RESEARCH ARTICLE APP-derived peptides reflect neurodegeneration in frontotemporal dementia n-Gala1,2 , Jordi Pegueroles1,2, Victor Montal1,2, Daniel Alcolea1,2 , Eduard Ignacio Illa Vilaplana 1,2  , Alexandre Bejanin1,2 , Sergi Borrego-Ecija 3 , Frederic Sampedro4, Andrea Subirana1, Marıa-Bele n Sanchez-Saudino  s , Ricard Rojas-Garcıa , Hugo Vanderstichele6, Rafael Blesa1,2, 1 5 Jordi Clarimo  n , Anna Antonell3, Albert Llado 1,2  3, Raquel Sa nchez-Valle3, Juan Fortea1,2,7 & Alberto Lleo  1,2 1 Memory Unit, Department of Neurology, Hospital de la Santa Creu i Sant Pau, Biomedical Research Institute Sant Pau, Universitat Auto noma de Barcelona, Barcelona, Spain 2 Centro de Investigacion Biomedica en Red de Enfermedades Neurodegenerativas, CIBERNED, Madrid, Spain 3  Clınic per a la Recerca Biom Alzheimer’s Disease and Other Cognitive Disorders, Neurology Department, Hospital Clınic, Fundacio edica, Institut d’Investigacions Biomediques August Pi I Sunyer, University of Barcelona, Barcelona, Spain 4 Movement Disorders Unit, Neurology Department, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain 5 Neuromuscular Diseases Unit, Department of Neurology, Hospital de la Santa Creu i Sant Pau, Universitat Auto noma de Barcelona, Barcelona, Spain 6 ADx NeuroSciences NV, Gent, Belgium 7 Barcelona Down Medical Center, Fundacio  Catalana de Sındrome de Down, Barcelona, Spain Correspondence Abstract , Memory Unit, Department of Alberto Lleo Neurology, Hospital de la Santa Creu i Sant Objective: We aimed to investigate the relationship between cerebrospinal fluid Pau, Sant Antoni Mª Claret, 167, 08025 levels (CSF) of amyloid precursor protein (APP)-derived peptides related to the Barcelona, Spain. Tel: +34935565986; Fax: amyloidogenic pathway, cortical thickness, neuropsychological performance, +34935565602; E-mail: alleo@santpau.es and cortical gene expression profiles in frontotemporal lobar degeneration (FTLD)-related syndromes, Alzheimer’s disease (AD), and healthy controls. Methods: We included 214 participants with CSF available recruited at two Funding Information The Catalan frontotemporal initiative (CATFI) centers: 93 with FTLD-related syndromes, 57 patients with AD, and 64 healthy is funded by the Health Department of the controls. CSF levels of amyloid b (Ab)1-42, Ab1-40, Ab1-38, and soluble b Government of Catalonia (grant PERIS fragment of APP (sAPPb) were centrally analyzed. We compared CSF levels of SLT002/16/00408 to Alberto Lleo  and Raquel APP-derived peptides between groups and, we studied the correlation between Sanchez-Valle). The principal investigator of CSF biomarkers, cortical thickness, and domain-specific cognitive composites in the CATFI study is Dr. Alberto Lleo. This work each group. Then, we explored the relationship between cortical thickness, CSF was also supported by research grants from levels of APP-derived peptides, and regional gene expression profile using a the Instituto de Salud Carlos III, Spain (grants PI11/02526, PI14/01126 and PI17/01019 to brain-wide regional gene expression data in combination with gene set enrich- Juan Fortea, PI13/01532 and PI16/01825 to ment analysis. Results: The CSF levels of Ab1-40, Ab1-38, and sAPPb were Rafael Blesa, PI15/01618 to Ricard Rojas- lower in the FTLD-related syndromes group than in the AD and healthy con- Garcıa, PI18/00435 to Daniel Alcolea, PI14/ trols group. CSF levels of all APP-derived peptides showed a positive correla- 1561 and PI17/01896 to Alberto Lleo ; AC14/ tion with cortical thickness and the executive cognitive composite in the FTLD- 00013 to Raquel S anchez-Valle) and the related syndromes group but not in the healthy control or AD groups. In the CIBERNED program (Program 1, Alzheimer  and SIGNAL study, cortical regions where we observed a significant association between cortical Disease to Alberto Lleo www.signalstudy.es), partly funded by Fondo thickness and CSF levels of APP-derived peptides, we found a reduced expres- Europeo de Desarrollo Regional (FEDER), sion of genes related to synaptic function. Interpretation: APP-derived peptides Union Europea, “Una manera de hacer in CSF may reflect FTLD-related neurodegeneration. This observation has Europa.” This work has also been supported important implications as Ab1-42 levels are considered an indirect biomarker by a “Marato  TV3” grant (20141210 to Juan of cerebral amyloidosis. Fortea, 044412 to Rafael Blesa, 20143710 to Ricard Rojas-Garcıa and 20143810 to Raquel Sanchez-Valle) and by Generalitat de Catalunya (2014SGR-0235 to Alberto Lleo , 2518 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. I. Ill an-Gala et al. APP-Derived Peptides in FTD PERIS SLT002/16/00408 to Daniel Alcolea and SLT006/17/00119 to Juan Fortea), and Fundacion BBVA (grant to A. Lleo ) and a grant from the Fundacio  Bancaria La Caixa to Rafael Blesa. Ignacio Illan-Gala and Sergi Borrego-Ecija are supported by the Rio Hortega grant from “Accio n estrategica en Salud 2013-2016” and the European Social Fund. Ignacio Ill an-Gala is supported by the Global Brain Health Institute (Atlantic Fellow for Equity in Brain Health). Anna Antonell received funding from Departament de Salut de la Generalitat de Catalunya (PERIS 2016- 2020 SLT002/16/00329). Received: 9 October 2019; Accepted: 25 October 2019 Annals of Clinical and Translational Neurology 2019; 6(12): 2518–2530 doi: 10.1002/acn3.50948 of b- and c-secretases.10 In addition to Ab1-42, the prote- Introduction olytic process of APP also generates other shorter Ab spe- Frontotemporal lobar degeneration (FTLD) is a neu- cies, such as Ab1-40 and Ab1-38, which can be used as ropathological umbrella term encompassing multiple pro- markers of APP metabolism. Previous studies have teinopathies with common patterns of neurodegeneration reported that different Ab peptides are reduced in FTLD- in frontotemporal regions.1 In contrast to Alzheimer’s related syndromes.11–14 Another APP-derived peptide gen- disease (AD), there are no specific pathophysiological erated from APP is the soluble b fragment of APP biomarkers for FTLD and current diagnostic criteria rely (sAPPb).10,15 sAPPb is generated by BACE1, the main b- on the identification of particular syndromes and neu- secretase in the central nervous system, and levels are also rodegenerative patterns on neuroimaging.1 Cerebrospinal reduced in CSF in FTLD-related syndromes11,16–20 and cor- fluid (CSF) biomarkers have been studied in neurodegen- relate with cortical thickness in frontotemporal regions.18,20 erative disease as a way to track different pathophysiologi- However, it remains unknown whether Ab1-42 and cal processes in the central nervous system.2 Levels of other Ab peptides generated in the amyloidogenic APP core AD biomarkers, amyloid b1-42 (Ab1-42), total Tau pathway are related to neurodegeneration in FTLD-related (t-Tau), and phosphorylated Tau (p-Tau), can be useful syndromes or AD.21 We hypothesized that, similarly to in FTLD-related syndromes to exclude AD.3 Specifically, what we have observed for sAPPb, the decrease in the low Ab1-42 levels in CSF are considered a biomarker of CSF levels of Ab1-42, Ab1-40, and Ab1-38 may reflect cerebral amyloidosis in current research frameworks4 and neurodegeneration in FTLD but not in AD, where amy- therefore its presence in patients with FTLD-related syn- loid-related pathophysiological changes in APP processing dromes is usually interpreted as a sign of either atypical may obscure this association.22 The identification of dif- AD mimicking FTLD or comorbid AD pathology in ferent patterns of APP-derived peptides in CSF in FTLD FTLD cases.5 and AD is important for the in-vivo identification of However, several studies in patients with FTLD have FTLD and its differentiation from other non-neurodegen- shown substantial disagreement between CSF levels of erative diseases and could also provide information about Ab1-42 and amyloid positron emission tomography.6,7 In central pathophysiological mechanisms in FTLD. In this addition, previous studies have also reported lower CSF multimodal biomarker study, we compare the CSF levels levels of Ab1-42 in pathology-proven FTLD cases without of Ab1-42, Ab1-40, Ab1-38, and sAPPb in FTLD-related comorbid AD,3 as well as in genetically confirmed FTLD,8 syndromes, AD, and healthy controls, and we explore suggesting that in FTLD, Ab1-42 levels could be affected their relationship with cortical thickness and neuropsy- by mechanisms independent of AD pathology.9 chological measures of cognitive impairment. To further Ab peptides are generated by proteolytic cleavage from explore the mechanisms underlying the relationship the amyloid precursor protein (APP) by sequential action between APP-derived peptides and neurodegeneration in ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2519 APP-Derived Peptides in FTD I. Ill an-Gala et al. FTLD, we investigated the transcriptional architecture of Image acquisition, processing, and analysis cerebral regions where cortical thickness correlated with APP-derived peptides in the FTLD group.23 A subset of 135 participants had 3T magnetic resonance imaging (MRI) available for quantitative neuroimaging anal- yses. Of these, 88 participants were scanned on a 3T Philips Materials and Methods Achieva, 7 participants were scanned on a different 3T Phi- lips Achieva, and 40 participants were scanned on 3T Sie- Study participants and classification mens 3T TrioTim. Briefly, surface-based cortical Patients with AD and FTLD-related syndromes were diag- reconstruction was performed using FreeSurfer v5.1 software nosed at each center by experts in behavioral neurology package (http://surfer.nmr.mgh.harvard.edu) to compute according to current diagnostic criteria.24–28 From 338 eli- individual cortical thickness maps, as previously reported.36 gible participants with CSF available at both centers, we Each individual cortical thickness map was morphed to Free- selected a subgroup of 219 participants (98 FTLD-related surfer’s standard surface space and smoothed applying a syndromes, 57 AD, and 64 healthy controls) with a CSF Gaussian kernel of 15 mm full width at half maximum as sample stored in the same polypropylene tube. Moreover, customary in surface-based analyses. Regarding neuroimag- to avoid the inclusion of FTLD-related syndromes and ing analyses, a vertex-wise general linear model (as imple- healthy control participants with comorbid AD or atypical mented in FreeSurfer v5.1; Athinoula A. Martinos Center for forms of AD that could alter the profile of APP-derived Biomedical Imaging, Charlestown, MA, USA) was used to peptides, we excluded five participants in the FTLD-re- assess the correlation between CSF biomarkers and cortical lated syndromes group with a CSF biomarker profile sug- thickness for each group independently. Specifically, for each gestive of AD (as defined by low CSF levels of Ab1-42 surface vertex, a general linear model was computed using levels and increased CSF levels of t-Tau or p-Tau) accord- cortical thickness as dependent variable and CSF values as ing to local validated thresholds at each center (916 and independent variable. All these analyses were adjusted by 500 pg/mL for Ab1-42; 456 and 450 pg/mL for t-Tau and age, sex, and magnetic resonance equipment. To control for 63 and 75 pg/mL for p-Tau, in Hospital de Sant Pau and false positives, a Monte-Carlo simulation with 10000 repeats HCP, respectively).29–31 In the AD group, we included as implemented in FreeSurfer (FWE <0.05) was tested. only mild cognitive impairment and dementia of the Alz- heimer’s type participants with low CSF levels of Ab1-42 Genetic studies levels and increased CSF levels of t-Tau or p-Tau using the same thresholds.29–31 Moreover, healthy controls APOE was genotyped in all participants according to pre- included in this study did not have a CSF biomarker pro- viously described methods.37 Participants in the FTLD-re- file suggestive of preclinical AD using the same criteria lated syndromes group were screened for most common and thresholds. For further details on sample composi- FTLD-related mutations as previously described.20 tion, please refer to Supplementary material. Statistical methods Neuropsychological measures We assessed normality of the variables by means of the Cognitive functioning was assessed using a previously Kolmogorov–Smirnov test. Between-group differences in described neuropsychological evaluation32 and Z-scores baseline characteristics were assessed using t-test, were calculated for different cognitive domains as previ- ANOVA, or Kruskal–Wallis test for continuous variables, ously described (Supplementary material).33 and v2 for categorical data. Between-group differences in CSF biomarkers were assessed with robust ANOVA (F welch) and analysis of covariance (ANCOVA) analyses CSF sampling and analyses with age and APOEɛ4 status as covariates. For ANOVA All biomarkers were analyzed at the Sant Pau Memory and ANCOVA analyses, variables not following a normal Unit Laboratory with commercially available ELISA kits distribution (Ab1-42, Ab1-38, and sAPPb) were log-trans- of Ab1-40, Ab1-38, and sAPPß (EUROIMMUN, L€ ubeck, formed to achieve normality. For the study of the correla- Germany; IBL, Minneapolis, MN, respectively) following tions within CSF biomarkers and between CSF previously reported methods and manufacturer’s instruc- biomarkers and cognitive measures, we applied simple tions.30,31,34 To account for inter-center variability in pre- correlations (Pearson’s coefficient) and partial correlations analytical protocols, we applied a previously validated (including age, sex, and education as covariates for cogni- harmonization method.35 For further details on CSF anal- tive measures) with bootstrapping-based 95% confidence yses, please refer to Supplementary material (Table S1). intervals (bias corrected and accelerated for 1000 2520 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. I. Ill an-Gala et al. APP-Derived Peptides in FTD samples). To explore the significant differences between Results the magnitude of the correlations of CSF biomarkers between groups, we applied a R-to-R transform. All statis- Demographics and clinical characteristics of tical tests were two-sided and statistical significance was participants set at 5% (a = 0.05). Statistical analyses were performed with the IBM SPSS Statistics 25 (IBM corp.) software. For The main clinical and demographic characteristics of the ANOVA and ANCOVA analyses, all P values were cor- study are shown in Table 1. There were no differences in rected for multiple comparisons (Bonferroni). age at CSF sampling and MMSE score between the FTLD- related syndromes and AD groups. However, the FTLD-re- lated syndromes group was more functionally impaired Cortical gene expression and showed a higher frequency of men than the AD group. Microarray data from a total of 3702 regional tissue sam- The healthy control group was younger at CSF sampling ple from post-mortem brain tissue of six individuals from than the FTLD-related syndromes and AD groups. As the Allen Institute Human Brain Atlas were used to ana- expected, APOEɛ4 carriers were overrepresented in the AD lyze cortical gene expression (http://human.brain-map. group compared to FTLD-related syndromes and controls. org/). Data were processed according to previously pub- In the FTLD-related syndromes group, there were 11 cases lished methods (for more details, see Supplementary with known pathogenic mutations in FTLD-related genes material).38 (6 C9orf72, 4 GRN, and 1 VCP). Gene set enrichment analysis CSF levels of APP-derived peptides are reduced in FTLD-related syndromes After the identification of cortical regions where cortical thickness showed a significant correlation with the CSF CSF levels of Ab1-42, Ab1-40, Ab1-38, and sAPPb were levels of APP-derived peptides, we quantified the differ- reduced in the FTLD-related syndromes group when com- ences in gene expression between these areas and the back- pared to the healthy control group (Fig. 1A–D). As shown ground expression levels in the rest of the cortex, in Table 2, the CSF levels of Ab1-42 were significantly dif- calculating the Cohen’s delta for each gene. The resulting ferent between groups even after accounting for age at CSF Cohen’s deltas were first normalized using Z-score and sampling and the presence of APOEɛ4. As expected, the then used to build a histogram distribution of the differ- lowest levels of Ab1-42 were observed in the AD group. ences between ROI and background expression. Then, a However, Ab1-42 levels were also decreased in FTLD-re- threshold of 2.58 9 SD (99% confidence interval) was set lated syndromes when compared to healthy controls. to obtain the list of genes showing a differential expression. As reported in previous studies,6,7 in a substantial pro- Then, we introduced these genes in the GO Enrichment portion of FTLD-related syndromes, participants (n = 40; Analysis39,40 with Fisher’s exact test and false discovery rate 43% of the FTLD-related syndromes group) had low CSF correction to perform the enrichment analysis. To test levels of Ab1-42, below our validated threshold for cere- whether genes interacting with APP were overexpressed in bral amyloidosis.31 However, only 13 participants in the the cortical regions, we evaluated the genes encoding for frontotemporal lobar degeneration-related syndromes proteins with the highest confidence score of interaction (FTLD-S) group (17%) had higher CSF levels of either t- from STRING (https://string-db.org/). Tau or p-Tau and normal levels of Ab1-42. Importantly, we observed the same results when we excluded these cases from the FTLD-S group. Standard protocol approvals, registrations, The CSF levels of Ab1-40, Ab1-38, and sAPPb also dif- and patient consent fered between groups after accounting for age at CSF sam- The study was approved by the local ethics committee of pling and the presence of APOEɛ4. The CSF levels of Ab1- each participant center and was conducted in accordance 40, Ab1-38, and sAPPb were lower in the FTLD-related with the Declaration of Helsinki. All participants gave syndromes group than in the AD and healthy control their written informed consent to participate in the study. groups (Fig. 1B–D). As expected, the Ab1-42/Ab1-40 and Ab1-42/Ab1-38 ratios differed between groups (Fig. 1E–F). The Ab1-42/Ab1-40 and the Ab1-42/Ab1-38 ratios were Data availability lower in the AD group than in the FTLD-related syndromes The datasets analyzed during the current study are avail- and healthy control groups. The Ab1-42/Ab1-38 ratio was able from the corresponding author on reasonable higher in the FTLD-related syndromes group than in the request. healthy control group. We did not find significant ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2521 APP-Derived Peptides in FTD I. Ill an-Gala et al. Table 1. Demographic, clinical, and neuropsychological data. Demographic, clinical and genetic characteristics FTLD-related syndromes Alzheimer’s disease Healthy controls Statistics and P values n participants (HSP/HCB) 93 (76/17)ns 57 (45/12)ns 64 (53/11)ns v2(2) = 0.3 P = 0.854 Age at CSF, years, median (Q1, Q3) 70.7 (64.7, 77.4)1 72.3 (69.3, 76.5)1 64.3 (61, 69.6)2,3 F(2, 214) = 14.2 P < 0.001 Age at clinical onset, years, median (Q1, Q3) 66 (60, 72)3 69.5 (65, 73)2 – t(144) = 39.8 P = 0.001 Male, n (%) 54 (58)3 21 (37)2 31 (48) v2(2) = 6.4 P = 0.041 APOEɛ4+, n (%) 18 (19)3 29 (51)1,2 12 (19)3 v2(2) = 21.1 P < 0.001 Education, years, median (Q1, Q3) 11 (8, 15)ns 9 (8, 13)1 15 (9, 18)3 H(2) = 12.9 P = 0.003 MMSE score, median (Q1, Q3) 26 (21.5, 28)1 25 (19.5, 27)1 29 (28, 30)2,3 H(2) = 62.2 P < 0.001 IDDD, total score, median (Q1, Q3) 44 (39, 57)1,3 38 (33, 47)1,2 33 (33, 33)2,3 H(2) = 72.9 P < 0.001 n 3T MRI available (HSP/HCB) 54 (37/17)ns 38 (26/12)ns 43 (32/11)ns v2(2) = 0.5 P = 0.780 Neuropsychological evaluation available, n (%) 63 (68)ns 42 (74)ns 53 (83)ns v2(2) = 4.5 P = 0.108 Memory z-score, median (Q1, Q3) 3.6 ( 6.4, 1.7)1,3 5.2 ( 7.0, 3.6)1,2 0 ( 0.4, 0.5)2,3 H(2) = 99.3 P < 0.001 Executive z-score, median (Q1, Q3) 1.7 ( 2.7, 0.9)1 1.5 ( 2.3, 0.7)1 0 ( 0.4, 0.4)2,3 H(2) = 62.7 P < 0.001 Language z-score, median (Q1, Q3) 2.1 ( 2.9, 1.2)1 2.2 ( 3.1, 1.3)1 0 ( 0.4, 0.4)2,3 H(2) = 79.9 P < 0.001 Visuospatial z-score, median (Q1, Q3) 1.7 ( 3.5, 0.4)1 1.3 ( 2.5, 0.1)1 0 ( 0.1, 0.5)2,3 H(2) = 54.9 P < 0.001 Statistically significant results are bold. CSF, cerebrospinal fluid; FTLD, frontotemporal lobar degeneration; HSP, Hospital de Sant Pau; HCB, Hospi- tal Clınic de Barcelona; IDDD, interview for deterioration in daily living activities in dementia; MMSE, mini-mental state examination; MRI, mag- netic resonance imaging; n, number; Q1, first quartile; Q3, third quartile; 3T, 3 tesla. 1 Different than the healthy control group (P < 0.05). 2 Different than the FTLD-related syndromes group (P < 0.05). 3 Different than the Alzheimer’s disease group (P < 0.05). differences in the CSF levels of APP-derived peptides 42/Ab1-40 ratio in FTLD-S cases and controls suggests that between FTLD-related syndromes clinical subgroups. comorbid AD may not be the cause of the observed We did not find any significant differences in the CSF decrease in APP-derived peptides in the FTLD-S group. levels of APP-derived peptides between participants with increased certainty of underlying FTLD-Tau (n = 30, pro- APP-derived peptides correlate with cortical gressive supranuclear palsy-corticobasal degeneration spec- thickness in FTLD-related syndromes trum [PSP-CBD], and non-fluent variant of primary progressive aphasia [nfaPPA] participants) and FTLD-TDP As shown in Figure 2, we found a positive correlation (n = 21, including patients with semantic variant PPA, between APP-derived peptides and cortical thickness in patients with motor neuron disease and carriers of C9orf72, the FTLD-related syndromes group. Particularly, the CSF GRN, and VCP mutations). However, patients with levels of Ab1-42 showed a positive correlation with corti- C9orf72, GRN, and VCP mutations (n = 11) had lower cal thickness (namely, lower CSF values of Ab1-42 levels of Ab1-42 than patients in the sporadic FTLD-S reflected cortical thinning) in the prefrontal cortex, insula, group (n = 82; t(18.2) = 2.158; P = 0.045). anterior and dorsal cingulate, and superior frontal cortex Of note, we did not observe statistically significant differ- in both hemispheres. The CSF levels of Ab1-40 showed a ences in the CSF levels of APP-derived peptides between positive correlation with cortical thickness in the inferior APOEɛ4 positive (n = 18) and APOEɛ4 negative (n = 75) perirolandic cortex, superior temporal and insula, but this FTLD-S cases. This finding together with the similar Ab1- correlation was restricted to the left hemisphere. The CSF 2522 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. I. Ill an-Gala et al. APP-Derived Peptides in FTD Figure 1. Group comparison of CSF levels of APP-derived peptides. CSF levels of (A) Ab1-42, (B) Ab1-40, (C) Ab1-38, (D) sAPPb, (E) Ab1-42/Ab1- 40 ratio, and (F) Ab1-42/Ab1-38 ratio across groups. Only statistically significant differences are displayed. We applied correction for multiple comparisons (Bonferroni’s post-hoc test; *P < 0.001). The red-dotted line represents our previously validated CSF cut-point for Ab1-42. AD, Alzheimer’s disease; Ab, amyloid b; FTLD-S, frontotemporal lobar degeneration-related syndromes; HC, healthy controls; sAPPb, soluble b fragment of amyloid precursor protein; CSF, cerebrospinal fluid; APP, amyloid precursor protein. levels of Ab1-38 showed a positive correlation with corti- progressive aphasia group despite the small sample size cal thickness in the prefrontal cortex, insula, and periro- (n = 5). However, we were unable to obtain significant landic cortex of both hemispheres (Fig. 2). Finally, the correlations in the nfaPPA (n = 14) and PSP-CBS (n = 5) CSF levels of sAPPb also showed widespread positive cor- subgroups. Of note, these two subgroups were the ones relation with cortical thickness along the prefrontal cor- with the lowest neurodegenerative burden. Importantly, tex, insula, anterior and dorsal cingulate, perirolandic we did not observe any correlation between APP-derived cortex, but also posterior cortical regions such as the pre- peptides and cortical thickness in the AD and healthy cuneus (Fig. 2). We also explored the correlation of APP- control groups despite the existence of extensive areas of derived peptides in CSF in the different clinical subgroups cortical thinning in regions related with neurodegenera- of FTLD-related syndromes. We obtained similar results tion in both groups (data not shown). Of note, neither for the behavioral variant of frontotemporal dementia Ab1-42/Ab1-40 nor Ab1-42/Ab1-38 ratios correlated with subgroup with MRI available (n = 28). We also found cortical thickness in the FTLD-related syndromes or the significant correlations in the semantic variant of primary AD groups. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2523 APP-Derived Peptides in FTD I. Ill an-Gala et al. Table 2. Cerebrospinal fluid biomarker data. CSF biomarkers FTLD-related syndromes Alzheimer’s disease Healthy controls ANCOVA1 Ab1-42, pg/mL2, median (Q1, Q3) 989 (718, 1423)3,4 560 (467, 634)4,5 1294 (1030, 1631)3,5 F(2, 209) = 72.4 P < 0.001 partial g2 = 0.409 Ab1-40, pg/mL, median (Q1, Q3) 5224 (3937, 7593)3,4 7511 (5686, 9339)5 7842 (6341, 9075)5 F(2, 208) = 17.5 P < 0.001 partial g2 = 0.144 Ab1-38, pg/mL2, median (Q1, Q3) 1582 (1172, 2235)3,4 2199 (1817, 2983)5 2464 (1953, 2856)5 F(2, 209) = 29.6 P < 0.001 partial g2 = 0.221 sAPPb, ng/mL2, median (Q1, Q3) 442 (345, 582)3,4 693 (507, 998)5 731 (429, 934)5 F(2, 209) = 24.7 P < 0.001 partial g2 = 0.191 Ab1-42/Ab1-40 ratio median (Q1, Q3) 0.29 (0.25, 0.32)3 0.11 (0.09, 0.13)4,5 0.28 (0.23, 0.32)3 F(2, 208) = 133.1 P < 0.001 partial g2 = 0.561 Ab1-42/Ab1-38 ratio, median (Q1, Q3) 0.88 (0.79, 1.00)3,4 0.34 (0.28, 0.41)4,5 0.83 (0.69, 0.93)3,5 F(2, 207) = 149.7 P < 0.001 partial g2 = 0.591 Statistically significant results are bold. ANCOVA, analysis of covariance; Ab, amyloid b; CSF, cerebrospinal fluid; FTLD, frontotemporal lobar degeneration; sAPPb, soluble b fragment of amyloid precursor protein. 1 ANCOVA adjusted for age at CSF sampling and APOEɛ4 status. 2 These variables were not normally distributed across groups and were log-transformed to achieve normality before the statistical analyses. 3 Different than the Alzheimer’s disease group (P < 0.001; Bonferroni pairwise comparisons with bias-corrected accelerated bootstrapping for 1000 samples). 4 Different than the healthy control group (P < 0.001; Bonferroni pairwise comparisons with bias-corrected accelerated bootstrapping for 1000 samples). 5 Different than the FTLD-related syndromes group (P < 0.001; Bonferroni pairwise comparisons with bias-corrected accelerated bootstrapping for 1000 samples). the AD group, where only mild correlations were APP-derived peptides relate to measures of observed. sAPPb levels did not correlate with Ab1-38 cognitive function in FTLD-related levels in the AD group. Comparison of Pearson coeffi- syndromes cient values showed that the correlation of CSF levels of We next investigated the correlation between APP-derived Ab1-42 with Ab1-38 and sAPPb was higher in FTLD-re- peptides and domain-specific cognitive scores in each lated syndromes group than in the AD group. The corre- clinical group. Importantly, we only found correlations in lation of the CSF levels of Ab1-42 with Ab1-40 and Ab1- the FTLD-related syndromes group, and these correlations 38 was higher in the FTLD-related syndromes group than remained significant after accounting for age, sex, and in the healthy control group. education. As shown in Table 3, the executive function composite showed a significant correlation with Ab1-42, Transcriptional and gene set enrichment Ab1-40, Ab1-38, and sAPPb and the visuospatial function analysis cognitive composite correlated with Ab1-42 and Ab1-38. To further investigate the potential mechanisms underly- ing the reduction of APP-derived peptides in FTLD, we The correlation between different APP- explored whether APP and other genes related to APP derived peptides varied across groups processing were overexpressed in the cortical regions with We next explored whether the correlation between differ- significant associations. We found that neither APP nor ent APP-derived peptides differed between groups. As other APP-related genes (for further details see Table S2) shown in Table 4, correlations between peptides were in were differentially expressed in the cortical regions with general stronger in the FTLD group compared to the positive correlations in the FTLD-related syndromes other two groups. The CSF levels of sAPPb showed mod- group (Fig. 3A and Table S3). Finally, we investigated the erate correlations with Ab1-42, Ab1-40, and Ab1-38 in transcriptional architecture of cerebral regions where cor- FTLD-related syndromes and control groups but not in tical thickness showed a significant correlation with APP- 2524 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. I. Ill an-Gala et al. APP-Derived Peptides in FTD Figure 2. Correlation of APP-derived peptides with cortical thickness in the FTLD-related syndromes group. Correlation of APP-derived peptides with cortical thickness in the FTLD-related syndromes group. Only clusters that survived family-wise error correction P < 0.05 are shown. Only positive correlations were found (displayed in orange and red). Cortical thickness analyses were adjusted for age, sex, and MRI equipment. Ab, Amyloid b; sAPPb, soluble b fragment of amyloid precursor protein; APP, amyloid precursor protein; FTLD-S, frontotemporal lobar degeneration- related syndromes; MRI, magnetic resonance imaging. derived peptides. As shown in Figure 3B, in these regions, neuropsychological measures in FTLD-related syndromes. we observed a differential expression of genes related to Our findings extend data from previous studies that had synaptic processes, neuronal differentiation, and neuroge- related APP processing and FTLD17,41–43 and suggest that nesis (Table S4). Of note, among genes with a differential APP-derived peptides are involved in FTLD through amy- expression, we found some genes codifying for synaptic loid-independent mechanisms.10 In addition, our tran- proteins such as NPTX1 and CPLX2 (Table S2). scriptional analyses suggested that biological processes related to synaptic function could link neurodegeneration and APP processing in FTLD. Discussion Our findings that Ab1-42 levels are reduced in FTLD In this study, we describe for the first time a direct corre- have important implications for the field of CSF biomark- lation between CSF levels of Ab1-42, Ab1-40, Ab1-38, ers as Ab1-42 levels are considered an indirect biomarker and sAPPb peptides with cortical thickness and of cerebral amyloidosis in current AD research ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2525 APP-Derived Peptides in FTD I. Ill an-Gala et al. Table 3. Partial correlations of cognitive composites with APP-derived CSF biomarkers after adjusting by age, sex, and education in the FTLD-re- lated syndromes group. Ab1-42 Ab1-40 Ab1-38 sAPPb Memory n.s n.s n.s n.s Executive function 0.41 (0.19–0.61)P = 0.001* 0.31 (0.05–0.56) P = 0.017 0.35 (0.12–0.57) P = 0.006 0.34 (0.13–0.53) P = 0.007 Language n.s n.s n.s n.s Visuospatial 0.34 (0.09–0.56)P = 0.009 n.s 0.37 (0.13–0.58) P = 0.004 n.s Correlation between Ab(1-42), Ab(1-40), and Ab(1-38) and sAPPb in (A) AD group, (B) FTLD-related syndromes group, and (C) healthy control group. Results are shown as Pearson correlation coefficient (95% confidence interval). 95% confidence intervals were calculated by means of bias-corrected accelerated bootstrapping (1000 samples). Significant correlations (P < 0.05) are marked in bold. APP, amyloid precursor protein; Ab, amyloid b; CSF, cerebrospinal fluid; FTLD, frontotemporal lobar degeneration; n.s, non-significant correlation. *Remained significant after adjusting for multiple comparisons (Bonferroni). Table 4. Correlations between different APP-derived peptides across clinical groups. This is supported by previous reports in several neurode- generative and neuroinflammatory conditions. For exam- Ab1-42 Ab1-40 Ab1-38 ple, low CSF Ab1-42 levels have been related to disease Alzheimer’s disease severity and clinical progression in multiple sclerosis45 sAPPb 0.29 (0.04–0.52)b 0.39 (0.14–0.59) 0.37 (0.14–0.58) and Lewy body disease (Parkinson’s disease and Lewy Ab1-42 – 0.70 (0.56–0.81) 0.59 (0.40–0.73)b body dementia).46 In addition, the CSF levels of Ab1-42 Ab1-40 – – 0.82 (0.73–0.89) have been found decreased in neuropathologically con- FTLD-related syndromes firmed cases of Creutzfeldt–Jakob disease independently sAPPb 0.60 (0.38–0.77)a 0.58 (0.35–0.73) 0.60 (0.42–0.64) of cerebral amyloid deposition47 and in neuroinfectious Ab1-42 – 0.84 (0.77–0.90)c 0.89 (0.84–0.93)ac Ab1-40 – – 0.88 (0.82–0.92) diseases.48 This evidence supports the view that the CSF Healthy control levels of APP-derived peptides could reflect pathophysio- sAPPb 0.42 (0.18–0.63) 0.47 (0.25–0.65) 0.57 (0.37–0.72) logical mechanisms other than cerebral amyloid deposi- Ab1-42 – 0.60 (0.40–0.76)b 0.62 (0.40–0.80)b tion.10 Ab1-40 – – 0.89 (0.82–0.94) This study also helps to explain the improved agree- ment between amyloid PET and CSF measures observed Correlation between Ab1-42, Ab1-40, and Ab1-38 and sAPPb in (A) AD group, (B) FTLD-related syndromes group, and (C) healthy control with the Ab1-42/Ab1-40 ratio in non-AD dementias.6,7 group. Results are shown as Pearson’s correlation coefficient (95% con- Because both Ab1-40 and Ab1-38 are reduced in FTLD fidence interval). 95% confidence intervals were calculated by means of but not in AD, the Ab1-42/Ab1-40 and Ab1-42/Ab1-38 bias-corrected accelerated bootstrapping (1000 samples). Moderate-to- ratios can correct for the effect of FTLD-related neurode- high correlations (r > 0.5) are in bold. The correlations that are signifi- generation on the CSF levels of APP-derived peptides cantly different between groups are underlined; a: different from the yielding a measure that is more specifically linked to cere- AD group; b: different from the FTLD-related syndromes group; c: dif- bral amyloid deposition than levels of Ab1-42 alone. ferent from the healthy control group. APP, amyloid precursor protein; Ab, amyloid b; FTLD, frontotemporal lobar degeneration; sAPPb, soluble We found a direct correlation between the CSF levels b fragment of amyloid precursor protein. of Ab peptides and cortical thickness in FTLD-related syndromes but not in AD. Our results expand previous findings that sAPPb correlate with cortical thickness in frameworks.4,44 Our results suggest that CSF levels of FTLD,18,20,49 by showing correlation also with Ab1-42, Ab1-42 could be reflecting two distinct but related phe- Ab1-40, and Ab1-38. In line with this result, we also nomena: on the one hand, Ab1-42 levels are known to be found that CSF levels of APP-derived peptides correlated associated with amyloid cerebral deposition, but on the with neuropsychological measures in the FTLD-related other hand, Ab1-42 could be also related to synaptic loss syndromes group. The association between APP-derived in FTLD and possibly other non-AD neurodegenerative peptides and cortical thickness suggests that these pep- diseases. Overall, the notion that CSF levels of APP- tides act as neurodegeneration biomarkers in FTLD. This derived peptides (including Ab1-42) may be also reduced hypothesis is consistent with a recent study showing that in diseases other than AD argues against the role of Ab1- CSF levels of Ab1-42 correlated with gray matter density 42 levels in CSF as a measure of cerebral amyloidosis only and were comparable to neurofilament light as a predictor and suggests that it may also reflect other pathophysiolog- of longitudinal MRI changes in FTLD-related syn- ical processes (i.e., synaptic loss or neuroinflammation). dromes.43 Further studies are needed to determine the 2526 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. I. Ill an-Gala et al. APP-Derived Peptides in FTD Figure 3. Transcriptional architecture and gene set enrichment analysis. (A) Histogram representing the distribution of the effect size scores of the neuro-related gene expressions between regions with and without a significant relationship with APP-derived peptides in FTLD. The full list of genes with differential expression can be found in Table S2. (B) Bar graphs show the top FDR-corrected (P < 0.0001) GO enrichment analyses of genes differential expression in the brain regions where correlations were observed. APP, amyloid precursor protein; FTLD-S, frontotemporal lobar degeneration-related syndromes; FDR, false discovery rate. role of APP-derived peptides as diagnostic and prognostic The transcriptional analyses performed in this study biomarkers in FTLD as well as their added value in com- suggest that biological processes related to synaptic func- parison to other biomarkers such as neurofilament light tion could link neurodegeneration and APP processing in or the p-Tau/t-Tau ratio.50 FTLD. Interestingly, there is strong evidence that APP is The lack of correlation between APP-derived peptides expressed at the synapse and that APP-derived peptides and cortical thickness in the AD group deserves further dis- are involved in multiple synaptic functions.10 Further cussion. If CSF levels of APP-derived peptides reflect also links between neuronal activity and synaptic function of neurodegeneration, then it would be expected to find a rela- APP-derived peptides come from studies showing that Ab tionship between CSF levels, cortical thickness, and cogni- levels in CSF fluctuate over time in the same individuals tive measures in the AD group. Several factors could explain under physiological conditions, and a reduction of Ab why we failed to find these correlations in our study. First, production during sleep has been related to decreased the effect of cerebral amyloid plaque deposition on the cerebral activity.53 Taken together, previous evidence dynamics of production and clearance of Ab to the CSF may combined with our data support the idea that the reduc- obscure a correlation with cortical thickness in the AD tion of APP-derived peptides in CSF reflects synaptic dys- group. Additionally, we and others have shown that b-site function in FTLD and more studies are warranted to amyloid precursor protein-cleaving enzyme (BACE) activity determine the underlying mechanisms of our findings. is increased in sporadic AD,22,51 which may lead to a com- The main strengths of this study are the inclusion of pensatory increase in APP-derived peptides. Finally, it is participants with a high probability of underlying FTLD possible that our relatively small AD group do not accurately and AD and the multimodal biomarker approach that reflect the heterogeneity of AD neurodegeneration and that allowed us to explore the structural correlates of APP- only a subset of cases show lower levels of APP-derived pep- derived peptides and their transcriptional architecture. tides in CSF.52 Indeed, one large multicenter study reported However, this study has also some limitations. First, this that the data-driven subgroup of AD characterized by a dif- study lacks neuropathological confirmation and misdiag- fuse pattern of atrophy had the lowest levels of Ab1-42 in nosis could have occurred. However, in all participants, CSF.52 This effect could mask a neurodegeneration-medi- the diagnosis of underlying FTLD and AD was supported ated reduction of APP-derived peptides with the exception by CSF cutoffs and longitudinal follow-up. Although we of Ab1-42 which would be reduced due to sequestration in did not find differences in APP-derived peptides between amyloid plaques.2 Taken together, as suggested in recent specific clinical syndromes related to FTLD or partici- studies,21,52 it remains to be elucidated whether CSF levels of pants with presumed Tau and TDP-43 pathology in our Ab1-42 may also reflect neurodegeneration in AD in addi- study, further path-confirmed studies should compare the tion to cerebral amyloid deposition. CSF profile of APP-derived peptides between the different ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2527 APP-Derived Peptides in FTD I. Ill an-Gala et al. neuropathological subtypes of FTLD. Moreover, addi- Fundacio Bancaria La Caixa to Rafael Blesa. Ignacio Illan- tional studies should compare genetic and sporadic cases  Gala and Sergi Borrego-Ecija are supported by the Rio Hor- since results from this and other studies suggest the exis- tega grant from “Acci on estrategica en Salud 2013-2016” tence of different CSF biomarker profiles in genetic and the European Social Fund. Ignacio Illan-Gala is sup- cases.5,54,55 Second, we relied on cross-sectional data and ported by the Global Brain Health Institute (Atlantic Fellow thus, we were not able to assess the impact of APP-related for Equity in Brain Health). Anna Antonell received funding peptides in longitudinal clinical trajectories. Thus, more from Departament de Salut de la Generalitat de Catalunya studies are needed to evaluate whether APP-derived pep- (PERIS 2016-2020 SLT002/16/00329). tides can predict changes in biomarker trajectories and We acknowledge all the participants in this study and clinical outcomes of FTLD-related neurodegeneration. all the collaborators of the SPIN cohort. We also Finally, although we used CSF criteria to exclude patients acknowledge Soraya Torres and Laia Mu~ noz for technical with FTLD-related syndromes that had a CSF biomarker assistance. We thank EUROIMMUN for providing Ab1- profile of AD, we cannot exclude entirely that some of 38 and Ab1-40 ELISA assays for this study. the FTLD-related syndromes may have some concurrent AD pathology and therefore this study could not account Author Contributions for comorbid AD pathology. Finally, the exploratory tran- scriptional analyses were performed in subjects without I.I.-G. and A. Lleo contributed to the conception and design neurodegeneration and further studies in brain samples of the study; I.I.-G., J.P., V.M., D.A., E.V., A.B., S.B.-E, F.S., from subjects with FTLD are needed to confirm the A.S., M.B.S.-S., R.R.-G., R.B., J.C., A.A., A. Llad o, R.-S.-V., underlying biological processes of this association. J.F., and A. Lleo contributed to acquisition of data; I.I.-G., In summary, we report that patients with FTLD show a J.P., V.M., E.V. A.B., F.S., and J.C. contributed to analysis of global reduction of APP-derived peptides in CSF and that data and preparing the figures; I.I.-G., J.P., D.A., J.C., and A. this reduction correlated with FTLD-related neurodegen- Lleo contributed to drafting the text. eration and cognitive impairment. This pattern of APP- derived peptides should be considered when evaluating Conflict of Interests the presence of AD pathophysiology in FTLD-related syn- dromes. More studies are needed to advance our mecha- Nothing to report. nistic understanding of the relationship between APP- derived peptides and synapses. References 1. Elahi FM, Miller BL. A clinicopathological approach to the Acknowledgments diagnosis of dementia. Nat Rev Neurol 2017;13:457–476. 2. Molinuevo JL, Ayton S, Batrla R, et al. Current state of The Catalan frontotemporal initiative (CATFI) is funded by Alzheimer’s fluid biomarkers. Acta Neuropathol the Health Department of the Government of Catalonia 2018;136:821–853. (grant PERIS SLT002/16/00408 to Alberto Lle o and Raquel 3. Toledo JB, Brettschneider J, Grossman M, et al. CSF Sanchez-Valle). The principal investigator of the CATFI biomarkers cutoffs: the importance of coincident study is Dr. Alberto Lleo. This work was also supported by neuropathological diseases. Acta Neuropathol 2012;124:23–35. research grants from the Instituto de Salud Carlos III, Spain 4. Jack CR, Bennett DA, Blennow K, et al. NIA-AA Research (grants PI11/02526, PI14/01126 and PI17/01019 to Juan For- Framework: toward a biological definition of Alzheimer’s tea, PI13/01532 and PI16/01825 to Rafael Blesa, PI15/01618 disease. Alzheimers Dement 2018;14:535–562. to Ricard Rojas-Garcıa, PI18/00435 to Daniel Alcolea, PI14/ 5. Lle o A, Irwin DJ, Illan-Gala I, et al. A 2-step cerebrospinal 1561 and PI17/01896 to Alberto Lle o; AC14/00013 to Raquel algorithm for the selection of frontotemporal lobar Sanchez-Valle) and the CIBERNED Program (Program 1, degeneration subtypes. JAMA Neurol 2018;75:738–745. Alzheimer Disease to Alberto Lleo and SIGNAL study, www. 6. Janelidze S, Zetterberg H, Mattsson N, et al. CSF Ab42/ signalstudy.es), partly funded by Fondo Europeo de Desar- Ab40 and Ab42/Ab38 ratios: better diagnostic markers of rollo Regional (FEDER), Uni on Europea, “Una manera de Alzheimer disease. Ann Clin Transl Neurol 2016;3:154– hacer Europa.” This work has also been supported by a 165. “Marat o TV3” grant (20141210 to Juan Fortea, 044412 to 7. Leuzy A, Chiotis K, Hasselbalch SG, et al. Pittsburgh Rafael Blesa, 20143710 to Ricard Rojas-Garcıa and 20143810 compound B imaging and cerebrospinal fluid amyloid-b to Raquel Sanchez-Valle) and by Generalitat de Catalunya in a multicentre European memory clinic study. Brain (2014SGR-0235 to Alberto Lle o, PERIS SLT002/16/00408 to 2016;139(Pt 9):2540–2553. Daniel Alcolea and SLT006/17/00119 to Juan Fortea), and 8. K€am€al€ainen A, Herukka S-K, Hartikainen P, et al. Fundaci on BBVA (grant to A. Lle o) and a grant from the Cerebrospinal fluid biomarkers for Alzheimer’s disease in 2528 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. I. Ill an-Gala et al. APP-Derived Peptides in FTD patients with frontotemporal lobar degeneration and 23. Hawrylycz MJ, Lein ES, Guillozet-Bongaarts AL, et al. An amyotrophic lateral sclerosis with the C9ORF72 repeat anatomically comprehensive atlas of the adult human expansion. Dement Geriatr Cogn Disord 2015;39:287–293. brain transcriptome. Nature 2012;489:391–399. 9. Cirrito JR, Yamada KA, Finn MB, et al. Synaptic activity 24. Rascovsky K, Hodges JR, Knopman D, et al. Sensitivity regulates interstitial fluid amyloid-beta levels in vivo. of revised diagnostic criteria for the behavioural Neuron 2005;48:913–922. variant of frontotemporal dementia. Brain 10. M€ uller UC, Deller T, Korte M. Not just amyloid: 2011;134:2456–2477. physiological functions of the amyloid precursor protein 25. Dubois B, Feldman HH, Jacova C, et al. Advancing family. Nat Rev Neurosci 2017;18:281–298. research diagnostic criteria for Alzheimer’s disease: the 11. Bibl M, Mollenhauer B, Wolf S, et al. Reduced CSF IWG-2 criteria. Lancet Neurol 2014;13:614–629. carboxyterminally truncated Abeta peptides in 26. Armstrong MJ, Armstrong MJ, Litvan I, et al. Criteria for frontotemporal lobe degenerations. J Neural Transm the diagnosis of corticobasal degeneration. Neurology (Vienna) 2007;114:621–628. 2013;80:496–503. 12. Pijnenburg YA, Schoonenboom SN, Mehta PD, et al. 27. H€oglinger GU, Respondek G, Stamelou M, et al. Clinical Decreased cerebrospinal fluid amyloid beta (1–40) levels in diagnosis of progressive supranuclear palsy: the movement frontotemporal lobar degeneration. J Neurol Neurosurg disorder society criteria. Mov Disord 2017;32:853–864. Psychiatry 2007;78:735–737. 28. Gorno-Tempini ML, Hillis AE, Weintraub S, et al. 13. Verwey NA, Kester MI, van der Flier WM, et al. Classification of primary progressive aphasia and its Additional value of CSF amyloid-b40 levels in the variants. Neurology 2011;76:1006–1014. differentiation between FTLD and control subjects. J 29. Balasa M, Sanchez-Valle R, Antonell A, et al. Usefulness Alzheimers Dis 2010;20:445–452. of biomarkers in the diagnosis and prognosis of early- 14. Gabelle A, Roche S, Geny C, et al. Decreased sAbPPb, Ab38, onset cognitive impairment. J Alzheimers Dis and Ab40 cerebrospinal fluid levels in frontotemporal 2014;40:919–927. dementia. J Alzheimers Dis 2011;26:553–563. 30. Alcolea D, Martınez-Lage P, Sanchez-Juan P, et al. 15. Ghidoni R, Paterlini A, Albertini V, et al. A window into Amyloid precursor protein metabolism and inflammation the heterogeneity of human cerebrospinal fluid Ab markers in preclinical Alzheimer disease. Neurology peptides. J Biomed Biotechnol 2011;2011:697036. 2015;85:626–633. 16. Alexopoulos P, Guo L-H, Tsolakidou A, et al. 31. Alcolea D, Pegueroles J, Mu~ noz L, et al. Agreement of Interrelations between CSF soluble AbPPb, amyloid-b 1– amyloid PET and CSF biomarkers for Alzheimer’s disease 42, SORL1, and tau levels in Alzheimer’s disease. J on Lumipulse. Ann Clin Transl Neurol 2019;6:1815–1824. Alzheimers Dis 2012;28:543–552. 32. Sala I, Illan-Gala I, Alcolea D, et al. Diagnostic and 17. Alcolea D, Carmona-Iragui M, Suarez-Calvet M, et al. prognostic value of the combination of two measures of Relationship between b-secretase, inflammation and core verbal memory in mild cognitive impairment due to cerebrospinal fluid biomarkers for Alzheimer’s disease. J Alzheimer’s disease. J Alzheimers Dis 2017;58:909–918. Alzheimers Dis 2014;42:157–167. 33. Ossenkoppele R, Pijnenburg YAL, Perry DC, et al. The 18. Alcolea D, Vilaplana E, Suarez-Calvet M, et al. CSF behavioural/dysexecutive variant of Alzheimer’s disease: sAPPb, YKL-40, and neurofilament light in clinical, neuroimaging and pathological features. Brain frontotemporal lobar degeneration. Neurology 2015;138(Pt 9):2732–2749. 2017;89:178–188. 34. Vanderstichele HMJ, Janelidze S, Demeyer L, et al. 19. Alcolea D, Irwin DJ, Illan-Gala I, et al. Elevated YKL-40 Optimized standard operating procedures for the analysis and low sAPPb:YKL-40 ratio in antemortem cerebrospinal of cerebrospinal fluid Ab42 and the ratios of Ab isoforms fluid of patients with pathologically confirmed FTLD. J using low protein binding tubes. J Alzheimers Dis Neurol Neurosurg Psychiatry 2019;90:180–186. 2016;53:1121–1132. 20. Illan-Gala I, Alcolea D, Montal V, et al. CSF sAPPb, YKL- 35. Zhou HH, Singh V, Johnson SC, et al. Statistical tests and 40, and NfL along the ALS-FTD spectrum. Neurology identifiability conditions for pooling and analyzing 2018;91:e1619–e1628. multisite datasets. Proc Natl Acad Sci USA 2018;115:1481– 21. Ossenkoppele R, Mattsson N, Teunissen CE, et al. 1486. Cerebrospinal fluid biomarkers and cerebral atrophy in 36. Montal V, Vilaplana E, Alcolea D, et al. Cortical distinct clinical variants of probable Alzheimer’s disease. microstructural changes along the Alzheimer’s disease Neurobiol Aging 2015;36:2340–2347. continuum. Alzheimers Dement 2018;14:340–351. 22. Pera M, Alcolea D, Sanchez-Valle R, et al. Distinct 37. Calero O, Hortig€ uela R, Bullido MJ, Calero M. patterns of APP processing in the CNS in autosomal- Apolipoprotein E genotyping method by Real Time PCR, dominant and sporadic Alzheimer disease. Acta a fast and cost-effective alternative to the TaqMan and Neuropathol 2013;125:201–213. FRET assays. J Neurosci Methods 2009;183:238–240. ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association. 2529 APP-Derived Peptides in FTD I. Ill an-Gala et al. 38. Arnatkevic I ut_e A, Fulcher BD, Fornito A. A practical frontotemporal dementia: looking beyond atrophy. Brain guide to linking brain-wide gene expression and 2019;142:1121–1133. neuroimaging data. NeuroImage 2019;189:353–367. 50. Meeter LH, Kaat LD, Rohrer JD, van Swieten JC. Imaging 39. Ashburner J, Friston KJ. Voxel-based morphometry–the and fluid biomarkers in frontotemporal dementia. Nat Rev methods. NeuroImage 2000;11(6 Pt 1):805–821. Neurol 2017;13:406–419. 40. Mi H, Muruganujan A, Ebert D, et al. PANTHER version 51. Fukumoto H, Cheung BS, Hyman BT, Irizarry MC. Beta- 14: more genomes, a new PANTHER GO-slim and secretase protein and activity are increased in the improvements in enrichment analysis tools. Nucleic Acids neocortex in Alzheimer disease. Arch Neurol Res 2019;47(D1):D419–D426. 2002;59:1381–1389. 41. Ferrari R, Lovering RC, Hardy J, et al. Weighted protein 52. Ten Kate M, Dicks E, Visser PJ, et al. Atrophy subtypes in interaction network analysis of frontotemporal dementia. J prodromal Alzheimer’s disease are associated with Proteome Res 2017;16:999–1013. cognitive decline. Brain 2018;141:3443–3456. 42. Paolicelli RC, Jawaid A, Henstridge CM, et al. TDP-43 53. Lucey BP, Mawuenyega KG, Patterson BW, et al. depletion in microglia promotes amyloid clearance but Associations between b-amyloid kinetics and the b- also induces synapse loss. Neuron 2017;95):297–308.e6. Amyloid diurnal pattern in the central nervous system. 43. Ljubenkov PA, Staffaroni AM, Rojas JC, et al. JAMA Neurol 2017;74:207–215. Cerebrospinal fluid biomarkers predict frontotemporal 54. Woollacott IOC, Nicholas JM, Heslegrave A, et al. dementia trajectory. Ann Clin Transl Neurol 2018;5:1250– Cerebrospinal fluid soluble TREM2 levels in 1263. frontotemporal dementia differ by genetic and 44. Illan-Gala I, Pegueroles J, Montal V, et al. Challenges pathological subgroup. Alzheimers Res Ther 2018;10:79. associated with biomarker-based classification systems for 55. Minami SS, Min S-W, Krabbe G, et al. Progranulin Alzheimer’s disease. Alzheimers Dement (Amst) protects against amyloid b deposition and toxicity in 2018;10:346–357. Alzheimer’s disease mouse models. Nat Med 45. Augutis K, Axelsson M, Portelius E, et al. Cerebrospinal 2014;20:1157–1164. fluid biomarkers of b-amyloid metabolism in multiple sclerosis. Mult Scler 2013;19:543–552. 46. Parnetti L, Paciotti S, Farotti L, et al. Parkinson’s and Supporting Information Lewy body dementia CSF biomarkers. Clin Chim Acta Additional supporting information may be found online 2019;495:318–325. in the Supporting Information section at the end of the 47. Lattanzio F, Abu-Rumeileh S, Franceschini A, et al. Prion- article. specific and surrogate CSF biomarkers in Creutzfeldt- Jakob disease: diagnostic accuracy in relation to molecular subtypes and analysis of neuropathological correlates of p- Table S1. Analytical procedures. tau and Ab42 levels. Acta Neuropathol 2017;133:559–578. Table S2. List of genes with the highest confidence score 48. Spitzer P, Lang R, Oberstein TJ, et al. A specific reduction of interaction with APP gene from STRING. in Ab1-42 vs. a universal loss of Ab peptides in CSF Table S3. List of genes with differential expression at differentiates Alzheimer’s disease from meningitis and cerebral regions where cortical thickness correlated with multiple sclerosis. Front Aging Neurosci 2018;10:152. APP-derived peptides.  49. Illan-Gala I, Montal V, Borrego-Ecija S, et al. Cortical Table S4. Gene enrichment analysis microstructure in the behavioural variant of 2530 ª 2019 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals, Inc on behalf of American Neurological Association.