FTLD-TDP and FTLD-tau can each lead to clinical FTD syndromes, although the underlying pathologic substrate is difficult to predict on clinical grounds alone. Using autopsy-confirmed cases of FTLD-TDP and FTLD-tau as our training set, we identified novel CSF biomarkers that can improve the distinction between FTLD-TDP and FTLD-tau, including AgRP, ACTH, IL-17, Eotaxin-3, and Fas, with high sensitivity and modest specificity for FTLD-TDP. While the involvement of these candidate biomarkers in the development and progression in FTLD remains to be determined, these analytes offer promise in the antemortem differential diagnosis of FTLD-TDP vs FTLD-tau.
Prior to the current study, only TDP-43 itself has been examined as a potential biomarker for nonfamilial cases of FTLD-TDP.11,22
One study showed elevated plasma TDP-43 levels in 46% of clinical FTD cases and 22% of clinical AD cases, but the lack of pathologic confirmation in these groups limited confident interpretation of results.11
Levels of TDP-43 also appeared to be elevated in CSF samples from patients with FTD-ALS,12
but the significant overlap in TDP-43 levels between patients and healthy subjects emphasized the need for improved assays. Other studies have also sought to identify biomarkers in disorders associated with FTLD, including ALS and progressive supranuclear palsy (PSP). Potential biomarkers of ALS have included elevated levels of TDP-43,12
inflammatory proteins (GM-CSF,23
and multiple interleukins including IL-2, IL-6, IL-8, IL-15, and IL-1723,24
), axonal structural proteins (neurofilament light chain25
), and growth factors (FGF basic protein and VEGF23
), and decreased levels of cystatin C,26
insulin-like growth factor 1,27
and angiotensin II.29
While alterations in some biomarkers are likely specific for FTLD-TDP spectrum disorders, changes in many likely can occur in either FTLD-TDP or FTLD-tau. For example, neurofilament light chains were proposed as a biomarker for ALS,25
but elevated levels were independently found in PSP and CBS.30
Hence, any discovery or validation biomarker work in FTLD-TDP or FTLD-tau needs to incorporate both disorders comparatively with neuropathologic confirmation.
Some of the proposed ALS biomarkers above were specifically evaluated in our multiplex panel.23
MCP-1 was found by 2 previous studies23,24
to be elevated in ALS, and MCP-1 was increased in our FTLD-TDP cases compared to control subjects (p
= 0.01) and FTLD-tau (p
= 0.089). However, MCP-1 did not emerge as a reliable biomarker for FTLD-TDP in the current study despite the difference in levels. Further analysis showed that MCP-1 levels were highly correlated with Fas levels (R
= 0.698, p
= 0.005), and Fas was identified by multiple methods as a key discriminator between FTLD-TDP and FTLD-tau. Our multiplex approach thus identified Fas as a more robust proxy biomarker for FTLD-TDP than MCP-1 for a similar underlying biological process. As apoptosis induced by Fas and Fas-associated death domain protein is associated with significant macrophage recruitment and MCP-1 upregulation,31
our finding of elevated Fas levels in FTLD-TDP (compared to FTLD-tau and control subjects) would suggest that Fas-induced apoptosis is more associated with FTLD-TDP than FTLD-tau. Among the remaining analytes that distinguished FTLD-TDP from FTLD-tau, AgRP and ACTH are both hypothalamic neuropeptides, and their elevation in the CSF may reflect hypothalamic dysfunction. No specific hypothalamic dysfunction has been previously described in FTLD-TDP, but disinhibited behaviors common in bv-FTD and hypothalamic dysfunction such as an eating disorder can both be linked to amygdala abnormalities.32
Clinically, elevated AgRP may contribute to the common hyperoral behavior in clinical FTD through its appetite-promoting effect. Analytes identified only by Mann-Whitney U
test may also be biologically associated with FTLD. For example, CSF levels of ANG-2 were elevated in FTLD-TDP, and this elevation may be associated with respiratory status of patients with FTLD-TDP.33
IL-23 levels differed between FTLD subtypes by Mann-Whitney U
test, and IL-23 promotes the development of helper T cells that release IL-17 (identified by both analytical strategies).21
These T-helper 17 cells have been implicated in multiple sclerosis,33
and microglia can themselves release IL-17 in the presence of IL-23.34
As IL-23 is relatively increased in both FTLD-TDP and FTLD-tau, there may be a common IL-17 dysfunction in FTLD. Whether higher IL-23 levels are protective or harmful in FTLD remains to be determined, along with the biological significance of decreased IL-17 levels in FTLD-TDP despite the upregulated IL-23 levels.
The true diagnostic accuracy of the novel analytes cannot be determined without a fully validated test set, but we predicted the underlying FTLD pathology in a group of living patients with clinical bv-FTD, PPA, and CBS. Consistent with previous reports,4,35,36
we found SemD to have the highest proportion of patients predicted to have FTLD-TDP among all FTD phenotypes. Patients with bv-FTD were next most likely to have FTLD-TDP, and patients with CBS and PNFA were least likely to have FTLD-TDP. However, the proportion of patients predicted to have FTLD-TDP in the bv-FTD, CBS, and PNFA groups were on the higher ends of previously reported ranges.4,35
This could be due to the higher sensitivity at the cost of specificity observed in the autopsy cohort, or bias associated with referral or research participation. At the same time, the pattern of relative performance in category naming fluency and confrontational naming in autopsy-confirmed cases of FTLD-TDP6
was also noted among patients with bv-FTD, which is the most prevalent phenotype. While this pattern of relative neuropsychological performance was not observed in other phenotypes, the limited power within each non-bv-FTD phenotype may mask any such trend in two tasks. The classification results are thus in keeping with reported clinicopathologic correlations, although continued follow-up of these patients to autopsy or validation in an independent autopsy cohort would be necessary to determine the diagnostic performance of novel biomarkers reported here.
The limited sample size in the autopsy cohort may bias the results of our classification, and replication in independent cohorts and platforms will be necessary. There also exists pathologic heterogeneity within FTLD-TDP and FTLD-tau. For example, FTLD-TDP is associated with multiple combinations of pathologic inclusions,2
and different types of FTLD-tau (corticobasal degeneration vs PSP) may uniquely associate with certain analytes. While we aimed to identify analytes common to members within the main pathologic groups, analytes associated with specific pathologic subgroups may also be of biological and clinical significance. Additional analytes may also help distinguish between FTLD-TDP and FTLD-tau, such as levels of tau (which did not significantly differ between FTLD-TDP and FTLD-tau cases in the current study) or phosphorylated TDP-43 peptides in the CSF using more sensitive assays. Improved assays may also determine the utility of previously reported candidate ALS biomarkers, as we were unable to detect levels of GM-CSF, G-CSF, IL-2, IL-6, or IL-15 using the standard RBM protocols. Nevertheless, based on this novel exploratory study of FTLD biomarkers, we propose a stepwise workup of patients clinically diagnosed with FTLD spectrum disorders using first the more validated biomarker assays of CSF AD biomarkers (p-tau181
, tau, and Aβ42) to exclude cases of clinical FTD due to atypical AD. This can then be followed by measurements of novel FTLD biomarkers like those reported here to further distinguish between FTLD-TDP and FTLD-tau.