Label-free analysis of human cerebrospinal fluid addressing various normalization strategies and revealing protein groups affected by multiple sclerosis

The aims of the study were to: (i) identify differentially regulated proteins in cerebrospinal fluid (CSF) between multiple sclerosis (MS) patients and non-MS controls; (ii) examine the effect of matching the CSF samples on either total protein amount or volume, and compare four protein normalization strategies for CSF protein quantification. CSF from MS patients (n = 37) and controls (n = 64), consisting of other noninflammatory neurological diseases (n = 50) and non neurological spinal anesthetic subjects (n = 14), were analyzed using label-free proteomics, quantifying almost 800 proteins. In total, 122 proteins were significantly regulated (p < 0.05), where 77 proteins had p-value <0.01 or AUC value >0.75. Hierarchical clustering indicated that there were two main groups of MS patients, those with increased levels of inflammatory response proteins and decreased levels of proteins involved in neuronal tissue development (n = 30), and those with normal protein levels for both of these protein groups (n = 7). The main subgroup of controls clustering with the MS patients showing increased inflammation and decreased neuronal tissue development were patients suffering from chronic fatigue. Our data indicate that the preferable way to quantify proteins in CSF is to first match the samples on total protein amount and then normalize the data based on the median intensities, preferably from the CNS-enriched proteins.

Cerebrospinal fluid proteomics in multiple sclerosis

Multiple sclerosis (MS) is an immune mediated chronic inflammatory disease of the central nervous system usually initiated during young adulthood, affecting approximately 2.5 million people worldwide. There is currently no cure for MS, but disease modifying treatment has become increasingly more effective, especially when started in the first phase of the disease. The disease course and prognosis are often unpredictable and it can be challenging to determine an early diagnosis. The detection of novel biomarkers to understand more of the disease mechanism, facilitate early diagnosis, predict disease progression, and find treatment targets would be very attractive. Over the last decade there has been an increasing effort toward finding such biomarker candidates. One promising strategy has been to use state-of-the-art quantitative proteomics approaches to compare the cerebrospinal fluid (CSF) proteome between MS and control patients or between different subgroups of MS. In this review we summarize and discuss the status of CSF proteomics in MS, including the latest findings with a focus on the last five years. This article is part of a Special Issue entitled: Neuroproteomics: Applications in Neuroscience and Neurology.

Consensus definitions and application guidelines for control groups in cerebrospinal fluid biomarker studies in multiple sclerosis

The choice of appropriate control group(s) is critical in cerebrospinal fluid (CSF) biomarker research in multiple sclerosis (MS). There is a lack of definitions and nomenclature of different control groups and a rationalized application of different control groups. We here propose consensus definitions and nomenclature for the following groups: healthy controls (HCs), spinal anesthesia subjects (SASs), inflammatory neurological disease controls (INDCs), peripheral inflammatory neurological disease controls (PINDCs), non-inflammatory neurological controls (NINDCs), symptomatic controls (SCs). Furthermore, we discuss the application of these control groups in specific study designs, such as for diagnostic biomarker studies, prognostic biomarker studies and therapeutic response studies. Application of these uniform definitions will lead to better comparability of biomarker studies and optimal use of available resources. This will lead to improved quality of CSF biomarker research in MS and related disorders.

Protein profiling reveals inter-individual protein homogeneity of arachnoid cyst fluid and high qualitative similarity to cerebrospinal fluid

BACKGROUND

The mechanisms behind formation and filling of intracranial arachnoid cysts (AC) are poorly understood. The aim of this study was to evaluate AC fluid by proteomics to gain further knowledge about ACs. Two goals were set: 1) Comparison of AC fluid from individual patients to determine whether or not temporal AC is a homogenous condition; and 2) Evaluate the protein content of a pool of AC fluid from several patients and qualitatively compare this with published protein lists of cerebrospinal fluid (CSF) and plasma.

METHODS

AC fluid from 15 patients with temporal AC was included in this study. In the AC protein comparison experiment, AC fluid from 14 patients was digested, analyzed by LC-MS/MS using a semi-quantitative label-free approach and the data were compared by principal component analysis (PCA) to gain knowledge of protein homogeneity of AC. In the AC proteome evaluation experiment, AC fluid from 11 patients was pooled, digested, and fractionated by SCX chromatography prior to analysis by LC-MS/MS. Proteins identified were compared to published databases of proteins identified from CSF and plasma. AC fluid proteins not found in these two databases were experimentally searched for in lumbar CSF taken from neurologically-normal patients, by a targeted protein identification approach called MIDAS (Multiple Reaction Monitoring (MRM) initiated detection and sequence analysis).

RESULTS

We did not identify systematic trends or grouping of data in the AC protein comparison experiment, implying low variability between individual proteomic profiles of AC.In the AC proteome evaluation experiment, we identified 199 proteins. When compared to previously published lists of proteins identified from CSF and plasma, 15 of the AC proteins had not been reported in either of these datasets. By a targeted protein identification approach, we identified 11 of these 15 proteins in pooled CSF from neurologically-normal patients, demonstrating that the majority of abundant proteins in AC fluid also can be found in CSF. Compared to plasma, as many as 104 proteins in AC were not found in the list of 3017 plasma proteins.

CONCLUSIONS

Based on the protein content of AC fluid, our data indicate that temporal AC is a homogenous condition, pointing towards a similar AC filling mechanism for the 14 patients examined. Most of the proteins identified in AC fluid have been identified in CSF, indicating high similarity in the qualitative protein content of AC to CSF, whereas this was not the case between AC and plasma. This indicates that AC is filled with a liquid similar to CSF. As far as we know, this is the first proteomics study that explores the AC fluid proteome.

Arachnoid cysts do not contain cerebrospinal fluid: A comparative chemical analysis of arachnoid cyst fluid and cerebrospinal fluid in adults

Background

Arachnoid cyst (AC) fluid has not previously been compared with cerebrospinal fluid (CSF) from the same patient. ACs are commonly referred to as containing "CSF-like fluid". The objective of this study was to characterize AC fluid by clinical chemistry and to compare AC fluid to CSF drawn from the same patient. Such comparative analysis can shed further light on the mechanisms for filling and sustaining of ACs.

Methods

Cyst fluid from 15 adult patients with unilateral temporal AC (9 female, 6 male, age 22-77y) was compared with CSF from the same patients by clinical chemical analysis.

Results

AC fluid and CSF had the same osmolarity. There were no significant differences in the concentrations of sodium, potassium, chloride, calcium, magnesium or glucose. We found significant elevated concentration of phosphate in AC fluid (0.39 versus 0.35 mmol/L in CSF; p = 0.02), and significantly reduced concentrations of total protein (0.30 versus 0.41 g/L; p = 0.004), of ferritin (7.8 versus 25.5 ug/L; p = 0.001) and of lactate dehydrogenase (17.9 versus 35.6 U/L; p = 0.002) in AC fluid relative to CSF.

Conclusions

AC fluid is not identical to CSF. The differential composition of AC fluid relative to CSF supports secretion or active transport as the mechanism underlying cyst filling. Oncotic pressure gradients or slit-valves as mechanisms for generating fluid in temporal ACs are not supported by these results.

Pretreatment of Mass Spectral Profiles: Application to Proteomic Data

Mass spectral profiles are influenced by several factors that have no relation to compositional differences between samples: baseline effects, shifts in mass-to-charge ratio (m/z) (synchronization/alignment problem), structured noise (heteroscedasticity), and, differences in signal intensities (normalization problem). Different procedures for pretreatment of whole mass spectral profiles described by almost 50,000 m/z values are investigated in order to find optimal approaches with respect to revealing the information content in the data. In order to quantitatively assess the impact of different procedures for pretreatment of mass spectral profiles, we use factorial designs with the ratio between intergroup and intragroup (replicate) variance as response. We have examined the influence of smoothing, binning, alignment/synchronization, noise pattern, and normalization on data interpretation. Our analysis shows that the spectral profiles have to be corrected for heteroscedastic noise prior to normalization. An nth root transform, where n is a small, positive integer, is used to create a homoscedastic noise structure without destroying the linear correlation structures describing individual components when using whole mass spectral profiles. The choice of n is decided by a simple graphic procedure using replicate information. Log transform is shown to change the heteroscedastic noise structure from being dominant in high-intensity regions, to produce the largest noise in the low-intensity regions. In addition, log transform has a negative effect on the collinearity in the profiles. Factorial designs reveal strong interactions between several of the pretreatment steps, e.g., noise structure and normalization. This underlines the limited usability of looking at the different pretreatment steps in isolation. Binning turns out to be able to substitute smoothing of spectra by, for example, moving average or Savitsky-Golay, while, at the same time, reducing the data point description of the profiles by 1 order of magnitude. Thus, if the sampling density is high, binning seems to be an attractive option for data reduction without the risk of losing information accompanying the integration of profiles into peaks. In the absence of smoothing, binning should be executed prior to alignment. If binning is not performed, the order of pretreatment should be smoothing, alignment, nth root transform, and normalization.