The discovery of shared cerebrospinal fluid proteins between two independent cohorts of CFS subjects that were not present in two independent sets of healthy control subjects demonstrated that there was differential protein expression in this syndrome (table ). There was significant overlap of CFS, PGI and FM syndromes within the CFS group (figure ). The mix of co-morbid syndromes was different for cohorts 1 and 2, yet the proteomic studies identified a single CFS – related proteome.
Both the individual proteins and the patterns of proteins may offer insights into CFS pathophysiology, and serve as potential biomarkers for CFS diagnosis and the assessment of disease severity. The variable B1/5
epitomized the qualitative pattern recognition of this proteomic "biosignature". The statistical model was predictive of CFS status regardless of which of the 5 proteins, or how many of them, were detected. Prospective studies will be required to determine the sensitivity, and specificity of the proteomic pattern and optimal logistic model(s) for discriminating CFS and related syndromes from normal, idiopathic fatigue, illness-related fatigue states, affective disorder, chronic pain states such as fibromyalgia and regional pain syndromes, and inflammatory central nervous system disorders [1
]. These studies of larger numbers of subjects would be anticipated to refine the CFS – related proteome and the terms of the biosignature variable. These prospective studies are also required to test the objective nature of biosignature variable and other CFS – related proteins for predicting CFS status.
The basis for differences between CFS and HC was the ability of the mass spectrometer to detect certain proteins in CFS but not HC. This concept of "detectability" is critical to understanding the qualitative nature of our results and predictive model. Improvements in the lower limits of detection may identify a larger number of low abundance proteins. These may include the cerebrospinal fluid proteins that were relatively more abundant (and detectable) in CFS compared to HC subjects. This remains to be tested. Quantitative studies will be required to confirm the qualitative distinctions between the CFS and HC groups.
The origins and putative functions of the CFS – related proteome may provide clues to CFS pathophysiology. Cerebrospinal fluid is generated by (a) selective diffusion and transport of plasma components through the choroid plexus and brain parenchymal vessels, (b) synthesis by the choroid plexus and meningeal epithelia, (c) secretion from brain neuron, glia, and other parenchymal cells into the local interstitial fluid, and (d) release from injured and apoptotic parenchymal, mucosal and leukocytic cells [43
]. Plasma protein flux is regulated by transport through the blood-brain barrier and the efflux of protein from cerebrospinal fluid through arachnoid granulations into venous blood [43
]. Alternatively, the fluid may exit the meningeal space through perinasal and olfactory lymphatics [45
]. Dysfunction of hydrodynamic plasma influx or cerebrospinal fluid efflux may contribute to the variations in relative detectability of brain versus plasma proteins in the CFS proteome [46
]. Decreased plasma influx would lead to relatively higher abundances of proteins synthesized in the brain. These proteins may have become easier to detect in CFS. Brain parenchymal mast cells may regulate brain microvascular permeability, possibly through histamine release [47
]. If so, this could explain the benefits of tricyclic drugs such as doxepin and imipramine that have potent histamine receptor 1 antagonist activities [48
]. Histamine receptor 1 antagonists that do not cross the blood brain barrier have no benefit in CFS [49
Proteins of plasma, choroid plexus, or meningeal origin included α-2-mac [50
], orosomucoid 1 and 2 [51
], α1-antichymotrypsin [52
], complement factor 4B precursor [53
], and ceruloplasmin [54
]. These antiproteases, antioxidant, pro- and anti-inflammatory proteins suggested activation of the cerebrospinal innate immune system. Secreted central nervous system proteins included PEDF [55
], CNDP1 [54
], and autotaxin [56
ORM2 is a glycosylated plasma lipocalin with a hydrophobic pocket that binds a wide variety of drugs, hemin, progesterone and the CCR5 receptor on macrophage – lineage cells [57
]. Both ORM2 and ORM1 are acute phase reactants that are synthesized in the liver, but may also be synthesized at sites of brain injury or astroglial cell activation [59
]. Like haptoglobin [60
] and hemopexin [61
], ORM2 and ORM1 may contribute to heme and iron sequestration in the central nervous system in CFS. Iron sequestration is an important antioxidant and antibacterial innate immune defense function [62
]. Haptoglobin and apolipoprotein J act as extracellular chaperone proteins in vivo
]. They may exert anti-inflammatory actions by inhibiting the inappropriate self-association of "damaged" (misfolded) extracellular proteins.
The presence of heme sequestering proteins begged the question of whether free hemoglobin was present in cerebrospinal fluid. This was the case since hemoglobins α1 and α2, β, β Sickle, and δ were detected in the cerebrospinal fluid proteome shared by all subjects ' [see Additional file 1
]'. The source of hemoglobin in normal cerebrospinal fluid could have been the lumbar puncture. If so, the detection rate for hemoglobin would be expected to be similar for all groups, as was found. However, this did not explain the significantly more frequent detection of heme scavengers in CFS (55%) compared to HC (15%) samples (table ). Apolipoprotein B has been used as a marker for the acute introduction of plasma into the cerebrospinal fluid since this protein is not synthesized in the brain [64
]. Apolipoprotein B was detected only once (CFS) ' [see Additional file 1
]', indicating that the lumbar punctures were not a consistent cause of hemorrhage [64
]. Free hemoglobin levels (and mass spectrometric detection) may be related to haptoglobin isoforms [65
]. We have not evaluated this potential correlation in our population.
A number of central nervous system conditions may lead to localized bleeding with hemoglobin release with the induction of heme sequestering proteins. One large group meeting these characteristics are the cerebral amyloid angiopathies (CAA) (cerebrovascular amyloidosis) [66
]. CAA syndromes are defined by protein misfolding, perivascular amyloid deposition, weakening of vessels walls, microhemorrhages to severe cerebral infarction, and dementia or sudden death occurring in the 3rd
decades. The CFS spectrum of illnesses do not demonstrate higher than normal rates of these causes of death making it unlikely that any of the currently identified CAA syndromes were responsible. However, we hypothesize that a mild, focally transient or reversible form that does not lead to either permanently damaging or lethal hemorrhage or dementia may occur in CFS.
This hypothesis would explain many of the parallels between the proteins associated with CAA syndromes and the CFS – related proteome. Gelsolin is an actin "capping" protein that terminates actin polymerization [67
]. Gelsolin cleaved by capsase-3 leads to the "blebbing" during apoptosis. Mutant gelsolin isoforms lead to misfolding and the perivascular amyloid fibril deposition in Finnish type cerebrovascular amyloidosis [66
]. Gelsolin amino acid sequences 173–243 and 173–202 are α-helices that are converted to β – pleated sheet conformations in amyloidosis [68
]. The proteases responsible for the gelsolin cleavage that permit the change in secondary structure are unknown. An E693
Q mutation in amyloid β (A4) precursor-like protein 1 (APLP1) leads to hereditary Dutch type – cerebral hemorrhage with amyloidosis ' [see Additional file 1
]. This is separate from the involvement of this protein in Alzheimer's disease. Immunoglobulin lambda (Ig λ) light chains are relatively unstable, have a tendency to unfold and polymerize [70
], and have been associated with APLP1 in an intracerebral syndrome [71
], and systemic amyloid syndromes with and without B lymphocyte dyscrasias [72
Cystatin C, a leptomeningeal inhibitor of papain – like cysteine proteases, was found in HC and CFS samples. Cystatin C is a homodimer with several domains. The random coil polypeptide linking the terminal domain to the rest of the protein can be proteolytically cleaved [73
]. The "free" domains refold their tertiary structure from constrained α – helices to β – pleated sheets with lower free energies [74
]. These domains are then reattached to the opposite member of the homodimer. This "protein swapping" mechanism is analogous to a DNA recombination – like process [73
]. Inactive β – pleated sheet domains may then polymerize into amyloid deposits. Autosomal dominant Icelandic CAA is due to the disease-causing L68Q variant of human cystatin C [75
]. Cystatin C amyloid immunoreactive material has been found in cerebral cortical, white matter parenchymal and leptomeningeal vessels [74
]. Deposition was more prominent in the media of parenchymal vessels and in the adventitia of leptomeningeal vessels. Complexes of cystatin C or Ig λ with APLP1 have been found in extracellular deposits.
Transthyretin, a thyroxine transporting member of the albumin family, was a common brain – derived component of cerebrospinal fluid that was not part of the CFS proteome ' [see Additional file 1
]'. Misfolding of transthyretin (meningovascular amyloidosis), angiotensinogen, β2 – microglobulin, lysozyme, the Notch3 gene product, and the familial prion protein may each lead to amyloidosis [66
]. Acquired prion diseases may potentially contribute to the CFS – spectrum of illnesses, but the predilection for females and other epidiomological findings make this an unlikely pathological event.
Other components of the CFS – related proteome promote amyloid deposition. Complement factors C3, C4 and B become activated in amyloidosis and Alzheimer's disease [79
]. Apolipoproteins E, E4, and J, and microtubule-associated protein 2 have been associated with CAA syndromes and Alzheimer's disease [66
]. Apolipoprotein E4 may target the amyloid to vessel walls. Chromogranin B – immunoreactive material (table ) was found in 15% of plaques in Alzheimer's disease [81
]. There was a significant loss of chromogranin B – immunoreactivity in the dorsolateral, the entorhinal, and orbitofrontal cortex in Alzheimer's disease. The absence of chromogranin B in these anatomical locations could result in defective synaptic function and the loss of neurohormonal effects. Chromogranin B – immunoreactive material was selectively associated with prion protein deposits in Creutzfeldt-Jakob disease. In contrast, chromogranin A was seen only in amyloid β plaques of Alzheimer's disease [82
Chromogranin B is a highly multifunctional protein. It is a high capacity, low affinity calcium (Ca2+
) storage protein that complexes to the inositol 1,4,5-trisphosphate receptor (InsP3R) in the endoplasmic reticulum. Thus, chromogranin B may modulate Ca2+
]. Chromogranin B (CGB) is a prohormone that can be cleaved to release secretogranin I precursor (Sg1), GAWK and CCB peptides. Both chromogranins B and A are prohormones for the antimicrobial peptides vasostatin-1 and secretolytin [84
]. mRNA for chromogranin B was detected in human monocytes, and may be present in other macrophage/monocytic lineages such as astroglial cells in the brain.
Pigment epithelium-derived factor (PEDF) (table ) is another multipurpose CNS protein [55
]. Although a member of the serpin (serine antiprotease) protein family, it does not possess this activity. PEDF has antiangiogenic activity that may prevent or reduce neovascularization after retinal or cerebral hemorrhage. Both PEDF and a 44 amino acid long proteolytic fragment have potent anti-vascular permeability effects [86
]. PEDF controls the transit of neurons through the cell cycle, promoting their entry into a quiescent state [87
]. The protein may protect neurons and potentially glial cells from apoptosis. Protein levels in the eye and brain decrease with age, but this cannot explain the increase in detection rate in CFS since age was not significantly different between groups. Dickkopf-3, another protein that limits neural proliferation [88
], was detected in 46% of samples.
Autotaxin (TEFLSNYLTNVDDITLVPGTLGR) is a 23 amino acid peptide cleaved from the middle of ectonucleotide pyrophosphatase/phosphodiesterase 2 (ENPP2 gene) [56
]. The parent protein possesses both nucleotide pyrophosphatase and lysophospholipase D (lyso-LPD) activities [56
]. Aliases include phosphodiesterase-1α, phosphodiesterase I/nucleotide pyrophosphatase 2, and alkaline phosphodiesterase I. This "tumor mobility peptide" enhances metastasis of breast and other cancer [91
]. Its unclear if the pyrophosphatase, lysophospholipase D, or autotaxin functions are more important in the brain and in CFS.
Chromogranin B, PEDF, autotaxin, angiotensinogen, and other polypeptides are significant prohormones. The protease cascades that lead to their cleavage and the release of active neuropeptides are poorly understood. However, the peptide hormone effects must be potent since plasma and brain – derived protease inhibitors were detected more frequently in CMI than in HC (table ).
] and α-1-antichymotrypsin were detected in the CFS proteome (table ). IL-1 may induce a mutant promoter allele of α-1-antichymotrypsin that leads to increased central nervous system and the promotion of Alzheimer disease [93
]. Angiotensinogen also has serine protease inhibitor properties. Angiotensinogen was detected in about half of all samples and was not associated with CFS. Haplotype B may be related to microangiopathy – related cerebral damage (MARCD) that can lead to cognitive impairment and gait disturbances in the elderly [78
]. This protein is synthesized by glial cells [94
]. The C-terminal has serine protease inhibitor activity that inhibits angiogenesis. Angiotensins I, II and III are cleaved from the N-terminal. Angiotensin II and III may bind to angiotensin 4 receptors (also known as insulin-regulated aminopeptidase [94
]) in hypothalamic and brainstem nuclei to stimulate the sympathetic nervous system (increase systemic blood pressure), sodium and thermal regulation. The remaining 97% (des [Ang I]angiotensinogen) has no assigned function.
The presence of keratin 16 suggests dysfunction in the leptomeningeal and choroid plexus epithelial system in CFS. Keratin expression in the central nervous system has been incompletely studied, with most of the focus placed on neoplastic tissues [96
]. Keratin 16 is upregulated in epidermal diseases such as psoriasis [99
]. Epidermal growth factor, interferon-γ and ras
can stimulate Sp1
(AP1) proteins that activate the keratin 16 promoter. By analogy, we hypothesize that the presence of keratin 16 in the CFS – associated proteome was an indication of epithelial cell activation within the central nervous system in CFS.
BEHAB may play a role in central nervous system repair or remodeling processes [100
]. Its 2 isoforms bind to extracellular matrix hyaluronan. The full length isoform is secreted into the extracellular matrix. The shorter, splice variant may be linked to glycophosphatidylinositol and form a cell surface protein. The longer variant is highly expressed in childhood, then reaches low, steady state cerebrospinal fluid concentrations by age 20. The shorter variant maintains uniformly low levels throughout development. BEHAB mRNA is elevated 7-fold in gliomas suggesting that glial cells are the normal source in vivo
. BEHAB is also increased in response to brain injury. Glial or other cells with BEHAB anchored to their external plasma membranes may be attracted to putative areas of tissue injury where extracellular matrix became exposed, or where hyaluronan was secreted.
An alternative to the CAA hypothesis is glial cell activation [101
] with the release of innate immune and regulatory factors. Activation of leptomeningeal cells with the secretion of several of the proteins listed above is also possible. CFS syndromes may be initiated by unknown factors that activate these cells, or they may activate anti-inflammatory and innate immune defenses as a result of the original insult. These possibilities may be addressed in future studies and by comparison of the proteomes from CFS subjects with different durations or patterns of illness.
Several novel proteins were identified. These included Dickkopf-3 [88
], disco-interacting protein 2 [Drosophila
] (DIP2) [102
], neuronal PAS domain protein 2 (seasonal affective disorder – related) [103
], additional sex combs – like protein 1 (ASXL1) [104
], and neuroglobin [23
]. Several were represented by only a single, highly selective peptide ' [see Additional file 1
]'. Keratins 5, 6c, 6e, 14, 16, and 17 were the largest single protein family to be newly described [96
No proteins were significantly associated with the good health of the HC subjects.