Mass spectrometric analysis of accumulated TDP-43 in amyotrophic lateral sclerosis brains

TDP-43 is the major disease-associated protein involved in the pathogenesis and progression of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration with ubiquitin-positive inclusions linked to TDP-43 pathology (FTLD-TDP). Abnormal phosphorylation, truncation and cytoplasmic mis-localization are known to be the characteristics for the aggregated forms of TDP-43, and gain of toxic abnormal TDP-43 or loss of function of physiological TDP-43 have been suggested as the cause of neurodegeneration. However, most of the post-translational modifications or truncation sites in the abnormal TDP-43 in brains of patients remain to be identified by protein chemical analysis. In this study, we carried out a highly sensitive liquid chromatography-mass spectrometry analysis of Sarkosyl-insoluble pathological TDP-43 from brains of ALS patients and identified several novel phosphorylation sites, deamidation sites, and cleavage sites. Almost all modifications were localized in the Gly-rich C-terminal half. Most of the cleavage sites identified in this study are novel and are located in N-terminal half, suggesting that these sites may be more accessible to proteolytic enzymes. The data obtained in this study provide a foundation for the molecular mechanisms of TDP-43 aggregation and ALS pathogenesis.

Immunoblotting of the Sarkosyl-insoluble fractions from FTLD and ALS cases using phosphospecific antibodies clearly demonstrated that hyperphosphorylated full-length TDP-43 of 45 kDa, smearing substances, and fragments at 18-25 and 35kDa are the major species of TDP-43 26,31,32 . We identified at least three C-terminal banding patterns that distinguish diseases with TDP-43 proteinopathy and reported that the banding pattern in different brain regions and spinal cord of individual patients is indistinguishable. Corresponding patterns of protease-resistant phosphorylated TDP-43 are also seen among the pathological phenotypes 26,27,32 . Recently, we showed that insoluble TDP-43 aggregates in brains of ALS and FTLD-TDP patients have prion-like properties 33 .
TDP-43 phosphorylation, cleavage and cytoplasmic mislocalization are all associated with TDP-43 aggregation. However it remains to be clarified whether these modifications are the cause of the disease. Recent reports on post-translational modifications, including fragmentation, were based on studies of cellular or animal models overexpressing TDP-43 or its derivatives [34][35][36][37] . However, TDP-43 aggregate formation in these artificial models may be different from that in the brains of patients. Therefore, it is important and essential for elucidation of the pathogenesis to identify the pathological events in TDP-43 truly occurred in the brain of patients. In this study, we carried out detailed molecular analysis of pathological TDP-43 aggregates in ALS brains.

Results
Immunoblot analysis revealed that abundant abnormal TDP-43 was recovered in the Sarkosyl-insoluble fractions of brains from both ALS cases as shown in Fig. 1A. A phosphorylation-independent anti-TDP-43 detected normal TDP-43 of 43 kD in the fractions of both ALS and control brains, but also detected abnormal TDP-43 bands, including full-length phosphorylated TDP-43 of 45 kD, ~25 kD fragments and smears in ALS brains (Fig. 1A). The pS409/410 antibody, which specifically recognizes the abnormal phosphorylation of Ser409 and 410 strongly reacted with these pathological TDP-43 bands and additional fragments of 18 ~ 24 kD (Fig. 1A). No such bands were detected in the same fraction of control brain. Intensities of pS409/410 immunoreactive bands in both ALS cases were 5 ~ 10 times higher compared with those from the other cases previously analyzed in our lab (data not shown). Therefore, we thought these two cases are good for the protein chemical analysis of pathological TDP-43.
After SDS-PAGE, the gel portions corresponding to these bands were cut out as shown in Fig. 1B and digested with trypsin and chymotrypsin. Digests were analyzed with nano-flow LC-MS/MS. As shown in Tables 1 and 2, we identified 229 peptides from brain of case 1 and 92 peptides from brain of case 2; this difference may reflect the differences in total amounts of pathological TDP-43 in these brains and the numbers of fractions. These peptides covered about 90% of the intact molecule (Tables 1 and 2). Identified peptides (representative), modifications and cleavage sites are shown and summarized in Figs 2 and 3. Further cleavage site peptides are shown in supplemental Fig. S1. Other proteins contaminating the Sarkosyl-insoluble fractions are listed in supplemental Table S1.
Analysis of Case 1. We identified at least four intrinsically cleaved TDP-43 peptides in the pathological TDP-43, i.e., three N-terminal peptides (blue arrows, Fig. 3A) and one C-terminal peptide (pink arrow). However, no caspase-cleaved peptides were detected in this study (Table 3). Ten serine residues were phosphorylated, as shown in Fig. 3A and Table 4. These phosphorylation sites were the same sites as previously reported 27,38 . Furthermore, deaminations of Asn and Gln residues and oxidation of Met residues were identified. Almost all modification sites were found in the Gly-rich C-terminal half.
Analysis of Case 2. We identified six intrinsically cleaved TDP-43 peptides in this case, i.e., five N-terminal peptides (blue arrows, Fig. 3B) and one C-terminal peptide (pink arrow). As shown in Fig. 3B and Table 4, 15  serine residues were phosphorylated and 21 Asn/Gln residues were deamidated. Moreover, we found that the 79Lys residue was ubiquitinated and 82Lys was acetylated. However, 145Lys and 192Lys were not detected 39 . Phosphorylation sites were the same sites as previously reported 27,38 . Again, almost all of these modification sites were localized in the Gly-rich C-terminal half.
Common modifications in cases 1 and 2. All modifications are summarized in Table 4. As described above, almost all modifications were localized in the Gly-rich C-terminal half. Furthermore, common modifications focused on 180-330 residues region of these cases, suggesting that the 180-330 region of TDP-43 had the same structure in both case 1 and case 2.

Discussion
Accumulation of filamentous inclusions composed of abnormally phosphorylated full-length TDP-43 (45 kDa) and its fragments (35 and 17-27 kDa) is a defining feature of TDP-43 proteinopathies 26 . It has been reported that overexpression of full-length TDP-43 with or without mutations in cultured cells and animals leads to fragmentation of the protein, generating 35-kDa (CTF-35) and 25-kDa (CTF-25) C-terminal fragments via caspase-mediated TDP-43 cleavage 30,34,40 . Furthermore, activation of calpain, a Ca 2+ -dependent cysteine protease, by upregulation of Ca 2+ -permeable AMPA receptors generates C-terminally cleaved TDP-43 fragments (~35 kDa) and causes mislocalization of TDP-43 in motor neurons 36 . Those reports noted that the ~35 kDa fragments were localized in the cytoplasm and formed insoluble aggregates. However, no such TDP-43 peptides cleaved by caspase or calpain reported were not detected in the pathological TDP-43 in the brains of patients in this study. On the other hand, uncleaved peptides by caspase or calpain were detected as the major molecules in the digests. Furthermore, several novel intrinsically cleaved TDP-43 peptides were identified in the pathological TDP-43, i.e., three and five N-terminal peptides (Cys164-Lys165, Tyr214-Gly215 and Asn279-Pro280 in case 1, and Arg55-Leu56, Val75-Asn76, Val108-Leu109, Val130-Leu131 and Asn279-Pro280 in case 2) (Figs 3A,B and 4 and Table 3) and two C-terminal peptide (Asn291-Ser292 in case 1 and Gly400-Phe401 in case 2) (Figs 3A,B and 4), strongly suggested that these cleaved fragments under pathological conditions was different from that under TDP-43-overexpressing conditions. Cellular expression of TDP-43 is tightly regulated at the transcriptional and post-translational levels. Under pathological conditions, however, formation of TDP-43 aggregates within the cell nucleus or cytoplasm results in reduced free nuclear TDP-43, and therefore the TDP-43 binding-receptor sensor detects a fall in protein levels and responds with increased TDP-43 production; the produced TDP-43 is then sequestered by TDP-43 aggregates 2 . Hence, TDP-43 production under pathological conditions does not necessarily represent overexpression. It is possible that TDP-43 aggregate formation under pathological conditions is different from that under TDP-43-overexpressing conditions. It has been recently reported that C9ORF72 repeat expansions in mice cause TDP-43 pathology. In these mice, insoluble phosphorylated TDP-43 was predominantly present in monomeric form and high-molecular-weight or truncated pTDP-43 species were not detected 41 . Further, recombinant full-length human TDP-43 forms structurally stable, spherical oligomers that are neurotoxic in vitro and in vivo 42 . Indeed, cellular aggregate formation or accumulation of TDP-43 C-terminal fragments (CTFs) is not primarily responsible for development of the observed disease phenotype in mutant or wild-type TDP-43 mice [43][44][45] . These findings suggest that full-length TDP-43, not the cleavage fragments, participates at an early stage in TDP-43 pathology. Fragmentation may occur after the accumulation of TDP-43.
We identified some intrinsically cleaved TDP-43 peptides at the N-terminus and C-terminus of Sarkosyl-insoluble TDP-43 in this study. These cleavage sites were neither the caspase cleavage sites nor the calpain cleavage sites of TDP-43 as reported previously 46,47 . Several cleavage sites deduced from these N-terminal peptides exist in the region of residues 55-280 as shown in Table 3 and Fig. 4. Therefore, full-length TDP-43 may be gradually processed at N-terminal sites after the formation of insoluble aggregates as shown in Fig. 4. Our findings that cleavage site is correlated with the distribution of molecular weight size are consistent with this idea. Further studies using antibodies for these cleavage sites will clarify the gradual processing mechanism.
In our previous studies, we identified phosphorylation sites by immunochemical methods 27 and also identified casein kinase-1 phosphorylation sites on recombinant TDP-43 38 . Almost all Ser and Thr residues in C-terminal Gly-rich half were able to be phosphorylated in vitro experiment 38 . In the present work, however, a part of accumulated TDP-43 was phosphorylated, not all. Non-phosphorylated peptides were major products from aggregated TDP-43 digests from these ALS brains. We protein-chemically identified several novel phosphorylation sites by directly analyzing the accumulated TDP-43 from two ALS patients. Phosphorylation sites identified in case 1 and case 2 corresponded to the sites identified in casein kinase-1 phosphorylated TDP-43. Although 9 phosphorylation sites were common to both cases, more phosphorylation sites were identified in case 2, as shown in Fig. 3A It has known that non-phosphorylated intact TDP-43 forms fibrous aggregate 42 , suggesting abnormal phosphorylation is not first event in TDP-43 pathology. The level of phosphorylation may change according to the time when TDP-43 aggregates exist in cytoplasm. Phosphorylation-triggered degradation is a common strategy for elimination of regulatory proteins in many important signaling processes. Phosphorylation of TDP-43 may occur in elimination process. The impairment of this process due to TDP-43 aggregation will induce abnormal phosphorylation of TDP-43 aggregate. As the time has passed, the level of phosphorylation seems to become more. Therefore, the individual difference may occur in the level of phosphorylation. Furthermore, it seems reasonable to speculate that heavily phosphorylated peptides may not be detected because of low recovery from the column and low ionization efficiency in mass analysis. S409/ S410-phosphorylated TDP-43 was immunochemically included in almost all fractions as shown in Fig. 1. However, we found small amount of peptide including S409 with phosphorylation in case 2, while we found no peptide S409/S410 with phosphorylation in case1. A peptide including S409/S410, which derived from chymotrypsin digestion, has at least 5 phosphorylation sites. We have already reported that TDP-43 C-terminal region including these phosphorylation sites was heavily phosphorylated 26,27,32 . Therefore, we might not detect C-terminal region peptide including S409/S410 residues. It is possible that there are some differences in phosphorylation states of accumulated TDP-43 and in the detection efficiency in the mass analysis between the cases.
Indeed, there were individual differences in other modifications, as shown in Fig. 3A,B. However, we found that common modifications in case 1 and case 2 were focused in the 180-330 region (Fig. 3A,B, and Table 4). This suggests that the 180-330 region of TDP-43 had the same structure in both case 1 and case 2.  Table 3. N-terminal sequences of abnormally accumulated TDP-43 fragments in human brain.
Interestingly, inclusions within the brain of ALS and FTLD-TDP patients are readily labelled with antibodies that recognize the C-terminus of TDP-43, but not with N-terminal TDP-43 antibodies. In contrast, spinal cord inclusions are labelled with both N-and C-terminal TDP-43 antibodies, suggesting that they are composed of full-length TDP-43 48 . This regional heterogeneity in terms of C-terminal fragment formation suggests that these fragments may not be necessary for TDP43-mediated neurodegeneration 49 . Furthermore, TDP-43 fragments generated during neurodegeneration were not C-terminal fragments, but rather were derived from a central portion of human TDP-43 45 . These results indicate that a common structure region, residues 180-330 of TDP-43, is significant for neurodegeneration It has already reported that a Gly-rich region (287-322), which contains multiple glycine repeats, may contribute significantly to fiber formation as well as aggregation propensity 50,51 . Recently, it has also been reported that the 274-414 GQN-rich region (C-terminal region) has prion-like properties 33,35 , and that RRM2 domain of TDP-43 plays a key role in forming proteinaceous aggregates 52 . The common structure region, residues 180-330 of TDP-43, includes RRM2 and a part of the GQN-rich region. Therefore, this common structure region may be the core region of TDP-43 aggregates. The identification of N-terminal and C-terminal peptides in this study   suggests that most of the cleavage sites exist in the N-terminal and C-terminal non-core regions. Proteolytic enzymes may readily access these regions and cleavage may occur gradually from distal regions. Our data provide important insight into the mechanism of TDP-43 accumulation, though further studies will be needed to clarify the molecular pathology of TDP-43 in detail.

Methods
Fractionation of accumulated TDP-43 in ALS brain. Sarkosyl-insoluble abnormal TDP-43 was prepared from brains of two ALS cases with abundant and widespread TDP-43 pathologies. Brain samples (0.5 g) from patients with ALS (case 1, 77-year-old male, disease duration 2.5 years and case 2, 57-year-old female, disease duration 2 years) and from vascular dementia (disease control, 88-year-old female) were homogenized in 10 ml of homogenization buffer (10 mM Tris-HCl, pH 7.5, containing 0.8 M NaCl, 1 mM EGTA, 1 mM dithiothreitol). Sarkosyl was added to the homogenates (final concentration: 2%), which were then incubated for 30 min at 37 °C and centrifuged at 20,000 g for 10 min at 25 °C. The supernatants were centrifuged at 100,000 g for 20 min at 25 °C. The pellets were further washed with sterile saline and centrifuged at 100,000 g for 20 min. The resulting pellets were used as Sarkosyl-insoluble fraction. This study was approved by the research ethics committee of Tokyo Metropolitan Institute of Medical Science (number: , and carried out in accordance with the approved guidelines. Informed consent about the brain donation was obtained from all subjects.