TDP-43 in Cerebrospinal Fluid of Patients With Frontotemporal Lobar Degeneration and Amyotrophic Lateral Sclerosis

Background  Recently, TAR DNA-binding protein 43 (TDP-43) was identified as the major component of ubiquitin-positive tau-negative neuronal and glial inclusions in the most common form of frontotemporal lobar degeneration (FTLD) and in amyotrophic lateral sclerosis (ALS). It was demonstrated that different TDP-43 profiles correspond to clinical phenotypes of FTLD or ALS subgroups, and the differential diagnostic potential of TDP-43 was suggested.

Objectives  To examine TDP-43 in cerebrospinal fluid (CSF) and to analyze whether it could serve as a diagnostic marker.

Design  We characterized CSF TDP-43 by immunoblot using different TDP-43 antibodies and determined the relative TDP-43 levels in CSF samples from patients.

Setting  Academic research.

Patients  Twelve patients with FTLD, 15 patients with ALS, 9 patients with ALS plus FTLD, 3 patients with ALS plus additional signs of frontal disinhibition, and 13 control subjects.

Main Outcome Measures  Results of TDP-43 immunoblot.

Results  Polyclonal TDP-43 antibodies recognized a 45-kDa band in all analyzed samples. Two monoclonal and N-terminus–specific antibodies did not detect any specific bands, but C-terminus–specific antibodies detected a 45-kDa band and additional bands at approximately 20 kDa in all CSF samples. Relative quantification of 45-kDa bands revealed significant differences among the diagnostic groups (P =.046). Specifically, patients with ALS (P =.03) and FTLD (P =.02) had higher TDP-43 levels than controls but with a prominent overlap of values.

Conclusion  Although there is no evidence of pathologically altered TDP-43 proteins in CSF, TDP-43 levels in CSF might aid in characterizing subgroups of patients across the ALS and FTLD disease spectrum.

Frontotemporal lobar degeneration (FTLD) is the second most common cause of dementia affecting individuals younger than 65 years.¹ On a cellular pathologic level, most FTLD cases are characterized by the presence of ubiquitin-positive inclusions. Because of the nature of aggregated constituents of these inclusions, for which key roles in pathologic mechanisms are proposed, FTLD was categorized into 2 major groups with and without tau. In the FTLD-tau group, filamentous tau proteins form the disease inclusions.² Recently, pathologically phosphorylated and ubiquitinated TAR DNA-binding protein 43 (TDP-43) was identified as a major pathologic protein of sporadic and familial FTLD with ubiquitin-positive tau-negative inclusions (FTLD-U),^3,4 which constitutes more than 50% of all FTLD cases.⁵

The clinical picture of patients with FTLD-tau and FTLD-U is heterogeneous. Besides behavioral and personality changes, patients manifest language disturbances in primary progressive aphasia with progressive nonfluent aphasia and semantic dementia. In the case of corticobasal degeneration and progressive supranuclear palsy, behavioral abnormalities are accompanied by extrapyramidal features.¹ Frontotemporal lobar degeneration can be associated with neurodegeneration of motor neurons in motor cortex, brainstem, and spinal cord, leading to a syndrome with features of amyotrophic lateral sclerosis (ALS) and FTLD.⁶ In ALS, the most common type of motor neuron disorder (MND), upper and lower motor neurons are affected, and most patients die of respiratory failure on average about 3 years after symptom onset.^7,8 Furthermore, approximately 15% of patients with ALS develop dementia categorized as FTLD.⁹ TDP-43 is not only the main constituent of inclusions in FTLD-U with and without ALS but is also present in sporadic ALS, ALS with dementia, and SOD1-negative ALS.^3,10,11 This observation led to the hypothesis that TDP-43 is a common pathologic substrate in these diseases, implicating similar pathophysiologic mechanisms, despite significant clinical, genetic, and neuropathologic heterogeneity of FTLD-U and ALS.^3,12

So far (to our knowledge), no specific laboratory marker exists for disease progression or for differential diagnostic use in ALS or FTLD. Because pathologic changes in the brain and spinal cord can be reflected by altered levels of proteins or other analytes in cerebrospinal fluid (CSF), we analyzed CSF from patients having FTLD with and without ALS, from patients having ALS with and without FTLD or signs of frontal disinhibition, and from control subjects to determine if TDP-43 could be detected in CSF and if assaying CSF TDP-43 could be used as a biomarker for the diagnosis, staging, or care of patients with FTLD-U or ALS.

All 52 CSF samples analyzed in this study were obtained from patients attending the general outpatient clinic, the outpatient memory clinic, or the outpatient clinic for MND (Department of Neurology, University of Ulm, Ulm, Germany) from January 2006 to December 2007. Collection and analysis of CSF samples were approved by the ethics committee. Routine CSF data such as albumin concentration and IgG concentration were available for all samples. All individuals, or their relative in the case of patients with dementia, gave written informed consent to their participation in the study and underwent clinical, neurologic, and neuroradiologic examinations and a short neuropsychological screening, including the Mini-Mental State Examination¹³ to investigate global cognitive functioning. If deterioration had been suggested, a detailed psychometric test battery covering executive functions, memory, constructional abilities, premorbid verbal intelligence, and depression¹⁴ was administered to assess more specifically for cognitive impairment and frontotemporal degeneration. The diagnoses of all patients were made in accord with the consensus criteria for FTLD¹⁵ and on the basis of DSM-IV criteria and were established by neurologists (C.H., A.D.S., C.A.F.v.A, A.L., and M.O.) in cooperation with a neuropsychologist (I.U.), both blinded to the neurochemical outcome measures. Diagnosis of ALS was made according to the El Escorial criteria of Pradat and Bruneteau.¹⁶

The FTLD group consisted of 12 patients (7 men and 5 women). The mean (SD) age of the patients at the time of CSF sampling was 68 (8.6) years. The diagnosis of frontotemporal degeneration was made in 11 patients, and 1 patient had primary progressive aphasia subtype.

The diagnosis was supported in 11 of 12 patients by fludeoxyglucose F 18 positron emission tomography. The results demonstrated reduced cortical glucose metabolism in the frontopolar, frontomesial, or frontotemporal region.

The ALS group consisted of 15 patients (9 men and 6 women). The mean (SD) age was 48 (7.1) years. Eight patients were diagnosed as having laboratory-confirmed ALS, 5 patients had clinically probable ALS, 1 patient had definitive ALS with a spinal course, and 1 patient had definitive ALS with bulbar progress. Ten of 15 patients with ALS were classified as having spinal disease, 3 patients as having bulbar disease, and 2 patients as having flail arm syndrome.

The group of patients with ALS plus additional signs of frontal disinhibition (ALS plus DI) comprised 3 women having a mean (SD) age of 63 (14.0) years. These patients exhibited additional clinical signs of frontal disinhibition without fulfilling the diagnosis of FTLD.

The group of patients with ALS plus FTLD comprised 9 patients (5 men and 4 women). The mean (SD) age was 63 (7.1) years. Six patients were classified as having the spinal form and 3 patients as having the bulbar form of ALS. These patients fulfilled diagnostic criteria for FTLD.¹⁵

The group of controls comprised 13 patients (6 men and 7 women) with a mean (SD) age of 60 (8.0) years. The final diagnoses of the patients were as follows: complex focal seizures (n = 3), polymyalgia rheumatica (n = 2), polyneuropathy (n = 3), carcinoma (n = 1), neuropathia vestibularis (n = 1), depression (n = 1), migraine (n = 1), and dissociative disorder (n = 1).

Cerebrospinal fluid samples were stored at −80°C until analysis, at which time they were thawed for study. Identical volumes of 50 μL of native CSF were acetone precipitated.

IgG and albumin depletion was performed according to the manufacturer's instructions (GE Healthcare, Chalfont St. Giles, United Kingdom). Purified human IgG and albumin were obtained from Sigma-Aldrich Inc (St Louis, Missouri).

Murine neuroblastoma cells were lysed in radioimmunoprecipitation assay (RIPA) buffer (150mM sodium chloride, 20mM Tris [pH 7.4], 1% NP-40, 0.05% Triton X-100, 0.5% sodium desoxycholate, and 0.5M EDTA). The homogenate served as a control and as an internal Western immunoblot standard.

Mouse whole brain was homogenized in phosphate-buffered saline (PBS) (1 mL/0.1 g of tissue) solution containing aprotinin (1 μg/mL), phenylmethylsulfonyl fluoride (0.2mM), and leupeptin (0.5 μg/mL) and was sonicated for 30 seconds. After centrifugation at 20 000g for 10 minutes at 4°C, the supernatant was retained, and the protein concentration was determined by bicinchoninic acid assay (BCA; Sigma-Aldrich Inc, St Louis, Missouri).

Urea fractions were prepared from frozen frontal cortex of a patient with FTLD-U. The sequential extraction protocol has been published previously.³

Samples were reconstituted or mixed with sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (Roti-load 1; Carl Roth GmbH, Karlsruhe, Germany) to a final concentration of 2.5% mercaptoethanol. They were boiled for 5 minutes before electrophoresis.

Proteins were separated on Laemmli gels with 12% acrylamide in the separation gel and with 4% acrylamide in the stacking gel. Electrophoresis was performed at 25 mA per gel for about 90 minutes. Proteins were transferred to polyvinylidene difluoride membranes (Millipore Corporation, Bedford, Massachusetts) by semidry blot. Membranes were blocked with PBS and 0.075% polysorbate 20 (Tween-20) containing 5% dry milk powder (Bio-Rad, Hercules, California) and were then probed with anti–TDP-43 antibodies in blocking solution. Affinity-purified polyclonal rabbit antibody was raised against amino acids 1 through 260 of recombinant TDP-43 (1:2000 and 1:10 000 to 1:1000; Proteintech Group Inc, Chicago, Illinois). Monoclonal TDP-43 antibody clone 2E2-D3 specific for amino acids 205 through 255¹⁷ (1:1000; Abnova, Taipei City, Taiwan) was also used. Polyclonal rabbit antisera were raised against N-terminus amino acids 6 through 24 or against C-terminus amino acids 396 through 414 of TDP-43 (1:5000 for both). Immunoblots were incubated with primary antibody overnight at 4°C and for 1 hour at room temperature after 3 washing steps with peroxidase-conjugated goat antirabbit (Dianova, Hamburg, Germany) or antimouse (DAKO, Glostrup, Denmark) secondary antibody. A Western blot detection reagent (ECL Plus; GE Healthcare) was used as a substrate, and chemiluminescence was measured with a charge-coupled device camera (LAS-1000; Fujifilm, Tokyo, Japan).

The 45-kDa TDP-43 bands from CSF detected by rabbit TDP-43 antibody (Proteintech Group Inc) were quantified in relation to a fixed amount of mouse neuroblastoma RIPA cell homogenate serving as an internal standard. Band volumes (adjusted for membrane background) were determined using commercially available software (Quantity One, Bio-Rad) and were calculated relative to the volume of the 46-kDa band present in mouse neuroblastoma cell homogenate separated in parallel on the same gel. The TDP-43 band of each patient's CSF was quantified from duplicates or triplicates run on different gels and days. Western blots in which the standard band density differed more than 30% from the mean value of standard bands determined in all immunoblots were discarded.

Analyses for significant differences in a given variable among all tested groups (Kruskal-Wallis test) or between 2 groups (Mann-Whitney test) were calculated using commercially available statistical software (SigmaStat Software; SigmaStat, asknet AG, Karlsruhe, Germany). Correlation between variables was examined using Spearman rank correlation. P < .05 was considered statistically significant.

Two bands were regularly detected by rabbit polyclonal anti–TDP-43 antibody in immunoblots of CSF from patients and from controls (Figure 1). The upper band migrates at 45 kDa, which is similar to the 46-kDa band detected in mouse cell lysates and higher than the 28-kDa band detected in mouse brain homogenate. In the urea fraction extracted from FTLD-U brain tissue, in addition to a 47-kDa band we found an approximately 50-kDa band of the same intensity, 2 weaker bands at 44 and approximately 60 kDa, and a high relative molecular mass smear (Figure 1B). No signals of corresponding molecular weight were detected in CSF.

The 28-kDa band present in CSF from all patients and controls quantitatively disappears by depletion of albumin and IgG from native CSF before precipitation, SDS-PAGE, and immunoblot. Intensities of the 45-kDa band in CSF were unaffected by removal of IgG and albumin. Analyzing purified human IgG and albumin (the most abundant proteins of CSF) using the polyclonal TDP-43 antibody, a band with exactly the same molecular weight of 28 kDa was detected in IgG, whereas no band was detected in purified albumin (Figure 1). Therefore, the antibody cross-reacts with the IgG light chain. Applying the commercially available monoclonal antibody raised against amino acids 1 through 260 of recombinant TDP-43 and uncharacterized with regard to the binding site, we regularly detected a 26-kDa band, a 44-kDa band, and bands of approximately 47, 50, and 60 kDa, and high relative molecular mass smear in urea fractions. There was no specific immunoreactivity in any patient or control CSF using these monoclonal antibodies in dilutions of up to 1:1000 and using chemiluminescence exposure times of up to 20 minutes (Figure 2). Several weak bands were detected in CSF using clone 2E2-D3 monoclonal TDP-43 antibody, mainly representing IgG and albumin. In some immunoblots of CSF, we found a trace band at about 42 kDa (Figure 3B) but no bands between 43 and 50 kDa. To exclude that the 45-kDa band in CSF detected by polyclonal antibodies is unspecific and unrelated to TDP-43, we subjected CSF immunoblots to immunodetection with polyclonal antibodies specific for the N-terminus or C-terminus of TDP-43. The 46-kDa band in murine neuroblastoma cell homogenate was strongly recognized by both antibodies. No specific bands were detected in CSF using the N-terminus–specific antiserum (Figure 3C). In contrast, C-terminus–specific antibodies bound to a protein band at 45 kDa in CSF from the same 5 samples analyzed using the N-terminus–specific antibody. In addition, protein bands at 20 kDa in 12% and at less than 20 kDa in 15% sodium dodecyl sulfate–polyacrylamide gels were detected using the C-terminus–specific antibody (Figure 3D).

Intensities of the 45-kDa band recognized in CSF by TDP-43 polyclonal antibody were quantified in samples from 52 patients with FTLD, ALS, ALS plus FTLD, and ALS plus DI and from controls without dementia or MND. The band was stable at various preanalytic conditions. Neither storage of up to 2 days at 4°C after lumbar puncture nor 3 freeze-thaw cycles affected detected band intensities. Sample preparation affected neither size nor number of bands. Neither pretreatment of CSF with RIPA lysis buffer and subsequent acetone precipitation nor acetone precipitation of mouse cell lysates affected the molecular weight of the bands detected by polyclonal TDP-43 antibodies (data not shown).

The ALS and ALS plus FTLD groups comprised younger patients, especially compared with the FTLD group (Table 1). No correlation was noted between patient age and relative 45-kDa TDP-43 levels detected by rabbit TDP-43 antibodies in CSF immunoblots (Spearman rank correlation coefficient, −0.153; P =.28).

There was a wide variation of TDP-43 levels among our CSF samples (Table 2 and Figure 4). TDP-43 levels ranged from 7% to 164% (median, 60%) in the ALS group, 26% to 92% (median, 63%) in the FTLD group, 9% to 105% (median, 24%) in the ALS plus FTLD group, and 5% to 79% (median, 28%) in the control group. Two of 3 patients in the ALS plus DI group had low TDP-43 levels (16% and 17%), whereas 1 patient had a level of 100%. The mean (SD) values are given in Table 2. Statistical analysis revealed significant differences among all tested groups (P =.046). The ALS plus DI group was omitted from this analysis because of the small sample.

TDP-43 levels were increased in the ALS and FTLD groups compared with controls (P =.03 and P =.02, respectively). In the ALS plus FTLD group, intermediate levels of TDP-43 were found that were not statistically different from those in the ALS group (P =.15), FTLD group (P =.13), or control group (P =.89). There was no correlation with age, nor was there a relationship between TDP-43 level and the clinical phenotype of patients with ALS (bulbar vs spinal). There was no significant correlation between the CSF IgG concentration or the CSF to serum albumin ratio and the relative TDP-43 band intensity in any of the analyzed patient groups.

TDP-43, a nuclear protein that putatively functions in regulating transcription and alternative splicing, is a main component of inclusions in most familial and sporadic FTLD-U cases, including FTLD-U subgroups such as FTLD-U plus MND and FTLD-U manifesting clinically as semantic dementia or progressive nonfluent aphasia.^3,4,18 Moreover, examination of subgroups with MND revealed TDP-43–immunopositive neuronal and glial inclusions in sporadic ALS, ALS with dementia, and superoxide dismutase 1, soluble SOD1-negative familial ALS,¹¹ suggesting that FTLD-U and ALS represent a clinicopathologic spectrum of disorders sharing similar pathomechanisms.^3,11,19

Demonstration of TDP-43 immunoreactivity in neuronal and glial inclusions is useful in the differential diagnosis of ALS or ALS plus FTLD from other MNDs affecting upper and lower motor neurons and confirms subclinical MND in FTLD without clinical or pathologic evidence of MND.¹⁰ Proteins involved in pathophysiologically regulated pathways correlate with altered CSF concentrations and may be helpful in the differential diagnosis. In our immunoblot analysis applying polyclonal antibodies against TDP-43, a specific 45-kDa band was regularly present in all analyzed CSF samples. In immunoblots of urea fractions extracted from central nervous system tissue of patients with FTLD-U and ALS, Neumann et al³ described a disease-specific biochemical profile of TDP-43 with bands of 24 and of 26 kDa, phosphorylated full-length TDP-43 of 45 kDa and high molecular smear, and a physiologic TDP-43 isoform at 43 kDa. No truncated isoform of TDP-43 at approximately 25 kDa or bands comparable to phosphorylated full-length TDP-43 were detected in CSF, suggesting that the pathologic signature of TDP-43 described in tissue fractions of FTLD-U and ALS seems not to be reflected in CSF.

Surprisingly, both monoclonal TDP-43 antibodies applied herein failed to detect protein bands in immunoblots of up to 200 μL of CSF. This could mean that the monoclonal antibodies did not bind to CSF TDP-43, possibly because of low affinity or epitope masking or that binding may be dependent on sample preparation procedure. To characterize the 45-kDa protein recognized by polyclonal antibodies in more detail, we applied N-terminus– and C-terminus–specific polyclonal antibodies. Whereas N-terminus–binding antibodies did not specifically recognize protein bands in 50 μL of CSF, C-terminus–specific antiserum detected the 45-kDa protein band in a similar fashion to the band seen with polyclonal antibodies raised against amino acids 1 through 260. The fact that the TDP-43 band in CSF migrated lower than the full-length TDP-43 in mouse cells and brains, combined with the lack of staining using N-terminus–specific antibodies, suggests that the detected TDP-43 species in CSF is truncated at the N-terminus.

A yet undescribed protein band (to our knowledge) migrating at or above 20 kDa that reacts with C-terminus antibodies was present in CSF. Analysis of representative individuals from our diagnostic groups revealed that the 20-kDa TDP-43 fragment is generally present in CSF. The fact that TDP-43 can be found in CSF from individuals without dementia or ALS points to a physiologic process of TDP-43 release into the CSF.

TDP-43 levels in CSF tend to be elevated in disease states, which is in accord with increased TDP-43 gene expression in FTLD plus MND and in FTLD-U.²⁰ Relative quantification of 45-kDa bands was significantly different among our diagnostic groups (P =.046), based primarily on increased levels in diseased vs nondiseased individuals (P =.045). The absence of a correlation between TDP-43 levels in CSF and the CSF to serum albumin ratio points to an intrathecal origin of CSF TDP-43. On a statistical basis, determination of relative levels of 45-kDa TDP-43 allows discrimination of ALS and FTLD samples from controls (P =.03 and P =.02, respectively). The wide range of TDP-43 levels in FTLD CSF might be due to the heterogeneity of this disease group, which statistically consists of approximately 50% of patients without pathologic TDP-43 and 30% to 50% of patients with pathologic TDP-43. TDP-43 levels in ALS plus FTLD samples and in ALS plus DI samples were in the range of those of controls.

So far (to our knowledge), no disease specific markers for ALS or FTLD have been identified except genetic mutations pathogenic for ALS or FLTD.¹¹ Indeed, mutations in the TDP-43 gene (TARDBP) have been identified recently that are pathogenic for sporadic and familial ALS,^21,22 adding further credence to the view that mechanisms of neurodegeneration in ALS are linked to pathologic TDP-43.^3,11 Some studies^23-29 describe a slightly different total tau and phospho-tau protein profile, and varying levels of amyloid β (Aβ) 42, Aβ40, or Aβ38 are suggested to be characteristic of FTLD. In an analysis of FTLD subgroups, soluble tau was found to be reduced in dementia lacking distinctive histologic findings and in FTLD plus MND or MND–type inclusions.³⁰ In ALS, decreasing serum levels of S-100B protein were found in a follow-up investigation of ALS,^31,32 as well as elevated levels of neurofilament in the CSF.³³ However, these variations have limited differential diagnostic potential.

Our results suggest that disease-associated TDP-43 immunoreactivity in brain tissue is reflected by elevated levels of TDP-43 in CSF. Although we demonstrated a significant difference among the FTLD, ALS, and control groups, the TDP-43 immunoblot methods used herein did not distinguish among disease groups with differential diagnostic accuracy. Further CSF studies of additional patients (including those with postmortem-verified diagnoses) using quantitative TDP-43 assays (eg, enzyme-linked immunosorbent assay) are needed to determine if measurement of TDP-43 levels in CSF may be informative in multifactorial analyses with other biomarkers to aid in the diagnosis of different forms of FTLD and ALS during life, as this will expedite efforts to introduce specific disease-modifying therapies when they become available.

Correspondence: Markus Otto, MD, Department of Neurology, University of Ulm, Steinhövelstrasse 1, 89075 Ulm, Germany (markus.otto@uni-ulm.de).

Accepted for Publication: June 13, 2008.

Author Contributions:Study concept and design: Steinacker, Jesse, Lehnert, Neumann, and Otto. Acquisition of data: Steinacker, Hendrich, Sperfeld, von Arnim, Pabst, Uttner, Tumani, Ludolph, and Otto. Analysis and interpretation of data: Steinacker, Lee, Trojanowski, Kretzschmar, Neumann, and Otto. Drafting of the manuscript: Steinacker and Otto. Critical revision of the manuscript for important intellectual content: Steinacker, Hendrich, Sperfeld, Jesse, von Arnim, Lehnert, Pabst, Uttner, Tumani, Lee, Trojanowski, Kretzschmar, Ludolph, Neumann, and Otto. Statistical analysis: Steinacker and Otto. Obtained funding: Steinacker and Otto. Administrative, technical, and material support: Steinacker, Hendrich, Sperfeld, von Arnim, Lehnert, Pabst, Tumani, Kretzschmar, Ludolph, Neumann, and Otto. Study supervision: Steinacker, Sperfeld, Uttner, Lee, Trojanowski, Ludolph, and Otto.

Financial Disclosure: None reported.

Funding/Support: This study was supported in part by grant P-LS-Prot/42 from the Landesstiftung Baden-Württemberg (Drs Steinacker and Otto), by grants cNeupro and NeuroTAS from the European Commission (Dr Otto), by grant 01GI0704 from the Federal Ministry of Education and Research (Drs Kretzschmar and Neumann) and the Heidelberger Akademie der Wissenschaften (Dr von Arnim), and by grants AG-10124 and AG-17586 from the National Institutes of Health (Drs Lee and Trojanowski).

Additional Contributions: We thank all the physicians who referred patients to our clinic.