Cerebrospinal Fluid YKL-40 and Chitotriosidase Levels in Frontotemporal Dementia Vary by Clinical, Genetic and Pathological Subtype

Background: Chronic glial dysfunction may contribute to the pathogenesis of frontotemporal dementia (FTD). Cerebrospinal fluid (CSF) levels of glia-derived proteins YKL-40 and chitotriosidase are increased in Alzheimer’s disease (AD) but have not been explored in detail across the spectrum of FTD. Methods: We investigated whether CSF YKL-40 and chitotriosidase levels differed between FTD patients and controls, across different clinical and genetic subtypes of FTD, and between individuals with a clinical FTD syndrome due to AD versus non-AD (frontotemporal lobar degeneration, FTLD) pathology (based on CSF neurodegenerative biomarkers). Eighteen healthy controls and 64 people with FTD (behavioural variant FTD, n = 20; primary progressive aphasia [PPA], n = 44: nfvPPA, n = 16, svPPA, n = 11, lvPPA, n = 14, PPA-NOS, n = 3) were included. 10/64 had familial FTD, with mutations in GRN(n = 3), MAPT(n = 4), or C9orf72 (n = 3). 15/64 had neurodegenerative biomarkers consistent with AD pathology. Levels were measured by immunoassay and compared using multiple linear regressions. We also examined relationships of YKL-40 and chitotriosidase with CSF total tau (T-tau), phosphorylated tau 181 (P-tau) and β-amyloid 1–42 (Aβ42), with each other, and with age and disease duration. Results: CSF YKL-40 and chitotriosidase levels were higher in FTD, particularly lvPPA (both) and nfvPPA (YKL-40), compared with controls. GRN mutation carriers had higher levels of both proteins than controls and C9orf72 expansion carriers, and YKL-40 was higher in MAPT mutation carriers than controls. Individuals with underlying AD pathology had higher YKL-40 and chitotriosidase levels than both controls and those with likely FTLD pathology. CSF YKL-40 and chitotriosidase levels were variably associated with levels of T-tau, P-tau and Aβ42, and with each other, depending on clinical syndrome and underlying pathology. CSF YKL-40 but not chitotriosidase was associated with age, but not disease duration. Conclusion: CSF YKL-40 and chitotriosidase levels are increased in individuals with clinical FTD syndromes, particularly due to AD pathology. In a preliminary analysis of genetic groups, levels of both proteins are found to be highly elevated in FTD due to GRN mutations, while YKL-40 is increased in individuals with MAPT mutations. As glia-derived protein levels generally correlate with T-tau and P-tau levels, they may reflect the glial response to neurodegeneration in FTLD.


Background
Frontotemporal dementia (FTD) causes progressive changes in behaviour (behavioural variant FTD, bvFTD) or language (primary progressive aphasia, PPA), and some individuals have concurrent motor neuron disease (MND) or an atypical parkinsonian disorder such as progressive supranuclear palsy (PSP) or corticobasal syndrome (CBS) [1]. Pathologically, most individuals have frontotemporal lobar degeneration (FTLD) with tau inclusions (FTLD-tau) or transactive response DNA binding protein-43 (TDP-43) inclusions (FTLD-TDP), although some, particularly those with logopenic variant PPA (lvPPA), have underlying Alzheimer's disease (AD) pathology [2,3]. Around two-thirds of cases are sporadic, but one-third are familial, associated most commonly with mutations in progranulin (GRN), microtubule-associated protein tau (MAPT) or chromosome 9 open reading frame 72 (C9orf72) [1]. Biomarkers are currently lacking that reliably differentiate the pathological changes in vivo in sporadic FTD, can predict onset of disease and guide timely initiation of future treatments in familial FTD, or assess treatment response in future clinical trials.
Although it is now well-established that levels of several glia-derived proteins are raised in CSF during various stages of neurodegenerative diseases such as AD or MND, fewer studies have explored how levels are altered in FTD. Three glia-derived proteins have been most extensively explored in neurodegenerative diseases: soluble triggering receptor expressed on myeloid cells 2 (sTREM2), YKL-40 (also known as chitinase-3-like protein 1, CHI3L1) and chitotriosidase (also known as CHIT1). TREM2 is an innate immune receptor expressed by myeloid cells, including microglia and peripheral macrophages [31], and is involved in phagocytosis, survival and migration of microglia. A soluble fragment (sTREM2) is cleaved and detectable in CSF and blood [32]. YKL-40 is a pro-inflammatory molecule released predominantly by activated astrocytes (and to a lesser extent by microglia) into the CSF and by activated peripheral macrophages into blood, which stimulates production of cytokines, and regulates macrophage, microglial and astrocytic function, endothelial cell migration and tumour angiogenesis [33]. Chitotriosidase is a chitin-degrading enzyme expressed by activated microglia (but not by astrocytes) in CSF [34], and by peripheral macrophages in blood [35]. It induces activation of a pro-inflammatory microglial phenotype [34] and has a range of other immunomodulatory functions, including stimulation of chemotactic factors, fibrosis and tissue remodelling [36].
We have recently shown that although CSF sTREM2 levels are not raised in a mixed cohort of individuals with a diagnosis of FTD compared with controls, they are higher in certain subgroups of FTD, such as individuals with a clinical syndrome consistent with FTD but underlying AD pathology, and in symptomatic GRN mutation carriers [37]. This has implications for the use of gliaderived proteins as CSF biomarkers in clinical trials for a condition as diverse as FTD. CSF YKL-40 and chitotriosidase levels have not been compared across all the clinical and the main genetic subtypes of FTD or correlated within FTD subgroups with levels of validated CSF neurodegenerative biomarkers used in clinical practice: total tau (T-tau), phosphorylated tau-181 (P-tau) and amyloid beta 1-42 (Aβ42). DOI This study therefore set out to examine how CSF YKL-40 and chitotriosidase levels differ between individuals with a clinical diagnosis of FTD and cognitively normal controls, and between different clinical and genetic subtypes of FTD. We also aimed to clarify how CSF YKL-40 and chitotriosidase levels differ between individuals with similar clinical FTD syndromes but different underlying pathologies: FTLD versus AD, based on the CSF tau/ Aβ42 biomarker profile. We also aimed to establish whether YKL-40 or chitotriosidase levels are associated with levels of T-tau, P-tau or Aβ42 in CSF, and to ascertain the relationship between YKL-40 and chitotriosidase levels, across the spectrum of FTD. Finally, we aimed to assess how YKL-40 and chitotriosidase levels are associated with parameters that may affect glial function and are relevant for future clinical trials, such as age, disease duration and sex.

Participants
The cohort consisted of 64 consecutively recruited individuals with dementia meeting consensus diagnostic criteria for either bvFTD [38] or PPA [39], and 18 healthy cognitively normal con-trols (as per [37], with the addition of an extra control, recruited subsequent to that study). Cases with additional motor neurone disease were not included in the study. The study was approved by the local NHS Research Ethics Committee and the Health Research Authority. All individuals gave informed written consent.
Within the patient group, 20 had bvFTD, 16 nonfluent variant PPA (nfvPPA), 11 semantic variant PPA (svPPA), 14 logopenic variant PPA (lvPPA) and 3 had a PPA syndrome not otherwise specified (PPA-NOS; not fulfilling criteria of any of the other PPA phenotypes). All participants with FTD were genetically screened for all known FTD causative mutations, including the C9orf72 expansion. Ten individuals were found to have familial FTD, with mutations in GRN (n = 3: two C31fs, one S78fs), MAPT (n = 4: two 10 + 16, two R406W) or C9orf72 (n = 3). All familial cases had a diagnosis of bvFTD except 2 individuals with GRN mutations who had nfvPPA. Demographics of the cohort are displayed in Table 1. Disease duration was calculated as the time, in years, between age at clinical onset of symptoms and date of CSF collection. There was no difference in age at CSF collection between the FTD and control groups (p = 0.854), but the svPPA subgroup were younger than the nfvPPA (p = 0.018) and lvPPA (p = 0.031) subgroups. There was no significant difference in age between any of the genetic subgroups or when compared with controls. There was a higher proportion of males in the FTD group than in the control group (p = 0.014), and a higher proportion of males in the bvFTD subgroup versus all other clinical subgroups and controls (p = 0.006), other than the PPA-NOS group, where all 3 participants were male (Table 1). There was no significant difference in disease duration between any of the clinical subgroups (p = 0.105). CSF T-tau/Aβ42 ratio, Values are all mean (standard deviation) except for gender and CSF T-tau/Aβ42 ratio; IQR, interquartile range; n/a, not applicable. Genetic mutations seen in bvFTD were GRN (n = 1: one S78fs), MAPT (n = 4: two 10 + 16, two R406W) or C9orf72 (n = 3), and in nfvPPA were GRN (n = 2: two C31fs).

CSF Collection, Processing and Biomarker Analysis
For all participants, CSF was collected and stored using standardised procedures [37,40]. Samples were collected by lumbar puncture in polypropylene tubes, which were immediately transferred to the laboratory. Samples were then centrifuged and the supernatant aliquoted and stored at −80 ° C within 30 min of arrival. Levels of T-tau, P-tau and Aβ42 were measured in CSF using commercially available INNOTEST sandwich enzyme-linked immunosorbent assays (Fujirebio Europe, Gent, Belgium).
CSF YKL-40 levels were measured using the commercially available Human YKL-40 Immunoassay Kit on the Mesoscale Discovery (MSD, Rockville, MD, USA) platform, with all samples assayed in duplicate and measured on the same day by a single operator using the same reagents. Briefly, CSF samples were diluted 1 in 400 with dilution buffer, and the provided standard reconstituted 1 in 20 using dilution buffer and serially diluted 1 in 4 to produce concentrations ranging from 50,000 to 12.2 pg/mL. 150 μL blocking agent was added to each well, and plates sealed and incubated at room temperate shaking at 500 rpm for 1 h. Plates were washed 3 times with 300 μL per well of PBS-T, then 50 μL of either diluted CSF sample, standard or blank (dilution buffer) was added to each well (pre-coated with capture antibody), and plates were sealed and incubated at room temperate shaking at 500 rpm for 2 h. Plates were washed 3 times with 300 μL of PBS-T and 25 μL of detection antibody solution (diluted to 1 in 50) added per well, then sealed and incubated at room temperate shaking at 500 rpm for 2 h. After a further 3 washes with PBS-T, 150 μL of Read Buffer T (diluted 1 in 2) was added to each well and the plate immediately analysed on the SECTOR Imager using a 4-parameter logistic model with averaged replicates.
CSF chitotriosidase levels were measured using the commercially available CircuLex Human ELISA Kit (MBL International, MA, USA) with all samples assayed in duplicate and measured on the same day by a single operator using the same reagents. Briefly, CSF samples were diluted 1 in 5 with dilution buffer and the provided standard was diluted to produce concentrations ranging from 3,600 to 56.25 pg/mL. 100 μL of either diluted CSF sample, standard or blank (dilution buffer) was added to each well, and plates sealed and incubated at room temperature for 1 h shaking at 300 rpm, then washed 4 times with 350 μL wash buffer. 100 μL of HRP conjugated detection antibody was added and plates sealed and incubated at room temperature for 1 h shaking at 300 rpm, then washed 4 times with 350 μL wash buffer. 100 μL of substrate agent was added to each well and plates were sealed, covered in foil and incubated for 15 min at room temperature shaking at 300 rpm. Finally, 100 μL of stop solution was added to each well in the same order as the substrate agent, and plate absorbance read immediately on a microplate reader at dual wavelengths of 450/540 nm. The concentration of chitotriosidase in each sample was calculated using a four-parameter fitting method based on the standard curve, using values which were blank corrected and averaged over replicates.

Participant Stratification
We performed three separate group comparisons: 1 By clinical syndrome, comparing bvFTD, nfvPPA, svPPA, lvPPA, PPA-NOS, and controls. 2 By genetic group, comparing those with GRN mutations, MAPT mutations, C9orf72 expansions, and controls. 3 By pathological group, comparing those with likely Alzheimer's disease, those with likely FTLD pathology, and controls. We used levels of CSF T-tau and Aβ42 to calculate the T-tau/ Aβ42 ratio for each participant to perform this stratification, classifying all individuals with dementia based on the CSF Ttau/Aβ42 ratio, with a cut-off of ≥1.0 (AD biomarker-positive, indicating likely AD) and < 1.0 (AD biomarker-negative, indicating likely FTLD) [37,40] ( Table 2). The cognitively normal controls formed a comparison group with all having a CSF Ttau/Aβ42 ratio of < 1.0. No significant difference in age at CSF was seen between these three groups (p > 0.050), but disease duration at CSF was lower in the AD biomarker-positive subgroup than the AD biomarker-negative subgroup (p = 0.037). There were significantly more males in the AD biomarker-negative subgroup (73.4%) than in controls (38.9%) and the AD biomarker-positive subgroup (60.0%) (p = 0.032).

Statistical Analysis
For YKL-40, levels were detectable in CSF of all individuals, so analyses were performed on all 18 controls and 64 individuals with FTD. For chitotriosidase, 3 individuals (one control, one sporadic bvFTD and one sporadic nfvPPA) had persistently undetectable levels of chitotriosidase in CSF, despite assaying their samples again at 1 in 5 dilution and using neat CSF. Approximately 6% of the population possess a homozygous 24-bp duplication in exon 10 of the CHIT1 gene which leads to a complete enzymatic deficiency of chitotriosidase [41]. These 3 individuals were very likely to be carriers of this mutation, and their levels would bias com-  parisons of the groups if included (or assigned the lower limit of detection); hence, they were excluded from the chitotriosidase analysis, leaving 17 controls and 62 individuals with dementia.
All analyses were carried out using STATA14 (Stata Corporation, College Station, TX, USA), with a significance threshold of p < 0.05. Shapiro-Wilk tests of raw values and Q-Q plots of residuals from multivariable linear regressions were used to test assumptions of normality.
Assessment of residuals in multivariable linear regression analyses of YKL-40 across groups revealed these were normally distributed and so met the assumptions required for parametric multivariable linear regression analysis. However, the same assessment for chitotriosidase revealed that residuals were not normally distributed, and so chitotriosidase values were natural log (Ln) transformed, which then met assumptions required for multivariable linear regression analysis. Multivariable linear regressions were used to compare YKL-40 and Ln(chitotriosidase) levels between groups (FTD vs. controls and between clinical, pathological and genetic subgroups and vs. controls), adjusting for age and sex in all analyses, and for disease duration in analyses involving comparison of disease groups (but not for genetic subgroups due to small sample size). Post hoc pairwise tests were used to compare individual subgroups.
In each group (except genetic subgroups due to small sample size), multivariable linear regressions were used to investigate the association between: (a) YKL-40 or Ln(chitotriosidase) levels and levels of CSF T-tau, P-tau and Aβ42, adjusted for age and sex (for the control group) and for age, sex and disease duration (for disease groups). Due to a non-linear relationship between T-tau levels and YKL-40 and chitotriosidase levels, T-tau values were Ln transformed and all regression analyses performed using Ln(T-tau). (b) Ln(chitotriosidase) and Ln transformed YKL-40 levels (due a to non-linear relationship between raw values), adjusted for age and sex, and for disease groups, disease duration as well. (c) YKL-40 or Ln(chitotriosidase) and both age at CSF collection (adjusted for sex in the control group and both sex and disease duration in disease groups) and disease duration at CSF collection (for disease groups, adjusted for age and sex). Mann-Whitney U tests were used to compare protein levels between males and females in each group (but not within genetic subgroups due to small sample size).
GRN mutation carriers had much higher levels of chitotriosidase compared with controls (mean [SD] = 9,492 [5,143]
Most clinical subgroups showed positive slopes for the association between YKL-40 and T-tau or P-tau levels (Fig. 3d, e), but this only reached significance in certain Control Dementia Aβ42 levels, these did not reach significance (Fig. 4f).
Although both pathological subgroups (AD biomarker-positive and AD biomarker-negative) and controls   Fig. 3i). There were no significant associations between chitotriosidase levels and T-tau, P-tau or Aβ42 levels in any pathological subgroup or controls   Fig. 6c) groups, or in any of the clinical (Fig. 6d) or pathological subgroups.
There were no significant associations between either YKL-40 (Fig. 7a) or chitotriosidase (Fig. 7c) (Fig. 7b, d) or pathological subgroups. CSF YKL-40 and chitotriosidase levels did not differ significantly between males and females in the whole cohort or in FTD or control groups, or in any of the clinical or pathological subgroups.

Discussion
This study shows that levels of two glia-derived proteins, YKL-40 and chitotriosidase, are raised in the CSF of individuals with a clinical diagnosis of FTD compared with controls. However, levels are not consistently raised across all clinical subtypes, with highest YKL-40 levels in lvPPA and nfvPPA, and highest chitotriosidase levels in lvPPA. Individuals with a clinical syndrome consistent with FTD but a CSF neurodegenerative biomarker profile consistent with AD pathologically have particularly high levels of both proteins compared with controls, and higher levels than individuals with a diagnosis of FTD and non-AD like CSF biomarkers (likely FTLD), who have a non-significant trend to higher levels than controls. In a smaller subgroup analysis, both YKL-40 and chitotriosidase levels are highly elevated in FTD due to GRN mutations, and YKL-40 levels are also elevated in FTD due to MAPT mutations. Associations between YKL-40 and chitotriosidase levels, and with T-tau, P-tau and Aβ42 levels, vary depending on clinical diagnosis and CSF biomarker profile. CSF YKL-40 levels, but not chitotriosidase levels, are associated with age, and neither are associated with disease duration.
Raised CSF YKL-40 levels have previously been demonstrated in several FTD cohorts versus controls [33,[42][43][44][45][46][47][48][49], including in familial FTD [48], and compared with individuals with primary psychiatric diagnoses [50]. In contrast, few studies have examined CSF chitotriosidase levels in FTD. One study found higher chitotriosidase levels in FTD (in cases without CSF biomarker or pathological confirmation) compared with healthy controls [51], but another found similar levels to controls in a mixed familial FTD cohort [48]. Our results and results from previous biomarker and histological studies suggest significant glial activation is present in individuals with clinical diagnoses of FTD. However, previous biomarker studies have included individuals with different clinical syndromes, gene mutations and underlying pathologies (or co-pathologies) within FTD cohorts. This has limited our understanding of how glia-derived proteins vary in CSF across the spectrum of FTD, and how these biomarkers may be useful for clinical trials targeting different FTD subgroups. Our study therefore aimed to elucidate how levels of YKL-40 and chitotriosidase vary across the spectrum of FTD by examining these proteins at the subgroup level.
CSF YKL-40 and chitotriosidase levels were significantly raised in several, but not all, clinical subtypes of FTD and there was significant variability in levels within clinical subgroups. The highest YKL-40 levels were seen in lvPPA and nfvPPA, but the PPA-NOS subgroup also had high levels (likely not reaching significance when compared with controls due to small sample size). The highest chitotriosidase levels were in lvPPA, but other clinical subgroups showed non-significant trends towards higher levels than controls. There were no significant differences in levels between clinical subgroups. Very few studies have explored this previously: one study examining CSF YKL-40 levels in bvFTD, nfvPPA, svPPA, CBS and PSP compared with controls found raised levels in all syndromes (except PSP) and no significant differences between FTD subgroups, although individuals with lvPPA were not delineated from an accompanying typical AD group [43]. A study of CSF chitotriosidase levels in FTD found no significant difference between bvFTD and PPA, or between PPA subtypes, although PPA subgroups were much smaller than in our cohort and lacked biomarker or pathological correlation [51]. Different clinical subtypes of FTD may have widely differing pathologies, co-pathologies and disease mechanisms and hence differing degrees of glial activation which could affect YKL-40 and chitotriosidase release. In particular, patients with MND have much higher CSF YKL-40 levels [46,48,[52][53][54][55] and CSF chitotriosidase levels or activity [34,48,51,[54][55][56][57] than controls, and higher chitotriosidase levels than in FTD [51], so our study did not include individuals with FTD-MND to avoid confounding results. The majority (78%) of our lvPPA subgroup had a CSF biomarker profile consistent with AD (rather than FTLD), and our nfvPPA subgroup contained 2 individuals with AD-like CSF biomarkers and 2 individuals with GRN mutations. In contrast, most individuals with bvFTD or svPPA had non-AD-like biomarkers and either a smaller percentage of (bvFTD) or no GRN mutations. We therefore hypothesise that the particularly high YKL-40 and chitotriosidase levels in lvPPA and high YKL-40 levels in nfvPPA may be due to more pronounced glial activation in individuals with underlying AD pathology and GRN mutations.
We were able to explore this further by stratifying our FTD cohort by CSF neurodegenerative biomarker profile (T-tau/Aβ42 ratio) rather than by clinical diagnosis. This was helpful in a previous study to demonstrate that DOI: 10.1159/000506282 CSF sTREM2 levels are raised in individuals with a clinical diagnosis of an FTD syndrome but AD-like CSF biomarkers, particularly lvPPA, compared with controls, but not in those with likely FTLD [37]. By repeating this approach, we confirm that there are also much higher levels of YKL-40 and chitotriosidase in the CSF of individuals with an FTD syndrome but AD-like CSF biomarkers compared with controls, and higher levels than in individuals with non-AD-like CSF (i.e., likely FTLD). Patients with typical amnestic AD have elevated levels of glia-derived proteins in CSF compared with controls, including YKL-40 [42,47,[58][59][60][61][62][63][64][65][66][67][68][69], sTREM2 [70][71][72][73], chitotriosidase [51,59,74,75], glial fibrillary acidic protein [76][77][78] and S100beta [79][80][81]. This suggests there is pronounced astrocytic and microglial activation in association with AD pathology. The soluble phosphorylated tau species found in patients with amnestic AD are highly toxic to microglia, resulting in pronounced microglial dysfunction and dystrophy [82] and tau oligomers co-localise with microglia, astrocytes and pro-inflammatory cytokines in patients with AD and FTLD [26]. Patients with clinical FTD but underlying AD pathology would therefore also be expected to have very high levels of glia-derived proteins compared with controls, and data from this study and our previous study [37] support this.
Although CSF YKL-40 and chitotriosidase levels did not differ significantly between individuals with FTD and non-AD-like CSF biomarkers (likely FTLD) and controls, there was a trend towards higher levels of both proteins in this group. There was also significant intra-group variability in protein levels, particularly for chitotriosidase, suggesting glial activation may vary considerably according to the FTLD subtype or disease mechanism. Other studies have explored this by stratifying pathologically confirmed FTLD groups and found higher CSF YKL-40 levels in both FTLD-tau and FTLD-TDP than in individuals with subjective memory impairment, and in FTLD-tau compared with AD [43,44,49]. One study found higher YKL-40 levels in "pure" FTLD-tau (excluding AD pathology) than in patients with FTLD-TDP or AD [44], although others have found higher levels in FTLD-TDP (but not FTLD-tau) compared with controls [43,45]. Differences in the number of genetic FTD cases in FTLD subgroups or inclusion of individuals with copathology are likely to have contributed to these disparities between studies. FTLD-TDP cohorts have included patients with concurrent MND [45], or patients with GRN mutations [43] who we demonstrate have very high levels of both YKL-40 and chitotriosidase. FTLD-tau co-horts have included differing numbers of MAPT cases, who we show have particularly raised YKL-40 levels. This variability in levels according to pathology or disease mechanism has implications for the use of inflammatory proteins as fluid biomarkers in both research studies and clinical trials for a disease as pathologically diverse as FTD. It also emphasises the importance of detailed stratification of cohorts or use of CSF biomarker or pathological correlation in biomarker studies of FTD.
Although we were unable to divide our cohort into FTLD subtypes due to a lack of cases with pathological confirmation, we were able to include a small number of individuals with familial FTD, enabling an exploratory analysis of CSF YKL-40 and chitotriosidase levels in a small number of individuals with known pathology (FTLD-TDP: GRN or C9orf72 mutations or FTLD-tau: MAPT mutations), and also differing disease mechanisms despite similar pathology (GRN and C9orf72 mutations). GRN mutation carriers had very elevated YKL-40 and chitotriosidase levels, and MAPT mutation carriers had high YKL-40 levels compared with controls. Very few studies have examined glia-derived CSF biomarkers in individuals with genetic FTD. In a recent study, CSF YKL-40 levels, but not CSF chitotriosidase levels, were significantly elevated in 23 familial FTD cases (combining C9orf72, GRN or MAPT mutation carriers) compared with controls [48]. However, genetic subgroup analyses were not performed to analyse differences between mutation types and the genetic group contained a significant proportion of patients with C9orf72 expansions (15/23) (who we found to have similar levels of both proteins to controls), which may have influenced results. The highly elevated levels of YKL-40 and chitotriosidase in our GRN mutation group is consistent with multiple studies showing elevated levels of other inflammatory markers in GRN mutation carriers [6,19,37,[83][84][85]. GRN haploinsufficiency results in significant microglial dysfunction and activation [4,21,86], which could lead to excessive YKL-40 and chitotriosidase release as a mutation-specific effect, exacerbated by the general glial response to neurodegeneration, perhaps explaining the higher levels in GRN than C9orf72 mutation carriers. Patients with heterozygous GRN mutations also display significant lysosomal dysfunction [87,88], which could exacerbate chitotriosidase release into CSF. Plasma [35] and CSF [89,90] chitotriosidase are highly elevated in the lysosomal storage disorder Gaucher's disease, where macrophages are chronically activated [35,91] and plasma levels are already used for monitoring treatment response [92]. Serum YKL-40 levels are also raised (and serum GRN levels Dement Geriatr Cogn Disord 2020;49:56-76 DOI: 10.1159/000506282 are reduced) in Gaucher's disease, and recombinant GRN reduces serum YKL-40 levels in GRN knockout mice and in fibroblasts from patients with Gaucher's disease [91]. This strengthens the evidence for a link between GRN haploinsufficiency, glial activation, and lysosomal dysfunction, which could be detectable at an early stage, and reversible, in GRN mutation carriers.
The high YKL-40 levels observed in MAPT carriers are consistent with elevated CSF YKL-40 levels in FTLD-tau [43,49] and colocalisation of activated astrocytes with tau oligomers in P301S MAPT mouse models [26]. There are also many activated microglia surrounding phosphorylated-tau positive neurons in MAPT P301S mice [8] or patients with P301S mutations [9] and pronounced frontotemporal microglial activation in MAPT mutation carriers [10]. This suggests that certain FLTD-tau pathologies, as well as tau in AD, may promote YKL-40 release. CSF chitotriosidase levels were also slightly raised in MAPT carriers, but this did not reach significance compared with controls. This may have been due to the small group size, or perhaps there is greater astrocytosis than microglial activation associated with certain MAPT mutations.
In order to explore further how biomarkers of glial activation link to neurodegeneration, we examined relationships between YKL-40 and chitotriosidase levels and CSF neurodegenerative biomarkers that are used in clinical practice and which reflect neuronal injury and tau pathology (T-tau and P-tau) and amyloid pathology (Aβ42). Overall, both YKL-40 and chitotriosidase levels were positively associated with T-tau and P-tau levels in FTD, but this association only reached significance in certain clinical subgroups. For T-tau, there was a significant association with YKL-40 in bvFTD and nfvPPA and with chitotriosidase in bvFTD and lvPPA. For P-tau, there was a significant association with YKL-40 in bvFTD and lvPPA and with chitotriosidase in lvPPA. There was a small positive association between Aβ42 levels and YKL-40 in lvPPA, and although most subgroups seemed to have a negative association between Aβ42 and chitotriosidase, none reached significance. This variation in the strength of association between biomarkers may be explained by underlying pathologies (different FTLD subtypes or AD) being associated with varying degrees of glial activation, neurodegeneration and tau pathology, or differences in clinical subgroup sizes. Positive associations between levels of CSF YKL-40 or chitotriosidase and neurodegenerative biomarkers have been identified in many studies of typical AD, particularly for T-tau [42,47,58,[65][66][67][68]93] and P-tau [42,58,65,66,68,69,93,94]. Consistent with this, there was a strong association between levels of P-tau and both glia-derived proteins in our lvPPA subgroup, where most individuals had ADlike biomarkers and high levels of both YKL-40 and chitotriosidase, suggestive of significant hyperphosphorylated tau pathology and glial activation. Our results are consistent with strong associations between T-tau and YKL-40 identified in other studies of FTD [49,58], but to our knowledge no studies have explored associations between chitotriosidase and neurodegenerative biomarkers in FTD. Our findings suggest that chitotriosidase release may be similarly linked to neurodegeneration in FTD.
In individuals with FTD but AD-like CSF (AD biomarker-positive subgroup), there were positive slopes for the association between YKL-40 and both T-tau and Ptau levels, but neither association reached significance, and chitotriosidase levels were not significantly associated with any neurodegenerative biomarker. This contrasts with the strong associations between sTREM2 and both T-tau and P-tau levels in this group found previously [37], and in studies of amnestic AD. However, levels of YKL-40 and chitotriosidase were very high in most individuals within this subgroup, so a lack of variability combined with a relatively small sample size may have hampered our ability to detect weak associations between biomarkers. It is unclear why there was a positive association between Aβ42 and YKL-40 in this subgroup, although others have shown a similar association [47,66] or negative [94] or no significant [42] association with Aβ42 in AD. In individuals with likely FTLD (AD biomarker-negative subgroup), there was a significant positive association between YKL-40 and both T-tau and P-tau levels and a trend towards a positive association between chitotriosidase and T-tau levels. This suggests that glial activation may correlate with the degree of neuronal injury, and perhaps also tau pathology, in individuals with FTLD, supporting histopathological studies showing pronounced astrocytosis and microgliosis in FTLD, particularly tauopathies [9,10,26,28,95].
We also analysed associations between YKL-40 and chitotriosidase levels in our cohort, which to our knowledge has not been explored directly in FTD previously, although these seem to correlate moderately in AD [74] and in a combined familial MND and FTD cohort [48]. There was a strong positive association between levels of both proteins in FTD overall, and in most clinical subgroups, although this reached significance only in bvFTD and nfvPPA. Levels of YKL-40 and chitotriosidase were also highly associated in the AD biomarker-negative sub-DOI: 10.1159/000506282 group, but this did not reach significance in the AD biomarker-positive subgroup, again likely due to high levels of both proteins in most individuals and smaller sample size. These results suggest that astrocytic and microglial activation arise in tandem in FTD syndromes due to FTLD and perhaps AD pathology.
Finally, we examined relationships between CSF YKL-40 and chitotriosidase levels and relevant clinical parameters such as age, disease duration and sex, which could independently affect glial activation. Age was strongly associated with YKL-40 levels in the whole cohort and in the FTD group, consistent with previous studies of YKL-40 in AD [47,61,68] and FTD [43,45,49], and a strong association between sTREM2 and age in FTD [37]. This may reflect increased glial activation associated with aging, especially within the context of neurodegeneration, and emphasises the importance of future studies of gliaderived biomarkers in neurodegenerative disease cohorts exploring associations with age and, where applicable, adjusting group comparisons for age. There was no significant association between age and chitotriosidase levels in any group. In MND, plasma chitotriosidase activity was not associated with age [57], and most studies of CSF chitotriosidase in AD and FTD have found no association with age [48,59,74]. It is unclear why this differs from YKL-40, but perhaps age has less of an influence on microglial chitotriosidase release. We found no association between either YKL-40 or chitotriosidase and disease duration in FTD, in contrast to sTREM2, where we previously described a negative association between CSF sTREM2 levels and disease duration [37], but consistent with a study of YKL-40 in FTD, where there was no association with disease duration [43]. However, our current data are cross-sectional, so we were unable to explore longitudinal changes in YKL-40 or chitotriosidase levels to confirm whether these markers alter throughout the disease course. Lastly, we found no difference in YKL-40 or chitotriosidase levels between males and females in any group, suggesting limited influence of sex on release of these proteins into CSF in both healthy individuals and patients with FTD.
Limitations of this study include the small size of some of the subgroups, which may have limited our power to detect significant differences between groups. However, this is inherent to a rare disease such as FTD which has multiple phenotypes, and difficult to avoid when analysing biomarker levels across a broad spectrum of disease, while confining CSF collection and biomarker analysis to one site in order to minimise inter-centre variation. Our patient cohort contained both individuals with a clinical diagnosis of an FTD syndrome most likely due to FTLD (bvFTD, svPPA and nfvPPA) and those more commonly associated with AD pathology (lvPPA). The use of clinical diagnosis rather than pathological confirmation as an inclusion criterion meant a combination of different pathologies and mutations in the FTD group may have affected YKL-40 and chitotriosidase levels in this group overall. However, we were able to dissect out differences in protein levels between broad pathological entities (FTLD vs. AD) and gene mutations through stratification of the FTD group by CSF biomarker profile and through a preliminary analysis by mutation type. We also intentionally used a stringent cut-off of T-tau/Aβ42 ratio > 1.0 to minimise misclassification of individuals into the wrong pathological subgroup, as employed previously [37]. In addition, all individuals with FTD were phenotyped in detail, meeting recent diagnostic criteria for bvFTD [38] or PPA [39].

Conclusions
We show that levels of two glia-derived proteins, YKL-40 and chitotriosidase, are higher in the CSF of individuals with a clinical diagnosis of FTD than in cognitively normal controls. However, levels are higher in individuals with an FTD syndrome due to underlying AD pathology (particularly lvPPA) than due to FTLD. We display preliminary evidence that there are mutation-specific differences in YKL-40 and chitotriosidase levels, with particularly pronounced elevations of YKL-40 and chitotriosidase in GRN mutation carriers, and YKL-40 in MAPT mutation carriers, which may remain undetected in mixed genetic FTD cohorts. As CSF YKL-40, and perhaps chitotriosidase, levels correlate with neurodegenerative biomarkers, particularly T-tau, and with each other, in individuals with likely FTLD, these proteins may reflect extensive astrocytic and microglial activation arising in tandem with neurodegeneration in individuals with FTLD.
Future studies should analyse CSF YKL-40 and chitotriosidase levels within larger cohorts of individuals with FTD, across a variety of clinical subgroups, and ideally in pathologically confirmed cases across the full spectrum of FTLD subtypes, with separate sporadic and genetic subgroups. Inclusion of a larger number of cases with mutations in GRN, MAPT and C9orf72 would enable confirmation of our preliminary observations of higher protein levels in symptomatic GRN and MAPT mutation carriers. Assessment of levels in presymptomatic mutation carri-Dement Geriatr Cogn Disord 2020;49:56-76 DOI: 10.1159/000506282 ers could establish when these change prior to expected symptom onset. Exploration of relationships between baseline and longitudinal measurements of CSF YKL-40 and chitotriosidase levels, and other markers of the disease process (such as serum or CSF neurofilament light levels or frontal lobe atrophy rate) in individuals with FTD, and presymptomatic individuals, would be extremely valuable. This could improve understanding of how chronic neuroinflammation links to neurodegeneration, enable determination of whether these proteins can be used as biomarkers of disease stage, intensity and progression, and provide validation for their use in upcoming clinical trials.

Statement of Ethics
The study was approved by the local NHS Research Ethics Committee and the Health Research Authority. All individuals gave informed written consent. The research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

Disclosure Statement
H.Z. has served at scientific advisory boards of Roche Diagnostics, Wave, Samumed and CogRx, has given lectures in symposia sponsored by Biogen and Alzecure, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB, a GU Ventures-based platform company at the University of Gothenburg (all outside the submitted work). The other authors declare that they have no competing interests. No other authors have any conflicts of interest.