Like many of the other studies studied in the Pharmacogenomics Journal this study was an order of complexity higher than we have been used to seeing. First these researchers tried to differentiate CFS patients from Idiopathic Fatigue (IF) and from Healthy Controls using the gene expression data. Then they used the clinical variables and laboratory data to create two ‘target spaces’, one largely delineating symptom presentation and one delineating biological findings. Next they determined the genes whose expression was correlated with those target spaces. Essentially they used the clinical findings (fatigue assessments, neuropsychological tests, depression tests, symptom inventory scores) in an attempt to create a whole picture of the clinical aspects of CFS and then determined which genes were most expressed in that picture. They did the same with the lab data.

If the authors conclusions are correct, this study could have considerable implications for the design and analysis of other gene expression studies. One of their findings was startling enough that I will quote it directly. After indicating that a principal components analysis designed to produce groups with similar types of gene expression jumbled the CFS, IF and healthy controls together, they stated, "this suggests that the vast majority of the variation in the gene expression data are attributable to factors other than illness." Put simply – most of the gene expression data we are seeing has nothing to do with CFS. Their analyses indicate that no more than 10% of the data we are seeing has relevance to CFS.

These researchers zeroed in on 39 genes that did allow them to differentiate CFS patients from the other groups. Unfortunately they immediately ran into a road block; information on the gene functioning was available for only 17 of them. This appears to be an extraordinarily low number. I have no idea what it means. Perhaps not surprisingly, few of these genes had been highlighted by other studies.

Just when this group seemed about to bury our confidence in the efficacy of gene expression studies they resurrected it. While few of the same genes have been found in past gene expression studies, from a functional aspect these genes were quite similar to those we have seen before. Most promisingly, these genes were similar to those the 2005 Whistler study found. That study found altered expression in ion channnel and immune response genes after exercise in CFS

The results were coherent enough for these researchers to use them to posit a model of CFS pathology. They conjectured that increased free radical production due to immune activation in CFS damages the ion channels on the membranes of the cells. As support for this they noted that the top gene highlighted in this study (SESN1) is produced in response to oxidative stress.

These researchers weren’t done yet. They also attempted to determine which laboratory and clinical measures were associated with the multi-dimensional symptom ‘target space’ created. The results were most interesting. They found that 9/17 of laboratory measures most associated with increasing symptoms in CFS dealt with low sleep heart rate variability (HRV).

Like many of the other studies in this journal the researchers of this study took a different approach than we’ve seen before. Instead of attempting to differentiate CFS patients from controls using gene expression data the researchers attempted to determine which genes were most responsible for the fatigue and depression seen in CFS.

CFS patients were broken into groups encompassing the most fatigued and depressed and the least fatigued and depressed patients. Statistical tests then determined which of the 15,000 genes’ (~50% of the human genome) activity was altered in these groups. Genes were considered to be differentially expressed if they were 4x more active in one group than the other. Special attention was given to genes which were active in both highly fatigued and depressed CFS patients.

This study found 188 and 164 genes that were associated with fatigue and depression, respectively, and 24 genes that were common to both. The researchers speculated that the activation of these ‘super genes’ could play a key role in CFS. They tested this idea by determining if they could differentiate CFS patients from healthy controls by comparing the activity of these 24 genes and found that they could. Most of the rest of the paper focused on these 24 genes.

Twenty Four Genes at the Heart of CFS? – Like other studies these genes were involved in a number of activities. The authors identified 11 pathways of interest, a good portion of which have been highlighted in other gene expression studies. They include the immune response, apoptosis (cell suicide), ion channel functioning and metal ion binding, cellular signaling and neuronal activity.

The authors picked out three genes whose activity was very significantly altered in CFS patients. Interestingly, for proponents of Dr. Marshall’s theory, one of them (GUCA1B) is associated with an increased sensitivity to light (photophobia). One part of the protocol suggested by this theory calls for staying out of the sunlight. Another one, an estrogen receptor on peripheral blood leukocytes (ESR2), could explain the increased levels of lymphocyte activation sometimes found in CFS. High levels of the estrogen receptor on leukocytes should make them highly responsive to estrogen. Could something like this help explain why women – who have higher estrogen levels than men – have higher rates of CFS than men? The third gene (DFFA) is involved with tumor necrosis alpha (TNF-a) signaling pathways and DNA fragmentation during apoptosis (cell suicide).

Several of the other 24 genes are compelling for other reasons. Three genes involved in calcium and/or sodium ion channel function and transport suggest a channelopathy in CFS. Two genes, interestingly enough, are involved in the binding of magnesium, perhaps the most widely used supplement in CFS. The heat shock protein gene found has been implicated in protein misfolding, a process occurring in amyloidosis (protein aggregation in the blood vessels) which was recently implicated in CFS by Baraniuk’s cerebral spinal fluid proteome study.  Finally two genes involved in Acyl-coenzyme A binding and phosphate metabolism are suggestive of metabolic problems. The acyl-CoA gene’s role in metabolizing fat is intriguing given the increased waist/hip ratios seen in the Maloney allostatic load study (click here)

Finally, several genes appear to indicate that disrupted communication (‘cross-talk’) between the brain and the immune system occurs in CFS. These and other genes lead the authors to focus on the ‘focal adhesion’ pathway. Genes in this pathway interact at the part of the cell where its cytoskeleton interacts with proteins of the extracellular matrix. The authors found it compelling that five of the pathways that interact at this spot (cytokine-cytokine interactions, phosphatidylinositol signaling, actin cytoskeletal regulation, apoptosis, MAPK signaling system) appear to be altered in CFS. This suggests that a disruption in this part of the cell may play a key role in CFS.

The imprecise nature of the diagnostic criteria for CFS is believed to result in the formation of heterogeneous study groups which contribute to the weak or inconclusive study findings sometimes seen in CFS. This group bypassed the uncertainties regarding the disease definition by focusing their attention not on CFS per se. but on a central symptom found in it – fatigue.

They used the Multidimensional Assessment of Fatigue or MFI to assesses five aspects of fatigue; general, physical and mental fatigue, reduced motivation and activity. Then they attempted to correlate gene expression patterns with different types of fatigue. They wanted to know which genes were expressed differently in people with increased fatigue. They assessed gene expression of about half the genes in our genome.

This study found 839 genes whose expression was significantly correlated with one or more dimensions of self-assessed fatigue. About 1/3^rd of these were associated with more than one type of fatigue and 15 genes were associated with all five dimensions of fatigue. Most of the correlations, however, were modest. The authors ran into a roadblock when they tried to assess the role these genes play – the function of only about 25% of these genes is known. Again, this appears to be an extraordinarily low number.

These genes were mainly involved in several basic cellular processes; metabolism, transcriptional regulation and signaling, etc. When the authors looked at the genetic activity in three broad functional categories; biological processes, molecular functions and the cellular component they found that genes in the following categories were most active in the fatigued patients:

• Cellular development - muscle development and embryonic development
• Metabolic processes – Many metabolic categories were highlighted including polysaccharide metabolism (and biosynthesis), glucan metabolism, lipid metabolism and glycogen metabolism (and biosynthesis) plus purine, tyrosine and tryptophan (serotonin) metabolism.
• Glycogen is converted into glucose – the main energy source of the cell. Polysaccharides are carbohydrates such as starch that contain saccharides. Glucan is a polysaccharide that yields glucose as well. The altered activity of carbohydrate and lipid metabolism genes in the more fatigued patients perhaps buttresses Mahoney’s thesis that CFS patients that the ‘energy set point’ in CFS patients has been altered. A small number of genes involved in oxidative phosphylation, the end stage of the metabolic process in which aerobic respiration In the mitochondria produces ATP for the body, were also highlighted. The wide range of metabolic genes noted that are upstream of the oxidative phosphorylation process would appear to suggest, however, that the energy production problems in CFS, sometimes suggested to possibly be due to mitochondrial defects, may occur prior to that point.
• Immune system – Genes involved in humoral immune defense such as the complement cascade, apoptosis, and infection were highlighted.

Molecular Functions

• Transcription related genes. Transcription is the process by which mRNA is transformed into DNA. These often show up in other gene expression studies.
•  Cytoskeleton related genes – also have shown up in other gene expression studies. De Meirleir has found evidence of unusual actin cytoskeleton fragments in CFS patients (Click here). The actin cytoskeleton is involved in many processes of the cell including the immune response.
• Potassium ion channels – ion channel genes have also shown up in other studies; only two kinds of particular potassium ion genes were altered in CFS in this study.

• Cytoskeletal – spindle genes.
• Endocytic vesicles – are involved in phagocytosis, an immune function in which pathogens (or other substances) are put in pouches and brought into the cell where they are digested.
• Endoplasmic reticulum – are the tubules through which proteins are transported after they are produced by the ribosomes.
• Eukaryotic initiation factor (EIF-4F) – also commonly shows up in gene expression studies.

In common with past gene expression studies we saw evidence of cytoskeletal, immune, transcriptional activation and some ion channel activity in the fatigued patients in this one. We didn’t see evidence of membrane problems indicative of increased oxidative stress, much ion channel activity or much endocrine or neuronal activity. The most evocative finding was the wide array of metabolic genes activated – metabolic genes have not been commonly found in past studies.

Most of the processes elucidated are so fundamental, however, that they didn’t aid the researchers to build a model of fatigue. What researchers really want to see are genes that can be tied to specific cells and specific parts of the body. The authors state that despite finding patterns of gene expression correlated with fatigue that ‘the pathogenesis of fatigue is not elucidated’ by this study.

The authors noted there were several limitations to this study including the possible presence of different fatigue producing pathologies that may have obscured their findings. In a rather strongly worded statement reflecting what they believe is an ‘urgent’ situation, they also stated that a significant portion of the microarray data is inaccurate (!) because the manufacturers have not updated their probe information to reflect recent developments. With regard to the missing gene function data they believe that the next few years will see a ‘vast improvement’ in our knowledge of gene function - much of this studies value may lie in the future.

THE GENE EXPRESSION STUDIES – AN OVERVIEW

The first big hurdle for the CFS gene expression researchers was to show that they could be used to differentiate CFS patients from controls. This was successful. The next hurdles concerned our ability to make sense of the results and to replicate them. Thus far most of the gene expression results have been too complex (too scattered) for most researchers to feel comfortable using them to elucidate models of CFS pathophysiology.

It’s encouraging that two sets of authors began to build a model of CFS pathophysiology based on their results. With its focus on the focal adhesion pathways and their involvement in cytokine – nervous system – HPA axis activity the Fang study opened new ground for CFS research. Identifying new areas of research was one of the things we wanted from these studies. It is encouraging as well that Dr. Kerr is confident enough of his results to begin to devise a therapy based on them and it is encouraging that we are seeing the same general patterns involving the immune system, nervous system, ion channels, and cellular signaling crop up over and over again.

The warning MERGE gave us regarding the complexity of translating gene expression results into new models of CFS pathophysiology or treatment has turned out, however, to be true. The findings thus far are too complex, and given their variability, too inconsistent to allow CFS researchers to really hone in on the source or sources of CFS. The Broderick study suggested that only about 10% of the genes highlighted in the gene expression studies contribute to CFS pathology. Vernon’s statement that ‘molecular profiling has demonstrated several albeit subtle perturbations in peripheral gene expression" supports the contention that the gene expression results have had only moderate applicability to CFS thus far. Out of several hundred genes highlighted in the six studies examining PBMC cells only a few have expressed in more than one study and none in more than two.

Several factors determine how effective a gene expression study is. Researchers are looking for at least three things in these studies; genes whose expression is highly altered, sets of genes whose expression they can fit together to produce a model of disease, and, of course, they are looking for consistently found genes. It is my impression – that of a layman with no expertise in this difficult field - that none of these have happened yet in the CFS gene expression studies; the alteration of the gene expression in CFS is not particularly great; while some genes do make sense given what we know about CFS it is difficult to fit the entire package of genes together to produce a model of CFS pathophysiology, and that the results, at least with regard to individual genes, have been inconsistent.

The inconsistencies thus far seen could be due to several factors; different sample populations, different gene arrays, different methodologies. Right now the results of the gene expression studies appear to be much like the results of other studies in CFS; they are too intriguing to turn ones back one but are not conclusive enough to all researchers to really focus in on CFS. It appears that something – probably subsets - is obscuring the view these studies are giving us of CFS.

It should be remembered, however, that almost all the gene expression studies have focused on one type of cell called peripheral blood mononuclear cells (PBMC’s). Since these cells interact with many different systems of the body as they circulate in the bloodstream they are thought to provide a snapshot of the activities of multiple systems of the body. They can only give only a snapshot, however, and are able to convey only limited amounts of information. The complexity present in the PBMC gene expression results thus far makes the coherence of the Baraniuk cerebrospinal fluid proteome study all the more noteworthy (click here). It is possible that the cerebral spinal fluid may provide a better window through which to view CFS than the bloodstream.

The Future: The upcoming Gow and Kerr studies, one examining the entire genome, and the other employing much, much larger numbers of CFS patients than have been seen before will be of particular importance. Happily, the preliminary reports from the Kerr study indicate his results are consistent with his past study, and that he is looking for and finding protein analogues to his gene expression results. This very large study (reportedly 1000 CFS patients) will undoubtedly be a landmark in CFS gene expression studies and should tell us much about how important a role gene expression will play in CFS. Dr. Sullivan is also engaged in a twin gene and protein expression study using not only blood but cerebral spinal fluid. This study should be completed next year

Next Up – Our examination of the gene expression pie in CFS will next consist of an overview of the CAMDA conference findings. In this conference teams of researchers from around the world took their shot at making sense of the mass of clinical, gene expression, SNP and proteome data gathered by the CDC in the 2003 Wichita Kansas study. There were many intriguing findings including one that proposed to have found a biomarker for CFS.