A Molecular network approach reveals shared cellular and molecular signatures between CFS and other fatiguing illnesses, 2021, Comella et al

John Mac

Senior Member (Voting Rights)
Full title:
A Molecular network approach reveals shared cellular and molecular signatures between chronic fatigue syndrome and other fatiguing illnesses

Abstract
The molecular mechanisms of chronic fatigue syndrome (CFS, or Myalgic encephalomyelitis), a disease defined by extreme, long-term fatigue, remain largely uncharacterized, and presently no molecular diagnostic test and no specific treatments exist to diagnose and treat CFS patients.

While CFS has historically had an estimated prevalence of 0.1-0.5% [1], concerns of a long hauler version of Coronavirus disease 2019 (COVID-19) that symptomatically overlaps CFS to a significant degree (Supplemental Table-1) and appears to occur in 10% of COVID-19 patients[2], has raised concerns of a larger spike in CFS [3].

Here, we established molecular signatures of CFS and a corresponding network-based disease context from RNA-sequencing data generated on whole blood and FACs sorted specific peripheral blood mononuclear cells (PBMCs) isolated from CFS cases and non-CFS controls.

The immune cell type specific molecular signatures of CFS we identified, overlapped molecular signatures from other fatiguing illnesses, demonstrating a common molecular etiology.

Further, after constructing a probabilistic causal model of the CFS gene expression data, we identified master regulator genes modulating network states associated with CFS, suggesting potential therapeutic targets for CFS.


https://www.medrxiv.org/content/10.1101/2021.01.29.21250755v1
 
CFS patients did an exercise test, with blood samples taken before and on the following 3 days. RNA-seq, followed by differential expression analysis of sorted immune cells showed little difference.

Furthermore, no statistically significant difference between time points was observed in data collected from the Modified Fatigue Impact Scale (MFIS) questionnaire, Karnofsky performance scores, or clinical workups, although patients did report increased physical fatigue (0-10 rating) between timepoint 1 and timepoints 2 and 3

That sounds like a failure to replicate PEM. They suggest it may be due to an insufficiently intense exercise challenge. I'm not sure, maybe they didn't recruit enough participants with PEM.

Viral load distribution differed between cases and controls.


T and B cell clones:
We observed greater read support per clone or a less diverse population of T and B Cell clones in cases compared to controls (p < 0.0001)


They did a complicated analysis and after comparing the results to a database of various other fatiguing illness found this which is interesting:
The top CFS modules are enriched for many of these external disease signatures, such as MIS-C, Lyme disease, and COVID-19, but notably are not enriched for any of the autoimmune signatures (Fig. 4b). Similarities between CFS, COVID-19, and Chronic Lyme Disease have been suggested at a clinical level [3, 24, 25] and our data further suggests a shared molecular etiology among these diseases.


And this is also interesting:
To further characterize these CFS modules, we searched for biological processes and pathways that were also enriched in them (Fig. 4c, Supplemental Tables 8-10). NKM4 was the module most strongly associated with CFS (logOR=1.71, FDR=1.39e-66) and was also enriched for the “recovering COVID-19” signature (Wen Proliferating T-cells, logOR=2.02, FDR= 1.07E-15), highlighting a molecular link that may help explain why so many recovered COVID-19 patients seem to experience CFS-like symptoms [2]. The top biological pathway associated with NKM4 was MAPK Cascade (logFC=4.691, FDR=1.72e-4), consistent with in vitro findings of MAPK dysregulation in NK cells of CFS [26] and COVID19 [27]. Lyme disease signatures were enriched for the greatest number of top CFS modules, with 4 of the top 5 CFS-enriched modules also enriched for the chronic Lyme disease signature, suggesting a potential molecular link between CFS and chronic Lyme disease as has been previously clinically described [28].

They discuss some further similarities between these and other illnesses.

Moving on, they tried to identify "key driver genes".

In conclusion, we present an unbiased data driven, network-based approach that identified molecular signatures of CFS and implicates a number of highly coherent co-expression modules to CFS. For the top 5 modules, the complementary analysis shown here point to a common underlying biology that shares immune and metabolic dysregulation also present in other clinically similar diseases such as Lyme, MIS-C, Kawasaki, and recovering COVID-19. Moreover, the top KDs we identified as regulators of these CFS-associated modules are biased towards higher pLI scores, indicating that loss of function mutations in these genes cannot be well tolerated, confirming their critical importance to normal system function. These top KDs we identified for CFS offer interesting points of therapeutic intervention to explore, with the most promising being MXD1, STX3, DYSF, LYN, MLL2, NCOA2, PTPRE, REPS2, RP11- 701P16.2, TECPR2, and TUBB1
 
Last edited:
What MXD1 does: it seems to suppress the process of making ribosomes (which in turn make proteins).

STX3: involved in exocytosis (a way to transport molecules in bulk out of the cell). Also involved in the growth of axons and dendrites of neurons.

DYSF: involved in skeletal muscle repair. Repairs wounds in the cell membrane. Defects in this gene lead to various types of muscle dystrophy.

LYN: a key role in cell activation. Especially B cells. Also participates in insulin signalling.

MLL2: The protein co-localizes with lineage determining transcription factors on transcriptional enhancers and is essential for cell differentiation and embryonic development. It also plays critical roles in regulating cell fate transition, metabolism, and tumor suppression.

NCOA2: is a transcriptional co-activator of the glucocorticoid receptor and interferon regulatory factor 1 (IRF1).

TUBB1: structural constituent of the cell cytoskeleton.

Too tired to do the rest.
 
Last edited:
Yes, it would be interesting to know more about the cooperation between dr. Enlander and Mount Sinai and the hospital's research in ME and care for ME-patients. If I remember correctly, Enlander was going to open a ward at Mount Sinai due to a donation from a patient.

Also, a few years ago there were some expectations regarding a scientist at Mount Sinai, Erik Schadt, who was entering the ME-field. But then it went silent..

Silent no more...

Curiouser and curiouser.........I emailed Dr Enlander and got a reply:

"I am on the faculty at Mount Sinai medical Center. One of my patients gave me research donation of $1 million. I gave the medical center the donation to form a ME CFS center."
.
 
other fatiguing illnesses or diseases involving a strong inflammatory component such as Multisystem Inflammatory Syndrome in Children (MIS-C), Kawasaki Disease (KD), Macrophage Activation Syndrome (MAS), Neonatal Onset multisystem inflammatory (NOMID), Lyme disease, active Influenza (IAV), active COVID-19, early recovery stage after COVID-19, Mixed Connective Tissue Disease (MCTD), Sjögren’s Syndrome (SJS), Systemic Lupus Erythematosus (SLE), Systemic Sclerosis (SSC), Undifferentiated Connective Tissue Disease (UCTD), Primary Antiphospholipid Syndrome (PAPS) and Rheumatoid Arthritis (RA). The top CFS modules are enriched for many of these external disease signatures, such as MIS-C, Lyme disease, and COVID-19, but notably are not enriched for any of the autoimmune signatures (Fig. 4b). Similarities between CFS, COVID-19, and Chronic Lyme Disease have been suggested at a clinical level [3, 24, 25] and our data further suggests a shared molecular etiology among these diseases.

I still haven't had a deeper look yet.
 
I'm happy that we're finally getting a taste of 21st century science, as well as a comparison with other fatiguing illnesses which smaller ME/CFS research teams often aren't able to provide.

That sounds like a failure to replicate PEM. They suggest it may be due to an insufficiently intense exercise challenge. I'm not sure, maybe they didn't recruit enough participants with PEM.
The participants were well characterized, they had to fulfill 3 diagnostic criteria, 2 of which require PEM:
CFS was formally diagnosed using the Fukuda Criteria (Supplemental Table-4)[19], Canadian Consensus Guidelines [20], and updated international consensus criteria, but excluding any related conditions such as major depressive disorder, Collagen-vascular diseases (CVD), neuromuscular diseases (NMD), and significant cardiac or pulmonary comorbidities (Online Method).
It seems that the moderate intensity of the CPET is the culprit. Maybe they didn't want participants to crash badly? (Glad they were able to get so many timepoints / blood draws though.)
All participants were administered a moderate-intensity cardiopulmonary exercise test (CPET), with whole blood draws occurring immediately before the CPET and then 24, 48, and 72 hours post-CPET (Fig.1a).
However, the methodology for CPET in ME/CFS requires attaining maximal effort according to Workwell (highlights mine) [1]:
The goal of the test protocol is to incrementally challenge energy production such that the patient is able to complete at least 8 min but no more than 12 min of cycling (45). For moderately ill ME/CFS patients who complete this protocol, workload increments of 10–15 W/min, beginning at 0 watts, is appropriate to achieve an 8–12 min test to maximum effort duration. However, for a patient with a significant history of physical training, a 20 or 25 W/min protocol may be appropriate. The same protocol should be used for CPET1 and CPET2.
[...]
Test termination should comply with testing guidelines (45) and is indicated by attainment of maximal effort, or test termination due to patient safety. When testing for evidence of disability, insurers and independent medical examiners will closely scrutinize patient effort. Therefore, criteria for maximal effort should be reported which could include; plateau in oxygen consumption with increases in workload, RPE ≥ 18 (6–20 scale), respiratory exchange ratio (RER) ≥ 1.1, or peak blood lactate ≥ 8 mM. These criteria support evidence of maximum effort during CPET. The RER criterion is generally considered a more valid indicator of patient effort compared to the other indicators (55). Generally, satisfying two of three criteria is acceptable to determine that maximum effort was given by the patient (56).
I think they would have found biological changes had they repeated the CPET, even at moderate intensity. That's been the case for Maureen Hanson's team at Cornell who does 2-day CPET. This was a small study but I wish the investigators had had the possibility (funds) to do a 2-day CPET. I should note that from the text, it seems to me that they're not aware of the difference between single and repeated CPETs? The IOM report says:
By contrast, a single CPET may be insufficient to document the abnormal response of ME/CFS patients to exercise (Keller et al., 2014; Snell et al., 2013). Although some ME/CFS subjects show very low VO2max results on a single CPET, others may show results similar to or only slightly lower than those of healthy sedentary controls (Cook et al., 2012; De Becker et al., 2000; Farquhar et al., 2002; Inbar et al., 2001; Sargent et al., 2002; VanNess et al., 2007).

[1] Stevens S, Snell C, Stevens J, Keller B, VanNess JM. Cardiopulmonary Exercise Test Methodology for Assessing Exertion Intolerance in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome. Front Pediatr. 2018;6:242. Published 2018 Sep 4. doi:10.3389/fped.2018.00242
 
Last edited:
Tiny study:

To provide some insight into the molecular processes that underlie CFS, we carried out a study on 15 patients diagnosed with CFS and 15 age, sex, and BMI matched controls (Supplemental Table-2). CFS was formally diagnosed using the Fukuda Criteria (Supplemental Table-4)[19], Canadian Consensus Guidelines [20], and updated international consensus criteria,

While I like the comparison with other 13 illnesses, I think sample is just way too small to justify such a broad analysis (even though they do seem to have taken multiple comparisons seriously). So I am going to pass on this unless someone gives me a good reason to take a deeper look.
 
It seems that the moderate intensity of the CPET is the culprit. Maybe they didn't want participants to crash badly? (Glad they were able to get so many timepoints / blood draws though.)

The exercise wasn't trivial:

Patients during the first 5 minutes were asked to gradually increase their work rate until reaching 70% of age-predicted maximal heart rate [56], at which point this target heart rate was maintained. Ratings of perceived exertion (RPE) were obtained on a VAS from 1-10 every 5 minutes while undergoing testing (5, 10, 15 and 20min timepoints). Blood pressure measurements and lactic acid measurements (Lactate Pro Portable Analyzer (DKD, Japan) were performed every 5 minutes while exercising. The maximum duration of exercise testing was 25 minutes.

The maximal heart rate function was from another paper:
eMHR = 179 + 0.29 x age - 0.011 x age(2).

This function underestimated my (1st CPET) maximal heart rate by about 25BPM!
70% of the predicted eMHR for me is significantly below the first ventilatory threshold (as measured on 1st CPET), but the workload at that heartrate is not trivial, at least for ME/CFS patients.
 
Furthermore, no statistically significant difference between time points was observed in data collected from the Modified Fatigue Impact Scale (MFIS) questionnaire, Karnofsky performance scores, or clinical workups, although patients did report increased physical fatigue (0-10 rating) between timepoint 1 and timepoints 2 and 3

I think impact scales do not take into account the way you can struggle and still do things if you are feeling worse with ME. There is a point when you can't do anything but it is commoner to do it but just feel bad about it. When I was moderate, I often felt that ME did not stop me doing things it just took all the fun out of them.

The other thing is how much they measured things like sore throats, swollen glands, pain and itches, not to mention cognitive problems.

After all these years I still have trouble describing how I feel if I do too much. Often, I only realise how bad things were when I get back to my more normal state. I am more likely to say I feel fine when I feel bad!
 
The maximal heart rate function was from another paper:
eMHR = 179 + 0.29 x age - 0.011 x age(2).

This function underestimated my (1st CPET) maximal heart rate by about 25BPM!
70% of the predicted eMHR for me is significantly below the first ventilatory threshold (as measured on 1st CPET), but the workload at that heartrate is not trivial, at least for ME/CFS patients.

Out of curiosity I compared the formula they used with the one I have seen recommended: 220 - age

The maximal heart rate function was from another paper:
eMHR = 179 + 0.29 x age - 0.011 x age(2).
Took me a while to work out the 2 in brackets means squared. I'm used to seeing it represented as age^2.

That gives some significant differences between formula, especially at the younger end, with their estimate being lower, for example:

Age 20:
(220 - age) x 0.7 = 140
(179 + 0.29 x age - 0.011 x age^2) x 0.7 = 126

Age 50:
(220 - age) x 0.7 = 119
(179 + 0.29 x age - 0.011 x age^2) x 0.7 = 116
 
Way out of my depth here. If somebody could translate into plain English?
To visualize virome-wide variance differences between cases and controls,we performed principal component analysis (PCA) on whole blood (Fig. 2a). The first two PCs represented ~30% of the total variance, revealing a separation between cases and controls across PC2 (~10% of the total variance), suggesting differences in total viral loads between groups. To further characterize these differences, we tested for differences in the viral load distributions between cases and controls using a Wilcoxon rank (Fig. 2b) and Kolmogorov–Smirnov (Fig. 2c) tests. Both tests were significant at the p < 0.0001threshold.
What to make of this: significant differences in viral loads between groups? The graphs don't help, I can't make head nor tail of of Fig2a-c. Is the viral load higher in patients or in controls? Either way, what I recall from other studies looking for viruses is that they didn't find any very significant differences one way or the other, possibly a little less virus in patients? Are they measuring something different here to get a significant difference? Or just an artifact of the small numbers?
Differential Expression (DE) analyses showed no statistically significant(FDR<0.05)gene expression differences between any time points in either cases or controls for any of the cell types assessed.
[...]
consistent with previous findings that CPET does not appear to strongly effect molecular differences in CFS[21, 22] to a significant degree.
Looking at the references there does seem to be some consistency to the null results though one paper [22] also speculates that blood may not have been the best choice to test this on.

[21] Keech, A., et al., Gene Expression in Response to Exercise in Patients with Chronic Fatigue Syndrome: A Pilot Study.Front Physiol, 2016. 7: p. 421
https://pubmed.ncbi.nlm.nih.gov/27713703/ (an A Lloyd paper, have only read the abstract; not discussed on S4ME yet)

[22] Bouquet, J., et al., Whole blood human transcriptome and virome analysis of ME/CFS patients experiencing post-exertional malaise following cardiopulmonary exercise testing.PLoS One, 2019. 14(3): p. e0212193.
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0212193

Discussed here: https://www.s4me.info/threads/whole...monary-exercise-testing-2019-chiu-et-al.8692/
 
Way out of my depth here. If somebody could translate into plain English?

What to make of this: significant differences in viral loads between groups? The graphs don't help, I can't make head nor tail of of Fig2a-c. Is the viral load higher in patients or in controls? Either way, what I recall from other studies looking for viruses is that they didn't find any very significant differences one way or the other, possibly a little less virus in patients? Are they measuring something different here to get a significant difference? Or just an artifact of the small numbers?

The Kolmogorov–Smirnov test is a nonparametric test (does not assume parameterised statistical distributions). The figure does show higher viral load (log10cpm), but the Wilcoxon Ranked Test figure still shows largely overlapping distributions. The difference might still indicate something about the group of patients, but the test is not useful as a specific biomarker.
 
Way out of my depth here. If somebody could translate into plain English?

Although your post referred to a couple of specific sections of the paper, I thought it would be useful to mention that Cort Johnson has written an article about this paper that he published today entitled Study Suggests Similar Processes Are Driving Long COVID and ME/CFS. The article is in many ways an extended summary of the paper, and I think that it covers the main points very well. Cort also includes a short interview he had with the paper's lead author, Phillip Comella.

Although Cort's article is only a summary, I think it gives a very good idea of what the paper is all about without having to wade through the technically dense paper itself. As a result, I think that his article makes the results of this work available to a much wider audience, especially those of us with moderate or severe brain fog.

A version of the paper that has the illustrations inline, making it somewhat easier to read, can be found here.

As for the content of the paper, it certainly makes a lot of sense to me, especially since my particular version of ME/CFS clearly involves significant immune dysfunction.
While I like the comparison with other 13 illnesses, I think sample is just way too small to justify such a broad analysis (even though they do seem to have taken multiple comparisons seriously). So I am going to pass on this unless someone gives me a good reason to take a deeper look.

The sample size is definitely problematic here; this is what happens when there's not enough research money to go around. But I think that the authors have done an excellent job with what they have here, and their work appears to be solid enough to put a stake in the ground for future research. As the numbers of people affect by long COVID are only going to grow over time, it should be easier to get money for more extensive studies along these lines than it has been for ME/CFS. I have high hopes that our illness will be a real beneficiary of the work done on long COVID, as I expect the funding for the latter to far exceed what we have been able to get for ME/CFS.
 
To use your analogy; what we have, over the last 30 years or so, is enough stales in the ground to lay out a reasonable sized town, but we still have no buildings, not even a single brick has been laid, all we have is 'stakes in the ground' - most of them so covered in vegetation/piles of crud as to be invisible.

One of the consequences of very small stakes and a lot of time.
 
Back in the day when I was a BMS(biomedical scientist with genetics, immunology and haematology background) a sample size n=15 was often used as a pilot study to cheaply test the water.

I am hopeful that the progress of technology will slowly move in this direction to enable real time dynamic monitoring of transcriptome. Then my understanding is longitudinal study might produce answers?

I am an eternal optimist.

Maybe this paper will be another little piece of the puzzle?

I am fascinated by the paper but it is very technically dense as someone else commented. I reckon it will keep me off the streets for months and absorbed trying to understand it
 
Back
Top Bottom