In this plasma proteomics study, we measured 1303 proteins in over 260 human plasma samples to identify protein profiles for the specific diagnosis of FTD and its pathological subtypes. We found a difference in overall protein profile between FTD, AD and controls. Importantly, we identified a plasma protein panel that discriminated FTD from AD patients, but not FTD from controls. No plasma or tissue protein changes were detected between FTD pathological subtypes.
To our knowledge, we were the first to apply proteomics in blood plasma of genetically or pathologically confirmed FTLD patients (29). We found a difference in overall plasma protein profiles between FTD, AD patients and controls, which could be attributed to six proteins. All these proteins were upregulated in AD compared to FTD patients and controls, which suggests that none of the proteins identified are associated to FTD pathogenesis. In line with these findings, our bootstrap classification exercises identified a combination of 61 proteins demarcating FTD patients and non-demented controls with limited accuracy (AUC: 0.61) and large confident intervals, underpinning the insufficient diagnostic accuracy and further supporting the lack of specific plasma protein signals specifically associated to FTD. Considering that the number of FTD and AD cases analyzed were comparable, the lack of biomarker signals associated to FTD might be explained by the clinicopathological diversity of FTD. The different clinical, genetic and pathological phenotypes within the FTD spectrum may hurdle the identification of specific biomarkers, highlighting the need to include large cohorts in biomarker studies.
We identified a panel of 12 blood-based proteins discriminating FTD from AD with very high accuracy (AUC: 0.99). Three of these proteins (fibronectin fragments 3 and 4 and Von Willebrand Factor (vWF)), were among the proteins differentially regulated between AD, FTD and controls identified before. Our findings are supported by a previous AD aptamer-based study, where fibronectin fragment 4 and fibronectin were also selected in a panel of plasma proteins to discriminate AD patients from controls (16).The observed high diagnostic accuracy supports potential use of this blood-based biomarker panel for the differential dementia diagnosis. However, as an AUC of 0.99 is near to perfect, replication of these findings, preferably through external validation is needed. The four proteins that contributed most to the discriminatory panel (Fibronectin, Fibrinogen gamma chain, hnRNPK and vWF) based on the largest beta’s, will be discussed in more detail. The protein with the strongest beta was Fibronectin (FN), a glycoprotein that plays a role in tissue repair, and regulating cell attachment, motility, hemostasis and embryogenesis (30). Several studies reported higher amounts of high molecular FN forms in plasma, CSF and frontal and temporal cortex of AD patients compared to vascular dementia and controls (31–33), corroborating our results showing higher levels of fibronectin fragments 3 and 4 in AD patients compared to FTD patients and controls. Interestingly, increased expression of FN type III domain has shown to decrease Aβ secretion in a cellular model (34). These data together suggest an increase of fibronectin fragments in AD which might potentially convey a neuroprotective effect. The protein with the second highest beta was Fibrinogen gamma chain, a blood borne glycoprotein essential to form an insoluble fibrin matrix. It is associated to amyloid deposition (35) and brain atrophy (36). The lower levels of this protein in AD compared to FTD and controls (37) indicate that this marker is specifically associated to AD pathogenesis. Experimental and neuropathological studies indeed suggest that this protein may contribute to AD by altering thrombosis and fibrinolysis (38). hnRNP K is one of the major pre-mRNA-binding proteins, likely playing a role in the nuclear metabolism of hnRNAs and in the p53/TP53 response to DNA damage (39). A previous proteome study found an upregulation of this protein in frontal cortex of AD cases (40). Recent exciting evidence showed mislocalisation of hnRNA K in pyramidal neurons of the frontal cortex to be a novel neuropathological feature associated with both frontotemporal lobar degeneration and ageing (41,42). Future studies should therefore address the potential role of this protein in both FTD and AD to understand how it contributes to discriminate these disorders. The protein with the fourth highest beta was VWF, a glycoprotein with critical functions in hemostasis (43). It was identified by the global test and was also part of the protein panel discriminating AD and FTD. VWF has frequently been studied in AD since vascular damage plays a role in the pathogenesis of AD dementia. However, results of VWF levels in AD patients have been conflicting. One CSF proteomics study that aimed to discriminate AD from non-AD patients based, has shown discrepant results in CSF VWF levels between three independent cohorts (31). Other studies reported no difference in VWF levels in blood plasma, CSF or brain cells between AD and controls (44,45), and one large population study reported higher levels of VWF in blood plasma of AD patients (46). We recently observed increased levels of CSF VWF in our ongoing AD studies (Del Campo et al, in preparation). A possible speculative explanation for these discrepant findings could be that the cohorts that reported an increase in VWF levels, including ours, had more patients with mixed vascular and AD pathology, whereas other cohorts mostly included patients with pure AD pathology. It would be very relevant to investigate the markers identified here together with novel promising plasma biomarkers, such as plasma pTau levels, pTau levels, that show very good discrimination between AD and FTD patients, being specifically increased in AD (47–49).
We could not find differentially regulated proteins between Tau and TDP pathological subtypes in tissue or plasma, nor could we identify discriminatory plasma protein signatures between these subtypes. Throughout literature, it has been challenging to identify and validate protein alterations between both pathological subtypes. For CSF, two previous proteomic studies reported several differentially regulated CSF proteins (50) or a biomarker panel in CSF which enabled sensitive differentiation between TDP and Tau pathology (51), although independent multicenter validation and replication on different platforms is still needed. The lack of a biomarker (panel) for FTD subtypes with feasibility in clinical practice thus far, could have several possible explanations. First, a potential explanation is the heterogeneity within Tau and TDP pathological subtypes, such as the different isoforms of TDP and Tau pathology, which have not been accounted for in fluid biomarker studies so far (52,53). For instance, patients with the TDP-A isoform might have a different protein signature than patients with the TDP-C isoform. This heterogeneity will complicate the search for a single discriminatory protein panel for TDP vs Tau, and will require larger and more homogeneous sample sizes, which are scarce. An alternative explanation could be that both pathological subtypes might have similar downstream pathological pathways leading to FTD. For instance, local TDP and Tau pathology could potentially be initiating the same prominent cascades, represented in similar proteomic changes in body fluids, ultimately leading to the neurodegenerative changes seen in FTD. This could also explain why both pathological subtypes are seen across the clinical FTD spectrum (10). Lastly, in most FTD biomarker studies familial and sporadic cases are often grouped to achieve a large sample size. However, the question remains whether the familial form of FTD with GRN, C9orf72 or MAPT mutations is biologically similar to sporadic FTD patients with TDP or Tau proteins. Future studies where (plasma) protein profiles of familial and sporadic FTD subtypes are independently studied could provide more clarity.
Proteomics in body fluids such as blood plasma or CSF can provide valuable mechanistic information as to whether post-mortem pathological changes are also seen in earlier ante-mortem disease stages, or whether there are also systemic responses involved in CNS diseases. As suggested also by the Consensus report of The Reagan Working group in 1998 (54), comparison of biofluid results with the expression of those proteins in brain tissue would be the most direct proof for a relation with the brain pathology. Indeed, pathological correlates are the basis for the now widely used biomarkers in AD, such as amyloid beta and pTau. We observed in our exploratory analysis that the proteins differentially regulated in FTD brain tissue were not dysregulated in plasma, suggesting that the brain changes identified were not reflected in plasma. This might be explained by the redundancy of plasma proteins from the periphery, which may mask low concentration and subtle changes of CNS-derived proteins in plasma. It is important to note that biofluid based biomarker levels are dynamic and may change along the disease process (55). Thus, the different time point of collection (i.e. ante-mortem for plasma vs. post-mortem for tissue), may explain the lack of overlap. However, the small sample size of the tissue sections prohibits strong conclusions.
Among the limitations of our study is that despite the large number of plasma proteins analyzed, the aptamer-based proteomic platform is still a targeted analysis dependent on the protein library. Thus, we cannot exclude that the other relevant or powerful brain-disease related biomarkers are not present within the aptamer library (i.e. Somamer library (56)). Nevertheless, the hypothesis free approach allowed us to identify novel proteins in addition to previously described proteins. Another limitation is that there was some center bias, because especially AD and control samples were collected from several sites (five). However, the majority (two third) of the AD samples were from the two sites that provided also the FTD samples, and all centers collected their samples under the same standardized protocol. Another limitation is the lack of replication of our findings in an independent validation cohort, especially considering the high accuracy of our FTD vs AD discriminatory panel. Validation of the plasma panel is technically not feasible yet on the Somascan technology. Novel large proteomics studies using an independent platform (proximity extension assay) with versatility of building smaller panels in plasma of FTD patients are current underway in the course of the JPND bPRIDE project (neurodegenerationresearch.eu).
The strengths of our study are that all our FTD cases had confirmed diagnosis based on genetic and/or pathological confirmation. Because FTD is clinically heterogeneous and does not correlate strongly to its pathologic subtypes, cohorts with known pathologic subtypes are important to provide relevant insights into underlying disease mechanisms. Of note, some of the AD plasma samples analyzed in this study came from non-specialized memory clinics, and were diagnosed using clinical criteria without AD CSF biomarker confirmation.