Background: A biomarker for the diagnosis of childhood-onset ataxia and central nervous system hypomyelination (CACH)/vanishing white matter disease (VWM) would have clinical utility and pathophysiologic significance.
Methods: We used 2-dimensional gel electrophoresis/mass spectrometry to compare the cerebrospinal fluid proteome of patients with mutation-confirmed CACH/VWM with that of unaffected controls. We characterized selected spots by in-gel digestion, matrix-assisted laser desorption/ionization time-of-flight tandem mass spectrometry, and nanospray Fourier transform mass spectrometry.
Results: A specific transferrin spot pattern was detected in the CSF samples of the CACH/VWM group (n = 7), distinguishing them from the control group (n = 23) and revealing that patients with CACH/VWM have a deficiency of the asialo form of transferrin usually present in healthy cerebrospinal fluid. The glycopeptide structure, determined from isolated transferrin spots by use of in-gel digestion and extraction, was found to be consistent with earlier reports.
Conclusions: The transferrin isoform abnormality in the cerebrospinal fluid of patients with CACH/VWM appears unique and is a potential clinical diagnostic biomarker. The rapid, efficient diagnosis of this disorder would have a significant impact on clinical studies exploring new strategies for the management and treatment of this disease.
The diagnosis of unclassified leukodystrophies remains a challenge for the clinical neurologist (1)(2). Childhood-onset ataxia and central hypomyelination (CACH)1 (3), also known as vanishing white matter disease (VWM)(4), in particular, is thought to be one of the most frequent etiologies of undiagnosed leukodystrophy(2)(5). CACH/VWM diagnosis currently requires the recognition of frequently variable clinical features, magnetic resonance imaging findings, and sequencing of the 57 exons of the eIF2B genes associated with the disease(6). In many cases this process may delay or prevent accurate diagnosis. A rapid, clinically available biomarker would simplify the diagnosis of this disorder.
Cerebrospinal fluid (CSF) is a relatively accessible patient tissue that has historically been used to investigate neurologic disorders of infectious, inflammatory, neoplastic, and degenerative etiology. To a certain degree, CSF is shielded from nonneurologic protein sources by the blood–CSF barrier and has a relatively slow protein turnover. Its close contact with the brain’s extracellular space makes CSF a very promising source of biomarkers associated with neurologic disorders.
As a result of the development of effective approaches for proteome analysis, an increasing number of candidate biomarkers have been identified recently in the CSF of patients with neurologic disorders, including amyotrophic lateral sclerosis (7), multiple sclerosis(8)(9), Creutzfeld–Jacob disease(10), primary brain tumors(11), and adult degenerative disorders such as Alzheimer disease(12)(13)(14). Most of these studies used either the conventional 2-dimensional gel electrophoresis (2-DGE)/mass spectrometry (MS) approach or shotgun proteomic approaches in combination with clustering analyses(15)(16). Thus, proteomics is emerging as a reliable tool to screen for disease-associated biomarkers(17)(18)(19), but as yet has had only limited application to pediatric degenerative central nervous system disorders.
In the present study, we used a 2-DGE/MS-based approach to explore possible differences between the CSF proteomes of CACH/VWM patients and controls. We hypothesized that the technique would identify a differential protein pattern between patients affected by CACH/VWM and controls.
Materials and Methods
All CSF samples were collected in accordance with an Institutional Review Board–approved protocol at Children’s National Medical Center and the collaborating institutions. Affected and diseased control samples were obtained at the National Institute of Neurological Disorders and Stroke (NINDS)/NIH. Other control samples were obtained at Children’s National Medical Center. Only excess CSF drawn for other clinical or research purposes was used for these analyses. Confirmation of eIF2B mutation status was performed at the Children’s National Medical Center (20). In total, samples from 7 patients with mutation-confirmed VWM, from 11 children without neurologic disorders but with other medical reasons for lumbar puncture, and from 12 children with unrelated neurologic disorders (spinal muscular atrophy, white matter abnormality related to a chromosomal abnormality, Alexander disease, primary brain tumor, X-linked adrenoleukodystrophy, or cervical cord lesion) were collected in sterile polypropylene cryogenic vials. Immediately after collection, each CSF sample was checked for blood contamination in a clinical laboratory by microscopy (assessing a standard cell count of erythrocytes and leukocytes per visualized field). The CSF samples were then centrifuged at 300g for 10 min to remove any residual debris. The supernatants were transferred to clean Eppendorf polypropylene tubes and stored at −80 °C until analysis.
The protein concentration in each CSF sample was measured by the Bio-Rad protein assay. Typically, CSF samples had a protein concentration ranging from 0.16 to 0.4 g/L. Aliquots containing 100 μg of total protein were taken from each sample and processed for 2-DGE analysis as follows. Samples were desalted against 10 mmol Tris-HCl (pH 7) on p6 Bio-Spin columns. Each solution was then dried by centrifugation under reduced pressure, after which 180 μL of rehydration buffer (7 mol/L urea, 2 mol/L thiourea, 20 g/L CHAPS, 50 mmol dithiothreitol, and 5 mL/L ampholytes, pH 3–10) was added to the dry sample to solubilize and denature the proteins. First-dimension electrofocusing was performed on IPG strips (11 cm; pH 3–10) in an electrofocusing chamber (Bio-Rad) operated as follows: 12 h of rehydration, 250 V for 15 min, 1000 V for 1 h, and 10 000 V for 4 h. The second-dimension sodium dodecyl sulfate–polyacrylamide gel electrophoresis was performed on Criterion Tris-HCl (8%–16%) precast gels (Bio-Rad). Protein spots were visualized by use of Bio-Safe Coomassie stain (Bio-Rad). The gel was then scanned on a GS800 densitometer (Bio-Rad) and imaged as a TIFF file. The resulting gels arrays were compared by PDQuest 2-DGE image analysis software (Bio-Rad), and the volumes and intensities of spots of interest were determined.
Protein spots were excised with the tip of a clean polypropylene pipette and transferred to a microcentrifuge tube containing 100 μL of deionized water. The bands were destained by 2–3 washes with 100 μL of acetonitrile–water (50:50 by volume). Tryptic digestion was performed as described previously (21). The peptides were extracted, dried by vacuum centrifugation, redissolved in 10 μL of 1 mL/L trifluoroacetic acid (TFA), and desalted by use of C18 ZipTip micropipette tips (Millipore) according to the manufacturer’s User Guide. The peptides were eluted from the ZipTip in 10 μL of acetonitrile–1 mL/L TFA (70:30 by volume). Typically, 0.3 μL of peptide solution was mixed with 0.3 μL of matrix solution [50 mmol α-cyano-4-hydroxycinnamic acid in acetonitrile–1 mL/L TFA (70:30 by volume)] and spotted on the matrix-assisted laser desorption/ionization (MALDI) plate.
MS and tandem MS (MS/MS) analyses were performed on a 4700 ABI TOF-TOF mass spectrometer (Applied Biosystems) equipped with an Nd:YAG 200 Hz laser. The instrument was operated with delayed extraction in reflectron-positive ion mode. A mixture of standard peptides was used to externally calibrate the instrument. Protein identification was carried out with GPS explorer software (Applied Biosystems). Both MS and MS/MS data were used for protein identification. To detect intact glycopeptides with high molecular masses, the 4700 ABI instrument was operated in linear positive mode and tuned for masses of 2000 and 10 000 Da.
Accurate mass measurements were obtained by nanospray (22) on an Apex III Fourier transform mass spectrometer (FTMS)(23)(24) (Bruker Daltonics) equipped with a 7.0 T active shielded superconducting magnet and an Apollo atmospheric pressure ionization source. The home-built heated-metal desolvation capillary was kept at 120–150 °C. Each experiment was performed in positive ion mode and required loading 5–10 μL of analyte solution into a freshly pulled borosilicate needle, while a platinum wire was inserted from the back end to provide the necessary nanospray voltage. All data were acquired in broadband mode and processed by Bruker XMASS 6.0.1 software. MS/MS experiments were carried out by isolating the precursor ion of interest using correlated RF sweeps (CHEF)(25) followed by activation through sustained off-resonance irradiation (SORI) against an argon background to obtain collision-induced dissociation (CID).
characterization of the glycopeptides
Glycopeptides were tentatively characterized with a GlycoMod tool (http://us.expasy.org/tools/glycomod) by entering the observed masses (using a mass accuracy better than 10 ppm) and selecting predicted sugar moieties.
The identity of target glycopeptides was further confirmed through controlled digestion with exopeptidase and exoglycosidase. Briefly, aliquots from the in-gel digest described above were dried by vacuum centrifugation and redissolved in 10 mmol ammonium bicarbonate (pH 7.4). Selected aliquots were treated with α-2–3,6,8,9-neuraminidase (Calbiochem) to induce hydrolysis of sialic acid-containing peptides. The remaining aliquots were treated with carboxypeptidase Y (Roche Diagnostics) to verify the identity of the peptide backbone. All reactions were monitored by MALDI time-of-flight tandem MS (MALDI-TOF-TOF) in linear positive mode as described above.
A frequent pitfall of comparative 2-DGE is a relatively high degree of intersample variability, particularly in clinical samples. Studies that have investigated the CSF proteome found a minimum of 6% intersample variability (26). Other reports have highlighted an even greater person-to-person variability, particularly for samples obtained from elderly individuals(27). This variability could significantly hamper the identification of a reliable biomarker. Consequently, it was not surprising that 2-DGE maps obtained from the different CSF samples showed great variability that was manifested not only between the 2 groups but also between individuals within each group (Fig. 1⇓ ). For example, the relative intensities of spots containing α1-acidic glycoprotein were found to vary considerably among 4 different samples, and although this protein was abundant in the CSF of patient CACH/VWM1, it was almost undetectable in patient CACH/VWM2. Conversely, apolipoprotein-A1 was readily detected in the CSF of patient CACH/VWM2 but was almost undetectable in the other CSF samples. This person-to-person variability renders conclusions difficult when only a small number of samples are investigated.
Careful analysis of the 2-DGE spot patterns of a larger number of samples (7 patients and 23 controls) led to the detection of a unique pattern in CSF of CACH/VWM patients that was not present in the controls. This pattern consisted of a variation in the composition of transferrin isoforms from CACH/VWM patients vs controls: the control CSF provided 6–8 spots of transferrin with pI values ranging from 5.5 to 6.5, whereas the patient CSF showed a greater intensity of spots with more acidic pI values (pI 5.5–6) and a lower intensity of spots with more basic pI values (pI 6–6.5) than the controls (Fig. 1⇑ ). These 2 different patterns were reproduced among all samples from controls and CACH/VWM patients, respectively (Fig. 2⇓ ). In all patient samples, there were fewer basic transferrin spots than in control samples.
The number of samples analyzed was definitively sufficient to exclude the possibility that the observed differences between 7 VWM and 23 controls were caused by random events. In addition, CSF samples from 2 additional patients suspected to have VWM were analyzed and found to have spot patterns that were clearly closer to the control cluster. Independent diagnosis based on gene sequencing showed that these 2 patients did not actually have the known mutation for VWM. This experiment could be considered a blinded test and shows the robustness of our candidate biomarker.
characterization of gel-separated transferrin isoforms
The isolated transferrin isoforms were subsequently analyzed by use of a well-established mass fingerprinting strategy (21). Tryptic peptides generated by in-gel digestion of the different spots were shown by reflectron MALDI-TOF-TOF analysis to cover ∼69% of the protein sequence (Table 1⇓ ⇓ ). However, examination of the detected peptides vs the known sequence of human transferrin (Swiss-Prot accession no. P02787) revealed that the maps of both acidic and basic isoforms were missing regions encompassing the known glycosylation sites (Asn432 and Asn630; Table 1⇓ ⇓ ). Switching the detection mode from reflectron TOF-TOF to linear TOF afforded detection of a greater number of peaks with higher masses (Fig. 3⇓ ), which we presumed could include extensive carbohydrate structures. These experiments differentiated 2 groups of isoforms corresponding either to the acidic spots S1, S2, and S3 or to the basic spots S4, S5, and S6 (Fig. 3⇓ , top panel). In particular, the acidic isoforms were characterized by signals with m/z ratios of 4431 and 4723 that were not detected in the basic group, and conversely, a peak was observed at m/z 4163 for the basic group that was not detected for the acidic group (Fig. 3⇓ ). No significant differences were noted for within-group comparisons of S1, S2, and S3 or S4, S5, and S6.
Neuraminidase treatment of the larger tryptic products caused prompt elimination of the peaks at m/z 4723 and 4431, whereas m/z 4163 remained unaffected (Fig. 3⇑ , bottom panel). This enzymatic reaction confirmed the presence of sialic acid moieties in the acidic isoforms, a finding that was consistent with the isoelectric focusing migration of their corresponding electrophoretic spots. Furthermore, the observation of a new single peak at m/z 4139, which accompanied the disappearance of both m/z 4723 and 4431 after neuraminidase treatment, suggests the possible presence of either 2 or 1 sialic acid molecules in 2 individual glycopeptides sharing the same peptide backbone The peptide at m/z 4163 generated from the most basic transferrin spots did not seem to contain any sialic acid residues, and its mass remained unchanged even after prolonged treatment with neuraminidase.
Image analysis of the 2-DGE patterns by PDQuest provided a determination of the ratio of asialo- vs sialotransferrin in each sample. In the control samples analyzed in this study, asialotransferrin represented 8%–35% of the total CSF transferrin (mean, 19.2%; 95% confidence interval, 16.6%–21.7%), whereas it amounted to 0.5%–5% of total transferrin (mean, 2.5%, 95% confidence interval, −2.2% to 7.2%) in CACH/VWM samples. The ratios of asialo- to sialotransferrin in the control and CACH/VWM samples were statistically significantly different [t(28) = 6.42; P <0.0001]. When the control group was limited to the 5 patients with non-CACH/VWM leukodystrophies (Alexander disease, X-linked adrenoleukodystrophy, white matter disease with chromosomal disorder, or hypomyelination), asialotransferrin represented 14%–25% of the total CSF transferrin (see the scatter plot in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol51/issue11/ ); the ratio remained statistically significant [t(10) = 8.88; P <0.0001]. In agreement with the image analysis data, estimates based on the MS analysis of tryptic digests suggested asialoglycopeptide concentrations that were 8–10 times lower in patients than in controls (data not shown).
structural characterization of glycopeptides generated from gel-separated transferrin isoforms
Mass measurements were obtained by nanospray-FTMS analysis for each of the glycopeptides of interest (Fig. 4⇓ ). Multiply charged signals were readily observed for the sialo form at m/z 1573.97 (3+), 1466.38 (4+), 1434.61 (4+), 1227.24 (3+), and 1180.72 (4+). No corresponding signals were detected for possible products generated by the asialo form (as determined by neuraminidase treatment). On the other hand, species at m/z 1388.23 (3+) and 1347.08 (2+) were observed only for the asialo form and not for the sialo form. The observed monoisotopic molecular masses obtained from these signals were used in the GlycoMod Tool (http://us.expasy.org/tools/glycomod/) to predict the putative structures of the corresponding glycopeptides (Table 2⇓ ). A mass accuracy of 10 ppm or better allowed us to minimize the incidence of possible false-positive identification.
These data suggest that acidic transferrins (sialotransferrins) are likely to include both Asn432 and Asn630 bearing the same biantennary carbohydrate structure composed of 2 N-acetylglucosamine (GlcNAc), 3 mannose (Man), 2 GlcNAc, 2 galactose (Gal), and 2 N-acetylneuraminic acid (NeuNAc; sialic acid; see structure in Table 2⇑ ). On the other hand, basic transferrins (asialotransferrins) appear to contain a completely different form of carbohydrate side chain attached to Asn630, corresponding to an N-linked biantennary structure composed of 2 GlcNAc, 3 Man, 3 GlcNAc, and 1 fucose (Fuc; see putative structure in Table 2⇑ ). It should be noted that, in agreement with our results, this particular type of glycosylation was described previously for transferrin purified from pooled human CSF samples (28).
Close examination of tryptic digests obtained from several asialotransferrin spots was carried out to further investigate whether the same glycosylation could also be linked to the Asn432 residue. Although no evidence could be gathered to support this hypothesis, a peptide with a monoisotopic mass of 2692.177 Da that was detected in the asialotransferrin spots was completely absent from the corresponding digests of sialotransferrin spots. This peptide was detected only with electrospray ionization-FTMS and not with the MALDI-TOF/MS, and its intensity was too low to confirm its exact structure. However, its mass was in close agreement with that calculated for peptide [421–433] with Asn432 modified by an asialocarbohydrate composed of 2 GlcNAc and 5 Man (Table 2⇑ ). Although this kind of glycopeptide does occur, and it has been characterized previously in sera of patients with congenital disorders of glycosylation (29), its occurrence in CSF asialotransferrin still needs further investigation.
Generally, the results provided by high-resolution nanospray-FTMS and MALDI-TOF-TOF were in excellent agreement despite the different characteristics of these ionization techniques. An exception was the detection of m/z 4431 in the MALDI spectrum of acidic transferrin, which was not observed by nanospray analysis of the same tryptic digest. However, this signal could be produced by the loss of 1 sialic acid unit (−292 Da) from the peptide at m/z 4723 (corresponding to 4718.887 Da by nanospray-FTMS), which is consistent with possible in-source fragmentation processes, as reported previously for glycopeptides generated from digestion of serum transferrin (30). Others have tested several matrix solutions for MALDI-TOF analysis of sialic acid–containing peptides. It was found that 2′,4′,6′-trihydroxyacetophenone and 2,5-dihydroxybenzoic acid offered better detection and stability of acidic glycopeptides than did α-cyano-4-hydroxycinnamic acid(31). In our case, the use of dihydroxybenzoic acid produced only a slight improvement in the stability of sialic acid–containing glycopeptides. In addition, we noticed a general decrease in sensitivity for the higher m/z values, which is typical of dihydroxybenzoic acid vs α-cyano-4-hydroxycinnamic acid.
Further support for the proposed glycopeptide structures was provided by SORI-CID (see Materials and Methods) of selected precursor ions performed on the nanospray-FTMS instrument (Fig. 5⇓ ). Most of the fragment ions seen in the 2 spectra corresponded to cleavages occurring in the carbohydrate side chain. In general, larger fragment ions corresponded to the loss of 1 or 2 saccharide moieties from the glycopeptide, and the low-mass ions corresponded to fragments of the carbohydrate side chain. In particular, the sialoglycopeptide (4718.887 Da) produced a recognizable fragment at m/z 292 (Fig. 5⇓ , top panel), which is indicative of the presence of sialic acid in the carbohydrate chain. On the other hand, no trace of this fragment was observed in the spectrum of the asialoglycopeptide (4161.730 Da; Fig. 5⇓ , bottom panel). The very limited amounts of analytes obtained by in-gel digestion afforded CID spectra with low signal-to-noise ratios. The limited availability of CSF fractions did not allow for a comprehensive characterization of the putative glycan forms attached to the different sites.
Carboxypeptidase Y treatment followed by MALDI TOF-TOF analysis further confirmed the assignment of the glycopeptide with masses of 4718 Da detected in sialotransferrin spots. The peak signals of glycopeptides with masses of 3680 and 2692 Da were too weak to detect their product after carboxypeptidase treatment (data not shown).
The distribution of transferrin isoforms in serum has been well characterized, in part because of their inherent abundance and easy access, but relatively little is known about transferrin isoforms in the central nervous system and CSF. Our results confirm earlier observations that, unlike serum, normal CSF contains both sialo and asialo forms of transferrin (28)(32). Although the sialotransferrins in our samples appear to be closely related, if not structurally identical, to those described in serum, there are strong indications that asialotransferrin isoforms may be brain specific. The accurate glycopeptide mass of 4161.686 Da detected in the tryptic digests of asialotransferrin spots is consistent with a structure containing 2 GlcNAc, 3 Man, 3 GlcNAc, and 1 Fuc connected to Asn630, as reported by a group that used a total of 100 mL of pooled CSF samples to purify ∼100 μg of asialotransferrin(28). The same marker glycopeptide was not observed in comprehensive studies of glycosylation of serum transferrin, including studies of patients with congenital disorders of glycosylation whose sera contained abnormal asialotransferrin isoforms(29). Whether this same carbohydrate is also attached to Asn432 remains to be demonstrated. The asialotransferrin samples were found to contain small amounts of a product with an accurate mass of 2692.177 Da, which could possibly contain glycosylated Asn432 (Table 2⇑ , second asialo isoform). Although this variant of asialotransferrin glycosylation has been described in the serum of a patient with an unexplained congenital disorder of glycosylation(29), its occurrence in CSF asialotransferrin still awaits further investigation.
The origin of both asialo- and sialotransferrin isoforms in CSF is unknown. It is presumed that sialotransferrin may be present in CSF because of a small transfer of serum-derived transferrin across the blood–brain and blood–CSF barriers. On the other hand, the occurrence of asialotransferrin in CSF is not well understood. In all of the control and non-CACH/VWM samples analyzed in our study, asialotransferrin represented 8%–35% of total transferrin, whereas it amounted to only 0.5%–5% of the total in CACH/VWM samples (see the table in the online Data Supplement). It has been suggested that asialo forms could result from partial metabolic degradation of the serum sialotransferrin transferred to the brain compartment (33). Others have suggested de novo synthesis by brain cells(28)(34). Hence, we hypothesize that the decrease in CSF asialotransferrin in patients with CACH is attributable either to an abnormal turnover of CSF transferrin or to reduced de novo production from local cells. This decrease in CSF asialotransferrin has, to our knowledge, not been described in any other disorders.
In conclusion, the combination of 2-DGE and MS analysis has allowed us to highlight characteristic and unambiguous variations in the distribution of transferrin isoforms in healthy controls vs individuals with CACH/VWM. Seven of 7 patients with mutation-confirmed CACH/VWM showed low to nearly undetectable amounts of asialotransferrin in their CSF, unlike 23 unaffected controls. This observation is based on a small sample size, and a larger number of patient samples will be needed for full validation of this observation. However, the CSF-transferrin spot patterns proposed in this study may serve as valuable biomarkers for diagnosis or preliminary screening before gene sequencing. The presence of such biomarkers could be used to classify patients with this neurodegenerative disease and could provide the basis for renewed efforts aimed at understanding the underlying pathophysiology of the disorder.
This work was supported in part by a Children’s Health Research Center, NIH-supported grant (K12HD001399), by grants from the NIH (HD-P30-40677 Child Health Research Career Development Award; 1P30HD40677-01 Mental Retardation and Developmental Disabilities Research Center), and by the Parson Family Foundation. The participation of D.F. and K.A.K. was supported by the National Institutes of Health (R01-GM643208). The participation of R.S. was supported in part by the Intramural Research Program of the NIH, NINDS.
1 Position numbering is for the full-length transferrin precursor (the mature protein starts at position 20).
2 Monoisotopic masses.
3 Difference between observed and theoretical mass.
4 ND, sequence not detected.
1 FTMS-observed masses are monoisotopic; MALDI-observed masses are averages.
2 Symbolic representation of monosaccharide is adopted from Essentials of Glycobiology by Varki et al. (35): ▪, GlcNAc; ○, Man; •, Gal; ⧫, sialic acid (NeuNAc); ▵, Fuc.
3 The glycosylation site is shown in bold in the peptide sequence. The linkage residue is shown in parentheses.
4 Theoretical masses are monoisotopic.
5 Mass corresponds to the difference between FTMS-observed mass and theoretical mass.
6 ND, not detected.
↵1 Nonstandard abbreviations: CACH, childhood-onset ataxia and central hypomyelination; VWM, vanishing white matter disease; CSF, cerebrospinal fluid; 2-DGE, 2-dimensional gel electrophoresis; MS, mass spectrometry; TFA, trifluoroacetic acid; MALDI, matrix-assisted laser desorption/ionization; MS/MS, tandem mass spectrometry; FTMS, Fourier transform mass spectrometry; SORI, sustained off-resonance irradiation; CID, collision-induced dissociation; TOF, time-of-flight; GlcNAc, N-acetylglucosamine; Man, mannose; Gal, galactose; NeuNAc, N-acetylneuraminic acid (sialic acid); and Fuc, fucose.
- © 2005 The American Association for Clinical Chemistry