Genetics: Chromosome 20: ARFGEF2, CSE1L, STAU1

[jnmaciuch]

Sorry for the late reply. Yes they are all negative values so the deletion is predicted to close chromatin and lower active enhancer marks in the myeloma cell lines.

Thanks, appreciate the follow up! That's great, consistent with a non-functional TF binding site (and increased endothelial junction protein expression from that inverse relationship with ARFGEF2, if we want to speculate about functional effect).

 

[jnmaciuch]

So, what does all of this mean? I’m not sure. None of hypothesis’s currently seems more convincing that the others. I plan to look closer the eQTL, coloc and AphaGenome predictions but I suspect we’ll need some experimental data from this region to know what it actually points to. It’s the biggest hit from DecodeME so I think it deserves a follow-up!

This is all fantastic. The immune hits lining up to myeloid cells makes a lot of sense (HL-60 is from acute myeloid leukemia, specifically). From memory (please correct me if I'm wrong), I'm pretty sure the reference database is cell lines and circulating immune cells. So this is exactly what I'd expect to see if the culprit was actually microglia or macrophages.

From the Protein Atlas link it looks like there's a specific population of microglia that also highly express FOXP2. If there's common genomic regulation between FOXP2 and these SNPs, it's possible that both of your analyses are actually pointing to the same place.

 

[V.R.T.]

I suspect we’ll need some experimental data from this region to know what it actually points to. It’s the biggest hit from DecodeME so I think it deserves a follow-up!

This region is starting to look really interesting- does anybody know of any researchers in this area we might be able to get interested?

@Jonathan Edwards do you have any connections with research interests in this region of the genome?

 

[Jonathan Edwards]

@Jonathan Edwards do you have any connections with research interests in this region of the genome?

My impression that although we think this locus is super interesting none of us is quite sure what it is pointing to. Investigators tend either to focus on a type of pathology and any bit of the DNA that might be relevant to that, or they focus on a known gene product and its functions, like BTN2A1. They don't tend to focus on 'regions of genome' I suspect.

I have been looking at other things and am completely lost with this one. @James Cox might be able to shed light on the ramblings of the last 80 posts. Or @jnmaciuch might be able to give us a simple summary of how this might fit with a favourite hypothesis.

If it is microglia aI am increasingly getting involved with neurobiology people so I might have an microgliologist at hand. There are almost certainly a shedload of them at UCL.

 

[Tom Hill]

But as yet we have no real idea where the cells that feel ill, in the way we report, are.

This may have been covered, but in case it hasn’t: a curiosity of PEM is it involves some of the feelings of sickness but not others. Correct me if I’m wrong, pwME/CFS who are reading this, but while PEM typically causes a crummy “it feels like I’m ill” feeling, it doesn’t typically cause fever, “warmth seeking” behaviour or a sudden loss of appetite (which happen when you ARE fighting off a real infection). Certainly never has in my own case.

There was a Harvard study in 2022 that claimed to identify the neurons involved in specifically fever, warmth seeking and loss of appetite in mice, here. The patch of neurons is in the ventral medial preoptic area (VMPO) of the hypothalamus.

If I’ve gotten the implications of that study right, you would need a theory of PEM onset that activates the “I’m feeling ill” neurons of unknown location but not these ones that activate fever, warmth seeking and loss of appetite. If those neurons were right next to each other in humans it would be weird you don’t get simultaneous activations of both, unless there’s some unusual barrier between the two that shields one from signaling? Or we’d have to assume it’s a region far enough away from those fever/loss of appetite neurons?

 

[Jonathan Edwards]

If I’ve gotten the implications of that study right, you would need a theory of PEM onset that activates the “I’m feeling ill” neurons of unknown location but not these ones that activate fever, warmth seeking and loss of appetite.

Yes, I think that has been a widely agreed position in forum discussions over the years.

But note that these are the cells that generate the signals encoding 'I am feeling ill (but not feverish). A different group of cells will get those signals and experience the sense of being ill (sensing something is due to receiving signals, not sending them).

If those neurons were right next to each other in humans it would be weird you don’t get simultaneous activations of both, unless there’s some unusual barrier between the two that shields one from signaling?

The whole point of brain architecture is that signals do not 'leak' to nearby cells. Cells signalling a pain in one finger cannot afford to leak their signals to cells that deal with the next finger. So, no, we do not expect any cross-talk at all in brains. The exception to this is where ephaptic signals like the dopamine in the hypothalamic portal veins, impact lots of cells in an area. But that is unlikely to be a way of imparting specific information. It is more likely a way of sensitising cells to specific signals.

 

[Tom Hill]

Cells signalling a pain in one finger cannot afford to leak their signals to cells that deal with the next finger. So, no, we do not expect any cross-talk at all in brains.

Thanks so much for explaining this. I think I get it now in terms of what we’re looking for (re. how immune signaling will work on the one hand and neuron-to-neuron signaling circuits function on the other).

 

[ScoutB]

This may have been covered, but in case it hasn’t: a curiosity of PEM is it involves some of the feelings of sickness but not others.

Yes, it's an interesting question and we have a bunch of threads on it. Off the top of my head the sickness behavior tag is one way to find quite a few of them. (For what it's worth, many of us don't love the term sickness 'behavior', but it's what the literature uses at the moment.)

Along with the hypothalamus study you mention there's also this one (Brainstem ADCYAP1+ neurons control multiple aspects of sickness behaviour) that found a population of neurons in the brainstem (so maybe a step before the hypothalamus ones) that they could reactive to recapture much of the sickness response.

My sense reading these studies (and a few others) is that there a probably a lot of circuits in the brain governing the response to illness in different overlapping ways, and so there's lots to map before we can begin guessing which are most promising for our pattern.

you would need a theory of PEM onset that activates the “I’m feeling ill” neurons of unknown location but not these ones that activate fever, warmth seeking and loss of appetite. If those neurons were right next to each other in humans it would be weird you don’t get simultaneous activations of both

Agree with the first sentence here. On the second, surprisingly, the take away I've gotten from reading a number of these circuit neuroscience papers is that it's normal for neurons to be sending unrelated or even opposing signals to their next door neighbours, and activating at different times/different situations. [Edit: pretty much trying to say what JE said but didn't feel qualified to speak in generalities haha]

There was an interesting argument I read where a neuroscientist was advocating for optogenetics being better in some situations than lesion studies. He pointed out sometimes you lesion a brain region and it doesn't seem to make much of a difference to the process you thought it was involved in, but with optogenetics we've found out in a couple of these cases that's because the neurons there were working in opposition to each other (to make a refined decision of some sort) and that's why destroying all of them ended up looking roughly neutral via our crude measurements of behavior.

 

[Tom Hill]

Off the top of my head the sickness behavior tag is one way to find quite a few of them. (For what it's worth, many of us don't love the term sickness 'behavior', but it's what the literature uses at the moment.)

Along with the hypothalamus study you mention there's also this one (Brainstem ADCYAP1+ neurons control multiple aspects of sickness behaviour)

Thanks so much for these steers @ScoutB

 

[V.R.T.]

For whatever its worth, while I don't spike a full fever on a thermometer, I often get feverish, cold and shivery and then far too hot, lose my appetite etc during my worst periods of PEM.

But its not every time and only really since I became severe iirc.

 

[Hutan]

Feeling cold, seeking heat and having chills is a regular part of my PEM. But, having an infection of some sort typically is part of my PEM too, and even just getting a cold sore, which is often part of my PEM, is known to come with these symptoms. So, it's a bit hard to unpack.

 

[ME/CFS Science Blog]

eQTL Catalogue analysis
Had a closer look at eQTL that match the DecodeME signal in this region. This uses data from experiments that measured RNA (gene expression) and linked it to the DNA of the participants. eQTL are the DNA letters that change a gene's expression. Often a GWAS hit cannot be matched with an eQTL but if it does, it helps to show which gene it changes and in what cell type or tissue.

I used the eQTLCatalogue which offers multiple of these datasources in the same format through their API. Their latest version also includes the full data for the Genotype-Tissue Expression (GTEx) Project version 10, a popular database that tested multiple tissue that was also used in the DecodeME preprint. I've included all other datasets related to brain and the immune system using the overview they have available here:
eQTL-Catalogue-resources/data_tables/dataset_metadata_r8.tsv at master · eQTL-Catalogue/eQTL-Catalogue-resources

I've only looked at gene expression (quant_method ="ge" or "microarray"). The data is large but it's in parquet and tabix formats that allows you to download only the locus that you need rather than the full file. So that's how I ended up with 259 datasets for 36 genes, giving 3,705 gene × dataset tests. I've tested colocalization with the DecodeME data in this region, which gives the probability (PP.H4) that both results are driven by the same (single) causal SNP. I've pasted the top findings below, the full results are in the excel file attached.



There are multiple datasets x genes that reach a high PP.H4 but given the large number of tests we should be wary of false positives. Most of the eQTL datasets also have a small sample size (often below 300) and findings below the significance threshold of 5*10^-8.

Interestingly, the best match was for LPS-triggered monocytes in the Kim-Hellmuth 2017 study. I've plotted the data below




This is, however, only one of the 21 tests they did using different microarray probes for the gene and e.g. a test 6 hours instead of 90 minutes after LPS administration. These others didn't match that well. There's also another LPS-triggered monocyte dataset (Fairfax 2014) that didn't gave a good match.


The GTEx data for CSE1L in the nucleus accumbens (part of the striatum and basal ganglia) also forms a good match. Nonetheless, plotting the data I often have the impression that all eQTLs are driven by the broad stroke to the right of the core SNPS from the DecodeME finemapping. There's some noise in that so if test multiple of these datasets, sometimes it shifts a bit to the left and seems to match the DecodeME signal. So despite the high PP.H4 I remain skeptical that both point to the same causal SNP and that we really have a match here.

ARFGEF2, the closest gene, is notably absent from the top eQTL results here but based on its position and uncertainty about whether the eQTL really match, I wouldn't write it off.
The core SNPs of the DecodeME signal lie in the introns of a lncRNA (it's called ENSG00000294533) which isn't included in these datasets. But I figured it wouldn't mess up the results (it would simply be an intermediary) unless it indirectly influences a gene that is really far from the locus.

Hope this is useful to someone. It didn't really give me much clarity, unfortunately.

 

[jnmaciuch]

Hope this is useful to someone. It didn't really give me much clarity, unfortunately.

No this is great. You're getting corroboration from both the chromatin assays and the eQTLs that this seems to be an important regulatory island in myeloids and potentially neurons. When we're thinking about this kind of genomic region, we might already be shooting ourselves in the foot by assuming there's only one single SNP or one single affected gene that is the "real" causative association with the disease. And it's already looking for needles in haystacks, so there are plenty of reasons why you might not see the same exact hits across different datasets.

The concern would just be making sure that you're not putting too much weight on a single association that might be a technical artifact. The fact that two different probes in the LPS monocytes are showing the same trend plus another nearby gene in an independent monocyte dataset is already good reason to believe that the links to that cell type are real. And the fact that monocytes and myeloid-derived cell lines saturated the top hits for the cCREs is even better. The repeat associations with brain could be skewed by brain regions being overrepresented in the datasets compared to other tissue types, but if so many other DecodeME genes point to brain this could just be corroboration of that.

 

[Jonathan Edwards]

When we're thinking about this kind of genomic region, we might already be shooting ourselves in the foot by assuming there's only one single SNP or one single affected gene that is the "real" causative association with the disease. And it's already looking for needles in haystacks, so there are plenty of reasons why you might not see the same exact hits across different datasets.

This was something that struck me, having found it very hard to know how one interprets these linked clusters in terms of real life 'causation'. The detail here is beyond me but I am glad to see people beginning to see patterns that might make more sense. With luck, with more data coming in, this will all pay off. I admire the persistent work on this.

 

[jnmaciuch]

This was something that struck me, having found it very hard to know how one interprets these linked clusters in terms of real life 'causation'.

My thoughts here are also driven by the fact that at least two of the nearby genes have very similar cellular functions (ARFGEF2, CSE1L) so it would make sense if the SNPs are in a regulatory region for a suite of genes that get transcribed together downstream of some trigger. Could be wrong about this here but it does happen

I admire the persistent work on this.

Agreed. Thanks for your work on this @ME/CFS Science Blog

 

[V.R.T.]

The most notable results were for immune cells for the SNPs 20:48911205: A: C and 20:48909956:T:G. HL-60 refers to a leukemia cell line.

The immune hits lining up to myeloid cells makes a lot of sense (HL-60 is from acute myeloid leukemia, specifically). From memory (please correct me if I'm wrong), I'm pretty sure the reference database is cell lines and circulating immune cells. So this is exactly what I'd expect to see if the culprit was actually microglia or macrophages.

You're getting corroboration from both the chromatin assays and the eQTLs that this seems to be an important regulatory island in myeloids and potentially neurons.

The fact that two different probes in the LPS monocytes are showing the same trend plus another nearby gene in an independent monocyte dataset is already good reason to believe that the links to that cell type are real. And the fact that monocytes and myeloid-derived cell lines saturated the top hits for the cCREs is even better.

Am I interpreting it right that as well as neurons or astrocytes this locus is pointing us towards monocytes and bone marrow, perhaps macrophages* in the bone marrow?

In which case the question is again how do we image or biopsy bone marrow usefully? Are there new techniques?

I perhaps remember the consensus before was that biopsy can't be done reliably. [Edit: or was that lympth nodes? I think it might have been]

Is the best route forward apart from more genetics a mouse study similar to what @James Cox is doing with CA10 but on both ARFGEF2 and CSE1L?

This locus seems like it could really shed some light on the pathology, perhaps as much or more than the BTN hit. We need to hit the ground running on this if we can imo.

*they're a type of monocyte right? I promise I'll finish that Khan Academy course soon and stop getting so much wrong!

 

[jnmaciuch]

Am I interpreting it right that as well as neurons or astrocytes this locus is pointing us towards monocytes and bone marrow, perhaps macrophages* in the bone marrow?

I’m not sure about the specific links to bone marrow. Are you thinking that because of the HL-60 hits? That’s just a cancer cell line that is relevant here because those cells express many overlapping gene programs with monocytes.

Is the best route forward apart from more genetics a mouse study similar to what @James Cox is doing with CA10 but on both ARFGEF2 and CSE1L?

possibly. In-vitro could work too for some preliminary information. We already have some confirmation from the eQTL and cCRE databases that these loci have effects on chromatin accessibility and gene expression. It might be useful to confirm that these loci also affect gene expression levels in microglia or another brain-relevant cell type.

A mouse model could check if knocking out those genes changes something in the brain during or after infection. It would just be a bit of a leap because you’d need to use humanized mouse models for the specific infections that are most relevant to ME/CFS. But if there’s a strong signal it might still be worthwhile.

they're a type of monocyte right? I promise I'll finish that Khan Academy course soon and stop getting so much wrong!

Monocytes can become macrophages when they migrate into tissue, though a bunch of experiments in mice (and monkeys too if I’m remembering correctly) show that some populations of macrophages seem to be seeded during embryonic development and don’t necessarily come from monocytes in the blood. [Edit: also no shame at all in asking questions or being wrong. I’m in the same boat with you most of the time]

 

[V.R.T.]

m not sure about the specific links to bone marrow. Are you thinking that because of the HL-60 hits? That’s just a cancer cell line that is relevant here because those cells express many overlapping gene programs with monocytes.

Yes it was that and the mention of myeloid. I must have been mistaken. Monocytes more generally is still very promising. I don't think we can categorically say the DecodeME hits all point exclusively to brain anymore!

Thanks for the informative reply. It would be really good if there was a lab somewhere with the resources and motivation to do the experiments you describe.

A mouse model could check if knocking out those genes changes something in the brain during or after infection.

Could it not also look at how monocytes are affected more generally?

 

[jnmaciuch]

Could it not also look at how monocytes are affected more generally?

Sure, though my guess is you won't see anything of note unless there is some active immune response

 

[V.R.T.]

Sure, though my guess is you won't see anything of note unless there is some active immune response

As in an active immune response e.g. autoantibodies or other adaptive immune responses as opposed to something gone awry with innate immune signalling?

Or a response to a pathogen?

Or just any current immune pathology as opposed to a brain only no immune involvement model?