YKL-40 in the brain and cerebrospinal fluid of neurodegenerative dementias

Detection of human YKL-40 by western blot was performed using the Human Chitinase 3-like 1 Antibody AF2599 (R&D Biosystems). Detection of murine YKL-40 by western blot, and human and murine YKL.40 by immunohistochemistry and immunofluorescence was performed using the Chitinase 3-like 1 antibody PA5–43746 (Thermo Fisher).

Anti-GAPDH 9484, anti-β-actin 8226 and anti-vimentin 92547 antibodies were from Abcam. Mouse anti-GFAP MAB360 was from Millipore. Rabbit Anti-GFAP Z0334 and anti-amyloid beta (β-amyloid) M0872 were from Dako. Anti-PrP(12F10) A03221 was from SPI Bio. Thioflavin was from Sigma. Chimeric recombinant prion protein (PrP), composed of the Syrian hamster residues 23–137 followed by sheep residues 141–234 (of the R154 Q171 polymorphism), was from Thermo Fisher.

Brain tissue was obtained from the Institute of Neuropathology Brain Bank (HUB-ICO-IDIBELL Biobank) and the Biobank of Hospital Clinic-IDIBAPS following relevant guidelines of both Spanish legislation and the local ethics committee. Neuropathological examination and characterization were carried out on paraffin-embedded samples in every case. Detailed information on neuropathology and inflammatory profiling of the sCJD and AD cohorts has been previously reported [39‐42]. sCJD MM1 and VV2 cases were selected due to their higher prevalence, but different clinical phenotypes [43]. The presence of infectious, metabolic and neoplastic diseases was ruled out in control samples. No correlation between post-mortem delay or sample storage time and levels of proteins and mRNA analysed was observed. A full summary of the cases is available in Additional file 1.

In order to establish the potential diagnostic utility of CSF YKL-40 measurement in the context of differential diagnosis of neurodegenerative dementias, we utilized samples from three independent cohorts: i) cases recruited at the Clinical Dementia Center at the University Medical Center Göttingen (Germany) (cohort 1), ii) cases recruited at the Hospital Clinic de Barcelona (Spain) (cohort 2) and iii) cases recruited at hospitals of northern Greece that were further processed at the Laboratory of Pharmacology (School of Pharmacy, Aristotle University of Thessaloniki), as part of additional molecular analysis to enhance disease diagnosis (cohort 3). Lumbar puncture was performed for diagnostic purposes followed by analysis of CSF standard parameters. YKL-40, tau, 14–3-3 and Aβ42 were analysed at the point of the first diagnostic work-up. In cohort 1, the neurological control group was composed of patients with either clinically or pathologically defined alternative diagnoses [44, 45]. This group included cases of psychiatric disorders, ischemic stroke, epilepsy, autoimmune diseases, alcohol abuse, headache, vertigo, pain syndromes, acute hypoxia and non-related neurological conditions. Control cases did not present biomarker profiles indicating the presence of a neurodegenerative disease. Diagnosis of AD was based on the International Classification of Diseases, Tenth Revision (ICD-10) definition for AD (F.00-G.30). Rapidly progressive Alzheimer’s disease (rpAD) was defined by a velocity of cognitive decline >6pts/yr. on the Mini Mental Status Examination scale [46]. Velocity of decline was calculated using linear regression (least square method) in accordance with Villemagne et al. [47]. The diagnosis of DLB was based on the criteria of McKeith [17] and PDD as dementia in patients with PD. VaD diagnosis was based on ICD 10 definition (F01) and NINDS-AIREN criteria [48]. In cohort 2, controls were neurologically healthy individuals with no neurological clinical diagnosis and normal neuropsychological assessment. AD patients were diagnosed with probable AD using the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association criteria [49] and displayed the typical CSF biomarker profile, characteristic of AD (low levels of amyloid beta 1–42 (Aβ42) and elevated t-tau and p-tau levels). In cohort 3, controls were composed of subjects with cognitive impairment/dementia (of unknown kind) with excluded prion diagnosis. sCJD cases were classified as definite cases by neuropathological examinations or as probable cases according to diagnostic consensus criteria [50, 51]. Genetic prion cases were confirmed by genetic testing (PRNP D178N and E200K mutations). Additional file 2 provides information on demographics and CSF biomarkers results.

As a sCJD MM1 mouse model, the tg340 mouse line expressing about 4-fold level of human PrP M129 on a mouse PrP null background [52] was used. Control or sCJD MM1 brain tissues were inoculated as 10% (w/v) homogenates in 6–10 week-old mice in the right parietal lobe using a 25-gauge disposable hypodermic needle. Mice were observed daily and the neurological status was assessed weekly. The animals were sacrificed at pre-symptomatic (pre-clinical: 120 day post inoculation (dpi) and symptomatic (early clinical: 160 dpi and late clinical: 183 dpi) stages. Additionally, sCJD MM1 inoculum dilutions were performed to study prolonged disease times; animals were sacrificed at 210 dpi (10–1 dilution). Part of the brain was fixed by immersion in 10% buffered formalin to quantify spongiform degeneration and perform immunohistological analyses. The remaining brain was snap frozen and stored at −80 °C for protein and RNA studies. The 5xFAD mouse line harbouring the APP KM670/671NL (Swedish), APP I716V (Florida), APP V717I (London), PSEN1 M146 L (A>C), PSEN1 L286 V mutations was used as a model of AD pathogenesis. This widely-used model recapitulates several AD-related phenotypes displaying severe amyloid pathology and gliosis at about 2–3 months, cognitive alterations at 4–5 months and synaptic and neuronal loss at 9 of age [53, 54]. For scrapie infections with prion strains RML (immunohistochemical data) and 22 L (qPCR data), all animal experiments were in strict accordance with the principles of laboratory animal care (NIH publication No. 86–23, revised 1985) and the recommendations in the Guide for the Care and Use of Laboratory Animals of the German Animal Welfare Act on protection of animals. Animals were maintained under specific pathogen-free conditions. Mice were anesthetized with Ketamine/Xylazin hydrochloride prior to inoculation. Eight weeks old mice were inoculated intracerebrally (i.c.) with RML 5.0-prions (immunohistochemical data) and 22 L prions (qPCR data). Animals were checked daily and were taken at indicated time points or were allowed to progress to clinical prion disease (presenting with slowness, ataxia, weight loss, trembling and ungroomed fur). For scrapie infections with RML strain (qPCR data), all experiments were performed in the animal facility of the Neurological Department, AHEPA University Hospital. Approximately 8–12 weeks old mice were inoculated. Infectious material corresponding to 100 μl of 1% brain homogenate from terminally ill RML-infected C57Bl/6 mice (104.5 × LD50) was intraperitoneally administered in fourteen animals; ten age-matched unchallenged mice were used as controls. RML-infected C57Bl/6 mice developed the first clinical symptoms, corresponding to abnormal gait, at approximately 130 days post inoculation. More severe symptoms, including loss of muscle tone, hind limb weakness, ataxia, weight loss and ruffled coat, appeared at approximately 160 days post inoculation. Whole brain was harvested from three animals per group at pre-clinical (120 dpi) and end-point clinical (180 dpi) disease stages.

Organotypic cerebellar slice preparations from postnatal day 9–13 (P9–13) pups of C57BL/6JRj mice, further infection with 22 L scrapie strain, sample collection and detection of PK-resistant Prion Protein (PrPres) were performed as described previously [55].

Total tau was measured using INNOTEST™ hTAU Ag; (Fujirebio). Levels of amyloid β1–42 (Aβ42) were measured with an ELISA kit (INNOTEST™ AMYLOID (1–42); Fujirebio). YKL-40 was measured using the MicroVue YKL-40 EIA ELISA kit from Quidel following the manufacturer’s instructions. Protein 14–3-3 was detected with western blot as previously described [56].

RNA from different human and mouse brain regions was purified using the miRVANA RNA isolation kit following the manufacturer’s protocol. For human samples, regions utilized for RNA extraction and corresponding age and gender information is summarized in Additional file 1. RNA was extracted from the following murine brain regions: tg340MM1-CJDMM1: cerebral cortex, RML scrapie: whole brain, 22 L scrapie: cerebral cortex and cerebellum. RNAs were treated with DNase Set (Qiagen) to eliminate genomic DNA contamination. Sample concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific) and RNA integrity was assessed based on the RNA Integrity Number (RIN value) measured by an Agilent 2100 Bioanalyzer (Agilent). RNA samples were retrotranscribed using the High Capacity cDNA Archive kit (Applied Biosystems). Quantitative PCR assays were performed in duplicates on cDNA samples, in a LightCycler® 480 System (Roche). The reactions were set using 20xTaqMan Gene Expression Assays and 2xTaqMan Universal PCR Master Mix (Applied Biosystems) and conducted under the following conditions: 50 °C for 2 min, 95 °C for 10 min, 40 cycles at 95 °C for 15 s and 60 °C for 1 min. Fold changes were estimated using the 2-ΔΔCT equation. Fold changes in mRNA expression were determined relative to the control cases after normalization with housekeeping genes; statistical significances were calculated using GraphPad Prism 6.01. As housekeeping genes, GAPDH and GUSB as well as Gusb and Hprt were used as in human and mouse samples respectively. The Taqman probes used were: CHI3L1 (YKL-40) human: Hs01072228_m1, GFAP human: Hs00909240_m1, AIF1 (IBA-1) human: Hs00741549_g1, GAPDH human: Hs03929097_g1, GUSB human: Hs00939627_m1, Chi3l1 (YKL-40) mouse: Mm00801477_m1, Hprt mouse: Mm03024075_m1, GusB mouse: Mm00446958_g1.

Human and mouse tissues were lysed in lysis buffer containing: 100 mM Tris pH 7, 100 mM NaCl, 10 mM EDTA, 0.5% NP-40 and 0.5% sodium deoxycolate plus protease and phosphatase inhibitors (Sigma). Samples were centrifuged (14,000×g for 20 min at 4 °C) and supernatants were quantified for protein concentration using the Bradford method (Biorad). Samples were mixed with SDS-PAGE sample buffer, heated to 95 °C for 10 min, and resolved by SDS-PAGE analysis. Gels were transferred onto PVDF membranes, blocked and incubated with anti-YKL-40 antibodies (1:1000 dilution) as indicated above. Following incubation with HRP-conjugates, species-specific secondary antibody, the membranes were processed for specific immunodetection using the ECL reagent. GAPDH and β-actin antibodies at 1:5000 dilutions were used as loading controls. Fold changes (in arbitrary units) were determined from densitometric analysis relative to the control cases.

The immunohistochemical study was performed on 4-μm thick dewaxed brain sections. Endogenous peroxidases were blocked with peroxidase (Dako, Glostrup) followed by 10% normal goat serum. Following incubation with the anti-YKL-40 antibody (diluted 1:200) at room temperature overnight, the sections were incubated with EnVision1 system peroxidase (Dako) for 15 min at room temperature. The peroxidase reaction was visualized with diaminobenzidine (DAB) and H₂O₂. Primary antibody was omitted in some sections as immunostaining control and no signal was detected after incubation with the secondary antibody. Sections were counterstained with haematoxylin. For immunofluorescence, 4 μm thick dewaxed sections were stained with a saturated solution of Sudan black B (Merck, Germany) for 15 min to block the autofluorescence of lipofuscin granules present in cell bodies, and then rinsed in 70% ethanol and washed in distilled water. The sections were boiled in citrate buffer to enhance antigenicity, blocked for 30 min at room temperature with 10% fetal bovine serum diluted in phosphate buffered saline, and incubated at 4 °C overnight with primary antibodies. Double-labelling immunofluorescence was performed using a combination of YKL-40 (1:100) and GFAP (1:200 or 1:400) antibodies or PrP (12F10) (1:200) and either Aβ (1:50), YKL40 (1:100) or GFAP (1:400) antibodies. After washing, the sections were incubated with Alexa (Molecular Probes) fluorescence secondary antibodies (Alexa-Fluor 488 and 555 conjugated secondary antibodies). Nuclei were stained with DRAQ5™ (Biostatus Ltd., Shepshed, UK). After washing, the sections were mounted in Immuno-Fluore mounting medium (ICN Biomedicals, Irvine, CA), sealed and dried overnight. Confocal images were acquired using a microscope Leica DMIRE2 and Leica confocal software. Z-Stack image interval was 0.5 μm. Single plane images for quantification were obtained by Nikon Eclipse E800 microscope with ProgRes® CapturePro 2.7.7 software (JENOPTIK). Quantification of YKL-40 and/or GFAP positive cells was performed manually using Fiji ImageJ software in 13–15 areas per section in 3 different patients.

Immunofluorescence of frozen brain sections in 5xFAD mice was carried out as described [57]. For the analysis of human samples, two to four cases per diagnostic group were used and three to four sections from every case were analysed. For immunohistochemical analysis upon prion infection, tissues were fixed with 4% buffered formalin, prion infected samples were inactivated by immersion in 98% formic acid for one hour, post fixed in 4% buffered formalin overnight and processed for paraffin embedding. Sections (2 μm) were subjected to staining with the CHI3L1 antibody (YKL-40, 1:100) using a Ventana Benchmark XT (Ventana, Tuscon, Arizona, USA). Sections were incubated with primary antibody for 1 h, anti-rabbit Histofine Simple Stain MAX PO Universal immunoperoxidase polymer (Nichirei Biosciences, Wedel, Germany) was used as secondary antibody. Detection of secondary antibodies and counter staining was performed with an ultraview universal DAB detection kit from Ventana (Ventana, Tuscon, Arizona, USA). Representative images were taken with a Leica DMD108 digital microscope.

Detection of murine YKL-40 in tissue homogenates was performed using the Mouse Chitinase 3-like 1/YKL-40 PicoKine™ ELISA Kit (Bosterbio) following the manufacturer’s instructions. Inter- and intra assay coefficients of variation were below 15%. Dilution linearity tested from neat CSF to 1:8 dilution factor. Percent linearity was 85–112%. For the assays, CSF was diluted 1:3.

RT-QuIC was performed as described previously [58, 59] with minor modifications. Briefly, 85 μl of reaction buffer containing recombinant PrP (10 μg) was mixed with 15 μl of a 1:100 dilution of 10%w/v brain homogenates. Tissues for RT-QuIC analysis were lysed in PBS 0.1% SDS and clarified by centrifugation for 10 min at 10000×g. Reactions were set at a final volume of 100 μl and placed in 96-well black optical bottom plates (Fisher Scientific). Each sample was run in triplicate. Plates were sealed and incubated in a FLUO Star OPTIMA plate reader (BMG Labtech Ortenberg, GE) at 42 °C for 80 h with intermittent shaking cycles, consisting of 1 min double orbital shaking at the highest speed (600 rpm) followed by a 1 min break.

For two-group comparisons the Mann–Whitney test (non-parametric distribution) and unpaired t-test (parametric distribution) were used. In multiple comparisons, the Kruskal–Wallis test was used. Dunn’s multiple comparison test was used for post-hoc analysis. Values of p < 0.05 were considered significant. Pearson (parametric distribution) and Spearman (non-parametric distribution) tests were used to detect correlations between biomarkers, demographic data and mRNA expression levels. Two-way ANOVA followed by Bonferroni’s post-hoc was used in the quantification YKL-40 and GFAP positive cells. Receiver operating characteristic (ROC) curves and derived area under the curve (AUC; AUC = 0.5 no discrimination and AUC = 1 perfect discrimination) were calculated. The best cut-off value, sensitivity and specificity were estimated based on the Youden index [60] (point on an ROC curve providing the best balance of both sensitivity and specificity). ROC curves and statistical analyses were performed using the Graph Pad Prism 6.01 software.

For detection of a prion disease-associated mutation, genetic testing was performed on genomic DNA isolated from blood or brain tissue [61]. Informed consent for the prion protein gene (PRNP) analysis was obtained from each patient or legal guardian.

Expression of YKL-40 was initially analysed in sCJD as this neurodegenerative dementia presents the highest neuroinflammatory profile, including massive gliosis [39] and regulation of inflammatory mediators [38] compared to other neurodegenerative dementias [38, 40]. mRNA analysis showed that YKL-40 was highly increased in the frontal cortex (FC) and cerebellum (CB) of sCJD subtypes MM1 and VV2 (p < 0.001). No statistically significant differences were observed between sCJD subtypes in the two regions studied. Mean fold change values for controls versus sCJD comparisons were 6.1 in the FC and 11.5 in the CB (Fig. 1a). YKL-40 levels were normalized with GAPDH, and similar results were obtained for GUSB normalization (Additional file 3A). In agreement with mRNA data, YKL-40 protein levels were increased in the FC and CB of sCJD subtypes MM1 and VV2 (p < 0.001). Mean fold change values for controls versus sCJD comparisons were 4.5 in the FC and 4.6 in the CB (Fig. 1b).

Further, immunohistochemistry was used for the study of YKL-40 expression and distribution in control and sCJD brains. YKL-40 immunoreactivity in normal brain was restricted to fibrillar astrocytes (stellate astrocytes with long processes) in the white matter and was increased in both the cerebral cortex and white matter of sCJD brain tissue sections compared to controls (Fig. 2a). YKL-40 was localized predominately in reactive protoplasmic and perivascular astrocytes, in addition to increased signal in fibrillar astrocytes of the white matter. Increased YKL-40 immunoreactivity was also detected in the Bergmann glia in the CB of sCJD cases (Fig. 2b). No apparent correlation was observed between PrPSc amorphous deposits in the CB in VV2 and YKL-40-positive astrocytes (data not shown). Specific expression of YKL-40 in astrocytes was confirmed by a positive co-localization between YKL-40 and GFAP in FC (Fig. 2c). Interestingly, double PrP-GFAP and PrP-YKL-40 immunostaining in the FC of sCJD revealed the presence of astrocytic processes within the PrP plaque-like deposits (Additional file 4A and 4B).

Additionally, we found a positive correlation between mRNA levels of YKL-40 and the astrocytic marker GFAP (p < 0.01) in the FC of sCJD cases, while no correlation was observed with the microglial marker AIF1 (p > 0.05) (Fig. 2d). As positive but moderate correlation between GFAP and YKL-40 expression levels were observed, the association between YKL-40+ and GFAP+ astrocytes was further explored. Quantification of double-labelled GFAP and YKL-40 immunoreactive cells showed co-localization in a majority of cells (GFAP+/YKL-40+ = 251 ± 8 cells/mm2) whereas the number of cells showing only GFAP staining or only YKL-40 staining was significantly smaller (44 ± 15 cells/mm2 and 3 ± 2 cells/mm2, respectively (p < 0.001) (Additional file 5A and B).

To gain insight into the expression levels of YKL-40 in pre-clinical and early clinical sCJD stages, we took advantage of the tg340-PRNP-129MM mouse line inoculated with sCJD MM1 brain homogenate, which fully recapitulates the neuroinflammatory profile of human patients [40]. Analysis of the mRNA levels indicated that YKL-40 was increased at pre-clinical (120 dpi) and early-clinical (160 dpi) disease stages (p < 0.05). Differences between control and sCJD MM1-inoculated animals were more significant at clinical stages (180 dpi, p < 0.001 and 210 dpi, p < 0.01), resembling observations in sCJD post-mortem brains (Fig. 3a). Western blots from the same set of inoculated animals validated the presence of elevated YKL-40 at pre-clinical (p < 0.05) and clinical (p < 0.001) disease stages at the protein level as well (Fig. 3b).

To ascertain whether increased YKL-40 expression was also present in non-human prion diseases, wild-type mice inoculated with 22 L and RML scrapie strains were utilized. Increased YKL-40 mRNA levels were detected in the whole brain of RML-infected animals at late pre-clinical (120 dpi, p < 0.05) and clinical (180 dpi, p < 0.001) stages of the disease (Fig. 3c). Immunohistochemical analysis of RML-infected mice revealed the presence of YKL-40 positive astrocytes in the hippocampus and thalamus at clinical stages. In contrast to sCJD infected mice no staining was detected in cortical regions (Fig. 3d). At pre-clinical stages sparse positive YKL-40 positive cells were detected in the thalamic nuclei (Fig. 3d). In contrast, mice inoculated with the 22 L strain, which led to a faster clinical onset and a distinct neuropathology including more cortical PrPSc deposition pattern, presented increased YKL-40 mRNA levels in the cerebral cortex at pre-clinical (p < 0.05) and clinical (p < 0.001) disease stages (Fig. 3e). Additionally, YKL-40 mRNA levels were highly increased in the cerebellum of 22 L infected mice (Fig. 3f) (p < 0.001).

The observation of increased YKL-40 in the cerebellum of 22 L infected mice allowed to speculate that (22 L)-infected cultured organotypic cerebellar slices (COCS), which recapitulate most of the morphological and neuropathological hallmarks associated with prion infection and serve as an excellent ex vivo tool for the study of prion pathogenesis [55, 62], could recapitulate the alterations in YKL-40 expression observed in in vivo scrapie models. Ten weeks post-22 L infection, we detected astrogliosis by means of vimentin and GFAP expression (Fig. 3g). This was accompanied by positive seeding ability of pathogenic PrP by means of RT-QuIC analysis and presence of PrPres demonstrating the presence of prion infection on the COCS (Fig. 3h). At this time point, increased YKL-40 protein was detected in scrapie-infected COCS compared to mock treated cells (p < 0.01) (Fig. 3i).

Compared to controls, YKL-40 mRNA levels were increased in the FC of AD cases at early (AD I-III, p < 0.05) and late (AD IV-VI and rpAD IV-VI, p < 0.001) neurofibrillary tangle Braak stages (Fig. 4a). Mean fold change values for controls versus AD comparisons were 1.8 for AD II-III and, 2.9 for AD IV-VI and 3.2 for rpAD IV-VI. No differences were detected between AD and rpAD cases (p < 0.05). YKL-40 mRNA levels were normalized with GAPDH, and similar results were obtained for GUSB normalization (Additional file 3B). In agreement with mRNA data, YKL-40 expression at protein level was increased in the FC of AD IV-VI (fold change 2.58) and rpAD IV-VI (fold change 2.59) cases (p < 0.05) (Fig. 4b).

Moreover, YKL-40 expression was analysed by immunohistochemical staining in several AD brain regions. Increased YKL-40 expression was detected in the cerebral cortex, white matter and subpial layer of AD sections compared to controls, while it remained unchanged in the cerebellum of AD cases (Fig. 4c). Increased expression was also detected in the temporal cortex and throughout the hippocampus region of AD brain sections (Fig. 4d, and Additional file 6A). YKL-40 staining was mainly detected in astrocytes, similarly to the patterns observed in sCJD cases. In contrast with the diffuse distribution of YKL-40-immunoreactive astrocytes in sCJD, YKL-40-positive astrocytes in the cerebral cortex in AD predominated in clusters often surrounding a core of fibrillar β-amyloid (Fig. 4e). YKL-40-positive astrocytes were also present around blood vessels with β-amyloid (Fig. 4e and Additional file 6B). Double-labelling with YKL-40 and β-amyloid antibodies was used to demonstrate the presence of YKL-40 positive astrocytes surrounding senile and diffuse β-amyloid plaques in the hippocampus of AD cases (Fig. 4f and Additional file 7).

Fibrillar astrocytes in the subcortical white matter and deep white matter were also strongly YKL-40 immunoreactive. Specific expression of YKL-40 in GFAP-immunoreactive astrocytes in the cerebral cortex and white matter regions (Fig. 4g) in AD brains was confirmed by double-labelling immunofluorescence and confocal microscopy. Quantitative assessment of double-labelled GFAP and YKL-40 cells indicated a significant superior number of astrocytes showing co-localization of GFAP and YKL-40 (238 ± 23 cells/mm2), compared to those showing only GFAP or only YKL-40 staining (28 ± 6 cells/mm2 and 17 ± 3 cells/mm2, respectively) (p < 0.001) (Additional file 5C and D).

As an in vivo model of AD pathogenesis we used the 5xFAD mouse model. At 3 months of age, 5xFAD animals presented glial changes and amyloid plaques but do not show clinical signs or behavioural alterations [53, 54]. At this time point, we found increased YKL-40 and GFAP protein levels (p < 0.05) (Additional file 8A). At immunohistochemical level, widespread astrocytic YKL-40 expression was detected in 10-month-old 5xFAD animals, especially in the cerebral cortex and hippocampus (Additional file 8B) as well as in the thalamus, caudate putamen and midbrain, while sparse YKL-40 positive astrocytes could be detected in the molecular layer of cerebellum. At this stage, animals showed clear clinical and behavioural signs.

YKL-40 mRNA expression was further analysed in the FC of DLB cases. Despite increased mean values of YKL.40 detected in DLB and rapid forms of DLB (rpDLB) compared to controls, no statistically significant alterations were found between these groups (Fig. 5a). Similarly, no differences at the protein level were detected between control and DLB cases (Fig. 5b). Sparse YKL-40-positive astrocytes were observed in the cortical regions, white matter and subpial layer in DLB cases; sparse clusters of YKL-40-positive astrocytes were not uncommon due to the fact of the associated low burden of AD-related pathology in DLB (Fig. 5c).

When CSF YKL-40 was analysed in the differential diagnostic context of neurodegenerative dementias (cohort 1, Additional file 2), significantly higher YKL-40 levels were detected in sCJD cases (480 ± 178 pg/mL) compared to neurological controls (NC) (250 ± 77 pg/mL) and to the other dementia groups (p < 0.001) (Fig. 6a, Additional file 2). Compared to NC, YKL-40 was also increased in AD patients (386 ± 221 pg/mL, p < 0.001) but not in VaD (323 ± 96 pg/mL, p > 0.05) and DLB/PDD cases (314 ± 103 pg/mL, p > 0.05). To calculate the clinical accuracy of YKL-40 in discriminating between dementia and NC groups, we estimated the AUC values. As expected, the highest values were detected when comparing NC and sCJD cases (AUC: 0.92, 95% CI: 0.87–0.96) (Fig. 6b). Using an optimal cut-off at 315 pg/mL defined by the Youden index, an overall sensitivity of 85% and specificity of 84% could be achieved. For the rest of the dementias AUC values were far below those detected for sCJD cases: NC vs AD; AUC: 0.77, 95% CI: 0.68–0.84, NC vs VaD; AUC: 0.71, 95% CI: 0.60–0.84 and NC vs DLB: AUC: 0.70, 95% CI: 0.60–0.80 (Fig. 6b). Next we explored the ability of CSF YKL-40 to discriminate sCJD from the rest of dementia groups. While YKL-40 could not discriminate sCJD from AD with high accuracy (AUC: 0.70, 95% CI: 0.62–0.78), values were higher for sCJD vs DLB/PDD (AUC: 0.81, 95% CI: 0.72–0.89) and sCJD vs VaD (AUC: 0.77, 95% CI: 0.67–0.88) comparisons.

Further we analysed whether YKL-40 levels could be influenced by the rate of cognitive decline and/or disease duration. For this reason, we stratified AD cases into slow and fast progressive AD (AD versus rpAD) and sCJD samples according to codon 129PRNP polymorphism, which plays a role in the clinico-pathological phenotype of the disease and, as a consequence, in disease duration [43, 63, 64]. No differences in CSF YKL-40 levels were observed between AD (387 ± 244 pg/mL) and rpAD (384 ± 128 pg/mL) cases or between sCJD cases according to their codon 129 polymorphism: sCJD MM (462 ± 195 pg/mL), MV (519 ± 203 pg/mL) and VV (456 ± 79 pg/mL) (Fig. 6c, d).

Next we sought to investigate the potential correlation between YKL-40 and tau, disease duration and age in sCJD cases. A significant positive correlation of YKL-40 with CSF tau was found (p < 0.05) (Fig. 7a), but no correlation of YKL-40 levels with CSF Aβ42, disease duration and age at disease onset was detected (p > 0.05) (Fig. 7b and c).

In order to validate the presence of increased YKL-40 levels in sCJD, two independent cohorts were used. The first validation cohort (cohort 2, Additional file 2) included healthy controls (HC), AD and sCJD cases. Compared to HC (226 ± 82 pg/mL), YKL-40 levels were significantly increased in sCJD (411 ± 157 pg/mL, p < 0.001) and AD (311 ± 113 pg/mL, p < 0.01) (Fig. 8a, cohort 2 Additional file 2). Similar results were obtained for the second validation cohort (cohort 3, Additional file 2), where YKL-40 levels in sCJD were significantly higher (603 ± 198 pg/mL) than those detected in cognitive impairment/dementia cases (358 ± 129 pg/mL, p < 0.01) (Fig. 8b, cohort 3 Additional file 2). Finally, we aimed to determine the usefulness of CSF YKL-40 levels in the discrimination of prion disease from genetic etiology. For this purpose, we analysed genetic cases from two cohorts and compared them with their respective controls and sCJD groups. The study included Controls, sCJD, gCJD PRNP-E200K and Fatal Familial Insomnia (FFI) (PRNP-D178N mutation) cases (cohort 1 and cohort 2, Additional file 2). In both cohorts gCJD cases harbouring the E200K mutation showed similar YKL-40 levels compared to sCJD cases and were statistically different from controls (p < 0.001). In contrast, YKL-40 levels in FFI cases were lower than in sCJD cases, but significantly higher compared to control subjects (p < 0.05) (Fig. 9a and b and cohort 1 and cohort 2, Additional file 2).