Frontotemporal lobar degeneration (FTLD) is the primary cause of early onset dementia after Alzheimer’s disease (AD) [1]. Worldwide prevalence of FTLD is underestimated due to difficult diagnosis complicated by clinical, neuropathological and genetic heterogeneity [2-5]. FTLD is clinically subdivided into behavioral variant frontotemporal dementia (bvFTD) [6] and language variant primary progressive aphasia (PPA) which in its turn comprises three variants [7]. The most common genetic etiologies resulting in FTLD include mutations in tau (MAPT) and progranulin genes (GRN) and a repeat expansion in C9orf72 [8-13]. Molecular pathologies underlying FTLD include aggregation from tau (FTLD-tau) or fused-in-sarcoma proteins (FTLD-FUS), accounting for approximately 45% and <5% of patients respectively [14]. The major pathological subtype, accounting for approximately 50% of FTLD population, is FTLD-TDP where patients have brain inclusions of transactive response DNA-binding protein of 43 kDa (TDP-43) [15,16]. Under physiological conditions, TDP-43 is a predominantly nuclear protein and its role in transcription and splicing regulation is well characterized [17]. In FTLD-TDP, TDP-43 is redistributed to the cytoplasm where it forms intraneuronal inclusions. This leads to an apparent loss of nuclear TDP-43 function, while the accumulation itself is expected to be toxic [18,19]. Furthermore, aggregation of TDP-43 is also characteristic for amyotrophic lateral sclerosis (ALS) [16], and clinical and genetic overlap between both disorders has corroborated their association in an FTLD-ALS spectrum [20]. Noteworthy, other neurodegenerative disorders can present with TDP-43 pathology as secondary feature, and this is the case in 20-50% of patients with AD and related tauopathies [14,21]. A comorbid TDP-43 pathology is reported to worsen neurodegeneration independently of AD pathology, leading to a more severe clinical presentation of dementia [22].

While mutations in a specific gene induce an associated molecular pathology, no strict relationship exists between clinical FTLD subtype and underlying proteinopathy [15,23]. Indeed, clinical symptoms rather reflect affected brain regions, which is especially exemplified in the heterogeneity of clinical FTLD. Moreover, it should be noted that up to 25% of clinical FTLD is actually due to atypical presentation of AD pathology [14,24]. The two-way clinicopathological association between FTLD(−TDP) and AD shows there is an urgent need for biomarkers that allow early and differential diagnosis of FTLD. A promising approach is quantification of disease-specific biochemical markers present in biofluids (cerebrospinal fluid (CSF) and blood) [23]. At present, well-characterized and validated diagnostic markers specific to FTLD pathology do not exist, with the exception of decreased progranulin concentrations for GRN mutation-related FTLD, a subgroup of FTLD-TDP [25,26]. A high-ranking candidate to become a biomarker for all FTLD-TDP patients is TDP-43 itself. Because of low absolute levels, quantitative analysis of TDP-43 in biofluids will demand a very sensitive immunoassay, preferably specific for pathological TDP-43 [27]. The TDP-43 protein comprises two RNA-recognition motives (RRM1-&2) and a glycine-rich C-terminal domain (Figure 1) [17,28]. Pathological aggregation of TDP-43 is regulated by both N-terminal and C-terminal regions, but also includes modifications like truncation, ubiquitination and phosphorylation [16,29-32]. Reported truncation sites are located inside the RRM2 and include Arg208, Asp218 and Asp247 [33-35] while major phosphorylation sites are serine residues located near the C-terminal end of TDP-43 [32,36].

The primary objective of this systematic review is to identify which antibodies have previously been described to detect TDP-43 pathology. These antibodies are expected to be suitable for defining the characteristic profile of pathological TDP-43 in human brain and biofluids, using immunostaining and immunoblotting. Additionally, antibodies specific for TDP-43 pathology hold potential to be used in immunoassays for quantification of TDP-43 and its pathological isoforms in biofluids. In this regard, a secondary objective is to review existing immunoassays with reference to specific antibody pairs.

A systematic review of existing literature was performed by searching PubMed and Web of Science using the predefined keywords “TDP 43” and variations like “transactive response DNA binding protein” or “TAR DNA binding protein” in publication title, and keywords indicating the use of antibodies (such as “antibody”, “immunoblot”, “immunohistochemistry” “immunoassay” or “ELISA” or, to broaden the search, “aggregate” or “inclusion”) in title or abstract for PubMed and as topic for Web of Science. Only original research papers published between 06/10/2006 (first report of TDP-43 as pathological protein in disease [16]) and 31/12/2014 were considered. For each English research paper the methods section was checked to select papers able to answer our learning objective.

This review was undertaken to identify the most suitable antibodies for future research of TDP-43, while evaluating ELISAs already used for its quantification. As much as 29 antibodies met the selection criteria and are thus reported to detect TDP-43 pathology. In contrast, only five different ELISAs have been published and their antibody combinations leave room for improvement.

The majority of antibodies in the final selection are characterized by an epitope in the C-terminal region, and many of them are specific for phosphorylated TDP-43. These findings are expected, since C-terminal truncation and phosphorylation are hallmarks of aggregated TDP-43 in disease [16,30,32]. However, the observed result is not bidirectional, as several antibodies directed at pathological isoforms do not specifically reflect TDP-43 pathology. This is the case for those antibodies that only obtained double negative scores, but is probably also the case for some antibodies in the secondary selection. The fact that certain antibodies have only been used once has two possible reasons. First, for “older” antibodies in the list, it might be an indication for their lack of sensitivity and specificity (false positives). This is a feature addressed by groups generating in-house antibodies themselves, who make a selection of antibodies to use for further research after internal validation [32,36]. With this in mind, it is not unlikely that antibodies of lower affinity are also generated and made available by commercial companies. Alternatively, “recent” antibodies in the secondary selection have not yet had the chance to be used multiple times (false negatives). However, also in these cases there can be antibodies that will not be pursued further by the manufacturing group [47,54]. Of note, the occurrence of false negatives because of a lack of replication studies also applies to antibodies that have not been used in both immunoassays in one publication and therefore were not included in the scored antibody list.

Another reason for false positive and false negative results is the fact that antibodies are always used in a very specific setup. For example, antibodies used for studying exogenous expressed C-terminal fragments are not necessarily successful in detecting pathological TDP-43 in patients. Conversely, when antibodies are used to study patients without suspected TDP-43 pathology it is understandable that they will indeed not detect pathology. These remarks are supported by the fact that many antibodies in the primary selection do not consistently have a double positive score, and some are even reported with a double negative score. Finally, there is the potential occurrence of biased results, since three antibodies are considered “standard” antibodies for studying TDP-43 proteinopathies, and are thus widely used. These antibodies are 10782 (ProteinTech), 2E2-D3 (available from different manufacturers) and TIP-PTD-P01&-P02 (Cosmo Bio) [38,55]. This assumption is reflected in the systematic review, as together they account for almost half the reports using a well-characterized antibody. 10782 (115/506 reports, 22.7%) and 2E2-D3 (68/506 reports, 13.4%) have been available since the beginning of TDP-43 studies [16,56]. TIP-PTD-P01&-P02 (68/506 reports, 13.4%) was first reported by the group which generated it, in a publication initiating the branch of research on phosphorylated TDP-43 [32]. Despite being commonly used, these antibodies are not necessarily the best options when studying TDP-43 pathology. 10782 only has a double positive score in 56.0% of its scored reports (14/25). However, it has been successful in the development of an ELISA both as capture or detection antibody, and is capable of immunoprecipitation. These facts prove that the antibody has the desired affinity for total TDP-43. 2E2-D3 is reported with a double positive score in 50.0% of its scored reports (12/24), and this antibody has also been successfully used in ELISAs and immunoprecipitation. Moreover, its specificity for human TDP-43 makes it very suitable in research of TDP-43 in transgenic animal models [38]. Nevertheless, when looking to study pathological TDP-43, TIP-PTD-P01&-P02 has a better track record, with 70.6% of its scored reports having a double positive score (12/17).

A final point of discussion when selecting antibodies to study TDP-43 is the use of commercially available antibodies or the generation of antibodies in-house. Despite favorable antigenicity profiling of TDP-43, only a few well-characterized antibodies within the systematic search are available from companies (n = 24, Additional file 1: Table S2). Most of these antibodies are polyclonal (n = 19) of which 15 are directed at the C-terminal region (including five phosphorylation-specific). Of the well-characterized commercial monoclonal antibodies one maps in the N-terminal region, two recognize epitopes in the RRM2 of TDP-43 and two are phosphorylation-dependent. The lack of available antibodies specific for pathology has stimulated the generation of novel antibodies. A number of groups have been successful using this approach and five in-house generated antibodies are now also commercially available [32,36,57]. However, many more have only been used a few times by the original group, in collaborative efforts or have not been used again after their first report [30,47,54,58-64]. Noteworthy, most high-ranking antibodies in the selection with an epitope directed at pathologically modified TDP-43 are originally in-house generated. From a research perspective this is a logical starting point when undertaking the production of antibodies, as these antibodies are more likely to be suitable for the study of TDP-43 pathology. As a side note, in-house generation of antibodies requires the use of either peptide immunogens or recombinant protein immunogens. In practice, only peptides can be used to obtain antibodies specific for major protein modifications. Indeed, all reported antibodies directed at phosphorylated TDP-43 within this search are generated using a peptide immunogen. The predominance of antibodies generated with peptide antigens is also related to the propensity of recombinant TDP-43 protein to aggregate, which is especially the case for C-terminal fragments [33]. With regard to ELISA development, antibodies generated with recombinant protein antigens are more than suitable when developing an ELISA for whole-target quantification, but when looking to quantify different isoforms of one protein optimal results are to be expected with peptide antigens.

As far as reported immunoassays detecting TDP-43 and its isoforms in biofluids go, it is clear that improvements can and should be made. Merely looking at used combinations of antibodies, previous ELISAs were developed to reflect either total TDP-43 or phosphorylated TDP-43 levels. However, none of these assays accounted for the presence of C-terminal fragments, as these may have been quantified together with full-length (phosphorylated) TDP-43. To utilize TDP-43 as an effective biomarker applicable for human TDP-43 proteinopathies, it is very likely that all isoforms of TDP-43 will need to be quantified separately. From this point of view, the ELISA reported by Kwong et al. [47] is already a step in the right direction, although it will probably still detect full-length TDP-43 together with C-terminal fragments. To solely quantify full-length TDP-43 a less ambiguous combination of antibodies should be used, that is certain to exclude any C-terminal fragments. The same requirement applies for quantification of phosphorylated full-length TDP-43 without interference from phosphorylated C-terminal fragments. Although lone detection of C-terminal fragments seems difficult, a few antibodies are reported to be specific enough to potentially achieve this [47,60], and further efforts should be undertaken to quantify this pathological isoform. When this is made possible, an assay specific for phosphorylated C-terminal fragments can also be developed, completing the spectrum of TDP-43 pathology.

In the context of this systematic review it is necessary to comment on the recent publication by Feneberg et al. describing a limited role for TDP-43 as a diagnostic tool [52]. The main finding of this publication is that TDP-43 present in CSF is mainly blood-derived. This is based on an immunoblot analysis, which shows that the concentration of TDP-43 in blood is 200 times higher than that in CSF. Based on published ELISAs it is difficult to comment on the comparison of TDP-43 levels between CSF and plasma. Only one study has quantified TDP-43 in CSF and plasma of the same patients and here only relative concentrations are reported, so no conclusion can be made with regard to blood/CSF ratio [46]. However, the claim that TDP-43 is mainly blood-derived and is therefore of minor importance as diagnostic tool is weak. Other studies on biomarker levels measured in paired blood and CSF samples demonstrate that correlations are more complex than anticipated. For example, though progranulin is much more abundant in plasma and its plasma and CSF levels correlate, plasma progranulin levels can only account for a very small part of the variability of CSF progranulin levels [65].

A final remark, which is also supported by Feneberg et al. [52], is that exosomes present in CSF may contain TDP-43 which is not blood-derived and may therefore be more suitable to reflect TDP-43 pathology. Thus, even if it is confirmed that TDP-43 in CSF has limited diagnostic value, this would only result in the adjustment of future immunoassays to another target, namely exosome preparations. But first, further evidence needs to be gathered by independent researchers before this assumption can be considered valid.

This systematic review provides an overview of antibodies reported to detect TDP-43 pathology. These antibodies can be used in animal models and patient studies of TDP-43 proteinopathies, such as FTLD. Thus, researchers studying these disorders can choose highly sensitive and specific antibodies most suitable for their objective based on this review. Antibodies from the primary selection have consistently proven their functionality in detecting pathological TDP-43 in both immunostaining and immunoblotting. Antibodies from the secondary selection have the potential to detect TDP-43, but validation of their results is necessary before conclusions can be made regarding their future use. Additionally, high-ranking antibodies are leading candidates when selecting an antibody pair in future development of novel quantitative immunoassays for TDP-43 in biofluids or exosome preparations, as a possible biomarker for FTLD-TDP.