Searching for novel cerebrospinal fluid biomarkers of tau pathology in frontotemporal dementia: an elusive quest

Background Frontotemporal dementia (FTD) is a pathologically heterogeneous neurodegenerative disorder associated usually with tau or TDP-43 pathology, although some phenotypes such as logopenic variant primary progressive aphasia are more commonly associated with Alzheimer’s disease pathology. Currently, there are no biomarkers able to diagnose the underlying pathology during life. In this study, we aimed to investigate the potential of novel tau species within cerebrospinal fluid (CSF) as biomarkers for tau pathology in FTD. Methods 86 participants were included: 66 with a clinical diagnosis within the FTD spectrum and 20 healthy controls. Immunoassays targeting tau fragments N-123, N-mid-region, N-224 and X-368, as well as a non-phosphorylated form of tau were measured in CSF, along with total-tau (T-tau) and phospho-tau (P-tau(181)). Patients with FTD were grouped based on their Aβ42 level into those likely to have underlying Alzheimer’s disease (AD) pathology (n=21) and those with likely frontotemporal lobar degeneration (FTLD) pathology (n=45). The FTLD group was then subgrouped based on their underlying clinical and genetic diagnoses into those with likely tau (n=7) or TDP-43 (n=18) pathology. Results Significantly higher concentrations of tau N-mid-region, tau N-224 and non-phosphorylated tau were seen in both the AD group and FTLD group compared with controls. However, none of the novel tau species showed a significant difference between the AD and FTLD groups, nor between the TDP-43 and tau pathology groups. In a subanalysis, normalising for total-tau, none of the novel tau species provided a higher sensitivity and specificity to distinguish between tau and TDP-43 pathology than P-tau(181)/T-tau, which itself only had a sensitivity of 61.1% and specificity of 85.7% with a cut-off of <0.109. Conclusions Despite investigating multiple novel CSF tau fragments, none show promise as an FTD biomarker and so the quest for in vivo markers of FTLD-tau pathology continues.

AbsTrACT background Frontotemporal dementia (FtD) is a pathologically heterogeneous neurodegenerative disorder associated usually with tau or tDp-43 pathology, although some phenotypes such as logopenic variant primary progressive aphasia are more commonly associated with Alzheimer's disease pathology. Currently, there are no biomarkers able to diagnose the underlying pathology during life. In this study, we aimed to investigate the potential of novel tau species within cerebrospinal fluid (CSF) as biomarkers for tau pathology in FtD. Methods 86 participants were included: 66 with a clinical diagnosis within the FtD spectrum and 20 healthy controls. Immunoassays targeting tau fragments N-123, N-mid-region, N-224 and X-368, as well as a non-phosphorylated form of tau were measured in CSF, along with total-tau (t-tau) and phospho-tau (p-tau (181) ). patients with FtD were grouped based on their Aβ 42 level into those likely to have underlying Alzheimer's disease (AD) pathology (n=21) and those with likely frontotemporal lobar degeneration (FtLD) pathology (n=45). the FtLD group was then subgrouped based on their underlying clinical and genetic diagnoses into those with likely tau (n=7) or tDp-43 (n=18) pathology. results Significantly higher concentrations of tau N-midregion, tau N-224 and non-phosphorylated tau were seen in both the AD group and FtLD group compared with controls. however, none of the novel tau species showed a significant difference between the AD and FtLD groups, nor between the tDp-43 and tau pathology groups. In a subanalysis, normalising for total-tau, none of the novel tau species provided a higher sensitivity and specificity to distinguish between tau and tDp-43 pathology than p-tau (181) /t-tau, which itself only had a sensitivity of 61.1% and specificity of 85.7% with a cut-off of <0.109. Conclusions Despite investigating multiple novel CSF tau fragments, none show promise as an FtD biomarker and so the quest for in vivo markers of FtLD-tau pathology continues.

INTroduCTIoN
Frontotemporal dementia (FTD) is a common form of early-onset dementia, but it is pathologically heterogeneous, which precludes accurate diagnosis during life of the underlying molecular cause. 1 The majority of patients with FTD have either tau or TDP-43 inclusions at post mortem, but at present there are no biomarkers that can reliably separate these groups from each other or from healthy controls. Currently available cerebrospinal fluid (CSF) measures of tau do not seem to represent the burden of cerebral tau pathology and are variably affected in different forms of FTD. Furthermore, they can be abnormal in other proteinopathies. [2][3][4][5] However, studies indicate that the tau protein can be cleaved into multiple different fragments, which are actively secreted from cells and can therefore potentially be identified in CSF. 6 7 In this study, we assessed the potential of novel CSF measures of different tau species as candidate biomarkers for FTD.

MeThods Participants
86 consecutively recruited participants with available CSF from the University College London FTD cohort studies were included in the study: 66 patients and 20 healthy cognitively normal controls. The 66 patients met consensus diagnostic criteria for either behavioural variant FTD (bvFTD) (21, of whom one patient had associated motor neuron disease) 8 or primary progressive aphasia (PPA) 9 (45). In the PPA cohort, 11 had semantic variant, 16 had non-fluent variant (of whom two patients had associated progressive supranuclear palsy), 15 had logopenic variant and 3 did not meet criteria for any of the three variants, named PPA-not otherwise specified (PPA-NOS). 9 All patients were screened for mutations causative of FTD and 10 patients were found to have a pathogenic mutation: MAPT (4), GRN (3) and C9orf72 (3).

Measurement of CsF markers
CSF was collected, processed and stored at −80°C according to standardised procedures. 10 11 Initially, the concentrations of the currently available markers of CSF T-tau, P-tau (181) and Aβ 42 were determined using sandwich ELISAs (INNOTEST; Fujirebio Europe N.V., Gent, Belgium) following manufacturer's instructions.
forms of tau (performed as in Lewczuk et al 12 (figure 1b, table 1, online supplementary table) Some patients with bvFTD or PPA (particularly the logopenic variant) may have underlying Alzheimer's disease (AD) pathology rather than frontotemporal lobar degeneration (FTLD) pathology. The initial analysis therefore aimed to compare these groups with controls. We used the Duits criteria 13 to identify those patients who were likely to have underlying AD pathology (i.e. CSF concentrations of Aβ 42 <550 pg/mL). This atypical AD group of 21 patients included 14 with logopenic variant PPA, 3 with non-fluent variant PPA, 1 with semantic variant PPA, 1 with PPA-NOS and 2 with bvFTD. Forty-five patients had Aβ 42 >550 pg/mL and therefore were included in the FTLD group figure 2. The mean (SD) T-tau and P-tau (181) concentrations were significantly higher in the AD group (722.7 (529.1) pg/mL; 72.1 (42.3) pg/mL) compared with healthy controls (326.6 (90.8) pg/mL; 51.1 (11.9) pg/mL) and the FTLD group (464.7 (344.1) pg/mL; 51.3 (24.1) pg/mL).
None of the novel measures showed a significant difference between the AD and FTLD group.

Comparison of those with likely FTLd-tau versus FTLd-TdP-43 pathology (figures 1C and d; table 1, online supplementary table)
Individuals in the FTLD group were then grouped based on their likely underlying pathology into an FTLD-tau group (containing MAPT mutation carriers, those with a secondary clinical diagnosis of progressive supranuclear palsy, and one patient with bvFTD who had subsequently come to post mortem and was found to have corticobasal degeneration; n=7) and an FTLD-TDP-43 group (containing GRN and C9orf72 mutation carriers, those with a primary clinical Neurodegeneration Figure 2 Flow diagram of participants included in the analysis. AD, Alzheimer's disease; CBD, corticobasal degeneration; FtD, frontotemporal dementia; FtLD, frontotemporal lobar degeneration; pSp, progressive supranuclear palsy; svppA, semantic variant primary progressive aphasia. diagnosis of semantic variant PPA or a secondary diagnosis of motor neuron disease; n=18) (figure 2).
None of the measures showed a significant difference between the FTLD-TDP-43 and FTLD-tau groups.
We performed a subanalysis normalising tau markers for T-tau, based on previous literature which has shown an improved differentiation of tau and TDP-43 pathology using the ratio of P-tau (181) to T-tau. [3][4][5] The P-tau (181) /T-tau ratio was significantly lower for both the FTLD-TDP-43 group (mean (SD) 0.113 (0.032)) and the FTLD-tau group (0.126 (0.033)) compared with controls (0.160 (0.027)), but there was no significant difference between the FTLD-tau and FTLD-TDP-43 groups. Receiver operating characteristic (ROC) curve analysis measuring the ability of P-tau (181) /T-tau to differentiate probable FTLD-tau from FTLD-TDP-43 showed a sensitivity of 61.1% and specificity of 85.7% with a cut-off point of <0.109 and an AUC 0.63.
Of the novel tau species, both tau X-368 and tau N-224 had a significantly different ratio in the FTLD-tau group (mean (SD): 0.036 (0.010); 0.035 (0.037)) compared with controls (0.050 (0.009); 0.013 (0.009)). For tau X-368, the ratio was also lower in the FTLD-TDP-43 group (0.039 (0.013)) compared with controls, but there was no difference between the FTLD-tau and FTLD-TDP-43 groups. For tau N-224, there was also a significantly higher ratio for FTLD-tau compared with the FTLD-TDP-43 group (0.019 (0.010)). However, sensitivity and specificity of tau N-224 ratio to differentiate between likely FTLD-tau and FTLD-TDP-43 groups was only 61.1% and 57.1% with a cut-off point of <0.019 (AUC of 0.63). No significant differences were shown in the other novel tau measures.

Correlations of tau CsF biomarkers
All CSF tau markers were significantly correlated with each other figure 1E. However, the strongest correlations were for T-tau with P-tau (181) (ρ=0.87) and tau N-mid-region (ρ=0.84), and for tau N-mid-region with tau X-368 (ρ=0.86) and non-phosphorylated tau (ρ=0.85). Although significant, the correlations of Tau N-123 with the other tau species were fairly weak.

dIsCussIoN
In this study, we investigated the potential of novel CSF tau measures as biomarkers of tau pathology in FTD. However, no significant differences were seen between those with likely underlying AD pathology and FTLD pathology, or between those with likely FTLD-tau and FTLD-TDP-43 pathology.
Tau N-224 was one of only two markers higher in the FTLD-tau group compared with controls, and when normalised for total-tau, showed a significant difference between FTLD-tau and FTLD-TDP-43, but separated the groups with only poor sensitivity and specificity of <65% (AUC of 0.63). A similar sensitivity (61.1%) and higher specificity (85.7%) was found for the P-tau (181) /T-tau ratio at a cut-off point of <0.109 (AUC of 0.63), a marker previously described by other groups: Hu et al 3  Although the fragments we have measured do not show a diagnostic accuracy that is superior to the existing tau biomarkers, we find different patterns in the concentrations of the fragments between the pathological groups. This finding is in concordance with other studies which suggest that tau may be differentially processed and secreted in a regulated manner. 14 We hypothesise that specific tau fragments may be generated and secreted in different tauopathies, and here we provide evidence that three tau fragments are significantly increased in CSF in FTD compared with controls. It is likely that there are other fragments of tau, not analysed in this study, that are more specific to FTD, and further work is required to identify these.
This study has a number of limitations. The majority of the patients did not have pathological confirmation of the cause of their illness, and in addition, although there is a relatively large number of cases for a study of a rare disorder like FTD, the individual numbers are small in each subgroup. There are also potentially limitations in the assay sensitivities for the novel tau fragments such that improvement in these may lead to a clearer difference between cases and controls that is not currently apparent.
In conclusion, while a number of these novel tau species show significantly higher concentrations in those with underlying AD pathology, they do not show any added benefit above current tau biomarkers and are not useful as biomarkers of tau pathology in FTD. Further work in the development of biomarkers of tau and TDP-43 in FTD is needed, particularly in light of potential disease-modifying tau therapies currently entering clinical trials.